Sequence of the Sodium Ion Pump Methylmalonyl-CoA Decarboxylase from Veillonella parvula*

The genes encoding methylmalonyl-CoA decarbox- ylase from Veillonellaparvula were cloned on plasmids using oligonucleotides derived from N-terminal amino acid sequences as specific probes. The entire DNA sequence of the methylmalonyl-CoA decarboxylase genes together with upstream and downstream regions was determined. The genes encoding subunits a (mmdA), 6 (mmdD), t (mmdE), y (mmdC), and B (mmdB) of the decarboxylase were clustered on the chromosome in the given order. The previously unnoted t-chain (Mr 5,888) was clearly shown to be a subunit of the decar- boxylase by correspondence of the N-terminal amino acid sequence with that deduced from the DNA se- quence of mmdE. The a-subunit was 60% identical with the carboxyltransferase domain of rat liver pro-pionyl-CoA carboxylase, the &subunit showed 61% sequence identity with the &subunit of oxaloacetate decarboxylase from Klebsiella pneumoniae, and the biotin-containing y-subunit was 29-39% identical with biotin-domains of other biotin enzymes. The 6-subunit of methylmalonyl-CoA decarboxylase and the y-subunit of oxaloacetate decarboxylase did not show significant sequence homology. The gross structure of both proteins, however, was similar, consisting of a hydrophobic membrane anchor near the N terminus, a proline/alanine linker,

Sequence of the Sodium Ion Pump Methylmalonyl-CoA Decarboxylase from Veillonella parvula* (Received for publication, June 7, 1993, and in revised form, July 23, 1993) Jon B. Huder and Peter DimrothS From the Mikrobiologisches Institut der Eidgenossischen Technischen Hochschule, ETH-Zentrum, CH-8092 Zurich, Switzerland The genes encoding methylmalonyl-CoA decarboxylase from Veillonellaparvula were cloned on plasmids using oligonucleotides derived from N-terminal amino acid sequences as specific probes. The entire DNA sequence of the methylmalonyl-CoA decarboxylase genes together with upstream and downstream regions was determined. The genes encoding subunits a (mmdA), 6 (mmdD), t (mmdE), y (mmdC), and B (mmdB) of the decarboxylase were clustered on the chromosome in the given order. The previously unnoted t-chain (Mr 5,888) was clearly shown to be a subunit of the decarboxylase by correspondence of the N-terminal amino acid sequence with that deduced from the DNA sequence of mmdE. The a-subunit was 60% identical with the carboxyltransferase domain of rat liver propionyl-CoA carboxylase, the &subunit showed 61% sequence identity with the &subunit of oxaloacetate decarboxylase from Klebsiella pneumoniae, and the biotin-containing y-subunit was 29-39% identical with biotin-domains of other biotin enzymes. The 6subunit of methylmalonyl-CoA decarboxylase and the y-subunit of oxaloacetate decarboxylase did not show significant sequence homology. The gross structure of both proteins, however, was similar, consisting of a hydrophobic membrane anchor near the N terminus, a proline/alanine linker, and a remarkable accumulation of charged amino acids in the C-terminal part. The sequence of the small t-subunit could be aligned to the C-terminal region of the &subunit downstream of the prolinelalanine linker, where the two subunits were 47% identical. Of considerable interest for the mechanism of Na+ transport are the long stretches of complete sequence identity between the hydrophobic &subunits of methylmalonyl-CoA decarboxylase and oxaloacetate decarboxylase and the presence of two conserved aspartic acid residues within putative membrane-spanning helices.
