Реконструкция петли антикодона и форма D-стима приводят к конкретному распознаванию ANN-декодирования tRNA для модификации t6A.

Introduction

N6-threonylcarbamoyladenosine (t 6 A) is a modified nucleoside found at position 37 in the anticodon loop of transfer RNAs (tRNAs) decoding ANN codons (N being A, U, G or C) (Fig. 1A ) [ 1 , 2 ], and it is one of a core set of 18 universal post-transcriptional tRNA modifications found in all domains of life [ 3 – 5 ]. t 6 A37 is required for accurate decoding by shaping the architecture of the anticodon loop [ 6 , 7 ]. It promotes proper selection of start codons [ 8 ], stabilizes codon-anticodon pairing at the ribosomal decoding site to prevent +1 frameshifting, and is required for efficient aminoacylation of specific cognate tRNAs [ 9 – 12 ]. Deficiency of t 6 A leads to amino acid misincorporation [ 13 , 14 ], translational deregulation [ 15 – 18 ], and growth and developmental defects (reviewed in [ 1 ]). In humans, deficiency is linked to Galloway–Mowat syndrome [ 19 ], while in most bacteria, including pathogens, the modification is essential [ 9 , 20 ].

t 6 A37 has been regarded as one of the “molecular fossils” of the origins of life and evolution [ 21 , 22 ]. Its universal presence in all three domains of life, together with its essential role in stabilizing A/U-rich codon–anticodon interactions, suggests that t 6 A arose very early in parallel with the emergence of the genetic code and that its biosynthetic pathway was already established in the last universal common ancestor [ 23 , 24 ]. Consistent with this view, a recent report has shown that in giant viruses, threonylcarbamoyl modification can occur on ATP/ADP in a tRNA-independent manner, extending this chemistry beyond tRNA and providing insight into the evolution of the pathway [ 25 ]. t 6 A biosynthesis occurs in two steps (Fig. 1B ). Members of the TsaC/Sua5 family (TsaC2 in T. maritima ), with orthologs present in all three domains of life [ 26 , 27 ], synthesize the intermediate threonylcarbamoyl adenylate (TC-AMP) from L-threonine, bicarbonate/CO 2 , and ATP [ 28 ]. The threonylcarbamoyl (TC) moiety of TC-AMP is then transferred to A37 of the substrate tRNA by a transferase embedded in a larger protein complex that varies between bacteria and archaea/eukarya. In bacteria, it is the threonylcarbamoyl transfer (TCT) complex comprised of TsaB, TsaD, and TsaE [ 28 – 30 ]; in archaea [ 31 ] and the cytosols of eukaryotes [ 32 , 33 ] it is the KEOPS complex (Pcc1, Kae1, Bud32, and Cgi121); and in mitochondria it is the TsaD ortholog Qri7 [ 34 ]. The TsaD/Kae1/Qri7 protein family harbors the active site for the transfer reaction. In bacteria, TsaB and TsaD associate to form a TsaB/TsaD hetero complex with a 1:1 stoichiometry (e.g. Salmonella typhimurium and E. coli systems [ 35 , 36 ]), or a 2:2 stoichiometry where TsaB serves as the dimerization module (e.g. T. maritima (Fig. 1B ) [ 29 , 37 ] and Aquifex aeolicus [ 38 ]). This complex is the minimal tRNA binding platform, accommodates one tRNA molecule, and carries out the TC transfer reaction [ 29 , 37 ]. In the absence of tRNA, TsaE binds to this minimal complex in an ATP-dependent manner to form a 1:1:1 complex (e.g. S. typhimurium and E. coli [ 35 , 36 ]) or a 2:2:2 complex (e.g. T. maritima [ 29 , 37 ] and A. aeolicus [ 38 ]). Using competition assays and SAXS analysis of the tRNA-bound complex, we previously showed that TsaE and tRNA bind to overlapping sites on Tm TsaB 2 D 2 (and on E. coli TsaBD) [ 29 , 37 ], and that Tm TsaB 2 D 2 binds one tRNA molecule at a time [ 37 ]. TsaE-catalyzed ATP hydrolysis is required for multiturnover t 6 A synthesis in the T. maritima t 6 A system [ 29 , 37 ] by a still poorly understood process in which the TCT complex is apparently reset/re-activated at the end of each cycle.

The structural and sequence determinants in tRNA required for t 6 A modification have been investigated both in vivo and in vitro . Only full-length tRNAs are competent substrates, and the modification does not depend strictly on the presence of other modifications [ 2 , 12 , 30 , 39 ]. The anticodon-loop sequence 36-UAA-38 is a strict and universal determinant for modification [ 10 – 12 , 39 – 41 ]. The bases at positions 32 and 33 also serve as determinants but appear to vary among species. For example, C32 is required for modification by KEOPS from human, yeast, and C. elegans [ 12 ], but is dispensable for the human Qri7 system [ 41 ]. Likewise, U33 is essential for modification by yeast KEOPS [ 12 ] but not by human Qri7 [ 41 ]. Beyond the anticodon loop, eukaryotic and archaeal KEOPS have been shown to depend on specific base pairs in the D-stem [ 12 , 42 , 43 ]. The 3′CCA-end of tRNA, however, is not universally required; it is essential for M. jannaschii and human KEOPS but dispensable for yeast, Drosophila , and A. thaliana KEOPS [ 12 , 40 , 44 ].

Despite extensive biochemical evidence defining sequence determinants, their molecular basis has not been elucidated. In particular, how the 36-UAA-38 anticodon loop motif is recognized has remained unknown, and recent cryo-EM structures of tRNA-bound KEOPS either lacked a resolved anticodon loop or captured the tRNA outside the active site [ 42 , 43 ]. Thus, the structural basis for the strict specificity of these phylogenetically diverse enzymes to NNU-tRNAs has remained unclear. Here, we report 3.2 Å cryo-EM structures of T. maritima TsaB 2 D 2 bound productively to tRNA, providing the first complete structural view of how the enzyme engages the anticodon loop and explaining the universal specificity of t 6 A for ANN-decoding tRNAs.

Materials and methods

Preparation of Tm TsaB 2 D 2 for cryo-EM

Recombinant wild-type T. maritima TsaB and TsaD proteins ( Tm TsaB and Tm TsaD, respectively) were overexpressed in E. coli Overexpress™ C41 (DE3) cells and purified by Ni-NTA chromatography, followed by size exclusion chromatography (SEC), as described previously [ 37 ]. Protein purity was verified by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) analysis ( Supplementary Fig. S1A ). The Tm TsaB 2 D 2 complex was prepared by mixing 26 μM each Tm TsaB and Tm TsaD in buffer containing 50 mM Tris–HCl (pH 7.5), 50 mM KCl, and 1 mM dithiothreitol (DTT), followed by SEC purification of the complex on an XK 16/70 column packed with 130 ml of Superdex 200 prep-grade resin (Cytiva) and equilibrated in the same buffer. The complex eluted as a single, symmetric peak, and SDS–PAGE analysis of peak fractions showed the expected relative amounts of Tm TsaB and Tm TsaD ( Supplementary Fig. S1B ). The final complex was concentrated to 3.5 mg/ml using a centrifugal filtration device, flash-cooled in liquid nitrogen, and stored at −80°C until use.

Preparation of tRNA for cryo-EM

Escherichia coli tRNA Thr CGU , cloned in pEXT20 plasmid, was expressed in E. coli MRE600 cells as described previously [ 27 ]. Cells from a 1.0 l culture were harvested and suspended in 40 ml lysis buffer [0.5 M sodium acetate, pH 5.2, 10 mM ethylenediaminetetraacetic acid (EDTA)], followed by extraction with an equal volume of acid phenol-chloroform (pH 4.5) on a rocker for 2 h at room temperature. The mixture was centrifuged at 33 000 × g for 30 min, and the aqueous phase was extracted twice more with phenol-chloroform (15 min each), followed by a final wash with chloroform and centrifugation. The tRNA-containing aqueous layer was collected and subjected to anion-exchange chromatography as described by Spears et al . [ 45 ]. Briefly, the aqueous layer was pH-adjusted by adding MOPS (pH 7.0) to a final concentration of 0.1 M and applied to a Qiagen-tip 2500 anion-exchange column pre-equilibrated with 50 mM MOPS (pH 7.0), 15% isopropanol, and 1% Triton X-100. The column was washed with buffer containing 50 mM MOPS (pH 7.0) and 0.2 M NaCl, and the RNA was eluted with 30 ml of 50 mM MOPS (pH 7.0), 0.75 M NaCl, and 15% ethanol. The pooled RNA was further purified by reversed-phase chromatography using an XBridge Protein BEH C4 column (300 Å, 5 µm, 10 × 250 mm) connected to a Bio-Rad NGC Quest 10 Plus system. RNA was resuspended in buffer A (20 mM sodium acetate, 20 mM ammonium acetate, 380 mM NaCl, 10 mM MgCl 2 , pH 5.2) to a final concentration of 40 mg/ml, and incubated at 75°C for 10 min. A 25 µl aliquot was injected onto the column and eluted at 2 ml/min using the linear gradient: 0–5 min, 0% buffer B; 5–66 min, 0%–11% buffer B; 66–86 min, 100% buffer B (buffer B: buffer A supplemented with 50% methanol). The two fractions comprising the major tRNA peak were checked on a 10% denaturing polyacrylamide gel ( Supplementary Fig. S1C ), then pooled and precipitated with 70% ethanol at −20°C, and air dried. The purified tRNA was folded by suspending in 7 mM MgCl 2 and 1× RNAsecure TM (Thermo Fisher Scientific), heating to 75°C for 10 min in a thermocycler, and gradually cooling to 12°C over 75 min. Final tRNA quality was assessed by microfluidic capillary electrophoresis using an Agilent 2100 Bioanalyzer (Small RNA Kit), which revealed a single sharp peak within the expected tRNA size range (60–80 nt) and no detectable smaller or larger RNA species, indicating high sample integrity ( Supplementary Fig. S1D ).

Synthetic RNA carrying the sequence of T. maritima tRNA Lys CUU ( Supplementary Fig. S1E ) was purchased from IDT, dissolved in water, precipitated in 70% ethanol, resuspended in 1× RNAsecure TM , and folded as described for Ec tRNA Thr CGU . Binding of these tRNAs to Tm TsaB 2 D 2 was assessed by electrophoretic mobility shift assays (EMSA), and the apparent affinities are K d = 0.2 and 0.16 μM for in vivo -produced Ec tRNA Thr CGU and synthetic Tm tRNA Lys CUU , respectively ( Supplementary Fig. S1F and G ). Surprisingly, in t 6 A synthesis assays the in vivo -produced Ec tRNA Thr CGU exhibited ∼30% of the activity observed with the Ec tRNA Thr CGU in vitro transcript (prepared as described below), consistent with the presence of a fraction of unmodified species ( Supplementary Fig. S1H ).

