Yeast Mitochondrial Initiator tRNA Is Methylated at Guanosine 37 by the Trm5-encoded tRNA (Guanine-N1-)-methyltransferase*

The TRM5 gene encodes a tRNA (guanine-N1-)-methyltransferase (Trm5p) that methylates guanosine at position 37 (m1G37) in cytoplasmic tRNAs in Saccharomyces cerevisiae. Here we show that Trm5p is also responsible for m1G37 methylation of mitochondrial tRNAs. The TRM5 open reading frame encodes 499 amino acids containing four potential initiator codons within the first 48 codons. Full-length Trm5p, purified as a fusion protein with maltose-binding protein, exhibited robust methyltransferase activity with tRNA isolated from a Δtrm5 mutant strain, as well as with a synthetic mitochondrial initiator tRNA (tRNAMetf). Primer extension demonstrated that the site of methylation was guanosine 37 in both mitochondrial tRNAMetf and tRNAPhe. High pressure liquid chromatography analysis showed the methylated product to be m1G. Subcellular fractionation and immunoblotting of a strain expressing a green fluorescent protein-tagged version of the TRM5 gene revealed that the enzyme was localized to both cytoplasm and mitochondria. The slightly larger mitochondrial form was protected from protease digestion, indicating a matrix localization. Analysis of N-terminal truncation mutants revealed that a Trm5p active in the cytoplasm could be obtained with a construct lacking amino acids 1–33 (Δ1–33), whereas production of a Trm5p active in the mitochondria required these first 33 amino acids. Yeast expressing the Δ1–33 construct exhibited a significantly lower rate of oxygen consumption, indicating that efficiency or accuracy of mitochondrial protein synthesis is decreased in cells lacking m1G37 methylation of mitochondrial tRNAs. These data suggest that this tRNA modification plays an important role in reading frame maintenance in mitochondrial protein synthesis.

One of the most ancient tRNA modifications, present in all organisms as well as mitochondria and chloroplasts, is methy-lation of the N1 atom of guanosine at position 37 (m 1 G37) 3 (1). The m 1 G37 modification is catalyzed by a tRNA (guanine-N1-)methyltransferase (EC 2.1.1.31) encoded by trmD in bacteria or TRM5 in archaea and eukaryotes (1)(2)(3). Remarkably, the bacterial trmD gene is not homologous to TRM5; thus, the m 1 G37modifying enzyme evolved twice. Trm5p has been shown to be responsible for m 1 G37 methylation of at least eight cytoplasmic tRNAs in Saccharomyces cerevisiae (1,4). trm5 mutants that lack this modification exhibit a severe growth defect (1), consistent with the important role of m 1 G37 methylation in reading frame maintenance (5).
S. cerevisiae also has at least eight mitochondrially encoded tRNAs that carry the m 1 G37 modification (6), including the initiator tRNA (tRNA Met f ) (7) and tRNA Phe (8). The enzyme(s) responsible for modifying these mitochondrial tRNAs has not been identified. There is no apparent homolog of bacterial trmD in eukaryotic or mitochondrial genomes, so it has been proposed (1,9) that Trm5p might also be responsible for methylation of mitochondrial tRNAs. To date, however, only cytoplasmic and nuclear localization of Trm5p has been reported (10). Therefore, it is possible that yeast encode a separate mitochondrial m 1 G37 methyltransferase enzyme.
To address this question, we have cloned, purified, and characterized the S. cerevisiae TRM5-encoded protein. We show that it possesses tRNA methyltransferase activity on both natural and synthetic mitochondrial tRNA substrates and is specific for methylation of the N1 atom of guanosine at position 37. Furthermore, we show that Trm5p is localized to both the cytoplasm and mitochondria in yeast. Thus, the protein encoded by the TRM5 gene in yeast represents yet another example of dual protein localization from a single gene (11,12).
The Escherichia coli expression vectors pMALc2H 10 T (13) and pRK793 (14) were generous gifts from Dr. John Tesmer (University of Michigan). Recombinant plasmids with inserts were sequenced at the DNA Core Facilities of the Institute for Cellular and Molecular Biology at The University of Texas at Austin. E. coli Rosetta(DE3) (Novagen) and E. coli JM109 (Promega) cells were grown aerobically in Luria-Bertani medium at 37°C.
