Structural Insights into the Mechanism of Nuclease A, a ββα Metal Nuclease from Anabaena*

Nuclease A (NucA) is a nonspecific endonuclease from Anabaena sp. capable of degrading single- and double-stranded DNA and RNA in the presence of divalent metal ions. We have determined the structure of the Δ(2-24),D121A mutant of NucA in the presence of Zn2+ and Mn2+ (PDB code 1ZM8). The mutations were introduced to remove the N-terminal signal peptide and to reduce the activity of the nonspecific nuclease, thereby reducing its toxicity to the Escherichia coli expression system. NucA contains a ββα metal finger motif and a hydrated Mn2+ ion at the active site. Unexpectedly, NucA was found to contain additional metal binding sites ∼26 Å apart from the catalytic metal binding site. A structural comparison between NucA and the closest analog for which structural data exist, the Serratia nuclease, indicates several interesting differences. First, NucA is a monomer rather than a dimer. Second, there is an unexpected structural homology between the N-terminal segments despite a poorly conserved sequence, which in Serratia includes a cysteine bridge thought to play a regulatory role. In addition, although a sequence alignment had suggested that NucA lacks a proposed catalytic residue corresponding to Arg57 in Serratia, the structure determined here indicates that Arg93 in NucA is positioned to fulfill this role. Based on comparison with DNA-bound nuclease structures of the ββα metal finger nuclease family and available mutational data on NucA, we propose that His124 acts as a catalytic base, and Arg93 participates in the catalysis possibly through stabilization of the transition state.

EndoG (5), and Streptococcus pneumoniae EndA (6) are some of the other family members. The best characterized of these is the Serratia nuclease (7), a homodimer (8 -11) that contains two disulfide bonds per subunit (12) and is characterized by an unusual Mg 2ϩ binding site optimized for binding a hydrated metal ion (13,14).
The prokaryotic nucleases of this family generally contain an N-terminal signal peptide and are secreted into the extracellular environment (15). It is believed that most of them serve nutritional purposes and possibly also function as bacteriocides, similar to the colicins of Escherichia coli (16). Because nonspecific nucleases will generally be extremely toxic to the cells that produce them, various mechanisms have evolved to deal with their toxicity. Many of the nucleases in this class contain disulfide bonds that activate the nucleases upon secretion into an oxidizing environment but when reduced in the intracellular environment render the enzyme inactive (15). If the cysteine residues in the N-terminal signal peptide are not included, there is only one additional cysteine in NucA, and hence it is not anticipated to contain disulfide bonds, which could prevent the nuclease from being active in the cell. Instead, NucA is paired with a specific inhibitor, NuiA, analogous to the colicin immunity proteins in E. coli (17). NuiA protects the cell from nuclease action by forming a very stable NucA⅐NuiA complex (2,18,19). The three-dimensional structure of NuiA has recently been determined (20).
NucA is a monomeric enzyme with secondary structure composition, as judged by circular dichroism spectroscopy, and a metal ion cofactor requirement similar to the Serratia nuclease (2). It prefers Mn 2ϩ and Co 2ϩ over Mg 2ϩ and shows little activity with Ni 2ϩ and no activity with Zn 2ϩ . It has a pH optimum of 5.5-7.5 and a temperature optimum around 35°C. With 8.4 ϫ 10 6 Kunitz units/mg of protein (k cat ϭ 2,055 s Ϫ1 ), NucA is one of the most active nucleases known. For comparison, DNase I has 7.2 ϫ 10 5 Kunitz units/mg of protein (k cat ϭ 200 s Ϫ1 ) (21). The general sequence preferences of NucA, e.g. avoiding d(A)⅐d(T) tracts (2), are similar to other nonspecific nucleases, such as the Serratia nuclease (22,23).
