The Archaeal Topoisomerase Reverse Gyrase Is a Helix-destabilizing Protein That Unwinds Four-way DNA Junctions*

Four-way junctions are non-B DNA structures that originate as intermediates of recombination and repair (Holliday junctions) or from the intrastrand annealing of palindromic sequences (cruciforms). These structures have important functional roles but may also severely interfere with DNA replication and other genetic processes; therefore, they are targeted by regulatory and architectural proteins, and dedicated pathways exist for their removal. Although it is well known that resolution of Holliday junctions occurs either by recombinases or by specialized helicases, less is known on the mechanisms dealing with secondary structures in nucleic acids. Reverse gyrase is a DNA topoisomerase, specific to microorganisms living at high temperatures, which comprises a type IA topoisomerase fused to an SF2 helicase-like module and catalyzes ATP hydrolysis-dependent DNA positive supercoiling. Reverse gyrase is likely involved in regulation of DNA structure and stability and might also participate in the cell response to DNA damage. By applying FRET technology to multiplex fluorophore gel imaging, we show here that reverse gyrase induces unwinding of synthetic four-way junctions as well as forked DNA substrates, following a mechanism independent of both the ATPase and the strand-cutting activity of the enzyme. The reaction requires high temperature and saturating protein concentrations. Our results suggest that reverse gyrase works like an ATP-independent helix-destabilizing protein specific for branched DNA structures. The results are discussed in light of reverse gyrase function and their general relevance for protein-mediated unwinding of complex DNA structures.

Four-way junctions are non-B DNA structures that originate as intermediates of recombination and repair (Holliday junctions) or from the intrastrand annealing of palindromic sequences (cruciforms). These structures have important functional roles but may also severely interfere with DNA replication and other genetic processes; therefore, they are targeted by regulatory and architectural proteins, and dedicated pathways exist for their removal. Although it is well known that resolution of Holliday junctions occurs either by recombinases or by specialized helicases, less is known on the mechanisms dealing with secondary structures in nucleic acids. Reverse gyrase is a DNA topoisomerase, specific to microorganisms living at high temperatures, which comprises a type IA topoisomerase fused to an SF2 helicase-like module and catalyzes ATP hydrolysis-dependent DNA positive supercoiling. Reverse gyrase is likely involved in regulation of DNA structure and stability and might also participate in the cell response to DNA damage. By applying FRET technology to multiplex fluorophore gel imaging, we show here that reverse gyrase induces unwinding of synthetic four-way junctions as well as forked DNA substrates, following a mechanism independent of both the ATPase and the strandcutting activity of the enzyme. The reaction requires high temperature and saturating protein concentrations. Our results suggest that reverse gyrase works like an ATP-independent helix-destabilizing protein specific for branched DNA structures. The results are discussed in light of reverse gyrase function and their general relevance for protein-mediated unwinding of complex DNA structures.
Four-way junctions are complex DNA structures that play a major biological role as intermediates in DNA rearrangements of various kinds. In particular, the Holliday junction (HJ) 2 is the central intermediate in DNA recombination and repair of collapsed replication forks, whereas cruciforms or stem-loop structures are four-way junctions due to intrastrand base pairing of inverted repeats in DNA, RNA, or DNA-RNA hybrids. These structures have a regulatory role but may also pose a threat to genome stability and cause replication, transcription, and translation stall, calling for specific mechanisms for their removal. Two classes of well characterized enzymes are able to resolve HJ: resolvases, which are responsible for its cleavage and formation of crossover products (1,2), and specialized helicases, which promote ATP hydrolysis-dependent branch migration of the junction (3)(4)(5). In archaea, representatives of the two classes of enzymes are the resolvase Hjc (6 -8) and the helicase Hjm (9), respectively. Diverse non-enzymatic architectural proteins whose action induces structural modification of DNA, and in particular DNA supercoiling, share the ability to bind four-way structures (10). For instance, the Structural Maintenance of Chromosomes (SMC) subunits of the cohesin and condensin complexes and the human DEK protein (which is involved in acute myeloid leukemias and in several autoimmune diseases), both inducing positive supercoiling of DNA upon binding, show high affinity for cruciform DNA (11,12). In addition, at least some DNA topoisomerases were shown to interact with their substrates preferentially at the level of DNA crossovers and to cleave four-way junctions in vitro (13,14).
