tRNA-dependent Aminoacyl-adenylate Hydrolysis by a Nonediting Class I Aminoacyl-tRNA Synthetase*

Glutaminyl-tRNA synthetase generates Gln-tRNAGln 107-fold more efficiently than Glu-tRNAGln and requires tRNA to synthesize the activated aminoacyl adenylate in the first step of the reaction. To examine the role of tRNA in amino acid activation more closely, several assays employing a tRNA analog in which the 2′-OH group at the 3′-terminal A76 nucleotide is replaced with hydrogen (\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{tRN}\mathrm{A}_{2^{{^\prime}}\mathrm{H}}^{\mathrm{Gln}}\) \end{document}) were developed. These experiments revealed a 104-fold reduction in kcat/Km in the presence of the analog, suggesting a direct catalytic role for tRNA in the activation reaction. The catalytic importance of the A76 2′-OH group in aminoacylation mirrors a similar role for this moiety that has recently been demonstrated during peptidyl transfer on the ribosome. Unexpectedly, tracking of Gln-AMP formation utilizing an α-32P-labeled ATP substrate in the presence of \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{tRN}\mathrm{A}_{2^{{^\prime}}\mathrm{H}}^{\mathrm{Gln}}\) \end{document} showed that AMP accumulates 5-fold more rapidly than Gln-AMP. A cold-trapping experiment revealed that the nonenzymatic rate of Gln-AMP hydrolysis is too slow to account for the rapid AMP formation; hence, the hydrolysis of Gln-AMP to form glutamine and AMP must be directly catalyzed by the \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{GlnRS}{\cdot}\mathrm{tRN}\mathrm{A}_{2^{{^\prime}}\mathrm{H}}^{\mathrm{Gln}}\) \end{document} complex. This hydrolysis of glutaminyl adenylate represents a novel reaction that is directly analogous to the pre-transfer editing hydrolysis of noncognate aminoacyl adenylates by editing synthetases such as isoleucyl-tRNA synthetase. Because glutaminyl-tRNA synthetase does not possess a spatially separate editing domain, these data demonstrate that a pre-transfer editing-like reaction can occur within the synthetic site of a class I tRNA synthetase.

The specificity of protein synthesis depends upon the fidelity of aminoacyl-tRNA synthetases (aaRS). 1 These enzymes attach amino acids to the 3Ј terminus of transfer RNAs in a two-step reaction (1). First, the amino acid is activated by reaction with ATP, to yield an aminoacyl adenylate intermediate and pyrophosphate. In the second step, one of the two hydroxyl oxygens of the 3Ј-terminal A76 nucleotide of tRNA attacks the carbonyl carbon of the adenylate, producing aminoacyl-tRNA with release of AMP. Each synthetase must discriminate among both structurally similar amino acids and tRNAs, selecting only the cognate species from cellular pools with an overall accuracy of approximately one error per 10 4 -10 5 codons (2). While the subsequent interaction of aminoacyl-tRNA with elongation factors may also provide some selection (3), it is clear that the specificity of protein synthesis primarily arises from the tRNA synthetase-mediated step.
It is well established that some tRNA synthetases are unable to accurately discriminate among chemically similar amino acids based solely on interactions made in the synthetic active site (reviewed in Ref. 4). These enzymes possess an additional hydrolytic activity for deacylation of misaminoacylated tRNAs. This reaction occurs in a second active site that effectively excludes correctly aminoacylated products. For example, in IleRS the synthetic active site cannot efficiently exclude the smaller valine, resulting in synthesis of significant amounts of Val-tRNA Ile in addition to the cognate Ile-tRNA Ile (5)(6)(7). In a subsequent step, IleRS specifically deacylates Val-tRNA Ile , but not Ile-tRNA Ile , in a spatially separated editing site that is able to exclude the larger isoleucine on steric grounds (8).
Some class I tRNA synthetases (including IleRS) possess a large domain inserted between the two pseudosymmetric halves of the catalytic Rossmann fold. This inserted domain contains the editing site located some 30 Å from the synthetic active site (9,10). A crystal structure of IleRS bound to tRNA Ile showed that the 3Ј-end of the tRNA approaches the editing domain, suggesting a plausible structural mechanism in which the flexible 3Ј-end of misaminoacylated Val-tRNA Ile translocates across the enzyme surface from the Rossmann fold domain into the distant editing active site (10). Similar editing reactions have since been established for the homologous class I LeuRS and ValRS enzymes, which also possess large inserted editing domains (11)(12)(13)(14). Editing has also been described in a number of class II tRNA synthetases, including ProRS, ThrRS, and AlaRS (15)(16)(17).
The structure-based translocation mechanism appears well founded for post-transfer editing reactions in which misaminoacylated tRNA is the substrate. However, some tRNA syn-thetases such as IleRS catalyze an additional hydrolytic reaction known as pre-transfer editing, in which hydrolysis occurs instead at the level of the misactivated aminoacyl adenylate (7). Mutational analysis has led to the suggestion that IleRS also catalyzes pre-transfer editing of valyl adenylate (Val-AMP) within its editing domain (18). To accomplish this, a complex tRNA-dependent translocation mechanism to move the misactivated amino acid from the synthetic to the editing active site has been proposed (19). However, a clear structural basis for such a mechanism is lacking, because no confined passageway between active sites is apparent that might serve to prevent dissociation of Val-AMP from the surface of the enzyme during translocation. Thus, the precise site at which pre-transfer editing occurs in IleRS remains unresolved.
