Substrate Specificity of Bacterial Prolyl-tRNA Synthetase Editing Domain Is Controlled by a Tunable Hydrophobic Pocket

Background: Proofreading by aminoacyl-tRNA synthetases is a key translational quality control step.

Results: The alanine specificity of the prolyl-tRNA synthetase editing domain can be modulated by mutations within the conserved hydrophobic binding pocket.

Conclusion: Modulation of the editing domain specificity by mutation confirms a size exclusion-based mechanism.

Significance: Distinct editing mechanisms are required to clear mischarged tRNA species and ensure accuracy of translation.

Abstract

Aminoacyl-tRNA synthetases catalyze the covalent attachment of amino acids onto their cognate tRNAs. High fidelity in this reaction is crucial to the accurate decoding of genetic information and is ensured, in part, by proofreading of the newly synthesized aminoacyl-tRNAs. Prolyl-tRNA synthetases (ProRS) mischarge tRNA^Pro with alanine or cysteine due to their smaller or similar sizes relative to cognate proline. Mischarged Ala-tRNA^Pro is hydrolyzed by an editing domain (INS) present in most bacterial ProRSs. In contrast, the INS domain is unable to deacylate Cys-tRNA^Pro, which is hydrolyzed exclusively by a freestanding trans-editing domain known as YbaK. Here, we have used computational and experimental approaches to probe the molecular basis of INS domain alanine specificity. We show that the methyl side chain of alanine binds in a well defined hydrophobic pocket characterized by conserved residues Ile-263, Leu-266, and Lys-279 and partially conserved residue Thr-277 in Escherichia coli ProRS. Site-specific mutation of these residues leads to a significant loss in Ala-tRNA^Pro hydrolysis, and altering the size of the pocket modulates the substrate specificity. Remarkably, one ProRS INS domain variant displays a complete switch in substrate specificity from alanine to cysteine. The mutually exclusive aminoacyl-tRNA substrate specificities of the WT and engineered INS domains is consistent with the evolution of two distinct editing domains that function to clear Ala-tRNA^Pro and Cys-tRNA^Pro in vivo.

Introduction

Errors in protein synthesis often lead to misfolded or nonfunctional proteins that can be deleterious to cells (1). Several checkpoints exist to minimize such errors, including proofreading by aminoacyl-tRNA synthetases (aaRSs)³ (2–5), resampling of the aminoacyl-tRNA before its transport to the ribosome by elongation factors (6), and premature termination of the polypeptide chain by the ribosome (7). Because the errors associated with each step in translation are cumulative, uncorrected mistakes made during aaRS catalysis are more likely to get misincorporated into proteins. Hence, editing activities play critical roles in vivo (8, 9), and reduced proofreading by aaRSs is correlated with disease (10, 11).

AaRSs catalyze aminoacyl-tRNA synthesis in a two-step reaction. In the first step, amino acids are activated by ATP to form an aminoacyl-adenylate intermediate, and in the second step, the activated amino acid is transferred from the adenylate to the 3′ terminus of the tRNA. AaRSs can confuse cognate amino acids with isosteric noncognate amino acids. For example, prolyl-tRNA synthetase (ProRS) mischarges tRNA^Pro with alanine and cysteine, in addition to the cognate proline. Mistakes by aaRSs are generally cleared either by hydrolysis or selective release of the aminoacyl-adenylate intermediate (pre-transfer editing) or by hydrolysis of the mischarged aminoacyl-tRNA (post-transfer editing). Post-transfer editing is catalyzed by a distinct editing domain that is either appended to the ancient catalytic core of the aaRS or encoded by a freestanding domain.

In the ProRS system, several distinct homologous domains play a role in post-transfer editing. An editing domain (INS) that clears mischarged Ala-tRNA^Pro is appended to most bacterial ProRSs, whereas Cys-tRNA^Pro is cleared by a freestanding editing domain known as YbaK (12–14). Some species, including those that lack an INS domain, encode a freestanding domain known as PrdX that hydrolyzes Ala-tRNA^Pro (14, 15). At least three additional freestanding editing domain homologs of unknown function (YeaK, ProX, and PA2301) have been identified based on sequence similarity to INS and YbaK (16).

