A rationally engineered misacylating aminoacyl-tRNA synthetase

Abstract

Information transfer from nucleic acid to protein is mediated by aminoacyl-tRNA synthetases, which catalyze the specific pairings of amino acids with transfer RNAs. Despite copious sequence and structural information on the 22 tRNA synthetase families, little is known of the enzyme signatures that specify amino acid selectivities. Here, we show that transplanting a conserved arginine residue from glutamyl-tRNA synthetase (GluRS) to glutaminyl-tRNA synthetase (GlnRS) improves the K_M of GlnRS for noncognate glutamate. Two crystal structures of this C229R GlnRS mutant reveal that a conserved twin-arginine GluRS amino acid identity signature cannot be incorporated into GlnRS without disrupting surrounding protein structural elements that interact with the tRNA. Consistent with these findings, we show that cumulative replacement of other primary binding site residues in GlnRS, with those of GluRS, only slightly improves the ability of the GlnRS active site to accommodate glutamate. However, introduction of 22 amino acid replacements and one deletion, including substitution of the entire primary binding site and two surface loops adjacent to the region disrupted in C229R, improves the capacity of Escherichia coli GlnRS to synthesize misacylated Glu-tRNA^Gln by 16,000-fold. This hybrid enzyme recapitulates the function of misacylating GluRS enzymes found in organisms that synthesize Gln-tRNA^Gln by an alternative pathway. These findings implicate the RNA component of the contemporary GlnRS–tRNA^Gln complex in mediating amino acid specificity. This role for tRNA may persist as a relic of primordial cells in which the evolution of the genetic code was driven by RNA-catalyzed amino acid–RNA pairing.

Fidelity in the expression of the genetic code depends on the accurate synthesis of aminoacyl-tRNAs. These reactions are catalyzed by the aminoacyl-tRNA synthetases in a universal two-step process involving generation of an activated aminoacyl adenylate intermediate (1). Four tRNA synthetases [glutaminyl-tRNA synthetase (GlnRS), glutamyl-tRNA synthetase (GluRS), arginyl-tRNA synthetase (ArgRS), and lysyl-tRNA synthetase (LysRS)], all relatively small monomers, require the presence of tRNA to synthesize aminoacyl adenylate (2). For this first half-reaction these four enzymes may therefore be considered as ribonucleoprotein (RNP) particles, with the catalytic functionality provided by the protein subunit. In the second half-reaction, this protein subunit then catalyzes transfer of the amino acid portion of the adenylate to the RNA subunit of the particle.

Nearly 20 years ago, de Duve (3) suggested that the tRNA synthetase-mediated matching of amino acids with the distinguishing structural features of cognate tRNAs [the tRNA identity set (4)] might occur by way of direct contact between the enzyme-bound tRNA and aminoacyl adenylate intermediate. Crystal structures of many tRNA–synthetase complexes bound to aminoacyl adenylate analogs and free amino acid substrates have since shown that this proposal is not correct: no direct tRNA-amino acid contacts have been observed (5). However, the absence of direct interactions does not rule out an indirect role for tRNA in helping to confer amino acid specificity. The tRNA requirement for aminoacyl adenylate synthesis in GlnRS, GluRS, ArgRS, and LysRS suggests that this subset of the contemporary tRNA synthetases may be particularly likely to retain a key role for tRNA in mediating its own pairing, a role that may have been exclusive to RNA in primordial organisms.

GlnRS and GluRS are related members of the subclass 1b tRNA synthetases and possess topologically identical Rossmann fold active site domains that bind ATP, amino acid, and the tRNA acceptor–stem helix and single-stranded 3′ end (6–8). Phylogenetic analysis has demonstrated that GlnRS is an evolutionary latecomer to the tRNA synthetase family, and that it is derived from the eukaryotic subclass of GluRS enzymes (9). GlnRS is presently found only in the eukaryotic cytoplasm and in some Gram-negative prokaryotes. In other organisms, Gln-tRNA^Gln synthesis is accomplished by way of a two-step pathway: a nondiscriminating GluRS (GluRS^ND) first synthesizes Glu-tRNA^Gln, and this mismatched aminoacyl-tRNA is then transformed to Gln-tRNA^Gln by a tRNA-dependent amidotransferase (10). Interestingly, whereas most GluRS^ND enzymes synthesize both cognate Glu-tRNA^Glu and Glu-tRNA^Gln, Acidothiobacillus ferrooxidans and Helicobacter pylori each possess two GluRS enzymes, one of which appears to be dedicated solely to the synthesis of the mismatched product (11, 12).

