The tRNA Recognition Mechanism of Folate/FAD-dependent tRNA Methyltransferase (TrmFO)*

Background: RNA modification enzymes select specific RNAs as substrates. Results: A novel assay for folate-dependent tRNA methyltransferase (TrmFO) was developed that clarified positive and negative determinants of TrmFO. Conclusion: TrmFO recognizes a T-arm structure including the U54U55C56 sequence and G53-C61 base pair; A38 prevents incorrect methylation of U32. Significance: Studying how proteins recognize RNA is crucial for understanding RNA maturation processes. The conserved U54 in tRNA is often modified to 5-methyluridine (m5U) and forms a reverse Hoogsteen base pair with A58 that stabilizes the L-shaped tRNA structure. In Gram-positive and some Gram-negative eubacteria, m5U54 is produced by folate/FAD-dependent tRNA (m5U54) methyltransferase (TrmFO). TrmFO utilizes N5,N10-methylenetetrahydrofolate (CH2THF) as a methyl donor. We previously reported an in vitro TrmFO assay system, in which unstable [14C]CH2THF was supplied from [14C]serine and tetrahydrofolate by serine hydroxymethyltransferase. In the current study, we have improved the TrmFO assay system by optimization of enzyme and substrate concentrations and introduction of a filter assay system. Using this assay, we have focused on the tRNA recognition mechanism of TrmFO. 42 tRNA mutant variants were prepared, and experiments with truncated tRNA and microhelix RNAs revealed that the minimum requirement of TrmFO exists in the T-arm structure. The positive determinants for TrmFO were found to be the U54U55C56 sequence and G53-C61 base pair. The gel mobility shift assay and fluorescence quenching showed that the affinity of TrmFO for tRNA in the initial binding process is weak. The inhibition experiments showed that the methylated tRNA is released before the structural change process. Furthermore, we found that A38 prevents incorrect methylation of U32 in the anticodon loop. Moreover, the m1A58 modification clearly accelerates the TrmFO reaction, suggesting a synergistic effect of the m5U54, m1A58, and s2U54 modifications on m5s2U54 formation in Thermus thermophilus cells. The docking model of TrmFO and the T-arm showed that the G53-C61 base pair is not able to directly contact the enzyme.

The conserved U54 in tRNA is often modified to 5-methyluridine (m 5 U) and forms a reverse Hoogsteen base pair with A58 that stabilizes the L-shaped tRNA structure. In Gram-positive and some Gram-negative eubacteria, m 5 U54 is produced by folate/FAD-dependent tRNA (m 5 U54) methyltransferase (TrmFO). TrmFO utilizes N 5 ,N 10 -methylenetetrahydrofolate (CH 2 THF) as a methyl donor. We previously reported an in vitro TrmFO assay system, in which unstable [ 14 C]CH 2 THF was supplied from [ 14 C]serine and tetrahydrofolate by serine hydroxymethyltransferase. In the current study, we have improved the TrmFO assay system by optimization of enzyme and substrate concentrations and introduction of a filter assay system. Using this assay, we have focused on the tRNA recognition mechanism of TrmFO. 42 tRNA mutant variants were prepared, and experiments with truncated tRNA and microhelix RNAs revealed that the minimum requirement of TrmFO exists in the T-arm structure. The positive determinants for TrmFO were found to be the U54U55C56 sequence and G53-C61 base pair. The gel mobility shift assay and fluorescence quenching showed that the affinity of TrmFO for tRNA in the initial binding process is weak. The inhibition experiments showed that the methylated tRNA is released before the structural change process. Furthermore, we found that A38 prevents incorrect methylation of U32 in the anticodon loop. Moreover, the m 1 A58 modification clearly accelerates the TrmFO reaction, suggesting a synergistic effect of the m 5 U54, m 1 A58, and s 2 U54 modifications on m 5 s 2 U54 formation in Thermus thermophilus cells. The docking model of TrmFO and the T-arm showed that the G53-C61 base pair is not able to directly contact the enzyme.
The formation of m 5 U54 is catalyzed by tRNA (m 5 U54) methyltransferases (tRNA (uracil-5-)-methyltransferase; EC 2.1.1.35) (16). The enzymes can be divided into two types according to their methyl group donors: S-adenosyl-L-methionine (AdoMet) 2 -dependent methyltransferases and folate/FAD-dependent methyltransferases. AdoMet-dependent tRNA (m 5 U54) * This work was supported in part by a Grant-in-aid for Scientific Research methyltransferases are present in eubacteria, some archaea, and eukaryotes (4,5,16). Escherichia coli AdoMet-dependent tRNA (m 5 U54) methyltransferase (classical name, RNA uridine methyltransferase) is one of the best-studied tRNA modification enzymes. The enzymatic activity was initially detected in crude E. coli cell extract, after which several purification procedures were developed (17)(18)(19). The gene encoding the enzyme was identified as trmA (20). An overexpression system for TrmA in E. coli was developed, and the purification procedures were devised (21). The studies on recombinant TrmA have significantly contributed to the tRNA modification enzyme field. TrmA catalyzes methyl transfer to a 17-mer T-arm-like microhelix RNA and recognizes the conserved nucleotides U54 and C56 (22,23). Therefore, this enzyme catalyzes in vitro methyl transfer to part of 16S rRNA (24). In the reaction, TrmA forms a covalent bond complex with substrate RNA (25,26), and methyl transfer takes place by a single displacement mechanism (27).
