Different modification pathways for m1A58 incorporation in yeast elongator and initiator tRNAs

Abstract As essential components of the protein synthesis machinery, tRNAs undergo a tightly controlled biogenesis process, which include the incorporation of numerous posttranscriptional modifications. Defects in these tRNA maturation steps may lead to the degradation of hypomodified tRNAs by the rapid tRNA decay (RTD) and nuclear surveillance pathways. We previously identified m1A58 as a late modification introduced after modifications Ψ55 and T54 in yeast elongator tRNAPhe. However, previous reports suggested that m1A58 is introduced early during the tRNA modification process, in particular on primary transcripts of initiator tRNAiMet, which prevents its degradation by RNA decay pathways. Here, aiming to reconcile this apparent inconsistency on the temporality of m1A58 incorporation, we examined its introduction into yeast elongator and initiator tRNAs. We used specifically modified tRNAs to report on the molecular aspects controlling the Ψ55 → T54 → m1A58 modification circuit in elongator tRNAs. We also show that m1A58 is efficiently introduced on unmodified tRNAiMet, and does not depend on prior modifications. Finally, we show that m1A58 has major effects on the structural properties of initiator tRNAiMet, so that the tRNA elbow structure is only properly assembled when this modification is present. This observation provides a structural explanation for the degradation of hypomodified tRNAiMet lacking m1A58 by the nuclear surveillance and RTD pathways.


INTRODUCTION
Transfer RN As (tRN As) are essential components of the cellular protein synthesis machinery, but also serve additional functions outside translation (1)(2)(3)(4).To achie v e their wide range of functions within cells, tRNAs undergo a tightl y controlled bio genesis process leading to the formation of mature tRNAs (5)(6)(7)(8).The biogenesis of tRNAs typically includes the removal of the 5 -leader and 3 -trailer sequences from the precursor-tRNA transcripts, the addition of the 3 -CCA amino-acid accepting sequence, and the incorporation of a large number of posttranscriptional chemical modifica tions.These modifica tions occur a t specific sites in a tightly controlled manner, which ensures that the tRNA biogenesis process effecti v ely leads to the formation of functional tRNAs (9)(10)(11)(12)(13).All the cellular functions of tRNAs are, to various extents, affected by modifications.In particular, modifications in and around the anticodon are implicated in the decoding process ( 9 , 14-17 ), whereas modifications found in the tRNA cor e ar e collecti v ely implicated in the folding and stability of tRNAs (18)(19)(20)(21).Posttranscriptional modifications are thus central to tRNA biology.Ma tura tion defects, resulting in lack of modifications in the tRNA cor e, may r esult in alternati v e folding ( 22 , 23 ), and often reduce tRNA stability, leading to the degradation of hypomodified tRNAs by the ra pid tRN A decay (RTD) pathway (24)(25)(26) and the nuclear surveillance pathway (27)(28)(29).
Although modifications are typically introduced in tR-N As independentl y of each other, se v eral modification circuits have been identified in which one or more modifica tions stimula te or r epr ess the incorporation of another modification ( 11 , 30 , 31 ).This obviously dri v es a defined sequential order in the tRNA modification process.Most of the reported examples of an ordered modification process occur in the tRNA anticodon loop region (32)(33)(34)(35)(36), but modification circuits in the tRNA core have also been reported (37)(38)(39).
One such circuit in the tRNA core involves modifications in the T-loop of yeast tRNAs.Using NMR spectroscopy to monitor the ma tura tion of tRNAs in a time-resolved fashion in yeast extracts ( 40 ), we previously identified a sequential order in the introduction of T54, 55 and m 1 A58 in yeast tRNA Phe , with 55 being introduced first, then T54 and finally m 1 A58 ( 39 ).Using specific deletion strains, we uncovered a cross-talk between these three modifications, with the m 1 A58 modification strongly dependent on the two others.In a pus4 Δ strain, lacking 55, we indeed observed a se v ere slo w-do wn in the introduction of both T54 and m 1 A58.Similarly, in a trm2 Δ strain, lacking T54, we observed a slo w-do wn in the introduction of m 1 A58 ( 39 ).In addition, w e show ed, using liquid-chromato gra phy coupled with tandem mass spectrometry (LC-MS / MS), that le v els of m 1 A58 and T54 are affected in the pus4 Δ and trm2 Δ strains, in both yeast tRNA Phe and in total yeast tRNAs, in a manner compatible with the cross-talk observed with NMR spectroscopy in y east extr acts.This demonstr ated that these cross-talks in the T-loop are manifest not only in tRNA Phe but also in other yeast tRNAs.Overall, the slo w-do wn in the incorporation of modifications and the corr esponding decr ease in the modification le v els observ ed in the absence of a specific enzyme, namely in the pus4 Δ and trm2 Δ strains, was interpreted as a positi v e effect of the corresponding modification on the introduction of the other ones.We thus concluded that two modification circuits exist in the T-loop of yeast tRNAs, the long-branch 55 → T54 → m 1 A58 circuit and the direct-branch 55 → m 1 A58 circuit, without being able to conclude on the dir ect or indir ect natur e of the effect of 55 on m 1 A58 ( 39 ).Overall, this report on yeast tRNA Phe identified m 1 A58 as a late modification, introduced after earlier modifications such as 55, T54 and m 7 G46 ( 39 ).Howe v er, pre vious reports suggested that m 1 A58 is introduced early along the tRNA modification process in yeast, with m 1 A58 being introduced on initial pre-tRNA transcripts ( 5 ).Yeast initiator pre-tRNA i Met lacking m 1 A58, but containing the 5 -leader and part of the 3 -trailer sequences, is targeted by the nuclear surveillance and RTD pathways ( 27 , 28 , 41 , 42 ).In yeast tRNA i Met , the m 1 A58 modification is part of an unusual tRNA elbow structure involving non-canonical nucleotides A20, A54 and A60.This unusual substructure is assembled via an intricate network of interactions between the D-and T-loops and is likely conserved in eukaryotic initiator tRNAs ( 43 ).Altogether, these reports led to the model that m 1 A58 is introduced on pre-tRNA i Met transcripts, which stabilizes the tRNA i Met unique substructure, ther eby pr e v enting its degr adation.In addition, degr adation of tRNA i Met lacking m 1 A58 by the RTD pathway was recently shown to be conserved in the phylogenetically distant yeast species S. pombe and S. cerevisiae ( 42 ), suggesting that throughout eukaryotes the m 1 A58 modification is crucial to tRN A i Met biolo gy.Her e, aiming to r econcile the appar ent inconsistency regarding the incorporation of m 1 A58 in yeast tRNAs, namely as a la te modifica tion in elonga tor tRNA Phe and as an early modification in initiator tRNA i Met , we decided to examine the m 1 A58 modifica tion pa thways in yeast elonga tor and initia tor tRNAs (see Supplementary Figure S1 for the sequence and modifications of yeast tRNA Phe and tRNA i Met ).On the elongator tRNA Phe , we aimed at characterizing the molecular details related to the modification cir cuits pr esent in the T-loop and involving 55, T54 and m 1 A58, in order, in particular, to untangle direct from indirect effects.On the initiator tRNA i Met , we sought to investigate the introduction of m 1 A58 and its dependence on other modifications.In addition, we aimed at investigating the impact of the m 1 A58 modification on the structural properties of the tRNA i Met elbow region.Understanding the maturation process of initiator tRNA i Met , and in particular the m 1 A58 incorporation, which has consequences on its stability and quality control, is indeed crucial considering the central role of tRNA i Met in transla tion initia tion and hence gene expression.
For that, we first implemented a generic approach enabling the preparation of tRNAs containing specific modifications.We then used these specifically modified tRNAs to demonstra te tha t the incorpora tion of T54 in tRNA Phe is directly stimulated by 55, and that the incorporation of m 1 A58 in tRNA Phe is directly and individually stimulated by 55 and T54, with a remar kab le cumulati v e effect when they are present together, thereby reporting in detail the molecular mechanisms controlling the 55 → T54 → m 1 A58 modification circuit in yeast elongator tRNAs.We also show that m 1 A58 is efficiently introduced on unmodified tRNA i Met , and does not strictly need any prior modification, although m 5 C48,49 have a slight stimulatory effect on m 1 A58 incorporation.Finally, we show that the m 1 A58 single modification has major effects on the structural properties of yeast tRNA i Met , with the tRNA elbow structure being properly assembled only when this modification is present.This provides a structural basis to the degradation of hypomodified tRNA i Met lacking m 1 A58 by the nuclear surveillance and RTD pathways.

