Structure-based Analysis of the Metal-dependent Mechanism of H-N-H Endonucleases*

Controversy surrounds the metal-dependent mechanism of H-N-H endonucleases, enzymes involved in a variety of biological functions, including intron homing and DNA repair. To address this issue we determined the crystal structures for complexes of the H-N-H motif containing bacterial toxin colicin E9 with Zn 2 (cid:1) , Zn 2 (cid:1) (cid:1) DNA, and Mg 2 (cid:1) (cid:1) DNA. The structures show that the rigid V-shaped architecture of the active site does not undergo any major conformational changes on binding to the minor groove of DNA and that the same interactions are made to the nucleic acid regardless of which metal ion is bound to the enzyme. The scissile phosphate contacts the single metal ion of the motif through distortion of the DNA brought about by the insertion of the Arg-96-Glu-100 salt bridge into the minor groove and a network of contacts to the DNA phosphate backbone that straddle the metal site. The Mg 2 (cid:1) -bound structure reveals an unusual coordination scheme involving two H-N-H histidine residues, His-102 and His-127. The mechanism of DNA cleavage is likely related to that of other single metal ion-dependent endonucleases, such as I- Ppo I and Vvn, although in these enzymes the single alkaline earth metal ion is coordinated by oxygen-bear-ing amino acids. The structures also provide a rationale as to why H-N-H endonucleases are inactive in the presence of Zn 2 (cid:1) but active with other transition metal ions

Indeed, these families have diverged to such an extent that they share only 2 active site residues, a histidine that acts as a general base (His-103 E9 DNase/His-98 I-PpoI) and an arginine that is involved in phosphate binding (Arg-5 E9 DNase/Arg-61 I-PpoI) (14).
The identity of the biologically active metal ion for His-Cys enzymes such as I-PpoI is known to be magnesium (16), whereas this issue remains enigmatic for H-N-H enzymes. This is because of the four histidine residues of the motif, all of which are required for catalytic activity (4). In the absence of DNA, the H-N-H motif binds transition metals directly, with no detectable binding of magnesium (17). This is borne out by crystal structures for the E7 and E9 DNases where Zn 2ϩ and Ni 2ϩ , respectively, have been observed bound to the motif histidines (6,11). Ku et al. (18) have also reported that the E7 enzyme is a zinc-dependent enzyme. By contrast, we have been unable to detect Zn 2ϩ activity for any colicin DNase (E2, E7, E8, or E9), with Zn 2ϩ tending to be inhibitory (8,9,17). Nevertheless, colicins are able to utilize some first row transition metal ions such as Co 2ϩ and Ni 2ϩ to cleave DNA (we have speculated that apparent Zn 2ϩ activity may be because of Ni 2ϩ contamination) (19). These differences are related to the kinetics and thermodynamics of transition metal binding. Zn 2ϩ association includes an additional step that traps the enzyme in a high affinity but inactive state (K d ϳ nM) (19). Active transition metals are those that bind with weaker affinities (K d ϳ M). Through mutagenesis of H-N-H residues, Walker et al. (4) have established that Mg 2ϩ ions are the preferred metal for dsDNA cleavage by the E9 DNase, and this is consistent with the metal requirements for other H-N-H enzymes such as I-CmoeI and LtrA from Chlamydomonas moewussi and Lactococcus lactis, respectively (20,21).
To understand the mechanism by which these enzymes hydrolyze nucleic acid, in the present work we have used crystallography to address how Mg 2ϩ ions bind to the histidine-rich H-N-H motif of the E9 DNase domain in the context of dsDNA. We have also rationalized why Zn 2ϩ is inactive within the H-N-H motif, whereas other transition metals such as Ni 2ϩ endow the motif with activity.

