Mitochondria-specific RNA-modifying Enzymes Responsible for the Biosynthesis of the Wobble Base in Mitochondrial tRNAs

Human mitochondrial (mt) tRNALys has a taurine-containing modified uridine, 5-taurinomethyl-2-thiouridine (τm5s2U), at its anticodon wobble position. We previously found that the mt tRNALys, carrying the A8344G mutation from cells of patients with myoclonus epilepsy associated with ragged-red fibers (MERRF), lacks the τm5s2U modification. Here we describe the identification and characterization of a tRNA-modifying enzyme MTU1 (mitochondrial tRNA-specific 2-thiouridylase 1) that is responsible for the 2-thiolation of the wobble position in human and yeast mt tRNAs. Disruption of the yeast MTU1 gene eliminated the 2-thio modification of mt tRNAs and impaired mitochondrial protein synthesis, which led to reduced respiratory activity. Furthermore, when MTO1 or MSS1, which are responsible for the C5 substituent of the modified uridine, was disrupted along with MTU1, a much more severe reduction in mitochondrial activity was observed. Thus, the C5 and 2-thio modifications act synergistically in promoting efficient cognate codon decoding. Partial inactivation of MTU1 in HeLa cells by small interference RNA also reduced their oxygen consumption and resulted in mitochondria with defective membrane potentials, which are similar phenotypic features observed in MERRF.

The precise codon-anticodon pairing that occurs during translation requires a post-transcriptional modification at the anticodon first position (the wobble position) of the tRNA (1)(2)(3)(4). In mammalian mitochondrial (mt) 1 tRNAs, we recently dis-covered novel taurine-bearing modified uridines at the wobble position, namely, 5-taurinomethyluridine (m 5 U) in mt tRNA Leu(UUR) and 5-taurinomethyl-2-thiouridine (m 5 s 2 U) in mt tRNA Lys (5). We then showed that these wobble modifications are generated by the direct incorporation of taurine supplied by the medium. This is the first time it has been shown that taurine is a component of biological macromolecules.
We have also shown that the taurine-bearing modifications do not occur in mutant mt tRNAs that contain pathogenic point mutations that are associated with mitochondrial encephalomyopathies (6,7). Point mutations in mt tRNA genes are known to be responsible for a wide spectrum of human diseases that are caused by mitochondrial dysfunction. We found that the mt tRNA Leu(UUR) molecules obtained from pathogenic cells of MELAS (mitochondrial myopathy, encephalopathy, lactic acidosis, and stroke-like episodes) patients (8), who have either the A3243G or the T3271C mutation, lack the m 5 U modification (6). To examine decoding disorder of the mutant tRNA due to the wobble modification deficiency independent of the pathogenic point mutation itself, we used a molecular surgery to construct a mt tRNA Leu(UUR) lacking the taurine modification but without the pathogenic mutation. This "operated" mt tRNA Leu(UUR) without the taurine modification showed severely reduced UUG translation but no decrease in UUA translation. We thus concluded that the UUG codon-specific translational defect of the mutant mt tRNA Leu(UUR) is the primary cause of MELAS at the molecular level (9). This result could explain the complex I deficiency observed clinically in MELAS. In addition, mt tRNA Lys molecules bearing the A8344G mutation that were obtained from cybrid cells of patients with MERRF (myoclonus epilepsy associated with ragged-red fibers) (10) were found to lack the m 5 s 2 U modification (7). Thus, the MELAS and MERRF mitochondrial diseases are both associated with a lack of taurine modification of their mutant tRNAs. We examined the translational ability of the mutant mt tRNA Lys molecules obtained from cybrid cells of MERRF patients who carry the A8344G mutation. This analysis revealed that the tRNA Lys that lacks the m 5 s 2 U modification was incapable of translating both of its cognate codons, AAA and AAG, due to complete loss of codon-anticodon pairing on the ribosome (11). This suggests that the taurine wobble modification-deficiency results in a translational defect and that this is likely to contribute significantly to the mitochondrial dysfunc-tion associated with the pathogenic A8344G mt tRNA mutation. Notably, it has been suggested that it is the lack of the 2-thio modification of the MERRF mutant mt tRNA that is the main reason for the decoding disorder associated with this mutation, because it was observed that the 2-thio modification of mnm 5 s 2 U (5-methylaminomethyl-2-thiouridine) in Escherichia coli tRNA Lys is responsible for stabilizing their codonanticodon interactions (12). However, it has also been shown that the C5-taurinomethyl modification also apparently contributes to the efficient codon recognition of the tRNA Lys (5,13).