NucA and Serratia nuclease share ϳ22% sequence identity ( Fig. 1), which is most pronounced in the active site region and the central multistranded ␤-sheet core of the nuclease. According to a detailed mutational analysis, most amino acid residues important for phosphodiester bond hydrolysis by Serratia nuclease (9,19,24) have counterparts in NucA, suggesting that Serratia nuclease and NucA share a common mechanism of action, with differences in detail (19). It has recently become clear that the ␤␤␣ Me finger motif (in this terminology "Me" refers to the bound metal ion) is shared by a broad class of nucleases (25) which includes the highly specific homing endonuclease I-PpoI (26 -28), a member of the His-Cys box family (29), and the Vibrio nuclease Vvn from Vibrio vulnificus (30,31). Despite the overall similarities, the sequences suggest a number of significant differences, the reality of which can only be resolved by structural comparisons. For example, sequence alignment of NucA with the Serratia nuclease indicates that Arg 57 of Serratia, a residue thought to be important for stabilizing the transition state, is not present in NucA (19). Thus, NucA either uses a different mechanism for stabilizing the transition state, or there is a structural alignment that is not immediately apparent from the sequence alignment. The resolution of this and related questions can only be resolved by a complete structural determination.
Here we report the crystal structure of NucA in complex with Mn 2ϩ at a resolution of 1.9 Å. NucA bears a mixed ␣␤ fold topology. Comparison with other nucleases reveals a similar structural arrangement in the endonuclease active site containing two ␤-strands and one ␣-helix with a centrally located divalent metal ion, as expected for a member of the ␤␤␣ Me finger family of nucleases. Superpositions of the NucA structure with Vvn (31) with bound DNA suggest the position of the DNA binding site as well as specific catalytic roles for various residues. These results provide insight into the catalytic mechanism for NucA as well as structural variations relative to the Serratia nuclease.

MATERIALS AND METHODS
Protein Construct and Site-directed Mutagenesis-Expression of nucleases in E. coli cells generally requires some type of nuclease inactivation to limit cellular toxicity. In the present NucA studies, this was achieved by the introduction of a D121A mutation, which results in a reported 0.04% activity relative to the wild type enzyme (19). The full-length NucA sequence contains a signal peptide at the N terminus, which is responsible for the protein excretion to the periplasm. For this study we have used a construct in which N-terminal residues 2-24 were replaced by an A(H) 6 M His tag, and a point mutation was introduced at position 121 to produce a D121A mutant. To obtain the D121A mutant, site-directed mutagenesis was performed using QuikChange kit (Stratagene) and verified by DNA sequencing. Residue numbering is based on the complete sequence of the enzyme (Fig. 1).
Protein Expression and Purification-Recombinant NucA was produced by transforming the plasmid into E. coli BL21 star (DE3). Cells were grown to mid-log phase (A 600 ϭ ϳ0.6) at 37°C in LB medium containing 50 g/ml carbenicillin. NucA protein expression was induced with 0.4 mM isopropyl-␤-D-thiogalactopyranoside at 37°C for 6 h. Cells were harvested by centrifugation (at 7,000 ϫ g), resuspended in 50 mM Tris-HCl, pH 7.5, and 150 mM NaCl, and lysed by sonication with a Branson Sonifier 200 using a microtip probe. The lysate was centrifuged at 30,000 ϫ g for 40 min. The nuclease was extracted from the inclusion bodies (for details, see Ref. 2) in the pellets using 50 mM Tris-HCl, pH 7.5, 6 M urea, 10 mM imidazole by soaking overnight. The urea-containing solution was recentrifuged (30,000 ϫ g, 60 min, 4°C), and the supernatant was applied to a nickel-nitrilotriacetic acid resin (Qiagen) equilibrated with extraction buffer and eluted with 50 mM Tris-HCl, pH 7.5, 6 M urea, and 200 mM imidazole. The eluted fraction containing protein was then concentrated using a Millipore concentrator to a desired volume of 10 ml (ϳ10 mg/ml concentration) and rena-tured using flash dilution of 20-fold with moderate stirring at 4°C. The renaturing buffer containing 50 mM Tris-HCl, pH 7.5, 300 mM NaCl, 2 mM MgCl 2 , and 5 mM dithiothreitol was precooled to 4°C. The renatured enzyme was then concentrated using an Amicon concentrator and applied to the Superdex-75 gel filtration 26 ϫ 60-cm column and equilibrated with 50 mM Tris-HCl, pH 7.5, 300 mM NaCl, 5 mM dithiothreitol. The major peak of absorbance at 280 nm was found to be pure NucA protein as judged by SDS-PAGE. The peak fractions were concentrated using a Millipore concentrator, and the buffer was exchanged to 20 mM Tris-HCl, pH 7.5, 200 mM NaCl.
