The Structure of Yeast Glutaminyl-tRNA Synthetase and Modeling of Its Interaction with tRNA

Abstract

Eukaryotic glutaminyl-tRNA synthetase (GlnRS) contains an appended N-terminal domain (NTD) whose precise function is unknown. Although GlnRS structures from two prokaryotic species are known, no eukaryotic GlnRS structure has been reported. Here we present the first crystallographic structure of yeast GlnRS, finding that the structure of the C-terminal domain is highly similar to Escherichia coli GlnRS but that 214 residues, including the NTD, are crystallographically disordered. We present a model of the full-length enzyme in solution, using the structures of the C-terminal domain, and the isolated NTD, with small-angle X-ray scattering data of the full-length molecule. We proceed to model the enzyme bound to tRNA, using the crystallographic structures of GatCAB and GlnRS–tRNA complex from bacteria. We contrast the tRNA-bound model with the tRNA-free solution state and perform molecular dynamics on the full-length GlnRS–tRNA complex, which suggests that tRNA binding involves the motion of a conserved hinge in the NTD.

Introduction

Aminoacyl-tRNA synthetases are required in all three domains of life for covalently attaching amino acids to their cognate tRNA molecules for use in protein synthesis.¹ These enzymes are separated into two structural classes,^2,3 class I and class II, which operate via a two-step reaction. First, after binding ATP and the amino acid, an aminoacyl-adenylate is formed, followed by esterification of the amino acid moiety to either the 2′ OH (class I) or 3′ OH (class II) of tRNA. In addition, eukaryotic tRNA synthetases often contain additional domains appended to their N-terminal or C-terminal ends that are not found in their prokaryotic homologs.⁴ These domains are known to have various roles, including nucleic acid binding, protein–protein interactions, and hydrolytic editing mechanisms¹; however, many of these domains are of unknown function. While structures of several eukaryotic tRNA synthetases have been determined,⁵ virtually all of these lack the appended domain. The structures of some appended domains have been determined in isolation, but few models of full-length eukaryotic tRNA synthetases exist.

While in most cases one synthetase exists for each amino acid, an exception occurs for glutamine and asparagine.⁶ In eukaryotes and some bacteria, the traditional pathway of aminoacylation exists for glutamine, in which glutaminyl-tRNA synthetase (GlnRS) binds to tRNA^gln, glutamine, and ATP and first forms a glutaminyl-adenylate molecule that is then covalently attached to the 3′-end of tRNA^gln with the release of AMP. A different pathway exists in most bacteria and all archaea, where a non-discriminating glutamyl-tRNA synthetase (GluRS) attaches glutamic acid to both tRNA^glu and tRNA^gln. The misacylated Glu-tRNA^gln is then converted to Gln-tRNA^gln by the GatCAB amidotransferase enzyme in bacteria and some archaea or by the GatDE amidotransferases in other archaea. GatCAB performs a similar function for asparaginyl-tRNA^asn. Since this indirect pathway for aminoacylation exists in most prokaryotes, GlnRS is primarily a eukaryotic enzyme, and its presence in a small number of bacteria is believed to have occurred through a horizontal gene transfer event.⁷ Nonetheless, there are significant differences between the prokaryotic and eukaryotic GlnRS enzymes, as described below.

Structural data for two prokaryotic GlnRS species exist,^8,9 yet no structure has been reported for any full-length eukaryotic GlnRS. Saccharomyces cerevisiae GlnRS (Gln4) is an 809-residue protein that contains a 215-residue domain appended to its N-terminus. This domain is nearly ubiquitous among eukaryotic GlnRS species but absent in prokaryotic homologs.¹⁰ Recently we described the structure of the N-terminal domain (NTD) of Gln4, revealing that it has an extraordinary structural resemblance to the region of the B subunit of the GatCAB amidotransferase (GatB)¹¹ that binds to tRNA^gln.¹² Although deletion of the NTD distinctly affects catalytic activity, growth in yeast, and tRNA binding,¹² the manner in which tRNA binding occurs is still unknown.

Here we present the first crystallographic structure of the C-terminal domain (CTD) of Gln4 from crystals of the full-length protein. Based on this structure, the structure of the isolated NTD,¹² and small-angle X-ray scattering (SAXS) data of the full-length enzyme, we present a model of the full-length enzyme in solution. We extend this to model the full-length enzyme bound to tRNA^gln from the crystallographic structures and homology with known transamidosome and GlnRS–tRNA complex structures^11,13 yielding new insights into the structural rearrangements occurring in eukaryotic GlnRS–tRNA^gln complex formation.