In Veillonella parvula (previously named Veillonella alcalescens) the free energy of methylmalonyl-CoA decarboxylation is conserved by conversion into an electrochemical gradient of Na' ions (1,2). The responsible methylmalonyl-CoA decarboxylase is a membrane-bound biotin-containing Na' pump (3,4). The enzyme shares a number of properties with the other members of the family of Na+-transporting decar-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
to the GenBankTM/EMBL Data Bank with accession numberfs) The nucleotide sequencefs) reported in thispaper has been submitted L22208. 3 To whom correspondence should be addressed. boxylases: oxaloacetate decarboxylase and glutaconyl-CoA decarboxylase (1, 2). All these decarboxylases contain a peripheral membrane-bound subunit of M, 60,000-65,000 that catalyzes a carboxyl transfer from the substrate to the prosthetic biotin group. In methylmalonyl-CoA decarboxylase (3) and glutaconyl-CoA decarboxylase ( 5 ) , the biotin is bound to a separate biotin carrier protein subunit of apparent M , 18,000-19,000, while oxaloacetate decarboxylase contains the biotin on the C-terminal domain of the a-subunit (6,7). In addition, all Na' transport decarboxylases contain a highly hydrophobic subunit (@) that migrates as a polypeptide of M, 32,000-33,000 on SDS gels (3,5,8). The true molecular weight of the &subunit of oxaloacetate decarboxylase from Klebsiella pneumoniae and Salmonella typhimurium as derived from the DNA sequence, however, is 45,000 (7). The @-subunit is supposed to contain several membrane-spanning domains (7, 9). Its function probably is the catalysis of the decarboxylation of the carboxylated biotin carrier and coupled Na+ translocation across the membrane (1,2,8). Oxaloacetate decarboxylase contains a third subunit (y) of M , 8,900, which probably contains one transmembrane a-helix in its N-terminal region and a more hydrophilic C-terminal part that could extend into the aqueous phase (7, 9). The fourth subunit found in methylmalonyl-CoA decarboxylase (3) and glutaconyl-CoA decarboxylase ( 5 ) ( 6 ) could be related to the oxaloacetate decarboxylase y-subunit. The function of this subunit is not yet known.
The sequence of the N-terminal domain of the oxaloacetate decarboxylase a-subunit revealed striking homology to the 5 S subunit of transcarboxylase from Propionibacterium shernanii that catalyzes the same carboxyl-transfer reaction (6,7). The C-terminal biotin-containing domain strongly resembled the biotin-containing subunits of transcarboxylase and other biotin-containing enzymes. An extended stretch of mostly alanine and proline residues in the N-terminal part of the biotin domain could provide the a-subunit with the flexibility required for the flip-flop movement of the prosthetic biotin group between the catalytic centers of the carboxyltransferase (N-terminal domain of the a-subunit) and the lyase (probably on the @-subunit). Neither the pnor the ysubunit of oxaloacetate decarboxylase was homologous to any known sequence (6,7).
In our efforts to elucidate structure and function of the sodium ion transport decarboxylases, we have cloned and sequenced the genes encoding the methylmalonyl-CoA decarboxylase from V. parvula. A comparison of the deduced amino acid sequences with those of the oxaloacetate decarboxylase provides insights into potentially important amino acids, notably within the membrane-bound @-subunits. EXPERIMENTAL PROCEDURES Materials". parvula (ATCC 17745) was grown anaerobically on sodium lactate as described (4). Methylmalonyl-CoA decarboxylase was purified from V. paruula by affinity chromatography on a monomeric avidin-Sepharose column (10). Restriction enzymes were from Boehringer Mannheim (Mannheim, Federal Republic of Germany). Oligonucleotides were custom-synthesized by Microsynth (Windisch, Switzerland). The plasmids pBluescript KS(+) and pBR322 and the host strain Escherichia coli DH5 a were from laboratory stock.
Amino Acid Sequence Analysis-The subunits of methylmalonyl-CoA decarboxylase were separated by SDS-polyacrylamide gel electrophoresis according to the procedure described in Ref. 11 and electroblotted onto a hydrophobic polyvinylidene difluoride membrane (Millipore) (12). Blot staining was performed with 0.1% COOmassie Brilliant Blue G250. The filter pieces with the respective subunits were used directly for N-terminal sequence analyses using a protein sequenator (model 470/A; Applied Biosystems) with on-line phenylthiohydantoin-derivative detection using high performance liquid chromatography (model 120; Applied Biosystems).
