Structural basis for the selective methylation of 5-carboxymethoxyuridine in tRNA modification

Abstract Posttranscriptional modifications of tRNA are widely conserved in all domains of life. Especially, those occurring within the anticodon often modulate translational efficiency. Derivatives of 5-hydroxyuridine are specifically found in bacterial tRNA, where 5-methoxyuridine and 5-carboxymethoxyuridine are the major species in Gram-positive and Gram-negative bacteria, respectively. In certain tRNA species, 5-carboxymethoxyuridine can be further methylated by CmoM to form the methyl ester. In this report, we present the X-ray crystal structure of Escherichia coli CmoM complexed with tRNASer1, which contains 5-carboxymethoxyuridine at the 5′-end of anticodon (the 34th position of tRNA). The 2.22 Å resolution structure of the enzyme-tRNA complex reveals that both the protein and tRNA undergo local conformational changes around the binding interface. Especially, the hypomodified uracil base is flipped out from the canonical stacked conformation enabling the specific molecular interactions with the enzyme. Moreover, the structure illustrates that the enzyme senses exclusively the anticodon arm region of the substrate tRNA and examines the presence of key determinants, 5-carboxymethoxyuridine at position 34 and guanosine at position 35, offering molecular basis for the discriminatory mechanism against non-cognate tRNAs.


INTRODUCTION
Transfer RN A (tRN A) deli v ers a codon-specific amino acid to ribosome playing a central role as an adaptor during translation ( 1 ).Most RNA molecules undergo posttranscriptional modifications; over 150 distinct modifications have been identified in RN A, w here 80% of those are found in tRNAs ( 2 ).On average, bacterial and eukaryotic tRNA contains 8 and 13 modifications per molecule ( 3 ), respecti v ely.Lack of proper post-transcriptional modifications in tRNA affects mRNA decoding capabilities ( 4 ), cellular growth ( 5 ) and regulatory function ( 6 ).In humans, certain irregular tRNA modifications have been associated with neurodegenerati v e and metabolic diseases ( 7 ).
The tRNA-modifying activity and the specificity of E. coli CmoM was first reported by Suzuki and his colleagues, who identified that mcmo 5 U34 was present in tRNA Ala1 , tRN A Ser1 , tRN A Pro3 and tRN A Thr4 , w hereas cmo 5 U34 was in tRNA Leu3 and tRNA Val1 ( 17 ) (Figure 1 B).Authors also demonstra ted tha t the terminal methyl group of the mcmo 5 U contributed to pre v enting frameshift during decoding GCG codon, although no detectable growth defects could be observed with cmoM -deficient mutant cells.CmoM belongs to Class I SAM-dependent methyltransferase family and the SAM-bound structure of CmoM from E. coli str.O157:H7 is available (PDB ID: 4HTF).Howe v er, it is not clear how the enzyme interacts with a specific tRNA substrate and installs the methyl group on the hypomodified wobble uridine as this structure lacks RNA component.
Her e, we r eport an X-ray crystal structure of E. coli CmoM complexed with sinefungin, an analog of SAM, and cellularly expressed E. coli tRNA Ser1 containing cmo 5 U at the wobble position.The 2.22 Å resolution complex structure clearly unveils detailed molecular interactions between cmo 5 U and the enzyme at atomic le v el, providing critical clues to the molecular basis for discriminating the cognate versus non-cognate tRNAs.Combined with the structural information, we further verified the specificity elements within both tRNA and the enzyme by examining the cellular and in vitro activities of CmoM variants.

