Human TRMT2A methylates tRNA and contributes to translation fidelity

Abstract 5-Methyluridine (m5U) is one of the most abundant RNA modifications found in cytosolic tRNA. tRNA methyltransferase 2 homolog A (hTRMT2A) is the dedicated mammalian enzyme for m5U formation at tRNA position 54. However, its RNA binding specificity and functional role in the cell are not well understood. Here we dissected structural and sequence requirements for binding and methylation of its RNA targets. Specificity of tRNA modification by hTRMT2A is achieved by a combination of modest binding preference and presence of a uridine in position 54 of tRNAs. Mutational analysis together with cross-linking experiments identified a large hTRMT2A–tRNA binding surface. Furthermore, complementing hTRMT2A interactome studies revealed that hTRMT2A interacts with proteins involved in RNA biogenesis. Finally, we addressed the question of the importance of hTRMT2A function by showing that its knockdown reduces translation fidelity. These findings extend the role of hTRMT2A beyond tRNA modification towards a role in translation.

Removal of the catalytic activity of TrmA does not impact the viability of E. coli , although deletion of the whole gene is lethal ( 13 ).In contrast, in budding yeast, Trm2 is not essential for survival ( 9 ).Hence, these phenotypes suggest a supporting role for the enzyme in tRNA ma tura tion and other as yet unknown biological processes beyond m 5 U catalysis ( 14 ).For instance, E. coli TrmA has recently been reported to be a tRN A cha per one assisting pr oper tRNA folding ( 15 ).Moreover, TrmA methylates not only tRNA, but also the T-loop of transfer mRNA ( 16 ) and 16S rRNA in vitro ( 17 ).
Detailed biochemical and structural studies on E. coli TrmA have contributed to a better understanding of the general enzymatic mechanism f or m 5 U f ormation.Most methyltr ansfer ases, including TrmA as well as human TRMT2A (hTRMT2A), use SAM ( S -adenosylmethionine) as methyl donor ( 10 , 12 , 18 ).Previous mutational studies identified catal yticall y r elevant r esidues in TrmA.Enzymatic activity as well as RNA binding was reduced upon mutation of the catalytic cysteine (C324A).The putati v e SAM-binding site (G220D), the proton abstraction base for the catalytic reaction (E358K) and one of the residues interacting with target uridine (Q190A) were also important for RNA binding ( 18 ).Furthermor e, r esidues H125A and F106 reduced tRNA binding and methylation, since they disrupt the tertiary interaction of the T-and D-loop ( 15 ).Such biochemical studies were supported by a co-crystal structure of TrmA with the T-loop of tRNA ( 19 ), confirming that the conserved C324 covalently binds to the target uridine, thus driving m 5 U formation.
Onl y recentl y our knowledge about m 5 U tRNA methyltr ansfer ases was expanded to human TRMT2A.Transcriptome-wide mapping of hTRMT2A methylation sites on RNA using FICC-CLIP (fluorouracil-inducedcatalytic cross-linking-cross-linking and immunoprecipitation) hav e re v ealed that 87% of the cellular hTRMT2A targets are indeed tRNAs ( 10 ).Binding sequences with the motif CTT CG / AA wer e enriched, which constitutes a conserved tRNA T-loop sequence.However, hTRMT2A binding expands beyond tRNAs as observed by its interaction with the HIST1H4B or KCND2 mRNA, small subunit rRNA and other RNAs ( 20 , 21 ).In line with such a broader function, its mammalian paralog TRMT2B was shown not only to be the dedicated mitochondrial methyltr ansfer ase f or m 5 U54 f ormation in mitochondrial tRNAs, but also to promote m 5 U429 formation of mitochondrial 12S rRNA ( 11 , 12 ).It remains to be shown whether hTRMT2A also has a dual function such as methylating tRNA and rRNA.
The cellular role of hTRMT2A, m 5 U formation and possible roles beyond m 5 U formation are not well understood, although some functional observations have been made; while in one study, hTRMT2A ov ere xpression correlated with Her + positi v e cancer ( 22 ) in another study hTRMT2A was described as a cell cycle regula tor tha t suppresses cell proliferation ( 23 ).In a third study, hTRMT2A knockout caused accumulation of tRNA-deri v ed fragments (tRFs) ( 24 ).Although the hTRMT2A-bound RNAs have been identified ( 10 ), we lack an understanding of its protein interactome and hence an ov ervie w of the molecular pathways to which it contributes.Furthermore, to date, all molecular properties of hTRMT2A have been inferred from orthologs in single-cell organisms.
In this study, we systematically characterized the functional and molecular features of hTRMT2A.Moreover, we identified hTRMT2A's determinants of specificity.We observed that hTRMT2A has only a modest binding preference for its physiological targets in vitro and that target specificity is rather achie v ed by the r equir ement of a uridine in the correct steric position to allow for site-specific methylation.Assessment of the cellular human TRMT2A interactome allowed us to identify potential cofactors, suggesting novel roles for hTRMT2A in different cellular processes.Finally, we show that loss of hTRMT2A in a cellular context resulted in reduced translation fidelity.
For the Central and Fusion domains, lysis buffer-B was used [500 mM NaCl, 50 mM HEPES-NaOH pH 7.5, 0.1% (v / v) Tween, 20 mM imidazole, 10% gly cerol].Moreov er, all buffers contained HEPES-NaOH with pH 7.5, and for the final size exclusion chromato gra phy a Super de x 200 column (16 / 600) and SEC-B buffer were used (50 mM HEPES-NaOH pH 7.5, 500 mM NaCl, 1 mM DTT).The Fusion domain was otherwise purified in the same way as the hTRMT2A RBD.The SUMO-tag of the Central domain was not cleaved due to subsequent protein degradation.Instead, after overnight dialysis in dialysis buffer-B (50 mM HEPES-NaOH pH 7.5, 150 mM NaCl, 1 mM DTT), the protein was loaded onto a HiTrap Heparin column equilibrated with dialysis buffer-B and eluted with Hep-B elution buffer (50 mM HEPES-NaOH pH 7.5, 1000 mM NaCl, 1 mM DTT).Protein-containing fractions were processed as described above.Pooled protein peak fractions were concentrated to 2-5 mg / ml with a centrifugal filter (Amicon), flash-frozen in liquid nitrogen and aliquots were stored at -80 • C. Primers for cloning and mutagenesis are listed in Supplementary Table S1

Human TRMT2A FL cloning, expression in insect cells and purification
The Bac-to-Bac baculovirus expression system (Invitrogen) was employed for expression of hTRMT2A fulllength (hTRMT2A FL).The N-terminal SUMO tag-fused hTRMT2A FL and its mutants were cloned into the pFast-Bac vector.For transfection using the FuGene HD transfection reagent, 1.5 g of bacmid DNA and 3 l of transfection r eagent wer e pr e-incubated for 30 min in 200 l of Sf-900 medium at room temperature, then added dropwise to the Sf21 cells pre-seeded on 6-well plates with a density of 0.4 × 10 6 cells / ml.The cells were then incubated for 4 days at 27.5 • C.
The supernatant of transfected cells from two 6-well plates was used to infect 10 ml of Sf21 shaking cell culture at a density of 1.4 × 10 6 cells / ml.After 3-4 days of incubation at 27.5 • C, insect cells were harvested by centrifugation for 10 min a t 4 • C a t 2000 rpm (Hettich centrifuge Rotina 420R).Virus-containing supernatant P1 was sterile-filtered using a 0.22 m filter (Merck).To amplify the virus, 1.5 ml of P1 was added to 250 ml of Sf21 shaking cell culture at a density of 0.4 × 10 6 cells / ml.After another 3-4 days of incuba tion a t 27.5 • C , cells were harvested.The viruscontaining supernatant P2 was sterile filtered and stored at 4 • C. Protein was expressed in Hi5 insect cells by adding 20 ml of P2 to 500 ml of shaking culture at a density of 1 × 10 6 cells / ml.Proteins were expressed in Hi5 insect cells for 2-3 days at 27.5 • C.
Cells were harvested by centrifugation (2000 g , 15 min, 4 • C) and resuspended in lysis buffer-C [500 mM NaCl, 50 mM HEPES-NaOH pH 7.5, 0.1% (v / v) Tween, 10% glycerol, 20 mM imidazole].Next, insect cells were lysed using a Dounce homogenizer.The lysate was clarified by centrifugation (40 000 g , 30 min, 4 • C).Subsequent purification was perf ormed as f or the hTRMT2A Fusion domain with an initial HisTrap FF column, then cleavage of the SUMO tag was followed by a subtracti v e HisTrap FF column, dialysis overnight with dialysis buffer-A and a final size exclusion chromato gra phy run with a Super de x 200 column (16 / 600) in SEC-B buffer.hTRMT2A FL for cross-linking and mass spectrometry (MS) experiments was purified with SEC-C buffer (50 mM HEPES-NaOH pH 7.5, 500 mM NaCl, 0.5 mM T CEP), wher e DTT was exchanged by TCEP.Pooled protein peak fractions were concentrated to 2-5 mg / ml with a centrifugal filter (Amicon), flash-frozen in liquid nitrogen and aliquots were stored at -80 • C.

