Structural insights into the Thermus thermophilus ADP-ribose pyrophosphatase mechanism via crystal structures with the bound substrate and metal.

ADP-ribose pyrophosphatase (ADPRase) catalyzes the divalent metal ion-dependent hydrolysis of ADP-ribose to ribose 5'-phosphate and AMP. This enzyme plays a key role in regulating the intracellular ADP-ribose levels, and prevents nonenzymatic ADP-ribosylation. To elucidate the pyrophosphatase hydrolysis mechanism employed by this enzyme, structural changes occurring on binding of substrate, metal and product were investigated using crystal structures of ADPRase from an extreme thermophile, Thermus thermophilus HB8. Seven structures were determined, including that of the free enzyme, the Zn(2+)-bound enzyme, the binary complex with ADP-ribose, the ternary complexes with ADP-ribose and Zn(2+) or Gd(3+), and the product complexes with AMP and Mg(2+) or with ribose 5'-phosphate and Zn(2+). The structural and functional studies suggested that the ADP-ribose hydrolysis pathway consists of four reaction states: bound with metal (I), metal and substrate (II), metal and substrate in the transition state (III), and products (IV). In reaction state II, Glu-82 and Glu-70 abstract a proton from a water molecule. This water molecule is situated at an ideal position to carry out nucleophilic attack on the adenosyl phosphate, as it is 3.6 A away from the target phosphorus and almost in line with the scissile bond.

Nudix pyrophosphatases are widely distributed in nature and share a highly conserved amino acid sequence motif called the "Nudix motif " (GX 5 EX 7 REUXEEXGU, where U is one of the bulky hydrophobic amino acids I, L, or V), which adopts a unique loop-helix-loop structure (1). Enzymes in this family catalyze the hydrolysis of nucleoside diphosphates, linked to another moiety x. Their postulated role is to control the cellular concentration of toxic nucleoside diphosphate derivatives or physiological metabolites, accumulation of which could be harmful (1). ADP-ribose (ADPR) 1 is one such diphosphate derivative, which is produced enzymatically as part of the turnover of NAD ϩ , cyclic ADPR, ADP-ribosylated proteins, and poly-ADP-ribosylated proteins. Although certain proteins are posttranslationally modified by ADPR, high intracellular levels of ADPR could result in nonenzymatic ADP-ribosylation. This is a deleterious process that inactivates enzymes and could interfere with the recognition of enzymatic ADP-ribosylation (2). ADPR pyrophosphatases (ADPRases) catalyze the hydrolysis of ADPR to AMP and ribose 5Ј-phosphate to prevent ADPR accumulation.
ADPRase activity has been detected in all three kingdoms (3)(4)(5)(6)(7), but the specificity for ADPR over other substrates and the selectivity of metal ions required for activity vary between species. The mechanism underlying the different substrate specificity and the metal dependence is unknown at both the structural and functional levels. Elucidation of these properties requires the study of ADPRases from numerous sources.
In this article, we investigated the catalytic mechanism of ADPRase from an extreme thermophile, Thermus thermophilus HB8 (TtADPRase). In general, proteins isolated from T. thermophilus are heat-stable and suitable for physicochemical studies, including x-ray crystallography (8,9). TtADPRase catalyzes the hydrolysis of ADPR to AMP and ribose 5Ј-phosphate in the presence of Mg 2ϩ and Zn 2ϩ ions. The enzyme is also heat-stable and retains ADPRase activity at 75°C. The structures of the free enzyme, the Zn 2ϩ -bound enzyme, the binary complex with ADPR, the ternary complexes with ADPR and metal, and the product complexes containing AMP or ribose 5Ј-phosphate were determined. Based on structural and functional analyses, we propose a novel reaction mechanism for ADPR pyrophosphate hydrolysis that is different from those proposed based on the ternary complex structures of ADPRases from Escherichia coli (EcADPRase) and from Mycobacterium tuberculosis (MtADPRase) (10,11).

