Reversible RNA phosphorylation stabilizes tRNA for cellular thermotolerance

Post-transcriptional modifications have critical roles in tRNA stability and function1–4. In thermophiles, tRNAs are heavily modified to maintain their thermal stability under extreme growth temperatures5,6. Here we identified 2′-phosphouridine (Up) at position 47 of tRNAs from thermophilic archaea. Up47 confers thermal stability and nuclease resistance to tRNAs. Atomic structures of native archaeal tRNA showed a unique metastable core structure stabilized by Up47. The 2′-phosphate of Up47 protrudes from the tRNA core and prevents backbone rotation during thermal denaturation. In addition, we identified the arkI gene, which encodes an archaeal RNA kinase responsible for Up47 formation. Structural studies showed that ArkI has a non-canonical kinase motif surrounded by a positively charged patch for tRNA binding. A knockout strain of arkI grew slowly at high temperatures and exhibited a synthetic growth defect when a second tRNA-modifying enzyme was depleted. We also identified an archaeal homologue of KptA as an eraser that efficiently dephosphorylates Up47 in vitro and in vivo. Taken together, our findings show that Up47 is a reversible RNA modification mediated by ArkI and KptA that fine-tunes the structural rigidity of tRNAs under extreme environmental conditions.

Post-transcriptional modifications have critical roles in tRNA stability and function [1][2][3][4] . In thermophiles, tRNAs are heavily modified to maintain their thermal stability under extreme growth temperatures 5,6 . Here we identified 2′-phosphouridine (U p ) at position 47 of tRNAs from thermophilic archaea. U p 47 confers thermal stability and nuclease resistance to tRNAs. Atomic structures of native archaeal tRNA showed a unique metastable core structure stabilized by U p 47. The 2′-phosphate of U p 47 protrudes from the tRNA core and prevents backbone rotation during thermal denaturation. In addition, we identified the arkI gene, which encodes an archaeal RNA kinase responsible for U p 47 formation. Structural studies showed that ArkI has a non-canonical kinase motif surrounded by a positively charged patch for tRNA binding. A knockout strain of arkI grew slowly at high temperatures and exhibited a synthetic growth defect when a second tRNA-modifying enzyme was depleted. We also identified an archaeal homologue of KptA as an eraser that efficiently dephosphorylates U p 47 in vitro and in vivo. Taken together, our findings show that U p 47 is a reversible RNA modification mediated by ArkI and KptA that fine-tunes the structural rigidity of tRNAs under extreme environmental conditions. Recent advances in epitranscriptomics research have demonstrated the chemical diversity and biological importance of RNA modifications [1][2][3][4] . Thus far, about 150 types of RNA modification have been reported in various RNA molecules from all domains of life 7 . In particular, tRNAs contain the widest variety and largest number of modified nucleosides, with 80% of RNA modifications identified in tRNA molecules. Diverse RNA modifications are clustered in the anticodon loop, especially at positions 34 and 37 (refs. 1,8 ). These modifications have critical roles in stabilizing and modulating codon-anticodon interactions on the ribosome, ensuring accurate and efficient protein synthesis. Many RNA modifications are also found in the tRNA body composed of the D-loop, TΨC loop (T-loop) and variable loop (V-loop) 9,10 (Fig. 1a). These RNA modifications are required for correct folding and stability of the tRNA core structure. In particular, 2′-O-methyl modifications (Nm) confer conformational rigidity to the tRNA core region by fixing C3′-endo ribose puckering 9,11 .
In thermophilic bacteria and archaea, unique RNA modifications contribute to the thermal adaptation of tRNAs 5,6 . 5-Methyl-2-thiouridine (m 5 s 2 U or s 2 T) is found at position 54 in the T-loop of tRNAs from thermophiles 12 . m 5 s 2 U54 adopts a rigid conformation with C3′-endo ribose puckering, thereby stabilizing the tRNA body in high-temperature environments 11 . The 2-thiolation level of m 5 s 2 U54 increases as the growth temperature rises 13,14 . m 5 s 2 U54 contributes to the thermotolerance of Thermus thermophilus 15 . In Pyrococcus furiosus, the relative abundance of N 4 -acetylcytidine (ac 4 C) and its 2′-O-methyl derivative (ac 4 Cm) were markedly increased with rising growth temperature 14 . ac 4 C is a prevalent modification that is present in tRNAs, rRNAs and other RNAs in hyperthermophilic archaea 14,16 . ac 4 C favours the C3′-endo form and stabilizes tRNAs 17,18 . Loss of ac 4 C results in a growth defect in Thermococcus kodakarensis at high temperature, contributing to cellular thermotolerance 19 . In Bacillus stearothermophilus, 2′-O-methylation in tRNAs increases when the growth temperature rises 20 . Archaeosine (G + ) is a unique 7-deazaguanosine derivative found at position 15 in the D-loop of archaeal tRNAs 21 . On the basis of quantum mechanics calculations, G + 15 stabilizes the Levitt base pair with C48 (ref. 22 ). In line with this, biochemical and genetic studies have shown that G + confers thermal stability to tRNAs and contributes to thermotolerance 19,23 .
Here we report the identification of 2′-phosphouridine (U p ) in tRNAs, which, to our knowledge, is the first known instance of internal RNA phosphorylation. Biochemical, structural and genetic studies showed that U p 47 is a reversible RNA modification and confers thermal stability to tRNA, thereby contributing to cellular thermotolerance.
Nature | Vol 605 | 12 May 2022 | 373 First, we comprehensively analysed all post-transcriptional modifications of tRNA Val3 by mass spectrometry (RNA-MS) 25 Table 1). Among the modifications, we detected an unknown uridine derivative with molecular mass of 324 (tentatively named N 324 ) at position 47 of the RNA fragments digested with RNases (Fig. 1b, c). The relative intensity of the mass chromatograms showed that N 324 occurred at a frequency of 96.8% (Fig. 1d), indicating that N 324 is an abundant modification. We also detected N 324 in the seven other tRNA species (Extended Data Fig. 3a, b, Supplementary Note 1, Supplementary Table 2), indicating that N 324 is a prevalent and abundant (82-100%) modification in class I tRNAs bearing U47 in the V-loop, but not in class II tRNAs with a long V-loop (Extended Data Fig. 3a). We also detected N 324 in tRNA precursors (Extended Data Fig. 3c, Supplementary Note 1).
High-resolution mass analysis of the N 324 -containing fragment showed that the additional mass of N 324 attached to the uridine residue was 79.97067 Da, equivalent to one phosphate group (theoretical mass, 79.96632 Da), with a low error value of 4.4 millimass unit , indicating that N 324 is a phosphorylated uridine residue. This prediction explains why the N 324 nucleoside was not detected in our nucleoside analysis (Extended Data Fig. 2a), owing to N 324 being dephosphorylated during nucleoside preparation. To determine the phosphorylation site of N 324 , we prepared the N 324 -containing nucleotide and analysed its chemical structure through collision-induced dissociation (CID) and biochemical approaches (Supplementary Note 2, Extended Data Fig. 4a-h). We found that phosphorylation occurs on the 2′-OH group of the ribose moiety and concluded that N 324 is 2′-phosphouridine (denoted U p , where 'p' is superscript to discriminate it from 3′-phosphate) (Fig. 1e).

