Identification and Molecular Characterization of the β-Ketoacyl-[Acyl Carrier Protein] Synthase Component of the Arabidopsis Mitochondrial Fatty Acid Synthase*

Substrate specificity of condensing enzymes is a predominant factor determining the nature of fatty acyl chains synthesized by type II fatty acid synthase (FAS) enzyme complexes composed of discrete enzymes. The gene (mtKAS) encoding the condensing enzyme, β-ketoacyl-[acyl carrier protein] (ACP) synthase (KAS), constituent of the mitochondrial FAS was cloned from Arabidopsis thaliana, and its product was purified and characterized. The mtKAS cDNA complemented the KAS II defect in the E. coli CY244 strain mutated in both fabB and fabF encoding KAS I and KAS II, respectively, demonstrating its ability to catalyze the condensation reaction in fatty acid synthesis. In vitro assays using extracts of CY244 containing all E. coli FAS components, except that KAS I and II were replaced by mtKAS, gave C4-C18 fatty acids exhibiting a bimodal distribution with peaks at C8 and C14-C16. Previously observed bimodal distributions obtained using mitochondrial extracts appear attributable to the mtKAS enzyme in the extracts. Although the mtKAS sequence is most similar to that of bacterial KAS IIs, sensitivity of mtKAS to the antibiotic cerulenin resembles that of E. coli KAS I. In the first or priming condensation reaction of de novo fatty acid synthesis, purified His-tagged mtKAS efficiently utilized malonyl-ACP, but not acetyl-CoA as primer substrate. Intracellular targeting using green fluorescent protein, Western blot, and deletion analyses identified an N-terminal signal conveying mtKAS into mitochondria. Thus, mtKAS with its broad chain length specificity accomplishes all condensation steps in mitochondrial fatty acid synthesis, whereas in plastids three KAS enzymes are required.

Fatty acids are synthesized by the enzymatic reactions of acetyl-CoA carboxylase (ACCase) 1 and fatty acid synthase (FAS). FAS enzyme complexes are classified into two groups based on their structural forms and organization. Complexes consisting of multifunctional polypeptides encoded by one or two genes (type I) are present in the cytoplasm of animals and fungi (1,2), whereas those composed of monofunctional enzymes (type II) are present in most bacteria and plant plastids (3,4) as well as in mitochondria, as will be detailed below. The incipient reaction of fatty acid synthesis is catalyzed by ACCase to form malonyl-CoA from acetyl-CoA, the initial carbon source. The first FAS activity, malonyl-CoA:ACP transacylase (MCAT), transfers the malonyl group from CoA to acyl carrier protein (ACP) to form the donor substrate malonyl-ACP that provides the C 2 -units for elongation. Repetitive elongation cycles accomplished by FAS start with condensation of the acyl primer substrate with the C 2 -unit to give a ␤-ketoacyl-ACP. The introduced ␤-keto group is then removed by three reactions, a ␤-keto reduction, a ␤-dehydration, and an enoyl reduction. The resulting saturated acyl-ACP serves as a substrate for the next extension. Most frequently seven or eight cycles yield palmitoyl (C 16 )-ACP and stearoyl (C 18 )-ACP, respectively. Additional FAS activities transfer the acyl chains to other compounds depending upon their final destination; for example, acyltransferases and thioesterases channel C 16 and C 18 fatty acids to membrane lipids.
The ␤-ketoacyl-ACP synthase (KAS) components of the type II FAS enzyme complexes present in bacteria and plant plastids carry out the condensation steps in fatty acid synthesis using a Claisen reaction consisting of three parts: (i) transfer of ACP bound acyl primer substrate to the active site cysteine of KAS, (ii) decarboxylation of malonyl-ACP to form the acetyl-ACP carbanion, and (iii) condensation of the carbanion with the carbonyl carbon of the acyl primer substrate. As shown in Fig.  1 (A and B), three KAS enzymes contribute to construction of fatty acyl chains. KAS III is singular in using a CoA-activated primer substrate, acetyl-CoA, for the initial condensation with a C 2 -unit (5,6). All the subsequent extensions are with ACPactivated acyl chains. They are carried out by KAS I and II, which differ in their substrate specificities. In plastids (Fig. 1A) KAS I utilizes butyryl (C 4 )-to myristoyl (C 14 )-ACPs as substrates, and KAS II executes the last step to yield C 18 -ACP (7). KAS IV enzymes in some seeds, such as those from Cuphea sp., show a marked preference for medium (C 6 -C 14 ) acyl chains (8).
In Escherichia coli KAS I and II carry out all the elongations starting with C 4 -ACP, with the exceptions that KAS I is unique for the initial step and KAS II for the final step in the pathway for unsaturated fatty acids (Fig. 1B) (9). Although double bonds are inserted into the growing acyl chain in E. coli, they are inserted by desaturases into the synthesized acyl chains in plastids. Mutants of fabF encoding KAS II have no phenotype as the absence of cis-vaccenoyl (C 18:1 cis ⌬11) acyl chains is compensated for by increased amounts of palmitoleoyl (C 16:1 cis ⌬9) acyl chains ( Fig. 1B) (10). In combination with mutants of fabB encoding KAS I, KAS II defects are revealed by the failure of the double mutant to grow in the presence of oleic acid (C 18:1 cis ⌬9) (9). ACP, a protein cofactor, is one of the best characterized of the FAS components, and carries the acyl chains through the FAS reactions. That ACP functions in additional pathways has been recognized for some time, for example, in polyketide synthesis (11). Nevertheless, the discovery of ACP in mitochondria was unexpected (12)(13)(14)(15)(16). A potential role of mitochondrial ACP in fatty acid biosynthesis was suggested by incorporation of radioactivity from [2-14 C]malonate into acyl-ACPs in mitochondria isolated from Pisum sativum (17) and Neurospora crassa (18). The nuclear genes, ACP1 (15), CEM1 (19), MCT1 (20), OAR1 (20), and Ybr026p (21) of Saccharomyces cerevisiae, Etr1p (21) of Candida tropicalis, mtACP-1 of Arabidopsis thaliana (16), and HsNRBF-1p of Homo sapiens (22), have been implicated in mitochondrial fatty acid synthesis. Inactivation of these yeast genes resulted in a respiration-deficient phenotype (19 -21, 23). Etr1p, Ybr026p, and HsNrbf-1 have enoyl-ACP reductase activity and are localized in mitochondria. The products of the yeast genes ACP1, CEM1, MCT1, and ORA1 have homology to plastid and bacterial ACP, KAS, MCAT, and ␤-ketoacyl-ACP reductase (KR), respectively, but definitive proof of their role in mitochondrial fatty acid synthesis has not been obtained. mtACP-1 is imported into mitochondria and can function as a cofactor in fatty acid synthesis (16). No candidate genes for the ␤-hydroxyacyl-ACP dehydrase have yet been isolated. During the final stages of preparing this report, a publication appeared characterizing human ACP and MCAT proteins that were targeted to mitochondria and had the requisite activities to function in fatty acid synthesis (24).
