Molecular basis of tRNA substrate recognition and modification by the atypical SPOUT methyltransferase Trm10

Abstract

The evolutionarily conserved methyltransferase Trm10 modifies the N1 position of guanosine 9 (G9) in some tRNAs, but how the enzyme recognizes and modifies its substrate tRNAs remains unclear. Here, we used an S-adenosyl-L-methionine (SAM) analog to trap the Trm10-tRNA^Gly complex and enable determination of its structure in a post-catalytic state by cryogenic electron microscopy (cryo-EM). We observed three distinct complexes: two with a single Trm10 bound to tRNA that differ in their tRNA acceptor stem orientation (“closed” and “open”) and a minor population with two Trm10s engaging the same tRNA. The monomeric complexes reveal a positively charged surface that guides the G9 into the catalytic site with key conserved residues forming “pincer”-like interactions that stabilize the flipped methylated nucleotide. In the open tRNA conformation, the acceptor stem is rotated away from the enzyme, weakening the tRNA–protein contacts, consistent with a product-release conformation. The dimeric complex, which is supported by tRNA-dependent protein crosslinking, reveals one Trm10 positioned similarly to the monomeric complexes and engaged with G9, while the other Trm10 contacts distal tRNA regions, suggesting a potential role in facilitating a key conformational transition during the process of catalysis or modified tRNA release. Finally, molecular dynamics simulations comparing G9- and A9-containing complexes reveal that G9 is efficiently stabilized in the binding pocket unlike A9, identifying the structural basis for guanosine selectivity. Overall, these findings reveal the structural determinants of G9-specific tRNA methylation by Trm10 and suggest a unique mechanism of action among RNA-modifying SPOUT methyltransferases.

INTRODUCTION

Methylation is one of the most common chemical modifications in biology and plays important roles in gene expression, small molecule metabolism, and regulation of macromolecule structure and function (1–4). The SpoU-TrmD (SPOUT) enzyme family–one of five major classifications of S-adenosyl-L-methionine (SAM)-dependent methyltransferases–was designated upon identification of structural similarity between the transfer RNA (tRNA)-modifying enzymes TrmH (also known as SpoU) and TrmD (5–10). SPOUT (or Class IV) methyltransferases are characterized by a unique α/β fold and a deep trefoil knot in the C-terminal half of the catalytic domain. This protein knot is functionally important as it stabilizes SAM in a bent conformation necessary for methyl transfer by this family of enzymes (7, 11–13).

The SPOUT tRNA methyltransferase Trm10 modifies the N1 base position of the 9^th nucleotide in the core region of some tRNAs and is evolutionarily conserved across Eukarya and in many Archaea (14). Saccharomyces cerevisiae Trm10 was first predicted to be a member of the SPOUT family in 2007 (9), and this designation was experimentally verified through X-ray crystallographic studies of the C-terminal SPOUT domains of Trm10 from S. cerevisiae and Schizosaccharomyces pombe (15). In S. cerevisiae, Trm10 methylates G9 of 13 of 26 possible tRNA substrates (16). In contrast, Trm10 enzymes in the Archaea Sulfolobus acidocaldarius and Thermococcus kodakarensis were found to either methylate A9 or exhibit dual G9/A9 activity, respectively (17–19). Humans express three Trm10 enzymes which also possess diversity in their substrate specificities and localization. TRMT10A (the direct homolog of yeast Trm10), which methylates many but not all tRNAs with G9, and TRMT10B, which methylates A9 in tRNA^Asp, are both cytosolic enzymes, while the mitochondria-localized TRMT10C modifies both G9 and A9 of all tRNAs that contain a purine at this position (16, 20–22). Among the human enzymes, the biological importance of TRMT10A has been highlighted by studies linking loss of its activity to neurological and endocrine disorders (23–27). Similarly, in S. cerevisiae, loss of Trm10 leads to a variety of detrimental effects, such as increased susceptibility to 5-fluorouracil (28), compromised tRNA integrity (29), and decreased translational fidelity (30).

The structures of the yeast Trm10 catalytic C-terminal domain (CTD) confirmed the presence of a typical knotted SPOUT methyltransferase fold (15). However, unlike all other known RNA-modifying SPOUT enzymes which function as homodimers, Trm10 was observed as a monomer in the crystal structure. Subsequent structures of archaeal Trm10 enzymes from S. acidocaldarius and T. kodakarensis revealed that these enzymes also lack the expected dimeric SPOUT methyltransferase structure (17, 31). Identification that Trm10 functions as a monomeric enzyme was unexpected because the catalytic centers of other SPOUT RNA methyltransferases are formed at the dimer interface, with both protomers making essential contributions to the other’s active site. Currently, only one other monomeric SPOUT methyltransferase, Sfm1, has been identified but this enzyme is also unique in other ways, such as acting on a protein substrate and containing a distinct negatively-charged surface surrounding the active site (32), compared to the positive surface of RNA-modifying SPOUT enzymes. As such, it remains unclear how Trm10 recognizes and efficiently methylates its tRNA substrate without dimerization.

More recently, the first structures of a Trm10 enzyme (human TRMT10C) bound to substrate tRNA were determined as part of the mitochondrial protein-only RNase P (PRORP) complex (33, 34), and have begun to address this question albeit within the context of multiple additional protein binding partners. Within the PRORP complex, TRMT10C serves as one of the enzyme subunits responsible for pre-tRNA recognition, modification, and processing. These structures show the NTD of TRMT10C wrapping around the tRNA so that the substrate is enclosed on each side by the catalytic SPOUT CTD and the NTD, with the target A9 base flipped into the enzyme active site. However, other subunits of the RNase P complex also contact the pre-tRNA, including the pentatricopeptide repeat and nuclease, which have known roles in RNA recognition and are necessary for tRNA 5’-end cleavage by the nuclease subunit. Importantly, TRMT10C is the only SPOUT methyltransferase that requires participation in a larger complex for its enzymatic activity which may explain why it is able to function as a “monomer”. In contrast, other Trm10 enzymes do not require accessory factors for methyltransferase activity, and it therefore remains unclear how yeast Trm10 and its direct human homolog TRMT10A are able to recognize their correct tRNA substrate(s) as atypical stand-alone monomeric SPOUT methyltransferases.

Here, we use single-particle cryo-EM to gain structural insight into the mechanism of tRNA^Gly-GCC substrate recognition by S. cerevisiae Trm10. Multiple states of the Trm10-tRNA complex were identified, including two “monomeric” Trm10-tRNA complexes, distinguished primarily by the orientation of the tRNA acceptor stem, and a minor population with two Trm10 proteins on a single tRNA ((Trm10)₂-tRNA), which was corroborated by observation of tRNA-dependent recruitment of two Trm10 in protein crosslinking experiments. The three different states of Trm10-tRNA and molecular dynamics (MD) simulations reveal key interactions that are critical for substrate tRNA recognition and G9 modification, elucidating the molecular determinants of guanosine selectivity and supporting a new mechanistic model for Trm10 function as an atypical monomeric SPOUT methyltransferase.

