MCD Encodes Peroxisomal and Cytoplasmic Forms of Malonyl-CoA Decarboxylase and Is Mutated in Malonyl-CoA Decarboxylase Deficiency*

Malonyl-CoA decarboxylase (MCD) catalyzes the proton-consuming conversion of malonyl-CoA to acetyl-CoA and CO2. Although defects in MCD activity are associated with malonyl-CoA decarboxylase deficiency, a lethal disorder characterized by cardiomyopathy and developmental delay, the metabolic role of this enzyme in mammals is unknown. A computer-based search for novel peroxisomal proteins led to the identification of a candidate gene for human MCD, which encodes a protein with a canonical type-1 peroxisomal targeting signal of serine-lysine-leucineCOOH. We observed that recombinantMCD protein has high intrinsic malonyl-CoA decarboxylase activity and that a malonyl-CoA decarboxylase-deficient patient has a severe mutation in the MCD gene (c.947–948delTT), confirming that this gene encodes human MCD. Subcellular fractionation experiments revealed that MCD resides in both the cytoplasm and peroxisomes. Cytoplasmic MCD is positioned to play a role in the regulation of cytoplasmic malonyl-CoA abundance and, thus, of mitochondrial fatty acid uptake and oxidation. This hypothesis is supported by the fact that malonyl-CoA decarboxylase-deficient patients display a number of phenotypes that are reminiscent of mitochondrial fatty acid oxidation disorders. Additional support for this hypothesis comes from our observation that MCD mRNA is most abundant in cardiac and skeletal muscles, tissues in which cytoplasmic malonyl-CoA is a potent inhibitor of mitochondrial fatty acid oxidation and which derive significant amounts of energy from fatty acid oxidation. As for the role of peroxisomal MCD, we propose that this enzyme may be involved in degrading intraperoxisomal malonyl-CoA, which is generated by the peroxisomal β-oxidation of odd chain-length dicarboxylic fatty acids.

Malonyl-CoA decarboxylase activity (EC 4.1.1.9, Scheme 1) has been described in a wide array of organisms, including prokaryotes, birds, and mammals (1)(2)(3). However, the physiological role of this enzyme is somewhat unclear. The only eu-karyotic malonyl-CoA decarboxylase (MCD) 1 gene that has been cloned was goose (Anser anser), which expresses two MCD transcripts from a single gene (4). The longer of the two transcripts is ubiquitously expressed and encodes a mitochondrial enzyme, whereas the shorter form has a more restricted pattern of expression and encodes a cytosolic enzyme. The mitochondrial localization of goose MCD suggested that this enzyme may function in removing intramitochondrial malonyl-CoA that is produced by the adventitious activity of propionyl-CoA carboxylase on acetyl-CoA. As for the cytosolic form of the enzyme, it was hypothesized to participate in the synthesis of methyl branched-chain fatty acids (5).
Although these hypotheses may explain the localization and function of goose MCD, metabolic studies raise the possibility that there are additional physiologically relevant roles for this enzyme. In mammals, cytoplasmic malonyl-CoA is a potent inhibitor of carnitine palmitoyltransferase (CPT1) and, thus, of mitochondrial fatty acid oxidation (6). In adipogenic tissues such as liver and white fat, acetyl-CoA carboxylase produces cytoplasmic malonyl-CoA under fed conditions as a precursor for fatty acid synthesis, and fatty acid synthase consumes this cytoplasmic pool of malonyl-CoA. However, nonadipogenic tissues such as cardiac and skeletal muscle also produce significant quantities of cytoplasmic malonyl-CoA from acetyl-CoA carboxylase (7), and the inhibitory effects of malonyl-CoA on CPT1 are roughly 100-fold greater in these tissues than in liver (6). The synthesis of cytoplasmic malonyl-CoA by acetyl-CoA carboxylase in muscle is tightly regulated by the AMP-dependent protein kinase (8), which inactivates acetyl-CoA carboxylase and blocks malonyl-CoA synthesis under conditions which require that muscle use fatty acids as their primary energy source (i.e. fasted and/or exercised states). However, there is currently no model for how these tissues degrade the cytoplasmic malonyl-CoA that is produced by acetyl-CoA carboxylase, even though this process is critical for the activation of mitochondrial fatty acid oxidation. One possible model is that a cytoplasmic form of MCD could act to remove cytoplasmic malonyl-CoA under conditions that inhibit acetyl-CoA carboxylase, thereby allowing mitochondrial fatty acid oxidation to resume.