Cloning and protein purification of CmoM variants
E. coli cmoM gene implemented on pCA24N vector was obtained from the E. coli ASKA library (NBRP, Japan).Point mutations were introduced by in vivo mutagenesis as described previously ( 18 ).Primers used for cloning are listed in Supplementary Table S1.Sequences were verified by standard sequencing (Macrogen, Korea).E. coli BL21 (DE3) cells were transformed with the vectors harboring a cmoM variant, grown in LB containing 100 g / ml ampicillin a t 37 • C , and induced with 0.5 mM Isopropyl ␤-D-1thiogalactopyranoside (IPTG) at an OD 600 of ∼0.6.Cells were further incubated at 20 • C overnight and harvested by centrifugation.Cell pellets were resuspended in lysis buffer (25 mM HEPES (pH 7.5) and 150 mM NaCl) and sonicated for 30 min.The lysate was centrifuged at 39 000 g for 30 min and the supernatants were applied to the HisTrap HP column (GE Healthcare) charged with Ni 2+ .The column was equilibrated with wash buffer (25 mM HEPES (pH 7.5), 300 mM NaCl, and 20 mM imidazole) and eluted with a linear gradient of imidazole (0.02-0.5 M).Purified protein was analyzed by SDS-PAGE and buffer exchanged with the lysis buffer using Amicon centrifugal filters (Merck).An extinction coefficient of ε 280 = 42 400 M −1 cm −1 was used to calculate the concentration of the protein.

Determination of molecular weight of CmoM by size e x clusion chromatography
The molecular weight of recombinant CmoM was determined by size exclusion using Super de x 75gl column (GE Healthcare, Life Sciences) on NGC FPLC system (Biorad).100 g of the wild-type CmoM (calculated monomer molecular weight = 31.7 kDa) was loaded onto the column with mobile phase (25 mM HEPES (pH 7.5) and 150 mM NaCl) at a flow rate of 0.5 ml / min.Elution of ovalbumin (44 kDa) and MnmC (76 kDa) from the column was also tested under the identical condition for comparison.

Purification of in vivo transcribed tRNA
The plasmid pBSTNAV (Addgene, USA) was used as the tRNA e xpression v ector which contains LPP promoter and rrnC terminator.E. coli str.K-12 substr.MG1655 serT (tRNA Ser1 (TGA)) gene was amplified by polymerase chain reaction from synthetic DNA template (IDT, USA).The primers used are summarized in Supplementary Table S1.The pBSTNAV vector and the amplified PCR fragments were double digested in 37 • C water bath for 2 h using the restriction enzymes EcoRI (Enzynomics, Korea) and HindIII (Enzynomics, Kor ea).Digested products wer e purified by agarose gel electrophoresis.Next, ligation step was performed using T4 ligase (Enzynomics, Korea) at room temperature for 4 h, and the production of pBSTNAV-serT plasmid was confirmed by DNA sequencing (Macrogen, Korea), which was used to transform cmoM -deficient competent cell (NBRP, Japan) using heat-shock method.
For ov ere xpression of tRNA, cmoM -deficient E. coli cells were transformed with pBSTNAV-serT plasmid and grown in 500 ml LB containing 100 g / ml ampicillin at 37 • C for 24 h.Cells were collected by centrifugation and cellular RNA was extracted by sa tura ted phenol (pH 4.5) (Biosesang, Korea) following the manufacturer's protocol.3 M sodium acetate (pH 5.3) and isopropanol were added to precipitate RNA, followed by three times of washing with 75% (v / v) cold ethanol.RNA loading dye containing 0.025% (w / v) bromophenol blue and xylene cyanol FF in 95% (v / v) was added to the RNA solution and the mixture was applied to a 7 M urea-PAGE gel, which was run for 14 h at 40 W. The r ecombinantly over expr essed tRNA Ser1 could be resolved from endogenous tRNAs on the gel due to the difference in size.The band containing the desired tRNA was excised by scalpel and tRNA was eluted from the gel strip using Elutrap (W ha tman).An extinction coef ficient of ε 260 = 852 000 M −1 cm −1 was used to quantify tRNA Ser1 .Purified tRNA was denatured at 95 • C for 5 min followed by incubation on ice for 20 min before use.Crystallization, data collection, and structure determination of CmoM-tRNA Ser1 (cmo 5 UGA)-sinefungin complex Crystallization of CmoM-tRNA complex was performed by mixing 1:4 molar ratio of CmoM and in vivo transcribed tRNA Ser1 (cmo 5 UGA) extracted from cmoM -deficient cells.Sinefungin (Sigma-Aldrich) was added to the mixture at the final concentration of 3 mM, which was used for co-crystallization of ternary complex of CmoM-tRNA Ser1 (cmo 5 UGA)-sinefungin by sitting drop vapor diffusion method.An equal volume of the mixture was added to a reservoir solution (0.04 M magnesium acetate tetr ahydr ate, 0.05 M sodium cacodylate trihydrate (pH 6.0), and 30% (v / v) (+ / −)-2-methyl-2,4-pentanediol) at 25 • C. Hexagonal prism shaped crystals were formed in 5 days.Crystals were mounted in a nylon loop and flash-cooled in liquid nitrogen.X-ray diffraction data were collected on Dectris Eiger X9M detector at Pohang Accelerator Laboratory (PAL) beamline 5C, using the wavelength λ = 0.9795 Å .Structur e factors wer e scaled by XDS ( 19 ) and Aimless ( 20 ) under space group P6 1 22.To solve the phases of structure factors, molecular replacement was performed with Molrep ( 21 ) using the structure of SAM-bound CmoM from E. coli str.O157:H7 (PDB ID: 4HTF) as a starting model.Extra electron densities corresponding to tRNA were evident after initial refinements with a protein-onl y model, w hich guided us to build tRNA component using Modelcraft ( 22 ).Subsequent model building and refinement were performed iterati v ely using Modelcraft ( 22 ), Coot ( 23 ), PDB REDO ( 24 ), REFMAC5 ( 25 ), and Phenix Refine ( 26 ).Crystallographic statistics are available in Table 1 .

