The acyl-AMP ligase FadD32 and AccD4-containing acyl-CoA carboxylase are required for the synthesis of mycolic acids and essential for mycobacterial growth: identification of the carboxylation product and determination of the acyl-CoA carboxylase components.

Mycolic acids are major and specific long-chain fatty acids of the cell envelope of several important human pathogens such as Mycobacterium tuberculosis, M. leprae, and Corynebacterium diphtheriae. Their biosynthesis is essential for mycobacterial growth and represents an attractive target for developing new antituberculous drugs. We have previously shown that the pks13 gene encodes condensase, the enzyme that performs the final condensation step of mycolic acid biosynthesis and is flanked by two genes, fadD32 and accD4. To determine the functions of the gene products we generated two mutants of C. glutamicum with an insertion/deletion within either fadD32 or accD4. The two mutant strains were deficient in mycolic acid production and exhibited the colony morphology that typifies the mycolate-less mutants of corynebacteria. Application of multiple analytical approaches to the analysis of the mutants demonstrated the accumulation of a tetradecylmalonic acid in the DeltafadD32::km mutant and its absence from the DeltaaccD4::km strain. The parental corynebacterial phenotype was restored upon the transfer of the wild-type fadD32 and accD4 genes in the mutants. These data demonstrated that both FadD32 and AccD4-containing acyl-CoA carboxylase are required for the production of mycolic acids. They also prove that the proteins catalyze, respectively, the activation of one fatty acid substrate and the carboxylation of the other substrate, solving the long-debated question of the mechanism involved in the condensation reaction. We used comparative genomics and applied a combination of molecular biology and proteomic technologies to the analysis of proteins that co-immunoprecipitated with AccD4. This resulted in the identification of AccA3 and AccD5 as subunits of the acyl-CoA carboxylase. Finally, we used conditionally replicative plasmids to show that both the fadD32 and accD4 genes are essential for the survival of M. smegmatis. Thus, in addition to Pks13, FadD32 and AccD4 are promising targets for the development of new antimicrobial drugs against pathogenic species of mycobacteria and related microorganisms.

Mycolic acids are major and specific long-chain fatty acids of the cell envelope of several important human pathogens such as Mycobacterium tuberculosis, M. leprae, and Corynebacterium diphtheriae. Their biosynthesis is essential for mycobacterial growth and represents an attractive target for developing new antituberculous drugs. We have previously shown that the pks13 gene encodes condensase, the enzyme that performs the final condensation step of mycolic acid biosynthesis and is flanked by two genes, fadD32 and accD4. To determine the functions of the gene products we generated two mutants of C. glutamicum with an insertion/deletion within either fadD32 or accD4. The two mutant strains were deficient in mycolic acid production and exhibited the colony morphology that typifies the mycolate-less mutants of corynebacteria. Application of multiple analytical approaches to the analysis of the mutants demonstrated the accumulation of a tetradecylmalonic acid in the ⌬fadD32::km mutant and its absence from the ⌬accD4::km strain. The parental corynebacterial phenotype was restored upon the transfer of the wild-type fadD32 and accD4 genes in the mutants. These data demonstrated that both FadD32 and AccD4-containing acyl-CoA carboxylase are required for the production of mycolic acids. They also prove that the proteins catalyze, respectively, the activation of one fatty acid substrate and the carboxylation of the other substrate, solving the long-debated question of the mechanism involved in the condensation reaction. We used comparative genomics and applied a combination of molecular biology and proteomic technologies to the analysis of proteins that co-immunoprecipitated with AccD4. This resulted in the identification of AccA3 and AccD5 as subunits of the acyl-CoA carboxylase. Finally, we used conditionally replicative plasmids to show that both the fadD32 and accD4 genes are essential for the survival of M. smegmatis. Thus, in addition to Pks13, FadD32 and AccD4 are promising targets for the development of new antimicrobial drugs against pathogenic species of mycobacteria and related microorganisms.
Mycolic acids, long-chain 2-alkyl, 3-hydroxy fatty acids, are the hallmark of Corynebacterineae, a suprageneric actinomycete group that includes corynebacteria, mycobacteria, nocardia, rhodococci, and other related microorganisms. In mycobacteria, these fatty acids, called eumycolic acids, possess a very long chain (C 60 -C 90 ) and may contain various oxygen functions, cyclopropyl rings, and methyl branches, in addition to the 3-hydroxyl group (1)(2)(3)(4). In contrast, mycolic acids found in other genera consist of homologous mixtures of saturated and unsaturated acids and contain shorter chains, e.g. C 40 -C 50 in nocardomycolic acids and C 22 -C 36 in corynomycolic acids (5,6). Mycolic acids represent the major lipid constituents of the singular cell envelope of Corynebacterineae and are found either as esters of trehalose or esterifying the terminal pentaarabinofuranosyl units of arabinogalactan, a polysaccharide that forms (with peptidoglycan and mycolic acids) the cell wall skeleton of the cell envelope of the Corynebacterineae members (1)(2)(3)(4)(5)(6). Both types of mycolate-containing components have been shown to play a crucial role in the structure and functions of the cell envelope. Mycolic acids attached to the cell wall arabinogalactan are organized with other lipids to form a barrier that contributes to the very low permeability of the envelope of Corynebacterineae and the natural resistance of these microorganisms to various antibiotics (1,(7)(8)(9). Trehalose mycolates have been implicated in numerous biological functions, notably in mycobacterial virulence, in which the structure of the mycolates has been shown to be important for the initial replication and persistence of pathogenic mycobacterial species in their hosts (10 -12).
Numerous studies have been and are currently devoted to the structures and biosynthesis of mycolic acids, primarily because these substances are specific to the Mycobacterium genus, and this metabolism is also the only clearly identified target inhibited by the major antitubercular drug, isoniazid (1)(2)(3)(4). With the reemergence of tuberculosis infections caused by multidrug-resistant strains of Mycobacterium tuberculosis and thereby the need for the development of new tuberculous drugs (13), deciphering the biosynthesis pathway leading to mycolates still represents a major objective of researchers. However, despite the intensive efforts of biochemists over decades (1)(2)(3)12) and, more recently, the help of molecular genetics (4, 10 -12, 14), the biosynthesis pathway leading to mycolic acids is far from being completely understood. Nevertheless, it is currently admitted that, in mycobacteria, the biosynthesis of the very long-chain C 60 -C 90 mycolic acids involves two fatty acid synthases, namely, fatty acid synthase I and fatty acid synthase II (2,4,14). In contrast, only fatty acid synthase I, a synthase necessary for the production of C 16,18 , would be sufficient in bacterial species that synthesize shorter mycolic acids. In mycobacteria, fatty acid synthase I has been shown to be a bimodal system capable of elongating C 16, 18 fatty acids to yield C [22][23][24][25][26] saturated fatty acids that may be either incorporated directly into mycolates as the two-branched chain or used as substrates of the fatty acid synthase II system to form C 48 -C 64 fatty acids, also called the meromycolate chain. The finding that isoniazid strongly and specifically inhibits InhA, an enoyl-ACP-reductase that belongs to fatty acid synthase II (15,16), is consistent with the proposed biosynthetic pathway leading to the C 60 -C 90 mycolic acids found in mycobacteria and the restricted inhibitory activity of this drug on mycobacteria.
