Activity and structure of Pseudomonas putida MPE, a manganese-dependent single-strand DNA endonuclease encoded in a nucleic acid repair gene cluster

A recently identified and widely prevalent prokaryal gene cluster encodes a suite of enzymes with imputed roles in nucleic acid repair. The enzymes are as follows: MPE, a DNA endonuclease; Lhr-Core, a 3′–5′ DNA helicase; LIG, an ATP-dependent DNA ligase; and Exo, a metallo-β-lactamase-family nuclease. Bacterial and archaeal MPE proteins belong to the binuclear metallophosphoesterase superfamily that includes the well-studied DNA repair nucleases Mre11 and SbcD. Here, we report that the Pseudomonas putida MPE protein is a manganese-dependent DNA endonuclease that incises either linear single strands or the single-strand loops of stem-loop DNA structures. MPE has feeble activity on duplex DNA. A crystal structure of MPE at 2.2 Å resolution revealed that the active site includes two octahedrally coordinated manganese ions. Seven signature amino acids of the binuclear metallophosphoesterase superfamily serve as the enzymic metal ligands in MPE: Asp33, His35, Asp78, Asn112, His124, His146, and His158. A swath of positive surface potential on either side of the active site pocket suggests a binding site for the single-strand DNA substrate. The structure of MPE differs from Mre11 and SbcD in several key respects: (i) MPE is a monomer, whereas Mre11 and SbcD are homodimers; (ii) MPE lacks the capping domain present in Mre11 and SbcD; and (iii) the topology of the β sandwich that comprises the core of the metallophosphoesterase fold differs in MPE vis-à-vis Mre11 and SbcD. We surmise that MPE exemplifies a novel clade of DNA endonuclease within the binuclear metallophosphoesterase superfamily.

Nucleases, helicases, and ligases are indispensable agents of DNA repair found in all taxa. These enzymes come in various flavors that have distinctive structures and biochemical specificities. It is often the case that multiple nucleases, helicases, and ligases coexist in the same organism, whereby the enzyme paralogs or homologs specialize in the repair of particular types of DNA damage. In prokarya, the clustering of genes encoding nucleases, helicases, or ligases in operons can provide strong clues to the existence of a coherent repair pathway. In this vein, we recently identified a cluster of four genes encoding predicted nuclease, helicase, and ligase enzymes that is widely prevalent in diverse bacteria from distinct phyla (1). The four enzymes are as follows: (i) Lhr-Core, a 3Ј-5Ј DNA helicase; (ii) MPE, a DNA endonuclease of the binuclear metallophosphoesterase family; (iii) Exo, a putative exonuclease of the metallo-␤-lactamase family; and (iv) an ATP-dependent DNA ligase. Many other bacteria and archaea have a two-gene cluster encoding Lhr-Core and MPE (1).
To understand whether and how these enzymes might contribute to nucleic acid repair, we undertook previously to purify and characterize the Lhr-Core and MPE proteins specified by the Exo⅐Lig⅐Lhr-Core⅐MPE gene cluster of Pseudomonas putida (1). The 816-amino acid P. putida Lhr-Core polypeptide is homologous to the core ATPase/helicase domain of the 1507amino acid Mycobacterium smegmatis Lhr protein (2,3). We reported that P. putida Lhr-Core is a single-strand DNA-dependent ATPase/dATPase (K m ϭ 0.37 mM ATP; k cat ϭ 3.3 s Ϫ1 ), an ATP-dependent 3Ј-to-5Ј single-strand DNA translocase, and an ATP-dependent 3Ј-to-5Ј helicase. Lhr-Core unwinds 3Ј-tailed duplexes in which the loading/tracking strand is DNA and the displaced strand is either DNA or RNA (1).
The 216-amino acid P. putida MPE protein is a member of the metallophosphoesterase superfamily of enzymes that utilize a binuclear transition-metal ion center to catalyze phosphomonoester or phosphodiester hydrolysis (4). We reported that P. putida MPE is a manganese-dependent phosphodiesterase that released p-nitrophenol from bis-p-nitrophenyl phosphate (k cat ϭ 212 s Ϫ1 ) and p-nitrophenyl-5Ј-thymidylate (k cat ϭ 34 s Ϫ1 ) but displayed no detectable phosphomonoesterase activity against p-nitrophenyl phosphate. MPE is also a manganese-dependent DNA nuclease that sequentially converted a closed circle plasmid DNA to nicked circle and linear forms prior to converting the linear DNA to smaller fragments (1). This initial characterization of MPE suggested that it might be a new bacterial homolog of the archaeal and eukaryal DNA nuclease Mre11 and the homologous bacterial DNA nuclease SbcD, which are also binuclear metallophosphoesterases (5)(6)(7)(8)(9)(10)(11)(12).
In the present study, we examine in greater detail the nuclease activity of P. putida MPE and find it to be a manganese-dependent single-strand DNA endonuclease that incises either linear single strands or the single-stranded loops of DNA stem-loop structures. MPE has feeble activity on duplex DNA.
We determined crystal structures of MPE in two different space groups at 2.0 -2.2 Å resolution. The MPE active site includes two octahedrally coordinated manganese ions. The defining amino acids of the metallophosphoesterase superfamily serve as the enzymic metal ligands. We discuss how the structure of MPE differs significantly from Mre11 and SbcD.

