The structure of the TsaB/TsaD/TsaE complex reveals an unexpected mechanism for the bacterial t6A tRNA-modification.

The universal N6-threonylcarbamoyladenosine (t6A) modification at position A37 of ANN-decoding tRNAs is essential for translational fidelity. In bacteria the TsaC enzyme first synthesizes an l-threonylcarbamoyladenylate (TC-AMP) intermediate. In cooperation with TsaB and TsaE, TsaD then transfers the l-threonylcarbamoyl-moiety from TC-AMP onto tRNA. We determined the crystal structure of the TsaB-TsaE-TsaD (TsaBDE) complex of Thermotoga maritima in presence of a non-hydrolysable AMPCPP. TsaE is positioned at the entrance of the active site pocket of TsaD, contacting both the TsaB and TsaD subunits and prohibiting simultaneous tRNA binding. AMPCPP occupies the ATP binding site of TsaE and is sandwiched between TsaE and TsaD. Unexpectedly, the binding of TsaE partially denatures the active site of TsaD causing loss of its essential metal binding sites. TsaE interferes in a pre- or post-catalytic step and its binding to TsaBD is regulated by ATP hydrolysis. This novel binding mode and activation mechanism of TsaE offers good opportunities for antimicrobial drug development.

INTRODUCTION

Transfer RNAs (tRNAs) are subjected to considerable posttranslational processing (1) and today >90 chemical modifications in tRNAs have been characterized (2). The N-threonylcarbamoyladenosine (t6A) modification at position 37 of ANN-decoding tRNAs is one of the few modifications that are found in the three domains of life (3–5). Studies using oocytes demonstrated that A37 and U36 are strictly necessary for t6A formation (6). The t6A base in Escherichia coli tRNALys stacks with its adjacent A38 and forms a cross-strand stack with the first base of the codon on the mRNA, contributing to the translational fidelity (7). Mutations within the t6A modification pathway compromise anticodon–codon interaction, creating erroneous selection of start codons and aberrant frameshifts. They result in pleiotropic phenotypes and some mutations in human were recently associated with severe neurodegenerative and renal-neurological diseases (8,9). Although the modification is known for >40 years, the t6A biosynthesis enzymes and pathways have only recently been discovered (10–15). t6A biosynthesis is composed of two main steps (schematized in Figure 1): in the first, members of the universal TsaC/TsaC2 (YrdC/Sua5) protein family synthesize an unstable threonylcarbamoyladenylate (TC-AMP) intermediate from L-threonine, bicarbonate and adenosine triphosphate (ATP); the second step consists of the transfer of the threonylcarbamoyl (TC) moiety from TC-AMP onto A37 of substrate tRNA. In bacteria, The TC-transfer is carried out by the universal TsaD enzyme (YgjD in E. coli) assisted by two other proteins that are present in almost all bacteria: TsaB (YeaZ in E. coli) and TsaE (YjeE in E. coli). In archaea and eukaryotes, the TC-transfer reaction is carried out by the KEOPS multi-protein complex, composed of Kae1 (the TsaD orthologue), Bud32 (a small protein kinase), Cgi121 and Pcc1, complemented by a fifth protein (16–20) (Gon7 in yeast and C14ORF142 in human). Although TsaD/Kae1 is carrying the TC-transfer activity, the presence of the other subunits are mandatory for the reaction (exception made for Cgi121). In contrast, the yeast mitochondrial TsaD orthologue, Qri7, is capable of, if provided with TC-AMP, modifying tRNA in vitro without assistance of ancillary proteins (12).

Figure 1.

Reaction scheme for t6A synthesis in bacteria. TsaC uses l-threonine, bicarbonate and ATP for the synthesis of a very unstable TC-AMP intermediate. TsaBD and TsaE then collaborate to transfer the TC-moiety from TC-AMP onto the A37 of tRNA substrates.

The biochemical function of the ancillary proteins involved in the t6A pathway is largely unknown, especially the contributions of the TsaE ATPase (bacteria) and the Bud32 kinase (archaea/eukaryotes) to the t6A synthesis reaction remains enigmatic. TsaB is a paralogue of TsaD that has lost its catalytic function and TsaE belongs to a large class of nucleotide hydrolases. TsaD physically interacts with TsaB and TsaE and their genes are often clustering (15,21–23). The crystal structures of the heterodimeric TsaBD (22,24) showed that TsaB and TsaD associate via hydrophobic packing of 2 two pairs of N-terminal helices into a helical bundle, mimicking the structure of homodimeric TsaB (25,26) and Qri7 (12). Although Pcc1 is unrelated to TsaB, a similar helical bundle is also found at the Kae1–Pcc1 heterodimer interface (27). Except for some intracellular or symbiotic organisms (e.g. Mycoplasma), TsaE is present in all bacteria (3). TsaE has an intrinsically weak ATPase activity that is strongly activated by its binding to the TsaBD complex regardless of the presence of other substrates for the biosynthesis of tRNA t6A (22–24). The need of ATP-hydrolysis for t6A biosynthesis in bacteria is presently unclear. The structure of TsaE in complex with ADP revealed the presence of a Walker A motif and two switch regions that are characteristic of P-loop GTP hydrolases (28,29) In presence of non-hydrolysable ATP analogues, EcTsaB, EcTsaD and EcTsaE form a EcTsaBDE complex with 1:1:1 stoichiometry (22). From small angle X-ray scattering (SAXS) measurements and biochemical data, it was proposed that EcTsaE binds at the EcTsaBD interface (22). In a very recent manuscript, Luthra et al. reported the characterization of the TsaBDE threonylcarbamoyl transfer complex of T. maritima (TmTsaBDE) (23). Their SAXS analysis revealed that TmTsaBDE forms a symmetric hexameric quaternary assembly in solution with 2:2:2 stoichiometry (TmTsaB2D2E2). This hexamer turned out to be a dimer of TmTsaBDE heterotrimers, with TmTsaB acting as a dimerization module, reminiscent of Pcc1 in eukaryotic and archaeal KEOPS (30,31). TmTasB and TmTsaD were further shown to form a TmTsaB2D2 complex that was capable to bind one tRNA molecule, and accommodate a single TsaE subunit. Quantitative activity measurements demonstrated that TmTsaB2D2 alone can catalyse only a single round of t6A synthesis (23). An ensemble of kinetic and binding experiments led to the hypothesis that TmTsaE-catalyzed ATP hydrolysis occurs after the release of the t6A-modifed tRNA (23).