Cyanogen bromide fragments were prepared by isolating the SDS gel pieces with the respective subunits and treatment overnight with 500 pl of 70% formic acid and 1 mg of BrCN at 25 "C. The supernatant was removed and the residual peptides were extracted from the gel pieces by treatment with 2 X 0.5 ml of 50% formic acid, 25% acetonitrile, 15% isopropanol, and 10% water for 1 h at 37 "C. The combined supernatant and wash solutions were evaporated in a Speed Vac. The cyanogen bromide fragments were separated by SDS-gel electrophoresis, isolated from the gel, and sequenced.
Southern AnalyseslRestriction Map of the mmd Genes-Chromo-soma1 DNA of V. paruula was obtained as described (13). The DNA was digested with different restriction enzymes and the fragments were separated by gel electrophoresis in 0.7% agarose (20 X 20 cm). The DNA fragments were blotted onto a Hybond N membrane (Amersham Corp.) and fixed on the matrix by 5 min of irradiation with UV light (320 nm). Restriction fragments were analyzed by Southern hybridization (14) with oligonucleotides 5"CAYGAYAAY-ATIGA-3' (oligonucleotide a ) , 5'-ATGAAYAAYTTRAARGT-3' (oligonucleotide y-l), and 5'-GAYGCIATGAAYATG-3' (oligonucleotide 7-2). Oligonucleotides a and 7-1 were deduced from the Nterminal amino acid sequences of the a-and y-subunit, respectively, and oligonucleotide 7-2 was based on the highly conserved sequence motif Glu-Ala-Met-Lys-Met, which constitutes the biotin binding site of most known biotin enzymes (15). The oligonucleotides were labeled with [y-32P]dATP and used for hybridization as described (16). The information obtained from these experiments is summarized in the restriction map shown in Fig. 1.
Cloning Strategy-For cloning the mmd genes we used the information obtained from the restriction analyses ( Fig. 1). Chromosomal DNA (100 pg) was digested completely with EcoRI, and the restriction fragments were separated by gel electrophoresis in 0.7% agarose. The desired DNA fragment that hybridized with all three oligonucleotides is about 6.2 kb' long ( Fig. 1). Therefore, DNA in the range of 6.0-6.5 kb was isolated from the gel by electroelution. The DNA was digested with restriction enzymes BamHI and XbaI, which do not cut inside the desired EcoRI fragment but will reduce the number of false EcoRI fragments of the same size. EcoRI fragments of 6.0-6.5 kb were subsequently separated from the generated smaller fragments by electrophoresis and isolated from the gel as described above. The isolated EcoRI fragment was further digested with HindIII, which cuts twice inside the EcoRI fragment in sufficient distance to the EcoRI sites so that a "pure" HindIII/HindIII fragment resulted that could be used for cloning part of the mmd genes. Isolated HindIII fragments of 2.7 kb were cloned into the pBluescript KS(+) vector and transformed into the host strain E. coli DH5a (16). The plasmids of 40 recombinant white colonies were isolated (17) and analyzed by hybridization against oligonucleotide 7-1, which yielded 10 positive signals. Cloning of the desired DNA fragment was verified by sequence analysis using oligonucleotide 7-1 as a primer. An amino acid sequence identical to the known N terminus of the y-subunit was derived from the DNA sequence. The plasmid thus obtained was termed pJHl (Fig. 1). Also cloned from the chromosomal DNA were the flanking regions of the pJHl insert. The 2-kb EcoRI/ClaI fragment (pJH2O) was cloned into pBluescript KS(+), and the 3.1-kb NheIIEcoRI fragment (pJH40) was cloned into pBR322 ( Fig. 1). Both clones were identified via colony hybridization using the pJH1 insert as a homologous probe. The pJHl insert was DIG-labeled according to the manufacturer (Boehringer Mannheim).
DNA Sequence Analysis-The DNA region of the methylmalonyl- The abbreviations used are: kb, kilobase(s); ORF, open reading frame.
CoA decarboxylase genes was subcloned from plasmids pJH1, pJH20, and pJH40 and sequenced completely on both strands according to the dideoxynucleotide chain termination method described by Sanger et al. (18). Specific oligonucleotide primers were used to complete the sequence.