We are investigating the mechanisms of amino acid discrimination in Escherichia coli glutaminyl-tRNA synthetase (GlnRS), a well studied class I tRNA synthetase that is capable of a very high level of amino acid selectivity apparently without the use of an editing mechanism (20 -22). GlnRS requires tRNA to activate the amino acid, a mechanistic feature shared by ArgRS, GluRS, and class I LysRS (1). In the overall two-step aminoacylation reaction, GlnRS discriminates against noncognate glutamate by 10 7 -fold (21). However, the requirement that tRNA be present for aminoacyl adenylate formation makes study of the individual mechanistic steps difficult. To isolate the first step of glutaminyl adenylate synthesis for kinetic characterization, we thus employed an A76 -2ЈH tRNA Gln substrate analog (tRNA 2ЈH Gln ), which prevents transfer of amino acid to tRNA but which is nonetheless likely to promote induced-fit transitions known to be required for active-site assembly (23). Kinetic measurements together with a high resolution crystal structure of the modified GlnRS⅐tRNA 2ЈH Gln complex suggested that the 2Ј-OH group of the tRNA plays a direct role in facilitating the amino acid activation step. Unexpectedly, we also found that the complex of GlnRS bound to tRNA 2ЈH Gln catalyzes hydrolysis of cognate glutaminyl adenylate to produce glutamine and AMP. This reaction is stereochemically identical to the hydrolysis of noncognate aminoacyl adenylates ("pre-transfer editing") catalyzed by certain other tRNA synthetases. Because class I GlnRS must hydrolyze Gln-AMP in its synthetic active site, this finding raises the possibility that pre-transfer editing by homologous class I tRNA synthetases such as IleRS may similarly be catalyzed in a region that is spatially adjacent or overlapping with the synthetic site, rather than in the separate domain used for post-transfer editing.

EXPERIMENTAL PROCEDURES
Preparation of Synthetic tRNAs-RNA oligonucleotides comprising tRNA Gln half-molecules, with a break between nucleotides A37 and U38, were purchased from Dharmacon (24,25). Oligonucleotides were deprotected by incubation at 60°C in 100 mM acetic acid (pH 3.8), lyophilized, and stored at Ϫ20°C until ready for use. RNAs were then resuspended in sterilized MilliQ water, mixed to a final concentration of 80 M each, and denatured in a buffer containing 30 mM Pipes (pH 7.5), 30 mM KCl, by heating to 90°C for 3 min., followed by incubation on ice for 5 min. The sample was then heated to 70°C in ligase reaction buffer (50 mM Tris-HCl, pH 7.8, 25 mM KCl, 10 mM MgCl 2 , 10 mM DTT, 1 mM ATP) for 5 min and then slowly cooled to room temperature. T4 RNA ligase (New England Biolabs) was added to 0.1 unit/l, ATP was supplemented to 1 mM, and the mixture incubated for 5 h at 37°C. The final concentration of each tRNA half-molecule in the ligation reaction was 40 M. Ligation was confirmed by denaturing polyacrylamide gel analysis. The intact, full-length tRNA was then purified by repeated phenol/ chloroform/isoamyl alcohol extraction followed by ethanol precipitation, and dialyzed against water prior to further use.
Enzymatic Transcription of tRNA-E. coli tRNA 2 Gln (containing a catalytically neutral U1G mutation to promote efficient transcription initiation) was transcribed in high yield from a synthetic DNA template. The template was constructed from complementary synthetic oligonucleotides containing a short central overlapping region, which were extended to form the full-length duplex by treatment with the Klenow fragment of Escherichia coli DNA polymerase I. The DNA template incorporated 2Ј-O-methyl sugars at the two 5Ј-nucleotides in the noncoding DNA strand, resulting in a high proportion of enzymatically active tRNA transcripts (26). Transcription was performed as described previously (27). The transcribed tRNA was purified by passage through a 5-ml DE-52 (Whatman) column. Transcribed tRNA was recovered by ethanol precipitation and dialyzed against highly purified water prior to further use.
Refolding of tRNA-All tRNAs prepared by either method were refolded immediately before use in kinetic experiments. tRNA (30 -200 M) was heated for 3 min in water at 70°C, followed by addition of MgCl 2 to a final concentration of 15 mM and slow cooling to ambient temperature.
ATP-PP i exchange assay-E. coli glutaminyl-tRNA synthetase was purified as described before (28,29). Conditions for the ATP-PP i exchange assay for both cognate glutamine and noncognate glutamate activation reactions with wild-type tRNA are reported in the accompanying report (30). When the activity of GlnRS⅐tRNA 2ЈH Gln was measured, the concentration of GlnRS was varied from 100 to 250 nM, and the concentration of tRNA 2ЈH Gln was 10 M. Synthesis of Glutaminyl-adenylate-Formation of Gln-AMP was measured in the presence of the substrate analog tRNA 2ЈH Gln , in which the 2Ј-OH group of A76 is replaced with hydrogen. GlnRS was preincubated with tRNA in a buffer containing 50 mM Tris-HCl (pH 7.0), 10 mM  (21), followed by quantitation by phosphorimaging analysis. Distinct calibration curves were constructed for each experiment. All assays were corrected for a very small background rate of ATP hydrolysis in the absence of GlnRS and tRNA. Data were obtained by summing volume densities for AMP and Gln-AMP. At low substrate concentrations Gln-AMP could not be independently quantitated, because it was close to the detection limits.
Steady-state parameters K m and k cat for glutamine and ATP were determined by varying concentrations of these substrates from 0.5 to 50 mM and from 0.1 to 5 mM, respectively. In both cases, the substrate that was not varied was present at 3-fold the K m value. Kinetic parameters were obtained from at least three independent measurements. Initial velocities obtained by time-course analyses were plotted against substrate concentration and fitted to the Michaelis-Menten equation with KaleidaGraph. K m and k cat were determined directly from these plots.