Validation of tRNA identity and modification status of purified E. coli tRNA Thr CGU by chromatography-coupled mass spectrometry

To confirm the identity of the in vivo- expressed E. coli tRNA Thr CGU and map its modifications, liquid-chromatography tandem mass spectrometry (LC-MS/MS) was performed. The RNA (200 ng) was digested in a 10 µl reaction containing 10 mM ammonium formate (pH 7), 1 mM EDTA, and 200 U RNase T1 (a G-specific RNase), and incubated at 37°C for 10 min, followed by 50°C for 2 min, and then 37°C for 5 min. The sample was dried under SpeedVac, reconstituted in 1 µl butylamine-acetic acid buffer (pH 8.0), and analyzed by reversed-phase ion-pair liquid chromatography on an Agilent 1200 Nanoflow LC system. Separation was performed on a homemade 0.15 mm × 150 mm C18 column using 10 mM butylamine-acetic acid buffer in water (pH 8) as eluent A, and methanol as eluent B, with a 15%–40% B gradient at a flow rate of 1 µl/min. Data-dependent MS/MS (ddMS 2 ) characterization was performed on an Agilent 6550 Q-TOF system equipped with an Agilent nano spray source and operated in negative ion mode. MS 1 spectra were collected over m/z 250–2000 at a scan rate of 2 spectra/s. The three most intense species in the first quadrupole were selected for fragmentation using a formulated collision engine (slope 1.4, offset 14). MS 2 spectra were acquired at 2 spectra/s with active exclusion after three spectra from the same MS 1 . MS 1 signals in the 250–400 m/z range were excluded from MS 2 analysis. Data were processed using ProteoWizard msConvert [ 46 ], and analyzed with NucleicAcidSearchEngine within OpenMS [ 46 – 48 ] against E. coli K-12 MG1655 mature tRNA sequences downloaded from GtRNAdb [ 49 ] ( https://gtrnadb.org ) as the reference library. The full list of fragment ion types (a-B, a, b, c, d, w, x, y, z) was considered for peak matching. Precursor and fragment mass tolerance were both set to 10 ppm. Apart from common fragments such as the 3′-terminal “CACCA,” which are shared by tRNA Thr CGU and many other tRNAs, several fragments matched to sequences shared by tRNA Thr CGU 1–1 and tRNA Thr CGU 2–1 (GtRNAdb nomenclature), and one fragment matched to a region shared by tRNA Thr CGU 1–1 and tRNA Thr GGU 1–1 ( Supplementary Table S1 ). No detected fragments mapped uniquely to tRNAs other than tRNA Thr CGU 1–1. Six distinct modifications were assigned to six positions on this tRNA: s 4 U at position 8, D at positions 16 and 17, t 6 A/ct 6 A at position 37, m 7 G at position 46, and m 5 U at position 54 ( Supplementary Fig. S2A ). Dihydrouridine at position 20, which is known to occur in this tRNA [ 50 ], was not detected. Since pseudouridine cannot be differentiated from uridine by mass spectrometry, the presence of pseudouridine at positions 39 and 55 (Ψ39 and Ψ55) was not confirmed. The annotated MS 2 spectra of “U[t 6 A]AUG” and “U[ct 6 A]AUG,” together with the annotated peak of the free t 6 A and ct 6 A bases, confirm the presence of both modifications at position 37 ( Supplementary Fig. S2B and C ). Altogether, these data confirm the identity of the enriched species as E. coli tRNA Thr CGU 1–1, with t 6 A and ct 6 A detected at position 37. Other tRNA species, if present, are below the detection limit of this assay.

Complex assembly and cryo-EM grid preparation

Complexes of Tm TsaB 2 D 2 with Tm tRNA Lys CUU or with Ec tRNA Thr CGU were prepared following the same protocol. Folded tRNA was mixed with pre-formed and SEC-purified Tm TsaB 2 D 2 in equimolar amounts to a final complex concentration of 2 µM (∼10-fold above the apparent K d , as measured by EMSA, Supplementary Fig. S1D and E ) in buffer containing 50 mM Tris (pH 7.5), 50 mM KCl, 5 mM MgCl 2 , and 1 mM DTT, and incubated on ice for 1 h. Cryo-EM grids were prepared using a Leica EM GP2 automated plunge freezer. A 3-µl aliquot of the complex was applied to a glow-discharged 300-mesh gold flat grid (Electron Microscopy Sciences Au-flat 1.2 µm/1.3 µm) treated at 15 mA for 45 s. Sample application and blotting were performed in a chamber maintained at 21°C at 99% relative humidity. Grids were blotted for 5 s and immediately plunged into liquid ethane.

Cryo-EM data collection and processing

For each structure, cryo-EM data were collected from a single grid on a 300 kV Titan Krios transmission electron microscope equipped with a Gatan K3 direct electron detector. Images were acquired at 105 000× nominal magnification in super-resolution mode, a physical pixel size of 0.8266 Å (super-resolution pixel size is 0.4133 Å), and defocus range −0.8 to −2.2 μm. Movies were collected at a 37° stage tilt using SerialEM v4.1.2, at a dose rate of 18 e − /pixel/s, with a total electron dose of 50 e − /Å 2 applied over 50 frames. Data processing was performed using cryoSPARC v4.5.3 [ 51 ]. Raw movies were motion corrected using Patch Motion Correction and binned two-fold to the physical pixel size 0.8266 Å for all subsequent processing. Following Patch CTF Estimation and junk detection, micrographs were curated manually.

Processing of the TmTsaB 2 D 2 /EctRNA Thr CGU dataset . Particles were initially picked on a subset of 1200 micrographs using the automated Blob Picker with a Gaussian template (35–160 Å diameter). 2D classification, ab initio 3D reconstruction, and heterogeneous refinement into four 3D classes identified one class (73 K particles) with clear features of the ribonucleoprotein complex. This subset was subjected to homogeneous refinement and non-uniform refinement, yielding an initial consensus map at 4.4 Å resolution. 2D templates generated from this map were used for template-based particle picking across the full dataset. The resulting particles were classified by two rounds of 2D classification, and selected particles generated five ab initio volumes, which served as references for heterogeneous refinement. Particles corresponding to classes representing the ribonucleoprotein complex were pooled and further curated by 3D classification to improve compositional and conformational homogeneity. Per-particle defocus and global aberrations (beam tilt, trefoil, and anisotropic magnification) were refined, followed by reference-based motion correction. Particles were then refined by non-uniform refinement and masked local refinement using a mask encompassing TsaB 2 /TsaD/tRNA. The final 3D reconstruction reached a resolution of 3.2 Å (gold-standard FSC 0.143), with a 3D-FSC sphericity of 0.947, indicating near-isotropic resolution.

Processing of the TmTsaB 2 D 2 /TmtRNA Lys CUU dataset . Particles were initially picked with the automated Blob Picker with a Gaussian template (35–175 Å) and were directly used for ab initio 3D reconstruction into six 3D volumes, followed by heterogeneous refinement. One 3D volume (∼535 K particles) exhibited features of the ribonucleoprotein complex. Particles contributing to this volume were further curated by 2D classification, iterated with ab initio reconstruction and heterogeneous refinement into 3–5 classes, resulting in one good class (62 K particles) for subsequent homogeneous and non-uniform refinement to generate a 4.3 Å consensus map. 2D templates generated from this map were used for template-based picking. Following the same cleanup procedure described earlier, ∼51 K template-picked particle images were curated. Particles from blob picking and template picking were combined and duplicates removed. The particle stack was further cleaned by 3D classification to remove low-quality images. Following local and global CTF refinement, and reference-based motion correction, particles were refined by non-uniform refinement, then masked local refinement using a mask encompassing TsaB 2 /TsaD/tRNA. The final 3D reconstruction reached a resolution of 3.2 Å (gold-standard FSC 0.143) with a 3D-FSC sphericity of 0.95, indicating near-isotropic resolution. Local resolution maps were calculated using CryoSPARC’s Local Resolution Estimation tool, and 3D-FSC was calculated using the 3dfsc.salk.edu server [ 52 ]. All data collection and processing details are included in Supplementary Table S2 . The final reconstructions and raw data were uploaded to the Electron Microscopy Data Bank (EMDB) and EMPIAR [ 53 ].

Atomic model building and refinement

The maps were locally sharpened using DeepEMhancer [ 54 ] prior to model building. The starting model for Ec tRNA Thr CGU was extracted from its published crystal structure bound to threonyl-tRNA synthetase (PDB ID 1qf6) [ 55 ]. The starting model for Tm tRNA Lys CUU was generated in AlphaFold 3 [ 56 ]. The previously reported crystal structure of T. maritima TsaB 2 D 2 E 2 (PDB ID 6n9a) [ 29 ] served as a starting model for the protein, with the TsaE subunits removed. For each structure, the initial models were rigid-body docked into the cryo-EM map using phenix.dock_in_map [ 57 ] and manually adjusted in Coot [ 58 ]. In addition to global adjustments, the tRNA anticodon loop, the lever arm of TsaD, and the C-terminal tail of TsaB were remodeled to fit the experimental density. For the modified Ec tRNA Thr CGU , since the final map does not provide unambiguous density for most base modifications, the tRNA was modeled using canonical nucleotides except at position A37, where additional density extending from the adenine N6 atom consistent with a threonylcarbamoyl substituent was visible; hence, t 6 A37 was modeled in the conformation most compatible with the density. Likewise, the CCA end and D-loop nucleotides D16 and D17 lacked density and were not modeled. The final models were refined using phenix.real_space_refine [ 59 ] with secondary-structure restraints applied and validated using PHENIX validation tools and the wwPDB online validation server [ 60 ].

Preparation of mutant proteins

Point mutagenesis was carried out on Tm TsaB and Tm TsaD using the Q5 ® Site-Directed Mutagenesis Kit (New England Biolabs Inc.) and the primers listed in Supplementary Table S3 . All mutations were verified by sequencing. The mutant proteins were overexpressed in E. coli and purified as described for the wild-type protein, and their purity verified by SDS–PAGE analysis ( Supplementary Fig. S1A ).