Computer Analyses-Computer analyses of the DNA and protein sequences were performed with the software package from the Genetics Computer Group of the University of Wisconsin (GCG version 7.1) and the PC-Gene program from IntelliGenetics (Geneva, Switzerland). The program TOP-PRED was obtained from G. von Heijne (Stockholm, Sweden).

RESULTS
Cloning of the Structural Genes for Methylmalonyl-CoA Decarboxylase-"terminal amino acid sequence analysis of the methylmalonyl-CoA decarboxylase subunits was performed in order to synthesize specific oligonucleotide probes for Southern hybridization and cloning of the methylmalonyl-CoA decarboxylase genes. The N termini of the pand the 6subunits were blocked, but those of the a-and the biotincontaining y-subunit could be sequenced and yielded MATVQEKIELL for a and MKKFNVTVNGTAYDVEV-NEVKAA for y. Since the DNA sequence reported here is the first one of a Veillonella species, no information of the codon usage was available. Therefore, all possible codons for a given amino acid had to be considered in the design of the oligonucleotides. The oligonucleotides synthesized corresponded to the N-terminal sequence of the a-subunit (oligonucleotide a), to the N-terminal sequence of the y-subunit (oligonucleotide y-l), and to the conserved biotin binding motif of the ysubunit (oligonucleotide 7-2) (see "Experimental Procedures''). All three oligonucleotides hybridized with the same BamHI, EcoRI, Sad, and XbaI restriction fragments of V. paruula chromosomal DNA, indicating that at least the genes for the a-and y-subunit form a cluster on the chromosome. Based on the restriction map derived from the hybridization data ( Fig. I), the 6.2-kb EcoRI fragment was chosen for cloning the genes for the a-and y-subunits. However, all attempts to clone this fragment via colony hybridization failed due to the high unspecific background signal obtained with the oligonucleotides. Therefore, a new strategy was developed, including an "enrichment" of the 2.7-kb HindIII fragment as described under "Experimental Procedures." Clone pJH1   Presumptive ribosome binding sites are underlined. Amino acids confirmed by protein sequencing of subunits or peptides are printed in bold letters. The stop codons are marked by asterisks. The lysine residue of the y-subunit that becomes biotinylated (LysS5) is indicated by a bold asterisk. could thus be isolated, which encoded the N terminus of the y-subunit as shown by DNA sequencing. From the hybridization data with oligonucleotide a, it was evident that the 5' end of the gene for the a-subunit was not present on pJH1. Furthermore, DNA sequence analysis of the complete 2.7-kb HindIII insert revealed that the 3' end of the gene for the psubunit was also missing on that fragment. Thus, in order to complete the sequences, two additional fragments containing the upstream and downstream DNA of the HindIII fragment had to be cloned. For this purpose, the 2-kb EcoRI/ClnI and the 3.1-kb NheI/EcoRI fragments were chosen and cloned from chromosomal DNA via colony hybridization using the pJHl insert as a homologous probe. The resulting plasmids were called pJH2O and pJH40 (Fig. 1).

M H A M G P N V A G V I G T A V A A G T
DNA Sequence Analysis-A DNA region of 4.68 kb, starting at the EcoRI site of pJH20, was completely sequenced on both strands using pJH1, pJH20, pJH40, and appropriate subclones. The complete nucleotide sequence together with the deduced amino acid sequence is shown in Fig. 2. Five open reading frames (ORFs) were identified, which will be described in the order shown in Fig. 1.
ORFl started at position 403 with an ATG and ended at position 1930 with a TAA stop codon. Since the deduced Nterminal amino acid sequence of ORFl was identical to the one obtained by protein sequencing of the a-subunit, ORFl undoubtedly encodes this subunit and is therefore termed mmdA (mmd is mnemonic for methyl-malonyl-CoA-decarboxylase). mmdA encodes a protein of 509 amino acid residues with a calculated M, of 55,100. ORF2 was unequivocally shown to encode the 6-subunit of the methylmalonyl-CoA decarboxylase by sequencing a BrCN fragment of this subunit. The resulting sequence (Fig. 2) was identical to amino acids 87-100 of ORF2, which is therefore called mmdD.