In the present study, we have purified and characterized a mitochondrial FAS constituent, namely the condensing enzyme (mtKAS) of A. thaliana that catalyzes all the condensation reactions in mitochondrial fatty acid synthesis. The mtKAS had an N-terminal mitochondrial target sequence and its cDNA complemented the fabF mutation in the E. coli strain CY244. Assays of purified mtKAS revealed that malonyl-ACP served as both the primer and donor substrates for de novo fatty acid synthesis explaining the absence of KAS III in mitochondria. Moreover, the fatty acids synthesized in a heterologous system exhibited an unusual bimodal distribution that was attributable to the mtKAS component of a type II FAS.

EXPERIMENTAL PROCEDURES
Plant Material-A. thaliana (Columbia ecotype) was grown on vermiculite under continuous light at a photon flux density of 40 mol m Ϫ2 s Ϫ1 at 25°C. Bright yellow-2 (BY-2) cells of tobacco (Nicotiana tabacum) were obtained from Dr. T. Nagata (University of Tokyo, Tokyo, Japan). The cells were grown in modified Linsmaier and Skoog's medium as described by Nagata et al. (25).
Construction of a cDNA Library and Cloning of the mtKAS cDNA Using Polymerase Chain Reaction (PCR)-Poly(A) ϩ RNAs were isolated from leaves and roots of 3-week-old A. thaliana with an mRNA purification kit (Amersham Biosciences). The cDNA library was constructed with cDNAs, which were prepared from the poly(A) ϩ RNAs with a Whether acetyl (C 2 )-CoA or C 3 -ACP serves as primer substrate in the first step is specified. Dotted arrows indicate more than one elongation step. A, plant plastids. B, E. coli. KAS I and II can carry out all elongations starting from C 4 -ACP except those designated. fabB encodes KAS I, fabF KAS II, and fabH KAS III. C, mitochondria. Specifying C 8 and C 14 to C 16 -ACP highlights the bimodal distribution of fatty acids synthesized by mtKAS. Additional details given under Introduction and "Results." cDNA synthesis kit (Amersham Biosciences), and a phage vector gt11 (Stratagene) according to the protocols from the manufacturer. phage DNAs prepared from this library served as templates for PCR. The primers, 5Ј-TCCGCTTAAACCGCTTCATC-3Ј and 5Ј-TGCTGTCCTAA-CTAACATCT-3Ј, were designed to amplify the central region of a predicted mtKAS cDNA derived from T103.5. The 3Ј-terminal region of mtKAS cDNA was amplified using the primers, 5Ј-TTGACACCAGAC-CAACTGGTAATG-3Ј and 5Ј-TCTTCGACAAGAGGTTCATGCCTT-3Ј, that anneal to the cloning site of gt11 and to an internal region of the deduced mtKAS cDNA, respectively. The 5Ј-terminal region of mtKAS cDNA was amplified by the 5Ј-rapid amplification of cDNA ends method (5Ј-Full RACE Core Set from Takara). The phosphorylated primer, 5Ј-GCAACTGCCTTAGAG-3Ј, was used for synthesis of cDNA using total RNAs isolated from A. thaliana leaves with a purification kit (RNeasy plant mini kit from Qiagen). The following two sets of primers were used for the first and second PCR in the 5Ј-rapid amplification of cDNA ends, respectively: (i) 5Ј-TTTGTGCCTTATGGATCAAACCCT-3Ј and 5Ј-GCAGCAACTTTAGAAGAAAGCTGA-3Ј, (ii) 5Ј-GGTGAATTTG-ATGAAGCCCT-3Ј and 5Ј-AGAGTCAATCCTCTAATCCC-3Ј. Each of the three amplified fragments was subcloned into pCR2.1 (Original TA cloning kit; Invitrogen), and its nucleotide sequence was determined. Combining the nucleotide sequences of the 5Ј-, internal, and 3Ј-regions gave a full-length cDNA of the predicted A. thaliana mtKAS of 1,566 bp.
Homology Search and Phylogenetic Analysis-Sequences homologous to the predicted sequence of mtKAS were retrieved using a BLASTP program at National Center for Biotechnology Information (www.ncbi.nlm.nih.gov). These sequences from bacteria, plant plastids, yeast, fungi, and animals were aligned with that of mtKAS using CLUSTAL X version 1.8 (26), followed by adjustment with the naked eye. The known or deduced N-terminal regions corresponding to transit peptides and internal gaps were removed prior to analysis. Distance matrices were generated with the PROTDIST program in PHYLIP version 3.573c (27) using the PAM matrix of Dayhoff et al. (28). A phylogenetic tree was constructed by using the neighbor-joining method (29) of the PHYLIP NEIGHBOR program.
