Molecular and Biochemical Characterization of Rat γ-Trimethylaminobutyraldehyde Dehydrogenase and Evidence for the Involvement of Human Aldehyde Dehydrogenase 9 in Carnitine Biosynthesis*

The penultimate step in carnitine biosynthesis is mediated by γ-trimethylaminobutyraldehyde dehydrogenase (EC1.2.1.47), a cytosolic NAD+-dependent aldehyde dehydrogenase that converts γ-trimethylaminobutyraldehyde into γ-butyrobetaine. This enzyme was purified from rat liver, and two internal peptide fragments were sequenced by Edman degradation. The peptide sequences were used to search the Expressed Sequence Tag data base, which led to the identification of a rat cDNA containing an open reading frame of 1485 base pairs encoding a polypeptide of 494 amino acids with a calculated molecular mass of 55 kDa. Expression of the coding sequence in Escherichia coli confirmed that the cDNA encodes γ-trimethylaminobutyraldehyde dehydrogenase. The previously identified human aldehyde dehydrogenase 9 (EC 1.2.1.19) has 92% identity with rat trimethylaminobutyraldehyde dehydrogenase and has been reported to convert substrates that resemble γ-trimethylaminobutyraldehyde. When aldehyde dehydrogenase 9 was expressed in E. coli, it exhibited high trimethylaminobutyraldehyde dehydrogenase activity. Furthermore, comparison of the enzymatic characteristics of the heterologously expressed human and rat dehydrogenases with those of purified rat liver trimethylaminobutyraldehyde dehydrogenase revealed that the three enzymes have highly similar substrate specificities. In addition, the highest V max/K m values were obtained with γ-trimethylaminobutyraldehyde as substrate. This indicates that human aldehyde dehydrogenase 9 is the γ-trimethylaminobutyraldehyde dehydrogenase, which functions in carnitine biosynthesis.

Apart from the dietary intake of carnitine, most eukaryotes are able to synthesize this compound from trimethyllysine (5)(6)(7). This 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, after which the resulting ␤-hydroxytrimethyllysine is cleaved by a specific aldolase into ␥-trimethylaminobutyraldehyde and glycine. ␥-Trimethylaminobutyraldehyde is subsequently oxidized by ␥-trimethylaminobutyraldehyde dehydrogenase (TMABA-DH) 1 to form ␥-butyrobetaine (8). In the last step, ␥-butyrobetaine is hydroxylated at the ␤-position by a second hydroxylase, ␥-butyrobetaine hydroxylase, yielding L-carnitine (5,9,10). In rat and mouse, ␥-butyrobetaine hydroxylase is exclusively localized in the liver, whereas in man, 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 animals (9 -13). Kaufman and Broquist (6) were the first to demonstrate that ␥-trimethylaminobutyraldehyde is an intermediate in the carnitine biosynthesis of Neurospora crassa using isotope labeling experiments, and they suggested that an aldehyde dehydrogenase mediates its conversion to ␥-butyrobetaine. Perfusion experiments showed that ␥-trimethylaminobutyraldehyde is readily absorbed by rat liver and converted to carnitine via ␥-butyrobetaine, demonstrating the conservation of the dehydrogenation step in higher eukaryotes (14). Subsequently, Rebouche and Engel (10) showed that TMABA-DH activity was present in the cytosolic fraction of human liver, kidney, brain, heart, and muscle homogenates. In the same year, Hulse and Henderson (8) purified a cytosolic NAD ϩ -dependent aldehyde dehydrogenase from bovine liver showing maximum activity with ␥-trimethylaminobutyraldehyde, converting it into ␥-butyrobetaine.
Except for the human ␥-butyrobetaine hydroxylase, which has recently been identified in our laboratory (15), none of the enzymes of the carnitine biosynthetic route have been characterized at the molecular level. We therefore purified the aldehyde dehydrogenase responsible for the conversion of ␥-trimethylaminobutyraldehyde to ␥-butyrobetaine from rat liver and determined part of its amino acid sequence. Using this sequence information we identified the cDNAs encoding TMABA-DH from rat, human, and mouse. Finally, we expressed the cDNAs in Escherichia coli and compared the substrate specificities of the recombinant enzymes with those of the purified rat liver trimethylaminobutyraldehyde dehydrogenase.
