The ATP-mediated formation of the YgjD–YeaZ–YjeE complex is required for the biosynthesis of tRNA t6A in Escherichia coli

The essential and universal N6-threonylcarbamoyladenosine (t6A) modification at position 37 of ANN-decoding tRNAs plays a pivotal role in translational fidelity through enhancement of the cognate codon recognition and stabilization of the codon–anticodon interaction. In Escherichia coli, the YgjD (TsaD), YeaZ (TsaB), YjeE (TsaE) and YrdC (TsaC) proteins are necessary and sufficient for the in vitro biosynthesis of t6A, using tRNA, ATP, L-threonine and bicarbonate as substrates. YrdC synthesizes the short-lived L-threonylcarbamoyladenylate (TCA), and YgjD, YeaZ and YjeE cooperate to transfer the L-threonylcarbamoyl-moiety from TCA onto adenosine at position 37 of substrate tRNA. We determined the crystal structure of the heterodimer YgjD–YeaZ at 2.3 Å, revealing the presence of an unexpected molecule of ADP bound at an atypical site situated at the YgjD–YeaZ interface. We further showed that the ATPase activity of YjeE is strongly activated by the YgjD–YeaZ heterodimer. We established by binding experiments and SAXS data analysis that YgjD–YeaZ and YjeE form a compact ternary complex only in presence of ATP. The formation of the ternary YgjD–YeaZ–YjeE complex is required for the in vitro biosynthesis of t6A but not its ATPase activity.


INTRODUCTION
During maturation tRNAs undergo many posttranscriptional modifications, some of which are essential for cell life (1,2). The N 6 -threonylcarbamoyladenosine (t 6 A) modification at position 37 of ANN-decoding tR-NAs is one of the few modifications that are found in the three domains of life (3)(4)(5). The t 6 A base in Escherichia coli tRNA Lys stacks with its adjacent A 38 and forms a cross-strand stack with the first codon of the mRNA, contributing to the translational fidelity (6,7). Furthermore, several isopentenyl-adenosine derivatives at position 37, such as N 6 -methyl-N 6 -threonylcarbamoyladenosine (m 6 t 6 A) and cyclic t 6 A (ct 6 A), are derivatives of t 6 A in some tRNAs (8,9). Absent or defective t 6 A has been extensively implicated in compromised anticodon-codon interaction, erroneous selection of start codons and aberrant frameshift as well as numerous pleiotropic phenotypes (3,10). It has been shown that the biosynthesis of t 6 A proceeds in two main steps: in the first, members of the Sua5/YrdC protein family, present in bacteria, eukaryotes and archaea, utilize L-threonine, bicarbonate and adenosine triphosphate (ATP) to synthesize an unstable intermediate threonylcarbamoyladenylate (TCA) (11,12); in the second step the threonylcarbamoyl moiety of TCA is transferred onto A 37 of substrate tRNA (11)(12)(13). In bacteria, the transfer involves three proteins: YgjD (TsaD), YeaZ (TsaB) and YjeE (TsaE) from E. coli (the Bacillus subtilis orthologs are YdiE, YdiC and YdiB, respectively) (11,14). In archaea and yeast, the transfer reaction requires the kinase, putative endopeptidase and other proteins of small size (KEOPS) protein complex, composed of Kae1, Bud32, Cgi121 and Pcc1, complemented by a fifth fungi-specific protein Gon7 in yeast (13,(15)(16)(17). It is now well established that YgjD and its orthologs Kae1 and Qri7 are responsible for the catalysis of the transfer reaction. The contribution of the other protein partners is indispensable but their biochemical function is poorly understood (12)(13)(14)18). For example, the mutation of catalytic residues of Bud32, an atypical P-loop kinase (19,20), abolished the formation of tRNA t 6
The following mutations have been inserted using the Quikchange mutagenesis kit (Agilent Technologies): YgjD-F100E; YgjD-S97E; YgjD-S97R; YgjD-E12A; YgjD-V85E; YeaZ-220; YeaZ-R118A; YjeE-E108A; YjeE-T43A; YjeE-Y82A and YjeE-W109A. Primers (sequences are summarized in supporting materials) have been ordered from Eurofins Genomics, Les Ulis, France. Plasmid template (10 ng) has been incubated with 0.04 M of each forward and reverse primer, 1 M of dNTP mix (Thermo Scientific) and 1 unit of Phusion High Fidelity DNA polymerase (Thermo Scientific) in HF buffer. Twenty-five cycles of amplification have been carried out, corresponding to 1 min of denaturation at 95 • C, 1 min of annealing at 55 • C and 2 30 of extension at 72 • C. Polymerase chain reaction (PCR) tubes have been incubated for 5 more minutes at 72 • C as a final extension step. Parental plasmid has been digested for 2 h at 37 • C with DpnI restriction enzyme (Thermo Scientific) and 5 l of the PCR mix has been used to transform XL10 chimio-competent cells (Agilent Technologies). For the screening procedure, four to eight independent colonies have been grown in LB medium supplemented with the appropriate antibiotic and plasmid have been extracted using the Genejet plasmid miniprep kit (Thermo Scientific). Plasmids have been sequenced (Beckman Coulter Genomics) on the whole gene to check for the presence of the right mutation. In few cases an alternative protocol has been used where two separate PCRs have been carried out in parallel using only the forward and the reverse primer in two separate tubes and 100 ng of plasmid template. In this case 30 cycles of amplification have been done and the two tubes have been pooled, heated to 95 • C and slowly cooled down before DpnI digestion (37). The rest of the protocol was the same as the one mentioned above. In all cases, all the genes have been sequenced on their full length.
The expression and purification details of all proteins are presented in the Supplementary Materials sections.

