Identification of the RecR Toprim Domain as the Binding Site for both RecF and RecO

The RecR protein forms complexes with RecF or RecO that direct the specific loading of RecA onto gapped DNA. However, the binding sites of RecF and RecO on RecR have yet to be identified. In this study, a Thermus thermophilus RecR dimer model was constructed by NMR analysis and homology modeling. NMR titration analysis suggested that the hairpin region of the helix-hairpin-helix motif in the cavity of the RecR dimer is a binding site for double-stranded DNA (dsDNA) and that the acidic cluster region of the Toprim domain is a RecO binding site. Mutations of Glu-84, Asp-88, and Glu-144 residues comprising that acidic cluster were generated. The E144A and E84A mutations decreased the binding affinity for RecO, but the D88A did not. Interestingly, the binding ability to RecF was abolished by E144A, suggesting that the region surrounding the RecR Glu-144 residue could be a binding site not only for RecO but also for RecF. Furthermore, RecR and RecF formed a 4:2 heterohexamer in solution that was unaffected by adding RecO, indicating a preference by RecR for RecF over RecO. The RecFR complex is considered to be involved in the recognition of the dsDNA-ssDNA junction, whereas RecO binds single-stranded DNA (ssDNA) and ssDNA-binding protein. Thus, the RecR Toprim domain may contribute to the RecO interaction with RecFR complexes at the dsDNA-ssDNA junction site during recombinational DNA repair mediated by the RecFOR.

Maintaining genomic integrity by repairing damaged DNA is crucial for all organisms. DNA damage can arise during normal DNA metabolism such as the introduction of mismatches during replication and the deamination of bases, or it can be caused by exposure to exogenous factors such as ultraviolet radiation, ␥-radiation, and chemical mutagens. Such DNA damage is mostly repaired by base excision repair, nucleotide excision repair, and mismatch repair pathways, based on information provided by the complementary strand in a damaged duplex. However, double-stranded DNA (dsDNA) 2 break and base lesions in single-stranded DNA (ssDNA) gap regions having no complementary strands can also arise, mainly during replication. To cope with this category of DNA damage, organisms have developed recombinational repair pathways that minimize the loss of genetic information by using homologous DNA as a template for repair.
RecFOR and/or RecBCD pathways are required to initiate homologous DNA recombination in bacteria (1). The Escherichia coli RecFOR pathway is mainly used for ssDNA gap repair, whereas the RecBCD pathway is responsible for dsDNA break repair. However, the RecFOR pathway can also repair dsDNA breaks, as has been demonstrated in recBC sbcB mutants, which are deficient in the RecBCD pathway (2,3). In the RecFOR pathway, dsDNA breaks are unwound by the RecQ helicase and processed by the RecJ 5Ј-to 3Ј-exonuclease, and the resulting 3Ј-tailed ssDNA is coated with the ssDNA-binding protein (SSB). The RecF, RecO, and RecR proteins then mediate the loading of RecA protein onto the SSB-coated ssDNA, specifically at junctions with dsDNA (4). Afterward, RecA forms a nucleoprotein filament on ssDNA that interacts with free homologous dsDNA to promote heteroduplex formation between the ssDNA and the complementary strand of the dsDNA (5)(6)(7). It was shown recently that the function of RecF, RecO, and RecR proteins is conserved in higher eukaryotes. The human tumor suppressor protein BRCA2 homologue Brh2 directs homologous recombination specifically to the dsDNA-ssDNA junction, suggesting a functional similarity between the RecFOR and BRCA2 proteins (8,9). Moreover, Rad52 proteins in human and yeast, which are functional homologues of RecO, facilitate the binding of the eukaryotic RecA homologue Rad51 to ssDNA coated with the eukaryotic SSB homologue RPA (10 -12).
E. coli RecF, RecO, and RecR proteins are well studied. The RecR protein plays a critical role in recombinational DNA repair by forming complexes with RecO or RecF (13)(14)(15). The RecOR complex facilitates RecA filament formation on SSB-coated ssDNA and prevents the dissociation of RecA from ssDNA ends (16). The RecFR complex binds dsDNA and attenuates the extension of RecA filaments beyond the region of ssDNA in the gapped DNA (14). In addition, the RecFR complex is required for reassembly of the replication holoenzyme after * This work was supported in part by grants-in-aid from the Japan Society for the Promotion of Science and Core Research for Evolutional Science and Technology (CREST), Japan Science and Technology (JST) (to T. S., Y. I., and T. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1-S3. 1 To whom correspondence should be addressed: Bio-supramolecular Structure-Function Group, RIKEN Discovery Research Inst., 1-7-29, Suehiro-cho, Tsurumi-ku, Yokohama 230-0045, Japan. Tel.: 81-45-508-7224; Fax: 81-45-508-7364; E-mail: mikawa@riken.jp.