EXPERIMENTAL PROCEDURES
Preparation of Mutant TtADPRases-pT7Blue-ndx4, which contains the ndx4 gene, was used as a template of mutagenesis (12). pT7Blue-ndx4 was digested with BamHI and HindIII, and the fragment was inserted into pKF19K (Takara Shuzo Co., Kyoto, Japan) digested with the same enzymes. Site-directed mutagenesis of E82Q, E86Q, and E129Q was performed according to the Mutan-Super Express Km kit instruction manual (Takara Shuzo). Site-directed mutagenesis of D126N, E127Q, and D128N was performed by replacement of the wildtype sequence in pT7Blue-ndx4 plasmid with each mutated fragment. E82Q mutant fragment was inserted into pET-3a, whereas the other mutant fragments were inserted into pET-11a. Overproduction and purification of mutant proteins were carried out in a similar manner to that for wild-type enzyme (12).
Enzyme Assays-TtADPRase activities of wild-type, E82Q mutant, and E86Q mutant were analyzed by reversed-phase high performance liquid chromatography. The reaction solution contained 50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 or 250 M ZnSO 4 , 0.1 M KCl, 50 nM wild-type enzyme or 5.5 M for E82Q and E86Q mutants, and various concentrations of ADPR. The 100-l reaction mixture was incubated at 25°C for 3 min for the wild-type or for 11 h for the E82Q and E86Q mutants. Reactions were stopped by the addition of an equivalent volume of 100 mM EDTA. The mixture was applied to a C18 column (CAPCELL PAK C18 MG column, Shiseido, Tokyo, Japan) equilibrated with 20 mM sodium phosphate, pH 7.0, 5 mM tetrabutylammonium phosphate and 5% (v/v) methanol with AKTA explorer (Amersham Biosciences). The elution was performed at 1 ml min Ϫ1 with a 5-50% gradient of methanol in the equilibration buffer, and an absorbance at 260 nm was monitored. AMP and ADPR were eluted at elution volumes of 5 and 10 ml, respectively. The amounts of compounds were assessed from their peak areas.
TtADPRase activities of wild-type, D126N, E127Q, D128N, and E129Q mutants were analyzed by measuring inorganic orthophosphate production using the colorimetric procedure of Ames and Dubin (13). The 100-l reaction solution contained 50 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 0.1 M KCl, 50 nM TtADPRase, 2 units of calf intestine alkaline phosphatase, and various concentrations of ADPR. After incubation at 25°C for 3 min, 75 l of reaction mixture was mixed with 175 l of the ascorbic-molybdate mixture and incubated at 45°C for 20 min. Then, the quantity of phosphoric acid was measured at an absorbance of 820 nm. The kinetic parameters were obtained by fitting the initial rate of product formation to the Michaelis-Menten equation.
Crystallization-The method for crystallization and data collection of the free enzyme and the binary complex has been described previously (12). Overproduction of selenomethionyl (Se-Met) TtADPRase was performed as follows. E. coli B834(DE3) cells carrying the pET-11a-ndx4 plasmid were cultured in LB medium at 37°C for a few hours. The preculture was incubated at a dilution ratio of 100:1 (v/v) into LeMaster medium containing Se-Met with lactose as a carbon source and cultured at 37°C for 24 h. The Se-Met TtADPRase was purified in a manner similar to that for native protein (12). All crystals, except the binary complex form, were obtained under the same condition as that of the free enzyme, with slight adjustments in pH or the concentration of ammonium sulfate, and the addition of metal, substrate, or product. Crystals were flash-cooled in a nitrogen stream, and data were collected under cryo conditions.
Multiwavelength Anomalous Diffraction (MAD) Phasing, Model Building, and Refinement of the Free Enzyme-MAD data were collected at four wavelengths near the selenium absorption edge at the beamline BL44B2 at SPring-8 (14). Diffraction images were processed using the HKL2000 program (15). The structure was determined by the MAD method using programs from the CCP4 suite (16) and data collected at four wavelengths (edge, high remote, low remote, and peak). The initial model was built with the aid of the amino acid sequence using the program O (17), and refined from 15-to 2.5-Å resolution using the phase information from the diffraction data with the CNS program (18). The three selenium atom sites deduced from the difference Fourier map coincided with the three sulfur sites of the methionine residues in the model. The N-terminal region was disordered, and the first Met was not detected. Progress in the structural refinement was evaluated at each stage by the free R factor and by inspection of stereochemical parameters calculated by the PROCHECK program (19). Final phasing and refinement statistics are shown in Table I.