Reverse gyrase is a unique DNA topoisomerase, specific of microorganisms living above 80°C. In contrast to all other topoisomerases, this enzyme introduces positive supercoils into DNA molecules, an activity possibly involved in genome stabilization and protection against thermal denaturation (for reviews, see Refs. 15 and 16). Reverse gyrase comprises two evolutionary highly conserved protein modules: a C-terminal type IA topoisomerase domain and an N-terminal domain similar to SF2 helicases, including an ATP-binding domain. ATP hydrolysis is essential for the positive supercoiling reaction, although true helicase activity has never been demonstrated. Several results suggest that reverse gyrase might participate in the cell response to DNA damage. In the crenarchaeon Sulfolobus solfataricus, reverse gyrase is recruited to DNA after UV irradiation (17), interacts with the single strand-binding protein (SSB), and the translesion DNA polymerase, PolY (18,19), and is degraded after treatment with methyl methane sulfonate (MMS) (20). Reverse gyrase structure and its involvement in genome stability are reminiscent of the evolutionary conserved complexes comprising topoisomerase III and members of the RecQ helicase family, which have essential functions in recombination and repair (for a recent review, see Ref. 5).
We show here that reverse gyrase is able to bind and unwind branched DNA structures, such as four-way junctions and forks. Unwinding is independent on the ATPase and cleavage activity of the enzyme, thus suggesting the existence of a previ-ously undescribed, non-enzymatic mechanism to process such important structures.

EXPERIMENTAL PROCEDURES
Proteins-All chromatographic separations were performed on Ä KTAFPLC systems (GE Healthcare Buckinghamshire, UK); protein concentration was determined with a Bio-Rad protein assay kit (Bio-Rad Pacific), and purity was assessed by SDS-PAGE. All proteins were diluted in RG buffer (20 mM phosphate buffer, pH. 6.5, 150 mM NaCl, 20% glycerol). Recombinant His-tagged S. solfataricus TopR1 reverse gyrase (hereafter called RG) as well as deletion and site-specific mutants were purified as described previously (21).
Native reverse gyrase (nRG) was purified from S. solfataricus P2 cells. Eight liters of cell culture were grown at 80°C as described previously (22) until exponential phase (0.4 -0.6 A 600 ). Cells were harvested by centrifugation, and soluble protein extract was prepared as described (17). Soluble extract was incubated for 30 min at 37°C with 20 units/ml Benzonase (Novagen) dialyzed against heparin buffer (20 mM Tris-HCl, pH 7.0) and loaded onto a heparin column (HiPrep 16/10 heparin FF, GE Healthcare). Proteins were eluted with a linear gradient of NaCl (0 -1 M NaCl). Throughout purification, fractions were analyzed by SDS-PAGE and Western blotting using the anti-RG polyclonal antibody (19). Positive fractions were pooled, concentrated to a final volume of 2 ml, dialyzed against phenyl buffer (50 mM phosphate buffer pH. 6.5, 1.2 M NaCl, 0.8 M ammonium sulfate, 1 mM EDTA, 1 mM DTT), and loaded onto a phenyl-Sepharose 26/10 column (GE Healthcare). Proteins were eluted with a linear gradient of ammonium sulfate (0.8 -0 M), and positive fractions were pooled, concentrated, dialyzed against 20 mM phosphate buffer, pH. 6.5, 150 mM NaCl, and loaded onto a 10/300 GL Superdex S200 column (GE Healthcare). Positive fractions were pooled, concentrated, and stored at Ϫ20°C with the addition of 20% glycerol, and positive supercoiling activity was analyzed by two-dimensional gel electrophoresis as described (23).