An additional role for MCD could be in the oxidation of dicarboxylic fatty acids (DFAs). In contrast to long and medium chain fatty acids, which are oxidized primarily in the mitochon-dria, DFAs are oxidized only in peroxisomes (9). DFA oxidation is poorly understood, and it is not clear whether peroxisomes degrade DFAs completely to malonyl-CoA (for odd chain-length DFAs) and oxalyl-CoA (for even chain-length DFAs). However, if the ␤-oxidation of DFAs in peroxisomes is complete, a peroxisomal form of MCD could function to eliminate this metabolic end-product of odd chain-length DFA oxidation.
A key role for malonyl-CoA decarboxylase in mammalian metabolism is suggested by the severe phenotypes of patients who lack this enzyme activity. Malonyl-CoA decarboxylase deficiency, also known as malonic aciduria, is a genetic disorder characterized by developmental delay, cardiomyopathy, mental retardation, and in its more severe forms, neonatal death (10 -14). These patients have several phenotypes that are reminiscent of mitochondrial fatty acid oxidation deficiencies, including diet-induced and infection-induced vomiting, seizures, hypoglycemia, and organic aciduria, as well as cardiomyopathy. Here we report the identification of a novel human gene encoding malonyl-CoA decarboxylase. Its role in malonic aciduria, its pattern of expression, the subcellular distribution of its product, and its possible roles in cellular metabolism are discussed.

EXPERIMENTAL PROCEDURES
Cloning of Human MCD-We searched the ExPASy TrEMBL/EMBL data bases for all proteins terminating in possible forms of the PTS1. This data base scan identified goose malonyl-CoA decarboxylase as a candidate peroxisomal protein because it ends in the canonical PTS1 of serine-lysine-leucine COOH (15). To determine whether the human gene also encoded a protein with a PTS1, and thus might encode a peroxisomal protein, we searched for human homologues of goose MCD. The goose MCD protein sequence was used as query in a BLAST search of the human data base of expressed sequence tags (ESTs) for any genes that had the potential to encode proteins with high sequence similarity to goose MCD. Multiple human ESTs were identified, all of which appeared to represent a single human gene. The longest of these at the time, GenBank TM accession number N52074, corresponded to the cDNA clone 282566 and was obtained from Genome Systems (St. Louis, MO). This clone was sequenced in its entirety. Based on its similarity to the goose isoforms of MCD, this human MCD clone lacked the first 170 codons relative to goose cytosolic MCD and the first 230 codons relative to goose mitochondrial MCD.