RESULTS
The o ver all structur e of CmoM-tRNA Ser1 (cmo 5 UGA)sinefungin complex To investigate the molecular interaction between the enzyme and the hypomodified base, cmo 5 U34, we employed recombinantly expressed and purified tRNA Ser1 from c moM -deficient E. coli mutant for structural study (Figure 2 A).In addition, an analog of SAM, sinefungin, was used for co-crystallization of the ternary complex to mimic a substrate-bound state of CmoM.The Xray co-crystal structure of CmoM-tRNA Ser1 (cmo 5 UGA)sinefungin was determined to 2.22 Å r esolution (Figur e 2 B).The asymmetric unit of the X-ray crystal structure contains a tRNA molecule bound to a protomer of CmoM, mainly through the anticodon loop on the protein surface enriched with positi v e charges (Figur e 2 C).The pr eviously r eported SAM-bound structure (PDB ID: 4HTF) displays a dimeric form of the enzyme and an essentially identical dimer can be generated with a symmetry-related protomer from our structur e (Figur e 2 D).We examined the oligomeric state of CmoM via size-exclusion chromato gra phy, w here the results were consistent with the dimeric form (Supplementary Figure S1).Analysis of the dimeric interface of the protein re v eals that the dimerization is mainly dri v en by polar interactions involving a total of eight residues from each protomer (Supplementary Figure S2) ( 28 ).Overall, the residue interactions at the dimer interface in our complex structure are essentially identical to those in the SAM-bound structure.In detail, the sidechain of R234 interacts with the carbonyl oxygen of Q169, while the imidazole ring of H153 interacts with the hydroxyl group of Y241.The remaining eight residues form a complex hydrogen bond network at the core of the interface.The sidechain of R216 interacts with the backbone carbonyl oxygens from both D213 and L215, whereas the backbone amide group of R216 interacts with the carbonyl oxygen of V161.
CmoM is a member of Class I SAM-dependent methyltr ansfer ases (MTases), which adopts a Rossmann-fold with a ␤-sheet ar chitectur e of 3-2-1-4-5-7-6, wher e the 7th ␤strand lies antiparallel to other ␤-strands (Supplementary Figure S3) ( 29 ).Ten ␣-helices flanking the ␤-sheet form a sandwich-like ␣␤␣-fold as observed in numerous Class I MTases ( 30 ).Other characteristic features of this family include the presence of conserved elements involved in SAM binding, such as the glycine-rich motif at the ␤1 (G52-G53-G54) and the amino acid residues D73 at the end of ␤2 and R26 at the first one-third position of ␣2, which interact with the ribose hydroxyl groups and the methionine carboxylate, respecti v ely.Other residues such as D3, A101, and Q102 recognizes SAM's adenine whereas Y18 interacts with the methionine carboxylate (Supplementary Figur e S4).The sear ch for structural homologs with the DALI server ( 31 ) yielded an ubiquinone / menaquinone biosynthesis methyltr ansfer ase-related pr otein fr om Thermotoga maritima (PDB ID: 2AVN) as a top hit, with an z-score of 25.9 and a root-mean-square deviation (r.m.s.d.) of 2.7 Å , which shares a sequence identity of 21% with CmoM over 248 residues (Supplementary Figure S5).
The overall conformation of tRNA-complexed CmoM observed in our model is not significantly different from that of the tRNA-free state, where the superposition of both structur es r esults in an r.m.s.d. of 0.450 Å .Moreover, the micr oenvir onment ar ound sinefungin is highl y homolo gous to that of the SAM binding site in the tRNA-free structur e, wher e sinefungin fits identically as the conformation of SAM.Notably, a total of se v en Mg 2+ sites were identified around tRNA in our model.Only two of those are directly coordinated to the phosphate oxygen of C48 and A59, whereas the others are in contact with tRNA molecule indirectly through water shell (Supplementary Figure S6).