In contrast, folate/FAD-dependent tRNA (m 5 U54) methyltransferases have been less well explored. The enzymatic activity was found in crude cell extract of Enterococcus faecalis (classical name, Streptococcus faecalis), and m 5 U formation without AdoMet was confirmed (44). The enzyme, which contains FAD and requires N 5 ,N 10 -methylenetetrahydroforate (CH 2 THF) as a methyl group donor, was purified (45). However, for a long time, the gene encoding the enzyme remained unidentified until Urbonavicius et al. (46) biochemically identified the gene, namely trmFO, from B. subtilis. Their study suggested that TrmFO is found widely among Gram-positive and some Gramnegative bacteria, including T. thermophilus (46). In our previous study, we devised an in vitro assay system for TrmFO activity and solved the crystal structures of free, tetrahydrofolate (THF)-bound, and glutathione-bound forms of T. thermophilus TrmFO (47). During the course of this study, FAD redox states of B. subtilis TrmFO were reported (48). Furthermore, it has also been reported that B. subtilis TrmFO forms a covalent bond complex with tRNA (48,49). In the current study, we focused on the tRNA recognition mechanisms of T. thermophilus TrmFO. Preparation of Enzymes-The expression systems and purification procedures for T. thermophilus TrmFO and serine hydroxymethyltransferase (SHMT) were described in the supplemental material of our previous report (47). The expression systems and purification procedures of tRNA (Gm18) methyltransferase (TrmH), tRNA (m 7 G46) methyltransferase (TrmB), tRNA (⌿55) methyltransferase (TruB), and tRNA (m 1 A58) methyltransferase (TrmI) were described in our previous reports (33, 50 -52 C]serine in buffer A (50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 6 mM 2-mercaptoethanol, and 50 mM KCl)) was incubated for appropriate times (0 -60 min) at 60°C. Non-radioisotope-labeled serine and glycine were added to the sample (final concentrations, 500 M each), and then the mixtures were frozen in liquid nitrogen. The sample (1 l each) was spotted onto a cellulose thin layer plate (Merck; TLC Cellulose F, catalog number 1.05565.0001) and separated by solvent (phenol-saturated with water, 28% ammonia water (99:1)). The [ 14 C]serine and [ 14 C]glycine were monitored with a Fuji Photo Film BAS 2000 imaging analyzer. The standard markers (nonradioisotope labeled serine and glycine) were detected by a ninhydrin reaction.

Materials-L-[U-
Measurements of TrmFO Methylation Activity-Transfer RNA transcripts were prepared by T7 RNA polymerase as described previously (32,50,51). The transcripts were purified by Q-Sepharose Fast Flow column chromatography and 10% polyacrylamide gel (7 M urea) electrophoresis (PAGE). In order to optimize the assay conditions, we tested various concentrations of enzymes and substrates, as described under "Results." Data in Figs Gel Mobility Shift Assay-The gel mobility shift assay was performed as described in a previous report (31). Purified protein and 0.05 A 260 units of T. thermophilus tRNA Phe transcript were incubated in 20 l of buffer B (50 mM Tris base, 50 mM acetic acid, and 5 mM Mg(OAc) 2 ) at 4°C for 20 min. 4 l of loading solution (0.25% bromphenol blue and 30% glycerol) was then added to each sample, and samples were resolved on a 6% polyacrylamide gel (width, 90 mm; length, 90 mm; thickness, 1 mm) prepared with 1ϫ buffer B. Electrophoresis was performed at 4°C for 1 h at 100 V. To detect protein, the gel was stained with Coomassie Brilliant Blue. Methylation blue was used for RNA detection.
Fluorescence Measurement-Fluorescence measurement was performed with a fluorescence spectrophotometer (F-2500, Hitachi) at 25°C. The excitation wavelength was 295 nm. Fluorescence intensity at 320 nm was monitored. 2 M TrmFO in buffer A was titrated with tRNA Phe transcript or tRNA Phe without the T-arm.
Inhibition Experiment-The methyl transfer reaction of TrmFO was inhibited by methylated tRNA Phe transcript, tRNA Phe without the T-arm, or the tRNA Phe U54C variant. Methylated tRNA Phe was prepared as follows. The transcript (50 g) was methylated by TrmA Modification of tRNA Transcript by TrmH, TrmB, TruB, and TrmI-T. thermophilus tRNA Phe transcript (25 g) was individually methylated by TrmH (5 g), TrmB (7.5 g), and TrmI (7 g) in the presence of 200 M AdoMet for 3 h at 55°C. In the case of ⌿55 modification, the transcript (25 g) was incubated with TruB (7.5 g) without AdoMet for 3 h at 55°C. The transcripts were recovered by phenol/chloroform treatment and ethanol precipitation. The purities of transcripts were checked by 10% PAGE (7 M urea).
Computer Programs for the Structural Figure-The structural figure (see Fig. 11) was generated by using PyMOL (DeLano Scientific, Palo Alto, CA). The Protein Data Bank code of T. thermophilus TrmFO is 3G5R. The docking model was constructed by the manual placement of the T-arm of yeast tRNA Phe onto the TrmFO structure with the program COOT (53). Fig. 1 shows the U54 modification in T. thermophilus tRNA Phe . TrmFO catalyzes the methyl transfer reaction. In our previous study (47), we reported an in vitro assay system for TrmFO activity. Briefly, the system is composed of two enzymatic reactions. The first reaction utilizes SHMT to produce [ 14 C]CH 2 THF, which is not commercially available. Then the [ 14 C]CH 2 THF is consumed by TrmFO in the same reaction mixture. In the reaction, an electron donor (NADPH or NADH) is required. The [ 14 C]methylated RNA is separated by 10% PAGE (7 M urea) and is detected by autoradiography. This system is relatively convenient and semiquantitative. However, there are two problems for the study of kinetics with this assay. The first problem is that the usable ranges of TrmFO and substrate concentrations are narrow. In simple terms, the rate-limiting factor of the methyl transfer reaction may be the concentration of [ 14 C]CH 2 THF that is produced by the SHMT reaction. To measure the kinetic parameters for electron donors and tRNA, excess amounts of methyl donor ([ 14 C]CH 2 THF) are required. The second problem is that the gel assay system is not suitable for the many samples required for a kinetic study. To overcome these problems, we have devised a new in vitro TrmFO assay system.

