We obtained the two different PfuHjm crystals (Forms 1 and 2) in the nucleotide free-state, and determined their structures at 2.4 Å (Form 1) and 2.0 Å (Form 2) resolutions, respectively (Table 1). In the Form 1 crystal, the C-terminal 60 residues, about two thirds of the C-terminal domain, are missing in the final model, presumably because of structural disorder. On the other hand, in the Form 2, we could build the model of almost the entire molecule except for the C-terminal twenty residues. Although we could not obtain cocrystals with nucleotides, we soaked ATP analogs into the crystals, and successfully determined the nucleotide complex structures. The structures of the two apo-forms are quite similar, with a root-mean square deviation (rmsd) of 1.05 Å for the corresponding 651 C_α atoms. Nucleotide binding to the protein also causes no large structural change; The overall rmsd value between the apo- and ATPγS-soakd states is 0.30 Å in Form 1, and similarly that between the apo- and AMPPCP-bound structures is 0.40 Å in Form 2.

PfuHjm folds into five domains (domains 1 to 5) with dimensions of approximately 70 × 50 × 30 Å. The protein possesses a concave surface on the front-view side and a hole (about 10 Å diameter) at the center of the molecule (Figure 1a). The two N-terminal domains 1 (residues 1–197) and 2 (residues 198–399) form typical helicase domains with a cleft between them, as commonly observed in the helicase superfamily. The seven conserved helicase sequence motifs 20 21 line the cleft walls in an arrangement similar to that observed in other helicase structures 22 . Domain 1 contains the Walker A and B motifs that are widely conserved in nucleotide triphosphate hydrolases. The Form 1 ATPγS-soaked crystal exhibited electron density corresponding to the hydrolyzed product ADP, rather than the soaked ATPγS, in the nucleotide-binding pocket (Figure 1b). On the other hand, clear electron density for the bound triphosphate was observed in the Form 2 AMPPCP-soaked crystal (Figure 1c). Regardless of the crystal form and the nucleotide binding, the structures of the Walker-A motif and the surrounding region are very similar to each other.

The ATP-analog AMPPCP is bound to the binding pocket of domain 1, and it participates in several key interactions with the protein: The adenine moiety is surrounded mainly by four hydrophobic residues, Ile21, Phe24, Tyr25, and Leu54, and the two nitrogen atoms hydrogen bond with Gln62, in a bidentate manner. The triphosphate is wrapped up by the Walker A motif (Thr48 to Thr53), which contains the invariant lysine residue (Lys52). The γ-phosphate faces the two acidic residues in the Walker B motif (Asp145 and Glu146). Interestingly, in the Hjm structures, the conformations around the nucleotide binding sites are almost the same, including the side-chain conformations, independently of the nucleotide-binding states. Figure 1d shows a close-up view around the nucleotide binding sites of the three family members. A comparison of the nucleotide binding pocket of Hjm with those in the two Hel308 helicases revealed that the pocket of A. fulgidus Hel308 is partly disrupted: In the superimposed structure, the three amino acids of the A. fulgidus Hel308 sterically clash with the ATP-analog molecule bound to Hjm (Ile26 with the adenine moiety, and Ala50 and Ala51 with the β-phosphate), indicating that the A. fulgidus Hel308 segments should undergo a structural change upon nucleotide binding. On the other hand, the S. solfataricus enzyme exhibits a highly similar structure around the nucleotide binding site, and therefore seems to be ready to bind the nucleotide.

The C-terminal region is divided into three domains (domains 3–5). Domain 3 (residues 400–492) has a structural segment similar to the winged-helix (WH) motif. This motif is often used for the recognition and binding of double-stranded DNA (ds DNA) 23 . In the case of Hjm, however, it is unclear whether this segment is important for DNA binding, because the electrostatic potential surface has few notably positive areas in this region. Consistently, in the structure of the A. fulgidus Hel308-DNA complex, the corresponding segment was not involved in DNA binding. Domain 4 (residues 492–642) folds into a seven α-helix bundle structure. This fold seems to be unique within this helicase family, as thus far.