Data Collection and Refinement of the Wild-type Complexes-The crystals of the Zn 2ϩ -bound enzyme were obtained by mixing 1:1 the reservoir solution and the solution containing 2 mM protein and 2 mM ZnCl 2 . Data were collected at the wavelengths of peak (1.278 Å) and remote (1.292 Å) at the beamline BL44B2 at SPring-8. Diffraction images were processed using the HKL2000 program. Data collection and refinement statistics are shown in Table II.
The crystals of the ternary complex were obtained by mixing 1:1 the reservoir solution and the solution containing 2 mM protein, 2 mM ADPR, and 2 mM ZnCl 2 or GdCl 3 . X-ray absorption fine structure measurement determined the Zn 2ϩ absorption peak and low remote to be 1.282 and 1.287 Å, respectively, and the Gd 3ϩ absorption peak and low remote to be 1.711 and 1.716 Å, respectively, at the beamline BL44B2 at SPring-8. Then, diffraction data of the ternary complex were collected at these wavelengths. Additionally, data of the Gd 3ϩ ternary complex were also collected at the wavelength of 1.0 Å.
The crystals that contained AMP were obtained by mixing 1:1 the reservoir solution and the solution containing 2 mM protein, 10 mM AMP, 2 mM ribose 5Ј-phosphate, and 2 mM MgCl 2 . The crystals that contained ribose 5Ј-phosphate were obtained by mixing 1:1 the reservoir solution and the solution containing 2 mM protein, 2 mM AMP, 2 mM ribose 5Ј-phosphate, and 2 mM ZnCl 2 . X-ray diffraction experiments were carried out in the laboratory using Cu K␣ radiation. Data were collected with R-Axis VII (Rigaku) and were processed with Crystal-Clear 1.3.5 (Molecular Structure Corp., dxTREK version 8.OSSI). The structures were determined by molecular replacement with the AMoRe program (20) using the free enzyme as starting model and refined by CNS, and the metal sites were located in anomalous difference Fourier maps.
Data Collection and Refinement of Mutants E82Q and E86Q-Crystals of two mutant ternary complexes were produced by mixing 1:1 the reservoir solution and the solution containing 2 mM mutant protein, 2 mM ADPR, and 2 mM MgCl 2 or ZnCl 2 . The x-ray diffraction experiments and data collections of the Mg 2ϩ complexes were carried out in the laboratory in a manner similar to that for the product complex. To identify the Zn 2ϩ positions, data for the E82Q ternary complex with Zn 2ϩ were collected at the wavelengths corresponding to peak (1.282 Å) and remote (1.290 Å) at the beamline BL26B2 at SPring-8 (21) and data for the E86Q ternary complex with Zn 2ϩ were collected at the wavelengths corresponding to peak (1.278 Å) and remote (1.292 Å) at the beamline BL44B2 at SPring-8. The mutant structures were determined in a manner similar to that for the wild-type ternary complex. The same refinement procedures were applied, and statistics are shown in Table III.
Structure diagrams were drawn using the MOLSCRIPT (22), BOB-SCRIPT (23), and Raster3D programs (24). In figures, atoms of ␣and ␤-phosphate of ADPR are represented by letters A and B, not by ␣ and ␤.

Overall Structure
The dimer structure of the Gd 3ϩ ternary complex of TtAD-PRase is shown in Fig. 1A. The two identical monomers (one subunit is colored red and blue, and another subunit is colored pink and light blue) were related by a crystallographic 2-fold axis, which was consistent with the result of gel filtration chromatography (12). The structure of TtADPRase did not show large conformational change, regardless of whether it exists as the free enzyme, Zn 2ϩ -bound enzyme, binary complex with ADPR, ternary complexes with ADPR and Zn 2ϩ or Gd 3ϩ , or product complexes with AMP or ribose 5Ј-phosphate. When the C␣ atoms of these six structures were superimposed on the Gd 3ϩ ternary complex structure, the average root mean square deviation values between the Gd 3ϩ ternary complex structure and the other structures ranged from 0.20 to 0.26 Å.