Improvement of in Vitro TrmFO Assay System-
To supply an excess amount of [ 14 C]CH 2 THF to the reaction mixture, a sufficient amount of THF needs to be added as the source of [ 14 C]CH 2 THF. However, excess amounts of THF inhibit the TrmFO reaction, because THF is an analog of [ 14 C]CH 2 THF. In order to determine the optimal THF concen- tration, we performed the methyl transfer assay with various concentrations of THF. Supplemental Fig. 1 (top) shows one example of these pilot experiments; in this case, Ͼ10 M THF inhibited the TrmFO (0.8 M) reaction. After these pilot experiments, we determined the practical concentrations of THF and TrmFO to be 5 and 0.25 M, respectively. Next, we compared the electron donors (supplemental Fig. 1 (bottom)). This experiment revealed that NADPH is superior to NADH as an electron donor, although TrmFO can use both NADPH and NADH. Furthermore, we confirmed that 3 mM NADPH is enough for the TrmFO reaction (supplemental Fig. 1 After the 10-min period, sufficient [ 14 C]serine remained in the reaction mixture. In contrast, after the 20-min period, half of the [ 14 C]serine was consumed, and the same amount of [ 14 C]glycine was produced. According to the progress of the TrmFO methyl transfer reaction, [ 14 C]CH 2 THF was consumed, and the equilibrium between [ 14 C]serine and [ 14 C]glycine shifted to the [ 14 C]glycine side. After 30 min, more than 90% of [ 14 C]serine was converted to [ 14 C]glycine. Thus, we concluded that the initial velocity of TrmFO methylation can be measured for 10 min under the optimized conditions. The second problem, involving the time and effort needed for analyzing multiple samples, was solved by employing a filter assay system instead of PAGE. We initially tested Whatman 3MM paper and washing with 5% trichloroacetic acid. However, this conventional method was not good because a considerable amount of [ 14 C]CH 2 THF remained on the filter. The best system was found to be the combination of Whatman DE81 paper and washing with 250 mM sodium phosphate buffer (data not shown). This filter assay system enabled us to use [ 3 H]serine instead of [ 14 C]serine. The detection limit of methylated RNA was dramatically improved because of the difference of specific activities between [ 3 H]serine (892 GBq/mmol) and [ 14 C]serine (1.85 GBq/mmol).
Thus, we successively improved the in vitro TrmFO assay system for kinetic studies. In fact, we could determine the apparent kinetic parameters for T. thermophilus tRNA Phe transcript ( Fig. 2B) and electron donors (Fig. 2C). To our surprise, both K m values for tRNA and electron donors were relatively large when compared with those of other tRNA methyltransferases and oxidoreductases. This may be caused by the existence of SHMT. In general, CH 2 THF is unstable and scarcely exists in a living cell. In thermophilic bacterial cells, CH 2 THF may be directly passed from SHMT to TrmFO. Although the formation of a complex between SHMT and TrmFO was not observed by the gel filtration column chromatography (data not shown), SHMT may have a weak affinity for TrmFO. This interaction hinders binding of tRNA and electron donors to TrmFO and is reflected in the relatively large K m values. Furthermore, during the course of this study, Hamdane et al. (49) proposed a very interesting hypothetical mechanism for formation of a TrmFO-tRNA complex based on the locations of catalytic and tRNA-binding cysteine residues. According to their theory, TrmFO monomer binds to tRNA, and then a second TrmFO binds to the complex. If correct, complicated reaction intermediates should be considered. Therefore, it should be mentioned that the kinetic parameters in this paper are apparent values. However, these values give clues to allow consideration of the interaction between TrmFO and tRNA.
Methyl Group Acceptance Activities of Truncated tRNA Molecules-To address the recognition sites in tRNA, we prepared nine truncated T. thermophilus tRNA Phe transcripts (Fig.  3). The methyl group acceptance activities of these truncated tRNA Phe transcripts using TrmFO were visualized using the previous gel assay system. The transcripts ( analyzed by 10% PAGE (7 M urea). Thus, the data in Fig. 3 do not represent the initial velocities. Furthermore, the initial velocities for microhelix RNAs (I and J; see Table 1) were slow. To visualize the methyl group acceptance activity of microhelix RNAs, the concentrations of TrmFO and RNA were increased to 0.4 and 12.0 M, respectively, and the imaging plate was exposed for longer durations. The kinetic parameters determined using [ 3 H]serine are summarized in Table 1. The gel was stained with MB to visualize the RNA molecules (Fig. 3, bottom left), and then [ 14 C]methyl group incorporation was monitored by autoradiography (Fig. 3, bottom right). As shown in Fig. 3, the methyl group acceptance activity of tRNA Phe full-length (wild type) was clearly observed. Similarly, truncated tRNA molecules showed methyl group acceptance activity (Fig. 3, B-G). In . Methyl group incorporation into truncated tRNA transcripts. A, cloverleaf structure of T. thermophilus tRNA Phe (wild type). We prepared nine truncated tRNA transcripts (B-I). The methyl transfer assay was performed, and RNA was recovered by phenol/chloroform treatment and ethanol precipitation. The RNAs were separated by 10% PAGE (7 M urea) and then stained with MB (left). [ 14 C]Methyl group incorporation was monitored by autoradiography (right). The kinetic parameters are given in Table 1.
a ND, methyl group incorporation was not detectable. b The parameters for I and J were not correctly measured due to low V max values.