tRNA recognition by CmoM
Ther e ar e a couple of substantial changes in local conformation induced by the tRNA-binding (Figure 3 A).The N-terminal helix ( ␣1) swings towards the bound tRNA by a pproximatel y 30 Å to establish contact with phosphate groups of C32 and U33.Another notable change is observed in the organization of the loop encompassing residues 170-188, which is mostly disordered in the tRNAfr ee structur e.In the tRN A-bound state, this loop mainl y interacts with the phosphates of G30, U31, A36, and i 6 A37 through the formation of salt bridges.All in all, eight nucleotides of tRNA (G30 through i 6 A37) are in direct contact with CmoM, six of which are located on the anticodon loop (Figure 3 B,C).A total of 18 amino acid r esidues ar e found at the interface with tRNA, where eight of those sense the shape of the phosphate backbone on the anticodon loop through making salt-bridges or hydrogen bonds; i.e.K177 with G30, R178 with G30 and U31, K12 with C32 and U33, N16 with U33 and cmo 5 U34, K22 with cmo 5 U34, H158 and R246 with G35, and K176 with A36 and i 6 A37.Four amino acid residues make contact with the ribose ring of cmo 5 U34 and G35 by forming hydrogen bonds; i.e. the sidechains of Y150 and S181 with the 2 -hydroxyl group of cmo 5 U34, whereas those of N159 and N164 with the O4' of G35.
Only three nucleobases of the entire tRNA molecule, U33, cmo 5 U34, and G35, participate in ionic or polar interactions with the protein.U33 appears to interact solely with K218 by forming a hydrogen bond via the O2 of uracil (Figure 3 D).Meanwhile, the hypomodified base of cmo 5 U34, which has flipped out from the canonical stacking position, exhibits the most extensive interactions with the enzyme (Figure 3 E).A total of fiv e residues are involved in multiple ionic / polar interactions with the modified uracil base, where the 5-carbo xymetho xy group is thoroughly examined by R26, W124 and Y247.Notably, R26 of CmoM forms a salt bridge and hydrogen bond with the carboxyl and ether oxygen, respecti v ely.Furthermore, the carboxyl group on the modified uracil base engages in a hydrogen bonding network with the side chains of W124 and Y247.Ther efor e, these thr ee r esidues appear to be primarily responsible for the proper positioning of the nucleophilic carboxyl group on cmo 5 U34, which is approximately 3 Å away from the N ε of sinefungin.Lastly, K22 interacts with O2, and R209 interacts with N3 and O4 of cmo 5 U34.Meanwhile, the guanine base of G35 interacts with three amino acid residues (Figure 3 F).The backbone carbonyl oxygen of A162 is within hydrogen bonding distance from N1 and N2.N164 forms hydrogen bonds with N2 and N3 via its sidechain, while K218 interacts with O6 and N7.