Because all types of mycolic acids display a common 2-alkyl 3-hydroxy structural feature (1-6), the mycolic motif ( Fig. 1), the enzymes involved in the formation of this motif represent good potential targets for the development of new and specific drugs against Corynebacterineae. Two mechanisms have been proposed for the synthesis of the C 32 corynomycolic acid (17): (i) a "Claisen-like" condensation, in which a palmitoyl thioester is condensed with either another palmitoyl CoA or a palmitaldehyde to give a 2-alkyl, 3-oxo ester; and (ii) a "malonic" condensation, similar to the reaction catalyzed by ␤-ketoacyl synthase, in which a palmitoyl CoA is condensed with a tetradecylmalonyl-CoA, followed by decarboxylation. In both cases, an enzyme, the condensase, is required for the condensation of the two fatty acyl substrates to yield 2-alkyl, 3-keto ester that subsequently has to be reduced to produce the C 32 corynomycolic acid. We have recently identified Pks13 as the condensase (18). This enzyme contains the required enzymatic domains for the condensation reaction, and the gene encoding this protein has been found in all members of the Corynebacterineae group analyzed. In Corynebacterium glutamicum, the inactivation of the pks13 gene completely abolishes the production of corynomycolates and, as a consequence, alters the structure of the cell envelope of the mutant strain (18). In mycobacteria, we demonstrated that the pks13 gene is essential for the mycobacterial growth (18). The remaining important question, addressed in the present work, is which of the two condensation mechanisms (Claisen-like or malonic reaction) is used by Corynebacterineae to synthesize the mycolic acid motif. Interestingly, two genes that may encode proteins involved in the activation of the two substrates of the condensase flank the pks13 gene in all Corynebacterineae examined (18). One of these genes, fadD32, predicted to encode an acyl-CoA synthase, was recently shown to be an acyl-AMP ligase (19). The other, accD4, is predicted to encode a subunit of an acyl-CoA carboxylase, a class of enzyme catalyzing the formation of carboxylated acyl-CoA. The present study was undertaken in order to determine the roles of the candidate enzymes in mycolic acid biosynthesis and to evaluate their importance for the physiology of corynebacteria and mycobacteria.
Computer Analysis-M. tuberculosis strain H37Rv and M. leprae DNA sequences were obtained from the Pasteur Institute web site (www.pasteur.fr). Research of FadD32 and AccD4 orthologs on M. smegmatis, C. diphtheriae, C. glutamicum, and C. efficiens genomes was performed at the National Center for Biotechnology Information web site (www.ncbi.nlm.nih.gov/) using the BLAST program. Multiple sequence alignments were performed using Clustal W (22) at the Pasteur Institute web site. The sequence of AccD4 was compared with the ␣ and ␤ subunits of the carboxyltransferase of E. coli (encoded by gene accA and accD) using the Needleman-Wunsh program on the Pasteur Institute web site.
The Mascot search engine (Matrix Science, London, UK) was used for protein identification with tandem mass spectrometry (MS/MS) by searching in nonredundant NCBInr TrEMBL and Swiss-Prot databases. Because most of the M. smegmatis protein sequences were not found in these databases, eight M. smegmatis proteins corresponding to the candidate subunits for being an acyl-CoA carboxylase (namely, the orthologs of AccA1, AccA2, AccA3, AccD1, AccD2, AccD3, AccD4, and AccD5 from M. tuberculosis) were added to the databases. The identification resulting from a Mascot search was confirmed by manual interpretation of corresponding MS/MS data.
FIG. 1. A proposed scheme for the terminal steps of mycolic acid biosynthesis. R1 and R2 correspond to alkyl chains of various sizes that vary according to the Corynebacterineae species. X1 corresponds to the carrier molecule on which the newly synthesized ␤-ketoacyl and mycoloyl residues are transferred.
Construction of the C. glutamicum ⌬fadD32::km and ⌬accD4::km Mutants and Complementation Plasmids-The C. glutamicum ⌬fadD32::km and ⌬accD4::km mutants were produced using a strategy described previously (18). Two DNA fragments (each of 0.8 kb in length) overlapping the fadD32 gene at its 5Ј and 3Ј extremities were amplified by PCR from C. glutamicum total DNA using primers Faddel1 ϩ Fad-del5 and Faddel3 ϩ Faddel4 (Table I). These two fragments were inserted flanking a kanamycin resistance cassette (km) from plasmid pCGL243 (23) into the vector pMCS5 (MoBiTec, Göttingen, Germany) to give pMCS⌬fad. The same strategy was used to produce a plasmid in which two DNA fragments (each of 0.7 kb in length) overlapping the 5Ј and 3Ј extremities of the accD4 gene of C. glutamicum were cloned flanking the km resistance cassette. These two PCR fragments were obtained using primers Accdel1 ϩ Accdel2 and Accdel3 ϩ Accdel4 (Table I), and the plasmid was named pMCS⌬acc. The two plasmids were transferred into C. glutamicum by electroporation, and transformants were selected on plates containing Km. Transformants in which allelic replacement had occurred between the wild-type chromosomal fadD32 or accD4 genes and the mutated plasmid-borne alleles were identified on the basis of their rough colony morphology and characterized by PCR using primers Fad2, Fad4, Fad5, pks2, k7, and k10 for the mutant within the fadD32 gene and AccI, AccII, AccIII, Acc5, k7, and k10 for the insertion within accD4 (Table I). Two mutant strains, C. glutamicum ⌬fadD32::km and ⌬accD4::km (named CGL2034 and CGL2030, respectively), were retained for further studies.
To complement the ⌬fadD32::km mutation, a 2.3-kb DNA fragment carrying the fadD32 gene plus the 370 bp upstream of the putative start codon of this gene was amplified by PCR using primers Fad3 ϩ Fad4 (Table I) and C. glutamicum genomic DNA. The amplified DNA fragment was inserted between the BglII and SmaI sites of pCGL482, which contains a replicon functional in C. glutamicum (24), to give pCGL2319.
Because the intergenic region between accD4 and the upstream gene, pks13, is just 21 bp, these two genes may be part of the same operon. Because the promotor driving the expression of accD4 is not located just upstream the gene and is not yet characterized, we chose to clone the accD4 gene under the control of another promotor functional in C. glutamicum in order to complement the ⌬accD4::km mutation. A 1.6-kb DNA fragment was amplified by PCR using primers AccNco and pks6 (Table I) from the C. glutamicum genomic DNA. The primer AccNco was designed to introduce a NcoI restriction site overlapping the putative start codon of accD4. The PCR fragment was digested with enzymes NcoI and BglII. A second DNA fragment containing the well-characterized ptrc promotor, which was previously shown to be functional in C. glutamicum (25), was generated using primers pKKBam and pKKNco (Table I) and pKK388-1 plasmid (Clontech) as template DNA. This second PCR fragment was digested with NcoI and BamHI. The two PCR fragments were then inserted simultaneously between the BamHI and BglII restriction sites of plasmid pCGL482 to give plasmid pCGL2316. The two plasmids, pCGL2319 and pCGL2316, were used for electrotransformation of the C. glutamicum ⌬fadD32::km and ⌬accD4::km mutants, respectively. Transformants were selected on chloramphenicol-containing plates.
