Molecular basis of A. thaliana KEOPS complex in biosynthesizing tRNA t⁶A

Abstract

In archaea and eukaryotes, the evolutionarily conserved KEOPS is composed of four core subunits―Kae1, Bud32, Cgi121 and Pcc1, and a fifth Gon7/Pcc2 that is found in fungi and metazoa. KEOPS cooperates with Sua5/YRDC to catalyze the biosynthesis of tRNA N⁶-threonylcarbamoyladenosine (t⁶A), an essential modification needed for fitness of cellular organisms. Biochemical and structural characterizations of KEOPSs from archaea, yeast and humans have determined a t⁶A-catalytic role for Kae1 and auxiliary roles for other subunits. However, the precise molecular workings of KEOPSs still remain poorly understood. Here, we investigated the biochemical functions of A. thaliana KEOPS and determined a cryo-EM structure of A. thaliana KEOPS dimer. We show that A. thaliana KEOPS is composed of KAE1, BUD32, CGI121 and PCC1, which adopts a conserved overall arrangement. PCC1 dimerization leads to a KEOPS dimer that is needed for an active t⁶A-catalytic KEOPS–tRNA assembly. BUD32 participates in direct binding of tRNA to KEOPS and modulates the t⁶A-catalytic activity of KEOPS via its C-terminal tail and ATP to ADP hydrolysis. CGI121 promotes the binding of tRNA to KEOPS and potentiates the t⁶A-catalytic activity of KEOPS. These data and findings provide insights into mechanistic understanding of KEOPS machineries.

Graphical Abstract

Graphical Abstract.

Introduction

In Saccharomyces cerevisiae, Kae1, Bud32, Cgi121, Pcc1 and Gon7 associate to form a stable protein complex that was dubbed KEOPS (Kinase, Endopeptidase and Other Proteins of Small size) (1) or EKC (Endopeptidase-like Kinase Chromatin-associated) (2). Deletion of genes encoding KEOPS/EKC (KEOPS hereafter) subunits or mutations disrupting the integrity of KEOPS complex led to lethality of budding yeast (1,2), which also manifested a plethora of phenotypes that include shortened telomeres (1,3–6), accumulation of DNA double-strand breaks (DSBs) (7), transcriptional inactivation of pheromone- and galactose-responsive genes (2,5,8), and global translational dysregulation (9,10). Comparative genomics analysis revealed that KEOPS proteins are universally encoded in genomes of archaea and eukaryotes (11), with exception for Gon7/GON7 that is presently only found in yeast (1) or humans (12). No ortholog of Cgi121 is encoded in Drosophila genome (11,13,14). Knockouts of KEOPS genes severely affect growing and development of higher eukaryotes, such as reduced size of larval of fruit flies (14,15), microcephaly and early lethality of zebrafish and mice (16,17). In humans, pathogenic mutations of KEOPS subunits―OSGEP (Kae1), PRPK (Bud32), TPRKB (Cgi121), LAGE3 (Pcc1) and GON7 (Gon7)―are implicated in the heterogeneous autosomal recessive Galloway–Mowat syndrome (GAMOS) that is characterized by early-onset steroid-resistant nephrotic syndrome and microcephaly (16,18).

In vitro enzymatic reconstitution demonstrated that KEOPS participates in biosynthesis of tRNA N⁶-threonylcarbamoyladenosine (t⁶A) (12,19–21), which is an essential post-transcriptional modification needed for correct decoding and translational regulation (9,10,22–25). t⁶A belongs to a core set of 18 ‘universal’ post-transcriptional modifications that are found in tRNAs from three domains of life (26,27). t⁶A is uniquely installed at position 37 of tRNAs that decipher codons starting with adenine, namely, ANN codons (N being A, U, C or G) (26). In t⁶A-modified tRNAs, t⁶A extends its planar ring via intramolecular hydrogen bonds and π–π stacking interaction with U36 and prevents the intra-loop Watson–Crick pairing between U33 and A37 (28,29). Genetic and biochemical analysis revealed that the biosynthesis of tRNA t⁶A requires members from two last universal common ancestor protein families―TsaC/Sua5 (COG0009) (30) and TsaD/Kae1/Qri7 (COG0533) (10), which cooperate to carry out two consecutive reactions (11,31,32). In the first step, TsaC/Sua5 protein utilizes L-threonine, CO₂/HCO₃⁻ and ATP to generate an intermediate L-threonylcarbamoyladenylate (TC-AMP) (33,34); in the second step, TsaD/Kae1/Qri7 protein catalyzes the transfer of TC-moiety from TC-AMP onto N6 atom of tRNA A37, leading to tRNA t⁶A (34,35). However, such a two-component biosynthetic system only exists in eukaryotic mitochondria (34–36). In bacteria, the transfer of TC-moiety is catalyzed by TsaD, TsaB and TsaE (33,37,38), which form an ATP-mediated TsaD–TsaB–TsaE complex (TsaDBE hereafter) (38–42). In archaea and eukaryotes, the TC-transfer is catalyzed by Kae1/OSGEP in form of KEOPS complex (12,18,43–46). Interestingly, TsaC- and TsaD-like domains are fused in TsaN from Pandoraviruses and are capable of catalyzing N⁶-threonylcarbamoylation modification of adenosine di-/tri-phosphates at nucleotide level, leading to t⁶ADP and t⁶ATP (47). The structural similarity in TsaC/Sua5 proteins (47–51) or TsaD/Kae1/Qri7 proteins (18,39,41,47,52–54) implies a conserved catalytic mechanism underlying TC-AMP formation or TC-transfer reactions (32). However, biochemical analysis showed large variations in t⁶A modification efficiencies of different t⁶A-modifying enzymes towards same tRNA substrates, and vice versa (12,31,36,43,44,55). It appears that the TsaD/Kae1/Qri7 family proteins have evolved more auxiliary components to regulate t⁶A-modifcation frequencies of tRNAs in coping with increasing biological complexity (11,31,32).

Crystal structures of KEOPS subcomplexes allowed reconstruction of structural models for KEOPS from archaea (3,44), yeast (56) and humans (18,57). These three models of KEOPS showed that the core subunits adopt a conserved linear architecture depicted as Pcc1/LAGE3–Kae1/OSGEP–Bud32/PRPK–Cgi121/TPRKB (31,32). Gon7/GON7 is an intrinsically disordered protein and interacts solely with Pcc1/LAGE3 in KEOPS (18,56). Recently, a paralog of archaean Pcc1, dubbed Pcc2, interacts with Pcc1 in a manner that is analogous to Gon7/GON7 (46). Biochemical analysis demonstrated that the four-subunit KEOPS from archaea forms a dimer (20,44) whereas the five-subunit KEOPSs from yeast and humans cannot form a dimer due to the presence of Gon7/GON7 (12,18,56). The dimerization of Pcc1 leads to formation of archaean KEOPS dimer that is required for in vitro tRNA t⁶A biosynthesis (44,46). Subtraction of Pcc1 from archaean KEOPS leads to dead t⁶A-catalytic activity of the subcomplex Kae1–Bud32–Cgi121 (20). Deletion of Gon7/GON7 severely affects tRNA t⁶A biosynthesis by KEOPS in yeast and humans (12,18,34). Likewise, replacement of Pcc1 by Pcc2 in archaean KEOPS still sustains the dimeric state but leads to loss of t⁶A catalytic activity (46). Crystal structure of M. jannaschii (Mj) Cgi121–tRNA complex allowed generation of a structural model of archaean KEOPS–tRNA complex (45). According to this model, anticodon stem loop of tRNA is anchored in the concave formed between Pcc1 and Kae1, allowing tRNA A37 to protrude into the t⁶A-catalytic center of Kae1; D stem loop of tRNA simultaneously makes contacts with Kae1 and Bud32; 3′ CCA end is bound by Cgi121; TψC stem loop does not participate in direct interaction with KEOPS. This model delineates a plausible binding mode for KEOPS–tRNA and explains the roles of individual subunits (31,45). However, precise molecular interactions between KEOPS and tRNA are still not determined, e.g. the binding of tRNA A37 in catalytic site of Kae1 is still enigmatic. Nonetheless, a manual adjustment of tRNA crystal structure is needed to geometrically fit an extended surface of the four subunits of KEOPS (45), suggesting that binding of tRNA to KEOPS might mutually induce large conformational changes in structures of KEOPS and tRNA. Structural analysis and biochemical validations demonstrated that t⁶A-catalytic activity of KEOPS necessitates an ATP to ADP hydrolysis by Bud32 (19,20,45,56,57). At present, it's poorly understood as how KEOPS subunits cooperate to regulate KEOPS–tRNA assembly and t⁶A modification efficiency. Therefore, an atomic structure of a complete KEOPS complex is desirable to investigate the precise roles and molecular workings of KEOPS in tRNA t⁶A biosynthesis.

Comparative genomics analysis revealed that orthologs of the four core subunits of KEOPS―Kae1, Bud32, Cgi121 and Pcc1―are encoded in Arabidopsis thaliana genome (11). However, the biochemical functions and structures of A. thaliana KEOPS proteins have not been characterized. Moreover, liquid chromatography–mass spectrometry (LC–MS) analysis revealed t⁶A and its hypermodified derivative–N⁶-methyl t⁶A (m⁶t⁶A) in tRNAs isolated from Arabidopsis thaliana (58). Yet, the biochemical pathway of tRNA t⁶A biosynthesis remains uncharacterized. In this study, we reconstituted an enzymatic biosynthesis of tRNA t⁶A using purified recombinant proteins of A. thaliana KEOPS (KAE1, BUD32, CGI121 and PCC1) and YRDC. We reconstituted A. thaliana KEOPS complex and investigated molecular mechanisms of A. thaliana KEOPS in tRNA t⁶A biosynthesis. Here, we report the cryo-EM structure of A. thaliana KEOPS and structure–function relationship analysis of A. thaliana KEOPS–tRNA assembly, which extend our current mechanistic understandings of the ancient KEOPS machineries.

Materials and methods

Plasmid construction, protein expression and purification

We constructed expression plasmids of Arabidopsis thaliana (At) KEOPS proteins using chemically synthesized DNAs encoding AtYRDC (Gene ID: 836180), AtKAE1 (Gene ID: 828368), AtBUD32 (Gene ID: 832 680), AtCGI121 (Gene ID: 829592) and AtPCC1 (Gene ID: 835384). In total, we constructed plasmids of pET21a–AtYRDC-6His, pET26a–AtKAE1–BUD32–CGI121–PCC1-6His, pET21a–AtKAE1–BUD32-6His, pET26a–AtKAE1–PCC1-6His, pJ241–AtBUD32–CGI121-6His and pET21a–AtCGI121-6His, which express six histidine (6His)-tagged YRDC, KEOPS, KAE1–BUD32, KAE1–PCC1, BUD32–CGI121 and CGI121, respectively. In addition, we constructed 30 plasmids that express AtKEOPS variants with site-directed mutagenesis following the manufacturer's protocol (Takara). The primers are listed in Supplementary Table S1.