Preparation of E. coli tRNA Thr CGU in vitro transcript for t 6 A assays

DNA templates for in vitro transcription of the WT and mutant sequences of Ec tRNA Thr CGU were produced as described previously [ 61 ]. Transcription reactions were incubated at 37°C for 5 h in 100 mM HEPES (pH 7.5), 2.0 mM spermidine, 25 mM MgCl 2 , 40 mM DTT, 0.01% Triton X-100, 5.0 mM NTPs, 0.15 pM DNA template, 0.5 U/ml inorganic pyrophosphatase, and 80 μg/ml of the Δ(172–173) variant of the T7 RNA polymerase [ 62 ]. After transcription, a phenol-chloroform extraction was performed. The tRNA transcripts were then precipitated by the addition of 1/15 volume of 7.5 M ammonium acetate and 2.5 volumes of 100% ethanol, and cooling at −20°C for 2 h or overnight. RNA was collected by centrifugation at 16 000 × g for 30 min at 4°C and resuspended in 1× TE buffer (pH 7.5). The solution was mixed 1:1 with formamide and heated at 95°C for 5 min. The RNA was purified via gel electrophoresis on a 12% polyacrylamide gel containing 7 M urea. RNA was visualized via UV-shadow and extracted from the gel by crush and soak. The eluate was precipitated as described earlier, resuspended in 1 mM citrate buffer (pH 6.4), and folded by addition of 7 mM MgCl 2 , heating to 75°C for 10 min, and gradual cooling to 12°C over 75 min.

Determination of Tm TsaC2 concentration for initial velocity assays

Assays to test the Tm TsaC2 concentration at which TC-transfer kinetics dominate in the production of t 6 A-modified tRNA were performed in reaction volumes of 35 μl. Each assay contained 50 mM Tris–HCl at pH 8.00, 100 mM KCl, 5 mM DTT, 5 mM NaHCO 3 , 10 mM MgCl 2 , 2 mM ATP, 250 μM L-[U- 14 C]-threonine (30 μCi/μmol), 10 μM Ec tRNA Thr CGU transcript, 0.2 μM Tm TsaB, 0.2 μM Tm TsaD, 0.2 μM Tm TsaE, and 0.0313, 0.0625, 0.125, 0.25, 0.5, 1, 2, or 4 μM Tm TsaC2. Reactions were incubated at 60°C and quenched after 2 min in 350 μl of cold 10% TCA. The tRNA was allowed to precipitate on ice for at least 15 min and then was transferred to Whatman GF/C filters, and the reaction tubes were washed three times with cold 10% TCA. The filters were then washed with cold 10% TCA, followed by cold 95% ethanol. After the filters were dried, they were submerged in 5 ml of Econo-safe TM liquid scintillation counting cocktail, and radioactivity was measured with liquid scintillation counting using a Beckman LS 6500 multi-purpose scintillation counter. All assays were carried out in duplicate, and the results represent at least three independent trials.

t 6 A activity assays

t 6 A activity assays were performed under initial velocity conditions in 100 μl reactions containing 50 mM Tris-HCl at pH 8.00, 100 mM KCl, 5 mM DTT, 5 mM NaHCO 3 , 10 mM MgCl 2 , 2 mM ATP, 250 μM L-[U- 14 C]-threonine (30 μCi/μmol), 10 μM Ec tRNA Thr CGU transcript (WT and mutants), 0.2 μM Tm TsaB WT/Mutant , 0.2 μM Tm TsaD WT/Mutant , 0.2 μM Tm TsaE WT , and 1 μM Tm TsaC2 WT . Reactions were incubated at 60°C, and 30 μl aliquots were taken at 1, 2, and 3 min after reaction initiation with ATP and L-threonine. Each aliquot was quenched in 350 μl of cold 10% TCA, and tRNA was allowed to precipitate on ice for at least 15 min. tRNA was then transferred to Whatman GF/C filters, and the reaction tubes were washed three times with cold 10% TCA. The filters were then washed with cold 10% TCA followed by cold 95% ethanol. After the filters were dried, they were submerged in 5 ml of Econo-safe TM liquid scintillation counting cocktail, and radioactivity was measured with liquid scintillation counting using a Beckman LS 6500 multi-purpose scintillation counter. All assays were carried out in duplicate, and the results represent at least three independent trials.

Sequence alignments

Structure-based multi-sequence alignments of bacterial TsaB and TsaD proteins were conducted in PROMALS3D [ 63 ] using all published crystal structures of these proteins, AlphaFold models of select orthologs, and 540 TsaD or 550 TsaB non-redundant sequences obtained from UniProt ( https://www.uniprot.org/ ), and the alignments were used to generate sequence logos. For tRNA, 3631 sequences for bacterial tRNA Arg CCU , tRNA Arg UCU , tRNA Asn GUU , tRNA Ile GAU , tRNA Lys CUU , tRNA Lys UUU , tRNA Met CAU , tRNA Ser GCU , tRNA Thr AGU , tRNA Thr GGU , tRNA Thr CGU , and tRNA Thr UGU were obtained from the GtRNAdb website ( https://gtrnadb.ucsc.edu/ ) [ 49 ] and aligned in ClustalW. Logos were generated using the WebLogo v.3.7.9 online tool ( https://weblogo.berkeley.edu/ ). The sequence logo for non t 6 A-accepting bacterial tRNAs was likewise generated using sequences for tRNA Phe GAA , tRNA Ala GGC , tRNA Ala UGC , tRNA iMet CAU , tRNA Asp GUC , tRNA Glu UUC , tRNA Gln UUG , tRNA Gln CUG , tRNA Gly CCC , tRNA Gly GCC , tRNA Gly UCC , tRNA His GUG , tRNA Arg5 CCG , tRNA Ser3 GGA , tRNA Val1,2 GAC , tRNA Tyr GUA , tRNA Trp CCA , tRNA Cys GCA , tRNA Leu CAG , tRNA Leu GAA , tRNA Leu UAG , tRNA Leu GAG , tRNA Leu UAA , tRNA Pro UGG , tRNA Pro GGG , and tRNA Pro CGG .

Results

We determined cryo-EM structures of the T. maritima t 6 A synthase bound to a modified E. coli tRNA Thr CGU produced in vivo , and to an unmodified T. maritima tRNA Lys CUU synthetic transcript. The two structures are highly similar, indicating that the mode of binding and recognition is not dependent on the tRNA modification state. The two tRNAs share 66% overall sequence identity, with complete sequence conservation in the D-stem. We have previously characterized the Ec tRNA Thr CGU transcript as a competent substrate for the T. maritima TsaBDE system [ 27 , 29 ], and a high-resolution crystal structure is available (PDB ID 1qf6 [ 55 ]) to aid model building in the cryo-EM map. LC-MS/MS analysis of the purified Ec tRNA Thr CGU identified both t 6 A37 and ct 6 A37 ( Supplementary Fig. S2A–C ), indicating that the sample contains a mixture of modified species. Residual activity of this preparation in enzymatic assays further suggests the presence of a fraction of unmodified tRNA ( Supplementary Fig. S1H ). The cryo-EM map shows density extending from the N6 position of A37, with the shape and extent consistent with a threonylcarbamoyl moiety. At the achieved resolution of the cryo-EM reconstruction (3.2 Å), distinct modification states at A37 could not be resolved by focused 3D classification or 3D variability analysis, and among the possible modified species only t 6 A can be modeled in the density without steric clashes. The observed density therefore likely reflects a tRNA population enriched in t 6 A, as ct 6 A is readily hydrolyzed to t 6 A under the mildly alkaline conditions used for reconstitution of the protein–RNA complex [ 50 ]. Accordingly, the Ec tRNA Thr CGU -bound structure is interpreted as representing a product-like state, whereas the Tm tRNA Lys CUU -bound structure represents a substrate-bound state.

Overall structure

Cryo-EM analysis of Tm TsaB 2 D 2 bound to E. coli tRNA Thr CGU or T. maritima tRNA Lys CUU yielded 2D and 3D class averages exhibiting the expected butterfly shape of the multi-subunit protein, with one tRNA molecule bound to one protomer at the interface between the TsaD and TsaB subunits (see Supplementary Figs S3 and S4 for example 2D classes and data processing workflows). For both structures, the final reconstruction revealed interpretable density for the full-length tRNA, both TsaB subunits, and one TsaD subunit. The TsaD′ subunit (in the unbound protomer) exhibited very weak density in its N-terminal domain and no density (denatured) for the C-terminal domain. Focused local refinement using a soft mask that excluded TsaD′ improved the overall map quality and nominal resolution to 3.2 Å (Fig. 2A and Supplementary Fig. S5A ). Hence, the final maps and atomic models represent the Tm TsaB 2 D/tRNA complex. A local resolution map indicates a resolution range of 3.0–3.4 Å in the active site, the tRNA anticodon stem-loop, and D-stem regions (Fig. 2B ), and 3.3–3.6 Å for the Tm tRNA Lys CUU -bound structure ( Supplementary Fig. S5B ). The two structures are highly similar and superpose with r.m.s.d. 0.5 Å over 724 C α atoms and 1.5 Å over 71 tRNA phosphates ( Supplementary Fig. S5C ). The tRNA binds to a positively charged surface at the TsaD/TsaB interface (Fig. 2C ), occupying much of the TsaE binding site observed in the crystal structure of the Tm TsaB 2 D 2 E 2 complex [ 29 , 64 ], consistent with previous biochemical data indicating a shared binding site [ 37 ]. The majority of the tRNA binding surface is contributed by the TsaD subunit, with the tRNA/TsaD and tRNA/TsaB interfaces burying 1064.7 and 573.3 Å 2 , respectively.

Mode of tRNA binding and conformational changes relative to the TsaE-bound state

The enzyme recognizes the tRNA substrate via contact with the D-stem and the anticodon loop (Fig. 3 ), while no contact is made with the acceptor stem or T arm. The D-stem is contacted primarily by helix α7 of the TsaD subunit (residues Ala210–Glu223 in Tm TsaD numbers, Supplementary Fig. S6 ), which lies along the major groove. The anticodon loop is anchored at the TsaD/B interface by the α1–α2 helical hairpin of TsaD and the C-terminal helix of TsaB (Fig. 3A ). Deletion of the TsaB C-terminal helix (residues I195–R205) was previously shown to eliminate tRNA binding [ 29 ], consistent with its role in stabilizing the anticodon loop at this interface. The TsaD helical hairpin, comprising residues Ser31–His50 and previously referred to as the “lever arm,” was shown to undergo partial unfolding and a dramatic hinged displacement away from the TsaD active site upon binding of TsaE to TsaBD [ 29 ]. In the tRNA-bound structure, the hairpin is fully ordered and adopts a conformation hinged 87° toward the active site relative to its position in the TsaE-bound state (Fig. 3B ), resembling its orientation in the absence of TsaE, as observed in the crystal structures of E. coli, A. aeolicus , and S. typhimurium TsaBD (PDB IDs 6z81 [ 65 ], 8ifx [ 38 ], and 3zeu [ 36 ], respectively; Supplementary Fig. S7A ). The C-terminal helix of TsaB also swings by ∼34° toward the active site relative to its position in the TsaE-bound structure. Lastly, the C-terminal helical domain of TsaD rotates toward the N-terminal domain by 28°. These movements essentially compact the enzyme and result in closing on the tRNA, trapping the anticodon loop, and insertion of A37 in the active site (Fig. 3C and D ).