The evolutionary connection between GlnRS and GluRS suggests that the structural determinants conferring tRNA and amino acid specificity might also be related. Indeed, the Rossmann folds of Escherichia coli GlnRS and Thermus thermophilus GluRS are quite similar structurally (rmsd of 0.98 Å over backbone atoms of 76 aa from both halves of the fold), and the conformations of the acceptor stem and 3′ end of the tRNA bound to the enzyme surface are also highly conserved (refs. 7 and 8 and Fig. 1A). However, the overall sequence identity generally present between representatives of the GlnRS and GluRS families is only ≈20%, and the induced-fit rearrangements that occur upon tRNA binding also are not well conserved (7, 8, 13). Further, significant divergence also exists between the sets of nucleotides that specify tRNA identity in the two systems (14, 15). These considerations suggest that the enzyme determinants conferring specific matching of amino acid and tRNA may not be straightforward to uncover.

Fig. 1.

Comparison of GlnRS and GluRS crystal structures. (A) Superposition of the tRNA-bound structures of GluRS (purple ribbon) and GlnRS (magenta and gray ribbons depicting the first and second halves of the Rossmann dinucleotide fold). The acceptor stem and 3′ acceptor end of tRNA^Gln are shown in orange; these elements of tRNA^Glu are shown in red. QSI bound to GlnRS and ESI bound to GluRS are shown in dark and light green, respectively. (B) Closer detail showing interactions in the GluRS (light blue) and GlnRS (green) active sites. The only H bond made by the Gln side chain in GlnRS is with Tyr-211. Glu makes electrostatic H bonds with Arg-5, Arg-205, and Tyr-187.

We sought to exploit the large existing database of sequence and structural information in the GlnRS and GluRS systems in rational protein engineering experiments to deduce the structural determinants of specificity in these enzymes. By analogy to specificity conversion experiments in serine proteases and other systems, we defined, as a benchmark of understanding, the capacity to rationally interconvert the functional properties of the system by systematic mutagenesis (16). Only limited progress along these lines has previously been reported for any tRNA synthetase (17). Here, we show that the introduction of 22 amino acid replacements and one deletion in the Rossmann fold of E. coli GlnRS confers >16,000-fold improvement in the capacity of the enzyme to synthesize the mismatched Glu-tRNA^Gln. These experiments substantially define the origins of amino acid selectivity in this system and may provide unique reagents for investigating the possibility that structural features involved in early RNA–amino acid pairing may yet persist in contemporary tRNAs.

Results

Primary Binding Pockets of GlnRS and GluRS.

To examine the origins of amino acid discrimination in the GlnRS and GluRS enzyme families, we sought to convert E. coli GlnRS to a misacylating Glu-tRNA^Gln-synthesizing enzyme. The GluRS cocrystal structures show that substrate Glu makes direct electrostatic contacts with Arg-5 and Arg-205 at the base of the pocket and further accepts a hydrogen bond from the side-chain of Tyr-187 (ref. 7 and Fig. 1B). Both Glu carboxylate oxygens contact the positively charged arginines, thereby demonstrating a clear direct basis for discrimination against Gln (Fig. 1B). In contrast, the GlnRS structures show that the side chain NH₂ of substrate glutamine makes hydrogen bonds with a buried water molecule and Tyr-211 (the equivalent of GluRS Tyr-187), whereas the amide oxygen lies ≈4 Å from the guanidinium of Arg-30 (the equivalent of GluRS Arg-5) and accepts no evident hydrogen bonds (18, 19) (Fig. 1B). The glutamine interactions then are entirely with ambiguous hydrogen bond donor-acceptor groups. Further, the cocrystal structure of the E. coli GlnRS–tRNA^Gln complex bound to noncognate Glu showed that one carboxylate oxygen of Glu occupies the identical position to the NH₂ of cognate Gln (18), again suggesting that the local substrate contacts are insufficient to provide amino acid specificity. The noncognate structure also showed that the other carboxylate oxygen of Glu interacts directly with the guanidinium group of Arg-30, an electrostatic contact that appeared to also drive misorientation of the α-carboxylate oxygen nucleophile with respect to ATP and the 3′ A76 nucleotide of tRNA^Gln.