RNA in vitro transcription and purification
All RNAs were transcribed from a high-performance liquid chromato gra phy (HPLC)-purified forward primer, which contained the T7 RN A pol ymerase promoter region and a polyacrylamide gel electrophoresis (PAGE)-purified re v erse primer that consisted of the re v erse complement DNA template and the T7 RN A pol ymerase promoter region (Supplementary Table S2).In brief, 4 M forward primer and 3.3 M re v erse primer were annealed in the presence of 20 mM MgCl 2 for 5 min at 60 • C in a total volume of 1.2 ml and subsequently cooled to room temperature.The 4 mM of each NTP, 16 mM MgCl 2 , 80 mg / ml polyethylene glycol (PEG) 800, 0.5 mg / ml T7 RN A pol ymerase and 10 × TRX-buffer (40 mM Tris-HCl pH 8.1, 1 mM spermidine, 0.01% Triton X-100, 5 mM DTT) were added to the annealed primer pair in a final volume of 5 ml.After incubation for 3 h a t 37 • C , the precipita te was removed by centrifugation for 10 min at 4 • C at 14 000 g .Thereafter the RNA was precipitated with 0.1 volume of 3 M sodium acetate and 3.5 volumes of ethanol for > 30 min at -20 • C. RNA pellets wer e r esuspended in 0.5 ml of diethylp yrocarbona te (DEPC)-trea ted H 2 O, and 0.5 ml of denaturing loading dye was added [2 mM Tris-HCl pH 7.5, 20 mM EDTA, 8 M urea, 0.025% (w / v) bromophenol blue, 0.025% (w / v) xylene cy anol].Separation b y 8% denaturing Trisboric acid-EDTA (TBE)-PAGE was followed by visualization of transcripts by UV shadowing and excision from the gel.RNA was extracted by electroelution (W ha tman Elutrap, GE Healthcare) at 200 V in 1 × TBE according to the man ufacturer's instructions.F ractions containing the RNA were pooled, and dialyzed twice against 5 M NaCl and once against DEPC-H 2 O. RNAs were lyophilized, subsequently resuspended in DEPC-H 2 O and stored at -20 • C. General recommendations for working with RNAs w ere follow ed ( 25 ).

RNA r adiolabeling and electrophor etic mobility shift assays (EMSAs)
In vitro transcribed RNA was 5 dephosphorylated using FastAP ™ thermosensiti v e alkaline phosphatase (Thermo Scientific) according to the manufacturer's recommendations.Thereafter, RNA was purified with phenolchloroform-isoamylalcohol extraction followed by isopropanol precipitation.Purified RNA (13 pmol) was phosphorylated with [ ␥ -32 P]ATP (Hartmann Analytics) by T4 polynucleotide kinase (NEB) for 30 min at 37 • C. The enzyme was deactivated at 75 • C for 10 min and labeled RNA was purified with a NucAway Spin column (Ambion).To ensure proper RNA folding after each thawing cycle, RNA was denatured at 98 • C for 2 min and snap-cooled on ice prior to use.
For 20 l EMSA r eactions, incr easing protein concentrations (0-25 M) were incubated with 100 nM ␥ -32 P-labeled RNA in the presence of 2.5 ng / l poly(U) competitor (polyuridine single-stranded RNA, Sigma) for 20 min at room temperatur e in RNase-fr ee EMSA buffer [50 mM HEPES-NaOH pH 7.5, 150 mM NaCl, 4% (v / v) glycerol].This results in a final concentration of 5 nM labeled RNA and 0-1.25 M protein per r eaction.Complex es wer e separated by 4% nati v e TBE-PAGE in TBE running buf fer a t 80 V.After gel fixation with 30% (v / v) methanol and 10% (v / v) acetic acid solution for 10 min, gels were dried in a vacuum gel dryer (model 573, BioRad).For analysis, phosphor imaging plates were exposed to the radioacti v e dried gels for > 20 min and scanned with a FLA-300 system (Fujifilm).EM-SAs were performed in triplicate on different days.

Methyltr ansfer ase assay
hTRMT2A FL methylation activity was analyzed with the MTase-Glo ™ assay (Promega).The assay was performed according to the manufacturer's instructions and experiments described in ( 26 ) in white-opaque OptiPlate ™-384 (PerkinElmer).Experiments to determine K m values were performed with increasing substrate concentrations.A final concentration of 5 M hTRMT2A FL was incubated with a final concentration of 0-6 M in vitro transcribed tRNA and 20 M SAM in 1 × reaction buffer [20 mM Tris buffer pH 8.0, 50 mM NaCl, 1 mM EDTA, 3 mM MgCl 2 , 0.1 mg / ml bovine serum albumin (BSA), 1 mM DTT] for 60 min at room temperatur e.The r eaction was stopped by adding 0.5% trifluoroacetic acid (TFA; PanReac AppliChem).After 5 min incubation, 6 × MTase-Glo reaction solution diluted in MilliQ-H 2 O was added and incubated for 30 min at room temperature.Finally, MTase-Glo detection solution was added, incubated for 30 min in the dark at room temperature and luminescence was read using a microplate reader (Perkin-Elmer).The data were transformed to display relati v e luminescence by subtracting the value of a negati v e control reaction without RNA.Data of triplicate measurements were plotted with R (version 4.1.0)and fitted with the Michaelis-Menten equation to obtain K m values.

Cross-linking and mass spectrometry
For the first set of e xperiments, comple x reconstitution was performed; to do so, in vitro transcribed tRNA Gln and recombinantly purified hTMRT2A FL were reconstituted in reconstitution buffer (50 mM HEPES-NaOH pH 7.5, 50 mM NaCl, 0.5 mM TCEP) in a 1:4 ratio.For the second set of experiments, the complex was co-purified using size exclusion chromato gra phy.Briefly, tRN A Gln and hTRMT2A FL were incubated in a 1:1 ratio on ice for 30 min.Then, the complex was separated from single RNA and protein using an S6 (15 / 150) column.For UV cross-linking, the in vitro reconstituted and co-purified complex was irradiated at 254 nm for 10 min on ice using an in-house-built crosslinking apparatus as described previously ( 27 ).For chemical cr oss-linking, mechlorethamine [nitr o gen m ustared (NM) B1785, APExBIO Technology LLC] was added to a final concentration of 1 mM following incubation for 30 min at 37 • C. Cross-linked samples wer e pr ecipitated using ethanol and further processed as described in Kramer et al .( 27 ) In brief, 10 g of RNase A (EN0531, Thermo Fisher Scientific), 1 kU of RNase T1 (EN0531, Thermo Fisher Scientific) and 250 U of Pierce ™ uni v ersal nuclease (88700, Thermo Fisher Scientific) were used for tRNA digestion that was carried out for 2 h a t 37 • C .Subsequently, proteins were digested using trypsin (V5111, Promega) at a 1:20 (enzyme to protein) mass ratio.After protein digestion, another 10 g of RNase A and 1 kU of RNase T1 were added and the samples were incubated for 1 h at 37 • C. F ree n ucleotides and salts wer e r emoved using C18 columns (744601, Harvard Apparatus), and cross-linked peptides were enriched using in-house-packed TiO 2 columns (Titansphere 5 m; GL Sciences).Enriched peptides were dried and dissolved in 2% (v / v) acetonitrile, 0.05% (v / v) TFA.Liquid chromato gra phy-tandem MS (LC-MS-MS) analyses were performed on a Thermo Scientific Orbitrap Exploris 480 mass spectrometer coupled to a nanoflow liquid chromato gra phy system (Thermo Scientific Dionex Ultimate 3000).Peptide separation was performed over 58 min with a flow rate of 300 nl / min using a linear gradient and a buffer system consisting of 0.1% (v / v) formic acid (buffer A) and 80% (v / v) acetonitrile, 0.08% (v / v) formic acid (buffer B).Peptide-nucleotide hetero-conjugates were analyzed in positi v e mode using a data-dependent top 30 acquisition method.MS1 and MS2 resolution were set to 120 000 and 30 000 FWHM (full width at half-maximum), respecti v ely.AGC targets were set to 10 6 (MS1) and 10 5 (MS2), normalized collision energy to 28, dynamic exclusion to 10 s and maximal injection times to 60 ms (MS1) and 120 ms (MS2).MS data were analyzed and manually validated using the OpenMS pipeline RNPxl and Open MS TOPPASViewer ( 27 ).All cross-links are listed in Supplementary Table S3.

Surface plasmon resonance (SPR)
SPR studies were performed using a BIACORE S200 system (GE Healthcare).tRNA Gln was 3 biotin labeled with the Pierce ™ RNA 3 end biotin labeling kit (Thermo Fisher) according to the manufactur er's r ecommendations.SA-chip (GE-Healthcare) was prepared according to the manufacturer's instruction before biotin-labeled RNA diluted in DEPC-H 2 O was coupled.Analysis of RNA-protein interactions was performed at a flow rate of 30 l / min in running buffer [150 mM NaCl, 50 mM HEPES-NaOH 7.5, 0.05 % (v / v) Tween] at 25 • C and an a 10 Hz data collection rate.The analyte of interest was diluted in running buffer and single-cycle concentration series were injected on the chip with 240 s contact time per concentration and a final dissociation time of 900 s.To remove any residual attached protein after the last analyte injection, 2 × 2 min regeneration injections with 0.5% (w / v) sodium dodecylsulfate (SDS) were done in between runs.Concentration series reached from 62.5 nM to 4000 nM for hTRMT2A wild-type (WT) and 15.6-1000 nM for hTRMT2A CDM (ca talytic dea th mutant) protein.Da ta were analyzed with the BIAevaluation software (GE Healthcare).Obtained binding curves were double-r efer enced against the signal in the empty r efer ence channel and a buffer run.At equilibrium of the binding curves, the corresponding response was plotted against the analyte concentration.The K D was determined by fitting this curve to a steadysta te af finity model.All experiments were performed in triplicates.