Yeast strains
Yeast strains used in this study are listed in Supplementary Table S1.The wild-type S. cerevisiae BY4741 strain and the YKO collection kanMX strains carrying deletions of the genes for modification enzymes T rm1, T rm2, T rm4, T rm8, T rm10, T rm11, Pus4, Dus1, Dus3 and Rit1, were obtained fr om Eur oscarf and used f or tRNA preparations f or MS analysis.The proteinase-deficient S. cerevisiae strain c13-ABYS-86 and the deri v ed strain c13-ABYS-86-trm4 Δ were used for the preparation of yeast extracts for NMR experiments.All strain constructions were verified by PCR using appropriate oligonucleotides (listed in Supplementary Table S2).

E . coli strains
E. coli strains used in this study are listed in Supplementary Table S1.The E. coli BL21(DE3) CodonPlus-RIL yggH::kan (trmB) strain was constructed by transferring the yggH::kan cassette from the appropriate K-12 strain of the Keio collection ( 44 ) to a BL21(DE3) CodonPlus-RIL strain (Agilent) by phage P1 vir -mediated transduction ( 45 ) (Supplementary Table S1).Deletion of the yggH gene and its replacement by the kanamycin resistance cassette in the BL21(DE3) CodonPlus-RIL strain was checked with PCR using appropriate sets of primers (Supplementary Table S2).

Modification enzymes cloning
The gene encoding the full-length yeast Pus4 (M1 to V403 -Uniprot entry P48567) was cloned from BY4741 genomic DNA between the EcoRI and NotI sites of a modified pET28a vector (Novagen) encoding an N-terminal His 6 -tag cleavable with TEV protease (pET28-Pus4).The gene encoding the full-length yeast Trm2 (M1 to I639 -Uniprot entry P33753) was initially cloned from BY4741 genomic DNA between the EcoRI and NotI sites of a pGEX-6p-1 v ector (pGEX-Trm2).Howe v er, this construct was insoluble and poorly expressed in E. coli BL21(DE3) CodonPlus-RIL cells.Since the N-terminal part of Trm2 contains highly hydrophobic stretches of amino-acids, and does not correspond to the catalytic domain of the protein, a second construct corresponding to V116 to I639 was cloned between the BamHI and XhoI sites of a pRSFDuet-Smt3 vector leading to an N-terminal His 6 -SUMO-fusion of Trm2 (pSUMO-Trm2).The naturally present BamHI site in the yeast trm2 gene was first removed by a silent mutation of the codon encoding for D564 from GAT to GAC with site directed mutagenesis.The genes encoding yeast T rm6 / T rm61 heterodimer (Trm6: M1 to I478 -Uniprot entry P41814; Trm61: M1 to K383 -Uniprot entry P46959) were cloned from BY4741 genomic DNA between the BamHI and NotI sites for Trm6 and NdeI and XhoI sites for Trm61 of a pETDuet-1 vector (Nov agen) thereb y encoding an Nterminal His 6 -tag on Trm6 (pETDuet-T rm6 / T rm61).

Modification enzymes purification
Pus4, T rm2 and T rm4 wer e over expr essed in E. coli BL21(DE3) CodonPlus-RIL cells (Agilent) in LB media.T rm6 / T rm61 heterodimer was ov ere xpressed in E. coli BL21(DE3) CodonPlus-RIL yggh::kan cells lacking the E. coli enzyme catalyzing m 7 G46 modifications in tR-N As, namel y TrmB since initial expression and purification in E. coli BL21(DE3) CodonPlus-RIL cells lead to a T rm6 / T rm61 heterodimer contaminated with an m 7 G46 modification activity (see Supplementary Figure S2).The cells were grown at 37 • C to OD 600 ∼0.4,cooled down to 18-30 • C and induced at OD 600 ∼0.6 by adding (IPTG) to a final concentration of 0.4-0.5 mM.Cells were harvested 6-22 h after induction by centrifugation.Cell pellets were resuspended in the corresponding lysis buffer supplemented with an EDTA-free antiprotease tablet (Roche) and lysed by sonication.Cell lysates were centrifuged for 30 min at 35 000 g.All column chromato gra phy purifications were performed on a ÄKTA Pure purification system (Cytiva) at 4 • C. The cell lysate supernatant was loaded on a Ni-NTA column and the protein of interest was eluted with an imidazole gr adient.Fr actions containing the protein were pooled, concentrated with an Amicon 50 000 MWCO (Millipore) and further purified with a combination of hydrophobic and size exclusion chromato gra phy depending on the protein.Purified protein samples loaded on size exclusion chromato gra phy were eluted in the corresponding protein stora ge b uffer, confirmed for purity using SDS-PAGE (Supplementary Figure S3), concentrated with an Amicon (Millipore) to ∼5-10 mg / ml and stored at -20 • C. The protein concentrations were determined by absorbance at 280 nm using the corresponding mM extinction coefficient (See Supplementary Table S3 for specific details on the purifications).

RNA sample pr epar ation f or NMR and enzyme activity assays
Unmodified yeast tRNA Phe -WT, tRNA i Met -WT, tRNA Phe -U17, tRNA Phe -A20A60, tRNA Phe -A54, tRNA Phe -UAAA, tRN A i Met -U17, tRN A i Met -G20C60, tRN A i Met -U54 and tRNA i Met -UGCU were prepared by standard in vitro transcription following pre viously pub lished procedures, either with unlabelled NTPs or 15 N-labelled Us and Gs ( 39 , 46 ).We replaced the first Watson Crick base pair A1-U72 of tRNA i Met with a G1-C72 base pair in order to improve in vitro transcription efficiency.To pr epar e the single modified 55-tRNA Phe , 112 M of refolded tRNA Phe was incubated with 3.3 M of purified Pus4 for 40 min at 30 • C in an 800 l reaction mix.To prepare T54-tRNA Phe , 80 M of refolded tRNA Phe was incubated with 12 M of purified Trm2 and a ∼6-8-times excess of S-adenosyl-Lmethionine (SAM) in an 800 l reaction mix for 14 h at 30 • C. To pr epar e the double modified 55-T54-tRNA Phe , 80 M of 55-tRNA Phe was incubated with 8 M Trm2 and a ∼6-8-times excess of SAM in an 800 l reaction mix for 4 h at 30 • C. To pr epar e m 5 C48,49-tRNA i Met , 146 M of refolded unmodified-tRNA i Met was incubated with 23 M of purified Trm4 for 17 h at 30 • C in a 500 l reaction mix.All reactions were performed in the following ma tura tion buf fer (MB): 100 mM NaH 2 PO 4 / K 2 HPO 4 pH 7.0, 5 mM NH 4 Cl, 2 mM DTT and 0.1 mM EDTA.The tRNA reaction products were then purified by ion exchange chromato gra phy (MonoQ, Cytiva), dial yzed e xtensi v ely against 1 mM Na-phosphate pH 6.5, and refolded by heating at 95 • C for 5 min and cooling down slowly at room temperature.Buffer was added to place the tRNAs in the NMR buffer (10 mM Na-phosphate pH 6.5, 10 mM MgCl 2 ), and the samples were concentrated using Amicon 10000 MWCO (Millipore) to ∼80 M for further use in kinetic assays, or ∼1.4-1.5 mM for the NMR study of tRNA i Met ma tura tion in y east extr acts.

