The recent determination of the complete genome sequence of Corynebacterium diphtheriae, the aetiological agent of diphtheria, has allowed a detailed comparison of its physiology with that of its closest sequenced pathogenic relative Mycobacterium tuberculosis. Of major importance to the pathogenicity and resilience of the latter is its particularly complex cell envelope. The corynebacteria share many of the features of this extraordinary structure although to a lesser level of complexity. The cell envelope of M. tuberculosis has provided the molecular targets for several of the major anti-tubercular drugs. Given a backdrop of emerging multi-drug resistant strains of the organism (MDR-TB) and its continuing global threat to human health, the search for novel anti-tubercular agents is of paramount importance. The unique structure of this cell wall and the importance of its integrity to the viability of the organism suggest that the search for novel drug targets within the array of enzymes responsible for its construction may prove fruitful. Although the application of modern bioinformatics techniques to the ‘mining’ of the M. tuberculosis genome has already increased our knowledge of the biosynthesis and assembly of the mycobacterial cell wall, several issues remain uncertain. Further analysis by comparison with its relatives may bring clarity and aid the early identification of novel cellular targets for new anti-tuberculosis drugs. In order to facilitate this aim, this review intends to illustrate the broad similarities and highlight the structural differences between the two bacterial envelopes and discuss the genetics of their biosynthesis.
Corynebacterium diphtheriae is the aetiological agent of diphtheria. The organism produces a potent bacteriophage-encoded protein exotoxin, diphtheria toxin (DT), that inhibits eukaryotic protein synthesis via ADP-ribosylation of elongation factor 2. Parenteral administration of DT to susceptible animals results in the typical lesions associated with diphtheria, including myocarditis, polyneuritis and focal necrosis in organs such as the liver, kidneys and adrenal glands . This potentially fatal infectious disease is controlled in many developed countries by an effective immunisation programme. However, the disease has made a dramatic return in recent years, in particular, within the Eastern European Region; the largest, and still on-going, outbreak since the advent of mass immunisation occurring within Russia and the Newly Independent States of the former Soviet Union (NIS) in the 1990s . The complete genome sequence of a UK clinical isolate (biotype gravis strain NCTC 13129), representative of the clone responsible for this outbreak, has now been determined.1
Despite the advent of chemotherapy and its initial successes in the developed world, Mycobacterium tuberculosis remains an extremely important pathogen, continuing to thrive in developing countries and re-emerging in the industrialised nations. The global burden of tuberculosis remains enormous, mainly because of poor control in Southeast Asia, sub-Saharan Africa, and Eastern Europe, and because of high rates of M. tuberculosis and HIV co-infection in some African countries. On the basis of tuberculin skin test reactivity, it is estimated that around one in three humans are infected with M. tuberculosis. Recent estimates provided by the WHO suggest that around 8 million people contract TB annually, of whom 99% live in developing countries, with almost 2 million deaths [3, 4]. Thus M. tuberculosis is responsible for more human deaths than any other single infectious agent  accounting for 7% of all deaths and 26% of all preventable deaths .
Along with the recent increases in TB, outbreaks of multidrug resistant (MDR) strains of M. tuberculosis have also been observed [5–7] and these threaten to undermine effective control of TB. In Estonia for instance, up to 14% of all TB cases are now resistant to more than one anti-tuberculosis drug and the high rates of MDR-TB incidence in the world's two most populous nations, India and China, is concerning as these contribute 40% of all TB cases . The resurgence of TB, and in particular MDR-TB, now represents a major financial burden; for example the epidemic since the early 1990s has cost New York City hundreds of millions of dollars to manage and the cost of treating MDR-TB over susceptible strains rises 20-fold . Given this setting, the need for novel treatments and cellular targets is evident.
In the past decade the complete nucleotide sequences of around 90 bacteria have been determined and analysed. The availability of the complete genome sequences for the M. tuberculosis strains H37Rv  and now CDC1551  has proven a valuable asset, aiding the identification of several genes whose products perform essential metabolic functions by comparison of their sequences with those of better characterised organisms. This information combined with the rapid development of novel molecular techniques for mycobacteria has allowed the characterisation of virulence factors and the dissection of biochemical pathways, rapidly increasing our understanding of the M. tuberculosis metabolome.
One of the major factors contributing to the inherent resistance of M. tuberculosis to many antibacterials is the architecture of its unusual cell wall. The biosynthesis/assembly of this structure has also provided the cellular targets for several front-line antitubercular drugs. The mycobacteria belong to a suprageneric taxon, the mycolata, which also includes corynebacteria, nocardiae and rhodococci. These all share a similar cell wall construction with varying degrees of elaboration (Fig. 1). Despite the very different pathology involved in the progression of diphtheria and tuberculosis, an increased understanding of the biosynthesis and assembly of the cell envelope of M. tuberculosis by comparison with those of its relatives may provide vital clues towards the development of a new generation of anti-mycobacterial agents. Furthermore, the enzymological definition of the biosynthesis of conserved mycolata-specific cell wall motifs may eventually provide drugs to combat all mycolated pathogens.
General features and organisation of the cell wall skeleton
Although belonging to the Gram-positive branch of eubacterial classification, these bacteria possess an unusual lipid domain that modifies the outer regions of their cell wall . The visualisation of a second fracture plane in freeze-etch electron microscopy studies of some mycobacteria and corynebacteria reveal that these lipid-rich domains adopt organised structures suggesting the presence of a lipid bilayer membrane in the outer wall [11–15].