Results and Discussion

Overall structure

We solved the crystal structure of Gln4 to 2.15 Å resolution [Protein Data Bank (PDB) ID: 4H3S] by single-wavelength anomalous dispersion with a zinc anomalous signal identified from an initial excitation scan, coupled with molecular replacement using the crystal structure of the Escherichia coli GlnRS as a template (Fig. 1 and Table 1). Of the 809 amino acids in the primary sequence of Gln4, ~30% were not resolved in the electron density. The unresolved regions are residues 1–187 (which correspond to the NTD), residues 188–214 (which encode a nonconserved 26-residue region that is predicted to be unstructured and links the NTD and CTD; Fig. S1), and residues 672–678 (which encode a disordered loop within the CTD). The presence of the NTD in purified protein and in the crystals was confirmed by expressing Gln4 with a His₆ tag on its N-terminus, followed by purification using Ni²⁺ affinity chromatography, crystallization, and antibody detection with anti-His antibody on dissolved crystals. Residues 1–214 were still unresolved in structural data from these crystals.

Fig. 1.

X-ray crystal structure of Gln4. Cartoon representation of the Gln4 structure is shown color-coded according to domains and insertions relative to E. coli GlnRS. Residues 1–214 are missing in the electron density. Domains are labeled with corresponding amino acid numbers.

Table 1.

Crystallographic data collection parameters and statistics

Beamline	SSRL BL 11−1
Wavelength (Å)	1.169
Space group	P3121
Unit cell parameters
a, b, c (Å)	176.61, 176.61, 72.19
α, β, γ (°)	90, 90, 120
Resolution limits (outer shell) (Å)	52.49–2.15 (2.227–2.15)
No. of unique reflections	70,276 (6963)
Completeness (%)	99.86 (99.84)
Rsim (%)	0.068 (0.348)
〈I/σ(/)〉	23.26 (2.98)
Reflections
Working	70,276 (6963)
Test (%)	5
Atoms	10,566
Protein	5031
Water	447
Ligands	75
R factor	0.160 (0.222)
R free	0.175 (0.242)
RMSD
Bonds (Å)	0.003
Angles (°)	0.73
Average B-factor
Overall	48.10
Macromolecules	47.40
Solvent (water)	54.20
Ramachandran analysis (%)
Most favored regions	99
Outliers	0.17
Generously allowed regions	0
Disallowed regions	0
Clash score	5.27
PDB code	4H3S

Since the NTD of Gln4 is known to be structured¹² and is present in the full-length protein in the crystal, but no electron density is observed for this region, we conclude that the NTD is crystallographically disordered (i.e., the position and/or orientation of the NTD varies between unit cells). It seems likely that the NTD is located in large solvent channels approximately 145–175 Å wide observed along the z-axis (Fig. S2). Consistent with this, we note that the channels can comfortably accommodate the structure of the NTD,¹² which has dimensions of ~25 Å × 25 Å × 75 Å and a macromolecular volume of ~25,000 ų. Moreover, the first residue resolved in the electron density, Arg215, is located adjacent to the solvent channels, and the preceding residues, 188–214, are predicted to be unstructured (Fig. S1). It is therefore likely that this region is the cause of the crystallographic disorder of the NTD.

Based on the considerations presented above, using the naming convention previously described for E. coli GlnRS,¹⁴ the overall structure of Gln4 consists of a folded NTD (residues 1–187); a positively charged 26-residue region linking the NTD and CTD (residues 188–214); a catalytic domain consisting of a Rossmann fold domain (residues 215–324 and 439–498), an acceptor stem binding domain (residues 325–438), and a helical subdomain (residues 499–567); and an anticodon binding domain consisting of proximal (residues 568–574 and 690–809) and distal (residues 575–689) β-barrels (Fig. 1). The structure of the CTD of Gln4 (residues 215–809) is highly similar to E. coli GlnRS (PDB ID: 1NYL; RMSD = 1.86 Ź⁵) but has multiple insertions relative to the bacterial homolog. The Gln4 catalytic domain (residues 215–567) is highly conserved in sequence and structure, showing an E value of 1 × 10^–101 relative to the E. coli GlnRS catalytic domain using PSI-BLAST and an RMSD of only 1.03 Å. The Gln4 anticodon binding domain (residues 568–809) shares moderate sequence homology but high structural conservation showing an E value of 5 × 10^–32 and an RMSD of 1.13 Å. The structures of the Gln4 CTD and a second bacterial homolog, Deinococcus radiodurans GlnRS (PDB ID: 2HZ7),⁸ are also similar with an RMSD of 1.62 Å.