The subunit was divided into two distinctive structural domains (Fig. 1, A and B): an N-terminal domain (residues 1-34, colored red) and a C-terminal domain, which is also referred to as the Nudix domain (residues 35-170, colored blue). The N- terminal domain comprised anti-parallel ␤-sheet from ␤1 to ␤3, and the Nudix domain was ␣ ϩ ␤ fold, with mixed ␤-sheet of ␤4 to ␤9 surrounded by three ␣-helices of ␣1 to ␣3 (Fig. 1B). N-terminal residues (1-10) and residues (124 -130) in loop L4 were disordered, but the presence of ADPR and metal resulted in order in loop L4. The Nudix motif (residues 67-89) folded into a loop-helix structure, not a loop-helix-loop structure, stabilized by an electrostatic network formed by salt links between Arg-81 and three Glu residues, Glu-73, Glu-82, and Glu-85. In general, thermophile proteins tend to have a short loop length, because flexibility of the loop reduces stability. Smaller loops, such as the loop-helix structure described above, are advantageous to TtADPRase. The TtADPRase structure of Gd 3ϩ ternary complex was quite similar to those of the EcAD-PRase and MtADPRase (10,11), although the detailed structures differed. N-terminal domain ␤-sheet was shorter, loop L4 was moved outward, and loop L5 tip was situated upward by 4 Å. TtADPRase has other properties that might contribute to stability, including a lower content of chemically unstable amino acids such as Asn, Cys, and Met (25), and a smaller surface area in the N-terminal domain than for EcADPRase and MtADPRase.

Dimer Interactions
The

ADPR Binding (Corresponds to
Reaction State II in Fig. 8) The stereo structure and the scheme of hydrogen-bonding and stacking interaction between TtADPRase and ADPR are shown in Fig. 2 (A and B). The activity of TtADPRase was dependent on the presence of metal ions and was inhibited by acidic pH. ADPR was not hydrolyzed when crystals were prepared in solutions without metal ions for the binary complex with ADPR or prepared in an acidic pH for the ternary complex with ADPR and Zn 2ϩ or Gd 3ϩ (Table II). The binary complex without metal ion gave a better Fourier map ( Fig. 2A) than the ternary complex. ADPR bound to the dimeric interface, composed of a turn between ␤1 and ␤2, L1, L3, ␤6 and ␣3 of one subunit, L3*, ␤2*, ␤3* of another subunit (secondary structures and residue numbers of another subunit are indicated by asterisks). Movement of amino acid residues was not observed between the free enzyme and the substrate complexes (the binary complex and the ternary complexes), except for the side chain Arg-18 and Arg-27* in the active site. In the pres- ence of ADPR, Arg-18 moved toward ADPR to stack on the adenine of ADPR. Arg-27* moved away to avoid the positive charge repulsion of Arg-18 and made a hydrogen bond with O2H of the adenosyl ribose (Fig. 3, compare A and B). Such rearrangement of two residues on binding of ADPR resulted in the formation of a fence with Tyr-28*.
Bound ADPR resembled a horseshoe, in which two ends come together such that the adenine N7 and ␤-phosphate oxygen made hydrogen bonds with the same water molecule W1 (Fig.  2). The amide nitrogen and oxygen of Glu-29* made hydrogen bonds with adenine N1 and N6, respectively. In the binary complex structure, the terminal ribose was in ␣-configuration, and this terminal ribose O1H made hydrogen bonds with two water molecules W2 and W3. These water molecules also made hydrogen bonds with amide nitrogen atoms of Ala-154 and Thr-155 and oxygen atoms of the side chain of Glu-63 and Thr-155. The adenine was stacked in the hydrophilic pocket composed of Arg-18, Ile-19, and Tyr-28*. Two oxygen atoms of the ␤-phosphate made hydrogen bonds with Arg-54. On the other hand, the ␣-phosphate group made a hydrogen bond with the Leu-68 amide nitrogen. Pro-103* stacked on the terminal ribose. In the binary complex, Ser-102* made hydrogen bonds with O2H and O3H of the terminal ribose. The amide oxygen and nitrogen of Gly-104* made hydrogen bonds with adenine N6 and the terminal ribose O2H, respectively. Glu-108 made a hydrogen bond with the terminal ribose O3H. These many interactions for recognition with the terminal ribose might explain the high substrate specificity of ADPR over ADP-sugar.