Native gel migration assay
A 1.2% agarose gel was prepared in 30 mM Tris-HCl pH 8.0 and 50 mM Glycine. Protein samples supplemented with 25% glycerol were loaded onto the gel. Migration lasted for 50 min at 135 mV in a running buffer composed of 30 mM Tris-HCl pH 8.0 and 50 mM Glycine. For analysis of the ATP/ADP-dependent association of the proteins, both the gel and the loading protein samples were additionally supplemented with 1.0 mM of ATP or ADP and 2.0 mM MgCl 2 . The gel was running in an electrophoresis apparatus on ice with an ambient temperature at 8 • C. Proteins were revealed by Coomassie staining.

ATPase activity assay
The ATPase activity (conversion into ADP and phosphate) was measured using a Nicotinamide adenine dinucleotide dehydrogenase (NADH)-coupled assay (38). The hydrolysis of ATP into ADP was coupled to the oxidation of NADH to NAD + and the consumption of NADH was collected by monitoring the absorbance at A 340 by a Cary WinUV 200 spectrophotometer (Agilent Technology, Inc.). Assays were performed in a 10mm cuvette at 20 • C. The 150 L reaction buffer contained 50 mM HEPES pH 7.5, and an excess of pyruvate kinase and lactate dehydrogenase (Sigma). The reactions were monitored after the steady-state reaction was achieved and the linear curves of absorbance at A 340 were recorded for 30 min. The concentration of the proteins in the assays was 1 M except for YjeE and its variants, which were at 2 M. The reaction velocity was calculated by converting the decrease in the absorbance to consumption of NADH using an extinction coefficient of ε 340nm = 6220 M −1 ·cm −1 . ATPase reaction rates as a function of ATP concentration was fitted using Michaelis-Menten equation and the cooperactivity of ATPase activity of YjeE by the dimer YgjD-YeaZ was fitted by Allosteric sigmoidal kinetics equation using the Prism program (GraphPad). All the experiments were done in triplicate and control experiments were carried out in which either proteins or chemicals were omitted from the reaction.

tRNA overexpression and purification
The E. coli tRNA Lys(UUU) was subcloned in a pBlue-Script vector harboring T5 promoter/lac operator sequence (17). The pB-tRNA Lys(UUU) was transformed into E.coli XL10 competent cells and was grown in 10ml 2xYT medium at 37 • C overnight. The 10 ml preculture was transferred to 800 ml Overnight Express Instant TB medium (Novagen) and the overexpression was auto-induced during the culturing for 36 h at 37 • C. The cell culture was harvested and centrifuged for 25 min at 4400 rpm. The cell pellets were suspended in Trizol (Sigma), followed by extraction of the total RNAs according to the manufacturer's procedure (TRI Reagent, Sigma). The total RNA pellet was dissolved in water and applied for purification by a denaturing gel that is composed of 10% of polyacrylamide and 8M Urea. After migration for 4 h at 250 V, 25 mA in Tris-acetate-EDTA (TAE) buffer (40 mM Tris, 20 mM acetic acid and 1 mM ethylenediaminetetraacetic acid (EDTA)), the gel slice containing tRNAs as visualized by UV light shadowing at 258 nm and was cut out and incubated with elution buffer (10 mM Tris-HCl pH 7.5, 300 mM NaCl, 1 mM EDTA and 1% sodium dodecyl sulphate (SDS)) overnight at an ambient temperature. The eluted tRNAs were precipitated with absolute ethanol and NaCl. Finally, the tRNAs were further purified by size exclusion chromatography equilibrated with buffer that is composed of 10 mM HEPES pH 7.5, 100 mM NaCl. The size and purity of the tRNA was analyzed by 1.5% agarose gel.

The tRNA t 6 A assay
The in vitro biosynthesis of tRNA t 6 A, the digestion of tRNA and the high performance liquid chromatography/mass spectrometry (HPLC/MS) analysis of the t 6 A formation used published protocols and are presented in the Supplementary Materials sections.

YgjD forms a stable heterodimer with YeaZ
The four E. coli proteins YrdC, YgjD, YeaZ and YjeE are necessary and sufficient for the in vitro synthesis of tRNA t 6 A (14). YgjD and YeaZ are forming a heterodimer that is capable of interacting with YjeE (33). In order to get insight into the biochemical function of YrdC, YgjD, YeaZ and YjeE we decided to investigate their interactions in vitro. We first expressed and purified the four recombinant proteins. The C-terminal 6xHis-tagged YgjD is stable and is not sensitive to proteolysis as reported previously (33). We tested the capability of various combinations of the four proteins to form complexes by native gel shift experiments. A mobility shift is observed upon mixing YgjD and YeaZ, demonstrating they are spontaneously forming a complex ( Figure 1A, lanes [1][2][3][4]. No other combination of proteins provoked shifts in the migration. In order to further characterize the complex formed by YgjD and YeaZ, we subjected it to size-exclusion chromatography. These data confirm that YgjD and YeaZ alone are homodimers and that a heterodimer readily forms upon equimolar mixing of YgjD and YeaZ ( Figure 1B). The interaction between YgjD and YeaZ was quantified by ITC (Table 1 and Supplementary Figure S1). Despite the fact that both YgjD and YeaZ form homodimers the titration of YgjD with YeaZ in absence of nucleotides (ADP or ATP) resulted in a binding curve that could be fitted by a simple equilibrium scheme (YgjD-YeaZ complex with 1:1 stoichiometry and a Kd of 122 nM). These binding parameters are similar as those measured for the interaction between YgjD and YeaZ from S. typhimurium (Kd of 300 nM) (25).