Trm2 and Tr m6 / Tr m61 kinetic assays on different substrates
To measure initial velocities of m 1 A58 and T54 formation, 10 M of unmodified tRNA Phe , 55-tRNA Phe , T54-tRN A Phe , 55-T54-tRN A Phe , unmodified tRN A i Met and m 5 C48,49-tRNA i Met were incubated each in a 300 l reaction with enzyme concentrations varying from 50 to 300 nM depending on enzyme and substrate type, 18 M nonradioacti v e SAM and 50 nM of radioacti v e [ 3 H]-SAM (see Supplementary Table S4 for details on the r eaction mix es).Reactions were performed in the MB buffer and were incuba ted a t 30 • C f or 30 min except f or the reaction with T rm6 / T rm61 and the unmodified tRNA Phe that was incubated for 96 min.Aliquots of 50 l were taken of each reaction at 6, 12, 18, 24 and 30 min (for the 30 min reactions) and at 24, 48, 72 and 96 min (for the 96 min reaction) and the samples were quenched by adding 5% (v / v) cold trichloracetic acid (TCA).Quenched samples were filtered through W ha tman glass microfibers disks pre-soaked with 5% (v / v) TCA, washed four times with 5% (v / v) TCA and one final time with ethanol.The filter disks were dried, then 5 ml Optiphase 'HISAFE' 2 scintillation cocktail (PerkinElmer) were added, and the counts per minute (CPM) equivalent to the incorporated [ 3 H]-methyl were determined by scintillation counting.Then CPM values were converted to concentrations of modified tRNAs using [ 3 H]-SAM / CPM calibra tion standards.Enzyma tic r eactions wer e performed in triplicates or quadruplicates.Since T rm6 / T rm61 and Trm2 activity turned out to vary greatly between different substra tes, dif ferent enzyme concentrations were used to perform the kinetic assays.Ther efor e, we normalized the quantities of modified tRNAs to an equivalent of 50 nM of enzyme.Initial velocities (V i ) were determined by linear regression using Prism7 (GraphPad), i.e. data were fitted to a single linear function: y = V i .x while forcing the curve to pass through the origin, and standard errors (SE) on the V i were determined by taking into account the data spread.

T rm6 / T rm61 activity assays on yeast tRNA Phe and tRNA i Met variants
To measure m 1 A58 formation, 10 M of unmodified yeast tRNA Phe -WT, tRNA Phe -U17, tRNA Phe -A20A60, tRNA Phe -A54, tRNA Phe -UAAA, tRNA i Met -WT, tRN A i Met -U17, tRN A i Met -G20C60, tRN A i Met -U54 and tRNA i Met -UGCU were incubated each in a 100 l reaction with 600 nM of purified T rm6 / T rm61, 18 M non-radioacti v e SAM and 100 nM of radioacti v e [ 3 H]-SAM.Reactions were performed in the MB buffer at 30 • C and concentrations of modified tRNAs were measured at t = 60 min.Samples were then treated as described above.Enzymatic reactions were performed in six replicates ( N = 6).Standar d de viations were relati v ely uniform across the different tRNA substrates and corresponded to 20-33% of the average value for tRNA Phe variants and to 17-20% for tRNA i Met v ariants.CPM v alues were converted to concentrations of modified tRNAs using [ 3 H]-SAM / CPM calibration standards.

NMR spectroscopy
All NMR spectra of yeast tRNA Phe and tRNA i Met were measured at 38 • C on a Bruker AVIII-HD 700 MHz spectrometer equipped with TCI 5-mm cryoprobe with 5-mm Shigemi tubes in the NMR buffer (10 mM Na-phosphate pH 6.5, 10 mM MgCl 2 ) supplemented with 5% (v / v) D 2 O.To verify that the desired modifications were incorpora ted quantita ti v el y in yeast tRN A Phe , 1D jump-andreturn-echo NMR spectra ( 47 , 48 ) of the different tRNAs wer e measur ed and compar ed to pr eviously characterized samples ( 39 , 49 ).To analyse the effect of nucleotide swapping on the structural properties of yeast tRNA Phe and y east tRNA i Met , 2D ( 1 H, 15 N)-BEST-TROSY spectr a of unmodified yeast tRNA Phe -WT, tRNA Phe -U17, tRNA Phe -A20A60, tRNA Phe -A54, tRNA Phe -UAAA, tRNA i Met -WT, tRN A i Met -U17, tRN A i Met -G20C60, tRN A i Met -U54 and tRNA i Met -UGCU wer e measur ed at 38 • C in the NMR buffer.In addition, to evaluate the effect of specific modifications on the structural properties of yeast tRNA i Met , 2D ( 1 H, 15 N)-BEST-TROSY spectra of unmodified tRNA i Met , m 5 C48,49-tRNA i Met and m 1 A58-tRNA i Met were measured at 38 • C in the NMR buffer.Imino resonances of the m 1 A58-tRNA i Met were assigned using 2D jump-andreturn-echo ( 1 H, 1 H)-NOESY ( 47 , 48 ) and 2D ( 1 H, 15 N)-BEST-TROSY ( 50) experiments.For monitoring the maturation of tRNA i Met in yeast extract, wild-type and trm4 Δ y east extr acts wer e pr epar ed in the c13-ABYS-86 background, as previously described ( 40 ).NMR spectra were measured at 30 • C with unmodified 15 Met at 40 M in yeast extracts supplemented with NaH 2 PO 4 / K 2 HPO 4 pH 6.5 150 mM, NH 4 Cl 5 mM, MgCl 2 5 mM, DTT 2 mM, EDTA 0.1 mM, SAM 4 mM, ATP 4 mM, NADPH 4 mM and D 2 O 5% (v / v) ( 51 ).Each 2D ( 1 H, 15 N)-BEST-TROSY experiment of the series was measured with a recycling delay of 200 ms, a SW( 15 N) of 26 ppm and 96 increments for a total experimental time of 120 min.The data were processed using TOPSPIN 3.6 (Bruker) and analysed with Sparky ( http://www.cgl.ucsf.edu/home/sparky/ ).

Total tRNA samples from yeast for mass spectrometry
Total tRNA from S. cerevisiae BY4741 wild-type or mutant strains used for mass spectrometry analysis were prepared as described previously ( 39 ).For each strain, all cultures and tRNA preparations were performed in triplicate for statistical anal ysis.Yeast tRN A i Met was isolated from ∼1 g total tRNA samples with a first step of SEC and a subsequent purification using T1 Dynabeads (Thermo Fisher Scientific, Product no.65801D) and a DNA probe specific to tRNA i Met ([Btn]-AAA-T CG-GTT-T CG-AT C-CGA-GGA-CAT -CAG-GGT -TAT -GA, Sigma-Aldrich, Munich, Germany) as previously reported ( 39 , 52 , 53 ).

Purified tRNA i
Met samples were digested to single nucleosides following previously published procedures ( 39 ) and sta ble isotope-la belled internal standard (SILIS , 0.1 volume of 10X solution) from yeast was added for absolute quantifica tion ( 54 ).Quantifica tion of the m 1 A modifica tion in tRNA i Met was performed with an Agilent 1290 Infinity II equipped with a DAD combined with an Agilent Technologies G6470A Triple Quad system and electro-spray ionization (ESI-MS, Agilent Jetstream) following previously published procedures ( 39 , 54 ).Absolute abundance of m 1 A from wild-type yeast corresponded to 0.69 ± 0.04 m 1 A per tRNA i Met .The absolute quantities of m 1 A in the deleted strains were normalized to that of the wild-type strain to determine abundance relati v e to wild-type.Analyses of the variations compared to the wild-type strain were conducted from the determination of the confidence intervals at 95% (CI 95%) using Prism7 (GraphPad).

