Crystal Structure of tRNA Adenosine Deaminase (TadA) from Aquifex aeolicus*

The bacterial tRNA adenosine deaminase (TadA) generates inosine by deaminating the adenosine residue at the wobble position of tRNAArg-2. This modification is essential for the decoding system. In this study, we determined the crystal structure of Aquifex aeolicus TadA at a 1.8-Å resolution. This is the first structure of a deaminase acting on tRNA. A. aeolicus TadA has an α/β/α three-layered fold and forms a homodimer. The A. aeolicus TadA dimeric structure is completely different from the tetrameric structure of yeast CDD1, which deaminates mRNA and cytidine, but is similar to the dimeric structure of yeast cytosine deaminase. However, in the A. aeolicus TadA structure, the shapes of the C-terminal helix and the regions between the β4 and β5 strands are quite distinct from those of yeast cytosine deaminase and a large cavity is produced. This cavity contains many conserved amino acid residues that are likely to be involved in either catalysis or tRNA binding. We made a docking model of TadA with the tRNA anticodon stem loop.

Inosine (Fig. 1A) was found at the first ("wobble") position of the tRNA anticodon 40 years ago (1). Crick postulated (2) that inosine is able to pair with U, C, and A. The codons corresponding to each inosine-bearing tRNA are synonymous, which contributes to decreasing the number of isoacceptor tRNAs (3). In eukaryotes, the wobble positions of eight cytoplasmic tRNAs (seven in Saccharomyces cerevisiae) bear inosine (4), which is generated by the posttranscriptional hydrolytic deamination of adenosine (5). In most bacteria and plant chloroplasts, only tRNA Arg-2 (Fig. 1B) has the inosine modification (4).
The enzymes that catalyze inosine generation were cloned recently (6,7). In eukaryotes, a heterodimeric enzyme composed of two sequence-related subunits (Tad2p/ADAT2 1 and Tad3p/ADAT3) is responsible for this modification (6). In most bacteria and plant chloroplasts, a tRNA adenosine deaminase (TadA) edits the adenosine residue at the wobble position of tRNA Arg-2 (7). TadA shares homology with Tad2p and is considered to form a homodimer.
TadA contains the consensus motif (C/H)XEX n PCXXC of the cytidine deaminase (CDA) superfamily, which includes diverse deaminases acting on bases, nucleosides, nucleotides, and nucleic acids (8 -16). This consensus motif forms the core of the active site and chelates a zinc ion tetrahedrally, and a common zinc-assisted deamination mechanism has been proposed (8 -12). In contrast, the mechanisms of substrate recognitions are diverse. Based on the complex structures solved thus far, the recognition mechanisms have been reported for cytidine (8,9), cytosine (10), and guanine (11). The only solved structure of a deaminase acting on polymeric nucleic acids is that of yeast CDD1 (12), which edits both mRNA and cytidine (17). Based on the available structures, molecular models of other deaminases acting on nucleic acids has been made and their substrate recognition modes have been reported (12,18).
In this study, we solved the crystal structure of TadA from Aquifex aeolicus at a 1.8-Å resolution. This is the first structure of a deaminase acting on tRNA. TadA forms a dimer, whereas the yeast CDD1 forms a tetramer. We made a docking model of TadA with tRNA.

