A Conserved Nuclease Domain in the Archaeal Holliday Junction Resolving Enzyme Hjc*

Holliday junction resolving enzymes are ubiquitous proteins that function in the pathway of homologous recombination, catalyzing the rearrangement and repair of DNA. They are metal ion-dependent endonucleases with strong structural specificity for branched DNA species. Whereas the eukaryotic nuclear enzyme remains unknown, an archaeal Holliday junction resolving enzyme, Hjc, has recently been identified. We demonstrate that Hjc manipulates the global structure of the Holliday junction into a 2-fold symmetric X shape, with local disruption of base pairing around the point of cleavage that occurs in a region of duplex DNA 3′ to the point of strand exchange. Primary and secondary structural analysis reveals the presence of a conserved catalytic metal ion binding domain in Hjc that has been identified previously in several restriction enzymes. The roles of catalytic residues conserved within this domain have been confirmed by site-directed mutagenesis. This is the first example of this domain in an archaeal enzyme of known function as well as the first in a Holliday junction resolving enzyme.

Holliday junction resolving enzymes play a role in the pathway of homologous recombination, recognizing and cleaving the four-way DNA junctions that arise from strand exchange between homologous duplex DNA species. Junction resolving enzymes are ubiquitous in nature. These proteins have been identified in Eubacteria (RuvC (1,2) and RusA (3)), bacteriophage (T4 endonuclease VII (4) and T7 endonuclease I (5, 6)), fungal mitochondria (Cce1 (7)), and most recently Archaea (Hjc and Hje (8,9)). Whereas activities have been detected in nuclear extracts from yeast (10) and mammalian cells (11,12), the relevant genes have yet to be identified. Resolving enzymes function as dimers, resolving the four-way DNA junction by the introduction of paired nicks in opposing strands with a magnesium-dependent endonuclease activity. Despite these functional similarities, the junction resolving enzymes are structurally diverse with no detectable sequence similarity among any of the known examples. Structural studies have highlighted this diversity because the crystal structures of RuvC (13), T4 endonuclease VII (14), and T7 endonuclease I 1 have radically different folds. These observations have led to the suggestion that resolving enzymes have arisen several times during the course of evolution, perhaps by recruitment of nucleases with other cellular roles. This is almost certainly the case for the eubacterial enzyme RuvC, which shares a fold and metal binding site with members of the RNase HI superfamily (13,15).
The Archaea constitute a third domain of life that is distinct from both the Eubacteria and the Eucarya. Whereas the Archaea resemble their fellow prokaryotes in most respects, they share many similarities with the Eucarya in the information processing pathways including DNA replication, transcription, and translation (reviewed in Ref. 16), and the archaeal processes constitute a useful model system for the much more complex eucaryal equivalents. We are investigating the pathway of homologous recombination in the Archaea and have detected two Holliday junction resolving enzymes, Hje and Hjc, in the Crenarchaeote Sulfolobus solfataricus (8,17). The gene for Hjc has been identified and is conserved in all Archaea for which extensive genome sequence is available (8,9). We have cloned S. solfataricus Hjc and overexpressed the recombinant enzyme in Escherichia coli (8). Here we demonstrate manipulation of the global structure of the four-way junction substrate by Hjc and disruption of base stacking near the cleavage site. We report the identification of a structural motif in Hjc that is shared by a diverse group of nucleases and constitutes a binding site for the catalytic metal ions. Parallels between Holliday junction resolving enzymes and restriction enzymes that were first pointed out at a biochemical level for the yeast mitochondrial enzyme Cce1 (18) now achieve a firm structural basis in the archaeal Hjc enzyme.

Oligonucleotide Synthesis and Assembly of DNA Junctions
Oligonucleotides were synthesized and four-way DNA junctions assembled as described previously (18) using the sequences described in the following paragraphs.
Junction 3-This fixed four-way junction was prepared with arms of 15 or 25 bp 2 as described previously (45). Comparative gel electrophoresis experiments utilized a version of junction 3 with four arms of 60 bp in length. Six forms of junction 3 with two long and two short arms were derived from this junction as described previously (46). Junction 1-This is a fixed junction with 20 bp in each arm assembled from four oligonucleotides, each of 40 nucleotides in length as described previously (47).