Synthesis of ␥-Trimethylaminobutyraldehyde-4-Aminobutyraldehyde diethylacetal was trimethylated in ethyl acetate using methyliodide in the presence of 1,8-bis(dimethylamino)naphatalene (proton sponge). The iodide salt of 4-N-trimethylaminobutyraldehyde diethylacetal precipitated together with the protonated proton sponge. This precipitate was subsequently dissolved in distilled water by heating the mixture in a boiling water bath. After slow cooling to room temperature, only the protonated proton sponge crystallized, whereas 4-N-trimethylaminobutyraldehyde diethylacetal remained in solution. After removal of the proton sponge by filtration, the process was repeated five times in smaller volumes of distilled water to completely remove the remainder of the proton sponge. Hydrolysis of the resulting acetal in 0.1 M HCl for 30 min at room temperature gave ␥-trimethylaminobutyraldehyde. Water and HCl were evaporated in a rotavapor, and the ␥-trimethylaminobutyraldehyde was taken up in distilled water. In solution, ␥-trimethylaminobutyraldehyde was stable for at least three months at Ϫ20°C.
TMABA-DH Assay-TMABA-DH activity was determined either spectrophotometrically or fluorometrically at 37°C by monitoring the formation of NADH using a centrifugal analyzer (COBAS FARA; Roche Molecular Biochemicals). The assay mixture used in both methods contained 0.1 M sodium pyrophosphate buffer at pH 9.0, 0.5 mM NAD ϩ , and the enzyme sample in a final volume of 250 l. The reaction was started by adding ␥-trimethylaminobutyraldehyde to a final concentration of 100 M, unless otherwise indicated. In the spectrophotometric assay, the increase in absorbance at 340 nm was measured, and the activity was calculated using 6220 M Ϫ1 cm Ϫ1 as the molar extinction coefficient of NADH. In the fluorometric method, NADH formation was detected by measuring the fluorescence at 450 nm after excitation at 340 nm. Standard solutions of NADH were used for calibration.
Purification of TMABA-DH-Livers were taken from Wistar rats and homogenized by five strokes of a Teflon pestle in a Potter-Elvehjem glass homogenizer at 500 rpm in a 5 mM MOPS buffer, pH 6.0, containing 0.25 M sucrose and 2 mM EDTA. The crude homogenate was centrifuged for 10 min at 800 ϫ g at 4°C to remove nuclei and whole cells. The resulting post-nuclear supernatant was centrifuged for 3 h at 20.000 ϫ g at 4°C to obtain the cytosolic fraction. The cytosolic fraction was applied to an SP-Sepharose fast flow column (inner diameter ϭ2.8 cm, length ϭ 10 cm), which was pre-equilibrated with a 10 mM MES buffer, pH 6.0, containing 200 g/liter glycerol and 1 mM dithiothreitol (DTT). Bound proteins were eluted with a linear gradient from 0 to 100 mM NaCl in the same buffer. Fractions containing high TMABA-DH activity were pooled and loaded onto a red Sepharose CL-6B column (inner diameter ϭ 0.8 cm, length ϭ 7.5 cm), that was pre-equilibrated with a 10 mM MES buffer, pH 6.0, containing 200 g/liter glycerol, 1 mM DTT, and 25 mM NaCl. Bound proteins were eluted with a linear gradient from 25 to 500 mM NaCl in the same buffer. Fractions containing TMABA-DH activity were pooled and dialyzed against a 10 mM MES buffer, pH 6.0, containing 200 g/liter glycerol, 1 mM DTT, and 20 mM potassium phosphate. This dialysate was loaded onto an Econo-Pac CHT-II hydroxylapatite column (inner diameter ϭ1 cm, length ϭ 5 cm) equilibrated with the same buffer. Bound proteins were eluted with a linear gradient from 25 to 250 mM potassium phosphate. Fractions were tested for TMABA-DH activity and analyzed by sodium dodecyl sulfateacrylamide gel electrophoresis (PAGE) followed by silver staining. SDS-PAGE and silver staining were performed as described by Laemmli (17) and Rabilloud et al. (18), respectively. Protein concentrations were determined by the method of Bradford (19), using bovine serum albumin as standard.