Construction of Plasmids Expressing KAS Genes-Deletions of the full-length mtKAS cDNA were created using PCR with specific primers designed to remove varying numbers of residues from the N-terminal end of the mtKAS protein. The 5Ј-specific primers, 5Ј-CATGCCATGG-CGACATCTAATCTCCGT-3Ј, 5Ј-CATGCCATGGGGATCTCTACTTCT-TCTTCTTATCATTCA-3Ј, and 5Ј-CATGCCATGGGGCATTCACATCG-CCGTGTTG-3Ј, and the 3Ј-specific primer 5Ј-AACTGCAGTTAGATAG-AGGCAAAGAGCAAAG-3Ј were used for PCR. In constructing these primers, the sequences 5Ј-CATGCCATGG-3Ј or 5Ј-AACTGCAG-3Ј, including an NcoI site and a PstI site, respectively, were added to the 5Ј end of each primer. The obtained PCR products were digested with NcoI and PstI, and ligated into the same sites of the expression vector pKK233-2 (GenBank TM accession no. X70478). The obtained plasmids were designated pmtKAS-⌬0, pmtKAS-⌬20, and pmtKAS-⌬27, where the numbers after ⌬ represent the number of deleted amino acids. Except for pmtKAS-⌬0, all the constructs code for an extra Gly as the second residue. The pmtKAS-⌬27 was used to fabricate an additional deletion mutant, pmtKAS-⌬30, using the QuikChange site-directed mutagenesis kit (Stratagene) as described previously (30). The pmtKAS-⌬30 then served as a template for construction of pmtKAS-⌬34. The following primers plus their complements, 5Ј-CACACAGGAAACAGA-CCATGGGGCGCCGTGTTGTTGTCACTGG-3Ј and 5Ј-CACACAGGA-AACAGACCATGGGGGTCACTGGTCTAGGCATCG-3Ј, were used. Confirmation that the five desired constructs had been obtained was ascertained by complete sequencing.
The E. coli fabF gene (31, 32) encoding KAS II was amplified by PCR using the primer set 5Ј-CATGCCATGGTGTCTAAGCGTCGTGTAGTT-G-3Ј and 5Ј-AACTGCAGAGGGTGGCAAATGACAACTTAG-3Ј. Genomic DNA extracted from E. coli JM109 strain functioned as the template. The PCR product was cloned into pKK233-2 as described above to give the plasmid pfabF with an extra residue, Met, preceding the N-terminal Val. The plasmid pDM4, designated pfabB in this study, which includes the E. coli fabB gene (33) encoding KAS I was extracted from the strain DM86/pDM4 (34).
Complementation of the Double Mutant E. coli Strain CY244 -The E. coli strain CY244 bears mutations in both the fabB and fabF genes (35). The fabB15 allele (GenBank TM accession no. CAA09934) codes for KAS I with an amino acid substitution (Ala-329 to Val) that results in temperature sensitivity, whereas the fabF1 allele encodes two amino acid substitutions (Ser-220 and Gly-262 to Asn and Met, respectively) that result in lack of KAS II activity (32). The control plasmid pKK233-2 plus the plasmids pmtKAS, pfabB, and pfabF were transformed into CY244. Transformants were selected by plating on Luria-Bertani (LB) plates supplemented with 100 g ml Ϫ1 ampicillin at 30°C. Resulting colonies were then streaked onto LB plates supplemented with 100 g ml Ϫ1 ampicillin and either 0 or 100 g ml Ϫ1 oleic acid (C 18:1 cis ⌬9, Sigma), and incubated at 30 or 42°C for 3 days. A stock solution of K ϩ -oleate in ethanol was diluted with 10% Tergitol Nonidet P-40 (Sigma) and added to LB medium to give 100 g ml Ϫ1 oleic acid and 0.1% Tergitol.
Elongation Assay for KAS Enzyme Activity-Transformants of CY244 were grown in liquid LB medium supplemented with 100 g ml Ϫ1 ampicillin and 100 g ml Ϫ1 oleic acid at 42°C, except that the pKK233-2 transformant was grown at 30°C. Cells collected from 10-ml overnight cultures by centrifugation were suspended in 1 ml of cold break buffer consisting of 25 mM Tris-HCl (pH 7.8), 300 mM NaCl, 10 mM dithiothreitol (DTT), 5 mM EDTA, and 1 mg ml Ϫ1 lysozyme. Cells were ruptured by sonicating twice for 20 s on ice. After centrifugation at 27,500 ϫ g for 20 min, the supernatant was recovered as the soluble protein extract. To visualize the presence of KAS proteins, the soluble protein extracts were subjected to SDS-PAGE followed by either Coomassie Brilliant Blue staining or transfer to a PVDF membrane for Western blot analysis with antibodies against mtKAS (see below), plus KAS I and KAS II of E. coli (31). The reactions were stopped by addition of 1 ml of cold 5% trichloroacetic acid. The precipitated acyl-ACPs were collected by centrifugation, resolved by 1, 2, 3, or 4 M urea-PAGE on the basis of the chain length and presence of double bonds, electroblotted to a PVDF membrane, and subjected to autoradiography using a PhosphorImager as described (30). 14 C-Labeled standards included C 16 -ACP (30), acetyl-and malonyl-ACPs prepared as described below, plus C 12 -ACP prepared using a labeled C 12 fatty acid (Amersham Biosciences) and acyl-ACP synthase purified as described (36) according to Rock and Cronan (37). Identity of additional acyl-ACPs was deduced based on previous work (31). The sensitivity of the expressed KAS enzymes to cerulenin was investigated as previously detailed (31) with some modifications. Cerulenin (Sigma) (0 -10 nmol) dissolved in ethanol was transferred to Eppendorf tubes and dried under vacuum. Soluble protein extracts (10 l) were added to the tubes. After standing on ice for 15 min, 90 l of the reaction mixture was added and the elongation assay carried out as described above.