A generic approach to pr epar e tRNAs with specific modifications
In order to evaluate the effect of pre-existing modifications on the introduction of further ones, we have implemented a generic method for preparing tRNA samples with a single or a specific set of modifications (Figure 1 ).Our approach is divided into four successive steps, (1) tRNA in vitro transcription and purifica tion, (2) modifica tion enzyme expression and purification, (3) in vitro tRNA modification reaction and modified tRNA purification and ( 4) tRNA sample quality control by NMR spectroscop y (Figur e 1 ).To introduce se v eral modifications on a tRNA, steps 3 and 4 can be reiterated on a tRNA sample already carrying modification(s).
For the present study on yeast tRNA Phe and tRNA i Met , in addition to the unmodified tRNA Phe and tRNA i Met , we applied our methodology to produce: tRNA Phe samples carrying single modifications ( 55-tRNA Phe and T54-tRNA Phe ), or double modifications ( 55-T54-tRNA Phe ), and tRNA i Met samples carrying m 5 C48,49 or m 1 A58 modifications (m 5 C48,49-tRNA i Met and m 1 A58-tRNA i Met ).For this purpose, we first transcribed and purified, using anion exchange chromato gra phy, the yeast unmodified tRN A Phe and tRNA i Met (Figure 1 , step 1).We then ov ere xpressed and purified the yeast enzymes Pus4 that introduces 55, Trm2 that adds T54 (or m 5 U54), Tr m6 / Tr m61 that adds m 1 A58 and Trm4 that introduces m 5 C48 and m 5 C49 (see Materials and Methods and Supplementary Figure S1; Figure 1 , step 2).Next, preliminary activity tests with these different enzymes allowed us to estimate the enzyme to tRNA ratios and the incubation times needed to introduce the desired modifica tions quantita ti v ely.We thus incubated the unmodified tRNAs with the appropriate enzymes and cofactors for the r equir ed duration, and then purified the in vitro modified tRNAs using anion exchange chromatography (Figure 1 , step 3).Finally, we verified that the desired modifications were introduced quantitati v ely by performing a quality control of our samples with NMR spectroscopy (Figure 1 , step 4).

The introduction of T54 by Trm2 to the yeast tRNA Phe is stimulated by 55
The fact that we observed a slower incorporation of T54 in tRNA Phe in the pus4 Δ yeast extract, and that the amount of T54 in tRNA Phe as well as in the total tRNA population is drastically reduced in the pus4 Δ strain, suggested that the 55 modification had a positi v e effect on the introduction of T54 by Trm2 ( 39 ).Howe v er, we could not exclude that the defect in T54 incorporation was due to a negati v e effect of other modifica tion(s) tha t onl y become a pparent in the absence of 55 or that the genetic expression of Trm2 was affected in the pus4 Δ strain.Here, in order to unambiguously determine whether the introduction of T54 on the yeast tRNA Phe by Trm2 is directly dependent on the presence of 55, we conducted activity assays with Trm2 on unmodified tRN A Phe and 55-tRN A Phe (produced as described above).Trm2 was incubated with each of the tRNAs in the presence of the methyl-donor cofactor SAM carrying a radioacti v e methyl group ( S -adenosyl-L -methionine [methyl-3 H]), and aliquots were taken at different time points to determine the initial velocities ( V i ) of the methylation reactions (Figure 2 A, Table 1 , and Supplementary Figure S4).These activity assays clear ly demonstr a ted tha t the methylation reaction catalysed by Trm2 is about 6 times faster on the 55-tRN A Phe w hen compared to the unmodified tRNA Phe (Table 2 ).This shows that the catalytic efficiency of Trm2 introducing T54 to tRN A Phe directl y depends on the prior presence of 55 and establishes the direct positi v e link between 55 and the introduction of T54 by Trm2.

The introduction of m 1 A58 by T rm6 / T rm61 to the yeast tRNA Phe is stimulated by 55 and T54
Like wise, our pre vious wor k suggested a positi v e effect of the 55 and T54 modifications on the introduction of m 1 A58 by the T rm6 / T rm61 comple x ( 39 ).Howe v er, as e xplained above for Trm2, we could not exclude that the observed behaviours were due to alternati v e effects.In addition, considering the above-mentioned effect of 55 on T54, it was not possible to distinguish a direct effect of 55 on m 1 A58 from an indirect effect via T54.To definitely establish whether the introduction of m 1 A58 on the yeast tRN A Phe is directl y dependent on the presence of 55 and T54, we conducted activity assays with the T rm6 / T rm61 complex on unmodified tRN A Phe , 55-tRN A Phe , T54-tRN A Phe and 55-T54-tRN A Phe .The T rm6 / T rm61 complex was incubated with each of the tRNAs in the presence of a radioacti v e [methyl-3 H]-SAM cofactor, and aliquots were taken at different time points to deri v e the initial velocities (Figure 2 B, Table 1 , and Supplementary Figure S5a-d).With these activity assays, we observed that the introduction of m 1 A58 by T rm6 / T rm61 is 3.3 times more efficient when the T54 modification is present compared to the unmodified tRNA Phe , 7.1 times more efficient in the presence of 55 and 15 times more efficient if both T54 and 55 are present in the yeast tRNA Phe (Table 2 ).This demonstrates that T54 and 55 have individually a positive effect on the introduction of m 1 A58, as well as a cumulati v e positi v e effect if they are both sim ultaneousl y present.Therefore, the catalytic activity of T rm6 / T rm61 directly depends on the presence of both the T54 and 55 modifications.Additionally, our measurements indica te tha t 55 stimulates the introduction of m 1 A58 about two times more efficiently than T54.Step 3) The unmodified tRNA is modified in vitro with the purified modification enzyme in presence of cofactors and subsequently purified by anion exchange chromato gra phy.(Step 4) A quality control step is performed by 1D 1 H NMR in order to establish that the desired modifications were fully incorporated in the tRNA population.   1 and 2).The m 1 A58 modification is efficiently introduced on an unmodified tRNA i

Met
The results presented above showed that efficient introduction of m 1 A58 in tRN A Phe , strongl y depends on the prior presence of 55 and T54 in the T-loop.In addition, since the le v els of modifications observ ed for total yeast tRNAs and tRNA Phe , are similarly affected in the pus4 Δ and trm2 Δ strains, the stimulation effect of 55 and T54 on the introduction of m 1 A58 is certainly a common feature of sever al y east tRNAs ( 39 ).At first sight, it might seem paradoxical that the efficiency of an enzyme encoded by two essential genes, i.e. trm6 / trm61 ( 5 , 55 , 56 ), is highly dependent on the prior presence of modifications encoded by non-essential genes, i.e. pus4 and trm2 .Howe v er, the origin of the essentiality of the m 1 A58 modification has been studied in detail in yeast and has been shown to be related to its importance for the ma tura tion of initia tor tRNA i Met ( 55 ).Hypomodified initiator tRNA i Met lacking m 1 A58 are indeed targeted to degradation by RNA decay pathways ( 27 , 28 , 42 ).As a possible explanation to this paradox, we noted that yeast tRNA i Met does not carry T54 and 55 in the T-loop, but contains unmodified A54 and U55 (Supplementary Figure S1).Altogether, we anticipa ted tha t the initiator tRNA i Met would have its own pathway of modification in the T-arm, in which the m 1 A58 modification did not depend on pre-existing modifications.More generally, since tRNA i Met transcripts lacking m 1 A58 are degraded by RNA deca y pathwa ys, it seems reasonab le that le v els of m 1 A58 in tRNA i Met should not be altered in different strains or growth conditions.Modification le v els should indeed reflect the r equir ement of m 1 A58 for tRNA i Met stability.To examine these points, we measured, using LC-MS / MS, the le v els of m 1 A in tRNA i Met from pus4 Δ and trm2 Δ strains, and from dus1 Δ, dus3 Δ, rit1 Δ, trm1 Δ, trm4 Δ, trm8 Δ, trm10 Δ and trm11 Δ strains, involved in the introduction of modifications D16, D47, Ar(p)64, m 2  2 G26, m 5 C48,49, m 7 G46, m 1 G9 and m 2 G10, respecti v ely.These le v els were compared with the le v els of m 1 A in tRNA i Met from wild-type yeast cultured under the same experimental conditions.As expected, we observed no substantial changes in the amount of m 1 A in any of these deleted strains compared with the wild-type le v el (Figure 3 A).The slight variations observed between some deleted strains and the wild-type could reflect a certain stability of m 1 A58-depleted tRNA i Met and / or small variations in the degree of purity of tRNA i Met r ecover ed from the total tRNA population in the purification procedure.In any case, our data do not allow to conclude that these slight variations are significant.Overall, this shows that the lack of any other single modification does not pre v ent the formation of mature tRNA i Met carrying m 1 A58, and suggests that m 1 A58 can be correctly introduced on unmodified tRNA i Met .To evaluate the efficiency of m 1 A58 modification on an unmodified tRNA i Met , we conducted activity assays with pr epar ed from modification-enzyme-deleted strains using the wild-type le v els as r efer ence.Black dots r epr esent individual measur ements, data heights r epr esent the mean of the biological replicates.Error bars correspond to the confidence interval at 95% (CI 95%).Modifications were quantified in three independent biological replicates ( N = 3).( B ) Time course of the introduction of m 1 A58 in tRNA i Met depending on the prior presence (blue) or absence (red) of the m 5 C48,49 modifications.Modified tRNA quantities were measured for 5 time points in three independent experiments ( N = 3), and initial velocities ( V i ) were determined by linear regression (see Tables 1 and 2).
T rm6 / T rm61 on unmodified tRNA i Met produced by in vitro transcription as described for tRNA Phe (Figure 3 B, Table 1 , and Supplementary Figure S5e).We observed that the introduction of m 1 A58 by T rm6 / T rm61 is 11.5 times more efficient on the unmodified tRNA i Met than on the unmodified tRNA Phe (Table 2 ).This rate corresponds to an efficiency of about 0.8 times that measured on the doubly-modified 55-T54-tRNA Phe (Table 2 ).Our data ther efor e establish that, on the contrary to its introduction on unmodified tRNA Phe , m 1 A58 is efficiently introduced on unmodified tRNA i Met , with an efficiency that is comparable to that observed for an optimally modified tRNA Phe bearing both 55 and T54.