To obtain additional sequence data for the 5Ј end of the human MCD cDNA, we used an MCD-specific primer (5Ј-CTCCTTGACGACTCGCT-TTATGAGG-3Ј) and a library vector-specific primer (5Ј-ATACCATTA-CAATGGATG-3Ј) to amplify 5Ј fragments of the MCD gene from a human heart cDNA library (MCD expression is higher in muscle than in any other tissues). The largest band that was generated was approximately 900-bp long. Subcloning and sequencing of this clone revealed that it overlapped with our truncated MCD cDNA clone and added about 600 bp to the 5Ј end of the MCD sequence. The assembled 2,121-bp MCD sequence contained a 95-bp 5Ј untranslated region, a 1,362-bp open reading frame (ORF), and a 664-bp 3Ј untranslated region. Its deduced protein product of 454 amino acids initiated at the same relative position as the goose cytosolic MCD (5). Our most recent data base searches revealed the presence of 34 ESTs for human MCD. None of these are as long as our assembled "full-length" MCD cDNA. However, four of these (GenBank TM accession numbers AI1284158, AI379572, AI123407, and AA015957) do appear to contain the entire MCD ORF and start at positions Ϫ50, Ϫ32, Ϫ31, and Ϫ1, respectively, relative to the A of the putative initiator ATG in the sequence we have assembled. We also identified a human genomic DNA clone for the MCD locus by screening a bacterial artificial chromosome library of human genomic DNAs with a 190-bp probe from the 5Ј end of the MCD cDNA. An 8-kilobase XbaI fragment containing the 5Ј end of the MCD gene was generated from the MCD bacterial artificial chromosome clone, and the structure of the MCD gene upstream of the 5Ј end of the MCD cDNA was determined by sequence analysis of this DNA fragment. An oligonucleotide corresponding to the 22 bp immediately upstream of the 5Ј end of the MCD cDNA (5Ј-ACCATGCGAGGCTTCGGGGCCAGGCTTG-3Ј) was used in PCR reactions to test for the expression of a longer cDNA product that might contain a second, upstream ATG. 5Ј-RACE experiments were also performed but failed to generate any MCD cDNA clones longer than the cDNA reported here. Additional BLAST sequence searches were used to identify a putative Caenorhabditis elegans form of MCD (GenBank accession number Z46242).
Plasmids-A portion of the MCD ORF encoding amino acids 187 to 451 was amplified using the primers 5Ј-CCAGTCGACGGACATGAAG-CGCCGCGTTGGGCC-3Ј and 5Ј-CCAAGCGGCCGCTCAAAGCTTCCA-GTTCTTTTGAAACTGGGCCACTAGGC-3Ј and the cDNA clone 282566 as template. The resulting PCR product was cleaved with SalI and NotI and cloned between the SalI and NotI sites of pMBP, a variant of pMALc2 (New England Biolabs) that contains unique SalI and NotI sites for insertion of fragments downstream of the maltose-binding protein (MBP) ORF (16). The resulting plasmid, pMBP-MCD/187-451, was the source for a recombinant MCD fragment that was used for immunization of rabbits and generation of affinity purified anti-MCD antibodies. Another set of primers (5Ј-CCAGTCGACGATGGACGAGC-TGCTGCGCCG-3Ј and 5Ј-CCAAGCGGCCGCTCAAAGCTTCGAGTTC-TTTTGAAACTGGGCCACTAGGC-3Ј) was used to amplify the entire MCD ORF in the correct reading frame for fusion to MBP in pMBP. The template for this reaction was a human heart cDNA library. A product of the expected length was generated, cleaved with SalI and NotI and cloned between the SalI and NotI sites of pMBP. The insert in this plasmid (pMBP-MCD) was sequenced in its entirety, confirming the full-length sequence assembled from the overlapping MCD cDNAs described in the cDNA cloning section above. This plasmid was used for generating full-length recombinant MCD that was used for enzyme assays. To create pcDNA3-MCD, the EcoRI-NotI fragment containing the MCD cDNA was excised from pMBP-MCD and inserted between the EcoRI and NotI sites of pcDNA3 (Invitrogen, San Diego).
Protein Production and Antibody Production-Recombinant MCD lacking its first 186 amino acids and last 3 amino acids was expressed in fusion with MBP from an Escherichia coli strain DH10B (17) carrying the plasmid pMBP-MCD/187-451. Full-length MCD was expressed in fusion with MBP from DH10B cells carrying the plasmid pMBP-MCD. MBP-LacZ was expressed from DH10B cells carrying pMALc2. All three proteins were expressed and purified as described by Geisbrecht et al. (16), with the exception that cells were induced overnight at room temperature. MBP-MCD/187-451 was used for the immunization of rabbits and affinity purification of anti-MCD antibodies, as described (18). Recombinant MBP-MCD and MBP-LacZ were used for the analysis of MCD enzyme activity.
Enzyme Assays -Malonyl-CoA decarboxylase activity was assayed by coupling the MCD reaction to malate dehydrogenase and citrate synthase and by measuring the production of NADH spectrophotometrically, as described by Kim and Kolattukudy (2). Only initial rates were used in the calculation of K m and v max . Assays for catalase (19) and succinate dehydrogenase (20) have been described.