Oxygen Consumption Assays-Oxygen consumption rates were measured using a standard Clark-type electrode (model 5300A; YSI Inc.). 5-ml cultures were grown in YMD to mid-log phase (A 600 ϭ ϳ0.5) and transferred to the airtight chamber, and oxygen consumption was measured at 30°C for at least 10 min. Rates (nmol oxygen min Ϫ1 ) were normalized to the A 600 of each culture.
Preparation of Yeast Mitochondria-The cells were grown aerobically in a semisynthetic galactose medium containing (per liter): 3 g of yeast extract, 10 g of galactose, 0.8 g of (NH 4 ) 2 SO 4 , 0.7 g of MgSO 4 ⅐7H 2 O, 0.5 g of NaCl, 1 g of KH 2 PO 4 , 0.4 g of CaCl 2 , and 5 mg of FeCl 3 ⅐6H 2 O. Mitochondria were isolated by spheroplasting and differential centrifugation according to Daum et al. (17).
Cloning Yeast TRM5 for Protein Expression in E. coli-The wild-type TRM5 (YHR070w) ORF was amplified from genomic DAY4 DNA using primers TRM5-F and TRM5-R, containing XbaI and HindIII restriction sites, respectively (supplemental Table S1). The resulting PCR product was digested with XbaI and HindIII, ligated into the same sites in the vector pMALc2H 10 T, and transformed into E. coli JM109. The recombinant plasmid was transformed into E. coli Rosetta(DE3) cells to express the maltose-binding protein (MBP) and decahistidine tag fused to the full-length Trm5 protein (MBP-Trm5). Heterologous protein was expressed in these cells after induction with 0.5 mM isopropyl-␤-D-thiogalactopyranoside at 29°C for 3 h. The cells were harvested and washed.
Purification of MBP-Trm5-Cells from a 2-liter culture of E. coli expressing MBP-Trm5 were lysed by sonication in 15 ml of binding buffer to which benzamidine (1 mM) and phenylmethanesulfonyl fluoride (1 mM) were added. After low speed centrifugation (20 min at 16,000 rpm in an SS34 rotor, Sorvall centrifuge), the supernatant was applied to a His-bind matrix (5-ml bed volume; Novagen) equilibrated in 1ϫ binding buffer, which contains 5 mM imidazole in 20 mM Tris-HCl buffer (pH 7.8) with 500 mM NaCl. After extensive washing, loosely adsorbed protein was eluted with "wash buffer" (60 mM imidazole, 500 mM NaCl in Tris buffer), and then 250 mM imidazole in the same Tris-HCl/NaCl solution was applied to elute tightly bound protein. The protein fractions eluted with 250 mM imidazole were dialyzed against 500 ml of 20 mM Tris-HCl (pH 7.5) for about 4 h, then dithiothreitol was added to give 1 mM, and the protein was frozen in small aliquots at Ϫ80°C. Analysis by SDS-PAGE (18) showed nearly pure protein of the expected size. The maltose-binding protein alone was expressed from pMALc2H 10 T and purified as described above to serve as a control in the methyltransferase assays. Tobacco etch virus protease was purified as described (19).
tRNA Preparations-Mitochondrial tRNA was isolated from mitochondria prepared from S. cerevisiae DAY4 cells. RNA was extracted using 2 ml of TRI reagent (Molecular Research Center) for mitochondria from 4 liters of yeast cells. The manufacturer's instructions were followed up to the phase separation step, and then a high salt precipitation was added to eliminate ribosomal RNA. Isopropanol (0.5 ml) was added to the aqueous phase followed by 0.5 ml of 0.8 M sodium citrate, 1.2 M NaCl. After 10 min at room temperature, the sample was centrifuged for 10 min at 12,000 ϫ g at 4°C. Isopropanol (0.7 volumes) was added to the supernatant to precipitate the tRNA. After 10 min at room temperature, tRNA was collected by centrifugation (15 min at 12,000 ϫ g at 4°C). The tRNA pellet was washed with 70% cold ethanol and then resuspended in 100 l of 10 mM ammonium acetate (pH 5). About 50 g of yeast mitochondrial tRNA was obtained. Low molecular weight RNA containing both cytoplasmic and mitochondrial tRNA was isolated from haploid ⌬trm5 cells (strain DLY1). Haploid spores were first grown on a YPD agar plate for 9 days at 30°C; then the cells were grown in YPD broth in the presence of 300 g ml Ϫ1 geneticin for at least 48 h followed by growth in YPD without geneticin for 48 h. Cells from a 200-ml culture were harvested and extracted with TRI reagent (8 ml of TRI reagent was used for 4 -5 ml of packed cell pellet). The mixture was kept at 37°C for 5 min before chloroform was added. Then the manufacturer's procedure was followed with no high salt precipitation step. This procedure gave mainly tRNA (analyzed by polyacrylamide-urea gel electrophoresis) with a yield of 2.3 mg. For further purification, a second ethanol precipitation was carried out. The tRNA was resuspended in sterile 10 mM ammonium acetate and stored frozen in aliquots. The same method was used to isolate total tRNA from DLY1 harboring various TRM5 plasmids.
