Cloning and functional expression of a mammalian gene for a peroxisomal sarcosine oxidase.

Sarcosine oxidation in mammals occurs via a mitochondrial dehydrogenase closely linked to the electron transport chain. An additional H2O2-producing sarcosine oxidase has now been purified from rabbit kidney. A corresponding cDNA was cloned from rabbit liver and the gene designated sox This rabbit sox gene encodes a protein of 390 amino acids and a molecular mass of 44 kDa identical to the molecular mass estimated for the purified enzyme. Sequence analysis revealed an N-terminal ADP-βαβ-binding fold, a motif highly conserved in tightly bound flavoproteins, and a C-terminal peroxisomal targeting signal 1. Sarcosine oxidase from rabbit liver exhibits high sequence homology (25-28% identity) to monomeric bacterial sarcosine oxidases. Both purified sarcosine oxidase and a recombinant fusion protein synthesized in Escherichia coli contain a covalently bound flavin, metabolize sarcosine, L-pipecolic acid, and L-proline, and cross-react with antibodies raised against L-pipecolic acid oxidase from monkey liver. Subcellular fractionation demonstrated that sarcosine oxidase is a peroxisomal enzyme in rabbit kidney. Transfection of human fibroblast cell lines and CV-1 cells (monkey kidney epithelial cells) with the sox cDNA resulted in a peroxisomal localization of sarcosine oxidase and revealed that the import into the peroxisomes is mediated by the peroxisomal targeting signal 1 pathway.

In mammals a variety of H 2 O 2 -producing oxidases including D-amino-acid oxidase, D-aspartate oxidase, L-hydroxy-acid oxidase, acyl-CoA oxidase, and L-pipecolic acid oxidase are compartmentalized in peroxisomes. The H 2 O 2 generated from these reactions is then converted to H 2 O and O 2 by the peroxisomal matrix enzyme catalase (1). Several disorders have been described in which there is a defect in peroxisomal assembly that results in a partial or total absence of peroxisomal functions (for a review see Ref. 2). Patients with these peroxisomal disorders such as Zellweger syndrome, neonatal adrenoleukodystrophy, infantile Refsum disease, and hyperpipecolatemia all have elevated levels of L-pipecolic acid, an imino acid, which in human and monkey liver is oxidized by a peroxisomal L-pipecolic acid oxidase (3). Indeed, L-pipecolic acid oxidase activity was not detected in liver samples from patients with Zellweger syndrome (4). Primates dehydrogenate L-pipe-colic acid to ␦-piperideine-6-carboxylate which is spontaneously converted to ␣-aminoadipic acid ␥-semialdehyde.
The subcellular localization of this pathway seems to differ in other mammalian species. In rabbits, guinea pigs, dogs, and sheep L-pipecolic acid oxidation is primarily mitochondrial (5). However, during our studies examining the subcellular distribution of L-pipecolic acid oxidation in rabbits, a considerable amount of L-pipecolic acid oxidation was detected in the peroxisomes, in addition to the previously reported mitochondrial activity (6). Interestingly, this peroxisomal enzyme showed a high specific activity for sarcosine and also oxidized L-pipecolic acid and L-proline. This finding raised the question whether the purified enzyme is also a sarcosine oxidase.
In mammals, the oxidative removal of the methyl group from sarcosine is catalyzed by sarcosine dehydrogenase (EC 1.5.99.1) and dimethylglycine dehydrogenase (EC 1.5.99.2) in mitochondria. These enzymes were characterized as the two main folate-containing enzymes in rat liver mitochondria. Both enzymes are closely linked to the electron transport chain and form 5,10-methylenetetrahydrofolate. This "active" formaldehyde is predominantly used for the formation of serine from glycine by serine hydroxymethylase (7)(8)(9)(10)(11).
The enzyme investigated in this study differs from the classical mammalian sarcosine dehydrogenase described earlier.
The reaction takes place in peroxisomes rather than in mitochondria. The reaction mechanism is an H 2 O 2 -generating oxidation and not an electron transport chain-linked dehydrogenation. Although no mammalian sarcosine oxidases are known, several sarcosine oxidases from bacteria have been purified and characterized. They can be classified as monomeric enzymes (e.g. Arthrobacter sp. TE 1826 (12), Bacillus sp. NS-129 (13), Bacillus sp. B-0618 (14), Streptomyces sp. KB210 -8SY (15), and Cylindrocarpon didymum M-1 (16)) and as tetrameric enzymes (e.g. Arthrobacter ureafaciens (17), from Corynebacterium sp. U-96 (18) and Corynebacterium sp P-1 (19)). The monomeric enzymes have a molecular mass of 42-45 kDa which is similar to the size of the ␤-subunit of the tetrameric enzymes. All enzymes contain a covalently attached flavin (␤subunit of the tetrameric enzymes); the tetrameric enzymes also have a noncovalently bound flavin and at least the enzyme from Corynebacterium sp. P-1 probably contains an additional tightly bound NAD (20). While the covalent attachment is not unusual for bacterial oxidases, it has been rarely found in mammals. The list of mammalian enzymes with covalently attached flavins includes the mitochondrial sarcosine dehydrogenase and dimethylglycine dehydrogenase from rat and the peroxisomal L-pipecolic acid oxidase from monkey liver (3).
Like the mitochondrial dehydrogenases, the bacterial tetrameric sarcosine oxidases (21) can bind tetrahydrofolate and then yield 5,10-methylenetetrahydrofolate instead of formalde-hyde. It is not known whether the monomeric enzymes react with tetrahydrofolate.
A mammalian sarcosine oxidase from rabbit kidney has now been purified to investigate the association between the oxidation of the imino acid L-pipecolic acid and the methyl group acceptor sarcosine. Subsequently, the sarcosine oxidase gene was cloned from rabbit liver and expressed in Escherichia coli and in mammalian cells.
