Molecular and Biochemical Characterization of Rat epsilon -N-Trimethyllysine Hydroxylase, the First Enzyme of Carnitine Biosynthesis.

epsilon-N-Trimethyllysine hydroxylase (EC ) is the first enzyme in the biosynthetic pathway of l-carnitine and catalyzes the formation of beta-hydroxy-N-epsilon-trimethyllysine from epsilon-N-trimethyllysine, a reaction dependent on alpha-ketoglutarate, Fe(2+), and oxygen. We purified the enzyme from rat kidney and sequenced two internal peptides by quadrupole-time-of-flight mass spectroscopy. The peptide sequences were used to search the Expressed Sequence Tag data base, which led to the identification of a rat cDNA of 1218 base pairs encoding a polypeptide of 405 amino acids with a calculated molecular mass of 47.5 kDa. Using the rat sequence we also identified the homologous cDNAs from human and mouse. Heterologous expression of both the rat and human cDNAs in COS cells confirmed that they encode epsilon-N-trimethyllysine hydroxylase. Subcellular fractionation studies revealed that the rat enzyme is localized exclusively in mitochondria. Expression studies in yeast indicated that the rat enzyme is synthesized as a 47.5-kDa precursor and subsequently processed to a mature protein of 43 kDa, presumably upon import in mitochondria. The Michaelis-Menten constants of the purified rat enzyme for trimethyllysine, alpha-ketoglutarate, and Fe(2+) were 1.1 mm, 109 microm, and 54 microm, respectively. Both gel filtration and blue native polyacrylamide gel electrophoresis analysis showed that the native enzyme has a mass of approximately 87 kDa, indicating that in rat epsilon-N-trimethyllysine hydroxylase is a homodimer.

. Apart from the dietary intake of carnitine, most eukaryotes are able to synthesize this compound from trimethyllysine (5)(6)(7). The trimethyllysine is generated by the hydrolysis of proteins containing lysines that are trimethylated at their ⑀-amino group by a protein-dependent methyltransferase using S-adenosylmethionine as a methyl donor. In the carnitine biosynthetic pathway, trimethyllysine is first hydroxylated at the ␤-position by ⑀-trimethyllysine hydroxylase (TMLH 1 ), after which the resulting ␤-hydroxytrimethyllysine is cleaved by a specific aldolase into ␥-trimethylaminobutyraldehyde and glycine (6,8). Subsequently, ␥-trimethylaminobutyraldehyde is oxidized by ␥-trimethylaminobutyraldehyde dehydrogenase to form ␥-butyrobetaine (9). In the last step, ␥-butyrobetaine is hydroxylated at the ␤-position by a second hydroxylase, ␥-butyrobetaine hydroxylase, yielding L-carnitine (5,7,10). In rat and mouse, ␥-butyrobetaine hydroxylase is localized exclusively in the liver, whereas in humans, the enzyme is present in kidney, liver, and brain. Although most tissues are capable of converting trimethyllysine into ␥-butyrobetaine, liver and kidney are the main sites of carnitine biosynthesis in all mammals, including humans (10 -15).
After the recent identification of the cDNAs coding for ␥-trimethylaminobutyraldehyde dehydrogenase and ␥-butyrobetaine hydroxylase (16 -18), we focused our attention on the first enzyme of the carnitine biosynthesis, TMLH. Like ␥-butyrobetaine hydroxylase, TMLH is a non-heme ferrousiron dioxygenase that requires ␣-ketoglutarate, Fe 2ϩ , and molecular oxygen as cofactors (8, 19 -21). In this class of enzymes, the hydroxylation of the substrate is linked to the oxidative decarboxylation of ␣-ketoglutarate. In both humans and rat, the highest TMLH activity is found in kidney but is also present in liver, skeletal muscle, heart, and brain (13,20). Subcellular localization experiments using differential centrifugation indicated that the enzyme is predominantly localized in mitochondria (8,21) in contrast to the other three carnitine biosynthetic enzymes, which are cytosolic. We purified the hydroxylase responsible for the conversion of trimethyllysine to hydroxytrimethyllysine from rat kidney and determined part of its amino acid sequence by (quadrupole-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  time-of-flight) mass spectrometry. Using this sequence information we identified the cDNAs encoding trimethyllysine hydroxylase from rat, human, and mouse. Finally, we expressed the rat and human cDNAs in COS cells to confirm that the identified cDNAs encode TMLH.