A detailed comparison of the Gln4 CTD structure with the unliganded (PDB ID: 1NYL)¹⁵ and tRNA-bound (PDB ID: 1QTQ)¹³E. coli GlnRS structures revealed both similarities and differences in the individual residues previously found to be important substrate recognition elements in E. coli GlnRS.^16,17 The ATP-binding pocket formed by the HIGH and MSK (LSK in Gln4) is similar to that found in E. coli GlnRS, which are signature sequences conserved across class I aminoacyl-tRNA synthetases.¹ The replacement of Met268 in E. coli GlnRS with Leu496 in Gln4 does not appear to significantly change the structure of this pocket. In E. coli GlnRS, residues Arg133, Arg192, Lys194, Thr68, and Asn69 were shown to be important in forming the hairpin loop of the tRNA^gln CCA tail that is necessary to bring the 2′ hydroxyl of A76 into proximity with the glutamine adenylate molecule for glutamine transfer to tRNA.¹⁶ In Gln4, these residues (Arg357, Arg421, Leu422, Thr293, and Asn294, respectively) are conserved in position and orientation and therefore are likely to cause the CCA tail to form a similar hairpin loop. In the anticodon binding region of E. coli GlnRS, Arg341 interacts with anticodon base U35, resulting in distant, interdomain communication to the active site necessary for glutamine recognition.¹⁵ The corresponding residue in Gln4, Arg568, is in the same orientation and therefore could lead to similar interactions with U35. In E. coli GlnRS, Lys401 and Arg402 interact with G36, which is the discriminating base of the anticodon, and are crucial for recognition of cognate tRNA^gln and discrimination against noncognate tRNA^glu. In Gln4, while Arg402 is conserved as Arg631, Lys401 is replaced by Phe630. While both lysine and phenylalanine are large residues, it is unknown if the loss of the positively charged side chain results in different interactions with G36 in Gln4.

Multiple sequence analysis among eukaryotic GlnRS species revealed three insertions relative to E. coli GlnRS (Figs. 1 and 2). Insertion 1, located at residues 234–250 of Gln4, is present in all eukaryotes, including those species lacking an NTD, although its length and sequence are not conserved among different kingdoms. Comparison with the E. coli GlnRS structure reveals that this insertion is part of the Rossmann fold domain in the CTD. Insertion 2 is located at residues 364–368 and is situated in a zinc binding motif consisting of Cys346-X-Cys348-X₂₄-Cys372-X-His374 (Fig. S3). The zinc binding motif and Insertion 2, which are always found together, are present in fungal GlnRS species but are absent in other eukaryotes. Insertion 3 is a loop located at residues 751–770 and is also only conserved among fungi and absent in other eukaryotes.

Fig. 2.

Domain architecture of GlnRS. Domains for GlnRS are shown for fungi, eukaryotes (other than fungi), bacteria, the bacterium D. radiodurans, and archaeal GluRS. Major domains are labeled NTD, catalytic domain, anticodon binding domain, and the unique CTD appended to the D. radiodurans GlnRS. Insertion 1, Insertion 2, Insertion 3, and Loop 1 are labeled along with the zinc coordinating residues of the zinc binding motif, which is present only in fungal GlnRS and archaeal GluRS.

GlnRS selects both the correct amino acid substrate, glutamine, and the cognate tRNA^gln molecule. To gain insight into the tRNA^gln discrimination mechanism in Gln4, we performed further sequence analysis incorporating the archaeal non-discriminating GluRS, an evolutionary precursor of GlnRS that has the ability to recognize both tRNA^gln and tRNA^glu.¹⁸ This revealed that Insertion 2 is part of a larger nine-residue loop, Loop 1, which is absent in archaeal GluRS (Fig. 2). In E. coli GlnRS, Loop 1 is only four residues long and has been suggested to provide the ability to discriminate against the G1:C72 base pair in tRNA^glu and instead only recognizes tRNA^gln by disrupting the weak U1:A72 base pair.¹⁸ Our results suggest that Loop 1 may play a different role in fungi and other eukaryotes than it does in bacteria. Although Loop 1 is present in Gln4, it differs in amino acid sequence compared to E. coli GlnRS and, with Insertion 2, is five residues longer. Additionally, while the first base pair of tRNA^gln is U1:A72 in E. coli, the first base pair of tRNA^gln is G1:C72 in fungi and other eukaryotes. It is therefore unlikely that the extended Loop 1 seen in Gln4 functions similarly to E. coli GlnRS Loop 1 by disrupting the first base pair, since if the extended Loop 1 in Gln4 were to disrupt the strong G1:C72 base pair in tRNA^gln it might also disrupt the weak U1:A72 base pair in tRNA^glu and fail to discriminate between them. While our observations are based on length, it is unknown whether sequence-specific contacts in Loop 1 contribute tRNA discrimination ability to Gln4. However, in all eukaryotes other than fungi, Loop 1 is entirely absent (Fig. 2) and this region of eukaryotic GlnRS more closely resembles archaeal GluRS, suggesting that Loop 1 is not likely to be a critical tRNA discriminatory motif in eukaryotes with the possible exception of fungi.