Preparation of Lysates, in Vitro Translation, Subcellular Fractionation, and Immunoblots-Whole cell protein extracts for use in immunoblotting were prepared from cultured human HepG2 cells, a hepatoblastoma cell line, and 5756T cells, a human skin fibroblast cell line. A nearly confluent flask (150 cm 2 ) of cells was resuspended in a solution of 30 mM Tris, 1 mM EDTA, 1% SDS, pH 8.0. The suspension was incubated for 10 min on ice, then spun at 15,000 ϫ g for 10 min. The supernatant containing soluble protein was removed and resuspended in SDS-polyacrylamide gel electrophoresis buffer. Coupled in vitro transcription and translation of human MCD was performed with TNT in vitro translation reagents according to the manufacturer's instructions (Promega, Madison, WI) and the pcDNA3-MCD plasmid. Subcellular fractionation of rat liver was performed as described (21).
Transcript Analysis and Mutation Detection-Northern blot analysis was performed using standard protocols and human multitissue Northern blots from CLONTECH (Palo Alto, CA). Human fibroblast RNA was SCHEME 1. Reaction catalyzed by malonyl-CoA decarboxylase.

RESULTS
Cloning the Human MCD Gene-We previously described the development and application of context sensitive motif scanning for the in silico identification of novel peroxisomal proteins in the yeast Saccharomyces cerevisiae (24). We applied a similar approach to identify candidate peroxisomal proteins in data bases from higher eukaryotes. One of the many candidates that were identified was the goose malonyl-CoA decarboxylase (A. anser MCD). This enzyme contains a perfect match to the consensus sequence for the type-1 peroxisomal targeting signal (PTS1), serine-lysine-leucine COOH (15) and suggested that MCD might contribute to peroxisomal metabolic processes. However, it seemed that an analysis of human MCD would have greater general relevance, because there is a larger body of work on the physiological effects of malonyl-CoA in mammalian tissues (6) and because malonyl-CoA decarboxylase deficiency is associated with defects in this enzyme (10 -14).
We used a computer-based approach to identify the human malonyl-CoA decarboxylase gene. The BLAST algorithm was used to scan the data base of human expressed sequence tags for cDNAs capable of encoding proteins similar to goose malonyl-CoA decarboxylase. Multiple overlapping ESTs corresponding to a single gene were identified. The cDNA clone that appeared to have the longest 5Ј end was obtained from a commercial vendor and sequenced in its entirety. This clone appeared to be missing several hundred base pairs from the 5Ј end. Additional MCD cDNA clones containing another 600 bp at the 5Ј end were obtained from a human heart cDNA library, allowing us to assemble an apparent full-length cDNA for human MCD.
The compiled human MCD cDNA sequence ( Fig. 1) is 2,121-bp long and contains a 1,362-bp open reading frame. The presumptive initiator ATG has a good match to the consensus sequence for high efficiency translation initiation (25,26), particularly because it has purines at both the Ϫ3 and ϩ4 positions, relative to the A of the ATG. The deduced protein product is 454-amino acids long, has a predicted molecular mass of approximately 50 kDa, and starts at exactly the same relative position as the goose cytoplasmic MCD (Fig. 2). Furthermore, it also contains the canonical PTS1 of serine-lysine-leucine COOH . We also identified partial cDNA clones for the mouse and rat forms of MCD, and they also encode proteins that contain a PTS1 (data not shown). Additional data base searches led to the identification of the putative C. elegans MCD, which does not contain a peroxisomal targeting signal-like sequence and is considerably shorter than the vertebrate proteins at its N ter-minus (Fig. 2).
