Structural basis for lysidine formation by ATP pyrophosphatase accompanied by a lysine-specific loop and a tRNA-recognition domain

Materials and Methods

Crystallization. The expression vector, pET15-TilS, was constructed by cloning the A. aeolicus tilS gene into the plasmid pET-15b (Novagen). A culture of E. coli strain BL21 codon plus (DE3), transformed with pET15-TilS, was grown to an OD₆₀₀ of 0.5, and then the expression of TilS was induced with 1 mM isopropyl β-d-thiogalactopyranoside (IPTG), followed by 20 h of cultivation at 37°C. After centrifugation at 5,000 × g for 30 min, the harvested cells were suspended in 100 mM Tris·HCl buffer (pH 7.8) containing 10 mM KCl, 8 mM MgCl₂, 10 mM DTT, and 1 mM PMSF and were gently sonicated. After centrifugation at 20,000 × g for 30 min, the supernatant was heat-treated at 70°C for 30 min, and then the recombinant TilS was purified by chromatography on SP Sepharose, Phenyl Superose, and Resource S columns (Amersham Biosciences). The purified TilS was dialyzed against a crystallization solution composed of 10 mM Tris·HCl (pH 8.0) and 10 mM MgCl₂ and was concentrated. Crystals of A. aeolicus TilS were grown within 2 days at 20°C by the hanging-drop vapor diffusion method. Drops were prepared by mixing equal volumes of a 5 mg/ml TilS solution and the reservoir solution, containing 0.1 M sodium acetate (pH 4.6) and 3.5 M sodium formate. Selenomethionine-labeled TilS was prepared by the conventional method and was purified in the same manner as the WT.

Structure Determination and Refinement. The TilS crystals were harvested in 0.12 M sodium acetate buffer (pH 4.6) containing 4.2 M sodium formate and then were sequentially transferred into the buffers containing 10% and 20% glycerol at 10-min intervals for cryoprotection. The crystals were flash-cooled in a nitrogen stream at 100 K. All diffraction data sets were collected at the NW12 station at the Photon Factory (Tsukuba, Japan) and were processed with the hkl2000 suite (HKL Research, Charlottesville, VA). The data set at the peak wavelength of the multiple wavelength anomalous dispersion (MAD) experiment up to 2.63 Å was used to search for the locations of the selenium atoms, by using the program snb (11). Several trials by snb revealed all eight consistent peaks expected in the asymmetric unit. Subsequent phase refinements were performed with the program sharp (12), and the model was manually built into the electron density maps by using the program o (13). The model was refined against the reflections from the native crystal up to 2.4-Å resolution by using the program cns (14), as shown in Table 2, which is published as supporting information on the PNAS web site. Structure representations were prepared with pymol (15) and cuemol (http://cuemol.sourceforge.jp/en).

Kinetic Analyses of TilS Mutants. All of the mutations were introduced by PCR into the A. aeolicus tilS gene cloned in pET-15b. The TilS mutants were prepared as the WT enzyme (described above). After dialysis against 50 mM Tris·HCl (pH 7.8), 25 mM MgCl₂, and 2-mercaptoethanol, the WT and mutant TilS proteins were concentrated and were stored at –25°C as 50% glycerol stocks before the kinetic assays. The gene encoding A. aeolicus tRNA^Ile₂ was amplified by PCR and cloned into the plasmid pUC119 (Takara Shuzo, Kyoto) by using the BamHI and HindIII sites. A. aeolicus tRNA^Ile₂ was transcribed in vitro by T7 RNA polymerase at 37°C in a reaction mixture composed of 80 mM Hepes (pH 8.1)/12 mM spermidine/100 mM DTT/100 mM KCl/0.1 mg/ml BSA/10 mM each NTP/8 mM GMP/100 mM MgCl₂/0.24 units/μl RNase inhibitor (Takara Shuzo)/0.05 units/μl PPase (Sigma). The transcript was subjected to phenol/chloroform extraction and isopropanol precipitation and then dried. The tRNA transcript was purified by denaturing (urea) 10% PAGE, followed by HPLC using a Resource Q column (Shimadzu).