tRNA Recognition Mechanism of TrmFO
contrast, truncated tRNA Phe , in which the T-arm was deleted, had completely lost methyl group acceptance activity (Fig. 3H). Thus, the minimum requirement of TrmFO seems to exist in the T-arm structure of tRNA. To confirm this idea, we prepared synthetic microhelix RNAs (Fig. 3, I and J). The 19-nt T-arm mimics the T-arm structure of E. coli tRNA Phe and was previously used for the crystallization of E. coli TrmA and RNA complex (42). The 22-nt T-arm has an artificial sequence to reinforce the stem structure. As shown in Fig. 3, I and J, both microhelix RNAs were clearly methylated, revealing that the minimum requirement(s) of TrmFO exists in the T-arm structure. Although the kinetic parameters for microhelix RNAs were not measured correctly due to the low methyl transfer activity, the tendency to small K m values was confirmed. The other part of tRNA (e.g. the D-arm in the L-shaped tRNA) may hinder the initial binding process. During the course of this study, it has been reported that B. subtilis TrmFO methylates a 31-mer mini-RNA, which includes the T-arm sequence of B. subtilis tRNA Asp (49). The results from our current study are in good agreement with these observations. Methyl Acceptance Activities of T-arm Mutant Transcripts-These results prompted us to investigate the effects of mutations in the T-arm, and we therefore prepared 17 mutant transcripts (Fig. 4A). Because U55, C56, and A58 bases are conserved in the T-arm of all bacterial tRNAs, these nucleotides were substituted by other nucleotides (Fig. 4, U55C, C56U, A58G, and A58U). As shown in Fig. 4A, the variants U55C and C56U completely lost methyl group acceptance activity, demonstrating that U55 and C56 are absolutely required for TrmFO recognition. The results also demonstrated that the methyl group acceptance activities of A58G and A58U variants were clearly decreased. Kinetic studies revealed that the K m values for the A58G and A58U variants were not significantly changed. Thus, the substitution of A58 does not have an effect on the initial binding process between TrmFO and tRNA and the releasing process of methylated tRNA from the complex. A58 forms a reverse Hoogsteen base pair with U54, the target uridine (13,14). With the introduction of U54 into the expected catalytic pocket of TrmFO (47), the U54-A58 tertiary base pair should be disrupted. Therefore, the substitution of A58 mainly influences the V max value. The other nucleotides (G57, U59, and U60) in the T-loop were individually substituted as shown in Fig. 4 (G57C, G57A, U59C, U60C, and U60G). These variants have methyl group acceptance activities comparable with that of the wild-type transcript ( Fig. 4 and Table 1), suggesting that G57, U59, and U60 are not recognized by TrmFO. It should be mentioned that four variants (G57C, U59C, U60C, and U60G) showed a high concentration inhibition; at high concentrations (Ͼ3 M), the methyl group acceptance activities of these variants were decreased (supplemental Fig. 2). This phenomenon is explainable by considering exchange of tRNA from the TrmFO-tRNA complex. These variants probably form a normal L-shaped tRNA structure because G57, U59, and U60 do not form a tertiary base pair. Therefore, these variants seem to bind to TrmFO like the wild-type tRNA in the initial binding process. After the initial binding process, tRNA (at least the T-arm) should change structure to introduce U54 into the catalytic pocket. These nucleotides (G57, U59, and U60) may have After the methyl transfer reaction, methyl group incorporation into these variants was measured using the gel assay. The gel was stained with MB (middle) and then subjected to autoradiography (right). The WT lane shows the positive control (the wild-type tRNA Phe transcript). B, the G53-C61 base pair was substituted with C53-G61 and A53-U61 base pairs or was disrupted by introduction of mutations. The variant "DT-stem" has a completely disrupted D-stem. Methyl group acceptance activities of these eight transcripts were investigated by the gel assay (middle and right). The kinetic parameters are given in Table 2.
an indirect effect on this structural change process (or prestructural change process). The mutations of these nucleotides may perturb the local structure of the T-arm, and this perturbation may enhance the exchange of tRNA from the TrmFO-tRNA complex at high tRNA concentrations. The kinetic parameters of these variants in Table 2 were calculated from the velocities at low tRNA concentrations (Ͻ2 M), although the kinetics are not approximated by the Michaelis-Menten equation.
Because the G53-C61 base pair in the T-stem is conserved in almost all tRNAs (1-3), we prepared eight variants in which the G53-C61 base pair was substituted by the other base pair or was disrupted (Fig. 4B). As shown in Fig. 4, these mutations produced complete loss of methyl group acceptance activity, showing that the G53-C61 base pair works as the essential positive determinant for TrmFO. Taking these results together, we conclude that the U54U55C56 sequence and the G53-C61 base pair are absolutely required for TrmFO recognition.
The Affinity of TrmFO for RNA in the Initial Binding Process Is Weak, and Positive Determinants in the T-arm Are Probably Required for the Structural Change Process-The apparent K m value for tRNA is considerably large, and several tRNA variants show high concentration inhibition, suggesting that the affinity of TrmFO for RNA in the initial binding is weak. To address this issue, we performed a gel mobility shift assay, which has been ND Disruption of T-stem (DT-stem) ND a ND, methyl group incorporation was not detectable. b Values in parentheses were calculated from data of low substrate concentrations.