Non-enzymatic Hydrolysis of Gln-AMP-The rate of nonenzymatic Gln-AMP hydrolysis was measured by preincubating GlnRS (0.5 M) and tRNA 2ЈH Gln (5 M) for 10 min in the reaction mixture for Gln-AMP synthesis, in the presence of 83 M ␣-[ 32 P]ATP. This allows accumulation of enzymatically synthesized Gln-[ 32 P]AMP. Unlabeled ATP was then added at 10-to 1500-fold molar excess, and reaction time points were taken. The decay reaction was stopped by mixing 2 l of the reaction mixture with 2 l of 400 mM sodium acetate (pH 5.0), 0.1% SDS. TLC was then performed, and results were quantitated as described above.
Glutamate-dependent AMP Formation by GlnRS-Glutamatedependent AMP formation was measured by the ATP-hydrolysis assay as described above. For reactions performed with wild type in vitro transcribed tRNA Gln , the concentrations of glutamate, tRNA, and enzyme were 0.5 M, 25 M, and 0.5 M, respectively. For reactions performed with the tRNA 2ЈH Gln analog, the concentrations of tRNA were 5 or 25 M, and those of GlnRS were 0.5 or 5 M, whereas glutamate concentrations were varied up to 2 M. For steady-state kinetics in the presence of wild type tRNA Gln , ATP was present at 580 M and glutamate was varied between 50 mM and 1 M. When the control experiment with cognate glutamine was performed, 25 M wild type tRNA Gln and 0.5 M GlnRS were present in the reaction mixture. Glutamine was varied from 2 to 50 mM. Purified Ef-Tu was a generous gift from A.

Wolfson.
Post-transfer Editing-To measure enzymatic deacylation rates, tRNA Gln transcripts were end-labeled with 32 P at the 3Ј-terminal internucleotide linkage using the exchange reaction of tRNA-nucleotidyltransferase (21,31). Aminoacylation of tRNA Gln with glutamate or glutamine was first performed at 37°C in a buffer containing 50 mM Tris-HCl (pH 7.0), 10 mM MgCl 2 , 5 mM DTT, 5 mM ATP, 14 M GlnRS, 25 M unlabeled tRNA Gln , 1 M 32 P-labeled tRNA Gln , and either 20 mM glutamine or 0.5 M glutamate. Aminoacylated tRNAs were recovered and dialyzed for 5 h against 10 mM sodium acetate (pH 5.0) at 4°C. Glu-tRNA Gln and Gln-tRNA Gln (3-7 M) were then incubated at 37°C in a buffer containing 50 mM Tris-HCl (pH 7.0), 10 mM MgCl 2 , 5 mM DTT, and 50 -500 nM GlnRS. Reactions were stopped by mixing 1.5 l of reaction mixture with 3.0 l of 0.66 mg/ml P 1 nuclease (Fluka) in 0.15 M sodium acetate (pH 5.2). Following 10-min incubation at 20°C, 32 Plabeled 2Ј-aminoacyl-AMP was separated from 32 P-labeled AMP by TLC as described (21). The ratio of 2Ј-aminoacyl-AMP to AMP after nuclease digestion is equivalent to the ratio of aminoacylated versus nonaminoacylated tRNA in the reaction.
Two-step Aminoacylation-Noncognate aminoacylation of tRNA Gln with glutamate was performed using 32 P-labeled tRNAs prepared as described above. Reactions were carried out under conditions paralleling those for which possible glutamate editing was observed. Reaction time points for the slow noncognate aminoacylations were taken by hand, quenched, processed, and analyzed by TLC separation and phosphorimaging as previously described (21,32).
X-ray Crystallography-tRNA 2ЈH Gln for crystallization experiments was prepared by the two-step chemical synthesis and ligation approach (see above). Prior to crystallization, the tRNA was resuspended in a solution containing 10 mM Pipes (pH 7.5) and 10 mM MgCl 2 , dialyzed against this buffer, and then concentrated to 6.6 mg/ml in a 3000 molecular weight cutoff microconcentrator (Amicon). Crystals of the GlnRS-tRNA 2ЈH Gln complex bound to the ATP analog AMPCPP (in which the linkage between the ␣and ␤-phosphates is replaced by a methyl group) and glutamine were grown by microseeding with crystals of the GlnRS⅐tRNA Gln ⅐ATP ternary complex, as described (26). A 2-l aliquot of a diluted stock of crushed crystals in 2.25 M ammonium sulfate, 20 mM Pipes (pH 7.0), 10 mM MgSO 4 , 2 mM DTT, 10 mM glutamine was mixed with 2 l of the GlnRS⅐tRNA Gln complex containing 4 mg/ml GlnRS, a 2:1 mol ratio of tRNA to enzyme, and 5 mM AMPCPP. The crystals grew in 1-2 weeks by vapor diffusion over a reservoir containing 2 M ammonium sulfate, 10 mM Pipes (pH 7.5), 10 mM MgCl 2 , 2 mM DTT. After growth, crystals were stabilized for 5 min in a solution containing 2.25 M ammonium sulfate, 20% glycerol, 20 mM Pipes (pH 7.5), 2 mM DTT, 2 mM AMPCPP, 10 mM glutamine, and then transferred for 1 min to a second stabilizer containing 2 M ammonium sulfate, 20% (w/v) sucrose, 15% xylitol, 20% glycerol, 10 mM Pipes (pH 7.5), 2 mM DTT, 2 mM AMPCPP, 10 mM glutamine, prior to freezing.
