In Mouse α-Methylacyl-CoA Racemase, the Same Gene Product Is Simultaneously Located in Mitochondria and Peroxisomes*

α-Methylacyl-CoA racemase, an enzyme of the bile acid biosynthesis and branched chain fatty acid degradation pathway, was studied at the protein, cDNA, and genomic levels in mouse liver. Immunoelectron microscopy and subcellular fractionation located racemase to mitochondria and peroxisomes. The enzymes were purified from both organelles with immunoaffinity chromatography. The isolated proteins were of the same size, with identical N-terminal amino acid sequences, and the existence of additional proteins with α-methylacyl-CoA racemase activity was excluded. A racemase gene of about 15 kilobases was isolated. Southern blot analysis and chromosomal localization showed that only one racemase gene is present, on chromosome 15, region 15B1. The putative initial ATG in the racemase gene was preceded by a functional promotor as shown with the luciferase reporter gene assay. The corresponding cDNAs were isolated from rat and mouse liver. The recombinant rat protein was overexpressed in active form in Pichia pastoris. The presented data suggest that the polypeptide encoded by the racemase gene can alternatively be targeted to peroxisomes or mitochondria without modifications. It is concluded that the noncleavable N-terminal sequence of the polypeptide acts as a weak mitochondrial and that the C-terminal sequence acts as a peroxisomal targeting signal.

reactions, catalyzed by the peroxisomal oxidases that initiate ␤-oxidative cleavage of the side chain, are highly stereospecific for the (25S)-isomers (4,5). Thus, inversion of the configuration at C-25 is required to connect the two pathways. THCA-CoA was shown to be racemized efficiently by Amacr (1). Other physiologic substrates are methyl-branched fatty acids of dietary origin such as pristanic acid (2,6,10,14-tetramethylpentadecanoic acid), which is the ␣-oxidation product of phytanic acid, a metabolite of the chlorophyll component phytol (for review see Ref. 6), and possibly also endogenously synthesized isoprenoids. Amacr has also attracted the interest of pharmacologists for its participation in the biotransformation of 2-arylpropionic acids (2-methylarylacetic acids) used as nonsteroidal anti-inflammatory and analgesic drugs (Ibuprofen®), from the inactive (2R)-to the pharmacologically active (2S)-enantiomer (7).
Side chain cleavage of bile acids as well as ␤-oxidation of methyl-branched fatty acids take place in peroxisomes (8,9). However, Amacr activity is found to be distributed between peroxisomes and mitochondria in varying proportions, depending on the species. In human tissues, 80 -90% of the activity is associated with peroxisomes (2), in line with its presumed function. In mouse and Chinese hamster, a roughly equal distribution between the two organelles is seen (10), whereas in rats, Amacr activity is found almost exclusively in mitochondria (1). The molecular basis of this distribution is not yet known. Only one cDNA sequence for the enzyme (cAMACR) has so far been found in mice, rats (10), and humans (11) alike. To gain insight into possible mechanisms governing this distribution, we isolated the enzymes from highly purified mouse liver mitochondria and peroxisomes, respectively, and compared their properties. At the genomic level, only one gene encoding Amacr (AMACR) was found and characterized. The data are discussed in view of the dual compartmentalization of the Amacr activity in mouse liver.

EXPERIMENTAL PROCEDURES
Materials-Inorganic salts were from Merck; sucrose (molecular biology grade) and Tris were from Applichem (Darmstadt, Germany); Zymolase (Lyticase®), EDTA, and Percoll were from Sigma; and morpholinopropane sulfonic acid was from Aldrich. All other chemicals were of analytical grade or of the highest purity available. Ingredients for bacterial and yeast growth media were from Difco Laboratories (Detroit, MI). Restriction enzymes and the corresponding buffers were from MBI Fermentas (Heidelberg, Germany).
Oligonucleotides were synthesized with an Applied Biosystems model 394 oligonucleotide synthesizer at the Department of Biochemistry (University of Oulu) or were purchased from Carl Roth GmbH & Co. (Karlsruhe, Germany). The sequences of the oligonucleotides mentioned in the text are compiled in Table I.
For the preparation of the immunoaffinity column, IgGs from rabbitanti-Amacr-antiserum (2) were purified by a HiTrap ProteinA column (1 ml; Amersham Pharmacia Biotech), following the instructions of the manufacturer. The resulting IgGs were bound to 1.5 g of CNBr-activated Sepharose 4B (Amersham Pharmacia Biotech), according to the manufacturer's instructions.