The essential character of TsaE for bacterial survival, the presence of its encoding gene in all pathogenic bacterial genomes and the absence of human orthologues, makes TsaE an attractive antibacterial drug target (29,32). However, its role in the t6A process remains poorly understood, mainly because there is no high resolution structural information on its interactions with TsaBD. In this manuscript we present the X-ray structure of the TmTsaBDE complex bound to AMPCPP at 3.1 Å resolution. TmTsaE is bound at the interface of TmTsaBD, making asymmetric contacts with TmTsaB and TmTsaD. The AMPCPP nucleotide was found sandwiched between TmTsaE and TmTsaD with its γ-phosphate group being coordinated by side chains from both subunits. Crystallographic symmetry generates the hexameric TmTsaB2D2E2 complex that was proposed to be present in solution (23). Surprisingly, TmTsaE partially denatures the active site of TmTsaD and prevents simultaneous substrate tRNA binding, suggesting that TmTsaE is required for a reaction step before or after the TC-transfer from TC-AMP to tRNA.

MATERIALS AND METHODS

Expression and purification of the TmTsaBDE subunits

We prepared the TmTsaBD binary complex and the TmTsaE protein individually. A bicistronic expression vector (named BC14) aimed to co-express TmTsaD and TmTsaB-His was designed as described (33). All synthetic gene constructs were optimized according to E. coli codon usage using the EuGene software (34) (Supplementary Table S2) and obtained from Genscript (Piscataway, USA). The DNA sequences coding for TmTsaD and TmTsaB containing a 6His-tag at the C-terminus have been cloned into a pET21a backbone using the NdeI and XhoI restriction sites. The second vector (named BC15) aimed to express TmTsaE-Nt-his was cloned into a pET24d backbone between NcoI and XhoI restriction sites. This plasmid, was used to express TmTsaE with a N-terminal 6His-tag followed by a TEV proteolytic recognition site to remove the 6his-tag if necessary. TmTsaD and TmTsaB-His have been co-expressed in E. coli Rosetta pLysS strain transformed with the BC14 plasmid and grown at 37°C in 2xYT liquid medium supplemented with ampicillin and chloramphenicol until mid-log phase. Expression was induced by adding 0.5 mM isopropyl β-d-1-thiogalactopyranoside and the cells were incubated for three more hours at 37°C and then harvested by centrifugation, re-suspended in lysis buffer (20 mM Tris–HCl pH 7.5, 200 mM NaCl), and stored at −20°C until purification. The same protocol was followed for the expression of TmTsaE, using plasmid BC15 and tetracycline antibiotic instead of ampicillin.

For purification, cells were thawed at 25°C and lysed by sonication (Branson Sonifier 250) on ice-cold water and centrifuged at 20 000 g for 30 min. Supernatant was applied onto NiIDA resin (Qiagen) previously equilibrated with Tris–HCl buffer (25 mM Tris–HCl pH 7.5, 200 mM NaCl, 5 mM 2-mercaptoethanol). Resin was washed with 20 ml of the same buffer and the proteins were eluted using three fractions of 4 ml of the previous buffer supplemented with 100, 200 and 400 mM imidazole. Fractions containing the proteins of interest were pooled and concentrated by ultra-filtration and loaded on superdex 75 (GE-Healthcare) equilibrated in HEPES buffer (20 mM HEPES pH 7.5, 200 mM NaCl, 5 mM 2-mercaptoethanol).

Finally, fractions containing the proteins of interest were pooled and concentrated to 50 mg.ml-1 or 60 mg.ml–1 for TmTsaE and TmTsaBD respectively. Purity was checked by SDS-PAGE.