WT GlnRS catalyzes Glu-tRNA^Gln synthesis 10⁷-fold less efficiently than the cognate reaction (18). In considering the structural basis for this extraordinary level of specificity, we considered that the interaction of Arg-30 with Glu, but not Gln, in WT GlnRS might function as a negative selectivity determinant by excluding productive Glu binding in GlnRS. To test this hypothesis, we constructed and purified to homogeneity the C-terminal His₇-tagged GlnRS mutants R30A and R30K. Steady-state kinetics were used to measure Gln-tRNA^Gln and Glu-tRNA^Gln synthesis, using a highly sensitive assay in which the 3′ internucleotide linkage of the tRNA is ³²P-labeled (20). This approach obviates the need to use radiolabeled amino acid and permits the characterization of mutants with very weak amino acid binding affinities, a necessity for this study. R30A and R30K are reduced in catalytic efficiency (k_cat/K_M) for cognate Gln-tRNA^Gln synthesis by 10³- and 30-fold, respectively, with the majority of the deficiency in K_M in each case (Table 1). Neither R30A nor R30K showed detectable levels of misaminoacylation under steady-state conditions, even at Glu concentrations >1 M. Clearly, removal of the proposed R30 negative determinant does not confer enhanced Glu-accepting capacity upon GlnRS.

Table 1.

Kinetic analysis of wild-type GlnRS and first shell mutants

	Wild type*	R30A	R30K	C229R	C229R/Q255I	C229R/Q255I/S227A/F233Y
kcat, s−1[GLN]	3.2 ± 0.48	0.35 ± 0.1	2.60 ± 0.3	(2.5 ± 0.7) × 10−3	No activity	No activity
KM, mM[GLN]	0.26 ± 0.04	31.4 ± 3.1	6.3 ± 0.7	0.21 ± 0.10	No activity	No activity
kcat/KM, s−1·M−1[GLN]	(1.2 ± 0.1) × 104	11.1	413	11.9	No activity	No activity
kcat, s−1[GLU]	0.046 ± 0.013†	No activity	Weakly active‡	(3.2 ± 0.1) × 10−4	Weakly active¶	(5.0 ± 0.6) × 10−3
KM, mM[GLU]	No saturation(KM > 750 mM)	No activity	Weakly active	240 ± 10	Weakly active	230 ± 17
kcat/KM, s−1·M−1[GLU]	(9.5 ± 1.3) × 10−4	No activity	Weakly active	1.3 × 10−3 [1]§	Weakly active	2.2 × 10−2 [23]

*All values for wild-type GlnRS are taken from ref. 6. Error estimates for k_cat/K_M values for wild-type enzyme are derived from measurements of initial velocities under k_cat/K_M conditions ([S] < < K_M).

^†k_cat was derived from reactions in which the tRNA concentration was varied (6).

^‡Plateau glutamylation levels of 24% were reached under conditions of excess enzyme.

^§Numbers in brackets refer to fold-improvement compared with wild-type GlnRS.

^¶Plateau glutamylation levels of 13% were reached under conditions of excess enzyme.

Introduction of C229R into GlnRS.

Examination of GlnRS and GluRS sequences showed that the only fully conserved difference in the amino acid pocket is the exchange of Arg-205 in GluRS for Cys-229 in GlnRS. The conservation of both Arg-5 and Arg-205 in all GluRS, and the direct interaction of Glu with both residues, suggests that efficient aminoacylation with Glu in the Rossmann fold context requires two arginines. We recapitulated the GluRS binding environment by constructing the C229R mutant in GlnRS: this mutant possesses Arg-30 and Arg-229 in topologically equivalent positions to Arg-5 and Arg-205 of T. thermophilus GluRS. C229R GlnRS was decreased 10³-fold in Gln-tRNA^Gln synthesis; in contrast with the behavior of the R30A and R30K mutants, the effect in this case was entirely in k_cat (Table 1). The efficiency of Glu-tRNA^Gln synthesis was nearly identical to that of WT GlnRS. However, although no glutamate saturation was observed for WT GlnRS at concentrations up to 1 M, we found that the amino acid pocket of C229R was indeed more hospitable toward Glu: K_M of 240 mM was measured, albeit in the context of reduced k_cat.