Methyltransferase Assay-Methyl group incorporation from [ 3 H]methyl-S-adenosyl methionine ([ 3 H]AdoMet) into tRNA was measured by determining radioactivity in acid-precipitable product (20). The assay was carried out in a total volume of 30 l containing 100 mM Tris-HCl (pH 8), 5 mM MgCl 2 , 100 mM KCl, 1 mM dithiothreitol, 0.01% Nonidet P-40, 600 -700 pmol of ⌬trm5 tRNA, or 20 -30 pmol synthetic tRNA Met f unless otherwise stated, various amounts of enzyme, and 52 M [ 3 H]AdoMet, (300 -500 Ci mol Ϫ1 ). The synthetic tRNA Met f was heated for 5 min at 70°C and cooled at room temperature before being added to the reaction mixture. Incubation was at 37°C for the indicated time, usually 30 min. Then 25 l of the reaction mixture were added to 40 g of bovine serum albumin in a glass tube. Protein was precipitated with 2 ml of cold 10% trichloroacetic acid. The tube was kept in ice for 10 min before the precipitate was collected on a 34 glass fiber filter (Schleicher & Schuell) using a Millipore manifold. The filters were washed three times with 3 ml of 10% trichloroacetic acid and dried for 15 min at 150°C, and the bound radioactivity was determined by liquid scintillation counting. A reaction mixture without tRNA was included as negative control.
Synthesis of 1-Methylguanosine-Methyl iodide was used to methylate guanosine, producing m 1 G as described by Broom et al. (21). The structure of the synthetic product was confirmed using liquid chromatography-electrospray ionization-tandem mass spectrometry at the Mass Spectrometry 2) showed absorbance maxima at 256 and 270 nm, consistent with the previously reported spectrum (22).
HPLC Analysis of Nucleosides-The reversed phase HPLC protocol used to separate nucleosides was modified from one described by Pomerantz and McCloskey (23). Buffer A contained 250 mM ammonium acetate (pH 6.0), and buffer B contained 60% (v/v) acetonitrile in water. Samples (20 l) were applied to a reversed phase C 18 column (4.6 ϫ 250 mm, 5 m; Axxium) with a Security Guard ODS cartridge (4 ϫ 3 mm; Phenomenex) that was equilibrated in mobile phase containing 8% buffer B at a flow rate of 1 ml min Ϫ1 at 35°C. After 10 min, a gradient from 8 to 15% buffer B was applied. UV absorbance was monitored using a System Gold 168 photodiode array detector (Beckman). The data were collected, and chromatograms were integrated using the 32 Karat software (Beckman). For the analysis of radioisotopes, fractions (0.5 ml) were collected and diluted with 3 ml of ScintiVerse scintillation fluid (Fisher). Radioactivity measurements were made using a LS 6000SC liquid scintillation counter (Beckman). Using this HPLC method, nucleosides eluted with the following retention times: cytidine (3.4 min), uridine (3.5 min), 5-methylcytidine (4.7 min), 7-methylguanosine (4.7 min), guanosine (4.7 min), 1-methylguanosine (6.7 min), 2-methylguanosine (7.4 min), and adenosine (8.9 min). Incompletely hydrolyzed nucleotides and oligonucleotides eluted before 3 min.
Subcellular Fractionation and Immunoblotting-Yeast mitochondria were isolated from strain EY0986/TRM5-GFP. The post-mitochondrial supernatant was used as the cytoplasmic fraction. The samples were analyzed by SDS-PAGE on 10% gels, and immunoblotting was performed as described (24) using anti-GFP primary antibodies (1:500 dilution) or anti-Hsp60 antibodies (1:50,000 dilution). For proteinase K treatment, 50-g aliquots of mitochondria were resuspended in 100 l of isotonic buffer (0.6 M sorbitol, 20 mM HEPES-KOH, pH 7.4). Five l of proteinase K (1 mg ml Ϫ1 ) was added to samples and incubated for 15 min on ice. In some digestions, the membranes were solubilized by addition of Triton X-100 (final concentration, 0.5%) during the 10-min proteinase K treatment. Digestion was stopped by addition of 1 l of 100 mM phenylmethylsulfonyl fluoride in ethanol followed by a further 10-min incubation on ice. Mitochondria were collected by centrifugation at 12,000 ϫ g for 15 min, the pellets were resuspended in SDS sample buffer, and protein was analyzed by immunoblotting.
