|Home | About | Journals | Submit | Contact Us | Français|
Mycobacterium tuberculosis is known to synthesize α-, methoxy-, and keto-mycolic acids. We propose a detailed pathway to the biosynthesis of all mycolic acids in M. tuberculosis. Fatty acid synthetase I provides C20-S-coenzyme A to the fatty acid synthetase II system (FAS-IIA). Modules of FAS-IIA and FAS-IIB introduce cis unsaturation at two locations on a growing meroacid chain to yield three different forms of cis,cis-diunsaturated fatty acids (intermediates to α-, methoxy-, and keto-meroacids). These are methylated, and the mature meroacids and carboxylated C26-S-acyl carrier protein enter into the final Claisen-type condensation with polyketide synthase-13 (Pks13) to yield mycolyl-S-Pks13. We list candidate genes in the genome encoding the proposed dehydrase and isomerase in the FAS-IIA and FAS-IIB modules. We propose that the processing of mycolic acids begins by transfer of mycolic acids from mycolyl-S-Pks13 to d-mannopyranosyl-1-phosphoheptaprenol to yield 6-O-mycolyl-β-d-mannopyranosyl-1-phosphoheptaprenol and then to trehalose 6-phosphate to yield phosphorylated trehalose monomycolate (TMM-P). Phosphatase releases the phosphate group to yield TMM, which is immediately transported outside the cell by the ABC transporter. Antigen 85 then catalyzes the transfer of a mycolyl group from TMM to the cell wall arabinogalactan and to other TMMs to produce arabinogalactan-mycolate and trehalose dimycolate, respectively. We list candidate genes in the genome that encode the proposed mycolyltransferases I and II, phosphatase, and ABC transporter. The enzymes within this total pathway are targets for new drug discovery.
Tuberculosis (TB) is one of the world's deadliest diseases (19). The Centers for Disease Control and Prevention has reported that each year, eight million people around the world become sick with TB and there are over two million TB-related deaths worldwide. Moreover, one-third of the world's population is infected with TB. Thus, Mycobacterium tuberculosis (the causative agent of TB) is clearly the most predominant global human pathogen. Two decades ago, it was thought that TB was under control and that it was a matter of time before it would be eradicated. Today, this disease has reestablished itself due to several factors. The lack of drug compliance, the appearance of multiple-drug-resistant strains, and the AIDS epidemic are a few factors that have led to the resurgence in TB. Drug resistance arises following inadequate compliance, and AIDS patients with weakened immune system are very susceptible to M. tuberculosis and the usual cause of death.
The cell envelope of M. tuberculosis is distinctive and is associated with its pathogenicity. Features that are very prominent in the cell envelope are the presence of arabinogalactan-mycolate covalently linked to the cell wall peptidoglycan via a phosphodiester bond located on the inner leaflet of the outer membrane (9, 56, 63, 69) and of a free glycolipid called trehalose dimycolate (TDM), which accumulates in a cord-like fashion on the surface of the cells (72, 73). This provides a thick layer of lipid on the outer part of the cell and protects the tubercle bacillus from noxious chemicals and the host's immune system. Mycolic acids are the major constituents of this protective layer. They also play other important roles as structural components of the cell wall and envelope (5, 11, 55, 65). More specifically, the cyclopropane rings in mycolic acids of M. tuberculosis contribute to the structural integrity of the cell wall complex (39) and protect the bacillus from oxidative stress (hydrogen peroxide) (123).
Deletion of the proximal cyclopropane ring of α-mycolic acid (41) or of methoxy- and keto-mycolates (31) in M. tuberculosis leads to a significant attenuation in growth of the two mutants in the mouse model of infection (structures are shown below). A deletion of the keto-mycolates leads to restricted growth of this mutant in macrophages (124). Thus, the fine structure of mycolic acids is associated with virulence of M. tuberculosis.
As defined by Asselineau and Lederer (8) mycolic acids are β-hydroxy fatty acids with a long α-alkyl side chain. They exist as homologous series of fatty acids differing by 28 atomic mass units (a two-carbon unit), and in M. tuberculosis they are characterized by very hydrophobic C54 to C63 fatty acids with C22 to C24 α side chains. Three distinct structural classes of mycolic acids are found in M. tuberculosis, and they are α-, methoxy- and keto-mycolic acids, as shown in Fig. Fig.1.1. The α-mycolic acid is the most abundant form (>70%), whereas methoxy- and keto-mycolic acids are the minor components (10 to 15%) (82). The α-mycolic acid is a cis, cis-dicyclopropyl fatty acid. There are two structural variations of this mycolic acid, depending on the source. These variations are in the length of the terminal alkyl group and the number of methylene groups between the cyclopropane rings and the carboxyl group. Thus, α-mycolic acids from strains H37Ra and Test represent one group, and α-mycolic acids from strains Brevanne, DT, PN, C, and Canetti represent the other group (6, 35, 38, 66, 68, 82, 100). The α-mycolic acid from the H37Ra strain and the clinical strains are different. Both methoxy- and keto-mycolic acids have structural series containing either cis- or trans-cyclopropane rings (67).
In recent years, many excellent reviews on the biosynthesis of mycolic acids have appeared (7, 11, 30, 49, 50). However, these reviews lack the details on how mycolic acids are synthesized and processed into the final products. Considerable knowledge has accumulated in the last 20 years, and today we are sitting on the cusp of this problem. To move this field forward, we have now critically examined and reviewed numerous old and new data in the literature and used them to develop a novel and plausible scheme of how mycolic acids are synthesized and processed in M. tuberculosis. We have performed computational genetic analyses to try to fill in the gaps in the current knowledge. The validity of parts of this proposed pathway can now be tested experimentally, and a more complete picture should emerge.
A study of how Mycobacterium synthesizes C16 to C18 and C24 to C26 fatty acids in a bimodal product pattern by the malonyl coenzyme A (malonyl-CoA) pathway dates back to the work of Bloch and associates in the early 1970s (16). They showed that Mycobacterium smegmatis contains both type I fatty acid synthetase (FAS-I), a multienzyme complex found in eukaryotes and advanced prokaryotes, and type II fatty acid synthetase (FAS-II), a disaggregated synthetase system found in plants and bacteria. The FAS-I and FAS-II systems of M. tuberculosis were shown to be similar to those of M. smegmatis (G. S. Besra, unpublished results).
The mycobacterial FAS-I system elongates acetyl group by two-carbon units, using acetyl-CoA and malonyl-CoA as substrates to yield butyryl-S-Enz (Fig. (Fig.2).2). Further elongation leads to C16- and C18-S-Enz; these are converted to the CoA derivative and used primarily for the synthesis of membrane phospholipids. Continued elongation of these fatty acids by FAS-I of M. tuberculosis produces specifically the C20- and C26-S-Enz products, and these fatty acids are released as the CoA derivatives (Fig. (Fig.2).2). We propose that the C20 fatty acid is the starting point where the FAS-II systems take over for the synthesis of the very-long-chain mero segment of α-, methoxy-, and keto-mycolic acids (see below). Meroacid is defined in this review as the long α-alkyl chain of mycolic acid that is released as a meroaldehyde by pyrolytic cleavage and is oxidized by a separate reaction to the fatty acid (8). FAS-I also produces hexacosanoyl-CoA (C26), and this fatty acid becomes the short α-alkyl chain and methyl carboxyl segment of all mycolic acids of M. tuberculosis. McCarthy (62) and Weir et al. (114) observed uptake of long-chain fatty acids from the culture medium by mycobacteria and incorporation into triacylglycerol. In M. tuberculosis infection, these internalized C16 and C18 fatty acids could enter the FAS-I pathway and be used for the synthesis of mycolic acids.
The basic FAS-II module shown in Fig. Fig.33 is used for chain elongation in the synthesis of meroacids in M. tuberculosis. We suggest that it is one of three different modules of FAS-II (discussed below). The malonyl-S-acyl carrier protein (malonyl-S-ACP) (a cosubstrate in reaction 1 in Fig. Fig.3)3) is derived from malonyl-S-CoA by the enzyme malonyl-CoA:ACP transacylase (FabD). Kremer et al. studied the properties of this enzyme (51). Through successive reactions by β-ketoacyl-ACP synthase (KasA/KasB) and β-ketoacyl-ACP reductase, R-CO-S-ACP is converted to R-CHOH-CH2-CO-S-ACP (reactions 1 and 2). This product is then converted to trans-R-CH=CH-CO-S-ACP by β-hydroxyacyl-ACP dehydrase (reaction 3). Curiously, the gene encoding this enzyme has yet to be identified in the M. tuberculosis genome (30). The product of reaction 3 is reduced by 2-trans-enoyl-ACP reductase (InhA) to yield R-CH2-CH2-CO-S-ACP, which is two carbons longer than the starting substrate. This product is recycled by a separate FAS-II module to further increase the chain length by two carbons.