X-ray crystallographic data were collected at the Stanford Synchrotron Radiation Laboratory. Data were reduced with MOSFLM and the CCP4 suite of programs (33). All crystallographic refinements and difference map calculations were carried out using CNS (34). Models of bound amino acid residues and the surrounding water molecules were generated using O. The starting model for refinement consisted of a previously refined, unpublished structure of GlnRS bound to tRNA Gln and AMPCPP in a ternary complex. From this model, the AMPCPP and all solvent molecules in the active site were removed and the occupancies of atoms in the adenosine 76 sugar ring were set to zero. Following an initial round of rigid body and simulated annealing refinement, an F o Ϫ F c electron density map was calculated. Based on difference density, a 2Ј-deoxy sugar (at position 76) and several of the water molecules were replaced in the rigid-body refined model. Water molecules were placed in spherical difference density with proper hydrogen bonding distance to at least one electronegative atom in the tRNA or protein model. Simulated annealing refinement and map calculation were again carried out, yielding an F o Ϫ F c map with clear density for the glutamine molecule. This map was used to position the amino acid in the complete rigid-body refined model. Simulated annealing and B-group refinement yielded the final refined model.

Influence of the tRNA Gln A76 2ЈOH Group on Glutamine
Activation-To examine the role of the tRNA Gln 2-OH group at the 3Ј-terminal A76 position in glutamine activation, we employed a chemically synthesized substrate analog in which this moiety is substituted with hydrogen (tRNA 2ЈH Gln ). This deoxy analog possesses all other structural features of unmodified wild-type tRNA Gln (steady-state kinetic parameters for modified versus unmodified tRNA 2 Gln species are identical (29)). tRNA 2ЈH Gln was inactive in a filter-binding assay measuring incorporation of 14 C-labeled glutamine into tRNA, whereas control wild-type tRNA Gln prepared by identical methods gave acceptor activities between 600 and 1000 pmol of glutaminyl-tRNA Gln /A 260 (26,29). Plateau levels between 50% and 80% aminoacylatable tRNA for the wild-type in vitro transcript were also determined directly using a TLC-based assay (21,31).
Measurements of glutaminyl adenylate synthesis in the presence of tRNA 2ЈH Gln by the ATP-PP i exchange assay showed that the 2Ј-OH group has a very large effect on the efficiency of the reaction. The K m for glutamine was elevated to 15 mM, whereas the k cat was reduced by 10 3 -fold (k cat ϭ 0.045 Ϯ 0.004 s Ϫ1 ; Table I). Thus, the 2ЈOH group at the terminal ribose of tRNA Gln plays a significant role in the formation of glutaminyladenylate. This may be directly due to an influence on the precise positioning of the reactive moieties of glutamine and ATP, because the 2Ј-OH group makes a direct hydrogen-bond to the ␣-phosphate of ATP in the ternary GlnRS⅐tRNA⅐ATP complex (35).
Although the 2Ј-OH group may also assist in more global aspects of assembling the active site, this appears less likely in view of the many enzyme-tRNA interactions that apparently help to drive the conformational transitions upon substrate binding (23). To address this question experimentally, we crystallized and determined the x-ray structure of GlnRS bound in a quaternary complex to tRNA 2ЈH Gln , the ATP analog AMPCPP, and glutamine. The site-specifically modified tRNA was produced in milligram quantities using a two-step procedure in Rates of Glu-AMP synthesis were too slow to permit derivation of individual k cat and K m values. c Rates reported in the bottom row of the table were obtained using the ATP-PP i exchange assay. Rates reported in the top two rows are derived from the direct aminoacyl adenylate synthesis assay.
which half-molecules with a break in the anticodon loop were first chemically synthesized. In the second step the tRNA halves were ligated with T4 RNA ligase, and the intact tRNA was then purified (26). The structure refined at 2.5-Å resolution shows that the substitution of 2ЈH in place of 2Ј-OH at tRNA A76 has only small local effects on the conformation of the complex ( Fig. 1 and Table II; see further discussion below). This supports the notion that the hydrogen-bond donated by the 2Ј-OH of A76 to the ␣-phosphate of ATP plays a direct critical role in stabilizing the latter moiety for attack by an ␣-carboxylate oxygen of glutamine.

Glutaminyl-adenylate Is Partitioned between Hydrolysis and Dissociation in the Active Site of a GlnRS⅐tRNA 2ЈH
Gln Complex-The ATP-PP i exchange assay measures Gln-AMP synthesis by way of the reverse reaction, incorporation of 32 P from PP i into ATP. To explore a different approach, we measured Gln-AMP synthesis directly by incubating the GlnRS⅐tRNA 2ЈH Gln complex in the presence of 50 mM glutamine and 0.5-1.7 mM ␣-[ 32 P]ATP. This approach has the advantage of tracking the fate of Gln-[ 32 P]AMP directly. Reaction products were separated by TLC ( Fig. 2A), by which [ 32 P]AMP, Gln-[ 32 P]AMP, and [ 32 P]ATP were easily resolved for quantitation. Conditions for which essentially all enzyme was fully bound by the tRNA analog were established at 0.5 M GlnRS and 5 M tRNA 2ЈH Gln . Linear increases in the formation of both [ 32 P]AMP and Gln-[ 32 P]AMP were then observed under steady-state conditions. Surprisingly, the rate constants (k obs ) for steady-state product formation derived from these data showed that AMP is formed 5-fold more rapidly than Gln-AMP ( Fig. 2B and Table III). Under the same conditions the rate of nonenzymatic ATP hydrolysis to produce AMP was at least 50-fold lower than that of enzymecatalyzed AMP formation; this rate was subtracted from the rates reported in Table III. Additional controls demonstrated that the inherent ATPase activity by GlnRS, measured in reactions in which the tRNA is omitted, is about 50-fold lower than the rate of AMP formation shown in Fig. 2.