The Escherichia coli topoisomerase III gene was amplified from E. coli K12 genomic DNA using the oligonucleotides Topo III 5/Topo III3 (see Table 1). The oligonucleotides match the 5Ј-and the 3Ј-terminal ends of the coding sequence with the addition of a BamHI site at the 5Ј-end and a XhoI site at the 3Ј-end (underlined). The BamHI/XhoI fragment was cloned in PET29a in-frame with a sequence coding for a histidine tag at its N terminus, producing Pet-29a-Topo III. The Topo3 protein was overexpressed in E. coli BL21-AI strain. Cultures were grown at 37°C in 3 liters of Luria-Bertani (LB) medium supplemented with 50 mg/ml ampicillin, 0.1% glucose until 1.0 at A 600 . The culture was induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside and 0.2% arabinose and incubated over night at 37°C. Cells were harvested by centrifugation and resuspended in 30 ml of phosphate buffer (20 mM NaH 2 PO 4 , 20 mM Na 2 HPO 4 , pH 7.4) containing 500 mM NaCl, incubated at 37°C for 30 min with lysozyme (1 mg/g of wet cells) and Benzonase (25 units/g of wet cells), and broken by French press and sonication. The lysate was clarified, and ammonium sulfate was added to a final concentration of 35% saturation and centrifuged at 30,000 ϫ g for 20 min. The pellet was dissolved in 20 mM Tris-HCl, pH 8.0, 300 mM NaCl and dialyzed overnight against the same buffer. Samples were applied to a nickel-nitrilotriacetic acid column (GE Healthcare, HisTrap, 1 ml) and eluted with a linear gradient (0 -1 M) of imidazole. Fractions were analyzed by SDS-PAGE and Western blotting with the anti-His antibody (GE Healthcare); positive fractions were pooled, concentrated with the Amicon Ultra system (Millipore), and stored at 20°C with the addition of 40% glycerol. Activity was checked by DNA relaxation assay.
DNA Topoisomerase Assays-Positive supercoiling assays were performed at 70°C as reported (23) using either RG or nRG and pQE31 plasmid (Qiagen) as substrate. Samples were analyzed by two-dimensional agarose gel electrophoresis, analyzed, and quantified as reported (23). Relaxation assays were performed at 55°C as reported (24) using recombinant purified E. coli Topo3 and pQE31 (Qiagen). Samples were analyzed by monodimensional 1.2% agarose gels in 1ϫ Tris-borate-EDTA buffer, stained, and analyzed as described before (23).
Western Blots-Western blots were performed as described previously (19) using the anti-RG polyclonal antibody, which recognizes both recombinant TopR1 and recombinant TopR2 3 as well as reverse gyrase in S. solfataricus cell extracts (19).

TABLE 1 Oligonucleotides used in this work
Sequences and 5Ј/3Ј-modifications are indicated. Underlined sequences indicate the BamHI site at the 5Ј-end and the XhoI site at the 3Ј-end of Topo3, respectively.

Name Sequence (5 to 3) 5 modification 3 modification
For each substrate, stoichiometric amounts of purified oligonucleotides were annealed in Tris-EDTA buffer and heated for 5 min at 95°C followed by slow cooling to room temperature. The annealed substrates were loaded on an 8% (w/v) polyacrylamide gel in 0.5ϫ Tris-borate-EDTA buffer (pH 8.3). After gel migration, the bands were excised from the gel and incubated overnight at 37°C with shaking in soak solution (0.5 M ammonium acetate, 10 mM magnesium acetate, 1 mM EDTA, 0.1% SDS). Eluted DNA substrates were ethanol-purified and quantified using a Nanodrop TM 2000 (Thermo Scientific) instrument. Substrates used for EMSA and for KMnO 4 probing were prepared as described above using the same oligonucleotides but without fluorophores. Prior to annealing, the A2 oligonucleotide was radiolabeled at the 5Ј-end using [␥-32 P]ATP and T4 polynucleotide kinase.
Gel-FRET Assay-DNA substrates (20 nM) were incubated in HJ buffer (2.5 mM Tris-HCl pH.8.0, 0.25 mM ␤-mercaptoethanol, 5 mM NaAc, 0.5 mM MgCl 2 ) at indicated temperatures for 30 min without or with different amounts of the indicated proteins. To minimize variations among samples within each experiment, a single mix with all common components was set up; in both negative controls and samples, the total buffer concentration was kept constant by correcting with the appropriate amount of RG buffer. Reactions were terminated by adding 5ϫ stop solution (0.5% SDS, 40 mM EDTA, 0.5 mg/ml proteinase K, 20% glycerol), and samples were immediately loaded on an 8% polyacrylamide gel containing 0.1% SDS and run in 0.5ϫ Tris-borate-EDTA buffer at 150 V/cm. Reaction substrates and products were visualized by gel imaging on a VersaDoc 4000 TM (Bio-Rad) using the preset laser excitation and emission setting for Cy5 and Cy3 fluorophores: red LED and filter at 695 nm for Cy5, green LED and filter at 605 nm for Cy3, and green LED and filter at 695 nm for Förster resonance energy transfer (FRET). Each assay was performed at least three times.