ORF3 started at position 2377 with an ATG and stopped at position 2542 with TAA. This ORF encoded a protein of 55 amino acids with a calculated M, of 5,888. Hitherto, such a small protein has never been detected in purified methylmalonyl-CoA decarboxylase preparations (3). However, as described below, the protein deduced from ORF3 is actually a fifth subunit of this enzyme and ORF3 is therefore named mmdE.
ORF4 Although not verified by amino acid sequence analysis, the considerable homology to the p-subunit of oxaloacetate decarboxylase from K. pneumoniae and s. typhimurium (Fig. 7) (7) definitely identified ORF4 as the gene for the p-subunit of methylmalonyl-CoA decarboxylase, and it was therefore named mmdB.
Biochemical Verification of a Fifth Subunit of Methylmalonyl-CoA Decarboxyhe-As described in the previous section, DNA sequence analysis led to the discovery of a small ORF, located between mmdD and mmdC (Fig. 1). The deduced protein consists of 55 amino acids with a calculated M, of 5,888. A protein of this size has never been observed in purified preparations of methylmalonyl-CoA decarboxylase (3), presumably because of its small size. Therefore, the decarboxylase subunits were separated on a 16.5% rather than on a 12% SDS polyacrylamide gel. As shown in Fig. 3, a band of about 6 kDa is visible after silver staining but only if large amounts of protein are loaded on the gel. In order to verify that this band corresponds to the protein deduced from ORF3, the protein band was eluted from the gel and subjected to Nterminal amino acid sequence analysis. In the first cycle, only a weak signal was observed (typical for serine), but no signal for methionine was found. The following cycles yielded the sequence Asn-Ala-Thr-Thr-Thr, which corresponds to the deduced N terminus of ORF3. Thus, the existence of a fifth subunit of the methylmalonyl-CoA decarboxylase was proven. Apparently, the N-terminal methionine is cleaved off in the mature protein, which therefore starts with a serine residue. The calculated M, of the mature t-subunit is 5,758.
Comparison of Amino Acid Sequences of Methylmalonyl-CoA Decarboxylase Subunits with Those of Other Protein.+" search of the GenBank and EMBL data banks revealed relationships between methylmalonyl-CoA decarboxylase subunits and other proteins with various degrees of sequence identity. A remarkable degree of 60 or 52% identity was found between the a-subunit of methylmalonyl-CoA decarboxylase and the carboxyltransferase domain of rat liver propionyl-       CoA carboxylase (20) or the 12 S subunit of transcarboxylase from P. shermanii, respectively (15) (Fig. 4). These data are in accord with a common function of these proteins/domains, i.e. catalysis of the carboxyl transfer from methylmalonyl-CoA to protein-bound biotin to yield propionyl-CoA and the carboxybiotin derivative, or catalysis of the carboxyl transfer in the reverse reaction. In contrast, no significant sequence homology was found to the carboxyltransferase domain of oxaloacetate decarboxylase (6,7), in accord with the different substrate specificities of these catalysts.

PKKHGNIPL WRKHANVPL A K K P W K L P L L S E E E I~E E E K D L M I A T L N K R V A S L E S E L G S L Q S D~~E D~T A I S
No significant sequence homology could be found between the &subunit of methylmalonyl-CoA decarboxylase and the y-subunit of oxaloacetate decarboxylase (7,9). The hydrophobicity profile of the two subunits, however, was similar, indicating a hydrophobic and probably membrane-spanning N-terminal part and a more hydrophilic C-terminal part. Another common feature of the two subunits is a proline/ alanine linker in the region between residues 47 and 65 (residues 45-60 in case of the oxaloacetate decarboxylase ysubunit). The y-subunit of oxaloacetate decarboxylase consists of 83 amino acid residues and contains, close to the C terminus, a remarkable motive of 4 histidines in series (7,9), which could provide the binding site for the Zn2+ metal ion present in this enzyme (21). The &subunit of methylmalonyl-CoA decarboxylase is 32 amino acids longer and lacks the series of multiple histidines. Related to these findings may be the occurrence of divalent metal ions (Zn2+, Co2+, or Mn2+) in carboxyltransferase components of biotin enzymes reacting on the keto acid substrates pyruvate and oxaloacetate but not in carboxyltransferases reacting with thioester substrates (e.g. propionyl-CoA/methylmalonyl-CoA) (15,22).