The C-terminal domain 5 (residues 643–720) is the smallest and contains the HhH motif. The HhH motif is present in many DNA metabolizing proteins that recognize ssDNA 24 . Indeed, the corresponding element in the A. fulgidus Hel308 helicase interacts with DNA 18 . In the case of the S. solfataricus and M. thermautotrophicus Hel308, this domain exhibited a regulatory function to tune the processivity of its helicase activity as a molecular brake 19 25 . PfuHjm possesses a PCNA-interacting protein (PIP) box at the C-terminus, which is required for the physical interaction with PCNA, and the unwinding activity of PfuHjm for the fork-structured DNA is enhanced by PCNA in vitro 16 . However, the C-terminal segment was invisible in both the Form 1 and Form 2 crystals, suggesting that this segment is highly mobile.

Based on the A. fulgidus Hel308-DNA crystal structure, a DNA unwinding mechanism has been proposed for this helicase 18 . In this mechanism, the central helix of domain 4 acts as the "ratchet" formed by two key amino acid residues (Arg592 and Trp599 of A. fulgidus Hel308). These residues form stacking interactions on base moieties of the DNA, thus pushing out 3' tails of unwound DNAs from the tunnel, formed by the domains 1, 3, and 4, toward an exit near domain 5. An interesting feature is that the ratchet helix is located near the conserved helicase motifs, Ia and Ib of domain 1, and IV of domain 2, which are associated with ATPase activity.

PfuHjm shares 30% and 37% amino acid identity with the A. fulgidus and S. solfataricus Hel308 helicases, respectively (see Additional file 1: Multiple sequence alignment), and the overall folding of PfuHjm is very similar to those of those proteins throughout the molecule. PfuHjm was fitted to the protein of the A. fulgidus Hel308-DNA complex, with an rmsd of 2.06 Å, for the corresponding 561 C_a atoms, while rmsd values for individual fitting of each domain is 1.04, 0.99, 0.90, 1.13, and 1.22 Å, for domains 1 to 5, respectively. This indicates that the spatial arrangements of the five domains significantly differ between PfuHjm and the DNA-bound A. fulgidus Hel308, and hence each domain of PfuHjm was separately fitted to the corresponding domain of A. fulgidus Hel308.

First, we superimposed PfuHjm on A. fulgidus Hel308 only using domain 2, which recognizes the branch points of the substrate DNA, and then the other four domains were further moved separately to the best fitted positions. The shifts of the second fitting could correspond to the movements of each domain upon DNA binding. The second shifts, defined as the center of mass, were 0.91, 0.77, 1.01, and 1.03 Å for domains 1, 3, 4, and 5, respectively (Figure 2a). Therefore, as suggested previously for A. fulgidus Hel308 18 , the domain rearrangement of PfuHjm should be small upon branched DNA processing. According to the scheme of the helicase-DNA recognition revealed from the A. fulgidus Hel308-DNA complex crystal structure, we visually inspected which amino acids interact with DNA in the fitted PfuHjm-DNA binding model, and found out that such amino acids and their locations are substantially conserved (see Additional file 1: Multiple sequence alignment).

Furthermore, the prominent β-hairpin loop in domain 2, which melts the duplex DNA in A. fulgidus Hel308, is shorter by one residue in PfuHjm. However, the residues that contact the DNA in the A. fulgidus Hel308-DNA complex are substantially conserved in the sequence, and thus the protein-DNA interactions at this β-hairpin loop could be quite similar between the two enzymes. In the PfuHjm crystal structures, several segments exhibit high temperature factors, and the side chain atoms could not be assigned in the electron density maps. Among these, segments 332–335 and 347–351 are located on the possible DNA interacting surfaces. We presume that their conformational flexibility would be important for the continuous DNA translocating and unwinding reaction, which is coupled with ATP binding/hydrolysis.

The Form1 crystals were obtained using ammonium sulfate as a precipitant, and it was found that five sulfate ions were bound to the protein. Notably, all of the sulfate ions lie on possible DNA binding surfaces (Figure 2b). Similarly, it was reported that phosphate ions are bound in the A. fulgidus and S. solfataricus Hel308 structures 18 19 . Collectively, these results indicate that the sulfate/phosphate ions mimic DNA backbone phosphates. For instance, a sulfate ion is strongly bound to Arg306 and Arg309 in domain 2 of PfuHjm. Two point mutations (R306A or R309A) in PfuHjm significantly decreased the DNA binding ability (Fujikane and Ishino, unpublished data).