MATERIALS AND METHODS
The H103A E9 DNase was expressed and purified as described (4,8). Apo-H103A E9 DNase was obtained after treatment with 5 mM EDTA and dialysis against 200 mM NaCl followed by dialysis against water and lyophilization (17,19). The absence of metal in the H-N-H motif of purified protein was monitored routinely by fluorescence spectroscopy as described by Pommer et al. (17). Crystals of the H103A E9 DNase in complex with Zn 2ϩ were obtained by hanging drop vapor diffusion against 26% w/v polyethylene glycol MME 2000, 100 mM KSCN, and 100 mM sodium acetate, pH 5.8, at a protein concentration of 30 mg/ml. Crystals were flash-cooled in Paraton-N and data collection performed at 100 K using the beamline ID14.1 at the European Synchrotron Radiation Facility (Grenoble).
In the preparation of H103A E9 DNase in complex with dsDNA, 8-mer palindromic DNA (5Ј-GCGATCGC-3Ј) was first denatured at high temperature and then annealed by slow cooling. Bases are numbered in the text as 1-8 for one strand and 9 -16 for the complementary strand. A stoichiometric amount of duplex was added to the protein to form the complex (final concentration, 40 mg/ml). Crystals of the Zn 2ϩ complex were grown by hanging drop vapor diffusion against 18% v/v polyethylene glycol 400, 100 mM HEPES buffer, pH 7, and 6 mM ZnCl 2 , whereas 15% v/v polyethylene glycol 400, 100 mM cacodylate buffer, pH 6.5, containing 100 mM MgCl 2 was used for the Mg 2ϩ complex. Crystals were flash-cooled in mother liquor containing 28% v/v polyethylene glycol 400. X-ray diffraction data for both complexes were collected at the European Synchrotron Radiation Facility (Grenoble) beamline ID14.1.
All data sets were integrated and scaled using MOLSFLM (22) and SCALA (23), with the final statistics presented in Table I. Crystals of the H103A E9 DNase⅐Zn 2ϩ complex belong to the space group P2 1 2 1 2 1 and have two molecules of the complex in the asymmetric unit, with a Matthews coefficient of 2.4. Crystals of complexes with DNA bound belong to the space group C222 1 and have four molecules of complex in the asymmetric unit, giving a Matthews coefficient of 2.0. Structures were solved by molecular replacement using the program AmoRe (24), using the E9 DNase domain bound to Im9 (Protein Data Bank code 1EMV) as the search model. All the structures were refined without non-crystallographic symmetry restraints using REFMAC (23), and the rebuilding was done using XtalView (25) and Coot (23). The omit maps for the metal site in the complexes with DNA were calculated independently using CNS (26) with simulated annealing and REFMAC; the residues omitted were the metal, the scissile phosphate, and histidines 102, 127, and 131. The maps calculated with both programs looked the same, and therefore only one is presented.

RESULTS AND DISCUSSION
Structure Determination and Description-To obtain structures of the E9 DNase bound to metal ions, we focused on an inactive mutant in which His-103, the putative general base, was substituted for alanine. The H103A mutant shows wild type binding of transition metals but is unable to catalyze DNA hydrolysis (4). Structures of three complexes of H103A E9 DNase were obtained in which the H-N-H motif was bound with (i) Zn 2ϩ , (ii) Zn 2ϩ ⅐dsDNA, and (iii) Mg 2ϩ ⅐dsDNA. The dsDNA used in crystallization experiments was an 8-mer palindromic sequence. Hsia et al. (13) used the same DNA duplex in their recent structure of the apo-E7 DNase bound to dsDNA. The three E9 DNase complexes were crystallized and their structures solved by molecular replacement using the coordinates for the wild type E9 DNase (see "Materials and Methods") (6,30). The resolution for each of the structures of H103A E9 DNase bound to Zn 2ϩ , Zn 2ϩ ⅐dsDNA, and Mg 2ϩ ⅐dsDNA was 2.0, 2.4, and 2.9 Å, respectively. The refinement statistics for each complex are presented in Table I. In the case of the H103A E9 DNase bound to a single Zn 2ϩ ion, two molecules of the metal-bound complex were present in the asymmetric unit, whereas in both DNA⅐metal complexes the enzyme was bound to a single 8-mer duplex and a single metal ion, with four copies in each asymmetric unit.