The lysidine formation assay was performed at 60°C in a reaction buffer containing 100 mM Tris·HCl (pH 7.8), 10 mM MgCl₂, 10 mM KCl, and 10 mM DTT. The final concentrations of A. aeolicus TilS in all reaction mixtures were 0.25 μM for WT and 1, 2, and 4 μM for the mutants. To calculate the kinetic parameters, the concentrations of two of the three substrates, ATP (2 mM), tRNA^Ile₂ transcript (40 μM), and l-lysine (4 mM), were fixed, and the concentrations of the other substrates were varied (40–640 μM ATP, 2.5–156.8 μM tRNA^Ile₂ transcript, and 0.25–4 mM l-lysine). The reaction was terminated by spotting an aliquot on a Whatman 3-mm paper filter, which was previously soaked with 200 μl of 5% trichloroacetic acid (TCA) and dried. The filters were washed three times with ice-cold 5% TCA and twice with ethanol and were dried, and their radioactivities were counted by a scintillation counter. The assay was carried out at least twice for each mutant. Kinetic parameters were calculated by fitting to a Lineweaver–Burk plot. In this kinetic analysis, the concentration of l-lysine could not be increased to >4 mM, because the concentration of the commercially available [U-¹⁴C] l-lysine is 165 μM, which is lower than the K_m value (629 μM) for l-lysine of the WT enzyme. Accordingly, the commercially available [U-¹⁴C] l-lysine had to be diluted with unlabeled l-lysine to the extent that the resulting scintillation counts are not below the threshold level.

Lysidine Formation Assay of tRNA^Ile₂ Mutants. The genes encoding the A. aeolicus tRNA^Ile₂ mutants were synthesized by PCR with mutated primers. The subsequent purification steps were the same as those used for the WT A. aeolicus tRNA^Ile₂ transcript. The lysidine-forming activity of the tRNA^Ile₂ mutants were measured in 50 μl of reaction buffer (100 mM Tris·HCl, pH 7.8/10 mM MgCl₂/10 mM KCl/10 mM DTT) containing 2 mM ATP, 10 μM tRNA^Ile₂ transcript, 250–4,000 μM [U-¹⁴C] l-lysine, and 0.125 μM A. aeolicus TilS. After 2, 5, and 10 min, a 15-μl aliquot was removed. The assay procedures were the same as those used for the aforementioned kinetic analyses.

Results and Discussion

Overall Structure. We solved the crystal structure of A. aeolicus TilS (317 aa) and refined the model to 2.42-Å resolution with an R_cryst of 21.8% and an R_free of 25.3% (Fig. 5, which is published as supporting information on the PNAS web site). The crystals belong to the P4₁2₁2 space group, with unit cell parameters of a = b = 81.96 Å and c = 302.6 Å (Table 2). The asymmetric unit contained the TilS dimer (Fig. 1A). The gel filtration and ultracentrifugation analyses revealed that A. aeolicus TilS behaves as a protein with a M_r of 57–62 kDa, corresponding to 1.50–1.65 times the M_r of the monomer (Fig. 6, which is published as supporting information on the PNAS web site). Each subunit of A. aeolicus TilS has a hammer-like shape composed of the N-terminal domain, referred to as the NTD (amino acid residues 1–217; blue and green in Fig. 1 A and B), and the C-terminal domain, referred to as the CTD (amino acid residues 249–317; red in Fig. 1 A and B), which are connected by a long α-helical linker (orange in Fig. 1 A and B). The NTD consists of a core of six stranded β-sheet flanked by nine α-helices, forming a dinucleotide-binding fold with a characteristic central hole; the CTD is composed of antiparallel β-sheets and three α-helices (Fig. 1 A–C).

Recently, the crystal structure of E. coli MesJ protein was determined as a function-unknown protein (PDB ID code 1NI5) (Fig. 1 D and E). A recent proteomics analysis has revealed that the MesJ functions in lysidine formation both in E. coli and Bacillus subtilis, and thus the gene was renamed as tilS. Although the coordinates of E. coli TilS were deposited in the Protein Data Bank as a monomer, a symmetrical operation of the E. coli TilS structure implies that the E. coli enzyme can also form a dimer (Fig. 1E) in a similar manner as A. aeolicus TilS. Actually, our recent gel filtration analysis indicated that E. coli TilS forms a dimer (data not shown). The solvent-accessible surface areas buried in the dimer interface are 1,800.5 Ų and 1,775.0 Ų for A. aeolicus and E. coli TilSs, respectively. The structures of the NTD and CTD of A. aeolicus TilS are quite similar to those of the NTD and CTD1, respectively, of the E. coli enzyme [rms deviation (rmsd) = 1.19 Å over 159 Cα atoms for the NTD and rmsd = 1.19 Å over 50 Cα atoms for the CTD] (Fig. 1 A and E). However, the orientations against the NTD and the surface electronic potentials are quite different between the CTD of A. aeolicus TilS and the CTD1 of the E. coli enzyme (described below). Furthermore, E. coli TilS has an additional CTD, CTD2, after CTD1, compared with the A. aeolicus enzyme (Fig. 1D). These structural differences may reflect the distinct manners of tRNA recognition by the two TilS enzymes, as described below.