36-UAA-38 is a t 6 A determinant sequence across phyla

In addition to A37, the defining sequence features of t 6 A substrate tRNAs include the flanking bases in the anticodon loop, namely U36 at the third anticodon position and A38. 36-UAA-38 has been shown to be a strict determinant for t 6 A modification in the cytosols and mitochondria of eukaryotes [ 10 , 11 , 39 , 40 ]. Although this specificity has not been experimentally investigated in bacteria, it has been inferred from the conservation of 36-UAA-38 in bacterial t 6 A tRNAs. Indeed, we conducted a large-scale sequence alignment of bacterial tRNAs which clearly reveals strict conservation of the 36-UAA-38 motif in ANN-decoding tRNAs (t 6 A substrates), and its absence in non-ANN-decoding tRNAs (non-substrates, Fig. 4A ), consistent with this motif serving as a determinant across phyla. The following observations in the structure of tRNA-bound Tm TsaB 2 D explain this ancient and universal specificity.

Remodeling of the anticodon loop and structural basis for recognition of the 36-UAA-38 motif

The anticodon loop, approached by the enzyme from the minor groove side, is sharply bent toward the major groove of the anticodon stem and remodeled to accomplish positioning of A37 in the active site of TsaD (Fig. 4 and Supplementary Fig. S8 ). The tRNA backbone is kinked at nucleotides 32, 34, 36, and 38 such that the bases 32–38 adopt a zigzag pattern, alternating in orientation relative to the loop path (Fig. 4B and Supplementary Fig. S8A ). The kinks at 36 and 38 result in extrusion of the A37 base from the loop and into the active site, and this backbone conformation is further stabilized by an apparent hydrogen bond between O2′ of U36 and a non-bridging oxygen atom of phosphate 38. The first two anticodon bases C34 and G35 (U35 in Tm tRNA Lys CUU ) sit in shallow surface pockets that do not impose base-specific contacts (Fig. 4C ), explaining the indifference of t 6 A biosynthesis to the base identities at these positions. U36 is flipped into the loop space, breaking the weak U32:A38 non-Watson–Crick interaction normally found in the canonical U-turn conformation of the anticodon loop (Fig. 4D ). This positions the Watson–Crick edge of U36 within distance to make 2 hydrogen bonds with the Hoogsteen edge of A38, with the pair in cis- glycosidic bond orientation and a characteristic constricted C1′–C1′ distance of ∼7.9 Å. The U32 base docks into the major groove side of the U36•A38 base pair within hydrogen bonding distance from both bases, thus forming a planar U32−U36•A38 base triple with all pairs in the triple in cis -glycosidic-bond orientation (Fig. 4C ). The apparent hydrogen bonding network within the triple is as follows: N3 and O4 of U36 hydrogen bond with N6 and N7 of A38, respectively. N3 of U32 hydrogen bonds with O4 of U36, and O4 of U32 with N6 of A38. According to the Leontis–Westhof classification system [ 66 ], the base triple would be classified as cWH–cWH (or cHW–cHW), considering U36 as the central base. A search in the RNA Base Triples Database [ 66 ] returned no similar triples, so to our knowledge this is the first observation of a UUA base triple in this class. Likewise, the overall conformation of the anticodon loop is not observed in any tRNA structure in the PDB, whether in free form or bound to a protein (see example comparisons in Supplementary Fig. S7B ).

The U32−U36•A38 triple is accommodated by a markedly hydrophobic trough on the protein surface, ~23 Å long and 8 Å deep. The trough is formed by the interhelical loop of the α1–α2 helical hairpin of TsaD and the C-terminal helix of TsaB (Fig. 4C , left). On the opposite side, the U32−U36•A38 triple is further stabilized by base stacking interactions with the adjacent A31:U39 Watson–Crick base pair at the proximal end of the anticodon stem. Specifically, U39 stacks on the U36•A38 pair, and A31 stacks on U32 (Fig. 4C , right).

The residues lining the hydrophobic trough for the base triple are Phe39, Val43, Pro44, and Leu82 from TsaD; and Ile195, Ala196, and Trp200 from TsaB, all of which are invariant or highly conserved residues in the respective protein family (Fig. 5A and Supplementary Figs S6, S8B , and S9 ). A38 makes the most extensive stacking contacts with this trough, engaging side chains from both protein subunits, whereas U32 and U36 contact only TsaB and TsaD, respectively. U32 stacks directly against TsaB-Ile195. The U36 base and ribose rings stack against TsaD-Pro44 and TsaD-Leu82, respectively. The adenine ring of A38 stacks face-to-face against TsaD-Phe39, and edge-to-face against TsaB-Trp200, and is anchored by TsaD-Val43 and TsaB-Ala196 at the center of the trough, while its ribose O2′ makes a hydrogen bond with the backbone carbonyl oxygen of TsaD-Phe39.

The backbone phosphates of the anticodon loop receive numerous stabilizing salt-bridge interactions with basic side chains: the phosphate of U32 with TsaB-Arg114, and the A38 and U39 phosphates with Arg169 and Lys166, respectively, from helix α5 in the C-terminal domain of TsaD (Fig. 5A and Supplementary Fig. S8B ). TsaD-Arg169 side chain also hydrogen bonds with the backbone carbonyl oxygen atom of TsaD-Gly40, which is part of the conserved Ω 39 GGVVPE 45 motif (Ω is an aromatic residue) comprising the inter-helix loop of the α1-α2 helical hairpin of TsaD ( Supplementary Fig. S6 ). Thus, Arg169 bridges the tip of the helical hairpin of TsaD to the phospho-ribose backbone of the anticodon loop. The functional importance of this conserved helical hairpin is further supported by prior observations that disrupting the α1 helix in A. aeolicus TsaD nearly abolished activity [ 38 ].

Notably, the enzyme-induced sharp bend in the anticodon loop brings the phosphate of A38 into close proximity to the phosphate of A26 at the D-stem/anticodon-stem junction (Fig. 5B and Supplementary Fig. S8C ). The resulting ∼2.7 Å (∼3.5 Å in Tm TsaB 2 D/ Tm tRNA Lys CUU structure) separation between their non-bridging oxygen atoms is too constricted to accommodate a Mg 2+ ion but is consistent with hydrogen bonding. The conserved TsaD-Lys166 in contact with phosphate 38 may serve to protonate it (Fig. 5A and B ; Supplementary Fig. S8C ), and the cryo-EM map, reflecting the Coulomb potential, reveals intervening density consistent with a positive charge bridging these two phosphates, an observation plausible at the local resolution of 3.1 Å [ 67 ]. In addition, the Ec tRNA Thr CGU -bound structure shows density consistent with a Mg 2+ ion bridging phosphates 25 and 26, which would help neutralize local charge and stabilize this compact backbone arrangement, although comparable density is not observed at this position in the Tm tRNA Lys CUU -bound structure. Overall, the remodeling of the anticodon loop backbone and associated phosphate compaction, potential protonation, and stabilization by conserved basic residues and possibly Mg 2+ define a configuration with A37 positioned in the active site.

To validate the structure and probe the importance of these interactions in catalysis, we introduced amino acid substitutions at selected interacting residues and tested the activities of the resulting mutant proteins in the t 6 A synthesis reaction (Fig. 5C ). As the structure shows that the modified nucleosides in the tRNA body do not contact the enzyme or functional sites, their absence is not expected to affect binding or catalysis; accordingly, activity assays were performed using an in vitro transcribed Ec tRNA Thr CGU as a substrate. This substrate has been used extensively in our prior work and provides a robust readout of t 6 A synthesis ( Supplementary Fig. S1G ). Activity assays were carried out under initial velocity conditions with a molar stoichiometry of TsaC2:TsaBDE of 5:1 (10:1 TsaC2 to TsaB 2 D 2 ) to ensure that TsaD-catalyzed threonylcarbamoyl transferase kinetics dominated the observed rate of t 6 A formation ( Supplementary Fig. S10A and B ). All mutants expressed well in E. coli , yielded soluble proteins, and were purified to homogeneity ( Supplementary Fig. S1A ). The TsaD Pro44Ala , TsaD Pro44Gly , and TsaD Phe39Ala mutants—targeting residues involved in base triple stabilization—exhibited dramatically reduced t 6 A synthesis activities (to 6%, 4%, and 10% of wild-type levels, respectively). TsaB Arg114Ala and TsaD Arg169Ala , which disrupt anticodon loop backbone interactions, showed a reduction to 11% and 20% of wild-type activity, respectively. In addition to its putative role in stabilizing the anticodon loop, TsaD-Lys166 plays a previously established role in t 6 A synthesis by coordinating the γ-phosphate of ATP at the TsaD/TsaE interface [ 29 ], essential for the ATPase activity of TsaBDE that drives multi-turnover t 6 A synthesis. We have previously shown that Tm TsaB 2 D 2 Lys166Ala exhibits significantly reduced tRNA binding affinity and that t 6 A synthesis activity is reduced to below single-turnover levels [ 29 ], indicating a block at a step preceding ATP hydrolysis in the catalytic cycle, specifically the TC transfer step. This defect is consistent with the observed role of TsaD-Lys166 in stabilizing the anticodon loop on the enzyme.

In all tRNAs, bases at positions 32 and 38 form a universal, non-Watson–Crick pair that is essential for the formation of the canonical hairpin structure of the anticodon loop [ 68 ]. They also provide loop flexibility for recognition by enzymes and for correct shaping and positioning of the anticodon triplet in the ribosome decoding center [ 69 , 70 ]. In the base triple observed here, this pairing pattern is reorganized such that the strictly conserved U36 and A38 form a two-hydrogen-bond pair, while the less conserved U32 contributes one interaction with each base in the pair. In bacterial t 6 A tRNAs, position 32 displays a nearly equal proportion of U or C, and in rare cases an A (Fig. 4A ). To probe the limits of tolerance at this position, we introduced the least conservative substitution, U32G, as G32 does not occur in t 6 A tRNAs. The U32G tRNA mutant remains a competent substrate, exhibiting an initial velocity ∼40% that of the wild-type tRNA (Fig. 5D ), indicating that base identity at position 32 is not a strict determinant of recognition. Modeling C32, A32, and even G32 in the structure suggests that they can all be accommodated within the base triple geometry ( Supplementary Fig. S11 ), albeit with one fewer hydrogen bond relative to the observed U32–U36•A38 triple.

In contrast, substitution of U36 with C yields a tRNA mutant that fails to serve as an efficient substrate, exhibiting an initial velocity ∼2% that of the wild-type tRNA (Fig. 5D ), indicating that even a pyrimidine substitution at this position is not tolerated. This is consistent with the observed cWH geometry, as a C36•A38 pair would not satisfy the required hydrogen bonding pattern without protonation of C36, which is disfavored at physiological pH.