Bacterial Strains, Libraries, and Cells-The E. coli strains K802, DH5␣, and TG1 were used for all molecular screening and cloning procedures. To obtain the sox sequence, a rabbit liver cDNA library constructed in gt10 (kindly provided by Dr. M. Kilimann) and a rabbit liver genomic library cloned into EMBL3 SP6/T7 (Clontech, Palo Alto, CA) were screened.
The human cell lines were obtained and cultured as described by Moser et al. (22). The transformed derivatives were kindly provided by Dr. S. J. Gould. African Green Monkey CV-1 cells were obtained from ATCC (Rockville, MD; ATCC number CCL70).
Sarcosine Oxidase Assay-Sarcosine oxidase activity was determined by measuring the hydrogen peroxide formation in a horseradish peroxidase-coupled fluorometric assay, as described by Poosch and Yamazaki (23). The reaction mixture contained in a total volume of 550 l, 55 mM Tris, pH 8.4, 1 mM 4-hydroxyphenylacetic acid, 4 units of horseradish peroxidase (Boehringer Mannheim, types I or II), 9 mM sodium azide, 1 mM FAD, 100 l of enzyme solution (protein 1-80 g), and 9.8 mM sarcosine. The reaction was started with sarcosine, proceeded for various periods (15-60 min) at 37°C in the dark, and was stopped by the addition of 1.5 ml of 0.2 M glycine/sodium carbonate buffer, pH 10.5. The fluorescence at 415 nm (excitation 318 nm) was determined with the spectrofluorometer JY 3D (Jobin-Yvon, France) or with the spectrofluorometer LS 50 (Perkin-Elmer, Weiterstadt, Germany).
Protein concentrations were estimated by the method of Bradford (24) using the Coomassie Protein Plus reagent (Pierce). Bovine serum albumin was used as standard.
Purification of Sarcosine Oxidase-All purification steps were performed at 4°C under conditions that minimize light exposure of the enzyme. Kidneys from New Zealand White rabbits were cut in half to isolate the kidney cortex. The cortex (10 g) was minced and homogenized in 5 volumes (v/w) of homogenization buffer (250 mM sucrose, 6.8 mM HEPES, 1 mM EGTA, pH 7.5, containing 0.5 mM phenylmethylsulfonyl fluoride, 1 M pepstatin, and 1 M leupeptin) with 3 up-and-down passages of a loose-fitting pestle in an Potter-Elvehjem homogenizer at 800 rpm (Braun, Melsungen, Germany). The homogenate was centrifuged at 600 ϫ g for 10 min. The resulting pellet was resuspended in 3.5 volumes (v/w) homogenization buffer and again centrifuged at 600 ϫ g for 10 min. Both supernatants were combined and further centrifuged at 5,100 ϫ g for 15 min to pellet the M fraction which contained most of the mitochondria and peroxisomes. This pellet was gently rehomogenized in 6 volumes of homogenization buffer. To release the enzyme from the organelles, the M fraction was frozen in a thin layer at Ϫ70°C for at least 15 min, quickly thawed at 37°C, chilled, and centrifuged at 26,000 ϫ g for 15 min. The supernatant containing the solubilized enzyme was saved.
Heat Denaturation-The enzyme solution was brought to 46°C, stirred continuously for 10 min, and then chilled immediately. The denatured proteins were removed by centrifugation at 26,600 ϫ g for 15 min. The resulting supernatant was used for CM-cellulose chromatography.
CM-52 Cellulose Batch Chromatography-CM-52 cellulose (Whatman), equilibrated in 1 mM potassium phosphate buffer, 1 mM EGTA, pH 6.0, was mixed with 2 volumes of enzyme solution (ϳ70 ml) and shaken at 4°C for 30 min. Afterward, the gel suspension was transferred onto a glass frit (porosity G3), and a light vacuum was applied to remove the filtrate. The remaining gel was washed 5 times with 20 ml of equilibration buffer, and the enzyme was eluted in 7 fractions of 15 ml each (200 mM potassium phosphate buffer, 1 mM EGTA, pH 8.3). The enzymatically active fractions were combined and concentrated by di-afiltration in an Amicon chamber (PM 30 membrane, Amicon, Witten, Germany) to about 22 ml.
Butyl-Sepharose Chromatography-Subsequently, the enzyme solution was slowly mixed with 1 volume of 1.72 M potassium phosphate buffer, pH 7.8, and subjected to butyl-Sepharose chromatography.
Butyl-Sepharose CL 4B (Pharmacia, Freiburg, Germany), equilibrated with 1 M potassium phosphate buffer, 0.1 mM EGTA, pH 7.8 (column 1 ϫ 2 cm, 1.6-ml bed volume) was loaded with the enzyme solution (0.47 ml/min) and washed with 15-bed volumes of equilibration buffer containing 1 M pepstatin and 1 M leupeptin. The enzyme was eluted with a linear decreasing gradient of 1 M to 300 mM potassium phosphate in 0.1 mM EGTA, pH 7.8, 1 M pepstatin and 1 M leupeptin (95 ml of 1 M and 95 ml of 300 mM buffer). Enzymatically active fractions were combined and concentrated in an Amicon chamber equipped with a PM30 membrane (Amicon). The enzyme was stored at Ϫ70°C in the dark.