The E9 DNase domain is known to exhibit substantial conformational heterogeneity around residues 20 -25 and 65-72, regions that are not part of the enzyme active site but are in close proximity to each other. This has been observed by heteronuclear NMR for the unbound protein as two slowly interconverting conformers (31), with this heterogeneity persisting in the Im9-bound form (32), and for the E9 DNase⅐Im9 complex crystallographically by virtue of the high B-factors for these regions relative to the rest of the structure (30). The temperature factors for the same regions are also high in the present structures of H103A E9 DNase bound to Zn 2ϩ and the enzyme⅐DNA complex bound to either Zn 2ϩ or Mg 2ϩ . In addition, the N-terminal 3-6 residues of all monomers in both protein⅐DNA complexes have high B-factors. Moreover, monomer A in the H103A E9 DNase⅐dsDNA⅐Zn 2ϩ complex shows higher B-factors than the other monomers, particularly in the conformationally flexible regions. Finally, residues 35-47 and 66 -75 were not well defined in the electron density and so were not included in the final model for this monomer only. In summary, the complexes of H103A E9 DNase bound to metal ions and dsDNA display several disordered regions, most of which have been documented previously. These affect the general quality of the model, reflected in the slightly high R factor/ R free (Table I).
The overall structure for the Mg 2ϩ -bound dsDNA complex of H103A E9 DNase is shown in Fig. 1A, with representative electron density for this complex presented in Fig. 1B. As reported previously, the E9 DNase is a mixed ␣/␤ protein, with its 34-amino acid H-N-H motif located at the extreme C terminus of the enzyme (Fig. 1A, dark green) (6,30). The H-N-H motif (residues 100 -134), along with residues 94 -100 that are part of a preceding loop, forms the principle contacts with the DNA although other regions of the protein, at positions 5, 51, and 54, also interact with the nucleic acid. The structure for H103A E9 DNase bound to Zn 2ϩ in the present work is very similar to that of the Ni 2ϩ -bound wild type enzyme (the root mean square deviation for C␣ atoms is 0.7 Å). Binding of dsDNA to the H103A E9 DNase, loaded with Zn 2ϩ or Mg 2ϩ in the H-N-H motif, also only resulted in small structural changes to the enzyme with the root mean square deviations for C␣ atoms relative to the Zn 2ϩ -bound enzyme being 0.6 and 0.8 Å, respectively. This is in stark contrast to the DNA bound at the active site, the minor groove of which is distorted relative to B-DNA of the same sequence (Fig. 1A).
In addition to the changes induced in the DNA by binding to the H-N-H motif of the E9 DNase, packing of the protein monomers within the asymmetric unit disrupts the base pairing at one end of the bound DNA duplex (Fig. 2). This results in the flipping-out of the terminal bases of both strands to yield final conformations that are stabilized by contacts with neighboring monomers. These changes, although most likely linked to the distortions induced by binding to the H-N-H motif and the destabilization of the duplex that ensues, are unlikely to be mechanistically relevant.
Protein-Nucleic Acid Contacts between the Metal-bound H103A E9 DNase and dsDNA-A direct consequence of the unyielding nature of the V-shaped active site is that in order for a scissile phosphate of B-DNA to reach the metal center of the H-N-H motif it must undergo some distortion; this is borne out by the current structures (Fig. 1A). The ␣-helix (and accompanying metal ion) of the E9 DNase H-N-H motif is positioned parallel to the helical axis of the DNA minor groove, with residues from the preceding six amino acids inserted into the groove itself. Together, these regions contact predominantly the phosphate backbone and cause a significant bend toward the major groove and widening of the minor groove (from 5.9 to ϳ9 Å; see Fig. 1A). We note that in the case of the E7 DNase binding to the same DNA sequence (13) this central distortion is essentially equivalent to that induced by the E9 DNase, demonstrating that it is independent of metal ion binding and is an intrinsic property of the H-N-H motif itself. This also emphasizes that the frayed end of the duplex seen in the present structures (but not seen in the E7 structure) is due to crystal packing forces.