Metal Binding (Corresponds to Reaction
States I and II in Fig. 8) To identify the metal-binding sites, the structures of the Zn 2ϩ -bound enzyme and the Gd 3ϩ and Zn 2ϩ ternary complexes were determined. In the presence of Zn 2ϩ ions, k cat was 5 times larger than that of Mg 2ϩ , as shown in Table IV. Gd 3ϩ ion is an analog of Mg 2ϩ ion with 30% activity for Mg 2ϩ (data not shown). From the Zn 2ϩ -bound enzyme, two Zn 2ϩ ions were detected based on anomalous diffraction signals in the active site, as shown in Fig. 4A and Table V. ZnI was coordinated with tetrahedral geometry by Glu-82 and three water molecules. The signal of ZnII was very weak in 2F o Ϫ F c map, and ZnII was therefore not included in the model.
From the ternary complexes, two Zn 2ϩ ions or two Gd 3ϩ ions were identified in the active site, as shown in Fig. 4 (B and C). ZnI was coordinated with octahedral geometry by the bidentate carboxylate of Glu-82, three water molecules, and ␣-phosphate. ZnII was coordinated by the bidentate carboxylate of Glu-70. On the other hand, two Gd 3ϩ ions were located in the active site, although we detected four Gd 3ϩ ions in a subunit. GdI was coordinated by the bidentate carboxylate of Glu-82, two water molecules, and ␣-phosphate. GdII was coordinated with octahedral geometry by the carboxylate of Glu-86, the carbonyl oxygen of Ala-66, two water molecules, the bridging oxygen (O3␣), and ␤-phosphate. The other two Gd 3ϩ ions were located close to the positions of Glu-122 and Glu-140, respectively. Partial protonation of Glu residues may cause the longer distance to metal ions, because both ternary complex crystals were produced under pH 5.2. When the two ternary complex structures were compared with that of the binary complex, in the binary complex the two carboxylates of Glu-82 and Glu-86 were located away from the substrate, and the distances from O1␣ to the carboxylate oxygen of Glu-82 and from O3␣ to the carboxylate oxygen of Glu-86 were 3.8 Å and 5.9 Å, respectively. On the other hand, in the ternary structures, these Glu residues were closer to the substrate, and the corresponding distances were 2.9 and 4.8 Å for Zn 2ϩ , and 2.6 and 4.9 Å for Gd 3ϩ , respectively, because metal ions neutralized negative charge of the two carboxylates and phosphates.

Product Binding (Corresponds to
Reaction State IV in Fig. 8) Two product complexes were obtained (Fig. 5, A and B): one with AMP and Mg 2ϩ , and another with ribose 5Ј-phosphate and Zn 2ϩ . In the case of product complex with AMP and Mg 2ϩ (Fig.  5A), the adenine was recognized by Ile-19 and Tyr-28* through stacking interactions and adenine N1 made a hydrogen bond with Glu-29*, and adenine N6 made hydrogen bonds with Glu-29* and Gly-104*, although positions of adenine and ␣-phosphate of AMP were located away from residues in the Nudix motif.
In the case of another product complex with ribose 5Ј-phosphate and Zn 2ϩ (Fig. 5B), the ring-opened ribose O3H made hydrogen bond with N⑀ of His-33, and aldehyde O1 made hydrogen bind with O␥ of Ser-102*. ZnI and ZnII were coordinated by Glu-82 and Glu-70, respectively.   Table V. In both product complexes, Arg-18 and Arg-27* were located away from the active site. Consequently interactions between "fence" residues, consisting of Arg-18, Arg-27*, and Tyr-28*, and purine ring were absent (Fig. 3, compare B and C). These conformational changes observed in product complexes suggest the product-releasing mechanism of ADPRase.