The lipopolysaccharide-rich outer leaflet dominates the properties of the outer membranes of the Gram-negative bacteria. The covalent linkage of four to eight fatty acyl chains to the di-glucosaminyl head group  and the intercalation of these LPS molecules with divalent metal ions combine to stabilise the hydrophobic packing of the membrane's outer leaflet, thereby producing a more coherent and effective permeability barrier . The mycolata have exploited similar principles in the evolution of their permeability barrier. A family of branched hydroxy fatty acids, the mycolates, are esterified to an underlying arabinogalactan polysaccharide layer that is, in turn, covalently bound to the peptidoglycan core of the cell wall (Fig. 1). Together these covalently bound components have been dubbed the mycolyl-arabinogalactan–peptidoglycan (mAGP) complex. In mycobacteria, enough non-extractable mycolate (around 10% of the cellular mass) is present to constitute a complete monolayer around the cell, forming the inner leaflet of an outer membrane structure, an idea first presented by Minnikin . The hydrocarbon chains of the mycolates interact in a tightly packed, parallel arrangement perpendicular to the plane of the cytoplasmic membrane. X-ray diffraction  and differential scanning calorimetry studies  on a highly purified cell wall preparation from M. chelonae provided supporting evidence that they formed the inner leaflet of an asymmetric lipid bilayer. The outer leaflet of the bilayer and any gaps in the inner leaflet are most likely occupied by the various free glycolipids and phospholipids that can be extracted from the walls of the mycolata [15, 19–21]. This arrangement of lipids provides a formidable permeability barrier around the bacterial cell, effectively protecting it against detergents, hydrophobic drugs and other noxious substances. Indeed, one estimate of cell wall permeability in M. chelonae suggested it to be around 100-fold less permeable to penicillins than that of Escherichia coli. The first visual evidence for a coherent covalently attached lipid domain surrounding the mycobacterial cell has only recently been achieved by the staining of the envelopes of solvent-extracted cells with lipophilic fluorescent molecular probes . Although they are structurally diverse, the discovery of a range of porins/pore-forming proteins within the cell walls of these Gram-positive bacteria [24–37] is entirely consistent with the presence of a coherent membranous structure in the cell envelope; these relatively non-specific protein channels allowing the exchange of small hydrophilic solutes across the membrane down concentration gradients. External to the outer permeability barrier of the mycolata occurs a complex, polysaccharide-rich capsule-like layer . In several corynebacteria regular, protein-rich S-layers have been reported [14, 39, 40] but only a single instance of a crystalline outer protein layer has been reported in mycobacteria . Due to the relative paucity of biochemical data and the apparent differences between the mycobacteria and corynebacteria in the outer regions of their cell walls we shall concentrate solely on the structure, biosynthesis and genetics of the conserved cell wall skeleton. Another significant class of wall-associated molecules of great importance to the pathogenicity of the mycobacteria are the structurally and biosynthetically-related lipoglycans, the lipoarabinomannans, lipomannans and their glycolipid precursors, the phosphatidylinositolmannosides. Again, as significant differences in the complement of these lipoglycans are apparent between the corynebacteria and the mycobacteria and as they are the subject of a comprehensive forthcoming review by Besra and Takayama, they will not be discussed further.
Fine structure of the M. tuberculosis and C. diphtheriae cell wall
In this and subsequent sections we aim to highlight the gross similarities and detail the structural differences between the cell wall skeletons of these two actinomycete pathogens. A vast body of work has been carried out towards the structural definition and biosynthesis of mycobacterial cell walls and in particular that of M. tuberculosis. As rather less is known regarding its corynebacterial relatives we shall focus on M. tuberculosis and compare analogous structures, where they have been ascertained, for the corynebacteria.
Mycolates and corynomycolates
The dominant feature of the cell walls of the mycolata is the lipid-rich outer permeability barrier, which in the mycobacteria is based upon the mycolic acids. These α-alkyl branched, β-hydroxy long-chain fatty acids (Fig. 2) are found esterified to terminal arabinofuranosyl residues of arabinogalactan or as the free, solvent-extractable trehalose conjugates, trehalose mono- and dimycolates (TMM and TDM, respectively) [11, 42, 43]. Mycolic acids from mycobacteria differ from those of other mycolata in that they are the largest, containing 70–90 carbon atoms [11, 43–46]. Early structural analyses revealed that pyrolytic cleavage of mycolates yields two components, a fatty acid, corresponding to the shorter α chain, and a long chain aldehyde, termed the meroaldehyde, which derives from the long meromycolate chain of the intact lipid (Fig. 2) . There is a considerable degree of heterogeneity within the mycolic acids [43, 47]. Three of the major classes of mycolic acid are represented in M. tuberculosis. These are variously modified at the distal positions of the meromycolate chain; α-mycolates bear two cis-cyclopropane rings whereas keto and methoxymycolates possess their respective oxygen functions at the distal position with a proximal cyclopropane ring in either cis or trans configurations (Fig. 2) [11, 43]. Heterogeneity in terms of fatty acyl chain length is also evident within the mycolates. The α-branches occur as C24 and C26 chains in the approximate ratio of 1:9 with negligible amounts of C22 chains . The oxygenated mycolates are also larger than their α-mycolate counterparts. The latter commonly contain 76–86 carbons (in both chains) [44, 45] while methoxy and ketomycolates contain 83–90 and 84–89 carbons, respectively .
The structure of the AG-esterified mycolic acids lends itself to very efficient packing within the mycobacterial cell envelope. The disparity in length between the meromycolate and α-chains of the mycolic acids may allow the free lipids that complete the bilayer to interact with the inner leaflet in a more complex way than the standard end-to-end manner. It is thought that these lipids may intercalate with the distal ends of the meromycolates in this inter-leaflet hydrophobic region [11, 18–21]. With the exception of the β-hydroxy group, all other functionality, which would tend towards disruption of the highly ordered ‘linear’ packing that forms the basis of this extremely effective permeability barrier, is found towards the distal end of the meromycolate chain in this intermediate region. Alterations in the profile of distal modifications of the cell-bound mycolates and the frequency of trans-cyclopropanation play a role in the modulation of the fluidity of this outer membrane structure [43, 49, 50].