Organelle Separation in Nycodenz Gradients-To investigate the subcellular distribution of sarcosine oxidase, freshly prepared M fractions were separated in a Nycodenz gradient. One ml of M fraction was incubated with 1 ml of 50% (w/v) Nycodenz (Nycomed AS, Oslo, Norway) in 2 mM MOPS, 1 1 mM EGTA, pH 7.5, for 30 min to increase the density of the peroxisomal fraction and then loaded on top of a linear 30-ml Nycodenz gradient 18 -50% (w/v) in 2 mM MOPS, 1 mM EGTA, pH 7.5. The 18% Nycodenz solution contained 3% (w/v) sucrose. The sample was overlaid with homogenization buffer. Centrifugation was performed at 33,000 ϫ g for 90 min in a SS90 vertical rotor (Sorvall). Thirty fractions of approximately 1 ml were collected from the bottom of the gradient with a peristaltic pump. Peak fractions of mitochondria and peroxisomes were identified by marker enzyme analysis. Succinate cytochrome c dehydrogenase, a mitochondrial marker, was determined according to Parkes and Thompson (25) and catalase, a peroxisomal marker, was measured as described by Hü bl and Bretschneider (26).
Gel Electrophoresis, Isoelectric Focusing, and Immunoblotting-Protein samples were submitted to SDS-polyacrylamide gel electrophoresis in 10% gels as described by Laemmli (27), except that piperazinediacrylamide replaced bisacrylamide as the cross-linking agent. The gels were silver-stained as described by Wray et al. (28). Isoelectric focusing of purified proteins was performed with the Pharmacia Phast system. Five percent acrylamide gels were rehydrated in 9 M urea, 50 mM dithioerythritol, 10% sorbitol, 10% Nonidet P40, containing 20 l each of ampholytes pH 3-10, 5-8, and servalyte 4 -9 per ml buffer. Samples were diluted with the same volume of buffer containing 9 M urea, 50 mM dithioerythritol, 10% Nonidet P-40, and 20 l of ampholytes pH 3-10 per ml. Focusing proceeded at 540 V for approximately 410 V-h at 15°C. The gels were fixed in 20% trichloroacetic acid for 15 min and silverstained with the Phast system following the instructions of the manufacturer.
After SDS-PAGE the proteins were transferred onto nitrocellulose with a Trans-Blot Semidry Electrophoretic Transfer Cell (Bio-Rad) or with a Bio-Rad mini blot apparatus, according to the manufacturer. Methanol (20% final concentration) and SDS (1.3 mM final concentration) were added to the transfer buffer described by Bjerrum et al. (29). The nitrocellulose was blocked for 2 h in TBS (100 mM NaCl, 100 mM Tris-HCl, pH 7.4) with 5% (w/v) bovine serum albumin and subsequently washed with TBS containing 0.1% (v/v) Tween 20. Antibodies directed against FAD (kindly provided by Dr. M. Barber) (30) were used in a 1:500 dilution in TBS, 0.1% Tween. Antiserum against L-pipecolic acid oxidase from monkey liver was either diluted 1000-fold and used directly for immunoblotting or affinity purified against purified L-pipecolic acid oxidase from monkey liver and then used without further dilution. The procedure for affinity purification followed the protocol of Höhfeld et al. (31). Detection was performed with a 1:15 000 dilution of alkaline phosphatase-coupled goat anti-Ig from rabbit (Sigma) in a nitro blue tetrazolium/4-bromo-5-chloro-3-indolyl phosphate-coupled reaction, as described (32). Alternatively, the enhanced chemiluminescence kit (Amersham, Braunschweig, Germany) was used for immunodetection.
Spectophotometric Characterization of the Flavoprotein-Sarcosine oxidase dissolved in 10 mM potassium phosphate buffer, 0.1 mM EGTA, pH 7.8, was initially scanned in a Shimadzu double beam dual wave-length recording spectrophotometer UV 300. One ml of the enzyme solution was then precipitated in the dark with 100 l of 3 M trichloroacetic acid at Ϫ20°C for 1 h and then centrifuged at 14,000 ϫ g for 5 min. The yellow pellet was resupended in 1 ml of 5 M guanidine hydrochloride. Additional spectra were recorded for the supernatant and the redissolved pellet against corresponding blanks. The protein dissolved in 5 M guanidine hydrochloride was concentrated using a Centricon 30 microconcentrator (Amicon). The concentrated protein solution was diluted again with 5 M guanidine hydrochloride, and the recorded spectrum was compared with the spectrum of the corresponding filtrate.
Sequencing and RT-PCR-Nucleotide sequences were determined by primer walking using the Sequenase Version 2.0 and T7 DNA Polymerase (U. S. Biochemical Corp.) following the method of Sanger et al. (33). Alternatively, samples were sequenced on a DNA sequencer 373A (Applied Biosystems, Weiterstadt, Germany) utilizing the Prism Ready Reaction DyeDeoxy Terminator Cycle Sequencing Kit.
The complete cDNA for sarcosine oxidase was constructed by RT-PCR. 1 g of total rabbit liver RNA was reverse transcribed with 2 pmol of T3-1 primer (5Ј-GCGAGAAAGTAGTTGTGA-3Ј) and 200 units of Superscript RT (Life Technologies, Inc., Eggenstein, Germany). The synthesized cDNA was used as template for a PCR with two specially constructed primers A (5Ј-TAGAGCCTCGAGATGGCGGCTCAGAAA-GAT-3Ј) and B (5Ј-CCGTCTAGAATATCTCAGGGACACTCC-3Ј), which annealed at bp 1-18 and bp 1245-1262, respectively, and contained either a XhoI or a XbaI restriction site. The major part of the PCR product, the fragment between base pairs 189 and 1182, was exchanged against the corresponding fragment from the cDNA clone obtained by library screening to avoid Taq polymerase errors. The rest of the PCR fragment was confirmed by sequencing. This PCR fragment was cloned into the XhoI and XbaI sites of pKSϩ (Stratagene, La Jolla, CA) to yield the pBR1 construct.