Several experiments were performed to ascertain the origin of the excess AMP production in this assay. Although no incor-poration of [ 14 C]glutamine into tRNA 2ЈH Gln was detected, the low concentrations of commercially obtained radiolabeled amino acid available may have limited detection of aminoacylation at the 3Ј-OH position of tRNA. Such a reaction would generate AMP from glutaminyl adenylate after transfer of the amino acid portion of the intermediate to tRNA. To examine whether tRNA 2ЈH Gln is indeed a substrate for aminoacylation, we examined the products of a reaction performed in the presence of 50 mM unlabeled glutamine by acid gel electrophoresis (Fig. 3). No incorporation of glutamine into the 3Ј-ribose position was detected, even at very high GlnRS concentrations (Fig. 3, lanes j  and k), whereas wild-type tRNA Gln was almost fully aminoacylated under these conditions. Thus, the production of AMP is not a consequence of aminoacylation at the alternative position of the terminal ribose.
Two possibilities remain for the origin of AMP production in this reaction (Fig. 4). AMP might be generated via enzymecatalyzed hydrolysis of Gln-AMP to produce the glutamine and AMP products at the active site. Alternatively, Gln-AMP may first dissociate from the enzyme, followed by nonenzymatic hydrolysis of the labile mixed anhydride linkage. To distinguish these possibilities, the nonenzymatic rate of Gln-AMP production was measured. Labeled Gln-AMP was first prepared by incubating GlnRS in a reaction buffer containing tRNA 2ЈH Gln , glutamine, and ␣-[ 32 P]ATP (see "Experimental Procedures"). The reaction was then quenched by the addition of unlabeled ATP at 10-to 1500-fold molar excess with respect to ␣-[ 32 P]ATP, and the conversion of Gln-AMP to AMP monitored. The rate of Gln-AMP hydrolysis was insensitive to the concentration of ATP in this concentration range and yielded a rate constant roughly 100-fold lower than the rate of AMP formation in the enzymatic reaction (Fig. 2C). This rate is much too slow to account for the observed rate of AMP formation in the reaction, which is some 35-fold faster. Furthermore, when the reaction was performed under burst conditions (5 M GlnRS, 25 M tRNA 2ЈH Gln , 50 mM Gln, 1.7 mM ␣-[ 32 P]ATP), the amount of AMP formed was significantly higher than that of Gln-AMP even at first turnover (data not shown). This result also strongly suggests that the synthesized Gln-AMP partitions between dissociation and hydrolysis at the GlnRS active site when tRNA 2ЈH Gln is bound. Therefore, hydrolysis of Gln-AMP in the presence of tRNA 2ЈH Gln is an enzyme-catalyzed activity. The steady-state parameters for Gln-AMP synthesis by the

GlnRS⅐tRNA 2ЈH
Gln complex were also determined (Table I). Both k cat and K m (Gln) in this assay were nearly identical to those found by ATP-PP i exchange in the presence of tRNA 2ЈH Gln (Table  I). Interestingly, the k cat for synthesis of Gln-AMP by GlnRS was ϳ10-fold higher than in other tRNA synthetases for which the forward step was also measured (36 -38). In those enzymes, which synthesize aminoacyl-AMP in the absence of tRNA, it is known that the product release step is rate-limiting.
The enzyme-catalyzed hydrolysis of the cognate aminoacyl adenylate reported here as a property of the GlnRS⅐tRNA 2ЈH Gln complex has not previously been observed in other tRNA syn-thetases (36,37). We examined the crystal structure of the GlnRS⅐tRNA 2ЈH Gln ⅐AMPCPP⅐glutamine quaternary complex for clues toward the origin of this activity. As noted above, the overall structure of the quaternary complex is highly similar to that of the native structure. Side-chain interactions made by the glutamine are identical, and no additional water molecule that might play a hydrolytic role is evident. However, several differences can be observed in the immediate active site that may indicate the basis of a novel hydrolytic activity. First, the ␣-carboxylate group of glutamine is rotated away from its position in the native quaternary complex (Fig. 1). Perhaps for this reason, the ␣-phosphate of AMPCPP is well ordered in the tRNA 2ЈH Gln complex, whereas it is disordered in the native complex (in both cases, the ␤and ␥-phosphates are disordered). Second, the atomic B-factors for glutamine are substantially elevated in the modified complex: the average B-factor for the substrate increases from 28.3 to 59.8 Å 2 , whereas B-factors throughout the remainder of the active site, including for the 3Ј-end of the tRNA, are similar in the two structures. An increase in the mobility and alteration in the conformation of the free amino acid could reflect similar differences in the adenylate complex formed after the first step is completed. Both differences may reflect the capacity of a hydrolytic water molecule to penetrate the active site, giving rise to hydrolysis of the glutaminyl adenylate intermediate.