Met
Aiming to identify the sequence elements and associated structural properties implicated in the differences observed for m 1 A58 incorporation in tRNA Phe and tRNA i Met , we designed a set of tRNA variants with the objecti v e of transfering elongator sequence elements and associated structural properties to initiator tRNA i Met , and vice versa.Since m 1 A58 is part of the specific initiator elbow structure ( 43 ), residues involved in this unique substructure, namely A20, A54 and A60, were primarily targeted for mutations.In addition, since the absence of nucleotide U17 is also a characteristic of initiator tRNA i Met , we chose to add it in a tRN A i Met variant.All tRN A variants, with their specific mutation or set of mutations are schematically summarized on Supplementary Figure S6.
First, in order to evaluate the effect of the nucleotide swapping between tRNA Phe and tRNA i Met from a structural point of view, we conducted NMR analysis on each tRNA variant.The comparison of the NMR fingerprint of unmodified tRNA Phe and tRNA i Met re v ealed clear differ ences (Supplementary Figur e S6).The NMR spectrum of tRNA Phe -WT displays sharp and uniform NMR signals, characteristic of a stable, homo geneousl y folded tRN A. On the contrary, the NMR spectrum of tRNA i Met -WT exhibits a heterogeneous NMR signal profile with both weak and strong signals, as well as signals with atypic line shapes.These classic exchange-broaden signals reflect a less homogeneous f olding f or unmodified tRNA i Met that probab ly e xchanges between se v eral folding states, an exchange occurring in the intermediate regime relati v e to the NMR chemical shift time scale.The NMR fingerprints of the different variants re v ealed that for tRNA Phe , important structural changes are taking place in the tRNA Phe -A54 and tRN A Phe -UAAA variants, w hich tend to acquire a heterogeneous NMR spectrum profile (Supplementary Figure S6a).Con versely, f or tRNA i Met , structural changes are a pparent mostl y for the tRN A i Met -U54 and tRN A i Met -UGCU variants, which exhibit slightly less heterogeneous or atypic NMR line shapes (Supplementary Figure S6b).
Ne xt, we conducted acti vity assays with T rm6 / T rm61 on unmodified tRNA Phe and tRNA i Met variants.On one hand, we observed that T rm6 / T rm61 is ∼5.5 times less efficient on tRN A i Met -U54, w here A54 is replaced by U54, compared to tRNA i Met -WT (Figure 4 ).This shows that A54 is r equir ed for an efficient incorporation of m 1 A58 by T rm6 / T rm61 on unmodified tRNA i Met .We do not observe any other significant changes in the efficiency of m 1 A58 introduction on other tRNA i Met variants and in particular on tRNA i Met -UGCU (Figure 4 ).This is quite puzzling since tRNA i Met -UGCU also lacks the A54 residue.The additional mutations U17, G20 and C60 seem to neutralize the negati v e effect of the lack of A54.This demonstrates the inherent complexity and challenge associated with comprehending how nucleotides collaborate to establish intricate networks of interactions that shape the structure of tRNAs.On the other hand, the tRNA Phe -A54 variant, in which U54 is replaced by A54, does not show an increased efficiency of m 1 A58 incorporation (Figure 4 ).This is a good illustration that converting a good substrate into a poor substrate through the removal of a single key element is considerably simpler compared to the transformation of a poor substrate into a good one by introducing the same key element.In addition, we do not observe any changes in the efficiency of m 1 A58 introduction on tRNA Phe -U17 and tRNA Phe -A20A60 variants compared to tRNA Phe -WT.This shows that neither Met variants (in nM).Names of the tRNAs are indicated below the graph and correspond to a specific nomenclature (see Supplementary Figure S6 for correspondence and details).Black dots represent individual measurements.Modified tRNA quantities wer e measur ed for 1 time point at t = 1 h in six independent experiments ( N = 6).Data heights r epr esent the mean of the replicates.Error bars correspond to the confidence interval at 95% (CI 95%).
adding nor removing U17, A20 and A60 residues to either tRN A i Met or tRN A Phe has an y effect on m 1 A58 f ormation.Although determining the sequence and structural elements that govern T rm6 / T rm61 activity appeared complicated, we identified nucleotide A54, which interacts with the Hoogsteen face of the target A58, as a key element for the efficient introduction of m 1 A58 into tRNA i Met .