We speculated that the pathogenic point mutation hinders the biosynthesis of the 2-thio as well as C5-taurinomethyl modifications of m 5 s 2 U. It will be necessary to identify and characterize the enzyme genes that are responsible for introducing the 2-thio and C5-taurinomethyl groups onto the wobble position of tRNA Lys before we can elucidate the true molecular reason for the pathogenic nature of the mt mutations associated with MERRF, as well as with other mitochondrial diseases. This will hopefully lead to the development of therapeutic measures for these diseases. The biosynthesis of the m 5 s 2 U 2-thio and C5-taurinomethyl modifications of mt tRNA Lys probably involves complicated systems that require many genes, including several unidentified genes. In the case of the C5 modification of the wobble uridines, mnmE (trmE) and gidA are known to be required for the initial step in the modification of E. coli tRNAs, namely, mnm 5 U (5-methylaminomethyuridine) synthesis (14 -16). MSS1 and MTO1 were found to be the respective homologs of the mnmE and gidA genes in both human and yeast (17,18). The protein products of these genes were shown to localize in the mitochondria as a heterodimer complex, and their mutants are associated with respiratory defects (19). These observations suggest that the human MSS1 and MTO1 genes encode the enzymes that are responsible for the initial C5-taurinomethyl modification of mt tRNAs, although there is no direct evidence for this as yet. Concerning the 2-thio modification of the modified wobble uridines, it has been recently shown by Nakai et al. (20) that the yeast Nfs1p protein is involved in the 2-thio modification of both mitochondrial and cytoplasmic tRNAs. Nfs1p is a eukaryotic homolog of the bacterial IscS protein that serves to supply activated sulfur from cysteine to be utilized in iron-sulfur cluster biogenesis (21)(22)(23). IscS (cysteine desulfurase) was also found to be involved in the thio modifications of bacterial tRNAs (24,25). It can be assumed that the sulfur of the cysteine is initially taken up by IscS and is then transferred onto unidentified sulfur mediators. An enzyme that is directly responsible for the 2-thio modification of the tRNAs finally then introduces the sulfur from the mediator onto its target tRNA molecules.
In this study, we report our identification of a mitochondriaspecific tRNA-modifying enzyme that appears to be responsible for the 2-thio modification of m 5 s 2 U in tRNA Lys . We have denoted this enzyme as MTU1 (mitochondrial tRNA-specific 2-thiouridylase 1). We also describe our observations that suggest that a deficiency in the 2-thio modification of mt tRNA is a major molecular cause of the pathogenesis of the MERRFassociated mutation in the mt tRNA Lys molecule.

EXPERIMENTAL PROCEDURES
Yeast Strains and Media-Haploid Saccharomyces cerevisiae strains that lack single genes due to deletion by using the KanMX4 cassette (26) were purchased from the EUROSCARF collection (available at www.uni-frankfurt.de/fb15/mikro/euroscarf/). The BY4741 (MATa) strain was used as the wild type. 0 cells of the BY4741 strain were constructed as described previously (27). Complete deletion of the mt DNA was checked by Southern blotting (data not shown). Homologous recombination in meiosis was employed to obtain haploid cells with double gene deletions. Thus, the ⌬MTU1 strain (EUROSCARF Y03729, MATa) was mated with the ⌬MTO1 or ⌬MSS1 strain (the EURO-SCARF numbers are Y14603 and Y10598, respectively, MAT␣), and the diploid cells were selected on SD plates without methionine and lysine. These heterozygous strains were then sporulated and subjected to tetrad analysis. The tetrads displaying 2:2 segregation viability were selected on YPD (1% yeast extract (Difco), 2% bacto peptone (Difco), and 2% glucose) with G418, and the viable spores lacking the two genes were obtained as haploid cells with the MATa type. This was checked by mating tests. The growth properties and mitochondrial activities of these double deletion strains were compared with those of the single deletion strains with the MATa type, namely, BY4741 (MATa), ⌬MTU1 (EUROSCARF Y03729), ⌬MTO1 (EUROSCARF Y04603), and ⌬MSS1 (EUROSCARF Y00598). The media used to grow the yeast were the fermentable medium YPD, the non-fermentable medium YPG (1% yeast extract, 2% bacto peptone, and 3% glycerol), and the minimal medium SD (0.67% yeast nitrogen base without amino acids (Difco) and 2% glucose supplied with the amino acids required). For tetrad analysis, presporulation medium (0.5% yeast extract, 1% peptone, and 0.17% yeast nitrogen base without ammonium sulfate and amino acids, 50 mM potassium phosphate buffer, pH 5.0, 1% potassium acetate, and 0.5% ammonium sulfate) and sporulation medium (1% potassium acetate with the appropriate nutrients as required) were used.
Plasmid Construction and Site-directed Mutagenesis-The yeast MTU1 gene, including its promoter region (ϳ1 kbp of 5Ј-untranslated region), was obtained from the S. cerevisiae genome by PCR using the following primers: yMTU1-5ЈXhoI (5Ј-gaccactcgagtggctaattctctggaaggc-3Ј) and yMTU1-3ЈBamHI (5Ј-gattcggatccttatcatcactatgcatggg-3Ј). The PCR product was cloned into pYO325LEU2 (derived from pRS315) (28) to yield pYMTU1. The ORF region of human MTU1 (GenBank TM accession number AB178028) was obtained by nested RT-PCR from the total RNA of HeLa cells. The first RT-PCR was performed with the primers hMTU1-F1 (5Ј-aaggcgcggaagcggcggtagctgc-3Ј) and hMTU1-R1 (5Ј-cccagccatgggtcctgctctccag-3Ј), which amplified the MTU1 ORF along with both untranslated regions. The second PCR was carried out with the primers hMTU1-F2 (5Ј-ggcaatcgattcaagtcagataacaatgcaggccttgcggcacgtcgtgtg-3Ј) and hMTU1-R2 (5Ј-aaggttcggatcctcagagcaagggactcaggcct-3Ј), which yielded cDNA of the human MTU1 ORF attached with sufficient edge remaining that could be used in PCR ligation that would attach the yeast promoter region to its 5Ј-end. To do this, the promoter region of the yeast MTU1 gene was amplified with the primers yMTU1-5ЈXhoI and yMTU1-PR (5Ј-cacacgacgtgccgcaaggcctgcattgttatctgacttgaatcgattg-3Ј). The sequence of yMTU1-PR is complementary to the hMTU1-F2 sequence. The ORF region of human MTU1 was then ligated to the promoter region of the yeast MTU1 gene by using the primers yMTU1-5ЈXhoI and hMTU1-R2. The product was then cloned into pYO325LEU2 to yield pHMTU1. We employed a QuikChange site-directed mutagenesis according to the manufacturer's instruction to introduce the D38A mutation into the yeast MTU1 protein and the D18A mutation into the human MTU1 protein. The primers yMTU1D38A-F (5Ј-gatagtggcaatgtcatctggggtagcttcctctgtggcggcagcac-3Ј) and yMTU1D38A-R (5Ј-gtgctgccgccacagaggaagctaccccagatgacattgccactatc-3Ј) were used to introduce the yeast MTU1-D38A mutation. The primers hMTU1D16A-F (5Ј-ctgtccggcggcgtggccagcgccgtggcc-3Ј) and hMTU1 D16A-R (5Ј-ggccacggcgctggccacgccgccggacag-3Ј) were used for human MTU1-D18A mutation.