TRM5 Transcript Mapping-The 5Ј end of the TRM5 transcript was mapped by a primer extension-based method (RNA ligase-mediated rapid amplification of cDNA ends (RACE)) using the FirstChoice RNA ligase-mediated RACE kit from Ambion. Total RNA was isolated from yeast strain DAY4 using the RNeasy Mini Kit (Qiagen). DNA contamination was eliminated using a Turbo DNA-free kit (Ambion). Nested antisense primers specific to the TRM5 transcript were designed for use with the two nested 5Ј-RACE primers provided in the kit. The TRM5-specific inner primer (5R_AGA_IP; supplemental Table S1) was complementary to nucleotides ϩ183 to ϩ161. The TRM5-specific outer primer (5R_AGA_OP; supplemental Table S1) was complementary to nucleotides ϩ283 to ϩ264. The 5Ј-RACE inner primer and the TRM5specific inner primer had BamHI and HindIII sites, respectively, at their 5Ј ends to facilitate cloning. PCR fragments generated in the "inner" PCRs were cloned and sequenced as described previously (24).
Construction and Expression of TRM5 Truncation Mutants in Yeast-The complete sequence of the yeast TRM5 gene including 300 base pairs 5Ј of the first AUG codon was amplified from yeast genomic DNA (strain DAY4) by PCR using KOD Hot Start DNA polymerase and primers TRM5_FOR and TRM5_REV containing HindIII and BamHI restriction sites, respectively (supplemental Table S1). The resulting PCR product was gel-purified, digested, and ligated into the yeast low copy vectors, URA3-containing pRS416 and LEU2-containing pRS415 (25), to produce the pRS416-TRM5 and pRS415-TRM5 wild-type constructs (Table 1). SLIC (sequence and ligation-independent cloning) (26) was used to generate the construct with codons 1-19 deleted, using T7 forward primer versus T5_1/20_2 and T5_1/20_3 versus T3 reverse primer (supplemental Table S1). The two PCR products and the linearized pRS415 vector (digested with HindIII and BamHI restriction enzymes) were gel-purified and treated with 0.5 unit of T4 DNA polymerase. The two PCR products and the vector were annealed together at a 2:2:1 molar ratio with 20 ng of RecA protein at room temperature for 30 min, and the recombined plasmid (pRS415-⌬1-19TRM5; Table 1) was recovered from transformed E. coli cells. Splice overlap extension PCR (27) was used to generate the other truncations. For the construct with codons 1-33 deleted (pRS415-⌬1-33TRM5; Table 1), the TRM5 gene containing 300 bp of 5Ј-UTR in pRS416-TRM5 was amplified with primers T7 forward primer (upstream vector primer) and T5_1/34_2 (supplemental Table S1). The DNA including the region between codon 34 and the stop codon of TRM5 was PCR-amplified using primers T5_1/34_3 and T3 reverse (downstream vector primer). Primers T5_1/34_3 and T5_1/34_2 contain 18 bp of complementarity at their 5Ј ends for the second splice overlap extension PCR. The resulting PCR products from two separate reactions were purified by a PCR purification kit (Qiagen) and combined together for second PCR with T7 forward primer and T3 reverse primer. This product was gel-purified, digested with HindIII and BamHI, and cloned into pRS415 (25). The construct with codons 1-47 deleted was constructed similarly using primers T7 forward versus T5_1/48_2 and T5_1/48_3 versus T3 reverse for the first PCR. Primers T5_1/48_3 and T5_1/48_2 contain 18 bp of complementarity at their 5Ј ends for the second splice overlap extension PCR. The procedure described above was used for the second PCR, followed by cloning into pRS415. These truncation constructs were transformed into DLY1 harboring pRS416-TRM5 for subsequent analysis.