Although freestanding editing domains with homology to the editing domains of alanyl-tRNA synthetase (AlaRS) and threonyl-tRNA synthetase have also been identified, YbaK is the only known editing domain with a distinct substrate specificity relative to the homologous synthetase domain. We recently showed that YbaK uses a unique substrate-assisted mechanism of catalysis involving thiol-specific chemistry to cleave Cys-tRNA^Pro (16). We hypothesize that due to the similar sizes of cysteine and proline, the ProRS INS domain could not evolve a substrate binding pocket that hydrolyzed both Ala- and Cys-tRNA^Pro while excluding Pro-tRNA^Pro. Here, we explore the proposed steric exclusion mechanism of substrate discrimination by the INS domain using computational modeling and biochemical approaches. We identify a hydrophobic substrate binding pocket that is optimized for Ala-tRNA^Pro hydrolysis but can be engineered to allow larger amino acids such as cysteine to bind and be hydrolyzed efficiently. The tunable nature of the substrate binding pocket is consistent with a steric exclusion mechanism. Remarkably, an INS variant that hydrolyzed Cys-tRNA^Pro but not Pro-tRNA^Pro also failed to hydrolyze Ala-tRNA^Pro, in accordance with the evolution of distinct editing domains that act upon each mischarged substrate in vivo.

EXPERIMENTAL PROCEDURES

Materials

All amino acids and chemicals were purchased from Sigma unless otherwise noted. [³H]Alanine (54 Ci/mmol), [³H]proline (99 Ci/mmol), [³H]serine (33 Ci/mmol), and [α-³²P]ATP were from Amersham Biosciences and [³⁵S]cysteine (1075 Ci/mmol) was from PerkinElmer Life Sciences.

Molecular Dynamics (MD) Simulation of Enterococcus faecalis INS Domain

The crystal structure of E. faecalis ProRS (Protein Data Bank (PDB) code 2J3L, chain B, residues 224–407) was used as the starting structure for MD simulation. The protonation states of the titratable residues were determined by PropKa (17) at a pH of 7. The protein was charge-neutralized by addition of Na⁺ counterions, and an octahedral box of TIP3P (18) water molecules within 8 Å of the protein was created using the xleap package in AMBER10 (19). The all-atom MD simulations with explicit water molecules were performed using AMBER10 with the ff03 force field (20). The system was energy-minimized for 2000 steps, keeping the protein rigid, followed by a 5000-step minimization for the entire system. In a preproduction MD simulation of 100 ps, the temperature of the system was raised from 0 to 300 K with a small force constant on the protein to restrict any drastic changes. Final production MD simulations were performed for 15 ns under constant temperature and pressure. The MD trajectory was analyzed for stability using Visual Molecular Dynamics software. The atomic fluctuations were calculated using the ptraj module of AMBER10.

Molecular Modeling of 5′-CCA-Ala Bound to E. faecalis INS Domain

Following MD simulation of the free protein, 25 snapshots of the protein structure were extracted at equal time intervals across the 15-ns MD simulation and were used for molecular docking. The structure of the 5′-CCA-Ala ligand was generated using the xleap module of AMBER10. This ligand was docked onto the 25 structures of INS using AutoDock 4.0 (21). Conformation of all of the ligand torsions were sampled during docking using a genetic algorithm. Side chain torsions of several residues in the binding pocket predicted from existing structural and biochemical data as well as MD simulation of the free protein, including Ile-263, Lys-279, Val-281, Glu-316, Phe-330, Val-333, and His-366, were allowed to change while the rest of the protein torsions were fixed. Each docking simulation involved generation of 200 different conformers, which were then clustered using a root mean square deviation cutoff of 2.0 Å. Resulting clusters were manually inspected, and four docked structures, deemed probable complex structures based on biochemical data, were chosen for further analysis. These structures were minimized using the ff03 force field in AMBER10, followed by 15 ns of MD simulation, as described above. A binding energy for each ligand was calculated using the Molecular Mechanic-Poisson Boltzmann Surface Area module of AMBER10. The complex structure displaying the most stable trajectory and the best binding energy was further minimized and was used as the final model for the INS·CCA-Ala complex. Models of WT INS with noncognate substrates and H366A INS with cognate substrate were generated by manually mutating the residue of interest followed by energy minimization.

Protein Expression and Purification

His-tagged Escherichia coli wild-type (WT) ProRS (22), WT AlaRS (23), C666A/Q584H AlaRS (AlaRS-CQ) (23), and E. coli tRNA nucleotidyltransferase (24) were purified using the Talon cobalt affinity resin (Clontech) as described previously. The editing active site variants of E. coli ProRS were generated using the QuikChange site-directed mutagenesis kit (Stratagene). Mutations were confirmed by DNA sequencing (Genewiz). All E. coli ProRS variants were overexpressed in BL21(DE3) cells and purified as described for the WT protein. The concentration of E. coli AlaRS-CQ and tRNA nucleotidyltransferase was determined by the Bradford assay (25). The concentrations of WT and mutant E. coli ProRS and AlaRS were determined by active site titration (26).