Construction of the M. smegmatis Conditional Mutant Strains-The M. smegmatis conditional mutant strains were produced using the strategy described previously (18). Briefly, two DNA fragments, overlapping the accD4 and the fadD32 genes, were amplified by PCR from M. smegmatis total DNA using primers Acc4a ϩ Acc4b and F32a ϩ F32b, respectively. The fragment overlapping the fadD32 gene was digested with SpeI and EcoRV and inserted into the corresponding sites of plasmid pJQ200 (26) to give pDP59. A hygromycin resistance cassette (hyg) was inserted between the two BamHI sites of pDP65 (located within the fadD32 gene) to yield pDP60. The fragment overlapping the accD4 gene was first inserted into the pGEM-T vector using the pGEM-T vector system I kit (Promega). The hyg cassette was inserted into the unique XhoI site within the cloned accD4 gene to give plasmid pDP62. A DNA fragment containing the disrupted accD4::hyg allele was inserted between the SmaI and SpeI sites of pJQ200 to yield pDP63.
Plasmids pDP60 and pDP63 were transferred into M. smegmatis by electroporation, and transformants were selected on Hyg-containing plates. Transformants in which pDP60 or pDP63 had been integrated by single crossover between the wild-type and mutated copies of fadD32 or accD4 were characterized by PCR using primers F32c, F32d, F32f, H1, and H2 for fadD32 and primers Acc4c, Acc4d, Acc4f, H1, and H2 for accD4. For each construct, one clone giving the pattern corresponding to insertion of the plasmid by single homologous recombination event was retained for further analysis. These strains, obtained with the integration of pDP60 and pDP63, were renamed PMM58 and PMM59, respectively.
To produce the complementation plasmids, the fadD32 and accD4 genes were amplified by PCR from M. smegmatis total DNA using primers F32d ϩ F32f and Acc4d ϩ Acc4f, respectively ( Table I). The PCR fragments were inserted into the pGEM-T vector (Promega) to give plasmids pDP65 and pDP64 containing the fadD32 and accD4 genes, respectively. The fadD32 gene was recovered from plasmid pDP65 on a KpnI/SpeI restriction fragment and inserted between the corresponding sites of plasmid pMIP12. The resulting plasmid, named pDP66, contained the fadD32 gene under the control of the pBlaF* mycobacterial promotor (27). A cassette containing the pBlaF* promotor, the fadD32 gene, and a transcription terminator was obtained on a PacI/AseI restriction fragment and inserted into thermosensitive mycobacterial plasmid pDP24 derived from pCG63 (28) by deletion of a useless AhdI fragment. This new plasmid was named pDP69. A similar strategy was used with the accD4 gene. A PstI/SpeI fragment carrying the accD4 gene without promotor was inserted into the PstI and SpeI sites of pMIP12 to give pDP67. A PacI/AseI cassette from pDP67, containing the accD4 gene now under the control of the pBlaF* promotor, was then transferred into the pDP24 thermosensitive vector to give pDP70.
Plasmids pDP69 and pDP70 were transferred by electroporation into PMM58 and PMM59, respectively. Transformants were selected on plates containing Km and Hyg. The second crossover events at the chromosomal fadD32 and accD4 loci were selected by plating a liquid culture of PMM58:pDP69 and PMM59:pDP70 cultured at 30°C on plates containing Km, Hyg, and Suc, which were then incubated at 30°C. Colonies were screened by PCR using primers F32c, F32d, F32f, H1, and H2 for fadD32 and Acc4c, Acc4d, Acc4f, H1, and H2 for accD4. Two strains, named PMM61:pDP69, in which the wild-type chromosomal copy of fadD32 was replaced by the mutated fadD32::hyg allele, and PMM60:pDP70, in which the wild-type chromosomal copy of accD4 was replaced by the mutated accD4::hyg allele, were selected for phenotypic analysis.
Construction of Plasmids for Immunoprecipitation-To perform the immunoprecipitation experiments in M. smegmatis, two new plasmids containing either the fadD32 gene or the accD4 gene fused to the myc tag were constructed. Briefly, two double-stranded oligonucleotides were produced by annealing either myc-PA with myc-PB or myc-KA with myc-KB in 40 l containing 45 M of each single-stranded oligo-  (Table I), 10 mM Tris, pH 7.5, and 50 mM NaCl. This mixture was incubated for 5 min at 100°C and allowed to cool slowly to reach 4°C. The resulting double-stranded oligonucleotides contained the coding sequence of the myc epitope. They were inserted between the BamHI and PstI sites or the BamHI and KpnI sites of pMIP12 to give pMIP12mycP and pMIP12mycK, respectively. A KpnI/SpeI fragment from pDP66, containing the fadD32 open reading frame, was then inserted between the KpnI and SpeI sites of pMIP12mycK to give pDP80. Similarly, a PstI/SpeI fragment from pDP67, containing the accD4 open reading frame, was cloned into the pMIP12mycP plasmid cut with PstI and SpeI to give pDP81. In these two plasmids, pDP80 and pDP81, the open reading frames corresponding to fadD32 and accD4, respectively, were fused to a sequence encoding the myc epitope. These two plasmids were transferred into the M. smegmatis strains PMM58 and PMM59. The second crossover events at the chromosomal fadD32 and accD4 loci were selected by plating a liquid culture of PMM58: pDP80 and PMM59:pDP81 cultured at 37°C on plates containing Km, Hyg, and Suc. Ten and seven colonies obtained with PMM58:pDP80 and PMM59:pDP81, respectively, were screened by PCR using primers F32c, F32d, F32f, H1, and H2 for fadD32 and Acc4c, Acc4d, Acc4f, H1, and H2 for accD4. Two strains, named PMM61:pDP80, in which the wild-type chromosomal copy of fadD32 was replaced by the mutated fadD32::hyg allele, and PMM60:pDP81, in which the wild-type chromosomal copy of accD4 was replaced by the mutated accD4::hyg allele, were selected for the immunoprecipitation experiments.
Biochemical Analysis of C. glutamicum Strains-Cultures of C. glutamicum were labeled by incubation of exponentially growing cells with 0.5 Ci/ml [ 14 C]acetate (specific activity, 54 mCi/mmol; ICN, Orsay, France) to late growth phase (18). Fatty acids were prepared from the labeled cells and separated by analytical TLC on Durasil 25 using dichloromethane as the eluent as described previously (29). Labeled compounds were quantified on a PhosphorImager (Amersham Biosciences).
Fatty acid methyl esters were obtained by saponification of cells, followed by methylation with diazomethane (29). They were fractionated on a Florisil column irrigated with petroleum ether containing 0%, 1%, 2%, 3%, and 100% diethylether. The last eluted fraction, which contained polar fatty acid methyl esters, was then analyzed by gas chromatography (GC) coupled with mass spectrometry (MS). GC-MS analyses were performed on a Hewlett-Packard 5889 X mass spectrometer (electron energy, 70 eV) working in the electron impact mode, coupled with a Hewlett-Packard 5890 series II gas chromatograph fitted with a similar OV1 column (0.30 mm ϫ 12 m). The temperature program was from 60°C to 100°C, at 20°C/min, and then 100°C to 310°C at 8°C/min.