As shown in Fig. Fig.4,4, FAS-I provides the eicosanoyl-S-CoA to the FAS-II system for the synthesis of the meroacids. Choi et al. (20) and Scardale et al. (90) have shown that the initial reaction upon entry into the FAS-II system from FAS-I involves a special β-ketoacyl-ACP synthase III (mtFabH) (Fig. (Fig.4,4, reaction 1). How M. tuberculosis places double bonds into the meroacid precursors is an unresolved issue. We propose that β-hydroxyacyl-ACP dehydrase and 2-trans-enoyl-ACP isomerase together are responsible for the production of cis unsaturation. Thus, we propose reactions 2 to 4 as shown in Fig. Fig.4.4. Reaction 2 shows that β-hydroxyacyl-ACP reductase catalyzes the formation of β-hydroxy-C22-S-ACP from β-ketoacyl-C22-S-ACP. Then β-hydroxyacyl-ACP dehydrase and 2-trans-enoyl-ACP isomerase converts this product to cis-Δ3-C22:1-S-ACP (reactions 3 and 4). We call the system that catalyzes reactions 1 to 4 the FAS-IIA module. This is new and a key concept that we are using to describe how a cis unsaturation is introduced in the synthesis of mycolic acids in M. tuberculosis. This dehydrase activity is similar to that of the β-hydroxyacyl-ACP dehydrase (FabZ) found in Escherichia coli and Streptococcus pneumoniae (71, 85, 125). This enzyme is involved in converting β-hydroxyacyl-S-ACP to 2-trans-enoyl-S-ACP. It should be noted that Barry et al. (11) considered the possibility of the existence of dehydrase but dismissed it because of a lack of any enzyme in the genome of M. tuberculosis with significant homology to FabA. We have searched the genome of M. tuberculosis for both dehydrase and isomerase and have found a few candidate enzymes that might be related to S. pneumoniae FabZ and FabM (see below).
The cis-Δ3-C22:1-S-ACP is carried through elongation by the FAS-II modules (Fig. (Fig.33 and and4)4) at one cycle per FAS-II unit (five cycles for α-meroacid and eight cycles for methoxy- and keto-meroacid). For α-meroacid, the product is cis-Δ13-C32:1-S-ACP (Fig. (Fig.4,4, reaction 5, where x = 10). For methoxy- and keto-meroacid, the product is cis-Δ19-C38:1-S-ACP (x = 16). KasA (β-ketoacyl-ACP synthase A) is probably involved in these early elongation steps. This pathway is supported by a study by Takayama et al. (102). They isolated and characterized these anabolic products as well as their expected sequential intermediates of elongation of Δ5-C24:1, Δ7-C26:1, Δ9-C28:1, and Δ11-C30:1 fatty acids in log-phase cells of M. tuberculosis H37Ra. Although it has not been determined, the double bonds are expected to have cis configuration.
For α-meroacid, the cis-Δ13-C32:1-S-ACP is carried through reactions 6 and 7 to yield β-hydroxy cis-Δ15-C34:1-S-ACP (Fig. (Fig.4).4). Then β-hydroxyacyl-ACP dehydrase and 2-trans-enoyl-ACP isomerase convert this product to cis,cis-Δ3,15-C34:2-S-ACP (reactions 8 and 9). We call the system that catalyzes reactions 6 to 9 the FAS-IIB module. Further elongation of this product by the FAS-II modules (Fig. (Fig.3)3) yields a cis,cis-Δ19,31-C50:2-S-ACP (reaction 10). For methoxy-meroacid, the product is cis,cis-Δ19,37-C56:2-S-ACP (where x = 16 and y = 17), and for keto-meroacid, the product is cis,cis-Δ21,39-C58:2-S-ACP (where x = 16 and y = 19). KasB (β-ketoacyl-ACP synthase B) is probably involved in the latter elongation steps. Two distinct cyclopropane synthases, MmaA2 (40) and PcaA (41), then introduce cyclopropane rings into the distal and the proximal unsaturations, respectively, in cis,cis-Δ19,31-C50:2-S-ACP to yield the mature C52-α-meroacid as the ACP derivative (reaction 11, where x = 10 and y = 17). Methyl transfer reactions at this point in the pathway are supported by a study by Qureshi et al. (83). Using a cell extract of M. tuberculosis H37Ra, they showed that the [14C]methyl group of S-adenosyl-l-methionine (SAM) goes directly into the full-size meroacids. Later, Yuan et al. confirmed this by showing that the cyclopropane residues are introduced into the meroacid-S-ACP prior to the final condensation step (121). Watanabe et al. established a clear structural relationship between related isolated type 1 (Fig. (Fig.4,4, product of reaction 11) and type 3 (Fig. (Fig.4,4, product of reaction 10) meroacids in M. tuberculosis (113).
Takayama and associates isolated and characterized saturated C35 to C56 fatty acids (104), mono- and diunsaturated C22 to C47 fatty acids (103), cis,cis-3,4- and 15,16-dimethylenetetratriacontanoic acid (80), and C49 to C58 fatty alcohols (81) from M. tuberculosis H37Ra. These fatty acids were considered to be precursors of mycolic acids. Many of these fatty acids that are smaller than meroacids and contain cyclopropane rings, as well as the fatty alcohols, are now considered to be products of β-oxidation of mycolic acids (from the catabolic pathway). The expected presence of a homologous series of diunsaturated fatty acids >C47 in this proposed pathway (homologous series of products of reaction 10) has not been investigated.
The starting point for the biosynthesis of meroacids by the FAS-II system appears to be the C20-S-CoA that is derived from FAS-I. Studies by Takayama and associates support this conclusion (27, 101). They showed that isoniazid specifically inhibits the synthesis of monounsaturated tetracosanoic acid (C24:1) and higher homologs in M. tuberculosis H37Ra. They also showed that the hexacosanoic acid accumulates in the tubercle bacillus treated with the drug. Jacobs and associates and others showed that the target of inhibition of mycolic acid synthesis by isoniazid is 2-trans-enoyl-ACP reductase (InhA) of the FAS-II elongation pathway (10, 59, 111). As the above-described two pathways show (Fig. (Fig.33 and and4),4), inhibition of InhA would inactivate the entire FAS-II pathways, except for the first four steps at the very beginning of the first cycle, starting with β-ketoacyl-ACP synthetase III, then β-ketoacyl-ACP reductase, β-hydroxy-ACP dehydrase (Fig. (Fig.4,4, reactions 1 to 3), and finally the special 2-trans-enoyl-ACP isomerase activities. The last reaction introduces the cis unsaturation (Fig. (Fig.4,4, reaction 4). Thus, the synthesis of Δ3-C22:1 fatty acid would proceed normally and not be affected by the drug, but it cannot elongate further by the FAS-II module because InhA is inactivated (Fig. (Fig.3,3, reaction 4). This should result in the inhibition of the synthesis of Δ5-C24:1 fatty acid and its higher homologs. This is observed. Isoniazid does not affect the synthesis of hexacosanoic acid, because this fatty acid is synthesized by the drug-insensitive FAS-I system. It accumulates as a result of inhibition of the synthesis of meroacids and mycolic acids.
The precursor for the synthesis of oxygenated meroacids is compound 1 (Fig. (Fig.5),5), where y = 17 for methoxy-meroacid and y = 19 for keto-meroacid. It is also the product of reaction 10 in Fig. Fig.4.4. In the presence of SAM, MmaA4 converts compound 1 to compound 2 (29, 31) and MmaA1 in turn converts compound 2 to compound 3 (122). MmaA4 and MmaA1 are special methyltransferases. MmaA4 introduces the methyl branch and adjacent hydroxyl group at the distal cis unsaturation, and MmaA1 introduces the allylic methyl branch with trans unsaturation at the proximal cis unsaturation. The former becomes the precursor of cis-oxygenated meroacids, and the latter becomes the precursor of trans-oxygenated meroacids. The alternative for the introduction of a methyl branch would be the so-called propionate pathway. Etamadi and Lederer first proposed the mechanism of conversion of cis unsaturation into a trans unsaturation with migration and the introduction of an adjacent methyl branch (C alkylation or C methylation) in α-mycolic acid of M. smegmatis (34). Lederer has reviewed this mechanism (54). The precise mechanism involved in the reaction catalyzed by MmaA4 is not known.