Activity of Wild-type and Mutant TtADPRases
Based on the ternary structures, the mutants of Glu-82 or Glu-86, which are suggested to serve as metal ligands (Fig. 4), were constructed. The results of the kinetic analysis are summarized in Table IV. The k cat values of E82Q and E86Q mutants are approximately 5 orders of magnitude lower than that of wild-type enzyme, although the K m values were similar.
According to the EcADPRase and MtADPRase structures of ternary complex with substrate analogue and metal ions, Ec-Glu-162 and MtGlu-142 in loop L9 are believed to act as catalytic bases (10,11). When these ternary complex structures were superimposed on that of TtADPRase, the results showed that loop L4 of TtADPRase corresponds to loop L9 of EcAD-PRase and MtADPRase. Then, mutants of Asp-126, Glu-127, Asp-128, and Glu-129 in loop L4 were constructed. The k cat and K m of D126N, E127Q, D128N, and E129Q mutants were similar to those of wild-type enzyme.

Structures of E82Q and E86Q Mutants
To investigate, in terms of three-dimensional structure, how these mutants, E82Q and E86Q, inactivated the enzyme, we analyzed crystal structures of these mutants. The backbone structures of E82Q and E86Q mutants were similar to that of wild-type enzyme. The average root mean square deviation values of the C␣ atoms between the overall structure of the Zn 2ϩ ternary complex of wild-type enzyme and those of E82Q or E86Q mutants were 0.13 or 0.1 Å, respectively. The mutant complex structures with Mg 2ϩ and Zn 2ϩ were compared with those of Gd 3ϩ and Zn 2ϩ ternary complex of the wild-type enzyme, respectively (Fig. 6).
In the E82Q mutant structure that was determined using crystals produced in the solution with ADPR and Mg 2ϩ , Gln-82 N⑀2 moved 0.7 Å toward the amide oxygen of Leu-68 to form a hydrogen bond and lost its hydrogen bond with Arg-81 (Fig.  6A). ADPR, Mg 2ϩ ions, and water molecule (Wn) were not observed. Instead of ADPR, one sulfate molecule was located at the position corresponding to ␤-phosphate of ADPR in the Gd 3ϩ ternary complex of wild-type enzyme. The water molecule (WB) lost interaction with Mg 2ϩ ions, and its pK a value could be higher than that of wild-type enzyme.
In the E82Q mutant structure that was determined using crystals produced in the solution with ADPR and Zn 2ϩ , ADPR and ZnI were not observed, but one sulfate molecule was located at the position corresponding to ␤-phosphate of ADPR in the Zn 2ϩ ternary complex of wild-type enzyme (Fig. 6B). ZnII and two water molecules (WB, Wn) were present. Gln-82 moved 0.8 Å toward the backbone carbonyl oxygen of Leu-68 to form a hydrogen bond, and donated a proton to the water molecule (Wn) through a hydrogen bond. Thus, replacement of Glu-82 with Gln reduced the affinity for metal ion, and formed a new hydrogen bond between Gln-82 and Leu-68.
In the E86Q mutant ternary complex with ADPR and Mg 2ϩ (Fig. 6C), the conformation of residues in the active site was almost the same as in the Gd 3ϩ ternary complex of wild-type enzyme, although Gln-86 moved to form hydrogen bonds with Glu-82 and Ala-66. MgII and water molecule (Wn) were not observed. MgI and the water molecule (WB) were present; however, MgI was located away from the water molecule (WB).
In the E86Q mutant ternary complex with ADPR and Zn 2ϩ (Fig. 6D), Gln-86 also moved to form hydrogen bonds with Glu-82 and Ala-66. The adenosine moiety of ADPR was observed; however, the other moiety of ADPR was disordered. The water molecule (WB) was not observed. ZnI was coordinated by Glu-82 and the water molecule (Wn). ZnI was located at different position from that of the Zn 2ϩ ternary complex of wild-type enzyme (compare a large green ball with a small green ball around Glu-82). Thus by replacement of Glu-86 with Gln, the location of metal I (MgI and ZnI) altered, and new hydrogen bonds between Gln-86 and Glu-82 and between Gln-86 and Ala-66 were formed.