EXPERIMENTAL PROCEDURES
Materials-Trimethyllysine was purchased from Sigma Chemical Co. Q-Sepharose HP and Butyl-Sepharose 4 Fast Flow were obtained from Amersham Pharmacia Biotech (Uppsala, Sweden), and CHT-II hydroxylapatite was from Bio-Rad (Hercules, CA). All other reagents were of analytical grade. The pMAL-C2X vector was purchased from New England BioLabs (Herts, United Kingdom), the pcDNA3 vector was from Invitrogen (San Diego, CA).
TMLH Assay-Two methods were used to determine TMLH activity. In the first method, TMLH activity was determined radiochemically by measuring trimethyllysine-dependent release of [ 14 C]CO 2 that is produced from the decarboxylation of ␣-[1-14 C]ketoglutarate to succinate. The assay mixture (total volume, 250 l) contained 75 mM Tris/ MES(HCl) buffer, pH 6.7, containing 90 M ␣-ketoglutarate, 10 M ␣-[1-14 C]ketoglutarate, 2.5 mM sodium ascorbate, 5.0 mM calcium chloride, 0.5 mM dithiothreitol, 0.5 mM ammonium iron(II)sulfate, 2 mg/ml bovine serum albumin, and 2 mM trimethyllysine. The reaction was started by adding the enzyme sample to the reaction mixture and allowed to proceed for 30 min at 37°C, after which it was terminated by the addition of 100 l of perchloric acid. The released [ 14 C]O 2 was trapped in 0.5 ml of 2 M NaOH, essentially as described by Wanders et al. (22), and the NaOH was counted for radioactivity in a liquid scintillation counter.
In the second method, the amount of hydroxytrimethyllysine that was enzymatically produced from trimethyllysine was determined by HPLC tandem MS. The reaction mixture and incubation time were the same as in the radiochemical method, except for the ␣-ketoglutarate concentration, which was 2.5 mM instead of 0.1 mM. This method will be described in detail elsewhere. Briefly, the reaction mixture was applied on a Microcon centrifugal filter unit with a 30-kDa cut-off (Millipore, Bedford, MA) to remove most of the proteins. 100 l of the filtrate was derivatized with methyl chloroformate at alkaline pH, followed by ethyl acetate extraction. Part of the aqueous phase was injected into an ion pair HPLC system using heptafluorobutyric acid as ion pairing agent, and the hydroxytrimethyllysine was quantified by tandem MS.
Purification of TMLH-Kidneys were taken from male Wistar rats and homogenized by five strokes of a Teflon pestle in a Potter-Elvehjem glass homogenizer at 500 rpm in a 10 mM Tris/HCl buffer, pH 8.0, containing 100 g/liter glycerol, 100 mM KCl, and 5 mM dithiothreitol (DTT). The crude homogenate was centrifuged for 10 min at 800 ϫ g at 4°C to remove nuclei and whole cells. The resulting supernatant was sonicated 5 times for 30 s at 10 watts and centrifuged for 1 h at 33,000 ϫ g at 4°C. The supernatant was collected, and the pellet was resuspended in the same buffer, after which the sonication and centrifugation steps were repeated. The two supernatants were pooled, diluted 20-fold in a 5 mM MES buffer, pH 6.0, containing 100 g/liter glycerol and 5 mM DTT and incubated overnight at 4°C. The resulting protein precipitate was pelleted by centrifugation for 20 min at 20,000 ϫ g at 4°C and dissolved in a 20 mM ethanolamine buffer, pH 9.3, containing 100 g/liter glycerol, 25 mM KCl, and 5 mM DTT. After centrifugation for 20 min at 20,000 ϫ g at 4°C, the supernatant was applied to a Q-Sepharose HP column (diameter ϭ 2.6 cm; h ϭ 10 cm; flow: 3 ml/min), which was pre-equilibrated with a 20 mM ethanolamine buffer, pH 9.3, containing 100 g/liter glycerol and 2.5 mM DTT. Bound proteins were eluted with a linear gradient from 25 to 400 mM KCl in the same buffer. Fractions containing high TMLH activity were pooled and diluted 1:4 in a 20 mM ethanolamine buffer, pH 9.3, containing 100 g/liter glycerol, 200 mM ammonium sulfate, 2 mM sodium ascorbate, 1 M KCl, and 5 mM DTT. The solution was centrifuged for 10 min at 4,000 ϫ g at 4°C to remove protein precipitates, and the supernatant was loaded onto a Butyl-Sepharose 4 Fast Flow column (diameter ϭ 1.6 cm; h ϭ 11 cm; flow: 2.5 ml/min), pre-equilibrated with the dilution-buffer. Bound proteins were eluted with a linear gradient from 1 M KCl ϩ 200 mM ammonium sulfate to 0 M of both salts in a 20 mM ethanolamine buffer, pH 9.3, containing 100 g/liter glycerol, 2 mM sodium ascorbate, and 5 mM DTT. Fractions containing high TMLH activity were pooled and diluted 1:1 in a 20 mM ethanolamine buffer, pH 9.2, containing 100 g/liter glycerol, 2 mM sodium ascorbate, and 5 mM DTT and loaded onto an Econo-Pac CHT-II hydroxylapatite column (diameter ϭ 2 cm; h ϭ 5 cm; flow: 1 ml/min) equilibrated with the same buffer. Bound proteins were eluted with a linear gradient from 0 to 50 mM potassium phos-phate. Fractions were tested for TMLH activity and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by silver staining. SDS-PAGE and silver staining were performed as described by Laemmli (23) and Rabilloud et al. (24), respectively. Protein concentrations were determined by the method of Bradford (25), using bovine serum albumin as standard.
Characterization of the Purified TMLH-The Michaelis-Menten constants (K m ) of purified TMLH for trimethyllysine, ␣-ketoglutarate, and Fe 2ϩ were determined using the radiometric assay described above. For the determination of the K m of ␣-ketoglutarate and Fe 2ϩ , a fixed concentration of 2 mM trimethyllysine was used. The pH optimum was determined by using 75 mM bis-tris-propane buffer instead of the Tris/ MES buffer, at pH values ranging from 5.5 to 9.5 in steps of 1 pH unit.
Protein and Peptide Analysis-For MALDI-TOF MS analysis, protein-containing gel slices were S-alkylated, digested with trypsin (Roche Molecular Biochemicals, sequencing grade), and extracted according to Shevchenko et al. (26). Extracted peptides were purified and concentrated using Zip-Tips (Millipore). Peptides were eluted from the Zip-Tips with 10 l of 1% (v/v) formic acid, 60% acetonitrile. The peptide solution was mixed with an equal volume of 10 mg/ml ␣-cyano-4-hydroxycinnamic acid (Sigma Chemical Co.) solution in acetonitrile/ethanol (1:1, v/v). Aliquots of 0.5 l were spotted on the target and allowed to dry at room temperature. MALDI-TOF MS spectra were acquired on a Micromass TofSpec 2EC (Micromass, Wythenshawe, UK) equipped with a 2-GHz digitizer. The resulting peptide spectra were used to search a non-redundant protein sequence data base (Swiss-Prot/ TREMBL) using the Proteinprobe program. For ESI-Q-TOF MS the peptide solution (2 l) was introduced into a nanospray capillary, and positive mode spectra were recorded with a Q-TOF mass spectrometer (Micromass) equipped with a Z-spray source.
Cloning and Expression of TMLH in COS Cells-The complete open reading frame (ORF) of rat TMLH was amplified by the polymerase chain reaction (PCR) from rat kidney cDNA using Pwo DNA polymerase (Roche Molecular Biochemicals) and the following primers: a BamHItagged forward primer 5Ј-aaaggatccATGAAGAGAGGAGACATAGCT-CAC-3Ј and a NotI-tagged reverse primer 5Ј-ttttgcggccgcTTAGGCAT-GAAGACCTAGAATTC-3Ј. The human ORF of TMLH was amplified from human kidney cDNA using the following primers: a BamHI-tagged forward primer 5Ј-tataggatccATGTGGTACCACAGATTGTC-3Ј and an NotI-tagged reverse primer 5Ј-tatagcggccgcCTGTTAAGCCTGAAG-CCCCAAGA-3Ј. The PCR products were cloned downstream of the P CMV promoter into the BamHI and NotI sites of the mammalian expression vector pcDNA3. Both ORFs were sequenced to exclude sequence errors introduced by Pwo DNA polymerase during the PCR, after which the constructs were transfected to COS cells using the Lipofect-AMINE Plus reagent (Life Technologies, Rockville, MD) as described by the manufacturer. 48 h after transfection, cells were harvested by trypsinization and lysed in a 10 mM sodium phosphate buffer, pH 7.4, containing 140 mM NaCl, 200 g/liter glycerol, and 1 mM DTT by sonicating two times for 15 s at 8 watts. TMLH activity was determined by the HPLC tandem MS method described above.