The introduction of m 1 A58 by T rm6 / T rm61 to the yeast tRNA i
Met is slightly stimulated by m 5 C48,49 After studying the structural effects on the introduction of m 1 A58 by mutating nucleotides implicated in the unique tRNA i Met elbow structur e, we wonder ed whether mor e subtle alterations that may also affect the tRNA i Met local structure could modulate m 1 A58 incorporation.In particular, posttranscriptional modifica tions tha t are close in space to m 1 A58, and tha t participa te in the tRNA i Met tertiary interactions, could be considered as prime targets.Among tRNA i Met modifications, m 5 C48 meets these criteria.Indeed, m 5 C48 is relati v ely close to m 1 A58 in the tRNA i Met structure ( < 10 Å ), and m 5 C48 and m 1 A58 are together implicated in the particular tRNA elbow structure of tRNA i Met involving the previously mentioned noncanonical nucleotides A20, A54 and A60 ( 43 ).Mor e pr ecisely, m 5 C48 is involved in an intricate network of interactions with G15, A20 and A59, with A59 and A20 forming a relay with another network involving A60 and m 1 A58 (Supplementary Figure S7).Another aspect prompted us to examine the link between m 5 C modifications and m 1 A58.Indeed, since the yeast trm4 Δ mutant has been implicated in the RTD pathway in combinations with se v eral other mutations (24)(25)(26), and since hypomodified tRNA i Met is targeted by the nuclear surveillance pathway and the RTD pathway, we wondered whether the modifications introduced by Trm4 could have an impact on the introduction of m 1 A58 by T rm6 / T rm61, thereby affecting tRNA i Met stability.For these reasons, we investigated whether m 5 Cs have any effect on the introduction of m 1 A58 in tRNA i Met .Note that in yeast initiator tRNA i Met , Trm4 introduces m 5 Cs at two positions, namely m 5 C48 and m 5 C49 (Supplementary Figure S1).We therefore conducted activity assays with T rm6 / T rm61 on m 5 C48,49-tRNA i Met .We observed that T rm6 / T rm61 is about 40% more efficient in the presence of m 5 C48,49 as compared with the unmodified tRNA i Met (Figure 3 B, Table 1 , and Supplementary Figure S5f).This corresponds to an efficiency of about 1.1 times the one measured on the doubly-modified 55-T54-tRNA Phe (Table 2 ).Thus, e v en though the m 5 C48,49 modifications are not strictly r equir ed for m 1 A58 introduction by T rm6 / T rm61, their presence enhances the efficiency of m 1 A58 introduction in tRNA i Met in vitro .In order to get a clearer idea of the origin of m 5 Cs positi v e effects on m 1 A58 introduction, we analysed m 5 Ccontaining tRNA i Met with NMR spectroscopy.We produced an m 5 C48,49-tRNA i Met sample 15 N-labelled on its imino groups, thereby allowing for the measurements of 2D 1 H-15 N NMR spectrum, which corresponds to its NMRfingerprint and reflects folding homogeneity and structural integrity, as explained previously for the tRNA Phe and tRNA i Met variants.The comparison of the 1 H-15 N BEST-TROSY experiments of unmodified and m 5 C48,49-tRNA i Met samples re v ealed mar ked differ ences (Figur e 5 A, B).Additional signals appear on the spectrum of the m 5 C48,49-tRNA i Met , and a decrease in signal line broadening is observed.In addition, the signal heterogeneity present in the unmodified tRNA i Met , with weak and strong signals coexisting, is less pronounced in the m 5 C48,49-tRN A i Met spectrum, w hich shows a more homogeneous signal profile, with overall stronger signals than in the unmodified tRNA i Met (Figure 5 A, B).As previously explained, NMR signals of RNA imino groups are only observed on condition that the imino protons are pr otected fr om exchange with the solvent by hydrogen bonding in any type of base pairing.The decrease in signal heterogeneity in the m 5 C48,49-tRNA i Met ther efor e r eflects mor e stable base pairs, as well as a less dynamic and more homogeneous folding of this tRNA.The introduction of m 5 Cs by Trm4 there- fore induces local and / or global changes in the folding of tRN A i Met , w hich could explain the increased efficiency of m 1 A58 incorporation (Figure 3 B, Tables 1 and 2 ).

Met
Since m 1 A58 is involved in the particular tRNA elbow structure of tRNA i Met (see ( 43 ) and text above), and since the m 1 A58 modification is essential for tRNA i Met stability and pre v ents its degradation by the nuclear surveillance and RTD pathways ( 27 , 28 , 42 ), we examined the effect of this single modification on the structural properties of tRNA i Met .We thus produced a 15 N-labelled m 1 A58containing tRNA i Met sample following our generic approach, and measured its NMR-fingerprint (Figure 5 C).The comparison of the 1 H-15 N BEST-TROSY spectra of the unmodified and of the m 1 A58-tRNA i Met (Figure 5 A-C) re v ealed considerab le changes in the structural properties of tRNA i Met upon m 1 A58 modification.The pronounced signal heterogeneity present in unmodified tRNA i Met (Figure 5 A) is completely absent in m 1 A58-tRNA i Met (Figure 5 C), the NMR spectra of which display the characteristics of a stable and homo geneousl y folded tRNA.Thus, a single modification has major effects on the structural properties of yeast tRN A i Met , w hich can likel y explain w h y h ypomodified tRNA i Met lacking m 1 A58 is targeted by degradation pathways.
To get a deeper understanding of the structural changes arising upon m 1 A58 introduction, we performed the assignment of the imino resonances of the m 1 A58-tRNA i Met following standard methods (Figure 5 C), as previously described for other tRNAs ( 49 ).With this assignment at hand, we noticed that the imino signals of G18 and U55 are only visible in the spectrum of the m 1 A58-tRNA i Met (Figure 5 A-C).These nucleotides, and their respecti v e imino groups, are engaged in uni v ersally conserv ed tertiary interactions at the le v el of the elbow region of tRNAs, with the imino group of U55 forming a hydrogen bond with a non-bridging oxygen of the phosphate backbone of A58, and that of G18 forming a hydrogen bond with an e xocy clic carbonyl group of U55 ( 57 ).The detection of these imino groups in the NMR spectra of m 1 A58-tRNA i Met attests that their imino pr otons are pr otected fr om an exchange with the solvent, thereby demonstra ting tha t the tRNA elbow structure is well-assembled.The imino signals of G18 and U55 can be considered as a signature of a properly folded tRNA with a well-assembled elbow structure.Conversely, their absence in the NMR-fingerprint of the unmodified tRNA i Met and the m 5 C48,49-tRNA i Met (Figure 5 A, B) indicate that the tRNA elbow structure is not properly assembled in these tRNAs.

Met in yeast extract
The existence of a positive effect of m 5 Cs on m 1 A58 introduction in tRNA i Met in vitro (Figure 3 B), does not necessaril y impl y tha t this ef fect occurs in a cellular context.For example, if m 1 A58 is introduced before m 5 Cs, no effect of m 5 Cs on the introduction of m 1 A58 can possibly be observ ed.In or der to inv estigate whether this positi v e effect persists in a cellular context, we applied our recently developed methodology ( 39 , 40 ) to the monitoring of the introduction of m 1 A58 into tRNA i Met in yeast extracts.As seen above, the imino signals of G18 and U55 constitute an NMR signature of a properly assembled elbow structur e, and ther efor e can be r egarded as an indirect marker of m 1 A58 introduction in the case of tRNA i Met .We made use of this marker to monitor the introduction of m 1 A58 in tRNA i Met in wild-type and in trm4 Δ yeast extracts using time-resolved NMR.For that, 15 N-labelled unmodified tRNA i Met was incubated at 30 • C in yeast extracts supplemented with the modification enzymes cofactors, SAM and NADPH.A series of 1 H-15 N BEST-TROSY experiments wer e measur ed for wild-type and trm4 Δ y east extr acts (Figure 6 ).The observation of the imino signals of G18 and U55 along the tRNA i Met ma tura tion routes re v ealed that m 1 A58 is introduced slightly faster in the wild-type extract than in an extract depleted of Trm4.This shows that lack of m 5 C48,49 has a negati v e effect on m 1 A58 introduction by T rm6 / T rm61 in yeast tRN A i Met , w hich is perfectl y consistent with the in vitro kinetic assays on tRNA i Met (Figure 3 B).Overall, our data show that m 5 C modifications have a positi v e effect on m 1 A58 introduction in tRNA i Met both in vitro and in a cellular context.