Taken together, the PfuHjm structures strongly suggest that this helicase recognize branched DNAs in a similar manner to that in the A. fulgidus Hel308-DNA complex. Therefore, it is also likely that the DNA unwinding mechanism is conserved between them.

We were not successful in obtaining PfuHjm DNA complex crystals. Therefore, we used single particle electron microscopy to analyze the structure of a PfuHjm in complex with a 3' overhang DNA, and indeed, a 3D image was obtained at 23Å resolution (Figure 3). The complex has a main body with a protruded portion. The main body corresponds to PfuHjm, as the atomic structure of PfuHjm fits well into the electron density isosurface. Consequently, the protruded portion should correspond to the ds DNA lying outside of the protein molecule. It should be noted that the orientation of the ds DNA is different between the PfuHjm-DNA EM structure and the A. fulgidus Hel308-DNA crystal structure. The ds DNA in our complex is tilted by about 70 degrees, as compared to that in the A. fulgidus enzyme complex. The sequence and the secondary structure of DNA used in our study is slightly different from that of the Hel308 complex. However, it is unlikely that this caused the difference in DNA orientations. In fact, the double-stranded region of the DNA substrate, in both of the protein-DNA complexes, weakly interacts with the helicases through minor contacts. For instance, our previous electrophoresis mobility shift assay (EMSA) indicated that the apparent dissociation constant of PfuHjm against ds DNA was about 5 times higher than those against single-stranded or Y-shaped DNA 16 . Thus, the ds DNA may have happened to be fixed at the distinct positions, because crystallographic and EM analyses target different states of protein or protein-DNA complexes.

Apart from the Hel308 helicases, Hjm is closest to a bacterial RecQ helicase (1oywA) 26 in its N-terminal region (domains 1 and 2). On the other hand, the C-terminal halves of Hjm and the archaeal Hel308s adopt unique folds. However, we could detect local fold similarity of domain 3 to transcriptional factors (Arg repressor, 1aoy 27 , and transcription initiation factor IIF, 1onvA 28 ). Likewise domain 4 shares local similarity to the signal recognition particle protein (1hq1A) 29 , while the C-terminal domain 5 shares similarity to DNA excision repair protein (2a1jB) 30 and HJ DNA binding protein (1d8lA) 31 .

The Hjm structure appears to be composed of a unique combination of the domains used for DNA/RNA-binding or processing. The overall structural comparison among the SF2 helicases is shown in Figure 4b. When these structures are aligned using the well-conserved helicase domains, the configurations of the other domains are quite variable. This indicates that these enzymes share the two helicase domains that are fundamental for the helicase activity, while the structural and spatial arrangements of the other domains are designed to correspond to their individual DNA unwinding mechanisms and substrate specificities.

The DNA metabolizing proteins from archaea are both structurally and functionally similar to those from eukaryote, and therefore, the structures of archaeal proteins are useful to understand the complicated DNA transaction mechanisms in eukaryotes. In this study, we showed that the 3D structure of PfuHjm is similar to those of the A. fulgidus and S. solfataricus Hel308 helicases, implying that these structural features could be extended to this helicase family, which includes the human PolΘ and Hel308 and Drosophila Mus308 proteins. Human PolΘ is A-family DNA polymerase and works in translesion DNA synthesis 32 33 . This protein is unique because it has both helicase and DNA polymerase domains on a single polypeptide chain. A homology model of the helicase domain of human PolΘ, which was built using the program MOE (Ryoka Systems Inc.), is highly similar to the PfuHjm and Hel308 helicases (Fig. 4c; also see Additional file 2: Homology model of the human DNA polymeraseΘ helicase domain). The model seems to be reasonable in that, as in the case of PfuHjm, the putative DNA-interacting segments are both sequentially and spatially conserved in the human PolΘ helicase domain. In this domain, PolΘ contains seventeen cysteine residues that are not present in PfuHjm. The homology model indicates that twelve cysteine residues are exposed to the solvent, and that two of them form a disulfide linkage in a region corresponding to domain 2 of PfuHjm. Furthermore, several cysteine residues are conserved in PolΘ helicase domains in eukaryotes other than human (see Additional file 1: Multiple sequence alignment). It is tempting to speculate that these cysteines are used for sensing oxidative stress, because a genetic analysis showed that vertebrate PolΘ gene-deficient cells exhibited hypersensitivity to oxidative base damage induced by H₂O₂ 34 .