recombinational DNA repair at replication forks (17,18). The RecF protein itself binds DNA and has a weak ATPase activity (19), both of which are enhanced by RecR binding (20,21). Furthermore, specific loading of the RecA protein onto the dsDNA-ssDNA junction is mediated by the RecF, RecO, and RecR proteins acting in concert, with no interaction between RecF and RecO detected (4). Therefore, RecR is thought to be the key mediator of RecF and RecO association at dsDNA-ssDNA junctions. The crystal structure of Deinococcus radiodurans RecR (drRecR) was recently determined (22). As predicted from sequence analysis, drRecR consists of a helix-hairpin-helix (HhH) motif, a zinc finger motif, a Toprim domain, and a Walker B motif. drRecR forms a ring-like tetramer, and it has been suggested that the HhH motifs in the central hole are important for DNA binding. However, the RecR binding sites for the RecO and RecF have not been identified. In addition, E. coli RecR has been reported to form a dimer in solution (15,20), and its structural conformation has not been determined. Previously, we reported the backbone NMR assignments of Thermus thermophilus RecR (ttRecR) and suggested that it forms a symmetric homodimer in solution (23).
Here, we have constructed a dimer model of ttRecR and have analyzed its specific interactions with DNA, RecO, and RecF. We have also discussed the role of RecR in recombinational DNA repair based on these protein-protein and protein-DNA interactions.
Preparation of ttRecR Mutants-The genes encoding the deletion mutants ttRecR  and ttRecR 1-173 were amplified by PCR and cloned into the vector pET-28a (Novagen) after sequence confirmation. The N-terminal His-tagged ttRecR  and ttRecR 1-173 were overexpressed in E. coli BL21(DE3) and purified from cell extracts with the Magtration System 6GC (PSS Bio Instruments, Inc.) and Superdex 75 10/300 GL. To prepare the ttRecR E84A, D88A, and E144A mutants, point mutations were incorporated into the ttRecR expression plasmid using the QuikChange site-directed mutagenesis kit (Stratagene) and were confirmed by sequencing. The ttRecR mutants were expressed and purified using the same methods as for wild-type ttRecR.
Ultracentrifugation Analysis-Sedimentation equilibrium experiments were performed in a Beckman Optima XL-A equipped with a four-hole rotor (An60Ti) with six channel standard cells at 15,000 rpm for ttRecR and 7,000 rpm for the ttRecFR complex for 13 h at 20°C. The ttRecR and ttRecFR complex were dissolved to 1, 2, or 4 mg/ml in buffer (25 mM Tris-HCl, pH 7.5, 150 mM NaCl). For molecular mass analysis, the data obtained were tailored to an ideal, single-component model using a solution density of 1.05 g/cm 3 and partial specific volume of 0.71 cm 3 /g for ttRecR and 0.68 cm 3 /g for the ttRecFR complex.
Size NMR Spectroscopy-NMR experiments were performed at 313 K in a triple resonance cryoprobe fitted with a Z-axis pulsed field gradient coil using a 600 MHz Bruker DRX spectrometer. Data were processed on a Linux PC using the AZARA 2.7 software package (W. Boucher, www-.bio.cam.ac.uk/azara/). All spectra were analyzed on a Linux PC with a combination of customized macro-programs in the OpenGL version of ANSIG v3.3 software (25). For processing three-dimensional NMR data, a two-dimensional maximum entropy algorithm was applied for indirect dimensions (26). Peak assignments for ttRecR spectra were obtained by transfer from previously described data (23).
Hydrogen/Deuterium Exchange and Three-dimensional NOESY NMR Experiments-2 H/ 15 N-labeled ttRecR (0.4 mM) in 20 mM Tris-HCl, pH 7.5, 150 mM KCl, and 1 mM EDTA was lyophilized and dissolved in D 2 O (99.96%). Immediately after the addition of D 2 O, protection of amide protons against deuterium was identified by analyzing a series of two-dimensional 1 H-15 N heteronuclear single quantum coherence (HSQC) spectra. The distance constraints around ttRecR residues 170 -194 were determined by analysis of the nuclear Overhauser effect from three-dimensional 15 N-separated nuclear Overhauser effect spectroscopy (NOESY)-HSQC spectrum (27).
Titration Analysis-NMR samples of 15 N-labeled ttRecR were prepared at 0.4 mM in 20 mM Tris-HCl, pH 7.2, 150 mM KCl, and 1 mM EDTA. Unlabeled protein and DNA were directly added into the sample from a concentrated stock to prevent changes in concentration and pH. For DNA titration experiments, chemical shift mapping was performed by recording the chemical shift of each peak before and after adding DNA to 1.6 mM. The chemical shift change (⌬ av ) was calculated and normalized using the following formula: [(⌬ 1 H N ) 2 ϩ (⌬ 15 N) 2 ] 1/2 , where ⌬ 1 H N and ⌬ 15 N are the chemical shift differences (Hz) along the 1 H and 15 N axes, respectively (28). For titration experiments with ttRecO, 0.05, 0.1, or 0.2 mM ttRecO were added into 0.2 mM 15 N-labeled ttRecR. The relative intensity of HSQC signals in the presence of 0.2 mM ttRecO (64 scans/free induction decay) or in its absence (16 scans/free induction decay) were plotted against the number of amino acids and mapped onto the ttRecR dimer structure.