DISCUSSION
In this study, we implemented a generic approach for the preparation of specifically modified tRNAs in order to pursue a thorough investigation of the cross-talk between modifications 55, T54 and m 1 A58 in yeast tRNA Phe .We demonstrated a direct positi v e and cumulati v e effect of modifications 55 and T54 on the incorporation of m 1 A58 in this elongator tRN A. Conversel y, we report that m 1 A58 is efficiently introduced on unmodified initiator tRNA i Met without the need of any prior modification, re v ealing distinct pathwa ys f or m 1 A58 incorporation in yeast elongator and initiator tRN As.Finall y, we show that the m 1 A58 single modification has a considerable impact on the structural properties of yeast tRNA i Met .This provides an explanation with structural basis for the degradation of hypomodified tRNA i Met lacking m 1 A58 by the nuclear surveillance and RTD pa thways.Our stud y has important implica tions for understanding tRNA modifica tion pa thways and in particular for the investigation of modification circuits.These aspects are discussed below.
Genetic approaches are v ery effecti v e strategies for identifying cross-talk between different genes, and genes encoding modification enzymes are no exception ( 30 ).These are howe v er most effecti v e when used in conjunction with biochemical approaches, allowing for a detailed characterization of the molecular aspects contributing to the observed phenotypes.Using specific deletion strains, we previously identi-fied an interdependence between the 55, T54 and m 1 A58 modifications in yeast tRNA Phe from the observation of a slo w-do wn in the incorporation of certain modifications in the absence of other specific enzymes ( 39 ).With a biochemical approach, we now establish that the incorporation of T54 is directly stimulated by 55, and that the incorporation of m 1 A58 is directly and individually stimulated by 55 and T54, with a notable cumula tive ef fect when they are both present, thus reporting that the effects of the modifications are direct and not the result of other indirect effects.These modification circuits in the T-arm of yeast elongator tRNAs concern modifications T54, 55 and m 1 A58, which are among the most conserved modified nucleotides in all sequenced tRNAs ( 58 , 59 ).These modifications participate in maintaining the uni v ersal tRNA tertiary fold, mor e pr ecisely at the le v el of the elbow region, assemb led via conserved contacts between the T-and D-loops ( 57 , 60 ).The characterization of this circuit involving modifications of the tRNA core is therefore of general interest for understanding the relation between modifications and structure in tRNAs.
Simple chemical modifications, namely an isomerisation in case of 55, and a methylation in case of T54, can thus render a gi v en tRNA a substantially better substrate for subsequent modification enzymes.The 55 → T54 → m 1 A58 and 55 → m 1 A58 modification circuits reported her e ar e robust cir cuits with highly pronounced effects, with for instance an initial velocity of m 1 A58 incorpora tion tha t is increased by a factor 15 in presence of both 55 and T54 (Table 2 ).The presence of 55 alone also greatl y stim ulates the activity of T rm6 / T rm61, with a positi v e effect on m 1 A58 incorporation that is about twotimes larger than the positi v e effect of T54 (Table 2 ).This marked effect of 55 leads to undetectable levels of m 1 A58 along the ma tura tion route of tRNA Phe in pus4 Δ yeast extracts monitored by NMR spectroscopy ( 39 ).In addition, pr evious time-r esolved NMR study of tRNA Phe in pus4 Δ and trm2 Δ yeast extracts pointed towards the m 1 A58 incorporation being more affected by 55 than by T54, which is perfectly in agreement with the kinetic data reported here.This indica tes tha t the time-resolved NMR approach we have developed in cellular extracts ( 40 ), is not only reliable to identify cross-talks between modifications, but also to discriminate between weak and strong dependencies.
The question remains of the molecular origin of the differences in the ca talytic ef ficiencies of T rm6 / T rm61 regarding tRNA Phe and tRNA i Met , as well as of the molecular basis of such ordered modification circuit in tRNA Phe .In a circuit of modifications, the observed effect of the initial modification on the subsequent enzyme is reflected in an increased turnover rate, meaning either a better substrate binding, or a better ca talytic ef ficiency, or a better product release, depending on the enzyme considered ( 61 ).For RNA modification enzymes, the rate-determining step of the reaction has been reported to be the catalytic step ( 62 , 63 ), the product release ( 64 , 65 ), or conformational changes of both the RNA and pr otein, most pr obably to accommodate the target nucleotide into the acti v e site ( 66 ).Since yeast T rm6 / T rm61 exhibits high structural similarity with its human homolog ( 67 ), the structure of human T rm6 / T rm61 in complex with tRN A L ys(UUU) can be examined to understand the tRN A reco gnition and modification mechanism of T rm6 / T rm61 ( 68 ).This structur e r e v eals that unfolding of the tRNA tertiary structure is required to allow access to the methylation target A58.In particular, the interactions between the T-and D-loops are disrupted and the D-arm is moved away from its position as a result of interactions with the N-terminal ␤-barrel domain of Trm61.Nucleotides 55-60 in the T-loop also change their conformation to accommodate the A58 target into the Trm61 acti v e site ( 68 ).This structure suggests that a weak interaction between the D-and T-arms would probably lead to a more favor able substr a te accommoda tion for m 1 A58 modification by T rm6 / T rm61.This could explain why unmodified tRN A i Met , w here the tRN A elbow is not properl y assembled without the m 1 A58 modification (Figure 5 ), is efficiently modified by T rm6 / T rm61.On the contrary , gi v en the structural properties of the unmodified tRNA Phe (Supplementary Figure S6), this substrate would be less favourably recognized by T rm6 / T rm61.Furthermore, in the case of elonga tor tRNA Phe , the stabiliza tion of the T-arm structure via the modifications T54 and 55 has a positi v e effect on m 1 A58 incorporation by T rm6 / T rm61 (Figure 2 and Table 1 ).The comparison of the NMR spectra of 55-and T54containing tRNA Phe with that of the unmodified tRNA Phe , shows very limited chemical shift variations, suggesting that these modifications do not induce large global rearrangements in the structure of tRNA Phe ( 39 , 49 ).Howe v er, since nucleotides 55-60 in the T-loop largely change their conforma tion upon accommoda tion of the A58 target into the acti v e site of Trm61 ( 68 ), modifications 55 and T54 may directly affect this step of T-loop reorganization.These modifications could indeed lead to local and / or global changes in the dynamic properties of the tRNA substrate.Such a mechanism has recently been reported in the case of E. coli tRN A f Met , in w hich conforma tional fluctua tions on the local le v el ar e incr eased in the modified tRNA ( 69 ).Modifications could thus help reach otherwise inaccessible structural conforma tions tha t ar e mor e suited to substrate accommodation by the next modification enzyme, which would explain the increased efficiency of m 1 A58 incorporation in presence of T54 and 55.
Even though modification cir cuits ar e widespr ead and have been reported in several organisms, including S. cerevisiae , S. pombe , E. coli , T. thermophilus , drosophila, human and plants ( 32-38 , 70-74 ), the role of such ordered circuits of modifications remains an open question.For modification circuits in the anticodon-loop region, howe v er, it has been recently proposed that modifications introduced first act as additional recognition elements for the subsequent enzyme.This w ould pro vide the means for adding modifications with considerable variation in the anticodon-loop region ( 31 ).This hypothesis is quite con vincing f or modifications in the anticodon-loop region, but cannot explain the actual modification circuit in the T-loop of yeast elonga tor tRNAs.This modifica tion cir cuit in the tRNA cor e indeed involves modifications that are highly conserved.Until recently, modification circuits in the tRNA core region have been only reported in the case of the extremelythermophilic bacterium T. thermophilus ( 37 , 38 ), and are most likely not implicated in sequential orders of modifica tion incorpora tion, but ra ther in a fine tuning of modification le v els in relation to an adaptation to variations in growth temperature ( 75 ).The 55 → T54 → m 1 A58 modification circuit in yeast elongator tRNAs ther efor e constitutes the first description of an order ed cir cuit of modification involving modifications from the tRNA core region.Since their identification remains difficult, particularly because real-time monitoring of tRNA ma tura tion a t a single nucleotide le v el is technically challenging ( 76 ), we are convinced tha t modifica tion cir cuits in the tRNA cor e r egion are certainly more widespread than currently thought.Such cir cuits ar e likely to be identified in the near future through the use of nanopore sequencing technolo gies a pplied to tR-NAs ( 77 , 78 ).
One of the most striking features of our study concerns the changes in the structural properties of tRNA i Met upon m 1 A58 modification.NMR-fingerprints of unmodified tRNA i Met and m 1 A58-tRNA i Met indeed re v ealed important structur al rearr angements upon addition of a single methyl group.Even though the NMR spectra of unmodified tRNA i Met indicate a certain dynamic that probably leads to intermediate exchange on the NMR chemical shift time scale, the presence at the almost exact same chemical shifts of the imino groups of U42, U50, G68, G70, G12, G24 and G30 in the NMR-fingerprints of unmodified-and m 1 A58-tRNA i Met attest to the proper secondary structure assembly of this tRNA (Figure 5 ).All RNA helices, namely the T-, D-, anticodon-and the acceptor-stems, are thus likel y correctl y assemb led.Howe v er, the three-dimensional structure of the tRNA is not properly formed as demonstrated by the lack of signals attesting to a properly assembled tRNA elbow structure, namely imino groups of U55 and G18 (Figure 5 ).These structur al rearr angements of yeast tRNA i Met upon m 1 A58 modification are most prob-ably at the origin of the specific degradation of hypomodified tRNA i Met lacking m 1 A58 by the nuclear surveillance and RTD pathways ( 27 , 28 , 41 , 42 ), while the properly folded m 1 A58-tRNA i Met is pr otected fr om degradation.It is noteworthy that other hypomodified tRNAs lacking at least one tRNA cor e modification ar e targeted to degradation by the RTD pathway.In all reported cases, this degradation in the absence of one or two modifications is tRNA specific, meaning that only specific tRNAs are targeted to degradation.For instance, tRNA Val(AAC) lacking m 7 G46 and m 5 C49 is ra pidl y degraded by the RTD pathway in S. cerevisiae ( 24 ); tRN A Ser(CGA) and tRN A Ser(UGA) lacking either Um44 and ac 4 C12 or m 2,2 G26 and m 5 C48 are also ra pidl y degraded by the RTD pathway in S. cerevisiae ( 26 , 79 ); tRNA Tyr(GUA)  and tRNA Pro(AGG) lacking m 7 G46 are rapidly degraded by the RTD pathway in S. pombe ( 80 ).Comparing these reports with the case of tRNA i Met lacking m 1 A58, it is tempting to speculate that the modifications involved might be responsible for large structural effects and stabilize the tRNA tertiary structure in these particular cases.The same modifications would, in comparison, not much alter much the structure of non-targeted tRNAs, a hypothesis that would need to be tested experimentally in future structural work.
Another important point re v ealed by the monitoring of the m 1 A58 introduction in tRNA i Met in yeast extracts resides in the fact that our NMR-based methodology for monitoring tRNA ma tura tion in cell extracts has the ability to report both on the introduction of chemical modifications, and on structural changes occurring during maturation.This point was not full y a ppreciated in the NMR study of yeast tRNA Phe , since this tRNA is, to a certain extent, properly folded without modifications ( 39 ).Changes in the NMR spectra of tRNA Phe upon modification are modest ( 39 , 49 ), and mainly reflect the incorporation of new chemical gr oups, with pr obably also some minor structural rearrangements.The example of tRNA i Met has highlighted that NMR spectroscopy is an ideal method that can report, in a time-resolved fashion, on how the modification process affects tRNA structural properties.
In this work, we have described different modification pathwa ys f or m 1 A58 incorporation in yeast elongator and initiator tRNAs.Unmodified elongator tRNA Phe is an intrinsically poor substrate of T rm6 / T rm61, whereas unmodified tRNA i Met is an intrinsically good substrate of the same enzyme.This raises the general question of what makes a good versus a poor substrate for a modification enzyme?To look into this matter, it is important to bear in mind tha t modifica tions may not necessarily have the same beneficial effect on all tRNAs ( 81 ).For instance, a certain modification may be particularly important for a certain tRNA, which constitutes the evolutionary pr essur e for retaining this modification enzyme, but might be much less important, if significant at all, in other tRNAs.In this context, dealing with good and poor substrates r epr esents an ordinary challenge faced by modification enzymes.Indeed, tRNAs are to some extent sufficiently similar to be recognized and employed by the translation machinery, but need at the same time to be sufficiently different to be uniquely recognized by their cognate aminoacyl-tRNA synthetases.The modification enzymes ther efor e should handle a population of highly similar but unique tRNAs and the tRNA modifica tion pa tterns can be regarded as the result of millions of years of coevolution of modification enzymes with the tRNA population ( 11 , 82 ).In this context, we believe that a potential role of modification circuits could be to allow the modification of both good and poor tRNA substrates.In the case of yeast elongator and initiator tRNAs, w hich m ust have sufficientl y different structural properties to be recognized by elongation or initiation factors, the existence of the 55 → T54 → m 1 A58 modification circuit enables the incorporation of m 1 A58 in certain elongator tR-NAs, such as tRNA Phe , which are poor intrinsic substrate of T rm6 / T rm61.In conclusion, modification circuits might be a solution found to deal with the problem of having at the same time poor and good tRNA substra tes tha t all r equir e to be e v entually modified.