Characterization of the First Condensation Reaction-The mtKAS-⌬20 cDNA was re-cloned into pQE-30 (Qiagen). This resulted in an N-terminal sequence in which Met-Gly was replaced with Met-Arg-Gly-Ser plus a His tag of six residues followed by Gly-Ser-Met (38). Expression and protein purification using Ni-NTA technology were carried out as detailed previously (30,38) with overnight induction and using only the lysozyme and sonication steps to open the cells. To determine the ability of the Ni-NTA-purified mtKAS enzyme to initiate fatty acid synthesis by joining two C 2 -units together in the first condensation reaction, an assay was developed. The assay also reveals whether the primer substrate is acetyl-CoA or acetyl-ACP. ACP and MCAT were pre-incubated (30) in the presence of DTT and then mixed with the other components of the assay mixture. The 100-l assay mixture consisting of 10 M ACP, 0 -120 M acetyl-CoA Ϯ 22-132 kdpm [1-14 C]acetyl-CoA (Amersham Biosciences, specific activity 2.18 GBq/ mmol), 10.1 M malonyl-CoA Ϯ 10 kdpm [2-14 C]malonyl-CoA (PerkinElmer Life Sciences, specific activity 2.2 GBq/mmol), 1 mM DTT, with or without 1 mM NADPH in 50 mM potassium phosphate buffer (pH 6.8) as well as 1.2 g of Ni-NTA purified mtKAS or KAS II of E. coli, with or without 2.5 g of KR (38) and MCAT was incubated for 15 min at 37°C. Acyl-ACP products were analyzed as described above. Radiolabeled acetyl-and malonyl-ACP standards were prepared by carrying out decarboxylase assays in the presence and absence of E. coli KAS I mutant protein that has a substitution of amino acid Cys-163 with Ala and efficiently decarboxylates malonyl-ACP to acetyl-ACP (30).
Northern and Reverse Transcription (RT)-PCR Analyses-Poly(A) ϩ RNAs used for Northern blot analysis were extracted from 2-week-old A. thaliana leaves using an mRNA purification kit (Amersham Biosciences). Approximately 3 g of poly(A) ϩ RNAs were separated by electrophoresis in a 1% (w/v) agarose gel, and transferred to a nylon membrane (Hybond-Nϩ; Amersham Biosciences). Hybridization and detection were performed using a DNA labeling and detection system (ECL kit; Amersham Biosciences). Total RNAs used for RT-PCR were extracted from leaves, roots, and flowers of 4-week-old A. thaliana with an RNA extraction kit (RNeasy® Plant Mini Kit; Qiagen). The primer set used for amplifying and cloning of the mtKAS-⌬20 sequence was used for RT-PCR as described before (39). As a control the expression of the gene for ␣-tubulin of A. thaliana (40) was also checked with the same RNA preparations using the primer set 5Ј-CTACTGAGAGAAG-ATGCGAG-3Ј and 5Ј-CAACATCTCCTCGGTACATC-3Ј.
Expression of mtKAS-⌬20 in E. coli and Antibody Production-The mtKAS-⌬20 sequence was amplified using the primer set 5Ј-CGCCAT-GGCGACATCTAATCTCCGT-3Ј and 5Ј-CGCCATGGTTAGATAGAGG-CAAAGA-3Ј. In constructing these primers, the sequence 5Ј-CGCCAT-GG-3Ј including an NcoI site was added to the 5Ј end of each primer. The obtained PCR products were digested with NcoI and ligated into the same site of the expression vector pET-30a(ϩ) (Novagen). The obtained plasmid was used to transform E. coli BL21(DE3) (41). Expression of mtKAS-⌬20 in BL21(DE3), purification of mtKAS-⌬20 protein, and production of a polyclonal antibody against this protein were performed using a previously described procedure (39).
Cell Fractionation and Western Blot Analysis-The cytosol, microsome, mitochondrial, and chloroplast fractions were prepared from 3-week-old A. thaliana leaves and used for Western blot analysis as described (39).
Construction and Visualization of mtKAS-Green Fluorescent Protein (GFP) Fusion Protein-The mtKAS cDNA sequence encoding the 50 N-terminal residues was amplified by PCR with the primer set 5Ј-GC-GTCGACATGGCGACATCTAATCTCCGTAGA-3Ј and 5Ј-CATGCCAT-GGTTGTTTCAACGCCTCTACCAAG-3Ј. The sequences 5Ј-GCGTCGA-CAT-3Ј or 5Ј-CATGCCATGG-3Ј, including a SalI site and an NcoI, respectively, were added to the 5Ј end of each primer. The PCR product was digested with SalI and NcoI, and ligated into the SalI-NcoI site of the CaMV35S⍀-sGFP (S65T)-nos3Ј plasmid, designated pGFP (42). The obtained recombinant plasmid pmtKAStp-GFP and pGFP were introduced into tobacco BY-2 cells prepared from a 5-day-old culture, with a particle bombardment device (PDS-1000/He Biolistic® Particle Delivery System; Bio-Rad) according to instructions from the manufacturer. The applied conditions were 1,100 p.s.i. of helium gas pressure, a distance of 9 cm from macrocarrier to cell suspension, and a decompression vacuum of 28 inches Hg. Tungsten particles (1.1 m) were used as a carrier of plasmid DNAs. The bombarded cells were incubated at room temperature in the dark overnight and stained with a mitochondriaselective probe (Mito Tracker® Red CM-H 2 Xros; Molecular Probes) following the recommendations from the manufacturer. Cells were examined with a confocal laser scan microscope (LSM410; Carl Zeiss). Excitation wavelengths were set at 488 nm for GFP and at 543 nm for Mito Tracker Red. All obtained images were processed by Adobe Photoshop software.