Structure Modeling-A structural model of ttRecR dimer was generated with MODELLER version 7.7, a comparative homology modeling software (29). For homology modeling calculations, the drRecR structure was obtained from the Protein Data Bank, 1VDD (22). The integrity and quality of the models were assessed using the program PRO-CHECK, which generates Ramachandran plot of model structure. The molecular diagrams were depicted using PyMOL.
ATP Hydrolysis Assays-The ssDNA-dependent ATP hydrolysis activity of ttRecA in the presence of ttRecF pathway proteins was measured as described by Bork et al. (16) with minor modifications. The ssDNA was first incubated with SSB in 190 l of the following buffer: 50 mM Tris-HCl, pH 7.5, 50 mM KCl, 10 mM MgCl 2 , 1 mM dithiothreitol, 0.3 mM phosphoenolpyruvate, 8 unit/ml of pyruvate kinase, 13 units/ml of lactate dehydrogenase, 1 mM ATP, and 250 M NADH at 37°C for 5 min. 10 l of the ttRecO and ttRecR mixture were added to the SSB-ssDNA solution and incubated for 10 min before ttRecA was added to initiate the ssDNA-dependent ATPase reaction. The standard reaction included 20 M (with respect to nucleotides) ssDNA, 1 M ttRecA, 1 M ttSSB, 1 M ttRecR, and 1 M ttRecO. The kinetics of ATP hydrolysis was followed by measuring the absorption of NADH at 340 nm using an Ultrospec 4300 pro spectrometer (Amersham Biosciences).
Agarose Gel Retardation Assay-The DNA-ttRecFR complex formation was detected using an agarose gel assay according to the protocol described by Webb et al. (20). Each substrate DNA was incubated with wild-type or mutant ttRecR (5 M) and ttRecF (10 M) in 20 l of reaction mixture containing 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM MgCl 2 and 1 mM EDTA at 37°C for 10 min. Samples were analyzed by electrophoresis in a 3% agarose gel in TAE buffer (40 mM Tris-acetate, pH 8.0, 1 mM EDTA). DNA and DNAprotein complexes were visualized by Gel Star (Cambrex Bio Science).
CD Spectroscopy-Circular dichroism (CD) measurements were carried out on a Jasco spectropolarimeter, model J720-W. radiodurans RecR (drRecR) tetramer (Protein Data Bank, structure 1VDD). Individual subunits are colored in blue, green, red, and yellow. C, kinetics of 1 H-15 N HSQC spectra of 2 H/ 15 N-labeled ttRecR after exposure to D 2 O. All spectra were processed using identical parameters. D, mapping of detectable amide protons onto a drRecR monomer after 4 h of exposure to D 2 O. Red spheres represent amide protons that were protected from deuterium exchange. The left panel, on which seven residues are labeled, shows amide protons located on the surface of the drRecR monomer. The right panel depicts the N-terminal interface of drRecR monomers. E, the oligomeric states of ttRecR, ttRecR 1-173 , and ttRecR  proteins were determined by size-exclusion chromatography. The molecular mass was calibrated with the following size markers: albumin (67 kDa), ovalbumin (43 kDa), chymotrypsin (25 kDa), and RNaseA (14 kDa) as indicated. F, three-dimensional NOESY NMR analysis of the ttRecR C-terminal region. The anti-parallel ␤-sheet structure between ttRecR residues 163-165 and 192-194 is displayed as a topology diagram. Interstrand backbone nuclear Overhauser effects are depicted as double-headed red arrows. G, ribbon model structure of the ttRecR dimer. Domains are colored as follows: HhH motif region (red), zinc finger region (blue), Toprim domain region (green), and Walker-B motif (orange).

Prediction of ttRecR Dimer Structure by Homology Modeling-
To clarify the biophysical and structural properties of ttRecR, we initially examined its oligomeric state using analytical ultracentrifugation ( Fig.  1A). At a ttRecR concentration of 1 mg/ml (47 M), the molecular mass was calculated to be 48,448 Da, which corresponds to the mass of the ttRecR dimer. Using size-exclusion chromatography, we confirmed that ttRecR exists as a dimer from 1 M to 1 mM of the protein concentrations. Because the ttRecR concentration (1 M) was diluted (below 50 nM) during chromatography, these results suggest that ttRecR exists as a dimer at a broad range of concentrations (data not shown). In contrast, the crystallized drRecR formed a ring-like tetramer in which N-terminal HhH motifs (1-53 residues) contacted with each other and C-terminal regions (175-198 residues) swapped as described in Fig. 1B (22).