Figure 1 .
Figure 1.A generic approach to pr epar e tRNAs with specific modifications.(Step 1) The unmodified tRNA is transcribed in vitro and purified by anion exchange chromato gra phy.(Step 2) The desired modification enzyme is ov ere xpressed in E. coli and purified by immobilized metal affinity chromato gra phy (IMAC) and further purification steps if needed.(Step 3) The unmodified tRNA is modified in vitro with the purified modification enzyme in presence of cofactors and subsequently purified by anion exchange chromato gra phy.(Step 4) A quality control step is performed by 1D 1 H NMR in order to establish that the desired modifications were fully incorporated in the tRNA population.

Figure 2 .
Figure 2. Influence of pre-existing modifications on Trm2 and T rm6 / T rm61 activities on tRNA Phe .( A ) Time course of the introduction of T54 in tRNA Phe depending on the prior presence (orange) or absence (black) of the 55 modification.( B ) Time course of the introduction of m 1 A58 in tRNA Phe depending on pre-existing modifications: unmodified tRNA Phe (black), single modified T54-tRNA Phe (green) and 55-tRNA Phe (orange), and double modified T54-55-tRN A Phe (purple).Modified tRN A quantities wer e measur ed for 4 or 5 time points in at least three independent experiments ( N = 3 or 4), and initial velocities ( V i ) were determined by linear regression (see Tables1 and 2).

Figure 3 .
Figure 3. Influence of pre-existing modifications on m 1 A58 abundance and on T rm6 / T rm61 activity on tRNA i Met .( A ) Quantitative analysis of nucleoside modifications in yeast tRNA i Met with LC-MS / MS.Histograms showing the relati v e abundance of m 1 A58 modification in purified yeast tRNA i Met

Figure 4 .
Figure 4. Influence of specific n ucleotide s wa pping between tRN A Phe and tRNA i Met on T rm6 / T rm61 activity.Histogram comparing the quantity of m 1 A58 introduced by T rm6 / T rm61 in tRNA Phe and tRNA iMet variants (in nM).Names of the tRNAs are indicated below the graph and correspond to a specific nomenclature (see Supplementary FigureS6for correspondence and details).Black dots represent individual measurements.Modified tRNA quantities wer e measur ed for 1 time point at t = 1 h in six independent experiments ( N = 6).Data heights r epr esent the mean of the replicates.Error bars correspond to the confidence interval at 95% (CI 95%).

Figure 5 .
Figure 5.Effect of m 5 C48,49 and m 1 A58 on the structural properties of yeast tRNA i Met imino ( 1 H, 15 N) correlation spectra of 15 N-labelled tRNA i Met with dif ferent modifica tion sta tus measured a t 38 • C .( A ) unmodified tRNA i Met , (B) m 5 C48,49-tRNA i Met and ( C ) m 1 A58-tRNA i Met .The assignment of the imino resonances of the m 1 A58-tRNA i Met was obtained following standard methods.

Table 1 .
Initial velocities ( V i ) of Trm2 and Tr m6 / Tr m61 acting on yeast tRNAs presenting different modification profiles.Initial velocities were determined by linear r egr ession and normalized to an equivalent of 50 nM of enzyme.The reported errors correspond to the standard error (SE) of the slope determination (see material and methods)

Table 2 .
Ratios of initial velocities ( V i ) showing enzyme efficiency depending on the presence of pre-existing modifications on the yeast tRNA Phe and tRNA iMet .The reported errors of the ratios were calculated by taking into account the propagation of uncertainties Time-resolved NMR monitoring of m 1 A58 introduction in tRNA i Met in yeast extracts.( A ) Imino ( 1 H, 15 N) correlation spectra of a 15 N-labelled tRNA i Met measured in a time-resolved fashion during a continuous incubation at 30 • C in yeast wild-type extract over 16 h.( B ) Imino ( 1 H, 15 N) correlation spectra of a 15 N-labelled tRNA i Met measured in a time-resolved fashion during a continuous incubation at 30 • C in yeast trm4 Δ extract over 16 h.Each NMR spectrum measurement spreads over a 2 h time period, as indicated.