Synthesis and Analysis of Tetradecylmalonic Acid-The tetradecylmalonic acid was obtained by alkaline deacylation of tetradecylmalonic ester that was prepared according to the method described previously for the preparation of n-butylmalonic ester (30). Briefly, the malonic dimethyl ester (500 mg) was gradually added to a freshly prepared solution of 1.5 M (4 ml) sodium ethoxide, and then, after cooling to about 50°C, the tetradecyl bromide (1 g) was slowly added. After heating the reaction mixture under reflux over a period of about 1 h followed by neutralization with acetic acid, the greater part of the alcohol was distilled, and the ester of tetradecylmalonic acid was extracted by diethylether from the residue diluted with water (5 ml). The tetradecylmalonic acid obtained by alkaline deacylation of this ester was subsequently purified by recrystallization in petroleum ether. The 1 H and 13 C NMR spectra were recorded in CDCl 3 using a Bruker X500 operating at 500 and 125 MHz, respectively.
Immunoprecipitation of accD4 and Analysis of Associated Proteins-Mycobacterial cell extracts were prepared from a 50-ml culture of strains PMM61:pDP80 and PMM60:pDP81 of M. smegmatis. Bacteria were pelleted at 3500 rpm for 10 min, washed once with phosphatebuffered saline ϩ Tween 80 (0.05%), and resuspended in 1 ml of TIP buffer (50 mM Tris, pH 8, 150 mM NaCl, 1 mM dithiothreitol). Glass beads (600 l) were added to an equal volume of the bacterial suspension. Cells were disrupted by agitation for 3 ϫ 1 min at the maximum speed in a Mini BeadBeater (31). The extracts were recovered and centrifuged for 5 min at 12,600 rpm. The supernatants (1 mg of protein) were precleared for 1 h at 4°C with Sepharose-coupled G protein (200 l) on a rotating wheel followed by a brief centrifugation at 13,000 rpm. The precleared supernatant was incubated overnight at 4°C with 10 g of a monoclonal antibody raised against the myc epitope (clone 9E10; Sigma). Then, Sepharose-coupled G protein was added for 12 h at 4°C. After extensive washing with TIP buffer, the beads were incubated overnight with 220 l of TIP buffer containing 1 mg of a synthetic peptide (MASMQKLISEEDL) corresponding to the myc epitope (a generous gift from Dr. C. Marty-Detraves). Extracts were loaded on a 40% sucrose cushion, and beads were separated from the eluted proteins by centrifugation at 14,000 rpm for 1 min. The supernatants (ϳ220 l) were recovered and analyzed by SDS-PAGE. For the immunoblotting analysis, the polyacrylamide gels were transferred onto polyvinylidene difluoride membrane. Immunoblotting was performed with either monoclonal anti-myc antibody (clone 9E10; Sigma) followed by horseradish peroxidase-conjugated anti-mouse IgG or streptavidin coupled with horseradish peroxidase and detected by chemiluminescence (ECL; Amersham Biosciences). For Western blot analysis, 6 l of immunoprecipitants (from a total volume of 220 l) were loaded on the gel.
For identification of proteins by MS, 150 l of immunoprecipitants were first separated by SDS-PAGE (8% polyacrylamide). The polyacrylamide gel was stained with Coomassie Blue, and the protein bands were subjected to in-gel tryptic digestion by using modified trypsin (Promega). Tryptic peptides were subjected to nanoliquid chromatography-MS/MS analysis on an ESI-Qq-Tof mass spectrometer (QSTAR Pulsar; Applied Biosystems, Foster City, CA) operating in positive mode with a 2.1-kV nanospray voltage. Chromatographic separation of peptides was performed on a 75 m (inner diameter) ϫ 15-cm PepMap C18 precolumn after loading onto a 300 m (inner diameter) ϫ 5-mm PepMap C18 precolumn (LC Packings; Dionex) using a linear gradient of increasing acetonitrile in water (5-45%) over 40 min using 0.1% (w/v) formic acid as ion pairing agent. Throughout the running of the liquid chromatography gradient, MS and MS/MS data were recorded continuously based on a 5-s cycle time. Within each cycle, MS data were accumulated for 1 s, followed by two collision-induced dissociation (MS/MS) acquisitions of 2 s each on the two most abundant ions. Dynamic exclusion was employed to prevent repetitive selection of the same ions within a preset time. Collision energies were programmed to be adjusted automatically according to the charge state and mass value of the precursor ions. The Mascot search engine (Matrix Science) was used for protein identification with MS/MS by searching in nonredundant NCBInr TrEMBL and Swiss-Prot databases as described above.

RESULTS
Sequence Analysis of FadD32 and AccD4 -We recently identified Pks13 as the condensase, the enzyme that catalyzes the condensation reaction between two activated fatty acyl substrates to yield mycolic acid precursors (18). The pks13 gene is flanked, on the genome of every Corynebacterineae analyzed, by two genes named fadD32 and accD4 (18). The FadD32 protein of M. tuberculosis has been recently shown to belong to a family of enzymes, the acyl-AMP ligases, involved in the activation of fatty acids as acyl-adenylates before transfer onto polyketide synthases (19). This finding suggested that FadD32 catalyzes the activation of the meromycolate chain before its transfer onto Pks13 in mycobacteria. Analysis of the amino acid sequence of FadD32 from C. glutamicum, C. efficiens, C. diphtheriae, and Rhodococcus rhodochrous revealed that the region that typifies the acyl-AMP ligase is highly conserved (Fig. 2A). Therefore, we hypothesized that FadD32 was involved in the activation of the R1 chain ( Fig. 1) in all the Corynebacterineae by a mechanism similar to that proposed for M. tuberculosis (19).
In silico analysis of the sequence of the putative AccD4 protein of M. tuberculosis shows high sequence similarities with the carboxyltransferase subunit of acyl-CoA carboxylase. In E. coli, this carboxyltransferase is itself subdivided into two subunits, the ␣ and ␤ subunits, encoded by the accA and accD genes, respectively. In the Corynebacterineae analyzed, the accD4 gene encodes a protein that exhibits similarities with the ␣ subunit of the E. coli carboxyltransferase at its COOH-terminal end and with the ␤ subunit of the E. coli carboxyltransferase at its NH 2 -terminal end (Fig. 2B). Therefore, in Corynebacterineae, the accD4 gene putatively encodes a protein corresponding to an entire carboxyltransferase enzyme. This enzyme may be involved in the carboxylation of acyl-CoA to form alkylmalonyl-CoA, one of the expected substrates of the condensase enzyme.
Deletion of fadD32 and accD4 in C. glutamicum and Analysis of the Phenotypical Changes of the Mutants-To investigate the roles of the acyl-AMP ligase FadD32 (19) and of the putative carboxyltransferase subunit of acyl-CoA carboxylase AccD4 in the biosynthesis of mycolic acids, we deliberately chose to delete the wild-type copy of the corresponding genes in C. glutamicum ATCC13032. This strategy was based on the established fact that corynebacterial species can survive without producing mycolates (18), consistent with the existence of the species C. amycolatum (32). In contrast, the inhibition of mycolic acid production has been convincingly shown to be lethal for mycobacteria, explaining why isoniazid, an inhibitor of mycolic acid synthesis, exhibits a bactericidal effect on these bacteria (33,34). Accordingly, two C. glutamicum strains with mutations in either the fadD32 or accD4 gene were produced. For this purpose, a chromosomal fragment overlapping the 5Ј and 3Ј ends of either fadD32 or accD4 was amplified by PCR and cloned, flanking a km resistance cassette, into a vector unable to replicate in corynebacteria. The two constructs were independently transferred by electroporation into C. glutamicum, and Km R transformants were analyzed by PCR using various combinations of primers ( Fig. 3; Table I). Several clones gave amplification patterns consistent with allelic replacement of the wild-type copy of either fadD32 or accD4 by the mutated copy with a 1.5-and 1.4-kb internal deletion for fadD32 and accD4, respectively, into which the km cassette was inserted (Fig. 3).