Crystal structure of the heterodimer YgjD-YeaZ in complex of ADP
We cocrystallized the YgjD-YeaZ heterodimer with ATP and MgCl 2 and determined the structure at a resolution of 2.3Å. The overall structure of the YgjD-YeaZ heterodimer and the binding modes of nucleotides (ATP, ADP and AMP) are very similar to those of YgjD-YeaZ from S.  The corresponding thermograms are presented in Supplementary Figure  typhimurium, with an rmsd of 0.232Å for 571 superposed C ␣ atoms ( Figure 2A) (25). We will therefore only describe the most important structural differences between the two homologous complexes. The interface between YgjD and YeaZ resembles that of the most commonly observed YeaZ homodimer and of the Qri7 homodimer ( Figure 2A and G) (31,39). Interestingly the Kae1 interface region of the Kae1-Pcc1 complex is the same as for YgjD-YeaZ ( Figure 2H). Pcc1 is not related to YeaZ, but its binding to Kae1 strongly mimics that of YgjD and YeaZ. This suggests that, comparable to the essentiality of Qri7 homodimerization, YeaZ and Pcc1 could play a similar role through their interaction with YgjD and Kae1, respectively.

The YgjD/YeaZ interface creates an atypical ADP-binding site
As expected we observed clear electron density for a nucleotide bound to two metal ions in the active site of YgjD, which we interpreted as Mg 2+ and Fe 2+/3+ based on electron density levels and X-ray fluorescence scanning at the Feedge (data not shown). Fe 2+/3+ is liganded by the conserved His 111 , His 115 and Asp 300 residues. The ␤-phosphate group of the nucleotide completes the Fe 2+/3+ -coordination sphere ( Figure 2D and Supplementary Figure S2B). The Mg 2+ ion is located 6.2Å away from the Fe 2+/3+ ion and is liganded by the carboxylate side chains of Asp 11 and Glu 12 , and close (3.7Å) to the ␤-phosphate oxygen of the nucleotide. Two water molecules are involved in the coordination of the Mg 2+ ion ( Figure 2D and Supplementary Figure S2B). We did not observe electron density for the ␥ -phosphate group of ATP, which probably has been converted into ADP during the crystallization process. The ADP bound in YgjD is well superposable to the ATP bound in Pa-Kae1. The Fe 3+ ion in Pa-Kae1 is coordinated by a tyrosinate bond (24). This tyrosine is conserved in eukaryotic and archaean Kae1 orthologs but neither in bacteria, nor in the Qri7 mitochondrial ortholog. Interestingly, the Fo-Fc difference map calculated after structural refinement revealed an important residual electron density cloud situated at the YgjD-YeaZ interface ( Figure 2C). An ADP molecule could unequivocally fit the density and its refined B-factors were similar to those of the surrounding side chains. ADP is bound on top of the YeaZ helical bundle at the heterodimer interface. The ADP nucleotide interacts mainly with the N-terminal domain of YeaZ, but there are also a few interactions with YgjD. The ␣-phosphate moiety of this ADP is very close to the best conserved sequence motif 64 GPGS(Y/F)TG(I/L/V)R 72 of YeaZ, reminiscent of a phosphate-binding motif (40). The ␤-phosphate is forming an ionic interaction with YeaZ Arg118 and the ribose 2 -and 3 -OH groups are forming hydrogen bonds with the YeaZ His34 side chain and the main chain carbonyl of YeaZ Arg32 , both these residues are also very well conserved. The adenine base lies on top of a hydrophobic depression created at the interacting interface of YgjD-YeaZ (YeaZ Thr69 and YgjD Leu89 ) but does not form any specific H-bond interactions with the heterodimer. Upon superposing the YgjD-YeaZ heterodimer and the YeaZ homodimer, we noticed that there is a large conformational change in the YeaZ loop connecting ␣3 and ␣1 ( Figure 2B residues Cys 30 -Thr 35 ). While this loop is in an open conformation in the heterodimer YgjD-YeaZ, it adopts a closed conformation in the YeaZ homodimer that overlaps with the interfacial ADP binding site in the heterodimer YgjD-YeaZ (31,41). A similar conformational change was observed for a comparable loop in YeaZ in the St-YgjD-YeaZ heterodimer which did not have any nucleotide bound at the interface (25). Instead, St-YgjD-YeaZ has a tris(hydroxyethyl)aminoethane molecule bound which partially overlaps with the phosphate groups of ADP at the Ec-YgjD-YeaZ dimer interface.