To assess how the different environments in the binding pockets of GlnRS and GluRS influence the conformations of the two key arginine residues and their interactions with substrate, we cocrystallized GlnRS C229R bound to tRNA^Gln and either 5′-O-[N-(l-Gln)-sulfamoyl] adenosine (QSI) or 5′-O-[N-(l-Glu)-sulfamoyl] adenosine (ESI) and determined each structure to 2.6-Å resolution in a crystal lattice isomorphous to the WT GlnRS-tRNA^Gln complex (Table 2 and Fig. 2). The overall fold of the protein was retained in both mutant structures. The structure of C229R bound to tRNA^Gln and ESI showed that one Glu carboxylate oxygen accepts hydrogen bonds from Tyr-211 and Arg-229, and is also located 3.4 Å from the positively charged guanidinium group of Arg-30 (Fig. 2a). The other Glu carboxylate oxygen accepts a hydrogen bond from the O3′ sugar hydroxyl of the adenosine portion of ESI and does not interact directly with either arginine. Thus, even though Arg-30 and Arg-229 of C229R occupy similar positions to Arg-5 and Arg-205 of GluRS, respectively, the interactions made by the carboxylate of ESI showed less electrostatic complementarity in the GlnRS mutant (Fig. 2a). Here, the GluRS and C229R structures are compared as ternary complexes bound to cognate tRNA and the identical ESI analog in each case.

Table 2.

X-ray crystallographic data collection and refinement statistics

Structure	GlnRS(C229R):tRNAGln:QSI*	GlnRS (C229R):tRNAGln:ESI*
Space group	C2221	C2221
Cell constants, a, b, c, Å	237.7 93.3 114.8	237.1 92.9 115.5
Resolution, Å	60.0 – 2.6	60.0 – 2.6
Total observations	170,271	151,770
Unique hkls	38,621	38,933
Rmerge†	7.6 (41.6)	11.0 (46.9)
Completeness, %	93.5 (86.9)	94.5 (99.4)
Multiplicity, %	4.4 (4.1)	3.9 (3.9)
Rcryst‡	21.0% (60 −2.6 Å)	21.5% (60 −2.6 Å)
Rfree§	26.3% (60 −2.6 Å)	24.9% (60 −2.6 Å)
rms bonds, Å	0.007	0.006
rms angles, °	1.2	1.2
Waters	136	137
Sulfate anions	2	2

Numbers in parentheses refer to the highest 0.1-Å resolution shell.

*QSI and ESI are analogs of the glutaminyl adenylate and glutamyl adenylate intermediates, respectively, in which the O-P-O reactive linkage is replaced by N-S-O.

^†R_merge = (Σ_hΣ_i|<F_h> − F_hi|)/(Σ_hF_h), where < F_h> is the mean structure factor magnitude of i observations of symmetry-related reflections with Bragg index h.

^‡R_cryst = (Σ_hΣ_i|F_obs| − |F_calc|)/(Σ|F_obs|), where F_obs and F_calc are observed and calculated structure factor magnitudes.

^§R_free is calculated with a test set with 5% data removed.

Fig. 2.

Crystal structure of GlnRS C229R. Comparison of interactions made by the Glu side chain of ESI bound to GluRS (Left) and GlnRS C229R (Right).

In both C229R structures, the clearest response to the introduction of C229R is a reorientation of Gln-255 about its χ1 side-chain rotamer angle, to avoid a steric clash with the Arg-229 guanidinium group. In the new orientation, Gln-255 occupies the position of the equivalent Phe-230 of T. thermophilus GluRS and is more exposed to solvent at the surface of the Rossmann fold opposite to the active site (Fig. 1b). The reorientation of Gln-255 is accompanied by an outward shift of the backbone at that position, which propagates to the neighboring Tyr-256 residue. The phenolic ring of Tyr-256 packs against the N-terminal helix of GlnRS, extending from residues Thr-8 to Lys-23, at one edge of the Rossmann fold. In turn, this entire helix shifts outward by ≈0.5 Å in both mutant structures, perturbing a peripheral enzyme–tRNA contact between Gln-13 and Cyt-16 of the tRNA.

The reorientation of Gln-255 also places this residue in close contact with the backbone of a surface loop at positions Arg-237–Arg-238. In the structure of C229R bound to QSI, the steric overlap generates an outward shift of these and adjacent loop residues, perturbing electrostatic interactions made by the guanidinium groups of Arg-237 and Arg-238 with the tRNA acceptor stem backbone at positions Gua4 and Gua5 (Fig. 3). These contacts are adjacent to the G3–C70 pair that plays a crucial role in tRNA^Gln identity (8). Gln-255 reorients into a slightly different position in the C229R/ESI structure, and the disruption of tRNA acceptor stem contacts is not observed in that case. Nonetheless, the perturbations of the GlnRS-tRNA^Gln interface arising from introduction of the double-arginine GluRS amino acid identity signature suggest the existence of an embedded connection between the tRNA and amino acid specificities of the enzyme. They also suggest that the tRNA concentration dependence of C229R activity may be perturbed relative to WT GlnRS.