EMSA-RG binding assays were performed as described previously (21) using either HJ 32 Plabeled at the 5Ј-end of the A2 oligonucleotide or 32 P-labeled A2 oligonucleotide double-stranded oligonucleotides (21) (see Table 1). Incubation was for 10 min at 37°C. Each assay was performed at least three times. Radioactivity was determined by  Table 1. The 5Ј-end of strand A2 and the 3Ј-end of strand A1 were labeled with Cy5 and Cy3 fluorophores, respectively. The graphic shows HJ and predicted junction unfolding products. B, Cy5/Cy3-labeled HJ (20 nM) was incubated at 55°C for 30 min without (lane 1) or with increasing concentrations of RG as indicated (lanes 2-5). Samples were run, and the same gel was scanned under different excitation/emission conditions as described under "Experimental Procedures." The panels show the images obtained for Cy3, Cy5, and FRET and the merge of all three images, respectively. Reaction substrate and products (forks, ss, and cleavage products, which are indicated by scissors) are indicated. The complicated band pattern seen in the merge panel is due to different rates of migration of the two fluorophore-labeled oligonucleotides with each other and with respect to the unlabeled ones. The gel is representative of 20 independent experiments obtained with three different protein preparations. autoradiography with a Storm PhosphorImager and quantified with the IQ-Mac software (GE Healthcare).
Modification of DNA by Potassium Permanganate-DNA probing by KMnO 4 was performed following the published procedure (6) using HJ 32 P-labeled at the 5Ј-end of the A2 oligonucleotide and 0.6 M RG. After electrophoresis, radioactivity was determined by autoradiography with a Storm Phosphor-Imager (GE Healthcare).
Fluorescence Measurements-Steady state fluorescence measurements were performed on a K2 fluorometer (ISS, Champaign, IL) equipped with a two-cell temperature controlled sample holder. Fluorescence emission spectra between 560 and 650 nm were recorded at a fixed excitation wavelength of 550 nm, with a slit width of 1.0 nm. Each emission curve obtained with different RG concentrations was normalized to the value obtained for the free junction in the presence of corresponding volumes of RG buffer and was corrected for lamp fluctuations and instrumental variations. All measurements were performed at 55°C. The effect of temperature on the fluorescence emission of TAMRA-A8, BHQ2-A9, and Q-HJ was almost negligible.

RESULTS
Reverse Gyrase Unwinds Fourway Structures-We developed a FRET-based electrophoretic technique (gel-FRET) to enable monitoring of assembly and disassembly of a four-way junction. We designed a synthetic junction (HJ) containing a central 4-nucleotide core homology and four arms of 20 bp each, in which the 3Ј-and 5Ј-ends of one arm were labeled with the Cy3 and Cy5 fluorophores, respectively (supplemental Fig. 1SA and Table 1). With respect to conventional radioactive methods, this technique offers the possibility to discriminate among different products, follow the fate of different strands, and determine the pathway of HJ processing in a single experiment (Fig.  1A). Because reverse gyrase is a highly thermophilic enzyme, whose activity is virtually undetectable below 50°C (21), we determined the optimal incubation conditions for the HJ (supplemental Fig. S1B); we chose incubation for 30 min at 55°C, which secured stability of the HJ and prevented spontaneous annealing of the denatured junction. The archaeon S. solfataricus encodes two isoforms of reverse gyrase, called TopR1 and TopR2, which are indistinguishable in size, activity, and immunoreactivity. 3 4S) as well as the nRG directly from S. solfataricus (supplemental Fig. 4S).