To investigate the dimerization interface of the ttRecR dimer, we used NMR to perform a proton/deuterium (H/D) exchange experiment, which can identify residues exposed on the protein surface by detecting the exchange of observable amide protons for non-observable deuterium (30,31). The panels in Fig. 1C show the 1 H-15 N HSQC spectra of the ttRecR dimer at 15 min, 4 h, and 24 h after the sample was lyophilized and dissolved in D 2 O. Each peak in the panels corresponds to an amide proton from a ttRecR amino acid residue. Amide proton signals were lost due to H/D exchange in a time-dependent manner (Fig. 1C). However, even after 4 h of exposure to D 2 O, approximately one-third of the amide protons were protected from deuterium exchange, suggesting that these protons are not on the ttRecR dimer surface. Because ttRecR and drRecR amino acid sequences share 57% identity and 74% homology and because the secondary structure of the RecR dimer that was previously determined based on our NMR assignments was identical to that of the drRecR crystal structure (supplemental Fig. S1), the monomeric structures of ttRecR and drRecR are thought to be quite similar despite the differences in their oligomeric states. Therefore, we next sought to explore the interface of the ttRecR dimer by mapping non-exchange ttRecR amide protons to corresponding drRecR regions. When these amide protons were mapped onto the drRecR monomer structure (Fig. 1D, left panel, red spheres), seven residues with protected amide protons were observed at the protein surface: Leu-17, Gly-21, Glu-41, His-55, Ile-59, Gly-83, and Gly-89. With the exception of Gly-83 and Ile-59, these residues are located near the HhH motif, making it plausible that this region is responsible for RecR dimerization (Fig.  1D, right panel).
To confirm the above idea, we prepared two ttRecR mutants, ttRecR  , in which the N-terminal HhH motif had been truncated, and ttRecR 1-173 , in which the C-terminal region corresponding to the C-terminal swapping region of drRecR had been truncated. Using circular dichroism spectroscopy (supplemental Fig. S2), we confirmed that these mutant proteins are not unfolding. Then, we examined their oligomeric state using size-exclusion chromatography (Fig. 1E). ttRecR (21.2 kDa) and ttRecR 1-173 (19.2 kDa) were eluted at positions corresponding to 48.5 and 44.6 kDa, respectively, showing that they form dimers. In contrast, ttRecR 75-194 (15.3 kDa) was eluted at a position corresponding to 18.2 kDa, showing that it exists as a monomer. These results indicate that the N-terminal region (1-74), and not the C-terminal region (174 -194), is responsible for ttRecR dimerization.
Although the ttRecR C-terminal region (174 -194) was not the interface of the ttRecR dimer, the secondary structure of the C-terminal regions was conserved in both ttRecR and drRecR (supplemental Fig.  S1). To examine the local conformation of the C-terminal region of the ttRecR dimer, we subsequently carried out 15 N-separated three-dimensional NOESY-HSQC experiments on 2 H/ 15 N-labeled ttRecR. Crosspeaks due to nuclear Overhauser effect were observed between amide protons of the 192/165, 194/165, and 164/163 residue pairs, indicating that these proton pairs are located within a region of ϳ5 Å. These results suggest that residues 192-194 of ttRecR form an anti-parallel ␤-sheet with residues 163-165 (Fig. 1F). In the case of drRecR, the corresponding ␤-sheet is formed between two RecR molecules (Fig. 1B, C-terminal  swapping region). Because the C-terminal regions of the ttRecR dimer did not form an interface, we concluded that the ttRecR C-terminal regions fold back to their own subunit and have a similar structure to that of drRecR.
We employed two drRecR monomers capable of interfacing with each other via the N-terminal HhH motif (Fig. 1B, blue and green subunits). Then, the C-terminal swapping regions of residues 173-196 were replaced by the same region of subunits, which are colored in red and yellow, respectively, in Fig. 1B. We used this structure as a template for homology modeling of the ttRecR dimer. The homology modeling was performed using the MODELLER program with the experimentally determined secondary structure of ttRecR (23) as an additional constraint. Finally, a dependable ttRecR dimer model was constructed (Fig.  1G). The root mean square deviation of C␣ between the derived ttRecR structure and the template structure was 0.38 Å. The RecR dimer model was divided into three regions (Fig. 1G): 1) Residues 1-52, termed the HhH motif region after the HhH motif (residues 5-34) it contains; 2) Residues 53-76, termed the zinc finger region after the zinc finger motif (residues 55-75) it contains; and 3) Residues 77-194, termed the Toprim domain region and comprised of the Toprim domain (residues 77-164) and the Walker-B motif (residues 165-180). Unless otherwise stated, the designations RecR, RecO, RecF, RecA, and SSB refer to the T. thermophilus proteins after this section.