U p 47 stabilizes tRNA structure
Given that U p 47 is a thermophile-specific modification found in the tRNA core region, we investigated whether U p 47 stabilizes the tertiary structure of tRNA. To this end, we treated S. tokodaii tRNA Val3 with yeast Tpt1p (2′-phosphotransferase) to remove the 2′-phosphate of U p 47. We measured the melting temperature (T m ) of S. tokodaii tRNA Val3 with and without U p 47 (Fig. 1f). In the melting curves, the tRNA without U p 47 gradually melted at around 65 °C while its hyperchromicity increased with temperature, whereas the tRNA with U p 47 remained stable even at 70 °C. The T m values of the tRNA with and without U p 47 were 85.8 ± 0.5 °C and 79.2 ± 0.5 °C, respectively. These observations clearly demonstrate that a single U p 47 modification increases the thermal stability of tRNA Val3 by 6.6 °C.
We next performed an RNase probing experiment to examine the nuclease resistance of tRNA with and without U p 47. S. tokodaii tRNA Val3 and its Tpt1p-treated form were labelled with 32 P at their 3′ termini and were probed with RNase I at 65 °C (Fig. 1g)  Article the intact tRNA with U p 47, the Tpt1p-treated tRNA was degraded more rapidly, within 5 min, indicating that the tRNA without U p 47 was highly sensitive to RNase I. This observation demonstrates that U p 47 stabilizes and protects tRNAs from nucleolytic degradation.

Structural study of U p 47 in native tRNA
To determine the molecular basis for thermal stabilization of tRNA by U p 47, we crystalized S. tokodaii tRNA Val3 and determined its atomic structure at a resolution of up to 1.9 Å by X-ray crystallography (Fig. 2a Supplementary Fig. 2). One unit cell contains two tRNA molecules, denoted as molecule A and molecule B. Molecule A formed a canonical tRNA core structure (Fig. 2b, c, Extended Data Fig. 5a), whereas molecule B had an altered core structure with a non-canonical base triple (Fig. 2b, c, Extended Data Fig. 5b). We clearly observed electron densities for tRNA modifications, including U p 47 (Fig. 2a, Extended Data Fig. 5c, d). In both molecules, the 2′-phosphate of U p 47 was oriented towards the solvent side and did not interact with any residues (Fig. 2a-c). The ribose puckering of U p 47 adopted a C2′-endo conformation (Supplementary Table 3), as observed in the synthetic nucleotide 28 . The O4′ position in the ribose of U p 47 formed a hydrogen bond with the N 6 -amino group of A21 in both molecules (Fig. 2d). The uracil base of U p 47 faced the tRNA core (Fig. 2b, d). Because the uracil base at position 47 favours an outward orientation, as observed in well-known structures of yeast tRNA Phe and other class I tRNAs 29 , we observed backbone rotation of the V-loop at positions 46-48 caused by U p 47 (Fig. 2e, Extended Data Fig. 6a, b). When U p 47 was virtually introduced to yeast tRNA Phe , the 2′-phosphate clashed with T-stem residues at positions 49 and 50, inducing backbone rotation of the V-loop that orients the uracil base inwards and the 2′-phosphate outwards (Fig. 2e). In this rotation from yeast tRNA Phe to molecule A (Fig. 2b, c, e, Extended Data Fig. 6b), G46 changed its ribose pucker from C2′-endo to C3′-endo with altered torsion angles (δ, ε and ζ were changed by -58°, -28° and 68°, respectively) (Extended Data Fig. 6b, Supplementary Table 3). In addition, the U p 47 backbone was substantially rotated with the α and ζ angles changing by -113° and 167°, respectively (Extended Data Fig. 6b, Supplementary Table 3). The m 5 C48 backbone was also rotated, with the α and γ angles changing by -36° and -86°, respectively (Extended Data Fig. 6b, Supplementary Table 3).
Although molecule A had a canonical tRNA core structure stabilized by multiple tertiary interactions between the D-arm and V-loop, including the base triples s 4 U8-A14-A21, Ψ13-G22-G46, C12-G23-C9 and m 2 G10-C25-G45 (Fig. 2b, c), molecule B unexpectedly had a non-canonical core structure (Fig. 2b, c). In molecule B, G46 was dissociated from the base triple Ψ13-G22-G46 and stacked with U p 47   (Fig. 2f, Supplementary Video 1). The N 2 -amino group of G46 formed hydrogen bonds with A21 by inserting itself between the base triple and U p 47 (Fig. 2d). This interaction pushes A21 towards A14 to make additional hydrogen bonds that stabilize the s 4 U8-A14-A21 triple (Fig. 2d). Because U p 47 does not substantially change its backbone conformation (Extended Data Fig. 6b, Supplementary Table 3, Supplementary Video 1), the G46 base was stably trapped by U p 47 in molecule B (Fig. 2d, f). To compensate for this conformational change, C9 comes up from the lower layer (C12-G23-C9) (Fig. 2g, Supplementary Video 1) to form the non-canonical base triple Ψ13-G22-C9 (Fig. 2f, Supplementary Video 1). Thus, the molecule B structure has a non-canonical base triple that might be stabilized by U p 47. In this structural alteration, the torsion angles of A44 and G45 were slightly changed to make the backbone bulge outwards, flipping the G46 base out with the χ angle altered by -70° (Extended Data Fig. 6b, Supplementary Table 3). C9 changes its backbone, altering the α, β, γ and χ angles by 171°, -37°, -180° and 26°, respectively (Supplementary Table 3).
To further investigate the structural role of U p 47, we also solved a crystal structure for Tpt1p-treated S. tokodaii tRNA Val3 (Extended Data Fig. 7a). Both molecules A and B of the Tpt1p-treated tRNA showed the canonical structure with the standard core (Extended Data Fig. 7b-f). In both molecules, U47 was dissociated from the s 4 U8-A14-A21 base triple (Extended Data Fig. 7b-e) with backbone angles α, γ and ε altered by 153°, -109° and -37°, respectively (molecule A) (Extended Data In another aspect of the Tpt1p-treated tRNA, C9 was detached from the C12-G23-C9 base triple in both molecules (Extended Data Fig. 7f). These findings imply that U p 47 stabilizes the metastable tRNA core structure with a non-canonical base triple during thermal denaturation.

Identification of an RNA kinase for U p 47
To identify a gene responsible for U p 47 formation, we narrowed down the candidate genes in the S. tokodaii genome by performing a comparative genomic analysis of sequenced genomes using RECOG (http://mbgd.genome.ad.jp/RECOG/). According to our analysis of U p 47 distribution in archaeal species (Supplementary Note 4, Extended Data Fig. 8a-d), U p 47 is present in seven archaeal species, including in S. tokodaii, but is absent in two species (Fig. 3a). Among the 2,826 genes encoded in the S. tokodaii genome, only nine genes (Supplementary Table 4) were commonly found in all seven archaeal species with U p 47 ( Fig. 3b). Among them, five genes (Supplementary Table 4) were of uncharacterized function (Fig. 3b). We chose one gene encoding a putative protein kinase, STK_09530 (hypothetical serine/threonine kinase, COG2112), as a strong candidate (Fig. 3b). STK_09530 resides in an operon containing a gene for a tRNA nucleotidyltransferase (STK_09520), implying that it encodes an enzyme related to tRNA maturation. We then constructed a strain of T. kodakarensis lacking tk2051, an orthologue of STK_09530. The tRNA fraction obtained from the Δtk2051 strain was subjected to liquid chromatography followed by MS (LC-MS) nucleotide analysis. A pU p m 5 C dimer was clearly observed in the parental strain (wild type) of T. kodakarensis (KU216), but was absent in the Δtk2051 strain (Fig. 3c). Therefore, tk2051 is the gene responsible for U p 47 formation in cells. We designated the gene arkI (archaeal RNA kinase).