Nucleotide Sequence Accession Number-The sequence data for mtKAS cDNA has been deposited in the DDBJ, EMBL, and GenBank TM data bases under accession no. AB073746.

RESULTS
Isolation of a cDNA Encoding mtKAS-The T1O3.5 sequence in GenBank TM was previously suggested to code for a KAS participating in mitochondrial fatty acid synthesis in A. thaliana (43). To confirm this hypothesis, we cloned and characterized a cDNA, designated mtKAS, corresponding to T1O3.5. The mtKAS cDNA contained an open reading frame of 1,383 bp encoding a polypeptide of 461 amino acids (molecular mass 49,379 Da). Comparison of the sequence of the cDNA and T1O3.5 revealed that mtKAS is composed of 14 exons and 13 introns. Southern blot analysis confirmed that mtKAS is a single-copy gene in the A. thaliana genome (data not shown), as predicted from the genome sequence. The deduced amino acid sequence of mtKAS is compared with those of KAS I and KAS II of E. coli and A. thaliana plastids in Fig. 2 (46). Lys-384 is another residue present in the active site and peculiar to the CHH group of KAS enzymes (30,46). The mtKAS sequence is distinguished from the other sequences (Fig. 2) by an insertion of 17 residues, which occurs in a 3-residue ␤-strand forming part of the capping region (47). An analogous insert is present in all other mitochondrial KAS sequences consisting of 4 -8 residues (data not shown). The 27-residue N-terminal extension of A. thaliana mtKAS relative to E. coli KAS II has characteristics of a mitochondrial transit peptide, i.e. an overall positive charge resulting from the relatively high proportion of arginines and the lack of negatively charged residues.
To probe the evolutionary relationship of mtKAS to other KAS enzymes, a phylogenetic tree was constructed based on the amino acid sequences of CHH KAS isozymes from bacteria, plants, fungi, and animals excluding N-terminal extensions. The resulting tree (see Supplemental Fig. 1, available in the on-line version of this article) confirmed the close relationship between two distinct groups of plastidial enzymes, and that they were more closely related to bacterial KAS II than KAS I isozymes as previously noted (48). Five of the mtKAS sequences form a fifth group that is even more closely related to bacterial KAS II enzymes than the plastidial enzymes are. The sixth mtKAS, CEM1 from S. cerevisiae, Sc-mt, was interesting, as it was not included in any of the five groups.
Complementation of an E. coli fabF Mutant-To confirm that mtKAS cDNA encodes a functional KAS enzyme, the Arabidopsis mtKAS cDNA was expressed in a heterologous system, namely the E. coli mutant, CY244, which lacks KAS II activity encoded by fabF and has a temperature-sensitive KAS I encoded by fabB (Fig. 1B). Thus, CY244 grows and synthesizes both saturated and unsaturated fatty acids at 30°C but not at 42°C. As shown in Fig. 3A, when the CY244 mutant was transformed with the pKK233-2 plasmid, the resulting transformants did not grow at 42°C. By contrast, transformants with pmtKAS-⌬20, which encodes the mtKAS lacking 20 Nterminal residues, grew at 42°C, but only on the plate supplemented with the unsaturated fatty acid, oleic acid. The same result was observed when CY244 was transformed with pfabF carrying the wild type fabF gene. These observations infer that pmtKAS-⌬20, just as pfabF, restores the activity to synthesize only saturated fatty acids. Transformants with pfabB carrying the wild type fabB gene grew at 42°C, even in the absence of oleic acid, as expected. Complementation of the fabF defect in CY244 clearly demonstrated that the mtKAS cDNA encodes an E. coli KAS II-like protein, i.e. mtKAS is able to elongate saturated acyl chains from 4 to at least 16 carbons. Whether it is also able to extend C 16:1 to C 18:1 (Fig. 1B) is not revealed by the present study.
To examine the effect of the N-terminal extension on the activity of mtKAS, complementation tests with CY244 were carried out with mtKAS cDNAs encoding mtKAS having a full-length N-terminal extension and four N-terminal deletions. Comparative growth of transformants is shown in Fig.  3B. Transformants with pmtKAS-⌬0 (full-length), pmtKAS-⌬20, and pmtKAS-⌬27 grew well, whereas those with pmtKAS-⌬30 and pmtKAS-⌬34 showed at best only the same faint growth exhibited by the empty vector pKK233-2 on oleic acid containing medium at 42°C. Western blot analysis of soluble extracts of all the transformants readily detected mtKAS proteins corresponding in size to ⌬0, ⌬20, and ⌬27, but none in the vector ⌬30 and ⌬34 extracts (data not shown). These results demonstrate that the 27 N-terminal residues of mtKAS are not necessary for the function of mtKAS, and that when 30 or more residues are deleted the translated protein is unstable. Thus, for a functional mtKAS, at least three residues need to precede Val-33, which corresponds to the first residue of the first ␤-strand in the ␣-␤-␣-␤-␣ core structure of the E. coli KAS I and KAS II enzymes (47,49). Our experimental results reveal that the PSORT prediction of a cleavage site between Gly-47 and Val-48 for the Arabidopsis mtKAS target sequence, which is just after the loop following the first ␤-sheet in the core structure, is incorrect. Presumably the analogous PSORT predictions for the deduced sequences of human and mouse mtKAS are also incorrect.