Biochemical analyses of key amino acid residues in tRNA binding and catalysis
To verify the functional role of the amino acid residues that were identified to interact with tRNA in our structur e, site-dir ected mutagenesis was performed to introduce a single mutation and the biochemical activity of the mutant proteins was examined (Figure 4 ).Previously, Suzuki's group tested the biosynthesis of mcmo 5 U in tRNA Pro3 from cmoM-deficient strain complemented with plasmidencoded cmoM -variants harboring an alanine mutation at R26, D73, W124, Y150, R209, D213, R246 and Y247, which were identified around the SAM-binding site in the tRNA-fr ee structur e ( 17 ).Ther efor e, these r esidues wer e excluded from our inv estigation.Moreov er, w e follow ed the time-course of the accumulation of SAH, a coproduct of the methyltransfer reaction, to quantitati v ely measure the in vitro activity of the enzyme using recombinantly purified CmoM and tRNA Ser1 containing cmo 5 U at the wobble position.
The assay results highlight the importance of N164, which, along with A162 and K218, is one of the three r esidues r eco gnizing G35 base.When m uta ted to asparta te, the mutant enzyme showed a se v er diminishment in methyl transfer activity.Additionally, K22 also appears to be critical in our assay, which forms a hydrogen bond with the base and salt bridge with the phosphate of cmo 5 U34.Mutation of this residue to glutamate resulted in total inactivation, suggesting its importance in identifying the flipped-out base of cmo 5 U34.Other notable residues include K12 and N16, as the mutation of each led to a substantial decrease in activity, nearly 80% lower than that of the wild-type enzyme.Interestingly, both of these residues are located on the N-terminal helix, ␣1, which was observed to undergo a mov ement towar ds tRNA upon binding as shown in the structure.When we tested a mutant CmoM lacking ␣1, the truncated enzyme was completely inacti v e.This suggests that the binding of tRNA had been seriously compromised, underscoring the essential roles of the N-terminal helix.The other secondary structure that displays substantial conformational rearrangement induced by tRNA docking is the loop composed of amino acid residues from 170 to 188.Within this loop, K176, K177 and R178 are identified as interacting with the phosphate backbone of the tRNA substrate through the formation of salt bridges.While K176E and K177E did not substantially affect the enzyme activity, R178E resulted in an approximate 50% decr eased activity compar ed to the wild-type.Meanwhile, the alanine mutant of H158, which interacts with the phosphate of G35, displays a negligible effect on methylation activity.Similarl y, w hen K218, w hich interacts with both U33 and G35 bases, was altered to glutamate, the activity of the mutant was comparable to that of the wild-type.Interestingly, K218 is the only residue in the structure that contacts the U33 base, suggesting that U33 is not a crucial determinant for CmoM.