Protein Digestion, Western Blotting, and Automated Edman Degradation-10 g of the purified TMABA-DH was digested for 1 h at 37°C with 0.05 g of endoprotease Glu-C (Roche Molecular Biochemicals) in a 50 mM Tris-HCl buffer, pH 8.0, containing 0.01% SDS. Protein fragments were resolved on a 15% SDS-PAGE gel, and a Multiphor II Nova Blot electrophoretic transfer unit (Amersham Pharmacia Biotech) was used to transfer proteins onto a polyvinylidene difluoride-sequencing membrane (Millipore, Bedford, MA) as described by the manufacturer of the transfer unit. Proteins were visualized with Coomassie Brilliant Blue. N-terminal amino acid sequencing was performed using a Procise 494 protein sequencer.
Cloning, Expression, and Purification of the Rat TMABA-DH and Aldehyde Dehydrogenase 9 (ALDH9) in E. coli-The complete open reading frame (ORF) of TMABA-DH was amplified by the polymerase chain reaction from rat liver cDNA using Advantage cDNA polymerase (CLONTECH, Palo Alto, CA) and the following primers: an EcoRItagged forward primer 5Ј-tatagaattcATGAGCACTGGCACCTTCG-3Ј and a SalI-tagged reverse primer 5Ј-tatagtcgacTTTTCAAAARGCW-GAYTCCAC-3Ј. The degenerate nature of the second primer also allowed the amplification of the ALDH9 ORF from human liver cDNA using the same primer set. The polymerase chain reaction products were cloned downstream of the isopropyl-1-thio-␤-D-galactopyranosideinducible P TAC promoter into the EcoRI and SalI sites of the bacterial expression vector pMAL-C2X, to express the TMABA-DH and ALDH9 as a fusion protein with maltose-binding protein (MBP). The ORFs were sequenced to exclude sequence errors introduced by the polymerase chain reaction, after which the constructs were transformed to the E. coli strain BL21. Transformed cells were grown on LB medium to an A 600 of 0.7, and isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 1 mM to induce expression of the fusion protein.
After 2 h, cells were pelleted and lysed in one-tenth of the culture volume in a 10 mM sodium phosphate buffer, pH 7.4, containing 140 mM NaCl, 200 g/liter glycerol, and 1 mM DTT by sonicating 2 times for 15 s at 8 W. 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 of the expression system (New England Biolabs) and stored at Ϫ80°C in a 10 mM sodium phosphate buffer, pH 7.4, containing 140 mM NaCl, 200 g/liter glycerol, 1 mM DTT, and 3 mg/ml bovine serum albumin.
Characterization of TMABA-DH-The Michaelis-Menten constant (K m ) and maximal velocity (V max ) for ␥-trimethylaminobutyraldehyde and several other aldehydes were determined for the purified rat liver enzyme and the purified recombinant fusion proteins using the assay described above. The concentration of ␥-trimethylaminobutyraldehyde was determined with 2,4-dinitrophenylhydrazine as described by Ariga (20). ␥-Aminobutyraldehyde was freshly prepared from 4-aminobutyraldehyde diethylacetal as described by Kurys et al. (21). Its concentration was determined with o-aminobenzaldehyde as reported by Jakoby and Fredericks (22). Because of the instability of ␥-aminobutyraldehyde at alkaline pH, activity measurements with this compound as substrate were performed at pH 7.4 using a 0.1 M sodium phosphate buffer. For the determination of the K m of NAD ϩ and NADP ϩ , ␥-trimethylaminobutyraldehyde was used at a fixed concentration of 100 M.