The addition of RG to HJ and incubation at 55°C followed by gel electrophoresis and multiplex fluorophore imaging showed RG concentration-dependent reduction of the HJ and appearance of faster products that, based on their migration and fluorophore labeling, were identified as the three forks, two fulllength ss, and two ss cleavage products ( Fig. 1B and supplemental Fig. 2S; see also below). Intermediate oligonucleotide combinations were also present. Supersaturating concentrations of protein (P/DNA ratio Ͼ10) were required for  NOVEMBER 19, 2010 • VOLUME 285 • NUMBER 47 efficient processing of HJ, whereas at lower RG concentration, the forks were the main products, and ss bands became gradually more evident with increasing RG concentration, suggesting that the reaction proceeds through formation of relatively stable fork intermediates, which are then processed to ss products (see below). Consistent results were obtained using a 32 P-labeled HJ (Table 1), indicating that the fluorophores do not affect the normal behavior of RG and HJ in this reaction (data not shown).

Four-way Junction Resolution by Reverse Gyrase
When the nRG purified from S. solfataricus was used, identical results were obtained ( Fig. 2A), thus ruling out any suspicion that the HJ unwinding activity might be due to a contaminant or some artifact of our His-tagged protein preparation. The high amounts of both RG and nRG needed for efficient HJ processing might raise doubts on the physiological significance of this activity; we thus determined the intracellular amount of reverse gyrase by quantitative Western blot (Fig. 2B). Reverse gyrase turned out to be a rather abundant protein, accounting for about 0.37% of the total soluble S. solfataricus protein.
Because exact data on the absolute intracellular protein content of this archaeon are not available in the literature, we could assume that, consistently with the relative size of the two genomes, S. solfataricus proteome is about half that of E. coli, which is 10 6 -10 7 protein molecules/cell (25). Thus, a conservative estimation would predict that reverse gyrase ranges between 2 ϫ 10 3 (about 4% of 5 ϫ 10 5 ) and 2 ϫ 10 4 (4% of 5 ϫ 10 5 ) molecules/cell. This means that there is enough reverse gyrase in the cell to have 20 -200 protein molecules binding simultaneously to 100 four-way junctions, which is likely far above the number of these structures in a cell.
Fluorescence Measurements-The gel-FRET experiments allow end point analysis and require extensive purification steps after the reaction to remove the protein, which might in principle affect the results. To follow the reaction in real time,  avoid artifacts due to protein removal, and quantify RG activity using an independent technique, we carried out fluorescence measurements. To overcome possible fluorophore quenching by the protein, leading to fluorescence reduction, we decided to synthesize a molecular beacon, an identical HJ (Q-HJ) in which two strands were labeled with the TAMRA fluorophore and its quencher BHQ2, respectively ( Fig. 3A and supplemental Table  S1). In this case, separation of the labeled strands is monitored by fluorescence increase. Fig. 3B shows that the fluorescence emission spectrum of the Q-HJ was almost completely quenched due to the fact that BHQ2 and TAMRA are located within Förster distance. The addition of RG resulted in a marked increase of the fluorescence emission signal at 580 nm, indicating disruption of the Q-HJ and separation of the two labeled DNA strands. The inset of Fig. 3B reports the effect of increasing amounts of RG on the Q-HJ. Consistently with gel-FRET data, the fluorescence emission increased with increasing RG concentration, reaching a plateau slightly above at 2.0 M, which corresponds to a P/DNA ratio of 40.
Reverse Gyrase Does Not Work Like Helicases or Resolvases in HJ Processing-Based on its known biochemical activities, one could anticipate that, by virtue of its helicase-like domain, reverse gyrase might promote branch migration of the HJ, a typical ATP-dependent reaction catalyzed by several helicases.
In addition, exploiting the cleavagereligation activity of its topoisomerase domain (21), RG might also act as recombinases, namely by cleavage, exchange, and religation of two DNA strands. We thus sought to elucidate the HJ processing reaction mechanism.
Both RG and its isolated N-terminal domain show DNA-dependent ATPase activity, which is essential for the positive supercoiling reaction (19,21); however, HJ processing by RG was unaffected by the addition of 5 mM ATP (Fig. 4A), 0.5 mM MgCl 2 , which is required for ATPase activity, or 1 mM EDTA, which inhibits ATPase activity (data not shown). At 37°C, RG was unable to process HJ, and ATP addition was ineffective in stimulating the reaction (Fig. 4A), suggesting that the energy derived from nucleotide hydrolysis cannot compensate for the temperature. In addition, the ATPase-deficient K116A mutant of RG (19) efficiently processed HJ (Fig. 4C). Because it is possible that under our conditions we are looking at single turnover reactions, we wondered whether ATP might affect product release and enzyme recycling under different circumstances. However, ATP failed to stimulate HJ processing in reactions containing lower enzyme concentrations (supplemental Fig. 4SA); in addition, the kinetics of the reaction was very rapid and was not affected by ATP (supplemental Fig. 4SB). Thus, the nucleotide does not appear to stimulate enzyme recycling. Taken together, these results suggest that RG processes HJ by an ATP-independent mechanism, which is thus distinct from that used by helicases.