U p 47 confers cellular thermotolerance
Next, we investigated the physiological importance of U p 47 in T. kodakarensis. The ΔarkI strain grew as well as the wild-type strain (KU216) at the nearly optimal temperature of 83 °C (Fig. 3d), whereas it showed a weak temperature-sensitive phenotype with slower growth than the wild-type strain at 87 °C and 91 °C (Fig. 3d). We considered synthetic effects of U p 47 with other tRNA modification, thus constructing a ΔarkIΔqueE double-knockout strain, in which queE is responsible for archaeosine (G + 15) formation, because G + 15 thermally stabilizes tRNAs and contributes to cellular thermotolerance 19 . We confirmed the absence of U p 47 and G + 15 in tRNAs from the double-knockout strain ( Supplementary  Fig. 3). The ΔqueE strain grew well at 83 °C, slowly at 87 °C and not at all at 91 °C ( Fig. 3d), as reported 19 . The ΔarkIΔqueE strain grew slower than the wild-type, ΔarkI and ΔqueE strains at 83 °C (Fig. 3d). The strain exhibited a severe growth phenotype at 87 °C ( Fig. 3d) and was unable to survive at 91 °C (Fig. 3d). This finding indicates that U p 47 and G + 15 cooperatively stabilize the tRNA core structure at high temperatures, thereby contributing to cellular thermotolerance.

Kinetics of tRNA phosphorylation by ArkI
We prepared recombinant T. kodakarensis ArkI (TkArkI) and examined in vitro U p 47 formation. U p 47 was efficiently reconstituted only in the presence of ATP (Fig. 3e). We then performed kinetic measurement of U p 47 formation catalysed by TkArkI. The K m and V max values for tRNA were 97.3 nM and 9.9 nM min -1 , respectively (Fig. 3f), showing that TkArkI efficiently recognizes tRNA substrate. By contrast, the K m value for ATP was found to be 1.2 mM (Fig. 3f). This value is extremely high when compared with the values for known protein kinases. This finding indicates that TkArkI-mediated U p 47 formation might be regulated by sensing the cellular ATP concentration. We also characterized ArkI homologues from other archaeal and bacterial species (Supplementary Note 5, Supplementary Figs. 4, 5a-c).

Crystal structure of TkArkI
To find the structural basis of U p 47 formation, we crystallized TkArkI and determined its atomic structure at a resolution of 1.8 Å using X-ray crystallography (Fig. 4a, Extended Data Table 1). On the basis of its amino acid sequence, TkArkI belongs to a superfamily of eukaryotic protein kinases (ePKs) 30 . As observed for ePKs, TkArkI also consisted of two lobes, termed the N-terminal and C-terminal lobes, which were connected by a hinge (positions 96-109) (Fig. 4a, Extended Data Fig. 9). ePKs consist of 12 conserved subdomains that fold into the catalytic core. TkArkI had subdomains I-V in the N-terminal lobe and subdomains VIab, VII, IX and XI in the C-terminal lobe, but lacked subdomains VIII and X ( Compared with the canonical ePK, the characteristic sequences in the conserved motifs were altered in TkArkI. The HRD triplet in the catalytic loop (VIb) of ePKs was replaced with HGQ in TkArkI (Extended Data Fig. 9). In addition, the DFG triplet in the metal-binding loop (VII) of ePKs was replaced with DFE in TkArkI (Extended Data Fig. 9). In subdomain IX, TkArkI had an α-helix (α6) specific to ArkI homologues. In subdomain XI, TkArkI had a longer α-helix (α7), when compared with the same helix in mouse PRKACA. In the extended C terminus of α7, the YKR motif is conserved in ArkI-family proteins (Extended Data Fig. 9), indicating that this positively charged motif is involved in RNA binding.
Although we demonstrated that TkArkI is an ATP-dependent RNA kinase involved in the formation of U p 47 (Fig. 3e, f), we observed a clear electron density for guanosine in the cleft of the two lobes (Fig. 4a, c, d), which corresponds to the ATP-binding site of ePKs surrounded by the hinge and metal-binding, catalytic and P-loops (Supplementary Fig. 6a, b). We confirmed guanosine (and deoxyguanosine) as a ligand that tightly binds to TkArkI (Supplementary Note 6, Supplementary Fig. 7a-c). These observations indicate that TkArkI has binding affinity for guanosine and deoxyguanosine but uses ATP as a major phosphate donor. In the ATP-binding site of mouse PRKACA ( Supplementary Fig. 6a, b), the triphosphate of ATP coordinates two Article Mn 2+ ions and interacts tightly with the conserved motifs, especially the metal-binding loop and P-loop. However, in the guanosine-bound TkArkI structure, the P-loop was dislocated from the ligand-binding site (Fig. 4c, d). Thus, ATP does not bind the ligand-binding site of the observed structure. In homology modelling to ePKs ( Supplementary  Fig. 6c), ATP virtually bound to the active form of the ligand-binding site of TkArkI. It is likely that the P-loop and other motifs form the active pocket for ATP binding following tRNA binding to TkArkI. Although the biological relevance of guanosine binding to TkArkI is not known, guanosine may compete with ATP to regulate tRNA phosphorylation, similar to the mechanism by which nucleoside derivatives inhibit protein kinases 31,32 . Judging by its high K m value for ATP (1.2 mM) ( Fig. 3f), TkArkI might sense the cellular energy status and guanosine binding to TkArkI might have a regulatory role in U p 47 formation. Given that TkArkI was a recombinant protein expressed in Escherichia coli, we cannot rule out the possibility that guanosine was an artificial ligand bound to the inactive form of TkArkI. It is unclear whether guanosine actually binds to TkArkI within archaeal cells at high growth temperatures.
The electrostatic surface potential showed a large positive area on one side of the TkArkI structure (Fig. 4e). The positively charged surface covered the ATP-binding site in the N-terminal lobe and extended to the ArkI-specific elongated α7 helix in the C-terminal lobe (Extended Data Fig. 9). Instead of the missing subdomain VIII involved in recognition of substrate peptide in ePKs (mouse PRKACA), the basic surface in the C-terminal lobe might bind substrate tRNA through electrostatic interaction.
To characterize the conserved residues in TkArkI, we constructed 14 TkArkI mutants in which targeted residues were replaced by alanine (Fig. 4b, f). All mutants were expressed in soluble form and purified. The tRNA phosphorylation activity of each mutant was measured (Fig. 4g). In the ATP-binding site, K32A, G33A, K51A and E65A substitutions markedly reduced activity, whereas the R95A substitution caused a mild reduction in activity. In addition, a severe reduction in activity was observed in the H130A, Q132A and K137A mutants with substitutions in the catalytic loop. No activity was detected for the D149A mutant, in which the mutated residue is in subdomain VII involved in metal binding. These results clearly confirm the critical role of catalytic residues in kinase activity. The N160A and T162A substitutions in subdomain IX led to decreased activity. We mutated the YKR motif in the α7 helix, finding a severe reduction in activity with the K201A substitution and a mild reduction with the Y200A and R202A substitutions. These observations indicate the importance of the conserved residues and positively charged surface in the C-terminal lobe.