In comparison with the wild-type strain, the ⌬fadD32::km and ⌬accD4::km mutants exhibited significant phenotypical changes. The smooth shiny colonies of the wild-type strain were replaced by rough colonies in both mutants, a phenotype similar to that observed for the mycolate-less ⌬pks13::km mutant (Fig. 4A). As previously noted for this latter strain, the ⌬fadD32::km and ⌬accD4::km mutants, but not the wild-type strain, aggregated strongly in liquid cultures. These unusual phenotypic features of the ⌬fadD32::km and ⌬accD4::km mu-tants were fully reversed to the wild-type phenotype upon transformation of the mutant strains with a plasmid carrying either the entire fadD32 or full-length accD4 from C. glutamicum, indicating that the observed phenotypical changes were due to the deletion of fadD32 and accD4. Therefore, it followed that, although a functional fadD32 or accD4 gene was not required for the viability of C. glutamicum, as expected from our previous data on pks13 (18), the gene deletion resulted in striking phenotypes suggestive of a defect in mycolic acid production and, as a consequence, a cell envelope modification.
Lipid Analysis of the ⌬fadD32::km and ⌬accD4::km Mutants of C. glutamicum-To determine the origin of the phenotypic changes observed for the ⌬fadD32::km and ⌬accD4::km mutants, the fatty acids produced by the mutant strains were isolated and compared with those produced by the parental strain. Cultures of the wild-type, the ⌬fadD32::km, and ⌬accD4::km mutants of C. glutamicum, as well as the mutants complemented with the wild-type copy or either fadD32 or accD4, were grown to exponential growth phase and labeled with [ 14 C]acetate. Fatty acids were then released from the bacterial cells by saponification. TLC and GC analyses of the fatty acid methyl esters showed a production of non-hydroxylated C 16 -C 18 fatty acid derivatives comparable for all the strains but a complete absence of corynomycolates in the ⌬fadD32::km and ⌬accD4::km mutants of C. glutamicum (Fig.  4B). This observation was also supported by the lack of production by the mutant strains of palmitone, a degradation product of the 3-oxo intermediate resulting from the condensation reaction ( Fig. 1) that usually appears upon alkaline hydrolysis (data not shown). As expected, complementation of the mutant strains with the wild-type copy or either fadD32 or accD4 fully restored the production of mycolic acids (Fig. 4B). These observations were confirmed by GC-MS analysis (data not shown) and clearly demonstrated that the ⌬fadD32::km and ⌬accD4::km mutants of C. glutamicum were devoid of mycolic acids because of the deletion of either fadD32 or accD4.
Synthesis of Tetradecylmalonic Acid and Identification of Carboxylated Intermediates in the ⌬fadD32::km and ⌬pks13::km Mutants of C. glutamicum-Three successive reactions are expected to yield the 3-oxo intermediate resulting from the condensation reaction between two fatty acid molecules (Fig. 1). According to this model, the deletion of either fadD32 or pks13, but not accD4, may lead to the accumulation of carboxylated intermediates, provided that such an accumulation is not prevented by a possible strictly regulated enzymatic activity and/or a polar effect on the expression of the downstream gene accD4 due to the insertion of the km cassette in the fadD32 or pks13 gene. The polar effect that may be due to the insertion of the km cassette in fadD32 was ruled out because the expression of the wild-type fadD32 gene in the ⌬fadD32::km mutant restored the production of mycolates (Fig.  4B). In the case of the ⌬pks13::km mutant, however, the transfer of a plasmid containing the wild-type pks13 gene did not fully complement the mycolate synthesis defect (18), suggesting that the km cassette located within the pks13 gene has led to an accD4 expression defect in the ⌬pks13::km mutant.
To validate the proposed condensation model further, we searched for the carboxylated intermediates that may accumulate in the ⌬fadD32::km and/or ⌬pks13::km mutants. Because the putative intermediates may not be easily identifiable in fatty acid mixtures, we first synthesized the compound predicted from the expected carboxylation of the major C 16 fatty acid by an acyl-CoA carboxylase, i.e. a tetradecylmalonic acid. The alkylmalonate was purified by recrystallization (30) and characterized. GC-MS analysis of the dimethylester of tetradecylmalonic acid (Fig. 5A) showed a pseudomolecular (Mϩ1, 1.7%) ion peak at 329 m/z and fragments at 297 m/z (M-31, 0.6%), 145 m/z (␤ cleavage, 100%), and 132 m/z (McLafferty cleavage, 91%). The 1 H NMR spectrum of the compound (Fig.  5B) showed peaks attributable to the resonances of terminal CH 3 at 0.80 ppm (3H, triplet), of (CH 2 ) n at 1.18 ppm (24H, multiplet), ␤ CH 2 at 1.81 ppm (2.4 H, multiplet), ␣ CH at 3.29 ppm (1H, triplet), and OCH 3 at 3.65 ppm (6.5H, singlet). Analysis of the 13 C NMR spectrum (Fig. 5C) (Table I) are indicated. The expected PCR amplification products for the various strains are indicated below each genetic structure. B, PCR analysis of the ⌬fadD32::km and ⌬accD4::km mutants and wild-type (WT) strain of C. glutamicum.
Based on the characteristic mass value of the McLafferty fragment of the synthetic alkylmalonate (132 m/z instead of 74 m/z for straight chain fatty acids), it was possible to focus on this mass value and thereby to identify a peak eluted at 15.43 min on GC of polar fatty acid derivatives from the ⌬fadD32::km mutant whose mass spectrum contained an intense ion at 132 m/z (Fig. 6). A less intense signal of the 132 m/z ion peak was also detected in the mass spectra of the wild-type strain of C. glutamicum but was absent from that of the ⌬accD4::km mutant (Fig. 6). Interestingly, this ion peak was observed in the mass spectrum of the polar fatty acid methyl esters of the ⌬accD4::km mutant complemented with the wild-type accD4 gene (Fig. 6). Furthermore, the entire mass spectrum of the natural compound (Fig. 6), typified by the occurrence of the 132 m/z ion fragment, was superimposable with that of the synthetic tetradecyl malonic acid methyl ester (Fig. 5A). It was thus concluded that the wild-type C. glutamicum, but not the ⌬accD4::km mutant, contained an alkylmalonate derivative whose mass spectrum corresponded to that of the expected activated C 16 fatty acid derivative destined to be condensed with the product of FadD32 to yield the 3-oxo intermediate of major C 32 corynomycolic acids. This compound was accumulated in the ⌬fadD32::km mutant, and a higher homolog of the fatty acid, i.e. hexadecylmalonic acid, was also detected in this mutant as a minor peak eluted at 17.51 min on GC (Fig. 6). Consistent with our finding that the wild-type pks13 gene did not fully complement the mycolate synthesis defect of the ⌬pks13::km mutant (18), only small amounts of the alkylmalonate were detected in the ⌬pks13::km mutant (Fig. 6), probably because of a polar effect of the km insertion within pks13, a gene ending just 20 bp upstream of the start codon of accD4.