Fig. 3.

Superposition of WT GlnRS (blue, green) with C229R bound to QSI (red, orange). Arg-229 in the mutant forces the reorientation of Gln-255, which in turn propagates a conformational change to the interface of Arg-237/Arg-238 with the tRNA acceptor stem.

The longer-range perturbations in the C229R structures, although small, nonetheless suggest that the determinants specifying amino acid substrate specificity might be located outside of the primary binding pocket, at least in part. To test this hypothesis systematically, we first examined whether further substitution of other GluRS amino acids within the primary binding site might improve the capacity of C229R to accept glutamate. GlnRS arose from the eukaryotic subbranch of GluRS enzymes, and this subbranch has diverged significantly, throughout the entire structure, from a distinct prokaryotic GluRS subbranch represented by the T. thermophilus crystal structure (9). Therefore, we elected to introduce further mutations based on the sequence of a representative eukaryotic GluRS, the human enzyme.

The crystal structures show that residues equivalent to GlnRS amino acids Tyr-211, Asp-212, Asp-219, Ser-227, Tyr-233, Gln-255, and Glu-257, in addition to Cys-229 and Arg-30, each are located in or directly adjacent to the amino acid binding pocket. Among these, however, Arg-30, Tyr-211, Asp-212, Asp-219, and Glu-257 are conserved in both GlnRS and GluRS, and thus cannot be specificity determinants. To examine whether the four distinguishing amino acids at positions 229, 255, 227, and 233 contribute to amino acid substrate selectivity, we constructed the GlnRS mutants C229R/Q255I and C229R/Q255I/S227A/Y233F (Table 1). The C229R/Q255I mutant, chosen based on the observed large rearrangement of Gln-255 in response to the C229R mutation, is very weakly active toward both Gln and Glu substrates (Table 1). However, introducing Ala-227 and Phe-233 into the C229R/Q255I framework, to produce the quadruple mutant C229R/Q255I/S227A/Y233F, improves activity for synthesis of noncognate Glu-tRNA^Gln by 20-fold as compared with WT GlnRS. K_M toward Glu is identical to that measured for the single mutant C229R alone, suggesting that the additional substitutions do not improve the complementarity of the pocket for Glu, but instead repair local structural defects associated with introduction of the bulky Arg-229. These data demonstrate that the determinants of amino acid selectivity in GlnRS extend beyond the immediate amino acid binding site.

Extended-Loop GlnRS Mutants.

To examine which distal elements of the GlnRS structure are essential to confer amino acid selectivity, we constructed additional mutants in which amino acids in adjacent surface loops were converted to their equivalents in human GluRS. The 1° shell/L1 mutant features 12 amino acid replacements: the entire primary binding pocket, and the adjacent loop spanning amino acids Thr-214–Leu-231, are converted to their GluRS equivalents (Fig. 4). A second hybrid enzyme was also constructed, in which the 1° shell/L1 mutant was augmented by the further transplantation of a second GluRS-equivalent surface loop spanning amino acids 243–256 of E. coli GlnRS (1° shell/L1/L2 mutant; Fig. 4). Unlike enzymes possessing 1° shell replacements only, however, neither of these mutants could be expressed in the E. coli cytoplasm in BL21(DE3) cells, perhaps because of toxic effects arising from the synthesis of Glu-tRNA^Gln in vivo (21, 22). To circumvent this problem, we attempted to direct expression into the periplasm by placing a 17-aa gIII signal sequence at the N terminus of the constructs. We found that expression in each case required addition of 10 mM β-mercaptoethanol to the culture media at the time of induction. Interestingly, N-terminal sequencing of the 1° shell/L1/L2 mutant purified from whole-cell lysates showed that the signal sequence tag was not removed by signal peptidase. Although we have not investigated whether a fraction of the expressed mutant might be secreted but not cleaved, the relatively high expression levels (≈0.5 mg of purified enzymes per liter of culture) suggest that it is primarily retained in the cytoplasm, perhaps associating with the Sec translocation machinery in a manner that precludes interaction with tRNA and thus relieves toxicity (23). As a control, we showed that the C-terminal (His)₇-tagged WT GlnRS expressed with the same N-terminal 17-aa leader peptide yields active enzyme exhibiting identical k_cat and K_M to GlnRS lacking this leader (data not shown), a finding consistent with the observation that the N-terminal seven amino acids of the enzyme are disordered in both tRNA bound and unliganded states (8, 13). Thus, the presence of the additional N-terminal peptide on the 1° shell/L1 and 1° shell/L1/L2 mutants is unlikely to affect kinetic parameters.