The corynebacteria also produce cell-free and arabinogalactan-bound α-alkyl, β-hydroxy fatty acids termed corynomycolic acids. These are much simpler in structure than their mycobacterial counterparts with those from C. diphtheriae containing 30–36 carbon atoms and a non-reduced β-keto group [11, 15, 51, 52]. As in the mycobacteria, the solvent-extractable corynomycolates are also found as trehalose conjugates, trehalose mono- and dicorynomycolates (TMCM and TDCM, respectively). In most corynebacteria analysed, however, the arabinogalactan-bound lipids are not present in sufficient quantity to form a complete monolayer around the cell. It has, therefore, been proposed that TMCM and TDCM may contribute to the formation of a complete inner leaflet . As these materials are found in sufficient quantity in corynebacterial cell walls, are similar in structure to parallel C16 fatty acid residues (Fig. 2) as might be found in phospholipids and can also spontaneously form liposomes, Peuch et al.  reasoned that their properties are entirely consistent with this role. It seems, therefore, that the function of the arabinogalactan-bound corynomycolates may be to form a ‘tether’ around which the outer membrane can form and be retained (Fig. 1). The increase in lateral mobility of the inner leaflet lipids in this corynebacterial model may reduce the need for the derivatisation of corynomycolates to modulate membrane fluidity.
Beneath the mycolic acid domain of the mycobacterial cell wall lies arabinogalactan (AG), the polysaccharide that covalently binds the mycolates to the peptidoglycan via a unique linker unit at its reducing end . The peptidoglycan found within the mycobacteria and other mycolata is unremarkable and similar to that found in E. coli [54–56]. In M. tuberculosis, the muramic acid residues, however, can either be glycolylated or acetylated [57, 58] and the terminal d-glutamate residue of the stem peptide can be amidated at its carboxyl group . The galactan domain of AG is linked to the C-6 position of MurNGly residues of PG via the disaccharide bridge α-l-Rhap-(1 → 3)-d-GlcNAc-(1 → P). The galactan domain consists of a linear polymer of around 30 β-d-galactofuranose (β-d-Galf) residues linked via alternating (1 → 5) and (1 → 6) glycosidic bonds . Periodically arabinan chains (three per galactan chain) are linked to the 5-position of some 6-linked Galf residues , probably nearer to the reducing end of the galactan chain . These arabinan chains are almost entirely composed of α-d-Araf residues. The bulk of the arabinan domain is polymerised via (1 → 5) glycosidic linkages with branching being introduced by 3,5-α-d-Araf residues . The non-reducing termini are decorated with a [β-d-Araf-(1 → 2)-α-d-Araf]2-3,5-α-d-Araf-(1 → 5)-α-d-Araf motif and within a particular arabinan chain two-thirds of these terminal arabinofuranoside motifs are esterified with mycolates at all four primary hydroxyl groups (Fig. 2) [60, 62].
The application of molecular mechanics modelling to the investigation of cell wall structure is a recent innovation. Dmitriev et al. [63, 64] have suggested that both galactan and the glycan moiety of peptidoglycan may be helical in nature and arranged perpendicular to the plane of the cytoplasmic membrane. In their elegant new model, the coiled glycan chains of peptidoglycan are cross-linked by peptide bridges in a highly ordered grid arrangement with the galactan chains, attached by the Rha-GlcNAc-phosphate linker unit, lying parallel to them between the peptide cross-links (Fig. 1). Further evidence for a helical galactan has been provided by the treatment of AG with a crude glycosidase preparation secreted from a Cellomonas species . One of the products produced by the degradative process was a cyclic Galf tetrasaccharide (hexa- and octasaccharides were also present) suggesting the presence of a trans-galactofuranosidase. A related activity, which is responsible for the production of cyclodextrins from starch, requires its substrate to adopt a helical formation in order to bring the two residues to be cyclised within close proximity. Therefore a scenario in which galactan adopts a helical structure in which four residues form one turn can be envisaged. It is tempting to speculate that this predicted helical structure might underpin an important spatial determinant defining the arrangement of mycolates. The combination of the proposed coiling of the galactan with the branching of the terminal hexaarabinofuranosyl motifs may achieve the lateral spacing necessary to arrange the esterified mycolates as the basis for an efficient outer membrane.
Recent studies of the arabinogalactan of C. diphtheriae showed that the linkage profile of its AG was broadly similar to that of M. tuberculosis but with a significant incorporation of mannose residues . Although the location of these mannose residues remains unclear, mannose-capping of some arabinan domains as seen in mycobacterial lipoarabinomannans is possible. A unique feature of the C. diphtheriae AG was the absence of 3,5-linked Araf residues suggesting the lack of an α-d-(1 → 3)-arabinofuranosyltransferase activity and, as a consequence, a less elaborate linear terminal Araf motif for corynomycolate deposition (Fig. 2c) .
The genetics and enzymology of cell wall core biosynthesis
The biosynthesis of the glycan moieties of the mycobacterial cell wall can be summarised as the sequential addition of sugars to a growing polyisoprenoid phosphate (Pol-P) lipid-linked chain. The necessary transfer reactions are carried out by a series of glycosyltransferases using activated sugar donors, either sugar nucleotides or Pol-P monosaccharides (Fig. 3).
Polyisoprenoid glycosyl carrier lipids
The structures of polyisoprenoid glycosyl carrier lipids and their roles in the biosynthesis of complex glycoconjugates are well established in both prokaryotes [65, 66] and eukaryotes . These are a family of membrane lipids varying in the number and stereoconfiguration of their linearly linked isoprene units. A chain of 11 isoprene residues is commonly found in the cytoplasmic membrane of eubacterial and archaebacterial cells [68, 69].
M. smegmatis is unusual among eubacteria, as it employs two different polyprenyl phosphate carrier lipids, a partially saturated C35-octahydroheptaprenyl phosphate carrier and a C50-decaprenyl monophosphate [70–72]. Recently, however, it has been shown that M. tuberculosis synthesises only the unsaturated decaprenyl monophosphate (DPP) . Although, the stereochemistry of the M. smegmatis carriers has been determined , that of the M. tuberculosis DPP is not known. These carriers are central to cell wall biosynthesis and have been isolated in various glycosylated forms; mannosylated as donors for the mannose domains of the lipoglycans lipomannan and lipoarabinomannan [70, 71], arabinosylated towards cell wall arabinan deposition  and also as ribosyl  and mannosyl-mycolyl conjugates , the latter probably involved in mycolate deposition. A series of polyprenol-based glycolipids have also been characterised which correspond to the growing AG polymer based upon the polyprenol carrier [76, 77].