Expression and Purification of Recombinant Hjc Protein
Recombinant Hjc protein was expressed in E. coli strain BL21 (DE3) CodonPlus RIL (Stratagene), and the protein was purified as described previously (8). In brief, Hjc was purified by chromatography on an SP-Sepharose high performance 26/10 column (Hi-Load, Amersham Pharmacia Biotech) equilibrated with Buffer A (50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 1 mM dithiothreitol). A 500-ml linear gradient of 0 -1000 mM NaCl was used to elute cationic proteins. Fractions corresponding to a distinct absorbance peak were analyzed by SDS-polyacrylamide gel electrophoresis, pooled, concentrated, and loaded onto a 26/70 gel filtration column (Superdex 200 Hi-Load, Amersham Pharmacia Biotech) and developed with Buffer A containing 300 mM NaCl. Active fractions were pooled and shown to contain essentially homogeneous Hjc protein. This protein was used for all subsequent analyses.

Site-directed Mutagenesis
Site-directed mutagenesis of Hjc was carried out in the plasmid pUC119 using the QuikChange method (Stratagene). After mutagenesis, DNA sequencing was used to confirm that no spurious mutations had been introduced. The Hjc mutants were subcloned into pET19b and were expressed and purified similar to the wild-type enzyme.

Comparative Gel Electrophoretic Retardation Analysis of Hjc-DNA Junction Interactions
Samples of purified Hjc protein (1 M) were incubated with radioactive 5Ј-32 P-labeled four-way DNA junction 3 in binding buffer (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 0.1 mg/ml bovine serum albumin, and 0.2 mg/ml calf thymus duplex competitor DNA) including either 1 mM EDTA or 0.1 mM MgCl 2 in a 10-l total volume for 5 min at 20°C, prior to addition of loading buffer (0.25% bromphenol blue, 0.25% xylene cyanol FF, 15% Ficoll type 400) at a dilution of 1:6 (v/v). Samples were loaded onto 5% polyacrylamide gels and electrophoresed in Tris-borate-EDTA buffer or Tris-borate and 0.1 mM MgCl 2 with buffer recirculation. After electrophoresis, gels were dried on Whatman 3MM paper and exposed to x-ray film for documentation.

Equilibrium Binding of Wild-type and Mutant Hjc Proteins
Binding affinity was measured by gel electrophoretic retardation analysis using radioactive 5Ј-32 P-labeled junction 1 in EDTA binding buffer as described previously (8).

Assay of Hjc Activity
Assays were carried out using 1 M purified recombinant Hjc protein in reaction buffer (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 15 mM MgCl 2 ) using 80 nM 5Ј-32 P-labeled junction as a substrate. Calf thymus DNA (0.2 mg/ml) was added as a competitor to minimize nonspecific endonuclease activity and DNA-binding proteins. Reactions were initiated by the addition of magnesium to the assay mix in a 5-l total volume and incubated at 60°C. At set time points aliquots were removed, and the reactions were stopped by the addition of 4 l of formamide/EDTA loading mix with heating to 95°C. Products were analyzed by denaturing gel electrophoresis and phosphor imaging as described previously (48).

Modification of DNA by Potassium Permanganate
Radioactive 5Ј-32 P-labeled four-way DNA junction Z28 was incubated at room temperature for 5 min in the presence or absence of 1 M Hjc protein in binding buffer. The total reaction volume was 20 l. Reactions were initiated by the addition of 2 l of freshly dissolved 25 mM KMnO 4 and were stopped after 1 min with the addition of 1.5 l of ␤-mercaptoethanol. After ethanol precipitation, DNA samples were reacted with 100 l of 1 M piperidine for 30 min at 95°C. Piperidine was removed by vacuum desiccation, and the pellets were washed three times with 30 l of water and were vacuum desiccated after each water addition. The dried samples were resuspended in formamide loading mix and analyzed on 15% denaturing polyacrylamide gels.