Synthesis of Acyl-ACPs by mtKAS in Crude Extracts of E. coli CY244 -To investigate the ability of the mtKAS to carry out the condensation reactions, crude soluble protein extracts prepared from the CY244 transformants grown at 42°C in the presence of oleic acid were used for elongation assays. The control transformant with the vector pKK233-2 was grown at 30°C. Acetyl-CoA and [2-14 C]malonyl-CoA were incubated as potential substrates with each extract. Labeled acyl-ACPs produced during a 30-min incubation at 42°C were separated by electrophoresis with a polyacrylamide gel containing 2 M (Fig.  4A) or 4 M (Fig. 4B) urea, electroblotted to PVDF membrane, and visualized by autoradiography. Only acetyl-and malonyl-ACPs were detected in the extract of the vector control (Fig. 4A,  lane 1). Malonyl-ACP was generated by transfer of the malonyl group from malonyl-CoA to ACP by MCAT, and acetyl-ACP by decarboxylation of malonyl-ACP by KAS III. Both MCAT and KAS III are present in E. coli soluble protein extracts. The pmtKAS-⌬20 transformant extract, by contrast, synthesized a series of saturated acyl-ACPs with up to 18 carbons (Fig. 4B,  lane 2). Of these, medium chain C 8 -and C 14 -ACPs plus the long chain C 16 -ACP were detected as major bands, resulting in a bimodal distribution. With the pfabB and pfabF transformant extracts, as expected, predominantly long chain acyl-ACPs, both saturated and unsaturated, were produced (Fig. 4B, lanes  3 and 4). These results demonstrate that mtKAS can catalyze many of the condensation reactions in fatty acid synthesis, and that it is compatible with all relative components of the E. coli FAS. Whether mtKAS is able to carry out the first condensation step to yield C 4 -ACP was unresolved, however, because of the presence of E. coli KAS III in the extracts.
Cerulenin, known as an irreversible inhibitor of type II FAS, binds to the active site cysteine residue of KAS I and II enzymes, and blocks their condensation activities (33,50). Although plastidial and bacterial KAS I enzymes are more sen-sitive to cerulenin than KAS IIs, the structural basis for this difference is not yet known (46). To check the effect of cerulenin on the activity of mtKAS, extracts of CY244 transformants were incubated in the presence of 0 -10 nmol of cerulenin before adding the substrates. As shown in Fig. 4C (lane 2 versus lane  3), mtKAS activity was almost completely inhibited by the addition of 1 nmol of cerulenin. Thus, it is at least as sensitive to cerulenin as KAS I of E. coli (lanes 5-7 versus lanes 2-4). Treating E. coli KAS II with 5 nmol of cerulenin (lane 9) had a little effect on saturated fatty acid synthesis compared with that on mtKAS and E. coli KAS I (lanes 4 and 7) as expected. Even with 10 nmol of cerulenin, some elongation still occurred (lane 10). The present results disclose that mtKAS is very sensitive to cerulenin and imply that Cys-209 is essential for catalysis.
The mtKAS and E. coli KAS I and II proteins expressed in CY244 were unstable in the crude extracts during storage at 0°C in break buffer, which lacks glycerol. This was most marked for mtKAS, which initially synthesized prominent amounts of C 14 and C 16 acyl chains (Fig. 4B, lane 2). After being stored for a week at Ϫ20°C, however, mtKAS synthesized predominantly C 8 acyl chains (Fig. 4C, lane 2), an ability that was maintained for at least 8 weeks of storage (data not shown).
The First Condensation Reaction Carried out by mtKAS-To examine the ability of mtKAS to carry out the first condensation reaction, the mtKAS-⌬20 protein was purified with aid of a His tag. This is the first report to our knowledge of a plant CHH type KAS being readily purified as an active enzyme. As shown in Fig. 5A, the activity assay was designed to reveal whether mtKAS could condense two C 2 -units to give a C 4 acyl chain, and if so whether the primer substrate was activated by CoA as characteristic for KAS III or by ACP as characteristic for KAS I and II. In the course of the assay, malonyl-ACP is generated from malonyl-CoA by MCAT. Decarboxylation of malonyl-ACP to acetyl-ACP, the donor substrate, is the second part of the tripartite KAS reaction. Decarboxylation assays using similar concentrations of mtKAS as in the present assay revealed an activity similar to that exhibited by E. coli KAS I (data not shown and Ref. 48). Acetyl-and malonyl-ACP standards are shown in lanes 1, 2, 10, and 11. In the final step of the KAS reaction, ␤-ketobutyryl (␤-keto C 4 )-ACP is synthesized if labeled malonyl-CoA is present in the reaction mixture as illustrated for mtKAS (lanes 3-7). The addition of KR and NADPH to the assay resulted in the reduction of the ␤-keto C 4 -ACP to ␤-hydroxybutyryl (␤-OH C 4 )-ACP (lanes 3-5). The latter was not formed if either KR or the cofactor NADPH upon which it is dependent was omitted (lanes 6 and 7). Why the reduction is not more efficient is unknown. A 4-fold increase in KR had no effect on the reduction of ␤-OH C 4 -ACP (data not shown). The presence or absence of unlabeled acetyl-CoA had no noticeable effect on the production of ␤-OH C 4 -ACP (lanes [3][4][5], and incorporation of radioactivity into ␤-OH C 4 -ACP from [2-14 C]acetyl-CoA was not detected under the tested conditions (lanes 8 and 9). These results demonstrate that (i) mtKAS possesses the ability to carry out the first condensation reaction and (ii) malonyl-ACP is a sole carbon source for the first condensation reaction catalyzed by mtKAS, i.e. it provides both the primer and donor substrates.
The same series of assays was carried out using a purified His-tagged E. coli KAS II (45) in place of mtKAS (Fig. 5B). Two interesting differences emerged. First, only traces of ␤-ketoacyl C 4 -ACP were detectable as predicted from its known instability (lanes 3-5). Second, acetyl-CoA can serve as the primer substrate for KAS II albeit inefficiently, as witnessed by a minor amount of ␤-OH C 4 -ACP (lanes 8 and 9). This band was absent FIG. 3. Complementation of the E. coli fabB and fabF double  mutant strain, CY244, by expression of A. thaliana mtKAS  cDNAs. A, CY244 was transformed with four plasmids, and the transformants were streaked onto plates without (left) or with 100 g ml Ϫ1 oleic acid (right) and incubated at 42°C for 3 days. pKK233-2 used as the negative control is in quadrant 1, pfabB in 2, pfabF in 3, and pmtKAS-⌬20 in 4. B, CY244 was transformed with pKK233-2, pmtKAS-⌬0, pmtKAS-⌬20, pmtKAS-⌬27, pmtKAS-⌬30, or pmtKAS-⌬34, and the transformants patched to a plate containing 100 g ml Ϫ1 oleic acid and incubated at 42°C.
if either KR or NADPH was omitted from the assay mixture (lanes 6 and 7).