P. putida MPE incises single-strand DNA
A 30-min reaction of 1 pmol (0.1 M) of 5Ј 32 P-labeled 50-mer single-strand DNA oligonucleotide with 1 pmol of MPE in the presence of 1 mM manganese resulted in incision of a fraction of the input 50-mer DNA to form a ladder of smaller 5Ј 32 P-labeled DNAs that were resolved by urea-PAGE (Fig. 1). One cleavage product was especially prominent at limiting enzyme (denoted by an arrowhead in Fig. 1). Increasing the input MPE to 2.5, 5, and 10 pmol resulted in progressive depletion of the 50-mer substrate, which was complete at 10 pmol of MPE ( Fig. 1). By contrast, when 10 pmol of MPE was reacted with 1 pmol of 5Ј 32 P-labeled 50-mer duplex DNA, the majority of the labeled strand was intact, and there was only scant formation of shorter cleavage products (Fig. 1).
To gauge the effect of DNA strand length on endonuclease activity, we reacted 2.5 and 10 pmol of MPE with 1 pmol of 5Ј 32 P-labeled 50-, 40-, 30-, 20-, and 10-mer single-strand DNA oligonucleotides of identical 5Ј-terminal nucleobase sequence (Fig. 2). The trend seen was that shortening the DNA resulted in less effective endonuclease activity, as reflected in the level of residual uncleaved substrate at 2.5 pmol of MPE. The transition from 20-to 10-mer elicited a sharp decrement in activity, such that there was only scant cleavage of the 10-mer at 10 pmol of input MPE (Fig. 2). From the cleavage ladders in Fig. 2, we could identify the prominent cleavage product generated by reaction of 2.5 pmol of MPE with the 50-mer DNA as a 15-nucleotide fragment resulting from incision after a dG nucleotide at the tip of a palindromic sequence, 5Ј-GGTACCCG2GGGATCC, that has the potential to form an intramolecular hairpin.

MPE incises the single-strand loop of DNA stem-loop structures
In the experiment in Fig. 3, MPE (10 pmol) was reacted in the presence of manganese with 1 pmol of a 56-nucleotide 5Ј 32 Plabeled DNA stem-loop substrate consisting of a self-complementary 18-bp stem and a 20-nucleotide single-strand loop. In a parallel reaction, the 56-mer stem-loop substrate was incubated with M. smegmatis 5Ј-flap endonuclease/5Ј-exonuclease FenA, which cleaves the DNA principally at the first duplex phosphodiester adjacent to the 5Ј-flap (to yield a 5Ј 32 P-labeled 39-mer product) and at the 5Ј-terminal phosphodiester (to release a mononucleotide product) (13). The MPE and FenA reaction products were analyzed by urea-PAGE alongside a 5Ј 32 P-labeled 18-mer corresponding the 5Ј-terminal 18-nucleotide segment of the stem-loop DNA. We found that 10 pmol of MPE incised the stem-loop within the single-strand loop to yield a cluster of about 20 5Ј 32 P-labeled products migrating between the FenA flap cleavage product and the 18-mer marker (Fig. 3). 5Ј 32 P-labeled products migrating faster than the 18-mer were also present.
We proceeded to test two other 5Ј 32 P-labeled stem-loop structures, of 46 and 41 nucleotides, respectively, composed of the same 18-bp duplex stem segment but with shorter 10-nucleotide or 5-nucleotide single-strand loops (Fig. 3). MPE cleaved the 46-mer DNA primarily within the loop to form a cluster of about 10 5Ј 32 P-labeled products migrating between the FenA flap cleavage product (in this case a doublet) and the 18-mer marker. MPE cleaved the 41-mer DNA within the loop