Figure Figure66 shows the final steps in the synthesis of cis-methoxy-, trans-methoxy-, cis-keto-, and trans-keto-meroacids (as the ACP derivatives). For the synthesis of the cis-C59-methoxy- and trans-C60-methoxy-meroacids, MmaA3 (methyltransferase) introduces a methyl residue into the secondary alcohol of compound 2 (y = 17) and compound 3 (y − 1 = 16) (120, 124). However the cyclopropane synthase used for cis cyclopropanation of unsaturation at the proximal position of compound 2 (y = 17) is MmaA2 (40), whereas that used for trans cyclopropanation of unsaturation at the proximal position of compound 3 (y − 1 = 16) is CmaA2 (42). For the synthesis of cis-C60-keto- and trans-C61-keto-meroacids, the secondary alcohol at the distal position of compound 2 (y = 19) and compound 3 (y − 1 = 18) is oxidized to the oxo group by a proposed oxidation-reduction system (see below). MmaA2 catalyzes the cis cyclopropanation of unsaturation at the proximal position of compound 2 (y = 19) (40), whereas CmaA2 catalyzes the trans cyclopropanation of unsaturation at the proximal position of compound 3 (y − 1 = 18) (42). Note the parallelism of the pathways to synthesis of cis- or trans-methoxy meroacids and cis- or trans-keto meroacids. The structures of the four oxygenated meroacid products as the ACP derivatives are given in Fig. Fig.66.
Portevin et al. were the first to identify polyketide synthase 13 (Pks13) as the enzyme complex involved in the final assembly of mycolic acids in M. tuberculosis (76). This pathway involves Claisen-type condensation and is shown in Fig. Fig.7.7. A high degree of control is placed over the entire condensation reactions. A very specific FadD32 (fatty acyl-AMP ligase) converts each meroacyl-S-ACP derived from the FAS-II system (α-meroacyl-S-ACP from Fig. Fig.44 and methoxy- and keto-meroacyl-S-ACPs from Fig. Fig.6)6) to meroacyl-AMP. Trivedi et al. recently discovered this novel reaction and its significance (107). The C26-S-CoA derived from FAS-I (after release from the enzyme as the CoA derivative) is carboxylated by AccD4 and AccD5 (acyl-CoA carboxylases) to yield the 2-carboxyl-C26-S-CoA. The functions of AccD4 and AccD5 are discussed in another section. These two enzymes must be specific for the C24 and C26 fatty acyl groups. The two products are substrates for the Claisen-type condensation by Pks13.
The C52-α-meroacyl-AMP is used as one of the five substrates in this illustration (Fig. (Fig.7).7). This scheme is an adaptation of the activation and transfer of fatty acids on cognate Pks protein described by Trivedi et al. (107). The first step (reaction 1) is the binding of a C52-α-meroacyl group from its AMP derivative and a 2-carboxyl-C26-acyl group from its CoA derivative to the N-terminal phosphopantetheine-binding (PPB) domain and the near-C-terminal PPB domain of Pks13, respectively, both as thioesters. In the former reaction, the acyltransferase (AT) enzyme is presumed to be specific for the AMP derivative. Reaction 2 involves the transfer of an α-meroacyl group from the N-terminal PPB domain to the ketoacyl synthase (KS) domain, presumably catalyzed by the acyltransferase at the AT domain. These two acyltransferase reactions are suggested by Trivedi et al. (107) and are based on a study by Admiraal et al. (2). Reaction 3 is the Claisen-type condensation of the two fatty acyl groups. A nucleophilic attack of the carbonyl group of the α-meroacyl-S-KS by the acidic α-carbon of the 2-carboxyl-C26-S-PPB results in the formation of 3-oxo-C78-α-mycolate bound to the near C-terminal PPB domain and the release of CO2 and KS. Reaction 4 is the reduction step, where an unidentified acyl reductase converts the 3-oxo to the secondary alcohol to yield the mature C78-α-mycolate (reactions 3 and 4 were combined in this illustration). The role of the thioesterase (TE) at the C-terminal TE domain is not to release the mycolic acid from the Pks13 complex, since such a large free fatty acid would become intractable. It is suggested that the TE domain of Pks13 contains a type II thioesterase that “edits” and corrects improper acylation at the near-C-terminal PPB domain (48, 57, 92). The four oxygenated mycolic acids are synthesized by the same mechanism as the α-mycolic acid.
Walker et al. (112) first described the mechanism for the Claisen-type condensation in the synthesis of corynomycolic acid in Corynebacterium diphtheriae, based on a study by Gastambide-Odier and Lederer (37). Later, Takayama and Qureshi (100) used it to describe this reaction in M. tuberculosis. The present scheme is a highly updated version of the old scheme in the latter reference, using the latest data.
Upon examining Fig. Fig.33 to to77 together, one might get the impression that the pathways to synthesis of all mycolic acids in M. tuberculosis are physically interconnected to give one large array network. Although this is possible, it is probably not the case. We suggest that the mycolate synthetase complexes of M. tuberculosis are organized as assembly line arrays of FAS-I and FAS-II modules, a methyltranferase(s), Pks13, and other components to yield five separate and distinct types, as shown in Table Table1.1. In this description we envisage a physical string or line of units bound in a row. This is a new concept, taken from the type II polyketide synthase systems as models (97). We suggest that there are three different FAS-II-type modules available to be arranged in a linear fashion in the assembly line arrays as described in Table Table11 and Fig. Fig.4.4. The FAS-II module is used for elongation of the fatty acids. FAS-IIA and FAS-IIB modules are able to elongate and place cis unsaturation at the distal and proximal positions, respectively, on the meroacid chain. Clearly, the linear array of modules allows the tubercle bacillus to direct the synthesis of mycolic acids with great structural precision. These long assembly line arrays would be labile and might be stabilized by binding to a structural component of the cells, e.g., to the inner surface of the membrane or to its continuum in the cytoplasm called the mesosome (43). Such a linear array of 19 to 26 modular units would be sensitive to physical disruption; thus, it would be very difficult to prepare an active cell-free system from M. tuberculosis for the synthesis of mycolic acids, as numerous investigators have found.
The final stages in the synthesis of the cell wall of M. tuberculosis are transport and attachment of newly synthesized mycolic acids to the peptidoglycan-arabinogalactan complex of the cell wall and the formation of TDM. We describe how this might occur through a six-step process as shown in Fig. Fig.8.8. This part of the pathway for the most part is unknown. Only the function of antigen 85/fibronectin-binding protein as mycolyltransferase (Ag85/Fbp) is well established (reactions 5 and 6).
The final product of synthesis of mycolic acid in the cytoplasm is thought to be mycolyl-S-Pks13 (Fig. (Fig.7).7). The mycolic acid in this product must somehow be transferred to trehalose by an as-yet-unidentified mycolyltransferase(s) to yield trehalose monomycolate (TMM). Ag85 does not appear to have this mycolyltransferase activity. In an unpublished study by L. Y. Armitige, C. A. Rivera-Marrero, and K. Takayama, the individual inactivations of the fbpA and fbpB genes in M. tuberculosis H37Rv by homologous recombination (4) were shown to cause decreases in the synthesis of TDM and clear accumulation of TMM when compared to the wild type. These findings were from short-term pulse-labeling experiments using sodium [1-14C]acetate in submerged cultures. Since ΔfbpA and ΔfbpB mutants were clearly able to synthesize TMM, the Ag85 complex must not be involved in the early mycolyl transfer reactions.
We suggest a completely new pathway to the synthesis of TMM. We first attribute a function for the previously discovered 6-O-mycolyl-β-d-mannopyranosyl-1-phosphoheptaprenol (Myc-PL) (14, 26) as the initial carrier form of the mycolyl group. The mycolyl group is first transferred from mycolyl-S-Pks13 (mycolyl-S-PPB) to d-mannopyranosyl-1-phosphoheptaprenol by a proposed cytoplasmic mycolyltransferase I to yield Myc-PL (Fig. (Fig.8,8, reaction 1). Myc-PL then migrates to the inner surface of the cell membrane and, with its hydrophobic heptaprenol tail, docks next to an ABC transporter. The subsequent three reactions are localized at the transporter. In a tightly coupled system, the mycolyl group is then transferred to trehalose 6-phosphate by the proposed membrane-associated mycolyltransferase II (reaction 2) to form TMM-phosphate, the phosphate group is removed by the membrane-associated trehalose 6-phosphate phosphatase (reaction 3), and the resulting product (TMM) is immediately transported outside the cell by the ABC transporter (reaction 4). There should be virtually no accumulation of TMM in the cytoplasm. This proposed mechanism would allow a rapid and efficient transfer of TMM from the inside to the outside of the cell for the synthesis of cell wall arabinogalactan-mycolate and TDM. This would be a requirement for normal growth of M. tuberculosis. It prevents the conversion of TMM to TDM inside the cell by the ubiquitously present Ag85/Fbp (47, 89). Such a reaction is to be avoided, since it would effectively consume most of the TMM synthesized internally and make it unavailable for transport outside the cell. The advantage of forming a TMM-phosphate intermediate rather than TMM directly from trehalose is not known. However, the role of trehalose 6-phosphate as the acceptor of mycolic acids in the synthesis of TMM is supported by a study by Shimakata and Minatogawa (93).