RESULTS

Structure of the S. cerevisiae Trm10-tRNA^Gly-GCC complex

To determine the cryo-EM structure of the Trm10-tRNA^Gly-GCC complex, we used a SAM analog (“NM6”) that is transferred and covalently attached to the tRNA in its entirety to trap Trm10 on its tRNA substrate in an immediately post-catalytic state (figs. S1–S3) (35). Trm10 is bound to tRNA^Gly-GCC in a predominantly 1:1 ratio in two distinct monomeric complexes that are distinguished by the orientation of the tRNA acceptor stem (“open” or “closed”, tRNA_open and tRNA_closed, respectively; Fig. 1A,B). In both structures, only the Trm10 SPOUT domain, tRNA, and NM6 are well defined, with no map corresponding to the Trm10 NTD, which is therefore absent from both models. The structure of a third complex containing two Trm10 SPOUT domains, (Trm10)₂-tRNA^Gly-GCC, was also determined from a minor fraction of the particles (discussed further below). Again, no well-defined map was observed for either Trm10 NTD.

Fig. 1. Structure of the monomeric Trm10-tRNA complex.

A, DeepEMhancer (DEM)-sharpened map of open Trm10-tRNA^Gly complex in the tRNA_open conformation at 3.37 Å resolution (Trm10 and tRNA^Gly-GCC are shown in green and light green, respectively). The final model is shown within a semi-transparent white map outline. Inset, Zoomed-in view of NM6 (yellow) covalently linked to G9 within the DEM-sharpened map shown (threshold: 0.234; value range: −0.00338 to 2.19). B, As for panel A, but for the monomeric Trm10-tRNA complex with the tRNA_closed conformation at 3.63 Å resolution (Trm10 and tRNA^Gly-GCC shown in blue and cyan, respectively). For the inset in panel B, the map is DEM-sharpened with a threshold of 0.591 and a map value range of −0.00171 to 2. C, Alignment of Trm10 from both monomeric tRNA complexes and free Trm10 (white) shows that the overall structure remains largely unchanged. D, Superimposition of tRNA_closed, tRNA_open, and free tRNA^Phe (PDB 1EHZ) reveals significant differences throughout the molecule, particularly in tRNA_open.

In both monomeric complexes Trm10 is structurally similar to the previously reported yeast Trm10 SPOUT domain structure (PDB code 4JWJ; Fig. 1C), with root mean square deviation (RMSD) for alignment of 1.5 Å (with tRNA_open) and 1.4 Å (with tRNA_closed) over 159 and 186 Cα atoms, respectively. The positioning of Trm10 on the tRNA is also consistent with the position of the Trm10 paralog TRMT10C in relation to its tRNA substrate as part of the mitochondrial RNase P complex, further supporting the relevance of these structures for substrate recognition by Trm10 (fig. S4A) (33). Trm10 makes extensive contacts with the core of the tRNA and the D-arm, with its α1 helix interacting directly with G9, which is covalently linked to NM6 (Fig. 1A–C). In contrast, Trm10 makes more limited contact with the tRNA anticodon stem-loop (ASL), which has a poorer map quality compared to the rest of the structure (fig. S3D,E) and appears to be more dynamic during substrate recognition by Trm10, with the highest B-factors observed for the loop nucleotides (fig. S4B).

The Trm10 NTD is functionally important (15) but, as noted above, is not observed in the map for either monomeric complex, most likely due to the high flexibility of this region. We speculate that the NTD may engage with the tRNA in an essential step prior to methylation and is thus not observed in the post-catalytic structure captured using the NM6 analog. We therefore generated and aligned an AlphaFold (35) model of full-length Trm10 to the Trm10 SPOUT domain in the tRNA_closed complex (RMSD of 1.4 Å, over 181 Cα atoms; fig. S5A), showing the modeled NTD wrapping around the tRNA in a binding mode similar to the TRMT10C-tRNA complex (33). Additionally, calculation of the modeled Trm10 electrostatic surface potential reveals that this modeled NTD position places a second positively charged surface against the opposite side of the tRNA from that of the CTD (fig. S5B). Thus, these two domains can hold the negatively charged tRNA backbone in a pincer grip as a part of the Trm10 substrate recognition mechanism.

Trm10 induces structural changes throughout the bound tRNA

Superimposition of tRNA^Gly-GCC from the tRNA_open and tRNA_closed monomeric Trm10 complexes with tRNA^Phe (which serves as an established “standard” tRNA for which a high-resolution structure is available; PDB code 1EHZ) yields RMSDs of 6.7 Å and 5.3 Å, respectively, over 68 phosphate atoms. Structural divergence between tRNA^Phe and tRNA_closed/ tRNA_open arises from conformation changes in each of the four major tRNA domains: acceptor stem, D-arm, T-arm, and ASL (Fig. 1D). The ASLs in both tRNA_open and tRNA_closed are similar, suggesting that this region undergoes consistent structural adaptation with Trm10 bound.

To enable visualization of phosphate atom movements using modevector analysis (36), a “free” tRNA^Gly-GCC structure was generated using AlphaFold. Comparison of this free structure with tRNA_closed reveals shifts in phosphate positions primarily in the ASL as well as the D-arm (Fig. 2A, left). This observation is supported by nuclease footprinting of free and Trm10-bound tRNA^Gly-GCC, which shows a strong increase in accessibility to RNase T1 at G34, consistent with significant distortion in the ASL (fig. S6). Modevector analysis comparing the free and tRNA_open structures reveals similar changes in the ASL to those in tRNA_closed as well as more extensive shifts in the D-arm and additional changes in the T-arm and acceptor stem (Fig. 2A, center). These differences suggest a large-scale conformational change is induced in tRNA_open by Trm10. Finally, comparison of tRNA_closed and tRNA_open confirms that the major differences in these structures arise from their distinct acceptor stem orientations and movements in the adjacent D- and T-arm regions (Fig. 2A, right).

Fig. 2. tRNA_open and tRNA_closed exhibit distinct architectures.

A, Modevector visualization of the tRNA conformational changes (P atom position) comparing (left to right): free tRNA vs. tRNA_closed, free tRNA vs. tRNA_open, and tRNA_closed vs. tRNA_open. B, The acceptor stem of tRNA_open undergoes a rotation of 23° compared to tRNA_closed, accompanied by an 18 Å displacement of its 3′ end. C, The T-arm rotates by 41° and undergoes a maximum shift of 18 Å in tRNA_open to accommodate the reoriented acceptor stem. D, The D-arm also rotates by 18° and shifts at the most distant point by 8 Å.

The tRNA_open and tRNA_closed structures are most clearly distinguished by the distinct orientation of their acceptor stem relative to the tRNA body and bound Trm10 (Fig. 2B and fig. S7). In tRNA_open, the acceptor stem comprising base pairs G1-C70 to G7-C64 (along with the bases U57-U65 and U8 of the T-arm) is rotated by ~23° and shifted at its maximum displacement by ~18 Å from the corresponding position in tRNA_closed. The movement of the acceptor stem is enabled by reorganization of T-arm nucleotides C46-U58 which act as a hinge region and are also rotated (41°) and shifted (18 Å) from their corresponding positions in tRNA_closed (Fig. 2C). The T-arm movement also results in a rotation (18°) and shift (8 Å) of the D-arm (G10-A23) to maintain the overall architecture of tRNA_open (Fig. 2D). The structures of the two monomeric Trm10-tRNA^Gly-GCC complexes thus reveal structurally distinct open and closed conformation of the substrate tRNA which arise from coordinated reorganizations of the acceptor stem, T-arm, and D-arm. Collectively, these observations indicate that Trm10 can induce distinct tRNA conformations that may contribute to the mechanism of substrate tRNA recognition and G9 modification.