KptA acts as an eraser for U p 47
Tpt1p removes the 2′-phosphate from tRNA precursors during maturation 33 . Tpt1/KptA homologues are distributed across all domains of life 34,35 (Supplementary Fig. 4). Although Tpt1/KptA homologues are also present in thermophilic archaea and bacteria (Supplementary Fig. 4), natural RNA substrates with 2′-phosphate have not been identified.
Efficient removal of U p 47 by yeast Tpt1p prompted us to speculate that archaeal KptA is capable of removing the 2′-phosphate of U p 47 from tRNAs in the cell (Fig. 5a). To explore this possibility, we conducted in vitro dephosphorylation of U p 47 with T. kodakarensis KptA (TkKptA) in the presence of NAD + , with the results indicating that the 2′-phosphate of U p 47 was efficiently removed (Fig. 5b). In the same reaction conditions used for U p 47 formation by TkArkI, we measured the kinetic parameters of U p 47 dephosphorylation catalysed by TkKptA: the K m and V max values for tRNA were 180 nM and 27 nM s -1 , respectively (Fig. 5c). The K m value for dephosphorylation by TkKptA is comparable to that of phosphorylation by TkArkI, implying that TkKptA acts as an eraser for U p 47 in the cell.   We then examined the in vivo function of Tpt1/KptA homologues in U p 47 dephosphorylation, using E. coli as a model organism. Because E. coli tRNAs have m 7 G46 and acp 3 U47 modifications, which inhibit U p 47 formation in the V-loop, we used the E. coli ΔtrmBΔtapT strain as a host cell in which both of these tRNA modifications are absent and then expressed Nitrososphaera viennensis ArkI (NvArkI), because N. viennensis is a mesophilic archaeon and its ArkI homologue was predicted to have efficient activity in E. coli. The class I tRNA fraction prepared from this strain was subjected to shotgun analysis to detect the U p 47 modification. We clearly detected four U p 47-containing fragments derived from various E. coli tRNA species (Fig. 5d, Supplementary  Table 5). Each fragment was sequenced by higher-energy collision dissociation analysis, confirming the presence of U p at position 47 (Supplementary Fig. 8). Next, we introduced TkKptA under the control of an isopropyl β-d-1-thiogalactopyranoside (IPTG)-inducible promotor and quantified the peak intensity of each U p 47-containing fragment when TkKptA expression was induced by addition of 10 or 100 μM IPTG (Fig. 5d, e). All four U p 47-containing fragments had decreased abundance as a function of IPTG concentration, demonstrating that TkKptA erases U p 47 in E. coli. We obtained similar results with E. coli KptA (Extended Data Fig. 10a, b) and S. cerevisiae Tpt1p (Extended Data Fig. 10c, d). Together, these data demonstrate that Tpt1/KptA homologues dephosphorylate U p 47 of tRNAs in vivo.

Discussion
U p 47 is, to our knowledge, the first known instance of internal phosphorylation as a stable RNA modification (Supplementary Note 7). 2′-Phosphate at an internal residue appears transiently during tRNA splicing in fungi and plants 36,37 . However, this moiety is not formed by phosphorylation but rather through hydrolysis of 2′,3′-cyclic phosphate via the healing and sealing pathway 36,38 . Because the 2′-phosphate is removed by Tpt1p 33 , it is not present in mature tRNAs.
In S. tokodaii tRNAs isolated in this study, U p 47 was detected in nine class I tRNA species with high frequency (82-100%) (Fig. 1d, Extended Data Fig. 3a) but was absent in two class I tRNAs (tRNA Gln2 and tRNA Cys ) and two class II tRNAs (tRNA Leu4 and tRNA Ser3 ) (Extended Data Fig. 3a). Judging from the primary sequences of these species ( Supplementary  Fig. 9), it is likely that ArkI introduces U p 47 in tRNAs bearing a V-loop with five bases, as tRNA Gln2 and tRNA Cys have four and six bases in the V-loop, respectively. Supporting this finding, only the class I tRNA fraction was phosphorylated in total RNA by in vitro reaction (Supplementary Fig. 5b, c).
RNA hydrolysis is mediated by the 2′-OH group in the presence of divalent metal ions such as Mg 2+ . Especially at high temperatures, RNA is rapidly degraded. Similarly to 2′-O-methylation, the 2′-phosphorylation of U p 47 also serves to prevent tRNA degradation. This property partly explains the RNase resistance of tRNA conferred by U p 47 (Fig. 1g). It is known that U p adopts C2′-endo ribose puckering 28 , which confers flexibility to the RNA strand by extending the backbone structure 39 . Hence, U p 47 presumably acts as a defining mark for single-stranded RNA. In the process of tRNA folding, U p 47 might have a role in preventing the V-loop from being accidentally incorporated into stem structures, ensuring correct folding of the tRNA L-shape structure. Especially in thermophiles, tRNA might frequently misfold owing to its high G+C content. Thus, U p 47 deposition in the tRNA precursor might be required to loop out the V-loop region to ensure correct folding of the tRNA. Other modifications at position 47, acp 3 U 40 and dihydrouridine 7 , are used in bacteria and eukaryotes, respectively. acp 3 U directly prevents  Article the V-loop from being incorporated into stem structures by inhibiting base pairing. Dihydrouridine also adopts the C2′-endo conformation 41 and confers flexibility to the V-loop. It is interesting that similar functions are evolutionarily conserved in different V-loop modifications across the domains of life. Intriguingly, S. tokodaii tRNA Val3 was present as two isomers (molecules A and B) with different conformations in the core region (Fig. 2a). Molecule A has a standard core structure found in many tRNAs, whereas molecule B has a non-standard core structure. Because the Tpt1p-treated tRNA has the canonical structure with the standard core (Extended Data Fig. 7a-f), it is likely that the structural alteration is caused by U p 47. During thermal denaturation of tRNAs, the core region and D-arm are unwound first 42,43 . In molecule B, G46 is released from the base triple Ψ13-G22-G46 and stacks with the uracil base of U p 47 (Fig. 2d, f). Presumably, this unique conformation is a metastable structure of tRNA during heat denaturation. Curiously, in the structural transition from molecule A to molecule B (Supplementary Video 1), the torsion angle of G46 changes substantially, whereas that of U p 47 does not (Extended Data Fig. 6b). U p 47 catches the G46 base that is dissociated from the base triple to restrict further rotation of the V-loop, thereby stabilizing the metastable core structure of the tRNA to prevent its heat denaturation. In addition, C9 comes up from the lower layer (C12-G23-C9) to fill in for the missing G46, forming the non-canonical base triple Ψ13-G22-C9 (Fig. 2f). U p 47 does not fix the tRNA rigidly but rather maintains a metastable structure when the tRNA core thermally fluctuates, thereby preventing further collapse of the core structure, as well as increasing the chance of return to the canonical structure.
ArkI homologues are mainly distributed in thermophilic archaea but are also present in some bacteria (Supplementary Fig. 4). We confirmed the activity of tRNA phosphorylation for bacterial ArkI homologues ( Supplementary Fig. 5a, c). In silico analysis of protein kinases suggested that ArkI-family proteins were originally classified as members of the AQ578 family found in bacterial and archaeal genomes 44 ; the AQ578 family was proposed to have emerged by gene duplication in the early archaeal lineage. The bacterial AQ578 family might have been acquired by horizontal gene transfer of the archaeal homologue, suggesting that the strategy of stabilizing tRNA by internal phosphorylation might have spread across the domains of life.
The ΔarkI strain of T. kodakarensis exhibited weak temperature sensitivity (Fig. 3d), demonstrating that U p 47 by itself contributes to cellular thermotolerance. Because multiple tRNA modifications cooperatively stabilize the tRNA structure, we chose to analyse the G + 15 modification, showing a synthetic phenotype with U p 47 loss. We found that the ΔarkIΔqueE double-knockout strain was extremely susceptible to high temperature (Fig. 3d), suggesting that U p 47 and G + 15 cooperatively stabilize the tRNA core structure and contribute to cellular thermotolerance. U p 47 flexibly deals with the structural change due to thermal denaturation of the core structure, like a padlock, whereas G + 15 tightly fixes the core structure, like a screw bolt (Supplementary Note 3). On the basis of these findings, we propose a new mechanism of tRNA stabilization mediated by two distinct but concerted actions of tRNA modification.
In eukaryotic mRNAs and non-coding RNAs, N 6 -methyladenosine (m 6 A) has a critical role in RNA metabolism and function as a reversible RNA modification 45 . If U p 47 is a reversible modification, it is expected that tRNA function and stability are dynamically regulated by a writer and eraser, raising the possibility of epitranscriptomic regulation of tRNAs in translation. The mechanism closely resembles post-translational modification of proteins. Phosphorylation and dephosphorylation rapidly and dynamically control protein function [46][47][48] . Because tRNA is a stable molecule with a low turnover rate and long lifetime in the cell, it would be reasonable for tRNA function to be regulated by U p 47 modification. We found efficient dephosphorylation of U p 47 by TkKptA in vitro (Fig. 5b) and confirmed the in vivo activity of Tpt1/KptA homologues in E. coli cells (Fig. 5d, e and Extended Data Fig. 10a-d). In fact, tRNA stability is regulated by thermophile-specific tRNA modifications including m 5 s 2 U and ac 4 C, which become much more abundant as the growth temperature increases 14,49 but are not reversible. Reversible U p 47 modification would be beneficial for hyperthermophilic organisms in extremely harsh environments. Future studies will be necessary to investigate U p 47 frequency and the expression levels of ArkI and KptA under various growth conditions, including during rapid changes in growth temperature and introduction of environmental stresses.  Table 5) from an E. coli ΔtrmBΔtapT strain expressing N. viennensis ArkI: U p CGp (top panels), U p CUGp (second panels), U p CAGp (third panels), U p CACAGp (fourth panels) and m 5 UΨCGp as a control fragment (bottom panels). Relative abundance of the U p 47-containing fragments was measured in E. coli strains in which TkKptA was not expressed (left panels) or where TkKptA expression was induced with 10 μM (middle panels) or 100 μM (right panels) IPTG. e, Relative peak intensity of each U p 47-containing fragment detected in the tRNA fraction from E. coli strains cultured with 0, 10 or 100 μM IPTG. Data represent the average values of technical triplicates ± s.d.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-04677-2.