Both FadD32 and AccD4 Are Essential for the Viability of Mycobacteria-Mycolic acid biosynthesis is known to be essential for mycobacterial growth (18,33,34). Consequently, FadD32 and AccD4 are expected to be essential in mycobacteria if the enzymes are not redundant. To address this question, a genetic approach was used in the model strain M. smegmatis mc 2 155. Two non-replicative vectors containing the counterselectable marker sacB (35) and a mutated copy of either fadD32 or accD4 were inserted into the chromosome by single crossover between the wild-type chromosomal allele of fadD32 or accD4 and the mutated alleles, to yield PMM59 and PMM58, respectively (Fig. 7). Plating a culture of each strain at 30°C on medium containing 5% Suc and Hyg generated clones with mutations in the sacB gene but failed to select the second recombination event that would have produced a strain carry- ing only the mutated fadD32::hyg or accD4::hyg allele (data not shown). These results suggested that both genes are essential for mycobacterial growth, consistent with the demonstrated involvement of these genes in mycolic acid production in C. glutamicum (see above). To firmly establish that the two genes are essential in mycobacteria, we transferred a second wild-type copy of fadD32 or accD4 in a thermosensitive mycobacterial vector into PMM59 and PMM58, respectively. In these constructs, named pDP69 for fadD32 and pDP70 for accD4, the expression of the cloned genes was under the control of the mycobacterial promotor pBlaF* (27). In this genetic context, the selection of clones resistant to both Suc and Hyg gave several mutants (3 of 5 Suc R and 10 of 10 Hyg R colonies tested for fadD32 and accD4, respectively) in which a second recombination event had occurred in PMM58:pDP69 and PMM59:pDP70 between the two chromosomal alleles of fadD32 or accD4, leaving only the non-functional copy of either fadD32 or accD4 on the chromosome (Fig. 7). These recombinant strains, PMM61:pDP69 and PMM60:pDP70, contained a deletion and insertion in the fadD32 gene or an insertion in the accD4 gene on the chromosome and a functional fadD32 or accD4 gene on a thermosensitive plasmid. Streaking these recombinant strains on Hyg plates at 30°C or 42°C revealed that they were unable to form colonies at high temperature (Fig. 8). In liquid culture, the PMM60:pDP70 strain grew as well as the wild-type strain at 30°C, a permissive temperature for plasmid replication (Fig. 8). In the case of PMM61:pDP69, there was a weak growth defect at 30°C illustrated by a colony size smaller for PMM61:pDP69 than for PMM60:pDP70 and the wild-type strain (Fig. 8A) and a delay in obtaining a liquid culture of the same optical density at 30°C. When the cultures were shifted to 42°C, a non-permissive temperature for plasmid replication, the number of viable bacteria increased during the first 12-24 h, during which time the temperature-sensitive plasmid was cured, and then remained stable during the next 24 h before declining (data not shown). After 100 h of growth at 42°C, the number of viable bacteria was unchanged for PMM61:pDP69, compared with that inoculated, and 10 times higher for PMM60:pDP70 (Fig. 8). This higher cfu count for PMM60:pDP70 may be due to a higher stability of the pDP70 plasmid. In any case, these data demonstrated that, as expected, both fadD32 and accD4 are essential for the survival of M. smegmatis.
Identification of the Other Subunits of the AccD4-containing Acyl-CoA Carboxylase-To be functional, an acyl-CoA carboxylase requires, in addition to the carboxyltransferase subunit, a biotin carboxyl carrier protein and a biotin carboxylase (36). In M. tuberculosis, four genes (accA1, accA2, accA3, and pca) encode proteins containing the biotin carboxyl carrier and biotin carboxylase domains (37) that may thus be part of the AccD4-containing acyl-CoA carboxylase. All four genes are highly conserved in M. smegmatis, but only one of these genes, namely, Ml0726 (the ortholog of accA3 in M. tuberculosis), is found in M. leprae, indicating that the AccA3 protein is a good candidate for being the requested portion of the acyl-CoA carboxylase in mycobacteria. In corynebacteria, two candidate genes for being the biotin carboxyl carrier protein and biotin carboxylase exist, namely, Ncgl0670 and Ncgl0659. Ncgl0659 is a putative pyruvate carboxylase, whereas Ncgl0670, which is homologous to both Ml0726 of M. leprae and AccA3 of M. tuberculosis, is probably essential for bacterial growth and involved in the biosynthesis of fatty acids (38). To experimentally address the question of the biotin carboxyl carrier protein and biotin carboxylase subunits of the AccD4-containing acyl-CoA carboxylase involved in the formation of mycolate, we used a novative proteomic approach by looking for proteins that interact with AccD4. For this purpose, we constructed a plasmid, pDP81, carrying the accD4 gene fused with a myc tag. To demonstrate that the tagged AccD4 is functional, the pDP81 plasmid was transferred in M. smegmatis PMM59. The second recombination event at the accD4 chromosomal locus was selected by plating a culture of PMM59:pDP81 on Hyg-, Suc-, and Km-containing medium. Analysis of seven Suc R , Hyg R , and Km R colonies by PCR showed that a second recombination event had occurred in five of these clones between the two chromosomal alleles of accD4 (data not shown). Because these strains are viable despite the mutated copy of accD4 on the FIG. 6. Gas chromatography-mass spectrometry analysis of the fatty acid methyl esters from the wild type (WT), the ⌬fadD32::km and ⌬accD4::km mutant strains, and the ⌬accD4::km mutant complemented with a plasmid carrying accD4 of C. glutamicum. An equal amount (10 g) of the polar fraction of fatty acid methyl esters from the various strains was injected in the gas chromatograph, and the ion current due to the McLafferty ion fragment (132 m/z) expected for an alkylmalonate was measured. chromosome, it was concluded that the fused accD4-myc gene carried by the pDP81 plasmid encodes a functional carboxyltransferase. Using a similar protocol, a control strain was produced containing a mutated fadD32 gene on the chromosome and a fadD32-myc fused gene on a plasmid named pDP80. Once again, the viability of the constructed strain demonstrated that the fadD32-myc fused gene encodes a functional acyl-AMP ligase.
To identify proteins that specifically interact with the myctagged AccD4, we carried out immunoprecipitation of proteins extracts of M. smegmatis obtained from PMM61:pDP80 or PMM60:pDP81 using a monoclonal antibody directed against the myc epitope. The immunoprecipitated proteins were analyzed by SDS-PAGE stained with silver nitrate (Fig. 9A, lane  A). A single protein band was obtained with the control strain PMM61:pDP80 (Fig. 9A, lane B). This protein had the expected molecular mass of FadD32-myc (70 kDa) and reacted with the anti-myc antibody (Fig. 9, A and B). In contrast, two protein bands were detected in the protein extracts from PMM60:  (Table I) are indicated. The expected PCR amplification products for the various strains are indicated below the last genetic structure. B, PCR analysis of the conditional mutants PMM61:pDP69 and PMM60:pDP70 and their parental strains PMM58, PMM59, and mc 2 155 (WT). pDP81 (Fig. 9A, lane D). The minor band, exhibiting the lowest molecular mass, reacted in Western blot analysis using the anti-myc antibody, demonstrating that it contained AccD4-myc (Fig. 9B, lane D). The major protein band (Fig. 9A, lane D), which was expected to contain the biotin carrier protein and the biotin carboxylase subunit of the acyl-CoA carboxylase, was suspected to be biotinylated. Indeed, a Western blot analysis performed using streptavidin coupled with horseradish peroxidase revealed several bands in the crude mycobacterial extracts (Fig. 9C, lanes A and C), as expected from the occurrence of numerous biotin-containing enzymes involved in various biological processes. In contrast, a single protein band that corresponded to the highest molecular mass protein was revealed in the immunoprecipitated extracts of the PMM60: pDP81 strain of M. smegmatis (Fig. 9C, lane D) and absent from the immunoprecipitated extracts of PMM61:pDP80 (Fig.  9C, lane B). This experiment demonstrated that a biotinylated protein strongly and specifically interacted with AccD4.