RNA Preparation and mt tRNA Isolation-Total RNAs from yeast cells was extracted with phenol, precipitated with ethanol, and washed with 70% ethanol as described (29) and then fractionated with DEAE-Sepharose Fast Flow (2 ϫ 30 cm) (Amersham Biosciences) with a linear gradient of NaCl (200 -500 mM) and MgCl 2 (2-16 mM) in a buffer containing 20 mM HEPES-KOH (pH 7.5). The fractions enriched with tRNAs were pooled and precipitated with ethanol. The mt tRNAs was homogeneously isolated from the fraction by an improved solid-phase DNA probe method that we have named chaplet column chromatography (30). The 3Ј-biotin-labeled synthetic DNA probe for mt tRNA Lys employed in this study is as follows: 5-tggtgagaatagctggagttgaaccaa-3Ј. Eighty micrograms of DNA probe was immobilized on 200 l of avidin-Sepharose (50% slurry, Amersham Biosciences) and packed in a small chaplet column. The tRNA fraction dissolved in binding buffer (1.2 M NaCl, 30 mM HEPES-KOH, pH 7.5, and 15 mM EDTA) was circulated through the column by a peristaltic pump at a temperature of 65°C. After washing out nonspecifically bound tRNAs with wash buffer (0.6 M NaCl, 15 mM HEPES-KOH, pH 7.5, and 7.5 mM EDTA), each tRNA was eluted from the column with elution buffer (20 mM NaCl, 0.5 mM HEPES-KOH, pH 7.5, and 0.25 mM EDTA) at 65°C. The eluted mt tRNA Lys was further purified by electrophoresis through a 12% polyacrylamide gel containing 7 M urea.
Mass Spectrometry-An LCQ ion-trap (IT) mass spectrometer (Ther-moFinnigan) equipped with an electrospray ionization source and an HP1100 liquid chromatography system (Agilent) was used to analyze the total nucleosides and the RNase T 1 -digested tRNA. Nucleoside analysis of total RNA was performed as described previously (30,31). We identified each ribonucleoside by comparing its retention time in mass chromatogram, exact m/z values of proton adduct [MϩH] ϩ and base fragment ion BH2 ϩ with the data base for LC/MS analysis of ribonucleosides as described (53). As for cmnm 5 U, we confirmed its retention time and mass spectrum by comparing with chemically synthesized cmnm 5 U (kindly provided by Dr. Murao). The exact mass of each RNA fragment produced by RNase T1 digest was determined by deconvoluting its multiply charged ions in the mass spectrum. Together with the information on a species of modified nucleosides in the tRNA and direct RNA sequencing by Donis-Keller's method (54), we could confirm the exact sequence of each RNA fragment.
Purified tRNA (ϳ0.05 A 260 /2.5 g) was digested with RNase T 1 (2.5 units) in 20 mM NH 4 OAc (pH 5.3) at 37°C for 30 min and then analyzed by mass spectrometry. The oligonucleotides produced by RNase T 1 digestion were analyzed as the negative ion form by LC/MS as described (32). Purified or unpurified tRNAs (0.02 A 260unit ) were digested by nuclease P 1 for 3 h at 37°C in 25 l of reaction mixture containing 20 mM HEPES-KOH (pH 7.5) and 0.5 l of 1 mg/ml nuclease P 1 and 0.5 l of 0.05 unit/ml bacterial alkaline phosphatase.
APM Gel Electrophoresis and Northern Blotting to Detect the Thiouridine Modification-The presence of the thiouridine modification in the purified tRNAs was verified by the retardation of electrophoretic mobility in a polyacrylamide gel that contains 0.05 mg/ml ((N-acryloylamino)phenyl)mercuric chloride (APM) (30,33), which is a method that was originally developed by Igloi (34). Thus, 0.05 A 260 unit of total RNA was separated by polyacrylamide gel electrophoresis and blotted onto Hybond N ϩ membranes (Amersham Biosciences). Each tRNA fraction was detected with a 5Ј-32 P-labeled oligonucleotide probe whose sequence was based on a specific sequence in the corresponding tRNA. The following oligonucleotides were used: 5Ј-tggtgagaatagctggagttgaac-3Ј for mt tRNA Lys ; 5Ј-tggttgaatcggtttgattcgaac-3Ј for mt tRNA Gln ; 5Ј-tggtaaccttaatcggaatcgaac-3Ј for mt tRNA Glu ; and 5Ј-tggctcctcatagggggctcgaac-3Ј for cytosolic tRNA Lys2 UUU . After hybridization, the membranes were washed, dried, and then placed in contact with the imaging plate for 1 h to detect the cytosolic tRNAs and for more than 24 h to detect the mt tRNAs. The radioactivities were detected with a BAS5000 bioimaging analyzer (Fuji Photo Systems).