Manipulation of the Global Structure of the Four-way DNA Junction by Hjc-The
Holliday junction assumes a conformation known as the "stacked X structure" in the presence of divalent metal ions, folding by coaxial stacking of pairs of helices in an antiparallel 2-fold symmetric cross (reviewed in Ref. 19). This structure has been studied by a wide variety of techniques and has recently been confirmed by x-ray crystallography (20,21). Comparative gel electrophoresis has proven a powerful technique for the analysis of the global configuration of the four-way junction, both in solution and complexed with proteins. Using this technique, all Holliday junction binding proteins studied to date have been shown to alter the conformation of the Holliday junction on binding (reviewed in Ref. 22).
We examined the effect of Hjc binding on the global conformation of the four-way junction J3 using comparative gel electrophoresis. In EDTA, the free junction species adopted the four-slow/two-fast pattern characteristic of the 4-fold symmetric junction with arms related by 90°. In the protein-DNA complex, this pattern was altered significantly with bound species adopting a distinctive "butterfly" (slow-fast-intermediate-intermediate-fast-slow) pattern (Fig. 1). When the experiment was repeated in the presence of 0.1 mM Mg 2ϩ , the free junction adopted a slow-medium-fast-fast-medium-slow "smile" pattern, as observed previously for J3, which folds with B upon X and H upon R stacking (23). Again, the Hjc⅐DNA complexes adopted the butterfly pattern. These observations suggest that the four-way DNA junction is bound by Hjc in a 2-fold symmetric conformation with the B/X arms and H/R arms related by acute angles and the B/H and R/X pairs related by obtuse angles (Fig. 1). This X shape is reminiscent of that induced on J3 by the resolving enzyme RuvC, although in that case the structure deviated only slightly from 4-fold symmetry and tended toward acute angles between the B/H and R/X pairs (24). Whereas we cannot quantify the angles between the pairs of arms, the large differences in mobility of the 6 S species suggest that the angles deviate significantly from the 90°observed for Cce1 and RuvA and may approach or surpass a 120°/60°relationship for the obtuse and acute angles.
Using a single turnover kinetic assay, we measured the rates of cleavage of each of the four strands of junction J3 individually. Cleavage was found to be biased, with a 5-fold faster rate observed for the b and r strands compared with the h and x strands (Fig. 2). Relating this to the global structure of J3 in complex with Hjc, the more strongly cleaved strands are situated in the strands exchanging with acute angles, 3 nucleotides 3Ј of the branch point ( Figs. 1 and 2). Cleavages in the other two strands probably reflect a minor population of the junction that has been bound by Hjc in the other possible conformation with the B/X and H/R pairs of arms related by acute angles.
Permanganate Probing Suggests Disruption of Base Pairing at the Cleavage Site-Thymine residues in DNA are susceptible to modification by potassium permanganate but are protected from modification in the context of fully stacked duplex DNA. This modification renders the phosphodiester backbone susceptible to cleavage by piperidine. Modification by potassium permanganate is therefore a sensitive probe for base stacking in duplex DNA. Holliday junction binding proteins such as Cce1 and RuvA increase the sensitivity of thymines at the branch point to permanganate modification in the presence of magnesium ions, and this has been interpreted as evidence that the proteins unstack the DNA junction and possibly disrupt base pairing at the point of strand exchange (24,25).
Permanganate probing was used to analyze base stacking in the context of the Hjc-junction complex with junction Z28, which has eight thymines flanking the point of strand exchange on the r strand. In the absence of Hjc protein, all eight thymines were subject to modification by potassium permanganate, with the strongest reactivity at the 2 nucleotides flanking the point of strand exchange and symmetrical distribution with respect to the junction center (Fig. 3A, lane 3). The presence of magnesium in the reaction reduced the sensitivity of the flanking thymines to permanganate, consistent with junction stacking under these conditions (Fig. 3A, lane 6). In the presence of saturating concentrations of the Hjc protein, we observed a marked increase in reactivity with potassium permanganate in both EDTA and magnesium (Fig. 3A, lanes 4 and  7).
Quantification of the increase in reactivity for each of the thymine residues (from positions Ϫ4 to ϩ4) revealed a pronounced bias in the reactivity toward the 5Ј side of the junction center (Fig. 3B), suggesting the protein unstacks the duplex DNA in this region. Analysis of the cleavage of junction Z28 by Hjc reveals that, similar to junction J3, cleavage is biased, occurring predominantly in the h and x strands 3 nucleotides 3Ј of the branch point (Fig. 3C). The position of cleavage of the h strand coincides with the area of enhanced reactivity toward permanganate detected on the r strand, suggesting that Hjc may introduce structural distortion in the DNA helix near the cleavage site. The base pairing in this region of the DNA duplex may be weakened or disrupted by Hjc on binding. Such distortions are common features of DNA recognition by restriction enzymes such as EcoRV (26) and EcoRI (27).
A Conserved Motif in Hjc Resembles the Active Site of Type II Restriction Enzymes-Analysis of the residues conserved in a multiple alignment of all known Hjc sequences revealed a motif containing three acidic residues: Glu-12, Asp-42, and Glu-55 followed by lysine (Lys-57), similar to a conserved motif present in type II restriction enzymes (28) (Fig. 4). This motif constitutes part of the active site of the restriction enzymes, with the