Crystal structure determination

Purified TmTsaBD and TmTsaE were mixed at a 1:1:1 ratio of the three subunits up to a final concentration of 12 mg.ml−1 (0.148 mM) in HEPES buffer 20 mM pH 7.5, 200 mM NaCl, 5 mM 2-mercaptoethanol supplemented with 3 mM AMPCPP and 6 mM MgCl2. Crystals of TmTsaBDE were obtained at 293 K using the sitting-drop vapour diffusion method by mixing 100 nL of a protein solution with 100 nl of 13% PEG 8000, 0.1 M Imidazole pH 7.35. Crystals were cryoprotected with reservoir solution supplemented with 30% glycerol and 3mM AMPCPP, flash-frozen and stored in liquid nitrogen for data collection. X-ray diffraction data collection was carried out on beamline Proxima2 at the SOLEIL Synchrotron (Saint-Aubin, France) at 100K. Data were processed, integrated and scaled with the XDS program package (35). TmTsaBDE crystals belonged to space group P212121 with unit cell parameters of a = 84.31 Å, b = 113.94 Å, c = 177.62 Å. The structure of TmTsaBDE was solved by molecular replacement using the PHASER module (36) implemented in the CCP4 software package (37). Search models for TmTsaB and TmTsaD were obtained by the MODELLER program (38) using the crystal structure of the EcTsaBD heterodimer (PDB ID: 4YDU) and TmTsaE was modelled from the HiTsaE structure (PDB ID: 1FL9). The initial structure was refined using the PHENIX program (39) and completed by interactive and manual model building using COOT (40). Two copies of the TmTsaBDE trimer were present in the asymmetric unit. AMP-CPP molecules and MgCl2 ions could be modeled into the residual Fo-Fc electron density contoured at 3.0σ. Data collection and refinement statistics are gathered in Table 1.

Table 1.

Crystallographic data-collection and refinement statistics of TmTsaBDE

X-ray source	PROXIMA 2
Wavelength (Å)	0.9801
Temperature (K)	100
Unit-cell parameters (Å, °)	a = 84.31, b = 113.94, c = 177.62, α = β = γ =90.0
Space group	P212121
Resolution limitsa (Å)	48.44 – 3.14 (3.33–3.14)
Number of observationsa	138146 (21963)
Number of unique reflections	30297 (4627)
R-measa (%)	19.1 (149.7)
Completenessa (%)	99.1 (95.2)
I/σ (I)a	7.1 (1.03)
CC (1/2)a (%)	99.2 (43.9)
Number of non-hydrogen atoms (proteins/ligands)	10418/77
R/R free (%)	22.58/29.37
R.M.S.D. Bonds (Å)/angles (°)	0.010/1.258
Average B-factor (proteins/ligands)	80.67/80.91

aValues in parentheses refer to the highest resolution shell.

tRNA docking

The EcTsaDB-tRNA complex model was generated by rigidly docking a tRNA molecule onto the EcTsaBD complex using the HADDOCK Server (version 2.2) (41). The tRNA structure was extracted from the crystal structure of threonyl-tRNA synthetase-tRNA complex (PDB : 1QF6) (42) in which the E. coli tRNAThr UGC contains a m6t6A at position 37 in the anticodon loop. For docking, the m6t6-moeity and other modifications were removed from the E. coli tRNAThr UGC. The docking protocols and parameters follow the default settings for protein-RNA complexes in HADDOCK. As for the direct contact sites, critically, His111 and His115 in EcTsaD were specified as active site residues in the receptor molecule whereas A37 was specified as substrate in tRNA. By default, the active sites-surrounding residues were specified as passive. Resulting models were ranked based on scoring performance and clustered. The first model from the top-ranked cluster was chosen to represent the structures.

Sequence alignment and structure comparison

The sequence conservation was analysed using the program CONSURF (43) with TmTsaBDE as the query. The frequency of the occurrence of amino acids on the consensus sequence was produced by WebLogo (44). Structural comparison was performed with the PDBeFold web server (45). All graphic structure representations were produced with PYMOL (46). The electrostatic potential representation were rendered and produced by PDB2PQR (47) and APBS in the Chimera visualization system (48).

RESULTS AND DISCUSSION

Overall structure of the TmTsaBDE complex

We obtained diffracting crystals of TmTsaBDE only in the presence of AMPCPP and the structure was solved at 3.14 Å resolution. The asymmetric unit of the crystals contains two copies of a TmTsaBDE trimer, whose structures are very similar (all RMSD values for superimposed structures are gathered in Supplementary Table S1).

The overall structures of TmTsaB, TmTsaD and TmTsaE in the complex are the same as those of the individual proteins (TmTsaB) or orthologues (H. influenza (Hi)- and B. subtilis(Bs)TsaE and St(S. typhimurium), Ec(E. coli)TsaD and Kae1) (Supplementary Table S1) (28,29) (49). Sequence alignments with superposed secondary structure assignments of a few TsaB, TsaD and TsaE orthologues are represented in Supplementary Figure S1. As illustrated in Figure 2a, TmTsaB and TmTsaD embrace TmTsaE to form a compact ternary complex, in agreement with SAXS measurements on Ec- and TmTsaBDE in solution (22,23). TmTsaE is held in a grip between the helical insertion of the C-terminal domain of TmTsaD on one side and TmTsaB on the other. TmTsaE establishes two main contacts with TmTsaBD. The first involves the α3-helix (residues 90–100) of TmTsaE that binds to the N-terminal end of the four-helical bundle of the TsaBD interface. The α3-helix is partially disordered in BsTsaE and forms two short perpendicularly oriented helices in HiTsaE (illustrated in Supplementary Figure S2). This helix and the connection between β5 and β6 (residues 115–120) contact both N- and C-terminal domains of TmTsaB. The second main contact between TmTsaE and TmTsaBD is provided by the loops emanating from the C-terminal end of the parallel section of the TmTsaE β-sheet and the helical insertion in the C-terminal domain of TmTsaD, englobing the AMPCPP ligand (Figure 2A and Supplementary Figure S8). The totally conserved TmTsaE peptide S61PT(F/Y)T, dips into the active site of TmTsaD, containing two essential histidines (His109 and His113, highlighted in Figure 2A). This peptide contacts α4 and the hairpin connection between β6 and β7 of TmTsaD, forming H-bonds between well conserved residues (Supplementary Figure S1). The loop between β6 and β7 sandwiches the γP-group of ATP in its complex with StTsaBD (24). Since the S61PT(F/Y)T motif is also involved in a H-bond network with the γP-group of AMPCPP, this loop could be a communicator between the compound bound in the active site of TsaD and the ATP binding site on TsaE (see discussion in mechanism section).