The c-subunit has considerable sequence homology to the C-terminal part (residues 75-117) of the &subunit (Fig. 5), indicating that the c-subunit may have evolved from the 6subunit by gene duplication of the N-terminal portion. Upstream of the homologous region, the c-subunit has only 15 additional residues with poor homology to the hubunit.
Like other biotin-binding proteins, the y-subunit of methylmalonyl-CoA decarboxylase contains the biotin-binding lysine 35 residue upstream of the C terminus within the highly conserved sequence LEAMKM (15) (Fig. 6). Additional con-  Fig. 4. References are given in the text.
served amino acids are scattered over the entire C-terminal region following the Conserved alanine/proline pair at position 65/66, which is required for recognition by biotin ligase (23). A remarkable sequence of the methylmalonyl-CoA decarboxylase y-subunit is a proline/alanine linker between residues 22 and 58, which contains 23 alanine, 11 proline, 2 lysine, and 1 valine residue. The oxaloacetate decarboxylases from S. typhimurium and K. penumoniae also contain proline/alanine linkers, which start 15 residues more downstream than that of methylmalonyl-CoA decarboxylase and extend to the same C-terminal position (6,7). The biotin carrier protein of transcarboxylase of P. shermanii contains an accumulation of 9 glycine, 6 alanine, and 2 prolines out of 19 residues in this region (15), and the biotin carrier of Streptococcus mutuns has 4 glutamines, 2 prolines, and 3 alanines in a stretch of 14 amino acid residues (possible Q-linker; Ref. 24) at the same location. All these amino acid sequences are supposed to serve the same purpose, i.e. to provide the protein with flexibility for movement of the prosthetic biotin group between different catalytic centers on the respective biotin-containing enzyme complexes.
The N-terminal portion of the methylmalonyl-CoA decarboxylase y-subunits (residues 3-17) is highly homologous to the biotin-carrier protein of transcarboxylase containing 12 identical residues and 3 conservative exchanges. This region of the transcarboxylase subunit is known to be essential for binding to the 12 S carboxyltransferase subunit (25). The degree of sequence identity between the two biotin-carrier proteins, as well as between the 12 S subunit of transcarboxylase and the a-subunit of methylmalonyl-CoA decarboxylase, indicates a similar binding between carboxyltransferase and biotin carrier protein subunits in both complexes.
The amino acid sequences of the @-subunits from methylmalonyl-CoA decarboxylase and oxaloacetate decarboxylase are 61% identical (7) (Fig. 7). The homology extends over the entire sequence with the exception of a long deletion of 57 amino acids in the methylmalonyl-CoA decarboxylase @-subunit following residue 64. Additional deletions of 1 and 6 amino acids are found after residue 73 and 124, respectively,   and residues 12-15 of the methylmalonyl-CoA decarboxylase @-subunit are an insertion with respect to the oxaloacetate decarboxylase @-subunit. Of interest are several long stretches of amino acid sequence identity between the @-subunits of methylmalonyl-CoA decarboxylase and two different oxaloacetate decarboxylases, which indicate functionally important regions of these proteins. We have recently taken advantage of these highly conserved areas for amplifying part of the gene for the @-subunit of methylmalonyl-CoA decarboxylase from Propionigenium modestum by the polymerase chain reaction technique.' The hydrophobicity plot of the @-subunit (Fig. 8) indicates a very hydrophobic protein with several putative membrane-spanning a-helices. Depending on the computer programs used, secondary structure models with 8-11 helices can be predicted. As the @-subunit secondary structure is probably conserved, the model should also fit the data derived from the oxaloacetate decarboxylase sequences (7). We therefore prefer a secondary structural model for the P-subunit with 9 transmembrane a-helices.