Pr oximity-dependent biotin ylation
BioID experiments were conducted to identify the hTRMT2A WT and hTRMT2A CDM proximity interactome in HEK 293T cells as described earlier ( 28 ) with fiv e replicates per condition.At 70% cell confluency, protein expression of BirA*-tagged protein of interest or BirA* tag alone was induced by the addition of doxy cy cline.After 6 h, 1 ml of sterile-filtered 20 × Dulbecco's modified Eagle's medium (DMEM) / biotin was added to the cells to reach a final concentration of 50 M biotin.Cells were incubated for 16-18 h.Prior to trypsinization, cells were washed three times with phospha te-buf fered saline (PBS).Detached cells (8 × 10 7 to 1 × 10 8 per r eplicate) wer e centrifuged for 5 min a t 4 • C a t 300 g .Cells wer e r esuspended in 5 ml of lysis buffer [50 mM Tris-HCl pH 7.4, 500 mM NaCl, 0.2% SDS) freshly supplemented with 1:1000 benzonase (Thermo), 1 mM DTT and Pierce ® Protease Inhibitor (Roche).Thereafter, the cell suspension was incubated on a rotating wheel at 4 • C for 30 min.After 500 l of 20% Triton X-100 was added, the samples were sonicated for 1 min with a 30% duty cycle (Bandelin Sonopuls GM 70UV 70).Three sonication rounds were performed, and the samples were incubated on ice for 5 min in between them.Before the third sonication round, 4.5 ml of pre-chilled 50 mM Tris-HCl pH 7.4 was added.Sonicated samples were centrifuged for 10 min at 4 • C with 16 100 g .A 500 l aliquot of Dynabeads MyOne Streptavidin T7 magnetic beads (Thermo) per condition was equilibrated in 1.5 ml of lysis buffer and 1.5 ml of 50 mM Tris-HCl, pH 7.4.Beads were placed in a magnetic stand (DYNAL ®, Invitr ogen).Then, the pr otein-containing supernatant from the centrifugation step was added to the equilibrated beads.Solution and beads were incubated on a turning wheel overnight at 4 • C. The supernatant was removed from the beads and substituted with 8 ml of wash buffer 1 (2% SDS in MilliQ-H 2 O).The wash step with wash buffer 1 was r epeated.Next, beads wer e w ashed once with w ash buffer 2 (0.1 sodium deoxycholate, 1% Triton X-100, 1 mM EDTA, 500 mM NaCl, 50 mM HEPES-NaOH pH 7.5) and wash buffer 3 (0.5% sodium deoxycholate, 0.5% NP-40, 1 mM EDTA, 250 mM LiCl, 10 mM Tris-HCl pH 7.4).To remove detergent, beads were washed three times with wash buffer 4 (1 mM EDTA, 20 mM NaCl, 50 mM Tris-HCl pH 7.4).Washed beads wer e r esuspended in 1.5 ml of 50 mM Tris-HCl pH 7.4 and then put on a magnetic stand to remove supernatant.For elution, 50 l of 1 × sample buffer (50 mM Tris-HCl pH 6.8, 12% sucrose, 1% SDS, 0.004% bromophenol blue, 20 mM DTT) supplemented with 3 mM biotin was added to beads, mixed and incubated at 98 • C for 7 min.Subsequently, samples were placed on a magnetic stand, and the supernatant subjected to MS or analytical western blot analysis.Enriched and depleted proteins are listed in Supplementary Table S4.

Co-immunoprecipitation (Co-IP)
Co-IP experiments were performed to identify the hTRMT2A proteome in HEK 293T cells with fiv e replicates per condition.At 70% cell confluence, expression of the protein of interest or FLAG tag alone was induced with doxy cy cline.After 16-18 h, green fluorescent protein (GFP) expression was inspected with fluorescence microscopy and cells were washed twice with PBS.For cell harvest, 10 ml of ice-cold PBS was added and cells were scraped off the plate.The cell suspension was centrifuged for 5 min at 4 • C at 300 g .Pellets were flash-frozen in liquid nitrogen to ensure equal sample time points.
Cell pellets w ere thaw ed and resuspended in 1 ml of lysis buffer (50 mM Tris-HCl pH 7.8, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100) supplemented with one tablet of Pierce ® Protease Inhibitor (per 100 ml of lysis buffer).The suspension was transferred to a new tube and centrifuged for 10 min at 4 • C at 16 100 g .Meanwhile, 50 l of Anti-FLAG ® M2 Magnetic Beads (Merck) were equilibrated twice with 900 l of Tris-buffered saline (TBS; 50 mM Tris-HCl pH 7.8, 150 mM NaCl) and three times with 900 l of lysis buffer.In between wash steps, the beads were incubated for 3 min on a magnetic stand (DynaMag 2, DYNAL ®, Invitrogen).Then, 3.5 mg of protein from the supernatant of the centrifugation step was added to the equilibrated beads.Protein concentration was measured with the Bradford assay solution (Roth) according to instructions.Beads were incubated for 1 h at 4 • C in the shaking mode on a rotator wheel.Thereafter, unbound protein was removed by washing three times with 900 l of high salt wash buffer (50 mM Tris-HCl pH 7.8, 250 mM NaCl, 1 mM EDTA, 1 × Triton X-100).The third wash step was incubated for 5 min at 4 • C in shaking mode on a rotator wheel.To remove detergent, beads were washed three times with 900 l of 50 mM Tris-HCl, pH 7.8.Beads were distributed to two tubes for elution: 900 l for MS and 100 l for western blot analysis.Elution was performed by addition of 20 l of 2 × Laemmli buffer [4% (w / v) SDS, 20% glucose, 0.004% bromophenol blue, 125 mM Tris-HCl pH 6.8] and incubation for 10 min a t 98 • C .Finally, tubes were placed onto the magnetic stand and the supernatant was transferred to a new tube for respecti v e MS analysis and analytical western blot analysis.Enriched and depleted proteins are listed in Supplementary Table S5.

LS-MS / MS analysis of BioID and Co-IP samples
Proteins were proteolyzed with LysC and trypsin with the filter-aided sample preparation procedure (FASP) as described ( 29 , 30 ).Acidified eluted peptides were analyzed on a QExacti v e HF-X mass spectrometer (Thermo Fisher Scientific) online coupled to an UItimate 3000 RSLC nano-HPLC (Dionex).Samples were automatically injected and loaded onto the C18 trap cartridge and after 5 min eluted and separated on the C18 analytical column (Acquity UPLC M-Class HSS T3 Column, 1.8 m, 75 m × 250 mm; Waters) by a 95 min non-linear acetonitrile gradient at a flow rate of 250 nl / min.MS spectra wer e r ecorded at a resolution of 60 000 with an automatic gain control (AGC) target of 3e6 and a maximum injection time of 30 ms from 300 to 1500 m / z .From the MS scan, the 15 most abundant peptide ions were selected for fragmentation via HCD with a normalized collision energy of 28, an isolation window of 1.6 m / z and a dynamic exclusion of 30 s. MS / MS spectra wer e r ecorded at a resolution of 15 000 with an AGC target of 1e5 and a maximum injection time of 50 ms.Unassigned char ges and char ges of +1 and > 8 were excluded from the precursor selection.
Data pr ocessing pr otocol.Raw spectra were imported into Progenesis QI software (version 4.1).After feature alignment and normalization, spectra were exported as Mascot Generic files and searched against the human UniProt database (20 434 sequences) with Mascot (Matrix Science, version 2.6.2) with the following search parameters: 10 ppm peptide mass tolerance and 20 mmu fragment mass tolerance, one missed cleavage allowed, carbamidomethylation was set as fixed modification, methionine oxidation and asparagine or glutamine deamidation were allowed as variable modifica tions.A Mascot-integra ted decoy da tabase search calculated an average false discovery rate (FDR) of < 1% when sear ches wer e performed with a mascot per colator score cut-off of 13 and an appropriate significance threshold P -value.P eptide assignment w as re-imported into the Progenesis QI software and the abundances of all unique peptides allocated to each protein were summed up.The resulting normalized abundances of the individual proteins were used for the calculation of fold changes of protein ratios between conditions.Proteins with a sample / control ratio of at least 4 wer e consider ed differ entially abundant.Statistical analysis was performed on log 2 -transformed normalized abundance values using Student's t -test.Changes in protein expression between conditions wer e consider ed significant at P < 0.05.Principal component analysis was performed in Progenesis QI software.

GO-term analysis
'Biological processes' (BP) and 'cellular compartments' (CC) Gene Ontology (GO) ter m analysis was perfor med with data from BioID experiments using the GO enrichment analysis tool from GO employing the PANTHER databank ( 31 ).UniPr ot pr otein names of significantly enriched hits from BioID experiments were used for GO enrichment analysis.Significance of GO term enrichment was calculated as the FDR with Fisher's exact test.Using custom R-scripts, data were visualized as enrichment of respecti v e GO terms on the x-axis, the significance of enrichment with color code and the counts per GO term with node size.

Generation of BioID and Co-IP stably transfected cell lines
For stable transfection of HEK 293T cells, a piggyBac (PB) transposon plasmid was used ( 32 , 33 ).Co-transfection with the PB helper plasmid, coding for PB transposase, allows for stable insertion of the gene of interest into the chromosome.For transfection, HEK 293T cells were plated in a 6-well plate and grown to 80% confluency.According to the Lipofectamine ™ 3000 transfection kit (Invitrogen), 2.5 g of PB plasmid with the gene of interest and 2.5 g of helper plasmid were transfected into HEK 293T cells and incubated for 3 days.For antibiotic selection, complete medium [Gibco ™ DMEM + 10% fetal bovine serum (FBS) + 1% PenStrep] supplemented with hygromycin B (0.7 mg / ml, Invitrogen) was added and exchanged 2-3 times a week for stable integration of the PB plasmid into the genome.
The same expression level of the protein of interest was achie v ed by fluorescence-acti vated cell sorting (FACS).Protein expression in cells was induced with doxycycline 18 h before the FACS experiment.For FA CS , the cells were detached with trpysin, washed twice with PBS for 5 min and resuspended in sterile FACS buffer (0.5% BSA, 2 mM EDTA, 25 mM HEPES-NaOH pH 7.5 in PBS).Directly prior to the FACS experiment, the cells were passed through a 35 m cell strainer.The first gating step etracted dead cells and aggregates.A second gate was set to sort the cell population with a low GFP signal.Sorted cells were plated in 6-well plates and propagated with complete medium containing hygromycin B.

Generation of stable RNAi-mediated KD of hTRMT2A
Stable RNA interference (RNAi)-mediated KD of hTRMT2A was achie v ed as described earlier ( 34 ).HEK 293T cells were infected with commercially available lentiviral particles for hTRMT2A KD and a scrambled control cell line (MISSION ® shRNA Lentviral Transduction Particles: scrambled control WT = SHC002V, KD1 = NM 182984.2-856s1c1, KD2 = NM 182984.2-1574s1c1; Sigma-Aldrich).Cells with stable integration of short hairpin RN A (shRN A) were selected as puromycinresistant colonies.hTRMT2A silencing effect was probed by western blotting.