Synthesis of Glu-AMP by GlnRS-We measured rates of Glu-AMP formation by both the wild-type and GlnRS⅐tRNA 2ЈH Gln complexes using the alternative assay in which ␣-[ 32 P]ATP is used as substrate (Fig. 5 and Table I). For the wild-type com-  ) are the products of a two-step aminoacylation reaction, where Gln-AMP is the aminoacylated A76 nucleotide, which is presumed to migrate similarly to the Gln-AMP reaction intermediate (21). B, steady-state time course showing that AMP is formed faster than Gln-AMP. Observed rates from these plots are presented in Table III. C, nonenzymatic hydrolysis of Gln-AMP measured in solution, as described under "Experimental Procedures" (Table III).  plex, steady-state parameters for Glu-AMP synthesis were within 3-to 4-fold of those determined for Glu-AMP synthesis in the ATP-PP i exchange assay (Table I) (30). By contrast, k cat /K m for Glu-AMP synthesis by the GlnRS⅐tRNA 2ЈH Gln complex was further decreased by nearly 10 3 -fold. Thus, the A76 2ЈOH group is also critical to noncognate amino acid activation. Moreover, the GlnRS⅐tRNA 2ЈH Gln complex synthesized Glu-AMP with k cat /K m decreased by 10 4 -fold compared with cognate glutamine (Table I). This decrease was only slightly smaller than the 4 ϫ 10 4 -fold specificity observed in k cat /K m for the ATP-PP i exchange assay using wild-type tRNA Gln (30). Results from both assays thus support the conclusion that a significant portion of the discrimination against glutamate by GlnRS is achieved in the first step of aminoacylation. Additionally, although the 2Ј-OH group of A76 plays an important role in catalyzing adenylate synthesis, it does not significantly contribute to discrimination versus glutamate at this step.
In the Gln-AMP synthesis reactions, the appearance of Gln-AMP at a rate slower than AMP was clearly detectable (Fig. 2). By contrast, only AMP appeared in the Glu-AMP synthesis reactions regardless of whether such reactions were performed in the presence of tRNA Gln or tRNA 2ЈH Gln . In the case of tRNA 2ЈH Gln , which cannot be aminoacylated, this may reflect a faster rate of Glu-AMP breakdown as compared with its slow synthesis. However, a rationale for why Glu-AMP did not appear in the case of wild-type tRNA Gln reactions is less clear. In this case, the hydrolysis of Glu-AMP to produce glutamate and AMP would, if present, represent a true pre-transfer editing reaction. Such reactions are in general characterized by the consumption of excess ATP (with concomitant production of AMP) at levels that exceed those for stable aminoacylation under the same conditions (39). However, such an activity is not expected in the case of GlnRS, because the level of discrimination observed by the ATP-PP i exchange kinetics (Table I) (30) should suffice for accurate protein synthesis in the cell.
The amount of AMP generated in the Glu-AMP synthesis reactions in the presence of native tRNA Gln (Fig. 5) substantially exceeds the amount of tRNA present (see "Experimental Procedures"), suggesting that a portion of this AMP might indeed originate from a pre-transfer editing reaction. To investigate this further we measured the steady-state rates of Glu-tRNA Gln formation and deacylation under similar conditions. Glu-tRNA Gln formation was measured using a TLCbased assay (21) in which the 3Ј-terminal internucleotide linkage of the tRNA is labeled with 32 P using the exchange activity of tRNA nucleotidyltransferase. Misacylation reactions carried out at 25 M tRNA, 0.5 M glutamate, 0.5 M enzyme, and 580 M ATP yielded a reaction rate of 0.016 s Ϫ1 , very similar to the value of 0.02 s Ϫ1 measured in the ATP hydrolysis assay under identical conditions. The post-transfer deacylation assay was performed by incubating Gln-tRNA Gln or Glu-tRNA Gln in the presence and absence of GlnRS ( Fig. 6 and Table IV) and monitoring the hydrolysis of the aminoacyl linkage by TLC (see "Experimental Procedures"). Slow rates of deacylation had previously been suggested by the stable synthesis of Glu-tRNA Gln to 50 -60% plateau levels (21). These experiments showed that deacylation of Glu-tRNA Gln is not an enzyme-catalyzed activity and that the rate of deacylation (k obs ϭ 0.012-0.03 s Ϫ1 ) is similar to both the k cat of 0.04 s Ϫ1 for glutamate-dependent AMP production (Tables I and IV), as well as to the rate of Glu-tRNA Gln synthesis.
These data show that the rates of Glu-tRNA Gln synthesis and nonenzymatic deacylation are very similar to the rate of AMP synthesis, for reactions carried out under very similar experimental conditions. Thus, it appears likely that much, if not all, of the observed AMP production in the reactions using ␣-[ 32 P]ATP originates from multiple-turnover cycles of Glu-tRNA Gln synthesis and deacylation. Indeed, control reactions performed in the presence of cognate glutamine on a similar timescale also generated excess AMP due to post-deacylation recycling of tRNA Gln (data not shown). In an attempt to determine whether any AMP production originates from Glu-AMP editing, reactions were performed in the presence of the EF-Tu⅐GTP complex. EF-Tu should bind to aminoacylated tRNA to prevent hydrolytic cleavage of the labile aminoacyl ester bond (40). Unfortunately, thermodynamic compensation studies have shown that misaminoacylated Glu-tRNA Gln possesses extremely low binding affinity for the EF-Tu⅐GTP complex (3,41). Perhaps for this reason, we were unable to find reaction conditions in which EF-Tu was able to selectively bind Glu-tRNA Gln . Therefore, although a weak pre-transfer editing activity by GlnRS against glutamate remains conceivable, it has not been demonstrated by these experiments.
