Insights into Editing from an Ile-tRNA Synthetase Structure with tRNAIle and Mupirocin

See allHide authors and affiliations

Science  13 Aug 1999:
Vol. 285, Issue 5430, pp. 1074-1077
DOI: 10.1126/science.285.5430.1074


Isoleucyl–transfer RNA (tRNA) synthetase (IleRS) joins Ile to tRNAIle at its synthetic active site and hydrolyzes incorrectly acylated amino acids at its editing active site. The 2.2 angstrom resolution crystal structure of Staphylococcus aureus IleRS complexed with tRNAIle and Mupirocin shows the acceptor strand of the tRNAIle in the continuously stacked, A-form conformation with the 3′ terminal nucleotide in the editing active site. To position the 3′ terminus in the synthetic active site, the acceptor strand must adopt the hairpinned conformation seen in tRNAGln complexed with its synthetase. The amino acid editing activity of the IleRS may result from the incorrect products shuttling between the synthetic and editing active sites, which is reminiscent of the editing mechanism of DNA polymerases.

The aminoacyl-tRNA synthetases that aminoacylate tRNAs with Ile, Thr, and Ala must discriminate against the amino acid containing one less methyl group. Based on thermodynamic considerations, Pauling (1) estimated that isoleucyl-tRNA synthetase (IleRS) could distinguish between Ile and Val by only a factor of 5; however, fewer than 1 error in 3000 occurs (2). To obtain the observed specificity of IleRS, a proofreading activity that hydrolyzes incorrect product is involved (3). It has been proposed (4) that selectivity is achieved by a “double-sieve” mechanism in which amino acids larger than the correct one are sterically excluded at the synthetic active site; in contrast, only smaller amino acids are sterically allowed to bind at a hydrolytic editing active site. The editing and synthetic active sites reside on separate domains (5). In Thermus thermophilus apo-IleRS, they are separated by 34 Å and have different specificities for amino acid binding (6).

We address here how amino acid editing is achieved by an active site that is 34 Å distant from the site where the incorrect product is formed and why editing of either the aminoacyl adenylate or the aminoacyl-tRNA requires tRNAIle (7). One possibility is that the active sites might be brought closer together by a rotation of the editing domain upon addition of tRNA to form a closed cavity that contains the tRNA 3′ terminus and both active sites (8). However, the crystal structure of S. aureus IleRS complexed with tRNAIle and Mupirocin presented here suggests an alternative mechanism involving a tRNA-dependent shuttling of the incorrect product between the two active sites.

The determination of the tRNA-IleRS structure is summarized in Table 1. The IleRS structure can be divided into three regions of differing function (Fig. 1A). The region that recognizes an unusual anticodon loop conformation is formed by three domains at the COOH-terminus and one at the extreme NH2-terminus. The dinucleotide (or Rossmann) fold domain forms the core that both activates and transfers the amino acid. The editing domain (connective peptide 1, or CP1) removes any incorrectly incorporated amino acid. tRNAIle binds along the long axis of the cigar-shaped IleRS structure. An antiparallel pair of α helices binds the minor groove of the tRNA acceptor stem; however, nucleotides in the D-loop of tRNAIle that were reported to be important for editing (7) do not interact directly with monomeric IleRS. Interactions between IleRS and tRNAIle that specify recognition of the tRNAIle sequence, as well as interactions with Mupirocin, are discussed elsewhere (9).

Figure 1

(A) Structure of the isoleucyl-tRNA synthetase complex with tRNAIle and Mupirocin. The ordered portion of the tRNA is shown in gold and the two modeled nucleotides are in silver. The Mupirocin drug is in pink. The protein is colored by domain or by section: NH2-terminal region (orange), Rossmann fold (green), editing domain or CP1 (yellow), acceptor stem binding domain (gray), CP2 domain (blue), helical domain (pink), COOH-terminal junction (turquoise), and Zn-binding domain (red). (B) Experimental electron density for the 3′ end of the tRNAIle. Good density specifies the positions of Ade73 (A73) and Cyt74 (C74), but little density exists for the Cyt75 (C75) and Ade76 (A76) nucleotides, which are model-built to extend the A-form conformation.