Biosynthesis of polyisoprenoid carrier lipids
The non-mevalonate pathway for isoprene donor provision
Based on the successful incorporation of labelled mevalonic acid into isoprenoids of every structural class from a wide range of organisms, it was generally assumed that they were all synthesised using the prenyl donors, isopentenyl diphosphate (IPP) and its isomer dimethylallyl diphosphate (DMAPP), with both generated from acetyl-CoA via the classical mevalonate pathway. The first contradictory evidence was provided by Zhou and White  who suggested a route to IPP synthesis via pyruvate in E. coli. A labelling study using a varied panel of bacteria suggested the occurrence of two pathways to IPP synthesis in prokaryotes; Myxococcus fulvus, Staphylococcus carnosus, Lactobacillus plantarum and Halobacterium cutirubrum all utilising the mevalonate pathway while E. coli and Zymomonas mobilis used an alternative route . Rohmer et al.  demonstrated an alternative pathway to IPP in Z. mobilis, Methylobacterium fujisawaense, E. coli and Alicyclobacillus acidoterrestris, which they proposed utilised pyruvate and glyceraldehyde 3-phosphate (GAP). Much of this pathway has now been biochemically defined (Fig. 4). Initially, (hydroxyethyl)thiamin diphosphate, from the decarboxylation of pyruvate, is condensed with the C1-aldehyde group of GAP yielding 1-deoxy-d-xylulose 5-phosphate (DXP) by DXP synthase (DXS) [81–83]. A class B dehydrogenase , DXP reductoisomerase (DXR), catalyzes the transformation of DXP into 2-C-methyl-d-erythritol 4-phosphate (MEP) in two steps. In the postulated mechanism DXP is first rearranged into 2-C-methyl-erythrose-4-phosphate before its NADH-driven reduction to yield MEP. The aldehyde intermediate is apparently not released from the enzyme active site during the catalysis [85, 86]. Hoeffler et al.  synthesized the aldehyde and showed its conversion to MEP in the presence of NADPH and to DXP with NADP+. Three reactions are successively performed on MEP; firstly, conversion of MEP into 4-diphosphocytidyl-2-C-methyl-d-erythritol [88, 89] then phosphorylation of its C-2 hydroxyl group yielding 4-diphosphocytidyl-2C-methyl-d-erythritol 2-phosphate [90, 91] which is then converted to into 2-C-methyl-d-erythritol 2,4-cyclodiphosphate [92, 93]. Two of the last steps of the pathway were identified by a combination of genetic and biochemical methods using E. coli with disruptions in gcpE (ipsG) and lytB ispH). Labelling experiments in the gcpE− strain suggested that 2-C-methyl-d-erythritol 2,4-cyclodiphosphate is the substrate of IspG . Incubation of [3-14C]-2-C-methyl-d-erythritol 2,4-cyclodiphosphate with a crude cell-free system from an E. coli strain over-expressing gcpE resulted in the formation of 4-hydroxy-3-methylbut-2-enyl diphosphate [95, 96]. In agreement with this observation, the accumulation of this diphosphate was seen in the lytB− strain . Rohdich et al.  showed that over-expression of ispH led to the transformation of labeled deoxy-xylulose into a 5:1 mixture of labelled IPP and DMAPP. IpsH has been characterised as a 4-hydroxy-3-methylbut-2-enyl diphosphate reductase . Although the recombinant protein was inactive when purified, its use to supplement extracts of E. coli and Capsicum increased their low intrinsic activity suggesting that IspH requires as yet unidentified accessory proteins to form the isoprene donors IPP and DMAPP . Although the interconversion of IPP to DMAPP can be achieved via an IPP isomerase [100, 101] encoded by idi in E. coli, the independent synthesis of these two products suggesting an as yet undefined branching of the pathway has been genetically demonstrated .
Evidence that mycobacteria utilise this non-mevalonate pathway was initially provided by the failure to incorporate [14C]mevalonate into dihydromenaquinone by both M. tuberculosis and M. smegmatis. The pattern of 13C incorporation from labelled glucose into dihydromenaquinone by M. phlei and M. smegmatis unequivocally demonstrated synthesis via the non-mevalonate route . Bailey et al.  cloned Rv2682c, a gene encoding a protein sharing 38% identity with DXP synthase from E. coli. Its subsequent overexpression in E. coli resulted in the purification of an active recombinant enzyme. The existence of this enzymatic activity in M. tuberculosis strongly suggests the utilisation of the non-mevalonate route to IPP/DMAPP in the organism. None of the other genes that might contribute to IPP/DMAPP biosynthesis have yet been identified and characterised from mycobacteria. A dxs homologue occupies DIP1397 in C. diphtheriae genome suggesting that the non-mevalonate route to IPP synthesis may be used in this bacterium and may be common to all actinomycetes. Putative ipsDF homologues are clustered together in both bacteria in a similar genetic setting.
An ipsE homologue of M. tuberculosis, Rv1011, shares homology with DIP0876 and a common genetic context. Upstream of Rv1011 lies a homologue of ksgA from E. coli. In E.coli ksgA, encoding an rRNA modifying enzyme, is partly co-transcribed with pdxA, a 4-hydroxythreonine 4-phosphate dehydrogenase whose product is condensed with DXP towards the biosynthesis of the ubiquitous coenzyme, pyridoxal 5-phosphate, which is of fundamental importance to the metabolism of all cells. Positive growth rate-dependent regulation of the transcription of these genes has been demonstrated . It is not known whether a similar growth rate-dependent regulation is present in the mycolata. Intriguingly the upstream loci are occupied by members of the rpf family whose products are autocrine growth factors implicated in the resuscitation of actinomycetes from stationary phase [106, 107]. In E. coli also genes implicated in stationary phase survival occupy the loci upstream of kpsG. It is tempting to speculate that the co-regulation of DXP-related metabolic fluxes with factors aiding stationary phase survival and KsgA activity may be crucial to the global regulation of growth rate in the mycolata and maybe bacteria in general.