RESULTS
Purification of TMABA-DH from Rat Liver-In initial experiments, high TMABA-DH activity (ϳ3 nmol/min⅐mg) could be measured in crude rat liver homogenates. Subsequent measurement of TMABA-DH activity in subcellular fractions of rat liver showed that the activity was only present in the cytosolic fraction (results not shown). Therefore, rat liver cytosol was used as source of enzyme for the purification of TMABA-DH using liquid chromatography. An overview of the purification scheme is given in Table I. TMABA-DH activity was completely retained by all columns used and eluted as a single peak during all purification steps. 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 55 kDa was observed after the last purification step. The purified enzyme was highly unstable, except when stored at Ϫ80°C in the presence of 1 mM DTT and 200 g/liter glycerol. Even after several months of storage, no loss of activity could be measured.
Identification of the cDNA Encoding TMABA-DH-Attempts to directly sequence the protein by Edman degradation failed, suggesting that the N terminus of TMABA-DH is blocked. Therefore, the purified enzyme was subjected to digestion with the endoprotease Glu-C to generate peptides with a free N terminus. Peptide fragments were separated on SDS-PAGE, blotted onto a polyvinylidene difluoride-sequencing membrane, and visualized by Coomassie Brilliant Blue staining. Two peptide fragments were N-terminally sequenced, which resulted in the following sequences: EXINNGKSIFEA and EARLDVDTS (where X denotes an amino acid that could not be identified unambiguously). When the Swiss-Prot data base was screened with these sequences, the only homology found was with the human ALDH9 (EC 1.2.1.19, Swiss-Prot P49189). Subsequent searches in the Expressed Sequence Tag (EST) 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 ALDH9 cDNA (GenBank accession number U34252) (23,24). Based on the EST data, primers were selected to amplify the ORFs from rat and mouse liver cDNA. The rat and mouse amplicons both contained an ORF of 1485 base pairs, coding for a polypeptide of 494 amino acids with a predicted molecular mass of 55 kDa (GenBank accession numbers AF170918 and AF170919, respectively). The translated rat and mouse ORFs both have 92% positional identity with the human ALDH9 protein. The rat and mouse proteins are also highly homologous and share 96% positional identity. The two peptides obtained by Edman degradation were found to overlap, resulting in the following sequence: EXINNGKSIFEARLDVDTS. This sequence is identical to a 19-amino acid stretch in the N-terminal region of the rat sequence (95-114), demonstrating that the rat cDNA encodes TMABA-DH.

Expression of the Rat ORF as MBP Fusion
Protein in E. coli-The entire coding sequence of the rat cDNA was cloned into the pMAL-C2X expression vector and expressed in E. coli as a fusion protein with MBP. The fusion protein was purified from the E. coli lysate by affinity chromatography, and TMABA-DH activity was measured. The purified fusion protein exhibited high TMABA-DH activity, which confirmed that the cDNA encodes TMABA-DH.
Expression of the ALDH9-MBP Fusion Protein in E. coli-Human ALDH9 has high homology with the identified rat TMABA-DH and has been reported to dehydrogenate ␥-aminobutyraldehyde and betaine aldehyde (21,25), which are compounds with considerable structural resemblance to ␥-trimethylaminobutyraldehyde (Fig. 2). Therefore, the ALDH9 ORF was cloned in the pMAL-C2X vector to express ALDH9 as an MBP fusion protein. Sequencing of the ALDH9 ORF revealed three additional nucleotides in a GC-rich stretch of the previously reported ALDH9 cDNA, altering the predicted amino acid sequence. This information can be accessed through the Gen-Bank data base, accession number AF172093. The MBP-ALDH9 fusion protein was affinity-purified from E. coli lysate to determine whether ALDH9 was also active toward ␥-trimethylaminobutyraldehyde. The fusion protein exhibited high TMABA-DH activity, which indicates that ALDH9 is the human orthologue of rat TMABA-DH.