Additional searches of the human EST data base identified 34 MCD cDNA clones, 4 of which were deposited only recently and contained the entire MCD ORF. However, none were as long as the MCD cDNA described in this report. 5Ј-RACE also failed to provide evidence for a longer MCD transcript. Knowledge of gene structure can help in the analysis of gene transcripts, and we therefore identified an MCD genomic DNA bacterial artificial chromosome clone. A fragment of this clone that hybridized to the 5Ј end of the MCD cDNA clone was obtained, and sequence analysis of this clone revealed the presence of an in-frame ATG just 22-bp upstream of the 5Ј end of the MCD cDNA, as well as the absence of splice acceptor sites between this ATG and the ATG we designated as the beginning of the MCD open reading frame (Fig. 3).
Any transcripts originating upstream of this ATG would have the potential to encode a longer form of human MCD, and we used PCR techniques to search for such transcripts. We prepared first strand MCD cDNA from human fibroblast mRNA and used this as template in PCR reactions containing either of two antisense MCD oligonucleotides (MCD-RACE.1 and MCD-RACE.2) and the MCD-hypoATG oligonucleotide, which spans this hypothetical upstream ATG. No detectable products were generated from these reactions even though (a) these same two combinations of primers amplified a fragment of the correct size from MCD genomic DNA and (b) the same cDNA sample and antisense oligonucleotides could be used to amplify a fragment of the correct size in PCR reactions using a 5Ј primer (MCDstart), which spanned the ATG that we list at position ϩ1 of the cDNA sequence. Similar results were obtained when these various primer pairs were used with human heart cDNA as the template. Given the sensitivity of PCR detection techniques, these data indicate that MCD transcripts containing the hypothetical upstream ATG are either of very low abundance or do not exist, at least in human fibroblasts and heart tissue.
Although we find no evidence for MCD transcripts that contain the hypothetical upstream ATG, it is useful to consider whether an mRNA that contained this sequence would encode a protein analogous to goose mitochondrial MCD. We think that this is unlikely for two reasons. First, this putative upstream ATG is followed by a pyrimidine (Fig. 3), which lessens the probability that it would serve as an efficient initiator codon were it present in a mammalian mRNA (the ϩ4 position is almost always a purine in highly expressed transcripts (25,26)). Even more importantly, the region between the hypothetical upstream ATG and the ATG at position ϩ1 of the cDNA clone encodes a peptide sequence that lacks features of a mitochondrial leader sequence and shares only slight similarity to the N-terminal mitochondrial targeting signal of goose mitochondrial MCD.
As an independent test of whether the cDNA reported here is capable of encoding the full-length MCD protein we compared the mobility of endogenously synthesized MCD with that of MCD synthesized in vitro from the MCD cDNA clone. Affinity purified anti-MCD antibodies were generated and tested by immunoblot analysis of total cellular protein extracts from human fibroblasts and human hepatoblastoma cells. These antibodies detected a single polypeptide of approximately 50 kDa in both cell types (Fig. 4A), indicating that they are specific for MCD. We then synthesized MCD in vitro in a rabbit reticulocyte lysate and used immunoblot analysis to compare its mobility with that of endogenously synthesized human fibroblast MCD. A single protein was detected in each sample and their mobilities were indistinguishable from one another (Fig.  4B). Control experiments confirmed that the level of rabbit MCD present in the in vitro translation lysates was below the limit of detection. Taken together, these various lines of evidence suggest that the MCD present in human fibroblasts corresponds to the product of the MCD cDNA clone.
Recombinant MCD Has Malonyl-CoA Decarboxylase Activity-To test the hypothesis that the gene we had identified encoded human malonyl-CoA decarboxylase, we expressed the entire human MCD protein in bacteria as a fusion with MBP.
The recombinant MBP-MCD fusion protein was purified by affinity chromatography on an amylose resin. Assessment of MBP-MCD purity by SDS-polyacrylamide gel electrophoresis showed a predominant band at 90 kDa (Fig. 5), the size predicted for a fusion containing MBP (42 kDa) and human MCD (46 kDa). Purified recombinant MBP-MCD and purified MBP-LacZ (expressed and purified by the identical protocol and from the same strain of E. coli) were assayed for their ability to convert malonyl-CoA to acetyl-CoA. MBP-MCD showed significant malonyl-CoA decarboxylase activity, with a specific activity of 3 units/mg and a K m of 220 M for malonyl-CoA (Fig. 5). The MBP-LacZ protein lacked activity altogether, demonstrating that the activity of the MBP-MCD fusion protein was intrinsic to the portion derived from MCD and that E. coli MCD does not co-purify with MBP fusion proteins on amylose resin. Although human MCD displayed a high K m for malonyl-CoA, it is only slightly higher than the K m of the goose enzyme (100 M) (2).