Expression plasmids of AtKEOPS proteins were transformed into Escherichia coli (Rosetta DE3 pLysS or BL21) and cultured overnight at 37°C. The transformant preculture was inoculated 1/100 (v/v) in LB or 2YT medium and grown until the OD₆₀₀ reached 0.5–0.8, induced with addition of 0.5 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) for 3 hours at 37°C or for 20 hours at 16°C. S. cerevisiae (Sc) KEOPS and P. salinus (Ps) TsaN^1–392 were expressed using plasmids pJ241–ScKEOPS-6His (56) and pET24a–TsaN^1–392 (47). Cells were harvested by centrifugation and resuspended in Lysis buffer (20 mM Tris–HCl pH 7.5, 300 mM NaCl and 5 mM β-mercaptoethanol). Cells were lysed using sonification, followed by centrifugation (15 000 g) for 30 min at 8°C. The supernatant was applied to Ni-NTA affinity chromatography for initial purification, followed by size–exclusion chromatography (SEC) purification using gel-filtration columns (HiLoad 16/600 Superdex 200 or 75, GE Healthcare). SDS–PAGE analysis of the proteins was applied in each purification step. Protein concentrations were determined by NanoDrop2000 (ThermoFisher Scientific).

AtKEOPS complex was reconstituted using purified KAE1–PCC1 and BUD32–CGI121, which were mixed at a molar ratio of 1:2 and applied to SEC for isolation of complete KEOPS complex. AtKEOPS variants were prepared in same way using corresponding variants of KAE1–PCC1 or BUD32–CGI121. The oligomeric state of AtKEOPS complex, subcomplexes and variants were determined by SEC (Superdex 200 Increase 10/300 GL column, HiLoad 16/600 Superdex 200 column, GE Healthcare) with reference to ScKEOPS, of which the oligomeric state and molecular weight were determined by SAXS (56). Standard calibration and reference to other well-determined proteins were used for comparing elution volumes and estimating the molecular weight.

Bulk tRNA extraction

Arabidopsis thaliana RNAs were extracted with TRIzol (ThermoFisher Scientific) from seedlings grown for 3 days at 25°C on Murashige and Skoog medium supplemented with 0.8% agar and 1% sucrose, followed by precipitation using absolute ethanol at –80°C. Total RNAs were applied to 8 M urea–polyacrylamide gel (12%) electrophoresis (Urea–PAGE). Gel slices containing full-length tRNAs were cut out, followed by elution in buffer containing 500 mM NaAc pH 5.2 and precipitation in ethanol at –80°C. The tRNA pellets were dissolved in buffer containing 50 mM Tris–HCl pH 8.0, 5 mM MgCl₂ and 100 mM KCl. Refolding of tRNAs was performed by heating up to 95°C and gradient annealing at a rate of –1°C/min. Yeast total RNAs were extracted with TRIzol from S. cerevisiae sua5Δ strain cells that were grown in YPD medium at 28°C for 72 h (4). Bulk SctRNAs were purified following same protocols of separation, precipitation and refolding as for bulk AttRNAs. Concentration of tRNAs was determined by NanoDrop2000 (ThermoFisher Scientific).

In vitro transcription of tRNA

DNA templates of tRNA^Arg_CCU (Gene ID: 3771562), tRNA^Arg_UCU (Gene ID: 3767807), tRNA^Thr_CGU (Gene ID: 3768453), tRNA^Ile_AAU (Gene ID: 3770537), tRNA^Lys_UUU (Gene ID: 3768774), tRNA^Ser_GCU (Gene ID: 3769744), tRNA^Met_CAU (Gene ID: 3766659), tRNA^Asn_GUU (Gene ID: 3768376) and tRNA^Ile_UAU (Gene ID: 3768844) from Arabidopsis thaliana were prepared by overlap extension PCR using chemically synthesized primers, in which the forward primer contains a T7 promoter sequence at the 5′ terminus. tRNA genes and primers are summarized in (Supplementary Table S2). Run–off transcription was carried out using T7 RNA polymerase at 30°C for 8 hours in a reaction mixture containing 40 mM Tris–HCl pH 8.0, 5 mM NTP mix, 5 mM DTT, 1 mM spermidine, 0.5% Triton X-100, 33 mM MgCl₂, 3 μM T7 RNA polymerase, 10 μM pyrophosphatase and 15 mM GMP. RNA transcripts from in vitro transcription (IVT) were further purified by Urea–PAGE following protocols as for bulk tRNAs of Arabidopsis thaliana. The correct folding of IVT AttRNA was confirmed by Circular Dichroism spectra analysis (59). Concentrations of IVT tRNAs were determined by NanoDrop2000.

Enzymatic synthesis of tRNA t⁶A and LC–MS analysis of t⁶A nucleoside

For TC-AMP formation, 2 μM AtYRDC was incubated with 4 mM L-threonine, 20 mM NaHCO₃, 2 mM ATP and 5 mM MgCl₂ in a reaction buffer containing 20 mM Tris–HCl pH 7.5, 200 mM NaCl and 1 mM TCEP for 10 minutes at 25°C. For tRNA t⁶A formation, 2 μM AtKEOPS and 20 μM tRNA were added to the TC-AMP biosynthesis reaction system and the mixture was incubated for 90 minutes at 30°C followed by purification using 12% Urea–PAGE as described above. The dissolved t⁶A-tRNA was digested into single nucleoside using Nuclease P1 (0.1 U/ml, Sigma) and Alkaline Phosphatase (0.1 U/ml, Sigma). 50 μl sample of the mononucleosides was chromatographed using a C18 column (5 μm, 4.6 × 250 mm, Agilent) at a flow rate of 0.8 ml/min using a mobile phase composed of 0.1% trifluoroacetic acid aqueous and methanol. The nucleosides were isocratically eluted with 5% methanol for 5 min and gradiently eluted with 5–40% methanol for 15 min and 40–98% methanol for 5 min. Nucleosides were identified by the retention time at 254 nm and the mass spectrometry detection (Agilent 6125B). Data collection and analysis were performed using the OpenLab software v3.5 (Agilent). Based on the integrated peak areas of A, U, C, G and t⁶A, the t⁶A modification efficiency was obtained by dividing the peak area ratio (t⁶A/A) by the number (N) ratio of 1/(N_A– 1). Three independent replicates were performed for all assays and data graphs were generated using GraphPad Prism. Error bars in quantification data represent standard deviations for triplicate measurements.

Cryo-EM sample preparation, data collection and processing, model building and validation

800 μg/ml AtKEOPS was mixed with equal volume of tRNA^Arg_CCU at a molar ratio of 1:2. A drop of 3 μl sample was applied to a freshly glow-discharged Quantifoil gold grid (300 mesh, R0.6/1) and blotted for 2 s at 10°C under 100% humidity in a FEI Vitrobot Mark IV (ThermoFisher Scientific). The grids were stored in the liquid nitrogen until data acquisition. Cryo-EM data were collected on the Titan Krios G3i microscope operated at 300 kV (ThermoFisher Scientific) and equipped with a BioQuantum K3 Imaging Filter direct electron detector (Gatan). The energy filter was used in zero-loss mode with a slit width of 20 eV. All the movies were automatically recorded with EPU software (ThermoFisher Scientific) at a nominal magnification of 105 000 x in super-resolution mode with a pixel size of 0.43 Å. The defocus range was from –0.9 to –2.7 μm. All the movie stacks were collected from three cryo-EM sessions. The total electron dose for each movie stack was 52–64 e⁻/Ų fractionated into 40 fractions over 3.2 s.

The entire image processing was carried out with cryoSPARC v3.3 (60). All the movie stacks were subjected to patch motion correction and patch CTF estimation. A total of 14 622 micrographs were selected to do the following procedure. Firstly, 1 470 particles were manually picked from 100 micrographs with varied defocus values and subjected to 2D classification. Good particles corresponding to those 2D class average maps showing clear density distribution were used for topaz training. In total, 736 244 picks were extracted with a box size of 512 pixels. After several rounds of 2D classification, 327 681 picks were retained and subjected to ab initio reconstruction into three models. Heterogeneous refinement was carried out to further clean the data set. A total of 232 232 picks from two classes were combined to perform the homogeneous refinement, yielding a resolution of 3.95 Å. After that, a further round of non-uniform refinement was performed to improve the resolution to 3.65 Å. An overall mask was generated and used in the local refinement, which finally improved the resolution to 3.2 Å. The resolution for all reconstructions were evaluated by using the gold standard Fourier Shell Correlation (FSC) of 0.143. Data collection and reconstruction parameters are presented in Supplementary Table S3.

Prediction structures of AtKAE1 (AF-O49653-F1), AtBUD32 (AF-Q94K14-F1), AtCGI121 (AF-Q6NMZ4-F1) and AtPCC1 (AF-Q8GWD7-F1) were retrieved from AlphaFold Protein Structure Database (www.alphafold.ebi.ac.uk) (61). AtKEOPS subcomplexes were generated using crystal structures archaean KEOPS subcomplexes (3,44) and were rigidly fitted into the cryo-EM map using Phenix (62). The complete model was improved with real-space refinement using Phenix, followed by manual building with COOT (63). The final model was validated by MolProbity (64), and the refinement statistics are presented in Supplementary Table S3.

ATPase assay

The NADH-coupled ATPase assay was employed to analyze the hydrolysis of ATP to ADP (39,47). 200 μl reaction mixture was made of 4 mM phosphoenolpyruvate, 0.5 mM NADH, 6 U/ml pyruvate kinase, 9 U/ml lactate dehydrogenase, 2 mM ATP and 5 μM proteins in the presence or absence of 10 μM tRNA^Arg_CCU in buffer containing 50 mM Tris–HCl pH 7.5, 100 mM NaCl, 50 mM KCl, 5 mM MgCl₂ and 1 mM DTT. The absorbance at 340 nm was recorded at an interval of 30 seconds for a total of 60 min at 25°C with MULTISKAN GO microplate reader (ThermoFisher Scientific) using a 96-well plate. When ATP is hydrolyzed to ADP, NADH is gradually consumed, as reflected by the decrease in absorbance values. A standard curve of ADP against NADH absorbance was obtained to quantify the hydrolysis rate of ATP to ADP. For kinetic analysis, 2 μM protein and 0–50 μM ATP were used and kinetic parameters were calculated according to Michaelis–Menten equation. All the measurements were performed independently in triplicates. Error bars in the data represent standard deviations.