In summary, substrate recognition appears to be dictated by the shape and dimensions of the binding trough, together with the restrained geometry imposed on the RNA backbone by tertiary interactions with conserved basic residues of the TsaD/TsaB interface (TsaD-Lys166, TsaD-Arg169, and TsaB-Arg114), which restrict the conformational space available to the anticodon loop, confining it to a binding mode in which a structurally compatible base triple is formed. Within this framework, U36 and A38 appear to define a core pairing geometry of cWH, whereas the base at position 32 can vary while remaining compatible with that geometry and contributing to its stabilization. We propose that this structural constraint underlies the strict requirement for U36 and A38 in t 6 A tRNAs and their role in defining tRNA identity for modification.

Shallow surface pockets for the anticodon wobble and central bases explain their “neutrality” for t 6 A modification

The tRNA wobble nucleotide C34 adopts a flipped-out conformation, rotated ∼180° away from the adjacent bases U33 and G35 (Fig. 5E and Supplementary Fig. S8D ). This allows U33 and G35 to stack against each other, a configuration enabled by a sharp kink in the intervening sugar-phosphate backbone. C34 binds to a surface pocket on TsaB, with its ribose and phosphate moieties within hydrogen bonding distance from the side chains of TsaB-Thr68 and TsaB-Arg71, respectively (Fig. 5F and Supplementary Fig. S8E ). Thr68 is a conserved Thr/Ser in TsaB proteins, and TsaB-Arg71 is invariant ( Supplementary Fig. S9 ). The C34 nucleobase is held in a largely hydrophobic shallow pocket lined with TsaB-Leu67, TsaB-Ile195, TsaB-Ala113, TsaB-Tyr119, and the aliphatic part of TsaB-Arg114 side chain (Fig. 5F and Supplementary Fig. S8E ), all of which are conserved in TsaB proteins ( Supplementary Fig. S9 ). TsaB-Tyr119 is oriented for a potential, possibly water-mediated, hydrogen bond with the base (the side chain OH of Tyr119 is ∼3.6 Å from the pyrimidine N4 of C34), but due to the modest map resolution in this region, this interaction is not certain. We probed several of these interactions with mutagenesis. The TsaB Thr68Ala mutant, targeting a backbone interaction with C34, reduced t 6 A synthesis to 50% of wild-type levels (Fig. 5C ). The TsaB Tyr119Ala and TsaB Ile195Ala mutants, targeting the nucleobase binding pocket, reduced t 6 A synthesis to 70% and 58% of wild-type levels, respectively. These results are consistent with a minor contribution of the C34 pocket to anticodon loop recognition and productive binding.

In bacterial ANN-decoding tRNAs, position 34 can be U/C/G (Fig. 4A and Supplementary Fig. S12 ) and is often modified. An A at that position is rare because ANN codons ending with U are either rare or decoded by GNU anticodons via wobble pairing or by CNU anticodons wherein C is modified to lysidine ( Supplementary Fig. S13 ). To better understand why t 6 A modification is indifferent to the wobble base, we modeled U and G in place of C34 in the structure ( Supplementary Fig. S14 ). The models suggest that, despite its conservation, the C34 binding pocket is large enough to accommodate a purine and can accommodate U/G with the same contacts observed for C.

The G35 nucleobase is bound in a shallow surface pocket defined primarily by TsaD-Ile83, a conserved hydrophobic residue in TsaD orthologs (Fig. 5G and Supplementary Figs S8F and S9 ). G35 stacks against TsaD-Ile83 on one face and against U33 on the other. The nucleobase functional groups of U33 and G35 are exposed to the solvent, and neither base forms hydrogen bonds with the enzyme. The aliphatic side chain of nearby TsaB-Lys32 (a non-conserved residue) may provide an additional hydrophobic interaction with the C-H edge of U33. The lack of base-specific interactions in this pocket and its shallowness allows it to accommodate any canonical base, explaining the tolerance of t 6 A modification to base identity at position 35 in substrate tRNAs. Notably, U33 is conserved in bacterial tRNAs and is unrelated to t 6 A modification (Fig. 4A ), reflecting instead its critical role in enforcing the U-turn conformation of the anticodon loop required for ribosome binding [ 71 ]. Alanine substitution of TsaD-Ile83 reduced t 6 A synthesis activity to 62% of wild-type levels (Fig. 5C ), again indicating a minimal role of this binding pocket in recognition of the anticodon loop.

Accommodation of A37 and t 6 A37 in the active site

The two structures, capturing A37 in the unmodified and t 6 -modified forms, reveal the same configuration of the A37 binding pocket in the substrate- and product-bound states. In both structures, the adenine base is similarly lodged in a deep pocket in the active site of TsaD, lined on one side with the hydrophobic side chains TsaD-Cys10, TsaD-Leu82, and TsaD-Val42, and the glycine-rich beta-turn S 134 GGH 137 on the other side, with the adenine ring directly stacking between Cys10 and Gly135–Gly136 (Fig. 6A and B ). TsaD-Val42 in the inter-helical loop of the α1–α2 helical hairpin and TsaD-Leu82 are conserved hydrophobic residues ( Supplementary Fig. S6 ). TsaD-Cys10 is part of the conserved E 7 -S/T-SCD 11 sequence motif. Alanine substitutions of these residues resulted in significantly compromised t 6 A biosynthesis activity: 8% and 26% of wild-type activity for TsaD-Cys10 or TsaD-Leu82, respectively (Fig. 6C ). We also interrogated Glu162 with mutagenesis due to its conservation and location in a well-defined helix in the vicinity of the ribose of A37, where it is potentially poised for interaction with the ribose 2′OH. Indeed, the TsaD-Glu162Ala mutant exhibits severely impaired t 6 A activity (Fig. 6C ).

In the Ec tRNA Thr CGU -bound structure, clear density is observed in the cryo-EM map for the threonylcarbamoyl (TC) moiety of t 6 A37, which forms a network of interactions with active-site residues in the E 7 -S/T-SCD 11 and S 134 GGH 137 motifs. Contacts in these motifs include the threonine moiety with the side chain of Ser9 and the carbonyl O with the side chain of Ser134 and the backbone amide of Gly135. We have previously shown a role for Ser134 in TC transfer, as its Ala substitution yields a mutant that supports ATP hydrolysis in the presence of TsaE but exhibits severely reduced t 6 A synthesis activity to below single-turnover level, indicating a disruption specifically in the TC transfer step of the catalytic cycle [ 29 ].

TsaD proteins harbor a bound Zn 2+ or Fe 2+ ion in their active sites [ 29 , 35 , 38 , 64 , 65 ] that coordinates the carboxylate and phosphate moieties of TC-AMP [ 65 ] in addition to serving as a structural metal [ 29 ]. Surprisingly, in both tRNA-bound structures, no metal density is observed in the cryo-EM map, although the coordination sphere residues His109, His113, and Asp296 are well defined (Fig. 6A and B ). Instead, His109 and Asp296 side chains are within contact distance to coordinate the carboxylate group and the carbamoyl NH of the TC moiety of t 6 A37, respectively (Fig. 6B ). His137, located in an interstrand loop in the TsaD C-terminal domain, is an additional Zn 2+ ligand in Tm TsaB 2 D 2 E 2 [ 29 ] but is displaced from the zinc site in the tRNA-bound structures as part of the global tRNA-induced closure of the C-terminal domain. The loss of metal is not explained simply by this rearrangement of the metal site, as other bacterial TsaBD structures with a similar coordination geometry retain bound metal [ 35 , 38 , 65 ]. It is possible that the absence of metal density is due to radiation damage [ 72 ].

D-stem recognition via indirect readout

The enzyme contacts the minor groove of the D-stem via the TsaD subunit only (Fig. 3A ). The D-stem base pairs G10:C25, C11:G24, and U12:A23 are contacted by conserved structural elements located in the C-terminal domain of TsaD (Fig. 7A and Supplementary Fig. S15 ). These elements are (i) the first turn of TsaD helix α7 and part of the preceding loop, together comprising the conserved sequence motif F 207 SFSGLK 213 ( Supplementary Fig. S6 ); and (ii) the loop connecting strand β8 and helix α5, carrying the conserved sequence motif h 156 DDs 159 (h and s are a hydrophobic and a small residue, respectively; Supplementary Fig. S6 ). The cryo-EM map and fitted model suggest a network of hydrogen-bonding interactions between conserved side chains in these motifs and the N3 atoms of purines in each of the three base pairs, as well as with the ribose moieties of C25, C11, and U12. Base contacts with distances consistent with hydrogen bonds include: a network of hydrogen bonds between TsaD-Ser208, TsaD-Asp157, and the N3 atom of G10 in G10:C25 (Fig. 7B , left; Supplementary Fig. S15 , left); TsaD-Thr214 from helix α7 with the N3 atom of G24 in C11:G24 (Fig. 7B , middle; Supplementary Fig. S15 , middle); and TsaD-Tyr218 from helix α7 with N3 of A23 in U12:A23 (Fig. 7B , right; Supplementary Fig. S15 , right). The map and fitted model also suggest potential interactions with the RNA backbone. These include a network of hydrogen bonds between TsaD-Asp158, TsaD-Lys213, and O2′ of C25 (Fig. 7B , left, Supplementary Fig. S15 , left); TsaD-Ser208 with the O2′ of C11 (Fig. 7B , middle; Supplementary Fig. S15 , middle); and the carbonyl oxygen atom of TsaD-Gly211 with O2′ of U12 (Fig. 7B , right; Supplementary Fig. S15 , right). That these TsaD residues are important for the t 6 A reaction is evidenced by the impact of their mutation; with the exception of Thr214 and Tyr218, which are not conserved, mutation to Ala virtually abolishes catalytic activity (Fig. 7C ). Note that Lys213 was previously investigated [ 29 ] and, as with Lys166, was shown to serve dual roles in tRNA binding by Tm TsaB 2 D 2 and in coordinating the phosphates of ATP at the TsaD/TsaE interface during the reset step of the catalytic cycle. Mutation to Ala resulted in loss of both tRNA binding and activity [ 29 ], consistent with these roles.