Separation of Organelles and Immunoisolation of Amacr-Subcellular fractionation of mouse liver and CHO fibroblasts on Nycodenz gradients was performed essentially as described previously (14). For immunopurification of Amacr, livers of five mice were homogenized in 50 ml of sucrose solution (8.6% sucrose (w/w) in 1 mM Tris, 0.1 mM EDTA, pH 7.0) using a Potter homogenizer. Nuclei and undisrupted material were sedimented by centrifugation at 600 ϫ g for 10 min at 4°C. The supernatant was centrifuged for 25 min at 12,000 ϫ g in a JA-20 rotor, and the resulted pellet was resuspended in 18 ml of sucrose solution and loaded onto six centrifuge tubes containing 35 ml of Percoll solution (32.5% SIP, 67.5% sucrose solution) each. SIP consists of 9 volumes of Percoll-Suspension (Amersham Pharmacia Biotech) and 1 volume of 2.5 M sucrose, 20 mM MOPS, pH 7.2, 10 mM EDTA, 1% ethanol. The tubes were centrifuged in an AH-628 rotor at 80,000 ϫ g for 45 min at 4°C. After centrifugation, the tubes were emptied from the bottom, and the contents were collected into fractions of 1.2 ml. The fractions were assayed for succinate dehydrogenase (mitochondrial marker) and catalase (peroxisomal marker) activities, respectively. Fractions 1-3 (mitochondria) and fractions 26 -28 (peroxisomes) of all gradients were pooled, diluted 1:5 with sucrose solution (8.6% sucrose in 1 mM Tris, pH 7.0, containing 0.1 mM EDTA), treated with ultra sound, and centrifuged in a SW40-Ti rotor for 1 h at 150,000 ϫ g at 4°C. The supernatants were pooled and applied onto the immunoaffinity column (1 ϫ 5 cm) three times with a flow rate of 1 ml/min. Unbound proteins were washed out with 40 ml of 0.1 M Tris, 0.5 M NaCl, pH 8.0. Bound proteins were eluted with 60 ml of either 200 mM glycine, pH 2.8 (acidic elution), or 50 mM diethylamine, pH 11.5 (alkaline elution). The eluates of the latter were immediately neutralized with 4 ml of 1 M sodium phosphate, pH 5.1, containing 1 M NaCl, and concentrated on a 30 kDa cutoff membrane in an Amicon Centriprep tube (Amicon, Witten, Germany) to a volume of 5 ml.
Immunoelectron Microscopy-Livers were harvested from control and clofibrate-treated Balb/c mice (fed with 0.5% (w/w) clofibrate-supplemented chow for 2 weeks) after vascular perfusion fixation (17). Small tissue blocks (ϳ1 mm) were processed for thin sectioning by progressive lowering of temperature (18). Briefly, small tissue blocks were dehydrated in graded series of ethanol concurrent with decreasing temperature from 0 to Ϫ70°C before infiltration with MicroBed resin (Electron Microscopy Sciences, Fort Washington, PA) at Ϫ35°C; tissues embedded in the same resin were then polymerized under UV light at Ϫ35°C for 60 h. For immunogold labeling, thin sections (80 m) on nickel grids were incubated with a blocking medium containing 1% bovine serum albumin and 0.1% fish skin gelatin (Sigma) in phosphatebuffered saline and then exposed to polyclonal anti-Amacr antibody followed by washing in phosphate-buffered saline. Similar grids were incubated with catalase and mitochondrial reductase antibodies as control for the labeling. Antibody distribution was identified by gold particles (20 nm) conjugated to protein A.