Contacts made by multiple conserved Trm10 residues direct tRNA binding

Trm10 binds tRNA^Gly-GCC using a positively charged surface with an opening to the catalytic site of the enzyme positioned to accommodate G9 for modification (Fig. 3A–C). Several Trm10 residues contact the tRNA to precisely position the substrate (Fig. 3D–G and fig. S8). Trm10 residue R121 forms a cation- interaction with G10 (Fig. 3D,E), directly stabilizing the tRNA backbone conformation adjacent to the target nucleotide G9, which has its nucleobase flipped into the active site. Residues R127 and R128 also interact with the tRNA phosphate backbone of A27/A28 and C25/C26/A27 in the anticodon stem, close to the tRNA core (Fig. 3D,E). Both R121 and R127 exhibit significant conservation across fungal Trm10 enzymes (Fig. 3H and Table S1), and their substitution with glutamic acid in the context of S. cerevisiae Trm10 results in loss of methylation activity (37). Our structures thus reveal the molecular basis for the important contributions of these residues in stabilizing the distorted tRNA structure surrounding the target site in S. cerevisiae Trm10 (37, 38).

Fig. 3. Trm10 engages tRNA via a partially conserved basic surface.

A-C, Trm10 possesses a positively charged surface that facilitates interaction with the tRNA and guides the flipped G9 into the active site opening. D, In tRNA_closed, side chains of R127 and R128 contact the negatively charged phosphate group of A27 and C25/C26, respectively, to stabilize the anticodon stem of the tRNA. R121 contacts the base of G10 in the D-arm. A second group of residues−R147 and T155/W160/Y123−is positioned adjacent to the phosphate of G44 (part of the variable loop) and A27, respectively, also contributing to tRNA stabilization. The interactions indicated with dashed lines are those described in the main text. E, Same view as for panel D, but for the Trm10-tRNA_open complex showing that R127 and R128 interact with A28 and C26/A27, respectively. Similarly, R147 and T155/W160/Y123 remain in position to contact the RNA backbone but are adjusted to contact G43 and A28, respectively. Interactions in the acceptor stem region in the F, Trm10-tRNA_closed and G, Trm10-tRNA_open complexes show the side chains of amino acids K211 and K110 contacting C47 and C45 (tRNA_open)/C46 (RNA_closed), respectively. In tRNA_closed, R243 and K215 also contact U8 and C64, respectively, which are missing in tRNA_open due to the movement of the acceptor stem. H, Sequence comparison of S. cerevisiae Trm10 and 321 Trm10 homologs from fungi depicts the conservation of R121, Y123, R127, R128, R147, and W160. The symbols above the alignment represent whether the residues were confidently modeled in the tRNA_open (green) and/or tRNA_closed (blue) conformation. I, As for panel H, but for residues K110, K211, K215, and R243.

Several other positively charged, polar, and aromatic residues make additional nearby interactions with the tRNA. These include R147 which is positioned to interact with variable loop nucleotides via either the phosphate of G44 (tRNA_closed) or the ribose 2’-OH of G43 (tRNA_open). Additionally, the side chains of W160 and Y123 interact with the phosphate of either A27 or A28 in tRNA_closed and tRNA_open, respectively, and the peptide backbone at T155 packs against the RNA backbone in this region (Fig. 3D,E). Among these residues, Y123, R147, and W160 are also strongly conserved among fungal Trm10 homologs, while T155 exhibits very low conservation (Fig. 3H and Table S1), consistent with roles for the observed side chains in all but the latter residue. Notably, residues R121 and R147 appear to act cooperatively as a ‘pincer’ to hold the tRNA from opposite sides and, together with the other adjacent residues, precisely position the tRNA for G9 modification (fig. S4C).

At a second site of interactions with the tRNA adjacent to the acceptor stem and variable loop, the moderately conserved K110 residue interacts with either C46 (tRNA_closed) or C45 (tRNA_open) in the variable loop, while K211 is also positioned close to C47 as well as the phosphate of G9 in both tRNA conformations (Figs. 3F,G,I and 4, and Table S1). K211 is thus positioned to stabilize the reconfigured tRNA backbone of the flipped target base in a conformation that appears to be further stabilized by interaction of R243 and the phosphate of the adjacent U8 in tRNA_closed. Interestingly, the interaction with R243 is absent for tRNA_open as the acceptor stem moves away from Trm10 (Fig. 3G). Similarly, K215 forms a hydrogen bond with the phosphodiester backbone of C64 in tRNA_closed that is absent for tRNA_open due to the rotation of the T-arm (Fig. 3F,G).

Fig. 4. Conserved polar and hydrophobic residues in the active site stabilize the flipped G9 target nucleotide.

A, G9 is captured in a flipped orientation compared to the adjacent nucleotides U8 and G10, covaently linked to NM6. B, Sequence analysis of fungal Trm10 homologs reveals strong conservation of residues involved in stabilizing the flipped nucleotide. Symbols above the alignment indicate whether the residues were confidently modeled in tRNA_open (green) and/ or tRNA_closed (blue) structures. C-D, In tRNA_closed, side chains of polar residues, T248, Q118, S114, and K211 contact G9. In tRNA_open, these contacts are maintained, but with T247 in place of T248, and an additional interaction is observed between D210 and both G9 and NM6. E-F, G9 and NM6 establish packing interactions with primarily non-polar residues, as well as direct contacts with G206, L186, V245, and I208.

Similar to R121 and R147, in tRNA_closed residues K110, K211, and R243 appear to act as structural pincers to secure the tRNA (fig. S4C). However, in contrast to the moderate conservation of K110 and K211 in fungi, R243 is highly conserved in >95% of analyzed sequences (Table S1). We note that the tRNA substrate specificities of Trm10 enzymes from diverse fungi have not been determined, but that modification of different sets of tRNA species in different species is a characteristic feature of m¹G9 methylation (16). Therefore, it is possible that divergence in some of these amino acids correlates with different patterns of tRNA specificity across fungi. Nonetheless, electrostatic complementarity and residue conservation support a model in which Trm10 uses a set of basic and other residues to engage extensively with different regions of its tRNA substrate, with several of these residues acting as molecular pincers to stabilize the distorted RNA conformation and hold the substrate in place for accurate G9 modification.

Target nucleotide G9 is flipped into the active site

To make the N1 target atom accessible for modification, the G9 base is flipped out of the tRNA core and positioned in the active site (Figs. 4A and fig. S9). The conserved and functionally critical Q118 (15) stabilizes G9 in its flipped conformation through interactions with the nucleobase N3 and primary amine group (exocyclic N2) (Fig. 4B–D, and Table S1). Additionally, S114, which is highly conserved in fungi, is positioned to interact with both the G9 ribose 2’-OH and the phosphodiester backbone. In TRMT10C, N222 makes a corresponding interaction with the G9 ribose (33), suggesting a partially conserved nature of this interaction among the human proteins (fig. S10A). Finally, the G9 exocyclic amino group is positioned to interact with one of two Thr residues via the side chain OH groups of T248 (tRNA_closed) and T247 (tRNA_open) (Fig. 4C,D). However, both Thr residues are only modestly conserved fungal Trm10 enzymes and are typically replaced by non-polar hydrophobic residues in other members (Fig. 4B and Table S1). Thus, the observed hydrogen bonding interactions may be less critical than precisely enclosing the flipped base within the active site. Indeed, a notable feature in the active site of Trm10 is the presence of a highly conserved hydrophobic pocket, in which the base of G9 packs against I208, V209, and V245–Trm10 amino acids that are evolutionarily functionally conserved across fungal homologs (Figs. 4B,E,F, and Table S1). Additionally, in tRNA_open, the main chains of I208 and V245 form hydrogen bonds with G9 (Fig. 4F). These interactions could not be confidently modeled in tRNA_closed due to its poorer map quality in this region; thus, it is unclear whether these additional interactions are specific to the tRNA_open conformation.