To test whether RG might work like HJ resolvases, we used the Y965F mutant, carrying a substitution of the catalytic tyrosine (Fig. 4B), which impairs both DNA relaxation and positive supercoiling (19). Surprisingly, the Y965F protein was able to process HJ efficiently, although as expected, no cleavage products were obtained (Fig. 4D), suggesting that HJ resolution by RG does not require DNA cleavage. This conclusion was supported by mapping the cleavage sites by RG on the A1 and A2 strands of HJ, which showed major sites located 3-8 bases from the strand crossing; importantly, exactly the same cleavage sites were also found on ssDNA (data not shown). Taken together, these results suggest that cleavage is a secondary event occurring on ss produced in the reaction and is not involved in HJ processing. We thus conclude that HJ unwinding by RG occurs via a mechanism distinct from that used by resolvases.
Specificity of the Unwinding Reaction-Previously, we have obtained the two domains of RG in isolation (21) and have shown that the N-terminal domain is a DNA-dependent ATPase, whereas the C-terminal domain behaves like a canonical type IA DNA topoisomerase; both domains show the ability to bind ss, ds, and mixed substrates, as predicted from the presence of two putative zinc fingers motifs (Fig. 5A). When the two isolated domains are combined in the reaction, they make specific physical interactions and induce ATP-dependent positive supercoiling of DNA molecules (21). We tested the ability of the isolated domains to induce HJ unwinding (Fig. 5B); although the N-terminal domain was completely inactive, the C-terminal module was unable to unwind the HJ but produced a small amount of cleavage product. This latter is likely due to cleavage of the ssDNA present in the HJ preparation, an activity typical of type IA topoisomerases. To test the specificity of the HJ unwinding activity, we cloned and purified a His-tagged version of E. coli Topo3, a type IA DNA topoisomerase homologous to the C-terminal domain of reverse gyrase (24); the enzyme was suitable for our purposes because it was active in DNA relaxation experiments at 55°C (supplemental Fig.  3S, D and E). When used under the same conditions and high P/DNA ratio, Topo3 produced a small amount of shorter products similar to those observed for RG and its C-terminal domain, likely due to ss cleavage (Fig. 5C); however, it was unable to process HJ. In contrast, when the two domains of RG were combined together, HJ processing activity was restored (Fig. 5B). Taken together, these results suggest that the HJ unwinding activity is specific for reverse gyrase, requires both domains, and is not a general property of type IA DNA topoisomerases.
Reverse Gyrase Unwinds Different Branched Structures-Due to its 4-nucleotide homologous core, our HJ may shift between two extreme conformations (see Fig. 7C), and a number of studies revealed that changes in HJ structure may facilitate its recognition and resolution by junction-resolving enzymes (6,26). RG was able to process with similar efficiency a completely heterologous, "immobile" HJ (IM-HJ) identical to HJ except for the central core (supplemental Table S1), suggesting that the reaction is not specific for the mobile HJ or its AT-rich core sequence (Fig. 6A). Previous studies failed to demonstrate true helicase activity for RG or its N-terminal domain (18,21). However, when incubated under the same conditions reported above for HJ and IM-HJ and using a high P/DNA ratio (Ͼ20), RG efficiently unwound the Cy3-Cy5-labeled fork substrate in an ATP-independent reaction (Fig. 6B). In addition, the protein unwound with similar efficiency the single-fluorophore-labeled forks formed by oligonucleotides A1ϩA4 and A2ϩA3, respectively (supplemental Fig. 1S, data not shown). Under the same conditions, RG was not able to unwind a 40-bp fully ds substrate (Fig. 6C), thus suggesting that the reaction requires the presence of an ss-ds junction.