TMLH Expression in Yeast-The rat and human TMLH ORFs were amplified as described above using the same primers. For the amplification of the rat ORF starting from the second methionine, the following BamHI-tagged forward primer was used: 5Ј-ttttggatccATGCGCTTTG-ATTATGTCTGGC-3Ј in combination with the same reverse primer described above. The PCR products were cloned downstream of the galactose-inducible GAL1 promoter into the BamHI and NotI sites of the yeast expression vector pYES2. To assess the fidelity of the PCR process the ORFs were sequenced. The constructs were transformed to the Saccharomyces cerevisiae strain INVSC2 using the lithium acetate procedure (27). Transformed yeast cells were grown on minimal glucose medium (6.7 g/liter yeast nitrogen base, 3 g/liter glucose) to fully repress transcription of the GAL1 promoter. Cells were transferred to minimal lactate medium (6.7 g/liter yeast nitrogen base, 20 g/liter lactate) and galactose was added to a final concentration of 4 g/liter to induce protein expression. After overnight induction, spheroplasts were prepared using zymolyase (ICN Biomedicals, Costa Mesa, CA) according to Franzusoff et al. (28) and lysed in a 10 mM sodium phosphate buffer, pH 7.4, containing 140 mM NaCl, 200 g/liter glycerol, and 1 mM DTT.
TMLH Antibody Generation-The complete ORF of TMLH was amplified by PCR from human kidney cDNA using Pwo DNA polymerase and the following primers: a BamHI-tagged forward primer 5Ј-tataggatccATGTGGTACCACAGATTGTC-3Ј and a PstI-tagged reverse primer 5Ј-tatagacgtcCTGTTAAGCCTGAAGCCCCAAGA-3Ј. The PCR product was cloned downstream of the isopropyl-1-thio-␤-D-galactopyranoside-inducible P TAC promoter into the BamHI and PstI sites of the bacterial expression vector pMAL-C2X, to express the TMLH as a fusion protein with maltose-binding protein. The ORF was sequenced to exclude sequence errors introduced by PCR after which the construct was transformed into the Escherichia coli strain BL21. Transformed cells were grown in LB-medium with 100 g/ml ampicillin to an A 600 of 0.7 and isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 1 mM to induce fusion protein expression. After 2 h, cells were pelleted and lysed in 1/10 of the culture volume in a 10 mM sodium phosphate buffer, pH 7.4, containing 140 mM NaCl by sonicating two times for 15 s at 8 watts. The bacterial lysate was centrifuged for 10 min at 14,000 ϫ g, and the pellet was discarded. Fusion proteins were purified from the supernatant following the specifications of the manufacturer (New England BioLabs) and stored at Ϫ20°C. This fusion protein was used to raise an antiserum in a rabbit as described earlier (29).
Density Gradient Analysis-Kidneys were obtained from male Wistar rats and homogenized in 5 mM MOPS buffer, pH 7.4, containing 250 mM sucrose and 2 mM EDTA. A postnuclear supernatant was produced by centrifugation of the homogenate at 600 ϫ g for 10 min at 4°C and subfractionated by equilibrium density gradient centrifugation in a linear Nycodenz gradient as described previous (30). Glutamate dehydrogenase, catalase, ␤-hexosaminidase, phosphogluco isomerase, were used as markers for mitochondria, peroxisomes, endoplasmic reticulum, and cytosol, respectively. The activity of the marker enzymes was determined as described previously (31,32).