Preparation of tRNA fractions
For small-scale preparation (~100-ml culture), archaeal cells were resuspended in 3 ml solution D (4 M guanidine thiocyanate, 25 mM citrate-NaOH (pH 7.0), 0.5% (wt/vol) N-lauroylsarcosine sodium salt and 1 mM 2-mercaptoethanol) and mixed with an equal volume of water-saturated phenol and 1/10 volume of 3 M sodium acetate (pH 5.3). The mixture was shaken for 1 h on ice and mixed with 1/5 volume of chloroform, followed by centrifugation at 8,000g for 10 min at 4 °C. The supernatant was collected and mixed with an equal volume of chloroform, followed by centrifugation at 8,000g for 10 min at 4 °C. Total RNA was obtained from the resultant supernatant by isopropanol precipitation. The total RNA prepared in this manner was separated by 10% denaturing PAGE, followed by staining with SYBR Gold or toluidine blue. The visualized tRNA fraction including class I and class II tRNAs was cut out and eluted from the gel slice with elution buffer (0.3 M sodium acetate (pH 5.3) and 0.1% (wt/vol) SDS), followed by filtration to remove the gel pieces and ethanol precipitation for RNA-MS analysis of the tRNA fraction.
For large-scale preparation of tRNA fractions from S. tokodaii, cell pellets (53 g) were resuspended in 530 ml solution D and then mixed with 53 ml of 3 M sodium acetate (pH 5.3) and 425 ml neutralized phenol. The mixture was shaken for 1 h on ice to which 106 ml chloroform/ isoamyl alcohol (49:1) was added, followed by centrifugation at 4,500g for 20 min at 4 °C. The supernatant was collected and mixed with 106 ml chloroform/isoamyl alcohol (49:1), followed by centrifugation at 4,500g for 15 min at 4 °C. The aqueous phase was collected and then subjected to isopropanol precipitation. The collected RNA was resuspended in 53 ml water and mixed with 80 ml TriPure Isolation Reagent (Roche), followed by centrifugation at 10,000g for 20 min at 4 °C. The supernatant was collected and mixed with 36 ml chloroform/isoamyl alcohol (49:1), followed by centrifugation at 10,000g for 10 min at 4 °C. The aqueous phase was collected and precipitated with isopropanol. The prepared total RNA (608 mg) was dissolved in 250 ml of buffer consisting of 20 mM HEPES-KOH (pH 7.6), 200 mM NaCl and 1 mM DTT and then loaded on a DEAE Sepharose Fast Flow column (320-ml beads) and fractionated with a gradient of NaCl from 200 to 500 mM. Fractions containing tRNA were collected by isopropanol precipitation.

Isolation of individual tRNAs
Isolation of individual tRNAs from thermophilic organisms is extremely difficult owing to their high melting temperatures, which are the consequence of their high G+C content and complex modifications. We thus optimized our original method for RNA isolation by RCC 24 or chaplet column chromatography (CCC) 52  Visualized bands of mature and precursor tRNAs were cut out and eluted from the gel slices with elution buffer, followed by filtration to remove the gel pieces and precipitation with ethanol.
To crystalize native tRNA bearing U p 47, we conducted large-scale isolation of S. tokodaii tRNA Val3 using CCC 52 . The S. tokodaii tRNA fraction (2,000 A 260 units) was subjected to CCC with tandem affinity chaplet columns for tRNA Val3 , tRNA Ile2 and tRNA Phe . The isolation procedure was carried out as follows: hybridization at 66 °C in 6× NHE buffer, washing separately at 50 °C with 0.1× NHE buffer and elution at 72 °C with 0.1× NHE buffer. The eluted tRNAs were recovered by isopropanol precipitation. The sequences of the DNA probes are shown in Supplementary  Table 6. The isolated tRNA Val3 was further purified by anion exchange chromatography to completely remove tRNA Val2 , as described below.

RNA mass spectrometry
For tRNA fragment analysis by RNA-MS, 30 ng (900 fmol) of the isolated tRNA or 150 ng (4.5 pmol) of tRNA mixture was digested with RNase T 1 (Epicentre or Thermo Fisher Scientific) or RNase A (Ambion) and analysed with a linear ion trap-Orbitrap hybrid mass spectrometer (LTQ Orbitrap XL, Thermo Fisher Scientific) equipped with a custom-made nanospray ion source and a splitless nanoHPLC system (DiNa, KYA Technologies) as described previously 26,27 . To analyse Ψ sites, tRNA was treated with acrylonitrile to cyanoethylate Ψ 53 and subjected to RNA-MS. For dephosphorylation of the U p 47-containing fragment (Extended Data Fig. 4a, b), RNase T 1 digestion was performed in the presence of 0.01 U μl -1 bacterial alkaline phosphatase (BAP C75, Takara Bio). To precisely map tRNA modifications, RNA fragments were decomposed by CID in the instrument. The normalized collision energy of LTQ Orbitrap XL was set to 40%. Mongo Oligo Mass Calculator v2.08 (https://mods.rna.albany.edu/masspec/Mongo-Oligo) was used for assignment of the product ions in CID spectra.
For nucleoside analysis, 800 ng (24 pmol) of the isolated tRNA Val3 was digested with 0.09 U nuclease P 1 (Fujifilm Wako Pure Chemical Corporation) in 20 mM ammonium acetate (pH 5.2) at 50 °C for 1 h and mixed with 1/8 volume of 1 M trimethylamine-HCl (TMA-HCl) (pH 7.2) and 0.06 U phosphodiesterase I (Worthington Biochemical Corporation), followed by incubation at 37 °C for 1 h. To this mixture, 0.08 U BAP was added, and the sample was incubated at 50 °C for 1 h. After that, 9 volumes of acetonitrile were added, followed by LC-MS/MS analysis as described in refs. 25,54 with some modifications as follows. The samples were chromatographed with a ZIC-cHILIC column (3-μm particle size, 2.1 × 150 mm; Merck) and eluted with 5 mM ammonium acetate (pH 5.3) (solvent A) and acetonitrile (solvent B) at a flow rate of 100 μl min -1 with a multistep linear gradient: 90-50% solvent B for 30 min, 50% solvent B for 10 min, 50-90% solvent B for 5 min and then initialization with 90% solvent B. The chromatographed eluent was directly introduced into the electrospray ionization source of the Q Exactive Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific).
For nucleotide analysis, 800 ng (24 pmol) of the tRNA fraction or individual tRNA was digested with 0.09 U nuclease P 1 in 20 mM ammonium acetate (pH 5.2) at 50 °C for 1 h and then mixed with 9 volumes of acetonitrile for LC-MS. The digests were chromatographed with a ZIC-cHILIC column and analysed by Q Exactive Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific) or LTQ Orbitrap XL (Thermo Fisher Scientific) with a multistep linear gradient: 90-50% solvent B for 30 min, 50% solvent B for 10 min, 50-90% solvent B for 5 min and then initialization with 90% solvent B.
The acquired LC-MS data were analysed using Xcalibur 4.1 (Thermo Fisher Scientific) and were visualized with Canvas X (Nihon poladigital k.k).

