Crystal structure of alanyl‐tRNA synthetase editing‐domain homolog (PH0574) from a hyperthermophile, Pyrococcus horikoshii OT3 at 1.45 Å resolution

The genetic code is interpreted by 20 aminoacyl-tRNA synthetases (aaRSs). The aaRS enzymes mediate the aminoacylation reactions that match amino acids with nucleotide triplets encoded as anticodons in transfer RNA (tRNA).1, 2 During selection of amino acids, aaRS has to discriminate the cognate substrate from another very similar amino acid (e.g., alanine from valine). Some tRNA synthetases distinguish such closely related amino acids with an error rate of 1 in 3000.3, 4 Such high accuracy is achieved by an "editing mechanism," a postprocessing of the aminoacyl-tRNA catalyzed by the editing domain of aaRS enzymes. The 20 aaRSs can be divided into class I and class II according to the architecture of the active site.5 While the editing function is well understood in some class I aaRSs, little is known about the mechanism in class II enzymes. To elucidate the editing mechanism of class II enzymes, we determined the crystal structure of alanyl-tRNA synthetase (AlaRS) editing domain homolog from Pyrococcus horikoshii OT3 [Open Reading Frame (ORF) ID: PH05746 (http://www.bio.nite.go.jp/)] at 1.45 Å resolution. This is the high-resolution structural report of AlaRS editing-domain homolog, COG2872.7 The amino acid sequence of PH0574 shows 42%, 28%, 24%, and 19% identity with the editing domains in Pyrococcus furiosus AlaRS, Escherichia coli AlaRS, Staphylococcus aureus threonyl-tRNA synthetase (ThrRS), and E. coli ThrRS (pfAlaRS, ecAlaRS, saThrRS, and ecThrRS), respectively (Fig. 1). In ecThrRS, the residues His73 and His77 proved to be essential for the editing activity.8 Both the corresponding histidines (His9 and His13 in PH0574 protein) are conserved in all the editing domains and homologs of AlaRS and ThrRS (Fig. 1 and Ahel et al.9). Many aaRS editing-domain homologs have been cloned10, 11 and characterized.12 For example, Methanosarcina barkeri AlaX (mbAlaX), which has sequence identity of 25% with PH0574 in the region of Arg97-Ile224 (Fig. 1), hydrolyzes mischarged seryl-tRNA to a serine and a tRNA.9 From these inspections, we concluded that PH0574 possesses the same editing activity as mbAlaX. Multiple alignment of the editing domain of AlaRS and ThrRS enzymes. Residues involved in coordination of the zinc ion (red) and highly conserved residues (yellow) are indicated. The secondary structure elements in the crystal structure of PH0574 (above) and ecThrRS (below) are shown in the alignment. GenBank accession numbers given in parentheses are PH0574 (NP_142539), mbAlaX (ZP_00296079), pfAlaRS (NP_577999), ecAlaRS (NP_417177), saThrRS (NP_646443, PDB ID: 1NYR), ecThrRS (NP_416234, PDB ID: 1QF6). In this study, we revealed that the crystal structure of PH0574 has a Zn2+ ion coordinated by the conserved three histidines and one cysteine residue.13 The same coordination was found in saThrRS editing domain.14 We propose that AlaRS hydrolyzes mischarged tRNA by the catalytic zinc ion in the editing domain based on the constructed tRNA binding model of AlaRS. PH0574 gene was amplified by the polymerase chain reaction (PCR) using P. horikoshii OT3 genomic DNA as a template. The plasmid pET-11a (Novagen) carrying a gene encoding PH0574 was constructed by the super-rare-cutter system (Hayashizaki et al., manuscript in preparation). The protein was overexpressed in the BL21-CodonPlus(DE3)-RIL (Novagen) strain grown in Luria–Bertani broth for 20 h at 37°C. After cell disruption and removal of debris by centrifugation, the supernatant was heat-treated at 90°C for 11.5 min. PH0574 protein was purified by serial chromatographic separation on SuperQ TOYOPEARL 650M (TOSOH), RESOURCE Q (Amersham Biosciences), and CHT type I ceramic hydroxyapatite (Bio-Rad) columns followed by HiLoad 16/60 Superdex 75 pg (Amersham Biosciences) gel filtration. PH0574 with selenium-substituted methionine (Se-Met) was also prepared for structural determination by multiwavelength anomalous dispersion (MAD) phasing. The Se-Met protein was overexpressed in the B834(DE3)pLysS (Novagen) strain grown in LeMaster broth with Se-Met for 24 h, and purified in a similar way to the native protein. The crystals of PH0574 were obtained from a solution containing 37.6 mg/mL protein, 27.5%(w/v) polyethylene glycol (PEG 4000), and 100 mM 2-Morpholinoethanesulfonic acid (MES)-Na (pH 6.3) using a full-automatic protein crystallization and observation system, "TERA."15 Harvested crystals were flash-cooled and X-ray diffraction data sets were collected at 100 K using an R-AXIS V (RIGAKU) or a JUPITER 210 (RIGAKU) detector at BL26B1, SPring-8. All the diffraction images were processed, integrated, and scaled using HKL2000.16 The selenium sites were found and refined, and initial phases were calculated with SOLVE/RESOLVE.17, 18 Initial model building was carried out by Arp/wARP19 automatically. Additional model building was manually performed with Quanta (Accelrys). Refinement of the model was performed using CNS20 and CCP4.21 The anomalous difference Fourier map was also calculated by CCP4.21 The model quality was checked by PROCHECK,22 which showed all the main-chain torsion angles to be within the most favored regions (93.9%) and the additional allowed regions (6.1%). A summary of the statistics for structural determination is given in Table I. The three-dimensional (3D) structural comparison was performed by DALI.23 The tRNA binding to PH0574 was modeled using Quanta. Following the DALI result, the crystal structure of ecThrRS complexed with tRNA [Protein Data Bank (PDB) ID: 1QF6]24 was used for the modeling of the tRNA. A serine residue (only a