Electrophoretic mobility shift assay

For the native gel analysis, 20 μM AtKEOPS was incubated with 20 μM IVT AttRNAs in the buffer containing 20 mM Tris–HCl pH 7.5, 300 mM NaCl and 5 mM β-mercaptoethanol and 20% (v/v) glycerol. The mixture was loaded onto a non-denaturating gel (2% agarose, 50 mM Tris–base pH 8.0 and 100 mM glycine) and electrophoresis ran for 1 hour at 100 V at 4°C in pre-chilled buffer containing 50 mM Tris–base pH 8.0 and 100 mM glycine. tRNAs were visualized under UV light at 254 nm after staining with Ethidium Bromide, and proteins were stained with Comassie Brilliant Blue. 5′ 6-FAM (6-Carboxyfluorescein)-labelled AttRNA^Arg_CCU (5′-6FAM-tRNA^Arg_CCU) was chemically synthesized (Tsingke) and was re-folded before use. 0.1 μM 5′-6FAM-tRNA^Arg_CCU was incubated with AtKEOPS proteins (0–0.5 μM) for gel-shift assay on a non-denaturing gel (1.6% agarose, 50 mM Tris–base pH 8.5 and 100 mM glycine). The electrophoresis ran for 1 hour at 100 V at 4°C in pre-chilled buffer containing 50 mM Tris–base pH 8.5 and 100 mM glycine. The presence of 5′-6FAM-tRNA^Arg_CCU on gel was visualized at a wavelength of 488 nm by Pharos FX (Bio-Rad). 10-fold amount of the original input protein in each lane was analyzed and visualized on a separate SDS-PAGE.

Microscale thermophoresis (MST)

50 nM 5′-6FAM-AttRNA^Arg_CCU was incubated with AtKEOPS complex, subcomplexes or variants at increasing concentrations (0.09765625, 0.1953125, 0.390625, 0.78125, 1.5625, 3.125, 6.25, 12.5, 25, 50 and 100 μM) in MST buffer (20 mM PBS pH 7.5, 300 mM NaCl and 0.05% (v/v) Tween 20). Measurements were performed at 25°C in capillaries (MO-K022, NanoTemper Technologies) on the Monolith NT.115 (NanoTemper Technologies) using 20% LED and medium IR-laser power. Binding data was analyzed using MO.affinity analysis software (NanoTemper Technologies) and equilibrium dissociation constant (K_d) values were fitted by the K_d equation and model. All experiments were reproduced at least for three times using proteins from different batches. Error bars in data analysis represent standard deviations for triplicate measurements.

LC–MS/MS proteomic analysis

In the pull-down experiment, 0.5 g Arabidopsis thaliana seedlings were grounded in liquid nitrogen. Total proteins were recovered in 200 μl Lysis buffer supplemented with 1 mM PMSF, followed by centrifugation at 4°C. The supernatant was incubated with 100 μg purified AtKEOPS complex for 1 h. The mixture was applied to purification by Ni-NTA affinity chromatography and unbound proteins were eluted with lysis buffer supplemented with increasing concentrations of imidazole. The final eluted fractions containing KEOPS and potential interactors were denatured and reduced with 2 M guanidine hydrochloride and 10 mM TCEP at 56°C for 30 min. The samples were subsequently alkylated using 40 mM iodoacetamide at room temperature in the dark, followed by overnight digestion using 2 μg trypsin. Tryptic peptides were acidified with 10% formic acid and then loaded on C-18 (3M) stage tips, desalted with 0.1% formic acid and eluted with buffer (80% acetonitrile, 0.1% formic acid). Samples were dried in a vacuum centrifuge. Dried peptides were dissolved in 0.1% formic acid and chromatographed using a C18 column (75 μm inner diameter, 15 cm length, 2 μm particles) on the EASY-nLC1000 UHPLC (ThermoFisher Scientific) coupled to Orbitrap Exploris 480 mass spectrometer (ThermoFisher Scientific). Mass spectra were acquired in data-dependent mode and MS measurements were constructed in the positive-ion mode. The m/z range was set to 350–1550, orbitrap resolution is 60 000. MS raw files were processed with the MaxQuant software (version 1.6.14) using standard settings with the additional options match between runs, iBAQ (intensity-based absolute quantification) selected. Other parameters were set up using the default values and the false discovery rate was set to 0.01 for both peptide and protein identification. Relative abundance of the identified interactors was measured by iBAQ values.

Bioinformatic analysis

We performed the sequence alignment and calculated the sequence identities using Clustal Omega (65). AtKAE1 (Uniprot: O49653) was aligned to Homo sapiens (Hs) OSGEP (Uniprot: Q9NPF4), Saccharomyces cerevisiae (Sc) Kae1 (Uniprot: P36132), Methanococcus jannaschii (Mj) Kae1 (Uniprot: Q58530), Pyrococcus abyssi (Pa) Kae1 (Uniprot: Q9UXT7) and Thermoplasma acidophilum (Ta) Kae1 (Uniprot: Q9HLA5); AtBUD32 (Uniprot: Q94K14) was aligned to HsPRPK (Uniprot: Q96S44), ScBud32 (Uniprot: P53323) and MjBud32 (Uniprot: Q58530); AtCGI121 (Uniprot: Q6NMZ4) was aligned to HsTPRKB (Uniprot: Q9Y3C4), ScCgi121 (Uniprot: Q03705) and MjCgi121 (Uniprot: Q57646); AtPCC1 (Uniprot: Q8GWD7) was aligned to HsLAGE3 (Uniprot: Q14657), ScPcc1 (Uniprot: Q3E833), PaPcc1 (Uniprot: Q9V1Z9) and Pyrococcus furiosus (Pf) Pcc1 (Uniprot: Q8TZI1). Superposition of structures was performed with COOT and PyMOL (Schrödinger, LLC). Root-mean-square deviation (RMSD) value between two structures was calculated by PyMOL. The visualization and graphic representation of the structures were generated using PyMOL and UCSF Chimera (66). The electrostatic potential molecular surface was generated with Adaptive Poisson–Boltzmann Solver (67).

Results

Enzymatic reconstitution of tRNA t⁶A by YRDC and KEOPS from Arabidopsis thaliana

We first confirmed t⁶A and a number of other modifications in bulk tRNAs that were isolated from Arabidopsis thaliana seedlings (Supplementary Figure S1A). Comparative genomic analysis identified A. thaliana (At) orthologs of TsaC/Sua5/YRDC―AT5G60590 (YRDC), and four core subunits of KEOPS―AT4G22720 (KAE1), AT5G26110 (BUD32), AT4G34412 (CGI121) and AT5G53045 (PCC1) (11). No orthologs of GON7/Gon7 or Pcc2 were identified to be encoded in A. thaliana genome (11,46).

We first constructed an expression plasmid for A. thaliana YRDC using chemically synthesized DNA and purified recombinant YRDC (Supplementary Figure S1B) that was expressed bacterial cells. Liquid chromatography–mass spectrometry (LC–MS) analysis shows that YRDC is active in catalyzing the formation of threonylcarbamoyl-AMP (TC-AMP) in the presence of ATP, L-threonine and NaHCO₃ (Supplementary Figure S1B).

We constructed expression plasmids of A. thaliana KEOPS proteins using chemically synthesized DNAs and tried heterologous expression of KAE1–BUD32–PCC1–CGI121 (KEOPS), BUD32–CGI121 (BC), KAE1–PCC1 (KP), KAE1–BUD32 (KB), KAE1–BUD32–PCC1 (KBP) and CGI121 (C). We did not succeed in concomitantly expressing KAE1, BUD32, CGI121 and PCC1 using a polycistronic plasmid similar to that for S. cerevisiae (Sc) KEOPS (56). Alternatively, we assembled KEOPS by mixing KP and BC at a molar ratio of 1:2, followed by purification using size―exclusion chromatography (SEC) (Figure 1A). The SEC and SDS–PAGE analysis of KEOPS complex, subcomplexes (KBP, KP, BC, KB) and CGI121 are shown in Supplementary Figure S1C and Figure 1B, respectively. Notably, we observed a persistent degradation KAE1 via the N-terminal end, which was confirmed by MS and western blot against 6His tag. SEC analysis shows that the four-subunit AtKEOPS eluted out (at a calculated molecular weight of ∼200 kDa) earlier than the five-subunit ScKEOPS (a monomeric complex with a molecular weight of 117.7 kDa) (56), indicating that AtKEOPS exists as an eight-subunit dimer (K₂B₂C₂P₂) as that of MjKEOPS (44) (Supplementary Figure S1C). Compared SEC profiles suggest that either KAE1–PCC1 or KAE1–BUD32–PCC1 forms a dimer whereas BUD32–CGI121 or KAE1–BUD32 exists as a monomer (Supplementary Figure S1C).

Figure 1.

In vitro enzymatic reconstitution of tRNA t⁶A biosynthesis using recombinant YRDC and KEOPS proteins of Arabidopsis thaliana. (A) Size-exclusion chromatography (SEC) profile (HiLoad 16/600 Superdex 200, GE Healthcare) of A. thaliana KEOPS and BUD32–CGI121. KEOPS was reconstituted using purified KP (KAE1–PCC1) and BC (BUD32–CGI121), which were mixed at a molar ratio of 1:2. (B) SDS–PAGE analysis of KEOPS, KBP (KAE1–BUD32–PCC1), KP, BC, CGI121 and KB (KAE1–BUD32). * indicates KAE1 degradation (confirmed by LC–MS/MS). (C) LC–MS analysis of t⁶A formation in assay that contained 4 mM l-threonine, 20 mM NaHCO₃, 2 mM ATP, 20 μM bulk SctRNAs (isolated from sua5Δ strain), 2 μM AtYRDC and 2 μM AtKEOPS or ScKEOPS. (D) Nucleotide sequences of A. thaliana tRNAs that were in vitro transcribed (IVT) and used for tRNA t⁶A assays. Amino acid acceptor arm, D stem loop, anticodon stem loop and TψC stem loop of tRNAs are highlighted in pink, green, blue and gray, respectively. Anticodons are shown in red. (E) Native gel analysis of the interaction between 10 μM AtKEOPS and 20 μM IVT AttRNAs (tRNA^Arg_CCU, tRNA^Arg_UCU, tRNA^Thr_CGU, tRNA^Ile_AAU, tRNA^Lys_UUU, tRNA^Ser_GCU, tRNA^Met_CAU, tRNA^Asn_GUU, tRNA^Ile_UAU and tRNA^Ile_UAU-G37A). (F) Quantified t⁶A modification efficiencies of AtKEOPS towards these IVT AttRNAs. Percentage of t⁶A was normalized according to tRNA sequence. (G) LC–MS analysis of tRNA t⁶A formation by 2 μM AtYRDC and AtKEOPS, AtKBP or AtKP using 20 μM IVT AttRNA^Arg_CCU or AttRNA^Arg_CCU-ΔCCA as substrate. (H) Quantified t⁶A modification efficiencies of AtKEOPS, AtKBP and AtKP towards AttRNA^Arg_CCU or AttRNA^Arg_CCU-ΔCCA (G).