Activity and structure of P. putida MPE
to generate a cluster of five 5Ј 32 P-labeled products extending upward from the 18-mer marker. (Note that FenA was unable to cleave the 5Ј-flap when the loop was shortened to 5 nucleotides.) Products shorter than 18 nucleotides were also formed during the MPE reaction with the 46-mer and 41-mer stemloop DNAs.

Characterization of the MPE endonuclease
Further studies of MPE activity were performed with the 56-nucleotide 5Ј 32 P-labeled stem-loop DNA substrate. An enzyme titration experiment in Fig. 4A showed that at limiting levels of MPE (0.5 and 1 pmol), at which not all of the input substrate was consumed, the cleavage events were confined to, and relatively evenly distributed among, the internucleotide linkages in the 20-nucleotide loop segment. As the enzyme was increased, and all substrate was cleaved at least once, the distribution of cleavage events within the loop shifted to yield shorter 5Ј-radiolabeled products, and there also appeared cleavages within the 5Ј-duplex segment to form labeled products of less than 18 nucleotides. A similar pattern was observed when we followed the temporal profile of 56-mer stem-loop cleavage by 10 pmol of MPE (Fig. 4B), whereby at early times when a frac-

Activity and structure of P. putida MPE
tion of the input DNA was consumed, the enzyme incised at all sites within the loop segment, and at later times, when all substrate was cleaved at least once, the loop cleavage product distribution shifted to smaller size, and cleavages within the 18-nucleotide 5Ј duplex segment became evident. These results affirm MPE as an endonuclease that preferentially cleaves single-strand DNA. The absence of cleavages within the 3Ј 18-nucleotide stem duplex at early reaction times and the paucity of cleavages near the 5Ј terminus of the 56-mer DNA at saturating enzyme or at late times indicate that MPE is not an exonuclease that sequentially cleaves single nucleotides from the termini.
The divalent cation specificity of the MPE endonuclease was examined in the experiment in Fig. 4C, in which reactions were performed either in the absence of added metal or in the presence of 1 mM calcium, cadmium, cobalt, copper, magnesium, manganese, nickel, or zinc. No loop incision was detected in the absence of metal, and the metal requirement was satisfied optimally by manganese. Nickel was less active, as gauged by the extent of consumption of the 56-mer DNA. Cadmium and cobalt supported trace levels of loop cleavage. Calcium, copper, magnesium, and zinc were inactive (Fig. 4C). Additional insights into the metal specificity of MPE were provided by mixing experiments, in which reactions containing 1 mM manganese were supplemented with a 1 mM concentration of another divalent cation (Fig. 5). The mixture of nickel and manganese reduced endonuclease activity to the level of a reaction containing nickel only. Mixing manganese with cadmium or cobalt resulted in trace activity comparable with that of cadmium or cobalt alone. Copper and zinc abolished nuclease activity in the presence of manganese. These results suggest that each of these four "soft" metals might out-compete manganese for one or both metal-binding sites on the enzyme, wherein engaged they are either unable to support phosphodiesterase reaction chemistry (copper and zinc) or support a lower level of activity vis-à-vis manganese (nickel, cadmium, and cobalt). By contrast, magnesium and calcium had no such deleterious effect in combination with manganese, implying that these "hard" metals do not bind effectively to the MPE active site.