Motion of NTD in solution is limited in the absence of tRNA

We collected SAXS data on full-length Gln4 (Fig. S4) because SAXS is a solution-based technique that is sensitive to the overall shape and size of a molecule and can also provide insight into protein dynamics.^19,20 Analysis of the pair distribution function (the distribution of interatomic distances) yields a maximum particle dimension of ~147 Å and a radius of gyration (R_g) of 43.51 Å, in close agreement with that estimated by Guinier analysis,²¹ 43.32 Å. This is significantly larger than the maximum dimension and R_g of the CTD from the crystal structure (108 Å and 32.3 Å, respectively). The increased values are consistent with the presence of an appended NTD. The SAXS data of full-length Gln4 suggest an elongated particle shape due to the large maximum dimension compared to the R_g and the left-skewed pair distribution function (Fig. S4d).

Since the crystallographic data showed that the NTD was disordered, we used the SAXS data to probe the motion of the NTD in solution. Information about protein flexibility can be obtained from a Porod–Debye analysis of SAXS data, where the decay in intensity is studied as a function of angle and reported as the Porod exponent.²² In cases of globular, rigid molecules, the Porod exponent will approach 4, whereas for flexible, random chains, it will approach 2. A Porod–Debye analysis of full-length Gln4 resulted in a Porod exponent of 3.4, suggesting a well-folded protein with mild flexibility, consistent with the crystallographic data. The total volume occupied by the protein can be measured from the SAXS data and is known as the Porod volume.²³ The Porod volume of full-length Gln4 is ~153,000 ų, corresponding to an estimated molecular mass of 92.6 kDa,²² similar to the molecular mass estimated from the forward scattering [I(0)] of 89.1 kDa and the expected molecular mass of 93.1 kDa (these data also confirm that the full-length protein is intact). This demonstrates that the volume occupied by the NTD is approximately limited to the size of the domain and does not occupy a larger region of conformational space.

Ten ab initio envelope reconstructions of full-length Gln4 were created and averaged, exhibiting an average normalized spatial discrepancy (NSD) of 0.633 ± 0.022. NSD is a measure of the variance among a set of multiple SAXS envelopes. NSD values less than 1 reflect uniqueness and low flexibility,²⁴ whereas values greater than 1 reflect an inability to describe the protein using one unique shape. To generate a rigid-body model, we created 10,000 random conformations of full-length Gln4 by treating the NTD and CTD crystal structures as independent rigid bodies and allowing the linker (residues 188–214) to be flexible. The model whose calculated scattering best fit the experimental SAXS data was selected from this pool. This model is shown in Fig. 3 superposed onto the ab initio envelope. The rigid-body model fits well to the experimental scattering data with Χ² = 1.82. The location of the NTD of Gln4 appears to be similar to the appended domain modeled in the hybrid structure of GlnRS from D. radiodurans.⁸D. radiodurans is unique because it is the only bacteria known to contain a GlnRS with an appended domain. However, the appended domain of D. radiodurans GlnRS is located on the C-terminus, downstream of the core enzyme (Fig. 2), and contains no sequence homology to the appended domain of eukaryotic GlnRS, which is located on the N-terminus, upstream of the core enzyme. The similar spatial location of both appended domains, despite being extensions at different termini of the core enzyme, is made possible by the much longer 80-residue linker in D. radiodurans GlnRS, compared to the 26-residue linker in Gln4. Additionally, the appended domain of D. radiodurans GlnRS was shown to cross-react with an antibody that recognizes the GatB subunit of the GatCAB amidotransferase used in the indirect pathway of Gln-tRNA^gln formation used in most prokaryotes.⁸ Our previous study reported a high degree of structural similarity between the Gln4 NTD and the GatB subdomain,¹² suggesting that the appended domains of GlnRS from eukaryotes and D. radiodurans perform similar functions and are likely an instance of convergent evolution.

Fig. 3.

SAXS rigid-body model of full-length Gln4. (a) Two orientations rotated 90° relative to each other are shown in cartoon representation and colored according to secondary structure and superposed onto the ab initio envelope shown in gray. (b) The calculated scattering of the rigid-body model (black continuous line) fit to the experimental SAXS data (gray circles).

Ab initio reconstructions and rigid-body modeling do not take into account the dynamic information present in SAXS data but represent an average of all conformations present in solution. Ensemble modeling can overcome this by representing a protein structure as an ensemble of multiple conformations. Using the ensemble optimization method²⁴ with multiple conformations of the NTD did improve the overall fit to the scattering data (Χ² = 1.75); however, the improvement was only marginal. The R_g distribution of conformers present in the ensemble was similar to that seen for simulated data of the rigid-body model alone, suggesting little flexibility. The limited flexibility from the Porod–Debye analysis, the low NSD for multiple ab initio envelope reconstructions, and the marginal improvement from an ensemble model compared to the rigid-body model demonstrate that the mobility of the NTD in solution is limited.