Isolation and detection of pN 324 p
Five A 260 units of the S. tokodaii tRNA fraction was completely digested with nuclease P 1 . Digests containing pN 324 m 5 C dinucleotide were subjected to periodate oxidation with 10 mM NaIO 4 for 1 h on ice in the dark. The reaction was stopped by addition of 1 M l-rhamnose and incubation for 30 min. For β-elimination, an equal volume of 2 M lysine-HCl (pH 8.5) was added, and the sample was incubated at 45 °C for 90 min. The product containing pN 324 p was then subjected to anion exchange chromatography with a Q Sepharose Fast Flow column (GE Healthcare) equilibrated with 20 mM triethylammonium bicarbonate (TEAB) (pH 8.2). The eluate with 2 M TEAB was collected and dried by evaporation in vacuo. The pellet was dissolved with water and mixed with an equal volume of chloroform, followed by centrifugation at 20,000g for 5 min at 4 °C. The supernatant was recovered and dried again. This process was repeated five times. The resultant digest was mixed with 9 volumes of acetonitrile and subjected to LC-MS/MS using an LCQ-Advantage ion trap mass spectrometer (Thermo Scientific), equipped with an electrospray ionization source and an HP1100 LC system (Agilent Technologies). For LC, the digest was chromatographed with a ZIC-HILIC column (3.5 μm; pore size, 100 Å; internal diameter, 2.1 × 150 mm; Merck) and eluted with 5 mM formic acid (pH 3.4) (solvent A) and acetonitrile (solvent B) at a flow rate of 100 μl min -1 with a multistep gradient: 90-70% solvent B for 25 min, 70-10% solvent B for 15 min, 10% solvent B for 5 min and then initialized with 90% solvent B.

Expression and purification of recombinant proteins
Synthetic genes for arkI from T. kodakarensis, Methanocaldococcus fervens, P. oguniense, Aquifex aeolicus, Nautilia profundicola and Leptolyngbya sp. PCC7376 were designed with codons optimized for E. coli expression and synthesized by GENEWIZ or Thermo Fisher Scientific. Each gene was cloned into the pE-SUMO-TEV vector by the SLiCE method 55 . N. viennensis arkI was PCR amplified from genomic DNA with a set of primers (Supplementary Table 6) and cloned into the BamHI and NotI sites of pE-SUMO-TEV.
E. coli BL21(DE3) or Rosetta2(DE3) cells transformed with the pE-SUMO-TEV vector carrying each arkI gene were cultured in 250 ml or 1 l of LB containing 50 μg ml -1 kanamycin and 20 μg ml -1 chloramphenicol when necessary. His 6 -SUMO-tagged recombinant protein was expressed at 37 °C for 3-4 h by induction with 0.1 or 1 mM IPTG or 2% (wt/vol) lactose when the cells reached OD 610 = 0.4-0.6. P. oguniense ArkI was expressed in cells cultured overnight at 18 °C. The collected cells were resuspended in lysis buffer (50 mM HEPES-KOH (pH 8.0), 150 mM KCl, 2 mM MgCl 2 , 20 mM imidazole, 12% (vol/vol) glycerol, 1 mM 2-mercaptoethanol and 1 mM PMSF) and disrupted by sonication, followed by centrifugation at 15,000g for 15 min at 4 °C. The supernatant was boiled at 60 °C for 20 min (for ArkI homologues from T. kodakarensis, M. fervens, P. oguniense and A. aeolicus) and centrifuged at 15,000g for 15 min at 4 °C. The recombinant protein was affinity captured on an Ni-Sepharose 6 Fast Flow column (GE Healthcare) and then eluted with lysis buffer containing 300 mM imidazole, followed by gel filtration with a PD-10 column (GE Healthcare) to remove the imidazole. The recombinant protein for N. viennensis ArkI was purified using a HisTrap column (GE Healthcare) with a linear gradient of 0-500 mM imidazole, followed by dialysis using a Slide-A-Lyzer Dialysis Cassette (Thermo Fisher Scientific) to remove imidazole. The purified protein was subjected to Ulp1 digestion at 4 °C overnight to cleave the His 6 -SUMO tag and then passed through a Ni-Sepharose 6 Fast Flow column to remove the tag. Because ArkI homologues from M. fervens (MfArkI) and Leptolyngbya sp. PCC7376 (LeArkI) aggregated following tag removal, His 6 -SUMO tag-fused proteins of these homologues were used for the phosphorylation assay. Purified protein was quantified by the Bradford method using BSA as a standard.
For large-scale preparation of T. kodakarensis ArkI for crystallization, the E. coli BL21(DE3) strain carrying pE-SUMO-TkArkI was cultured in 2 l of LB containing 50 μg ml -1 kanamycin and TkArkI was expressed at 25 °C overnight by induction with 0.1 mM IPTG when the cells reached OD 610 = 0.4. The cells were collected and disrupted by sonication in lysis buffer (50 mM HEPES-KOH (pH 8.0), 150 mM KCl, 2 mM MgCl 2 , 20 mM imidazole, 12% (vol/vol) glycerol, 1 mM 2-mercaptoethanol and 1 mM PMSF). The protein was purified using a HisTrap column with a linear gradient of 20-520 mM imidazole. Fractions containing TkArkI were pooled and subjected to Ulp1 digestion at 4 °C overnight to cleave the tag, followed by passage through a Ni-Sepharose 6 Fast Flow column to remove the tag fragment. The flow-through fraction was filtered through a 0.45-μm PVDF membrane to remove the resin. The protein was further purified by affinity chromatography with a HiTrap Heparin HP column (GE Healthcare) using a linear gradient of 150-1,150 mM KCl. TkArkI was further purified by size exclusion chromatography using a Superdex 75 10/300 GL column (GE Healthcare) with buffer containing 20 mM Tris-HCl (pH 8.0), 150 mM NaCl and 10 mM 2-mercaptoethanol and then concentrated to 5.74 mg ml -1 and stored at -80 °C.
The T. kodakarensis kptA gene was PCR amplified from genomic DNA from T. kodakarensis with the primers listed in Supplementary Table 6 and cloned into pE-SUMO-TEV to give pE-SUMO-TEV-tkkptA. The E. coli Rosetta2(DE3) strain carrying pE-SUMO-TEV-tkkptA was cultured in 1 l LB containing 50 μg ml -1 kanamycin and 20 μg ml -1 chloramphenicol, and TkKptA was expressed at 37 °C for 3 h by induction with 0.1 mM IPTG when the cells reached OD 610 = 0.6. The recombinant TkKptA was purified as described above. The gene encoding Tpt1p was PCR amplified from the genomic DNA of S. cerevisiae BY4742 with the set of primers listed in Supplementary Table 6 and was cloned into pET21b (Merck) between the NdeI and XhoI sites. Recombinant Tpt1p was purified as described above.