Thermodynamics of nucleotide binding
In order to better understand the role of ATP in the transfer of the threonylcarbamoyl-moiety from TCA onto tRNA, Nucleic Acids Research, 2015, Vol. 43, No. 3 1809 we characterized the nucleotide-binding properties of the YgjD, YeaZ, YjeE and YgjD-YeaZ by ITC. Thermograms (Supplementary Figure S3) and thermodynamic values (Table 2) were all obtained in presence of 500 M Mg 2+ . The titration of YgjD or YgjD-YeaZ heterodimer with ATP or ADP did not produce any heat exchange signals (data not shown for ATP and Supplementary Figure S3G for ADP). There was no heat exchange upon mixing of YeaZ with any of the tested nucleotides (Supplementary Figure S3H and S3I). Hence, the binding of ADP to the YgjD-YeaZ interface could not be detected by ITC experiments. However, AMPCPP bound to YgjD with a Kd of 1.6 M and a H of 2.45 Kcal/mole, and to the YgjD-YeaZ heterodimer with a Kd of 0.7 M and a H of 0.9 Kcal/mole. The heterodimerization with YeaZ does not seem to significantly affect the affinity of YgjD for AMPCPP.
YjeE is structurally related to small GTPases (Supplementary Figure S2C and S2E) and exhibits intrinsically weak ATPase activity (34)(35)(36). However, the titration of YjeE with Guanosine diphosphate (GDP) or ATP in presence of Mg 2+ did not produce any heat exchange signals.  Table 2). This also confirms that YjeE preferentially binds to ADP rather than ATP and it also correlates with the fact that most small G-proteins preferentially bind to GDP compared to GTP (34,42).

The ATPase activity of YjeE is activated and regulated by the YgjD-YeaZ heterodimer
The t 6 A biosynthesis systems require ATPase activities. One ATP is chemically consumed for the synthesis of the TCA intermediate by YrdC/Sua5. A second ATP molecule is hydrolyzed during the threonylcarbamoyl transfer from TCA onto tRNA (13,14). The hydrolysis of the latter ATP molecule is not needed for the chemistry of the transfer reaction and might therefore play a regulatory role (13). To better understand the need for ATPase activity in bacterial biosynthesis of t 6 A, we measured the hydrolysis of ATP into ADP by YgjD, YeaZ, YjeE and YrdC using an NADHcoupled ATPase activity assay. Individual YgjD, YeaZ, YjeE and YrdC do not exhibit ATPase activity whereas YgjD exhibits very weak ATPase activity in presence of YeaZ or YjeE ( Figure 3A). However the ATPase activity of YjeE is strongly activated by the YgjD-YeaZ heterodimer and this effect is maximal for a 1:2 YgjD-YeaZ:YjeE stoichiometry. Kinetic characterization of the ATPase activity of the YgjD-YeaZ-2xYjeE complex against ATP concentration yielded a K m and k cat of 0.644 mM and 0.4 s −1 (Supplementary Figure S4), in comparison with reported values of YjeE alone of 1.4 mM and 0.003 s −1 respectively (34). The ATPase activity is not affected by the addition of either YrdC or tRNA Lys ( Figure 3A). We observed that YjeE exists in an equilibrium between monomer and dimer (Figure 6A and Supplementary Figure S4C). Consistent with this observation the ATPase activity of YjeE is also proportional to the percentage of monomeric YjeE (Supplementary Figure S4C), suggesting monomeric YjeE binds to and is activated by YgjD-YeaZ. Similarly, the monomeric form of the B. subtilis ortholog of YjeE, YdiB, also demonstrated higher ATPase activity than its dimeric form (43). The analysis of the reaction velocity showed a sigmoid dependence against the YjeE concentration. This does not necessarily indicate cooperative behavior and could for instance be due to the YjeE monomer-dimer equilibrium (Supplementary Figure S4D).

The ATP-mediated interaction between YjeE and the heterodimer YgjD-YeaZ
Prompted by the strong activation of the ATPase activity of YjeE by YgjD-YeaZ heterodimer, we further investigated the interaction between YgjD-YeaZ and YjeE by ITC and native gel shift experiments. YgjD or YeaZ alone do not bind YjeE ( Figure 1A and Supplementary Figure S1D-F), but the YgjD-YeaZ heterodimer and YjeE display a clear gel shift when combined ( Figure 4A and Supplementary Figure S6C). Very importantly, in all our binding experiments we observed interaction between YgjD-YeaZ and YjeE only in presence of ATP and Mg 2+ . ITC titration of YgjD-YeaZ with YjeE revealed heat exchange signals, but the ITC curve is complex and could not be interpreted by a simple binding process, probably due to ATP hydrolysis during the titration (Supplementary Figure S5F). We therefore quantified the binding of YjeE to YgjD-YeaZ in presence of non-hydrolysable AMPPNP and Mg 2+ (Supplementary Figure S5A). The resulting binding curve yielded a Kd of 0.60 M, a H of 10.16 kcal/mol and a 1:1:1 stoichiometry for the YgjD-YeaZ-YjeE complex ( Table 1). The 1:1:1 stoichiometry was confirmed by varying the YjeE concentrations in the gel shift experiments ( Figure 4A). We did not observe an interaction between YgjD-YeaZ and YjeE in presence of ADP ( Figure 4B and Supplementary Figure  S5D and E). We conclude that YjeE forms a complex with YgjD-YeaZ only in presence of ATP and that it contacts both YgjD and YeaZ.