A model of full-length Gln4 bound to tRNA^gln suggests a substantial conformational change of the NTD

We used the structure of E. coli GlnRS bound to tRNA¹³ and the structure of Thermotoga maritima GatCAB bound to tRNA¹¹ to obtain a model of full-length Gln4 bound to tRNA. First, the structure of E. coli GlnRS bound to tRNA^gln (PDB ID: 1QTQ) was superposed onto the Gln4 CTD (residues 215–809). The structure of the tRNA molecule was then extracted, providing an initial model of tRNA^gln complexed with the Gln4 CTD. Since the nucleotide sequence of E. coli tRNA^gln differs from yeast tRNA^gln, the tRNA molecule was computationally mutated to match that of yeast tRNA^gln and a geometric minimization was performed using ModeRNA.²⁵ Our model contains similar contacts between the tRNA^gln anticodon and the Gln4 anticodon binding domain that were shown to be important identity elements in E. coli GlnRS^9,13 due to the high degree of structural similarity between E. coli GlnRS and the Gln4 CTD.

To correctly place the NTD in the model, we utilized known structural homology of the NTD with subunit B of the GatCAB amidotransferase (GatB) from T. maritima (PDB ID: 3AL0), which was solved in complex with tRNA^gln. The tail subdomain (residues 119–187) of the Gln4 NTD was superposed to the tail subdomain of GatB due to the high level of structural homology between these two regions.¹² Finally, MODELLER²⁶ was used for de novo modeling of the predicted flexible linker (residues 188–214) (Fig. 4).

Fig. 4.

Homology model of full-length Gln4 bound to tRNA^gln. (a) Full-length Gln4 shown bound to tRNA^gln. (b) Enlarged and rotated model showing gap between NTD helical subdomain and tRNA molecule.

Our model of full-length Gln4 bound to tRNA^gln shows a significant change in the NTD position when compared to the tRNA-free, SAXS-derived conformation (Fig. 5a). The model shows an ~160° rotation and an ~40-Å translation of the NTD with respect to the solution conformation according to an analysis by DynDom.²⁷ To determine if this new conformation of the NTD exists in solution, we simulated the SAXS profile from our model of Gln4 bound to tRNA, after removing the tRNA molecule, and compared it to the experimental SAXS data. Fitting the simulated scattering of only the protein portion of the model of the protein–tRNA complex to the experimental SAXS data resulted in a poor fit, yielding Χ² = 12.25 compared to 1.82 for the rigid-body model created using the experimental SAXS data (Fig. 5b). The limited flexibility of the NTD, coupled with the poor fit of the simulated scattering of the protein portion of the model bound to tRNA^gln, suggests that, without tRNA bound, this conformation does not exist in solution.

Fig. 5.

Comparison of NTD position before and after tRNA binding. (a) Gln4 is shown in cartoon representation. The NTD position prior to tRNA binding determined by SAXS rigid-body modeling is shown in light gray. The NTD position upon binding tRNA predicted by homology modeling is shown in dark gray. The CTD is also shown in dark gray. (b) The simulated SAXS profile of the rigid-body model (broken line) and the homology model (continuous line; calculated without tRNA in the model) are shown fitted to the experimental SAXS data (circles). The better visual fit and lower Χ² value of the tRNA-free model shows that the SAXS data are able to clearly distinguish between which model exists in solution.

Conformational change in NTD subdomains is predicted upon interaction with tRNA^gln

In the compiled model of Gln4 bound to tRNA, the NTD subdomains appear to be in an “open” conformation resulting in the helical subdomain (residues 1–110) being too distant from the tRNA molecule to form stable contacts (Fig. 4b). This conformation more closely resembles that observed in the Staphylococcus aureus GatCAB structure solved without tRNA bound¹² than the conformation seen in tRNA-bound T. maritima GatCAB. Thus, we considered that the molecule might undergo a conformational change upon tRNA binding. To probe the dynamics of full-length Gln4 in complex with tRNA^gln, we carried out 70 ns of molecular dynamics simulation. The model stabilized within approximately 10 ns of simulation time and remained relatively unchanged for the remaining 60 ns (Fig. S5), showing that the simulation time captured the relevant dynamics.

The resulting molecular dynamics trajectory (Figs. 6 and 7 and Fig. S5) shows that the helical subdomain of the NTD rotates about a conserved hinge by 22.9°, forming electrostatic interactions between the positively charged side chains of lysine residues 19, 20, and 26, which have been previously implicated in tRNA binding integrity,²⁸ and the negatively charged phosphate backbone of tRNA. These electrostatic interactions along the inner face of the NTD were speculated¹² to be responsible for the nonspecific RNA binding activity.²⁹ This new conformation of the NTD is in agreement with the tRNA-bound conformation seen in the T. maritima GatB structure. The hinge motion in our model is strongly supported by our previous study demonstrating that mutations in conserved residues in the hinge region, connecting the helical and tail domains of the NTD, significantly reduce tRNA binding.¹² In addition, our modeling predicts that residues K29, K63, G64, T65, and D66 of the helical subdomain of the NTD make several contacts with the CTD, which includes residues P238 and M241 of Insertion 1. Since Insertion 1 was shown above to be exclusive to eukaryotes, we speculate that this insertion may provide a means of communication between the NTD and CTD.