tRNA Recognition Mechanism of TrmFO
previously used for binding assays of several modification enzymes (10,31,39). As shown in Fig. 5A, no shift band corresponding to the complex of TrmFO and tRNA Phe transcript was observed under the tested condition, although TrmFO can methylate this tRNA transcript. We tried the gel mobility shift assay under several conditions (e.g. in the presence or absence of SHMT, NADPH, and/or folate analogues). However, we did not detect a shift band derived from the complex formation under any of these conditions (data not shown). These results are in line with the idea that TrmFO has weak affinity for tRNA in the initial binding process. If TrmFO repeatedly binds and releases RNA in a short time during the initial binding, the weak affinity could be rationally explained. If this idea is correct, TrmFO would show affinity for non-substrate RNAs. To confirm this idea, we performed fluorescence quenching experiments (Fig. 5B). T. thermophilus TrmFO has six tryptophan residues. Fortunately, the addition of tRNA Phe transcript (circles in Fig. 5B) caused a decrease in fluorescence derived from tryptophan residues. This phenomenon suggests that one or multiple tryptophan residues move to a hydrophilic environment with the increase in tRNA Phe transcript. It should be mentioned that tRNA Phe transcript does not have absorption at 320 nm (data not shown). The 50% effective concentration of wild-type tRNA Phe was around 5 M, consistent with the idea that the affinity of TrmFO for RNA in the initial binding is weak. Next, we performed the same experi-ment with non-substrate RNA (triangles in Fig. 5B), in which the T-arm was deleted (Fig. 3H). The non-substrate RNA caused a decrease in fluorescence derived from tryptophan residues similar to the wild-type tRNA Phe transcript, demonstrating that TrmFO interacts with non-substrate RNA with the same affinity as for substrate RNA. To clarify whether or not the selection of substrate RNA takes place in the initial binding process, we performed an inhibition assay (Fig. 6). We prepared three types of non-substrate tRNA (methylated tRNA Phe , tRNA Phe without the T-arm, and the tRNA Phe U54C variant). As shown in Fig. 6A, the methylated tRNA did not strongly inhibit the methyl transfer reaction. This result suggests that the methyl group of m 5 U54 in the methylated tRNA causes steric hindrance with CH 2 THF in the catalytic pocket. Thus, the methylated tRNA seems to be released after the initial binding process. The weak affinity for tRNA in the initial binding process has an advantage for release of the methylated tRNA. Furthermore, because the methylated tRNA is more abundant as compared with the non-methylated precursor tRNA, this mechanism is rational for effective methylation in a living cell. The tRNA without the T-arm inhibited the methyl transfer reaction strongly as compared with methylated tRNA (Fig. 6B). This result suggests that tRNA without the T-arm is released during the structural change process. Thus, positive determinants in the T-arm mainly function during the structural change process. The tRNA Phe U54C variant strongly inhibited  4, 5, 6, and 7) have a disrupted three-dimensional core structure. In three variants (2, 6, and 7), U54 was replaced by C to investigate the methylation site. C, methyl group acceptance activities of tRNA Pro variants. The methyl group acceptance activities were monitored by the gel assay. The gel was stained with MB to visualize the RNAs, after which an autoradiogram of the gel was taken. DECEMBER 14, 2012 • VOLUME 287 • NUMBER 51 the methyl transfer reaction (Fig. 6C), suggesting that the U54C variant progresses the structural change process.

tRNA Recognition Mechanism of TrmFO
Taking these results together, we concluded that TrmFO does not exclude non-substrate RNA in the initial binding process and that affinity for RNA in the initial binding process is weak. Our results suggest that the positive determinants in the T-arm structure are probably required for progression of the structural change process.
TrmFO Can Distinguish the T-arm from the Anticodon Arm in T. thermophilus tRNA Pro -The above conclusion prompted us to ask one question, namely how does TrmFO distinguish the T-arm from the anticodon arm? In general, the nucleotide sequence of the T-arm is partially similar to that of the anticodon arm, although the T-arm and anticodon arm conformations are completely different (54). T. thermophilus tRNA Pro has a distinct anticodon arm, in which the essential recognition sequences of TrmFO can be found (Fig. 7A). If TrmFO simply recognizes the nucleotide sequences in the stem and loop structure, TrmFO should methylate both U54 and U32 in tRNA Pro . Fortunately, the RNA sequence of T. thermophilus tRNA Pro has been reported (55); U32, which corresponds to U54, is known to be unmodified (Fig. 7A). That is to say, this natural modification pattern suggests that TrmFO can distinguish the T-arm from the anticodon arm in T. thermophilus tRNA Pro .
Initially, we assumed that the three-dimensional core structure in tRNA Pro prevents the incorrect methylation of U32. Thus, we considered steric hindrance by the three-dimensional core. To investigate this idea, we prepared eight tRNA Pro variants (Fig. 7B). As shown in Fig. 7C, precursor tRNA Pro transcript was clearly methylated. When U54 in the precursor was replaced by C, the methyl group acceptance activity was completely lost (Fig. 7C, lane 2), showing that the methylation site is only U54. When the T-arm was deleted or both D-and T-arms were deleted, U32 was not methylated (Fig. 7C, lanes 3 and 4). Furthermore, when both G18-U55 and G19-C56 tertiary base pairs were disrupted, only U54 was methylated (Fig. 7C, lanes 5  and 6). Moreover, the substitution of the U54U55 sequence by a C54C55 sequence brought about the loss of methyl group acceptance activity, showing that the methylation of U32 had not occurred (Fig. 7C, lane 7). Although we could experimentally show that TrmFO is able to distinguish the T-arm from the anticodon arm, these results revealed that our initial idea was incorrect. Thus, the three-dimensional core of tRNA Pro does not work as the negative determinant for TrmFO.