Characterization of the Purified Rat Liver TMABA-DH and Comparison with MBP Fusion Proteins-To investigate if the purified rat liver TMABA-DH and the rat MBP-TMABA-DH could also handle the substrates reported for ALDH9 and to further characterize the substrate specificity of the three enzymes, their kinetic properties were determined. Table II shows the kinetic parameters of the purified rat liver TMABA-DH, rat MBP-TMABA-DH, and MBP-ALDH9 with NAD ϩ , NADP ϩ , ␥-trimethylaminobutyraldehyde, ␥-aminobutyraldehyde, betaine aldehyde, and a range of aliphatic aldehydes as substrates. The K m and relative V max values of the purified MBP fusion proteins for the different substrates show a similar profile as the purified rat liver TMABA-DH. NAD ϩ was by far the preferred oxidant for all substrates, although NADP ϩ could also be used. The three enzymes have the lowest K m for ␥-trimethylaminobutyraldehyde in combination with a high V max value. As a consequence, the V max /K m ratio is highest for ␥-trimethylaminobutyraldehyde when compared with the other substrates. The presence of a free amino group in ␥-aminobutyraldehyde instead of the trimethylated amino group in ␥-trimethylaminobutyraldehyde results in considerably lower efficiency. Betaine aldehyde, the carbon backbone of which is two atoms shorter than ␥-trimethylaminobutyraldehyde but that contains the trimethylated amino group, is readily oxidized to betaine as reflected in the high V max values. The three enzymes have a high K m for betaine aldehyde, however, which results in a substantially lower efficiency when compared with ␥-trimethylaminobutyraldehyde. For the aliphatic aldehydes in the C 2 -C 8 range, the decrease in the K m values is accompanied by a steady increase of V max , showing that the efficiency of the enzymes is higher when the chain length of the aliphatic aldehyde increases. The efficiency of the enzymes with the longer aldehydes, hexadecanal and octadecanal, is very low if not undetectable. DISCUSSION To identify the enzymes of the carnitine biosynthetic pathway at the molecular level, we previously purified rat liver  ␥-butyrobetaine hydroxylase, the last enzyme in carnitine biosynthesis and used protein sequence data in combination with the EST data base to identify the corresponding human cDNA (15). In this study the same approach was used to identify TMABA-DH, which mediates the penultimate step in carnitine biosynthesis. The enzyme was purified from rat liver to apparent homogeneity and used for peptide sequencing. The resulting peptide sequences were subsequently used to search the EST data base, and two ORFs were identified from rat and mouse encoding proteins with high homology to the previously reported human ALDH9. The following observations demonstrated that the identified rat cDNA truly encodes TMABA-DH. First, the peptide sequence obtained by sequencing of the purified rat TMABA-DH exactly matched a 19-amino acid stretch in the translated coding region of the rat cDNA. Second, the cDNA encodes a protein with a calculated molecular mass of 55 kDa, which is in accordance with the apparent molecular mass of the purified rat liver TMABA-DH. Third, heterologously expressed rat cDNA exhibited high TMABA-DH activity. Finally, the kinetic properties of the recombinant rat MBP fusion protein are highly similar to those of TMABA-DH purified from rat liver. Although we did not express the mouse ORF in E. coli, it has 96% positional identity with the rat TMABA-DH and, therefore, most likely represents the mouse orthologue of rat TMABA-DH.
The rat TMABA-DH has high positional identity (92%) with human ALDH9. ALDH9 is a cytosolic NAD ϩ -dependent dehydrogenase belonging to the human aldehyde dehydrogenase gene family (26). It has been extensively investigated because of its proposed function in the alternative synthesis of the inhibitory neurotransmitter ␥-aminobutyric acid (GABA) (23,(27)(28)(29). In this pathway, diamine oxidase oxidatively deaminates putrescine (1,4-diaminobutane) to ␥-aminobutyraldehyde, which is subsequently oxidized to GABA by ALDH9. The majority of the GABA in rat adrenal gland is produced via this alternative pathway, whereas the GABA in brain is predominantly synthesized from glutamate by glutamate decarboxylase (30). Both the physiological importance of the conversion of putrescine to ␥-aminobutyric acid and the function of GABA outside the central nervous system is not well understood and remains to be established.
More recently, ALDH9 has also been implicated in the synthesis of betaine. Betaine can serve as a methyl donor in the biosynthesis of methionine and has also been proposed to be involved in the regulation of the osmolarity in the kidney during antidiuresis (31)(32)(33). For the synthesis of betaine, choline is oxidized by choline dehydrogenase to betaine aldehyde, which is subsequently converted to betaine by ALDH9 (25,34). In human tissues, betaine aldehyde dehydrogenase activity is predominantly found in liver, adrenal gland, and kidney. Northern blot analysis has shown the presence of the ALDH9 mRNA in liver, kidney, skeletal muscle, heart, brain, pancreas, lung, and placenta (23,27,34).