Immunoblot Analysis-A Multiphor II Nova Blot electrophoretic transfer unit (Amersham Pharmacia Biotech) was used to transfer proteins onto a Nitrocellulose sheet (Schleicher & Schuell, Dassel, Germany) as described by the manufacturer. After blocking of nonspecific binding sites with 50 g/liter Protifar and 10 g/liter bovine serum albumin in 1 g/liter Tween-20/phosphate-buffered saline for 1 h, the blot was incubated for 2 h with a 1:200 dilution of rabbit polyclonal antibodies raised against human recombinant TMLH fused to maltose-binding protein (purified as described above) in the same buffer without Protifar. Goat anti-rabbit IgG antibodies conjugated to alkaline phosphatase were used for detection, according to the manufacturer's instructions (Bio-Rad).
Gel Filtration and Blue Native PAGE-For gel filtration analysis a Superdex 200 column (Amersham Pharmacia Biotech) was used. A 20 mM ethanolamine buffer, pH 9.3, containing 100 g/liter glycerol, 2 mM sodium ascorbate, and 5 mM DTT was used as eluant at a flow rate of 0.4 ml/min. All analyses were performed at 4°C. The column was calibrated under identical conditions with the following protein standards: thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), aldolase (158 kDa), bovine serum albumin (66 kDa), ovalbumin (45 kDa), chymotrypsinogen A (25 kDa), and ribonuclease A (14 kDa), all from Amersham Pharmacia Biotech. The 10 log of the molecular mass of the protein standards was plotted against the corresponding elution fractions, and the molecular mass of TMLH was calculated by interpolation.
Blue native PAGE was performed as described previously using a 6 -14% polyacrylamide gradient gel (33,34). Citrate synthase (87 kDa) from pig heart (Sigma) and bovine serum albumin (66 kDa) were used as protein standards.

RESULTS
Purification of TMLH from Rat Kidney-Because kidney contains the highest TMLH activity in the rat (35), this tissue was used as source of enzyme for the purification of TMLH using liquid chromatography. An overview of the purification scheme is given in Table I. TMLH activity was retained completely by all columns used and eluted as a single peak during all purification steps. The presence of ascorbate was essential for preserving the enzymatic activity during the later purification steps and subsequent storage at Ϫ80°C. Samples obtained after each purification step were analyzed by SDS-PAGE followed by silver staining (Fig. 1). A single protein band with an apparent molecular mass of 43 kDa was observed after the last purification step.
Identification of the cDNA Encoding TMLH-Attempts to directly sequence the protein by Edman degradation failed, suggesting that the N terminus of TMLH is blocked. Therefore, the purified protein was digested with trypsin and analyzed by MALDI-TOF MS. Because no match was found in the nonredundant data base (Swiss-Prot/TREMBL), two peptides were selected for sequencing by Q-TOF MS, which resulted in the following sequences: TLLVDGFYAAQQVLQR (1821.99 Da) and MWYFTSDFRS (1339.58 Da). When the non-redundant data base was searched with these peptides sequences, both showed high homology with the human hypothetical protein FLJ10727 (GenBank accession number: NP_060666). Subsequent searches in the EST (Expressed Sequence Tag) data base identified several rat, mouse, and human EST clones with high homology to the peptide sequences. The homologous human ESTs all corresponded to the FLJ10727 cDNA (GenBank accession number: AK001589). Interestingly, the translated FLJ10727 cDNA showed high homology with human, rat, and Pseudomonas sp. AK-1 ␥-butyrobetaine hydroxylase. Based on the EST data, primers were selected to amplify the ORFs from rat, human, and mouse kidney cDNA. The rat and mouse amplicons both contained an ORF of 1218 base pairs, coding for a polypeptide of 405 amino acids with a predicted molecular mass of 47.5 kDa (GenBank accession numbers: AF374406 and AY033513, respectively). When the theoretical trypsin digest of the translated rat ORF was compared with the MALDI-TOF MS spectrum of the purified TMLH, 12 of the 27 theoretical peptides could be matched, which corresponds with a protein coverage of 34%.
The human amplicon contained an ORF of 1266 base pairs, coding for a polypeptide of 421 amino acids with a predicted molecular mass of 49.5 kDa (GenBank accession number: AF373407). The human TMLH cDNA sequence was identical to the FLJ10727 cDNA. This sequence is derived from genomic clone NT025307, which has been mapped to Xq28. BLASTn analysis of the human genome data base using the human TMLH cDNA as query showed that the TMLH gene spans about 130 kb and consists of at least eight exons.