The A38 in the Anticodon Arm Works as a Negative Determinant for TrmFO Recognition-Next, we considered whether a negative determinant might exist in the anticodon loop. To address this issue, we designed four tRNA Pro U54C variants in which the anticodon arm sequence of tRNA Pro U54C was substituted by the T-arm sequence (Fig. 8, A-D). To confirm whether the methylation site was U32, we additionally prepared four variants in which U32 was substituted by C (Fig. 8, E-H). As shown in Fig. 8, A-D, all tRNA Pro U54C variants were methylated, although their methyl group acceptance activities were considerably weaker than that of the wild-type tRNA Pro , which has the normal methylation site U54. When U32 in these tRNA Pro U54C variants was substituted by C, the methyl group acceptance activity was completely lost (Fig. 8, E-H). Thus, the methylation site is U32. These results demonstrated that the A38 in the anticodon loop of tRNA Pro works as the negative determinant for TrmFO.
Hypothetical Mechanism of Prevention of Incorrect U32 Methylation by A38-In the T-arm, the methylation site U54 forms a reverse Hoogsteen base pair with A58, and the U54-A58 tertiary base pair stacks with the conserved G53-C61 base pair (Fig. 9A). In contrast, U32 in the anticodon arm of the tRNA Pro transcript probably forms a Watson-Crick base pair with A38, and the U32-A38 base pair stacks between the G31-C39 and U33-G37 base pairs (Fig. 9B, left). The substitution of A38 by U causes a structural change of this anticodon arm and results in the emergence of a structural equilibrium between the T-arm-like and open loop structures (Fig. 9B, middle and  right). This raises the possibility of the anticodon arm being recognized by TrmFO. Thus, the A38 sequence in T. thermophilus tRNA Pro would prevent incorrect U32 methylation by To search for the negative determinant for TrmFO, the anticodon loop sequence was replaced stepwise by the T-loop sequence (A-D). The variants (A-D) contained C54 to abolish U54 methylation. We additionally prepared four tRNA Pro variants in which U32 was substituted by C to determine the methylation site (E-H). The methyl group acceptance activities of these variants were tested by the gel assay (lower panels). The gel was stained with MB (left) and then subject to autoradiography (right). WT, the positive control, in which the wild-type tRNA Pro transcript was used as the substrate.

tRNA Recognition Mechanism of TrmFO
TrmFO. To investigate this hypothetical mechanism, we prepared three tRNA Pro variants (Fig. 9C). In these variants, G36 and A38 were substituted by A and U, respectively. These substitutions form a U54-A36 tertiary base pair and disrupt the U32-A38 Watson-Crick base pair, and the structural equilibrium in Fig. 9B is shifted to the T-loop-like structure side. In agreement with the hypothesis, the methyl group acceptance activities of these variants were dramatically improved (Fig.  9C). Because these variants do not have U54, the methylation site is U32. These results support the hypothesis in Fig. 9B. The substitution of G36 by A increased the velocity of incorrect U32 methylation by TrmFO. U32 and mutated A36 probably form a reverse Hoogsteen base pair, as seen with U54-A58 in the T-arm. This result suggests that coexistence of U32 and A36 in a single tRNA is to be avoided in order to prevent of incorrect U32 methylation. The T. thermophilus genome encodes 47 tRNA genes, and nine of these have a U32 sequence. However, there is no tRNA that has both U32 and A36 (1-3). Thus, natural tRNA sequences are such that they avoid incorrect U32 methylation.
Effects of Other tRNA Modifications on TrmFO Methylation-In our previous publications (51,52), we reported the existence of a network formed between modified nucleotides and modification enzymes in T. thermophilus tRNA modification. Briefly, the existence of the m 7 G46 modification enhances the velocities of Gm18 and m 1 G37 formations, in vivo and in vitro (51). In contrast, the existence of the ⌿55 modification decreases the velocity of Gm18, s 2 U54, and m 1 A58 formations in vivo (52). In these studies, we showed that the deletion of the m 7 G46 or ⌿55 modification in native tRNA Phe does not affect the m 5 U54 content (51,52). Furthermore, we demonstrated that the m 5 U54 modification does not have any effect on the FIGURE 9. Hypothetical mechanism of prevention of incorrect U32 methylation by A38. A, typical T-arm structure. In the T-arm structure, U54 forms a reverse Hoogsteen base pair with A58. This U54-A58 base pair stacks with the G53-C61 base pair. B, structural change model of T. thermophilus tRNA Pro anticodon loop. In the T. thermophilus tRNA Pro transcript (wild type), U32 forms a Watson-Crick base pair with A38. This U32-A38 base pair stacks between the G31-C39 and U33-G37 base pairs. If A38 is replaced by U (A38U variant), the U32-A38 base pair is disrupted. This causes the emergence of an equilibrium between the T-arm-like structure and an open loop structure. The T-arm-like structure is recognized by TrmFO as substrate because this anticodon loop includes all positive determinants for TrmFO. Thus, A38 prevents incorrect U32 methylation by TrmFO, demonstrating that A38 works as a negative determinant for TrmFO. C, to reinforce the hypothetical mechanism, we prepared three tRNA Pro variants. Variant 1 has three point mutations; G36, A38, and U54 were replaced by A, U, and C, respectively. Variant 2 has three mutations; the T-arm was deleted, and G36 and A38 were replaced by A and U, respectively. Variant 3 has four mutations; both the D-and T-arm were deleted, and G36 and A38 were replaced by A and U, respectively. The methyl group acceptance activities of these variants were tested by the gel assay (right panels). The gel was stained with MB and subject to autoradiogram. WT, the positive control, in which the wild-type tRNA Pro transcript was used as the substrate.