FIG. 4. A conserved metal binding domain in Hjc. A, Sulfolobus
Hjc sequence from residues 1-60 is shown aligned with that of three type II restriction endonucleases, EcoRV, BglI, and PvuII. Four residues absolutely conserved in all Hjc sequences (Glu-12, Asp-42, Glu-55, and Lys-57) are highlighted in black. These align with residues in all three restriction endonucleases that are known to constitute the binding pocket for the catalytic metal ions at the active site. The secondary structural elements of the three restriction enzymes taken from the crystal structures and the predicted secondary structure of Hjc generated by the program PredictProtein (30) using a multiple alignment of all the Hjc proteins are shown below the respective sequences. three acidic residues forming the metal binding pocket for the two catalytic magnesium ions (29) and the conserved lysine predicted to play a role in stabilizing the transition state during catalysis (26).
To investigate the predicted secondary structure of Hjc in this region, we submitted a multiple alignment of seven Hjc proteins to the PredictProtein structure prediction program (30). The results are outlined in Fig. 4 with the conserved domain of Hjc aligned with the corresponding domains of EcoRV, PvuII, and BglI. The predicted secondary structure for Hjc consists of an ␣-helix containing the first acidic residue followed by three ␤-sheets, in remarkable agreement with the known secondary structures of the three restriction enzymes (31).
Roles of Conserved Residues Probed by Site-directed Mutagenesis-To test our model for the metal binding site of the Hjc protein, we created site-directed mutants of the three conserved acidic residues (E12Q, D42N, E55Q) and the conserved lysine (K57A). The mutant proteins were expressed in E. coli and purified as for the wild-type enzyme. The D42N mutant was poorly expressed in E. coli, and only small amounts of protein could be purified. The other three mutants were expressed at similar levels to the wild-type protein and retained the ability to bind specifically to the four-way junction J1 with an affinity comparable with the wild-type Hjc protein (Fig. 5A). To compare the catalytic activities of the wild-type and mutant enzymes, we incubated the proteins with the four-way junction substrate Z28 in cleavage buffer at 60°C and separated cleavage products from uncut substrate by denaturing gel electrophoresis (Fig. 5B). A weak residual activity in the mutant K57A represented a 1000-fold decrease in activity compared with the wild-type enzyme. Mutants E12Q, D42N, and E55Q displayed no detectable catalytic activity. All four of these mutations thus appeared to have had a specific effect on catalysis rather than on substrate binding, consistent with the functional assignment made by analogy to the equivalent residues in EcoRV and other nucleases. DISCUSSION Hjc contains a motif first identified as constituting part of the active site of the type II restriction enzyme EcoRV. Subsequent crystal structures of the restriction enzymes PvuII (32) and BglI (33) demonstrated that the metal binding pocket is structurally conserved in an ␣␤-barrel catalytic domain in the three enzymes, even though the overall folds are very different. The type II restriction enzymes EcoRI, BamHI, and Cfr101 also possess this ␣␤-barrel domain (reviewed in Refs. 34 and 35), and more recently it has been identified in MutH (35), exonuclease (36), and type I and III endonucleases (37). All of these enzymes have a catalytic (usually N-terminal) domain in the form of an ␣␤-barrel with different domains that mediate subunit interactions and DNA recognition (35,38). The quaternary structures and functions of these proteins are remarkably divergent. Type II restriction enzymes are dimeric and recognize and cleave specific nucleotide sequences. MutH is monomeric, and its activity is regulated by the MutL and MutS proteins following mismatch DNA recognition (35). However, -exonuclease, a processive 5Ј33Ј exonuclease that degrades one strand of a DNA duplex after the formation of a double strand break, has a toroidal homotrimeric structure (36). Perhaps the most striking example of the separation of recognition and catalysis is seen in the restriction enzyme FokI, which has the catalytic ␣␤-barrel domain separated from the DNA recognition domain by a linker (39). Nature has thus repeatedly utilized a catalytic metal ion binding domain efficient in phosphodiester bond hydrolysis coupled with an array of specific DNA and protein recognition elements to solve many diverse problems in nucleic acid metabolism.
We have shown by site-directed mutagenesis that mutations in residues Glu-12, Asp-42, Glu-55, and Lys-57 of Hjc all result in at least a 1000-fold decrease in catalytic activity, though the mutants retain the ability to bind the four-way junction. By comparison, mutagenesis of the equivalent residues in EcoRV results in similar large decreases in catalytic activity without significantly affecting substrate binding (40 -44). This is the first example of this domain in a Holliday junction resolving enzyme and the first in an archaeal protein of known function.