Expression of 1,600 nucleotides, which was close to the size of the 1,566 nucleotides of the cloned mtKAS cDNA (data not shown). This discrepancy in size presumably arises from the absence of the 3Ј downstream region and poly(A) ϩ tail in the cDNA.
To estimate the expression level of the mtKAS gene in organs, RT-PCR analysis was carried out using total RNAs prepared from leaves, roots, and flowers. In all tested organs, a DNA fragment corresponding to mtKAS cDNA was detected at relatively similar levels (Fig. 6, upper panel) as was the control gene for ␣-tubulin (Fig. 6, lower panel). These results suggest that the mtKAS gene is expressed at the same level in these three organs.
Intracellular Localization of mtKAS in A. thaliana-As mentioned above, mtKAS has a putative mitochondrial transit peptide on its N terminus, implying its presence in this organelle. To explore this possibility, we prepared cytosolic (Fig. 7, lane  1), microsomal (lane 2), chloroplast (lane 3), and mitochondrial (lane 4) fractions from A. thaliana leaves and investigated the intracellular localization by Western blot analysis using an antibody against mtKAS. A single band at ϳ47 kDa corresponding in size to the mtKAS lacking ϳ20 residues was detected only in the mitochondrial fraction (Fig. 7, upper panel).
To check the purity of the mitochondrial proteins, the same fractions were analyzed with an antibody against pea H-protein, a subunit of the glycine decarboxylase complex (glycine cleavage system) present in mitochondrial matrix (51). The lower panel in Fig. 7 shows that H-protein was primarily in the mitochondrial fraction. These results support the contention that mtKAS is located in mitochondria.
To provide further evidence of the mtKAS target site, we prepared a recombinant plasmid that encodes the 50 N-terminal residues of mtKAS fused to GFP. This recombinant plasmid, pmtKAStp-GFP, was introduced into BY-2 tobacco cells by particle bombardment. As a control, the same construct lacking the mtKAS sequence (pGFP) was also introduced. The transformed BY-2 cells were examined with a laser scanning microscope. As shown in Fig. 8, the green fluorescence (panels A and B) represents the site where GFP is present, and the red fluorescence (panel C) from Mito Tracker Red identifies the mitochondria. In the cells in which pmtKAStp-GFP was introduced, the green fluorescence from GFP overlapped extensively with the red fluorescence from mitochondria (panel D), indicating that the mtKAS-GFP fusion protein was targeted into mitochondria in BY-2 cells. By contrast, in the cells in which pGFP was introduced, the green fluorescence from GFP was observed in the cytoplasm and nucleus (Fig. 8A). These results demonstrate that mtKAS is located in mitochondria and suggest that the N-terminal region of mtKAS functions as a mitochondrial transit peptide. DISCUSSION Although there is no longer any doubt that de novo fatty acid synthesis takes place in mitochondria of eukaryotic cells, as well as in plant plastids and the cytoplasm of animal cells, numerous questions remain about the functions of the synthe-sized fatty acids and about the enzymes of the FAS complex and their genes. Least is known about the condensing, ketoacyl reductase and dehydrase enzymes. As a step toward remedying these deficiencies, we have cloned and carried out the initial characterization of the mitochondrial condensing enzyme component, mtKAS, from A. thaliana. In vivo the Arabidopsis mtKAS gene complemented the fabF defect in E. coli CY244 that bears mutations in both fabB and fabF, demonstrating a role of mtKAS in fatty acid synthesis as a condensing enzyme. mtKAS mRNA was present in leaves, roots, and flowers. The biosynthesis of fatty acids in mitochondria would occur in these organs if the fatty acids were vital for a mitochondria to function in respiration, as suggested by the respiration-deficient phenotypes of knockout mutants (19 -21, 23, 24). Western blot and intracellular targeting analyses with the mtKAS-GFP fusion protein revealed that the amino terminus of mtKAS serves as a signal to target protein into mitochondria. Deletion analyses demonstrated that the maximum length of the presequence was 29 residues. Thus, mtKAS falls into the most common class of mitochondrial precursor proteins destined for the matrix, having a 20 -50-residue N-terminal transit peptide that forms an amphipathic helix (52). Phylogenetic analysis disclosed that the mitochondrial KAS enzyme from A. thaliana and its homologues from M. musculus, H. sapiens, N. crassa, and S. pombe are more closely related to bacterial KAS II enzymes than to those from plastids. This is in accord with the hypothesis that mitochondria and plastids each originated from a different bacterial endosymbiont. With respect to cerulenin sensitivity, however, mtKAS is similar to E. coli KAS I and not to KAS II.