Pulse-labeling of Mitochondrial Protein Synthesis-The [ 35 S]methionine incorporation into mitochondrial proteins encoded by mt DNA was basically performed as described previously (35). Thus, yeast cells incubated with sulfur-lacking medium were harvested and cycloheximide (final concentration, 150 g/ml) was added to inhibit cytoplasmic protein synthesis. The cells were then labeled with [ 35 S]methionine (54 mCi/mmol, Amersham Bioscience) at a concentration of 10 7 cells/ml/80 Ci for 10 min at 30°C. The total proteins from the cells were separated by SDS-PAGE, and the gel was dried and exposed to an imaging plate. The radiolabeled mitochondrial proteins were detected with a BAS5000 bioimaging analyzer (Fuji Photo Systems).
Subcellular Localization of MTU1 Protein-The ORF of human MTU1 was amplified from pHMTU1 by PCR using the primers hMTU1-5ЈXhoI (5Ј-ggccactcgagatgcaggccttgcggcacg-3Ј) and hMTU13ЈBamHI (5Ј-ggttcggatcccggagcaagggactcaggc-3Ј) and cloned between the XhoI and BamHI sites of pEGFP-N1 (Clontech) to yield pHMTU1/EGFP, which expresses a reporter MTU1 fused to the N terminus of EGFP. The resultant plasmid was transfected into 2 ϫ 10 5 HeLa cells by using a LipofectAMINE reagent (Invitrogen) according to the procedure recommended by the manufacturer. After 24 h of transfection, the mitochondria were stained by MitoTracker Red (Molecular Probes), and the red fluorescence of MitoTracker Red and the green fluorescence of EGFP were simultaneously observed by confocal fluorescence microscopy (Leica DEMIRBE).
Visualizing Mitochondrial Membrane Potential-72 h after the transfection of siRNA targeting MTU1, the cells were stained with both MitoTracker Red (final concentration, 50 nM) and MitoTracker Green (final concentration, 100 nM). As a control, we used cells that had been transfected with an siRNA-targeting luciferase. The cells were washed with phosphate-buffered saline, and 1 ml of medium was added. The cells were then visualized by confocal fluorescence microscopy (Leica DEMIRBE).
Oxygen Consumption-The rate of oxygen consumption by yeast cells (A 600 ϭ 0.4) cultured in 4 ml of YPD was measured for 15 min at 30°C or 37°C by using a Clark type electrode (YSI 5300A Biological Oxygen Monitor System). The HeLa cells were trypsinized and adjusted to 5 ϫ 10 5 cells/ml. The rate of oxygen consumption was monitored for 45 min at 37°C.

RESULTS
Identification of Mitochondrial tRNA-specific 2-Thiouridylase (MTU1)-It has been reported that the Escherichia coli iscS and mnmA (trmU and asuE) genes encode enzymes that are responsible for biosynthesizing the 2-thio group of mnm 5 s 2 U (37, 38). Moreover, recombinant IscS has been shown to cooperate with MnmA to synthesize the 2-thio modification of tRNA in vitro (39). MnmA is believed to act in 2-thiouridylation by recognizing the substrate tRNA. Thus, we speculated that the mitochondrial mnmA homolog may be a candidate gene that introduces the 2-thio group of m 5 s 2 U in mt tRNA Lys . We obtained the mnmA homologs in yeast, Caenorhabditis elegans, Drosophila, mice, and humans by BLASTP searches of protein databases. Sequence alignment revealed considerable conservation between these sequences (Fig. 1). The N-terminal regions contain a highly conserved SGGXDS sequence that is predicted to be a P-loop motif, which is a common ATP-binding motif found in the ATP pyrophosphate (pyrophosphate synthetase) family (40).
YDL033c is the S. cerevisiae yeast homolog of mnmA. It shows 29 and 24% homology to the sequences of the E. coli mnmA gene and its human homolog, respectively. It is a nonessential gene whose deletion strain is available (26). Thus, to examine the involvement of YDL033c in the 2-thio modification of the mt tRNAs for Lys, Glu, and Gln and the cytosolic tRNA Lys , total RNAs obtained from the YDL033c-deletion strain were subjected to polyacrylamide gel electrophoresis containing (N-acryloylamino)phenylmercuric chloride (APM) combined with Northern blotting (33, 34) (see "Experimental Procedures") (Fig. 2). The four tRNAs from the wild type strain show specific retardation in the APM-containing polyacrylamide gel due to the strong affinity of the 2-thio group in these tRNAs with the mercuric compound in the gel (Fig. 2, A-D, lane  1). We have confirmed that no other thio modifications in these tRNAs are responsible for their gel-retardation in this experiment (data not shown). In the case of the tRNAs from the deletion strain, the retarded bands for the three mt tRNAs were not observed (Fig. 2, A-C, lane 2). However, the cytosolic tRNA Lys molecule was still specifically retarded. This suggests that YDL033c is responsible for the 2-thio modification of mt tRNAs.
To confirm these observations, the mt tRNA Lys molecules (Fig. 3A) were isolated from the wild type strain and the ⌬MTU1 deletion strain and subjected to LC/MS analyses.