We first tested the t⁶A-catalytic activity of AtKEOPS using bulk SctRNAs that were isolated from S. cerevisiae sua5Δ cells. LC–MS analysis shows that SctRNAs were deficient in t⁶A but acquired t⁶A in the presence of AtKEOPS and AtYRDC (Figure 1C). In addition, the catalytic activity of AtKEOPS towards SctRNAs was comparable to that of ScKEOPS (Figure 1C). In summary, the active recombinant proteins expressed in E. coli cells demonstrate that folding and function of KEOPS do not require essential post-translational modifications for in vitro biosynthesis of tRNA t⁶A. We transcribed and purified 9 ANN-decoding tRNAs―tRNA^Arg_UCU, tRNA^Arg_CCU, tRNA^Thr_CGU, tRNA^Ser_GCU, tRNA^Lys_UUU, tRNA^Ile_AAU, tRNA^Ile_UAU, tRNA^Met_CAU and tRNA^Asn_GUU―that are encoded in A. thaliana nuclear genome (Figure 1D). Electrophoretic mobility shift assay (EMSA) confirmed an interaction between AtKEOPS and each of these in vitro transcribed (IVT) tRNAs (Figure 1E). We then performed t⁶A assays using these IVT AttRNAs and analyzed t⁶A formation by LC–MS analysis (Supplementary Figure S1D). Normalized levels of t⁶A against A, U, C or G according to the nucleotide sequence showed great variations in t⁶A modification efficiencies of AtKEOPS towards these IVT AttRNAs (Figure 1F). AtKEOPS exhibits strong activities on tRNA^Arg_CCU and tRNA^Arg_UCU, and gradually lowering activities on tRNA^Thr_CGU, tRNA^Ile_AAU, tRNA^Lys_UUU and tRNA^Ser_GCU. In comparison, tRNA^Met_CAU, tRNA^Asn_GUU and tRNA^Ile_UAU were not t⁶A-modified by AtKEOPS. To exclude the possible problem in preparation of IVT tRNAs, we also deterimined different t⁶A-modification variations in these IVT AttRNAs by ScKEOPS (Supplementary Figure S1E and F). Our analysis suggests that the t⁶A modification frequencies are dependent on sequences of AttRNAs. In our case, AttRNA^Ile_UAU and AttRNA^Met_CAU lack a 36-UAA-38 motif (Figure 1D), which is a determinant of t⁶A modification (55). We mutated 36-UGA-38 motif of AttRNA^Ile_UAU (Figure 1F) into a 36-UAA-38 motif and generated a tRNA^Ile_UAU variant―AttRNA^Ile_UAU-G37A. Indeed, AttRNA^Ile_UAU-G37A is t⁶A-modified by AtKEOPS (Figure 1F). As for the inactivity of AttRNA^Asn_GUU, we confirmed the overall folding by CD spectra analysis (Supplementary Figure S1D). Promoted by a recent discovery of a novel function of P. salinus TsaN in generating t⁶ATP/t⁶ADP using TC-AMP and ATP/ADP (47), we also analyzed the formation of t⁶ATP/t⁶ADP by KEOPS via detecting t⁶A―the dephosphorylated products of t⁶ATP/t⁶ADP. Our LC–MS analysis shows that AtKEOPS is inactive in generating t⁶ATP/t⁶ADP in the presence of ATP and TC-AMP while P. salinus TsaN^1–392 (TsaD domain) catalyzes the formation of t⁶ATP/t⁶ADP in conjunction with AtYRDC (Supplementary Figure S1B).

As AtKEOPS possesses the highest activity on AttRNA^Arg_CCU, we set out to test the t⁶A-catalytic activities of AtKEOPS subcomplexes in assays using IVT AttRNA^Arg_CCU as substrate (Figure 1G and 1H). Our data demonstrates that KAE1–BUD32–PCC1 subcomplex is still capable of catalyzing the t⁶A formation of AttRNA^Arg_CCU, but the activity is decreased by around 50% compared that of KEOPS. KAE1–PCC1 subcomplex is no longer active in catalyzing the t⁶A modification of AttRNA^Arg_CCU (Figure 1F). In addition, we show that AtKEOPS still catalyzes t⁶A-modification of tRNA^Arg_CCUΔCCA in which the 3′ CCA end was chopped (Figure 1G and 1H). In sum, our data show that BUD32 is strictly needed by AtKEOPS to catalyze the biosynthesis of tRNA t⁶A while CGI121 is dispensable for in vitro tRNA t⁶A biosynthesis. In contrast, MjCgi121 directly binds tRNA via the 3′ CCA end and plays a pivotal role in tRNA binding and t⁶A-catalytic function of MjKEOPS (45).

Cryo-EM structure of A. thaliana KEOPS complex

We set out to determine the structure of AtKEOPS–tRNA^Arg_CCU by single particle cryo-electron microscopy (cryo-EM). We were not able to isolate a preformed AtKEOPS–tRNA^Arg_CCU complex by SEC (data not shown). Alternatively, a mixture of freshly purified AtKEOPS and AttRNA^Arg_CCU (at a molar ratio of 1:2) was applied to cryo-EM grids for structural analysis. We optimized the vitrification and collected a dataset of 14 622 micrographs on a Titan Krios microscope operated at 300 keV. We analyzed the complete dataset by applying manual picking and topaz training as well as iterative rounds of particle picking. With 232 232 picks, we determined a map of AtKEOPS at an overall resolution of 3.2–5.5 Å (Supplementary Figure S2 and Supplementary Table S3). We retrieved prediction models of AtKEOPS subunits from AlphaFold Protein Structure Database (61) and generated fitting models for subcomplexes of KAE1–PCC1, KAE1–BUD32, BUD32–CGI121 and PCC1–PCC1 based on crystal structures of MjKae1–Pcc1 (44), MjKae1–Bud32 (3), ScBud32–Cgi121 (56) and P. furiosus (Pf) Pcc1–Pcc1 (44), respectively. The reconstruction shows six KEOPS subunits―KAE1, PCC1, PCC1, KAE1, BUD32 and CGI121, but no presence of a tRNA molecule (Figure 2A). The dissociation of BUD32–CGI121 from KAE1–PCC1 in one KEOPS protomer was probably due to the structural damage at air–water interface during sample preparation. Imposing local refinement (Supplementary Figure S2B and C), we managed to determine almost a complete structure of AtKEOPS complex with exception for the C-terminal tail of KAE1 (residues 341–353), N-terminal end (residues 1–13) and C-terminal tail (residues 222–226) of BUD32, and N-terminal end of PCC1 (residues 1–15). The map was blurred for two flexible loops (residues 36–48 and 173–179) of KAE1, which were modelled with references to crystal structures of its orthologs (Supplementary Figure S3A). We modelled a Fe²⁺ ion in the binding sphere of H113–H117–D298 motif in KAE1 based on crystal structures of Kae1/OSGEP proteins (3,18,52). The final model is composed of a complete AtKEOPS protomer (PCC1–KAE1–BUD32–CGI121) and an incomplete protomer that only contains KAE1–PCC1 (Figure 2A). Based on the dimeric state of AtKEOPS complex in solution as determined by biochemical characterizations (Figure 1B and Supplementary Figure S1C), we aligned the structure of a complete KEOPS protomer to the structure of KAE1–PCC1 and completed a model of BUD32–CGI121 in relation to KAE1–PCC1 in the other protomer. In the reconstructed model of AtKEOPS dimer, the two protomers interact via an interacting interface of PCC1 dimer and adopt a V-shaped architecture (Figure 2B).

Figure 2.

Cryo-EM structure of A. thaliana KEOPS complex. (A) Cartoon representation of the cryo-EM structure of A. thaliana KEOPS complex (KAE1–PCC1–PCC1–KAE1–BUD32–CGI121) associated with cryo-EM map that is represented in transparent surface. PCC1, KAE1, BUD32 and CGI121 are color-coded in purple, red, blue and orange, respectively. (B) Reconstructed model of A. thaliana KEOPS dimer. BUD32–CGI121 from protomer 1 was modelled to complete protomer 2 via structural alignment of PCC1–KAE1. (C) Structural comparison of S. cerevisiae KEOPS (Gon7–Pcc1–Kae1–Bud32–Cgi121), H. sapiens KEOPS (GON7–LAGE3–OSGEP–PRPK–TPRKB) and A. thaliana KEOPS (PCC1–KAE1–BUD32–CGI121). Crystal structures of Bud32–Cgi121 (PDB: 4WW9)/PRPK–TPRKB (PDB: 6WQX) and Gon7–Pcc1 (PDB: 4WX8)/GON7–LAGE3–OSGEP (PDB: 6GWJ) were aligned to structures of their orthologs in A. thaliana KEOPS complex. Structures of Gon7/GON7 at the extremity are represented in cylindrical helices. (D) Structural alignment of KAE1 in A. thaliana KEOPS complex (PCC1–KAE1–BUD32–CGI121) and Kae1 in archaean KEOPS complex (Pcc1–Kae1–Bud32–Cgi121), of which a composite model was generated using crystal structure of M. jannaschii Kae1–Bud32–Cgi121 (PDB: 3ENH) and crystal structure of M. jannaschii Kae1–P. furiosus Pcc1 (PDB: 5JMV). Two different conformations of the C-lobes and C-terminal tails of A. thaliana BUD32 and M. jannaschii Bud32 are indicated with arrows. (E) A model of A. thaliana KEOPS–tRNA complex was generated by structural juxtaposition of A. thaliana KEOPS dimer and archaean KEOPS–tRNA, of which the model was generated using a composite model of archaean KEOPS complex and a crystal structure of M. jannaschii Cgi121–tRNA^Lys_UUU (PDB: 7KJT). The model of archaean KEOPS (D) was omitted for clarity. D stem loop (DSL), anticodon stem loop (ASL), TψC stem loop and 3′ CCA end of tRNA are labeled.