Plasmids-In addition to pBR1, the constructs pBR2-4 were used to express full-length sarcosine oxidase in E. coli and for transfection experiments in mammalian cell cultures. Cloning of the XhoI/NotI fragment from pBR1 into the SalI/NotI sites of pGEX4T-3 (Pharmacia) resulted in the pBR2 construct which encodes a N-terminal fusion with glutathione S-transferase (GST). To obtain the pBR3 construct, coding for an N-terminal fusion with maltose binding protein (MBP), the XhoI/XbaI fragment of pBR1 was cloned into the SalI/HindIII sites of pMALc2 (Invitrogen, NVLeek, The Netherlands). The XbaI and HindIII sites were filled in with Klenow enzyme (Boehringer Mannheim) to yield blunt ends. The XhoI/XbaI fragment of pBR1 was also cloned into the corresponding sites of the mammalian expression vector pcDNA3 (Invitrogen) to yield pBR4.
Plaque and Northern Hybridization-For library screening phage plaques were transferred to nylon filters. Prehybridization and hybridization with a digoxigenin (DIG)-labeled probe (a 1500-bp long cDNA fragment of a human pipecolic acid oxidase clone 2 labeled with DIG-dUTP following Feinberg et al. (34)) was carried out at 42°C in 7% SDS, 2% blocking reagent (Boehringer Mannheim), 50% formamide, 0.1% N-lauroylsarcosine, and 5 ϫ SSC in 50 mM sodium phosphate buffer, pH 7.0. The filters were washed twice with 2 ϫ SSC, 0.1% SDS at room temperature for 10 min and then twice with 0.1 ϫ SSC, 0.1% SDS at 60°C for 30 min. For chemiluminescent detection of positive clones, the filters were processed with alkaline phosphatase-coupled DIG antibodies and Lumigen PPD (Boehringer Mannheim) and exposed to x-ray film. The genomic library was screened with a probe (bp 1-188) from a pipecolic acid oxidase cDNA clone 2 , labeled with [␣-32 P]dATP. Hybridization and washing followed the procedure described in Sambrook et al. (35) for high stringency conditions. Phage DNA was isolated (36) and the inserts subcloned into pKSϩ.
Total RNA from rabbit kidney and liver was isolated by guanidine thiocyanate/phenol/chloroform extraction (37), separated with a glyoxal agarose gel (38), and transferred onto a nylon membrane. The membrane was probed with an ␣-32 P-labeled human L-pipecolic acid oxidase cDNA 2 (39) for 24 h at 42°C and washed under the same conditions already explained for the plaque hybridization.
Expression of Sarcosine Oxidase in E. coli-The plasmids pBR2 or pBR3, respectively, were used to transform E. coli TG1 cells. Expression was induced with 0.3 mM IPTG at 37°C for pBR3 and with 0.1 mM IPTG at 30°C for pBR2 for 7 h. Lysis of bacteria which expressed the MBP fusion protein is described in Sambrook et al. (35). The protein was purified by affinity chromatography with an amylose resin (New England BioLabs, Schwalbach, Germany). For purification of the GST fusion protein cells were lysed by sonication followed by an incubation in lysis buffer (phosphate-buffered saline pH 7.5/1% Triton X-100) according to the instructions of Pharmacia. The fusion protein was further purified with a glutathione-Sepharose 4B resin as described by the manufacturer (Pharmacia).
Expression of sox in Mammalian Cells-CV-1 cells and human fibroblast cell lines were transfected with 3 g of pBR4 plasmid in the presence of 30 g of lipofectamine (Life Technologies, Inc.) according to the manufacturer's suggestions. After 48 h cells were analyzed by indirect immunofluorescence as described by Slawecki et al. (40). Best results were obtained at a dilution of 1:400 for affinity-purified ␣-pipecolic acid oxidase antibodies and 1:100 for ␣-catalase antibodies (The Binding Site, Heidelberg Germany). The fluorescein isothiocyanateand tetramethylrhodamine isothiocyanate-conjugated secondary antibodies (Dianova, Hamburg, Germany) were diluted 1:100. Micrographs were taken with Tmax 400 and Ektachrome 400 films (Kodak).  etic acid as fluorophore. The purification procedure for sarcosine oxidase from rabbit kidney is summarized in Table I. We obtained the M fraction containing peroxisomes, mitochondria, and lysosomes by differential centrifugation. Although preliminary experiments determining the subcellular distribution of sarcosine oxidase suggested a peroxisomal localization for sarcosine oxidase, an additional centrifugation step to purify the peroxisomes resulted in a lower yield of sarcosine oxidase, probably due to breakage of the organelles and loss of the enzyme into the supernatant. Sarcosine oxidase was purified 666-fold with an overall yield of 12.8% (Table I). The purest fraction separated into two closely migrating bands by SDS-PAGE (Fig. 1). These bands appear to have nearly identical pI and molecular mass in two-dimensional gel electrophoresis. The molecular mass of sarcosine oxidase was estimated at 44 kDa under denaturing conditions (Fig. 1). Isoelectric focusing in 6 M urea and 10% Nonidet P40 gave a pI of 7.8. The optimal enzyme activity in Tris buffer at 37°C was found at a pH of 8.6. Sarcosine Oxidase Oxidizes Sarcosine, L-Pipecolic Acid, and L-Proline-Substrate studies at a fixed concentration of 9.8 mM, which are summarized in Table II, identify sarcosine as the major substrate. The catalytic constants of each substrate are compared with the catalytic constant for sarcosine. Other prominent substrates are L-pipecolic acid and L-proline with rates of 30 and 21%, respectively, compared with sarcosine. The D-isomers of aspartic acid, alanine, proline, and pipecolic acid reacted to a minor extent, but we could not rule out a small contamination of the protein preparation with D-aspartate oxidase. No H 2 O 2 was formed with dimethylglycine. This substrate pattern is quite different from that for D-amino acid oxidase and D-aspartate oxidase (41) but similar to that of mammalian L-pipecolic acid oxidase isolated from monkey liver (3). L-pipecolic acid oxidase predominantly catalyzes the oxidation of L-pipecolic acid but also reacts with L-proline and sarcosine with rates of 23 and 10%, respectively.