The recombinant PfuHjm protein was produced and purified as described previously 15 . The gene encoding the protein was cloned into the pET21d vector, and the constructed plasmid, pHJM100, was introduced into E. coli BL21 codonPlusTM (DE3)-RIL cells (Stratagene). The transformed cells were grown in LB medium containing 50 μg/mL ampicillin and 34 μg/mL chloramphenicol at 37°C to an OD₆₀₀ of 0.35, and then protein expression was induced by 1 mM IPTG for 5 h. The cells were harvested and disrupted by sonication in buffer A (50 mM Tris-HCl, pH 8.0, 0.5 M NaCl, 0.5 mM EDTA, 1 mM DTT, and 10% glycerol). The soluble fraction was collected by centrifugation (12 000 g, 15 min) and then was incubated at 80°C for 20 min. Polyethylenimine was added to the supernatant to a final concentration of 0.15% (v/v), to remove the nucleic acids. The soluble fraction was clarified by centrifugation and precipitated by 80%-saturated ammonium sulfate. The proteins were resuspended in buffer B (50 mM Tris-HCl, pH 8.0, 1.25 M (NH₄)₂SO₄, 0.5 mM EDTA, 1 mM DTT, and 10% glycerol), loaded onto a hydrophobic column (HiTrap Butyl, GE Healthcare), and eluted with H₂O. The pooled fraction was dialyzed against buffer C (10 mM K-phosphate, 7 mM β-mercaptoethanol, 0.01 mM CaCl₂, and 10% glycerol) and was loaded onto a CHT-II hydroxyapatite column (Bio-Rad), which was developed with a linear gradient of 0.01 to 1 M K-phosphate. The fraction pool containing the PfuHjm protein of interest was subsequently dialyzed against buffer D (50 mM Tris-HCl, pH 8.0, 0.5 mM EDTA, 1 mM DTT, and 10% glycerol), and was loaded onto an anion exchange column (MonoQ 5/5, GE Healthcare). The column was developed with a 0 to 1 M NaCl linear gradient, and the purified protein was eluted at 0.32–0.37 M NaCl. The purified protein was concentrated to 8 mg/ml for crystallization. The calculated extinction coefficient of 101,190 M^-1cm^-1 at 280 nm was used for the determination of the protein concentration. To prepare the selenomethionine (SeMet) derivative of PfuHjm, pHJM100 was transformed into the methionine auxotrophic strain E. coli BL21(DE3) Codonplus RIL-X (Stratagene). The SeMet derivative was expressed by IPTG induction in a minimal medium containing seleno-L-methionine at a final concentration of 25 μg/ml, and was purified using the same procedure as for the wild type protein.

PfuHjm was crystallized by the hanging drop vapor diffusion technique with the micro-seeding at 293 K. The first diffraction quality crystals (Form 1) were obtained using a reservoir containing 100 mM citrate (pH 5.0) and 1.6 M ammonium sulfate. The crystals belonged to the space group C2, with unit cell constants a = 118.6 Å, b = 85.0 Å, c = 95.0 Å, and β = 121.0°, and contained one Hjm molecule per asymmetric unit. The SeMet protein was crystallized under the same conditions as for the wild-type Hjm. Tantalum (Ta₆Br₁₄)- and platinum (K₂PtCl₄)- derivatized crystals were prepared by soaking. ATPγS-soaked crystals were prepared by soaking native crystals in reservoir solution containing 1 mM ATPγS. Crystals were harvested with the reservoir solution containing 20% (v/v) glycerol for X-ray diffraction data collection at 100 K. Data sets of the native crystal and a Pt-derivative were collected on BL-6B of the Photon Factory, Tsukuba, Japan. The Ta derivative data were collected on BL40-B2, and those for the ATPγS-soaked crystal and the Se-Met derivative were obtained on BL41-XU of SPring-8 (Harima, Japan). Data sets were processed by DENZO/SCALEPACK or the HKL2000 package 35 .