Table 1

Crystallographic data and structure refinement statistics. The IleRS from S. aureus (24) was overexpressed, purified, and cocrystallized (25) withE. coli tRNAIle, synthesized in vitro, and Mupirocin, which is an inhibitor. The nucleotides important for specifying tRNAIle identity (26) are identical in the E. coli and the S. aureus sequences of tRNAIle. Crystals were in space group P212121, diffracted to 2.2 Å resolution, and exhibited two unit cell sizes: a small cell (a= 70 Å, b = 100 Å, c = 180 Å), which formed in the presence of mercury ion, and a large cell (a= 70 Å, b = 100 Å, c = 186 Å). The structure was solved with a combination of multiple wavelength anomalous dispersion (MAD) and multiple isomorphous replacement (MIR) (27). The 18 selenium sites per asymmetric unit were located with the program Solve (28) on MAD data that had been locally scaled with MADPRB (29). Refinement in crystallographic and NMR system (30) with a maximum-likelihood Hendrickson-Lattman (MLHL) target yielded an overallR free value of 28.1% for reflections between 10 and 2.2 Å from the large cell crystal form and an overallR free value of 34.3% for reflections between 10 and 2.9 Å from the small-cell crystal form (27). Native data were collected at Cornell High Energy Synchrotron Source A1 beamline and at a CuKα source; four-wavelength MAD data were collected at Brookhaven National Laboratory (BNL) light source X12C, X4A, and X8C and beamlines at a CuKα source; Dy derivative data were collected at BNL X12C; and Pt derivative data were collected at home CuKα. Large unit cell Hg derivatives were prepared by soaking crystals in 0.1 mM HgI4 for 12 hours; small unit cell Hg derivatives were prepared by soaking crystals in 1 mM HgI4or 0.1 mM HgCl2 for 12 hours. The small unit cell crystals of selenomethionine-incorporated complexes were created by soaking large unit cell crystals in 0.1 mM HgCl2 overnight. Pt derivatives were prepared by soaking crystals in 0.1 mM cis-Pt(NH3)2Cl2 for 12 hours. Dy derivatives were prepared by soaking crystals in a stabilization solution 10% saturated with DyCl3 for 2 hours.R sym = Σ|I − 〈I〉|/ΣI, where I is the measurement of intensity. Figure of merit = cos〈Δφ〉. Phasing power = [Σ|F PH(calc)|2/Σ|F PH (obs) − F P (calc)|2]1/2, where F PH and F P refer to structure factors of derivatives and native data.R = Σ|F obs − F calc|/ΣF obsfor all reflections and R free is theR value for 5% of the reflections that were excluded in refinement, where F obs is the observed best native amplitude and F calc is the calculated one from atomic models.

View this table:

The synthetic and editing active sites are located at the bottoms of two deep clefts and are separated by 34 Å. To mark the synthetic active site we positioned a sulfamyl analogue of glutaminyl adenylate (QSI) from the structure of the glutaminyl-tRNA synthetase (GlnRS) ternary complex (10) into the IleRS structure (11); to accomplish this, we superimposed 114 α-carbon atoms in the Rossmann fold domains of the two enzymes with a root-mean-square deviation (rmsd) of 1.3 Å. The binding sites of Mupirocin and the imported adenylate analogue overlap (9). In the previously determined, ligand-free IleRS structure (6), the editing domain binds Val, and the active site region is identified by Tyr392 and His394, whose mutation alters the specificity of the editing activity (12). In this ternary structure, however, one wall of the editing active site cleft showed well-defined electron density, whereas electron density for the second wall was weak, which suggests some disorder.