Nucleoside analysis revealed that the mt tRNA from the wild type cells contains cmnm 5 s 2 U (5-carboxymethylaminomethyl-2-thiouridine) and its precursor cmnm 5 U (5-carboxymethylaminomethyluridine) as the wobble modification (Fig.  3B, left panel). In the mt tRNA Lys from the deletion strain, however, there was a marked increase in the cmnm 5 U peak, and the cmnm 5 s 2 U peak disappeared (Fig. 3B, right panel). The purified mt tRNAs Lys molecules from the wild type and mutant strains were also subjected to RNase T 1 digestion, which cleaves off the anticodon-containing fragment (Fig.  3A), and this fragment was then analyzed by LC/MS (Fig.  3C). In the wild-type mt tRNA Lys , an anticodon-containing RNA fragment consisting of cmnm 5 s 2 U (5Ј-⌿CUcmnm 5 s 2 UUUt 6 AAGp-3Ј (3105.9 Da)) was prominently detected, whereas a fragment containing cmnm 5 U instead of cmnm 5 s 2 U (5Ј-⌿CUcmnm 5 UUUt 6 AAGp-3Ј (3089.9 Da)) was detected as a minor product. In contrast, in the tRNA molecules from the deletion strain, the cmnm 5 U-containing fragment was detected as the major product and no cmnm 5 s 2 Ucontaining fragments could be seen. Thus, YDL033c encodes an enzyme that is responsible for synthesizing the 2-thio group on cmnm 5 s 2 U in the three mt tRNAs for Lys, Glu, and Gln. Thus, we have named this gene MTU1 (mitochondrial tRNA-specific 2-thiouridylase 1).
Direct Evidence of the Requirement of MTO1 and MSS1 in the C5 Modification of the Wobble Uridine in mt tRNAs-In the modification of the E. coli tRNAs, the initial step of mnm 5 U synthesis is known to require the mnmE (trmE) gene, because the disruption of mnmE resulted in defective mnm 5 U synthesis (14,41). The gidA gene also appears to be involved, as suggested by indirect genetic evidence (15,16). Although MSS1 and MTO1 were found to be the respective homologs of the mnmE and gidA genes in both human and yeast (17,18), it has  never been demonstrated that these mitochondrial proteins are responsible for biosynthesizing the C5 modification of wobble uridines in mt tRNAs. To test this notion, yeast mt tRNAs Lys (Fig. 3A) molecules were isolated from MTO1 and MSS1 deletion strains and subjected to LC/MS analyses. First, total nucleoside analysis revealed that the mt tRNAs Lys from both deletion strains contains s 2 U (2-thiouridine) instead of the cmnm 5 s 2 U found in the wild type tRNA Lys molecules (data not shown). RNA fragment analyses by LC/MS then also showed that the anticodon-containing RNA fragment of the mt tRNAs Lys molecules from the ⌬MTO1 and ⌬MSS1 strains bear s 2 U (5Ј-⌿CUs 2 UUUt 6 AAGp-3Ј (3018.9 Da)) but not cmnm 5 s 2 U (Fig. 3C). These results directly show that MSS1 and MTO1 genes are both involved in the biosynthesis of the 5-carboxy-methylaminomethyl group of cmnm 5 s 2 U of mt tRNA Lys .
Synergism of MTU1 with MTO1 and MSS1 in Promoting Efficient Mitochondrial Decoding-To determine the physiological effects of deleting yeast MTU1, the respiratory activity of the ⌬MTU1 strain was examined by analyzing its growth on non-fermentable (YPG) and fermentable (YPD) medium plates and by determining its oxygen consumption. As a control, 0 cells grew normally on fermentable medium (YPD) but did not grow on the non-fermentable YPG plate (Fig. 4A). Growth defects of ⌬MTU1 on the YPD plate were not seen, but significant growth defect could be observed on the YPG plate. This can be explained by the fact that ⌬MTU1 showed strongly defective oxygen consumption (Fig. 4B). In addition, when we analyzed the mitochondrial protein synthesis in the wild type and ⌬MTU1 strains by [ 35 S]Met pulse-labeling experiments, we found it was severely reduced in the ⌬MTU1 strain (Fig.  4C). These results suggest that the 2-thio modification of mt tRNAs is essential for mitochondrial protein synthesis, which directly reflects the respiratory activity of mitochondria.
Because MSS1 and MTO1 appear to be responsible for the biosynthesis of the C5 modification in cmnm 5 s 2 U, we assessed whether the 2-thio group and the C5 modification act synergistically in the decoding activity of mt tRNAs. The ⌬MSS1 and ⌬MTO1 strains show a growth phenotype that is similar to that of the wild-type cell on both YPD and YPG plates (Fig. 4A), as has been previously reported for cells with paromomycin-sensitive genetic backgrounds (19,42). However, both ⌬MSS1 and ⌬MTO1 did show decreased rates of oxygen consumption (Fig.  4B), which indicates that defective C5 modification causes mitochondrial dysfunction to some extent, although the effect is weaker than that of ⌬MTU1. We then constructed two double knock-out strains, namely, ⌬MTU1/⌬MSS1 and ⌬MTU1/ ⌬MTO1, which should contain mt tRNAs with unmodified wobble uridines, because C5 and 2-thio modifications occur inde- pendently. Both strains showed a severe growth defect on even the YPD plate, and no growth was observed on the YPG plate (Fig. 4A). Moreover, almost no respiratory activity could be seen in either strain (Fig. 4B). The severe phenotypes of these double knock-out strains demonstrate that both the 2-thio and C5 modifications are required for efficient decoding in mitochondrial translation.