Ribosome isolation and MS analysis of rRNA modification status
Ribosome isolation from hTRMT2A WT, KD1 and KD2 HEK 293T cells was performed according to a protocol published earlier ( 35 ).Briefly, 80% confluent cells were harvested in ice-cold PBS and centrifuged at 4 • C. The cell pellet was resuspended in an equivolume of buffer A (250 mM sucrose, 250 mM KCl, 5 mM MgCl 2 , 50 mM Tris-HCl pH 7.4).For cell lysis, 0.7% (v / v) NP-40 was added and incubated on ice for 10-15 min.To separate nuclei, the cell lysate was centrifuged for 10 min at 750 g and 4 • C. To pr epar e the fraction without mitochondria, the supernatant was centrifugated for 10 min at 12 500 g and 4 • C. Post-mitochondrial fraction (PTM) supernatant was transferred to a new tube and the KCl concentration was adjusted to 500 mM using a 4 M stock solution.The adjusted PTM was car efully layer ed on top of a 2 ml sucrose cushion (1 M sucrose, 500 mM KCl, 4 mM MgCl 2 , 50 mM Tris-HCl pH 7.5).Samples were ultracentrifuged for 2 h at 250 000 g and 4 • C. The supernatant was discarded and the translucent ribosome pellet w as w ashed with 200 l of cold DEPC-treated H 2 O. Finally, ribosomes were resuspended in 3 × 100 l buffer C (25 mM KCl, 5 mM MgCl 2 , 50 mM Tris-HCl pH 7.4).
For rRNA isolation, 100 l of ribosomes diluted in buffer C wer e mix ed with 15 l of DEPC-H 2 O and 15 l of 3 M NaOAc.Next, 1 volume of phenol-chloroformisoamylalcohol (25:24:1) was added.After centrifugation for 1 min at 16 100 g , the upper aqueous phase was mixed with 1 volume of chloroform and again centrifuged.The r ecover ed upper phase was precipitated in the presence of 0.5 l of RNA-grade glycogen and 2 volumes of ethanol for > 20 min a t -20 • C .RNA was pelleted for > 20 min at 16 100 g and 4 • C. The dry pellet was resuspended in 20 l of DEPC-H 2 O.
For isolation of 18S and 28S rRN A, purified rRN A was separated on a 0.8% agarose gel for 80 min at 120 V in 1 × TAE buffer.The bands for 18S and 28S rRNA were cut from the gel and electroeluted for 200 min at 150 V in 1 × TAE buffer.After electroelution, the rRNA was precipitated with 2 volumes of 100% ethanol for 30 min at -20 • C and further pelleted for 30 min at 13 200 rpm at 4 • C. To increase the yield, 1 l of RNA-grade glycogen per 1 ml of the solution was added.The pellet was dissolved in water and the samples were frozen in liquid nitrogen for further storage a t -80 • C .All r eagents used wer e RNase and sodium ion free.The integrity and purity of rRNA were checked with Ta peStation anal ysis done using the Agilent Ta peStation 4150 with the Agilent High Sensitivity RN A ScreenTa pes according to the manufacturer's instructions.

rRNA digestion to the nucleoside level
From 500 ng to 5 g of pre-purified rRNA subunits (20 l of each) were digested to single nucleosides by using 2 U of benzonase nuclease ( > 90% purity, Merck), 0.2 U of phosphodiesterase I (VWR) and 2 U of alkaline phosphatase (QuickCIP, NEB) in MgCl 2 (1 mM) containing Tris (pH 8, 5 mM) buffer.In a final volume of 30 l, 1 g of pentostatin, 5 g of tetrahydrouridine (Merck) and a final concentration of 10 M butylated hydroxytoluene were added to each reaction for protection against degradation of the released nucleosides.After incubation for 2 h at 37 • C, 20 l of LC-MS Buffer A (see below) was added to the mixture and then filtered through 0.2 m Supor Natural PP filters (Pall Corporation, AcroPrep Advance350, 96-well pla te) a t 3000 g and 4 • C for 30 min before measurement by LC-MS / QQQ.

LC-MS / QQQ analysis of nucleosides
For quantitati v e MS, an Agilent 1290 Infinity equipped with a variable wavelength detector (VWD) combined with an Agilent Technologies G6490 Triple Quad LC / MS system with electrospray ionization (ESI-MS, Agilent Jetstream) was used.Operating parameters: positi v e ion mode, cell accelerator voltage of 5 V, N 2 gas temperature of 120 • C and N 2 gas flow of 11 l / min, sheath gas (N 2 ) temperature of 280 • C with a flow of 11 l / min, capillary voltage of 3000 V, nozzle voltage of 0 V, nebulizer at 60 psi, high-pr essur e radiofrequency at 100 V and low-pressure radiofrequency at 60 V.The instrument was operated in dynamic MRM mode (Supplementary Table S6).For separation, an Uptisphere C18-HDO column (3.0 m, 150 × 2.1 mm from Interchim, UP3HDO-150 / 021) was used.Running conditions were 35 • C and a flow rate of 0.35 ml / min in combination with a binary mobile phase of 5 mM NH 4 OAc aqueous buffer A, brought to pH 4.9 with glacial acetic acid (200 l / l), and an organic buffer B of 2 mM NH 4 COOH in 80% acetonitrile (Roth, Ultra LC-MS grade, purity ≥ 99.98).The gradient started at 100% solvent A for 0.5 min, followed by an increase of solvent B to 10% over 5.5 min.From 6.0 min to 8.5 min, solvent B was increased to 20% then to 80% in 1 min and maintained at 80% for 1.5 min befor e r eturning to 100% solvent A in 0.5 min and a 2.2 min re-equilibration period.A 10 l aliquot of each sample was co-injected with 1 l of stable isotope-labeled internal standard (ISTD) which was aspirated automatically before each injection from the instrument itself.The sample data were analyzed by the quantitati v e MassHunter Software from Agilent using the integr ated calibr ation function.The calibration solutions ranged from 0.05 to 100 pmol for each canonical nucleoside and from 0.002 to 5 pmol for each modified nucleoside (12 calibration le v els, 1:2 dilution).
The measured molar amount of each canonical nucleoside (pmol) was divided by their expected occurrence in the respecti v e rRNA subunit reported in the literature [# per molecule from modomics ( 4 Cells on coverslips wer e fix ed for 10 min at room temperature using 3.7% formaldehyde solution in PBS.After two PBS washing steps, cells were permeabilized for 5 min with 0.5% Triton X-100 in PBS.Next, cells were blocked for 30 min using blocking buffer (1% goat serum in PBS-T).Primary antibody diluted in PBS was added for 1 h at room temperatur e.After thr ee washing steps with PBS-T, cells were incubated with secondary antibody diluted in blocking buffer for 1 h at room temperature in the dark.Cells were washed three times with PBS-T.Next, the cell nuclei were stained with 4 ,6-diamidino-2-phenylindole (DAPI) at a concentration of 0.5 g / ml DAPI (Thermo) in PBS for 5 min at room temperature and washed twice with PBS.Finally, coverslips were mounted using the ProLong ™ Diamond Antifade Mountant (Invitrogen).Cells were imaged with the Leica DMi8 fluorescence microscope and the Leica imaging software.Images were analyzed with the Fiji ImageJ software.The primary antibody used was mouse anti-BirA (1:10; Novus Biologicals, #5B11c3-3) and the secondary antibody was goat anti-mouse IgG Alexa Fluor 647 (1:2000; Invitrogen, #A-21235)

Dual luciferase translation assay
The dual luciferase assay to probe translation fidelity was performed using three Renilla-Firefly fusion protein constructs.In two Renilla sequences, mutations were introduced at residues crucial for luciferase activity ( 36 ).hTRMT2A WT (scrambled control) as well as KD1 and KD2 (scrambled control) wer e cultur ed (Gibco ™ DMEM + 10% Gibco ™ FBS + 1% Gibco ™ Antibiotic-Antimycotic) and seeded at a density of 20 000 cells / well in a 96-well plate suitable for luminescence measurements (greiner bio-one #655075).Cells were transfected with Renilla and Firefly fusion constructs (WT, D120N, E145Q) using T r ans IT ®-LT1 Transfection Reagent (Mirus Bio) according to the manufacturer's instructions.After 48 h of incuba tion a t 37 • C and 5% CO 2 , the culture medium was exchanged with fresh growth medium, and luciferase activity was measured with the Dual-Glo ® Luciferase Assay Kit (Promega) according to the manufacturer's instructions using the Tecan Spark ® plate reader.Integration time was 10 s.Data r epr esent the mean ± standard deviation (SD) of six biological replicates, in technical triplicates.Statistical analysis was performed with the Wilco x on rank test and Student's t -test in R.