EXPERIMENTAL PROCEDURES
Protein Preparation-The A. aeolicus tadA gene was PCR-amplified from the genomic DNA and subcloned into pET28b (Novagen) between the NdeI and SalI sites. The recombinant protein consists of the 151 amino acid residues from A. aeolicus TadA and 20 additional vectorencoded His tag residues (MGSSHHHHHHSSGLVPRGSH) at the N terminus for affinity purification. Escherichia coli strain BL21(DE3) CodonPlus (Stratagene) was transformed with the plasmid. For protein overexpression, the cells were grown in LB medium at 37°C to an A 600 of 0.6 and then the expression was induced with 1 mM isopropyl-␤-Dthiogalactopyranoside for 3 h. The cells were harvested and sonicated in 20 mM Tris-HCl buffer (pH 8.5) containing 500 mM NaCl, 10 mM imidazole, 1.4 mM 2-mercaptoethanol, and a protease inhibitor mixture, Complete EDTA-free (Roche Applied Science). The insoluble cell debris was removed by centrifugation at 15,000 ϫ g for 10 min at 4°C. The supernatant was heat-treated at 72°C for 20 min to denature the E. coli proteins. The heat-treated mixture was centrifuged at 15,000 ϫ g for 15 min at 4°C. The supernatant was applied to a 10-ml column of nickelnitrilotriacetic acid Superflow (Qiagen) equilibrated with 20 mM Tris-HCl buffer (pH 8.5) containing 500 mM NaCl, 10 mM imidazole, and 1.4 mM 2-mercaptoethanol. The protein was eluted in one step with 20 mM HEPES-NaOH buffer (pH 7.5) containing 300 mM NaCl, 250 mM imidazole, and 1.4 mM 2-mercaptoethanol. Four volumes of 20 mM HEPES-* This work was supported by a grant-in-aid for Scientific Research S from the Japan Society for Promotion of Science (JSPS), a grant-in-aid for Scientific Research in Priority Areas from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan, the RIKEN Structural Genomics/Proteomics Initiative (RSGI), and the National Project on Protein Structural and Functional Analyses, MEXT. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The  (28) and were corrected manually based on the secondary structures determined using the program PROCHECK (24). The secondary structures of A. aeolicus TadA and yeast CDD1 are shown at the top and the bottom, respectively. The numbering of the sequence is that of A. aeolicus TadA. The figure was generated with ESpript (29,30). Consensus residues of the CDA superfamily are shown in white within red-filled rectangles. The conserved residues at the putative active site of TadA, which are mentioned in Fig. 7A, are in blue boxes.

FIG. 1. Modification of the adenosine residue into inosine at the wobble position.
A, the chemical structures of adenosine and inosine are shown. The adenosine residue is modified into inosine by hydrolytic deamination. B, secondary structure of the A. aeolicus tRNA Arg-2 (ACG) transcript. The dots represent base pairs. The anticodon stem-loop moiety (residues 26 -43), which is sufficient for E. coli TadA to deaminate adenosine (7), is colored green with the adenosine residue at the wobble position (position 34) colored red.

SDS-PAGE, and 1 mg of protein was obtained per liter of LB medium.
Crystallization and Data Collection-Prior to crystallization, the protein was concentrated to 8.4 mg/ml by ultrafiltration. For crystallization, the hanging drop vapor diffusion method was used by mixing 1 l of the protein solution with 1 l of 50 mM MES-NaOH buffer (pH 5.6) containing 200 mM KCl, 10 mM MgCl 2 , and 5% polyethylene glycol 8000 and equilibrating the mixture against 500 l of the reservoir solution composed of 45 mM MES-NaOH buffer (pH 5.6) containing 400 mM NaCl, 180 mM KCl, 9 mM MgCl 2 , and 4.5% polyethylene glycol 8000 at 20°C. Crystals were grown in 3 days to dimensions of ϳ0.5 ϫ 0.3 ϫ 0.15 mm 3 . The diffraction data set of the crystal was collected at beamline BL26B1 at SPring-8 (Harima, Japan) to a 1.8-Å resolution. Before flash-cooling, the crystal was transferred into a cryoprotective solution composed of 55 mM MES-NaOH buffer (pH 5.6) containing 35% (v/v) glycerol, 220 mM KCl, 11 mM MgCl 2 , and 5.5% polyethylene glycol 8000. The data were processed and reduced using the HKL2000 program (19). The crystal (space group P2 1 ; unit-cell parameters, a ϭ 43.2 Å, b ϭ 152.0 Å, c ϭ 54.1 Å, ␤ ϭ 113.4°) contains four molecules per asymmetric unit with a solvent content of 41%.
Phase Determination and Structure Refinement-We solved the structure by the molecular replacement method. The initial model was the N-terminal 94 amino acid residues of yeast cytosine deaminase (CD) (Protein Data Bank code 1UAQ), and its sequence was replaced by the sequence of A. aeolicus TadA by the 3D-PSSM server. The N-terminal amino acid residues (Met 15 -Val 108 of yeast CD and Leu 7 -Ile 98 of A. aeolicus TadA) share 31% sequence identity (Fig. 2). We carried out molecular replacement using the program MOLREP (20,21). Three monomers were found and were first subjected to rigid body refinement using the data set up to a 2.5-Å resolution. After several rounds of Cartesian coordinate energy minimization, simulated annealing, and B factor refinement using the program CNS (22) as well as manual revision and building of the model using the program O (23), R and R free decreased to 34.6 and 36.7%, respectively. Two of the three monomers formed a dimer, so we used this dimer as a search model and repeated the molecular replacement. Two dimers then were found, and the structure refinement was performed in the same way using the data set from a 50.0 -1.8-Å resolution. The model refinement finally converged to an R value of 19.8% and an R free value of 24.8% with good stereochemistry. A Ramachandran plot analysis using the program PROCHECK (24) showed that 89.8% of the residues are in the most favored regions and 10.2% are in the additionally allowed regions. Data collection and modelrefinement statistics are shown in Table I. The electron density of the main chain was clear for all of the TadA residues with the exception of the partially disordered residues Pro 119 -Asn 122 and the two C-terminal residues. In addition, 3, 8, and 9 of the 20 vector-encoded His tag residues were visible at the N terminus in molecules A, B, and C, respectively.