The structural uniformity and variation of KEOPSs

KEOPSs function as molecular machineries (31). Up to date, no atomic structures of a complete KEOPS complex are available except for a composite model of KEOPS from archaea (3) and crystal structures of KEOPS subcomplexes from yeast (56) and humans (18,57). We compared these KEOPS structures with AtKEOPS to gain insights into functions. Overall, AtKEOPS subunits manifest a substantial structural conservation to their orthologs from archaea, yeast and humans (Supplementary Figure S3). RMSD values of structural alignment coupled sequence identities are summarized in Table 1. Crystal structures of HsLAGE3–OSGEP binary complex (18) and HsPRPK–TPRKB binary complex (57) are perfectly superposed with structure of AtKEOPS (Figure 2C). Likewise, crystal structures of ScPcc1 (56) and ScBud32–Cgi121 binary complex (56) are satisfactorily aligned with structure of AtKEOPS (Figure 2C) except for Kae1 whose structure has not been determined. As for archaean KEOPS, crystal structures of PfPcc1–MjKae1 binary complex and MjBud32–Cgi121 binary complex are also satisfactorily superposed to PCC1–KAE1 and BUD32–CGI121 in the cryo-EM structure of AtKEOPS, respectively (Supplementary Figure S3E). In sum, the overall arrangement and inter-subunit interfaces of KEOPS subunits are extremely conserved among KEOPSs. However, structural juxtaposition of AtKEOPS and MjKEOPS as a whole reveals a marked variation in the C-lobes of BUD32 and Bud32 (Figure 2D). In general, the C-lobe of AtBUD32 is located farther away (at least 4 Å) than that of MjBud32 from the C-terminal domain of KAE1. Consequently, C-lobe makes fewer contacts with KAE1 (Figure 2D). Superposition of AtPCC1 dimer and PfPcc1 dimer exhibits conserved V-shaped architectures but varied conformations of AtKEOPS dimer and MjKEOPS dimer (Supplementary Figure S3F).

Table 1.

Statistics of the structural alignment coupled sequence identities for KEOPS subunits from A. thaliana, H. sapiens, S. cerevisiae and archaean species (M. jannaschii, P. abyssi, T. acidophilum and P. furiosus)

	Comparing protein (PDB ID)	Sequence identity (%)	RMSD (Å)
KAE1	H. sapiens OSGEP (6GWJ)	68.95	1.074 (379 Cα)
	M. jannaschii Kae1 (5JMV)	44.19	1.236 (288 Cα)
	P. abyssi Kae1 (2IVP)	46.91	1.248 (292 Cα)
	T. acidophilum Kae1 (3ENO)	41.31	1.515 (297 Cα)
BUD32	H. sapiens PRPK (6WQX)	43.80	1.852 (178 Cα)
	S. cerevisiae Bud32 (4WW9)	28.31	1.689 (151 Cα)
	M. jannaschii Bud32 (3ENH)	30.85	1.996 (154 Cα)
CGI121	H. sapiens TPRKB (6WQX)	29.65	1.208 (127 Cα)
	S. cerevisiae Cgi121 (4WW9)	16.86	1.471 (116 Cα)
	M. jannaschii Cgi121 (7KJT)	13.10	3.026 (119 Cα)
PCC1	H. sapiens LAGE3 (6GWJ)	33.33	1.057 (73 Cα)
	S. cerevisiae Pcc1 (4WXA)	25.00	1.014 (60 Cα)
	P. abyssi Pcc1 (7A67)	15.85	1.469 (47 Cα)
	P. furiosus Pcc1 (5JMV)	14.63	1.611 (63 Cα)

Unfortunately, tRNA^Arg_CCU was not resolved in the cryo-EM structure of AtKEOPS. We made use of MjKEOPS–tRNA model to gain insights into the binding of tRNA to AtKEOPS. Structural juxtaposition of MjKEOPS–tRNA to AtKEOPS dimer generated a model for a dimer of AtKEOPS–tRNA (Figure 2E). According to this model, tRNA could be accommodated on an extended surface of the four subunits of AtKEOPS. According to the docked model, anticodon stem loop of tRNA is sandwiched between KAE1 and PCC1; D stem loop protrudes between KAE1 and BUD32; 3′ CCA end extends to nucleotides-binding pocket of CGI121. Though steric clashes in some regions are evident, e.g. KAE1 and anticodon stem loop, this model might represent a general binding orientation of tRNA to AtKEOPS based on MjKEOPS–tRNA model (45).

KEOPS dimer is required to form a t⁶A-catalytic KEOPS–tRNA assembly

Based on high t⁶A-catalytic activity of AtKEOPS on AttRNA^Arg_CCU, we chose to use 5′ 6-FAM (6-Carboxyfluorescein)-labelled tRNA^Arg_CCU (5′-6FAM-tRNA^Arg_CCU) for interaction analysis. Our gel analysis shows a strong interaction between AtKEOPS and 5′-6FAM-tRNA^Arg_CCU (Figure 3A). In contrast, neither PCC1–KAE1 nor CGI121 is capable of binding 5′-6FAM-tRNA^Arg_CCU. In addition, our analysis revealed a relatively weak interaction between 5′-6FAM-tRNA^Arg_CCU and PCC1–KAE1–BUD32 or BUD32–CGI121 (Figure 3A). We further measured the binding affinities with microscale thermophoresis (MST) (Figure 3B) and determined equilibrium constants (K_d) for the interactions between 5′-6FAM-tRNA^Arg_CCU and AtKEOPS proteins (Table 2). We determined a K_d value of 18.08 μM for the interaction between 5′-6FAM-tRNA^Arg_CCU and AtKEOPS. In comparison, K_d values for PCC1–KAE1–BUD32 and BUD32–CGI121 towards 5′-6FAM-tRNA^Arg_CCU are 37.99 μM and 61.58 μM, respectively. Consistent with EMSA analysis, no binding events were observed with 5′-6FAM-tRNA^Arg_CCU and PCC1–KAE1 or CGI121 by MST measurements (Figure 3B). These binding data demonstrate that CGI121 does not bind tRNA but promotes the binding of tRNA to KEOPS whereas BUD32 is strictly needed for binding of tRNA to AtKEOPS.

Figure 3.

Characterization of the interaction between A. thaliana KEOPS and tRNA. (A) EMSA analysis of the interaction between 0.1 μM 5′-6FAM-tRNA^Arg_CCU and 0–0.5 μM A. thaliana KEOPS complex and subcomplex: KAE1–BUD32–PCC1 (KBP), KAE1–PCC1 (KP), BUD32–CGI121 (BC) and CGI121 (C). The gels show migration of 0.1 μM 5′-6FAM-tRNA^Arg_CCU in the absence of KEOPS proteins at indicated concentrations. (B) Microscale thermophoresis (MST) measurements of the interactions between 50 nM 5′-6FAM-tRNA^Arg_CCU and A. thaliana KEOPS proteins at indicated concentrations (97 nM–100 μM). KEOPS dimer, wild-type KEOPS; KEOPS monomer, P^{R70D/R73D/A74E}KBC. Error bars represent standard deviations for triplicate measurements. (C) A rigidly adjusted model of A. thaliana KEOPS–tRNA complex in which tRNA (colored in orange) from archaean KEOPS–tRNA was rotated by ∽15° along with structural alignment (shifted by ∽16 Å) of the C-terminal lobe of M. jannaschii Bud32 (represented in cylindrical helices) to that of A. thaliana BUD32. (D) Geometrical complementarity and electrostatic potential surface representation of A. thaliana KEOPS–tRNA model. (E) SEC (Superdex 200 Increase 10/300 GL, GE Healthcare) profiles of A. thaliana KEOPS dimer (P₂K₂B₂C₂), KEOPS monomer (P^{R70D/R73D/A74E}KBC) and S. cerevisiae KEOPS. The insert shows the interacting interface of A. thaliana PCC1 dimer and sites of mutation. (F) EMSA analysis of the interactions between 0.1 μM 5′-6FAM-tRNA^Arg_CCU and 0–0.4 μM AtKEOPS dimer or KEOPS monomer (P^{R70D/R73D/A74E}KBC). The upper panel shows the migration of 5′-6FAM-tRNA^Arg_CCU and the lower panel shows 10-fold amount of the original protein input in each lane as visualized by a separate SDS–PAGE. (G) LC–MS analysis of the time-course formation of t⁶A in assay that contained 2 μM A. thaliana KEOPS dimer or KEOPS monomer, 5 μM A. thaliana YRDC and 20 μM IVT AttRNA^Arg_CCU. The assay contained 4 mM L-threonine, 20 mM NaHCO₃ and 2 mM ATP.

Table 2.

Summary of the equilibrium dissociation constants (K_d) for the binding of 5′-6FAM-tRNA^Arg_CCU to A. thaliana KEOPS proteins as determined by microscale thermophoresis (MST). K, KAE1; B, BUD32; C, CGI121; P, PCC1; dimer, wild-type KEOPS complex (P₂K₂B₂C₂); monomer, KP^{R70D/R73D/A74E}BC; mutant-1, Y36R/I37K/T38R/P39R/P40R/G41W/H42D; mutant-2, G43E/F44R/L45K/P46R/R47D/E48K; n.d., not determined

KEOPS	K d (μM)	KAE1 mutants	K d (μM)	BUD32 mutants	K d (μM)
dimer	18.08 ± 3.87	KH117ABCP	53.75 ± 17.70	KBI47KCP	221.59 ± 216.78
monomer	74.37 ± 13.51	KD298RBCP	97.90 ± 39.83	KBK51ECP	29.82 ± 10.66
KBP	37.99 ± 13.56	KI17FBCP	315.36 ± 16.72	KBK55ECP	104.29 ± 15.22
KP	n.d.	KK202RBCP	111.31 ± 21.71	KBN58RCP	9.81 ± 2.73
BC	61.58 ± 15.02	KR284CBCP	128.18 ± 5.68	KBT162RCP	10.47 ± 2.99
CGI121	n.d.	KA231GBCP	138.89 ± 51.80	KBS163RCP	329.68 ± 11.10
KBCPR73D	352.7 ± 39.96	KY305ABCP	71.18 ± 13.80	KBL165KCP	302.42 ± 200.43
–	–	Kmutant-1BCP	5.27 ± 1.20	KBR220stopCP	148.39 ± 26.97
–	–	Kmutant-2BCP	n.d.	KBR220ACP	88.20 ± 15.58

Our interaction analysis demonstrates a central role of BUD32 in tRNA binding. However, the model of AtKEOPS–tRNA does not support a direct contact between BUD32 and tRNA (Figures 2E and 3C). In contrast, the model of MjKEOPS–tRNA coupled mutational validations demonstrated that the C-lobe of MjBud32 interacts with D stem loop and amino acid acceptor stem of MjtRNA^Lys_UUU (45). We presumed that the C-lobe of AtBUD32 is also primarily involved in binding of tRNA. We manually adjusted the position of tRNA by means of aligning the C-lobe of MjBud32 in rigid complex with MjtRNA^Lys_UUU to the C-lobe of AtBUD32, leading to an adjusted model of AtKEOPS–tRNA (Figure 3C). In this model, tRNA was rotated by ∽15° and shifted by ∽16 Å towards the C-lobe of AtBUD32, giving better geometrical complementarity and electrostatic interactions (Figure 3D). Such a manual rotation of MjtRNA^Lys_UUU from crystal structure of MjCgi121–tRNA^Lys_UUU was also performed to fit the geometry of MjKEOPS (45). It further suggests that assembly of KEOPS–tRNA complex might involve large conformational changes in KEOPS subunits and tRNA.