Fig. 4.

Superposition of GlnRS (green) and GluRS (magenta) indicating the location of the surface loops (L1 and L2) within the second half of the Rossmann fold. Key amino acids at GlnRS positions 211, 219, 227, 229, 233, 255, and 257 are indicated in green in the sequence alignment.

Kinetic analysis demonstrated that the 1° shell/L1 mutant shows no detectable activity for synthesis of Gln-tRNA^Gln (Table 3). Remarkably, however, the catalytic efficiency of this enzyme for synthesis of mispaired Glu-tRNA^Gln was increased ≈3,000-fold as compared with the activities of both WT GlnRS and GlnRS C229R. Although most of the improvement in this enzyme is at the level of k_cat, K_M toward Glu was also decreased by 5-fold as compared with C229R and C229R/Q255I/S227A/Y233F, demonstrating substantially improved complementarity for the noncognate substrate. These data show that distal portions of the GlnRS structure are essential to conferring amino acid specificity. To examine the properties of this enzyme further, we asked whether the engineered pocket might accommodate cognate Gln if the primary Arg-229 GluRS determinant were reverted to Cys-229, as in WT GlnRS. Interestingly, this 1° shell/L1/R229C enzyme does reacquire the ability to synthesize cognate Gln-tRNA^Gln, with k_cat/K_M reduced by ≈10³-fold as compared with WT GlnRS (Table 3). The Gln K_M was elevated by only 10-fold, demonstrating that GlnRS may not require a precise constellation of amino acid residues to accommodate its Gln substrate. This inference is consistent with the apparently nonspecific enzyme-Gln interactions observed in the crystal structures (18, 19).

Table 3.

Kinetics of extended-loop GlnRS mutants

	1° shell/L1	1° shell/L1/R229C	1° shell/L1/L2
kcat, s−1 [GLN]	No activity	0.034 ± 0.003	No activity
KM, mM[GLN]	No activity	3.18 ± 0.96	No activity
kcat/KM,s−1·M−1 [GLN]	No activity	10.7	No activity
kcat, s−1 [GLU]	0.12 ± 0.03	No activity	0.09 ± 0.02
KM, mM[GLU]	44 ± 12	No activity	5.8 ± 0.5
kcat/KM,s−1·M−1 [GLU]	2.7[2,800]	No activity	15.5[16,300]

All values for wild-type GlnRS are taken from ref. 6. The 1° shell/L1 mutant is C229R/Q255I/S227A/Y233F plus: T214A/H215C/C216P/S218V/A220S/L221I/I224V/L231T. The 1° shell/L1/L2 mutant is the 1° shell/L1 mutant plus: V243I/L244I/D245E/N246A/I247L/T248G/P250R/ΔV251/H252K/R254Y/Y256W. Numbers in brackets refer to fold-improvement compared with wild-type GlnRS. Values correspond to data averaged over both 30 and 60 μM tRNA concentrations, which were indistinguishable within experimental error.

The 1° shell/L1/L2 mutant, featuring 22 amino acid substitutions and one deletion with respect to WT GlnRS, possesses k_cat/K_M for noncognate Glu-tRNA^Gln synthesis that is improved by a further 6-fold as compared with the 1° shell/L1 enzyme, with all of the improvement manifesting in lower K_M toward Glu (Table 3). As observed for the 1° shell/L1 enzyme, this mutant is unable to synthesize Gln-tRNA^Gln. The overall improvement in k_cat/K_M for Glu-tRNA^Gln synthesis is now >16,000-fold, and the enzyme remains an 800-fold less efficient catalyst when compared with cognate Gln-tRNA^Gln synthesis by WT GlnRS.

The large number of amino acid substitutions in the 1° shell/L1 and 1° shell/L1/L2 mutants suggests the possibility that the alteration of amino acid specificity may well have influenced the ability of these enzymes to interact with tRNA. To explore this question, we carried out steady-state glutamylation experiments at 50 nM enzyme/2.8 μM tRNA, 500 nM enzyme/30 μM tRNA, and 500 nM enzyme/60 μM tRNA, for each mutant. Each enzyme showed reproducible and identical steady-state kinetic behavior at 30 and 60 μM tRNA (Table 3), but the behavior at 2.8 μM tRNA varied in different enzyme preparations. For the 1° shell/L1/L2 mutant, we also found that the enzyme activity decreased when stored under our standard conditions in a neutral pH buffer at −20°C in 50% glycerol. These results suggest that tRNA binding may enhance the stability of the enzyme. The relatively high required tRNA concentrations provide another demonstration of the interdependence of tRNA and amino acid binding sites in GlnRS and show that improved designs toward more efficient misacylation must also consider repair of the tRNA interface.