To further characterize the proteins co-immunoprecipitated with AccD4-myc, the various bands obtained on SDS-PAGE were digested with trypsin directly on the gel. The resulting peptides were extracted from the gel and analyzed by mass spectrometry and proteomic approaches. As expected, the minor protein band (Fig. 9A, lane D) that exhibited the lowest mobility on SDS-PAGE and reacting with the myc antibody (Fig. 9B, lane D) was found to contain AccD4. More importantly, the peptide sequences obtained from the mass spectrometry analysis of the major protein band that reacted with streptavidin (Fig. 9C, lane D) corresponded with certainty (protein coverage of 30 -40%) to two M. smegmatis proteins (Fig.  9D), namely, the orthologs of AccA3 and AccD5 from M. tuberculosis. Indeed, the finding that AccA3 forms a complex with AccD4 was not surprising because it was in agreement with our bioinformatic analysis data, which identified AccA3 as the best candidate for being the missing subunit of the AccD4-con-taining acyl-CoA carboxylase in mycobacteria (see above). The observation that AccD5 is also part of the complex was more surprising but in agreement with a recent finding of Gande et al. (39), who also showed that the ortholog of AccD5 in C. glutamicum is required for the production of corynomycolate. It must be noted that no peptide sequence that may correspond to orthologs of the remaining candidate subunits for being an acyl-CoA carboxylase (namely, the orthologs of AccA1, AccA2, AccD1, AccD2, and AccD3) was found in the protein digests, showing the specificity of the immunoprecipitation reaction. Taken together, these experiments identified AccA3 and AccD5 as two other subunits of the AccD4-containing acyl-CoA carboxylase. DISCUSSION The present study was undertaken in order to identify new enzymes involved in mycolic acid biosynthesis, a key metabolism in Corynebacterineae that is a target for isoniazid, a powerful antituberculous drug. In their seminal work on the biosynthesis of C 32 mycolic acids, Gastambide-Odier and Lederer (17) have proposed two models for the condensation of two C 16 to yield the corynomycolic acid. In the malonic condensation mechanism, a carboxylated acyl-coenzyme A is condensed to a second activated acyl chain to yield a 3-oxo intermediate, which would then be reduced to form mycolic acid (Fig. 1). We have recently provided strong experimental support for this model by identifying the condensase, the enzyme responsible for the final condensation step in mycolic acid biosynthesis. This enzyme is encoded by the pks13 gene, which seems to form an operon together with two other genes, fadD32 and accD4. We postulated that both AccD4 and FadD32 would be involved in the activation of the condensase substrates because these proteins exhibit similarities with acyl-CoA synthases and subunits of acyl-CoA carboxylases, two classes of enzymes that may be involved in the substrate activation required for the final condensation reaction. Consistent with this hypothesis is the presence of the fadD32 and accD4 genes in the genomes of all the Corynebacterineae examined (18). The recent demonstration that the protein FadD32 of M. tuberculosis belongs to a family of enzymes, the acyl-AMP ligases, involved in the activation of fatty acids as acyl-adenylates before transfer onto polyketide synthase (19) further supported our hypothesis. To establish the functions of accD4 and fadD32, we adopted the strategy of deleting the putative genes of interest from the genomes of bacteria and measuring the consequences of the deletions both in terms of production of mycolic acids and phenotypical changes that typify cells devoid of these structurally important cell envelope constituents. Because previous works have established the essentiality of gene products involved in the biosynthesis of mycolic acids in mycobacteria and not in the phylogenetically related corynebacteria, we first deleted the accD4 and fadD32 genes in C. glutamicum. As expected, the two resulting mutant strains were deficient in mycolic acid production and exhibited the colony morphology that typifies the mycolate-less mutants of corynebacteria. The parental corynebacterial phenotype was restored in both mutants after expression in trans of the wild-type fadD32 and accD4 genes. Furthermore, the ⌬fadD32::km mutant, but not the ⌬accD4::km strain, accumulated a tetradecylmalonic acid, the expected major product of AccD4-containing acyl-CoA carboxylase in corynebacteria, a phenotype that was again fully reversed to that of the wild type by complementation of the mutant with a functional fadD32 gene. These data clearly demonstrated the following: (i) AccD4 is involved in the carboxylation of the fatty acid that will be found as the ␣ chain of the mycolic acid, (ii) Fad32 is required for the activation of the meromycolate chain, probably through the formation of an acyl-adenylate as shown by Trivedi et al. (19); and (iii) these two activation reactions are required for the final condensation step by Pks13 to produce mycolic acid precursors. To confirm and extend our finding, we also attempted to generate knockout mutants in M. smegmatis using conditionally replicative plasmids. We showed that insertion/deletion within either fadD32 or accD4 was lethal for M. smegmatis unless functional merodiploid strains were used, thus indicating that the gene products are essential in mycolic acid biosynthesis and viability in M. smegmatis. As such, these enzymes represent two novel good targets for the development of new antituberculous drugs in the context of re-emergence of multiresistant strains of the tubercle bacillus. Besides, the specific implication of Fad32 in mycolic acid biosynthesis illustrates the complexity of lipid metabolism in mycobacteria that have devoted as many as 36 non-redundant FadD proteins to the activation of fatty acid substrates (37). In agreement with this observation is the result of the inactivation of fadD26 that generates attenuated tubercule bacilli (40) that lack phthiocerol dimycocerosates but not other types of lipids such as mycolic acids (41).