Preparation of tRNA and Aminoacyl-tRNA Substrates

WT E. coli tRNA^Pro and G1:C72/U70 E. coli tRNA^Pro were prepared by in vitro transcription as described (27). Preparation of radiolabeled Ala-, Ser-, Cys-, Pro-, and 2-aminobutyric acid (2-Abu)-tRNA^Pro was carried out as described (12, 16). Briefly, [³H]Ala-tRNA^Pro (G1:C72/U70 variant to facilitate charging by AlaRS) was prepared by incubating 2 μm E. coli WT AlaRS, 8 μm G1:C72/U70 tRNA^Pro, and [³H]alanine (12.9 μm) in buffer A (50 mm HEPES (pH 7.5), 4 mm ATP, 20 mm KCl, 20 mm β-mercaptoethanol, 25 mm MgCl₂, and 0.1 mg/ml BSA) for 4 h at 25 °C. [³H]Ser-tRNA^Pro (G1:C72/U70) was prepared by incubating 4 μm E. coli AlaRS-CQ, 10 μm E. coli G1:C72/U70 tRNA^Pro in reaction buffer A containing [³H]serine (19 μm) for 1.5 h at 25 °C. [³⁵S]Cys-tRNA^Pro was prepared by incubating E. coli ProRS (8 μm) and E. coli tRNA^Pro (8 μm) with [³⁵S]cysteine (0.9 μm) and unlabeled cysteine (50 μm) in a reaction buffer containing 20 mm Tris-HCl (pH 7.5), 20 mm KCl, 10 mm MgCl₂, 25 mm dithiothreitol, and 2 mm ATP for 1 h at 37 °C. [³H]Pro-tRNA^Pro was prepared by incubating 2 μm E. coli WT ProRS, 8 μm tRNA^Pro, and [³H]proline (8.8 μm) in buffer A for 2 h at 25 °C. 2Abu-3′-[³²P]tRNA^Pro was prepared by incubating 8 μm E. coli ProRS, 8 μm 3′-[³²P]-tRNA^Pro, and 450 mm 2-Abu in buffer A for 1.5 h at 37 °C. 3′-[³²P]-tRNA^Pro was prepared using E. coli tRNA nucleotidyltransferase and WT tRNA^Pro as described previously (28). Following aminoacylation reactions, the substrate aminoacyl-tRNAs were phenol-chloroform-extracted, ethanol-precipitated, and stored at −80 °C for use in deacylation assays.

Deacylation Assays

To compare the loss in deacylation activity of WT and mutant ProRS, 5 μm enzyme concentration was used in all assays. Deacylation of Ala-, Ser-, Cys-, Pro-, and 2-Abu-tRNA^Pro was carried out at 25 °C, whereas deacylation of Cys-tRNA^Pro was carried out at 25 °C or 37 °C according to published protocols (13, 16). Reactions were performed using ∼1 μm aminoacyl-tRNA in 150 mm KPO₄ reaction buffer at pH 7.0. A background reaction was carried out in which enzyme was omitted. For [³H]Ala-, [³H]Ser-, [³⁵S]Cys-, and [³H]Pro-tRNA^Pro, the reactions were monitored by precipitating the tRNA on Whatman filter pads followed by scintillation counting (12). For 2-Abu-3′-[³²P]tRNA^Pro, reactions were quenched by adding 0.5 unit/μl P1 nuclease (Sigma-Aldrich) in 200 mm NaOAc (pH 5.0) on ice. After digestion for 20 min at room temperature, 1 μl of quenched mixture was spotted onto a PEI-cellulose TLC plate pre-run with water. Separation of 2-Abu-AMP and AMP was accomplished by developing the TLC plates in 0.1 m ammonium chloride and 5% acetic acid. The radioactivity was analyzed using a Typhoon PhosphorImager, and data were analyzed using ImageQuant TL software (GE Healthcare). The level of deacylation was determined from the ratio of 2-Abu-AMP remaining over the total AMP plus 2-Abu-AMP, as described previously (28). The fraction of aminoacyl-tRNA remaining was plotted as a function of time following subtraction of the buffer-only hydrolysis background reaction and fitted to a single-exponential equation to obtain k_obs. All assays were carried out in triplicate.