The high homology with rat TMABA-DH and the structural resemblance of the substrates of ALDH9 with ␥-trimethylaminobutyraldehyde prompted us to study whether human ALDH9 is in fact the human TMABA-DH. The finding that the recombinant human MBP-ALDH9 fusion protein exhibits high TMABA-DH activity suggests that human ALDH9 is, indeed, the human TMABA-DH.
Since ALDH9 has been reported to oxidize betaine aldehyde and ␥-aminobutyraldehyde, the kinetic properties of the two recombinant MBP fusion proteins and the purified rat liver

TABLE II
Substrate specificity of purified TMABA-DH and heterologously expressed enzymes Enzyme assays were performed as described in "Experimental Procedures," and the results are the means of three independent measurements. V max is expressed as the percentage of the maximal velocity measured with ␥-trimethylaminobutyraldehyde. In the measurements with NAD(P) ϩ , 100 M ␥-trimethylaminobutyraldehyde was used as the substrate. Measurements with ␥-aminobutyraldehyde were performed in a 0.1 M sodium phosphate buffer, pH 7.4. The V max values for this substrate are based on measurements with 100 M ␥-trimethylaminobutyraldehyde as the substrate at the same pH.

Substrate
Rat liver TMABA-DH Rat-MBP-TMABA-DH MBP-ALDH9 TMABA-DH were also determined for these and other substrates. Like the ALDH9 MBP fusion protein, both rat liver TMABA-DH and the rat TMABA-DH MBP fusion protein oxidized ␥-aminobutyraldehyde and betaine aldehyde. However, when considering both affinity and maximal velocity, ␥-trimethylaminobutyraldehyde is clearly the best substrate for all three enzymes. Recently, the three-dimensional structure of cod liver ALDH9 has been determined at a 2.1-Å resolution by x-ray crystallography (35). This protein has 70% positional identity with the human ALDH9 and is considered to be the cod orthologue of human ALDH9. The structural information revealed that the active site of cod ALDH9 is capable of handling larger aldehydes than betaine aldehyde, which is in accordance with our results, which reveal that the three forms of ALDH9 studied in this paper show the highest V max /K m ratio for straightchain aldehydes with a length of 7/8 carbon atoms. The preference for longer aldehydes also explains the relatively low K m values for ␥-trimethylaminobutyraldehyde and ␥-aminobutyraldehyde opposed to the high K m value for betaine aldehyde. The high V max value of the enzymes for betaine aldehyde is difficult to explain on the basis of the data presented here. Additional research is needed to understand this phenomenon.
Further investigation of the active site of cod ALDH9 showed that there is no negatively charged residue in the substrate pocket that interacts with the trimethylated amino group of betaine aldehyde. Instead, a hydrophobic interaction has been proposed between a tryptophan residue and the trimethylamino group of betaine aldehyde (35). If the nature of the interaction with the amino group of the substrate is hydrophobic instead of electrostatic, the reason for a higher K m value for ␥-aminobutyraldehyde than for ␥-trimethylaminobutyraldehyde could be that the positively charged amino group in ␥-aminobutyraldehyde is shielded by three methyl groups in ␥-trimethylaminobutyraldehyde.
The high activity of the heterologously expressed human ALDH9 with ␥-trimethylaminobutyraldehyde and the highly similar substrate specificity of ALDH9 and rat TMABA-DH strongly suggest that the human ALDH9 is the human TMABA-DH. This is supported by the presence of high betaine aldehyde dehydrogenase activity in human kidney and liver and the ALDH9 mRNA in tissues that contain high TMABA-DH activity (10,23,34). Although our data do not exclude an additional function of ALDH9 in GABA and/or be-taine synthesis, the results presented in this paper indicate that ALDH9 is the predominant, if not exclusive aldehyde dehydrogenase that functions in carnitine biosynthesis.