The covalent attachment formed between the SAM analog NM6 and the G9 nucleobase N1 position during catalysis is clearly observed in the map (Fig. 1B,D). NM6 adopts a bent conformation, typical for the cosubstrate when bound to SPOUT methyltransferases (39), and its position closely matches that of SAH observed in the previously reported yeast Trm10 SPOUT structure (fig. S10B–C) (15). NM6 is bound within a highly conserved hydrophobic pocket, which also forms part of the trefoil knot structure that accommodates the bent conformation of the SAM analog (40). In the tRNA_closed structure, side chain packing interactions are observed with L186, L232, and L246, and backbone interaction with G206 (Fig. 4E). The map for the Trm10-tRNA_open also allowed modeling of additional interactions with the SAM analog made by the main chain atoms of residues L186 and L232, and positioning of the I234 side chain to enclose the adenine ring of NM6 (Fig. 4F). Finally, the carboxylate side chain of D210 interacts with NM6 in tRNA_open, while the amino acid backbone atoms stabilize G9 (Fig. 4D). These structures thus reveal how a collection of conserved polar and hydrophobic residues in Trm10 together coordinate to position both the flipped G9 base and the SAM analog to enable the enzyme to carry out methylation efficiently and accurately.

Q118 mediates selective stabilization of G9 for methylation

To better understand how Trm10 selectively recognizes and modifies G9, we compared a portion of the active site of TRMT10C in complex with tRNA^Ile (PDB code 8CBO; contains G9) or tRNA^His,Ser (PDB code 8CBK; contains A9) (34), and the Trm10-tRNA_open complex in which interactions with G9 and NM6 were best defined by our maps. Additionally, to enable these comparisons, NM6 was removed and replaced by SAM (Trm10-tRNA_open-G9), and a second model was then generated in which G9 was replaced with A9 (Trm10-tRNA_open-A9). For both TRMT10C and Trm10, the amino and carboxyl groups of the Q226/Q118 side chain contact G9 (Fig. 5A). A similar interaction is maintained by TRMT10C with A9 but is absent in the modeled Trm10-tRNA_A9 complex due to the movement of the nucleobase away from Q118 (Fig. 5B). In TRMT10C, both G9 and A9 are contacted by N348 with additional interactions made by N350 or D314, respectively (Fig. 5A,B). The structurally corresponding Trm10 residues, R243, V245, and D210, are unable to make similar interactions due to the hydrophobic side chain of V245 and distinct orientations of D210 and R243, which place them beyond hydrogen bonding distance.

Fig. 5. Q118 guides selective modification of G9 by Trm10.

A, Active site comparison of TRMT10C-pre-tRNA^Ile (left, G9, violet, PDB 8CBO) and S. cerevisiae Trm10-tRNA_open-G9 (right) showing the conserved contact between Q226/Q118 and G9. B, This interaction is retained with A9 in the TRMT10C–pre-tRNA^His,Ser complex (left, A9, violet, PDB 8CBK) but absent in the modeled Trm10-tRNA_open-A9 complex (right,) due to nucleobase movement away from Q118. Additional TRMT10C residues involved in interaction are N348 (both G9 and A9), N350 (G9), D314 (A9), whereas the corresponding Trm10 residues (R243, V245, D210) fail to form equivalent H-bonds due to unfavorable polarity and geometry. C, MD simulations of Trm10–tRNA_open-G9 and Trm10–tRNA_open-A9 complexes show that the two side chain interaction distances of Q118 with G9 are significantly shorter and within H-bonding distances compared to that with A9. D, Potential energy profile of the Q118-G9 interaction is lower than that with A9, suggesting a more stable interaction with guanosine.

We next used MD simulations of the Trm10-tRNA_open-G9 and Trm10-tRNA_open-A9 complexes and monitored the distance distribution between the side chain of Q118 and G9/A9 over the course of the trajectory. This analysis reveals that the Q118-G9 interaction is significantly more stable, with a distribution centered on a shorter distance and retaining consistent hydrogen bonding or electrostatic interaction (Fig. 5C). In contrast, the Q118-A9 distance distribution is shifted toward longer distances, indicating a weaker or less persistent interaction over the simulation (Fig. 5C). This comparison thus suggests that Q118 engages more consistently in favorable interactions with G9 compared to A9. We also calculated the total potential energy profiles of each complex, with the G9-containing system displaying a lower average energy and reduced variation over the simulation (Fig. 5D). This result is again consistent with the presence of G9 resulting in a more stable conformational ensemble, likely due to the role of Q118 in anchoring the nucleotide in a position suitable for modification. In contrast, the lack of recognition of A9-containing tRNAs by S. cerevisiae Trm10 can be explained by the less favorable interaction with the nucleotide, along with a higher energy landscape. Notably, this proposed role is also consistent with the larger effect on the observed rate of m¹G9 modification (42-fold) vs. the rate of m¹A9 modification (4-fold) upon alteration of the analogous residue Q122 (Q118 in S. cerevisiae) to alanine in the context of the bifunctional T. kodakarensis Trm10 (18).

Overall, these findings support a model in which Q118 selectively recognizes and stabilizes G9 for modification, a hallmark of fungal Trm10 enzyme specificity. Further, Trm10 enzymes from Archaea and Eukaryotes with distinct specificities (A9 or dual G9/A9) also employ the analogous residue to Q118, but specificity appears to be supported by critical interactions made by other residues surrounding the target nucleotide. For example, we speculate that the selective recognition mechanism of A9 in human TRMT10B, the structure of which is unknown, may arise via interactions with Q148 (Q118 in S. cerevisiae) in combination with other residues in a manner similar to TRMT10C.

A subpopulation of complexes containing two Trm10 enzymes is observed with a single tRNA

All RNA-modifying SPOUT family methyltransferases other than Trm10 have been shown to act as dimers (6, 39). Interestingly, although the previously published yeast Trm10 enzyme only structures (15) and our Trm10-tRNA complexes show predominantly monomeric enzyme, a small subset of 2D classes in our dataset appeared to contain two Trm10 proteins for each tRNA molecule, i.e., (Trm10)₂-tRNA^Gly-GCC (fig. S2A). The number of particles comprising these classes (~5.2%) was significantly less than for monomeric Trm10, suggesting that a second Trm10 may only be present as a transient binding partner with the monomeric Trm10-tRNA complex. A 3D reconstruction of the (Trm10)₂-tRNA^Gly-GCC complex revealed one Trm10 SPOUT bound to the tRNA core and positioned similarly to the enzyme in the monomeric Trm10-tRNA complexes, while the second Trm10 enzyme is bound near the ASL (Fig. 6A,B). Although the map provided sufficient density to confidently dock and model the second Trm10 SPOUT domain, the quality of the overall map is affected by the significant orientation preference compared to the monomeric Trm10s (fig. S3G–I), limiting our ability to definitively identify specific residues in tRNA or protein involved in the additional interaction.

Fig. 6. Trm10 forms a transient tRNA-dependent dimeric complex.