RESULTS

Several homologous domains, including the INS domain of bacterial ProRS and the freestanding editing domains, YbaK and PrdX, have been shown to enhance the fidelity of proline codon translation by editing mischarged tRNA^Pro. To understand the molecular basis of substrate selection and discrimination by the INS domain, we first used computational approaches to generate a model of the bacterial ProRS INS domain bound to an Ala-tRNA substrate analog, 5′-CCA-Ala.

MD Simulation of the Free INS Domain

Crystal structures represent one of many possible conformations that proteins may adopt and are often not well suited for the molecular docking of a large substrate. MD simulations of the free protein not only allow the sampling of these dynamic motions, but also provide information on the flexible regions of the protein that may be involved in substrate binding. We performed a 15-ns all-atom MD simulation of the E. faecalis ProRS INS domain (PDB code 2J3L, residues 224–407). Atomic fluctuations of backbone atom C, Cα, and N for each residue, calculated as the root mean square deviation from the starting structure over the course of the simulation, identified several flexible regions of the protein (Fig. 1A). In addition to the linker regions of the INS domain (Fig. 1B), flexible regions were generally localized around the Lys-279 residue, earlier identified as critical for catalysis (22, 29). Residue Gly-331, part of the highly conserved GXXXP motif hypothesized to play an analogous role to that of an oxyanion hole (30), was also found to be highly flexible. Residue His-366, which is equivalent to residue His-369 in E. coli ProRS and previously implicated in Ala specificity of the INS domain (22), does not show high flexibility but lies close to more flexible regions of the protein. The congruence between the flexible regions of the protein and the catalytically important residues is consistent with putative substrate binding in this region of the protein.

FIGURE 1.

Relative flexibility of residues in the INS domain of E. faecalis ProRS. A, root mean square of the atomic fluctuation of the backbone atoms for each residue of the INS domain over 15-ns MD simulation. Horizontal line represents the average fluctuation for all residues. Residues with greater than average fluctuation are shown in magenta for the linker region and in light blue for the others. B, ribbon representation of the INS domain with flexible residues colored as in A. Catalytically important residues Lys-279, Gly-331 (part of GXXXP loop), and His-366 are shown as stick models. First residues of some flexible regions are labeled.

Modeling of the INS·CCA-Ala Complex

A truncated substrate analog of Ala-tRNA, 5′-CCA-Ala, was docked into the putative active site of the protein. Twenty-five snapshots of the protein structure were extracted at equal time intervals from the MD trajectory of the free INS domain. A large grid, encompassing all of the flexible regions of the protein, was used for the docking of the substrate. All torsions of the substrate and side chain torsions of several bulky residues in the flexible regions of the protein, including Ile-263, Lys-279, Val-281, Glu-316, Phe-330, Val-333, and His-366, were conformationally sampled. The substrate generally clustered near the flexible regions, but the computed binding energy did not show a significant preference for any conformation. Because previous studies have shown Lys-279 and His-366 (His-369 in E. coli ProRS) residues to be important in catalysis and substrate specificity, respectively, a weak constraint that the Cα carbon atom of the substrate alanine should be within 10 Å of these residues was applied. Four unique conformations of the substrate were selected for further analysis (supplemental Fig. 1). These structures of the INS·CCA-Ala complex were energy minimized and subjected to 5 ns of equilibration MD and 10 ns of production MD. Complex A (supplemental Fig. 1) showed a stable trajectory during the MD simulation, as well as the best binding energy calculated by Mechanic-Poisson Boltzmann Surface Area, and was further optimized and selected as the final structure of the INS·CCA-Ala complex.

The model of the INS·CCA-Ala complex (Fig. 2A) reveals several key interactions of the substrate with the protein (Fig. 2, B–D). The A76 phosphate group interacts with the Lys-279 residue via charge-charge interaction. This interaction may be crucial for the positioning of the 3′-end of the tRNA in the INS active site and would explain the essential role of Lys-279 residue in catalysis. The adenine base of the substrate is predicted to form three H-bonds with the protein, one between N7 and the Leu-282 backbone amine and two between the N6 hydrogens and the Leu-282 and Ala-317 carbonyl groups. The adenine base also interacts with residue Phe-330 via a stacking interaction (Fig. 2B). These interactions may play an important role in orienting the substrate in a catalytically competent manner in the active site. The C75 and C74 bases bind on the surface of the INS domain and H-bond to residues Glu-316 and Asp-299, respectively. The 5′-end of the substrate remains exposed to the solvent in a manner that would be compatible with binding of full-length tRNA.

FIGURE 2.