Structural model of the ternary complex YgjD-YeaZ-YjeE in solution
We were unable to crystallize the YgjD-YeaZ-YjeE complex in presence of ATP or AMPPNP. We then performed SAXS experiments in order to obtain structural data of the complexes in solution ( Figure 6, Supplementary Figure S8, Table S2). We first analyzed the scattering curves of the individual YgjD, YeaZ, YgjD-YeaZ and YjeE components in presence of ATP and Mg 2+ . Gel-filtration and SAXS data confirmed that the isolated YgjD and YeaZ form homodimers in solution and that YgjD-YeaZ is a heterodimer. As explained in the supplementary data, the Xray scattering curves of YgjD, YeaZ and YgjD-YeaZ can be satisfactorily described from the crystal structures using the SASREF program (Supplementary Figure S8A, S8G and S8C, respectively). The SAXS curve (Supplementary Figure S8B) of the mutant YgjD S97E is compatible with a monomer showing that the mutation disrupts the dimer of YgjD ( Figure 1B). The best fit was obtained by using a model with a slightly widened interface between the Nand C-terminal domains compared to the wild-type structure in complex with YeaZ. We quantified the monomer- The corresponding thermograms are presented in Supplementary Figure S3. a in presence of 500 uM MgCl 2 . dimer distribution of YjeE by gel filtration analysis at various time lapses. YjeE adopts an equilibrium that shifts toward the dimeric form over time and less than 20% forms monomers 18 h after purification ( Figure 6A and Supplementary Figure S4C). By using a HPLC gel filtration step coupled to the SAXS measurements, we were capable of obtaining data on the separated monomeric and dimeric fractions of YjeE. The calculated scattering curve obtained from a homology model of monomeric Ec-YjeE (see Supplementary Materials) is in very good agreement with the scattering curve (Supplementary Figure S2D). In the case of the dimer, an envelope was obtained by using the ab initio program GASBOR (Supplementary Figure S2F and S2G). Two subunits of Ec-YjeE could be fitted into the envelope, but with only the SAXS data we could not propose a precise orientation of the subunits of the dimer. We then used gel filtration coupled to SAXS to follow the formation of the YgjD-YeaZ-YjeE complex in presence of ATP. The gel filtration profile of a mixture of YgjD-YeaZ and YjeE in a 1:4 stoichiometric ratio yielded three peaks that correspond to YgjD-YeaZ-YjeE, homodimeric and monomeric YjeE, respectively ( Figure 6A). We also confirmed by SDSpolyacrylamide gelelectrophoresis the presence of YgjD, YeaZ and YjeE in a 1:1:1 molar ratio in the first peak represented in Figure 6A (data not shown). Analysis of the same protein mixture in presence of ADP showed no evidence for the formation of a stable ternary YgjD-YeaZ-YjeE complex (Supplementary Figure S8D). The X-ray scattering data revealed that YgjD-YeaZ-YjeE has exactly the same maximal extension than YgjD-YeaZ (D max = 90Å, Supplementary Figure S8H) and that the envelope obtained using GASBOR ( Figure 6B) is compact. These results suggest that YjeE binds at the YgjD-YeaZ interface, rather than forming a linear complex. This interpretation of the SAXS curves is compatible with biochemical data showing that only the YgjD-YeaZ heterodimer binds to YjeE and not the respective homodimers. We then attempted to create a structural model compatible with SAXS scattering curves and biochemical data.
Since the presence of ATP is mandatory for the formation of the YgjD-YeaZ-YjeE complex, we reasoned that the ATP-binding site of YjeE might be involved in the interaction with YgjD-YeaZ. On the other hand, we noticed that the ADP bound at the YgjD-YeaZ interface makes contacts with both YeaZ and YgjD ( Figure 2C) and that its orientation is inverted compared with the ADP bound to YjeE ( Figure 6C). We therefore constructed a model of the ternary YgjD-YeaZ-YjeE complex using the SASREF program by requiring ADP of YgjD-YeaZ to be bound to YjeE in the same manner as the ADP to YjeE (PDB ID:1HTW). The calculated SAXS curve of the YgjD-YeaZ-YjeE model is in excellent agreement with the experimental scattering curve ( Figure 6B). In this model, YjeE sits on top of the helical bundle formed by ␣2 and ␣3 from YgjD and ␣1 and ␣2 from YeaZ and the three partners in the complex contribute to the formation of the ATP site. YjeE therefore interacts directly both with YgjD and YeaZ. The regions surrounding the interfacial ADP site are very well conserved suggesting they are important for function ( Figures 2E and 6F). The structure of the YjeE in complex with ATP is not known and there might exist significant structural rearrangements in the switch regions between the ADP and ATP bound forms.

Exploration of structural data by site-directed mutagenesis
We carried out a number of mutations to study the structure function relationships. We summarized the strategy and results in Table 3.

Active site mutant of YgjD
The importance for t 6 A activity of the two conserved histidines (His 111 and His 115 in Ec-YgjD) that coordinate a Feion in Kae1/Qri7 was already investigated (12,13,18). We wanted to analyze the importance of the Mg 2+ ion, by mutating its Glu 12 carboxylate ligand ( Figure 5B) into alanine. The crystal structure of YgjD E12A -YeaZ in complex with ATP determined at 2.3Å confirmed the absence Mg 2+ in the active site. Interestingly, this mutant has ATP bound at the active site compared to ADP for the wild YgjD-YeaZ type although no binding of AMPCPP to YgjD E12A was observed by ITC (Supplementary Figure S3J). YgjD probably has a very weak ATPase activity, which is inhibited in the YgjD E12A mutant, explaining why we observed ATP at the active site of the mutant. Moreover, there was no electron density observed for ADP at the YgjD E12A -YeaZ interface, suggesting this site could have a preference for ADP rather than ATP. We then investigated the effect of this mutant on the in vitro biosynthesis of tRNA t 6 A. Overexpressed tRNA Lys was first incubated with YgjD, YeaZ, YjeE and YrdC in presence of ATP, Mg 2+ , L-threonine and bicarbonate. At the end of the reaction, tRNA was enzymatically digested, the nucleosides were separated by C18 reverse phase HPLC and analyzed by mass spectrometry. As seen in Supplementary Figure S7, the presence of the modified adenine could clearly be demonstrated using wild-type proteins (Supplementary Figure S7C). We observed that the in vitro t 6 A activity of YgjD E12A is totally abolished (Figure 5 and Supplementary Figure S7D). However the AT-Pase activity of the ternary complex YgjD E12A -YeaZ-YjeE is not affected by this mutation ( Figure 3B). These data further confirm that YgjD is responsible for the t 6 A coupling reaction and YgjE for ATP hydrolysis.