Crystallization and Data Collection-The NucA ⌬(2-24),D121A mutant was concentrated to 10 mg/ml and dialyzed in 20 mM Tris, pH 7.5, and 200 mM NaCl. Crystals of this protein were obtained using the hanging drop vapor diffusion technique at 4°C by mixing 2 l of the protein solution with 2 l of the reservoir solution consisting of 100 mM MES, pH 6.5, 10 mM ZnSO 4 , 300 mM NaCl, and 35% polyethylene glycol 550. For the data set without manganese ion present, crystals were transferred and soaked in 100 mM MES, pH 6.5, 300 mM NaCl, and 35% polyethylene glycol 550. For the data set with manganese, crystals were transferred and soaked in 100 mM MES, pH 6.5, 10 mM MnCl 2 , 300 mM NaCl, 35% polyethylene glycol 550, and 10 mM of the dinucleotide 5Ј-CG-3Ј.
For data collection, crystals were flash cooled by submersion in liquid nitrogen and placed on the goniometer in a stream of nitrogen gas cooled to Ϫ180°C. Data were collected using a RU-H3R generator equipped with Osmic mirrors and a Raxis IV area detector. NucA crystallized in the orthorhombic space group P2 1 2 1 2 1 with one molecule/ asymmetric unit.
Structure Determination and Refinement-A model for molecular replacement was created from the crystal structure of Serratia endonuclease (PDB code 1G8T (14)) by replacing nonidentical side chains with alanines. The program Molrep (32) from CCP4 (33) was used to solve the molecular replacement problem. The model was then refined by iterative cycles of model building using the program O (34) and refinement using the program CNS (35). The quality of the final structures was assessed using the programs Procheck (36) and MolProbity (37). The statistics for the data collection and results from refinement are reported in Table I. The structure of the Mn 2ϩ enzyme has been submitted to the Protein Data Bank (PDB code 1ZM8).

RESULTS
Overall Structural Characteristics of NucA-The crystal structure of NucA reveals an ␣␤ mixed globular fold (Fig. 2a). There are 13 helices and 8 strands giving rise to a large, six-stranded ␤-sheet and a small, two-stranded ␤-sheet. The active site of the enzyme contained one hydrated divalent metal ion, and one sulfate ion coordinated to the active site residues. An interesting outcome of this structural determination was the observation of a second and third metal ion binding site located ϳ 26 Å from the metal ion in the active site (Fig. 2a).
Structure of the Active Site in the Manganese-substituted Enzyme-To obtain a structure of NucA with Mn 2ϩ in the active site, crystals grown in the presence of Zn 2ϩ were transferred into solution containing no Zn 2ϩ and only Mn 2ϩ as a divalent cation. The bond lengths and B-factors obtained from the refinement are consistent with the identification of this ion as Mn 2ϩ , and we have modeled it as such. The active site revealed an octahedral coordination of the divalent ion, with an inner coordination sphere that includes the side chain carbonyl oxygen of Asn 155 , four water molecules, and an apparent sulfate ion. The Mn-O bond distances range from 2.04 Å for the bond to Asn 155 , to 2.36 for one of the bonds to water, with an average length of 2.19 Å (Table II). The mode of interaction with the divalent cation observed in NucA is typical of a series of other nucleases and ATPases with active sites that are optimized for binding a hydrated ion, e.g. Serratia nuclease, I-PpoI, and MutL (13,38). In all of these examples, the active sites are optimized to interact with a hydrated metal ion rather than making direct contact with the metal ion, so that most of the protein interactions with the active site divalent ion are mediated