Despite the conservation of the D-stem sequence across bacterial t 6 A substrates (Fig. 7D and Supplementary Fig. S12 ), and the critical role of the D-stem protein contacts discussed earlier, TsaB 2 D 2 is largely insensitive to the nucleotide sequence, as deduced by base-pair mutations (Fig. 7E ). Mutation of G10:C25 to A10:U25 resulted in the loss of only ∼20% activity, and while flipping the base-pair orientation to C10:G25 or U10:A25 resulted in a greater loss of activity, residual activity was still ∼30%–40% of the wild-type activity. Similarly, mutation of C11:G24 to U11:A24, and U12:A23 to A12:U23, saw reductions in activity of only ∼35%. Other D-stem mutants were not accessible to investigation because their sequences possessed alternative secondary structures of comparable stability to the tRNA cloverleaf (RNA secondary structure prediction program Mfold [ 73 ]). Together, these results indicate that although TsaB 2 D makes direct contacts with D-stem bases, these interactions do not mediate sequence-specific readout but instead suggest recognition through indirect readout of D-stem helix geometry and shape. Our structure adds to a growing number of examples in which α-helical elements contribute to substrate recognition through shape sensing of the tRNA D-stem. Examples include the mitochondrial tRNA methyltransferase TRMT10C (PDB ID 8cbk [ 74 ]), human lysyl-tRNA synthetase (PDB ID 9dpl [ 75 ]), and the ELP3 subunit of the human Elongator complex (PDB ID 8ptx [ 76 ]).

Discussion

We determined 3.2 Å cryo-EM structures of the T. maritima t 6 A synthase TsaB 2 D in complex with two different ANN-decoding tRNA species, one unmodified and one natively modified, representing substrate- and product-bound complexes, respectively. Both structures reveal a well-ordered anticodon loop that is remodeled by the TsaB/D interface for specific recognition of the 36-UAA-38 motif and placement of A37 in the TsaD active site. The structures also offer insights into conserved and divergent features of recognition in KEOPS, as well as mechanistic insights, which we explore in the following sections.

Comparison with KEOPS suggests a universal mechanism for recognition of tRNA identity by t 6 A biosynthesis systems

The structures show no interaction between Tm TsaB 2 D and the CCA end of tRNA, consistent with previous findings that the CCA tail is dispensable for tRNA binding and t 6 A biosynthesis by E. coli TsaBDE [ 44 ]. This contrasts with archaeal/eukaryotic KEOPS, which requires CCA recognition via the Cgi121 subunit [ 44 ]. Recently reported cryo-EM structures of archaeal KEOPS bound to tRNA Lys UUU captured two states: one with an intact tRNA but unbound anticodon loop and A37 far from the active site, and another with a distorted tRNA in which the anticodon domain is unraveled and the anticodon loop unresolved [ 42 , 43 ]. Neither structure reveals the direct contacts with D-stem bases observed in the Tm TsaB 2 D/tRNA complex, as the tRNA in those structures is positioned differently, and the D-stem is displaced from helix α7 of TsaD ( Supplementary Fig. S16 ). Nonetheless, base substitutions in the D-stem were shown to inhibit the t 6 A modification activity of KEOPS [ 42 , 43 ], although the substitutions tested did not maintain the Watson–Crick pairing (e.g. G9:C25 to G9:A25, or the double mutation C10:G24 to U10:G24 and U11:A12 to C11:A23 in M. jannaschii tRNA Lys UUU numbers), and thus may have compromised the structural integrity of the D-stem rather than revealing base-specific recognition.

Further, in the distorted tRNA-bound KEOPS structure, G26 is flipped-out of the tRNA core and contacts a conserved arginine in the C-terminal tail of Bud32 [ 42 , 43 ]. In contrast, our structures show a canonical tRNA core and no flipped-out nucleotides in this region. The corresponding nucleotide in our structures is A26, whose backbone phosphate lies at the center of a network of interactions with conserved lysines and, in the Ec tRNA Thr CGU -bound structure, a Mg 2+ ion that stabilize the local, highly constricted RNA backbone around A37. Further, changing G26 to C/U/A either increased or had no effect on the t 6 A activity of KEOPS [ 42 , 43 ], consistent with a role involving backbone remodeling rather than base-specific interaction. It is possible that the G26 protrusion observed in the distorted tRNA bound to KEOPS represents an intermediate state that occurs during loop remodeling by the enzyme.

Neither of the available KEOPS/tRNA complex structures reveals how the anticodon loop is recognized at the Kae1/Pcc1 interface. However, the TsaD recognition motifs observed in Tm TsaB 2 D are conserved in Kae1 (as well as in the human ortholog OSGEP and the mitochondrial ortholog Qri7, Supplementary Fig. S17 ), suggesting a conserved mode of recognition. These include the G 40 GVVP 44 motif at the tip of the α1–α2 helical hairpin, which forms the hydrophobic trough for the U32-U36•A38 base triple (Fig. 5A and Supplementary Fig. S8B ). For example, TsaD-Pro44, on which U36 stacks, corresponds to Pro41 in M. jannaschii Kae1 ( Mj Kae1), and the Mj Kae1 Pro41Ala mutation was shown to abolish the t 6 A activity of KEOPS [ 42 , 43 ]. Also conserved are TsaD-Lys166 and TsaD-Arg169, which stabilize the anticodon loop backbone at the TsaD/TsaB interface (Fig. 5A and B ; Supplementary Fig. S8B and C ), and we showed that their alanine substitutions impair activity ([ 29 ] and Fig. 5C ). The equivalent Mj Kae1 Gln160Asp and Mj Kae1 Arg163Glu mutants were similarly shown to abolish t 6 A activity of KEOPS [ 42 – 44 ]. Likewise, the corresponding S. cerevisiae Qri7 Arg207Ala mutant failed to rescue the kae1Δ growth defect in yeast and is inactive in vitro [ 34 ].

The functional homology between TsaB and the KEOPS Pcc1 subunit with respect to anticodon loop binding is less obvious. The contribution of TsaB to anticodon loop binding is its C-terminal helix located in the C-terminal lobe, which is entirely missing in Pcc1. Pcc1 exhibits partial structural similarity to the N-terminal lobe of TsaB but lacks sequence homology. To explore how Pcc1 might engage the anticodon loop, we modeled the anticodon loop from the Tm TsaB 2 D/ Ec tRNA Thr CGU structure (with similar results using the Tm tRNA Lys CUU -bound structure) onto the Kae1/Pcc1 complex (PDB ID 5jmv [ 77 ]), by superimposing the TsaD and Kae1 subunits (Fig. 8A ). The model suggests that the N-terminal turns of Pcc1 helix α2 could serve the role of the TsaB C-terminal helix in trapping the anticodon loop. Further, Pcc1-Arg63 ( P. furiosus numbering), an invariant residue essential for the t 6 A activity of KEOPS [ 44 ], is positioned near phosphate 32, potentially mimicking the anticodon loop backbone stabilization provided by TsaB-Arg114. Notably, TsaB-Arg114 resides in the C-terminal lobe of TsaB, a region missing in Pcc1. The use of analogous arginines in different structural contexts illustrates a specific strategy by which these phylogenetically distinct proteins have achieved functional convergence for anticodon loop recognition.

Mechanistic insights into TC transfer from TC-AMP

To gain insight into the catalytic mechanism of t 6 A formation, we generated a docking model of TC-AMP and Zn 2+ in the active site of Tm tRNA Lys CUU -bound TsaD (the substrate-bound complex) based on Zn 2+ and the TC-AMP analog BK951 in the crystal structure of Ec TsaBD (Fig. 8B ) [ 65 ]. As noted above, His137 serves as a ligand to Zn 2+ in the Tm TsaB 2 D 2 E 2 structure [ 29 ], but in the tRNA-bound structure, it has moved out of the coordination sphere. In the docking model, it is similarly out of the way, which allows coordination by the carboxylate and phosphate oxygens of TC-AMP (Fig. 8B ). In this arrangement TC-AMP is poised for TC transfer with N6 of A37 ∼2.5 Å from the carbamoyl carbon of TC-AMP. Nucleophilic attack by N6 could be aided by proton transfer to Asp296, a Zn 2+ ligand, which is ∼4.5 Å from N6. The transient tetrahedral intermediate could then be stabilized by H-bond donation from Ser134, followed by intermediate collapse and cleavage of the bond to the phosphate oxygen of AMP. This latter step could be facilitated by proton transfer from Asp296 to the phosphate oxygen.

Overview of the catalytic cycle

Previous biochemical studies established that in the absence of TsaE/ATP, TsaBD supports a single turnover of t 6 A synthesis, whereas TsaE-catalyzed ATP hydrolysis is required for multiple turnover. This led to a working model for the bacterial t 6 A catalytic cycle in which TsaBD binds tRNA and TC-AMP, catalyzes TC transfer, and releases product tRNA without the involvement of TsaE, placing the role of TsaE after product release (Fig. 8C ). The structures presented here define the tRNA-bound state in this cycle. Upon tRNA binding, the enzyme adopts a closed conformation in which the TsaD C-terminal domain encloses the tRNA, and the α1–α2 helical hairpin of TsaD, together with the C-terminal helix of TsaB, pin the anticodon loop, positioning A37 in the TC-transfer site. Following product formation, the t 6 A-modified tRNA is released, leaving the enzyme in a catalytically inactive state. TsaE-ATP binding induces an open, pre-reset conformation in which these structural elements are displaced from the active site, and subsequent ATP hydrolysis and TsaE release restores the closed conformation for the next cycle. In this model, several aspects remain unclear, including the basis of complex inactivation, the role of TsaC/C2, if any, in delivering TC-AMP to the TC-transfer site, and whether its binding is coordinated with tRNA. These questions await further investigation.

Conclusions

The structure of the bacterial t 6 A synthase bound to tRNA reveals that specificity is achieved through remodeling of the anticodon loop, enabling indirect readout of the 36-UAA-38 motif via base triple formation with U32, while accommodating the wobble and middle anticodon bases in nonspecific pockets. Additional contacts with the D-stem provide a second site of indirect readout, and mutagenesis confirms their functional importance. Together, these features account for the selective targeting of ANN-decoding tRNAs and the strict requirement for A38.

Supplementary Material

Acknowledgements

We thank Profs. Dmitry Lyumkis and Steve Reichow for general guidance in cryo-EM procedures, Prof. Kevin Corbett for guidance on cryo-EM data processing, and Dr Nathan Burrows for assistance in cryo-EM grid screening.

Author contributions : Ana Kutchuashvili (Data curation [equal], Formal analysis [equal], Investigation [equal], Validation [supporting], Visualization [supporting], Writing—original draft [equal], Writing—review & editing [supporting]), Erik Scheleen (Data curation [supporting], Investigation [supporting], Visualization [supporting], Writing—review & editing [supporting]), George Guynes (Investigation [supporting]), Junzhou Wu (Data curation [supporting], Investigation [supporting], Writing—review & editing [supporting]), Chi-Kong Chan (Data curation [supporting], Investigation [supporting], Writing—review & editing [supporting]), Peter C. Dedon (Funding acquisition [supporting], Methodology [lead], Resources [supporting], Supervision [supporting], Writing—review & editing [supporting]), Dirk Iwata-Reuyl (Funding acquisition [equal], Methodology [equal], Resources [supporting], Supervision [equal], Writing—review & editing [supporting]), and Manal A. Swairjo (Conceptualization [lead], Formal analysis [lead], Funding acquisition [equal], Methodology [equal], Project administration [lead], Resources [lead], Supervision [lead], Validation [lead], Writing—original draft [equal], Writing—review & editing [lead]).