The three pathways to synthesis of trehalose in Mycobacterium species and Corynebacterium glutamicum are OtsA/OtsB, TreY/TreZ, and TreS (108). Wolf et al. (117) showed that the ΔotsA ΔtreY double deletion in C. glutamicum completely eliminated the synthesis of TMM and TDM. Single deletions of these two genes caused about a 40% reduction (from the wild-type level) in the synthesis of TMM and TDM (our interpretation of their results). The conclusion here is that both the OtsA/OtsB and TreY/TreZ pathways are involved in the synthesis of TMM and TDM. The OtsA/OtsB pathway generates trehalose 6-phosphate, whereas the TreY/TreZ pathway generates free trehalose. Thus, in the latter case, we can assume that the product would have to be phosphorylated before mycolyl transfer can occur. The ΔotsA ΔtreS ΔtreY triple deletion mutant of M. smegmatis was not able to grow unless supplemented with trehalose (119). This shows that mycobacteria have an absolute requirement for trehalose, presumably to process the mycolyl groups.
We suggest that TMM is transported outside the cell (into the outer membrane) by the ABC transporter (reaction 4). This is a new concept and has experimental support. Inhibitors of the ABC transporter (reserpine and verapamil) were shown to inhibit the incorporation of radiolabel from sodium [1-14C]acetate into cell wall arabinogalactan-corynomycolate in Corynebacterium matruchotii (C. Wang and K. Takayama, unpublished results).
By the action of the extracellular mycolyltransferase called Ag85/Fbp/PS1 (PS1 is the mycolyltransferase in C. glutamicum) (13, 45, 77, 78), the final products of cell wall arabinogalactan-mycolate and TDM are formed from TMM (reactions 5 and 6). The former product formation is based on genetic evidence. Since M. tuberculosis is known to excrete copious amounts of Ag85 into the culture medium (1), it is readily available for these two reactions outside the cell.
The genes and enzymes believed to be involved in the biosynthesis and processing of mycolic acids in M. tuberculosis are listed in Table Table2.2. Most of the genes and their functions have been identified. These include the single gene for FAS-I, most of the genes for the FAS-II system, and all of the genes for the methyltransferases and cyclopropane synthases. There are two enzymes in the proposed pathway to synthesis of meroacids by the FAS-II system whose corresponding genes have yet to be identified or studied in detail. These are the β-hydroxyacyl-ACP dehydrase and the 2-trans-enoyl-ACP isomerase (presumed to be present in the FAS-IIA and FAS-IIB modules). Almost all of the enzymes in the mycolic acid-processing pathway need to be identified and studied. Computational genetic analyses were performed to identify candidate genes encoding these unidentified proteins.
A single gene, fab (Rv2524c), encodes the multifunctional FAS-I system of M. tuberculosis. The active gene expresses a protein that forms a homodimer containing all of the necessary functions of a de novo fatty acid synthesis system (96). By comparing the M. tuberculosis FAS with known FASs, the functional catalytic domains have been identified. The domains of M. tuberculosis FAS are organized in the following order: acyltransferase, enoyl reductase, dehydratase, malonyl/palmitoyl transferase, acyl carrier protein, β-ketoacyl reductase, and β-ketoacyl synthase. This system provides the short-chain fatty acyl-CoA substrates for further elongation by FAS-II. It also provides the hexacosanoyl-S-CoA for the final Claisen-type condensation (Fig. (Fig.77).
The substrates for the FAS-II enzymes are ACP derivatives. The exception is the first cycle involving FAS-IIA. The gene acpM encodes the protein ACP in M. tuberculosis, designated AcpM (64). The function of AcpM in the M. tuberculosis FAS-II system is supported by its clustering with malonyl-CoA:ACP transacylase (fabD) and β-ketoacyl synthases (kasA and kasB) (23), the identification of a broad range of long-chain fatty acids bound to AcpM (64), and its function with the type II enzymes in E. coli (91). Overexpression of AcpM in E. coli produced holo-AcpM, apo-AcpM, and palmitoylated-AcpM (51). AcpM shared 93% sequence identity with the Mycobacterium leprae ACP, which is the only ACP-like protein in M. leprae. Analysis of the M. tuberculosis genome revealed the presence of at least two other genes for the putative ACP (Rv0033 and Rv1344) that contain the phosphopantetheine-binding site and share 30 to 33% sequence identity with the AcpM gene (acpM). Their functions are not known.
The solution structure of AcpM consists of an amino-terminal region, folded similarly to the structures of other bacterial ACPs, and a highly flexible and structurally undefined carboxyl terminus (118). Like in other bacterial ACPs, the Asp40Ser41Leu42 motif is conserved in AcpM. The 4′-phosphopantetheine prosthetic group is covalently linked to the serine residue in this motif, and the growing acyl chain is esterified to the terminal sulfhydryl group of 4′-phosphopantetheine. It has been proposed that the unique AcpM carboxyl-terminal domain may interact with the very-long-chain fatty acid intermediates in mycolic acid biosynthesis (118).
FAS-II requires the conversion of malonyl-CoA to malonyl-S-ACP as a substrate for further elongation of fatty acids. This reaction is catalyzed by mtFabD. Analysis of the M. tuberculosis genome revealed that acpM is genetically linked to fabD, kasA, and kasB. The latter two encode the β-ketoacyl synthases (KasA and KasB, respectively). These FAS-II components belong to the same transcription unit. Purified recombinant FabD displayed strong malonyl-CoA:ACP transacylase activity and a preference for holo-AcpM as the substrate in vitro (51). When complemented with the M. tuberculosis fabD expression unit, a temperature-sensitive E. coli mutant deficient in malonyl-CoA:ACP transacylase activity was able to grow at the nonpermissive temperature (51). mtFabD shared 81.2% sequence identity with the M. leprae homolog. The active-site G89X90S91X92G93 motif, characteristic of serine-dependent acylhydrolases (18), is conserved in FabD and its homolog in M. leprae. The active-site serines of these enzymes are involved in the nucleophilic attack and the formation of the acyl-enzyme intermediate. Analysis of the M. tuberculosis genome revealed that FabD shares 28% sequence identity with FabD2. However, the characteristic active-site motif GXSXG is absent in FabD2. It has been proposed that FabD2 may use an alternative nucleophilic group (51). A gene homologous to fabD2 was not found in the M. leprae genome. Since M. leprae is regarded as a minimal genome (22), the lack of fabD2 in M. leprae indicates that FabD is the essential enzyme possessing malonyl-CoA:ACP transacylase activity in M. tuberculosis. Whether the fabD2 gene product catalyzes the malonyl-CoA:ACP transacylase reaction awaits further investigation.
The protein encoded by fabH (FabH) forms a pivotal link between the last cycle of FAS-I and the first cycle of the FAS-II system (involving the FAS-IIA module) in M. tuberculosis. mtFabH initiates chain elongation of C20-S-CoA released from FAS-I as the substrate to generate a β-ketoacyl-S-AcpM product. The function of mtFabH is supported by conservation of active-site residues (Cys112, His244, and Asn274) characteristic of FabH-condensing enzymes and its ability to catalyze the condensation of acyl-CoA with malonyl-ACP and produce β-ketoacyl-ACP (20). Purified recombinant FabH exhibited a preference for acyl-CoA as a substrate. Compared with the general FabB/FabH type of β-ketoacyl synthases, the M. tuberculosis FabH exhibited high substrate specificity. Purified FabH preferentially condenses lauroyl-CoA (C12) and myristoyl-CoA (C14) to generate myristoyl-AcpM and palmitoyl-AcpM (C16), respectively, using malonyl-AcpM as the cosubstrate (90). Chain elongation of C22 to C24 fatty acids by FabH does not occur in these in vitro studies. The crystal structure of M. tuberculosis FabH revealed several unique features, including a CoA/malonyl-ACP binding channel and a long acyl chain-binding channel (90). The M. tuberculosis FabH shares 37% sequence identity to E. coli FabH, and surprisingly, its homolog was not found in the M. leprae genome.