MCD Is Bimodally Distributed to the Cytoplasm and to Peroxisomes-As noted earlier, human, mouse, and rat MCD all end in a PTS1. To test whether mammalian MCD is actually associated with peroxisomes, we prepared a post-nuclear supernatant from a rat liver homogenate and then fractionated this sample by Nycodenz density gradient centrifugation. Equal amounts of each fraction were assayed for marker enzymes of peroxisomes (catalase) and mitochondria (succinate dehydrogenase), as well as by immunoblot with affinity purified anti-MCD antibodies (Fig. 6). MCD was detected in perox- isomal fractions but was also present at the top of the gradient in the cytosolic fractions. Furthermore, the amount of MCD at the top of the gradient exceeds the amount of peroxisomal enzymes released by homogenization and fractionation, as estimated here by the levels of catalase in these low density fractions. These results suggest that MCD is bimodally distributed to both the peroxisome and cytoplasm, with most of the protein localized to the cytoplasm. It should be noted that we did not detect MCD in the mitochondrial fractions. This last result is consistent with our inability to detect a longer human MCD transcript that would encode a protein similar to the goose mitochondrial MCD. Similar results were obtained from the analysis of the human HepG2 liver cells (data not shown).
Malonyl-CoA Decarboxylase Is Highly Expressed in Muscle-We also analyzed the expression of MCD mRNA in different human tissues. Multitissue Northern blots containing poly(A) ϩ RNA from 16 different tissues were hybridized with a radiolabeled MCD-specific probe (Fig. 7). A 2.3-kilobase MCD transcript was detected in all tissues that we examined. However, the abundance of the MCD mRNA appeared to vary considerably, with extremely strong expression in cardiac and skeletal muscle. The levels of MCD mRNA were much lower in liver, kidney, and pancreas, and low but detectable in all other tissues. The 2.3-kilobase size of the MCD mRNA is consistent with the 2,121-bp size of the MCD cDNA, given that the average length of the poly(A) tract is 100 -200 bp. It is interesting to note that MCD expression is highest in cardiac and skeletal muscle, the two tissues that have the greatest dependence on fatty acids as an energy source and express forms of CPT1 that are extremely sensitive to inhibition by malonyl-CoA.

MCD Is Mutated in Malonyl-CoA Decarboxylase Deficiency and Maps to the Long Arm of Chromosome 16 -Previous stud-
ies have reported that loss of malonyl-CoA decarboxylase activity is the cause of malonic aciduria (10 -14). To test whether mutations in MCD might be responsible for loss of malonyl-CoA decarboxylase activity in these patients, we analyzed the sequence of the MCD gene from a severely affected malonic aciduria patient, MA002 (14). A skin fibroblast cell line was derived from MA002, RNA was extracted from this cell line, and overlapping MCD cDNA fragments were generated by PCR using this cDNA as template. As expected, we identified a mutation in the MCD cDNA from this patient, a deletion of two T residues at position 947 and 948 of the cDNA (data not shown). This mutation is referred to as c.947-948delTT and is expected to encode a truncated protein lacking its C-terminal 142 amino acids, roughly one-third of the protein. The C-terminal third of human MCD shares significant amino acid sequence similarity with both the goose and worm forms of MCD (Fig. 2), making it extremely unlikely that a protein lacking these amino acids would retain MCD activity.