RESULTS AND DISCUSSION
Structure Determination-We determined the crystal structure of A. aeolicus TadA at a 1.8-Å resolution by the molecular replacement method. The crystal contains four molecules, A, B, C, and D, per asymmetric unit (Fig. 3, A and B). The root mean square deviations of all of the atoms are 1.14, 1.06, and 0.75 Å when comparing A with B, C, and D, respectively. The monomer structure (Fig. 3A) consists of a central ␤-sheet (␤1-␤5) with two ␣-helices (␣1, ␣5) on one side of the sheet and three where I j (hkl) and ͗I j (hkl)͘ are the intensity of measurement j and the mean intensity for the reflection with indices hkl, respectively. b R factor ϭ ⌺͉F obs Ϫ kF calc ͉/⌺ hkl F obs , where k is a scale factor and R free is the R factor for the test set of reflections not used during refinement (5% data set). ␣-helices (␣2-␣4) on the other side. 〈 long loop exists between the ␤4 and ␤5 strands and is designated as the ␤4 -␤5 loop (residues Lys 104 -Arg 124 ).
Zinc Ion-In the F o Ϫ F c Fourier map contoured at a 9level, each of the monomers contained one strong spherical density, which was assigned as a zinc ion based on the x-ray absorption fine structure data (Fig. 4). Because no zinc ion was added in the buffer during the protein purification and crystallization, A. aeolicus TadA contains endogenous zinc ions. The zinc ion is tetrahedrally coordinated (Fig. 3, A and C) by His 52 N ␦1 (2.1 Å), Cys 82 S ␥ (2.3 Å), Cys 85 S ␥ (2.3 Å), and a water molecule, Water O (2.3 Å). Glu 54 O ⑀ interacts with the zinc-bound water (2.5 Å) (Fig. 3, A and C). These residues are conserved among the CDA superfamily members (Fig. 2). The active-site architectures are also similar within the CDA superfamily (8 -12), and the deamination mechanisms may be the same. The zinc ion and Glu 54 are proposed to activate the zinc-bound water to form a hydroxide ion. Glu 54 may shuttle a proton from the water to the adenosine residue. The zinc-bound hydroxide ion is proposed to attack the C 6 atom of the adenosine residue (Fig. 1A) nucleophilically.
Dimerization State of TadA-In the crystal, molecules A-B and molecules C-D form two apparent dimers along the noncrystallographic 2-fold axes (Fig. 3B). Each dimer is almost spherical with the exception of the C-terminal protrusions and the zinc-containing cavities (Fig. 3D). The dimerization interface is mainly composed of helices ␣3 and ␣4 and the ␤4 -␤5 loop. The interface is extensive and buries 1300 Å 2 of the total monomer surface area of 8100 Å 2 . Thirty residues are involved in the dimerization including eight conserved hydrophobic residues (Met 55 , Ile 58 , Met 84 , Ala 88 , Val 112 , Phe 113 , Ile 115 , and Leu 121 ). In addition, four intersubunit salt bridges are formed between Glu 44 O ⑀ and Lys 68 N (2.5 Å) and between Asp 48 O ␦ and Lys 59 N (2.7 Å). When Asp 64 in E. coli TadA, which corresponds to Asp 48 in A. aeolicus TadA, was replaced by Glu, the mutant enzyme was fully active in vivo but lost its activity in vitro (7,25). This residue may contribute to the structural stabilization of the protein, either by itself or through interaction with other protein(s).
The A-B and C-D dimers touch (Fig. 3B)  idues reportedly retained its editing activity (7,25). Therefore, the seven corresponding residues at the C terminus of A. aeolicus TadA may be dispensable for editing (Fig. 2), indicating that the interaction between molecules A and D is not biologically important. All of the interactions between the A-B and C-D dimers may be because of crystal packing.