The interfacial ADP binding site
We wanted to test the pertinence of the ADP bound at the YgjD-YeaZ interface ( Figure 2C) using two mutants, YeaZ R118A and YgjD V85E . The crystal structure of YgjD V85E -YeaZ in the presence of ATP showed no ADPlike electron density at the dimer interface. Nonetheless, YgjD-YeaZ R118A still interacts with YjeE in presence of AMPPNP, as determined by ITC (Table 1 and Supplementary Figure S5C). The ATPase activity of the ternary YgjD V85E -YeaZ-YjeE and YgjD-YeaZ R118A -YjeE complexes remained unaffected ( Figure 3B) and the latter is fully active for the in vitro biosynthesis of t 6 A ( Figure 5A).

The YgjD-YeaZ interface
In order to estimate the importance of heterodimer formation, we generated mutations at the YgjD-YeaZ interface. Ser 97 and Phe 100 of YgjD are both involved in packing of the helices and are surrounded by hydrophobic residues (Figure 6C). ITC, gel filtration and native gel data showed that YgjD S97E , YgjD S97R and YgjD F100E , are no longer capable of forming homodimers nor heterodimers with YeaZ (Table 1 and Supplementary Figure S1C for YgjD S97E by ITC and Supplementary Figure S6A and S6C for YgjD S97E , YgjD S97R and YgjD F100E by gel filtration and native gel shift, respectively), suggesting that the same regions of YgjD are involved both for homo-and heterodimer formation. We demonstrated by ITC and native gel shift that a mixture of YgjD S97R and YeaZ is not able to associate with YjeE in presence of AMPPNP or ATP (Supplementary Figures S5I and S6C). None of these variants are capable of activating the ATPase activity upon mixing with YeaZ and YjeE ( Figure 3B), suggesting that the formation of the YgjD-YeaZ heterodimer is mandatory for the AT-Pase activation of YjeE. However, the in vitro t 6 A activity of YgjD S97E decreased by only ∼40% (Figure 5A and Supplementary Figure S7E).

YjeE ATP binding site mutations
YjeE displays structural features typical for GTPase proteins (34,35), characterized by the presence of switch regions that undergo conformational changes upon nucleotide hydrolysis, triggering signal transduction to other proteins or nucleic acids (42). Two such switch regions were identified in the structure of YjeE: switch I is involved in coordination of the Mg 2+ ion and switch II is in proximity to the ␤and ␥ -phosphates of bound nucleotide (Supplementary Figure S2E). To further explore the role of YjeE we designed two types of mutations: those with the goal of affecting nucleotide binding and ATPase activity (residues Thr 43 and Glu 108 ) and mutations of residues in the putative switch regions (Trp 109 and Try 82 ). Tyr 82 from switch II is in a connection between ␤3 and ␤4 strands and should be close to the ␥ -phosphate of ATP whereas Trp 109 from switch I is next to the Mg 2+ -binding residue Glu 108 and stacks against Tyr 82 (Supplementary Figure S2E). The ATP␥ S-or ADPbinding properties of these mutants were quantified by ITC. While YjeE Y82A and YjeE W109A still bound to ATP␥ S and ADP with a Kd of 8.90 and 4.83 M, respectively (Table 2 and Supplementary Figure S3E and F), the binding of ADP  to both YjeE T43A and YjeE E108A was abrogated (Supplementary Figure S3K and S3L). The ATPase activity of the YjeE Y82A has dropped to 40% and that of the other mutants fell below 10% compared to the wild-type YjeE ( Figure 3B). These data strengthen the hypothesis that YjeE is the bona fide ATPase activated by the YgjD-YeaZ heterodimer.
We further tested whether these mutants are capable of forming a complex with YgjD-YeaZ. Native gel shift experiments ( Figure 4A) showed that the YjeE T43A and YjeE E108A mutants no longer interact with YgjD-YeaZ while the binding capacity of the switch mutants YjeE Y82A and YjeE W109A remained unaffected. The capacity of YjeE to bind ATP therefore seems to be correlated with its bind-ing to YgjD-YeaZ. ITC measurements further showed that the binding capacity of YjeE W109A to YgjD-YeaZ is comparable to that of wild-type YjeE in presence AMPPNP (Table 1 and Supplementary Figure S5B). The X-ray scattering curve of the YgjD-YeaZ-YjeE W109A mixture is identical to that of the wild-type complex, suggesting that its overall structure is not affected (Supplementary Figure S8F).
We then wanted to find out if the ATPase activity of YjeE is essential for the in vitro biosynthesis of tRNA t 6 A. Interestingly, although the YjeE E108A and YjeE W109A mutants no longer exhibit ATPase activity ( Figure 3B), the in vitro biosynthesis of t 6 A by YjeE W109A remained unaffected while it dropped by 60% for YjeE E108A (Figure 5A   plementary Figure S7). We conclude that for the in vitro biosynthesis of tRNA t 6 A the ATPase activity of the ternary complex YgjD-YeaZ-YjeE seems dispensible.