The translated ORFs of rat and mouse both have 88% positional identity with the human TMLH protein. The rat and mouse proteins are also highly homologous and share 92% positional identity.
Expression of the Rat and Human TMLH cDNA-When the putative rat and human TMLH cDNAs were expressed in either E. coli (as maltose-binding fusion protein) or S. cerevisiae, no TMLH activity could be detected in lysates of these cells. Therefore, both ORFs were cloned into the eukaryotic expression vector pcDNA3 and transiently transfected to COS cells. As a negative control, the pcDNA3 vector without insert was included in the transfection experiment. After 48 h, TMLH activity was measured in the lysates of the transfected cells employing the TMLH assay, which is based on the measure-  Fig. 2A). Subsequent immunoblot analysis using the TMLH antibody showed a band with the same molecular mass as the purified TMLH in cells transfected with the rat and human ORF, which was hardly detectable in lysates of cells transfected with pcDNA3 without insert. Additionally, the amount of immunoreactive material was proportional to the TMLH activity (Fig. 2B).
Subcellular Localization of TMLH-To investigate the subcellular localization of TMLH, a density gradient analysis was performed with rat kidney homogenate. All the TMLH activity was associated with the particulate fraction, which was loaded on a Nycodenz density gradient. The activity profile in the gradient exactly coincided with that of the mitochondrial marker glutamate dehydrogenase, confirming the mitochondrial localization of TMLH (Fig. 3). When we analyzed the gradient-fractions by immunoblot analysis using antibodies raised against recombinant TMLH, the pattern of the immunoreactive material corresponded exactly with the TMLH activity profile (Fig. 3).
Processing of TMLH-With a calculated molecular mass of 47.5 kDa, the size of the translated ORF of rat TMLH is not in agreement with that of the purified protein, which has an apparent molecular mass of 43 kDa. Additionally, the COS cell transfection experiment showed that the produced human and rat TMLH both have the same apparent molecular mass as the purified rat TMLH, although the calculated molecular masses are 47.5 and 49 kDa, respectively. The 5Ј-end of the rat cDNA, as well as the mouse and human cDNAs, contained a second putative start codon. The use of this methionine would result in a protein of 42 kDa, and we therefore expressed the shorter rat protein in S. cerevisiae to investigate whether translation starts at this methionine. Immunoblot analysis of the yeast lysate with the TMLH antibody, however, clearly showed that the size of the expressed protein is smaller than the purified rat TMLH, indicating that this methionine is not used as start codon (Fig. 4). Another possibility is that TMLH is synthesized as a 47.5-kDa precursor, which is processed after import into the mitochondrion. The protein sequences of rat, mouse, and human TMLH indeed contain a putative N-terminal mitochon-  drial targeting sequence as determined by the Predotar version 0.5 prediction program. 2 Immunoblot experiments support this hypothesis, because expression of the full-length rat TMLH in S. cerevisiae resulted in a protein of 47.5 kDa (the predicted molecular mass of the translated rat ORF), but also showed a band with the same molecular mass as the purified rat protein (Fig. 4). Together, these results suggest that a 47.5-kDa precursor protein is synthesized and subsequently processed between the first and second methionine, resulting in a mature protein of ϳ43 kDa.
Characterization of the Purified TMLH-The enzyme has a broad pH optimum between 6.5 and 7.5 at 37°C, which is in agreement with previous results (20). K m values of trimethyllysine, ␣-ketoglutarate, and Fe 2ϩ were determined for the highly purified enzyme from Lineweaver-Burk double-reciprocal plots and were 1.1 mM, 109 M, and 54 M, respectively (Fig.  5). The K m value of trimethyllysine is in agreement with the results of Sachan et al. (21), who determined a K m value of 1.6 mM for the partially purified rat liver enzyme. Two other groups have determined K m values of 0.1 mM (20) and 0.13 mM (36) for the rat and bovine liver enzymes, respectively, which are considerably lower than the K m value determined in this study. The K m values found previously for ␣-ketoglutarate (480 and 220 M) and Fe 2ϩ (21 and 60 M) are in agreement with our results (20,36).