Contributor Information

Ana Kutchuashvili, Department of Chemistry and Biochemistry, San Diego State University, San Diego, CA 92182, United States.

Erik Scheleen, Department of Chemistry, PO Box 751 Portland State University, Portland, OR 97207, United States.

George Guynes, Department of Chemistry and Biochemistry, San Diego State University, San Diego, CA 92182, United States.

Junzhou Wu, Singapore-MIT Alliance for Research and Technology, 138602, Singapore.

Chi-Kong Chan, Department of Biological Engineering, MIT, Cambridge, MA 02139, United States.

Peter Dedon, Singapore-MIT Alliance for Research and Technology, 138602, Singapore; Department of Biological Engineering, MIT, Cambridge, MA 02139, United States.

Dirk Iwata-Reuyl, Department of Chemistry, PO Box 751 Portland State University, Portland, OR 97207, United States.

Manal A Swairjo, Department of Chemistry and Biochemistry, San Diego State University, San Diego, CA 92182, United States.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

National Institutes of Health GM110588 to M.A.S. and D.I.-R., a grant from the Singapore-MIT Alliance for Research and Technology under the Antimicrobial Resistance Interdisciplinary Research Group to P.C.D., C.-K.C. is supported by a Human Frontier Science Program Cross-Disciplinary Fellowship. A portion of this research was supported by NIH grant R24GM154185 and performed at the Pacific Northwest Center for Cryo-EM (PNCC) with assistance from Marzia Miletto. Funding to pay the Open Access publication charges for this article was provided by NIGMS grant GM110588.

Data availability

The cryo-EM density map, atomic coordinates, and raw cryo-EM data for Tm TsaB 2 D bound to Ec tRNA Thr CGU have been deposited in EMDB, PDB, and EMPIAR under accession codes EMD-77046, 13FN ( https://doi.org/10.2210/pdb13fn/pdb ), and EMPIAR-13544, respectively. The cryo-EM density map, atomic coordinates, and raw cryo-EM data for Tm TsaB 2 D bound to Tm tRNA Lys CUU have been deposited in EMDB, PDB, and EMPIAR under accession codes EMD-77048, 13FO ( https://doi.org/10.2210/pdb13fo/pdb ), and EMPIAR-13543, respectively. Raw LC-MS/MS data have been deposited to the ProteomeXchange Consortium via the PRIDE ( https://www.ebi.ac.uk/pride/ ) partner repository with the project accession # PXD071262.

References

1. Zhang W, Westhof E. The biology of tRNA t(6)A modification and hypermodifications—biogenesis and disease relevance. J Mol Biol. 2025;437:169091. 10.1016/j.jmb.2025.169091.

2. Su C, Jin M, Zhang W. Conservation and diversification of tRNA t(6)A-modifying enzymes across the three domains of life. Int J Mol Sci. 2022;23:13600. 10.3390/ijms232113600.

3. Machnicka MA, Olchowik A, Grosjean H et al. Distribution and frequencies of post-transcriptional modifications in tRNAs. RNA Biol. 2014;11:1619–29. 10.4161/15476286.2014.992273.

4. Jackman JE, Alfonzo JD. Transfer RNA modifications: nature’s combinatorial chemistry playground. Wiley Interdiscip Rev RNA. 2013;4:35–48. 10.1002/wrna.1144.

5. Grosjean H, de Crécy-Lagard V, Marck C. Deciphering synonymous codons in the three domains of life: co-evolution with specific tRNA modification enzymes. FEBS Lett. 2010;584:252–64. 10.1016/j.febslet.2009.11.052.

6. Murphy FVt, Ramakrishnan V, Malkiewicz A et al. The role of modifications in codon discrimination by tRNA(Lys) UUU . Nat Struct Mol Biol. 2004;11:1186–91. 10.1038/nsmb861.

7. Agris PF. Bringing order to translation: the contributions of transfer RNA anticodon-domain modifications. EMBO Rep. 2008;9:629–35. 10.1038/embor.2008.104.

8. Thiaville PC, de Crecy-Lagard V. The emerging role of complex modifications of tRNALysUUU in signaling pathways. Microb Cell. 2015;2:1–4. 10.15698/mic2015.01.185.

9. Thiaville PC, El Yacoubi B, Kohrer C et al. Essentiality of threonylcarbamoyladenosine (t 6 A), a universal tRNA modification, in bacteria. Mol Microbiol. 2015;98:1199–221. 10.1111/mmi.13209.

10. Wang Y, Zeng Q-Y, Zheng W-Q et al. A natural non-Watson–Crick base pair in human mitochondrial tRNAThr causes structural and functional susceptibility to local mutations. Nucleic Acids Res. 2018;46:4662–76. 10.1093/nar/gky243.

11. Lin H, Miyauchi K, Harada T et al. CO2-sensitive tRNA modification associated with human mitochondrial disease. Nat Commun. 2018;9:1875. 10.1038/s41467-018-04250-4.

12. Wang J-T, Zhou J-B, Mao X-L et al. Commonality and diversity in tRNA substrate recognition in t6A biogenesis by eukaryotic KEOPSs. Nucleic Acids Res. 2022;50:2223–39. 10.1093/nar/gkac056.

13. Wu X, Yuan H, Wu Q et al. Threonine fuels glioblastoma through YRDC-mediated codon-biased translational reprogramming. Nat Cancer. 2024;5:1024–44. 10.1038/s43018-024-00748-7.

14. Zhang Y, Zhou JB, Yin Y et al. Multifaceted roles of t6A biogenesis in efficiency and fidelity of mitochondrial gene expression. Nucleic Acids Res. 2024;52:3213–33. 10.1093/nar/gkae013.

15. Joshi K, Bhatt MJ, Farabaugh PJ. Codon-specific effects of tRNA anticodon loop modifications on translational misreading errors in the yeast Saccharomyces cerevisiae . Nucleic Acids Res. 2018;46:10331–9. 10.1093/nar/gky664.

16. Rojas-Benítez D, Eggers C, Glavic A. Modulation of the proteostasis machinery to overcome stress caused by diminished levels of t6A-modified tRNAs in Drosophila . Biomolecules. 2017;7:25.28272317

17. Liu Y, Yang X, Zhou J et al. OSGEP regulates islet β-cell function by modulating proinsulin translation and maintaining ER stress homeostasis in mice. Nat Commun. 2024;15:10479. 10.1038/s41467-024-54905-8.

18. Rojas-Benítez D, Ibar C, Glavic Á. The Drosophila EKC/KEOPS complex: roles in protein synthesis homeostasis and animal growth. Fly. 2013;7:168–72. 10.4161/fly.25227.

19. Arrondel C, Missoury S, Snoek R et al. Defects in t6A tRNA modification due to GON7 and YRDC mutations lead to Galloway–Mowat syndrome. Nat Commun. 2019;10:3967. 10.1038/s41467-019-11951-x.

20. Handford JI, Ize B, Buchanan G et al. Conserved network of proteins essential for bacterial viability. J Bacteriol. 2009;191:4732–49. 10.1128/JB.00136-09.

21. Weiss MC, Sousa FL, Mrnjavac N et al. The physiology and habitat of the last universal common ancestor. Nat Microbiol. 2016;1:16116. 10.1038/nmicrobiol.2016.116.

22. Schneider C, Becker S, Okamura H et al. Noncanonical RNA nucleosides as molecular fossils of an early Earth—generation by prebiotic methylations and carbamoylations. Angew Chem Int Ed Engl. 2018;57:5943–6. 10.1002/anie.201801919.

23. Grosjean H, Westhof E. An integrated, structure- and energy-based view of the genetic code. Nucleic Acids Res. 2016;44:8020–40. 10.1093/nar/gkw608.

24. Galperin MY, Koonin EV. ‘Conserved hypothetical’ proteins: prioritization of targets for experimental study. Nucl Acids Res. 2004;32:5452–63. 10.1093/nar/gkh885.

25. Jin M, Zhang Z, Yu Z et al. Structure–function analysis of an ancient TsaD–TsaC–SUA5–TcdA modular enzyme reveals a prototype of tRNA t6A and ct6A synthetases. Nucleic Acids Res. 2023;51:8711–29. 10.1093/nar/gkad587.

26. Thiaville PC, Iwata-Reuyl D, de Crecy-Lagard V. Diversity of the biosynthesis pathway for threonylcarbamoyladenosine (t 6 A), a universal modification of tRNA. RNA Biol. 2014;11:1529–39. 10.4161/15476286.2014.992277.

27. El Yacoubi B, Lyons B, Cruz Y et al. The universal YrdC/Sua5 family is required for the formation of threonylcarbamoyladenosine in tRNA. Nucleic Acids Res. 2009;37:2894–909. 10.1093/nar/gkp152.

28. Lauhon CT. Mechanism of N6-threonylcarbamoyladenonsine (t 6 A) biosynthesis: isolation and characterization of the intermediate threonylcarbamoyl-AMP. Biochemistry. 2012;51:8950–63. 10.1021/bi301233d.

29. Luthra A, Paranagama N, Swinehart W et al. Conformational communication mediates the reset step in t 6 A biosynthesis. Nucleic Acids Res. 2019;47:6551–67. 10.1093/nar/gkz439.

30. Deutsch C, El Yacoubi B, de Crécy-Lagard V et al. Biosynthesis of threonylcarbamoyl adenosine (t 6 A), a universal tRNA nucleoside. J Biol Chem. 2012;287:13666–73. 10.1074/jbc.M112.344028.

31. Perrochia L, Guetta D, Hecker A et al. Functional assignment of KEOPS/EKC complex subunits in the biosynthesis of the universal t 6 A tRNA modification. Nucleic Acids Res. 2013;4:9484–99. 10.1093/nar/gkt720.

32. Srinivasan M, Mehta P, Yu Y et al. The highly conserved KEOPS/EKC complex is essential for a universal tRNA modification, t 6 A. EMBO J. 2011;30:873–81. 10.1038/emboj.2010.343.

33. Daugeron M-C, Lenstra TL, Frizzarin M et al. Gcn4 misregulation reveals a direct role for the evolutionary conserved EKC/KEOPS in the t 6 A modification of tRNAs. Nucleic Acids Res. 2011;39:6148–60. 10.1093/nar/gkr178.

34. Wan LC, Mao DY, Neculai D et al. Reconstitution and characterization of eukaryotic N6-threonylcarbamoylation of tRNA using a minimal enzyme system. Nucleic Acids Res. 2013;41:6332–46. 10.1093/nar/gkt322.