Noncognate tRNA Cys -induced Glutaminyl-adenylate Synthesis-As expected, we were unable to detect Gln-AMP synthesis in the absence of tRNA even at very high enzyme or substrate concentrations. Therefore, to explore the importance of the tRNA body in promoting correct active site assembly, we performed similar reactions with a noncognate tRNA 2ЈH Cys substrate. The GlnRS⅐tRNA 2ЈH Cys complex catalyzed Gln-AMP synthesis, but at a rate some 400-fold slower than the GlnRS⅐tRNA 2ЈH Gln complex (k cat /K m (glutamine) ϭ 1.1 Ϯ 0.4 ϫ 10 Ϫ2 s Ϫ1 M Ϫ1 ). Thus, even a noncognate tRNA is capable of driving at least a partial reorganization of the GlnRS active site, although certain specific nucleotide contacts are clearly necessary for precise alignment. The reactions were quantitated entirely based on AMP synthesis; we were unable to observe Gln-AMP production at any time point. This is presumably because the slow reaction necessitated much longer reaction times well beyond the 16min half-life of nonenzymatic Gln-AMP hydrolysis ( Table III). Dissociation of intact Gln-AMP from the enzyme followed by nonenzymatic hydrolysis in solution might then explain the lack of observed Gln-AMP in these reactions. Although we cannot rule out the possibility that the GlnRS⅐tRNA 2ЈH Cys complex can hydrolyze Gln-AMP, it seems more likely that the poorly assembled active site in this case lacks this activity. Acid gel analysis confirmed that tRNA 2ЈH Cys is not glutaminylated at the 3Ј-OH group (Fig. 3). A similar effect was observed in human GlnRS⅐tRNA system where bulk yeast tRNA, possessing poor acceptor activity in glutaminylation, promotes the activation of glutamine as efficiently as a good tRNA substrate (42). In contrast to GlnRS, ArgRS requires a cognate tRNA with good amino acid acceptor activity to stimulate its ATP-PP i exchange reaction (43). Thus, whereas both enzymes catalyze amino acid activation in a strictly tRNA-dependent manner, the specific tRNA sequence has greater influence on assembly of the active site in ArgRS. DISCUSSION A Role for tRNA in the Synthesis of Glutaminyl Adenylate-Comparison of the rates of ATP-PP i exchange in the presence of tRNA Gln and tRNA 2ЈH Gln shows that the 2Ј-hydroxyl group at A76 plays a crucial role in facilitating amino acid activation (30) ( Table I). With intact tRNA Gln , the k cat of 47 s Ϫ1 for ATP-PP i exchange (30) is similar to values obtained for other tRNA synthetases (16,43,44,45). However, substitution of tRNA 2ЈH Gln in the assay results in a 10 3 -fold decrease in k cat and 5-fold increase in glutamine K m . Previously, the use of tRNA 2ЈH Gln resulted in no measurable activity in ATP-PP i exchange, but these experiments were done at reduced glutamine concentrations and with modified tRNA produced by tRNA nucleotidyltransferase, which can generate heterogeneous mixtures (46).
Although there are no enzyme residues in direct contact with the A76 2ЈOH group (35), this moiety is positioned 2.7 Å from a nonbridging oxygen of the ATP ␣-phosphate. Because the hydrogen atom is located such that the O-H . . . O hydrogen-bonding angle is ϳ90°, the interaction is unlikely to possess significant hydrogen-bonding character. In a model for the transition state of glutamine activation, the rearrangement of the phosphate toward a pentacovalent configuration likely improves the hydrogen-bonding character of this interaction (35). Regardless, the removal of the A76 2Ј-OH group in the active site apparently suffices to destabilize the mutual orientation of the ATP ␣-phosphate and glutamine carboxylate oxygen sufficient to produce a nearly 10 4 -fold drop in catalytic efficiency. Such destabilization implies a direct catalytic role for tRNA in the first step of the reaction.
Consistent with the requirement for tRNA Gln for synthesis of glutaminyl adenylate, comparisons of the unliganded and tRNA-bound GlnRS structures show that this substrate plays an important role in facilitating required induced-fit transitions of the active site (23). This includes repositioning of key residues that bind both the adenine ring and phosphates of ATP (Arg-260 and Lys-270), and the ␣-amino and side-chain amide groups of glutamine (Asp-66 and Tyr-211). Although it is possible that the A76 2Ј-OH group is important to these largescale rearrangements, the position of the hydroxyl group in direct contact with the reactive moieties of glutamine and ATP makes a direct catalytic role more likely. The finding that noncognate tRNA Cys is capable of driving a partial assembly of the active site is consistent with this hypothesis. Additionally, the crystal structure of GlnRS bound to tRNA 2ЈH Gln , glutamine, and AMPCPP shows that no large-scale rearrangements occur in response to replacement of the 2Ј-OH with hydrogen (Fig. 1). The involvement of tRNA in synthesis of aminoacyl adenylate may be considered analogous to the catalytic role of RNA in other ribonucleoprotein assemblies (47). Indeed, the identical 2Ј-OH group of tRNA-A76 bound in the P-site has recently been shown to play a key role in catalyzing peptidyl transfer on the ribosome (48). Substrate-assisted catalysis has also been demonstrated for E. coli RNase P, where the 2ЈOH group at nucleotide Ϫ1 of the precursor tRNA substrate preferentially contributes to enzyme-substrate binding in the transition state (49). It is also of interest to note that in vitro selection experiments have demonstrated that RNA alone is able to catalyze the activation of an amino acid (50). The involvement of tRNA in the catalysis of aminoacyl adenylate formation may reflect the ancient ancestry of tRNA synthetases, and possibly their emergence from a primordial RNA world in which amino acid activation was an entirely RNA-based activity.
Hydrolytic Activity in the Synthetic Active Site-Some tRNA synthetases possess proofreading or editing mechanisms to assist in the overall fidelity of protein synthesis (5). It is known that misactivated amino acids can be eliminated by three routes: (i) dissociation of the noncognate complex to give free aminoacyl adenylate, which then undergoes hydrolysis in solution (kinetic proofreading (40)); (ii) hydrolysis of the aminoacyl adenylate when bound to the enzyme in a tRNA-dependent or tRNA-independent manner (pre-transfer editing); and (iii) enzyme-catalyzed deacylation of misaminoacylated tRNA (post-transfer editing) (51). Improved insight into editing mechanisms is obtainable through use of the adenylate synthesis assay using ␣-[ 32 P]ATP, as described herein. The fate of synthesized aminoacyl adenylate can be followed directly, and both the aminoacyl-[ 32 P]AMP and its hydrolytic product [ 32 P]AMP can be observed after appropriate separation by TLC. In comparison to the more commonly used assay employing ␥-[ 32 P]ATP, the main advantage of this approach is that kinetic proofreading and intrinsic enzymatic hydrolysis can be distinguished. By contrast, placing the label in the ␥-position of the ATP allows monitoring of ATP hydrolysis to produce pyrophosphate but not the hydrolysis of aminoacyl-AMP after its synthesis.