The presence of the non-mevalonate dependent pathway to IPP/DMAPP in plants and pathogenic bacteria, but not in man, has attracted considerable interest in its potential as a target for novel antibiotics and herbicides. Fosmidomycin represents a specific high-affinity inhibitor of IspC, the DXP reductoisomerase. Recently, its anti-malaria activity in man has been demonstrated in clinical trials  and the structural basis of fosmidomycin action revealed by the resolution of the crystal structure of its complex with IspC from E. coli. As the supply of polyprenyl isoprenoid carrier lipids has been suggested to regulate the growth rate of bacteria, this pathway thus represents an attractive target for novel anti-tuberculosis chemotherapeutics . This may be particularly relevant as the slow-growing M. tuberculosis produces polyprenol carriers at much slower rates than its fast-growing, saprophytic counterpart M. smegmatis.
Interestingly, intermediates from the non-mevalonate pathway also play a role in the activation of γδ T cells. Although IPP was the first ligand described for the Vγ9/Vδ2 T cell receptor, 4-hydroxy-3-methylbut-2-enyl pyrophosphate, which accumulates in lytB− mutants of E. coli, evokes a vastly greater response . A similar, but incompletely characterized  molecule purified from various mycobacteria is also implicated in the activation γδ T cells .
Isoprenyl pyrophosphate synthases
The linear polyprenyl chains of isoprenoid carrier lipids are constructed by a family of isoprenyl pyrophosphate synthases (IPPS). Each transfer reaction is initiated with the elimination of the diphosphate ion from an allylic diphosphate. The allylic cation formed is subject to nucleophilic attack by IPP with stereospecific removal of a proton to form a new carbon-carbon bond and a new double bond in the product. By consecutive repetition of these 1′-4 head to tail condensations between IPP and the growing allylic prenyl diphosphate product, IPPSs ultimately synthesise a prenyl diphosphate of a defined length and stereochemistry [112, 113].
In the first round, the allylic primer is DMAPP and prenyl chain elongation forms the C10 geranyl diphosphate (GPP). In M. tuberculosis and M. smegmatis it is suggested that ω,E-GPP acts as the primer for two distinct IPPSs, a membrane-associating Z-IPPS that forms ω,E,Z-farnesyl-PP (C15) and a cytoplasmic enzyme that forms ω,E,E,E-geranylgeranyl-PP (C20). . In both bacteria, the ω,E,Z-farnesyl-PP is further elongated by another membrane-associated IPPS to form decaprenyl diphosphate, the immediate precursor to DPP carrier lipid. In M. smegmatis, the ω,E,E,E-geranylgeranyl-PP is extended for the synthesis of octahydroheptaprenyl diphosphate. In M. tuberculosis, which does not synthesise the latter carrier lipid, ω,E,E,E-geranylgeranyl-PP was converted to a novel octaprenyl species at a very low rate. This polyisoprenoid has never been recognised as a carrier lipid and its role or fate is not known . Two IPPSs from M. tuberculosis have now been cloned and characterised [114, 115]. The product of Rv1086 adds one isoprene unit to ω,E-GPP forming ω,E,Z-FPP, the putative substrate for Rv2361c. This enzyme adds a further seven isoprene units to form decaprenyl diphosphate (Fig. 3) . Orthologues of these appear to be encoded by DIP0929 and DIP1712, respectively. The isoprenoid carrier lipids of the corynebacteria have not yet been characterised but initial evidence suggests the presence of a mannosylated-C55 polyprenylphosphate in C. glutamicum (K.J.C. Gibson, G.S. Besra and L. Eggeling, unpublished results). Thus the chain length specificity of DIP1712 may direct the synthesis of an undecaprenol diphosphate akin to those of other eubacteria rather than the shorter counterparts seen in mycobacteria.
Although the biosynthesis of peptidoglycan (PG) in E. coli is relatively well defined [59, 116], little biochemical data has been gathered regarding the biosynthesis of mycobacterial PG but it is assumed to be very similar to that described for E. coli. The reader is referred to an excellent recent review describing the comparative biochemistry of the cross-linking of bacterial peptidoglycans .
Lipid II synthesis
The immediate precursor to polymerised PG is lipid II, a polyisoprenoid-bound disaccharide . In E. coli the structure of lipid II is thus, N-acetylmuramic acid (MurNAc) is bound via its C-1 carbon and a pyrophosphate bridge to its undecaprenyl carrier lipid. The MurNAc is linked via its C-4 carbon to GlcNAc and via its C-3 carbon to the stem polypeptide, which participates in PG cross-linking . In M. tuberculosis, the formation of UDP-MurNAc from UDP-GlcNAc is thought to be catalysed by the products of the genes designated murA and murB, Rv1315 and Rv0482, respectively. Demonstration of an increase in UDP-GlcNAc-dependent phosphate release from phosphoenolpyruvate in extracts of M. smegmatis over-expressing Rv1315 is consistent with MurA fulfilling the role of UDP-GlcNAc 1-carboxyvinyltransferase suggested by its homology with its E. coli counterpart . The next step in PG biosynthesis is the formation of UDP-MurNGly-pentapeptide. Initially the lactate moiety of UDP-MurNGly is modified by the addition of l-alanine with the peptide being sequentially extended via additions of d-glutamate, diaminopimelate and a d-alanyl-d-alanine dipeptide, reactions catalysed by MurC, MurD, MurE and MurF, respectively. Homologues of the genes encoding these enzymes are found clustered together in M. tuberculosis H37Rv. Also present in the cluster are a homologue of murX, whose product is thought to transfer phosphoryl-murNGly-(pentapeptide) from the nucleotide donor to its DPP acceptor to form decaprenyl-diphosphoryl-MurNGly-(pentapeptide) and homologues of ftsQ, ftsW and ftsZ, genes essential for cell division and septum formation [59, 116]. Of these genes, however, only the product of murC (Rv2152c) has been functionally characterised , although MurD is apparently enzymatically similar to its E. coli counterpart . Although the other mur homologues await functional definition, a degree of confidence in their assignment can be gained from their genetic organisation. The arrangement is very similar to that seen in the division cell wall dcw cluster of E. coli [59, 116] and is also conserved in C. diphtheriae (Fig. 5).