Purification of Sarcosine
The catalytic efficiency, the ratio of k c /K m , is more suitable to determine which substrate is predominantly metabolized when several competing substrates are present (42). As shown in Table III, the k c for sarcosine is about 7 times higher than the k c for L-pipecolic acid, but the K m value for L-pipecolic acid (5.4 mM) is much lower than that for sarcosine (66.7 mM). Thus, the calculated catalytic efficiencies of 0.158 mM Ϫ1 min Ϫ1 for Lpipecolic acid and 0.093 mM Ϫ1 min Ϫ1 for sarcosine suggest that L-pipecolic acid is a slightly favored substrate.
The effect of benzoate, a known competitive inhibitor of other peroxisomal oxidases (D-amino-acid oxidase and L-pipecolic acid oxidase) was investigated with L-pipecolic acid and sarcosine as substrates. When L-pipecolic acid was the substrate, a K i of 2.04 mM was calculated for benzoate, with competitive inhibition (Fig. 2B). This value is similar to the K i of 0.75 mM previously reported for L-pipecolic acid oxidase (3). The K i value for benzoate with sarcosine as substrate was estimated at 5.46 mM. However, in this case the inhibition type was noncompetitive ( Fig. 2A).  Sarcosine Oxidase Contains a Covalently Bound Flavin-Purified sarcosine oxidase had a yellow color, and its sarcosine oxidizing activity was not dependent on FAD addition. Its absorption spectrum with a peak at 450 nm was typical for a flavoprotein (data not shown). When the purified enzyme was precipitated with 5% trichloroacetic acid, the protein pellet was yellow, indicating that the flavin was tightly associated with the protein. After resuspending the pellet in 5 M guanidine HCl, the purified enzyme absorbed at 356, 372, and 450 nm with a shoulder at 480 nm (Fig. 3B), a pattern typical for tightly bound flavins (43). After subsequent ultrafiltration (cut off, 30 kDa) the filtrate did not exhibit a typical flavin spectrum, but the retained enzyme, resuspended in 5 M guanidine HCl, had the same spectrum as the original solution. The fact that the flavin remained bound to the enzyme, even after precipitation with trichloroacetic acid or after resolubilizing in 5 M guanidine HCl, strongly suggested a covalent association.
Subsequently, sarcosine oxidase from rabbit kidney, along with pipecolic acid oxidase from monkey liver (covalently bound flavin), and D-amino-acid oxidase from pig kidney (noncovalently bound FAD) were separated by SDS-PAGE, blotted onto nitrocellulose, and examined with antibodies directed against flavins (30). Both, sarcosine oxidase and L-pipecolic acid oxidase reacted with the antibodies, suggesting that their flavin was covalently bound, whereas D-amino-acid oxidase, which has a noncovalently bound flavin, did not react with the antibodies (Fig. 4). The same samples were analyzed with antibodies raised against L-pipecolic acid oxidase from monkey liver (3). Sarcosine oxidase from rabbit kidney was recognized by the antibodies, but the antibodies did not cross-react with D-amino-acid oxidase (Fig. 4). Sarcosine oxidase migrated slightly faster than L-pipecolic acid oxidase during SDS-PAGE.
Sarcosine Oxidase Is a Peroxisomal Enzyme-Because the antibodies against L-pipecolic acid oxidase cross-reacted with sarcosine oxidase, they could be used to determine the subcellular localization of sarcosine oxidase. For these studies, a heavy mitochondrial fraction (M fraction) from rabbit kidney cortex was further separated in a Nycodenz gradient. Peak fractions of peroxisomes, mitochondria, and the supernatant fractions, identified by marker enzyme analysis, were separated by SDS-PAGE and blotted onto nitrocellulose. When the blot was analyzed with anti-L-pipecolic acid oxidase antibodies, a single protein band with a molecular mass of 44 kDa was FIG. 3. Absorption spectrum of purified and recombinant sarcosine oxidase. A, the recombinant GST-sarcosine oxidase fusion was precipitated with 10% (w/v) trichloroacetic acid, and the pellet was diluted to a protein concentration of 0.2 mg/ml with phosphate-buffered saline, pH 7.5, and the spectrum was recorded against the same buffer. B, sarcosine oxidase pooled after butyl-Sepharose chromatography was precipitated with 10% (w/v) trichloroacetic acid, dissolved in 5 M guanidine hydrochloride, and ultrafiltrated. The retenate was diluted in 5 M guanidine hydrochloride to the original volume, and the spectrum was recorded against 5 M guanidine hydrochloride in a Shimadzu UV 300 spectrophotometer. Protein concentration was 0.17 mg/ml. identified in the peroxisomal fractions (Fig. 5). All of the faint cross-reacting bands in the mitochondrial fractions and in the supernatant fractions also appeared in control blots incubated with preimmune serum from rabbits.
Molecular Cloning of Rabbit Liver Sarcosine Oxidase-The similarity between L-pipecolic acid oxidase from monkey liver (3) and sarcosine oxidase from rabbit kidney, especially the substrate specificity and the immunological cross-reactivity with L-pipecolic acid oxidase antibodies, encouraged us to investigate the sarcosine oxidase gene from rabbit liver.
When we probed a Northern blot with RNA from rabbit kidney and liver with a partial cDNA clone for human pipecolic acid oxidase, 2 hybridizing RNAs (approximately 2.3 kilobases (Fig. 6)) were detected in both kidney and liver with a higher expression level in kidney.