Computed model of E. faecalis ProRS INS domain bound to CCA-Ala substrate analog. A, electrostatic potential surface of the protein. B, ribbon diagram of the protein showing interactions with the substrate. C, interactions of Ala in the substrate binding pocket. D, methyl side chain of substrate Ala binding in a hydrophobic pocket formed by residues Ile-263, Val-266, and Ile-277. Corresponding residues in E. coli ProRS, if different, are shown in parentheses.

The substrate alanine binds in a hydrophobic cleft in the INS domain. The amine group makes H-bonds with the backbone carbonyls of residues Gly-331 and Ile-333, as well as with the His-366 side chain. The substrate carbonyl group is directed toward the positively charged GXXXP loop (Figs. 2C and 3A). The methyl side chain of the alanine substrate binds in a hydrophobic pocket formed by the side chains of highly conserved residues Lys-279, Ile-263, and Val-266 (Leu-266 in E. coli ProRS) (Fig. 2D and supplemental Fig. 2). The partially conserved residue Ile-277 (Thr-277 in E. coli ProRS) interacts with Ile-263 and Val-266 and completes the hydrophobic pocket into which the substrate alanine binds; however, it does not interact directly with the substrate. Overall, this alanine binding pocket is relatively large, and based on the docking model, we predict that this pocket would be able to fit slightly larger hydrophobic substrates.

FIGURE 3.

Optimized structures of WT and H366A E. faecalis ProRS INS domain bound to cognate and noncognate substrates. WT ProRS INS domain in complex with CCA-Ala (A), CCA-Cys (B), CCA-Pro (C), CCA-Ser (D), and H366A ProRS INS domain in complex with CCA-Pro (E). Distances from the substrate carbonyl oxygen to the backbone N atoms of Gly-331, Ser-332, and Ile-333 of the GXXXP loop and from the substrate amino group to His-366 (A–D only) are indicated. H-bonding interactions of two active site water molecules are also shown. C74 and C75 nucleotides are not shown for clarity. Substrates are labeled in green and red for active and inactive substrates, respectively. Direction and magnitude of displacement of Cα atom of substrates, relative to the Cα of Ala substrate in WT ProRS, are shown in magenta. A small displacement in computational models for Ser deacylation by WT ProRS and Pro deacylation by H366A ProRS is consistent with their slightly reduced deacylation in experimental studies.

Model of INS Domain Bound to Noncognate Substrates

Biochemical data show that the INS domain selectively hydrolyzes Ala-tRNA^Pro but not Cys-tRNA^Pro or Pro-tRNA^Pro (12, 27). To understand the molecular basis of this selectivity, we generated models of the INS domain bound to 5′-CCA-Cys (Fig. 3B) and 5′-CCA-Pro (Fig. 3C) by changing the alanine residue of the docked substrate to cysteine or proline, followed by energy minimization. The model of INS·CCA-Ala (Fig. 3A) shows that the carbonyl group of the substrate points to the backbone amine groups of the GXXXP loop (residues 331–333). The oxyanionic tetrahedral intermediate formed during catalysis is likely stabilized by this low electron density oxyanion hole. As mentioned above, this orientation of the substrate is further stabilized by H-bonds of the substrate amine group (Fig. 2B).

There are two water molecules in our models that may play catalytic or structural roles. One water is situated close to the Cα of the substrate alanine and forms H-bonds with the 2′-OH (Fig. 3A) and Gly-261 carbonyl (data not shown). This water may play a catalytic role as it is positioned for nucleophilic attack on the carbonyl center of the substrate. Another water molecule on the active site floor (Fig. 3), which is also present in the crystal structure of E. faecalis ProRS and is highly stable in MD simulations, does coordinate with the Ser-280 side chain and with backbone atoms of Ile-278 and Gly-359. This water likely plays a structural role and is important in maintaining the overall configuration of the INS active site.

Upon substitution of alanine with cysteine, the large sulfhydryl side chain pushes the substrate 0.6 Å away from the floor of the binding pocket (Fig. 3B), resulting in a change in orientation of the substrate carbonyl relative to the GXXXP loop, as well as altering the position of the putative catalytic water molecule relative to the 2′-OH of the substrate. These relatively minor structural changes may result in destabilization of the transition state. Larger structural changes occur upon substitution with proline (Fig. 3C), resulting in a loss of one H-bond of the substrate amine, as well as loss of interaction of the substrate carbonyl with the GXXXP loop. The Cα of the substrate proline is also displaced by 0.7 Å relative to the Cα of the alanine substrate. These structural perturbations would not only result in weaker binding of the substrate but also a destabilized transition state.