35. Zhang W, Collinet B, Perrochia L et al. The ATP-mediated formation of the YgjD–YeaZ–YjeE complex is required for the biosynthesis of tRNA t 6 A in Escherichia coli . Nucleic Acids Res. 2015;43:1804–17. 10.1093/nar/gku1397.

36. Nichols CE, Lamb HK, Thompson P et al. Crystal structure of the dimer of two essential Salmonella typhimurium proteins, YgjD & YeaZ and calorimetric evidence for the formation of a ternary YgjD–YeaZ–YjeE complex. Protein Sci. 2013;22:628–40. 10.1002/pro.2247.

37. Luthra A, Swinehart W, Bayooz S et al. Structure and mechanism of a bacterial t 6 A biosynthesis system. Nucleic Acids Res. 2018;46:1395–411. 10.1093/nar/gkx1300.

38. Lu S, Jin M, Yu Z et al. Structure-function analysis of tRNA t(6)A-catalysis, assembly, and thermostability of Aquifex aeolicus TsaD(2)B(2) tetramer in complex with TsaE. J Biol Chem. 2024;300:107962. 10.1016/j.jbc.2024.107962.

39. Morin A, Auxilien S, Senger B et al. Structural requirements for enzymatic formation of threonylcarbamoyladenosine (t 6 A) in tRNA: an in vivo study with Xenopus laevis oocytes. RNA. 1998;4:24–37.

40. Zheng X, Su C, Duan L et al. Molecular basis of A. thaliana KEOPS complex in biosynthesizing tRNA t6A. Nucleic Acids Res. 2024;52:4523–40. 10.1093/nar/gkae179.

41. Zhou J-B, Wang Y, Zeng Q-Y et al. Molecular basis for t6A modification in human mitochondria. Nucleic Acids Res. 2020;48:3181–94. 10.1093/nar/gkaa093.

42. Chuquimarca SMO, Beenstock J, Daou S et al. Author correction: structures of KEOPS bound to tRNA reveal functional roles of the kinase Bud32. Nat Commun. 2025;16:6810. 10.1038/s41467-025-61672-7.

43. Ona Chuquimarca SM, Beenstock J, Daou S et al. Structures of KEOPS bound to tRNA reveal functional roles of the kinase Bud32. Nat Commun. 2024;15:10633. 10.1038/s41467-024-54787-w.

44. Beenstock J, Ona SM, Porat J et al. A substrate binding model for the KEOPS tRNA modifying complex. Nat Commun. 2020;11:6233. 10.1038/s41467-020-19990-5.

45. Spears JL, Gaston KW, Alfonzo JD. Analysis of tRNA editing in native and synthetic substrates. Methods Mol Biol. 2011;718:209–26.

46. Chambers MC, Maclean B, Burke R et al. A cross-platform toolkit for mass spectrometry and proteomics. Nat Biotechnol. 2012;30:918–20. 10.1038/nbt.2377.

47. Wein S, Andrews B, Sachsenberg T et al. A computational platform for high-throughput analysis of RNA sequences and modifications by mass spectrometry. Nat Commun. 2020;11:926. 10.1038/s41467-020-14665-7.

48. Röst HL, Sachsenberg T, Aiche S et al. OpenMS: a flexible open-source software platform for mass spectrometry data analysis. Nat Methods. 2016;13:741–8.

49. Chan PP, Lowe TM. GtRNAdb 2.0: an expanded database of transfer RNA genes identified in complete and draft genomes. Nucleic Acids Res. 2015;44:D184–9. 10.1093/nar/gkv1309.

50. Miyauchi K, Kimura S, Suzuki T. A cyclic form of N6-threonylcarbamoyladenosine as a widely distributed tRNA hypermodification. Nat Chem Biol. 2013;9:105–11. 10.1038/nchembio.1137.

51. Punjani A, Rubinstein JL, Fleet DJ et al. cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination. Nat Methods. 2017;14:290–6. 10.1038/nmeth.4169.

52. Tan YZ, Baldwin PR, Davis JH et al. Addressing preferred specimen orientation in single-particle cryo-EM through tilting. Nat Methods. 2017;14:793–6. 10.1038/nmeth.4347.

53. Iudin A, Korir PK, Somasundharam S et al. EMPIAR: the Electron Microscopy Public Image Archive. Nucleic Acids Res. 2023;51:D1503–11. 10.1093/nar/gkac1062.

54. Sanchez-Garcia R, Gomez-Blanco J, Cuervo A et al. DeepEMhancer: a deep learning solution for cryo-EM volume post-processing. Commun Biol. 2021;4:874. 10.1038/s42003-021-02399-1.

55. Sankaranarayanan R, Dock-Bregeon AC, Romby P et al. The structure of threonyl-tRNA synthetase-tRNA(Thr) complex enlightens its repressor activity and reveals an essential zinc ion in the active site. Cell. 1999;97:371–81. 10.1016/S0092-8674(00)80746-1.

56. Abramson J, Adler J, Dunger J et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature. 2024;630:493–500. 10.1038/s41586-024-07487-w.

57. Liebschner D, Afonine PV, Baker ML et al. Macromolecular structure determination using X-rays, neutrons and electrons: recent developments in Phenix. Acta Crystallogr D Struct Biol. 2019;75:861–77. 10.1107/S2059798319011471.

58. Emsley P, Lohkamp B, Scott WG et al. Features and development of Coot. Acta Crystallogr D Biol Crystallogr. 2010;66:486–501. 10.1107/S0907444910007493.

59. Afonine PV, Poon BK, Read RJ et al. Real-space refinement in PHENIX for cryo-EM and crystallography. Acta Crystallographica Section D. 2018;74:531–44. 10.1107/S2059798318006551.

60. Young JY, Westbrook JD, Feng Z et al. OneDep: unified wwPDB system for deposition, biocuration, and validation of macromolecular structures in the PDB archive. Structure. 2017;25:536–45. 10.1016/j.str.2017.01.004.

61. Sherlin LD, Bullock TL, Nissan TA et al. Chemical and enzymatic synthesis of tRNAs for high-throughput crystallization. RNA. 2001;7:1671–8.

62. Lyakhov DL, He B, Zhang X et al. Mutant bacteriophage T7 RNA polymerases with altered termination properties. J Mol Biol. 1997;269:28–40. 10.1006/jmbi.1997.1015.

63. Pei J, Kim B-H, Grishin NV. PROMALS3D: a tool for multiple protein sequence and structure alignments. Nucleic Acids Res. 2008;36:2295–300. 10.1093/nar/gkn072.

64. Missoury S, Plancqueel S, Li de la Sierra-Gallay I et al. The structure of the TsaB/TsaD/TsaE complex reveals an unexpected mechanism for the bacterial t 6 A tRNA-modification. Nucleic Acids Res. 2018;46:5850–60. 10.1093/nar/gky323.

65. Kopina BJ, Missoury S, Collinet B et al. Structure of a reaction intermediate mimic in t6A biosynthesis bound in the active site of the TsaBD heterodimer from Escherichia coli . Nucleic Acids Res. 2021;49:2141–60. 10.1093/nar/gkab026.

66. Abu Almakarem AS, Petrov AI, Stombaugh J et al. Comprehensive survey and geometric classification of base triples in RNA structures. Nucleic Acids Res. 2012;40:1407–23. 10.1093/nar/gkr810.

67. Bick T, Dominiak PM, Wendler P. Exploiting the full potential of cryo-EM maps. BBA Adv. 2024;5:100113. 10.1016/j.bbadva.2024.100113.

68. Auffinger P, Westhof E. Singly and bifurcated hydrogen-bonded base-pairs in tRNA anticodon hairpins and ribozymes. J Mol Biol. 1999;292:467–83. 10.1006/jmbi.1999.3080.

69. Demeshkina N, Jenner L, Westhof E et al. A new understanding of the decoding principle on the ribosome. Nature. 2012;484:256–9. 10.1038/nature10913.

70. Olejniczak M, Uhlenbeck OC. tRNA residues that have coevolved with their anticodon to ensure uniform and accurate codon recognition. Biochimie. 2006;88:943–50. 10.1016/j.biochi.2006.06.005.

71. Ashraf SS, Ansari G, Guenther R et al. The uridine in “U-turn”: contributions to tRNA-ribosomal binding. RNA. 1999;5:503–11. 10.1017/S1355838299981931.

72. Hattne J, Shi D, Glynn C et al. Analysis of global and site-specific radiation damage in cryo-EM. Structure. 2018;26:759–66.e754. 10.1016/j.str.2018.03.021.

73. Zuker M. Mfold web server for nucleic acid folding and hybridization prediction. Nucleic Acids Res. 2003;31:3406–15. 10.1093/nar/gkg595.

74. Meynier V, Hardwick SW, Catala M et al. Structural basis for human mitochondrial tRNA maturation. Nat Commun. 2024;15:4683. 10.1038/s41467-024-49132-0.

75. Devarkar SC, Budding CR, Pathirage C et al. Structural basis for aminoacylation of cellular modified tRNALys3 by human lysyl-tRNA synthetase. Nucleic Acids Res. 2025;53:gkaf114. 10.1093/nar/gkaf114.

76. Abbassi NE, Jaciuk M, Scherf D et al. Cryo-EM structures of the human Elongator complex at work. Nat Commun. 2024;15:4094. 10.1038/s41467-024-48251-y.

77. Wan LC, Pillon MC, Thevakumaran N et al. Structural and functional characterization of KEOPS dimerization by Pcc1 and its role in t6A biosynthesis. Nucleic Acids Res. 2016;44:6971–80. 10.1093/nar/gkw542.

78. Cardone G, Heymann JB, Steven AC. One number does not fit all: mapping local variations in resolution in cryo-EM reconstructions. J Struct Biol. 2013;184:226–36. 10.1016/j.jsb.2013.08.002.

Associated Data

Supplementary Materials

Data Availability Statement

The cryo-EM density map, atomic coordinates, and raw cryo-EM data for Tm TsaB 2 D bound to Ec tRNA Thr CGU have been deposited in EMDB, PDB, and EMPIAR under accession codes EMD-77046, 13FN ( https://doi.org/10.2210/pdb13fn/pdb ), and EMPIAR-13544, respectively. The cryo-EM density map, atomic coordinates, and raw cryo-EM data for Tm TsaB 2 D bound to Tm tRNA Lys CUU have been deposited in EMDB, PDB, and EMPIAR under accession codes EMD-77048, 13FO ( https://doi.org/10.2210/pdb13fo/pdb ), and EMPIAR-13543, respectively. Raw LC-MS/MS data have been deposited to the ProteomeXchange Consortium via the PRIDE ( https://www.ebi.ac.uk/pride/ ) partner repository with the project accession # PXD071262.