A, CryoSPARC (CS)-sharpened map of the (Trm10)₂-tRNA complex (Trm10 in orange and yellow; tRNA^Gly-GCC in pink) at 3.89 Å (threshold: 0.019; map value range: −0.0897 to 0.164). B, The final model is shown within a semi-transparent white map. Inset, Close-up view of the NM6 (yellow) attached to G9 with the corresponding CS-sharpened map shown (threshold: 0.021). C-E, Comparison with tRNA_open (green) and tRNA_closed (cyan), and indicates that the dimeric Trm10-bound tRNA adopts intermediate conformations in its acceptor stem, D-arm, and T-arm. F, The two Trm10 protomers employ an extended positively charged surface to surround the tRNA substrate. G, BS3 crosslinking shows that Trm10 dimer formation is dependent upon presence of tRNA but not SAM.

The two Trm10s in the dimeric complex are structurally similar to the previously published Trm10 structure (fig. S11A), with an RMSD of ~3.0 Å for both Cα atom alignments. In contrast, alignment of the tRNA bound to the two Trm10s to tRNA^Phe shows an RMSD of 7.8 Å over 63 phosphate atoms, with significant differences distributed across the acceptor stem, D-arm, and T-arm (fig. S11B). Comparing the tRNAs from the two monomeric complexes and the (Trm10)₂-tRNA complex reveals the tRNA bound to two Trm10 enzymes to be in an intermediate conformation in terms of the deformations observed in the acceptor stem, D-, and T-arm of tRNA_closed and tRNA_open (Fig. 6C–E). This observation suggests that the tRNA conformation in the (Trm10)₂-tRNA complex may represent an intermediate stage in the process of recognition of tRNA by Trm10. Protein electrostatic surface calculation reveals that the second Trm10 also employs a positively charged surface to interact with the tRNA, such that the tRNA is sandwiched between positive surfaces from the two proteins and further supporting the plausibility of the positioning of the second Trm10 molecule (Fig. 6F). Finally, superimposing the AlphaFold full-length Trm10 structure on the Trm10 directly engaged with G9 shows no clash between its NTD and the SPOUT domain of the second monomer, indicating that the two protein molecules can simultaneously bind to the tRNA without steric hindrance (fig. S5C). It is also notable that the second Trm10 shows fewer direct contacts with the tRNA than the Trm10 protomer common to both the dimeric and monomeric complexes, suggesting that its dynamic NTD might play a more prominent role in this additional interaction. However, like the NTD from the first Trm10, this region remains unresolved in our map due to poor density.

Finally, to further support the observation of two Trm10 enzymes binding to the tRNA and exclude the possibility that the observation is an artifact of the structure determination process, we performed a protein crosslinking analysis using the bifunctional reagent BS3 for Trm10 in solution in the presence and absence of both tRNA^Gly-GCC and SAM. Exclusively monomeric Trm10 was observed both with or without the cosubstrate in the absence of tRNA, whereas a prominent band corresponding to the presence of two Trm10s appeared in a tRNA- dependent manner, regardless of the presence of SAM (Fig. 6G). This result demonstrates the necessity of the tRNA substrate to promote the binding of two Trm10 molecules and can explain the absence of Trm10 in previous structural studies of the enzyme in which tRNA was absent (15). In summary, our structural and biochemical studies support a model in which two Trm10 molecules are necessary only when tRNA is present, and that this recruitment may reflect a transient step in the enzyme-substrate recognition and modification process.

DISCUSSION

Trm10 is an evolutionarily conserved tRNA methyltransferase that in yeast modifies the N1 base position of G9 in the tRNA core. In this work, we used cryo-EM to generate three distinct structures of the Trm10-tRNA complex: two monomeric complexes distinguished by the conformation of the bound tRNA, and a dimeric (Trm10)₂-tRNA complex. The monomeric Trm10 complexes mark the first detailed structural insight into the process of tRNA substrate recognition and modification by this stand-alone monomeric SPOUT methyltransferase, as it is captured on the tRNA in a state immediately after modification. Additionally, the unexpected dimeric (Trm10)₂-tRNA complex revealed how two molecules of Trm10 may act in unison on a single tRNA. The structures allowed us to identify important interactions between Trm10 and tRNA that contribute to substrate recognition, stabilization of the flipped G9 nucleobase for modification, and insight into the mechanism of selective recognition of guanosine at position 9 by yeast Trm10.

The two Trm10 monomeric structures, sequence conservation analysis, and previous studies allow us to rationalize functional data from yeast and human proteins. For instance, the highly conserved D314 in the active site of human TRMT10C, which is also conserved as D210 in S. cerevisiae, has been shown to be functionally important in TRMT10C (22, 34). Importantly, the lack of a direct interaction between the side chain of analogous D210 and the N1 atom of G9 in any of our structures is consistent with and rationalizes the previous demonstrations that this residue does not function as a general base during catalysis in either S. cerevisiae or T. kodakarensis Trm10, and possibly reflects the distinct catalytic mechanism of non-mitochondrial eukaryotic and archaeal Trm10 enzymes (31, 38). Furthermore, T247 interacts with G9 in its flipped conformation and alteration of the equivalent residue (T244) in S. pombe (15) reduced enzyme activity by 65%, underscoring the importance of properly positioning G9 for modification (Fig. 4D). Similarly, the V206A variant of S. pombe Trm10 (equivalent to V209 in S. cerevisiae) shows a 80% reduction in activity (15), indicating a key contribution of the G9 base stacking interaction for enabling target nucleotide modification (Fig. 4E,F).

Previous biochemical studies on residues implicated in tRNA binding showed that a double K153/R147E substitution affects tRNA binding in S. pombe (15) and our structures reveal R147 interacting with G43/G44, which would be disrupted by this substitution (Fig. 3D,E). Additionally, substitution of K208 in S. pombe (equivalent to K211 in S. cerevisiae) to alanine results in a 72% loss in activity compared to the wild-type enzyme (15), supporting a role for K211 in Trm10 activity via stabilization of the flipped G9 and interaction with C47 (Figs. 3F,G and 4C,D). Previous studies have shown that deletion of residues K236-R240 in S. pombe nearly abolished Trm10 activity (15). Notably, R243 in S. cerevisiae, which is equivalent to R240, interacts with U8 in the closed conformation and thus appears to be involved in stabilization of tRNA architecture around the modified G9 base (Fig. 3F).

Our structures can also rationalize the underlying basis of disease-linked changes. Q118 is highly conserved in the active site of Trm10, and is important for enzymatic activity in S. pombe (15) ans humans (TRMT10C) (41), and substrate selectivity in T. kodakarensis–observations reinforced and explained by our structures and MD simulations (Figs. 4C,D and 5). A Q124K (Q118 in yeast) substitution in human TRMT10A has been reported in patients, but no associated physiological disorders have been observed (42). At G206, a residue important in stabilizing the bent cosubstrate conformation, clinical studies show that patients with mutations encoding an Arg or Ala substitution in human TRMT10A exhibit microcephaly and intellectual disability (24, 43). Likewise, a nonsense mutation that results in truncation of TRMT10A at P233, another residue involved in positioning the cosubstrate, also results in neurological disorders (44). Trm10 residues K110, R121, and R127 are conserved in both S. cerevisiae and S. pombe and their substitution with glutamic acid was observed to affect either tRNA binding (S. pombe) (15) or methyltransferase activity (S. cerevisiae) (37), indicating the importance of these contacts. Additionally, clinical studies have shown substitutions of TRMT10A K116 (equivalent to K110 in S. cerevisiae) (45), R127 (equivalent to R121 in S. cerevisiae) (23), and R133 (equivalent to R127 in S. cerevisiae) (27) to be associated with microcephaly, intellectual disability, and epilepsy.