Conformation of full-length tRNA Ser1
All 88 nucleotides in tRNA Ser1 could be clearly modeled in the pr esent structur e (Figur e 5 A and Supplementary Figure S7).To our knowledge, this r epr esents the first 3-D structure of a cellularly transcribed, full-length tRNA Ser1 .The most notable feature observed in the CmoM-bound tRNA conformation is the rearrangement of the anticodon loop region, particularly the flipping out of cmo 5 U34, whereas anticodon bases are in general stacked on one another as observed in other free-or ribosome-bound tRNA structures ( 32-34 ) (Figure 5 B).Furthermore, our tRNA structure exhibits an atypical U-turn motif within the anticodon loop, where two characteristic hydrogen bonds r equir ed for the formation of the anticodon U-turn ( 35 ) are not observed; i.e. the interaction between N3 of U33 and the phosphate of N36, as well as that between the 2 -OH of U33 and the nucleobase of N35.This deformation of the anticodon loop is likely a result of the accompanying flip-out of cmo 5 U34.Another interesting structural feature of the CmoM-bound tRNA Ser1 is the presence of a long variable arm composed of 16 nucleotide residues; C44, G45, U46, C47, C47a, C47b, G47c, A47d, A47e, A47f, G47g, G47h, G47i, A47j, U47k, and C48 (Figure 5 C).This e xtended variab le arm is significantly longer than those found in other tRNA species, which are typically composed of 4-5 nucleotides ( 36 ).The variable arm is well-defined in our structure mainly due to the lattice contacts provided by symmetry-related molecules at the unit cell interface.The stem region of the variable arm is composed of four Watson-Crick (U46-A47j, C47-G47i, C47a-G47h and C47b-G47g) and two non-canonical (G45-U47k and G47c-A47f) base pairs, which is capped with a short turn composed of four nucleotides (G47c to A47f).

Native tRNA modifications
Because we used cellularly deri v ed tRNA for crystallization, we attempted to model all known post-transcriptional modifications in the structur e. Ther e ar e se v en different types of modified nucleotides in E. coli tRNA Ser1 in addition to the hypermodified wobble uridine ( 37 ) ; e.g.4-thiouridine (s 4 U8), 2 -O -methylguanosine (Gm18), dihydrouridine (D20 and D20a), 2 -O -methylcytidine (Cm32), 2-methylthio-N 6 -isopentenyladenosine (ms 2 i 6 A37), 5methyluridine (m 5 U54), and pseudouridine (P55).The presence of certain modifications was evident in the F o -F c differential Fourier density ma p, especiall y bulky ones such as 5-caro xymetho xy moiety on U34 and N 6 -isopentenyl group on A37 (Supplementary Figure S8).Interestingly, we could not model 2-methylthio group on A37 because of the complete lack of electron density, e v en though this residue is known to contain both N 6 -isopentenyl and 2-methylthio moieties (ms 2 i 6 A).Meanw hile, it was generall y ambiguous to model relati v ely small-sized modifications based on the electron density with the gi v en resolution, for example , D , Cm, Gm and P.
To examine whether the tRNA sample contains all the expected modifications, we analyzed the content of the recombinant tRNA Ser1 using mass spectrometry.The tRNA sample was digested to nucleosides or nucleotides and then subjected to ultra performance liquid chromatography (UPLC) coupled with electrospray ionization mass spectrometry.To obtain a comprehensi v e profile of posttranscriptional modifications present in the sample, the nucleosides were analyzed in positi v e mode, while the nucleotides containing a 5 -phosphate were analyzed in negati v e mode.Pseudouridine was excluded from our analyses as it cannot be distinguished from uridine through simple mass spectrometry.We confirmed the presence of the 5 -monophosphates of s 4 U, D, cmo 5 U and m 5 U in negati v e mode, while Gm, Cm and ms 2 i 6 A were observed in positi v e mode (Supplementary Figure S9).Surprisingly, we also detected N 6 -isopentenyladenosine (i 6 A) in both positi v e mode, likely as a precursor of ms 2 i 6 A at position 37, and negati v e mode as its 5 -monophosphate form.To determine the relati v e abundance of tRNA species containing ms 2 i 6 A or i 6 A, the intensities of the characteristic fragments, N 6 -isopentenyladenine and 2-methylthio-N 6isopentenyladenine, were measured using multiple reaction monitoring (MRM) mode analysis.Based on the comparison of peak areas for each modified base, it is estimated that i 6 A-containing species are a pproximatel y 5.3-times more abundant than ms 2 i 6 A-containing one (Supplementary Figure S10).