Catalytic NTD Derived from the N-type ATP-PPase. The NTD of A. aeolicus TilS has the amino acid sequence ³²SGGVDS³⁷. This sequence is identical to the fingerprint sequence, SGGXDss (where X is any hydrophobic amino acid and lowercase letters are highly conserved amino acids), of the ATP-binding P-loop in the N-type ATP-PPase subfamily, such as GMP synthetase, NAD synthetase, asparagine synthetase, and argininosuccinate synthetase, which commonly adenylates a substrate and then attacks the adenylated substrate with a nucleophilic nitrogen from a second substrate (16–19). Furthermore, the catalytic domain of the ATP-PPase subfamily has been shown to adopt a dinucleotide-binding fold, which indicates that the TilS NTD was derived from the N-type ATP-PPase. Particularly, the TilS NTD shares the highest structural homology with the ATP-PPase domain of GMP synthetase (rmsd = 2.397 Å over 106 Cα atoms) (Fig. 1F) (16). Furthermore, the TilS NTD shares a high sequence similarity with the ATP-PPase domain of GMP synthetase (Fig. 7, which is published as supporting information on the PNAS web site). GMP synthetase includes an amidotransferase domain appended to the N-type ATP-PPase domain, which allows the synthesis of guanine 5′-monophosphate from a xanthosine 5′-monophosphate (XMP) by means of two steps, the adenylation of the C2 carbon of XMP and the nucleophilic attack by an ammonium derived from the amide group of glutamine (20, 21). By analogy, it is suggested that TilS first adenylates the C2 carbon of cytosine 34 of tRNA^Ile₂ and then catalyzes the nucleophilic attack of the adenylated C2 carbon by the activated terminal amino group of the l-lysine substrate (Fig. 2A). Because the soaking or cocrystallization experiment did not allow us to see the electron density corresponding to ATP and l-lysine, we superposed the TilS NTD on the AMP/PP_i-bound ATP-PPase domain of E. coli GMP synthetase (Fig. 2B) (16)

Intriguingly, in our docking model, the TilS NTD interacts with AMP and PP_i in the same manner as the ATP-PPase domain of GMP synthetase. The backbone carbonyl and amide groups of Phe-61 hydrogen-bond to the N6 amino group and the N1 atom, respectively, of the adenine ring of AMP, and the hydroxyl groups of Ser-32 and Ser-37 hydrogen-bond to the phosphate oxygens of PP_i. The side chain of Arg-113 in one monomer of A. aeolicus TilS was bent to stack with the adenine ring of AMP, as the side chain of Phe-315 of GMP synthetase, and the hydroxyl group of Tyr-114 hydrogen-bonds to O4′ of the AMP ribose. In E. coli TilS, the corresponding Arg-97 stacks with the ribose ring as well. Thus, our docking model suggests that some amino acid residues comprising the central hole of NTD, such as Ser-32, Asp-36, Ser-37, Phe-61, Arg-113, Tyr-114, and His-133, could participate in ATP binding. Therefore, we prepared the Ala mutants of Ser-32, Asp-36, Ser-37, Arg-113, Tyr-114, and His-133 and analyzed their kinetic parameters for ATP, l-lysine, and the A. aeolicus tRNA^Ile₂ transcript (Table 1). We tried to prepare the S32A mutant, but it was not expressed well in E. coli. The hydroxyl group of Ser-32 forms an intramolecular hydrogen bond with the backbone nitrogen of Gly-34 to stabilize the sharp turn of the P-loop, which forms an inner core of the TilS NTD. This interaction may explain why the S32A mutant was not properly folded and not expressed in E. coli. The other five mutants either completely impaired the lysidine-forming activity or increased the K_m for ATP 7- to 14-fold, although the H133A mutant decreased the K_m for ATP 2.5-fold, and the other parameters remained essentially unchanged. His-133 may participate in the release of AMP from the ATP-binding pocket after lysidine formation for efficient enzymatic turnover. These results support our AMP/PP_i docking model. Furthermore, the K_m values for l-lysine of the S37A and Y114A mutants increased 6- and 5-fold, respectively, concomitantly with the increase of the K_m for ATP, which implies that the central hole of NTD places the l-lysine-binding site near the ATP-binding site. It was reported that the closely related GMP synthetase exhibits a sequential and ordered reaction: ATP, xanthosine 5′-monophosphate (XMP), and the nitrogen source bind to the enzyme in this order, and the XMP adenylation occurs in the absence of an ammonium (21). Therefore, ATP, tRNA^Ile₂, and l-lysine are likely to sequentially bind to TilS in this order to fulfill the base-modification reaction.