KasA and KasB initiate the subsequent rounds of acyl extension by the FAS-II module (Fig. (Fig.3)3) (64). There are two genes, kasA (Rv2245) and kasB (Rv2246), that encode two separate enzymes, β-ketoacyl-ACP synthase A and β-ketoacyl-ACP synthase B. Both KasA and KasB have been expressed in E. coli, and the purified enzymes have been shown to extend acyl-ACP thioesters by condensation with malonyl-ACP (52, 91). Both enzymes can elongate long-chain fatty acids of >C14 (52, 91). Although there is some redundancy, it has been proposed that KasA catalyzes the initial elongation reactions and KasB extends the elongation to full length. Cell lysates of M. smegmatis overexpressing KasA generated C40 monounsaturated fatty acids, whereas cell lysates of M. smegmatis overexpressing both KasA and KasB generated C54 multiunsaturated fatty acids (94). This is consistent with the pathway described in Fig. Fig.4.4. When KasB was deleted by transposon mutagenesis in Mycobacterium marinum, the mutant produced mycolic acids that were two to four carbons shorter than the wild type, with a significant reduction in keto-mycolate (36). These results indicate that KasA and KasB function independently on separate sets of substrates and that KasB is required for the full extension of the mycolate. One might suggest that KasA is involved in elongation reaction 5 whereas KasB is involved in elongation reaction 10 on the pathway to synthesis of cis,cis-diunsaturated meroacid precursors and α-meroacid (Fig. (Fig.44).
KasA and KasB share 66% sequence identity. KasA and KasB share 92 and 91% sequence identity with their respective homologs in M. leprae. KasA and KasB share 28 and 29% sequence identity with the KS domain of Pks13, respectively (Fig. (Fig.77).
MabA carries out the reduction of β-ketoacyl-ACP to β-hydroxyacyl-ACP in all FAS-II modules. Marrakchi et al. purified recombinant MabA and performed an in vitro assay (61). They found that this enzyme catalyzes NADPH-specific reduction of long-chain β-ketoacyl-ACP (C8 to C20). The crystal structure of tetrameric MabA revealed a large substrate-binding pocket, which could adequately accommodate long acyl chains. MabA shares conserved fold of the short-chain dehydrogenases/reductases (21).
In the M. tuberculosis genome, mabA (Rv1483) is located adjacent to inhA and presumably is transcribed in the same operon. MabA shares 69% sequence identity with its homolog in M. leprae. Numerous homologs of MabA were identified by BLAST searches of M. tuberculosis genome (61). These included FabG2, FabG3, FabG4, FabG5, and a number of hypothetical proteins. Determination of whether these proteins are functional in the FAS-II pathway awaits further investigation.
β-Hydroxyacyl-ACP dehydrase is expected to be present in all FAS-II modules. In E. coli both FabA and FabZ participate in the conversion of β-hydroxyacyl-ACP to trans-2-enoyl-ACP (25, 71). FabA is a bifunctional β-hydroxyacyl-ACP dehydrase/trans-2-enoyl-ACP isomerase, and it diverts the flow of acyl-ACP into the unsaturated fatty acid by converting trans-2-enoyl-ACP to cis-3-enoyl-ACP. FabZ is the primary dehydrase in the elongation of unsaturated acyl-ACP, and it is more active than FabA toward long-chain saturated acyl-ACP (44). As previously reported, a search of M. tuberculosis genome failed to yield FabA-like dehydrase/isomerase (11, 23). In this study, sequence similarity searches were performed by using BLAST with default parameters provided at the National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/BLAST).
We then turned to the S. pneumoniae β-hydroxyacyl-ACP dehydrase FabZ. S. pneumoniae lacks homologs of the fabA gene, and FabZ is the dehydrase that participates in the elongation of saturated and unsaturated fatty acids (125). We compared the sequence of FabZ from S. pneumoniae with sequences of proteins in the genome of M. tuberculosis by using the NCBI BLAST program. This revealed that a 183-amino-acid hypothetical protein encoded by Rv0098 shared 36% identity and 54% similarity with S. pneumoniae FabZ in a 33-amino-acid segment containing active-site residues with an E value of 2.6. Amino acid sequence alignment of Rv0098 with S. pneumoniae FabZ and E. coli FabA and FabZ showed that active-site residues His72 and Glu97 are conserved in Rv0098 (Fig. (Fig.9).9). Computational analysis showed that Rv0098 is among 219 core mycobacterial genes that are essential for mycobacteria and that it shows no similarity with proteins from other microorganisms (58). The Rv0098 gene product is required for survival of M. tuberculosis in mouse lung macrophages (88). Genes homologous to Rv0098 were found in M. leprae, Mycobacterium bovis, and M. smegmatis. The protein encoded by Rv0098 shared 68% identity with a 183-amino-acid conserved hypothetical protein in M. leprae. We thus propose that Rv0098 is a good candidate for the gene that encodes β-hydroxyacyl-ACP dehydrase (FabZ) in M. tuberculosis.
The other possible candidate gene for M. tuberculosis FabZ is Rv0130. When the M. tuberculosis genome was compared with the predicted acyl dehydratase (NCgl0284) of C. glutamicum ATCC 13032, the hypothetical protein Rv0130 was found to share 44% sequence identity with NCgl0284 in a 147-amino-acid segment with an E value of 2E−33. Sequence analysis predicted that Rv0130 is synthesized as an intracellular protein and belongs to the MaoC dehydratase gene family (http://pfam.wustl.edu/; Washington University, St. Louis, Mo.); however, active-site residues of E. coli FabA and FabZ dehydratases were not conserved in Rv0129. Interestingly, Rv0130 is clustered with the mycolyltransferase gene fbpC (Rv0129c).
We expect 2-trans-enoyl-ACP isomerase to be present only in the FAS-IIA and FAS-IIB modules. In S. pneumoniae the Sp0415 gene adjacent to the FAS-II gene cluster was annotated as an enoyl-CoA hydratase/isomerase family protein. Biochemical and genetic analysis of this gene product established that the previously unknown enzyme, designated FabM, functions as an isomerase in the FAS-II system (60). In a reconstituted fatty acid biosynthesis system, FabM was capable of isomerizing 2-trans-enoyl-ACP to 3-cis-enoyl-ACP but could not dehydrate β-hydroxyacyl-ACP (60). Our analysis of the M. tuberculosis genome by using the NCBI BLAST program revealed that a group of 21 enzymes share strong sequence homology with S. pneumoniae FabM (C. Wang and K. Takayama, unpublished results). These enzymes were annotated as the probable enoyl-CoA hydratases (23). Sequence analysis predicted that these enzymes are synthesized as intracellular proteins and belong to the hydratase/isomerase superfamily (Pfam000378; http://pfam.wustl.edu/). Of these enzymes, EchA10 (Rv1142c) and EchA11 (Rv1141c) shared the highest sequence homology with the S. pneumoniae FabM (Fig. (Fig.10).10). EchA10 and EchA11 shared 26 to 28% identity and 42 to 44% similarity with the S. pneumoniae FabM, with E values of 2E−23 and 1E−21, respectively. The proton donor of the enoyl-CoA hydrase/isomerase family was identified as Asp150 in EchA10/EchA11 and as Asp143 in S. pneumoniae FabM (70). Other echA genes shared 23 to 27% identity and 40 to 43% similarity with the S. pneumoniae FabM, with E values ranging from 0.001 to 6E−20. Computational methods have been developed to identify genes that are involved in common biochemical pathways and facilitate the inference of protein function (99). EchA10 and EchA11 are not functionally linked, whereas each gene product is functionally linked with other EchA family members (99). Although EchA10 and EchA11 share the highest sequence homology among EchA family members, these two enzymes may have distinct substrate specificities and are not functionally redundant. Determination of whether the echA10 and echA11 genes encode the 2-trans-enoyl-ACP isomerase for the synthesis of cis-unsaturated meroacid precursors, as proposed in this review, awaits further study.