Protein N-terminal Sequence Analysis-The proteins were separated by SDS-PAGE in a 12.5% gel (19) and transferred by electroblotting onto a polyvinylidene difluoride membrane (ProBlott®, Perkin-Elmer, Applied Biosystems Division, Foster City, CA) in 10 mM CAPS (pH 11)/10% methanol (v/v) with a constant potential of 50 V for 120 min (20). After staining with Coomassie Brilliant Blue, the protein bands of interest were cut off and loaded into the sequencer. Protein N-terminal sequencing was performed with a Procise 494A sequencer (Perkin-Elmer, Applied Biosystems Division, CA).
cDNA Analysis-Rat liver total RNA (30 g) was prepared from rat liver with the Biozym PUREscript RNA isolation kit (Biozym Diagnostik, Oldendorf, Germany), according to the instructions of the manufacturer. For the first strand synthesis of cAMACR, the TitanOne Tube reverse transcription-PCR system from Roche Molecular Biochemicals was used according to the instructions of the manufacturer, with 1 l of 100 M sequence-specific primer RRaceII (for oligonucleotide sequences see Table I). The first strands were polyadenylated and amplified by a PCR reaction using the polyT primer and primer HisR. The PCR products were purified, and a second nested PCR was performed with the same polyT primer and M9R as nested cAMACR-specific primer. The purified PCR products were ligated into pCR2.1 with the Invitrogen original TA cloning kit (Invitrogen, De Schelp, Netherlands) and transformed into competent INV-␣ cells, following the instructions of the manufacturer. Plasmids from 50 colonies were isolated and treated with EcoRI. The size of the excised inserts was determined by agarose gel electrophoresis. The three largest inserts were sequenced, using plasmid-specific primers (M13forw and M13rev).
Expression of Recombinant Rat Liver Amacr-For amplification of the complete rat liver cAMACR, a PCR was performed, using 1 l of rat liver Marathon-Ready cDNA (CLONTECH, Palo Alto, CA), primers RS-F and HisR, and Stratagene native Pfu DNA polymerase. The resulting product was purified from the agarose gel, cloned in pCR2.1, and transformed into INV-␣. The corresponding plasmids were amplified, purified, and digested with NdeI. The insert was cloned into NdeI-digested and dephosphorylated pET3a (pET Expression System 3, Novagen, Madison, WI) and transformed into INV-␣ cells. Plasmids containing the insert in the right orientation were transformed into competent BL21(DE3)-LysS. Plasmid C1-9, containing the coding sequence for rat liver racemase, was produced in the same manner, except that oligo R-YN-2-Kozak was used as forward primer and oligo R-YC-1-Mut was used as reverse primer for the PCR reaction.
For the expression of Amacr in yeast cells, the Pichia expression kit (Invitrogen) was used. For the generation of the expression casette, the coding cAMACR was amplified by PCR, using plasmid C1-9 as template, RY-F-wt as forward primer, and RY-C2-wt as reverse primer. The product was cloned into pCR2.1 plasmid and transformed into INV-␣ cells. The plasmids were amplified, isolated, and treated with NdeI. The resulting insert was isolated from a 1% agarose gel and ligated into NdeI-digested and dephosphorylated pHilD2 vector. The plasmid was amplified in INV-␣ cells, isolated, linearized with NotI, and transformed into Pichia pastoris strain GS115. The expression of recombinant Amacr from 90 Mut S transformants was analyzed according the

5Ј-TTG GAT CGT GAT TGG GAT ATT T-3Ј
manual of the manufacturer. Isolation of the Amacr Gene (AMACR)-Genomic DNA was extracted from kidney of BALB/c mouse (21) and used as template in PCR. Amplification (Expand TM High Fidelity PCR system; Roche Molecular Biochemicals) with primers ST10.1 and SR19.0, resulted in a 5-kb fragment, which was subcloned into the pUC18 vector (Sure Clone ligation kit; Amersham Pharmacia Biotech) and partly sequenced from the 5Ј end. A pair of oligonucleotide primers, INT14 and ST10.1, was designed to give in PCR a 600-bp fragment from the exon-intron boundary. This set of primers was sent to Genome Systems Inc. (St. Louis, MO) for screening of the mouse ES-129/SvJ I genomic library. Received bacterial artificial chromosome (BAC) clones were BACM-235K4, BACM-13L1, and BACM-151H13. BAC DNA was isolated with the K-100 Magnum kit (Genome Systems) and digested with restriction enzymes HindIII and EcoRI. Digested DNA was subjected to Southern blotting (21), and the presence of AMACR in BAC clones was detected with three probes made by PCR and random prime labeling to correspond to the 5Ј end of the mouse cAMACR with primers RaseI and RS4.0, the middle part with primers RU1.0 and K100, and the 3Ј end with primers K51 and SR19.1.