This same partial cDNA clone from human pipecolic acid oxidase was chosen to screen a rabbit liver cDNA library. A distinct cDNA clone (bp 157-2083) including the polyadenylation signal but lacking a putative start codon was obtained. Since we were not able to identify the 5Ј-end of the gene applying the rapid amplification of cDNA ends protocol (44), we probed a genomic rabbit liver library with a cDNA fragment (188 bp) from the very 5Ј-end of the human liver L-pipecolic acid oxidase gene. Several clones containing the putative transla-tion start site and additional 5Ј bp were isolated. A full-length cDNA was compiled by RT-PCR. We designated the rabbit sarcosine oxidase gene as sox. The sequence of the complete cDNA and the deduced amino acid sequence are shown in Fig.  7. The cDNA consists of a 12-bp 5Ј-untranslated region, an open reading frame of 1170 bp, and 913 bp of 3Ј-untranslated region and encodes for a protein Sox with 390 amino acids and a molecular mass of 44 kDa. The length of the cDNA is consistent with both the size of the native enzyme and with the size of the detected mRNA (Fig. 6), allowing 150 -200 bp for the poly(A) tail. The N terminus of the protein has sequence homology to an ADP-␤␣␤-binding fold, typical for proteins that bind FAD, NAD ϩ , or NADP ϩ (45,46). The last three amino acids, AHL, represent a peroxisomal targeting signal 1 (PTS1), characteristic for mammalian proteins that are translocated into peroxisomes (47).
The amino acid identities over the whole protein between rabbit Sox and the four monomeric bacterial sarcosine oxidases are between 25 and 28%. Apart from the ADP-␤␣␤-binding fold, three other almost identical regions were identified among the proteins. These segments are labeled 1-4 in Fig. 8. While the binding fold is characteristic for many enzymes, the three other homology regions are unique for monomeric sarcosine oxidases. Lower identities (14.9, 14.6, and 12.3%) were found with the ␤-subunit of the heterotetrameric sarcosine oxidase from Corynebacterium sp. P-1 (49), with the N terminus of dimethylglycine dehydrogenase from rat liver (50), and with the amino acid deaminase from Proteus mirabilis (51). High identities occurred with a not yet identified gene product (accession number U23529) from Caenorhabditis elegans. Interestingly, the encoded protein of C. elegans showed an N-terminal duplication of the first 300 amino acids of Sox. The identities to rabbit Sox are 25.1% for the first part and 27.6% for the C-terminal complete part (Fig. 9). As with the rabbit sarcosine oxidase, the C. elegans protein contains a PTS1 signal at the C terminus (AHL for rabbit Sox and SKI for the C. elegans gene product).
Comparison of Purified Sarcosine Oxidase to the Recombinant Gene Product of the Rabbit Liver sox Gene-To ensure that the right gene had been cloned, three characteristics of the purified sarcosine oxidase, the covalently attached flavin, the unusual substrate specificity, and the peroxisomal localization were investigated with recombinant sarcosine oxidase synthesized in E. coli.
Bacterial Expression of sox Revealed That Sarcosine Oxidase Has a Covalently Attached Flavin-Several approaches were used to express sox in E. coli. After transformation of E. coli with the pBR2 plasmid, a GST-fusion protein with a molecular mass of 80 kDa could be purified with a glutathione-Sepharose column. The recombinant protein had a slightly yellowish color. The spectrum of the recombinant protein showed absorption maxima at 380 and 450 nm (Fig. 3A) which were similar to those of the purified enzyme (Fig. 3B) and consistent with a bound flavin (43). When the protein was trichloroacetic acidprecipitated as described for the purified enzyme, no flavin was detected in the supernatant. The spectrum recorded for the recombinant enzyme (Fig. 3A) is similar to the spectra described for sarcosine oxidases (52). Interestingly, the shoulder at 480 nm, which has been described for the monomeric bacterial sarcosine oxidases, was more pronounced with the purified enzyme (52) (Fig. 3B).
Although the fusion protein could be synthesized with the covalently bound flavin, we were not able to isolate the protein in an enzymatically active form. Even after induction of the protein expression at a low IPTG concentration and at a temperature of 30°C, the formation of inclusion bodies (53) could not be prevented. The GST-fusion protein was only solubilized and purified after treatment with detergents.
Sarcosine Oxidase Synthesized as Fusion with Maltose-binding Protein Has Enzymatic Activity toward Sarcosine, L-Pipecolic Acid, and L-Proline-A different fusion protein was used to isolate enzymatically active sarcosine oxidase. After transformation of E. coli with the pBR3 construct, an MBP-fusion protein with a molecular mass of 86 kDa was partially purified with an amylose resin. After separation by SDS-PAGE the protein levels observed for the sarcosine oxidase fusion were lower than for the fusion protein found when the pMALc2 plasmid alone was used for expression in E. coli, even if the induction was carried out for 7 h (Fig. 10). Immunoblotting with the previously used antibodies against L-pipecolic acid FIG. 8. Sequence alignment between Sox from rabbit liver and four monomeric sarcosine oxidases from microorganisms. The alignment was performed with Lasergene (DNASTAR, London, UK) by the Clustal method using a PAM250 residue weight table. Amino acids identical in four out of five sequences are overlaid with black boxes. The PTS1 motif at the C terminus of rabbit Sox is marked by asterisks. The ADP-␤␣␤-binding fold and three other distinct sequence motifs, unique for sarcosine oxidases, are numbered 1-4 and overlined.
oxidase revealed an 86-kDa fusion protein cross-reacting with the antibodies, as well as a lower migrating protein band, likely due to partial degradation of the MBP-Sox protein (Fig. 10). Fractions purified by affinity chromatography with an amylose resin were investigated for oxidase activity using different substrates. The fusion protein oxidized the substrates sarcosine, L-pipecolic acid, and L-proline with different kinetics as shown by their K m and k c values (Table IV). Interestingly, the lowest K m was determined for L-pipecolic acid as substrate (1.9 mM), but k c was lower than that for sarcosine and L-proline. Sarcosine and L-proline had K m values of 6.7 and 8.0 mM, respectively. A marked inhibition was noted with sarcosine at higher substrate concentrations (above 13 mM). Catalytic efficiencies were best for L-pipecolic acid, followed by L-proline and sarcosine. When the pMALc2 vector alone was expressed in E. coli, the similarly processed protein product showed no activity with any of the three substrates.