Expression of Trm5p as a Fusion
Protein-Attempts to express S. cerevisiae Trm5p in E. coli fused to N-or C-terminal polyhistidine tags produced mostly insoluble protein (data not shown). Therefore the full-length TRM5 ORF (499 amino acids) was cloned into the vector pMALc2H 10 T to express N-terminal maltose-binding protein, decahistidine tag, and tobacco etch virus protease cleavage site fused to Trm5p (MBP-Trm5). E. coli cells expressed this fusion protein at high levels in soluble form. MBP-Trm5 was purified by nickel affinity chromatography, resulting in nearly homogeneous preparation of fusion protein that had an apparent mass of 97 kDa, similar to its calculated mass of 101.4 kDa (data not shown).
Methyltransferase Activity of the MBP-Trm5 Fusion Protein and Its Cleavage Product-The fusion protein could be partially cleaved with tobacco etch virus protease to generate Trm5p and MBP polypeptides. The Trm5p fragment migrated on an SDS-PAGE gel with an apparent mass of 61 kDa, close to its predicted mass of 57.2 kDa, whereas the MBP fragment had an apparent mass of 42 kDa, similar to its predicted mass of 44.2 kDa. The MBP-Trm5 protein had substantial methyltransferase activity in assays using ⌬trm5 tRNA as substrate (supplemental Table S2). However, preincubation with buffer alone or tobacco etch virus protease reduced the activity considerably; this appears to be due to the inherent instability of the enzyme. Thus, all subsequent experiments described here were carried out with the uncleaved fusion protein (MBP-Trm5). As a control experiment, maltose-binding protein alone was purified from empty pMALc2H 10 T vector. The MBP protein had no methyltransferase activity when incubated with ⌬trm5 tRNA and AdoMet (data not shown).
A linear relationship was observed between the amount of substrate (⌬trm5 tRNA) and the incorporation of methyl groups when a saturating amount of enzyme and 50 M [ 3 H]AdoMet were used in the assay (Fig. 1A). From initial rate conditions with a small amount of enzyme and saturating concentrations of the substrates, a specific enzymatic activity of 30 Ϯ 1.3 nmol of methyl incorporated per min per mg fusion protein was calculated. MBP-Trm5 did not catalyze the methylation of wild-type S. cerevisiae total tRNA (data not shown), presumably because it was already fully methylated.
In Vitro Synthesis of Yeast Mitochondrial Initiator tRNA and Methylation by MBP-Trm5-To produce a synthetic transcript of yeast mitochondrial tRNA Met f with its native 5Ј uridylate nucleotide, a self-cleaving hammerhead ribozyme construct was produced using eight overlapping oligonucleotides (28, 29) (supplemental materials). T7 RNA polymerase efficiently transcribed the template, and the ribozyme self-cleaved to produce 25 g of tRNA from 200 -300 ng of template. After denaturation at 70°C for 5 min, followed by slow cooling to room temperature, the synthetic yeast mitochondrial tRNA Met f was a good substrate for methylation by MBP-Trm5. In reactions containing high levels of AdoMet and enzyme, up to 50% of the tRNA was methylated, as shown in Fig. 1B. The extent of this reaction is comparable with that reported for a homologous archaeal enzyme methylating a synthetic transcript (3). With-out the tRNA refolding procedure, methyl group incorporation was very poor into the synthetic yeast mitochondrial tRNA Met f .

Identification of the Trm5 tRNA Methylation Site and
Product-Two complementary approaches were taken to demonstrate that the MBP-Trm5 protein catalyzed the methylation of guanosine at tRNA position 37, on N1 of the guanine base: primer extension and chromatographic analysis of yeast mitochondrial tRNA Met f nucleosides to identify m 1 G. Because the m 1 G modification interferes with Watson-Crick base pairing, reverse transcriptase is unable to incorporate a complementary cytidylate, blocking further primer extension (30). Primer extension experiments were carried out as described under "Experimental Procedures." Two different primers specific for yeast mitochondrial tRNAs were used: one for the initiator tRNA Met f (supplemental Fig. S1) and another for tRNA Phe , which also in its native form has a m 1 G37 following A36 (8). Total tRNA isolated from wild-type or ⌬trm5 cells was annealed to 32 P-labeled primer, then the reverse transcriptase reaction was carried out, and the samples were analyzed by denaturing gel electrophoresis (Fig. 2, lanes 1-3 (tRNA Met primer) and lanes 4 -6 (tRNA Phe primer)). With the tRNA Met f primer, the bands are weaker than with the tRNA Phe primer; this may reflect lower abundance of initiator tRNA Met f in the ⌬trm5 tRNA sample. Reverse transcriptase extends the annealed primers up to m 1 G37 in the wild-type tRNAs, terminating opposite the A at position 38 (Fig. 2, lanes 1 and 4). In contrast, full-length transcripts are present in reactions containing undermodified ⌬trm5 tRNA (Fig. 2, lanes 3 and 6). When ⌬trm5 tRNA was first methylated in vitro by MBP-Trm5, a significant portion of the transcripts terminated at the same position observed in reactions with wild-type tRNA (Fig. 2,  lanes 2 and 5). A similar result was observed comparing modified and unmodified synthetic tRNA Met f transcripts (supplemental Fig. S2). The samples methylated by MBP-Trm5 show a mixture of full-length and m 1 G37-terminated transcripts, consistent with the proportion of modified tRNAs shown in Fig. 1B. This has been observed with other synthetic tRNAs (31).