Deletion of the C-terminal region of YeaZ
As the C-terminus of YeaZ (220-231) is not observed in the crystal structure of the YeaZ homodimer but is wellpositioned by interaction with YgjD in the crystal structure of the YgjD-YeaZ heterodimer (Figure 2A), truncated YeaZ 1-219 was produced to test the contribution of the Cterminal YeaZ 220-231 in formation of heterodimer YgjD-YeaZ. Binding was followed by ITC and native gel shift and it was demonstrated that YeaZ 1-219 still interacts with YgjD with slightly higher affinity (Kd of 34 nM compared to 122 nM for intact YeaZ) (Supplementary Figure S6B, Table 1 and Supplementary Figure S1B). The X-ray scattering curve for the YgjD-YeaZ 1-219 -YjeE complex was identical to that of YgjD-YeaZ-YjeE (Supplementary Figure S8E), confirming the C-terminal tail of YeaZ 220-231 is not essential for the formation of the ternary complex.

The YgjD-YeaZ heterodimer is essential for activity
The reaction scheme of the biosynthesis of tRNA t 6 A in bacteria is composed of two main steps. In the first half of the reaction, YrdC utilizes ATP, L-threonine and bicarbonate to synthesize an unstable intermediate TCA, whose threonylcarbamoyl-moiety is subsequently transferred onto the adenosine at position 37 of the substrate tRNAs by the cooperative action of three proteins YeaZ, YgjD and YjeE. The yeast mitochondrial ortholog of YgjD, Qri7, is capable of transferring the threonylcarbamoyl-moiety of TCA onto tRNA without assistance of other protein partners (12). This provides strong evidence that YgjD is responsible for the coupling reaction while YeaZ and YjeE assist its action (11,12,14). How exactly YeaZ and YjeE contribute to the tRNA t 6 A modification in vivo remains unknown. We showed here that E. coli YgjD forms a stable heterodimer with YeaZ, and that this dimer forms a complex with YjeE in presence of ATP, confirming previous studies on the Salmonella thyphimurium orthologs (24). The YgjD-YeaZ interface is very similar to that of the YeaZ and Qri7 homodimers (12,41). Mutations disrupting the YgjD-YeaZ heterodimer abrogated interaction with YjeE and are largely inactive in the in vitro biosynthesis of tRNA t 6 A. A similar dependence of the t 6 A biosynthesis upon homodimerization was observed for Qri7 (12). Interestingly we observed that YgjD alone also forms homodimers but these are unable to bind YjeE and are inactive. The YgjD S97E , YgjD S97R and YgjD F100E mutants neither form homodimers nor heterodimers with YeaZ, suggesting the YgjD homodimer is structurally similar to those of Qri7 and YeaZ. Despite the similarities between the interaction modes of YgjD, YeaZ and Qri7, the complementation of E. coli ygjD requires both the ygjD and yeaZ orthologs from B. subtillus, showing a species-specific recognition between YgjD by YeaZ (10). The coupling of TCA to A 37 of cognate tRNA in archaea and yeast is carried out by the KEOPS complex that consists of a linear arrangement of the Kae1, Bud32, Cgi121 and Pcc1 proteins (16,21). Pcc1 is required for the in vitro biosynthesis of tRNA t 6 A in both archaea and yeast cytoplasm (12,13,18). In the structural model of archaean KEOPS, Pcc1 binds to the N-terminal lobe of Kae1 whereas Bud32 binds to the C-terminal lobe. Pcc1, although unrelated to YeaZ, forms a complex with Kae1 whose interface is structurally mimicking that of YgjD-YeaZ ( Figure 2H). The binding of a protein partner (Qri7, Pcc1 or YeaZ) to the N-terminal lobe of Qri7/Kae1/YgjD seems to be essential for t 6 A activity and is likely involved in the binding of the tRNA substrate. We were not able to detect the binding of tRNA to the YgjD-YeaZ-YjeE complex nor to any of the individual proteins by electrophoretic mobility shift assay (data not shown), meaning that the tRNA binding is too weak to be detected and occurs only transiently.
Nucleic Acids Research, 2015, Vol. 43, No. 3 1815 The ATPase YjeE is strongly activated by YgjD-YeaZ The role of the essential YjeE protein in the transfer of threonylcarbamoyl-moiety from TCA onto A 37 of tRNA in bacteria remains unknown (11). Previous two hybrid data showed that YgjD, YeaZ and YjeE form a network and that the interaction between YeaZ and YjeE was mutually exclusive with the formation of the YgjD-YeaZ complex. S. typhimurium YjeE was reported to bind both to YgjD-YeaZ and to YgjD in presence of ADP or ATP (25). Our experiments however demonstrate that only the YgjD-YeaZ heterodimer strongly interacts with YjeE and that this interaction is dependent on the presence of ATP but not ADP. The resemblance Hi-YjeE with small GTPases suggested that YjeE might work as a molecular switch triggered by ATP hydrolysis (36). We showed that Ec-YjeE binds more strongly to ADP than to ATP␥ S or AMPPNP and that its intrinsic weak ATPase activity is strongly activated by binding to YgjD-YeaZ (14,34). Mutants of YgjD (YgjD S97E , YgjD S97R or YgjD F100E ) that are unable to form heterodimers with YeaZ are incapable of activating the AT-Pase activity of YjeE. The YjeE W109A (in switch I) and YjeE Y82A (in switch II) mutations did not affect ATP or YgjD-YeaZ binding, but they lost their ATPase activity by 100 and 65% respectively. Interestingly, the in vitro t 6 A biosynthesis of the YjeE W109A mutant remained intact. The YjeE T43A and YjeE E108A mutations abrogated ADP-or ATP-binding and hence they were not able to interact with YgjD-YeaZ. The activity of YjeE E108A in biosynthesis of tRNA t 6 A is reduced by more than 60%. The binding capacity of YjeE to YgjD-YeaZ rather than its ATPase activity seems to be primordial for tRNA t 6 A modification. This confirms the observation that the B. subtilis YgjD, YeaZ and YjeE orthologs function in biosynthesis of tRNA t 6 A in absence of ATP (11). However the ATPase activity of YjeE is necessary for cell growth since it was observed that the AT-Pase inactive K41A mutant of the B. subtilis ortholog YdiB had the same negative effect on cell growth than the knockout mutant (43). This suggests that YjeE may be involved in other important cellular processes.