Sequencing of 3Ј end of the gene was done on the subcloned 2-kb HindIII fragment of clone BACM-13L1 with oligonucleotide primers SR19.1 and LO2. Upstream sequencing was done directly on the BACM-13L1 clone. Oligonucleotide primers used in sequencing were designed according to sequences obtained from previous reactions. Sequencing was performed on both DNA strands.
PCR reactions were used to determine the size of the introns. Intron 1 was amplified with primers ALE28 and ALE18, intron 2 with ALE22 and ALE17, intron 3 with RU3.0 and RS4.0, and intron 4 with primers K50 and ST 10.1. In addition, introns 1 and 2 were sequenced through. DNA sequencing was done with an automated ABI Prism 377 DNA sequencer (Perkin-Elmer).
Chromosomal Localization-Chromosomal localization of the gene was detected by Genome Systems, Inc. The fluorescent in situ hybridization was carried out by using the digoxigenin-labeled BACM-13L1 clone as probe. Hybridization was done to normal metaphase chromosomes derived from mouse embryo fibroblasts.
Southern Hybridization-Genomic DNA was extracted from kidney of BALB/c mouse with the SDS-proteinase K method (21), and 33 g of isolated DNA was digested with EcoRI or HindIII or with both enzymes. Southern blotting and hybridization was done on Hybond-N ϩ nylon membranes (Amersham Pharmacia Biotech) at 65°C (21, 22) using 32 P-labeled mouse cAMACR as a probe. The filter was washed twice before autoradiographing at Ϫ70°C for 24 h (Kodak X-Omat AR).
Northern Hybridization-For Northern hybridization, a mouse Multiple Tissue Northern TM blot (CLONTECH) nylon membrane was used, on which 2-g aliquots of poly(A ϩ ) RNA from various mouse tissues had been blotted. Hybridization was done according to the manufacturer's instructions in ExpressHyb TM hybridization solution with 32 P-labeled mouse cAMACR as probe. The hybridized fragments were autoradiographed at Ϫ70°C for 72 h.
Promotor Activity-Primers PR-3 and PR-4 were designed to amplify a 1670-bp-long piece of the 5Ј noncoding region of AMACR by PCR (Pfu polymerase; Stratagene, La Jolla, CA). The amplified fragment was cloned into the SmaI-digested pGL3-Basic reporter vector (Promega, Madison, WI). Correct orientation of the insert was ascertained by sequencing. HepG2 cells were obtained from the American Type Culture Collection (Manassas, VA) and maintained in Dulbecco's minimal essential medium containing penicillin (25 units/ml), streptomycin (25 units/ml), gentamicin (50 g/ml), 1% (v/v) nonessential amino acids, and 10% (v/v) fetal bovine serum. Cells were transfected by using LipofectAMINE-transfection reagent according to the manufacturer's instructions (Life Technologies Inc.). Briefly, 4 ϫ 10 5 cells were plated onto a 60-mm dish 24 h before transfection with a solution containing 4 g of luciferase reporter plasmid with amplified promoter or the same promoter region but in inverted orientation or 2 g of luciferase reporter plasmid without promoter or 2 g of pGL3-control vector (Promega). In addition, in all above-mentioned transfections, 0.6 g of pSV-␤-galactosidase control plasmid (Promega) was used as normalization vector. Cell extracts were prepared using reporter lysis buffer (Promega), and luciferase and ␤-galactosidase activities were assayed according to the manufacturer's instructions (Promega and CLON-TECH) with Labsystems Luminoscan RS luminometer (Labsystems, Helsinki, Finland). All luciferase values were normalized with ␤-galactosidase activity in the same extract.
Sequence Analysis-DNA sequences were analyzed with the DNASIS program (Hitachi Software Engineering Co., Yokohama, Japan). Transcription factor binding sites were searched for by using a MatInspector V2.2 data base (23).
Other Methods-Western blots with the rabbit antiserum against rat liver Amacr were done as described (2). Protein content was determined by the dye binding method of Bradford (24). The use of animals was approved by the University of Oulu Committee on animal experimentation RESULTS Subcellular Fractionation and Amacr Distribution-Mouse liver homogenate was subjected to subcellular fractionation on a Nycodenz gradient. As shown in Fig. 1 (lower panel), Amacr activity was found in three regions of different density. Two regions were found in the gradient, comigrating with the marker enzymes for peroxisomes (catalase) as well as for mitochondria (succinate dehydrogenase), and one region was in the fractions floating at the top of the gradient. Furthermore, when subjecting the fractions to Western blot analysis using rabbit antibodies against rat liver Amacr, the intensity of the visualized bands correlated with the distribution of the Amacr activity. The size of the visualized polypeptides was the same in each region (Fig. 1, upper  panel).