Figure 2.

Structure of TmTsaBDE. (A) Ribbon diagram of TmTsaBDE (TmTsaB: blue, TmTsaD: red, TmTsaE: yellow). The main structural elements involved in protein interfaces are indicated following the colour codes of the individual proteins (TmTsaBD : α1 and α2 from TmTsaB and TmTsaD; TmTsaDE: α4 and α6 from TmTsaD and loops from TmTsaE). The two histidines (His109 and His113) in the active site of TmTsaD responsible for metal binding and the S61PTFT motif of TmTsaE are indicated by an arrow. AMPCPP at the interface between TmTsaD and TmTsaE is represented by sticks. The Mg-ion boud to AMPCPP is represented as a grey sphere. (B) AMPCPP binding site. TmTsaE in yellow and TmTsaD in red, AMPCPP is in sticks, Mg-ion as orange sphere. Side chains in H-bond interaction with the AMPCPP γP group are represented as sticks.

The TmTsaB C-terminus is elongated and disordered beyond residue 194 in one copy of the asymmetric unit, while in the other it forms a helix comprised between residues 190 and 203. We showed that the C-terminal peptide of EcTsaB is not required for formation of the binary EcTsaBD complex (22), but it might contribute to the stability of EcTsaBDE. The regions involved in these contacts are very poorly conserved, suggesting that they are not a major determinant for TsaE binding.

Although TsaB and TsaD have the capacity in vitro of forming homodimers whose quaternary structures are very similar to the TsaBD heterodimer, individually they do not interact with TsaE (22,23). Although there is no structure of the TsaD homodimer available, it likely resembles its mitochondrial orthologue Qri7, whose homodimer assembly is very similar to that of TsaB (12). TmTsaE is engaged in a totally asymmetric interaction with TmTsaB and TmTsaD, explaining why TsaE binds neither TsaB nor TsaD homodimers (22).

AMPCPP binds at the interface between TmTsaD and TmTsaE

After initial refinement of the molecular replacement solution, a consistent cloud of residual electron density was identified at the expected nucleotide-binding site of TmTsaE. As shown in Supplementary Figure S4, this density could be fitted by AMPCPP, whose presence was mandatory for obtaining diffracting crystals. Residual electron density was also present for a Mg2+ ion, coordinated by the β- and γ-phosphate groups of AMPCPP, Thr42Oγ and Glu108COO−. AMPCPP is located at the C-terminal edge of the parallel strands of the central β-sheet. Most remarkably, AMPCPP is sandwiched between TmTsaE and the helical insertion of the C-terminal domain of TmTsaD that partially seals off its polyphosphate. The AMPCPP nucleotide adopts an unusual syn-conformation and overlaps with ADP in complex with HiTsaE and BsTsaE (28,29) (Supplementary Figure S2). TmTsaE binds AMPCPP in a manner globally observed in other P-loop proteins (Figure 2B). The N6 atom of adenine hydrogen bonds to main chain atoms of Ser132 and Glu11 and the base-ring is stacked between the side chains of Glu11 and Arg134, which also hydrogen bonds with the ribose O4. The ribose O2 hydrogen bonds with the Glu11 carboxylate and the ribose O3 establishes a H-bond with the main chain carbonyl of TmTsaD Phe116. The phosphate moieties of AMPCPP make classic hydrogen contacts with the TmTsaE P-loop (G38AGKTT) at the N-terminus of α2. EcTsaE does not bind to EcTsaBD in presence of ADP (22) and the affinity of TmTsaE for TmTsaBD drops fourfold in absence of ATP (23). It was also shown that the basal low ATPase activity of EcTsaE is boosted by the presence of EcTsaBD (22). These observations can be explained by the fact that the position of the AMPCPP γP-group is stabilized by salt bridge formation with the conserved Lys166 and Lys213 from TmTsaD (Figure 2B). Lys166Nζ also hydrogen-bonds with Thr63OγH that is part of the totally conserved SPT motif in contact with the TmTsaD active site. The carboxylate group of the totally conserved TmTsaE Asp80 is at ∼5 Å from the γP of AMPCPP, ideally positioned to act as a general base for ATP hydrolysis. It was noticed that this Asp is part of a conserved HxD(L/V)YR motif that was reminiscent of Hanks type kinases (29,51). Mutation of the homologous Asp80 in BsTsaE annihilated its (auto)phosphorylation activity.