Here we show that Gln-AMP partitions between dissociation and hydrolysis in the active site of GlnRS⅐tRNA 2ЈH Gln complex. The AMP was demonstrated to originate from an intrinsic hydrolytic activity of the enzyme, with a rate 5-fold greater than the dissociation of intact Gln-AMP. Nonenzymatic hydrolysis (kinetic proofreading) was ruled out by measuring the rate of Gln-AMP hydrolysis in solution under the same conditions. The observed kinetic constant (k obs ) for dissociation of intact cognate Gln-AMP was similar to that found for dissociation of cognate aminoacyl-AMP from other tRNA-synthetases (36,37). However, the hydrolytic activity toward cognate aminoacyl-AMP intermediate observed in the GlnRS⅐tRNA 2ЈH Gln complex has not, to our knowledge, previously been observed. Although we could not unambiguously demonstrate the existence of a hydrolytic activity toward Glu-AMP by GlnRS, the hydrolysis of Gln-AMP to produce glutamine and AMP by the GlnRS⅐tRNA 2ЈH Gln complex is stereochemically identical to the pre-transfer editing of noncognate aminoacyl adenylates exhibited by some other tRNA synthetases. In GlnRS, this activity must occur within the confines of the synthetic site in the catalytic Rossman fold domain. Presumably, the tRNA analog disrupts or loosens the active-site structure to allow access of a hydrolytic water molecule that does not penetrate in the case of the native complex. This notion is supported by the crystal structure of the tRNA 2ЈH Gln complex, which shows reorientation of the glutamine ␣-carboxylate group as well as elevation in the B-factors for this substrate. However, in the homologous class I IleRS, ValRS, and LeuRS enzymes, it has been proposed that both pre-and post-transfer editing occur in a spatially separated "editing" domain that is inserted into the Rossman fold (12,18,52,53). The proposal is supported by crystallographic and mutational studies suggesting that both pre-and posttransfer editing substrates can bind to the editing domain, and that distinct residues are involved in each pathway (18,52). For post-transfer editing, the translocation of the 3Ј-end of misaminoacylated tRNA by ϳ30 -35 Å from the synthetic to the editing site presents little difficulty because the amino acid is covalently bound (10). By contrast, structural models for pretransfer editing are much less well rationalized.
The proposed model for pre-transfer editing by IleRS (and perhaps other class I tRNA synthetases) is highly complex (53,54). Pre-transfer editing is initiated with a "priming" step, whereby the misactivated amino acid is first covalently attached to tRNA and then shifted to the remote editing site. This "priming" post-transfer editing step then induces conformational changes in the enzyme that allow several subsequent pre-transfer editing hydrolysis reactions to occur in the editing site. Shuttling of noncognate adenylates from the synthetic to the editing site is proposed to occur while the enzyme is "caught" in the conformation induced by the priming step.
The main difficulty with this model is that there is no evident pathway by which the misactivated and highly labile adenylate can be kept sequestered from bulk solvent and stabilized from hydrolysis en route to the editing active site. Crystal structures of IleRS, LeuRS, and ValRS do not suggest the nature of such a pathway regardless of the orientation of the tRNA acceptor end on the enzyme. It is also difficult to conceive of a biological rationale for the evolution of such a complex mechanism. Misactivated aminoacyl adenylates are not toxic to the cell and are readily hydrolyzed after dissociation from the enzyme.
A pointer to the resolution of the pre-transfer editing enigma emerges from an early study of ValRS from yellow lupin seeds, which is able to catalyze pre-transfer editing in the absence of tRNA (38). Our observations demonstrating that glutaminyl adenylate is hydrolyzed by the class I GlnRS⅐tRNA 2ЈH Gln complex within the synthetic Rossman fold domain are also relevant. We suggest then that pre-transfer editing in IleRS, LeuRS, and ValRS occurs within the structurally homologous synthetic active sites of these class I tRNA synthetases, whereas their editing domains are dedicated to post-transfer editing only. As suggested for the analogous Gln-AMP hydrolytic activity of the GlnRS-tRNA 2ЈH Gln complex, binding of noncognate amino acids in the active sites of IleRS, LeuRS, and ValRS, followed by condensation with ATP, could result in local enzyme rearrangements allowing penetration of a hydrolytic water molecule. Of course, the editing site would not be identical to the synthetic active site, but would be created in the process of the misactivation reaction and would operate to exclude the (larger) cognate amino acid. For example, interactions made by the addi-tional methylene group of isoleucine compared with valine in IleRS could help to maintain a rigid enzyme active-site structure. When noncognate valine binds, the decreased rigidity of the enzyme⅐substrate complex could then allow penetration of a hydrolytic water molecule that is otherwise excluded. Because such a pre-transfer editing mechanism would likely be inefficient, it would operate as a crude filter and might generate only a small increase in overall fidelity. Such a modest improvement in a primordial IleRS, however, could provide a driving force for evolution of the more sophisticated post-transfer editing mechanism at a later time. Pre-transfer editing activities that still persist in contemporary enzymes could thus be viewed as remnants of the initial reaction.