Native Molecular Mass Determination of Purified TMLH-Gel filtration analysis showed that the native enzyme has a molecular mass of ϳ87 kDa, suggesting that TMLH has a dimeric configuration (Fig. 6A). This result was supported by blue native PAGE analysis (Fig. 6B), which showed that TMLH has a similar size as citrate synthase (87 kDa). MALDI-TOF analysis demonstrated that the protein band of ϳ87 kDa only contained TMLH, suggesting that TMLH is a homodimer.

DISCUSSION
To identify the genes encoding the enzymes of the carnitine biosynthetic pathway we previously purified rat liver ␥-trimethylaminobutyraldehyde dehydrogenase and ␥-butyrobetaine hydroxylase, the penultimate and ultimate enzymes in carnitine biosynthesis, respectively. We used protein sequence data in combination with EST data base searching to identify the corresponding rat and human cDNAs (16,17). In this study the same approach was used to identify TMLH, which mediates 2 Available at www.inra.fr/Internet/Produits/Predotar. the first step in carnitine biosynthesis. The enzyme was purified from rat kidney to near homogeneity and used for peptide sequencing. Subsequently, the resulting peptide sequences were used to search the EST data base, and ORFs were identified from rat, mouse, and human origin with high homology to the peptide sequences. The following observations demonstrated that the identified rat cDNA truly encodes TMLH. First, the peptide sequences obtained from the purified kidney TMLH exactly matched two stretches of sequence from the translated coding region of the rat cDNA. Second, the heterologously expressed rat cDNA exhibited high TMLH activity. Third, the peptide pattern of the purified rat TMLH determined by MALDI-TOF analysis matched the theoretical trypsin digest of the translated rat ORF. Finally, the antibody raised against recombinant TMLH recognized the purified enzyme.
Because the human homologue of the rat TMLH has 88% positional identity with the rat protein and exhibited high TMLH activity in the heterologous expression system, the corresponding cDNA encodes human TMLH. Data base searching showed that the TMLH cDNA is identical to the FLJ10727 cDNA and that the TMLH gene is localized at Xq28. Although we did not express the mouse ORF, it has 92% positional identity with the rat TMLH, and therefore most likely represents the mouse homologue of TMLH.
Purified TMLH behaves as an 87-kDa enzyme in both gel filtration and blue native PAGE analysis. Because a single protein of 43 kDa was present in the final purification sample and the MALDI-TOF analysis of the blue native PAGE sample demonstrated that the dimer consisted of a single protein, TMLH appears to be homodimer. The last enzyme of carnitine biosynthesis, ␥-butyrobetaine hydroxylase, has considerable homology with TMLH and has also been reported to function as a homodimer (37)(38)(39). Analysis of the non-redundant data base with the BLASTp algorithm using rat TMLH as query, only retrieved ␥-butyrobetaine hydroxylase sequences from several organisms. No homology was found with other ␣-ketoglutarate-dependent dioxygenases, suggesting that TMLH and ␥-butyrobetaine hydroxylase belong to a separate subclass of dioxygenases.
TMLH has been reported to be localized in mitochondria (8,21), although this conclusion was drawn from relatively crude experiments involving differential centrifugation. Therefore, the subcellular localization of TMLH in rat kidney was reinvestigated by subcellular fractionation using density gradient analysis. The TMLH activity profile and the distribution of immunoreactive material clearly showed that TMLH is localized exclusively in mitochondria. The expression studies of TMLH in S. cerevisiae suggest that translation of TMLH starts at the first available start codon, which results in the formation of a 47.5-kDa precursor protein. This precursor is subsequently processed to the mature 43-kDa protein, presumably upon import into mitochondria where the mitochondrial import machinery removes the N-terminal presequence.
The mitochondrial localization of TMLH is remarkable, because the other three enzymes of the carnitine biosynthesis are localized in the cytosol. The submitochondrial localization of TMLH will have implications for the substrate flow and regu-lation of the carnitine biosynthesis. If TMLH is localized in the mitochondrial matrix, the existence of transport system to shuttle substrate and product over the inner mitochondrial membrane would be required. In contrast, if TMLH is present in either the inner membrane space or the outer mitochondrial membrane, no transport system would be needed because the outer mitochondrial membrane is permeable for small molecules.