DISCUSSION
Nucleotide Sequence-We show here that the mmdA, mmdD, mmdE, mmdC, and mmdB genes encoding subunits a , 6, t, y, and @, respectively, of methylmalonyl-CoA decarboxylase are clustered on the genome of V. purvulu in the given order. Putative ribosome binding sites are located upstream of each gene (Fig. 2). Several putative E. coli 070 promotor consensus sequences (consisting of the -35 and -10 regions) are detectable within 400 base pairs upstream of the mmdA gene (two of these are indicated in Fig. 2 by overlining). The upstream region also contained the 3'-part of an open reading frame (residues 1-250 of the sequenced part of the genome). The derived amino acid sequence was not homologous to any protein in the data base.
A palindromic sequence of 2 x 8 nucleotides, 32 base pairs downstream of the stop codon of the @-subunit, that might be functioning as a terminator is indicated in Fig. 2 by overlining. An open reading frame in this downstream region starts at position 4422 with TTG and extends to the end of the sequence. No homology of the deduced protein sequence to any known protein sequence could be found. A translation of this ORF into protein seems unlikely because more than 21% of * P. Burda, unpublished observation. the amino acids would be aromatic. In addition, the G+C content in the region of the ORF is 10% lower (-32%) than the G+C content of the mmd genes and the ORF found upstream of the a-subunit, and TTG is a rarely used start codon in Gram-negative bacteria.
Our attempts to clone all mmd genes together in a high copy number plasmid in E. coli have not been successful, possibly because expression of methylmalonyl-CoA decarboxylase is lethal to an E. coli cell. The enzyme also catalyzes the decarboxylation of malonyl-CoA (3) and will thus interfere with fatty acid biosynthesis. It is interesting that the genes encoding methylmalonyl-CoA decarboxylase and those encoding oxaloacetate decarboxylase are clustered on the chromosome and that the genes encoding the P-subunits of both enzymes are transcribed last (7,9). In contrast, the gene encoding the a-subunit of glutaconyl-CoA decarboxylase from Acidaminococcus ferrnentans is separated on the genome from the genes encoding the additional subunits of the decarboxylase (26).
The Protein-Biochemical evidence has indicated structural and functional relationships among the biotin-containing Na+-transporting decarboxylases oxaloacetate decarboxylase, methylmalonyl-CoA decarboxylase, and glutaconyl-CoA decarboxylase (1,2). These relationships have been refined by the complete primary structure of methylmalonyl-CoA decarboxylase reported here. All these decarboxylase complexes consist of a tightly membrane-bound @-subunit, which contains a binding site for Na' (8, 27,28) and is therefore most likely responsible for Na+ translocation across the membrane. The more than 60% identity of the amino acid sequences between methylmalonyl-CoA decarboxylase and oxaloacetate decarboxylase @-subunits is clear evidence for the same function of these proteins in each of these enzyme complexes. The decarboxylases also contain a peripheral asubunit acting as carboxyltransferase. The sequences of these carboxyltransferases are homologous to those of biotin-containing carboxylases or transcarboxylase with the same substrate specificity. Transcarboxylase in fact contains two carboxyltransferases (5 and 1 2 S subunits) with sequence homology to the a-subunit of oxaloacetate decarboxylase and methylmalonyl-CoA decarboxylase, respectively (15). The biotin carrier protein is a distinct domain of the oxaloacetate decarboxylase a-subunit (6,7) but occurs as a separate protein moiety in the decarboxylases acting on thioester substrates (3,5). It is interesting that these biotin binding subunits contain alanine/proline linkers in their N-terminal region that are far more extended than putative linker sequences in the biotin carriers of carboxylases and transcarboxylase (6,7,15). The reason for these differences may be a requirement in the decarboxylases for a more extended movement of the prosthetic biotin group between the two catalytic centers of the carboxyltransferase and the lyase (decarboxylase). The 6subunit of methylmalonyl-CoA decarboxylase seems to be anchored in the membrane by a transmembrane helix in the N-terminal region. Following this, the protein contains a proline/alanine linker, a very hydrophilic strech of 9 amino acids (containing 2 asparagine, 2 glutamines, and 3 aspartates), a hydrophobic region of 12 amino acids, and a hydrophilic C-terminal tail. No sequence homology of this protein was found to the y-subunit of oxaloacetate decarboxylase (7,9), but the predicted secondary structures were surprisingly similar. Whether this similarity reflects a functional relationship is presently unknown. Remarkable is the presence in methylmalonyl-CoA decarboxylase of a small polypeptide (Mr 5,758, t-subunit) that shows strong sequence homology to the C-terminal portion of the &-subunit. While this indicates an origin of this subunit by gene duplication, a correlation of the e-subunit with function is presently not available.