hTRMT2A sho ws lo w binding specificity for its physiological target tRNAs
To understand how the specificity of tRNA methylation is achie v ed by hTRMT2A, it is imperati v e to assess both the RNA binding and catalytic activity of hTRMT2A ( 10 ).For in vitro characterization of the TRMT2A binding prefer ences, we pr epar ed r ecombinant human TRMT2A fulllength (hTRMT2A FL, Supplementary Figure S1B) and dif ferent in vitr o transcribed RNAs (Supplementary Figure S1A).We assessed binding with an EMSA, where 32 Plabeled RNA was pre-incubated with increasing concentrations of hTRMT2A FL and poly(U) as a competitor to preclude non-specific binding.To explore how important the uridine in position 54 of a gi v en tRNA is for binding, we included two tRNAs with U54 (tRNA Gln and tRNA Phe ) and two tRNAs lacking U54.The latter consisted of tRNA Ala with A54 and a mutant version of tRN A Gln w here U54 had been changed into G (tRNA Gln U54G ).hTRMT2A FL bound all these tRNAs with apparent equilibrium dissociation constants ( K D ) in the low nanomolar range (Figure 1 A).A somewhat stronger binding was observed for tRN A Gln , w here a shifted band was already observed at a protein concentration of 60 nM.Of note, for all tRNAs except for tRNA Phe , we observed supershifts that imply multiple hTRMT2A enzymes binding to their respecti v e RNA.Howe v er, since these supershifts only occurred at higher, micr omolar pr otein concentrations, we assume that they r epr esent unspecific oligomerization without physiological importance.Also, for the negati v e control pol y(U)-RN A, the binding appeared only as a smear and not as a specific bandshift, indicati v e of a r ather unspecific inter action (Figure 1 B).
To dissect the structural r equir ements for hTRMT2A FL binding, we also performed EMSAs with the tRNA Gln Tloop and two structured fragments deri v ed from mRNA: a fragment of the KCND2 mRN A, w hich consists of the pre viously pub lished stem-loop target of hTRMT2A ( 10 ), and the unrelated E3 stem-loop of the S. cerevisiae ASH1 mRN A, w hich is not known to be methylated by hTRMT2A, but contains the previously identified sequence motif in the predicted loop recognized by hTRMT2A (Supplementary Figure S2C) ( 10 , 37 ).The tRNA Gln T-loop showed a weaker shift, with most of the RNA remaining in its free form, which suggests the formation of complexes with lower stability (Figure 1 B).Both stem-loopcontaining mRNAs displayed binding, howe v er in a high nanomolar to micromolar K D range (Figure 1 B), which indica tes lower af finity than observed with tRNAs.KCND2 showed a somewhat less defined binding than ASH1 E3, indicated by the lack of a distinctly shifted band.Howe v er, an additional band at higher molecular weight could point in the direction of a secondary binding e v ent.
In summary, these results suggest that hTRMT2A FL binding to tRNAs is independent of the presence of a U54 as a potential methylation site and that the enzyme binds in vitro stem-loop structured RNAs with modestly lower affinities than tRNAs.While binding affinities are in the expected range ( 15 ), binding specificity of hTRMT2A for its ph ysiological meth ylation targets appears to be low in vitro .

Enzymatic activity of hTRMT2A FL
To correlate our binding studies with enzymatic activities, we tested the methyltr ansfer ase activity of hTRMT2A FL on selected RNAs using the luminescence-based MTase Glo ™ assay (Figure 1 C).In this assay, relati v e luminescence corr elates dir ectly with enzymatic activity.hTRMT2A FL was incubated with increasing RNA concentrations to genera te sa tura ted K m curves for Michaelis-Menten fitting.A reaction without RNA was used as background control.We observed that hTRMT2A methylates tRNA Gln and tRNA Phe with a K m of 1.52 ± 0.05 and 0.72 ± 0.06 M, respecti v el y, w hich is in the range of pr eviously r eported K m values for methyltr ansfer ases ( 38 ).In contrast, the two tRNAs lacking U54 (tRNA Ala and tRNA Gln U54G ) were methylated orders of magnitude less efficiently (Figure 1 D, F).Taken together, binding assays and enzymatic assays indicate that the specificity of target tRNA methylation is mainly mediated by a uridine residue in the correct position and not by the affinity for these tRNAs per se .
KCND2 mRNA experiments resulted in higher overall luminescence, potentially deriving from a secondary methylation site in vitro (Supplementary Figure S2B).In agreement with unspecific binding (Figure 1 B), howe v er, KCDN2 mRNA was r equir ed in high concentrations and did not reach sa tura tion, indica ting non-physiological enzymatic activity (Figure 1 E, F).The presence of the RNA recognition motif of hTRMT2A in the unrelated, stemloop structured ASH1 E3 mRNA was sufficient for similar methylation to those of tRNA substrates, with a K m of 0.78 ± 0.15 (Figure 1 F; Supplementary Figure S2).Thus, hTRMT2A cannot methylate only substrate tRNA, but also other stem-loop structured RNAs.

tRNA is bound by hTRMT2A via multidomain interactions
Based on the results from binding and methylation experiments, we chose tRNA Gln to dissect the RNA-interacting domains of hTRMT2A.For this purpose, we generated se v er al hTRMT2A fr agments, whose domain borders were chosen based on experimental evidence ( 34 ), sequence alignments and homology models.We purified hTRMT2A FL, the RBD (amino acids 69-147) and the central domain (Central, amino acids 237-414).Unfortunately, the predicted entire methyltr ansfer ase domain (amino acids 179-581) was unsuitable for in vitro studies due to protein degrada tion.Instead, we crea ted a more stable protein fragment, in which the RBD is fused directly to a part of the methyltr ansfer ase domain (Fusion, amino acids 69-240 / 413-590) (Figure 2 A; Supplementary Figure S1C, D).In EMSAs, all fragments showed only weak mobility shifts in the micromolar protein range (Figure 2 C) and hence much weaker binding than hTRMT2A FL (Figure 2 B).This indicates that none of the individual domains --RBD, Central, and Fusion --is sufficient for full tRNA Gln binding and that they might have to interact cooperati v ely for physiological binding affinity.

tRNA binding r equir es inter actions with all hTRMT2A domains
To map the hTRMT2A-tRNA Gln interaction surface with near-amino acid resolution, we performed NM-and UV radiation-induced cross-linking of the hTRMT2A-tRNA Gln complex followed by MS.Experiments were conducted with the in vitro reconstituted protein-RNA complex (Supplementary Figure S3A).To confirm that hTRMT2A FL forms a stable complex with tRNA Gln , we also performed size exclusion chromato gra phy with tRNA Gln and hTRMT2A FL separately and with the hTRMT2A-tRNA Gln comple x (1:1 ratio).Stab le comple x formation was apparent as a shift of the elution peak to smaller elution volumes (from 17.41 ml to 15.46 ml), corresponding to a higher molecular weight (Figure 3 A).These co-purified complex es wer e also assessed by MS approaches (Figure 3 A; Supplementary Figure S3B).
The in vitro reconstituted and co-purified samples were cross-linked, trypsin and RNase digested, and analyzed by MS.Out of the four collected datasets (UV, NM, in vitro reconstituted and purified complex), we consistently found in at least three of them the following amino acids to be cross-linked to RNA (Supplementary Table S3): C270 to a C base, C463 to A or U, and H386 to U / C or G.Both C111 and H83 cross-linked to all four nucleotides (Figure 3 B).In some instances, these amino acids were also crosslinked to incompletely digested dinucleotides such as UU.Mapping of both cross-linked amino acids C111 and H83 onto the previously published X-ray structure of the RBD from hTRMT2A FL [( 34 ) PDB-ID: 7NTO] shows tRNA interaction with the RBD ␤-sheet surface (C111) and with one of the fle xib le loops (H83) (Figure 3 C).For RBDs, the ␤-sheet surface has been identified as the major RNA binding pla tform, although loop-media ted RBD-RNA interactions have also been reported ( 39 , 40 ).The electrostatic surface potential illustrates the position of cross-linked C111 and H83 in relation to the surface charge, where especially the aromatic ring of H83 is part of a positi v ely charged patch (Figure 3 D).
The cross-linked amino acids C270, H386 and C463 are distributed across the methyltr ansfer ase domain.A sequence alignment with E. coli TrmA (Supplementary Figure S4) re v ealed that C463 of hTRMT2A corresponds to A241 of E. coli TrmA.In the E. coli TrmA T-loop costructure [( 19 ) PDB ID: 3BT7], residue A241 is close to the catalytic center and in proximity ( < 10 Å ) to the stem of the T-loop.Proximal to the cross-linked residue C270 lies G263, which corresponds to R51 of E. coli TrmA, a residue shown to modestly contribute to tRNA binding ( 15 ).
In two of four datasets, 17 additional cross-linked r esidues wer e identified (Supplementary Figur e S3C).Most notably amongst them is the catalytic cysteine C538 crosslinked to uridine.Since C538 is known to make a covalent bond to the U54 base during catalysis ( 10 , 19 ), this highlights the specificity of our cross-linking MS data.Taken together, these results cross-validate our previous observation from EMSAs that multiple domains are involved in and r equir ed for RNA binding.