Using the sequence alignments shown in Fig. 4, we used the crystal structure of the metal binding domain of EcoRV to generate a model of the catalytic domain of Hjc (Fig. 6A). The were constructed and purified and tested for their ability to bind to and cleave the four-way DNA junction. A, electrophoretic analysis of DNA junction binding affinity. Radioactive 5Ј-32 Plabeled junction 1 (1 nM) was incubated with increasing concentrations of wild-type or mutant Hjc at room temperature in binding buffer in the presence of 1 mM EDTA. Free junction and DNA-junction complexes were separated by electrophoresis in 5% acrylamide gels. The fraction of DNA-junction bound to protein was calculated for each concentration of Hjc by phosphor imaging and plotted against the logarithm of the protein molarity (with Hjc assumed to be dimeric). The data were fitted to a model for the binding process (see "Materials and Methods"), from which the binding affinities were calculated. All measured equilibrium dissociation constants agreed to within a factor of 3. E, wild-type Hjc; q, mutant E12Q; ‚, mutant E55Q; OE, mutant K57A. B, activity of wildtype and mutant Hjc enzymes is shown. Mutants E12Q, D42N, and E55Q gave no detectable cleavage products after incubation with junction Z28 labeled on the h strand for 30 min at 60°C (lanes 1-3). Very weak activity was detected for mutant K57A under the same conditions (lane 4) whereas the wild-type (w-t) enzyme displayed strong cleavage activity after a 20-s incubation (lane 5). Lanes 1-4 show junction Z28H that was incubated for 30 min at 60°C with mutants E12Q, D42N, E55Q, and K57A, respectively. Lanes 5 and 6 show junction Z28H that was incubated at 60°C with wild-type Hjc for 20 s and 30 min, respectively. model encompasses the N-terminal half of the Hjc protein consisting of an ␣-helix and a three-stranded antiparallel ␤-sheet. The three acidic residues group in reasonable positions to act as ligands for two magnesium ions (magenta spheres in the model), and the orientation of the conserved lysine is consistent with a role in catalysis. By analogy with the other nucleases containing the metal ion binding domain, the Cterminal half of the protein, which is less conserved in Hjc proteins from different archaeal species, may play a role in dimerization and/or DNA recognition. Fig. 6B shows a model for the complex of the catalytic domain of Hjc, positioned on the minor groove side of a four-way DNA junction, which is represented in the X conformation suggested by comparative gel electrophoresis. The presumed catalytic magnesium ion has been positioned within 5 Å of the known site of cleavage on the phosphodiester backbone. Using this constraint, the model suggests a possible mode of binding of the Hjc protein with the four-way DNA junction substrate. Basic residues in Hjc are in reasonable proximity to the negatively charged phosphodiester backbone, and in particular, the model suggests possible roles for two arginines, Arg-13 and Arg-28. These are the only basic residues other than the catalytic Lys-57 that are absolutely conserved in all Hjc sequences, and in the model of the complex they exist in suitable positions to form salt bridges with phosphates on an adjacent arm of the junction. On the basis of the model and the conservation of these residues we predict that they are important for DNA binding and possibly for recognition of the branched structure of the four-way DNA junction.
In conclusion, we have demonstrated that the archaeal Holliday junction resolving enzyme Hjc manipulates the global structure of the four-way DNA junction into a 2-fold symmetric X conformation on binding. Paired nicks in the phosphodiester backbone are introduced on the minor groove face in a region of duplex DNA 3 bp from the point of strand exchange, accompanied by local distortion of base stacking in the duplex. Hjc utilizes an ␣␤-barrel domain for the chemical step in the resolution of four-way DNA junctions, with four conserved catalytic residues found in many other nucleases of diverse function. The C-terminal half of the Hjc protein may complete the ␣␤-barrel domain and provide an interface for dimerization. Thus the Hjc dimer may consist of two nuclease domains capable of introducing nicks in duplex DNA on either side of the branch point of a four-way junction, with other structural elements providing structural specificity and forming the dimer interface.
FIG. 6. Model of the catalytic domain of Hjc and its interaction with the Holliday junction. Catalytic domain models were generated using the program MODELLER (49) with the structure generated based on homology to the requisite domain of EcoRV (1B94). Twenty model structures were generated and averaged using the program X-PLOR (50). The junction was generated within INSIGHT II (Molecular Simulations, Inc.). The complete structure was assembled with the rigid body minimization using distance restraints between the catalytic metal ions, the conserved arginine residues, and the DNA backbone. A, model of the catalytic domain of Hjc. The structure is represented by a cartoon of the predicted secondary structure. Proposed catalytic residues and metal ions are shown as stick and space filling representations, respectively. B, model of the Hjc-junction complex. Model catalytic domains are represented by ribbons with the positions of the magnesium ions and the conserved residues Arg-13 and Arg-28 indicated as stick representations. The model DNA junction structure is shown in gray with a backbone ribbon highlighting the position of junction cleavage in red.