The latter stages of PG biosynthesis are carried out by a set of penicillin-binding proteins (PBP) [59, 116]. High molecular mass PBPs (HMM-PBP) are two domain proteins that can be divided into two classes depending upon the structure and activity of their N-terminal domain. In both cases the C-terminal domain is responsible for the transpeptidation reactions that cross-link the glycan chains. In Class A HMM-PBPs the N-terminal domain carries a glycosyltransferase activity, thus they are bifunctional enzymes capable of transglycosylation and transpeptidation reactions towards PG polymerisation . The N-terminal domain of Class B HMM-PBPs are probably involved in interactions with other membrane proteins . Several PBPs have now been identified in mycobacteria [59, 121–125]. Among these are two putative class A HMM-PBP of M. tuberculosis, encoded by Rv0050 (ponA) and Rv3682. A soluble version of one of these murein polymerases, the ponA product (s-PBP1*), has been constructed and shown to bind benzylpenicillin . Recent analysis  has suggested that the other class A HMM-PBP encoded by Rv3682 is a penicillin-resistant murein polymerase. C. diphtheriae appears to carry orthologues of both HMM-PBPs in DIP2294 and DIP0298, respectively. The cross-links between the stem peptides of E. coli, the mycobacteria and the corynebacteria during exponential growth occur between the fourth residue (d-Ala) and the third (meso-diaminopimelate) of another. During stationary phase several bacteria including M. tuberculosis alter their PG cross-linking pattern incorporating increasing amounts of (3 → 3) linked peptides along with the (4 → 3) linked variety (Fig. 1). This increasing frequency of (3 → 3) interpeptide bridges has been associated with penicillin resistance and it is likely that these linkages are created by the penicillin-resistant HMM-PBPs such as Rv3682 .
M. tuberculosis and C. diphtheriae both appear to possess a single representative of three subclasses of Class B HMM-PBPs. Rv2163c and DIP1604 both appear to encode class B3 HMM-PBPs  and their loci are associated with the dcw clusters of the respective organisms (Fig. 5). Similarly, DIP55 and Rv0016 have been assigned to subclass B-Like I  and these are clustered with orthologues encoding cell division and cell shape-determining factors and such as FtsW and PknAB. Another class B HMM-PBP is apparently encoded by Rv2864c and DIP1497 and it is likely that these do not bind penicillins. Consequently these have been assigned to the penicillin-resistant subclass B-like II .
M. tuberculosis contains three genes encoding ‘free-standing’ PBPs that may be involved as auxiliary cell cycle proteins in (4 → 3) linked PG synthesis . These comprise genes encoding putative D-alanyl-D-alanine carboxypeptidases dacB1 and dacB2, Rv3330 and Rv2911, respectively, and Rv3627c. Like M. leprae, C. diphtheriae contains two similar genes , DIP0637 showing greater homology with dacB1B2 while DIP2005 with Rv3627c.
Biosynthesis of arabinogalactan P-Pol glycoconjugates
The first evidence suggesting the course of the biosynthesis of the entire linkage unit-AG complex was revealed by the elaboration of a series of glycolipids by cytoplasmic membrane preparations of M. smegmatis and M. tuberculosis. Membrane preparations from both species catalysed the incorporation of radioactivity from UDP-[14C]GlcNAc into two glycolipids (GL1 and GL2) . This incorporation of 14C was tunicamycin-sensitive and the glycolipid products were mild-acid labile and mild-alkali stable, all features consistent with their assignment as polyprenyl phosphate-based glycolipids. This data suggested that the initial step in the synthesis of mycobacterial cell wall AG involves the formation of a polyprenyl-P-P-GlcNAc unit (GL1), either the C35-P or C50-P isoprenoid carriers. Incorporation of [14C]Rha from dTDP-[14C]Rha only occurred into GL2, identifying it as polyprenyl-P-P-GlcNAc-Rha . Addition of a cell wall enzyme preparation resulted in the formation of more polar glycolipids, GL3 and GL4. The inclusion of UDP-[14C]Galp resulted in exclusive labelling of GL3 and GL4 indicating stepwise growth of the galactan chain on the polyprenol-P-P-GlcNAc-Rha unit . Solubilisation of the polymerised product resulting from the labelling experiments and its subsequent analysis pointed to the formation of even higher intermediates, GL5, GL6, etc., eventually resulting in a polymer that was consistent with a highly polymerised, lipid-linked version of GL1-4 [77, 126]. The polymer was found to consist of 35–50 residues, and glycosidic linkage analysis showed evidence for t-Galf, 5-linked-Galf, 6-linked-Galf and 5,6-linked-Galf, indicative of the alternating 5- and 6-linked linear galactan components of mycobacterial cell walls with a small amount of branching. The incorporation of radiolabel from synthetic decaprenyl-P-[14C]Araf into the same polymer  suggests that the total synthesis of the arabinan domain of AG may occur while still linked to the polyprenyl-P carrier. Our current understanding of mycobacterial cell wall core biosynthesis, including the provision of key sugar donors and carrier lipids is portrayed in Fig. 3.