Discussion

The experiments demonstrate that the specificity of E. coli GlnRS for glutamine is a distributed property of the Rossmann fold, rather than being determined by local interactions in the immediate binding pocket. Remarkably, the distal protein loops that are required for the specificity conversion function primarily by improving the complementarity of the primary binding pocket for glutamate, as manifested by the large reduction in K_M for this noncognate substrate (Tables 1 and 3). Improved K_M is necessary for successful specificity redesign in this system, because the 10⁷-fold discrimination is manifested primarily in this parameter (6, 18). k_cat for Glu-tRNA^Gln synthesis is only improved by 2- to 3-fold in the 1° shell/L1 or 1° shell/L1/L2 mutants, yet remains just 10- to 50-fold below canonical values typically observed for cognate tRNA synthetase reactions. Because catalytic residues that interact directly with the reactive moieties of ATP, amino acid, and tRNA are unaltered in any of the mutants, it is likely that the remaining 800-fold kinetic deficit in the 1° shell/L1/L2 mutant has its origins in a failure to fully promote the precise juxtaposition of groups required for optimal catalytic efficiency. Another possibility is that the mutant is limited by release of aminoacylated product, as is the case for WT GlnRS (24). Interestingly, the replacement of several distal connective loops was also required to convert the amino acid specificity of rat anionic trypsin to that of chymotrypsin (25). Crystal structures of the hybrid enzymes in that system showed that the distal residues function by subtly altering the shape of the active-site cleft in the region that directly binds substrate (26). A similar structural mechanism likely operates here.

Our findings do not yet establish a minimal component of protein structure necessary and sufficient to provide glutamine specificity. Some of the substitutions toward GluRS made in the L1 and L2 loops, particularly of surface residues not interacting with more proximal portions of the protein, are likely not to be required for the 16,000-fold enhancement so far achieved. Of greater interest is the question of which further substitutions might confer the remaining 800-fold deficit. One possibility is that mutations in other adjoining areas of the protein might provide necessary and still-missing buttressing interactions to the primary shell. Several peptides within the acceptor binding domain of the enzyme (residues 100–210), and in the helical subdomain that follows the second half of the Rossman fold (residues 271–318), contact the regions already mutated and could augment or repair the interfaces between the GluRS and GlnRS portions of the hybrid enzyme.

A second and perhaps more provocative possibility is suggested by the observation that introduction of the GluRS twin-arginine identity signature into the primary amino acid binding pocket disrupts the GlnRS-tRNA interface (Fig. 3). This finding augments other information that also demonstrates that amino acid and tRNA specificities in GlnRS are coupled through the protein structure: the requirement for tRNA to catalyze aminoacyl adenylate formation, the dependence of tRNA binding affinity on the identity of the bound amino acid, and the weakening of glutamine affinity upon disruption of distant contacts between the enzyme and the anticodon (6, 24). Possibly, then, specific tRNA nucleotides, perhaps those in the adjacent acceptor stem, provide the still-required specificity-determining interactions. Through indirect readout, mutation of the tRNA acceptor stem toward tRNA^Glu sequences might modulate the sugar–phosphate backbone interactions to repair the structural alteration generated by C229R. Such a finding would suggest that structural elements of a primordial RNA-based synthetase remain in the contemporary GlnRS RNP and would validate the fundamental concept that tRNA plays an active role in its own specific aminoacylation.

Materials and Methods

Mutagenesis, Expression, and Enzyme Preparation.

Mutations were introduced into the E. coli GlnRS gene by using the Quikchange protocol (Stratagene). Introduction of the correct mutations was confirmed by DNA sequencing. His-tagged GlnRS enzymes were expressed intracellularly in BL21(DE3) pLysS cells (24). Purification of the mutants was by affinity chromatography on a nickel resin; mutants were recovered at > 99% purity as judged by SDS/PAGE.