To study the subcellular distribution of the Amacr activity further, experiments were carried out with CHO fibroblasts. When control CHO fibroblasts were fractionated, a distribution of the Amacr activity similar to that in mouse liver was found (Fig. 2a). When the fractionation was carried out with CHO peroxisome-deficient mutant cells, which are devoid of peroxisomes, Amacr activity was still comigrating with the mitochondrial fractions and showed the same specific activity as in the mitochondrial fractions of the control CHO fibroblasts (Fig. 2b). Neither particle bound catalase activity was found in these cells, in agreement with earlier findings (13), nor particle bound Amacr activity in the fractions with densities corresponding to the peroxisome-containing fractions of the control cells.
Immunoelectron Microscopical Investigations-For further investigation of the subcellular distribution of Amacr, sections of mouse liver were subjected to immunoelectron microscopy with the antiserum against rat liver Amacr (2) using the protein A-gold labeling technique (Fig. 3). The gold particles were located in mitochondria and peroxisomes from control and clofibrate-treated mice (10 days) without any observed change in the labeling intensity. For control purposes, the tissue sections were treated with antibodies against rat 120-kDa ⌬ 2 -⌬ 4dienoyl-CoA reductase (12) and rat catalase for labeling mitochondria and peroxisomes, respectively. In both cases, the antibody recognized the expected compartment as shown in Fig. 3.
Immunopurification of Mitochondrial and Peroxisomal Mouse Liver Amacr-An affinity column was prepared with the rabbit antibodies against rat Amacr (2). When the soluble extract of ultrasound-treated mouse liver homogenate was applied to the column, all Amacr activity was bound to the column, and none could be detected in the flowthrough fractions (data not shown), documenting that the antibodies used recognize all soluble Amacr(s). Subsequently, peroxisomes and mitochondria were isolated from mouse liver on a self-generating Percoll gradient (Fig. 4a), and the soluble extracts were applied separately to the immunoaffinity column. More than 99% of the Amacr activities were retained on the column.
When the bound proteins from each of the experiments were eluted with acidic buffer and subjected to SDS-PAGE and Western blot analysis, protein staining of the SDS-PAGE gels showed one major protein band in each lane (Fig. 4b), at a position corresponding to the size of the purified Amacr protein (42 kDa). In Western blots, the immunologically cross-reactive bands showed the same size as that from mouse liver homogenate (not shown but see compare with Fig. 1). No Amacr activity could be detected in this eluate. When alkaline buffer instead of acidic buffer was used for elutions, SDS-PAGE and Western blot analysis revealed similar results, and also Amacr activity was detectable, albeit very low, (0.83 and 0.42% of the applied activity for mitochondrial and peroxisomal fractions, respectively). This showed that mitochondrial and peroxisomal Amacrs are equally well bound to and released from the affinity column but are rapidly inactivated under the harsh conditions employed. Similary, the cytosol (supernatant of the differential centrifugation at 12,000 ϫ g) was applied to and eluted from the affinity column with comparable results.
The immunoisolated proteins of the alkaline elution were, after SDS-PAGE, blotted on a polyvinylidene difluoride membrane, and the Amacr bands were excised and subjected to N-terminal amino acid sequencing. Each of the immunopurified samples from peroxisomes or mitochondria or cytosol gave the same single homogeneous N-terminal sequence: Val-Leu-Arg-Gly-Val-Arg-Val-Val-Asp-Leu-Ala-Gly-Leu-Ala-Pro (NAmacr).
cDNA Analysis-NAmacr was not found in the polypeptide encoded by the known mouse cAMACR (10). When searching the data banks with NAmacr, however, expressed sequence tag (EST) clones encoding the identical polypeptide were identified. These EST clones, among them clone AA085247, were labeled as originating from human beings. However, parallel work on the human Amacr in our laboratory had revealed that they were mislabeled and were in fact of mouse origin (11). The cDNA sequence of AA085247 contained an ATG embedded in a Kozak consensus sequence followed by a sequence encoding NAmacr and 24 additional base pairs. The last 10 bp of these matched with the 5Ј end of the published cAMACR (Ref. 10 and Fig. 5). Altogether, the previously known cAMACR was extended by 62 bp at the 5Ј end. Together with the initial Met and additional 21 new amino acid residues, the open reading frame of the revised cAMACR encodes a polypeptide of 381 amino acid residues with a predicted molecular mass of 41,718 Da.