Chemical Shift Mapping of DNA Binding Sites onto the RecR Dimer-We sought to determine whether the RecR dimer interacts with DNA and to identify its DNA binding sites. In this assay, we added unlabeled DNA substrates to 15 N-labeled RecR protein and compared the 1 H-15 N HSQC spectrum of RecR alone (black) with that in the presence of DNA (green). Fig. 2A shows the superimposed patterns of the two spectra for 15 N-labeled RecR in the absence or presence of 15 bp dsDNA (upper Because each 1 H-15 N cross peak in the two panels is derived from the amide proton of a RecR amino acid as described in Fig. 1C, the amino acids corresponding to the cross-peaks affected by the addition of DNA can be presumed to be involved in DNA binding. To examine the amino acid residues involved in DNA binding, we expressed the data shown in Fig. 2A as a change in chemical shift/residue (Fig. 3A, left panels). Major chemical shift changes of residues Ͼ20 Hz were observed for the Gly-21, Gly-19, and Ala-25 with 53, 41, and 20 Hz, respectively ( Fig. 2A,  upper panel, and Fig. 3A, upper left panel). In contrast, when ssDNA was added to RecR, no major changes in chemical shift were observed ( Fig.  2A, lower panel, and Fig. 3A, lower left panel).
We then color coded the amino acid residues that exhibited a change in chemical shift Ͼ20 Hz as red and those that exhibited a change between 5 and 20 Hz as yellow and mapped them onto the molecular surface of the RecR dimer (Fig. 2B). The result shows that the two Gly-19 and two Gly-21 the residue number is also shown. The cross-peaks with relative intensities below 10% are indicated in red. C, residues exhibiting significant differences in relative intensity (⌬ int ) are color coded and mapped onto the structure of the RecR dimer (⌬ int Ͻ10%, red; 10% Ͻ⌬ int Ͻ20%, yellow). Residues that could not be analyzed are in blue. The ribbon model at right was generated by a 30°rotation around the vertical axis followed by a 90°rotation around the horizontal axis. The circled areas show the proposed RecO binding sites. D, electrostatic potential at the molecular surface of the RecR dimer (left panel). The potential was calculated using the GRASP program. Positive and negative potentials are indicated in blue and red, respectively. The conserved acidic cluster region of the RecR Toprim domain is circled. The right panel shows a ribbon diagram of the RecR dimer. Glu-84, Asp-88, and Glu-144, three conserved acidic residues comprising the acidic cluster region, are labeled and depicted by space-filling model with red.
residues showing high levels of chemical shift changes gather on the interface between the two HhH motifs of the RecR dimer, forming a patch with other residues whose chemical shifts are highly affected by the addition of dsDNA. As shown in Fig. 2, B and C, residues Gly-19 and Gly-21 correspond to the hairpin region of the HhH motif, which is located at the center of the inner surface of the RecR dimer. In addition, two Tyr-178 residues in the Toprim domain are situated in the RecR inner surface, forming a cavity with the hairpin region that can accommodate dsDNA. These results suggest that RecR binds dsDNA in this cavity via Gly-19 and Gly-21 of the HhH motif of the RecR dimer (Fig. 2C, circled area).
The Toprim domains of topoisomerase or primase have an acidic cluster region that catalyzes nucleotidyl transferase activity in an Mg 2ϩ -dependent manner. Because RecR also has a Toprim domain, it has been speculated that it recognizes specific features of damaged DNA via a similar mechanism (33). Therefore, we further investigated RecR-DNA interactions using 3Ј-tailed and 5Ј-tailed DNA substrates possessing a base-paired 5Ј or 3Ј terminus at junction, respectively, in the presence or absence of Mg 2ϩ . As shown in Fig. 3, RecR did not show major chemical shifts at the Toprim domain region, either in the absence (Fig. 3A) or the presence (Fig. 3B) of the Mg 2ϩ ion. These results suggest that the role of the Toprim domain of RecR differs from that of topoisomerase or primase. The chemical shift change patterns for dsDNA, ssDNA, and dsDNA-ssDNA junc-tions all differ, but the last is particularly distinctive as the site where dsDNA and ssDNA interactions simultaneously occur.
Investigation of the RecO Binding Site on the RecR Dimer-The E. coli RecOR complex has been reported to facilitate the loading of RecA onto SSB-coated ssDNA (15). Because similar results were obtained for T. thermophilus, 3 we tried to examine the RecO binding site of the RecR dimer using NMR titration analysis. When we sequentially added 0.025, 0.025, 0.05, and 0.1 mM RecO into 0.2 mM 15 N-labeled RecR dimer, we observed a significant decay in signal intensity for all the cross-peaks of the RecR dimer (data not shown). This decay is not due to aggregation of NMR samples, because RecR and RecO did not form any precipitate after NMR measurements and no aggregates were detected in the NMR sample using size-exclusion chromatography. The most likely reason for these phenomena is the increased molecular weight of RecR due to complex formation with RecO. To compensate for the decreased sensitivity of RecR signals in the presence of 0.2 mM RecO, we repeated the scans four more times and compared the two 1 H-15 N HSQC RecR dimer spectra (Fig. 4A). Although most cross-peaks were recovered, some were still weak in the presence of RecO (Fig. 4A, right panel). Although we cannot exclude other possibilities, we speculate that these weaknesses are caused by the interaction of RecO with RecR. Therefore,   JULY 7, 2006 • VOLUME 281 • NUMBER 27 the relative intensity of each cross-peak/residue was plotted as a histogram (Fig. 4B). The relative intensity of cross-peaks for residues 84,85,89,124,144,145,146,148,153,154,168,173,174, and 175 dropped to Ͻ10% when RecO was added. All these residues are located in the Toprim domain of the RecR dimer. Then, we color coded residues exhibiting a relative intensity below 10% as red and those between 10 and 20% as yellow and mapped them onto the structure of the RecR dimer (Fig. 4C). The colored residues were observed to cluster primarily on the inner region of each Toprim domain in the RecR dimer, suggesting that this region may be responsible for RecO binding (Fig. 4C, right  panel, two circled areas). Interestingly, the proposed RecO binding sites and the acidic cluster regions of the RecR Toprim domain largely overlapped (Fig. 4, C and D, left panels).