To map the termination site of reverse transcription after methylation with MBP-Trm5, the primer extension reaction was modified to include dideoxynucleoside triphosphate terminators (supplemental materials). The tRNA Met f primer was annealed to ⌬trm5 tRNA, and then the reverse transcriptase reaction was carried out in the absence or presence of the four different ddNTPs. The termination product observed with wild-type yeast tRNA corresponds to dideoxy T incorporation complementary to A38 (supplemental Fig. S3). As expected, the  reverse transcriptase is unable to incorporate a nucleotide complementary to m 1 G37, so primer extension terminates.
To confirm that the product of Trm5 methylation is m 1 G, synthetic tRNA Met f was incubated with MBP-Trm5 and [ 3 H]AdoMet in a standard reaction. The radiolabeled product was enzymatically hydrolyzed (supplemental materials), and the constituent nucleosides were separated by reversed phase HPLC. A UV-absorbing compound in the hydrolysate co-elutes with a synthetic m 1 G standard (Fig. 3) and has a similar absorbance spectrum (data not shown). Fractions of the eluate were monitored by liquid scintillation counting, and the majority of radioactivity corresponded to the m 1 G peak. The radioactive species eluting early in the chromatogram could be due to m 1 G nucleotides that were incompletely hydrolyzed or to AdoMet degradation products that co-precipitated with tRNA from the reaction mixture.
Hydrolysates of total tRNA isolated from the haploid ⌬trm5 yeast contained significant levels of m 1 G (data not shown). This m 1 G is likely due to the activity of the m 1 G (9) tRNA methyltransferase encoded by the TRM10 gene (30).
Mitochondrial Localization of Trm5p-These results confirm that the Trm5p methyltransferase is capable of methylating the N1 of guanosine at position 37 in mitochondrial tRNAs (tRNA Met f and tRNA Phe ), at least in vitro. These mitochondrial tRNAs, encoded and synthesized in the mitochondrion, are known to be modified in vivo. If Trm5p is responsible for this modification in vivo, the enzyme must also exist in mitochondria. To address this possibility, S. cerevisiae strain EY0986/ TRM5-GFP was used (Table 1). This strain carries a GFPtagged TRM5 gene integrated at its normal chromosomal locus, under control of the endogenous TRM5 promoter. EY0986/ TRM5-GFP expresses Trm5p with GFP fused to its C terminus. The cells were grown in a semisynthetic galactose medium, fractionated by differential centrifugation into cytoplasmic and mitochondrial fractions, and analyzed by immunoblotting with antibodies against GFP. Fig. 4A shows that Trm5-GFP is detected in both the cytoplasmic and mitochondrial fractions, with the cytoplasmic protein migrating slightly faster than the mitochondrial protein. Antibodies against the mitochondrial matrix marker Hsp60 confirm the lack of contamination between the two fractions (Fig. 4A). Mitochondrial Trm5-GFP was protected from digestion by proteinase K, unless the membranes were solubilized by Triton X-100, similar to the matrix marker Hsp60 (Fig. 4B). Taken together, these data indicate that the single TRM5 gene encodes a protein that localizes both to the cytoplasm and mitochondrial compartments, with the mitochondrial form residing in the matrix.