Activity and structure of P. putida MPE Crystal structure of P. putida MPE
Crystals grown from a solution of 0.5 mM MPE and 5 mM MnCl 2 had two distinct morphologies: a hexagonal form (crystal form 1) that diffracted to 2.0 Å resolution in space group P6 1 22 and a rodshaped form (crystal form 2) that diffracted to 2.2 Å resolution in space group P2 1 2 1 2 ( Table 1). The structure of MPE in the form 1 crystal was solved by SIRAS phasing on manganese (R work /R free ϭ 0.192/0.243). The structure of MPE in the form 2 crystal was then solved by molecular replacement using the structure of the crystal form 1 as a search model (R work /R free ϭ 0.204/0.264). Both crystal forms contained a single MPE protomer and two manganese ions in the asymmetric unit. MPE was monomeric in each crystal. The MPE folds in the two crystal forms superimposed with a z score of 33.4 and an RMSD 2 of 1.2 Å at 194 C␣ positions, as determined by pairwise comparison in DALI (14). One of the manganese ions in the form 1 crystal was situated on a crystallographic 2-fold axis in a position that we construed to be off-pathway (to be discussed below). Because both manganese ions in the form 2 crystal were located in the enzyme active site, we will focus henceforth on the form 2 structure of MPE.
The MPE tertiary structure, shown in stereoview in Fig. 6A, consists of 16 ␤ strands, three ␣-helices, and three 3 10 helices. The secondary structure elements are displayed above the MPE amino acid sequence in Fig. 6B. The strands are organized into an extended ␤ sandwich. The topologies of the two ␤ sheets of the sandwich are ␤12⅐␤21⅐␤32⅐␤41⅐␤51⅐␤61⅐␤71 and ␤82⅐␤91⅐␤101⅐␤131⅐␤122⅐␤151⅐␤162. An additional three-strand ␤ sheet (␤122⅐␤111⅐␤141) emanates from the top surface of the second sheet of the ␤ sandwich (Fig. 6A). The active site containing the binuclear manganese cluster (green spheres) is formed by constituents of the interstrand loops on the forward surface of MPE (Fig. 6A). The seven amino acids that bind the metals, Asp 33 , His 35 , Asp 78 , Asn 112 , His 142 , His 156 , and His 158 , are depicted as stick models (Fig. 6A) and denoted by red dots below the amino acid sequence (Fig. 6B).

MPE active site and surface electrostatics
A stereoview of the active site highlighting the binuclear metal complex is shown in Fig. 7A. The metal-binding mode is  We showed previously that mutation of Asp 78 to alanine squelches MPE phosphodiesterase activity with the generic substrate bis-p-nitrophenyl phosphate and its endonuclease activity on supercoiled plasmid DNA (1). The metal-bridging water may correspond to the nucleophile in the phosphodiesterase reaction, insofar as its proximity to the two metals will significantly lower the pK a of the water and activate it for attack on the DNA backbone. In the same vein, we speculate that the other two waters of the M1 and M2 coordination spheres occupy the positions of two of the phosphate oxygens of the scissile phosphodiester. The His 113 side chain located above the metal complex (Fig. 7A) is con-served in many phosphodiesterase/nuclease enzymes of the metallophosphoesterase superfamily; this histidine serves as a general acid catalyst of DNA cleavage, by donating a proton to the 3Ј-OH leaving group. A surface electrostatic model of MPE highlights a swath of positive potential on either side of the negatively charged metal-binding pocket (Fig. 7B). The positive surface suggests a binding site for the negatively charged backbone of the single-strand DNA substrate of the MPE endonuclease reaction. MPE amino acids that contribute to the positive surface include (proceeding from top to bottom in the view in Fig. 7B) Arg 86 , Arg 83 , His 81 , Lys 38 , Arg 115 , Arg 172 , Arg 170 , and Arg 164 . Of these, Arg 115 , His 81 , and Lys 38 are located near the active site (Fig. 7A).