Rabbit Sox Is a Peroxisomal Protein in Mammalian Cells-To investigate the localization of Sox in mammalian cells, the cDNA was cloned into the mammalian expression vector pcDNA3 creating pBR4. Three different human skin fibroblast cell lines, which were already transformed with SV40 large T antigen 3 and CV-1 cells (monkey kidney epithelial cells), were transfected with pBR4, using Lipofectamine. Two days after transfection, the cells were analyzed by indirect immunofluorescence. All cell lines expressed the gene with transfection rates of 10 -20%. Because rabbit sarcosine oxidase has a PTS1 signal at the C terminus, a peroxisomal localization of the enzyme was expected. After normal fibroblasts (GM5756) were transformed with pBR4 and subjected to indirect immunofluorescence using the antibodies against L-pipecolic acid oxidase (3), a punctate staining pattern was obtained (Fig.  11A). Double staining to include antibodies against the peroxisomal matrix protein catalase revealed that Sox colocalized with catalase, suggesting that it is found in or at the peroxisomes (Fig. 11B). The second investigated cell line 005-T is a transformed fibroblast cell line from a patient lacking a PTS1 receptor (complementation group 2 of the peroxisomal biogenesis disorders (40,54)) that results in a peroxisomal import defect for PTS1 and PTS2 proteins. When pBR4 was transfected into these cells, Sox was synthesized, but the protein was found throughout the cytoplasm of the cells (Fig. 11C). Again, this protein gave the same staining pattern as catalase, which is not imported into the peroxisomes of 005-T cells (Fig. 11D). The third cell line, which was from a patient with classical rhizomelic chondrodysplasia punctata (RCDP), has an isolated peroxisome import defect for the PTS2 protein thiolase but normal PTS1 import (40,55). Transfection of pBR4 resulted in a peroxisomal localization (Fig. 11E) of sarcosine oxidase (note the same subcellular distribution for catalase in Fig. 11F) in this cell line. These results indicate that sarcosine oxidase is a peroxisomal protein imported into the organelles by the PTS1dependent pathway. The same results were obtained when different plasmids encoding an N-terminal histidine-tagged sarcosine oxidase (pBR5) or a protein fusion with maltosebinding protein (pBR6) were transfected. Expression of sarcosine oxidase cDNA in CV-1 cells also resulted in a peroxisomal localization of the gene product. DISCUSSION During our studies of L-pipecolic acid oxidase, we discovered that rabbits have a similar enzyme, but this enzyme also oxidizes sarcosine. When the gene for this protein was cloned, we found that its amino acid sequence showed the most homology to the monomeric sarcosine oxidases.
Previously, it had been reported that while in human and monkey liver, L-pipecolic acid is oxidized in peroxisomes (3); in rabbit liver and kidney L-pipecolic acid is primarily oxidized in mitochondria (6). However, our studies with rabbit kidneys suggested that there might be L-pipecolic acid oxidation in peroxisomes as well. When we reexamined the oxidation of 14 C-radiolabeled L-pipecolic acid in subcellular fractions of a Nycodenz gradient by measuring the formation of aminoadipic acid (6), approximately 80% was attributable to mitochondrial activity, and 20% could be accounted for as peroxisomal activity 4 . When oxidase activity was determined by H 2 O 2 formation from L-pipecolic acid, almost all activity was found in the peroxisomal fraction, confirming two different systems for the oxidation of L-pipecolic acid in rabbit kidney. Like L-pipecolic acid, sarcosine has been reported to be oxidized by mitochondria in certain mammals (7,9).
Since antibodies raised against pipecolic acid oxidase from   (3) cross-reacted with rabbit sarcosine oxidase, as well as with the recombinant enzyme, we were able to use them in immunoblot analysis of fractionated rabbit kidney to show that rabbit Sox was solely peroxisomal. This peroxisomal localization was further supported by the finding that the rabbit sarcosine oxidase sequence encoded for a C-terminal PTS1 (tripeptide AHL), as well as the sequence for the hypothetical C. elegans protein (tripeptide SKI) (accession number U23529), but the sequences for the bacterial sarcosine oxidases (12-15) did not contain a peroxisomal targeting signal. Even if skin fibroblasts are not typical cells for the expression of amino-acid oxidases, the availability of cell lines with different peroxisomal import defects made them an excellent tool to study the intracellular localization of Sox in mammalian cells. Expression of cDNA for the rabbit sox gene resulted in an exclusively peroxisomal localization of the corresponding gene product in normal fibroblasts and in kidney CV-1 cells. In contrast, the peroxisomal biogenesis disorder cell line 005-T, which has no detectable PTS1 receptor, synthesized the Sox when transfected with the corresponding plasmid but did not import it into the peroxisomes. Instead, it remained in the cytoplasm like the peroxisomal matrix protein catalase. This finding indicated that the import of sarcosine oxidase into peroxisomes was dependent on an intact PTS1 receptor. In contrast, when sox was expressed in a cell line from a patient with RCDP, with a distinct defect in the import of PTS2 proteins, the protein was targeted to peroxisomes as in normal fibroblasts, further suggesting that rabbit sarcosine oxidase behaves like a typical PTS1 protein.