An alternative method of introducing a double bond(s) to form Δ5-C24:1 and Δ19,31-C50:2 fatty acid precursors of α-meroacid (Fig. (Fig.4,4, reactions 3 and 4 and reactions 8 and 9) and Δ19,37-C56:2 and Δ21,39-C58:2 fatty acid precursors of methoxy- and keto-meroacids (Fig. (Fig.4,4, reaction 10) is by the action of desaturases (oxidative desaturation). Analysis of the M. tuberculosis genome reveals three potential aerobic desaturases encoded by desA1, desA2, and desA3 (23). Overexpression of desA3 in M. bovis BCG and biochemical analysis of the purified enzyme showed that DesA3 desaturase is responsible for the synthesis of oleic acid (cis-Δ9-C18:1) (75). From these results, it appears unlikely that DesA3 is involved in the synthesis of mycolic acid. The possible roles of DesA1 and DesA2 in the synthesis of mycolic acid have not been evaluated.
In the FAS-II elongation module, an NADH-dependent 2-trans-enoyl-ACP reductase encoded by inhA (Rv1484) catalyzes the final step of a cycle of elongation, the reduction of 2-trans-enoyl chains (>C12) to yield the saturated chains (28, 79). InhA from M. smegmatis displays long-chain fatty acid elongation activity, which promotes the elongation of C16-ACP to form C18- to C30-ACP products (59). The preference of InhA for long-chain substrates is consistent with its involvement in the elongation of meroacid precursors in M. tuberculosis.
The M. tuberculosis InhA protein shares 90% sequence identity with its homolog in M. leprae. Sequence analysis revealed that four conserved hypothetical proteins (encoded by Rv3485c, Rv0927c, Rv3530c, and Rv3559c) share 24 to 26% sequence identity with InhA. Further studies are needed to determine their roles in the synthesis of mycolic acid.
The first gene to be identified to modify mycolic acid in M. tuberculosis was cmaA1, which encodes a cyclopropane mycolic acid synthase. It was cloned based on its homology to the E. coli cyclopropane fatty acid synthase gene (123). The same group (Barry and associates) cloned another gene, cmaA2, and it shares 52% identity with cmaA1 (39). By using the cmaA1 gene as a probe, a cluster of genes encoding four highly homologous methyltransferases (>52% identity) was cloned in M. tuberculosis (120). These genes were designated mmaA1 to mmaA4 to indicate their involvement in methoxy-mycolic acid biosynthesis when introduced into M. smegmatis. Dubnau et al. cloned a homologous gene cluster of methyltransferases (cmaA to cmaD) from M. bovis BCG (32). Two additional potential cyclopropane synthases were identified in the M. tuberculosis genome and annotated as umaA1 and umaA2 (23). The function of umaA1 is not known. The gene umaA2 was identified to encode the proximal cyclopropane synthase of α-mycolic acid and was renamed pcaA (41). These eight genes share between 50 and 75% identity, and the enzymes are all thought to utilize SAM in methyl transfer reactions (121). In vitro assays of recombinant methyl transferases indicate that the immediate substrates of these enzymes are AcpM-bound meroacid precursors (Fig. (Fig.44 and and6)6) (121). Biochemical functions of this gene family were defined by genetic manipulation of each synthase in M. tuberculosis and M. smegmatis as well as by systematic analysis of the meroacids.
Analysis of an mmaA2 deletion mutant of M. tuberculosis revealed that α-mycolic acid lacks a distal cyclopropane group and instead contains a cis unsaturation (40). Thus, mmaA2 is required for the distal cyclopropane modification of α-mycolic acid. An early study suggested that the cmaA1 product might also function as a distal cyclopropane synthase for α-mycolic acid. When cmaA1 was overexpressed in M. smegmatis, it was shown that the distal cis unsaturation was transformed into a cyclopropane ring (123). However, a recent study demonstrated that cmaA1 is not required for α-mycolic acid modification. The α-mycolic acid of an M. tuberculosis mutant with a cmaA1 deletion was unaffected, and cmaA1 had no discernible role in mycolic acid modification (40). It is likely that the product of cmaA1 nonspecifically catalyzed distal cyclopropane ring synthesis when highly overexpressed in M. smegmatis.
A pcaA deletion mutant of M. tuberculosis produced α-mycolic acid at a reduced level compared to the wild-type control (41). In addition, a hybrid mycolic acid accumulated with a cis unsaturation at the proximal position in place of the cis-cyclopropane ring (41). These results suggest that pcaA encodes a cyclopropane synthase that is responsible for the transformation of cis unsaturation to a cis-cyclopropane ring at the proximal position of α-mycolic acid. The accumulation of the hybrid mycolic acid in the pcaA deletion mutant of M. tuberculosis supports the view that meroacid precursors with cis unsaturation are first produced and subsequently modified as shown in Fig. Fig.44 (113).
The fact that deletion of pcaA did not completely eliminate the synthesis of α-mycolic acid indicates that another enzyme may share the specificity of PcaA in terms of modifying α-mycolic acid at the proximal position. A previous study suggested that cmaA2 might function as a proximal cyclopropane synthetase for α-mycolic acid. When overexpressed in M. smegmatis, cmaA2 introduced cis-cyclopropane rings at the proximal positions of α-mycolic acid and epoxy-mycolic acid (39). However, deletion of cmaA2 in M. tuberculosis did not alter the α-mycolic acid content (42).
When the mmaA4 gene of M. tuberculosis was inactivated, the mutant synthesized only α-mycolic acid and not oxygenated mycolate (29, 31). Genetic complementation restored the production of oxygenated mycolate (29). In addition, monoethylenic (at the distal position) monocyclopropanated (α2) mycolate accumulated in this mutant and presumably is the substrate of MmaA4 (29). M. smegmatis does not contain methoxy- and keto-mycolates. Overexpression of mmaA4 in M. smegmatis led to the synthesis of keto- and hydroxy-mycolates while synthesis of diunsaturated α-mycolates was significantly reduced (29). Overexpression of mmaA4 or cmaA (the homolog of mmaA4 in M. bovis BCG Pasteur) in M. smegmatis induced the production of methyl-branch hydroxy-mycolic acids (32, 120). These results indicate that MmaA4 is responsible for the production of methyl-branch hydroxy-mycolic acids, which are the common precursors for oxygenated mycolates in M. tuberculosis as shown in Fig. Fig.55 and and66.
The methyl-branch hydroxy-meroacid precursor produced by MmaA4 is converted to the methyl-branch keto-meroacid precursor through an oxidation-reduction process (Fig. (Fig.6).6). The enzyme responsible for this transformation has not been identified. Since the proposed enzyme, provisionally called NAD(P)H-dependent hydroxyacyl-ACP dehydrogenase, participates in the same biochemical pathway as methyltransferases, the M. tuberculosis genome was analyzed to identify genes that are functionally linked with methyltransferases (99). Computational analysis showed that an oxidoreductase/oxygenase encoded by Rv0161 is functionally linked with all methytransferases, indicating that Rv0161 is a possible candidate enzyme responsible for the synthesis of methyl-branch keto-meroacid precursors. The proposed NAD(P)H-dependent hydroxyacyl-ACP dehydrogenase may belong to the alcohol dehydrogenase family. We have searched the genome of M. tuberculosis for alcohol dehydrogenases and have found a few candidate enzymes encoded by Rv0162c and Rv3057c (C. Wang and K. Takayama, unpublished results). Rv0162c encodes a probable zinc-type long-chain alcohol dehydrogenase. Rv3057c encodes a probable alcohol dehydrogenase/reductase, and it shares similarity with various oxidoreductases involved in polyketide synthesis.
When mmaA3 was transferred into M. smegmatis, it appeared to have no independent function. However, when coexpressed with mmaA4, mmaA3 appears to O-methylate the hydroxyl group newly formed by MmaA4 by using a SAM-derived methyl group (120). Amino acid alterations of CmaB (the homolog of MmaA3 in M. bovis BCG Pasteur) are likely responsible for the failure of M. bovis BCG Pasteur methyltransferase to produce methoxy-mycolic acid in M. smegmatis (32). Overexpression of mmaA3 in M. tuberculosis resulted in the loss of keto-mycolates and a significant increase in methoxy-mycolate production (124), strongly supporting the idea that hydroxy-mycolate is the common intermediate for both keto- and methoxy-mycolate (as shown in Fig. Fig.55 and and6).6). Thus, mmaA3 encodes the O-methytransferase responsible for the transfer of the methyl group from SAM to the 37-hydroxyl groups to form the methoxy group (Fig. (Fig.66).