We set out to find out whether AtKEOPS dimer is needed for tRNA t⁶A biosynthesis. To disrupt the PCC1 dimer, we simultaneously mutated the Arg70, Arg73 and Ala74 of PCC1 (Figure 3E) and generated an AtKEOPS variant―PCC1^{R70D/R73D/A74E}–KAE1–BUD32–CGI121 (P^{R70D/R73D/A74E}KBC) (Supplementary Figure S4A and B). SEC analysis demonstrates that P^{R70D/R73D/A74E}KBC exists exclusively as a monomer with an elution volume roughly overlapping with that of the five-subunit ScKEOPS (Figure 3E). Hereafter, P^{R70D/R73D/A74E}KBC is dubbed as AtKEOPS monomer and the dimer refers to wild-type AtKEOPS. The interaction between AtKEOPS monomer and 5′-6FAM-tRNA^Arg_CCU becomes apparently weaker than that of AtKEOPS dimer and 5′-6FAM-tRNA^Arg_CCU (Figure 3F). MST measurements determined a K_d value of 74.37 μM for the interaction between AtKEOPS monomer and 5′-6FAM-tRNA^Arg_CCU (Figure 3B). Our LC–MS analysis shows that AtKEOPS monomer is no longer active in catalyzing t⁶A biosynthesis (Figure 3G). To find out whether the loss of t⁶A-ctalytic activity is due to disruption of dimer but not the loss of Arg70, Arg73 or Ala74 (as functional sites), we mutated Arg73 of PCC1 and generated an AtKEOPS variant―PCC1^R73D–KAE1–BUD32–CGI121 (P^R73DKBC). SEC analysis shows that P^R73DKBC elutes out between AtKEOPS dimer and monomer (Supplementary Figure S4A and B), suggesting that P^R73DKBC exists as mixture of dimer and monomer. Another explanation of the SEC profile is that the R73D mutation possibly induces large conformational changes in the dimeric architecture. EMSA analysis and MST measurements exhibit a significantly weaker interaction (K_d= 352.7 μM) between P^R73DKBC and 5′-6FAM-tRNA^Arg_CCU (Supplementary Figure S4C and D). However, P^R73DKBC still sustains around 25% t⁶A-catalytic activity of AtKEOPS dimer (Supplementary Figure S4E). In sum, our structural analysis and functional validation show that AtKEOPS dimer is needed to form a t⁶A-catalytic KEOPS–tRNA assembly.

Characterization KAE1 reveals regulatory sites related to t⁶A-catalytic activity of KEOPS

KAE1 adopts a typical two-subdomain fold that is conserved in all TsaD/Kae1/Qri7/OSGEP structures (Supplementary Figure S3A). The t⁶A-catalytic site is located between the two subdomains and a divalent metal ion (Fe²⁺/Zn²⁺) is essential for TC-AMP binding and t⁶A catalysis (32,41,54). Based on the model of KEOPS–tRNA complex (Figure 3C) and loss-of-function mutations in HsOSGEP, we chose to characterize the functional sites of AtKAE1 (Figure 4B): (i) an extremely conserved H–H–D motif (H117 and D298) that coordinates the metal ion Fe²⁺ (32); (ii) I17, K202 and R284 that are equivalent to missense mutations (I14F, K198R, R280C) of HsOSGEP in GAMOS patients (16); (iii) non-conserved A231 and Y305 in the vicinity of the catalytic site of KAE1. We generated these mutations and purified corresponding KEOPS variants: K^H117ABCP, K^D298RBCP, K^I17FBCP, K^K202RBCP, K^R284CBCP, K^A231GBCP and K^Y305ABCP (Supplementary Figure S5A and B). Our gel interaction analysis shows that 5′-6FAM-tRNA^Arg_CCU is almost completely bound to these KEOPS variants when mixed at a molar ratio of 1:5 (Supplementary Figure S5C). However, quantitative analysis by MST reveals that the interactions between 5′-6FAM-tRNA^Arg_CCU and these KEOPS variants are weaker than that between 5′-6FAM-tRNA^Arg_CCU and wild-type KEOPS (Figure 4C), as indicated by increased K_d values for these KEOPS variants (Table 2). Notably, I17F severely interferes with the binding of tRNA to KEOPS. We performed assays on t⁶A catalysis of these KEOPS variants (Supplementary Figure S5E and F) and show different effects of these mutations on t⁶A-catalytic activity of KEOPS (Figure 4D): 1) K202R mutation confers less effect on t⁶A-catalytic activity of KEOPS; 2) either I17F mutation or A231G mutation results in half of the t⁶A-catalytic activity; 4) H117A, D298R or R284C leads to dead t⁶A-catalytic activity; 5) Y305A mutation also causes a dead t⁶A-catalytic activity as only trace amount of t⁶A was detected. In light of the structural model (Figure 4B), our analysis of the interaction and t⁶A-catalysis of these mutations of KAE1 confirms an essential role of H117 and D298 in t⁶A catalysis. Interestingly, no degradation of KAE1 was observed in K^H117ABCP and K^D298RBCP (Supplementary Figure S5B). Our analysis suggests that I17, Y305 and A231 might modulate the configuration of t⁶A-catalytic center whereas K202 might participate in tRNA binding and R284 might modulate the function of BUD32.

Figure 4.

Characterization of functional sites of KAE1. (A) Local sequence alignment of A. thaliana (At) KAE1, H. sapiens (Hs) OSGEP, S. cerevisiae (Sc) Kae1 and M. jannaschii (Mj) Kae1. (B) The sites of single-mutation (I17, H117, K202, A231, R284, D298 and Y305) and the connecting loop (residues 36–48) of α1 and β1 are labeled in the structure of KAE1 (colored in red). (C) Microscale thermophoresis (MST) measurements of the interactions between 50 nM 5′-6FAM-tRNA^Arg_CCU and AtKEOPS variants bearing mutations in KAE1. mutant-1, Y36R/I37K/T38R/P39R/P40R/G41W/H42D; mutant-2, G43E/F44R/L45K/P46R/R47D/E48K. Error bars represent standard deviations for triplicate measurements. (D) Comparison of t⁶A modification efficiencies of IVT AttRNA^Arg_CCU by AtKEOPS variants bearing mutations in KAE1 in conjunction with 2 μM AtYRDC.

The precise binding of anticodon stem loop of tRNA to catalytic sites of TsaD/Kae1/Qri7 proteins still remains undetermined and poses an obstacle in mechanistic understanding of KEOPS. Our model of KEOPS–tRNA shows that the connecting loop (residues 36–48) between β1 and α1 of KAE1 adopts a conformation that creates steric clashes with incoming anticodon stem loop of tRNA (Figure 3C). Sequence and structural alignment show that these connecting loops in eukaryotic Kae1/KAE1/OSGEP proteins are highly conserved (Figure 4A) and conformationally divergent (Supplementary Figure S3A). However, the functions of these loops of TsaD/Kae1/Qri7 family proteins remain unexplored. We generated two AtKEOPS variants with multiple mutations in the loop connecting β1 and α1 of KAE1―K^{Y36R/I37K/T38R/P39R/P40R/G41W/H42D}BCP (K^mutant-1BCP) and K^{G43E/F44R/L45K/P46R/R47D/E48K}BCP (K^mutant-2BCP). We substituted non-charged residues with positively-charged residues of arginine and lysine in these two mutants. In K^mutant-2BCP, we additionally substituted positively-charged Arg47 with Asp and negatively-charged Glu48 with Lys, respectively. Our purification and SEC analysis demonstrate that these multiple mutations do not affect solubility of KAE1 and overall structure of KEOPS (Supplementary Figure S5A and B). Our gel analysis shows that both K^mutant-1BCP and K^mutant-2BCP are capable of binding 5′-6FAM-tRNA^Arg_CCU (Supplementary Figure S5D). Notably, MST measurements demonstrate an enhanced interaction between K^mutant-1BCP and 5′-6FAM-tRNA^Arg_CCU (K_d= 5.27 μM) but no interaction between K^mutant-2BCP and 5′-6FAM-tRNA^Arg_CCU (Figure 4C). Both K^mutant-1BCP and K^mutant-2BCP lose the t⁶A-catalytic activity (Supplementary Figure S5E, Figure 4D), suggesting that the connecting loop between β1 and α1 of KAE1 is essentially required for tRNA t⁶A biosynthesis by KEOPS. In particular, our model of KEOPS–tRNA and interaction analysis suggest that the first half of the loop (residues 36–42) might participate in tRNA binding (Figure 4B), as K^mutant-1BCP has acquired increased affinity towards tRNA (Figure 4C).

KEOPS is modulated by BUD32 via the C-terminal tail and ATP to ADP hydrolysis

Our model of KEOPS–tRNA coupled interaction analysis features an essential role of BUD32 in direct binding of tRNA to KEOPS (Figure 3D, Figure 5A). The model suggests that the 3′ amino acid acceptor arm (nucleobases 66–71) of tRNA is located in close proximity to α1 and α5 of BUD32 (Figure 5A). To confirm the participation of α1 and α5 of BUD32 in tRNA binding, we generated KEOPS variants bearing mutations in α1 (I47K, K51E, K55E and N58R) and in the connecting loop of β8 and α5 (T162R, S163R and L165K) of BUD32. We purified these BUD32-mutated KEOPS variants―KB^I47KCP, KB^K51ECP, B^K55ECP, KB^N58RCP, KB^T162RCP, KB^S163RCP and KB^L165KCP (Supplementary Figure S6A and B). We first measured the interactions between 5′-6FAM-tRNA^Arg_CCU and KEOPS variants by EMSA and MST. EMSA analysis shows that K55E mutation severely affects the interaction between 5′-6FAM-tRNA^Arg_CCU and KEOPS; K51E, T162R or S163R mutation confers a milder effect; I47K, N58R or L165K mutation exerts no effect (Supplementary Figure S6C). However, quantitative analysis by MST demonstrates that I47K, K55E, S163R, L165K or K51E mutation leads to weaker interaction between 5′-6FAM-tRNA^Arg_CCU and KEOPS while N58R or T162R mutation slightly promotes the binding of 5′-6FAM-tRNA^Arg_CCU to KEOPS (Figure 5B, Table 2). It seems that substitution of Lys55 by negatively-charged Glu leads to markedly weak binding affinity of KEOPS towards 5′-6FAM-tRNA^Arg_CCU whereas substitution of Asn58 and Thr162 by positively-charged Arg enhances the binding affinities, suggesting that Lys55, Asn58 and Thr162 participate in tRNA binding. We then measured the t⁶A-catalytic activities of these KEOPS variants (Supplementary Figure S6D, Figure 5C) and demonstrate that K55E mutation substantially compromises t⁶A-cataalytic activity of KEOPS; K51E or T162R mutation mildly affects the t⁶A-catalytic activity; I47K, N58R, S163R or L165K mutation confers negligible effect (Figure 5C). Nonetheless, K55E mutation causes a drop in both binding affinity and t⁶A-catalytic activity, strongly documenting a role for Lys55 of BUD32 in tRNA binding.