Different effects of ATP nucleotides on the binding of TsaE to TsaBD have been observed for different organisms. TmTsaE is able to bind TmTsaBD in absence of added nucleotides with an affinity constant of 1.3 μM, which decreases about 4-fold in presence of ATP, while ADP has no effect (23). Two active-site mutants defective in ATP binding, TmTsaET42A and TmTsaEE108A, preserved their binding to TmTsaBD (23). EcTsaE does not bind to EcTsaBD in absence of ATP but binds with 1:1:1 stoichiometry and a 0.6 μM Kd in presence of Mg2+ and AMPPNP (22). EcTsaET43A and EcTsaEE108A mutants (also defective in binding to ATP or ADP) did not interact with EcTsaBD (22). The strength and ATP dependence of the binding of TsaE to TsaDB probably is species dependent and variations might exist between for instance hyperthermophilic and mesophilic bacteria.

Comparison of TsaE with small GTPases suggested that Tyr82 and Trp109 are part of putative switch regions (Supplementary Figure S2) (22,50). Trp109Nϵ1 (switch I) hydrogen-bonds with the γP of AMPCPP and Tyr82 (switch II) stacks against TsaD Lys166. EcTsaEY82A and EcTsaEW109A had ∼7-fold diminished affinity for ATPγS but EcTsaEW109A kept WT binding capacity to EcTsaBD in presence of AMPPNP. Although the EcTsaEW109A mutant had diminished ATPase activity it was active in t6A biosynthesis. It should be noticed that the conformation of switch regions in small GTPases is usually strongly influenced by the nature of the bound nucleotide (GTP versus GDP) (50), but this is not the case for TmTsaE.

To conclude, AMPCPP binding in TmTsaBDE is conferred by residues both from TmTsaD and TmTsaE. Residues from TmTsaD complete the coordination of the γP group of ATP and hence are likely responsible for the activation of the hydrolytic activity of TmTsaE upon binding to TmTsaBD. Our structure clearly explains in general why ATP is needed for (or increases) the binding of TsaE to TsaBD and why this complex boosts the ATPase activity of TsaE.

TmTsaE binding induces an inactive conformation of TmTsaD

Global superposition of Tm- and EcTsaD indicates that the relative positions of the N- and C-terminal domains changes considerably upon formation of the TmTsaBDE complex (Figure 3A). The α3 and α9 helices of TmTsaD at the bottom of the active site shift their positions by ∼4 Å forming the hinge of a rotation that separates the C- and N-terminal domains and opens up the active site cleft. At its periphery the N-terminal domain of TmTsaD moves by ∼11 Å. These observations corroborate reported SAXS data obtained from TmTsaB2D2 in solution, which were best fitted by an equilibrium population between an open and closed form of TmTsaD (23). The distance between the two TmTsaD units in the ‘open form’ of TmTsaB2D2 was estimated to be 85 Å, a value that agrees with the present crystal structure of TmTsaB2D2E2. The closed form of TmTsaB2D2E2 likely corresponds with the structure of TsaD as observed in the St- and EcTsaBD binary complexes.

Figure 3.

Structure comparison between EcTsaBD and TmTsaBDE. (A) Superimposition of EcTsaBD (PDB ID 4YDU; EcTsaB in cyan and EcTsaD in gray) and TmTsaBDE (same colour code as for Figure 2A). The C-terminal domains of TsaD were superimposed. The ADP (bound to EcTsaBD) and the AMPPNP bound to TmTsaE are shown as sticks. As indicated the N-terminal domain position of TsaD shifts by ∼11 Å between EcTsaBD and TmTsaBDE. (B) Zoom into the TsaD active site. Superposition of EcTsaBD (gray shadow) and TmTsaBDE (red shadow). Fe- and Mg-binding side chains are shown using the same color code. Fe- and Mg-ions are in pink and grey spheres respectively and ADP is in sticks.

Apart from pushing the two domains of TmTsaD apart, TmTsaE binding partially ‘melts’ its active site pocket. Electron density was missing for residues 31–47 of TmTsaD for both copies in the asymmetric unit, suggesting they became disordered. This peptide region has a well-defined conformation in binary TsaBD complexes: residues 31–39 form a helix (α1) that is collinear with strand β3 and residues 41–47 form the two first helical turns of α2. The superimposition of EcTsaBD and TmTsaBDE onto the TsaD N-terminal domains (Supplementary Figure S5), shows that the structure of this peptide as present in EcTsaBD would create an important steric clash with TmTsaE. A second region of TmTsaD that becomes disordered in TmTsaBDE, is contained between residues 292 and 294. The corresponding peptide in EcTsaBD provides a carboxylate ligand (EcAsp300) to the Fe-ion. The induction of partial disorder and conformational changes into the TmTsaD active site by the presence of TmTsaE significantly affects the positions of many of the crucial residues involved in AMP and metal binding. Figure 3b illustrates ADP bound in the active site of EcTsaBD (22,24) superposed onto the active site of TmTsaD. The α- and β-phosphates are ligands of a metal ion (Fe in EcTsaBD), which is further coordinated by His111 (TmHis109), His115 (TmHis113) and Asp300 (TmAsp296). A Mg-ion is positioned at 4 Å away from the α- and β-phosphates of ADP and is coordinated by the Asp11 (TmAsp11) and Glu12 (TmGlu12) carboxylates. The opening of the active site pocket and the induced disorder render the metal binding sites of TmTsaD ineffective (Figure 3B): (i) His109 and His113 shifted over >3 Å and are no longer in a configuration capable of binding a metal ion; (ii) the peptide containing the third metal-coordinating residue Asp296 became disordered; (iii) the two Mg-coordinating residues (Asp11 and Glu12) also moved by ∼3 Å away from the polyphosphate moiety position of the nucleotide. Mutations of the homologous histidines in Kae1, renders yeast strains hardly viable and abolishes the t6A-synthesizing activity (27,31). The EcTsaDE12A mutant is totally inactive, although it is still capable of binding EcTsaE and EcTsaB (22), suggesting that the integrity of the Mg-binding site of TsaD is essential for activity but not for TsaE binding. We conclude that the binding of TmTsaE opens and partially melts the active site of TmTsaD, resulting in an inactive conformation of TmTsaD.