When the amino acid sequence for rabbit sarcosine oxidase was submitted to a BLAST search (48), the highest homology was seen with an unknown gene product from C. elegans, followed by the monomeric bacterial sarcosine oxidases. The gene for heterotetrameric enzyme from Corynebacterium sp. P-1 is organized as an operon containing genes that encode for all four subunits and the glyA gene, which encodes serine hydroxymethyltransferase (14). The ␤-subunit of this sarcosine oxidase shows less homology to Sox from rabbit than the monomeric enzymes but has a close relationship to the N terminus of mitochondrial dimethylglycine dehydrogenase from rat liver. The relationship to this mitochondrial dehydrogenase is supported by the finding that the ␣-subunit of the tetrameric FIG. 11. Subcellular localization of recombinant Sox from rabbit liver and catalase in normal transformed human fibroblasts and in fibroblasts from patients with peroxisomal biogenesis disorders. Normal fibroblasts (GM5756) (A and B), 005-T cells (no import of PTS1 and PTS2 proteins) (C and D), and cells from a patient with RCDP (no import of PTS2 proteins) (E and F) were transfected with pBR4 in the presence of lipofectamine. For indirect immunofluorescence the cells were double-stained with antibodies raised against pipecolic acid oxidase (A, C, and E) and catalase (B, D, and F) followed by fluorescein isothiocyanate-and tetramethylrhodamine isothiocyanate-conjugated secondary antibodies. Sox as well as catalase are imported into the peroxisomes in normal fibroblasts (A and B) and cells from the RCDP patient (E and F) but remained in the cytoplasm of 005-T cells (C and D). The bar indicates 10 m. enzyme has additional homology to the C terminus of dimethylglycine dehydrogenase (49). The tetrameric sarcosine oxidases share more similarities with the mitochondrial dimethylglycine dehydrogenases (the sequence for sarcosine dehydrogenase is not known), whereas the monomeric sarcosine oxidases are closer to the peroxisomal enzyme in mammalian cells.
The close relationship between purified and recombinant rabbit sarcosine oxidases, L-pipecolic acid oxidase from monkey liver and the bacterial sarcosine oxidases is further reflected in their substrate specificity. The tetrameric sarcosine oxidase from Corynebacterium sp. P-1 is able to metabolize L-proline and L-pipecolic acid but at turnover rates 220-fold less than that for sarcosine (56). No information is available as to whether L-proline or L-pipecolic acid are substrates for monomeric sarcosine oxidases.
When the substrate specificities of native rabbit kidney sarcosine were compared with that of the recombinant fusion protein investigated in this study and with that of pipecolic acid oxidase from monkey liver (Table V), all three enzymes utilized the same amino acids as substrates but the kinetics were different. The common link between these substrates is the imino moiety (Fig. 12) and the similarity of the reaction mechanisms. Oxidation of these substrates by sarcosine oxidase or pipecolic acid oxidase would yield glycine and formaldehyde for sarcosine, ⌬-piperideine-6-carboxylate for L-pipecolic acid, and ⌬-pyrroline-5-carboxylate for L-proline. In contrast, the oxidation by D-amino-acid oxidase would result in the formation of ⌬-piperideine-2-carboxylate from D-pipecolic acid.
Sarcosine oxidase from rabbit liver exhibits an ADP-␤␣␤binding fold satisfying the 11 consensus sequence requirements postulated by Wierenga et al. (57). The aspartate in position 1 does not fit the consensus, but in several well characterized FAD-binding sites an aspartate has been observed in this position (49,58). The absorption spectrum of the GST-Sox fusion protein and of the native protein suggests that sarcosine oxidase has a tightly attached flavin as coenzyme. The flavin attachment site has not yet been identified. As in the monomeric sarcosine oxidases, the rabbit enzyme lacks the DHVA tetrapeptide identified as the flavin attachment site (His-175) in Corynebacterium sp. U-96 and Corynebacterium sp. P-1 (both tetrameric enzymes) (49,59). However, in all monomeric sarcosine oxidases, in the C. elegans protein, and in rabbit Sox, a conserved histidine (His-49 for rabbit Sox) aligns with the flavin attachment site (His-84) of dimethylglycine dehydrogenase (50, 60). Newer results from Willie and Jorns (20) and Willie et al. (52) indicate that in the case of the enzyme from Corynebacterium sp. P-1 (19) this flavin is FMN, whereas two monomeric sarcosine oxidases contain FAD.
It might be remarkable that the fourth fragment of high homology at the C terminus (Fig. 2) shows high similarity to the N-terminal ADP-binding fold but did not fit the consensus of Wierenga et al. (57) exactly.
Several lines of evidence indicate that the purified rabbit enzyme is identical to the sox gene product. Both the native enzyme from rabbit kidney and the recombinant enzyme were localized in or targeted to peroxisomes. The calculated size from the sox sequence was identical to that estimated by SDS-PAGE of the native enzyme. The same polyclonal antibodies recognized both the purified and the recombinant enzyme. Both enzymes utilized the same substrates with similar catalytic efficiencies for sarcosine and L-pipecolic acid. A covalently attached flavin was found in both enzyme preparations. Some features such as the peroxisomal localization and the involvement of a nucleotide could be predicted from the primary sequence of the sox gene and were confirmed with the native protein.
In summary, L-pipecolic acid, L-proline, and sarcosine can be degraded by the same enzyme via an identical reaction mechanism. This enzyme belongs to a family of sarcosine oxidases which are all characterized by a flavin moiety which is covalently bound. This common type of flavin binding to the protein suggests that it may be required for this reaction mechanism. In contrast, the D-amino-acid oxidases, which have a similar but chirally opposite mechanism, do not contain a covalently bound flavin. Future studies of this reaction mechanism should help in elucidating just how flavin binding to enzymes is associated with particular reaction mechanisms. a Recombinant sarcosine oxidase (fusion with maltose-binding protein), this study, substrate concentration 9.8 mM.
b Purified sarcosine oxidase from rabbit kidney, this study, substrate concentration 9.8 mM.
c Purified L-pipecolic acid oxidase from monkey liver, Mihalik et al.