It is worth noting that most of the tRNA interactions identified in our structures are observed with the backbone (i.e., ribose and phosphate) moieties and are thus consistent with the lack of specific sequence-based identity elements for distinguishing substrate and non-substrate RNA, as has been demonstrated repeatedly for Trm10 enzymes (16, 18, 21). Rather, previous SHAPE analysis has shown that substrate recognition is dependent on inherent tRNA flexibility and the ability of Trm10 to induce distinct conformational changes in the tRNA upon binding (37). For instance, the D-arm was shown to be highly reactive in SHAPE studies which is reflected in our modevector analysis (Fig. 2A). Additionally, our studies indicated an unwinding of the helix in the ASL and a translational shift in the acceptor stem, while SHAPE showed an overall stabilization of both the regions. This suggests that Trm10 binding induces a conformational change that results in a more constrained ASL structure, possibly through interaction with the NTD as observed in our AF model, with reduced nucleotide flexibility and solvent accessibility in SHAPE, even though global backbone geometry is altered. Similarly, the lack of flexibility in the acceptor stem indicates a sterically constrained movement of the tRNA while sampling the different conformations observed in the tRNA_open and tRNA_closed forms. The relatively weaker patterns of conservation observed for Trm10 residues that interact with tRNA in the current structure may reflect the use of evolved changes in these amino acids to achieve the different patterns of tRNA specificity across Trm10 from diverse organisms (Table S1).

Interestingly, as more diverse orthologs of eukaryotic Trm10 enzymes are considered, the relative conservation of many of these residues also decreases. Human TRMT10A, the functional ortholog of S. cerevisiae Trm10, exhibits the most similarity to fungal Trm10s among the amino acids implicated above in tRNA interactions, except for K211, which is His in human TRMT10A (Table S1). Perhaps not surprisingly, human TRMT10C lacks obvious conservation at several of these residues, which likely reflects its distinct composition as part of a multi-protein complex during its interaction with mitochondrial tRNA. Human TRMT10B represents somewhat of an intermediate among vertebrate enzymes, with some amino acids sharing similarity with fungal Trm10 and human TRMT10A, while others show distinct patterns of conservation. Therefore, the distinct substrate specificity for a single tRNA exhibited by this ortholog remains unexplained. Overall, the partial functional conservation of the tRNA-interacting residues suggests many of these are collectively important for tRNA binding. The absence of some individual contacts in selected Trm10 proteins is likely compensated by interactions of other residues or potentially reflect differences in substrate selectivity. Thus, the observation could provide a compelling explanation for this so far unresolved aspect of Trm10 substrate selectivity.

Structural alterations in tRNA domains due to enzyme binding have been reported in studies of several other tRNA-modifying enzyme-substrate complexes (46–49), and particularly in tRNA methyltransferases, such as Trm5 (50), TrmD (51), and Trm6-Trm61 (52). While Trm5 and TrmD binding primarily affect the tRNA structure at the ASL, Trm6-Trm61, aspartyl-tRNA synthetase (AspRS) (47), and Sep-tRNA:Sec-tRNA synthase (SepSecS) (49) induce large distortions in the acceptor stem and T-arm regions. However, in contrast to the opening of the tRNA acceptor stem in the Trm10-tRNA_open complex, binding of AspRS, SepSecS, and Trm6-Trm61 induces a closure of the acceptor stem towards the tRNA body, representing an ‘induced fit’ mechanism of positioning the target nucleotide in the catalytic pocket for modification. Thus, while these enzymes induce structural rearrangements surrounding their site of modification to facilitate catalysis, the conformational changes induced by Trm10 in the post-catalytic state are more distant from its modification site in the tRNA core.

The substantial tRNA distortions may also reflect a Trm10-tRNA complex in a state poised for product release, a process that may require overall structural relaxation in the complex. In other systems, the free energy of an enzyme-substrate complex is lower in the catalytically active phase compared to the inactive phase or where the complex is not fully formed for catalysis (53, 54). Accordingly, the Trm10-tRNA_closed complex exhibits lower calculated interaction energy compared to the tRNA_open form, and a more extensive buried surface (2330 vs 2080 Ų in the complexes with tRNA_closed and tRNA_open, respectively). Similar to the finding that the conformational change in dihydrofolate reductase during product release results in an increase in the free energy (55), our observations suggest that the tRNA_open conformation may be part of product release post-catalysis. Similarly, a previously reported inactive AspRS-tRNA complex formed due to the absence of acceptor stem interaction with the enzyme (47), parallels our open Trm10-tRNA conformation and suggests functional inactivation before product (tRNA) release. Based on these observations and our three Trm10-tRNA structures, we propose an overall mechanism of catalysis and tRNA release, in which a monomeric Trm10 first binds the tRNA in a closed form for G9 methylation (Fig. 7), which is also similar to that seen in the TRMT10C-tRNA complex (34). Following catalysis, the modified tRNA transitions to an open conformation, resulting in the release of the substrate from Trm10; however, post-catalytic dissociation of the complex is blocked in the current structure by the covalent attachment of NM6 to the modified tRNA. This model also offers a plausible explanation for why Trm10 binds to, but does not modify, the type II tRNAs. The structural transition observed in tRNA_open would be hindered by the extended variable loop in type II tRNAs, preventing the movement of the T-arm, and thereby restricting the tRNA to the closed conformation as suggested previously (16). Finally, we also speculate that the (Trm10)₂-tRNA complex may represent a transient functional state that either precedes the closed state by helping to position G9 for modification, or after modification, with the second Trm10 protomer binding the closed complex to displace the acceptor arm and weaken interactions to promote substrate release. The hypothesis that the second protomer plays a role in orienting G9 is supported by the structures of Trm6-Trm61 and Trm7-Trm734 bound to tRNA, in which Trm6 and Trm734 act as a tRNA-binding scaffold, while Trm61 and Trm7 provide the SAM binding motif and position the nucleotide to be modified (52, 56).

Fig. 7. Proposed model for Trm10 catalysis mechanism and tRNA release.

Free Trm10 (gray) and unmodified (G9) tRNA (gray) assemble in Trm10-tRNA_closed conformation (blue), promoting incorporation of the m¹G9 modification (red sphere). Following catalysis, structural reorganization of the tRNA to the Trm10-tRNA_open conformation (green) enables release of free Trm10 and the m¹G9-modified tRNA. Additionally, the transient (Trm10)₂-tRNA complex may either promote the formation of a modification-competent tRNA conformation in the closed complex (left) or, following catalysis, promotes opening of the tRNA structure to allow product release (right).

In conclusion, our studies reveal structures of the highly conserved atypical SPOUT methyltransferase Trm10 bound to a substrate tRNA in the absence of other binding partners and offer the first evidence of Trm10 activity employing a unique, transient tRNA-dependent dimeric complex to modify its substrate. Additionally, these studies shed light on the unique mechanism of selective recognition of G9 for modification, offering a mechanistic blueprint for substrate recognition by this atypical fungal tRNA SPOUT methyltransferase.