The structure was determined by the MIRAS method. All the heavy atom sites were located on isomorphous Patterson maps, and the heavy atom parameters were refined by the program SHARP 36 . The experimental phases were improved by density modification techniques, with the programs DM and SOLOMON in the CCP4 suite 37 . The initial atomic model was built, based on this modified map, with the program O 38 . About 70% of the amino acid residues were located using the modified map. The combination of the experimental MIRAS phases with those calculated from a partial model further improved the quality of the electron density map, leading to the construction of the other parts. Crystallographic refinement was performed with the program CNS 39 . The final model of the Form 1 apo crystal consisted of 660 amino acid residues, except for the disordered region (mainly the C-terminal 60 residues). The structure of the ATPγS-soaked crystal was determined by using the apo-form as the initial model, and was refined to convergence. Careful inspection of the electron density maps revealed that the bound nucleotide was the hydrolyzed product ADP, rather than the soaked ATP-analog.

The second crystals (Form 2) were obtained under different crystallization conditions, using a reservoir solution containing 80 mM Tris-HCl (pH 8.5), 160 mM CaCl₂, and 11% (w/v) PEG4000. The micro-seeding technique was also used to obtain diffraction quality crystals. These crystals also belonged the space group C2, as did Form 1, but had significantly different unit cell constants (a = 122.3 Å, b = 81.2 Å, c = 85.2 Å, and β = 111.9°), suggesting distinct crystal packing. The complex with AMPPCP was prepared by soaking the Form 2 apo crystals into reservoir solution containing 0.5 mM AMPPCP. Diffraction data sets for the Form 2 apo crystal were collected at 100 K on BL38-B1 of SPring-8, and those for the AMPPCP complex crystal were collected at BL-6B of the Photon Factory. These structures were determined by molecular replacement, using the program CNS and the Form 1 apo structure as a probe. The Form 2 structures are better ordered in the crystals, and the almost the entire molecule, except for the C-terminal 20 residues with the PIP-box sequence, was visible in the electron density map. Crystallographic refinements were reiterated to obtain satisfactory convergence. All of the crystallographic statistics are summarized in Table 1. The atomic coordinates have been deposited in the Protein Data Bank, under the accession codes 2ZJ2, 2ZJ5, 2ZJ8, and 2ZJA, for the Form 1 apo, Form 1 ADP complex, Form 2 apo, and Form 2 AMPPCP complex, respectively.

The 3' overhang DNA was prepared by forming a hairpin structure from a synthetic oligonucleotide (5'- AGCACTGCTATTCCCTAGCAGTGCTAGATGCACGAC-3'). The Hjm protein was mixed with DNA (1:1 protein/DNA ratio) and was incubated in a buffer containing 50 mM Tris-HCl pH8.0, 0.15 M NaCl, 0.5 mM EDTA, 1 mM DTT, and 10% glycerol, at room temperature for 20 min. The complex was purified by gel filtration chromatography on a Superdex 200 PC 3.2/30 column (GE Healthcare), using a SMART system (GE Healthcare). An aliquot of the complex solution was applied to a carbon support film, and was negatively stained with 2% uranyl acetate. The specimens were examined with a JEM 1010 electron microscope (JEOL), operated at an accelerating voltage of 100 kV. Images were recorded by BioScan CCD camera (Gatan). A minimum dose system (MDS) was used to reduce the electron radiation damage of the sample. The step size of a pixel of the image was calibrated to be 5.1 Å, using TMV as a reference sample. Image processing was performed using the software packages EMAN 40 and IMAGIC 41 . Individual particle images were boxed out, using the GUI-based program boxer in EMAN. The class average images of the Hjm-DNA complexes were obtained by several cycles of a multireference alignment and classification procedure for image sets. The programs in IMAGIC were used to calculate these class averages. The initial 3D map was obtained by common-line method and subsequent iterative refinement was performed using REFINE routine of EMAN. The resolution of the 3D map was estimated by the 0.5 criterion of the Fourier shell correlation. The visualization of the 3D map and fitting of the crystal structure into the map were performed, using the Chimera software 42 .

The homology model of the helicase-like domain of human DNA polymerase Θ (UniProt code Q6VMB5) was constructed by using the Homology module of the MOE application (Ryouka Systems Inc.), which was based on the methods of Levitt 43 and Fechteler et al. 44 .