Fig. 6.

NTD undergoes conformational change after binding to tRNA. (a) Plot of backbone RMS fluctuation as a function of residue. The RMS has been calculated as the deviation from the starting structure. (b) Structure of Gln4 before (blue) and after (red) molecular dynamics simulation. The tRNA molecule has been removed for clarity.

Fig. 7.

Comparison of NTD structure before and after molecular dynamics simulation. The position of the NTD before the simulation is shown in light gray and that after the simulation is shown in dark gray. The black continuous arrow shows the degree and direction of angular motion calculated by DynDom.

The molecular dynamics simulation shows a significant distortion of the tRNA molecule as a result of the NTD binding to tRNA. At the start of the simulation, base U73 is pointed away from the inner core of the enzyme, making no specific contacts with the protein. Approximately 3 ns into the simulation, upon distortion of the tRNA molecule, this base flips inwards toward the enzyme, resulting in contacts with Leu463. This conformational change in the tRNA, caused by binding of the NTD, results in these previously unknown interactions between Leu463 and U73. Our previous study reported that the presence of the NTD across eukaryotes is strongly correlated with the existence of base U73 of the acceptor stem, immediately preceding the CCA tail.¹² Furthermore, it was shown in E. coli GlnRS that any alterations to base 73 led to substantial increases in K_M for glutamine.¹⁷ Our molecular dynamics simulation showing these concerted conformational changes, combined with the phylogenetic correlation of the NTD and base U73 and the increase in K_M for glutamine when mutating base 73 in E. coli GlnRS, may explain the result that deletion of the NTD in Gln4 leads to increased K_M for glutamine.¹²

Previously it was shown that GatB provides tRNA discrimination ability to GatCAB through the tail domain.³⁰ This discrimination occurs through an interaction with U20_B of tRNA^glu that is absent in tRNA^gln in prokaryotes. Analysis of tRNA^gln and tRNA^glu sequences in yeast reveals that base U20_B is present in both tRNA sequences, suggesting that this base does not participate in tRNA discrimination interactions with Gln4. Our model of Gln4 bound to tRNA^gln reveals no evidence of significant interactions between tRNA^gln and the NTD that would otherwise be negated by replacement with tRNA^glu. This is consistent with previous studies demonstrating nonspecific RNA binding activity of the NTD.^28,29 Therefore, in contrast to the tRNA discrimination properties of GatB, our model provides no evidence that the Gln4 NTD makes any direct contacts with tRNA^gln that would confer any additional capacity for discrimination against tRNA^glu, suggesting that discrimination occurs via specific contacts with the CTD.

Conclusion

The structure of the CTD of Gln4 presented here was shown to be highly similar to the E. coli GlnRS structure. However, multiple insertions relative to E. coli GlnRS revealed insights into one structural motif common to all eukaryotes and two motifs specific to fungi. Since most eukaryotic GlnRS species lack the bacterial unpairing loop that is proposed to play a role in tRNA discrimination, then another mechanism must be employed to discriminate between tRNA^glu and tRNA^gln in eukaryotic GlnRS species. Therefore, in eukaryotes, there may be a compensating mechanism to discriminate between G1:C72 of tRNA^gln from U1:A72 of tRNA^glu or, as in the case of non-discriminating archaeal GluRS,¹⁸ the first base pair does not play a significant role in tRNA discrimination.

Our structure-based models of the first full-length eukaryotic GlnRS with and without tRNA^gln bound suggest that CTD binding to tRNA results in a large conformational reorientation of the NTD allowing for interactions between the NTD and the tRNA. Given the distinct increases in K_M and K_D for tRNA^gln following deletion of the NTD,^10,12 the solution model of full-length Gln4 presented here suggests that the NTD plays a direct role in tRNA binding. Our molecular dynamics simulation revealed that the helical and tail subdomains of the NTD undergo a hinge motion after binding to tRNA, allowing for tighter binding between the NTD and tRNA. Our structural results and modeling suggest the intriguing possibility that the NTD communicates with the CTD through Insertion 1, which is found in all eukaryotes. The absence of such an interaction may explain the loss of the NTD in bacterial GlnRS evolution. In addition, since the NTD and the active site of Gln4 are too distant to interact directly and since deletion of the NTD also increases K_M for glutamine and ATP, it seems plausible that the effects on glutamine and ATP are due to the concerted conformational changes in Gln4 that occur upon tRNA binding as was observed in E. coli GlnRS.^10,12,13,15 Further experimental studies will be required to confirm these proposed interactions in eukaryotic GlnRS.