by water molecules. For NucA, water molecule 1 (w1) is positioned by its interaction with His 124 , water molecules w2 and w3 are positioned by Glu 163 (Fig. 2b), and water molecule w4 by an interaction with Gln 150 . These positioning residues are themselves held in place by an additional network of hy-  2. Overall fold of NucA. a, stereo view of the crystal structure of NucA ⌬(2-24),D121A mutant showing the two-stranded and sixstranded ␤-sheets (navy) and 13 ␣-helices (lavender) forming the framework of the molecule. The two metal binding sites bind a total of three manganese ions, shown in green, and a sulfate ion, shown in yellow. b, stereo view showing the active site residues (in khaki) and the divalent metal ion (in green) with its octahedral coordination to Asn 155 , a sulfate ion, and four water molecules (water molecules 1-4 named as w1-w4, respectively) illustrated by orange dotted lines. The dotted lines shown in navy represent the network of hydrogen bonds near the active site. The sulfate ion is shown in yellow. drogen bonds, e.g. the second carboxyl oxygen of Glu 163 is hydrogen-bonded to the N⑀ of Trp 159 . The only direct bond formed between the divalent ion and the protein involves the Asn 155 side chain carbonyl oxygen atom. In the native enzyme, the position of the Asn 155 side chain would probably be stabilized by the interaction with Asp 121 , analogous to the interaction between Asn 119 and Asp 86 in the Serratia nuclease. However, the Asp 121 side chain is not present in the D121A mutant that was studied. Apparently, the interaction between the Asn 155 side chain and the Mn 2ϩ is sufficient to define the position of this side chain in the absence of the Asp 121 interaction. The sulfate ion observed in the active site interacts with the Mn 2ϩ , with the side chain amide of Asn 155 , and with the guanidino group of Arg 93 . If the sulfate is considered as a representative for a nucleotide phosphate group, this structure implies that the Asn 155 and Arg 93 residues would also interact directly with a phosphate oxygen of the substrate. Additionally, Arg 93 is hydrogen bonded to Asp 95 as a part of the additional hydrogen bond network around the active site which may exist to help stabilize the active conformation of Arg 93 . Finally, we note that although the dinucleotide 5Ј-CG-3Ј was present in the crystallization solution, no electron density could be clearly identified for the dinucleotide.
Additional Metal Binding Site-An unexpected result of the crystal structure determination reported here is the presence of a second and third metal binding site located near the C terminus of the protein and ϳ26 Å from the active site of the enzyme. This site also represents an important crystal contact because both metal ions also coordinate with a Glu 136 residue from a symmetry-related molecule. The site includes four acidic residues: Asp 246 , Glu 249 , Asp 255 , and Glu 269 , as well as carbonyl oxygen ligands derived from Phe 256 and Gln 265 . These residues together bind two divalent ions (Fig. 3). The coordination of metal ion 2 (Mn2) is approximately octahedral, involving interactions with a single carboxyl oxygen from Asp 246 , Glu 249 , and Glu 269 , a water molecule, and the two carboxyl oxygen atoms from Glu 136 of the symmetry-related NucA molecule. The average metal-ligand bond length (Table II) (40) and ColE9 (41), Vvn nuclease (31), and the apoptotic nuclease CAD/DFF40 (42); (c) structure-specific nucleases such as T4endoVII (43); (d) type II restriction endonucleases such as KpnI (44); and (e) homing endonucleases such as I-PpoI (38) and I-HmuI (45).