Genetic transfer of mmaA2 into M. smegmatis led to the conversion of the proximal cis unsaturation in mycolic acids to a cis-cyclopropane ring (120). An MmaA2-deficient mutant of M. tuberculosis produced both complete and incomplete (cis unsaturation-containing) methoxy-mycolic acids. The amount of complete methoxy-mycolic acid was reduced to one-half of that in the wild type (40). The ratio of cis- to trans-keto-mycolic acid was not altered in this mutant. These results indicate that another enzyme may share the specificity of MmaA2 to transform cis unsaturation to a cis-cyclopropane ring in oxygenated meroacids at the proximal position. When the cmaA2 gene is transferred into M. smegmatis, the organism produces mycolic acids containing a cis-cyclopropane ring but not a trans-cyclopropane ring (39). It is possible that CmaA2 has a redundant role with MmaA2 in catalyzing cis-cyclopropane synthesis of methoxy-mycolates. However, MmaA2 may be the preferred cyclopropane synthetase for methoxy- and keto-mycolate synthesis (Fig. (Fig.6)6) (40).
When mmaA1 was transferred into M. smegmatis, it appeared to have no independent function (120). When mmaA1 was overexpressed in M. tuberculosis, it resulted in an increase in the amount of the oxygenated mycolate-containing trans-cyclopropane ring and the appearance of oxygenated mycolates containing trans unsaturation with an allylic methyl branch (122), which suggests that MmaA1 converts cis unsaturation into trans unsaturation with the simultaneous introduction of an allylic methyl branch to oxygenated meroacid precursors. Because overexpression of mmaA1 in M. tuberculosis produced an excess of both trans-unsaturated and trans-cyclopropanated mycolic acids, MmaA1 action is presumably an early step in trans-cyclopropane synthesis (Fig. (Fig.55).
Inactivation of cmaA2 in M. tuberculosis abolished trans-cyclopropanated mycolates and caused the accumulation of unsaturated oxygenated mycolates (42). These results indicate that cmaA2 encodes the trans-cyclopropane synthase of oxygenated meroacids (Fig. (Fig.66).
Claisen-type condensation was proposed by Walker et al. (112) as the mechanism for the condensation of two molecules of palmitic acids to form the C32-corynomycolic acid in a cell-free system of C. diphtheriae. Takayama and Qureshi (100) used this mechanism to describe the final step in the synthesis of mycolic acid in M. tuberculosis. With the discovery by Portevin et al. (76) that Pks13 is involved in the Claisen-type condensation as the final step in the synthesis of mycolic acid in M. tuberculosis, details of this condensation reaction can now be described as shown in Fig. Fig.7.7. This condensing enzyme is encoded by pks13 (Rv3800c). In M. tuberculosis, the enzyme Pks13 is a member of the type I polyketide synthase (Pks) gene family. Sequence analysis revealed that Pks13 contains two nonequivalent phosphopantetheine-binding, ketoacyl synthase, acyl transferase, and thioesterase domains (Fig. (Fig.7).7). Pks13 shares 83% sequence identity with its homologous gene in M. leprae (ML0101). A C. glutamicum mutant with a deletion in the pks13 gene was unable to synthesize corynomycolic acid and produced large amount of precursors, C16 to C18 fatty acids (35a). Since Pks13 is essential for the viability of Mycobacterium, mycolic acid synthesis was analyzed in a conditional mutant of M. smegmatis containing a deletion in pks13 on the chromosome and a functional pks13 gene on the thermosensitive plasmid. A marked reduction of mycolate production (60%) was observed when the mutant was grown at a nonpermissive temperature compared to growth at the normal temperature. The identity of the reductase, which converts the 3-oxo-mycolic acid to mycolic acid, remains unclear. A number of genes in the M. tuberculosis genome share sequence homology with the gene for the known acyl reductase in the FAS-II system, β-ketoacyl-ACP reductase (MabA). Determination of whether any of these enzymes can function as the 3-oxo-mycolic acid reductase awaits further investigation.
Gande et al. performed gene deletion experiments to determine which enzymes are required for corynomycolic acid synthesis in C. glutamicum (35a). They found that Fas-IA is required for the generation of C16 fatty acid, FadD is required for the formation of C16-S-CoA (or possibly C16-AMP), AccD2 and AccD3 are required for the formation of the carboxylated C16-S-CoA, and Pks is required for the Claisen-type condensation. Phylogenomic analysis showed that two clusters of carboxyltransferases (AccD2 and AccD3 in C. glutamicum) are well conserved in Corynebacteriaceae. The equivalent systems and orthologs in M. tuberculosis would be FAS-I/FAS-II (to generate the meroacids), FadD32, AccD4, AccD5, and Pks-13, respectively (Fig. (Fig.7).7). A more detailed comparison of the two systems (Claisen-type condensation in corynebacteria and mycobacteria) would be of great interest, since they should be similar.
Mycolyltransferase I and mycolyltransferase II are the proposed enzymes that would transfer newly synthesized mycolic acid in thioester linkage to Pks13 (Fig. (Fig.7),7), first to mannopyranosyl-1-phosphoheptaprenol to yield Myc-PL and then to trehalose 6-phosphate to yield the TMM-phosphate (6-O-mycolyl-treh-6′-P) (Fig. (Fig.8,8, reactions 1 and 2). We are addressing an outstanding question of how the key carrier of the mycolyl group, TMM, is synthesized. These reactions are hypothetical and have not been studied.
A recent study demonstrated that an acyltransferase (PapA5), which catalyzes diesterification of phthiocerol and phthiodiolone with mycocerosic acid, is located in the phthiocerol and phthiodiolone dimycocerosate ester synthesis gene cluster (74). The genes papA1, papA2, and papA3 from M. tuberculosis, encoding proteins presumably catalyzing O esterification of trehalose with the methyl-branched polyketides, are associated with loci encoding Mas-like polyketide synthases. We examined the pks13 locus and found that Rv3802c may function as the proposed mycolyltransferase I by catalyzing the transfer of newly synthesized mycolic acid from Pks13 to mannopyranosyl-1-phosphoheptaprenol (Fig. (Fig.8,8, reaction 1). Rv3802c encodes an essential protein for growth (87), and it is located immediately upstream of fadD32 and pks13 and downstream of the mycolyltransferase gene fbpA (Fig. (Fig.11).11). Sequence analysis predicted that Rv3802c is synthesized as a membrane protein and belongs to the serine esterase cutinase gene family (Pfam01083; http://pfam.wustl.edu).
We performed an NCBI BLAST search of the M. tuberculosis genome and identified two homologous sequences of the Ag85 complex encoded by Rv0947c and Rv1288. Rv0947c encodes a 76-amino-acid protein (corresponding to the N-terminal sequence of the Ag85 complex) and shares 56% identity and 70% similarity with FbpA in the 48-amino-acid overlap, with an E value of 1E−07 (C. Wang and K. Takayama, unpublished results). An NCBI BLAST search of mycobacterial genomes revealed that the gene in M. bovis (Mb0972c) is identical to Rv0947c.
Rv1288 shares 26 to 31% sequence identity and 41 to 46% similarity with the Ag85 complex (FbpA, FbpB, and FbpC) at the amino acid level (Fig. (Fig.12).12). Rv1288 encodes a 456-amino-acid protein and shares 26% identity and 41% similarity with FbpA in a 128-amino-acid overlap, with an E value of 8E−05. An NCBI BLAST analysis of mycobacterial genomes revealed that Rv1288 has homologous genes in M. bovis, M. smegmatis, M. marinum, M. leprae, and Mycobacterium avium. Further bioinformatics analysis revealed that the N-terminal amino acid sequence of the protein encoded by Rv1288 (amino acids 5 to 153) shows three LysM (lysine motif) domains. LysM was originally identified in enzymes that degrade bacterial cell walls and is also present in proteins known to be associated with bacterial cell walls (15). A recent study showed that the LysM domain mediates direct binding to peptidoglycan on the cell wall (98). The C-terminal amino acid sequence of the protein encoded by Rv1288 (amino acids 73 to 443) indicates an esterase D family member (12). This family of esterases includes Ag85 and C. glutamicum mycolyltransferase PS1 (78). The active-site residues Ser124, His260, and Glu228, forming the catalytic triad, are conserved in Rv1288. We propose that the Rv1228 gene product is likely a candidate protein for mycolyltransferase II (Fig. (Fig.8,8, reaction 2). It is expected that the substrate-binding sites would be different from those of Ag85. The potential cell wall localization of the protein encoded by Rv1288 suggests that this enzyme may play a role in cell wall mycolate-peptidoglycan formation. Studies of the effect of genetic deletion of Rv1288 and inhibition of its gene product may shed light on its role in the processing of mycolic acids.