TmTsaE blocks the active site entrance for a tRNA substrate

The biochemical characterization of tRNA binding to TmTsaBDE was recently reported (23). Both TmTsaB2D2 and TmTsaB2D2E are able to bind tRNA with μM affinity, but no tRNA binding was observed for TmTsaB2D2E2. These data corroborated our observations that tRNA did not bind to EcTsaBDE (22). Our structural data explain why TsaE and tRNA are competing for binding to TsaBD: the position of TsaE prevents the access of the anti-codon loop of the bulky tRNA substrate to the TsaD carbamoylation site (centered on the Fe/Mg cluster). The surface projection of the electrostatic potential (Figure 4A) shows that TmTsaBD contains an extensive positively charged groove surrounding the active site pocket, obviously a good candidate for tRNA binding. We modeled a complex of a rigid tRNA substrate with EcTsaBD using the HADDOCK software (41). The tRNA in the HADDOCK model extensively interacts with the positive surface patch present on TsaBD (Figure 4B) and most importantly sterically overlaps with TmTsaE in the TmTsaBDE complex. Electrostatic surface calculation of TmTsaE reveals a negatively charged surface patch centered on the α3 helix that is positioned onto a positively charged patch at the TmTsaBD interface. Although the residues at the interface are not completely conserved, the electronegative character of this region is maintained in the Ec-, Bs- and TmTsaE homologues (Supplementary Figure S6). TsaE can be considered as an electrostatic competitor of the tRNA substrate.

Figure 4.

TmTsaE and tRNA binding sites overlap. (A) Left panel: the electrostatic potential surface of TmTsaE and ribbon presentation of TmTsaBD; right panel: the electrostatic potential surface of TmTsaBD and ribbon presentation of TmTsaE (negatively and positively charged surfaces are in red and blue respectively). Ribbon presentation uses same colour code as Figure 2A. (B) Left panel: the HADDOCK model of EcTsaBD (in ribbon) bound to tRNA; right panel: view of the active site of the modelled EcTsaBD-tRNA-TC-AMP complex. TC-AMP is in sticks. tRNA-A37 is in green sticks and some of the metal ligands are in red sticks. Fe-ion in pink and Mg-ion in gray sphere.

The competition between tRNA and TsaE also explains why the strong ATPase activity upon binding of TsaE to TsaBD is observed independently of tRNA (22,23). In agreement with conclusions from biochemical data (22,23), the TmTsaBDE structure definitely confirms that TsaE must operate in a reaction step that is unrelated to the proper carbamoyl-transfer from TC-AMP onto tRNA A37.

Quaternary structure of TmTsaBDE

The structure of the TmTsaB dimer was previously determined by a structural genomics consortium (25). The crystal packing in that study suggested two possible dimerization modes for TmTsaB. In the first, the β3 strands of the two N-terminal domains align to form an antiparallel extended β-sheet. In the second dimer type, the α1 and α2 helices of the N-terminal domains of each subunit are packed into a four-helical bundle, closely mimicking the helical arrangement observed at the TsaBD interface. Although the latter dimerization has a smaller contact area mode, it was found exclusively in all other TsaB homologue structures (26,52). SAXS analysis of TmTsaBDE demonstrated it forms a heterohexamer in solution consisting of two copies of each subunit (TsaB2D2E2) (23). The two copies of the TmTsaBDE trimer in the asymmetric unit do not associate into a hexamer. However, TmTsaB forms a crystal symmetry generated dimer that involves the β3 strands of the two N-terminal domains. Using TmTsaB as the dimerizing subunit, we generated a TsaB2D2E2 heterohexamer (Figure 5A) that is very close to the one proposed from SAXS data (23). We confirm that the quaternary structure of the TmTsaBDE complex is different from trimeric Ec- and StTsaBDE (22–24). Structure imposition reveals that Ec- and StTsaB have a 19 amino acid insertion in the C-terminal domain compared to TmTsaB, that forms an extra strand and a long helical connection. This insertion is positioned close to the TmTsaB interface region observed in the structures of TmTsaB and TmTsaBDE. Superimposition of the Ec- and StTsaB monomer structures onto the TmTsaB dimer creates a considerable steric clash between the inserted regions of the two subunits (not shown). Therefore, Ec- and StTsaB would not be able to adopt the TmTsaB type dimer structure, explaining why Ec- and StTsaBDE are trimers and not hexamers in solution. This is probably the case for the majority of species, since most TsaB sequences have this C-terminal insertion and we expect that the trimeric TsaBDE complex is the standard configuration.