Materials and Methods

Protein expression and purification

To obtain highly purified Gln4 protein and its derivatives, we cloned open reading frames into the previously described LIC vectors BG2483 or BG2663 under P_GAL1 control³¹ as previously described¹² and expressed them in yeast strain BCY123.³² To purify full-length Gln4 from a Gln4-pt-tagged construct (QB721B), we performed large-scale growth, affinity purification on IgG Sepharose, removal of GST-3C protease, concentration of samples, and sizing on Superdex HiLoad 16/60 (GE Healthcare 17–1069, 10 mm × 300 mm bed dimension) as previously described.³³ This purified Gln4 construct was used for crystallographic and structural analyses.

To ensure that full-length polypeptide was purified for crystallography, we added an N-terminal Met-Ala-His₆ tag at the N-terminus of GLN4, during PCR amplification of the GLN4 gene using QB832ADFP and QB832ADRP primers (Supplementary Table 1). Full-length GLN4 was amplified with QB1012ADFP and QB832ADRP.

For crystallography of the full-length His₆-Gln4 protein, the His₆-Gln4-pt was purified from strain QB832AD (21.6 L at 6.5 OD/L) on 24 mL IgG Sepharose and eluted with 64 mL 3C cleavage buffer.³³ After elution from IgG Sepharose and removal of GST-3C protease but prior to sizing, the sample containing the full-length Gln4 protein with the N-terminal His₆ tag (QB832AD) was diluted with an equal volume of buffer T [25 mM Hepes (pH 7.5), 850 mM NaCl, 10% (v/v) glycerol, 2 mM β-mercaptoethanol (BME), and 2 mM phenylmethylsulfonyl fluoride], mixed with 3 mL prewashed Talon resin [washed in Wash buffer, 20 mM Hepes (pH 7.5), 0.5 M NaCl, 5% (v/v) glycerol, and 2 mM BME], incubated for 1 h at 4 °C, followed by centrifugation at 2000 rpm for 2 min and removal of supernatant, after which the Talon resin was washed once for 10 min with 40 mL Wash buffer containing 0.5 M NaCl [20 mM Hepes (pH 7.5), 0.5 M NaCl, 5% (v/v) glycerol, and 2 mM BME], centrifuged at 2000 rpm for 2 min, followed by two more washes of the resin with Wash buffer containing 1 M NaCl, then a wash with Wash buffer containing 0.5 M NaCl, followed by a wash with Wash buffer containing 0.5 M NaCl and 10 mM imidazole. The protein was eluted from the Talon resin with four sequential washes of the resin with Wash buffer that contains 0.5 M NaCl containing 250 mM imidazole, each of which was mixed for 10 min prior to the low-speed spin. Three fractions (elutions 1 and 2 as well as the 10 mM imidazole wash) were combined and diluted with 75 mL of no-salt buffer to bring the NaCl to 0.2 M, followed by concentration to 5 mL using an Amicon (Millipore UFC901024), and loaded onto a Superdex 200 sizing column as previously described.³³

Crystallization

Initial crystallization conditions were identified using a high-throughput microbatch-under-oil method with a 1536-condition, incomplete-factorial-based screen.³⁴ Conditions that produced crystals appearing suitable for optimization were prioritized according to their ease of cryoprotection³⁵ and optimized using a drop volume ratio versus temperature technique.³⁶ Crystals of full-length Gln4 were prepared for diffraction by combining 3.5 μL of protein solution [13.9 mg/mL protein in 200 mM NaCl, 5% (v/v) glycerol, 2 mM DTT, 0.025% (w/v) NaN₃, and 20 mM Hepes buffer (pH 7.5)] with 2.0 μL of precipitant solution [50 mM NH₄Br, 50 mM KC₂H₃O₂, 100 mM Hepes (pH 7.5), and 20%(w/v) polyethylene glycol 20,000], incubated at 14 °C. The optimized 16:9 protein-to-precipitant ratio and 14 °C temperature were determined from a drop volume ratio versus temperature technique. Crystals appeared after 4 weeks.³⁶

Single crystal data collection and structure solution

Crystals of the full-length protein were harvested and cryoprotected, then shipped to Stanford Synchrotron Radiation Lightsource (SSRL). Single crystal X-ray data were collected remotely on beamline 11-1.³⁷ An initial excitation scan revealed the presence of zinc. Data were collected to 2.15 Å, integrated with XDS,^38–40 and reduced with Scala,⁴¹ part of the CCP4 package.⁴² Initial molecular replacement with PHENIX⁴³ using an ~40% sequence homology E. coli Gln-tRNA synthetase (PDB ID: 1GTS)¹⁴ failed, but a combined molecular replacement/single-wavelength anomalous dispersion approach with the zinc signal was successful. An iterative process of PHENIX refinement and manual model building through Coot⁴⁴ was employed with validation using MolProbity.⁴⁵ The coordinates were deposited as PDB ID: 4H3S. Data collection, processing, and refinement statistics are given in Table 1. The N-terminal region, residues 1–214, was unresolved in the electron density map. SDS-PAGE gels indicated that it was present in the crystals. A similar procedure was followed to determine the structure of the His tag purified protein. Western blot analysis again confirmed that the NTD residues were present in the crystals. PyMOL⁴⁶ was used to generate symmetry mates, analyze solvent channels and crystallographic packing, and produce images.