DISCUSSION
The most critical determinant on tRN A for reco gnition by CmoM is cmo 5 U34.In the structure, the calculated buried surface area of this nucleotide alone accounts for approximately 30% of the total interface between the protein and tRNA (317.5 out of 1072.6 Å 2 ).It is plausible that the flipped-out hypomodified base from the canonical stacked conformation is stabilized by the e xtensi v e intermolecular interactions with eight amino acid residues from the enzyme.Notably, R26, W124 and Y247 appear to fine-tune the orientation of the nucleophilic carbo xymetho xy group on the wobble uridine to facilitate the efficient methyl transfer step.Indeed, the previous mutagenesis study showed that the enzyme activity is abolished with R26A and W124A, or partially decreased with Y247A ( 17 ).Due to its low pKa value, it is likely that the carbo xymetho xy group does not require a general base to become an efficient nucleophile during methyltransfer reaction, which is further ensured by the ionic / polar interactions with those residues (Figure 6 ).In E. coli , it has been demonstra ted tha t tRN A Ala1 , tRN A Ser1 , tRN A Pro3 and tRN A Thr4 , w hich contain mcmo 5 U34, have guanosine at position 35, while tRNA Leu3 and tRNA Val1 , which contains cmo 5 U34, have adenosine at the corresponding position.Ther efor e, G35 is likely an important determinant for CmoM in distinguishing its substrates, although a low-le v el mcmo 5 U formation in tRNA Leu3 and tRNA Val1 has been reported at a frequency below 10% ( 17 ).Our structure of CmoM-tRNA Ser1 (cmo 5 UGA)-sinefungin complex provides supporting evidence by unveiling detailed The extended variable arm is a characteristic feature of class II tRNAs, including tRNA Ser , tRNA Leu , tRNA Sec and tRNA Tyr ( 38 ).Howe v er, structural information on these tRNAs with a long variable arm is quite limited.Our cocrystal structure features the full-length tRNA Ser1 containing a 16 nt-long variable arm, although it is irrelevant to the function of CmoM.Meanwhile, it was shown that a certain stem length of the variable arm is r equir ed for recognition by seryl-tRNA synthetase (SerRS) ( 39 ).Likewise, the long variable arm of tRNA Sec is crucial for the serine ligation activity of SerRS ( 40 ).Structural comparison of these tRNAs re v eals an ov erall similarity, while the variab le arm of E. coli tRNA Ser1 resembles that of Aquifex aeolicus tRNA Sec (PDB ID: 3W3S) more than that of Thermos thermophilus tRNA Ser (PDB ID: 1SER), both of which are complexed with SerRS (Supplementary Figure S11).Structural and functional characterization of more di v erse class II tRNAs will be valuable in understanding the conserved roles of long variable arms, which confer a conformational di v ersity to the canonical L-shaped scaffold.
Our co-crystal structure shows both expected and unexpected posttranscriptional modifications present in tRNA.Surprisingly, the structural data was consistent with i 6 A at position 37 instead of the expected ms 2 i 6 A. Analysis of hydrol yzed tRN A sample used for structure determination indica tes tha t it is a mixtur e of both with a r elati v e abundance of 5.3:1.In the biosynthesis of ms 2 i 6 A, MiaA and MiaB ar e r esponsible f or attaching isopenten yl moiety at N6 and methylthio group at C2 of adenine ring, respecti v ely.The cause of incomplete ms 2 i 6 A formation in our sample is unknown.One possibility is that it resulted from an artifact of using a robust ov ere xpression system, where massi v e production of recombinant tRNA occurs under the strong promotor of the expression plasmid, while the rate of MiaBdependent methylthiolation does not increase comparably.An alternati v e scenario is the potential non-constituti v e na-ture of methylthiolation at A37, although it has not been reported whether MiaB-dependent modification is inducible or not.Further investigation into the cellular and biochemical regulation of MiaB will be required to address the validity of this hypothesis.

DA T A A V AILABILITY
The atomic coordinates and structure factors for E. coli CmoM-tRNA Ser1 (cmo 5 UGA)-sinefungin are deposited to Protein Data Bank (PDB) under accession code 8JOZ.

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.