ATP-mediated formation of YgjD-YeaZ-YjeE
Our biochemical and structural data suggest that YjeE contacts both YgjD and YeaZ to form a compact heterotrimeric complex. The YgjD-YeaZ interface contains a well-conserved surface patch that is centered on a bound ADP molecule. We used this ADP moiety as an anchor for docking the YjeE protein onto the YgjD-YeaZ interface. The resulting model of the ternary complex brings into light many interesting features. First, the interfacial ADPbinding site is composed of residues from the three proteins and may explain theATPase activating effect of YjeE upon interaction with YgjD-YeaZ. Second, YjeE interacts both with YeaZ and YgjD, as expected for its binding specificity for heterodimeric YgjD-YeaZ. Third, the active site of YjeE binds to well-conserved surface patches of the YgjD-YeaZ complex. Fourth, in our model of the ternary complex compatible with the SAXS data, YjeE and YgjD form a crater large enough to accommodate the substrate tRNA. The catalytic metal site of YgjD is at the bottom of this crater. YjeE is therefore well positioned in the ternary complex to be involved in substrate tRNA binding. The rigid superposition that was used for the construction of the model of the ternary complex created some small steric clashes between Ala 130 and Gln 131 from YjeE and Pro 327 and Arg 328 of YgjD, and between Pro 61 -Thr 64 of YjeE and Gly 120 and Ala 169 of YeaZ, all of which reside in potentially flexible loops. It should be noticed that the YgjD-YeaZ-YjeE complex forms in the presence of ATP but that the complex was modeled using the structure of the YjeE-ADP complex. Small G proteins usually adopt rather different conformations between the GDP and GTP bound forms. It seems therefore plausible that the ATP bound form of YjeE may undergo conformational changes in the switch regions that might relieve the present steric clashes in our model. The mutations of two residues that contact this ADP molecule (YeaZ R118A and YgjD V85E ) did not confer any effect on the heterodimerization of YgjD and YeaZ, nor on the activation of the ATPase activity of YjeE. The intact t 6 A biosynthesis but corrupted ATPase activities of the YjeE W109A mutant showed that formation of the ternary YgjD-YeaZ-YjeE complex is primordial.

CONCLUSION
All organisms use Sua5/YrdC for the biosynthesis of the intermediate TCA, but the bacteria, eukaryotes/archaea and eukaryotic mitochondria developed different strategies for the condensation of threonylcarbamoyl-moiety from TCA onto adenosine at position 37 of tRNA. Although these strategies involve different proteins, they have several points in common. All the tRNA t 6 A modification systems use a closely related threonylcarbamoyl-transferase, Qri7/Kae1/YgjD, sharing a common structure and active site configuration, centered on a very conserved metal cluster. These enzymes all need support from a protein subunit bound to the N-terminal lobe. Although, the structures of these subunits are different for Qri7/Kae1/YgjD, the interfaces between the subunits are almost identical. The most plausible hypothesis about the function of noncatalytic subunits is their involvement in tRNA recognition, but few data are available on tRNA interaction for any of the t 6 A biosynthesis systems. Kae1 and YgjD further need Bud32/Cgi121/Pcc1/(Gon7) and YeaZ/YjeE to carry out their activity in biosynthesis of t 6 A, respectively. Although Bud32 is annotated as a protein kinase, structural and functional studies and comparison with its paralog RIO kinase led to the hypothesis that Bud32 may act as an ATPase within the KEOPS complex rather than as a protein kinase. RIO kinase plays an essential role in pre-40S ribosomal subunit maturation and its ATP hydrolysis triggers 40S subunit biogenesis. YjeE and Bud32 probably play regulatory roles in the biosynthesis of tRNA t 6 A through their interactions with and/or hydrolysis of ATP. YjeE and Bud32 are unrelated proteins and they interact very differently with YgjD and Kae1, respectively. Further studies will have to establish the functional relationships of these proteins within the t 6 A biosynthesis system. The crystal structure of the ternary complex YgjD-YeaZ-YjeE in complex with tRNA will also shed light on more complete mechanistic understanding of the transfer of threonylcarbamoyl-moiety from TCA onto A 37 of tRNA.