The replacements of Glu-82 and Glu-86 with Gln resulted in the loss of metal ions or the change of metal location, and resulted in the formation of new hydrogen bonds. As a result of such alteration of charge distribution in the active site, two water molecules (Wn, WB) could disappear or have pK a values different from those of the wild-type enzyme. Such situation was reported in mutagenesis studies of RNase HI from E. coli (26). These results indicate that the configuration composed of two water molecules (Wn and WB) and metal ions, orientations and positions of which are determined by Glu-82 and Glu-86, is important for activity.

TtADPRase Catalytic Mechanism
Based on the eight structures (Fig. 7, A--D) and functional analyses of TtADPRase, the following mechanism for the ADPR hydrolysis reaction is proposed and its reaction scheme is represented in Fig. 8.
Metal Binding of TtADPRase (Reaction State I in Fig. 8)-Even in the absence of ADPR, metal I (MI) can bind to Glu-82, as shown in the Zn 2ϩ -bound structure (Fig. 7A).
Ternary Complex with Substrate and Metal (Reaction State II in Fig. 8)-In the presence of ADPR, metal II (MII) can bind the ligands, as shown in the ternary structures with ADPR and metal ions (Fig. 7, C and D). Glu-82 and Glu-70 could abstract a proton from the water molecule (Wn), and could poise Wn for nucleophilic attack on the adenosyl phosphate. The distance between the nucleophile Wn and P␣ of ADPR is 3.6 Å. The angle formed between this P␣-O and P␣-O3␣, which is a scissile bond, is 145°, suggesting an in-line mechanism for the cleavage reaction. Arg-54 interacts with ␤-phosphate and the terminal ribose O5, which may stabilize the negative charge of the substrate. In the presence of a metal, the nucleophile Wn is located at an ideal position to be activated by Glu-82, and this water molecule can attack on the adenosyl phosphate. In the absence of metal ions, Glu-82 and Glu-86 are distant from the substrate as a result of ion-ion repulsion, and the distance between Glu-82 and the nucleophile Wn is increased to 4.5 Å (Fig. 7B); therefore, Glu-82 cannot abstract a proton from the nucleophile Wn.
Transition State (Reaction State III in Fig. 8)-In the transition state, the adenosyl phosphorus is expected to be bound to five oxygen atoms with trigonal bipyramidal geometry (Fig. 8,  III). To determine the site of nucleophilic substitution, the enzymatic reaction was conducted in the presence of H 2 18 O and the products were analyzed by electrospray mass spectrometry. The addition of 18 OH to AMP occurred, indicating that the FIG. 6. The active site structures of the E82Q and E86Q mutants. E82Q mutant structures were superimposed onto the corresponding metal ternary complex structures of wild-type enzyme with ADPR and Gd 3ϩ (A) or Zn 2ϩ (B). E86Q mutant structures were also superimposed onto the corresponding metal ternary complex structures of wild-type enzyme with ADPR and Gd 3ϩ (C) or Zn 2ϩ (D). The mutant structures are shown as ball-and-stick models. The structures of wild-type enzymes are not shown except for ADPR molecules (orange lines or brown lines), because they are similar to those of the mutants. Large balls represent water molecules (red balls) and metal ions (green balls) in the mutant structures, and small balls represent those in the wild-type structure. The atoms that were not observed in the mutant structures are written in gray letters. Hydrogen bonds and metal coordination in the mutant structures are shown by dashed blue lines. nucleophilic water (Wn) attacks the adenosyl phosphorus (data not shown). The pK a value of a water molecule that coordinates metal is lower than that of a normal water molecule (27). In this structure, the water molecule (WB) bound to metal ions (MI) serves as a general acid and donates a proton to the bridging oxygen O3␣.