Residue His-366 has been previously implicated in Ala specificity of the INS domain. An H369A variant of E. coli ProRS displays significantly reduced Ala-tRNA^Pro deacylation activity and gains Pro-tRNA^Pro deacylation capability (22). Loss of the H-bond between His-366 and the substrate amine may explain the loss in alanine deacylation activity. To understand the basis of Pro-tRNA^Pro deacylation by the mutant, we generated a model of H366A INS bound to CCA-Pro (Fig. 3E). This shows that proline is easily accommodated in the active site of the mutant enzyme, whereas the critical interactions between the substrate and the INS active site are not perturbed as they are in the case of WT INS (Fig. 3C). In WT ProRS, there is strong interaction between the substrate Ala-tRNA^Pro and the His-366 residue and a weaker interaction with Val-266. In contrast, in the case of the H366A mutant, the substrate Pro-tRNA^Pro lacks a close interaction with Ala-366, but appears to interact with the Val-266 residue strongly (supplemental Fig. 3). The better fit of proline in the mutated active site pocket together with this new interaction may explain the hydrolysis of Pro-tRNA^Pro by the H366A mutant.

Substrate Specificity of the WT INS Domain

The model of INS·CCA-Ala described above suggests that the INS domain might accommodate and hydrolyze aminoacyl-tRNA substrates larger than alanine. To test this hypothesis, we examined post-transfer editing of Ser- and 2-Abu-tRNA^Pro by E. coli ProRS. As expected, E. coli ProRS can readily hydrolyze Ala-tRNA^Pro (k_obs = 0.088 ± 0.0004 min⁻¹) but fails to hydrolyze larger substrates Cys- and Pro-tRNA^Pro (Fig. 4). Interestingly, hydrolysis of 2-Abu-tRNA^Pro (k_obs = 0.014 ± 0.005 min⁻¹) and Ser-tRNA^Pro (k_obs = 0.012 ± 0.001 min⁻¹) is observed, albeit with 6–7-fold lower efficiency. Hydrolysis of Ser- and 2-Abu, a nonnatural amino acid with an additional methyl group compared with alanine, is consistent with the expansive hydrophobic substrate binding pocket observed in our model. A model of INS·CCA-Ser (Fig. 3D), generated as described for CCA-Cys and CCA-Pro substrate analogs, suggests that the hydroxyl side chain of serine is stabilized by H-bonding to the structural active site water molecule (coordinated to Ser-280 residue) without any significant disruption of the carbonyl or amine geometries (compared with Fig. 3A). These results suggest that the larger size and less polar character of the cysteine sulfhydryl group may be responsible for the lack of hydrolytic activity against Cys-tRNA^Pro.

FIGURE 4.

Deacylation substrate specificity of WT E. coli ProRS. WT INS domain can deacylate Ala-tRNA^Pro (▾), 2-Abu-tRNA^Pro (△), and Ser-tRNA^Pro (♦) but cannot deacylate Cys-tRNA^Pro (□) or Pro-tRNA^Pro (▴). All experiments were performed at 25 °C. Substrate aa-tRNA and ProRS concentrations were 1 μm and 5 μm, respectively. Data points and error bars represent average and S.D., respectively, of three independent experiments. The extent of hydrolysis of each substrate by buffer (0.15 m KPO₄, pH 7.0) was subtracted prior to plotting the data.

Altered Substrate Specificities of INS Domain Mutants

Our model of INS·CCA-Ala predicts that mutation of the bulky hydrophobic residues comprising the substrate binding pocket to smaller residues, such as alanine, would result in significant loss of interaction with the active substrates and thus reduced deacylation activity. At the same time, these changes may allow larger substrates to bind in the active site and be deacylated. To test these predictions, we generated several site-directed mutants of E. coli ProRS and tested their activity against Ala-, 2-Abu-, Ser-, Cys-, and Pro-tRNA^Pro. Ile-263, Val-266, and Ile-277 residues in E. faecalis ProRS correspond to Ile-263, Leu-266, and Thr-277 residues in E. coli ProRS, respectively (Fig. 2D). The structure of E. coli ProRS generated by homology modeling using SWISS-MODEL (data not shown) suggests that Thr-277, although polar in nature, contributes to the hydrophobic pocket via its terminal methyl group. Accordingly, we generated single E. coli ProRS mutants I263A, L266A, and T277S for our analysis, as well as a I263A/L266A double mutant. In the case of nonconserved residue Thr-277, a change to serine rather than alanine was made based on inspection of an E. coli ProRS homology model generated using 3D-JIGSAW (data not shown) (31). This model predicts an intramolecular interaction of the Thr-277 hydroxyl group, which points away from the active site pocket, whereas the methyl side chain points into the proposed active site. Thus, Thr-277 was mutated to serine to preserve the predicted interaction of the hydroxyl group while altering the active site pocket size.