Removal of the 2′-phosphate of U p 47 by Tpt1p
Removal of the 2′-phosphate of U p 47 by yeast Tpt1p was performed as described 33  RNase probing of tRNA S. tokodaii tRNA Val3 (25 pmol) with or without U p 47 was labelled with 32 P at the 3′ terminus by ligation with [5′-32 P]cytidine 3′,5′-bisphosphate (PerkinElmer). The labelled tRNA was separated on a 7.5% (wt/vol) polyacrylamide gel containing 7 M urea, 1× TBE and 10% (vol/vol) glycerol and was purified by gel extraction. Labelled tRNA was mixed with the S. tokodaii tRNA fraction as a carrier to a concentration of 100,000 counts per minute (c.p.m.) per A 260 unit and was precipitated with ethanol. The pellet was dissolved in water to a concentration of 0.1 A 260 units per μl. For the RNase degradation assay, the labelled tRNA (0.1 A 260 units, 10,000 c.p.m.) was incubated at 65 °C in a reaction mixture consisting of 10 mM HEPES-KOH (pH 7.6), 0.5 mM MgCl 2 , 100 mM NaCl and 0.1 U μl -1 RNase I (Promega). At time points of 1, 3, 5, 10, 15 and 30 min after starting the reaction, aliquots were taken from the mixture and mixed well with chilled phenol/chloroform/isoamyl alcohol (25:24:1, pH 7.9) to stop the reaction, followed by centrifugation at 15,000g for 15 min at 4 °C. The supernatant was collected and treated with an equal volume of chloroform, followed by centrifugation at 15,000g for 5 min at 4 °C. The supernatant was mixed with 2× loading solution (2× TBE, 7 M urea, 13.33% (wt/vol) sucrose, 0.05% (wt/vol) xylene cyanol and 0.05% (wt/vol) bromophenol blue) and subjected to 10% denaturing PAGE. The gel was exposed to an imaging plate, and radioactivity was visualized by using an FLA-7000 imaging analyser (Fujifilm). Graphs were generated using Microsoft Excel.

Crystallization of S. tokodaii tRNA Val3
S. tokodaii tRNA Val3 (500 μg), isolated as described above, was refolded in annealing buffer (50 mM HEPES-KOH (pH 7.6), 5 mM MgCl 2 and 1 mM DTT) by incubation for 5 min at 80 °C and cooling to 25 °C with a rate of 0.1 °C s -1 . tRNA Val3 was further purified by anion exchange chromatography using a Mono Q 5/50 GL column (GE Healthcare) with a linear gradient of 200-1,000 mM NaCl. The major peak was collected, precipitated with isopropanol, dissolved in water and precipitated with ethanol. Tpt1p-treated tRNA Val3 was prepared with the same procedure as described above. The purified tRNA was dissolved in buffer consisting of 10 mM Tris-HCl (pH 7.1) and 5 mM MgCl 2 to a concentration of 50 μM. One microlitre of tRNA solution was mixed with 1 μl Natrix 2 no. 32 (80 mM NaCl, 12 mM spermine-4HCl, 40 mM sodium cacodylate·3H 2 O (pH 7.0) and 30% (vol/vol) MPD) (Hampton Research) on silicon-coated glass and crystalized by the hanging drop vapor diffusion method at 20 °C.

Crystallization of T. kodakarensis ArkI
The concentration of TkArkI was adjusted to 5 mg ml -1 before crystallization. One microlitre of the protein solution was mixed with 0.5 μl reservoir solution, containing 25% (vol/vol) ethylene glycol. TkArkI was crystallized by the hanging drop vapor diffusion method at 20 °C.

Data collection and crystal structure determination
The datasets were collected at beamline BL-17A at the Photon Factory at KEK, Japan. For data collection for the tRNA Val3 crystals, the crystals were cryoprotected with a portion of the reservoir solution. For data collection for the native TkArkI crystal, the crystal was cryoprotected with solution containing 25% (vol/vol) ethylene glycol, 2 mM MgCl 2 and 1 mM ATP. For data collection for the iodide-derivative TkArkI crystal, the crystal was briefly soaked in and cryoprotected with solution containing 300 mM potassium iodide and 22.5% (vol/vol) ethylene glycol, and the diffraction dataset was collected at a wavelength of 1.5 Å. The datasets were indexed, integrated and scaled using xds 56 . The initial phase of tRNA Val3 was determined by molecular replacement with Phaser 57 . The structure of T. thermophilus tRNA Val (PDB, 1IVS) 58 was used for the model. The initial phase of TkArkI was determined by the SAD method using the anomalous signal of iodide ions. The iodine sites were located by SHELX 59 , and the initial phase was calculated by Phaser. Subsequent density modification and initial model building were performed with RESOLVE 60 . The model was further modified with Coot 61 and refined with Phenix 62 . Crystal structures and their electron density maps were visualized using PyMOL, Cuemol or Coot. Torsion angles of the tRNAs were analysed with DSSR software 63 .

Analysis of ligands bound to TkArkI
TkArkI purified by affinity chromatography with a HiTrap Heparin HP column (GE Healthcare) (100 pmol) was mixed with [ 15 N]adenosine (10 pmol) and [ 15 N]guanosine (10 pmol) as tracer molecules, followed by addition of 4 volumes of methanol, an equal volume of chloroform and 3 volumes of water and vigorous mixing. The denatured protein was removed by centrifugation at 15,000g for 1 min at 4 °C. The supernatant was dried in vacuo and dissolved in 20 μl water. Half of the extract was analysed by LC-MS. The tracer molecules were prepared by dephosphorylation of [

Construction of gene knockout strains of T. kodakarensis
Knockout strains of T. kodakarensis were constructed by pop-in/ pop-out recombination as described previously 64 . The 5′ and 3′ flanking regions (about 1,000 bp) of T. kodakarensis arkI and kptA were PCR amplified from genomic DNA with a set of primers (Supplementary Table 6) and inserted into the pUD3 vector bearing the pyrF marker 65 to yield pUD3-arkI and pUD3-kptA. The T. kodakarensis KU216 strain (ΔpyrF) was transformed with pUD3-arkI or pUD3-kptA, and the uracil-prototrophic transformants generated by pop-in recombination were selected on an ASW-AA-S 0 plate without uracil. The selected strains were then cultured on an ASW-AA-S 0 plate supplemented with 5-FOA to obtain uracil-auxotrophic, 5-FOA-resistant transformants formed by pop-out recombination. The knockout strains of arkI or kptA were selected among the transformants by genomic PCR with a set of primers (Supplementary Table 6). The double-knockout strain of arkI and queE (ΔarkI/queE::Tn) was constructed by deletion of arkI from FFH05 (queE::Tn) isolated from a random mutagenesis library 19 . T. kodakarensis strains used in this study are listed in Supplementary  Table 7.
Growth phenotype analysis T. kodakarensis KU216 (wild type), FFH05 (queE::Tn), ΔarkI and ΔarkI/queE::Tn strains were precultured in MA-YT-Pyr medium at 83 °C overnight and inoculated into 8 ml fresh MA-YT-Pyr medium with an initial OD 600 of 0.01. The cells were cultured at 83 °C, 87 °C or 91 °C, and cell growth was monitored every 2 h by measuring OD 600 with an S1200 diode array spectrophotometer. Graphs were generated using Microsoft Excel.