Product Complex (Reaction State IV in Fig. 8)-In the product-bound state, some interactions between enzyme and product were lost (Fig. 6, A and B), and a "fence" was opened (Fig.  3C). This conformation was quite different from those of the ternary complexes with ADPR and metal, and suggested that Arg-18 and Arg-27* are involved in product release. DISCUSSION Recent structural studies of EcADPRase and MtADPRase proposed that a water molecule, which bridges two metal ions and is deprotonated by EcGlu-162 or MtGlu-142 in loop L9, is at an ideal position for carrying out the nucleophilic attack on the adenosyl phosphate. Although TtGlu-127 in loop L4 corresponds to EcGlu-162 and MtGlu-142 (Fig. 1B), the residues in loop L4 are distant from the substrate. Additionally, the mutagenesis studies showed that D126N, E127Q, D128N, and E129Q mutants retained activity (Table IV), suggesting that residues in loop L4 are not essential for the reaction. In the TtADPRase structures of the ternary complex with ADPR and Gd 3ϩ or Zn 2ϩ , a water molecule is present close to Glu-82 and Glu-70, and is in a favorable position for nucleophilic attack on the adenosyl phosphate. The structural and kinetic analyses of E82Q and E86Q suggested that Glu-82 and Glu-86 in the Nudix motif contribute to the binding of metal ions and water molecules, and thus are essential for the reaction.
Degradation of pyrophosphate derivatives by Nudix family proteins depends on metal ions. TtADPRase requires the presence of Mg 2ϩ or Zn 2ϩ ions for its activity. Although metal coordination and position are different between the Zn 2ϩ and Gd 3ϩ ternary complexes, as shown in Fig. 4 (B and C), the positions and orientations of the nucleophilic water (Wn), catalytic water molecule (WB), and catalytically essential residues are similar. TtADPRase did not show activity in the presence of other metal ions, such as Ca 2ϩ , Ni 2ϩ , and Mn 2ϩ . Positions and orientations of the nucleophilic water (Wn), catalytic water molecule (WB), and catalytically essential residues in the presence of the other metal ions would be different from those in the presence of Mg 2ϩ or Zn 2ϩ . The nucleophile water and catalytic acid water are represented by Wn and WB, respectively. Substrates P␣, P␤, and O3␣ are represented by PA, PB, and O3A, respectively. The water molecule that forms a hydrogen bond with Glu-85 is represented by WA. A, the Zn 2ϩ -bound enzyme structure. B, the binary complex structure. C, the ternary complex with ADPR and Gd 3ϩ . D, the ternary complex with ADPR and Zn 2ϩ . The Gd 3ϩ ternary complex structure is assumed to be analogous to a Mg 2ϩ ternary structure. The nucleophilic attack and proton abstraction events are shown by pink arrows in C and D.
Genome analysis of T. thermophilus HB8 suggested the absence of an ADPR-transferase like protein and the presence of an ADPR-glycohydrolase like protein, involved in ADPR transfer in mono-ADP-ribosylation (8,9). It implies that the possibility that ADPR molecules are constantly produced by mono-ADP-ribosylation is low. Recently, Sauve et al. (28) reported that the novel molecule O-acetyl-ADPR (OAADPR) is produced during NAD-dependent protein deacetylation by Sir2. Sir2 enzymes are broadly conserved from bacteria to humans and function in transcriptional silencing, DNA repair, and life span extension (29 -31). Rafty et al. (32) found that ADPRases such as yeast YSA1 and mouse NudT5 hydrolyzed OAADPR to AMP and acetylated ribose 5Ј-phosphate when studied in vitro. In TtADPRase binary complex with ADPR, hydroxyl groups of the terminal ribose interact with the residues through the water molecules. Acetyl group of OAADPR can be placed in the space occupied by the water molecules. A Sir2 homolog has been identified in the T. thermophilus HB8 genome providing a source of OAADPR, which is expected to be a natural substrate for TtADPRase. FIG. 8. Schematic representation of the reaction mechanism of ADPR hydrolysis. The reaction scheme was proposed based on the active site architecture according to the structural and functional analyses of TtADPRase. Reaction state I is the metal-bound state, II is the ternary complex with substrate and metal, III is the transition state, and IV is the product release state. The nucleophilic attack and proton abstraction events are shown by pink arrows.