The RecR Toprim Domain Is the RecF and RecO Binding Site
Introduction of Mutation into the RecR Acidic Cluster Region-The acidic cluster region of the RecR Toprim domain consists of three conserved acidic residues, Glu-84, Asp-88, and Glu-144 (Fig. 4D, right  panel). In the presence of an equimolar amount of RecO, the Glu-84 and Glu-144 signals were lost, though Asp-88 could not be analyzed because of the overlapping cross-peaks (data not shown). The results from these NMR titration experiments suggest that the acidic cluster region may be a binding site for RecO. To investigate the functional significance of these regions, we generated the RecR mutants E84A, D88A, and E144A by substituting each residue with alanine. The elution profiles obtained by size-exclusion chromatography for purified RecR E84A, D88A, and E144A matched that obtained for wild-type RecR, suggesting that each mutant protein folds properly and forms a dimer (data not shown).
The RecR E144A and E84A Mutations Affect Complex Formation with RecO-To clarify the functional role of the RecR Glu-84, Asp-88, and Glu-144 residues in RecO binding, the RecR E84A, D88A, and E144A mutant proteins and wild-type RecR were incubated with RecO, and the resulting RecOR complex formation was analyzed using native PAGE (Fig. 5A). RecO did not migrate into the gel because RecO has a positive charge (pI 10.7) in the running buffer, whereas RecR (pI 5.5) was observed as a single band (Fig. 5A, lanes 1 and 2). The mobility of E84A, D88A, and E144A differed slightly from that of wild-type RecR, perhaps because of the substitution of alanine for an acidic residue. When RecR and RecO were present at a molar ratio of 5:1, the band corresponding to the RecOR complex was observed (Fig. 5A, lane 3). When the molar ratio of RecR and RecO was 1:1, the band of RecR became faint, whereas the band of the RecOR complex was stained more clearly (Fig. 5A, lane  4). In contrast, when Glu-144 was mutated to alanine, the complex formation with RecO was clearly abrogated even with a molar ratio of 1:1 (Fig. 5A, lanes 11-13). The band for the RecOR complex was faintly visible when Glu-84 was mutated to alanine (Fig. 5A, lanes 5-7), and the D88A mutation had almost no effect on RecOR complex formation (Fig.  5A, lanes 8 -10). To confirm the specificity of the RecR-RecO binding observed by native PAGE, we electrophoresed the mixture of RecR and lysozyme (pI 11.8) possessing a similar isoelectric point as RecO. As a result, RecR was not affected and migrated to the original position (Fig.  5A, lane 14). These results suggest that Glu-144 and Glu-84 residues of RecR are important for RecO binding.
We next examined the effect of these mutants on RecA activity (Fig.  5B). RecA forms a filament on ssDNA that hydrolyzes ATP in a timedependent manner (Fig. 5B, gray line A). This ssDNA-dependent ATPase activity is suppressed when SSB is preincubated with ssDNA (Fig. 5B, gray line AϩSSB). However, the addition of RecO and RecR prior to addition of RecA to the reaction mixture enhances the loading of RecA onto SSB-coated ssDNA and restores its ATPase activity (Fig.  5B, gray line AϩSSBϩOR[WT]). Because RecO or RecR alone could not restore the ATPase activities (Fig. 5B, gray line AϩSSBϩO and AϩSSBϩR), the complex formation of RecR and RecO seems to be important. As expected from the results shown in Fig. 5A, the substitution of the E144A mutant for wild-type RecR in the reaction mixture clearly inhibited RecA-mediated ATPase activity, causing levels of hydrolyzed ATP to drop from 70 to 20 nmol after a 30-min reaction (Fig.  5B, line AϩSSBϩOR [WT] and AϩSSBϩOR[E144A], respectively). A lesser reduction in ATPase activity, from 70 to 58 nmol of hydrolyzed ATP, was observed for the E84A mutant. D88A did not have a suppressive effect. Because RecFOR proteins function at the nucleation step of the RecA protein, it would be important to examine the effect of RecR mutants on the initial rate of RecA-mediated ATPase activity. Therefore, we calculated the rate of ATP hydrolysis at 5 min and described it as a histogram (Fig. 5C). In accordance with the results of the complex formation by native PAGE, the mutants that could form RecOR complex tended to restore the ATPase activity of RecA. These results suggest that the Glu-144 residue is significantly important for RecO binding and that Glu-84 also participates in RecO binding.