Consistent with our prediction, Ala-tRNA^Pro (Fig. 5A) and 2-Abu-tRNA^Pro (Fig. 5B), both substrates for WT ProRS, were either deacylated very poorly or not deacylated by the mutants examined. In contrast Ser-tRNA^Pro, another substrate of WT ProRS, was hydrolyzed by all the active site variants, in some cases more efficiently than the WT enzyme (Fig. 5C). Strikingly, the I263A/L266A variant displayed an ∼8-fold increased rate of hydrolysis (k_obs = 0.091 ± 0.006 min⁻¹) relative to WT ProRS. This behavior can be explained by the model of INS·CCA-Ser (Fig. 3D), which shows that the hydroxyl group of serine forms an H-bond with a structural active site water. Mutations in the hydrophobic pocket would generate more space for additional water molecules to enter the active site and form an H-bonding network, thus allowing the stabilization of the active site water coordinated to Ser-280, as well as the bound serine substrate. Consistent with this idea, single mutations in residues closer to this water (e.g. Thr-277 and Leu-266) result in enhanced Ser-tRNA^Pro hydrolysis, whereas changing a residue farther away (e.g. Ile-263) does not (Fig. 5C).

FIGURE 5.

Deacylation substrate specificity of WT and mutant E. coli ProRSs. Deacylation of Ala-tRNA^Pro (A), 2-Abu-tRNA^Pro (B), Ser-tRNA^Pro (C), Cys-tRNA^Pro (D), and Pro-tRNA^Pro (E) are shown. Deacylations by WT (●), I263A (▿), L266A (■), T277S (◊), and I263A/L266A (▴) E. coli ProRS are shown. All experiments were performed at 25 °C except for Cys deacylation, which was at 37 °C. Substrate aa-tRNA and enzyme concentrations were 1 μm and 5 μm, respectively. Data points and error bars represent average and S.D., respectively, of three independent experiments. The extent of hydrolysis of each substrate by buffer (0.15 m KPO₄, pH 7.0) was subtracted prior to plotting the data.

Cys-tRNA^Pro deacylation by the INS mutants was also tested (Fig. 5D). Interestingly, efficient deacylation of Cys-tRNA^Pro by the I263A variant (k_obs = 0.068 ± 0.005 min⁻¹) was observed, and weak deacylation by the L266A variant was also detected (k_obs = 0.01 ± 0.001 min⁻¹) at 37 °C, the temperature typically used for Cys-tRNA deacylation by YbaK (16). To compare the observed rates for Cys-deacylation with the rates obtained for other substrates at room temperature, cysteine deacylation assays using the I263A and L266A variants were also carried at room temperature (supplemental Fig. 4). The k_obs of 0.07 ± 0.007 min⁻¹ was similar to that obtained at 37 °C for the I263A mutant, whereas L266A showed negligible deacylation at room temperature (k_obs of 0.002 ± 0.0005 min⁻¹). These same variants failed to deacylate Ala-tRNA^Pro (Fig. 5A). The complete switch of substrate specificity from Ala- to Cys-tRNA^Pro suggests that the size of the active site pocket was successfully engineered to accommodate the larger cysteine side chain in a manner that mutually excludes Ala-tRNA^Pro deacylation (Table 1).

TABLE 1.

Substrate specificities of WT and mutant E. coli ProRS

Results are relative to the rate of Ala-tRNA^Pro deacylation by WT ProRS, which is designated as +++. Symbols indicate rate within 2-fold of WT ProRS (+++), 5–10-fold lower than WT ProRS (++), or 11–20-fold lower than WT ProRS (+). A minus indicates no detectable deacylation (rate estimated to be at least 70-fold reduced relative to Ala-tRNA^Pro hydrolysis by WT E. coli ProRS).

Aa-tRNAPro substrate	INS domain mutations
	WT	I263A	L266A	T277S	1263A/L266A
Pro	−	−	−	−	−
Ala	+++	−	−	−	−
Cys	−	+++	−	−	−
2-Abu	++	−	+	−	−
Ser	++	+	+++	++	+++

Similar to WT ProRS, none of the mutants showed significant deacylation activity against Pro-tRNA^Pro (Fig. 5E and Table 1). Binding of CCA-Pro to WT INS (Fig. 3C) suggests that the bulky proline ring results in disruption of the catalytically competent orientation of the substrate carbonyl. The observed lack of proline deacylation by the INS variants suggests that merely increasing the binding pocket size is not sufficient to allow binding of a catalytically competent Pro-tRNA^Pro substrate.