The Western blot analysis of both rat and mouse Amacr demonstrated that they are of the same size (10). These data, together with the revised mouse cAMACR, suggest that also the earlier published cAM-ACR for rat liver racemase was not complete either. To investigate this, the reverse transcription-PCR with rat liver total RNA as template was carried out, using RRaceII for first strand synthesis and then HisR and polyT as primers, followed by nested PCR with M9R and polyT. A cDNA fragment of 570 bp was obtained; nucleotide sequencing of this revealed additional 246 bp, preceding the 5Ј end of the hitherto published rat liver cAMACR (10). The extended rat liver cAMACR includes an ATG as putative initial codon, followed by an open reading frame that corresponds to the one for mouse liver (Fig. 5). The identity between the rat and mouse liver Amacrs was 89 and 90.7% at the amino acid-and nucleotide-coding region level, respectively.
The open reading frame of rat liver cAMACR was cloned into the EcoRI sides of the pHIL-D2 expression vector, transformed into P. pastoris and overexpressed. The Amacr activity, measured with [2-3 H]pristanoyl-CoA as substrate, was 0.5 nmol ϫ min Ϫ1 ϫ mg Ϫ1 in the soluble supernatant of overexpressing cells but was below the detection limit (10 Ϫ3 nmol ϫ min Ϫ1 ϫ mg Ϫ1 ) in uninduced yeast cells. The immunoblotting analysis of the soluble fraction from overexpressing yeast cells and from rat liver homogenate showed that the detected polypeptide band was of the same size as that found in rat liver.
Searching for Mouse AMACR-Screening of the mouse genomic library ES-129/SvJI with the obtained random primed mouse cAMACR (corresponding to nucleotides 2-1503) (10) did not result in correct positive clones. The first exon-intron junction (later identified as exon 4/intron 4 junction) was found by PCR amplification with the primers ST10.1 and SR19.0 using genomic DNA extracted from BALB/c mouse was isolated and sequenced and was found to contain the sequence identical to the 3Ј end of cAMACR up to the poly(A ϩ ) tail. 766 nucleotides toward the 5Ј end, the sequence was interrupted by intronic sequences. This intron/exon junction was confirmed by direct sequencing of the BACM-13L1 clone from the 5Ј direction. Further direct sequencing of the BAC clone toward the 5Ј end revealed that AMACR contains three additional introns (Fig. 5). All exon-intron boundaries followed the GT-AG rule (Table II). The sizes of the introns 1-4 were 1.3, 1.2, 4, and 5 kb, respectively, as determined with PCR (Table II), and the AMACR gene spans a stretch of 13 kb in the mouse genome (Fig. 6). The characterized gene included with full match the sequence of EST clone AA085247. From the putative initial ATG toward the 5Ј direction, 1680 bp were sequenced. A GC-rich region was located at the positions Ϫ53 to Ϫ90 that can serve as an SP1-binding site (Fig. 7). No TATA box was present, but other consensus sequences for transcription factors including sites for GATA1, NF1, AP-1, AP-2, and AP-4 were found.
When the 245 bp preceding the start codon were compared with the 5Ј-UTR of rat Amacr cDNA, a fairly good similarity (81.4%) was found, assuming a 148-bp deletion in the rat sequence between bases Ϫ15 and Ϫ16 and four smaller deletions (altogether 43 bp) further upstream in the mouse sequence (Fig. 5).
Promotor Analysis-To test the functionality of the genomic fragment between the base pairs Ϫ1670 to Ϫ1 as a promotor, it was cloned and ligated to the 5Ј end of the luciferase gene in the pGL3-Basic reporter vector. Transient transformation of human HepG2 cells with this construct exhibited luciferase values that were 135 Ϯ 49-fold (mean Ϯ S.D.; n ϭ 10) higher than those obtained with the empty reporter construct, indicating that the fragment encompasses sequences that can function as promotor elements in intact cells.