In vitro transcription of tRNA
For in vitro transcription of T. kodakarensis tRNA Val3 and its G5-C68 variants by T7 RNA polymerase 66 , template DNAs were constructed by PCR using synthetic DNA (Supplementary Table 6). The tRNAs were transcribed at 37 °C overnight in a reaction mixture consisting of 40 mM Tris-HCl (pH 7.5), 24 mM MgCl 2 , 5 mM DTT, 2.5 mM spermidine, 0.01% (vol/vol) Triton X-100, 0.8 μg ml -1 T7 RNA polymerase, 1 μg ml -1 pyrophosphatase, 30 nM DNA template, 2 mM ATP, 2 mM CTP, 2 mM UTP, 2 mM GTP and 10 mM GMP, followed by extraction with phenol/chloroform treatment and desalting with PD-10 columns (GE Healthcare). In vitro transcripts prepared in this way were separated by 10% denaturing PAGE, followed by staining with toluidine blue. The stained bands were cut out and eluted from the gel slice with elution buffer, followed by filtration to remove the gel pieces and ethanol precipitation.
In vitro phosphorylation of tRNA by ArkI U p 47 formation by TkArkI was carried out at 70 °C for 20 min in a reaction mixture (30 μl) containing 50 mM HEPES-KOH (pH 7.5), 1 mM MgCl 2 , 1 mM MnCl 2 , 1 mM DTT, 10% (vol/vol) glycerol, 0.5 mM ATP, 0.9 μM tRNA fraction (from the T. kodakarensis ΔarkI strain) and 1 μM TkArkI. After the reaction, the tRNA was extracted by acidic phenol/chloroform, desalted on a NAP-5 column (GE Healthcare) and precipitated with isopropanol. For RNA-MS, the prepared tRNA was dialysed against water on a nitrocellulose membrane (0.025-μm VSWP, MF-Millipore, Merck) for 2 h (drop dialysis). To examine GTP as a phosphate donor, 0.5 mM ATP or GTP was added to the reaction mixture and U p 47 formation was performed with 0.5 μM TkArkI for 5 min, followed by RNA-MS analysis. The activities of TkArkI variants were measured by γ-phosphate transfer from [γ-32 P]ATP to tRNA similarly to the kinetic studies of TkArkI (see below). tRNA phosphorylation was performed at 70 °C for 15 min in an 8-μl reaction mixture. For PAGE analysis, 4 μl of the reaction mixture was mixed with 4 μl of 2× loading solution, resolved by 10% denaturing PAGE and exposed to an imaging plate to visualize radiolabelled RNA with an FLA-9000 imaging analyser (Fujifilm). The gel image was analysed using Multi Gauge (Fujifilm). Bar graphs with independent plots were prepared with R (R Foundation). For phosphorylation of total RNA, the reaction was performed at 70 °C for 30 min in an 8-μl reaction mixture consisting of 50 mM HEPES-NaOH (pH 7.5), 1 mM MgCl 2 , 1 mM MnCl 2 , 1 mM DTT, 10% (vol/vol) glycerol, 100 μM [γ-32P] ATP (3,000 mCi mmol-1; PerkinElmer), 1.8 μM TkArkI and 50 ng μl-1 total RNA fraction (from the T. kodakarensis ΔarkI strain). Then, 0.5 μl of 50 mM EDTA (pH 8.0) was added, and 4 μl of reaction mixture was mixed with 2× loading solution, resolved by 10% denaturing PAGE and visualized as described above.
In vitro dephosphorylation of tRNA by T. kodakarensis KptA Dephosphorylation of U p 47 by TkKptA was carried out at 60 °C for 1 h in a reaction mixture (30 μl) containing 20 mM Tris-HCl (pH 7.4), 0.5 mM EDTA (pH 8.0), 1 mM NAD + , 2.5 mM spermidine, 0.1 mM DTT, 0.9 μM T. kodakarensis tRNA fraction and 0.1 μg μl -1 recombinant TkKptA. After the reaction, the tRNA was extracted by acidic phenol/chloroform, desalted on a NAP-5 column (GE Healthcare) and precipitated with isopropanol. For RNA-MS, the prepared tRNA was desalted by drop dialysis as described above.
TkKptA-mediated dephosphorylation of U p 47 was quantified by measuring the reduction in radioactivity for tRNA. In vitro-transcribed T. kodakarensis tRNA Val3 was phosphorylated by TkArkI with [γ-32 P]ATP as described above and then purified by gel extraction and isopropanol precipitation. In addition, the same tRNA was phosphorylated by TkArkI with unlabelled ATP. By mixing labelled and unlabelled tRNAs, the specific activity of the labelled tRNA was adjusted to 6,250 c.p.m. per pmol in buffer consisting of 50 mM HEPES-KOH (pH 7.6), 5 mM MgCl 2 and 1 mM DTT. The labelled tRNA was incubated at 80 °C for 5 min and then cooled at room temperature, followed by isopropanol precipitation. The labelled tRNA was dissolved in water to a concentration of 8 μM (50,000 c.p.m. per μl). Dephosphorylation of the labelled tRNA by TkKptA was performed at 70 °C in a reaction mixture (30 μl) consisting of 50 mM PIPES-NaOH (pH 6.9), 125 mM NaCl, 1 mM MgCl 2 , 1 mM MnCl 2 , 1 mM DTT, 10% (vol/vol) glycerol, 1 mM NAD + , 0.05 mg ml -1 BSA (Takara), 1 nM TkKptA and 12.5-800 nM 32 P-labelled tRNA. At each time point (2 and 5 min), 8-μl aliquots were spotted on Whatman 3MM filter paper, which was immediately soaked in 5% (wt/vol) trichloroacetic acid. The filter paper was washed three times for 15 min with ice-cold 5% (wt/vol) trichloroacetic acid, rinsed for 5 min with ice-cold ethanol and dried in air. Radioactivity on the filter paper was measured by liquid scintillation Article counting (Tri-Carb 2910TR, PerkinElmer). Kinetic parameters were calculated using Prism 7.
In vivo dephosphorylation of U p 47 by KptA N. viennensis arkI was PCR amplified and cloned into pMW118 (Invitrogen) under the control of the synthetic constitutive J23106 promoter 67,68 , followed by insertion of sequences encoding a His 6 tag and a 3×Flag tag at the C terminus of the N. viennensis arkI gene, yielding pMW-J23106-nvarkI (Supplementary Table 7). T. kodakarensis kptA, E. coli kptA and S. cerevisiae tpt1 were PCR amplified and cloned into pQE-80L (Qiagen). The ampicillin resistance cassette (Amp r ) was replaced with a chloramphenicol resistance cassette (Cam r ), yielding pQE-80LC-tkkptA, pQE-80LC-eckptA and pQE-80LC-sctpt1, respectively (Supplementary Table 7). The E. coli ΔtrmBΔtapT (Kan r ) strain was transformed with pMW-J23106-nvarkI and further transformed with pQE-80LC-tkkptA, pQE-80LC-eckptA or pQE-80LC-sctpt1. The transformants were inoculated in 3 ml LB supplemented with 20 μg ml -1 chloramphenicol, 50 μg ml -1 kanamycin and 100 μg ml -1 ampicillin and cultured at 37 °C until mid-log phase. When the OD 610 reached 0.6, IPTG was added to a final concentration of 10 or 100 μM to induce expression of the KptA/Tpt1p homologue and cells were cultured for 3.5 h. A 1.5-ml aliquot of the culture was taken, and the tRNA fraction was extracted and analysed by shotgun RNA-MS as described above. Primers, E. coli strains and plasmids used are listed in Supplementary Tables 6, 7. Bar graphs with independent plots were prepared with R (R Foundation).

Drawing of chemical structures
Chemical structures were drawn with chemical structure drawing tools, including ACD/ChemSketch (ACD/Labs) or ChemDraw (PerkinElmer).

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
Coordinates and structure factors have been deposited in the Protein Data Bank under accession codes 7VNV, 7VNW and 7VNX. Source data are provided with this paper.