We also assessed the chromosomal location of the human MCD gene. A sequence tagged site, STS WI-11775, was generated from the 3Ј untranslated region of the human MCD cDNA and localized by radiation hybrid mapping between the markers D16S422 and D16S402 on the long arm of chromosome 16. This localization of the MCD gene is consistent with the autosomal inheritance of malonic aciduria. DISCUSSION We recently described a computer-based search for novel peroxisomal proteins in the yeast S. cerevisiae (24). We have extended these studies to the analysis of sequence data bases of higher eukaryotes. 2 One of the candidate peroxisomal proteins that was identified in this latter search was goose malonyl-CoA decarboxylase, which contains a PTS1 at its C terminus. Loss of malonyl-CoA decarboxylase activity has been implicated in human disease (10 -14), and we pursued the identification and characterization of the human MCD gene to improve our understanding of peroxisome function and to resolve the molecular basis of malonyl-CoA decarboxylase deficiency. Computer- based approaches led to the identification of a single human gene that encodes a protein with high sequence identity (70%) with goose MCD and contains a PTS1, serine-lysineleucine COOH (15). This gene encodes a protein with intrinsic MCD activity and is mutated in a patient with malonyl-CoA decarboxylase deficiency (malonic aciduria), indicating that it is the human MCD gene.
Malonyl-CoA decarboxylase has been suggested to be a mitochondrial enzyme (27) but its metabolic role in mitochondria has never been clearly established. The identification of human MCD provided us with an opportunity to test whether the MCD product was localized to mitochondria and/or whether it resided in other cellular compartments. Affinity purified anti-MCD antibodies were generated and used to localize MCD. We detected both peroxisomal and cytoplasmic isoforms of MCD in both rat and human liver cells. Surprisingly, we did not detect any MCD protein in mitochondrial fractions. Similar results were observed in human cells (data not shown). Given that the functions of mammalian MCD must be consistent with its subcellular distribution, it is interesting to consider possible roles for the cytoplasmic and peroxisomal forms of human MCD.
In adipogenic tissues, cytoplasmic malonyl-CoA is produced as a precursor for fatty acid synthesis. In this context, the inhibition of CPT1 by malonyl-CoA serves to prevent a futile cycle of fatty acid synthesis in the cytoplasm and fatty acid oxidation in the mitochondrion. However, the fact that acetyl-CoA carboxylase is expressed in virtually all cells, and thus, generates malonyl-CoA in many tissues that do not make fatty acids, suggests that malonyl-CoA is a general regulator of fatty acid oxidation in all tissues. This hypothesis is supported by the fact that the inhibitory effects of malonyl-CoA on CPT1 are 100 times greater in muscle than in liver (6). Not surprisingly, acetyl-CoA carboxylase is a tightly regulated enzyme and is inactivated by the AMP-dependent protein kinase when the ATP/AMP ratio falls (8,28). This inhibitory regulation of acetyl-CoA carboxylase is thought to lower the steady-state levels of malonyl-CoA and lead to activation of CPT1 and mitochondrial fatty acid oxidation under conditions that require muscle tissue to utilize fatty acids rather than glucose as their primary energy source. Although acetyl-CoA carboxylase plays a key role in this regulatory event, the degradation of cytoplasmic malonyl-CoA is equally important. Prior studies have been unable to explain how cells rid themselves of cytoplasmic malonyl-CoA, because MCD has been thought to be an exclusively mitochondrial enzyme and because malonyl-CoA cannot cross the mitochondrial membrane. Our observation that there is a significant pool of cytoplasmic MCD suggests that this enzyme could contribute to the regulation of cytoplasmic malonyl-CoA levels and mitochondrial fatty acid oxidation. The high level of MCD expression in cardiac and skeletal muscle, tissues that require fatty acids for much of their energy and are extremely sensitive to the effects of malonyl-CoA, lends support to this hypothesis. It will be interesting to determine whether the activity of cytoplasmic MCD is regulated in a manner complementary to that of acetyl-CoA carboxylase.