Catalytic death mutant of hTRMT2A FL
To better dissect the function of hTRMT2A, we decided to create a mutant version of this enzyme with impaired enzymatic activity.Based on the EMSAs with subdomains of hTRMT2A (Figure 2 ) and our cross-linking experiments ( A ) Comparative EMSAs with hTRMT2A WT, CDM, CBM and CNM display similar high affinity binding profiles to tRNA Gln in the nanomolar K D range.The hTRMT2A CDM shows supershifted binding e v ents.Binding assays were performed with 32 P-labeled tRNA Gln .For experiments, poly(U) competitor and labeled RNA were pre-incubated with increasing concentrations of hTRMT2A for 20 min.Reaction without protein served as the control.Free RNA and RNA-protein complex were separated by 4% nati v e PAGE.Each experiment was performed in triplicate.( B ) Single-cycle SPR results show a similar K D for hTRMT2A WT and hTRMT2A CDM.For all experiments, biotinylated tRNA Gln was coupled to an SA chip.A protein concentration series from 62.4 to 4000 nM (WT) and 15.6 to 1000 nM (CDM) was injected, resulting in the illustrated, doub le-referenced curv es.Triplicate measur ements ar e shown, as ar e r especti v e appar ent K D values.Mor e details ar e available in Supplementary Figure S5.( C ) K m curve results for hTRMT2A WT, CDM, CBM and CNM mutants from methylation assays confirm strongly reduced enzymatic activity of the hTRMT2A CDM mutant.For experiments, constant protein concentrations and increasing tRNA Gln concentrations were used.Data r epr esent the mean ± SD of thr ee r eplicates.Curves wer e fitted with the Michaelis-Menten equa tion illustra ted as lines.K m values are listed.
(Figure 3 ), we anticipated that an enzymatic activity mutant of hTRMT2A would r equir e surface mutations on the RBD to reduce interaction with target RNA and across the methyltr ansfer ase domain to abolish catalytic activity.Thus, these areas were chosen for mutation.For selecting the exact amino acids, sequence conservation (Supplementary Figure S4) and the reported impact of mutations on RNA binding in E. coli TrmA were considered ( 18 ).Finally, we selected three, conserved and solvent-exposed amino acids on the ␤-sheet surface of the RBD (E76A, K104E and F113A) and fiv e in the methyltr ansfer ase domain including the main catalytic base C538 for mutation (K242A, F409A, Q411A, D510A and C538A).This hTRMT2A mutant version was termed the c atalytic d eath m utant (CDM).The hTRMT2A CDM was used in comparati v e RNA binding and enzymatic acti vity e xperiments with hTRMT2A WT, the previously reported catalytic base mutant E581A (CBM) and the catalytic nucleophile mutant C538A (CNM) ( 10 ) (Supplementary Figure S1B).In EMSAs, CBM and CNM m utants showed slightl y less binding to tRNA Gln than WT protein when comparing their fractions of unbound RNA at higher protein concentrations (Figure 4 A).Surprisingl y, the CDM m utant displayed similar binding to the WT, but an increase in super-shifted binding e v ents.This highlights the difficulty in abolishing RNA binding of a protein with a large interaction surface such as hTRMT2A, reported through our cross-linking experiments.
Since EMSAs only provide an estimate of binding parameters, we quantified binding affinities of hTRMT2A WT and CDM by SPR experiments with surface-coupled tRNA Gln .The experiments indicated rather complex binding e v ents possib l y involving m ultiple interaction sites, which pre v ented precise determination of binding kinetics.Howe v er, in agreement with EMSAs, hTRMT2A WT exhibited high affinity binding with an apparent K D of 306 ± 60 nM, and hTRMT2A CDM showed a similar, albeit slightly lower affinity (apparent K D of 428 ± 71 nM) (Figure 4 B; Supplementary Figure S5).These results indica te tha t hTRMT2A CDM exhibits a similar binding affinity for tRNA Gln as compared with hTRMT2A WT.Howe v er, despite similar tRNA binding of hTRMT2A CDM, its enzymatic activity is almost abolished, as clearly visible from methylation assay results (Figure 4 C).
Further, methylation by hTRMT2A CBM saturated at lower RNA concentrations and a modestly lower K M of 0.64 ± 0.06 M when compared with hTRMT2A WT (Figure 4 C), which is in line with irre v ersib le trapping of RNA and blocking of the catalytic center.The hTRMT2A CNM showed a twice as high K m , which indicates less efficient methylation.Since hTRMT2A CDM showed the strongest reduction of tRNA Gln methylation, but similar RNA binding profiles, we selected this mutant for cellular studies.Moreover, this mutant was of particular interest as the E. coli homolog of hTRMT2A, TrmA, was shown to exhibit an enzymatic activity-independent secondary function as a tRN A cha perone ( 15 ).By including a mutant with impaired catalytic activity but similar tRNA binding affinity in cellular studies, we hoped to gain insight into this prospecti v e function for hTRMT2A.

Proximity ligation-based interactome of hTRMT2A
To better understand the cellular functions of hTRMT2A, we aimed at an unbiased assessment of its protein interactome and hence its cellular functions by employing proximity-dependent biotinylation (BioID) ( 28 ).In this assay, a protein of interest is fused to the promiscuous biotin ligase BirA* which biotinylates nearby proteins, allowing for the detection of proximal proteins with a distance of up to 10 nm (Supplementary Figure S6A).In our study, expression of BirA*-tagged hTRMT2A protein from an ov ere xpression plasmid (Supplementary Figure S7A) was induced in biotin-supplemented cells.Cells were lysed, and biotinylated proteins were captured using streptavidincoupled beads and analyzed by MS (Supplementary Figure S8A).To assess the correct subcellular localization of the BirA* fusion proteins to the nucleus and cytoplasm ( 23 , 41 ) (proteinatlas.org), we performed immunostainings against FLAG-tag from the ov ere xpression fusion protein.Immunostainings illustrated that FLAG-tagged hTRMT2A primarily localizes to the nucleus, but also in the cytoplasm.These stainings confirm indistinguishable subcellular localization of FLAG-BirA*-tagged, ov ere xpressed hTRMT2A variants (Supplementary Figure S9).
BirA* biotin ylated man y cellular proteins as detected by streptavidin western blotting (Supplementary Figure S6B).The fold change ratios of the MS results from WT / BirA* and CDM / BirA* were plotted against significance values (Figure 5 A).After a ppl ying cut-off thresholds (log 2 ratio > 2, -log 1 0 P -value > 3), 222 and 177 proteins were enriched in WT and CDM samples over BirA* background control, respecti v ely.Both datasets shared a subset of 115 proteins.Since hTRMT2A WT and CDM displayed similar RNA binding, a large overlap of 115 proteins was expected and constitute candidates for interaction partners.GO term analyses for biological processes and cellular compartments re v ealed enrichment of rRNA processing and ribosome biogenesis for protein function and suggested a primary localization of the identified proteins in the nucleolus (Figure 5 B).These GO terms are consistent with the im-munostainings of hTRMT2A (Supplementary Figure S9), further validating the BioID experiment.
Mor eover, we compar ed differ ences in the interactome of hTRMT2A WT and CDM by plotting the hTRMT2A CDM / WT ratios against significance le v els (Supplementary Figure S10A).Notably, proteins that are associated with RNA and nucleic acid metabolic processes were enriched in hTRMT2A WT over CDM.This demonstrates that hTRMT2A CDM interactors are depleted in nucleic acid-associated functions, which might be due to its abolished enzymatic inactivity and thus potentially points towards proteins associated with a non-enzymatic, secondary function of hTRMT2A.Howe v er, these differences were not very pronounced.Futures studies will be required to dissect if hTRMT2a indeed has such a secondary RNAregulatory function.

Cross-validation of protein hTRMT2A interactors by Co-IP
To cross-validate the interaction partners identified by BioID with a second pr oteome-wide appr oach, Co-IP experiments were performed.Co-IP detects protein interactions that are either direct or protein-/ RNA-mediated indir ect (Supplementary Figur e S11A).Expr ession of FLAGtagged hTRMT2A protein and an only FLAG-peptide was induced.Cells were lysed, and tagged proteins were directly captured using FLAG antibod y-coa ted beads and analyzed by MS (Supplementary Figure S8B).Successful elution of FLAG-hTRMT2A and its interacting pr oteins fr om beads was confirmed with western blotting (Supplementary Figure S11B).After analysis of the MS results, we plotted WT / FLA G and CDM / FLA G fold change ratios against significance values (log 2 ratio > 2, -log 10 P -value > 3).This re v ealed enrichment of 74 and 72 proteins in hTRMT2A WT and CDM samples over FLAG background control, respecti v ely (Figure 6 A).To identify potential hTRMT2A interaction candidates, the ratios of significantly enriched proteins from the Co-IP hTRMT2A WT and CDM datasets were plotted against each other.Both datasets shared a subset of 39 proteins (Supplementary Figure S10B).
The further overlap of hTRMT2A WT data from BioID and Co-IP experiments resulted in 13 common proteins, that constitute confident protein interaction partners (Figur e 6 B).Thr ee of these pr oteins, SSB (Lupus La pr otein), TARBP2 and SPINDOC, are also enriched in hTRMT2A CDM data from BioID and Co-IP experiments.These three common proteins constitute cross-validated, highestconfidence interaction candidates of hTRMT2A.

hTRMT2A does not methylate rRNA
The GO term annotations for biological processes (BioID) and the identification of ribosomal proteins such as RPL4 or NCL as possible hTRMT2A interactors motivated us to explore a potential functional link between hTRMT2Adependent methylation and rRNA processing / ribosome biogenesis.Previous cryoelectr on micr oscopy (cryoEM) studies of the human ribosomes had re v ealed all chemical modifications in rRNAs ( 42 ).Two reported sites of m 5 U modifications gained our particular interest: m 5 U4083 in the 60S subunit and m 5 U815 in the 40S subunit.To our knowledge, neither of these m 5 U modifications had been previously associated with a dedicated methyltr ansfer ase.Recently, FICC-CLIP experiments confirmed a cross-link of hTRMT2A with the small subunit rRNA ( 10 ).Moreover, the hTRMT2A homolog hTRMT2B has been shown to methylate tRNA and 12S rRNA in human mitochondria ( 12 ).Of note, another MS-based study did not identify m 5 U at respecti v e rRNA positions, as a result of which its presence is still discussed ( 43 ).Thus, we wanted to clarify whether m 5 U is found at respecti v e cytosolic rRNA positions and whether hTRMT2A is the dedicated m 5 U methyltr ansfer ase.For this purpose, two hTRMT2A KD HEK 293T cell lines (KD1 and KD2) were generated using stable transfection of shRNA.Knockdowns were validated with western blotting and show a clear reduction of TRMT2A protein le v els, with a higher reduction in KD1 ( 34 ) (Supplementary Figure S7B, C).For experimental as-sessment of rRNA methylation status, cytosolic ribosomes from hTRMT2A WT, KD1 and KD2 cells were isolated.After rRNA extraction from ribosomes, 18S and 28S rRNA were separated via gel electrophoresis.Tape station analysis showed high RNA purity and effecti v e separation of 18S and 28S rRNA as well as no contamination by tRNA (Supplementary Figure S12).Pure 18S and 28S rRNAs were subjected to MS analysis and the number of modifications per rRNA was calculated (for details see the Materials and Methods).The amount of Cm, Um and modifications per 18S and 28S rRNA corr esponds well with pr eviously reported numbers ( 4 ) (Supplementary Figure S13A).Substochiometric m 5 U of 18S and 28S rRNA from hTRMT2A WT was detected in line with ∼1-4% methylation of one uridine (Figure 7 B).This constitutes a much lower le v el than is usually found for modifications that occur on one site in rRNA (Figure 7 A; Supplementary Figure S13B).Further, the very low level of m 5 U modifications resembled that of modifications reported to be not present in rRNA (Supplementary Figure S13C).Finally, both knockdowns do not further reduce the already low le v els of m 5 U modification.In summary, these findings suggest that m 5 U cannot be found at physiolo gicall y rele vant le v els in either 18S or 28S rRNA of HEK 293T cells.