Figure 1 .
Figure 1.xo-type wobble uridine modifications in bacterial tRNA.( A ) Summary of the biosynthesis of xo 5 U34 modification in bacterial tRNA.Colors in chemical structures match those of the enzymatic activity.( B ) Gene sequences that encode tRNA anticodon arm containing cmo 5 U34 and mcmo 5 U34.Anticodon is colored in red.

Figure 2 .
Figure 2. Crystal structure of E. coli CmoM-tRNA Ser1 -sinefungin complex.( A ) Sequence of cellularly transcribed tRNA Ser1 (cmo 5 UGA) from cmoM E. coli including known modifications, which was used for crystallography and biochemical assays in this study.( B ) Overall structure of E. coli CmoM complexed with tRNA Ser1 (cmo 5 UGA) and sinefungin.CmoM and tRNA Ser1 (cmo 5 UGA) are shown in marine and orange ribbons, respecti v ely.Sinefungin is r epr esented in spher e, wher e carbon is sho wn in grey, o xygen in red, and nitrogen in blue.( C ) Docking site of tRNA (orange) on CmoM is shown, where electrostatic potential is mapped on the surface of the enzyme (positi v e in b lue, negati v e in red, and neutral in white).( D ) Superposition of reconstituted dimer of CmoM-tRNA complex (marine and orange for CmoM and tRNA, respecti v ely) and tRNA-free CmoM (PDB ID: 4htf, yellow) structures.

Figure 3 .
Figure 3. Intermolecular interactions between CmoM and tRNA.( A ) Superposed structures of tRNA-complexed (marine) and tRNA-free (PDB ID: 4htf , yello w) CmoM, where the inset highlights a large shift of ␣1 helix upon binding of tRNA.( B ) Acti v e site residues of CmoM (blue) that interact with tRNA anticodon loop region (orange).Sinefungin is shown in grey sticks.( C ) A schematic diagram tRNA showing polar interactions with the enzyme at each r esidue, wher e the phosphate is r epr esented by cir cle, ribose by pentagon, and base by square.Amino acids associated with the phosphate backbone (blue), sugar (orange), and base (black) are grouped in brackets for individual nucleotides.Amino acid residues interacting with nucleobase of ( D ) U33, ( E ) cmo 5 U34 and ( F ) G35 are shown, where dashed lines represent hydrogen bonding interaction.Distance between the amine of sinefungin and the carboxyl group of cmo 5 U34 is labeled in Å .

Figure 4 .
Figure 4. Mutagenesis studies of amino acid residues participating in ionic / polar interactions with tRNA.( A ) The conversion rates of cmo 5 U to mcmo 5 U by the wild-type and mutant CmoM are plotted, which have been determined through in vitro assays using SAM and tRNA Ser1 extracted from E. coli ΔcmoM cells.Error bars r epr esent the standar d de viation from thr ee independent measur ements at each time point.( B ) Roles of amino acid residues selected for mutagenesis and relati v e acti vities of mutants are summarized, where ++ stands for the wild-type comparable, + for moderate and -for minimal activity.

Figure 5 .
Figure 5. Features of cellularly transcribed tRNA Ser1 .( A ) Structure of the CmoM-bound tRNA with individual regions shown in different colors; acceptor stem in purple, D-arm in salmon, anticodon arm in red (anticodon in dar k grey), variab le arm in b lue, and T-arm in yellow.( B ) Superposition of anticodon stem loops of tRNA Ser1 (orange) and S. cerevisiae tRNA Phe (PDB ID: 1EHZ, grey), which highlights the conformational change in tRNA Ser1 , especially the flipping-out cmo 5 U34.( C ) A close-up view of the long variable arm of tRNA Ser1 .Base pairing interactions are represented in black dashed lines.

Figure 6 .
Figure 6.A proposed reaction mechanism for the SAM-dependent cmo 5 U methylation by CmoM.Kinetically essential residues such as R26 and W124 are positioned to fine-tune the orientation of the nucleophile and pre v ent the binding of the protonated form to promote the nucleophilicity.Subsequently, the carboxylate of cmo 5 U34 attacks the S -methyl group of SAM to form mcmo 5 U34 and SAH.