The peroxisomal form of MCD is likely to play a different role in cellular metabolism than its cytoplasmic counterpart. Currently, there is no direct evidence that malonyl-CoA is produced inside peroxisomes and, thus, there is no bona fide role for MCD in peroxisomal metabolic processes. However, the ␤-oxidation of DFAs occurs exclusively in peroxisomes (9), and malonyl-CoA is the ultimate ␤-oxidation product of odd chainlength DFAs. Peroxisomal MCD may catalyze the final step in the oxidation of these compounds, converting intraperoxisomal malonyl-CoA to CO 2 and acetyl-CoA. As with the acetyl-CoA that is produced by peroxisomal fatty acid ␤-oxidation (29), the acetyl-CoA produced by MCD could be converted to acetylcarnitine and exported for subsequent use in the cytoplasm or mitochondria.
Although previous studies have implicated defects in MCD as the cause of malonyl-CoA decarboxylase deficiency (10 -14), the identification of the MCD gene allowed us to test whether mutations in this gene are indeed the cause of this disease. We sequenced the MCD gene from a severely affected malonic aciduria patient (MA002 (14)). The fact that this patient had a frameshift mutation (c.947-948delTT) that effectively deleted the C-terminal one-third of MCD strongly suggests that mutations in MCD are the cause of this disease.
The demonstration that MCD is the gene responsible for malonyl-CoA decarboxylase deficiency allows us to consider the phe-FIG. 6. Subcellular localization of MCD. Rat liver was homogenized and a post-nuclear supernatant was subjected to differential centrifugation on a 15-40% Nycodenz gradient. Each gradient fraction was assayed for succinate dehydrogenase (white bars) and catalase activities (shaded bars) (upper panel). Equal quantities of each fraction were separated by SDS-polyacrylamide gel electrophoresis, transferred to membranes, and probed with purified antibodies to human MCD (lower panel). notypes of malonyl-CoA decarboxylase deficiency patients in the context of our model for MCD function. The hypothesis that cytoplasmic MCD plays a significant role in degrading cytoplasmic malonyl-CoA suggests that its loss would result in elevated malonyl-CoA levels and inappropriate inhibition of mitochondrial fatty acid oxidation. These effects may be most severe in muscle, the tissue with greatest MCD expression, highest sensitivity of CPT1 to malonyl-CoA (6), and no other means for removing cytoplasmic malonyl-CoA. Case reports of malonic aciduria patients support this hypothesis (10 -14). These patients display cardiomyopathy and diet-induced and infection-induced seizures, vomiting, hypoglycemia, and organic aciduria, an array of phenotypes that are reminiscent of mitochondrial fatty acid oxidation disorders (30). As for the peroxisomal form of MCD, we do not know how its loss may impact the phenotypes of malonyl-CoA decarboxylase deficiency patients. We might predict that loss of this enzyme would result in accumulation of dicarboxylic fatty acids. Although these patients display a marked organic aciduria, with accumulation of malonic, succinic, glutaric, suberic, and adipic acids, this organic aciduria is also a hallmark of patients with defects in mitochondrial fatty acid oxidation deficiencies. Thus, we cannot conclude from the clinical data whether there is or is not a general impairment of peroxisomal DFA oxidation. Furthermore, there is no a priori basis for concluding that a defect in peroxisomal DFA oxidation would contribute to human disease.
The clinical similarities between malonyl-CoA decarboxylase deficiency patients and patients with classical mitochondrial fatty acid oxidation disorders has been noted previously (10 -14). However, the report that MCD is an exclusively mitochondrial enzyme (27) has become an accepted tenet of the field and may have precluded a coherent explanation for MCD function and dysfunction in metabolism and disease. Our localization of MCD to the cytoplasm and peroxisomes does not exclude the possibility that there may be some mitochondrial MCD but does indicate that there may be novel metabolic roles for MCD in cellular metabolism. In addition, our results have provided new insight into the molecular and metabolic basis of malonyl-CoA decarboxylase deficiency. We propose here that the phenotypes of malonic aciduria patients are chiefly caused by the loss of cytoplasmic MCD and the resulting dysregulation of mitochondrial fatty acid oxidation. Further investigations should help us determine the mechanisms by which MCD, together with acetyl-CoA carboxylase, controls steady-state levels of cytoplasmic malonyl-CoA and mitochondrial fatty acid oxidation.