Linker unit synthesis
The enzyme catalysing the initial reaction in the glycosylation of the DPP carrier, i.e. the transfer of Glc-NAc-1-phosphate from its UDP donor, remains unidentified. However, the complementation of a wbbL mutant of E. coli, deficient in Rha transfer in lipopolysaccharide biosynthesis, with Rv3265c implicates its product as the probable rhamnosyltransferase involved in linker unit synthesis . The enzymes providing the dTDP-Rha donor substrate have also been identified and expressed in E. coli [129, 130]. RmlA to RmlD have been characterised as an α-d-glucose-1-phosphate thymidylyltransferase, dTDP-d-glucose 4,6-dehydratase, dTDP-4-keto-6-deoxy-d-glucose 3,5 epimerase and dTDP-Rha synthase, respectively . The lack of rhamnose in human glycochemistry identifies all of these as good candidate targets for novel anti-tubercular chemotherapeutics. Indeed, the inactivation of rmlD in M. smegmatis was found to be lethal in the absence of a rescue plasmid carrying a functional copy of rmlD thereby demonstrating that dTDP-Rha is an essential component for mycobacterial growth and supporting the enzymes towards its synthesis as novel chemotherapeutic targets. Homologues of rmlABCD have all been identified within the C. diphtheriae genome. In C. diphtheriae the rml genes occupy three adjacent loci with DIP0361 representing a fusion of rmlC and rmlD, the former encoding the N-terminus of the fused product. In M. tuberculosis rmlBC (Rv3464-Rv3465) are clustered, as are rmlD and wbbL (Rv3266c and Rv3265c, respectively). The gene encoding the rhamnosyltransferase of C. diphtheriae is not yet apparent.
The provision of the nucleotide sugar donor for galactan deposition has also been defined in M. tuberculosis. Two reactions are required to reach UDP-Galf from UDP-Glcp, the first being catalysed by UDP-Glcp epimerase to form UDP-Galp, which is then converted to UDP-Galf by UDP-Galp mutase. Weston et al. purified a protein with UDP-glucose 4-epimerase activity from M. smegmatis. N-terminal sequence analysis of the protein suggested it to be related to the product of Rv3634. A similar strategy was used to identify an M. smegmatis orthologue of Rv3809c as a UDP-Galp mutase, Glf. The identity of the protein encoded by Rv3809c as the M. tuberculosis Glf was confirmed by molecular cloning and analysis of crude extracts containing the recombinant protein .
Four glycosyltransferase activities may be required to carry out the synthesis of the large galactan-LU-P-Pol glycoconjugates; a UDP-Galf:Rha-GlcNAc-P-Pol Galf transferase activity, β-d-Galf transferases forming the alternating (1 → 5) and (1 → 6) internal glycosidic linkages of the galactan and possibly a transferase which deposits the terminal Galf residue. Several candidate genes from M. tuberculosis were investigated as potential Galf transferases. One of the major factors determining the folding pathways of proteins is hydrophobic interaction between neighbouring amino acids. In hydrophobic cluster analysis (HCA) the amino acid sequence of a protein is plotted as a theoretical α-helix and clusters of hydrophobic amino acid residues are highlighted on this template. The shapes of these clusters suggest secondary structure features, a vertical cluster representing a hydrophobic β-strand and horizontal features suggesting α-helices. Thus the pattern of hydrophobic clusters in a protein may relate information regarding its fold [133–135]. As sequence homology between glycosyltransferases is often very low, HCA is particularly useful in categorising these enzymes into functional classes . Genes from M. tuberculosis H37Rv showing homology to glycosyltransferases were analysed by HCA and the plots analysed for recognised features of inverting glycosyltransferases , i.e. those which carry out sugar transfers in which the configuration of the anomeric carbon is inverted. These features include a domain A with alternating hydrophobic α-helices and β-strands in which the DXD motifs, commonly found in glycosyltransferases, are positioned close to the C-terminal end of strong vertical hydrophobic clusters . Several candidates were identified within the M. tuberculosis H37Rv genome satisfying these criteria. One in particular, Rv3808c, appeared to be a good candidate in that it occupied the locus adjacent to glf, encoding the UDP-galactopyranose mutase. As these genes overlapped by 4 nucleotides, it was considered likely that they comprised part of an operon. The putative transferase also appeared to contain the signature QXXRW motif of the less well defined Domain B, which is found only in processive enzymes, i.e. those which carry out multiple sugar transfers . Molecular cloning and over-expression of Rv3808c was achieved by two groups [77, 138]. Mikusova et al.  successfully over-expressed Rv3808c in M. smegmatis, observing an increased yield of a bona fide galactofuran in the over-producing strain. Kremer et al.  designed neoglycolipid acceptor substrates incorporating a Galf disaccharide moiety, the sugars either being linked β-(1 → 5) (G5G) or β-(1 → 6) (G6G). An identical profile of radioactive trisaccharide and tetrasaccharide products were formed when these were incubated with membrane preparations of E. coli over-producing Rv3808c as was seen when the acceptors were used to probe the galactosyltransferase activities of native mycobacterial membrane preparations. Glycosidic linkage analysis of these products showed that the new linkage in the G5G-derived trisaccharide was (1 → 6) and that the G6G equivalent was (1 → 5). Moreover, the tetrasaccharide product, which was abundant using this acceptor, also contained a new (1 → 6) linkage suggesting that, wholly consistent with the repeating nature of native galactan, the product of Rv3808c was a processive enzyme capable of depositing alternating (1 → 5) and (1 → 6)-linked Galf residues from UDP-galactose. This enzyme, now designated GlfT, is likely to be responsible for the bulk of galactan deposition in the cell wall of M. tuberculosis, but it still remains unclear if it directs the deposition of the initial galactofuranose residue on the rhamnose residue of the linker unit. The importance of galactan synthesis to the growth of mycobacteria was demonstrated by the disruption of glf in M. smegmatis by allelic replacement. The resulting knockout mutant was only achieved in the presence of two recovery plasmids bearing the M. tuberculosis glf and glfT genes suggesting a polar effect on glfT expression. Functional copies of both genes were essential for mycobacterial growth thus emphasising the importance of galactan biosynthesis as a target for novel chemotherapeutics .