little bigger than the cognate alanine residue) was attached at the 3′ end of the tRNA. Residue numbers 2 and 71 of the tRNA were then digitally mutated to those of the P. horikoshii tRNA. Energy minimization was performed during and after modeling using Quanta. The figures were prepared using Molscript,25 Raster3D,26 and PyMOL.27 PH0574 gene cloned for this study shares sequence homology with the region of Arg556-Ile682 of ecAlaRS.28 This region of ecAlaRS contributes to the editing function of the enzyme (Fig. 1),29 suggesting a similar role for PH0574, which is a monomeric protein consisting of only 157 amino acid residues, with a molecular weight of 18,152. The final model with an R-factor of 20.6% and an Rfree of 22.7% at 1.45 Å resolution contained three monomers, each with a Zn2+ ion and 599 water molecules per asymmetric unit [Fig. 2(A)]. PH0574 has a two-domain structure consisting of seven antiparallel β-sheets and six helices. The large domain (green) consists of residues 1–69 and 124–157, with three helices, two α-helices and one 310-helix, surrounded by four antiparallel β-strands. A 310-helix (H6) is located between β5 (small domain) and β6 (large domain). The small domain (orange) consists of residue 70–123, with three antiparallel β-strands surround two short helices. The Zn2+ ion was assigned to the high electron density in the 2Fo − Fc map located at the hinge region [Loop 3 in Fig. 2(A)] with 26.5, 32.6, or 35.5 σ, based on the environment around the atom with the extremely strong anomalous signal. Figure 2(B) represents the 2Fo − Fc map contoured at 2 σ (blue) and the anomalous difference Fourier map contoured at 20 σ (orange) using the diffraction data set collected at 1.0 Å but no peak in the data set collected at 1.54 Å. The native data set was also used for the anomalous difference Fourier map calculation. Despite using a slightly shorter wavelength than the zinc anomalous peak (wavelength 1.28 Å), three strong peaks in the anomalous difference Fourier map located at the Zn2+ positions having 39.2, 53.6, and 48.2 σ in each monomer, respectively, were observed in the asymmetric unit. The assignment of the Zn2+ ion is consistent with the functional and structural studies of the essential zinc ion in ThrRS,14, 30 while zinc was the most abundant metal in the crystallization protein sample of PH0574 by a preliminary semiquantitative microwave induced plasma–mass spectrometry (MIP-MS) analysis. The large and small domains of the proteins are bridged by this Zn2+ ion, which is coordinated by the conserved His9, His13, Cys116, and His120 residues. These coordinating residues form a regular tetrahedron around the Zn2+ ion [Fig. 2(B and C)]. This type of tetracoordinate histidine and cysteine residues was also seen in saThrRS (PDB ID: 1NYR),14 with a root-mean-square deviation (RMSD) of 0.5 Å for these four residues. The important difference is that a water molecule is coordinated to the Zn2+ ion at a distance of 3.6 Å in the case of saThrRS. No water molecules are involved in Zn2+ coordination in PH0574. In contrast, many water molecules are observed in the cleft around the Zn2+ ion. The hydrophilic environment is suitable for the recognition of the hydrophilic amino acid in the mischarged tRNA. (A) Ribbon diagram of PH0574, which consists of a large (green) domain and a small (orange) domain. The Zn2+ ion is located between the two domains. (B) Close-up view of the Zn2+ ion binding site. A strong peak in the anomalous difference Fourier map was observed only at the zinc position. Many water molecules were observed around the Zn2+ ion, although none of these are involved in coordination. The anomalous difference Fourier map contoured at 20 σ (orange) and 2Fo − Fc map contoured at 2 σ (blue) are shown. (C) Zinc ion coordinate diagram. The coordination of the Zn2+ ion is mediated by the conserved histidine and cysteine residues both in PH0574 (green) and in saThrRS (blue, parentheses, PDB ID: 1NYR). (D) Structure of the putative active site drawn with a van der Waals surface with charge. The cavity accommodates an acyl serine. PH0574 shows similar dimensions and domain architecture to ecThrRS editing domain (Fig. 1)8, 24 with a DALI Z-score of 18.6 and an RMSD of 2.3 Å between two structures. The sequence homology to mbAlaX (Fig. 1), together with the structural similarity to the editing domain of saThrRS,14 suggests that PH0574 fulfills the editing function that AlaRS performs, although this remains to be experimentally determined. Based on the structure, we suggest that the cleft near the bound Zn2+ ion is the putative PH0574 active site. A model of a serine in a mischarged aminoacyl-tRNA was placed at the active site [Fig. 2(D)]. In the model, the carbonyl oxygen of the acyl serine lies pointing toward the Zn2+ ion at a distance of 3.8 Å. This was supported by a full seryl-tRNA binding model of PH0574, whose construction was based on the structure (available from the authors) of the tRNA complexed with ecThrRS (PDB ID: 1QF6),24 according to the direction by Dock-Bregeon et al.8, 30 Thus, the Zn2+ ion may activate the carbonyl group in the deacylation catalysis. This analysis of the structure of PH0574 has shown the coordination pattern of the Zn2+ ion and revealed a possible active site. After the deposition of our PH0574 structure coordinate and the structure factor (PDB ID: 1V4P) for this work, another structure of PH0574 at 2.62 Å resolution in the different space group, P212121, became available (PDB ID: 1V7O),31 and the structure is the same (RMSD of Cα atoms: 0.38 Å) substantially except that no Zn2+ ion was included. We thank all the staff of the Highthroughput factory and SPring-8 BL26 who have supported this work.