MPE crystal form 1
The structure of MPE in space group P6 1 22 had several notable features from the crystallographic standpoint. First, Figure 6. Overview of MPE crystal structure. A, stereoview of the MPE tertiary structure in crystal form 2, depicted as a cartoon model with magenta ␤ strands, cyan ␣-helices, and blue 3 10 helices. The secondary structure elements are labeled. Manganese ions are shown as green spheres. Amino acids that bind the metal ions are depicted as stick models. B, the secondary structure elements of MPE are displayed above the amino acid sequence. The seven metal-binding amino acids are indicated by red dots.

Activity and structure of P. putida MPE
whereas it clearly contained two manganese ions in the asymmetric unit, as affirmed by the anomalous difference peaks over the metal atoms (Fig. 8, A and B), one of the metals was located on a 2-fold crystallographic symmetry axis (the z axis looking down on the images in Fig. 8) and was coordinated by ligands in neighboring MPE protomers on either side of the axis. The axial manganese (labeled Mn2 in Fig. 8B) was octahedrally coordinated to two His 156 side chains (via N⑀, rather than by N␦ that engages Mn2 in the form 2 structure), two Glu 143 side chains (which is not a metal ligand in the form 1 structure), two waters bridged to Glu 143 , and two waters on the symmetry axis (Fig.  8B). Asn 112 and His 142 , which are Mn2 ligands in the form 2 structure, do not interact with a metal in crystal form 1. The other manganese, equivalent to Mn1 in the form 2 MPE structure described above, was located in the active site in an octahedral coordination complex with Asp 33 -O␦, His 35 -N⑀, Asp 78 -O␦, His 158 -N⑀, and two waters. We surmise that an off-pathway complex of the Mn2 ion aided in the formation of the hexagonal crystal lattice.
Second, the main difference in the MPE fold in the two crystal forms entailed a shift of the 85 ARAP 88 loop segment immediately preceding the ␤5 strand (by 5.9 Å at the Pro 88 C␣ atom, 6.9 Å at the Ala 87 C␣ atom, 7.4 Å at the Arg 86 C␣ atom, and 6.2 Å at the Ala 85 C␣ atom). Third, Pro 120 in the loop following the ␣3 helix adopts a trans conformation in the form 2 crystal versus a cis conformation in the form 1 crystal.
The relationship of MPE to Mre11 and SbcD is pertinent because the latter two are manganese-dependent DNA nucleases. Mre11 is the nuclease subunit of a DNA end-processing complex, which also includes an SMC-like ATPase subunit, Rad50, that functions in cellular pathways of double-strand break repair (e.g. homologous recombination and nonhomologous end-joining) (reviewed in Ref. 18). SbcD and its SMC-like ATPase partner SbcC form an SbcCD complex (19), a homolog of Mre11⅐Rad50, that triggers replication-dependent double-strand breaks at palindromic sequences in the bacterial chromosome (20). The  Fig. 6A). A third manganese ion is situated at the crystallographic symmetry axis. B, close-up view of the active sites and metal contacts in the symmetry-related MPE protomers (colored as in A). Waters are depicted as red spheres. Anomalous difference density for the manganese atoms, contoured at 5, is shown in pink mesh.