Chromosomal Localization of the Mouse Amacr-To determine the chromosomal localization of the mouse AMACR, the BACM-13L1 clone was labeled and hybridized on metaphase chromosomes derived from mouse embryo fibroblasts. The identification of the mouse chromosomes was based on their G banding pattern. In addition, mouse chromosome 15 was identified with a probe specific for the telomeric region of that chromosome. Only double spot signals were considered as specific hybridization signals. Of 80 metaphase spreads, 73 showed the mouse racemase gene location on chromosome 15, region 15B1 (Fig. 8).
Southern and Northern Blotting-When genomic mouse liver DNA was digested with EcoRI or HindIII or both enzymes and the products were analyzed by Southern blot hybridization with the mouse cAMACR as probe, two fragments were detected in the case of EcoRI digestion with sizes of approximately 13 and 5 kb. Digestion with HindIII revealed five bands of approximate sizes of 6.2, 5.8, 4.0, 2.3, and 1.7 kb. The double digestion yielded five fragments with approximate sizes of 6.0, 4.0, 2.1, 1.7, and 0.9 kb (Fig. 9). This restriction fragment pattern and the restriction map constructed by the DNASIS program (Fig. 6) suggest that the size of AMACR is approximately 15 kb.
For Northern blotting, 2-g aliquots of mRNA, isolated from various mouse tissues (heart, brain, spleen, lung, liver, skeletal muscle, kidney, and testis), were hybridized with the mouse cAMACR as probe, as described under "Experimental Procedures." A signal corresponding to a size of 1.9 kb was detected in the samples from liver and kidney tissues (Fig. 10A). Northern hybridization with a ␤-actin probe showed that there were two signals of ␤-actin in heart and skeletal muscle, namely a 2-kb one and slightly smaller one of 1.6 -1.8 kb (25,26), as well as a weaker one in testis. The other tissues have one mRNA of approximately 2 kb (Fig. 10B).

DISCUSSION
The present work demonstrates that the mouse genome contains only one AMACR, encoding a polypeptide that can be targeted alternatively into mitochondria or peroxisomes, the subcellular compartments containing active Amacr. This statement is based on the combination of observations at the genomic and cDNA level and on the exclusion of other proteins than Amacr with Amacr activity in mouse liver.
Fluorescence in situ hybridization experiments performed on mouse metaphase chromosomes visualized only one region for AMACR (chromosome 15, region 15B1). The analysis of EcoRI/ HindIII digests of genomic DNA with Southern blotting gave an estimated gene size of 15 kb. This size is sufficient to accommodate the complete AMACR, whose size was about 13.5 kb, when estimated by direct sequencing and PCR analysis of the mouse genomic BACM-13L1 clone. These data indicate that the haploid mouse genome contains only one copy of AMACR.
By screening of mouse EST libraries and PCR analysis or screening of mouse liver cDNA libraries (10), one cAMACR could be identified, and no other sequences with significant similarities were found. Furthermore, Northern blot analysis of mouse liver and kidney RNA visualized only one signal of about 1.9 kb, in good agreement with the full size of the mouse cAMACR. The cAMACR identified in this study can be assumed to be complete because: (i) the genome fragment preceding the open reading frame (Ϫ1670 to Ϫ1) is a functional promotor, as shown by the luciferase reporter gene assay, (ii) when the 5Ј-flanking nucleotide sequence was translated in the frame, 22 stop-codons were encountered, and (iii) the identified putative initial ATG was embedded in a Kozak consensus sequence.
In accordance with the finding of the existence of one AM-ACR and Amacr mRNA, only one gene product was found. As shown by Western blot analysis of tissue homogenate, cytosol, and purified mitochondria and peroxisomes, in all cases only one protein band was visualized, corresponding to the same molecular mass. The N-terminal amino acid sequencing of the Amacr proteins, immunoisolated from cytosol and purified mitochondria and peroxisomes, resulted in NAmacr, which matched fully with the amino acid sequence predicted by the  cAMACR minus the leading methionine. This experiment, together with the genomic and cDNA analysis, showed that the Amacr polypeptides recognized by the anti-Amacr antibody in different subcellular compartments were the same. When applying mouse tissue extracts to the affinity column prepared with the anti-Amacr-antibody, the Amacr activity was completely removed from the flowthrough. This excludes the possibility of the existence of Amacr activities not detectable by the antibody.