Mutants that could not be expressed intracellularly in BL21(DE3)pLysS cells, because of apparent lethality, were expressed instead as an N-terminal fusion with the leader sequence of the bacteriophage fd gene III protein (Invitrogen) (27). Insertion of the gIII leader sequence at the N terminus of GlnRS was accomplished by ligation of a complementary pair of oligodeoxynucleotides into the NdeI site of pQRSH. The oligodeoxynucleotides contain a silent HindIII restriction site used to determine the orientation of the NdeI insert. Expression and purification of GlnRS mutants directed to the periplasm was performed in the same manner as for intracellular expression, except that 10 mM β-mercaptoethanol was added to the culture at the time of induction. N-terminal sequence analysis of the purified 1° shell/L1/L2 mutant was performed commercially by Alphalyse Europe on an ABI Procise 494 sequencer.

tRNA Preparation.

E. coli tRNA₂^Gln containing a catalytically neutral U1G mutation was synthesized by in vitro transcription. The duplex DNA template was first synthesized from two single-stranded oligodeoxynucleotides containing a complementary overlap duplex region as described (28). tRNA preparations used in this study could be glutaminylated to plateau levels of 70% or more.

Aminoacylation Kinetics.

tRNA was ³²P-labeled at the 3′ terminal internucleotide linkage by using the exchange reaction of tRNA nucleotidyltransferase as described (18, 20, 24). To ensure maximal substrate activity, labeled and unlabeled tRNA were mixed to the appropriate final concentration and heated at 65°C for 3 min. MgCl₂ was added to 7.5 mM, and the reactions were slow-cooled to ambient temperature. Aminoacylation reactions were quenched in a buffer containing 0.4 M sodium acetate (pH 5.2) (P1 nuclease digestions) or 0.6 M sodium acetate (pH 4.6), 100 μM zinc chloride (S1 nuclease digestions). P1 or S1 nuclease digestions were performed by adding 1–5 μl of the reaction mixture to a microtiter well containing 3–5 μl of 0.1 mg/ml P1 or S1 nuclease (Fluka) prepared in the buffers indicated and incubating for 10 min at ambient temperature. Aminoacylated tRNA (as 3′ aminoacylated A76) and unreacted substrate (as unmodified AMP) were separated by polyethyleneimine TLC developed in 10% (wt/vol) ammonium chloride, 5% (vol/vol) glacial acetic acid. Data were quantified by phosphorimaging analysis.

Steady-state assays were performed at 10 mM MgCl₂, 20 mM Tris (pH 7.5), 1 mM DTT, 5–10 mM ATP, and appropriate concentrations of glutamine or glutamate. Enzyme concentrations were maintained 20- to 1,000-fold below those of tRNA. The tRNA concentrations used were 15 μM for R30A and R30K, 2.8 μM for C229R and 1° shell/L1/R229C, 60 μM for C229R/Q255I and C229R/Q255I/S227A/F233Y, and 2.8, 30, and 60 μM each for the 1° shell/L1 and 1° shell/L1/L2 enzymes.

Crystallization and X-Ray Structure Determinations.

tRNA was prepared for crystallization by dialysis into 10 mM Pipes (pH 7.5)/10 mM MgCl₂, followed by microconcentration to 6 mg/ml. QSI or ESI were then added to the tRNA solution to concentrations of 1.8 mM (QSI) or 5.4 mM (ESI). The tRNA/analog solution was then mixed with equal volumes of a 6.3 mg/ml solution of GlnRS prepared in 5 mM Pipes (pH 7.0), 5 mM β-mercaptoethanol. Crystals of the GlnRS(C229R)–tRNA^Gln complexes bound to the analogs were grown and cryoprotected as described (29, 30). QSI (2 mM) or ESI (2 mM) were included in place of the amino acid and ATP analog used in the prior study (30).

X-ray crystallographic data were collected at the Stanford Synchrotron Research Laboratory (Menlo Park, CA). Data were reduced with MOSFLM and the CCP4 suite of programs. The C229R mutant bound to tRNA and either QSI or ESI crystallized isomorphously to the native complex; both structures were solved by difference Fourier analysis. All crystallographic refinement and difference map calculation was carried out by using CNS (31), using a starting model consisting of the WT GlnRS:tRNA^Gln:QSI complex from which the QSI ligand, all active-site solvent molecules, and the side chains of C229 and Q255 had been removed. Modeling was performed by using O (32).

Sequences of glnS and gluS genes were aligned by using ClustalW. The 76 structurally equivalent residues in the Rossmann folds were defined as follows: E. coli GlnRS: 26–32, 40–64, 76–89, 211–228, 236–247 (Protein Data Bank ID code 2RE8); T. thermophilus GluRS: 1–7, 15–39, 51–64, 187–204, 212–223 (Protein Data Bank ID code 2CV2).