The active site of the ␤␤␣ Me finger nucleases supplies one or two Mg 2ϩ ligands as identified by crystal structure analyses: Asn 119 in Serratia nuclease (13,14); Asn 155 in NucA, His 544 , and His 569 in ColE7 (40); His 102 and His 127 in ColE9 (41); Glu 79 and Asn 127 in Vvn nuclease (31); Asp 262 and His 308 in CAD/ DFF40 (42); Asp 40 and Asn 62 in T4endoVII (43); Asn 119 in I-PpoI (38); Asp 74 and Asn 96 in I-HmuI (45). Based on structure model building and biochemical data, the Mg 2ϩ ligands are Asn 174 in EndoG (39) and Asp 148 and Gln 175 in KpnI (44). The different modes of metal ion cofactor binding allow dividing the ␤␤␣ Me finger nucleases into subgroups: Serratia nuclease, NucA, I-PpoI, and, based on computations, EndoG with an Asn residue as the sole Mg 2ϩ ligand; Vvn nuclease, T4endoVII and I-HmuI with a carboxylate and a carboxamide as Mg 2ϩ ligands; the colicins with two His residues as Mg 2ϩ ligands and; CAD/ DFF40 with a carboxylate and a histidine as Mg 2ϩ ligands takes up a position between the last two groups. In all cases, direct coordination of the metal ion by protein residues is incomplete, so that the active site is optimized for binding a hydrated metal ion.
The closest known structural homolog of NucA among the ␤␤␣ Me finger nucleases and nonspecific nuclease family containing the DRGH recognition motif is the nonspecific nuclease from S. marcescens. Although Serratia nuclease and NucA have only 22% sequence identity (Fig. 1), it is the closest structure identified by the DALI server (46). The overall folds of the two enzymes are very similar (Fig. 5a). Despite this structural homology of Serratia nuclease with NucA, there are several significant differences. In particular, the Serratia nuclease, along with most other nucleases in this family, has a dimeric structure, whereas NucA is monomeric (19). The dimer interface in the Serratia nuclease includes a short loop from Arg 136 to Ile 139 , a short loop and ␤-strand region from Asn 178 to Tyr 185 , and a longer stretch of residues at the C terminus of the protein (ϳresidues 225-245). As can be seen from the alignment in Fig. 1, this region of the protein is generally not well conserved and includes many gaps. His 184 , which forms a hydrogen bond with the Ser 229 and with the Pro 180 carbonyl in the interchain region, corresponds to Arg 222 in NucA. It has previously been noted that the H184R mutant of the Serratia nuclease is monomeric (11). The interstrand salt bridge (Asp 225 -Arg 136 ) in the Serratia structure would not exist in NucA, which has Asn and Gln residues at these positions. Similarly, the (Lys 233 -Glu 239 ) interstrand salt bridge in Serratia would not be present in NucA. Of course, alternate stabilizing interactions could in principle be present in NucA; however, both biochemical (47) and the present crystal structure data support a monomeric structure for NucA.
The active site of NucA is observed to be homologous with that of the Serratia nuclease. A Mn 2ϩ ion is coordinated directly to the side chain carbonyl group of Asn 155 , four additional water molecules, and a sulfate oxygen forming a hexadentate octahedron. This pattern of coordination is analogous to that of the active site Mg 2ϩ ion observed in the Serratia nuclease (Fig. 5b). Compared with the surface residues of the Serratia nuclease only four of the basic residues are conserved or conservatively replaced in NucA, among which Arg 122 and Arg 167 , corresponding to Arg 87 and Arg 131 in the Serratia nuclease, are also important for activity based on earlier mutational analysis as mentioned in Ref. 19.
One region of considerable sequence disparity between the two nucleases is found at the N terminus (residues 5-33 for NucA; residues 36 -70 for the Serratia nuclease) (Fig. 5c). In Serratia, this region includes two cysteine residues that form a disulfide bond that has been proposed to regulate the activity of the nuclease (12). Despite the complete lack of sequence homology and the missing disulfide bond, both structures occupy a similar region of the protein, filling a generally hydrophobic pocket located on the face opposite the active site. NucA contains an ␣-helix in place of the disulfide in Serratia nuclease. The extra secondary structure in this region of NucA may help stabilize the active conformation, whereas the presence of the disulfide in this region leads to active Serratia nuclease. The structural importance of this region in NucA was supported by our observation that a ⌬59 construct of NucA failed to yield protein that could be refolded.