Comparisons of the A. aeolicus TadA Structure to Those of Other CDA Superfamily Members-Sequence alignments (Fig.  2) reveal that the N-terminal halves are conserved among the CDA superfamily members. Residues 1-100 of A. aeolicus TadA share sequence identities of 40% with residues 1-103 of Bacillus subtilis guanine deaminase (GD), 24% with residues 9 -111 of yeast CD, 16% with residues 1-106 of B. subtilis CDA, and 15% with residues 9 -116 of yeast CDD1. Yeast CDD1 deaminates both mRNA and cytidine (17). The C-terminal halves have no sequence conservation.
A structural homology search by DALI (26) shows that A. aeolicus TadA displays similarities to yeast CD, B. subtilis GD (Protein Data Bank code 1TIY), B. subtilis CDA (Protein Data Bank code 1JTK), and yeast CDD1 (Protein Data Bank code 1R5T) with Z-scores of 20.2, 16.8, 9.7, and 9.6, respectively. Especially, the N-terminal halves, which are shown in pink (Figs. 5 and 6), share high structural similarity. The root mean square deviations of the N-terminal C ␣ atoms of yeast CD, B. subtilis GD, B. subtilis CDA, and yeast CDD1, superposed on the corresponding residues of A. aeolicus TadA, are 2.55, 1.89, 3.17, and 3.30 Å, respectively. This high structural conservation of the N-terminal motifs would be necessary to coordinate a zinc ion in the active site.
We next describe the comparisons of the oligomerization states and of the C-terminal halves of architectures. A. aeolicus TadA (Fig. 5A) and yeast CD (Fig. 5B) form similar dimeric structures. Their dimer interfaces comprise helices ␣3 and ␣4 and the regions between ␤4 and ␤5 (the loop in A. aeolicus TadA and the helix ␣5 in yeast CD). In addition, the C-terminal motifs of the monomers have similar structures (Fig. 6). First, the ␤4 and ␤5 strands run parallel to each other. Second, the N and C termini are located on the same side of the ␣/␤/␣ threelayered fold. Third, both of the C-terminal helices (␣5 in A. aeolicus TadA and ␣6 in yeast CD) contact the ␣1 N-terminal helices.
B. subtilis GD forms an intertwined dimer through C-terminal domain swapping (Fig. 5C). The dimer interface is composed of the helices ␣3 and ␣4 and the C-terminal domain. Helices ␣5 and ␣6, which are located between the ␤4 and ␤5 strands, make the domain swapping possible and have important roles in guanine recognition (11). As a consequence of the swap, the C-terminal motif of one subunit interacts with the other subunit in a manner similar to that of A. aeolicus TadA and yeast CD (Fig. 6C).
Yeast CDD1 and B. subtilis CDA form tetrameric structures. One subunit of yeast CDD1 (Fig. 5D) interacts with the other three subunits via the ␣2-␣6 helices. B. subtilis CDA also has a tetrameric structure similar to that of yeast CDD1. The tetramers are not formed by the dimerization of two TadA-like dimers. Therefore, the quaternary structure of yeast CDD1 (Fig. 5D) completely differs from that of A. aeolicus TadA (Fig.  5A). Also, the C-terminal motifs of B. subtilis CDA (Fig. 6D) and yeast CDD1 (Fig. 6E) share no structural similarities with that of A. aeolicus TadA (Fig. 6A). First, the ␤4 and ␤5 strands run antiparallel to each other and only short turns exist between the ␤4 and ␤5 strands. Second, the N and C termini are located on the opposite sides of the ␣/␤/␣ three-layered fold.
The oligomerization state does not determine whether nucleic acids are accommodated. Based on the structure of yeast CDD1, a model of human activation-induced cytidine deaminase has been made (12) on the assumption that its quaternary structure is similar to a tetramer. On the other hand, an analysis of the human activation-induced cytidine deaminase sequence with the 3DJury method suggested that the dimeric yeast CD was the best template (27). One problem with using yeast CD as a template was that the intramolecular active site was too small to accommodate large nucleic acid molecules (12). However, the structure of A. aeolicus TadA indicates that activation-induced cytidine deaminase might form a TadA-like dimer for nucleic acid binding.