One of the striking observations from a comparison of the Zn 2ϩ and Mg 2ϩ structures for dsDNA bound to H103A E9 DNase is the similarity of interactions made by the enzyme to the nucleic acid. In both the Mg 2ϩ and Zn 2ϩ H103A E9 DNase complexes the same scissile phosphate oxygen atom (P6) between T 5 and C 6 is coordinated by the metal ion, with the same hydrogen-bonding interactions being made toward the surrounding phosphates (Fig. 3). Hence, although Mg 2ϩ ions, but not Zn 2ϩ ions, support DNA cleavage by the E9 DNase the global features of their respective complexes with nucleic acid are essentially identical. However, subtle differences do exist between the two complexes that explain the differing activities of the metal ions, which are largely restricted to the vicinity of the metal ion. We return to this issue below.
Given the poorer resolution (2.9 Å) of the data for the Mg 2ϩbound complex the following is a summary of the main contacts to dsDNA in the Zn 2ϩ complex (2.4 Å). A total of ϳ700 Å 2 of protein surface area is buried on complexation. The distortion of the DNA is because of the insertion of the Arg-96-Glu-100 salt bridge into the minor groove of the DNA, causing it to bend and widen appreciably. This conformation is stabilized by a network of hydrogen bonds between the DNA and residues Arg-5, Arg-54, and Asp-51 that straddle the scissile phosphodiester at P6 (Fig. 3). The E7 DNase makes a similar set of interactions to dsDNA, although interestingly these are centered around a different scissile bond (P4). Arg-5 and Arg-54 make three hydrogen bonds to DNA phosphates P6 and P7. Arg-96 forms a hydrogen bond to the hydroxyl of a base on the opposing strand (T 13 ) in the minor groove (Arg-96 N1 with T 13 O 2 ). The equivalent residue in the E7 DNase makes a watermediated contact with a guanine base (13). Asp-51 of the E9 DNase also makes a base-specific hydrogen bond (Asp-51 OD2 with C 8 N4), as well as engaging in a salt bridge with Arg-54 (Fig. 3). The hydrogen bonds either side of the scissile phos- phodiester in the E9 DNase complex are linked to each other by virtue of the salt bridge between Arg-96 and Glu-100. In the Zn 2ϩ -bound structure, Glu-100 orients His-127, a critical ligand to metal ion, although this interaction is less clear-cut in the Mg 2ϩ -bound structure, with the metal in each case contacting the scissile phosphate at P6 (Fig. 3). In summary, a network of bonds from the H-N-H motif of the E9 DNase ensures that deformation of dsDNA is linked to the approach of the scissile phosphodiester to the metal ion embedded within the motif. Consistent with the important roles of Arg-5, Arg-54, Arg-96, and Glu-100 in this deformation, alanine mutants at these positions abolish DNase activity toward dsDNA (4). 2

Comparison of Alkaline Earth and Transition Metal Ion Binding Sites in the Colicin E9 H-N-H Motif-
The present study demonstrates that in the presence of DNA the H-N-H motif binds a single divalent cation. Single metal coordination has also been observed for the His-Cys enzyme I-PpoI (16), the DNA junction resolving enzyme T4 endonuclease VII (33), and the nonspecific endonuclease Vvn (15). Because all are mem-bers of the ␤␤␣-Me superfamily, it is clear that this is the norm for ␤␤␣-Me enzymes and that this contrasts the two-cation mechanisms prevalent in, for example, type II restriction enzymes (2). Based on the properties of the E9 DNase we have proposed that its H-N-H active site is capable of binding Mg 2ϩ or Ca 2ϩ (both can be used by the enzyme) as well as transition metal ions such as Zn 2ϩ (9). The histidine-rich nature of H-N-H enzymes is compatible with their ability to bind transition metal ions. Such a site is not, however, optimal for Mg 2ϩ binding in enzymes, especially endonucleases, where oxygen ligands are generally utilized for the coordination of the metal ion (34).