The RecR E144A Mutation Suppresses the RecFR Complex Formation-We also used native PAGE to analyze the RecFR complex formation of the RecR acidic cluster mutants E84A, D88A, and E144A (Fig. 6A). RecF did not migrate into the gel, because RecF has a positive charge (pI 9.09) in the running buffer (Fig. 6A, lane 1). As shown in Fig. 6A, lanes 3 and 4, wild-type RecR formed a complex with RecF. The mutation of the Glu-144 residue to alanine inhibited complex formation (Fig. 6A, lanes 11-13). In contrast, the E84A and D88A mutations did not affect RecFR complex formation (Fig.  6A, lanes 5-7 and 8 -10, respectively).
We further analyzed the effect of these mutations on dsDNA binding using an agarose gel retardation assay (Fig. 6B). dsDNA binding of RecR was not observed in this assay, perhaps because of its weak affinity (Fig.  6B, lane 2), although we revealed interactions between RecR and dsDNA by NMR. RecF was capable of binding dsDNA, but most of the RecF formed aggregates that were trapped in the well of the agarose gel (Fig. 6B, lane 3). When RecF, RecR, and dsDNA were all present, RecR formed a complex with RecF and bound dsDNA (Fig. 6B, lane 4). When the E144A mutant was substituted for wild-type RecR in the reaction mixture, the band corresponding to the RecFR-dsDNA complex could not be observed and the RecF formed aggregates, just as it had when it had alone been present (Fig. 6B, compare lanes 3 and 7). Both E84A and D88A mutants allowed RecFR-dsDNA complex formation (Fig. 6B,  lanes 5 and 6, respectively), although RecF showed a slight tendency to aggregate in the presence of the E84A mutant (Fig. 6B, lane 5). These results suggest that the Glu-144 residue of RecR plays a significant role in RecFR complex formation and in dsDNA binding by the complex and that the Glu-84 residue of RecR also helps contribute to dsDNA binding. Interestingly, the tendency of the RecR mutants for RecF binding paralleled those for RecO binding.
RecR and RecF Form a 4:2 Heterohexamer-We next used size-exclusion chromatography to further investigate the complex formation of RecR, RecF, and RecO. The RecR dimer, RecF, and RecO eluted as a single peak, in general agreement with their calculated molecular masses of 42, 38, and 25 kDa, respectively (Fig. 7, A, B, D, and F). The trailing shoulder peaks of RecO may be because of an interaction between RecO and the column resin. When a mixture of RecF and RecR at a molar ratio of 1:2 was subjected to chromatography, a single peak corresponding to a molecular mass of 157 kDa was preferentially observed (Fig. 7, C and F). When analyzed by sedimentation equilibrium analysis, this fraction was found to contain a protein complex of 158 kDa (supplemental Fig. S3) and a molecular mass approximately the same as that described above.
When a mixture of RecF, RecR, and RecO at a molar ration of 1:2:1 was subjected to chromatography, two peaks of 158 and 25 kDa were observed, which molecular masses were identical to those of the RecFR complex and RecO, respectively (Fig. 7, C-E). Next, fractions 1-13 (elution volume 8 -16 ml) were analyzed using SDS-PAGE (Fig. 7G). The initial peak contained RecF and RecR (Fig. 7G, lanes 5-7), but not RecO, with RecO eluting at its original position (Fig. 7G, fractions 11-13). The molecular ratio of RecR to RecF in the RecFR complex was estimated using densitometric analysis in which we compared the densities of two of the bands in SDS-PAGE (Fig. 7G, lanes 5-7) with those of known amounts of RecR and RecF. The molar ratio of RecR to RecF was found to be 2:1. On the basis of these findings, we propose that RecR and RecF form a 4:2 heterohexamer (160 kDa) that is not dissociated by the addition of RecO to the solution. When subjected together to size-exclusion chromatography, RecR and RecO were eluted separately at their original position. This result suggests that RecO and RecR are unable to form or maintain the complex through the elution process (data not shown).

DISCUSSION
The Toprim domain of RecR was found to be important for binding RecO and RecF. Previously, the acidic cluster regions of the Toprim domain in topoisomerase or primase were reported to be the active center for nucleotidyl transferase activity, which needs the interaction of a conserved glutamate and two aspartate residues with Mg 2ϩ (32)(33)(34). However, the two aspartate residues are not conserved in the Toprim domain of RecR (35). Recently, structural analysis of drRecR showed that the RecR Toprim domain commonly has a RecR-specific acidic cluster region that consists of three acidic residues (Glu-84, Asp-88, and Glu-144 in the case of T. thermophilus RecR) (22). In this study, we used NMR titration analysis to reveal that the Toprim domain of RecR is not involved in DNA binding through Mg 2ϩ . Instead, we revealed that the region was involved for binding RecO and RecF. When the Glu-144 residue of RecR was mutated to the uncharged residue alanine, the ability of RecR to bind RecO and RecF was severely diminished, showing that Glu-144 plays a significant role in the formation of both the RecOR and RecFR complexes. We also demonstrated that Glu-144 is important for RecOR or RecFR complex functions, as described in Figs. 5 and 6. Because Glu-144 is the most externally positioned of the three acidic residues and located at the greatest distance from the HhH motif of the dsDNA binding site (Fig. 4D, right panel), its location may be favorable for interaction with RecF and RecO.