Reverse Gyrase Distorts HJ Structure upon Binding-Many HJ processing proteins show binding selectivity or higher affinity for four-way structures. Reverse gyrase binds ds, ss, and mixed DNA substrates with increasing affinity and a certain degree of cooperativity (21,27,28). To correlate ATP-independent unwinding activity with binding affinity, we performed EMSA experiments (Fig. 7A). The efficiency of binding of RG for HJ FIGURE 6. Substrates of ATP-independent unwinding. A, processing of immobile HJ. IM-HJ (20 nM) was produced by annealing the A1, A2, A5, and A6 oligonucleotides shown in Table 1 and was incubated at 55°C for 30 min without (lane 1) or with increasing concentrations of RG, as indicated (lanes 2-4). The panels show the images obtained for Cy3 and Cy5, respectively. The gel is representative of two independent experiments. B, ATP-independent unwinding of a fork substrate. A gel-FRET assay with fork substrate (oligonucleotides A1ϩA2 in supplemental Table S1, 20 nM) and RG (0.8 M) is shown. Incubation was for 30 min at 55°C. The panels show the images obtained for Cy3, Cy5, and FRET, respectively. The gel is representative of four independent experiments. C, RG does not unwind a ds substrate. A gel-FRET assay was performed with the 40-bp ds substrate (oligonucleotides A2ϩA7 in Table 1 was similar to that for ds oligonucleotides and significantly lower than that for a fork, which is the preferred substrate (21).
Many HJ-binding proteins cause structural distortion of HJs, which stabilizes HJs in their open form, facilitating unfolding of the junction (6,26). To determine whether RG might induce modification of the HJ structure, we used the permanganate probing technique (Fig. 7B). The experiment was performed at 37°C to prevent unwinding of the junction and assess the effect of the protein binding. In the presence of RG, high reactivity to permanganate was observed specifically at the level of two thymines, located 4 bases apart at the center of the oligonucleotide A2 and thus falling exactly at the point of strand crossing in the two possible HJ conformations (Fig. 7C). This experiment indicated that RG distorts HJ structure by disrupting base pairing specifically at its core, suggesting that HJ processing might be facilitated by decrease of duplex DNA stability brought about by contacts between RG and DNA, starting from the point of strand crossing.

DISCUSSION
The uniqueness of reverse gyrase phylogenetic distribution, structure, and activity has been repeatedly underlined (16,15). We have described another activity of reverse gyrase, the ability to unwind substrates containing helical junctions independent of the ATPase and DNA cleavage activity. Unwinding requires high temperature, both domains of reverse gyrase, the presence of either a crossover or an ss-ds junction in the substrate, and saturating amounts of protein; upon binding, reverse gyrase distorts the HJ specifically at its crossing point, inducing local base unpairing. The E. coli Topo3 enzyme could not substitute for reverse gyrase in this reaction, suggesting that it is not a general property of type IA topoisomerases. The experiments presented here imply a non-catalytic activity of reverse gyrase in which structural recognition, distortion, and junction unwinding are all tightly connected. We suggest that RG works as a branched structurespecific helix-destabilizing protein; due to its peculiar structure and fluctuation between stacked (open) and unstacked form, the HJ core is prone to distortion by reverse gyrase, which induces base melting or stabilizes "breathing" regions. These may in turn facilitate binding of other DNA molecules and possibly further distortion followed by strand destabilization. On the fork, the ss-ds junction acts FIGURE 7. Binding of RG to HJ distorts its structure. A, RG binding to HJ was analyzed by EMSA. Increasing concentrations of RG (60, 120, 240, 480 nM) were incubated at 37°C with the following 32 P-labeled substrates (20 nM): HJ, 40-bp ds (oligonucleotide A2ϩA7 in supplemental Table S1), fork (oligonucleotides A1ϩA2 in supplemental Table S1), and 80-bp ds (Valenti et al. 21). The graph shows the quantification of results expressed as the percentage of shifted DNA versus the P/DNA ratio. For each DNA ligand, the fraction of shifted DNA versus the amount of protein used is plotted. Binding assays were performed in triplicate, and the results were averaged. Values are the mean Ϯ S.E. of three independent experiments. B, HJ probing by potassium permanganate. 32 P-labeled-HJ (20 nM) was incubated with RG at 0.6 M (lanes 1 and 3) or without protein (lanes 2 and 4) at 37°C for 30 min; after incubation, samples were reacted with KMnO 4 followed by piperidine cleavage (lane 1 and 2) or directly incubated with piperidine (lane 3); lanes 5 and 6 show molecular weight markers of 22 and 18 nucleotides, respectively. Controls were mock-treated exactly as samples. The gel is representative of three independent experiments. C, the two extreme conformations of HJ with its 4-base homologous core (underlined). Arrows indicate the thymines specifically cleaved by piperidine after KMnO 4 treatment in the presence of RG. NOVEMBER 19, 2010 • VOLUME 285 • NUMBER 47 analogously to the HJ core, triggering the reaction. In this model, the energy required to unwind DNA is provided by protein-DNA or protein-protein interactions instead of ATP hydrolysis.