tRNA Recognition Mechanism of TrmFO
velocities of m 7 G46 and ⌿55 formations in vitro (51,52). However, during these studies, the in vitro TrmFO assay system had not been optimized. Therefore, we have investigated whether the other tRNA modifications have an effect on the TrmFO activity. We prepared four tRNA modification enzymes, tRNA (Gm18) methyltransferase (TrmH) (56), tRNA (m 7 G46) methyltransferase (TrmB) (57), tRNA (⌿55) synthase (TruB) (58), and tRNA (m 1 A58) methyltransferase (TrmI) (59) (Fig. 10A). T. thermophilus tRNA Phe transcript was individually modified by these enzymes for 3 h at 55°C and then recovered by phenol/ chloroform treatment and ethanol precipitation. The contents of Gm18, m 7 G46, and m 1 A58 modification per tRNA molecule were determined by measuring the methylation of transcripts by the enzymes using [ 14 C]AdoMet and were calculated to be 0.86, 0.81, and 0.89, respectively. The content of ⌿55 modification per tRNA molecule was determined to be 0.9 -1.00 by HPLC nucleoside analysis (data not shown). As shown in Fig.  10B, Gm18 and m 7 G46 modifications did not have a significant effect on the velocity of m 5 U54 formation as catalyzed by TrmFO. In the previous studies (51,52), the m 7 G46 and ⌿55 modifications were seen not have an effect on in vivo m 5 U54 formation. In the case of the m 7 G46 modification, the current in vitro result is in good agreement with our previous in vivo result, whereas our in vitro experiment showed that the ⌿55 modification slightly accelerates the TrmFO reaction (Fig.  10B). In contrast, the m 1 A58 modification produced by TrmI clearly accelerated the initial velocity of TrmFO methylation; this effect was confirmed by the repeated experiments (Fig.  10B). This observed effect is probably caused by the reinforcement of the U54-A58 tertiary base pair. The m 1 A58 modification is a positive determinant for the 2-thiouridylation complex (TtuA, TtuB, TtuC, and the other protein factors), which catalyzes s 2 U54 formation (8,9). Therefore, our current results and other reports (8,9) together suggest that m 5 U54 methylation by TrmFO, m 1 A58 methylation by TrmI, and the s 2 U54 sulfur transfer reaction by the 2-thiouridylation complex have a synergistic effect on m 5 s 2 U54 formation.
The Docking Model of TrmFO and T-arm-In our previous site-directed mutagenesis study (47), we found that the Arg-97, Lys-282, and Lys-287 residues in TrmFO are important for the methyl transfer reaction. Based on the current and previous experimental results, we constructed the docking model of TrmFO and T-arm of yeast tRNA Phe (Fig. 11). The positive determinants (U54U55C56 sequence and G53-C61 base pair) are shown in Fig. 11A. These positive determinants are conserved in all T. thermophilus tRNAs. The positive determinants, U54, U55 (⌿55 in Fig. 11B), and C56 nucleotides, could be placed on the TrmFO surface. In contrast, the positive determinant, G53-C61 base pair was embedded in the T-arm structure. Because the variant, which has a G53-U61 base pair, was not methylated (Fig. 4), TrmFO can distinguish the G53-C61 and G53-U61 base pairs. In general, cytidine can be discriminated from uridine by the 4-amino group. Therefore, a G-C base pair can be discriminated from a G-U base pair by the major groove binding of TrmFO (Fig. 11C). However, we could not construct a docking model in which the G53-C61 base pair makes direct contact with the enzyme due to steric hindrance. The second important difference between G-C and G-U base pairs is the location of the pyrimidine base in the stem structure (Fig. 11C). This difference influences the ribose-phosphate backbone structure. The ribose and phosphate of C61 could be placed on the enzyme surface in the docking model (Fig. 11B). Therefore, the G53-C61 base pair recognition by TrmFO may be not mediated by the direct recognition of bases but by ribose-phosphate backbone recognition. As described above, positive determinants in the T-arm probably function in the structural change process. To clarify the precise mechanism of G53-C61 recognition, further study will be required.

DISCUSSION
In this study, we have focused on tRNA recognition by folate/ FAD-dependent tRNA methyltransferase, TrmFO. In general, the activity measurement of folate-dependent methyltransferase is not so easy, because [ 14 C]CH 2 THF is not commercially available. In the early study, [ 14 C]CH 2 THF was chemically synthesized from THF and [ 14 C]formaldehyde (44). Because this method is difficult for most biochemical researchers, TrmFO remained a relatively unstudied enzyme. Urbonavicius et al. (46,60) detected TrmFO activity by 32 P-internally labeled tRNA. Although this system is convenient, measurement of velocities is not so easy. Therefore, we initially altered our in vitro TrmFO assay system, which we had previously reported (47). We optimized the concentrations of enzymes and substrates and determined the practical ranges. In addition, a filter assay system was introduced instead of the gel assay system. These alterations enabled us to measure the kinetic parameters of TrmFO for electron donors and tRNA transcripts, as shown in Fig. 2. Our assay system may be applicable to the assay of other folate/FAD-dependent tRNA methyltransferases (e.g. the MnmE-MnmG (previous name, GidA) complex, which is involved in the first steps (nm 5 U34 and cmnm 5 U34 syntheses) of mnm 5 U34 formation) (for a recent review, see Ref. 61). The MnmE-MnmG complex changes the pathway by supply of sub-strates (ammonium ion and glycine) and is regulated by GTP (61). Therefore, the construction of an in vitro MnmE-MnmG complex assay system is likely to be more difficult than for TrmFO.