To ascertain the presence of the individual metal ions in our structures and more critically evaluate the conformation of surrounding residues, omit maps were calculated (Fig. 4). The metal ion within the H103A E9 DNase⅐dsDNA⅐Zn 2ϩ complex is held in tetrahedral geometry, with three protein ligands (His-102, His-127, and His-131) and a phosphate oxygen atom from the scissile bond (Fig. 4A). An identical tetrahedral arrangement is seen for Zn 2ϩ bound to the H103A E9 DNase in the absence of DNA, where the phosphate oxygen is replaced by water. The three imidazole nitrogen-Zn 2ϩ distances (2.0 -2.1 Å) 2   are consistent with those of other transition metal ions bound to enzyme active sites. By contrast, the distance between the Zn 2ϩ and the phosphate oxygen is shorter, at 1.9 Å. This is the same as that seen for the Zn 2ϩ -O bond in the structure of the E7 DNase bound with Zn 2ϩ and phosphate (instead of DNA) at its H-N-H motif (12). Similar Zn 2ϩ -O bond distances of ϳ2 Å have been reported previously, for example in the E. coli repair enzyme endonuclease IV (35).
The major differences between the H103A E9 DNase⅐-dsDNA⅐Zn 2ϩ and H103A E9 DNase⅐dsDNA⅐Mg 2ϩ complexes are localized to the metal centers. Although all four molecules contained metal ion in the DNA⅐Zn 2ϩ complex, only two of the four copies of the H103A E9 DNase⅐dsDNA⅐Mg 2ϩ complex in the asymmetric unit (molecules A and C) contained density consistent with a Mg 2ϩ ion (Fig. 4B). The unambiguous assignment of Mg 2ϩ in endonuclease active sites is often difficult because of the weak nature of the binding. Mg 2ϩ does not bind to the E9 DNase in the absence of DNA, and the high concentrations of Mg 2ϩ required for optimal activity (10 -20 mM) (8) further suggest that binding will also be weak even in the presence of DNA. This being the case, we have to consider that the density adjacent to the H-N-H histidines in the Mg 2ϩ ⅐DNA complex may be a water molecule. We suggest that this is unlikely for four reasons. First, the H103A E9 DNase⅐dsDNA⅐Mg 2ϩ crystals were grown in a high concentration of Mg 2ϩ ions (100 mM). Second, Kuhlmann et al. (30) have determined the structure for apo-E9 DNase, with a water molecule replacing the metal ion in the H-N-H motif. The structure shows that the conformations and hydrogen-bonding interactions of the H-N-H histidines are different from those seen in the current structure. Third, the hydrogen-bonding interactions of the molecule seen in the H-N-H motif are not compat- The first nucleotide (G 1 ) is not visible in any of the electron densities. The bases at positions 2 and 16 (C 2 and C 16 ) are rotated ϳ180°away from the helical axis, their highly distorted positions stabilized by contacts with neighboring monomers B and D (cream and blue, respectively). Additional stabilization of these distorted conformations is through direct DNA-DNA hydrogen bonds between neighboring complexes (not shown). All monomers in both Zn 2ϩ and Mg 2ϩ complexes make equivalent crystal contacts.

FIG. 3. A hydrogen-bonding network connects DNA distortion to approach of the scissile bond to the H-N-H motif metal center.
Stereo representation of the main hydrogen-bonding interactions between the E9 DNase and the DNA minor groove in the Zn 2ϩ complex, with identical interactions observed in the Mg 2ϩ -bound complex. The DNA and protein residues are cream and green, respectively, the Zn 2ϩ ion is magenta, and hydrogen bonds are plotted as dotted lines. The scissile bond of the DNA can only approach the metal ion if the DNA is distorted. This is accomplished by the insertion of the Arg-96-Glu-100 salt bridge into the minor groove and stabilization of this configuration through a number of hydrogen-bonding interactions to the DNA phosphates through residues Arg-5, Asp-51, and Arg-54.
ible with those of a water molecule. Fourth, the temperature factors for the Mg 2ϩ ion are of the same order as those of surrounding atoms.