Analysis of TRM5 Truncation Mutants-There are several mechanisms known that can generate dual localization of a protein encoded by a single gene (11,12). In most cases, alternative translational start site use results in alternative N-terminal sequences of the proteins, which in turn targets the proteins to different compartment (e.g. cytoplasm versus mitochondria). Inspection of the 5Ј end of the proposed TRM5 ORF reveals four in-frame AUG codons (positions 1, 20, 34, and 48) within the first 48 codons (Fig. 5). To try to identify the translational start site(s) used in vivo, we constructed three N-terminal truncation mutants of Trm5p, in which amino acids 1-19, 1-33, or 1-47 were deleted. The 5Ј-UTR up to position Ϫ1 adjacent to the first AUG codon remained intact in each construct. Each construct was expressed from the single-copy vector pRS415 (LEU2 vector), driven by the TRM5 promoter. These pRS415 constructs were transformed into the ⌬trm5 haploid mutant strain (DLY1) harboring a wild-type copy of TRM5 on pRS416 (URA3 vector). Leuϩ Uraϩ transformants were then streaked onto yeast minimal glucose (YMD) or ethanolϩglycerol (YPEG) plates containing 5-fluoroorotic acid to evict the wildtype TRM5 gene on the pRS416 plasmid. Fig. 6A shows that the ⌬1-19 and ⌬1-33 constructs support growth on glucose as well as the full-length TRM5 construct, whereas cells expressing the ⌬1-47 construct failed to grow. Likewise, all the constructs support growth on the nonfermentable ethanol ϩ glycerol plate except the ⌬1-47 construct (Fig. 6A). Thus, the first 33 amino acids of the proposed TRM5 ORF are not necessary for the expression of functional Trm5p, because trm5 null mutants exhibit a severe growth defect on glucose (1). The simplest interpretation of these data is that translation initiates at the third AUG codon (codon 34) to produce a Trm5p that functions in the cytoplasm to methylate cytoplasmic tRNAs.
To try to determine the translation start site of the mitochondrial form of Trm5p, we examined the methylation status of mitochondrial tRNAs in DLY1 cells expressing the ⌬1-19 and ⌬1-33 truncation constructs. The termination product indicative of methylation at position 37 was undetectable in mitochondrial tRNA Met f and tRNA Phe isolated from the two trun-cation mutants (Fig. 7). Overexposure of the gel revealed a faint signal in the tRNA Met f isolated from the ⌬1-19 mutant (lane 3), but this could not be replicated in subsequent experiments. These data suggest that initiation at either the first or second AUG codon can produce a functional mitochondrial form of Trm5p, although use of the second AUG (codon 20) is much less efficient. Locations and size (kDa) of molecular mass markers are shown on the right. Full-length Trm5p (499 amino acids) fused to GFP (238 amino acids) would have a molecular mass of ϳ83.3 kDa. A duplicate gel was immunoblotted with antibodies against mitochondrial marker Hsp60 as a control. B, mitochondria (50 g protein) were incubated with or without proteinase K (PK, 1 mg ml Ϫ1 ) in the presence or absence of 0.5% Triton X-100 and immunoblotted with antibodies against GFP. A duplicate gel was immunoblotted with antibodies against mitochondrial matrix marker Hsp60 as a control. The arrows beginning at nucleotides Ϫ24, Ϫ18, ϩ12, and ϩ21 indicate transcriptional start sites based on 5Ј-RACE (this work). The diamond at Ϫ94 indicates the 5Ј end of the TRM5 transcript mapped using a high density oligonucleotide array (36). FIGURE 6. Growth phenotypes and oxygen consumption of Trm5p truncation mutants. A, a ⌬trm5 haploid strain (DLY1) harboring full-length TRM5 on a URA3 plasmid (pRS416-TRM5) was transformed with various TRM5 constructs in LEU2 plasmids (pRS415). Leu ϩ transformants were then streaked onto 5-fluoroorotic acid (5-FOA)/YMD or 5-fluoroorotic acid/YPEG plates and incubated at 30°C for 4 or 5 days, respectively. Sector 1, full-length TRM5 (pRS415-TRM5); sector 2, pRS415-⌬1-33TRM5; sector 3, pRS415-⌬1-19TRM5; sector 4, pRS415-⌬1-47TRM5; sector 5, pRS415 empty vector (negative control). B, oxygen consumption of strain DLY1 harboring various TRM5 constructs grown in minimal medium with glucose as carbon source (YMD). All of the rates (nmol oxygen min Ϫ1 ) were normalized to the A 600 of the culture, and the mean oxygen consumption rate (nmol min Ϫ1 A 600
Although neither the ⌬1-19 or ⌬1-33 truncation mutants showed a growth defect on YPEG plates, the lack of detectable methylation prompted us to take a closer look at mitochondrial function in these two mutants. Oxygen consumption rates were measured on DLY1 cells expressing either the full-length or the ⌬1-19 and ⌬1-33 truncation constructs. As shown in Fig. 6B, cells expressing the ⌬1-33 construct exhibited a significantly lower rate of oxygen consumption compared with cells expressing full-length TRM5. Cells expressing the ⌬1-19 construct exhibited a smaller, nonsignificant decrease in oxygen consumption rate.