Several of the metal ion ligands present in the second and third metal ion binding site of NucA are also present in the Serratia nuclease; however, as indicated in the alignment of Fig. 1, Asp 255 and Glu 269 are not conserved in the Serratia sequence. No other metal binding sites were observed in two reported crystal structures of Serratia nuclease (13,14). A survey of other related (DRGH motif-containing) nucleases also revealed no significant conservation of the residues involved in the second and third metal binding site except for Glu 249 , indicating that it is unique to NucA. Although the location of the second and third metal ion binding sites at a crystallographic interface introduces some uncertainty regarding the physiological significance of these sites, the presence of four anionic residues contributed by a single NucA molecule makes for interesting speculation. Potentially, this site could play a structural or regulatory role, it could be involved in anchoring or binding NucA to some other cellular component, similar to its role in creating a crystal contact, or it could possess an additional catalytic activity. The availability of the two metal ions for an intermolecular interaction would be consistent with the ability to interact with other substrates, suggesting that it might play a functional role. Mechanism of Phosphodiester Bond Hydrolysis by NucA-In the present study we report the first crystal structure of NucA. This structure substantiates previous suggestions that NucA is a member of the ␤␤␣ Me family of nucleases (19), providing insight into its catalytic mechanism. Drawing analogies to the proposed mechanisms for other ␤␤␣ Me nucleases I-PpoI (38), ColE9 (41), and I-HmuI (45), the hydrolytic reaction pathway would proceed through an in-line displacement reaction mechanism whereby His 124 , acting as a general base, would deprotonate a water molecule which would attack the scissile phosphate coordinated to the divalent metal ion through a nonbridging oxygen and the leaving group 3Ј-oxygen (Fig. 6). In both the Serratia nuclease (13,14) and NucA structures, the proposed general base histidine forms a hydrogen bond with an inner coordination sphere water of the metal ion. Superpositions with I-PpoI (38) and I-HmuI (45), which were both crystallized in the presence of DNA, suggest that this water occupies the position of the nonbridging phosphate oxygen chelated to the metal ion. Thus, it is possible that upon substrate binding this water is replaced by the nonbridging oxygen of the phosphate, and the His 124 side chain undergoes a slight conformational change to its catalytically active conformation, similar to the position of the catalytic histidine residue in the I-PpoI structure (Fig. 6). The divalent metal ion, along with the Arg 93 guanidino group, would help stabilize negative charge buildup in the transition state through interactions with the nonbridging oxygen as well as enhancing the leaving group characteristics of the 3Ј-bridging oxygen.
Based on comparisons with the Vvn structure (31), the singly coordinated divalent metal only interacts with the nonbridging oxygen, suggesting that its role is to stabilize charge buildup on the scissile phosphate. It is suggested that a water molecule coordinated to the metal ion may help stabilize charge developed on the 3Ј-leaving group. It is interesting to note that a superposition of the structure of the Vvn product complex (31) positions the previous scissile phosphate directly on top of the sulfate coordinated to the Mn 2ϩ ion in the NucA structure (Fig.  7). Whether or not this suggests that the mechanism of NucA is more similar to that proposed for Vvn is unclear. It should be noted, however, that in the Vvn structure (31) the divalent metal ion in the complex is Ca 2ϩ , which has significantly larger bond distances than Mg 2ϩ or Mn 2ϩ , leading to a distorted active site. In either mechanism, Arg 93 is in position to stabilize the transition state (Fig. 6) by interacting with the second nonbridging oxygen on the scissile phosphate or perhaps through stabilization of charge developed on the bridging oxygen of the 3Ј-leaving group. Arg 93 is located in a position spatially similar to Arg 57 in the Serratia nuclease (13,14) and Arg 99 in Vvn (31). We note, in addition, that in the DNA-bound model of NucA, the position of DNA on the surface of NucA near the catalytic site is consistent with the positive charge of the electrostatic potential map of the surface of NucA (Figs. 4  and 7).