MATERIALS AND METHODS

Trm10 expression and purification

Full-length wild-type Trm10 from S. cerevisiae with an N-terminal 6xHis-tag was expressed from the pET-derived plasmid pJEJ12–3 in E. coli BL21(DE3)-pLysS grown in lysogeny broth, as described previously (14). Briefly, protein expression was induced by the addition of 1 mM β-D-1-thiogalactopyranoside at mid-log phase growth (OD₆₀₀ ~0.6), and growth continued at 37°C for an additional 5 hours. All steps during lysis and initial purification were performed in 20 mM HEPES pH 7.5, 4 mM MgCl₂, 1.0 mM b-mercaptoethanol (BME), 10 mM imidazole, and 5% glycerol. To ensure removal of co-purifying SAM, cells were lysed in this buffer with 1 M NaCl and 0.5% Triton X-100 added, and the lysate was dialyzed three times against the same buffer but containing 2 M NaCl. A final dialysis step was used to reduce the NaCl to 0.25 M for protein purification by sequential Ni²⁺-affinity (HisTrap HP), heparin-affinity (HiPrep Heparin 16/10), and gel filtration (Superdex 75 16/600) chromatographies on an ÄKTApurifier10 system (GE Healthcare). Trm10 was eluted from the gel filtration column in 20 mM Tris (pH 7.5) buffer containing 100 mM NaCl, 1 mM MgCl₂, 5 mM BME, and 5% glycerol and flash frozen in liquid nitrogen before storage at −80°C.

RNA in vitro transcription and purification

tRNA^Gly-GCC was in vitro transcribed from BstNI linearized plasmid DNA using T7 RNA polymerase as previously described (57). Briefly, in vitro transcription was performed for 5 hours at 37°C in 200 mM HEPES-KOH (pH 7.5) buffer containing 28 mM MgCl₂, 2 mM spermidine, 40 mM dithiothreitol (DTT), 6 mM each rNTP, and 100 mg/mL DNA template. Following addition of 40 mM EDTA to clear pyrophosphate-magnesium precipitates and dialysis against TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0), tRNA^Gly-GCC was purified by denaturing polyacrylamide gel electrophoresis (50% urea, Tris-borate-EDTA buffer) and eluted from the gel using the crush and soak method in 0.3 M sodium acetate, followed by ethanol precipitation, as previously described (57).

Trm10-tRNA complex formation and cryo-EM specimen preparation

The SAM analog 5’-(diaminobutyric acid)-N-iodoethyl-5’-deoxyadenosine ammonium hydrochloride (“NM6”) was prepared as previously described (58) and purified by semi-preparative reverse-phase HPLC. In situ activation of NM6 results in a Trm10 cosubstrate that is covalently attached by the enzyme to tRNA (37, 59). Prior to the preparation of the Trm10-tRNA complex for cryo-EM, glycerol was removed from the Trm10 sample by dialysis against 20 mM Tris buffer (pH 7.5) containing 100 mM NaCl, 1 mM MgCl₂, and 5 mM BME. NM6 was dissolved in the same buffer. tRNA^Gly-GCC was incubated at 65°C in TE buffer for 10 minutes and then slowly cooled to room temperature. The Trm10-tRNA-NM6 complex was formed by mixing Trm10 (60 μM), tRNA (30 μM), and NM6 (0.6 mM) in a 2:1:20 molar ratio, followed by incubation at 30°C for 30 minutes. The sample was diluted in the Trm10/NM6 buffer to a final complex concentration of 2.25 μM or 4.5 μM. The diluted complex (3 μl) was applied to freshly glow-discharged grids (UltrAufoil R 0.6/1, 300 Mesh, Au), with blotting for 2–3 s at 100% humidity at room temperature before freezing in liquid ethane using a Vitrobot Mark IV System (Thermo Scientific). Grids were stored in liquid nitrogen until used for data collection.

Cryo-EM image collection, processing and analysis

Data were collected on a Titan Krios microscope (FEI) operating at 300 keV with a K3 direct electron detector (Gatan) at the National Center for Cryo-EM Access and Training (NCCAT). A total of 26,656 micrographs were collected with a defocus range of −0.8 to −2.0 μm at 105,000x magnification with a 0.412 Å/pixel size. The dataset contains micrographs that were collected with tilting angles of 0°, 30°, and 45° to capture the complex in a wider distribution of orientations. Micrographs were collected as 50 frames with a dose rate of 29.91 e⁻/Ų/s and a total exposure of 2.0 seconds, for an accumulated dose of 59.82 e⁻/Ų.

The complete workflow for cryo-EM structure determination of the three structures in CryoSPARC (60) is summarized in fig. S2. First, the micrographs were imported and motion corrected by Patch Motion Correction, followed by estimation of contrast transfer function (CTF) parameters using Patch CTF. Blob Picker was used for particle picking, and incorrectly selected particles were discarded after reference-free two-dimensional (2D) class averaging. The best 2D classes were used for template-based particle picking, followed by multiple rounds of reference-free 2D class averaging, resulting in 5,634,445 and 244,276 particles (256-pixel box size) corresponding to monomeric Trm10-tRNA and dimeric (Trm10)₂-tRNA complexes, respectively. Multiple ab-initio 3D reconstructions were generated from these particles with C1 symmetry, and good-quality maps with observable features for both Trm10 and tRNA were used as a reference for 3D heterogeneous refinement. For the (Trm10)₂-tRNA complex, heterogeneous refinement of the ab-initio map did not yield improved results, and the results were therefore discarded.

For both monomeric Trm10-tRNA complexes, the best maps from heterogeneous refinement were used for non-uniform (NU) refinement followed by CTF refinement, reference-based motion correction, and further NU refinement. This process yielded a final global map resolution of 3.37 Å based on gold-standard refinement Fourier Shell Correlation (0.143 cutoff) for the Trm10-tRNA_open complex (fig. S3A). For the Trm10-tRNA_closed complex, the resultant NU refined map was 3D classified, and the best classes were chosen for further NU refinement, resulting in a final global map resolution of 3.63 Å (fig. S3B).

For the (Trm10)₂-tRNA complex, NU refinement was performed with the best maps from the ab-initio refinement, followed by reference-based motion correction and 3D classification. The best classes were chosen for NU refinement, generating a final map of 3.89 Å resolution (fig. S3C). Finally, the NU refined maps from the three structures were sharpened in CryoSPARC, and the half-maps were sharpened using DeepEMhancer (61). CryoSPARC local resolution and orientation diagnostics tools were used to generate local resolution maps and Sampling Compensation Factor values, respectively, for all maps (fig. S3D–I).

Model building and refinement

The NU refined maps were used for initial model building of the three Trm10-tRNA complexes. A yeast Trm10 SPOUT domain structure (PDB code 4JWJ), with ligands and water molecules removed, was docked as a single copy in the maps for two monomeric complexes, while two copies of the same structure were used for (Trm10)₂-tRNA. As there is no available structure of tRNA^Gly-GCC, we generated a 3D model using AlphaFold 3 (35) with the sequence of tRNA^Gly-GCC and a single model was docked into each of the three maps using Chimera (v1.17) (62). The combined model was refined in PHENIX (63) using rigid body, global minimization, simulated annealing, local grid search, and B-factor refinement. Manual adjustment of the refined model and de-novo modeling of NM6 modified-G9 was done in COOT (64) using the DeepEMhancer (61)- and CryoSPARC-sharpened maps, followed by a B-factor refinement in PHENIX. All three structures were validated using PHENIX. Comprehensive details on data collection, processing, model construction, refinement, and validation are provided in Table 1.