TRM5 Transcript Mapping-To determine whether the TRM5 gene produces multiple transcripts, we used RNA ligasemediated RACE to attempt to map the 5Ј end of the TRM5 transcript. The final PCR products were cloned into pBluescript II. Sequencing of 10 clones revealed four start sites: Ϫ24, Ϫ18, ϩ12, and ϩ21 (Fig. 5).
In primer extension-based methods, reverse transcriptase can pause or prematurely terminate in regions of high GC content or extensive secondary structure in the target mRNA. Therefore, we also tried a complementary transcript mapping method, ribonuclease protection analysis. In this method, a single-stranded RNA probe, covering nucleotides Ϫ300 to ϩ180 of the TRM5 gene, is hybridized to the 5Ј end of the mRNA, forming a duplex RNA that is resistant to digestion by single strand-specific ribonucleases. The length of the protected fragment is then determined by gel electrophoresis. However, we were unable to reproducibly detect protected fragments, apparently because of the presence of stable secondary structure(s) in the probe RNA.

DISCUSSION
At least eight mitochondrially encoded tRNAs are known to be methylated at G37 in vivo (6), yet the enzyme(s) responsible for modification of these mitochondrial tRNAs has not been previously identified. We have shown here that the tRNA(guanine-N1-)-methyltransferase encoded by the yeast TRM5 gene is responsible for methylation of G37 in mitochondrial initiator tRNA (tRNA Met f ). Similar to the human Trm5p (9), the recombinant yeast enzyme uses S-adenosyl-L-methionine as a methyl donor and is specific for G residues at position 37 in natural tRNAs. In particular, yeast Trm5p efficiently methylates G37 in mitochondrial tRNA Met f , which contains U at position 36. Thus, like the human Trm5p and in contrast to prokaryotic TrmD (9,32), the yeast enzyme appears to tolerate any nucleotide at position 36. The product of the reaction catalyzed by yeast Trm5p was shown to be m 1 G by HPLC.
Because Trm5p is also responsible for m 1 G37 modification of cytoplasmic tRNAs (1), the enzyme must have a dual localization to both cytoplasm and mitochondria. This was confirmed in yeast using a GFP-tagged version of Trm5p and antibodies against GFP. Proteinase K digestion showed that the mitochondrial form resides in the mitochondrial matrix, where it can access the mitochondrially encoded tRNAs. Early work in yeast indicated the existence of two tRNA(guanine-N1-)-methyltransferases in that organism, specific for tRNA sites 9 and 37 (2,33). These two distinct enzymes are now known to be encoded by TRM10 (30) and TRM5 (1), respectively. Sindhuphak et al. (2) even detected two peaks of m 1 G37 methyltransferase activity on a tRNA affinity column, foreshadowing the two forms of Trm5p (cytoplasmic and mitochondrial) reported here.
The predicted amino acid sequence of Trm5p reveals four potential AUG start codons in the N terminus (codons 1, 20, 34, 48) (Fig. 5). Growth analysis (Fig. 6A) revealed that a Trm5p active in the cytoplasm (as determined by normal growth on glucose) could be obtained with a construct lacking amino acids 1-33, whereas production of a Trm5p active in the mitochondria (as determined by the methylation status of mitochondrial tRNAs) required these first 33 amino acids. Oxygen consumption assays (Fig. 6B) revealed decreased respiration efficiency for both the ⌬1-19 and ⌬1-33 constructs, although only the ⌬1-33 construct was significantly reduced. These data suggest a model in which the mitochondrial form initiates at the first AUG codon, producing a protein with an N-terminal mitochondrial targeting sequence. Helical wheel analysis indicates that residues 1-19 could fold into an amphipathic ␣-helix, with a hydrophobic face and a basic face (cti.itc.virginia.edu/ϳcmg/ Demo/wheel/wheelApp.html). Our data cannot deduce whether or where the presequence is cleaved upon translocation into the mitochondrion. PSORT II analysis (psort.hgc.jp/ form2.html) predicts mitochondrial localization only for the full-length ORF; initiation at any of the downstream AUG codons predicts a cytoplasmic localization. In our model, the shorter cytoplasmic form initiates at the third AUG codon, producing a protein lacking a mitochondrial targeting sequence. This model is consistent with the slightly larger size of the mitochondrial form compared with the cytoplasmic form on SDS