The 3′ terminus of the tRNAIle acceptor stem is continuously stacked A-form RNA and is visible in the experimental electron density map to residue Cyt74 (Fig. 1B). We have modeled the last two disordered nucleotides to extend the acceptor strand as a continuously stacked helix, which positions the 3′ Ade directly into the editing active site. This extended, stacked conformation has previously been observed in the crystal structure of uncomplexed yeast tRNAPhe (13) as well as the structures of tRNAs complexed with class II synthetases (Fig. 2A). Moreover, given the observed positions of Ade73 and Cyt74, it is not possible to model build the 3′ end of the tRNAIle to place the 3′ terminus in the synthetic active site. By building the two tRNA 3′ terminal nucleotides as A-form RNA, the ribose of Ade76abuts His392 and Tyr394, residues that have been directly implicated in the editing specificity (12) (Fig. 1B). Thus, the complex crystallized here appears to resemble an “editing complex” instead of the “transfer complex” observed in the structure of GlnRS complexed with tRNAGln and QSI or ATP (10, 14).

Figure 2

(A) Superposition of the acceptor stems of complexed tRNAIle (gray thick line), complexed tRNAGln (black thin line), and uncomplexed tRNAPhe (light thick line). The acceptor strand of tRNAIle is similar to the continuously stacked conformation of tRNAPhe but differs from the hairpinned conformation of tRNAGln. (B) and (C) Superposition of tRNAGln on tRNAIle on a solvent contact surface representation of IleRS. The acceptor strand of tRNAIle, represented as a ribbon, is observed at the editing active site. Homology modeling of the tRNAGln hairpin conformation (ball and stick representation) places its 3′ terminus in the synthetic active site 34 Å away. Two orientations are shown to display the active site pockets and the cleft running between them.

Comparison of the structures of the IleRS and GlnRS complexes shows that aminoacylation of tRNAIle requires that its 3′ end adopt the alternative, hairpinned conformation found for tRNAGln (10, 14). To build a model of a tRNAIle conformation that would allow aminoacylation, we superimposed (15) 12 backbone phosphates of the tRNAGln acceptor stem on the equivalent region of tRNAIle, which gave an rmsd of 1.2 Å (Fig. 2A) and adjusted the phosphate backbone of nucleotides 73 to 76 to avoid steric clashes with the IleRS enzyme (Fig. 2B). In this hairpinned conformation, the bulged Cyt74 fits into a small crevice in the surface of the enzyme and the phosphate backbone is near several basic residues. Most notable among these putative interactions is one between the phosphate of Cyt75 and Arg455 of the Trp-Cys-Ile-Ser-Arg sequence, which is conserved among all synthetases of this class I subclass (16). The hairpinned conformation of the tRNAIle terminus is likely to differ from that of tRNAGln because the Gua73 of tRNAGln makes a sequence-specific contact with a backbone phosphate that cannot be made by the Ade73 of tRNAIle. Nevertheless, an analogous hairpinned conformation of the acceptor strand is both necessary and feasible when its 3′ terminus is being aminoacylated.

In addition, the conformation of the conserved Lys-Met-Ser-Lys-Ser (KMSKS) peptide differs between the GlnRS and IleRS complexes in a manner that may be important for editing. In the GlnRS complex, the Lys of the Val-Met-Ser-Lys-Ser sequence (Lys270) binds the α-phosphate of bound adenosine triphosphate (ATP), presumably to assist in catalysis (14). In the IleRS complex, the equivalent Lys598 is more than 15 Å away from the α-phosphate of the model-built aminoacyl-adenylate analogue and instead binds to the carboxylate tail of the inhibitor Mupirocin. Perhaps more significantly for editing, the backbone amide of Lys595 of the K595MSKS sequence, as well as the backbone carbonyl of Gly593, binds to the tRNA backbone, thereby stabilizing the extended tRNA conformation and producing an open active site. Because this conserved sequence is in a different orientation in the ligand-free structure of IleRS (6), the conformational change is induced by the binding of tRNA and may be related to the tRNA dependence of editing.

The editing domain could hydrolyze a misactivated or misacylated substrate in one of two different ways: either the editing catalytic site could move to within a few angstroms of the acyl bond of the erroneously acylated substrates or the substrate could shuttle between the two active sites. We suggest that the former possibility is sterically precluded, whereas the second is structurally and kinetically plausible. The fundamental problem with bringing the synthetic and editing active sites together to act on substrate bound to one site is the virtual impossibility of getting the catalytic sites close enough to each other, as they both lie on the bottom of separate clefts. Furthermore, the chemistry of the two reactions requires that the catalytic sites approach the substrate from the same side. Hence, the misacylated substrate must dissociate from the synthetic site and move to the editing site.