hTRMT2A is important for translation fidelity
As hTRMT2A is the dedicated methyltr ansfer ase for tRNA m 5 U formation, lack of this modification was hypothesized to influence tRNA folding and ther efor e translation ( 44 ).Howe v er, for the hTRMT2A paralog hTRMT2B, a lack of m 5 U in tRN A / rRN A did not affect translation efficiency or mito-ribosome stability ( 12 ).Indeed, we have recently shown that in a hTRMT2A-deficient situation, protein synthesis is not impaired ( 34 ), which suggests normal translation efficiency.Ther efor e, we asked whether another parameter of tr anslation, tr anslation fidelity, might be affected by hTRMT2A KD.Of note, hypomodification of RNA other than tRNA, such as tRNA fragments (tRFs) or mRNA, could also impact translation fidelity ( 20 , 24 , 45 ) To test this hypothesis, we used a gain-of-function reporter previously described to detect small increases in mistr anslation r ates ( 36 ).The reporter construct contains a fusion protein of Renilla and Firefly luciferases (deri v ed from Renilla reniformis and Photinus pyralis , respecti v ely).Single amino acid substitutions within the Renilla luciferase (E145Q or D120N) render the enzyme inacti v e.In a condition with increased translation error rate, the Renilla luciferase would stochastically be synthesized as a functional protein.The dual luciferase set up allowed us to control for protein expression and therefore measure mistranslation accurately.
The reporter constructs were transfected into hTRMT2A WT and KD cells, and luminescence was measured.Indeed, for the E145Q Renilla mutation, both KD1 and KD2 cell lines showed significantly increased normalized luminescence ratios, which are in line with an increased mistranslation in hTRMT2A KD over WT.Of note, a higher increase in luminescence ratios was seen for the KD1 cell line, which displayed a stronger reduction of hTRMT2A le v els than KD2 in western blot analysis (Supplementary Figure S7B, C).Howe v er, no significant increase in mistranslation was observed in the same experiment using the D120N mutation (Figure 7 C).These observations suggest that E145Q is more prone to mistranslation than D120N.

DISCUSSION
In this study, we characterized the properties of the human tRNA methyltr ansfer ase TRMT2A with respect to RNA binding and catalysis.We also identified its RNA-binding interface.Furthermore, we provide the first unbiased assessment of the protein interaction network of hTRMT2A.These insights suggest novel implications of hTRMT2A and m 5 U modification in translation.

hTRMT2A is a promiscuous enzyme that binds target RNAs with low specificity
Previously, it has been shown with FICC-CLIP experiments that hTRMT2A binds to the sequence motif GTTCG(A)A ( 10 ), which suggests a clear sequence preference of hTRMT2A in vivo .In the same study, no further dissection of structural r equir ements, orthogonal assays or mutational studies of the sequence motif were performed.Thus, we chose positi v e hits from these experiments to ( B ) Barplot of rRNA MS analysis of m 5 U in 28S and 18S rRNA shows very low modification le v els that do not change upon hTRMT2A KD. rRNA was obtained from ribosomes that were isolated from hTRMT2A WT and KD1 and KD2 cells.Data r epr esents the mean ± SD of thr ee biological r eplica tes.( C ) Transla tion fidelity assay with dual luciferase WT, E145Q and D120N mutants, and their respecti v e luminescence readout indicated ov er the gr aph.Lucifer ase plasmids were tr ansfected into hTRMT2A WT, KD1 and KD2.For the E145Q luciferase mutant, the in KD1 and KD2 was significantly increased over that in the WT, indicating a higher translation error rate.WT and the D120N luciferase mutant did not show significant changes.Data r epr esent the mean ± SD of six biological replicates in technical triplicates.Significance was tested with Wilco x on test for non-normally distributed values with P > 0.05 = ns, P < 0.05 = significant.complement available data with our orthogonal in vitro experiments.As an initial result, we observed that the binding affinity of hTRMT2A for tRNAs was in the expected range.The major parameter that increased methylation efficiency was the presence of a uridine positioned in the T-loop of the tRNA.Howe v er, in our e xperimental set-up, binding and methylation were not restricted to tRNA as exemplified by the unrelated ASH1 E3 stem-loop structured RNA containing a hTRMT2A-binding motif (Figure 1 ).In conclusion, our in vitr o da ta point towards an enzyme which binds and methylates not only tRNA, but also other structured RNAs, thus displaying no e xclusi v e pr efer ence for tRNA.
Previousl y, hTRMT2A was anal yzed regarding its binding (iCLIP) and methylation (FICC-CLIP) of RNA ( 10 ).Her e, it was appar ent that hTRMT2A did not methylate all bound RNA substrates, as exemplified in our data for tRN A Gln U54G and tRN A Ala binding and methylation (Figure 1 ).These observations could point in the direction of other, m 5 U formation-independent functions of hTRMT2A, for example as a tRNA chaperone ( 15 , 46 ).
Apart from the consistent r esults pr esented her e, we r eported unspecific binding and methylation of a KCND2 mRNA fragment ( 10 ).Hence, it might be that greater specificity is achie v ed in vivo by masking of unspecific methylation sites by other factors or an increase of specificity in RNA binding by cofactors of TRMT2A.Further experiments will be r equir ed to clarify whether KCND2 binding and methylation are indeed a physiolo gicall y rele vant e v ent in vivo .

In silico model of hTRMT2A-tRN A Gln comple x supported by experimental evidence
So far, there is no complete structural model of hTRMT2As available.Our data from cross-linking experiments followed by MS resolved the hTRMT2A-tRNA Gln interaction surface to a near-amino acid le v el.To better understand binding of hTRMT2A to tRNA, we used the AlphaFold homology model of hTRMT2A ( 47 ), removed protein areas with low confidence pr ediction scor es ( < 70 pLDDT, https: //alphafold.ebi.ac.uk/entry/Q8IZ69 ) and mapped the identified cross-links onto the model.Superposition of this homology model on the crystal structure of TrmA bound to a tRNA T-loop allowed us to confidently superpose a full tRNA Phe (PDB ID: 4TRA) into the homology model (Figure 8 ).The deri v ed hTRMT2A-tRNA Phe model is consistent with experimental observations presented.For example, all three hTRMT2A domains, the RBD, Central and methyltr ansfer ase domain, are in close contact with the modeled tRNA.Hence, this multidomain interaction of full tRNA is in full agreement with our EMSA experiments, w here tRN A binding by hTRMT2A FL was much stronger than binding by its individual subdomains (Figure 2 ).In addition, observed cross-links are localized around the catalytic center of the model (Figure 3 ; Supplementary Figure S3) and map to the region into which the tRNA T-loop is predicted to be inserted for methylation ( 19 ).Importantly, the cross-link of amino C463, found in all four datasets, resides near the modeled tRNA T-loop stem.In our homology model of hTRMT2A, the cross-linked C270 is placed between the T-loop and D-loop of the tRNA to potentially interrupt tertiary interactions and render U54 accessible for methylation, as reported for E. coli TrmA ( 15 ).Both amino acids are also found cross-linked to a dinucleotide UU, which suggests their proximity to the T-loop U54-U55 or the D-loop U18-U19 of tRNA.The cross-link of residue H386, found in three of four datasets, is positioned close to the catalytic center, but further away from the modeled tRNA (Figure 8 A).Se v eral mor e r esidues found in fewer than three datasets such as Y331, W17, and C486 positioned close to the catalytic center can be explained with our model.Other cross-links would r equir e structural r earrangements to be close enough to the RNA such as C191, C225, C329, K559, C551 and C225.As the RBD is connected via an unstructured linker (amino acids 148-178) to the methyltr ansfer ase domain, positioning the RBD with respect to the RNA bears considerable uncertainty (Supplementary Figure S14).Howe v er, our cross-linking data indica te tha t the RBD binds RNA with its ␤-sheet surface as well as with the fle xib le loop regions close to cross-linked H83.To account for these experimental data, the position of the RBD was adjusted manually with respect to tRNA (Supplementary Figure S14, zoom).The manually curated orientation of the RBD appears to be suitable to form the cross linking data-suggested interface and thus stabilize the hTRMT2A-tRNA complex.
Interestingly, our model is also consistent with hTRMT2A's limited substrate selectivity with respect to RNA binding and methylation, because large parts of the tRNA such as the anticodon stem and loop are not bound by hTRMT2A.Deviations in these structures might be tolerated, which would explain the binding of RNAs other than substrate tRNA such as stem-looped structured RNA or mRNA [Figures 1 and 8 , ( 20 )].
Despite the very consistent insights that arise from combining our experimental data with an advanced homology model, a full comprehension of the hTRMT2A-tRNA interface would r equir e an experimentally determined highr esolution co-structur e.

The hTRMT2A interactome and future implications
With BioID and Co-IP, we combined the strength of both methods to dissect the hTRMT2A interactome.BioID is superior in recapitulating the physiological nanoscale environment of hTRMT2A, as biotinylation of proteins occurs within intact cells.Moreover, transient interactions can be captured.Howe v er, BioID does not allow us to distinguish between direct interactors and proximal, nonbinding proteins, for instance in a very cro w ded environment such as the nucleolus.In contrast, Co-IP confidently identifies proteins that are part of the same cocomplex es including dir ect and indir ect interactions of the bait protein.Moreover, Co-IP carries the risk of detecting unphysiological interactions that occur only after cell lysis.For RBPs with lower specificity, in particular, the risk of false positi v es is high.Having these differences in mind, we propose that most confident protein interactors can be identified by merging Co-IP and BioID data.A total of 13 proteins were found both in BioID and Co-IP datasets from hTRMT2A WT cell lines: NCL, TRIM26, hTRMT2A, SSB, TARBP2, LARP7, PPM1G, UPF1, IGF2BP1, NPM1, SET, SPINDOC and RPL4.Some hits (SSB and TARBP2) wer e pr eviously identified in ortho gonal studies, w hich may serv e as positi v e controls and indicate that our combinatory interactome study is )]. 18S rRNA: #C = 493, #U = 356, #G = 540, #A = 403 (RNA Source ID = J01866) 28S rRNA: #C = 1646, #U = 687, #G = 1790, #A = 781 (RNA Source ID = K03432)In this way, the measured amount of injected rRNA subunit in pmol is calculated by each of the four canonical nucleosides separately.Those results were averaged and used as normalization for the modifications to determine the number of modifications per rRNA molecule in the respecti v e sample.n rRNA subunit = ) experiments were performed to determine hTRMT2A localization in HEK 293T cells.