Further analysis of the M. tuberculosis genomic sequence revealed two conserved proteins encoded by Rv0519c and Rv0774c that shared low sequence homology with the protein encoded by Rv1288 (11.0% identity and 30.4% similarity) (C. Wang and K. Takayama, unpublished results). The protein encoded by Rv0774c shares 27% sequence identity and 43% similarity with Rv1288 in a 66-amino-acid overlap, with an E value of 1.3. The protein encoded by Rv0519c shares 29% sequence identity and 48% similarity with Rv1288 in a 114-amino-acid overlap, with an E value of 2.1. The highly homologous proteins encoded by Rv0519c and Rv0774c share 58% sequence identity in a 299-amino-acid overlap, which indicates that these two proteins may have similar cellular functions. Like in Rv1288, the proteins encoded by Rv0519c and Rv0774c belong to the esterase D family (12). An NCBI BLAST search of mycobacterial genomes revealed that Rv0519c and Rv0774c have homologous genes in M. bovis, M. smegmatis, M. marinum, and M. avium. A catalytic triad (Ser, His, and Arg) was identified by sequence alignment of proteins encoded by Rv0519c and Rv0774c with mycolyltransferases (FbpA, FbpB, and FbpC) (Fig. (Fig.13).13). Rv0519c is predicted to encode a 300-amino-acid basic membrane protein, whereas Rv0774c encodes a 312-amino-acid exported protein. The potential cellular localization of these enzymes is consistent with the proposed functions of transferring mycolic acid from Myc-PL to trehalose 6-phosphate on or close to the inner membrane surface. These two proteins are additional candidates for mycolyltransferase II (Fig. (Fig.8,8, reaction 2).
We wish to identify the gene that encodes the TMM-6-phosphate phosphatase. Two possible trehalose 6-phosphate phosphatase genes (otsB1 and otsB2) were identified in M. tuberculosis (23). A recent study demonstrated that recombinant OtsB2 exhibited highly specific trehalose 6-P phosphatase activity, while OtsB1 had no detectable phosphatase activity (33). Our analysis of the M. tuberculosis genomic sequence revealed that a probable hydrolase gene (Rv3400) and the N-terminal hydrolase domain of otsB1 (Rv2006) share significant sequence homology with otsB1 and otsB2 (46 to 48% identity and 58 to 59% similarity) (C. Wang and K. Takayama, unpublished results). This is shown in Fig. Fig.14.14. An NCBI BLAST search of mycobacterial genomes revealed that otsB1 is unique to the M. tuberculosis and M. bovis genomes, while homologous genes of Rv3400 were found in M. bovis, M. smegmatis, M. marinum, M. leprae, and M. avium. Rv3400 encodes a 262-amino-acid acidic protein that is predicted to be an intracellular protein. This enzyme belongs to a subgroup of phosphatases that use aspartate residues as the active-site nucleophile. A characteristic N-terminal Asp-X1-Asp-X2-Thr/Val motif is conserved in Rv3400 (24, 105). Variations in the C-terminal motif have been reported for this class of phosphatases (95). Computational analysis revealed that Rv3400 is functionally linked with otsB1 (99). The gene products of otsB1 and Rv3400 are potential candidates for TMM-P phosphatase (Fig. (Fig.8,8, reaction 3).
We hypothesize that the proposed TMM transporter in M. tuberculosis is actually an ABC transporter (Fig. (Fig.8,8, reaction 4). Analysis of the M. tuberculosis genome by Braibant et al. revealed the presence of at least 26 complete (including at least one nucleotide-binding domain and one membrane-spanning domain) and 11 incomplete ABC transporters (17). Of the 37 potential ABC transporters, two are identified with no known functions and another four are similar to multidrug resistance proteins of eukaryotes and prokaryotes. The physiological role of multidrug transporter has been controversial. It has been proposed that they may be involved in the transport of substrates that share structural properties with drugs. Several studies have shown that multidrug transporters are able to transport a variety of lipids. The human multidrug resistance P-glycoprotein MDR1 mediates transport of short-chain lipid analogs (109) as well as the translocation of native glycosphingolipid precursor glucosylceramide (110). The human multidrug resistance P-glycoprotein MDR3 is able to transport phosphatidyl choline across the canalicular membrane of the hepatocyte (109). A recent study showed that MsbA (lipid A transporter in E. coli) and LmrA (multidrug transporter in Latobacillus lactis) have overlapping specificities for lipid A and chemotherapeutic and cytotoxic drugs (84). These results suggest a possible role of multidrug transporters in transporting lipids (like TMM) to the outer membrane. Further biochemical and genetic studies are needed to determine whether any one of the four MDR transporters in M. tuberculosis (encoded by Rv1273c/Rv1272c, Rv1348/Rv1349, Rv0194, and Rv1819c) or the two transporters with no known functions (encoded by Rv1747 and Rv1687c/Rv1686c) is able to transfer TMM from the inside to the outside of the cell membrane.
The antigen 85 complex is composed of Ag85A (FbpA), Ag85B (FbpB), and Ag85C (FbpC) as the predominant secreted protein in M. tuberculosis. The physiological function of Ag85 is to transfer mycolic acid from a TMM to another TMM and cell wall arabinogalactan to form TDM and cell wall arabinogalactan-mycolate, respectively, outside the cells. The corresponding genes encoding these proteins are fbpA (Rv3804c), fbpB (Rv1886c), and fbpC (Rv0129c).
The three Ag85 complex proteins share 65 to 80% identity. The crystal structure of the secreted forms of Ag85C (FbpC) and Ag85B (FbpB) revealed a canonical α/β hydrolase fold; a catalytic triad of Ser, Glu, and His (conserved in FbpA, FbpB, and FbpC); a probable TMM-binding site (a hydrophobic pocket and tunnel); and a probable site for the interaction with human fibronectin (3, 86).
A related gene, fbpD (also called fbpC1), is located adjacent to fbpA but does not belong to the same operon. FbpD (MPT51 protein) is highly related to the Ag85 complex and displays 35 to 39% identity. However, none of the residues of the catalytic triad are conserved within FbpD. The highly conserved carboxyesterase consensus sequence GXSXXG is absent in the FbpD sequence. Consistent with sequence analysis, recombinant FbpD does not exhibit in vitro mycolyltransferase activity (53). The crystal structure of MPT51 revealed that it folds like Ag85 complex members as an α/β hydrolase but does not contain a TMM-binding site (115). These findings excluded FbpD as a potential candidate for the proposed mycolyltransferase I or mycolyltransferase II (Fig. (Fig.88).
The significance of this review can be summarized as follows.
(i) The cyclopropane rings of mycolic acids in M. tuberculosis contribute to the structural integrity of the cell wall complex. Moreover, the cyclopropane rings protect the bacillus from oxidative stress. Thus, the structure of mycolic acids determines the degree of protection that M. tuberculosis receives against the hostile environment of the host.
(ii) Deletion of the proximal cyclopropane ring of α-mycolate or deletion of the methoxy- and keto-mycolates leads to a significant attenuation of growth of the two mutants in the mouse model of infection. Deletion of keto-mycolates in M. tuberculosis leads to restricted growth of this mutant in macrophages. Thus, the distribution and fine structure of mycolic acids determine the virulence of M. tuberculosis. It can be concluded that mycolic acids are very important components of the pathogenic M. tuberculosis.
(iii) Mycolic acid synthesis is the target of well-known antituberculosis drugs, i.e., isoniazid, ethionamide, and thiocarlide (116). This suggests that all reactions on the pathway to synthesis and processing of mycolic acids are viable targets for new drug discovery (target validation).
(iv) Much progress has been made in the past 20 years or so to elucidate how mycolic acids are synthesized. However, how newly synthesized mycolic acids are processed and deposited into the cell wall needs further study. As this review shows, it is now possible to present a much more detailed pathway to the synthesis and processing of mycolic acids in M. tuberculosis than ever before. This review provides a clear direction for further studies. Complete elucidation of the biochemistry and genetics of this pathway would help in the program to find new and more effective drugs against multiple-drug-resistant strains of M. tuberculosis.
This work was supported by a Merit Review grant from the Research Service of the Department of Veterans Affairs to K.T. G.S.B. is a Lister Institute-Jenner Research Fellow and is supported by the Medical Research Council (United Kingdom) and The Wellcome Trust.
Laurens Anderson is acknowledged for his critical review of the manuscript. We thank Kathryn Kleckner for preparing the figures.