A subdomain composed of a loop and a helix is appended to the C terminus of the TilS NTD (green in Fig. 1). This subdomain is absent from the other N-type ATP-PPase and is specific to TilS (Fig. 7). The subdomain includes a conserved sequence, ¹⁹⁰eDXS/TNXXXXJXRnXJR²⁰⁵ (where J is any hydrophobic amino acid), which forms an edge of the aforementioned characteristic central hole. In contrast, in the other N-type ATP-PPase domain, the central hole is replaced by a valley-shaped pocket. Therefore, this appended subdomain may have essential functions specific to TilS.

Catalytic Site for C34 Adenylation. In addition to the aforementioned residues involved in the ATP binding, several charged or polar amino acid residues that are strictly conserved in the TilS enzymes are clustered at the edge of the central hole of the NTD, suggesting that the reaction step after the ATP binding could occur near the central hole. Mutant analyses of these residues revealed their involvement not only in the binding of ATP and l-lysine, but also in the recognition of C34 of tRNA^Ile₂ (Table 1). The substitution of Ala for Glu-140 and Arg-205 similarly affected the kinetic parameters. The E140A mutant increased the K_m values for ATP and tRNA by 7- and 5-fold, respectively, and decreased the k_cat by 6-fold, whereas it barely affected the K_m for l-lysine (Table 1). The R205A mutant also showed 15-fold and 4-fold higher K_m values for ATP and tRNA, respectively, and a 16-fold lower k_cat, than the WT, with the K_m for lysine substantially unchanged (Table 1). Therefore, Glu-140 and Arg-205 may be involved in the first-step reaction, the C34 adenylation (ATP binding, tRNA binding, and/or transition-state stabilization). The R174A mutant showed a 6-fold increase in the K_m for tRNA, whereas all of the other kinetic parameters remained essentially unchanged (Table 1), suggesting that Arg-174 is involved in the tRNA recognition. In contrast, the mutation of Glu-106 to Ala caused 2-fold and 18-fold reductions in the K_m values for tRNA and k_cat, respectively; the K_m values for l-lysine and ATP remained unchanged (Table 1). These results suggest that Glu-106 functions in the release of the lysidine-modified tRNA^Ile₂ product, thus contributing to the effective turnover of the TilS reaction. On the basis of these kinetic analyses, we propose a docking model for C34 of tRNA^Ile₂ (Fig. 2B). In this docking model, the base moiety of C34 is recognized by Glu-140, Arg-174, and Arg-205. The side chain of Arg-174 is fixed by the carboxyl group of Asp-36 (Fig. 2B) and may also hydrogen-bond to the PP_i product, by analogy to the fact that, in the closely related GMP synthetase, the side chain of Lys-381 corresponding to Arg-174 is oriented by the equivalent of Asp-36 and forms a hydrogen bond to the PP_i product. Arg-201, which when mutated to Ala abolished the lysidine-forming activity, might have a role in flipping and fixing the C34 base, which stacks within the anticodon loop in the canonical tRNA.

The Appended Subdomain Essential for l-Lysine Recognition. As described above, the NTD, which is involved in the C34 adenylation, is followed by a subdomain that consists of a loop and a helix. This subdomain is characteristic of TilS and may have TilS-specific functions. Actually, Arg-201 and Arg-205 are supposed to have TilS-specific functions, including interacting with and recognizing C34 of tRNA^Ile₂ (see the previous paragraph). Therefore, we analyzed the kinetic parameters of the other conserved residues, Asp-191, Thr-193, and Asn-194, on the subdomain loop (Table 1). The substitution of Ala for the absolutely conserved Asp-191 and Asn-194 abolished the lysidine-forming activity (Table 1). The replacement of Thr-193 with Ala increased the K_m for l-lysine by 10-fold and reduced the k_cat by 7-fold but hardly affected the K_m values for ATP and tRNA^Ile₂ (Table 1). The maximum concentration of l-lysine used in the kinetic experiment was 4 mM (see Materials and Methods). Because the K_m value (6.1 mM) of the T193A mutant is >4 mM, the real k_cat value of T193A might be higher under appropriate experimental conditions with a sufficient l-lysine concentration. Therefore, we propose that Thr-193, and probably Asp-191 and Asn-194, are involved in l-lysine recognition.