Of special interest for the mechanism of Na' translocation is the structure of the P-subunit. A secondary structural model is shown in Fig. 9. It is based on predictions made by hydrophobicity analysis under consideration of the positive-inside rule (29) (Fig.  8 ) and also takes into account arguments discussed previously for postulating the secondary structure of the P-subunit of oxaloacetate decarboxylase from K. pneumoniae and S. typhzrnuriurn (7). It is assumed that the structures of the different P-subunits were conserved during evolution. The N terminus is supposed to reach into the periplasm because it contains 2 negatively charged and no positively charged residues, while the end of the first transmembrane helix (residues 17-40) of the ,&subunit of methylmalonyl-CoA decarboxylase contains an arginine and that of oxaloacetate decarboxylase contains 2 lysines side by side. These positively charged amino acids are supposed to function as stop transfer signals (30). In this first helix 12 residues are identical in the three different p-subunits and the others are conservative exchanges. It follows another highly conserved area of 19 amino acid residues containing 7 hydrophobic residues. In our model, this region was not postulated to traverse the membrane, because the putative helix would be rather short and/ or would contain a charged glutamate residue. Nevertheless, the highly conserved sequence of this area indicates that it is functionally important. In the Veillonella P-subunit there then follows a large gap (Fig. 5 ) . The second membrane helix was postulated to run from Ilea' to Ala'". A short loop of three amino acids connects helix I1 with helix 111, which runs from Thrlo5 to Leu126. After a short loop the polypeptide traverses the membrane again (Alai33 to T h P ) . T h e conserved amino acids Lys-Leu-Ala-Pro-His-(Glu) provide the linkage to helix V, which ends with G~u '~~. A loop with many charged residues leads to helix VI (IleZw to Gly233), which is not as highly conserved as the other putative membrane-spanning a-heliices. This region of the protein is highly hydrophobic in all the P-subunits sequenced. An aspartate in the middle of putative helix VI in oxaloacetate decarboxylase, however, is replaced by serine in the methylmalonyl-CoA decarboxylase sequence. It follows another region with many charged amino acids that is certainly not integrated in the lipid phase and a more hydrophobic region that could coil into helix VI1 to Ala271). Interestingly, the putative helix VI1 of both oxalo- acetate decarboxylase @-subunits contains a lysine residue, which is replaced by threonine (position 268) in the methylmalonyl-CoA decarboxylase sequence. It is conceivable that the non-conserved charged amino acids in putative helices VI and VI1 of the oxaloacetate decarboxylase P-subunit form a salt bridge within the membrane. The next helix (VIII) is rather clearly defined by its strong hydrophobicity (Ile2s2 to Gly301). The following region contains one of the most highly conserved areas of the whole protein (Asn313 to Ser331), where all 19 amino acid residues are identical. This area is flanked on both sides by less conserved segments with a number of charged residues. We do not want to speculate whether the mainly hydrophobic conserved region traverses the membrane. If so, an arginine would be located within the membrane-bound part. The polypeptide chain could traverse the membrane again (helix IX) from Phe343 to leaving the short C-terminal peptide Ala-Met-Leu-Ser-Glu-His extending into the aqueous phase.
Most of the helices predicted by this model are uncharged. A conserved aspartate residue is found in the middle of putative helix I1 and helix IV in highly conserved areas of the protein. These aspartic acids may be important residues for Na' translocation. As discussed above, the proposed model contains a number of disputable elements that must in the future be investigated by topological studies.