Activity and structure of P. putida MPE
Mre11⅐Rad50 and SbcCD complexes have ATP-dependent double-strand DNA exonuclease activity and ATP-independent single-strand endonuclease activity. Like MPE, SbcCD is adept at cleaving the single-strand loops of DNA hairpin structures (21).
A number of structural features distinguish MPE from Mre11 and/or SbcD. These are highlighted in Fig. 10, which shows a side-by-side superposition of the structures of P. putida MPE, human Mre11 (10), and E. coli SbcD (12), aligned with respect to their metal-binding sites, with the structural elements in Mre11 and SbcD that are not present in MPE shaded gray. First, the 216-aa MPE protein is smaller than the human Mre11 catalytic core (416 aa) or E. coli SbcD (400 aa). MPE lacks the so-called capping domain, comprising a ␤ sheet and two ␣-helices, that is appended to the C terminus of the metallophosphoesterase fold of Mre11 orthologs and SbcD. The capping domain is implicated in interaction of Pyrococcus Mre11 with duplex DNA (6) and of Thermotoga Mre11 with Rad50 (11). P. putida MPE and its homologs are not linked genetically with an SMC-like ATPase protein and thus may have no need for a capping domain.
Second, the monomeric quaternary structure of MPE in solution (as gauged by gel filtration (1)) and in both crystal forms reported here (as analyzed in PISA) contrasts with the homodimeric quaternary structures of Mre11 and SbcD (5)(6)(7)(8)(9)(10)(11)(12). The homodimerization interface of SbcD and archaeal and fungal Mre11 proteins consists of a four-helix bundle formed by the ␣2 and ␣3 helices from each protomer (as indicated in Fig. 10, right). Whereas MPE has a counterpart of the ␣2 helix (albeit with no amino acid sequence identity to the SbcD ␣2 helix), it lacks an equivalent of the ␣3 helix. Human Mre11 has a distinctive homodimer interface (10) involving the ␣3 helix and several loops (Fig. 10, middle), none of which are present in MPE. Third, Mre11 and SbcD have additional ␣-helices and loops below the central ␤ sandwich that are not present in MPE (Fig. 10). Finally, the topology of the ␤ sheets that form the central sandwich differs in MPE versus Mre11 and SbcD. For example, Mre11 and SbcD have no equivalent of the ␤1 strand of MPE. The topology of the rest of this ␤ sheet (which is ␤21⅐␤32⅐ ␤41⅐␤51⅐␤61⅐␤71 in MPE) is different in human Mre11, where the strand order is ␤141⅐␤132⅐␤11⅐␤21⅐␤31⅐␤41 (10). Mre11 and SbcD also have no equivalent of the MPE ␤15 and ␤16 strands within the other sheet of the ␤ sandwich or of the ␤11 and ␤14 strands. The topology of the rest of the second sheet in the sandwich (which is ␤82⅐␤91⅐␤101⅐␤131⅐␤122 in MPE) is different in human Mre11, where the strand order and direction is ␤71⅐␤91⅐␤101⅐␤121⅐␤112 (10).

Conclusions and speculations
We surmise from the results presented above that P. putida MPE exemplifies a novel clade of bacterial DNA endonucleases within the binuclear metallophosphoesterase enzyme superfamily. MPE's activity in cleaving single-strand DNA and single-strand regions in duplex DNA is compatible with a variety of potential DNA repair functions, especially considering its signature genetic clustering with a DNA helicase and, in many taxa, a DNA ligase. We speculated previously on possible MPE roles in vivo (1). Here our focus was on characterization of the MPE nuclease activity and MPE structure. The outstanding issue for MPE (as it is still for Mre11 and SbcD (18)) is how the scissile DNA strand is engaged in the nuclease active site to promote phosphodiester hydrolysis via a proper orientation of water nucleophile and DNA leaving group. The exposed positive electrostatic surface above the MPE metal-binding site provides a likely landing pad for single-strand DNA, but the structure solved here is not instructive with respect to the trajectory of the DNA across the enzyme or any potential conformational changes in MPE that might be triggered upon DNA binding. With respect to the latter point, we speculate that the surface disordered loop 43 RALHQPVPR 51 between ␤4 and ␣1, starting next to the Tyr 42 residue (denoted by the asterisk in the surface view in Fig. 7B) might become ordered and encircle the scissile DNA strand after it sits down on the positive patch. Resolving these mechanistic issues will hinge on obtaining an MPE⅐(Mn 2ϩ ) 2 ⅐DNA cocrystal in a state mimetic of a Michaelis complex. This in turn depends on identifying a minimal perturbation of the enzyme or the DNA substrate that preserves metal binding and DNA binding but prevents DNA cleavage.

Purification and crystallization of P. putida MPE
Recombinant MPE was produced in E. coli as a His 10 Smt3 fusion and purified from a soluble extract by serial nickel-affinity, tag removal, and gel filtration steps as described previously (1). Protein concentration was determined with the Bio-Rad dye reagent using BSA as the standard. MPE crystals were grown at 22°C by sitting-drop vapor diffusion. The MPE protein preparation (12.5 mg/ml in 20 mM Tris-HCl, pH 8.0, 300 mM NaCl, 1 mM DTT, 10% glycerol) was adjusted to 5 mM MnCl 2 and incubated for 20 min. Aliquots of the protein solution (1 l) were then mixed with an equal volume of precipitant solution containing either 0.1 mM Tris-HCl, pH 7.4, 20% PEG-MME-2000 (crystal form 1) or 0.1 mM Tris-HCl, pH 7.2, 16% PEG-MME-2000 (crystal form 2). Crystals grew to their full size in 1 or 2 days. Single crystals were harvested and transferred briefly to precipitant solution containing 25% glycerol prior to flash-freezing in liquid nitrogen.