Identification and Characterization of the Human MTU1 Gene-The human homolog of MTU1 was introduced into the yeast ⌬MTU1 strain to see whether it can complement yeast MTU1. The ⌬MTU1 strain bearing the human MTU1 gene (⌬MTU1ϩpHMTU1) was now able to grow on YPG plates (Fig.  5A), and its rate of oxygen consumption was increased slightly relative to the ⌬MTU1 strain (Fig. 5B). The 2-thio modification of the mt tRNA Lys in this strain was then examined by APM gel electrophoresis. This revealed that the introduction of human MTU1 resulted in a slight but apparent recovery of the 2-thio modification in the ⌬MTU1 strain (Fig. 5C). The relatively weak complementation observed may be the result of the experimental system employed here, as we found that placing the yeast MTU1 gene back into the ⌬MTU1 strain by transfecting the cells with pYMTU1 also did not completely restore the 2-thio modification or the oxygen consumption in the ⌬MTU1 strain (Fig. 5C). Thus, it seems that this experimental design does not lead to complete complementation. Nevertheless, these experiments still clearly show that the human MTU1 gene works as a 2-thiouridylase in yeast cells.
It has been reported that certain genes responsible for tRNA modification work not only as tRNA-modifying enzymes but also as molecular chaperons for tRNAs, because the phenotype resulting from their gene deletion could be restored by introducing the enzyme genes that have been inactivated by mutating the active center of the enzymes (43). To determine whether the mitochondrial activity associated with MTU1 depends on its 2-thio modification of the mt tRNAs or another function, we constructed inactive yeast and human MTU1 mutants by, respectively, introducing the D38A and D16A point mutations into their P-loops. These mutant genes were then subjected to the complementation test, but neither promoted YPG plate growth or oxygen consumption of the ⌬MTU1 strain (Fig. 5, A  and B). This indicates that the highly conserved Asp residue in the P-loop works as an essential residue for both yeast and human MTU1 and that the enzymatic function of MTU1 is required for its role in mitochondrial activity.
The subcellular localization of human MTU1 was then examined by transiently expressing EGFP-fused MTU1 (MTU1-EGFP) in HeLa cells. Judging from the N-terminal sequences of human and yeast MTU1, the probability that they localize in the mitochondria is 0.53% and 0.87%, respectively, according to PSORT II Prediction analysis (44) (available at psort.nibb. ac.jp/). As shown in Fig. 5D, MTU1-EGFP indeed localizes mainly in the mitochondria. Recently, it has been reported that mouse 5-methylaminomethyl-2-thiouridylate-methyltransferase (TRMU) homolog localizes in mitochondria, although there is no direct evidence of this gene to be responsible for mt tRNA modification (45). This is consistent with the observation that MTU1 is a mitochondrial tRNA-specific 2-thiouridylase.
Knocking Down Human MTU1 Results in Mitochondrial Dysfunction-We have shown that human MTU1 complements yeast MTU1 and localizes in mitochondria. To obtain direct evidence that human MTU1 is responsible for the 2-thio modification of m 5 s 2 U in mt tRNAs, five siRNAs (MTU1-1 through -5, see "Experimental Procedures") that target human MTU1 were designed by an algorithm that predicts efficacious siRNA sequences. 2 HeLa cells were transfected with each siRNA or with siRNA targeting luciferase as a control and harvested 72 h later. The total RNAs were then subjected to semi-quantitative reverse transcription (RT)-PCR to estimate the knockdown activity of each siRNA. Each of the five MTU1 siRNAs caused a potent reduction in the MTU1 mRNA levels but had no effect on the glyceraldehyde-3-phosphate dehydrogenase mRNA levels, which served as a control. We chose MTU1-1 siRNA for further experiment. Judging from the band intensity of the RT-PCR products, MTU1 mRNA levels clearly decreased by MTU1-1 siRNA (Fig. 6A). APM gel-Northern analysis was then carried out to measure the degree of 2-thio modification of mt tRNA Lys in the MTU1 siRNA-transfected cells. A large part of the 2-thiolated tRNA Lys molecules from control luciferase siRNA-transfected cells that were retarded in the APM gel disappeared when MTU1-targeting siRNAs were introduced (Fig. 6B). The rate of oxygen consumption of the MTU1 siRNAtransfected cells was also clearly decreased as compared with the wild-type cells (Fig. 6C), although this is shown as raw data, because it is difficult to obtain large number of cells that are transfected by siRNA. We also examined the mitochondrial membrane potential of the knockdown cells by staining HeLa cells transfected by MTU1-or luciferase-targeting siRNAs with both MitoTracker Red and Green, which are fluorescent indicators for staining mitochondria (Fig. 6D). MitoTracker Red is an indicator of mitochondrial membrane potential (⌬⌿). The mitochondria of the control luciferase siRNA-transfected cells showed good staining with both dyes, and when the two images were superimposed they revealed a well developed meshwork. In contrast, the MTU1-knockdown cells had granular-shaped mitochondria that stained poorly with MitoTracker Red. Thus, knocking down human MTU1 results in mitochondria with a defective membrane potential, which is consistent with a phenotypic feature of cells from MERRF patients who carry the A8344G mutation (46,47). DISCUSSION In this study, we describe our identification of MTU1 as a mitochondrial-specific 2-thiouridylase that is responsible for the wobble modification of yeast and human mt tRNAs. The disruption of yeast MTU1 results in a 2-thio modification defect in the three mt tRNAs (for Lys, Glu, and Gln) that were examined. Pulse-labeling experiments with the ⌬MTU1 strain indicated that the 2-thio modification of the mt tRNAs is required for efficient mitochondrial protein synthesis. The human homolog of MTU1 complemented the ⌬MTU1 strain as its introduction caused the 2-thio modification of mt tRNAs and the oxygen consumption of the strain to recover slightly. Moreover, siRNA-induced knocking down of human MTU1 in HeLa cells resulted in a 2-thio modification defect in their mt tRNAs. This is the first time a reverse genetic approach utilizing siRNA to study a mammalian RNA modification enzyme has been reported. This specific repression of the human MTU1 gene resulted in reduced oxygen consumption and defective mitochondrial membrane potentials. This nicely explains our previous observation that the wobble modification deficiency of mt tRNA Lys bearing the 8344 mutation results in poor decoding of cognate codons (11). However, it should be noted that the effects of knocking down MTU1 may not completely correspond to the consequences of the 8344 mt tRNA Lys mutation, because MTU1 suppression may cause three mt tRNAs for Lys, Glu, and Gln to be inadequately modified. Consequently, it is likely that the knockdown of MTU1 causes the wobble base of the mt tRNAs for Lys, Glu, and Gln to lack the 2-thio modification and therefore to only carry the m 5 U modification. The MERRFassociated 8344 mutation, on the other hand, results in the complete absence of any modifications of mt tRNA Lys only (i.e. both the C5 and 2-thio modifications are absent). Nevertheless, our observations indicate that the 2-thio modification defi- ciency of mt tRNA induces strong mitochondrial dysfunction and that this may at least partly explain the mitochondrial disorders seen in MERRF patients.