Figure 1 .
Figure 1.hTRMT2A FL (amino acids 1-625) binds and methylates RNA with low specificity.( A ) EMSAs of tRNA with differing nucleotides at position 54 show for all tRNAs high affinity binding in the low nanomolar K D range.( B ) EMSA of truncated tRNA Gln T-loop displays unspecific binding with unbound RNA at high protein concentrations.EMSA with mRNA deri v ed from S. cerevisiae ASH1 E3 element (1771-1821; 51 nt) shows binding in the nanomolar K D range.EMSA with mRNA deri v ed from human KCND2 (chr7: 366813-366875; 62 nt) shows binding in the micromolar K D range, whereas a control experiment using poly(U) (21 nt) shows no binding.Binding assays were performed with 32 P-labeled RNA.For experiments, poly(U) competitor and labeled RNA were pre-incubated with increasing concentrations of hTRMT2A FL for 20 min.Reaction without protein served as control.Free RN A and RN A-protein complex were separated by 4% nati v e PAGE.Each e xperiment was performed in triplicate.( C ) Principle of the MTase Glo ™ methyltr ansfer ase assa y f or methylation experiments.( D ) Results of tRNA methylation experiments.( E ) Results of mRNA methylation experiments with tRNA Gln as control.Data r epr esent the mean ± SD for thr ee r eplicates.Curves wer e fitted with the Michaelis-Menten equa tion illustra ted as lines.( F ) Listed K m values from Michaelis-Menten fitting.For methylation experiments, hTRMT2A FL and methyl donor SAM wer e pr e-incubated with increasing concentrations of in vitro transcribed RNAs for 1 h.Reaction without RNA served as control.

Figure 2 .
Figure 2. hTRMT2A domains bind tRNA Gln cooperati v ely.( A ) Cartoon depicting protein fragments of hTRMT2A used for EMSAs.( B ) EMSA with hTRMT2A FL shows high affinity binding to tRNA Gln in the nanomolar K D range.( C ) EMSAs with the hTRMT2A RBD, Central and Fusion domain display only low affinity binding to tRNA Gln in the micromolar K D range.Because the SUMO-tagged Central domain was used, a SUMO-tag control was included.Binding assays were performed with 32 P-labeled tRN A Gln .For experiments, pol y(U) competitor and labeled RNA were pre-incubated with increasing concentrations of hTRMT2A protein for 20 min.Reaction without protein served as control.Free RNA and RNA-protein complex were separated by 4% nati v e PAGE.Each experiment was performed in triplicate.

Figur e 3 .
Figur e 3. tRN A Gln contacts various amino acids from all hTRMT2A domains.( A ) Size exclusion chromato gra phy pr ofile fr om a size exclusion chr omatography run with an S6 (15 / 150) column with hTRMT2A FL and tRNA Gln , as well as the stable protein-RNA 1:1 complex.SDS-PAGE and denaturing PAGE of protein, RNA and their complex are shown.Pooled fractions for cross-linking experiments are highlighted in red.( B ) Cross-linked amino acids are mapped on a linear r epr esentation of hTRMT2A FL and those consistently cross-linked in three or four of four datasets are shown.( C ) Cross-linked amino acids (yellow) are mapped on the hTRMT2A RBD X-ray structure (PDB ID: 7NTO), and the predicted RNA-binding surface is colored orange.( D ) Electrostatic surface potential of hTRMT2A RBD and the position of cross-linked amino acids H83 and C111 (dashed lines, stick r epr esentation).The v e (b lue) and negati v e (r ed) charge profile was pr epar ed with the PyMOL APBS plugin.Figur es wer e pr epar ed with PyMOL (version 2.0.4).

Figure 4 .
Figure 4.The hTRMT2A CDM mutant exhibits impaired enzymatic activity.( A ) Comparative EMSAs with hTRMT2A WT, CDM, CBM and CNM display similar high affinity binding profiles to tRNA Gln in the nanomolar K D range.The hTRMT2A CDM shows supershifted binding e v ents.Binding assays were performed with 32 P-labeled tRNA Gln .For experiments, poly(U) competitor and labeled RNA were pre-incubated with increasing concentrations of hTRMT2A for 20 min.Reaction without protein served as the control.Free RNA and RNA-protein complex were separated by 4% nati v e PAGE.Each experiment was performed in triplicate.( B ) Single-cycle SPR results show a similar K D for hTRMT2A WT and hTRMT2A CDM.For all experiments, biotinylated tRNA Gln was coupled to an SA chip.A protein concentration series from 62.4 to 4000 nM (WT) and 15.6 to 1000 nM (CDM) was injected, resulting in the illustrated, doub le-referenced curv es.Triplicate measur ements ar e shown, as ar e r especti v e appar ent K D values.Mor e details ar e available in Supplementary FigureS5.( C ) K m curve results for hTRMT2A WT, CDM, CBM and CNM mutants from methylation assays confirm strongly reduced enzymatic activity of the hTRMT2A CDM mutant.For experiments, constant protein concentrations and increasing tRNA Gln concentrations were used.Data r epr esent the mean ± SD of thr ee r eplicates.Curves wer e fitted with the Michaelis-Menten equa tion illustra ted as lines.K m values are listed.

Figure 5 .
Figure 5. hTRMT2A WT CDM interactome from BioID experiments.( A ) Volcano plots of hTRMT2A WT and hTRMT2A CDM BioID experiments show enriched proteins and was performed in quintuplicates.Log 2 fold change ratio of hTRMT2A WT (BirA*-tagged) / BirA* background and hTRMT2A CDM (BirA*-tagged) / BirA* background was plotted against -log 10 P -value.Ratio cut-offs were > 2 and < 0.5, and significance cut-off with a P -value < 0.05.Hits in agreement with these thr esholds ar e highlighted in yellow (WT) and b lue (CDM), respecti v ely.The top 10 hits are labeled with protein names.( B ) The top 10 most enriched GO terms for biological processes and cellular compartments of hTRMT2A WT (upper panel) and hTRMT2A CDM (lower panel) BioID experiments show enrichment of rRNA processing and ribosome biogenesis processes and the nucleolus as the most enriched cellular compartment.GO terms were ranked according to the FDR as computed with Fisher's exact test.Graphs illustrate GO terms plotted against log 2 fold enrichment.Node color r epr esents -log 10 FDR.Node size r epr esents counts per GO term.GO terms were visualized in R.

Figure 6 .
Figure 6.interaction partners of hTRMT2A.( A ) Volcano plots hTRMT2A WT and hTRMT2A CDM Co-IP experiment show enriched proteins and were performed in quintuplicates.Log 2 fold change ratio of hTRMT2A WT (FLA G-tagged) / FLA G background and hTRMT2A CDM (FLA G-tagged) / FLA G background was plotted against -log 10 P -value.Ratio cut-offs were > 2 and < 0.5, and significance cut-off of P -value < 0.05.Hits in agreement with these thresholds are highlighted in yellow (WT) and blue (CDM), respecti v ely.The top 10 hits are labeled with protein names.( B ) Overlap of significant hits from hTRMT2A WT BioID and Co-IP data shown as a plot of the respecti v e ratios (b lack).Ov erlap of significant hits from hTRMT2A WT and CDM, BioID and IP experiments (red) shown in the same plot.Venn diagram summarizes overlap of BioID and Co-IP datasets.

Figure 7 .
Figure 7. hTRMT2A has no impact on rRNA methylation, but affects translation fidelity.( A ) Barplot of rRNA MS analysis of m 5 C in 28S and 18S rRNA shows two modifications in 28S, as reported earlier.(B ) Barplot of rRNA MS analysis of m 5 U in 28S and 18S rRNA shows very low modification le v els that do not change upon hTRMT2A KD. rRNA was obtained from ribosomes that were isolated from hTRMT2A WT and KD1 and KD2 cells.Data r epr esents the mean ± SD of thr ee biological r eplica tes.( C ) Transla tion fidelity assay with dual luciferase WT, E145Q and D120N mutants, and their respecti v e luminescence readout indicated ov er the gr aph.Lucifer ase plasmids were tr ansfected into hTRMT2A WT, KD1 and KD2.For the E145Q luciferase mutant, the in KD1 and KD2 was significantly increased over that in the WT, indicating a higher translation error rate.WT and the D120N luciferase mutant did not show significant changes.Data r epr esent the mean ± SD of six biological replicates in technical triplicates.Significance was tested with Wilco x on test for non-normally distributed values with P > 0.05 = ns, P < 0.05 = significant.

Figure 8 .
Figure 8. Model of the hTRMT2A-tRNA complex.( A ) hTRMT2A AlphaFold homology model (gray) overlaid with the T-loop from the E. coli TrmA structure (PDB ID: 3BT7; orange).Consistently (blue spheres, three of four datasets) and less consistently (bright blue spheres, < 3 of four datasets), cross-linked amino acids were mapped onto the model.( B ) Hybrid model consisting of the hTRMT2A AlphaFold homology model (gray) and an overlaid experimental full tRNA structure (PDB ID: 4TRA) based on the superposition with the E. coli TrmA T-loop co-structure (PDB ID: 3BT7).Cross-linked amino acids from MS experiments are highlighted (coloring as in A) and mapped onto the hybrid model.Models were generated with PyMOL.