Although the Mg 2ϩ ion in the H103A E9 DNase⅐dsDNA⅐Mg 2ϩ complex is present in monomers A and C, the details of its coordination in monomer C are better resolved. The position of this metal ion is displaced by ϳ1 Å relative to the metal ion in the DNA⅐Zn 2ϩ structure, which is greater than the root mean square deviation between the two complexes (Fig. 5). The coordination of this metal ion is clearly very different from that of the Zn 2ϩ ion in the H103A E9 DNase⅐dsDNA⅐Zn 2ϩ complex. However, at the current level of resolution only five of the anticipated six ligands can be seen to contact the Mg 2ϩ ion. Equatorial positions are taken by two oxygen atoms from the scissile phosphate (O1P and O3Ј) and two nitrogen atoms from the H-N-H histidine residues (His-102 and His-127), with one of the axial positions taken by a water molecule (Fig. 4B). We presume that the sixth position, likely to be another water molecule, is not resolved in the current structure. To our knowledge, this is the first time that the coordination shell of a Mg 2ϩ ion in an enzyme active site has been shown to include two histidine residues, although we note that similar bis-His coordination is seen for manganese ions binding in the DNA repair nuclease Mre11 from Pyrococcus furiosus (36). Intriguingly, His-131, the third li-gand in the Zn 2ϩ complex, moves by Ͼ2 Å and is replaced by a water molecule that resides midway between it and the Mg 2ϩ ion (Fig. 4B). We note that although only two of the four monomers in the H103A E9 DNase⅐dsDNA⅐Mg 2ϩ structure contain magnesium, water does not replace the metal ion in the apparently vacant monomers. Furthermore, the conformations of the three critical histidine residues (His-102, His-127, and His-131) are very similar in all four monomers, especially that of His-131, which is in the same displaced conformation. This suggests that a magnesium ion could be present in all monomers, but its occupancy is either too low or the B-factors too high to be seen in all but two of them.
Although unusual, there is precedent for the ligation of Mg 2ϩ by histidines in enzymes. Dudev and Lim (34) have pointed out that although tetrahedrally oriented histidine active sites ligate alkaline earth metal ions poorly, Mg 2ϩ binding could be accomplished if a negatively charged side chain is incorporated to create a site with octahedral geometry. The scissile phosphate oxygen atoms in the DNA-bound complex provide this negative charge. Such mixed coordination of Mg 2ϩ , as well as Ca 2ϩ ions, has been seen in other enzymes, notably xylose isomerase and diisopropylfluorophosphatase (37,38). In xylose isomerase, one of two Mg 2ϩ ions is coordinated by a histidine residue as well as carboxylate side chains, whereas in diisopropylfluorophosphatase one of two Ca 2ϩ ions is similarly coordinated. The metal-imidazole distances in both enzymes are ϳ2.6 Å; those for His-102 and His-127 to the Mg 2ϩ in the present structure are 2.5 and 2.2 Å, respectively.
In the original description of the Mg 2ϩ -dependent mechanism of the H-N-H class of endonucleases, Pommer et al. (9) postulated that a single histidine residue (His-127) contacted the metal ion, by analogy with the His-Cys/␤␤␣-Me enzyme I-PpoI that binds a single Mg 2ϩ ion in the presence of DNA (16). This ligand is an asparagine in I-PpoI (Asn-119), and its position is equivalent to His-127 (14). With the recent publication of the structure for Vvn endonuclease (15), it is now evident that ␤␤␣-Me enzymes fall into two subgroups, those utilizing a single protein ligand to the Mg 2ϩ (or Ca 2 ), as seen in I-PpoI and Serratia, and those with two ligands, as seen in Vvn where asparagine and glutamic acid bind the metal ion. We now see that H-N-H enzymes such as the E9 DNase fall into the latter group; the two residues are histidines, with their positions superimposable on the two Mg 2ϩ -coordinating amino acids of Only the phosphate coordinated to the metal ion in each structure is shown. Atoms are color-coded for the different complexes. White, H103A E9 DNase⅐dsDNA⅐Zn 2ϩ complex; black, H103A E9 DNase⅐dsDNA⅐Mg 2ϩ complex; gray, wild type E9 DNase bound with Ni 2ϩ and a phosphate molecule. The transition metal ions superimpose with each other, whereas the Mg 2ϩ ion is displaced by ϳ1 Å. His-131 is in a different location in each complex, with the largest change occurring in the Mg 2ϩ complex where a water molecule (not shown) resides midway between the metal ion and the imidazole ring (see Fig. 4B).