Diffraction data collection and structure determination
X-ray diffraction data were collected from single crystals at the Advanced Photon Source beamline 24-ID-C at a wavelength of 1.7712 Å near the manganese X-ray absorption edge. Form 1 and form 2 crystals diffracted to 2.0 and 2.2 Å resolution, respectively. Indexing and merging of the diffraction data were performed in HKL2000 (22). The form 1 crystal was in space group P6 1 22. The phases for solving the form 1 structure were obtained by single isomorphous replacement with anomalous scattering (SIRAS) as implemented in SHELX (23). Two strong manganese anomalous scattering peaks were identified per asymmetric unit containing one MPE protomer. One of the manganese atoms in the form 1 crystal was situated on a crystallographic 2-fold symmetry axis. Iterative model building was performed in O (24). Refinement was accomplished with PHE-NIX (25). The model of the MPE protomer in the form 1 crystal, refined to 2.0 Å resolution (R work /R free ϭ 0.192/0.243), consisted of three polypeptide segments (Asn 2 -Ala 40 , Thr 53 -Arg 164 , and Leu 171 -Leu 216 ) punctuated by disordered surface loops (aa 41-52 and 165-170) for which there was no interpretable electron density. The form 2 crystal was in space group P2 1 2 1 2 with one MPE protomer in the asymmetric unit. The form 2 crystal structure was solved by molecular replacement by using the form 1 structure as the search model. The model of MPE protomer in the form 2 crystal, refined to 2.2 Å resolution (R work /R free ϭ 0.204/0.264), comprised four polypeptide segments (Asn 2 -Tyr 42 , Gly 52 -Gly 165 , Arg 170 -Gly 209 , and Arg 212 -Leu 216 ) separated by disordered surface loops. Data collection and refinement statistics for both crystal forms are compiled in Table 1.

Nuclease substrates
The 5Ј 32 P-labeled single-strand DNA and stem-loop DNA substrates were prepared by reaction of synthetic oligonucleotides with T4 polynucleotide kinase (Pnk) and [␥-32 P]ATP. The kinase reaction mixture was heated to 95°C to inactivate T4 Pnk. The DNA was separated from free ATP by electrophoresis through a nondenaturing 18% polyacrylamide gel and then eluted from an excised gel slice by overnight incubation at 4°C in 200 l of 10 mM Tris-HCl, pH 7.4, 1 mM EDTA. To form the 50-mer DNA duplex, the radiolabeled pDNA strand and template DNA were annealed at a 1:1.5 molar ratio in 10 mM Tris-HCl, pH 7.4, 1 mM EDTA by incubating for 10 min at 65°C, 15 min at 37°C, and then 30 min at 22°C.

MPE DNA nuclease assay
Reaction mixtures (10 l) containing 20 mM Tris-HCl, pH 8.0, 30 mM NaCl, 1 mM DTT, 1 mM MnCl 2 , and 1 pmol (0.1 M) of 5Ј 32 P-labeled DNA oligonucleotide and MPE as specified in the figure legends were incubated at 37°C. The reactions were quenched at the times specified by adding 10 l of 90% formamide, 50 mM EDTA. The samples were heated at 95°C for 5 min and then analyzed by electrophoresis through a 40-cm 18% polyacrylamide gel containing 7.5 M urea in 44.5 mM Tris borate, pH 8.3, 1 mM EDTA. The products were visualized by autoradiography.
Author contributions-A. E. and S. S. conceptualization; A. E., Y. G., and S. S. investigation; A. E. and Y. G. writing-review and editing; S. S. funding acquisition; S. S. writing-original draft; S. S. project administration.