The substitution of Ala for Trp-188, located just upstream of the appended subdomain, increased the K_m values for l-lysine, ATP, and tRNA^Ile₂ 8-fold, 5-fold, and 3-fold, respectively, whereas it did not affect the k_cat (Table 1). However, because the K_m value for l-lysine (5.2 mM) of W188A is >4 mM, the K_m values for ATP and tRNA^Ile₂ might be lower under appropriate experimental conditions, as in the case of the T193A mutant. Trp-188 forms a hydrophobic core, together with Val-38, Val-39, Tyr-171, and Ala-181. This hydrophobic core fixes the spatial arrangement of the second α-helix of the NTD, which bears the P-loop, and the α-helix of the subdomain, which follows the subdomain loop with Asp-191, Thr-193, and Asn-194 (Fig. 2). Also, in E. coli TilS, the corresponding Trp-174 forms a hydrophobic core in the same way. Nevertheless, the rmsd of the lysine-binding loop between the E. coli and A. aeolicus TilS structures is as large as 7 Å (data not shown), suggesting that the lysine-binding loop is quite flexible and that the conserved Trp residue is structurally important for the formation of the l-lysine-binding site as well as the C34 adenylation site. The mutation of Trp-188 may indirectly reduce the lysine-binding activity by augmenting the flexibility of the lysine-binding loop so it cannot bind l-lysine efficiently. The substantial flexibility of the lysine-specific loop may explain both the high K_m value (630 μM) for l-lysine and why no electron density for l-lysine was observed in the soaking experiment.

A sequence alignment of TilS and GMP synthetase revealed that eight residues are absolutely conserved (Fig. 7). Among them, Asp-36, Asp-137, and Glu-140 (residue numbering according to A. aeolicus TilS) are located near the phosphate group of the superposed AMP and the C2 carbon of the docked C34 of tRNA^Ile₂ (Fig. 2B), and, particularly, Asp-36 has been suggested to play a catalytic role in the amino transfer reaction in GMP synthetase (16). The mutations of Asp-36 and Asp-137 abolished the lysidine-forming activity, whereas the E140A mutant retained some activity (Table 1). Therefore, these structural and kinetic results suggest that Asp-36 and/or Asp-137 are involved in deprotonation of the ε-amino group of l-lysine. Based on these results, we constructed a lysine-docking model where Asp-191, Thr-193, and Asn-194 are involved in the recognition of the α-amino and α-carboxyl groups and Asp-36 and Asp-137 hydrogen-bond to the terminal amino group of the l-lysine substrate (Fig. 2B).