Figure 5.

Quaternary structure of TmTsaBDE. (A) Ribbon representation of the TmTsaB2D2E2 hexamer, a dimer of TmTsaBDE trimers, that was extracted from the crystal packing. This assembly agrees with the hexamer which was proposed in solution, based on SAXS experiments (23) (TmTsaB: blue, TmTsaD: red, TmTsaE: yellow). TmTsaB forms the central dimerization unit, corresponding to one of the dimer configurations in the crystal structure of TmTsaB alone (superposed in grey, PDB ID 2A6A), but was not observed in crystal structures of homologous TsaB structures. (B) Superimposition of StTobZ (PDB ID 3VEN; StTobZ-TsaD like in gray and StTobZ-YrdC like in black) and TmTsaBDE (same color code as for Figure 2A).

As discussed above, bacterial TsaBDE adopts different quaternary assemblies according to species. This type of heterogeneity was also observed in the archaeal and eukaryotic systems. Archaeal KEOPS is a dimer of heterotetramers (Pcc1-Kae1-Bud32-Cgi121) (30) while in eukaryotes, a fifth KEOPS subunit (Gon7 in yeast) prevents Pcc1 dimerization and the functional unit is a heteropentamer. In humans, C14ORF142 plays a similar role as Gon7 in yeast and prevents the dimerization of the LAGE3 (the Pcc1 orthologue) subunit (19). In absence of C14ORF142, LAGE3-OSGEP-TPRK-TPRB forms a dimer of heterotetramers.

Activation mechanism of the t6A activity by TsaE

Kinetic experiments suggested that, in absence of TmTsaE, TmTsaB2D2 is able to catalyse a single turnover. However, the presence of TmTsaE is required to attain multiple cycles of t6A activity (23). TmTsaE mutants that have lost their ATPase or TmTsaBD binding activities are incapable of sustaining multiple-round catalysis of the TmTsaD enzyme. On the other hand, the present structure of TmTsaBDE, supported by recent biochemical data (23), proves that TmTsaE and tRNA-substrate cannot simultaneously bind to TmTsaBD. TmTsaE must therefore carry out an essential step on the enzymatic pathway occurring before or after the transfer of the threonylcarbamoyl-moiety from TC-AMP onto A37. Luthra et al. suggested that after synthesis of t6A-tRNA, TmTsaD might be left in an inactive state and TmTsaE-mediated ATP hydrolysis might ‘reset’ TmTsaD for the next reaction cycle (23). Obviously, to provide access for the tRNA substrate to the active site of TmTsaD, TmTsaE must bind and dissociate from TmTsaBD during the reaction cycle, a process that is likely regulated by its ATPase activity. The affinity of TmTsaE for TmTsaBD is in the micromolar range, and diminishes in absence of ATP (23). For EcTsaE we did not observe binding to EcTsaBD without ATP (22). Apart from playing a regulatory role for tRNA binding, TmTsaE also affects the active site conformation of TmTsaD. TmTsaE perturbs the integrity of the TmTsaD metal-binding sites which are essential for activity and hence induces an inactive state of the TmTsaD enzyme. It is not obvious why TmTsaD should go through a ‘denaturation’ step during the reaction cycle. This kind of step could possibly be related to the release of reaction products or intermediates. However, affinities of TsaD for nucleotides and tRNA are moderate (22–24) and there is no evidence for the formation of stable intermediate/product complexes that would need assistance to release them from the enzyme. TsaE must therefore be involved in another step of the reaction cycle, for instance binding of the TC-AMP substrate. Inspection of TmTsaBDE shows that TC-AMP has full access to the active site. Since the lability of the TC-AMP intermediate hampers its use in structural studies, we made a model of the EcTsaBD -TC-AMP complex by superposing its AMP-moiety onto the EcTsaBD-ADP complex (22) (Figure 4b). This model places the carbamoyl group of TC-AMP close to the Fe-ion. We then compared this model with substrate complexes of TobZ, which is a fusion protein of an TsaC- and a TsaD-like module (53). TobZ from Streptoalloteichus tenebrarius is responsible for the O-carbamoylation of tobramycin (an aminoglycoside related to kanamycin) to form nebramycin 5′. Its TsaC-module catalyses the formation of a carbamoyl-AMP intermediate, whose carbamoyl moiety is subsequently transferred to tobramycin by the TsaC-module. It was noticed that the active sites of TsaD and TobZ-TsaD are very similar (53). In TobZ complexes with carbamoyl-AMP and the acceptor tobramycin, the carbamoyl group is liganded to the Fe-ion and the tobramycin-OH group approached the Fe-bound phosphate (53). Superimposition of TobZ bound to carbamoyl-AMP and the acceptor tobramycin with our model of the TsaD/TC-AMP complex showed a perfect overlap between the carbamoyl-AMP moieties. The Fe-cluster can hence be proposed as a carbamoylation centre for all t6A TC-transfer systems. Our HADDOCK model of the tRNA-TsaBD complex positions the N6-amino group of the A37 anticodon base close to the carbamoyl group of TC-AMP and is well positioned for a nucleophilic attack of the TC-AMP carbonyl (Figure 4b). We hypothesize that the altered conformation of TsaD induced by TsaE could still allow binding of TC-AMP but not its chemical conversion. Superposition of TobZ onto TsaBDE, shows that the TsaC-like module of TobZ is bound to the TsaD-like module at the opposite side of TsaD compared with TsaE. TobZ contains a large central cavity that connects the active sites of the TsaC- and TsaD-like modules. This tunnel prevents diffusion of the unstable carbamoyl-phosphate intermediate into the bulk solution and channels it directly to the active site of the TsaD-like module (53). Similarly, binding of TsaC to TsaBDE in a manner that resembles the association of the TobZ TsaC and TsaD modules, would create a complex that could allow efficient transfer of TC-AMP from the active site of TsaC to that of TsaD. The role of TsaE might be to ‘prepare’ TsaD for interaction with TsaC. Complex formation between the four t6A synthesis protein partners of T. maritima was observed by native gel shift experiments although a stable quaternary complex could not be isolated by gel filtration (23). For enzymatic purposes however, this complex would not need to be stable, but should only be sufficiently long-lived to permit TC-AMP transfer between TsaC and TsaD. Once the TC-AMP is bound, ATP could be hydrolysed by TsaE, promoting its dissociation from TsaBD, followed by binding of the tRNA substrate, renaturation of the TsaD metal binding sites and threonylcarbamoyl transfer to A37. The region around the active site that could be involved in TC-AMP channelling has conserved sequence patches (Supplementary Figure S7). The highly conserved SPT(F/Y)T loop of TsaE has its side chains in the coordination sphere of the γP of AMPCPP and is close to the threonylcarbamoyl moiety of TC-AMP in the TsaD active site. It is therefore well positioned to serve as a sensor to coordinate ATP hydrolysis and TC-AMP binding. A scheme representing the hypothetical role of TsaE during the t6A reaction cycle is represented in Figure 6.