RecF and RecO may compete for their overlapping RecR binding sites. Indeed, we found that the RecOR-enhanced binding of RecA to SSB-coated linear ssDNA (as shown in Fig. 5B) was reduced by the addition of RecF in a concentration-dependent manner (1-10 M RecF, data not shown). Our results are partly supported by the work of Bork et al. (16), which showed that the E. coli RecF protein competes with RecO for RecR protein association with the RecA filament by analyzing its ATPase activity in the presence of SSB-coated ssDNA.
The manner in which RecR proteins bind DNA has not been well described, although the drRecR tetramer has been reported to bind both dsDNA and ssDNA (22). The dsDNA binding of the RecR dimer could not be observed by a gel retardation analysis (Fig. 6B, lane 2) as shown previously in similar assay using E. coli RecR dimer (20). However, NMR titration analysis showed that the hairpin region of the HhH motif in the RecR dimer is responsible for dsDNA binding. The dissociation constant for the binding of RecR to dsDNA was considered to be in the mM range, because the chemical shift changes of both the Gly-19 and Gly-21 residues of RecR (0.4 mM) were not saturated by the addition of 1.6 mM dsDNA (data not shown). Although the HhH motif is a positively charged region and a nonspecific interaction between the motif and a negatively charged dsDNA might occur, we believe the binding to be specific for the following reasons. 1) We showed that two neutral residues of Gly-19 and Gly-21 are most affected. 2) Our results accord well with findings that two glycine residues in the hairpin region of the HhH motif were critical for interaction with dsDNA (36,37). Moreover, glycine-isoleucine-glycine in the hairpin region has been suggested to bind to dsDNA through hydrophobic interactions with the bases in the groves (38).
3) The hairpin region of the HhH motif is highly conserved among known RecRs, including ttRecR, drRecR, and E. coli RecR (supplemental Fig. S1). 4) The mutational work of drRecR also shows that the HhH motif is critical for DNA binding (22). On the basis of these findings, we concluded that the hairpin region of the HhH motif is an important region for dsDNA binding of RecR. It should be noted, however, that dsDNA binding of RecR could not be observed at low protein concentrations. The dsDNA binding by RecR alone may not be significant in vivo. Because RecR is able to bind dsDNA tightly by forming a complex with RecF as described in this study (Fig. 6B, lane 4) and in a previous report in E. coli (14,20), the HhH motif of RecR may be important when it exists as the RecFR complex.
RecR forms a stable dimer in solution, which is supported by the dimerization property of E. coli RecR (15,20). In contrast, ultracentrifugation analysis has revealed that both drRecR and Helicobacter pylori RecR form a tetramer in solution (22). Moreover, crystallized drRecR forms a ring-like structure involving both the N-and C-terminal interfaces. The structural properties of the C-terminal regions (see Fig. 1), i.e. their capacity for interacting with each other to form tetramers, may determine why some RecR proteins form dimers and others tetramers. Given that the drRecR tetramer binds dsDNA more tightly than does the RecR dimer, our finding that RecR and RecF spontaneously form a 4:2 heterohexamer complex that stably binds dsDNA suggests that, like drRecR, the RecR dimer functions as a ring-like tetramer in the presence of RecF. Although the conformation of the four RecR monomers in the RecFR heterohexamer is not known at present, RecR proteins may have the ability to form a ring-like structure that enables them to bind dsDNA tightly.
RecR preferentially formed a RecFR complex in the presence of RecF, RecO, and RecR (Fig. 7). The RecF or RecFR complex has been reported to bind dsDNA-ssDNA junctions (4,39). Although it has been suggested that RecO (or the RecOR complex) binds the RecFR complex (or RecF) via RecR at the dsDNA-ssDNA junction site (4), the process of RecFOR assembly at the dsDNA-ssDNA junction is not yet understood in detail. In this study, RecR and RecF formed a 4:2 heterohexamer in solution that was unaffected by adding RecO, indicating a preference by RecR for RecF over RecO. RecO is thought to be localized at ssDNA regions that are coated with SSB, because E. coli RecO prefers ssDNA to dsDNA (40) and also has a high affinity for SSB (15). These properties were conserved in the T. thermophi- lus RecO. 3 Therefore, the RecO localized at the ssDNA region may interact with the Toprim domain of RecR that forms a complex with RecF at dsDNA-ssDNA junctions to ultimately form the RecFOR complex (or the RecOR complex) at the site, as described in Fig. 8.