The editing mechanism of the IleRS may involve a competition between the two active sites for the same substrate (see Fig. 4). Editing of the misacylated tRNA may utilize the two known conformations for the 3′ end of tRNA: one, the hairpinned conformation seen in the GlnRS complex with tRNAGln, places the Ade76 in the synthetic active site, and the other, the continuously stacked conformation seen in uncomplexed yeast tRNAPhe, places Ade76 in the editing active site. If the rate of transfer of the aminoacylated Ade76 between the two active sites is faster than the dissociation of the tRNA, then a shuttling mechanism for the misacylated tRNA is possible.

Editing of a valyl-adenylate by a shuttling mechanism requires a rate of its dissociation from the synthetic active site that is fast relative to the rate of amino acid transfer. Further, the rate of adenylate dissociation or the rate of its hydrolysis must be increased by the binding of tRNAIle (7) to explain the dependence of valyl-adenylate editing on the binding of tRNAIle. Perhaps tRNA binding facilitates valyl-adenylate dissociation by shifting the KMSKS loop away from the adenylate, thereby increasing its dissociation rate. Additionally, the IleRS structure in the absence of ligands shows the editing domain overlapping the tRNA binding site (6), whereas the IleRS complex described here shows that the editing domain rotates by about 47° relative to the rest of the molecule, presumably in response to interactions with the tRNA (Fig. 3). In the tRNA complex a deep channel spans the two active sites (Fig. 3A), whereas a ridge lies between the active sites in the unliganded enzyme structure (Fig. 3B).

Figure 3

Change in orientation of the editing domain (yellow) relative to the catalytic Rossmann fold domain (green) produced by the binding of tRNA (white). The Rossmann fold domains of the liganded and unliganded enzymes were oriented similarly by least-squares superposition of the α-carbon atoms. (A) A cleft lies between the Mupirocin (ball and stick representation) and the expected position of the editing active site. The direction of the axis about which the editing domain rotates by 47° in going to the ligand-free form is shown. (B) The editing domain in the apo IleRS (6) overlaps the tRNA binding site.

The editing of either the misacylated tRNA or adenylate depends on nucleotides in the D-loop that must be the tRNAIle sequence (7). Because this D-loop sequence does not interact with the protein in this crystalline complex, the explanation for the dependence of editing on this sequence is not immediately apparent. Either the protein undergoes a major conformational change, perhaps bringing the editing domain in contact with the D-loop sequence, or the influence of this sequence is expressed through the tRNA structure. In either case, the misacylated substrates would have to travel between the two active sites.

A similar dynamic competition between the synthetic and editing sites is also used in DNA polymerase editing (17). In Klenow fragment of Escherichia coli DNA polymerase I, double-stranded DNA binds to the polymerase active site and a melted out single-stranded primer terminus binds to the editing active site (17). The shuttling of the DNA substrate between the two active sites of Klenow fragment can occur by dissociation from one site and reassociation with the other or by processive sliding (18). We propose that the editing of both misincorporated nucleotides by DNA polymerase and misaminoacylation of tRNA or ATP by IleRS may proceed by analogous shuttle mechanisms (Fig. 4).

Figure 4

Schematic cartoon comparing editing in DNA synthesis to editing in aminoacylation. In the synthetic mode of DNA replication, the 3′ end of the double-stranded DNA lies in the polymerase active site marked by a small circle between the fingers and thumb. In the editing mode, a single-stranded primer strand lies in the 3′ exonuclease active site (pocket), where nucleotides are removed. Similarly, in the synthetic mode of aminoacyl transfer, the acceptor strand of the tRNA forms a hairpinned conformation to pack against residues of the Rossmann fold; in the editing mode, the tRNA adopts the extended stacked conformation to place the amino acid in the editing domain (pocket).

  • * Present address: Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, 240 Longwood Avenue, Boston, MA 02115, USA.


View Abstract

Stay Connected to Science

Navigate This Article