Table 1:

Cryo-EM data collection, refinement, and model validation for the Trm10-tRNA complexes

	Trm10-tRNAclosed	Trm10-tRNAopen	(Trm10)2-tRNA
Deposition
Coordinates (PDB)	9XZQ	9XZR	9XZS
Map (EMDB)	EMD-72368	EMD-72369	EMD-72370
Data collection and processing
Microscope	TFS Titan Krios	TFS Titan Krios	TFS Titan Krios
Camera	Gatan K3	Gatan K3	Gatan K3
Voltage, kV	300	300	300
Magnification ()	105,500	105,500	105,500
Electron exposure, e−/Å2	59.82	59.82	59.82
Defocus range, μm	−0.8 to −2.0	−0.8 to −2.0	−0.8 to −2.0
Pixel size, Å	0.412	0.412	0.412
Symmetry	C1	C1	C1
No. particles,
initial	5,634,445	5,634,445	244,276
final	1,398,183	1,440,392	156,897
Map resolution (FSC 0.143), Å	3.63	3.37	3.89
Refinement and model
Model resolution (FSC 0.143, unmasked), Å	3.9	3.4	4.4
CCmask	0.76	0.73	0.54
Non-hydrogen atoms	3090	2981	4513
Protein residues	189	177	360
RNA residues	71	71	71
SAM analog NM6 (AN6 “ligand”)	1	1	1
B factors (min./ max./ mean), Å2
Protein	170.4/399.5/252.9	94.4/299.3/176.8	100.8/364.0/225.0
RNA	202.1/458.0/262.8	154.4/357.2/213.5	182.4/1016.1/359.8
SAM analog NM6 (AN6 “ligand”)	249.8/249.8/249.8	154.4/154.4/154.4	190.1/190.1/190.1
RMS deviations
Bond lengths, Å	0.016	0.013	0.011
Bond angles, °	1.643	1.402	1.271
Validation
MolProbity score	1.87	2.57	2.40
Clashscore	8.67	12.9	9.7
Rotamer outliers, %	2.3	3.1	2.4
Ramachandran plot (protein)
Favored, %	97.3	89.1	87.9
Allowed, %	2.7	10.9	11.3
Disallowed, %	0.0	0.0	0.8

Structural figures were prepared using PyMOL (v3.1) or Chimera X (v1.9) (65, 66), using the DeepEMhancer- and the CryoSPARC-sharpened maps. Analysis of the tRNA conformational differences between free and Trm10-bound tRNA^Gly-GCC were obtained using the modevectors.py script available from PyMOLWiki (36). Buried surface area calculations were performed using the PDBePISA webserver (v1.52) (67). Average residue interaction energy, which includes nonbonded electrostatic and van der Waals terms, was calculated over all residue pairs (protein–protein, RNA–RNA, and protein–RNA) in Maestro (BioLuminate module, Schrödinger 2024–3) using the OPLS4 force field.

Structural modeling of target nucleotide interactions and MD simulations

Models of the Trm10-tRNA complex investigating G9/A9 specificity in Trm10 were generated using the Schrödinger software suite (version 2024–4). A monomeric, open Trm10-tRNA complex with a G9 to A9 mutation was made using the Maestro mutagenesis tool, and the resulting variant structure (Trm10-tRNA_open-A9) energy minimized to relieve steric clashes and optimize geometry. Wild-type and mutant complexes included tRNA nucleotides 1 to 71 and Trm10 residues 87 to 262 (i.e., excluding the disordered NTD), and NM6 was replaced with SAM in its established binding site from a previously reported structure (PDB 4JWJ).

Each complex was prepared for MD simulation using the Protein Preparation Wizard in Maestro to assign bond orders, add hydrogens, and set protonation states appropriate for pH 7.5 using Epik. MD was then set up with the OPLS4 force field and solvation in an orthorhombic TIP3P water box, and system neutralization at physiological ionic strength was accomplished with 150 mM NaCl. System relaxation followed the default multistep relaxation protocol in Desmond, including restrained minimization, restrained and unrestrained equilibration stages, and gradual heating from 0 K to 310.15 K. The final equilibration phase was carried out for 10 ns under NPT conditions at 310.15 K and 1 atm pressure, using the Nose-Hoover thermostat and Martyna–Tobias-Klein barostat. For each system, production simulations were run for 100 ns in the NPT ensemble, with temperature maintained at 310.15 K and pressure at 1 atm. Three independent replicates were performed with unique initial velocities, and coordinates were saved every 100 ps for analysis.

Nuclease (RNase T1) footprinting

³²P-5’-end labeled in vitro transcribed tRNA was subjected to partial degradation with RNase T1 (0.01 U/ml; Ambion) either in the absence or presence of Trm10 (at 1 or 5 M final concentration) for 1 hour at 37 °C. Reactions were stopped by the addition of phenol:chloroform:isoamyl alcohol (25:24:1), purified by phenol extraction and ethanol precipitation, and purified RNA fragments were run on a 10% polyacrylamide denaturing (8M urea) gel. Gels were dried and exposed to a phosphor screen and scanned using TyphoonTM imaging system (GE Healthcare) and quantified using ImageQuantTM TL software (GE Healthcare). All reactions were done in duplicate and run on the same gel to reduce variability. Alkaline hydrolysis ladders were prepared by incubating the labeled tRNA in 50 mM sodium carbonate for 10 minutes at 75 °C and stopped by the addition of denaturing gel loading buffer.

Bis(sulfosuccinimidyl)suberate (BS3) protein crosslinking

Trm10 and tRNA^Gly-GCC were dialyzed against 20 mM HEPES pH 7.5 buffer containing 150 mM NaCl, and 5 mM BME. Immediately before use, BS3 crosslinking reagent was prepared in water to a final concentration of 12.5 mM and added in a 50-fold molar excess to samples containing 6 μM Trm10 in the presence or absence of tRNA^Gly-GCC and SAM. Samples were incubated at room temperature for 30 minutes and quenched with Tris pH 7.5 to a final concentration of 50 mM. Samples were analyzed using a 9% SDS-PAGE gel and visualized by staining with Coomassie blue.

Trm10 protein family sequence analysis

The Trm10 protein sequence from S. cerevisiae was used as a query to search for tRNA methyltransferase Trm10-type domain protein family members in fungi in the UniProtKB database. The search yielded 1687 sequences for which Uniref50 IDs were assigned. After removing redundant entries based on UniRef50 clustering, 321 unique fungal sequences were retained for multiple sequence alignment in Geneious Prime using the Blosum62 matrix, followed by calculation of residue identity.

Funding Statement

National Institutes of Health grant R01 GM130135 (JEJ, GLC)

National Science Foundation GRFP (SES)

NIH Common Fund grant U24 GM129539 (NYSBC)

Simons Foundation award SF349247 (NYSBC)

NY State Assembly (NYSBC)

Data and materials availability:

Structural coordinates and EM maps have been deposited in the Protein Data Bank (PDB) and Electron Microscopy Data Bank (EMDB), respectively, with accession codes 9XZQ and EMD-72368 (Trm10-tRNA_closed), 9XZR and EMD-72369 (Trm10-tRNA_open), and 9XZS and EMD-72370 ((Trm10)₂-tRNA). All other data are available in the main text or the supplementary materials.