Our current kinetic study showed that NADPH is superior to NADH as an electron donor for TrmFO (Fig. 2). In contrast, it has been reported that MnmG (GidA) uses only NADH as an electron donor (62). Thus, although both folate/FAD-dependent tRNA methyltransferases (TrmFO and MnmG) have a similar FAD-binding domain (47,62,63), the binding modes of electron donors are completely different. To understand this difference, we compared the structures of TrmFO and MnmG (data not shown). Because the isoalloxazine ring of FAD in MnmG is embedded in the protein (62,63), the phosphate group of NADPH may cause steric hindrance to gain access to the FAD. In contrast, the isoalloxazine ring of FAD in TrmFO locates near the enzyme surface, and many positively charged amino acid residues exist around the FAD (47). These structural differences may cause electron donor specificities.
The minimum substrate of TrmFO was found to be the 19-nt T-arm-like microhelix RNA (Fig. 3). Likewise, the minimum substrate of TrmA (AdoMet-dependent tRNA (m 5 U54) methyltransferase) has been reported to be a 17-nt T-arm-like microhelix RNA (22,23). Although this local structure recognition is common between both tRNA (m 5 U54) methyltransferases, the mechanisms of recognition are considerably different. Because TrmA mainly recognizes the ribose-phosphate backbone structure of the T-arm, the positive determinants are only U54 and C56 (23). The crystal structure of the TrmA and tRNA Recognition Mechanism of TrmFO DECEMBER 14, 2012 • VOLUME 287 • NUMBER 51 microhelix RNA complex revealed that TrmA mainly contacts the U54 base and ribose-phosphate backbone of the T-arm (42). In contrast, TrmFO recognizes the conserved nucleotide sequences (U54U55C56 and G53-C61 base pair) in the T-arm. The positive determinants for TrmFO are rather similar to those for tRNA (⌿55) synthase (TruB) (28) and the 2-thiouridylation complex for s 2 U54 (7). TruB recognizes the U54U55 sequence and A58 (28), whereas the 2-thiouridylation complex recognizes the U54U55C56 sequence and A58 (7). In comparison, TrmFO has an absolute requirement for the recognition of the G53-C61 base pair while recognizing A58 (U54-A58 tertiary base pair) weakly.
During the course of study, Hamdane et al. (49) reported that B. subtilis TrmFO forms a covalent complex with tRNA via Cys-226. According to their report (49), we also attempted to detect the T. thermophilus TrmFO-tRNA complex; however, we unable to observe the formation of the covalent complex by electrophoresis, in which the gel was stained with methylene blue and Coomassie Brilliant Blue (data not shown). The stability of the covalent complex would therefore seem to be quite different between B. subtilis and T. thermophilus TrmFO enzymes. The striking similarity of the amino acid sequences of both enzymes suggests that these TrmFO enzymes have the same catalytic mechanism (46). In previous work, we proposed the other cysteine (Cys-51 in T. thermophilus TrmFO, which corresponds to Cys-53 in B. subtilis TrmFO) to be the site of covalent bond formation based on the crystal structure (47). In contrast, Hamdane et al. experimentally verified the covalent bond formation site as Cys-226 in B. subtilis TrmFO (49). Therefore, the covalent bond formation site in our previous hypothetical mechanism should be modified to fit in with the proposal of Hamdane et al. (49). If Cys-226 in B. subtilis TrmFO is the covalent bond formation site, TrmFO may form a dimer structure during the enzymatic reaction (49). To clarify this issue, further study will be necessary.
The formation of a covalent intermediate between substrate RNA and enzyme was also observed in the TrmA reaction; TrmA forms a covalent bond complex not only with tRNA but also with rRNA (24,64). The formation of a complex between E. coli TrmA and 16S rRNA is essential for cell viability (64). It has been reported that B. subtilis trmFO gene disruption strains can survive (46). However, the enzyme (at least B. subtilis TrmFO) has the potential to form a covalent bond with other RNA species as well as tRNA because the RNA recognition mechanism of TrmFO is relatively simple, as described in this paper.
In previous works (51,52), we reported the existence of a network between modified nucleotides and modification enzymes in T. thermophilus tRNA modification. In the current study, we confirmed that the U54 modification catalyzed by TrmFO is not significantly influenced by the presence of the Gm18 and m 7 G46 modifications, whereas the ⌿55 modification may slightly contribute to m 5 U54 formation. In contrast, the m 1 A58 modification has a clear positive effect on TrmFO activity. In the previous study, we found that 10 -30% of U54 in tRNA Phe from cells cultured at 55°C is unmodified U (or s 2 U) (52). In contrast, U54 in tRNA Phe from cells cultured at 70°C is nearly fully modified to m 5 U54 (or m 5 s 2 U54) (52). Therefore, in vivo TrmFO activity seems to be mainly regulated by culture temperatures. At high temperatures (Ͼ70°C), the m 5 U54 methylation catalyzed by TrmFO, the m 1 A58 methylation catalyzed by TrmI, and the s 2 U54 sulfur transfer reaction catalyzed by the 2-thiouridylation complex seem to have a synergistic effect on m 5 s 2 U54 formation. To clarify the role of the m 5 U54 modification on the other modifications, extensive analysis of a T. thermophilus trmFO gene disruption strain is necessary.