It has been reported that tRNA-modifying enzymes have other function. An E. coli deletion strain of truB encoding pseudouridylase for position 55 shows growth defect in coculturing with wild type strain. However, the phenotype of this strain is recovered by introducing a plasmid containing truB mutant whose product has no catalytic activity (43). In addition, E. coli trmA encoding m 5 U54-methyltransferase is an essential gene, but its catalytic activity for methylation is not required for cell viability (48). These facts suggest that some RNA-modifying enzymes are multifunctional proteins. To test this possibility for yeast MTU1, point mutations D16A and D38A in the P-loop were respectively introduced into human and yeast MTU1 proteins. Because these mutants could not complement the ⌬MTU1 strain and restore the 2-thio modification, the conserved Asp residue in the P-loop is essential for enzymatic function of MTU1, which seems to be required for restoration of mitochondrial activity. However, introduction of a plasmid containing wild type MTU1 did not completely restore the 2-thio modification and oxygen consumption of ⌬MTU1 strain, indicating that there is an unknown cis element for MTU1 expression to be investigated. MTU1 bears the highly conserved P-loop motif (SGGXDS) in its N-terminal domain. This motif occurs widely in the pyrophosphate synthetase family ( Fig. 1) (40). The P-loop is known to participate in the binding and hydrolysis of the ␣-␤ phosphate bond of ATP. Other tRNA-modifying enzymes with the P-loop motif also exist. One is TilS, an essential tRNA-modifying enzyme that was recently identified by our group to be responsible for synthesizing lysidine at position 34 of the bacterial tRNA Ile molecules that are specific for the AUA codon (31). Also bearing the P-loop motif is ThiI, a thiouridine synthetase that is responsible for the synthesis of 4-thiouridine at position 8 (s 4 U8) of bacterial tRNAs (49,50). Both enzymes require ATP as the energy source for the reaction, as does E. coli mnmA, the homolog of MTU1 (39). More recently, TtcA was identified as a P-loop-containing modification enzyme for 2-thiocytidine at position 32 (s 2 C32) (51). Thus, the P-loop motif is commonly found in enzymes responsible for RNA modifications that target the carbonyl carbon of bases in tRNAs, which suggests a similar reaction mechanism is involved in these modifications. According to the reaction mechanism employed by GMP synthetase, which belongs to the pyrophosphate synthetase family (52), it is thus likely that the first step of 2-thiouridine synthesis is the activation of the C-2 carbonyl group of the wobble uridine by the addition of AMP to form an adenylate intermediate. Nucleophilic attack on the adenylate by an activated sulfur carried on an unidentified protein probably completes the reaction. The details of the mechanism of 2-thiouridine formation will be clarified in the near future.
It has been shown previously that a mutation in MTO1 caused yeast cells to exhibit a respiratory defect only when coupled with a paromomycin-resistant mutation in mitochondrial 15 S rRNA (19). This phenotype was similar to that induced by a mutation in MSS1 (42). Indeed, Mto1p and Mss1p were shown to form a heterodimer complex (19). It has not been reported for E. coli that the products of the gidA gene (the bacterial MTO1 homolog) and the mnmE (MSS1 homolog) directly interact. However, genetic evidence has suggested that gidA is responsible for C5 modification along with mnmE (14 -16). Nevertheless, direct evidence for this has not been reported. In our study reported here, we show by analyzing the tRNA molecules from the ⌬MTO1 yeast strains that MTO1 (gidA) are indeed responsible for the initial step of C5 modification. Disruption of MTO1 or MSS1 resulted in a normal growth phenotype, even upon culture on YPG plates, which is consistent with previous reports (19,42). However, both the ⌬MTO1 and ⌬MSS1 strains showed reduced oxygen consumption, which suggests that a decoding disorder arises from a deficiency in C5 modification of mt tRNAs. Notably, when MTO1 or MSS1 were disrupted along with ⌬MTU1, a severe reduction in respiratory activity resulted. This indicates that the C5 and 2-thio modifications of the wobble uridine act synergistically in promoting cognate codon decoding efficiency. Because the human homologs of MSS1 and MTO1 have been shown to be localized in the mitochondria (17,18), it appears that these proteins are responsible for the initial steps taken to synthesize the m 5 U or m 5 s 2 U modifications of mt tRNAs. It is also likely that there are some unidentified genes other than MSS1 and MTO1 that participate in these modifications, including a putative taurine-transferase for the biosynthesis of m 5 U (5). To understand how the C5 modification deficiency can contribute to the molecular pathogenesis of MELAS, it will be necessary to identify the enzyme genes that are responsible FIG. 7. Biosynthetic pathway that introduces the m 5 s 2 U (cmnm 5 s 2 U) modification of mitochondrial tRNAs. MSS1 and MTO1 are involved in the initial step of m 5 s 2 U (cmnm 5 s 2 U) synthesis on mt tRNAs. Mitochondrial taurine or glycine is subsequently incorporated into mt tRNAs by unidentified transferases to build m 5 s 2 U or cmnm 5 s 2 U, respectively. Nfs1 is responsible for the initial step in the 2-thiolation of m 5 s 2 U (cmnm 5 s 2 U) in mt tRNAs and mcm 5 s 2 U in cytoplasmic tRNAs (20). The sulfur from cysteine is transferred to unknown sulfur mediators by Nfs1p. MTU1 then acts as a mitochondria-specific 2-thiouridylase for m 5 s 2 U (cmnm 5 s 2 U) by using the activated sulfur from the mediators.