Vvn. The counterparts of His-102 and His-127 of the E9 DNase are Glu-79 and Asn-127 in Vvn. Interestingly, sequence alignments of H-N-H motif-containing enzymes show that the two metal-coordinating histidines of the E9 DNase are occasionally substituted for oxygen-bearing amino acids, such as asparagine, glutamine, and aspartic acid.
Proposed Mechanism for Mg 2ϩ -dependent Cleavage of H-N-H Enzymes and Why Zn 2ϩ Is Inactive-The Mg 2ϩ ion in I-PpoI serves to polarize the scissile phosphate, stabilize the phosphoanion transition state, and activate a water molecule that protonates the leaving group (16). We had proposed previously that for Mg 2ϩ -based activity of the E9 DNase the metal ion would serve two of these roles, with that of general acid being taken by His-102 (9). However, it is now clear that this residue, along with His-127, is engaged in Mg 2ϩ coordination. Because there are no other protein side chains within the vicinity of the scissile bond, it seems more likely that this role is taken by a water molecule, possibly that which sits midway between the Mg 2ϩ ion and His-131 (Fig. 4B). His-103 (mutated to alanine in the present study) almost certainly serves as the general base in the activation of the hydrolytic water molecule, the corresponding histidine in I-PpoI being His-98. When His-103, from the Ni 2ϩ ⅐phosphate complex of the E9 DNase (6), is superimposed onto the present structures its position is appropriate for such a role (Fig. 5). Moreover, a water molecule is correctly positioned in the E9 DNase⅐Zn 2ϩ ⅐dsDNA complex to act as the attacking nucleophile to the phosphodiester (data not shown).
We now turn to the question of the transition metal-dependent activity of the H-N-H motif within the E9 DNase and the controversy surrounding zinc. We had proposed previously that for activity in the presence of transition metals two non-protein-bound coordination sites at the metal center were required, one for ligation of the scissile phosphate oxygen and the other for the activation of a water molecule for leaving group stabilization (9). The present structure for the Zn 2ϩ -bound enzyme shows that this is not possible because three histidine side chains tightly coordinate the metal ion (Fig. 4A). Although we do not have a structure for an active transition metal such as Ni 2ϩ bound to an E9 DNase⅐dsDNA complex, the structure for the E9 DNase bound with Ni 2ϩ and a phosphate molecule is informative (6). In this case, the third metal-ligating histidine, His-131, disengages from the metal ion, with its distance lying midway between that seen in the Mg 2ϩ and Zn 2ϩ structures of the present study (Fig. 5). This may allow a water molecule to replace the histidine and so serve as a protonating device for DNA hydrolysis, in other words by a similar mechanism to that proposed for Mg 2ϩ -based activity.
In conclusion, the current work has provided a structural basis for understanding why high affinity zinc binding inactivates the H-N-H motif and has helped explain why only weakly bound transition metals are active in the motif. It has also provided the first glimpse as to how the motif binds and utilizes Mg 2ϩ ions to cleave dsDNA by a mechanism that likely underpins the action of all known H-N-H/␤␤␣-Me endonucleases. The current structures of the H-N-H motif support previous work on this class of enzyme that suggested the motif to be an adaptable catalytic center, able to utilize both alkaline earth and transition metal ions. Contrary to expectations, however, the mechanism of DNA cleavage for the different classes of metal ions may be the same.