The enzyme activities of acyl-CoA degradative pathways generally have multiple subcellular localization as exemplified by the mitochondrial and peroxisomal ␤-oxidative enzymes. The proteins of these different pathways are encoded by separate genes and are often presented as paralogs, even in the same compartment. The Amacr is exceptional not only among the acyl-CoA metabolizing gene products but also in general, because unlike other enzymes, which are modified when present in different compartments, the same Amacr can be targeted to two subcellular compartments. In most cases, proteins with different compartmentation were found to acquire different localization signals by alternative transcription or splicing or, in some cases, different translation start points or proteolytic modifications (for review see ref. 27). No indication for any of these processes was found for the Amacr. Another known dual locating mammalian protein metabolizing CoA thio esters in mitochondria and peroxisomes is the ⌬ 3,5 -⌬ 2,4 -dienoyl-CoA isomerase. In this case however, an N-terminal amino acid sequence is cleaved off upon mitochondrial targeting, whereas the peroxisomal targeting occurs via the C-terminally located peroxisomal targeting signal type 1 (PTS1). As a consequence, the mitochondrial polypeptide is some 4 kDa smaller than the peroxisomal isoform (17). An intriguing question is which factor determines the alternative targeting of Amacr either to mitochondria or to peroxisomes. Amacr contains the polypeptide -KANL at the C terminus, which has previously been shown to act as functional PTS1 in catalase (28). The helical wheel of the N terminus of Amacr shows the typical features of a mitochondrial targeting signal (MTS), but it contains a negatively charged residue (glutamate) at the position 11. Although uncommon in MTSs, a negative charge is sometimes accepted; the ␤ subunit of the human pyruvate dehydrogenase has even two glutamates in positions 10 and 14 (29). Conceivably, the Amacr MTS is only poorly recognized by the mitochondrial import machinery, allowing the folding of the protein in the cytoplasm. In this case, the C-terminal PTS1 will, after completion of synthesis, be recognized by the PTS1 receptor Pex5p, and the enzyme will be imported into the peroxisome. The quality of the MTS and the frequency of its recognition by the mitochondrial import system would then determine the relative distribution of the Amacr protein between the mitochondrial and peroxisomal compartments.
This model would also explain why in cells from patients with Zellweger syndrome, which are devoid of functional peroxisomes, Amacr activity is reduced to the 10 -20% normally found in mitochondria (2). Any Amacr protein not sequestered by the mitochondrial import system early in its synthesis could not be imported into mitochondria later but would remain in the cytosol to be degraded rapidly.
During the past few years, data have been emerging that provide an alternative route to the direct targeting of enzymes from the cytoplasm to their final destination. Evidence was presented that some proteins are imported into mitochondria and then re-exported and transferred to other compartments (for review see Ref. 30). It cannot be excluded that the Amacr protein follows a similar route. Because its MTS appears to be one of the few that are not cleaved inside the mitochondria, the sizes of the differently located proteins do not yield any information on the possible pathways taken. The transgenic mouse model being developed in future should also provide a possibility to study the influence of the different localization signals, after transfection with native and mutated cAMACRs.
The physiological role of Amacr in mammalian peroxisomes can be easily inferred. As indicated in the introduction section, it is required in the bile acid synthesis from THCA-CoA and the degradation of pristanoyl-CoA, processes occurring in peroxisomes. In contrast, the physiologic significance of the mitochondrial Amacr activity is still unclear. There is evidence that ␤-oxidation of branched chain fatty acids in peroxisomes does not go to completion but stops at 4,8-dimethylnonanoic acid, which, after conjugation with carnitine, is imported into mitochondria for complete degradation (31). Because phytanic and pristanic acids have the (R)-configuration at the inner methyl branch points (32) and the ␣-methylacyl-CoA dehydrogenases in mitochondria, like the peroxisomal oxidases, appear to be specific for the (2S)-enantiomers (14), the racemase would be required in mitochondria, too.
Additional physiological functions of Amacr in the metabolism of other, as yet unknown, endogenous substrates must also be considered. Patients with inborn defects of the ␤-oxidation pathway of branched chain fatty acids, e.g. of the peroxisomal multifunctional enzyme type 2, have, in addition to the expected hepatic dysfunction, severe neurological symptoms from birth (33,34). Obviously, there must be branched chain substrates of endogenous origin for this pathway that occur in all tissues. Conceivably, the racemase may be required for the degradation of at least some of them, in peroxisomes but possibly also in mitochondria. It is hoped that the transgenic mouse model, which is currently being developed, will provide some answers to these questions, too.