Although acetyl-CoA was shown in 1962 (53) to participate in fatty acid synthesis, it was only recently identified as the primer substrate for KAS III in E. coli (5) and spinach (6). KAS III has the same ␣-␤-␣-␤-␣ fold as KAS I and II, and also carries out a Claisen condensation, although its sequence has little homology to KAS I and II and its active site comprises a cysteine, a histidine, and an asparagine (CHN) rather than CHH as in KAS I and II (46). Because the second substrate utilized by KAS III is malonyl-ACP, it is not surprising that both a CoA and an ACP docking site in KAS III for delivering the substrates to the active site have been identified in crystallographic studies (54,55), whereas a single ACP docking site is predicted for KAS I and II. The presence of only one KAS enzyme in A. thaliana mitochondria infers an ability to carry out both the priming and the elongating reactions. Our results demonstrate that mtKAS has this capacity. Specifically, (i) the purified His-tagged mtKAS synthesized C 4 -ACP using malonyl-ACP as the primer substrate and (ii) in vivo and in the in vitro elongation assays, mtKAS complemented the KAS II defect in the E. coli strain CY244 extending C 4 -ACP in iterative reactions to yield C 18 -ACP. This broad substrate specificity of mtKAS (Fig. 1C) is similar to that of the KAS domain of the cytosolic mammalian type I FAS (1), for example, and dissimilar to the restricted substrate specificities characterizing most plastid and bacterial KAS enzymes in type II FAS systems.
The seminal characterization of E. coli KAS II and its comparison to KAS I revealed that both enzymes could use acetyl-ACP as the priming substrate (56). Subsequently, the same was shown for spinach KAS I (7). Thus, mtKAS is not unique in being able to use acetyl-ACP as well as longer acyl-ACPs in fatty acid synthesis, but its efficiency in using acetyl-ACP is presumably improved compared with that of the mentioned E. coli and plastidial enzymes. To address the question of the specificity of mtKAS for its primer substrate, additional assays employing the purified His-tagged mtKAS were carried out. The results disclosed that labeled acetyl-CoA in the presence of malonyl-ACP failed to serve as a primer substrate under the experimental conditions used, inferring that mtKAS lacks acetyl-CoA transacylase activity. By comparison, in analogous assays, the purified His-tagged E. coli KAS II incorporated minor amounts of label into C 4 -ACP from acetyl-CoA, revealing that, in contrast to KAS I (57), it has acetyl-CoA:ACP transacylase activity. The apparent inability to use acetyl-CoA as a primer substrate exhibited by the purified A. thaliana mtKAS in vitro fits with the observations that mitochondrial extracts from pea leaves incorporated only insignificant amounts of label from acetate into fatty acids, which could have been caused by traces of plastid contamination (17), and that the extracts lack detectable acetyl-CoA:ACP transacylase activity (58). We are unable to explain why, in the latter study, the label from acetyl-CoA was "readily incorporated" into fatty acids in the presence of malonyl-ACP (58). Given that mitochondria, in contrast to plastids, lack ACCase activity (17,59), it is not surprising that acetyl-CoA is an unlikely carbon source for mitochondrial fatty acid synthesis and that malonate assumes this role. Although the source of the malonate is still a question, the enzyme activities to produce malonyl-ACP from malonate are present in mitochondria (58).
Our results lead to the conclusion that malonyl-ACP provides both the primer and donor substrate in the first condensation reaction carried out by A. thaliana mtKAS (Fig. 1C). If the active site structure of mtKAS resembles that of the E. coli KAS I and II enzymes (46), the following scenario can be envisaged. The first malonyl-ACP enters the decarboxylating pocket resulting in a C 2 primer substrate, acetyl-ACP, that is transferred to the active site cysteine of the acyl binding pocket, and the concomitant release of the ACP from the enzyme. Decarboxylation of a second malonyl-ACP yields the extender C 2 -unit, which is also transferred to the active site cysteine effecting synthesis of a C 4 acyl chain. The latter is released from the cysteine and linked to the phosphopantetheine arm of an ACP molecule in the docking site, enabling its exit from the active site. That one compound provides both primer and extender units also occurs during synthesis of the derailment product triacetic acid lactone in E. coli (60,61). In this case the initial C 4 acyl chain serves as a substrate for an additional extension, giving a C 6 acyl chain that upon release from the enzyme cyclizes, yielding triacetic acid lactone.
As a step toward explaining the function of FAS in mitochondria, the nature of the fatty acyl chains synthesized in this organelle has been the subject of several earlier investigations. Studies using Neurospora mitochondria and soluble extracts revealed major amounts of C 8 and C 14 acyl chains (18), whereas those with pea mitochondria disclosed C 10 to C 14 acyl chains (17). Subsequently, using a more sensitive detection system, C 8 , C 16 , and C 18 were identified as the predominant acyl chains synthesized by FAS in soluble mitochondrial pea extracts (58). Our results with the A. thaliana mtKAS expressed in a heterologous system also reveal a bimodal distribution, in this case C 8 versus C 14 and C 16 . Such a distribution was only obtained when the KAS components of the E. coli FAS complex were replaced by mtKAS. This infers that the bimodality, noted in earlier work with mitochondrial extracts (17,18), may well reflect activity of the mtKAS component in the extracts. Even though the FAS complex in E. coli produces C 8 acyl chains for lipoic acid, C 14 acyl chains for lipid A, and C 16 and C 18 acyl chains for membrane phospholipids, we have never detected in our assay systems using KAS I and II any suggestion for a phenomenon similar to the bimodality characterizing mtKAS. Our combined results infer that, although the KAS component of the mitochondrial FAS plays a major role in determining the observed fatty acid distributions in this organelle, acyltransferases, such as PlsB and LpxA, plus lipoic acid synthase (LipA) are the predominant factors in E. coli (62)(63)(64).
An intriguing question for the future will be to probe the structural basis of the observed bimodality. Interestingly, the capacity of mtKAS to synthesize longer acyl chains is more readily lost than that to synthesize the shorter ones as the time of storage before assaying increases. Although today it is accepted that the C 8 acyl chains are precursors of lipoic acid in vivo, the function(s) of the longer ones remain to be clarified. The longer acyl chains are deduced to play a vital role, as C 14 acyl chains are found in Neurospora on subunits of the mitochondrial cytochrome c oxidase (65) and NADH dehydrogenase (66) and because mutants of BPL1 (biotin ligase protein) in S. cerevisiae with normal levels of lipoic acid are respirationdeficient as the result of a failure in mitochondrial fatty acid synthesis (67).