Recognition and Discrimination of tRNA^Ile₂. The posttranscriptional modification from C34 to lysidine 34 in tRNA^Ile₂ not only changes the codon specificity from AUG to AUA, but also converts the aminoacylation specificity from recognition by methionyl-tRNA synthetase to that by IleRS (9). Therefore, TilS must discriminate unmodified tRNA^Ile₂ from unmodified tRNA^Met species bearing the anticodon CAU. When we compare the nucleotide sequence of tRNA^Ile₂ with those of tRNA^Met isoacceptors (initiator tRNA^Met_f and elongator tRNA^Met_m) in A. aeolicus, only the G3·C70 and C4·G69 base pairs in the acceptor stem and the C29·G41base pair in the anticodon stem are specific to tRNA^Ile₂, which are replaced by the C3·G70, G4·C69, and G29·C41 base pairs, respectively, in two tRNA^Met isoacceptors (Fig. 8, which is published as supporting information on the PNAS web site). To identify the nucleotide elements for TilS discrimination of tRNA^Ile₂ from tRNA^Met species, the lysidine-forming activities were investigated for C3G70, G4C69, and G29C41 mutants of tRNA^Ile₂ (Fig. 3). The C3G70 and G4C69 mutants of tRNA^Ile₂ exhibited lysidine-forming activity similar to that of the WT tRNA^Ile₂, whereas the mutation of the C29·G41 base pair abolished the lysidine-forming activity (Fig. 3). These results revealed that the C29·G41 base pair is crucial for the discrimination of tRNA^Ile₂ by A. aeolicus TilS. Therefore, we manually docked the Thermus thermophilus tRNA^Glu complexed with the cognate gulutamyl-tRNA synthetase (PDB ID code 1N77) onto the A. aeolicus TilS so that the base pair corresponding to C29·G41 can bind to the positively charged area of the CTD and C34 can enter into the catalytic pocket in the NTD (Fig. 4 A and B). The used tRNA^Glu retains canonical l-shape and is suitable for the present docking analysis. The sequence difference between tRNA^Ile₂ and tRNA^Glu is not critical because the aim of the model construction is only to evaluate the distribution of the nucleotide elements recognized by TilS. The tRNA fits well to the molecular surface of A. aeolicus TilS. This docking model shows that C20 approaches closely so it can be recognized by the TilS CTD (Fig. 4 A and B). Therefore, we analyzed the lysidine-forming activity for the C20U mutant and tRNA^Ile₂ transcript with an additional uridine inserted between G19 and C20, as in the cases of tRNA^Met_f and tRNA^Met_m, respectively (Fig. 8). Both of these mutations reduced the lysidine-forming activity to the level of tRNA^Met_m (described below), which suggests that C20 is essential for A. aeolicus TilS to distinguish tRNA^Ile₂ from two tRNA^Met isoacceptors and strongly supports the integrity of our docking model (Fig. 4 A and B). The efficiency of the lysidine-forming activity for tRNA^Met_m was less than half of that for tRNA^Ile₂ at their plateau levels (Fig. 3), which indicates that unmodified tRNA^Met_m was loosely recognized by A. aeolicus TilS in vitro. However, in vivo, because the anticodon wobble position of tRNA^Met_m is modified into ac⁴C, competition with the modification enzyme may prevent tRNA^Met_m from being modified by TilS (22, 23). Therefore, these results suggest that TilS specifically modifies tRNA^Ile₂ in vivo.

In E. coli, it has recently been revealed that the C4·G69, C5·G68 (in the acceptor stem), and G27·U43 (in the anticodon stem) base pairs, which are specific to tRNA^Ile₂ as compared with tRNA^Met, play a key role in recognition by the cognate TilS (Y.I., A.S., T. Ote, J. Kato, Y.S., and T.S., unpublished data). Whereas the electrostatic potential is limited to one part of the CTD in A. aeolicus TilS, E. coli TilS has two positively charged areas, in a distinct part of the CTD1 and in the α-helical linker connecting the NTD and the CTD1 (Fig. 4). On the basis of these data, we also built the tRNA-docking model of E. coli TilS (Fig. 4C). As described above, E. coli TilS has CTD2 (equivalent to CTD of A. aeolicus TilS) (Fig. 1 B and D). The positively charged interdomain cleft between CTD1 and CTD2 grasps the acceptor stem of E. coli tRNA^Ile₂ (Fig. 4C). These functional and structural analyses clearly indicate that the tRNA recognition manners fundamentally differ between the A. aeolicus and E. coli TilS enzymes; A. aeolicus TilS does not recognize the acceptor stem of tRNA^Ile₂, in striking contrast to the E. coli enzyme, but strictly recognizes the C29·G41 base pair in the anticodon stem. These differences in the tRNA recognition mechanism between the two TilS enzymes are consistent with the fact that A. aeolicus TilS efficiently modifies E. coli tRNA^Ile₂ with the C29·G41 base pair but that E. coli TilS does not modify A. aeolicus tRNA^Ile₂ with G5·C68 instead of the C5·G68 base pair (Fig. 8) (data not shown). E. coli IleRS recognizes the C4·G69 and C29·G41 base pairs as well as the U12·A23 base pair, the anticodon trinucleotides, t⁶A37, and A73 (the discriminator base) of tRNA^Ile to discriminate it from the other tRNA species (24, 25). Particularly, the C29·G41 base pair is one of the strongest recognition elements (24), whereas C4·G69 serves as an antideterminant against methionyl-tRNA synthetase (26). Therefore, the E. coli and A. aeolicus TilS enzymes discriminate tRNA^Ile₂ from tRNA^Met by a strategy similar to that used by IleRS, but in individually distinct manners. Elucidation of the precise tRNA recognition mechanism should be facilitated by the structure determination of the TilS·tRNA^Ile₂ complex.

Supplementary Material