The RNA-binding Site of A. aeolicus TadA-We next examined the substrate-binding site differences between A. aeolicus TadA and yeast CD to determine which motifs contribute to the specific binding of their substrates, tRNA and cytosine, respectively. The active-site cavity of A. aeolicus TadA (Fig. 5A) is composed of the ␣2 and ␣5 helices, and the zinc ion from one subunit, the ␤4 -␤5 loops from both subunits, and loop 1 (between ␣2 and ␤3) and loop 2 (between ␣3 and ␤4) from the other subunit. Both of the ␤4 -␤5 loops of the A. aeolicus TadA dimer are extended and cooperate with each other to form the putative RNA-binding site (Fig. 5A). On the other hand, the corresponding regions (the ␣5 helices) in the yeast CD dimer are involved in the intersubunit interaction but are far from the cytosine-binding site (Fig. 5B). The C-terminal ␣5 helix of A. aeolicus TadA protrudes outward, whereas the corresponding C-terminal ␣7 helix of yeast CD bends inward, which makes the cavity narrower and suitable for a small cytosine base (10). In summary, the diversified regions between the ␤4 and ␤5 strands and the C-terminal helices have important roles in the specific recognition of the substrates by A. aeolicus TadA and yeast CD.
The Recognition of tRNA Arg-2 -The anticodon stem loop structure (Fig. 1B) is reportedly sufficient for E. coli TadA to deaminate the adenosine residue at the wobble position (posi- tion 34) (7). A mutational study showed that the bases 33-36 (UACG) are recognized base-specifically and that the stem is recognized base-nonspecifically (7). Fig. 7A shows the side chains of the conserved residues in the active-site cavity of A. aeolicus TadA, which should hold the anticodon loop of tRNA so that the adenosine residue lies just above the zinc-bound hydroxide ion.
The asparagine residue at the C-terminal end of the ␤2 strand is widely conserved within the CDA superfamily (Fig. 2) and is known to interact with cytidine, cytosine, and guanine (9 -11). The corresponding residue, Asn 41 , in A. aeolicus TadA is located at the bottom of the active site and may contribute to the recognition of the adenosine residue at position 34. The adenosine residue normally base-stacks with nucleotide residues 35-38 but should be unstacked for the deamination. The conserved hydrophobic residues (Val 23 , Val 25 , Val 77 , Leu 79 , and Leu 139 ) colored yellow in Fig. 7A cluster on the wall of the cavity and may accommodate the unstacked anticodon.
The stem moiety may be recognized by the helix ␣5, which includes some conserved positively charged amino acid residues (Fig. 2). Lys 144 and Arg 147 in A. aeolicus TadA lie on the same side of the helix (Fig. 7A). We thought that these positively charged side chains may interact with the negatively charged tRNA phosphate backbone. The seven C-terminal residues including Arg 147 in A. aeolicus TadA are reportedly not essential for editing (7,25). Thus, as long as TadA has Lys 144 , it may be able to bind with tRNA. Fig. 7, B and C, show a preliminary docking model of TadA and the anticodon stem loop moiety (bases 26 -43) of yeast tRNA Phe (Protein Data Bank code 1EHZ). We docked them with no conformational change. The anticodon stem loop is well embedded in the putative RNA binding cavity (Fig. 7B). The surface of A. aeolicus TadA is color-coded according to its elec-trostatic potential (red, Ϫ10kT/e; blue, ϩ10kT/e) (Fig. 7C). The positive charges at the C-terminal helix are located near the tRNA phosphate backbones.
Several conserved residues exist at the entrance of the cavity (Fig. 7A). Lys 105 , Phe 142 , and Phe 143 of one subunit and Lys 68 , Tyr 69 , Arg 93 , His 123 , and Asn 124 from the other subunit may also interact with the tRNA. Lys 68 and Tyr 69 are conserved as a combination of a positively charged residue and an aromatic residue (Fig. 2). The side chains of these two residues are fixed by a cation-interaction. The conserved Arg 93 residue is also located adjacent to Tyr 69 in A. aeolicus TadA. These residues may contribute to specific interactions with the anticodon bases. In the loop region, Asn 122 and His 123 are not very close to the putative catalytic site but the loop might change its conformation and bring them closer upon tRNA binding.
In summary, the large cavity formed around the catalytic site with the zinc ion seems to be important for accommodating the anticodon loop stem of tRNA Arg-2 . This structural feature clearly distinguishes TadA from yeast cytosine deaminase, which has a smaller substrate. FIG. 7. The putative tRNA Arg-2 -binding site of A. aeolicus TadA. A, stereoview of the conserved residues in the cavity of A. aeolicus TadA. One subunit is colored pink and the other is colored cyan. B and C, preliminary docking model of TadA and the anticodon stem-loop moiety (bases 26 -43) of yeast tRNA Phe . The ribbon model of the anticodon stem-loop is colored yellow. TadA is represented as a surface model. The color scheme in B is the same as in A. The surface of A. aeolicus TadA in C is color-coded according to its electrostatic potential (red, Ϫ10 kT/e; blue, ϩ10 kT/e). The molecular surface was produced using the program MSMS.