Figure 5.

Functional modulation of A. thaliana KEOPS by BUD32. (A) The model of A. thaliana KEOPS–tRNA shows residues of BUD32 (colored in blue) in proximity to amino acid acceptor arm of tRNA. (B) Microscale thermophoresis (MST) measurements of the interactions between 50 nM 5′-6FAM-tRNA^Arg_CCU and KEOPS variants bearing mutations in BUD32. (C) Comparison of the t⁶A modification efficiencies of IVT AttRNA^Arg_CCU by KEOPS variants bearing mutations in BUD32. Error bars represent standard deviations for triplicate measurements. (D) The interacting interface of KAE1 (colored in red) and BUD32 (colored in blue) in the model of A. thaliana KEOPS–tRNA. ATP and Mg²⁺ ions were projected in the ATPase-catalytic site of BUD32 according to the crystal structures of H. sapiens PRPK (PDB: 6WQX) and S. cerevisiae Bud32 (PDB: 4WW9). The C-terminal tail (²²²RTMIG²²⁶) as represented in dashed cylinder was not observed in cryo-EM structure of A. thaliana KEOPS complex. tRNA A37 and Fe²⁺ ion indicate catalytic center of KAE1. (E) Local sequence alignment of the C-terminal tails of A. thaliana (At) BUD32, H. sapiens (Hs) PRPK, S. cerevisiae (Sc) Bud32 and M. jannaschii (Mj) Bud32. (F) MST measurements of the interactions between 50 nM 5′-6FAM-tRNA^Arg_CCU and AtKEOPS, AtKB^R220stopCP or KB^R220ACP. Error bars represent standard deviations for triplicate measurements. (G) Comparison of the t⁶A modification efficiencies of A. thaliana KEOPS, KB^R220stopCP or KB^R220ACP towards IVT AttRNA^Arg_CCU. (H) Hydrolysis rate of 2 mM ATP to ADP by 5 μM A. thaliana KEOPS complex, subcomplexes or variants in the presence or absence of 10 μM IVT AttRNA^Arg_CCU. (I) Kinetic plots of hydrolysis of ATP to ADP by 2 μM A. thaliana KEOPS, K^D298RP or KB^D137ACP in the presence or absence of 10 μM IVT AttRNA^Arg_CCU. (J) Comparison of the t⁶A modification efficiencies of A. thaliana KEOPS, KB^D137ACP or KB^D156ACP towards IVT AttRNA^Arg_CCU.

Our cryo-EM structure of AtKEOPS reveals distinct conformation of the C-lobe of BUD32 in relation to KAE1 compared to that of MjKEOPS. Strikingly, the C-terminal tail of BUD32 adopts a unique conformation (Figure 3C). In both structures, the C-terminal tails of BUD32/Bud32 protrude towards the catalytic site of KAE1/Kae1. However, the structure of the tail sequence (²²⁰RKRTMIG²²⁶) was not observed in our structure (Figure 5D) nor in structures of BUD32 orthologs (3,56,57). Sequence alignment reveals that the tail sequences are highly conserved in eukaryotic BUD32/PRPK/Bud32 proteins but less conserved in archaean Bud32, which lacks a MxG motif (Figure 5E). The C-terminal tails of BUD32/PRPK/Bud32 proteins contain a positively-charged RKR motif (Figure 5E). To analyze the function of the C-terminal tail of BUD32, we generated a series of KEOPS variants, including a variant lacking ²²⁰RKRTMIG²²⁶ (KB^R220stopCP) and eight variants with single mutation―KB^R220ACP, KB^K221ACP, KB^R222ACP, KB^T223ACP, KB^M224ACP, KB^I225ACP, KB^G226ACP and KB^G226RCP (Figure 5D). We purified good quality proteins of these KEOPS variants (Supplementary Figure S7A and B). Our EMSA analysis shows that the binding of 5′-6FAM-tRNA^Arg_CCU to KEOPS is not essentially affected or disrupted by deletion of ²²⁰RKRTMIG²²⁶ or single mutation in it (Supplementary Figure S7C). MST measurement demonstrates that the interaction between 5′-6FAM-tRNA^Arg_CCU and KB^R220ACP or KB^R220stopCP was weaker than wild-type KEOPS (Figure 5F). The K_d values for KEOPS, KB^R220stopCP and KB^R220ACP are 18.08 μM, 148.39 μM and 88.20 μM, respectively. We performed tRNA t⁶A assay using IVT AttRNA^Arg_CCU and these variants (Supplementary Figure S7D), and detected no t⁶A with KB^R220stopCP and only trace amount of t⁶A with KB^R220ACP, respectively (Figure 5G). In contrast, we detected large amount of t⁶A that is comparable to wild-type KEOPS in assays using KB^K221ACP, KB^M224ACP, KB^I225ACP, KB^G226ACP or KB^G226RCP. In parallel, we detected decreased levels of t⁶A in assays using KB^R222ACP or KB^T223ACP. These data demonstrate that Arg220 of BUD32 plays an essential role in regulating the t⁶A-catalytic activity of KEOPS.

Previous studies demonstrated that Bud32/PRPK regulates the function of KEOPS via an ATP to ADP hydrolysis-based mechanism (20,45). We first measured the ATPase activity of AtKEOPS complex and subcomplexes using an NADH-coupled ATPase assay. We determined an efficient conversion of ATP to ADP in the presence of KAE1–BUD32, PCC1–KAE1–BUD32 or KEOPS, but neither KAE1–PCC1 nor BUD32–CGI121 (Supplementary Figure S6E, Figure 5H). Comparison of the ATP hydrolysis rates (reaction velocity in steady state) demonstrate that BUD32 catalyzes the ATP hydrolysis and KAE1 stimulates the ATPase activity of BUD32, which is further potentiated by CGI121 (Figure 5H). Structural juxtaposition of AtBUD32 and HsPRPK in complex with ATP and Mg²⁺ projects ATP and two Mg²⁺ ions in the catalytic site of AtBUD32 (Figure 5D), which reveals that the conserved D137 and D156 might participate in coordination of ATP and Mg²⁺ (Figure 5D). We mutated D137 and D156 of BUD32 and generated two KEOPS variants―KB^D137ACP and KB^D156ACP (Supplementary Figure S6A and B). We show that KB^D137ACP and KB^D156ACP is no longer active in hydrolyzing ATP to ADP (Figure 5H). Meanwhile, KB^R220stopCP exhibits an ATPase activity that is comparable to the KEOPS (Figure 5H), suggesting that the C-terminal tail of BUD32 does not directly participate in ATP hydrolysis. We further determined that IVT AttRNA^Arg_CCU markedly potentiates the ATP hydrolysis by KEOPS (Figure 5H). We measured catalytic kinetics of ATP hydrolysis by 2 μM KEOPS (Figure 5I): V_max= 0.039 ± 0.0015 μM·s^–1, K_m= 0.95 ± 0.053 μM, K_cat= 0.019 ± 0.00073 s^–1, K_cat/K_m= 0.020 μM^–1·s^–1. In parallel, we determined catalytic kinetics of 2 μM KEOPS in the presence of 10 μM IVT AttRNA^Arg_CCU: V_max= 0.11 ± 0.00021 μM·s^–1, K_m= 1.22 ± 0.029 μM, K_cat= 0.053 ± 0.00011 s^–1, K_cat/K_m= 0.043 μM^–1·s^–1. The kinetic parameters indicate that tRNA potentiates hydrolysis of ATP to ADP by KEOPS but slightly reduces the binding affinity of ATP to KEOPS. Thus, it suggests that binding of tRNA to KEOPS directly induces a conformational change in the catalytic site of BUD32. We further determined that either KB^D137ACP or KB^D156ACP still sustains a very low level of t⁶A-catalytic activity (Supplementary Figure S6D and Figure 5J), suggesting that deprivation of the ATPase activity of BUD32 severely interferes with turnover of t⁶A-catalysis by KEOPS.

Proteomic analysis of KEOPS–interacting proteins

Recently, Daugeron et al. identified no orthologs of Gon7/Pcc2 in Arabidopsis thaliana using advanced comparative genomic analysis (46). Our in vitro assay also confirmed a t⁶A-catalytic activity of AtKEOPS on IVT tRNAs. Nonetheless, we performed proteomic analysis of KEOPS-interacting proteins using LC–MS/MS. We simply incubated purified KEOPS with total soluble proteins isolated from Arabidopsis thaliana seedlings and purified potential interactors of 6His tagged KEOPS by Ni-NTA chromatography. Unbound proteins were removed by repeated washing using lysis buffer until no proteins appeared in the wash fraction prior to final elution of KEOPS. SDS–PAGE analysis showed that the final elution sample contained four subunits of KEOPS and very weak bands of proteins (data not shown). The final elution sample were digested into tryptic peptides and applied to LC–MS/MS for proteomic analysis. In total, 671 proteins were identified in the MS samples and listed in order according to the intensity-based absolute quantification (iBAQ) values (Supplementary Table S4). The list contains 664 characterized proteins with functions implicated in a wide range of cellular processes and 7 uncharacterized proteins (Uniprot IDs: Q9FKA5, A0A1P8B3M2, Q9C6U3, O64818, Q9FZG2, Q93W28 and F4J0L7). Notably, Q9FZG2 has a molecular size of 10.67 kDa, which is close to that of GON7/Gon7. Interestingly, AlphaFold prediction structure of Q9FZG2 (entry: Q9FZG2, www.alphafold.ebi.ac.uk) shows that Q9FZG2 is an intrinsically disordered protein with a structured core comprising two antiparallel β-stands connected to an α-helix.

Discussion

Evolutionarily conserved KEOPS catalyzes the transfer of TC-moiety from TC-AMP to N6 atom of adenine at position 37 of tRNA, leading to tRNA t⁶A. Up to date, characterizations of KEOPSs (Methanococcus jannaschii, Saccharomyces cerevisiae and Homo sapiens) revealed structures, overall arrangement and functions of KEOPS subunits (31). However, the structure of a complete KEOPS and molecular workings of KEOPS still remain unresolved. In the present study, we determined a cryo-EM structure of KEOPS complex from Arabidopsis thaliana and performed structure–function relationship analysis of A. thaliana KEOPS. The data analysis and findings are discussed below.