for the biosynthesis of m 5 U and to be able to reconstitute this whole biosynthesis pathway in vitro. Fig. 7 shows a summary of the m 5 s 2 U biosynthetic pathway. Our analyses of the tRNA Lys modifications in the ⌬MTO1, ⌬MSS1, and ⌬MTU1 yeast (Fig. 3C) indicate that both the C5 and 2-thio modifications appear to proceed independently. Concerning the 2-thio modification, Nfs1, an eukaryotic homolog of bacterial iscS, has been found to be a master enzyme gene involved in the 2-thio modification of both mitochondrial and cytoplasmic tRNAs (20). Nfs1p works as a supplier of activated sulfur from cysteine to be utilized in iron-sulfur cluster biogenesis, and it is also known to be involved in thio modifications of tRNAs. Moreover, it has been reported that recombinant IscS and MnmA (the MTU1 homolog) cooperate to synthesize 2-thio modifications in vitro by using ATP and cysteine as substrates (39). However, we have found that additional genes are indispensable for the biosynthesis of 2-thio modifications in E. coli tRNAs. 3 Thus, there must be sulfur mediators that act between Nfs1p and MTU1. MTU1 may directly recognize mt tRNAs and synthesize the 2-thio group by using ATP and the sulfur served by the mediator. With regard to the C5 modification, we have shown that MTO1 and MSS1 are directly involved in the initial step of its biosynthesis. It is likely that subsequent steps that involve as yet unknown genes finally lead to the synthesis of m 5 s 2 U through the incorporation of taurine supplied from plasma (5).
We have shown that MTU1 is not involved in synthesizing 2-thio modification of mcm 5 s 2 U in cytoplasmic tRNAs, showing the presence of an unknown enzyme in cytoplasm. It has been reported that yeast Trm9 gene encodes methyltransferase which is responsible for C5 modification of mcm 5 U and mcm 5 s 2 U (53). Although disruption of Trm9 gene resulted in complete loss of mcm 5 s 2 U in tRNA Glu , it is not clear whether it is an unmodified uridine or s 2 U in this report. Recently, our group has demonstrated that tRNA Glu from the Trm9 deletion strain has an unmodified wobble uridine, suggesting that 2-thio modification occurs in coupling with C5 modification in yeast cytoplasm. 4 It is a distinct biosynthesis with bacterial and mitochondrial 5-methyl-2-thiouridine type modifications (xm 5 s 2 U).
In bacterial and mitochondrial tRNAs for Lys, Gln, and Glu, the U at the wobble position is modified to xm 5 s 2 U such as mnm 5 s 2 U, cmnm 5 s 2 U, mcm 5 s 2 U, and m 5 s 2 U (1, 3-5, 54, 55). Because these tRNAs are responsible for decoding two codon sets that end in purine (R) (i.e. NNR), the xm 5 s 2 U modifications participate in preventing the misreading of the pyrimidine (Y)-ending near cognate codons (NNY) (56). xm 5 s 2 U is largely fixed in the C3Ј-end form (56,57). Due to this conformational rigidity, the xm 5 s 2 U modification prefers to base pair with A and prevents misreading of NNY codons (56,58,59). In addition, it was reported that 2-thio modification of mnm 5 s 2 U in E. coli tRNA Lys confers efficient ribosome binding (12,55,60). Furthermore, the C5-taurinomethyl group of m 5 U was shown to be required for the efficient decoding of the UUG codon by stabilizing the U:G wobble-base pairing on the ribosomal A site (9,61).
We have speculated that the MERRF A to G point mutation at the 8344 position of mt tRNA Lys may act as a negative determinant that inhibits the biosynthesis of both the C5 and 2-thio modifications involved in generating m 5 s 2 U (5, 7). Because we have found that human MTU1 is a mitochondrial specific 2-thiouridylase that participates in generating m 5 s 2 U of mt tRNAs, it will be possible to determine whether and how MTU1 can recognize MERRF mt tRNA Lys molecules once the biosynthetic pathway leading to the 2-thio modification can be reconstituted in vitro. If we find that the 8344 mutation is in a critical position for MTU1 recognition, it may be that MTU1 recognizes the whole architecture of mt tRNAs, because position 8344 in mt tRNA Lys is far from the wobble position. Furthermore, if we could successfully design mutant MTU1 molecules that efficiently recognize MERRF mt tRNA Lys molecules, it may be possible to restore the 2-thio modification of MERRF mt tRNA Lys by introducing the mutant MTU1 from a plasmid vector. This could potentially lead to the development of a gene therapy for MERRF.