Figure 6.

Hypothetic reaction scheme for TsaBDE. Binding of TsaE and ATP to TsaBD converts TsaD into an inactive conformation (symbolized by shady red colour). This conformation of TsaD then interacts with TsaC followed by transfer of TC-AMP (green stick) from TsaC to TsaD (pathways a or b). TsaE hydrolyses ATP and dissociates from TsaBD, liberating the binding site on TsaBD of the tRNA substrate. After transfer of the TC-moiety from TC-AMP to tRNA, the modified tRNA dissociates from TsaBD.

Comparison between bacterial TsaE and archaeal/eukaryotic Bud32

The threonyl-carbamoyl transferase multiprotein complexes in eukaryotes/archaea and bacteria have only the TC-transferring enzyme in common, Kae1 and TsaD respectively. Kae1 is flanked in the KEOPS complex by two subunits: Pcc1 binds to its N-terminal and Bud32 to its C-terminal domain (27,54). The helical bundle at the Pcc1-Kae1 interface mimics the TsaBD interface, despite the fact that Pcc1 and TsaB are unrelated. Pcc1 and TsaB likely play similar roles in the t6A reaction pathway. Superposition of TmTsaDE onto Kae1-Bud32 (Supplementary Figure S8) shows that TsaE and Bud32 both interact with the helical insertion of the C-terminal domain of TsaD and Kae1 respectively. TsaE however binds at the opposing end of TsaD compared to Bud32 and Kae1. Moreover, Bud32 does not block the access of tRNA to the active site of Kae1. The role of Bud32 in the t6A reaction pathway has not been clarified so far. Bud32, resembles the small protein kinases of the Rio family, involved in the biogenesis of ribosomes (55). An ATPase-dependent role of the Rio proteins was suggested to regulate their dynamic association with pre ribosomal subunits (56). Inactivation of the active site of Bud32 is deleterious for yeast cells and also for t6A activity of the KEOPS complex (14). It remains an open question whether TsaE and Bud32 are functional analogues or whether they play a different role in the t6A reaction cycle.

CONCLUSION

Our data provide a structural rationale for the biochemical observations collected on Ec- and TmTsaBDE, e.g. the competition between tRNA and TsaE for binding to TsaD. The structure of TmTsaBDE confirms TsaE is not directly involved in the TC-transfer to tRNA, but suggests that it ‘prepares’ TsaD at the start of the reaction cycle. TmTsaE converts TmTsaB2D2 into an open form that could accept the TC-AMP intermediate from TmTsaC2. After ATP hydrolysis, TmTsaE dissociates from TmTsaBD which switches to a closed active form, capable of binding metals and tRNA substrate. Then, modified tRNA leaves the active site and TmTsaBD enters into a new reaction cycle. The open (inactive) and closed (active) form of TmTsaBD might be in equilibrium in absence of TmTsaE as suggested by SAXS data (23). It is tempting to correlate these two states with the observation that TmTsaBD, in absence of TmTsaE, catalyses a single turn-over tRNA modification. This equilibrium probably is species-dependent since it was not observed for EcTsaBD (22).

It is of note that the TsaBDE complex offers an attractive target for antimicrobial drug development. Chemical compounds that either inhibit TsaE to bind to TsaBD or significantly stabilize the TsaBDE complex would have deleterious effects on t6A activity and bacterial survival.

DATA AVAILABILITY

The coordinates of the TmTsaBDE complex have been deposited at the PDB under the code 6FPE.

Click here for additional data file.