Bioinformatics

The DISOPRED2 prediction of protein disorder server was used to predict disordered residues.³² Primary sequences were taken from the UniProt database.⁴⁷ Structures of GlnRS from E. coli (PDB ID: 1NYL) and D. radiodurans (PDB ID: 2HZ7) and GluRS from Methanobacterium thermoautotrophicus (PDB ID: 3AII) were taken from the PDB.^5,8,15,18 Sequence data and structural data were combined for use in the PROMALS3D multiple sequence and structure alignment server,⁴⁸ and the resulting alignment was used to identify structural motifs discussed in the text. tRNA sequences were taken from the tRNAdb 2009 online database.⁴⁹ BLAST (Basic Local Alignment Search Tool) was used to calculate expectation values (E values) for comparing sequences.^50,51

Small-angle X-ray scattering

Small-angle X-ray solution scattering data were collected on beamline 4-2 of the SSRL.⁵² Data were collected from the full-length protein at a wavelength of 1.3 Å for eight consecutive 2-s exposures collected at five different concentrations ranging from 1.0 to 7.7 mg/mL. Data were collected from the flow-through buffer of the final purification column [200 mM NaCl, 5% (v/v) glycerol, 2 mM DTT, 0.025% (w/v) NaN₃, and 20 mM Hepes buffer (pH 7.5)] and subtracted from the total protein solution scattering. The data were integrated with SasTool⁵² and then examined with PRIMUS.⁵³ Analysis of eight consecutive time frames ensured that no radiation damage took place over the course of the experiment. The SAXS data for different protein concentrations were investigated for aggregation using Guinier plots.²¹ Radius of gyration estimates were derived by the Guinier approximation $I\left(q\right)=I\left(0\right)\exp (-q^2{R}_{\mathrm{g}}^2∕3)$ with qR_g < 1.3 using the radius of gyration function of PRIMUS where q = 4πsinθ/λ. GNOM⁵⁴ was used to compute the pair distribution function, P(r), and to determine the maximum particle dimension, D_max. No evidence of concentration dependence was seen when comparing all five concentrations. Chicken egg white lysozyme was used as a protein standard to estimate the molecular weight from I(0) extrapolated from the scattering curve using GNOM in reciprocal space. Porod–Debye analysis and Porod volume estimations were performed using a pre-release version of the software SCÅTTER²² and confirmed with PRIMUS. Ten ab initio shape reconstructions were generated by DAMMIF⁵⁵ and averaged with DAMAVER.⁵⁶ Ensemble modeling was carried out with the ensemble optimization method²⁴ using default parameters. Fifty identical runs were performed using 20 conformers in the ensemble selected from among 10,000 random conformations of full-length Gln4, while treating the NTD and CTD crystal structures as independent rigid bodies and allowing the linker (residues 188–214) to be flexible. Rigid-body modeling was performed similarly using only one conformer in the ensemble, which selected the best single conformation from among all 10,000 random conformations.

Molecular dynamics

The initial model used for molecular dynamics simulation was generated as described in the text. All structural alignments were performed using the “fit” function of PyMOL using carbon α atoms. Molecular dynamics simulations were performed in GROMACS with the AMBER99SB force field.^57,58 The initial model was solvated using a cubic SPC/E water model⁵⁹ and neutralized with ions prior to minimization via steepest descents. Distance restraints were added to keep the zinc ion in place. The model was then equilibrated under an isothermal–isochoric ensemble for 100 ps at 300 K followed by equilibration under an isothermal–isobaric ensemble for 100 ps to ensure stability prior to the simulation. Simulations were then performed at the Center for Computational Resources⁶⁰ on 512 processors. Total simulation time was 70 ns. The DynDom Protein Domain Motion Analysis Web server²⁷ was used to compare models. RMSD from the starting structure was calculated for each frame of the trajectory after fitting to the starting structure. RMS fluctuation was calculated as the deviation from the starting structure after fitting to the starting structure.

Abbreviations used

NTD

N-terminal domain

CTD

C-terminal domain

SAXS

small-angle X-ray scattering

PDB

Protein Data Bank

NSD

normalized spatial discrepancy

BME

β-mercaptoethanol

SSRL

Stanford Synchrotron Radiation Lightsource