The observation that FadD32 and AccD4, like Pks13 (18), proteins involved in the last condensation step of mycolic acids biosynthesis, are essential for the growth of M. smegmatis is in apparent conflict with two reports that indicate that tempera-ture-sensitive strains isolated from M. smegmatis cultures after chemical mutagenesis are viable without the continued production of full-length mycolic acids and accumulate longchain meromycolate-like fatty acids (42,43). Biochemical analysis of one of the spontaneous mutant strains has shown that it also lacks the characteristic glycopeptidolipids and lipooligosaccharides of the parent strain (44). In addition, the latter mutant strain exhibits an abnormal ultrastructure that has not been observed in mycolate-less bacteria (45). Temperaturesensitive mutants isolated by chemical mutagenesis are known to be highly versatile, and all our attempts to maintain these strains failed and resulted in the selection of either revertants or contaminants. Thus, additional studies that should include genotypic analyses are clearly needed to clarify the status of the strains before drawing any definite conclusion. Another study that may contradict the fact that mature mycolic acids are essential for in vitro growth of mycobacteria is the work of Mdluli et al. (46), who have reported continued growth of M. avium treated with isoniazid concentrations that inhibit mycolate synthesis. Again, the resulting bacteria exhibited altered colony morphologies and an uncommon cell wall ultrastructure different from what was observed in the temperature-sensitive mycolate-less strain (44) and C. amycolatum (45), whose ultrastructure does not significantly differ from FIG. 9. Identification of AccA3 and AccD5 as the other subunits of the AccD4-containing acyl-CoA carboxylase. A, SDS-PAGE of proteins co-immunoprecipitated with the tagged AccD4. The gel was stained with silver nitrate. B, immunoblotting analysis of the immunoprecipitants using a monoclonal anti-myc antibody. C, analysis of the immunoprecipitants using streptavidin. SDS-PAGE of proteins was performed with 8% polyacrylamide gel. The same membrane was used for the two immunoblot analyses. Lane A, crude protein extract from PMM61:pDP80; lane B, immunoprecipitants from PMM61:pDP80; lane C, crude protein extract from PMM60:pDP81; lane D, immunoprecipitants from PMM60: pDP81. For the crude protein extracts, 125 g of the extracts were loaded on the gel. For the immunoprecipitants, 6 l of the extracts (from a total volume of ϳ220 l) were loaded. D, identification of the AccA3 and AccD5 proteins by mass spectrometry analysis and proteomic approaches. The amino acid sequences of the AccA3 and AccD5 proteins from M. smegmatis are presented. The peptides identified by MS/MS from the protein co-immunoprecipitated with AccD4-myc are indicated in bold.
that of mycolate producer strains (1,45). Importantly, we were unable to reproduce the data obtained on M. avium that also contrast strongly with those of Vilchèze et al. (33), who clearly established a correlation between the inhibition of mycolate biosynthesis and the viability of M. smegmatis. These latter data are supported by several recent and genetically wellcontrolled studies that have convincingly demonstrated the essentiality of gene products involved in the biosynthesis of the mycobacterial key cell wall components, notably mycolic acids and arabinogalactan (18,47,48).
As far as the mechanism of the condensation reaction that yield mycolic acids is concerned, it has been postulated from structural considerations that the C 32 corynomycolic acid could result from a condensation of two molecules of C 16 (17). Indeed, the intervention of two molecules of palmitic acid in this synthesis has been first proved by incubating C. diphtheriae cells with [1-14 C]palmitic acid. In the resulting corynomycolic acid, the labeling was specifically found in carbon atoms 1 and 3 (17). Moreover, a particulate cell-free preparation of C. diphtheriae incubated with [1-14 C]palmitate was shown to produce a ␤-keto ester specifically labeled on carbon atoms 1 and 3 (49). Two distinct mechanisms have been proposed for the condensation reaction: (i) a Claisen-like condensation in which a fatty acyl thioester is condensed with another CoA derivative or with palmitaldehyde to give a 2-alkyl, 3-oxo ester or a 2-alkyl, 3-hydroxy derivative; and (ii) a condensation of a palmitoyl CoA with a tetradecylmalonyl-CoA followed by decarboxylation, in a manner similar to the action of the ␤-ketoacyl synthases during fatty acid chain elongation, to produce a 2-alkyl, 3-keto ester, which is subsequently reduced to give corynomycolic acid. The difference between the two mechanisms resides in the existence of a carboxylation step necessary to produce the malonyl CoA derivative. Experiments designed to differentiate between the two mechanisms have led to contradictory conclusions. Whereas the condensation reaction in cell-free preparations of C. diphtheriae was inhibited by avidin, indicating the occurrence of a carboxylation step through the intervention of a biotin enzyme in the reaction sequence (49), avidin showed no effect on the condensation reaction in cell-free extracts of C. matruchotii (50). More recently, labeling experiments using [2,2-2 H]palmitic acid have shown that the deuterium atom ( 2 H) was located at position C-2 in the mature corynomycolic acid. Accordingly, the authors have proposed a mechanism involving a highly activated enolate intermediate to explain the reaction (51) and concluded that palmitate condensation in whole cells of C. matruchotii does not involved an intermediate carboxylation. This conclusion implies, however, that the malonyl intermediate reacts further in a true malonic condensation, i.e. with the formation of a carbanion by the loss of one deuterium. Importantly, it has been convincingly shown in yeast that the malonyl intermediate does not react in this way but rather by a concerted decarboxylation with conservation of the deuterium (52). Consistent with this observation, the present work, by proving the involvement of the AccD4 carboxyltransferase and the accumulation in the ⌬fadD32::km mutant of tetradecylmalonic acid, clearly discriminates between the proposed mechanisms and proves that a carboxylation step to form a malonyl derivative takes place before the condensation reaction in mycolic acid biosynthesis.
Acyl-CoA carboxylases are complex enzymes composed of several catalytic domains. In E. coli, they contain subunits encoded by four genes: accA, accB, accC, and accD (36). Genes accA and accD encode the ␣ and ␤ subunits of the carboxyltransferase, respectively; accC encodes the biotin carboxylase; and accB encodes the biotin carboxyl carrier protein (36). In Corynebacterineae, our sequence analysis data revealed that AccD4 encodes a protein containing both the ␣ and ␤ domains of a carboxyltransferase. To identify the other subunits of the entire acyl-carboxylase enzyme, we combined comparative genomics, molecular biology approaches, and proteomics. A comparison of the genome sequences of various mycobacteria and corynebacteria suggested that the genes named accA3 in M. tuberculosis, Ml0726 in M. leprae, and Ncgl0670 in C. glutamicum were likely to encode the biotin carboxyl carrier subunit and biotin carboxylase subunit of the AccD4-containing acyl-carboxylase. The identity of the protein was then clearly established in co-immunoprecipitation experiments by demonstrating that AccD4 interacts with the M. smegmatis ortholog of AccA3 of M. tuberculosis. In these co-immunoprecipitation experiments, we also identified the ortholog of the M. tuberculosis carboxyltransferase AccD5 as another partner of AccD4. This finding was surprising in view of what is known with regard to E. coli, but it is not unexpected because while this work was under review, Gande et al. (39) have published a series of experiments demonstrating that the ortholog of M. tuberculosis AccD5 in C. glutamicum, called AccD2 by these authors, is required for corynomycolate biosynthesis. Thus, the occurrence of two different AccD protein subunits in an acyl-CoA carboxylase appears to be a unique feature of Corynebacterineae. We now have a scenario for the final condensation reaction of the mycolic acid biosynthesis in which five proteins are involved (Fig. 1): the condensase Pks13, the acyl-AMP ligase FadD32, and the acyl-CoA carboxylase formed by AccD4, AccD5, and AccA3. In their model, Gande et al. (39) have also proposed that the carboxylation of one acyl-chain (the merochain) occurs after transfer of this chain onto Pks13. However, our results do not support this part of the model because an alkyl-malonyl intermediate is formed in a corynebacterial mutant deficient in the synthesis of Pks13. Rather, our finding indicates that the carboxylation of the mero-chain occurs before the transfer onto the condensase, Pks13.
To conclude, our results further extend our understanding of the biosynthesis of mycolic acids, the key lipid components of the mycobacterial cell envelope. New enzymes were identified and shown to play a role in the activation of the substrates of the condensase. These proteins, as well as the condensase, are essential for the viability of mycobacteria and specific for a restricted number of bacterial species. Therefore, they represent new and attractive targets for the development of novel drugs for the treatment of mycobacterial infections in humans.