Four-way Junction Resolution by Reverse Gyrase
It was previously reported that the stoichiometric binding of reverse gyrase to an open circular DNA causes a decrease of the DNA linking number (after nick closure by a thermophilic ligase); this result was interpreted as local duplex unwinding induced by reverse gyrase binding (29). Interestingly, the activity was shown to be ATP-and DNA cleavage-independent, requiring high P/DNA ratio and the presence of both domains. Although the biological meaning of reverse gyrase-induced plasmid linking number decrease was not addressed, this activity is reminiscent of the junction unwinding described here because at high temperatures, plasmids may expose ss-ds junctions.
The analysis of the three-dimensional structure of reverse gyrase suggested the presence of a number of predicted DNAbinding regions spanning both N-terminal and C-terminal domains, including zinc fingers and extended regions of positive potential (30). The result that the two separate RG domains are not effective in HJ processing might suggest that cooperation among these binding sites might be required for this reaction.
The fact that HJ and fork unwinding is independent of the two reverse gyrase catalytic activities might seem unsound; however, both DNA topoisomerases and helicases have been reported to function independently of their catalytic activities under specific circumstances. For instance, human Topo III␣ (a type IA DNA topoisomerase) stimulates HJ unwinding by the RecQ homolog BLM, yet the catalytic activity of Topo III␣ is dispensable for the enhancement of this reaction (31). Furthermore, ATPase-deficient alleles of the yeast helicase gene SGS1 are functional in limiting crossovers induced by a site-specific double-strand break, a function likely requiring the ability to process HJ intermediates (32).
The high amounts of reverse gyrase needed for efficient HJ processing might raise questions about the physiological significance of this activity; however, we have shown that in S. solfataricus, reverse gyrase is rather abundant and that there is enough intracellular protein to deal with at least 100 branched structures/cell. Although ATP-independent unwinding of HJ has not been reported so far, several ssDNA-binding proteins, including human RPA and S. solfataricus SSB, show the ability to destabilize duplex DNA in an ATP-independent manner requiring P/DNA ratios ranging from 10 Replication Protein A (RPA) to 125 (SSB) (33,34). High P/DNA ratios are also required for the enzymatic resolution of HJ by the resolvase Hjc (6). In addition, so-called nucleic acid chaperones, such as rotavirus NSP2 and the HIV NS2 proteins, are also able to destabilize DNA-RNA or RNA-RNA duplexes. In these cases, 100 -500-fold molar excess of protein relative to the nucleic acid strand is typically needed to observe optimal activity (35, 36); both classes of protein work by coating the nucleic acid strand rather than through a catalytic mechanism. Interestingly, reverse gyrase has been previously suggested to have nucleic acid "chaperone" activity, which was ATP-independent (37).
The junction melting property of reverse gyrase might be involved in one or more important cell processes. It may obviously participate in HJ resolution during recombination or restart of arrested replication forks but also remove cruciform structures, thus assisting replication or helicase/translocase tracking along DNA. In addition, because reverse gyrase is able to bind RNA, 3 it might also remove secondary structures in DNA-RNA hybrids. Interestingly, cruciform extrusion is induced by DNA supercoiling (38); thus, based on its in vitro activities, reverse gyrase might both generate (changing the DNA supercoiling) and resolve (through the ATP-independent activity described here) four-way structures, possibly regulating four-way structure formation during DNA transactions.