We successfully purified recombinant KEOPS proteins and reconstituted a stable KEOPS complex that is composed of KAE1, BUD32, CGI121 and PCC1. We show that the four-subunit A. thaliana KEOPS is active in catalyzing tRNA t⁶A biosynthesis in conjunction with recombinant A. thaliana YRDC (Figure 1C). Though no ortholog of GON7/Gon7 in Arabidopsis thaliana has been identified using bioinformatic analysis (46), we identified a large number of potential KEOPS interactors using pull-down assay and LC–MS/MS proteomic analysis (Supplementary Table S4). Of note, the uncharacterized Q9FZG2 (At1g47820) and O64818 (At2g23090) possess GON7/Gon7 features―comparable molecular sizes and partial intrinsically disordered structures. The structures and functions of Q9FZG2 and O64818 remain to be experimentally determined. Moreover, the presence of other proteins in MS sample eluted with KEOPS suggests that KEOPS or subunits might interact weakly or transiently with identified proteins. However, such a hypothesis remains to be experimentally validated in future.

A number of studies demonstrated large variations in t⁶A modification frequencies of KEOPS towards different tRNA substrates (20,44,45,55). Our in vitro assay shows that the t⁶A-catalytic activities of KEOPS towards IVT tRNAs are different (Figure 1F). KEOPS exhibits very low t⁶A-catalytic activity on tRNA^Ser_GCU and no activity on tRNA^Asn_GUU, tRNA^Met_CAU and tRNA^Ile_UAU, of which the latter two substrates lack a 36-UAA-38 motif. Introduction of a 36-UAA-38 motif into tRNA^Ile_UAU restores the t⁶A modification by KEOPS (Figure 1F). As per the inactivity of tRNA^Asn_GUU and low activity tRNA^Ser_GCU, other native modifications might be needed to stabilize the active conformations of these two tRNAs for more efficient t⁶A modification. The other explanation is that the in vitro folding of IVT tRNAs does not adopt a t⁶A-productive conformation, though our SEC, CD and interaction analysis shows that these IVT tRNAs adopt comparable overall 3D structures.

We determined a K_d value of 18 μM for the binding of KEOPS with IVT tRNA^Arg_CCU and attempted to determine a cryo-EM structure of KEOPS–tRNA^Arg_CCU. Unfortunately, it turned out that tRNA was absent in the cryo-EM structure. Moreover, BUD32–CGI121 of one protomer of the KEOPS dimer was not observed in the structure (Figure 2A). We presume that the interaction between KEOPS and tRNA^Arg_CCU was too weak to form a stable complex in the EM sample, as we were not able to purify a KEOPS–tRNA^Arg_CCU by SEC. In another scenario, a fraction of tRNA^Arg_CCU bound to BUD32–CGI121 that was dissociated from KEOPS in the EM sample, as the K_d for the interaction between BUD32–CGI12 and tRNA^Arg_CCU is 62 μM. Our cryo-EM structure of AtKEOPS complex reveals a conserved overall arrangement of KEOPS subunits (31). The model of MjKEOPS–tRNA provides a good framework for mechanistic understanding of KEOPS (45). It shows that tRNA binds to an extended surface of KEOPS and makes contacts with the four subunits (31,45). We generated a rigid model for A. thaliana KEOPS–tRNA by means of structural juxtaposition of AtKEOPS and MjKEOPS–tRNA (Figure 2E). We further manually refined the geometric and electrostatic fitting of tRNA to KEOPS with reference to conformation of the C-lobes of MjBud32/AtBUD32, which is directly involved in tRNA binding (45). The resulting model of KEOPS–tRNA reveals a good geometric complementarity and electrostatic interaction (Figure 3D). The model features a central role of BUD32 in tRNA binding. Our mutational analysis of BUD32 supports a direct participation of α1, on which the K55E mutation leads to markedly weak interaction between KEOPS and tRNA^Arg_CCU (Figure 5B) and decreased t⁶A-catalytic activity of KEOPS (Figure 5B). In addition, a number of other mutations in the connecting loop of β8 and α5 of BUD32 also interferes with the interaction between KEOPS and tRNA^Arg_CCU. Moreover, we show that the loop connecting β1 and α1 of KAE1 participates in binding of tRNA to KEOPS (Figure 4B). We determined that CGI121 or PCC1–KAE1 is not capable of binding tRNA but BUD32–CGI121 and KAE1–BUD32–PCC1 can independently bind tRNA (Figure 3A and B). Subtraction of CGI121 from KEOPS or deletion of 3′ CCA end of tRNA^Arg_CCU only reduces the t⁶A-catalytic activity of KEOPS (Figure 1H). In contrast, MjCgi121 alone or MjBud32–Cgi121 is capable of binding tRNA but MjKae1–Bud32–Pcc1 does not bind tRNAs (45). PaPcc1–Kae1 is a binding core for tRNAs and PaCgi121 alone does not bind tRNAs (20). The functional differences among different KEOPS proteins imply specific mechanisms underscoring the molecular interaction and regulation of KEOPS–tRNA assembly. However, mutational analysis of the interactions and t⁶A-catalytic activities justifies at least the binding orientation of tRNA to KEOPS in our model of AtKEOPS–tRNA.

Our cryo-EM structure of A. thaliana KEOPS revealed a dimeric structure of KEOPS, which is consistent with the dimeric state of AtKEOPS as determined by SEC. AtKEOPS dimer is mediated by homodimerization of PCC1 in a similar manner as for archaean KEOPS (Supplementary Figure S3E) (44). We generated a four-subunit KEOPS monomer (P^{R70D/R73D/A74E}KBC) via disrupting the dimerization of PCC1. KEOPS monomer is capable of binding tRNA (Figure 3F). However, the binding affinity for KEOPS monomer and tRNA^Arg_CCU (K_d= 74 μM) is close to that for BUD32–CGI121 and tRNA^Arg_CCU (K_d= 62 μM), which are greatly lower than the binding affinity of wild-type KEOPS dimer towards tRNA^Arg_CCU (Table 2). Nonetheless, KEOPS monomer is no longer active in biosynthesizing tRNA t⁶A (Figure 3G). Our model of AtKEOPS–tRNA shows that one KEOPS protomer could binds one tRNA, whose anticodon stem loop potentially makes contacts with PCC1–KAE1 from the other protomer (Figure 2E, Figure 3D). Therefore, we presume that correct binding of tRNA to one KEOPS protomer is facilitated by the other protomer. However, we cannot exclude the possibility that KEOPS dimer binds only one molecule of tRNA. In this case, binding of tRNA to one protomer precludes the binding of second tRNA to the other protomer. Such a similar mechanism is adopted by mitochondrial Qri7 dimer (34) and bacterial TsaD2–TsaB2 tetramer (38). Unfortunately, we could not determine the stoichiometry and number of binding sites by MST and EMSA measurements. In sum, our data shows that AtKEOPS dimer is required to form a t⁶A-catalytic KEOPS–tRNA assembly.

ATP to ADP hydrolysis by BUD32 is involved in the turnover of t⁶A-catalysis by KEOPS. Such a mechanism is also adopted by bacterial TsaDBE complex (38,41). The ATPase catalytic site is sandwiched between the N-lobe and C-lobe of Bud32/PRPK (56,57). We show that ATP to ADP hydrolysis by BUD32 is stimulated by KAE1 and further strongly potentiated by tRNA^Arg_CCU (Figure 5H). Deletion of C-terminal tail or abrogation of ATPase activity of AtBUD32 does not interfere with binding of tRNA to AtKEOPS but leads to dead t⁶A-catalytic activity. But how the C-terminal tail and ATPase activity of BUD32 affect the t⁶A activity of KEOPS remains a long-standing question. Here we determined that Arg220 at the C-terminal tail of BUD32 is critical for the t⁶A-catalytic activity of Kae1. Our structure shows a distance of ∽25 Å between Arg220 of BUD32 and t⁶A-catalytic site of KAE1. Therefore, it's unlikely that Arg220 of BUD32 reaches into the t⁶A-catalytic site of KAE1 and modulates the TC-transfer reaction. We presume that Arg220 or the positively charged C-terminal tail of AtBUD32 might play a pivotal role in stabilizing anticodon stem loop of tRNA. In structure of AtKEOPS, the closest distance between α7 of KAE1 and ATPase catalytic site of BUD32 is less than 8 Å. We hypothesize that α7 of KAE1 interacts with α1 and α5 of BUD32 and stimulates the ATPase activity of BUD32. In turn, the hydrolysis of ATP to ADP drives the relocation of α7 of KAE1. By doing so, the two lobes of BUD32 undergo substantial conformational changes between the ATP-bound state and ADP-bound state. Movement of the C-terminal tail along with the relocation of the C-lobe of BUD32 might participate in binding and release of tRNA. Therefore, the C-terminal tail of BUD32 offers an opportunity to couple ATP hydrolysis with correct binding of tRNA to KEOPS, conferring an allosteric regulation of t⁶A-catalytic cycle of KEOPS.

Conclusion

We have determined a conserved role of A. thaliana KEOPS in tRNA t⁶A biosynthesis. Our structure–function relationship analysis of KEOPS–tRNA assembly suggests that the four-subunit KEOPS dimerize via PCC1 in support of a correct binding of tRNA to KEOPS, which requires the essential contribution of BUD32 and is further promoted by CGI121. BUD32 seems to be key regulator of KEOPS. The C-terminal tail of BUD32 functions as a ‘trigger’ to modulate the t⁶A-catalytic activity of KAE1. KAE1 stimulates the ATPase activity of BUD32 in exchange for turnover of the t⁶A-catalysis, which is driven by ATP to ADP hydrolysis by BUD32. Still, there are several fundamental questions to be addressed in order to mechanistically understand the inner workings and cellular roles of KEOPS machineries. In particular, the specific recognition of ANN-decoding tRNAs by KEOPSs and the chemistry of t⁶A catalysis by Kae1/KAE1/OSGEP await to be fully elucidated. A high-resolution structure of a complete KEOPS–tRNA complex is undoubtedly desirable to solve these mysteries.

Data availability

The cryo-EM electron density map and the atomic coordinates of A. thaliana KEOPS model have been deposited in the Protein Data Bank (https://www.wwpdb.org) under the EMDB ID: EMD-36808 and PDB ID: 8K20.

Supplementary data

Supplementary Data are available at NAR Online.

Funding

Science Fund for Distinguished Young Scholars of Gansu Province, China [23JRRA1018]; National Natural Science Foundation of China [32000847]; Fundamental Research Funds for Central Universities of Lanzhou University [lzujbky-2023-26]. Funding for open access charge: Science Fund for Distinguished Young Scholars of Gansu Province.

Conflict of interest statement. None declared.