In several nucleases for which nuclease-DNA co-crystal structures are available (ColE7, His 545 (40); ColE9, His 103 (41); Vvn nuclease, His 80 (31); I-PpoI, His 98 (38); and I-HmuI, His 75 (45)) a histidine residue is ideally positioned to act as a general base to activate a water molecule for the nucleophilic attack on the phosphorus atom. Most probably, this is the case in other nucleases of ␤␤␣ Me finger nuclease family. This assumption is supported by results of mutational analyses (Serratia nuclease, His 89 (24); NucA, His 124 (19); EndoG, His 143 (39); ColE7, His 103 (48); T4endoVII, His 41 (49); KpnI, His 149 (44)) and the analysis of the pH dependence of phosphodiester bond cleavage (pH half-maximal activity ϳ 6 for Serratia nuclease (24) and NucA (2)). In general, most of these nucleases do not require an additional activation or orientation of the histidine residue involved in the catalysis. It is of particular interest that, in contrast with the other active site residues, His 124 is not more rigidly positioned, e.g. by a hydrogen bond to the N⑀ of the ring. In general, a more rigid position would enhance its ability to interact with water w1, which is in the inner Mn 2ϩ coordination sphere. The absence of such an interaction supports the idea of some positional variability, as illustrated, for example, in Fig. 6 as part of its catalytic function.
The position of Arg 93 in NucA and its interaction with the sulfate ligand, as well as the simulated DNA complex based on the Vvn structure (Fig. 7), suggest that it may play a role in positioning the substrate and in stabilizing the transition state structure. A comparison of the active sites of some ␤␤␣ Me finger nucleases suggests that in many nucleases a basic amino acid residue also is involved in neutralizing the extra negative charge on the phosphate oxygen in the transition state. Replacing Arg 57 in Serratia nuclease (24), Arg 5 in ColE9 (50), and Arg 61 in I-PpoI (51) by alanine demonstrates that these arginine residues are essential for nuclease activity (in the case of I-PpoI a large effect is only seen at pH 10). In T4endoVII three histidine residues are candidates for transition state stabilization: their substitution by other amino acid residues leads to an inactive or almost inactive enzyme (43,52,53). A central role for Arg 93 in the catalytic mechanism of NucA is supported by the observation that the NucA R93A variant has only 0.12% residual activity (19) (for comparison the Serratia nuclease R57A variant has 0.6% residual activity (24)). We therefore conclude that in the NucA-catalyzed reaction, Arg 93 (Fig. 6) is involved in transition state stabilization. CONCLUSION This work presents the three-dimensional structure of the nonspecific nuclease, NucA, from Anabaena sp., determined using x-ray crystallographic methods. NucA is one of the most active nonspecific nucleases that has been identified and is able to cleave both single-and double stranded DNA and RNA. In contrast with several other family members for which structural information is available, NucA is monomeric. However, it exhibits an ␣␤ mixed fold that is similar to several other known nucleases, most closely resembling the Serratia nuclease. Despite limited sequence identity, the structure allows classification of NucA as a member of the ␤␤␣ Me family, which includes both specific and nonspecific nucleases. The crystal structure also shows the presence of a second metal ion binding site near the C terminus of the protein which is ϳ26 Å from the active site of the enzyme. This site, which includes four acidic residues, binds two Mn 2ϩ ions, but its physiological significance is unknown at present. Based on the crystal structure of NucA in the presence of manganese, the available mutational data, and a model of a NucA⅐DNA complex derived by homology with the previously determined Vvn⅐DNA complex (31), a single-metal ion mechanism is proposed for the DNA hydrolysis catalyzed by NucA. In this mechanism His 124 would act as a general base, and both Arg 93 as well as the active site divalent ion could stabilize the transition state, analogous to proposals for other ␤␤␣ Me finger family nucleases. FIG. 7. Modeled DNA-NucA interaction. Stereo view of the modeled interaction of a duplex DNA octamer with selected residues from NucA is shown. The model is based on the alignment of the NucA structure (Arg 122 -Ile 125 ) residues with Vvn (Trp 78 -Val 81 ) and NucA (Met 147 -Arg 156 ) with Vvn (Leu 119 -Gly 128 ) residues in the Vvn⅐DNA structure (1OUP). The coordination of the active site Mn 2ϩ (green) to four water molecules as well as to Asn 155 and to a phosphate oxygen of the cleaved DNA strand (shown in pink) is indicated. The active site residues are shown in gray, and the complementary strand of the DNA is shown in blue.