DISCUSSION

Three-dimensional crystal structures of post-transfer editing substrate analogs bound to the ProRS editing domain or to freestanding homologous domains are not yet available. Our recent model of YbaK bound to a 5′-CCA-Cys substrate analog (16) and the model of the INS domain bound to 5′-CCA-Ala presented in this study together with new biochemical data provide the first structure-based understanding of the distinct substrate specificities of these editing domains. The only available structures with bound substrate are those of Thermus thermophilus ProX, a homologous domain of unknown function, bound to pre-transfer editing substrate analogs 5′-O-[N-(l-alanyl)-sulfamoyl]adenosine and 5′-O-[N-(l-cysteinyl)-sulfamoyl]adenosine (PDB codes 2Z0K and 2Z0X). Although these substrates differ significantly from the post-transfer substrate analogs modeled here, these structures are consistent with aspects of our models. A comparison of our INS·CCA-Ala model with the T. thermophilus ProX·Ala-SA structure (supplemental Fig. 5) shows that the substrate alanine occupies a similar active site pocket in both the domains. In addition, the substrate carbonyl points toward the GXXXP loop in both structures. However, the ribose moiety is flipped by ∼180°, presumably due to differences in connectivity of the adenine base to alanine (via the ribose in CCA-Ala and via the phosphate in 5′-O-[N-(l-alanyl)-sulfamoyl]adenosine), causing the adenine base and the phosphate group of the substrate to occupy distinct positions and orientations in the two structures.

Based on the INS·CCA-Ala model, we predicted correctly that the INS domain could hydrolyze somewhat larger substrates such as 2-Abu in addition to alanine. Ser-tRNA^Pro was also hydrolyzed by WT ProRS, and by increasing the size of the active site pocket, the efficiency of this reaction could be increased. Mutant forms of E. coli ProRS that result in a complete switch in substrate specificity from alanine to cysteine (I263A) or to serine (I263A/L266A) (Table 1) were also generated.

Several different mechanisms have been proposed for discrimination of substrates by post-transfer editing domains. For example, isoleucyl-tRNA synthetase uses the so-called “double-sieve” mechanism to activate isoleucine and smaller valine in the synthetic active site (“coarse” sieve), but uses size exclusion to reject cognate Ile-tRNA in the editing active site (“fine” sieve) (32, 33). Discrimination between isosteric substrates valine and threonine by valyl-tRNA synthetase is dependent on the chemically distinct properties of their side chains (34). Recently, it was shown that a coordinated zinc ion in the editing domain of AlaRS may help provide specificity for noncognate serine, which is larger than alanine (35, 36). Structural studies of the ThrRS editing domain suggest that the discrimination against threonine in the editing domain is achieved by positioning of the substrate rather than steric exclusion (37). We recently showed that YbaK, which only cleaves Cys-tRNAs, achieves substrate discrimination by taking advantage of the unique side chain chemistry of the substrate cysteine sulfhydryl moiety (16). The results presented in this study suggest that cysteine discrimination by the INS domain is achieved via size exclusion, which can be counteracted by mutations in the substrate binding pocket. Discrimination against proline, on the other hand, is largely dependent on functional orientation of the substrate, in addition to size exclusion. For example, H369A E. coli ProRS does hydrolyze Pro-tRNA^Pro, albeit with significantly reduced efficiency relative to Ala-tRNA^Pro hydrolysis by WT ProRS (22).

Taken together, the results of this study shed new insights into substrate selection and specificity of the bacterial ProRS INS domain. Although the INS domain can be engineered to deacylate Cys-tRNA^Pro, the mechanism of catalysis is unlikely to be the same as that of YbaK. The latter is believed to exclude water from the active site and uses a substrate-assisted mechanism of catalysis (16). Consistent with this mechanism, selenocysteine-tRNA is a substrate for YbaK, whereas Ser-tRNA is a competitive inhibitor of Cys-tRNA deacylation (16). In addition, manipulating the size of the active site pocket of YbaK fails to alter the substrate specificity.⁴ In contrast, cleavage in the INS active site is likely initiated by an active site water molecule, and substrate specificity can readily be modulated (Table 1). The fact that the engineered I263A INS domain deacylates Cys- but not Ala-tRNA^Pro is consistent with the need for distinct editing domains to perform these proofreading functions in vivo.