Unlike the situation in M. tuberculosis in which glfT lies immediately downstream and overlaps with glf, the corresponding loci are separated by a further four open reading frames in C. diphtheriae (Fig. 6). DIP2199–DIP2202 appear to encode a functional unit devoted to the production of the osmoprotectant glycine-betaine from phosphorylcholine. DIP2199 shares significant homology with phosphoryl choline phosphatases and it is possible that phosphoryl choline, released from a source such as phosphatidylcholine by a phospholipase C, could be utilised by this enzyme to form choline. DIP2202 appears to encode a choline dehydrogenase that could convert choline to glycine-betaine aldehyde and DIP2200 potentially encodes a glycine-betaine aldehyde dehydrogenase that could complete the biosynthesis of the osmoprotectant from the aldehyde. A gene that appears to encode a transporter protein with significant homology to characterised choline transporters occupies the neighbouring locus, DIP2201. M. tuberculosis possesses homologues of all of these genes but these are spread across the chromosome and may not provide a functional glycine-betaine biosynthetic pathway. The positioning of these genes in C. diphtheriae however, does suggest a functional relationship but the relevance of their positioning between glf and glfT loci is not clear and it is entirely possible that their placement is irrelevant to cell wall assembly. It is tempting to speculate, however, that cell wall assembly may require a low water activity and that this is more difficult to maintain in the corynebacterial system requiring the provision of an osmoprotectant to exclude water from the sensitive process. The arrangement of the open reading frames suggests the possibility of their divergent transcription from a promoter between DIP2201 and DIP2202. The effect of this insertion upon the presumably important co-regulation of glf and glfT in C. diphtheriae is not known.
In vivo pulse-chase labelling experiments in M. smegmatis suggested that the Araf residues ultimately deposited in AG derive directly from decaprenyl-Araf (DPA)  and that the action of ethambutol, an anti-tubercular agent known to inhibit AG and arabinomannan biosynthesis , led to the accumulation of DPA. Moreover, the use of synthetic [14C]DPA  as an in vitro Araf donor in a cell-free assay system and the subsequent characterisation of polymerised arabinan showed that radiolabel was deposited within all of the recognised cell wall arabinan motifs suggested DPA as the major Araf donor in mycobacteria . The possibility of both UDP-Ara  and GDP-Ara  has been suggested in M. smegmatis and these cannot be ruled out as minor cell wall Araf donors. The biosynthetic origin of the arabinosyl residues was traced from variously labelled forms of glucose and was suggested to proceed via the nonoxidative pentose shunt route . As attempts to assay the obvious route from the ribulose-5-phosphate to arabinose via arabinose phosphate isomerase failed, an alternative route was investigated. Wolucka et al. [72, 75] had previously demonstrated the presence of Pol-P-ribose in mycobacteria including M. tuberculosis and proposed that it may serve as a precursor of an as yet unknown ribose-containing cell wall or capsular polymer. A 2′-epimerase activity able to convert the ribose form to DPA could not be detected suggesting that it did not act as direct biosynthetic precursor of DPA. Scherman et al. subsequently showed that both DPA and Pol-P-Rib are derived from 5-phosphoribose pyrophosphate with a 2′ epimerase mediating the Rib → Ara conversion at an intermediate stage . The enzymes that direct the formation of DPA in M. tuberculosis have not yet been identified.
The products of the emb locus of M. avium were identified as the targets for the antimycobacterial drug ethambutol. Overexpression of embA and embB from M. avium conferred ethambutol resistance in M. smegmatis. The inhibitory effect of ethambutol on the incorporation of Araf from DPA into arabinan in vitro was partially recovered in extracts from cells overexpressing embB. In accord with these observations, many ethambutol resistance-associated mutations occur within embB and in particular within a predicted ‘periplasmic’ region of the protein [146–148]. M. tuberculosis possesses three closely related emb genes, clustered embCAB. A similar genetic organisation occurs in M. smegmatis and recent gene knock out studies in this organism have shed light on this apparent redundancy . Individual mutants inactivated in embC, embA and embB were characterised. All three strains were viable but of them, the embB− mutant was most profoundly affected. Cell wall integrity seemed to be drastically altered as morphological changes were evident and the cells displayed increased sensitivity to hydrophobic drugs and detergents. The arabinose content of the AG was diminished for both the embA and embB− strains. Nuclear magnetic resonance studies showed that these mutations resulted in considerable effects upon the formation of the terminal hexaarabinofuranosyl motives, specifically the addition of the β-d-Araf-1 → 2-β-d-Araf disaccharide to the 3 position of the 3,5-linked Araf residue resulting in a linear terminal motif. The AG formation of the embC− strain, however, seemed unaffected yet arabinan deposition in the lipoglycan lipoarabinomannan (LAM) was abolished. These data suggest that the mycobacterial Emb proteins are intimately involved in the process of cell wall arabinan deposition and that EmbA and EmbB are crucial to the formation of the hexaarabinofuranosyl motives of AG that are crucial for the esterification of mycolic acids. As the expression of active Emb proteins in a heterologous system is still to be achieved, their role as arabinosyltransferases remains unconfirmed. As these proteins contain 11–13 predicted transmembrane helices, the possibility arises that they could play a part in arabinan export rather than polymerisation, or even combine both functions as a bifunctional protein. Another scenario is that EmbA and EmbB could combine an export role with the recruitment of appropriate, and as yet undefined, arabinosyltransferase activities specifically to AG synthesis. A similar role for EmbC could be envisaged in LAM biosynthesis.
The scenario in C. diphtheriae is vastly simplified by the occurrence of a single Emb homologue (DIP0159), which is most closely related to EmbC of the mycobacteria. As C. diphtheriae AG contains no 3,5-linked Araf residues this appears entirely consistent with the lack of EmbAB functions, i.e. the deposition of the β-d-Araf-1 → 2-β-d-Araf disaccharide to the C-3 position of the 3,5-linked Araf residue as in mycobacteria. As C. diphtheriae does contain arabinan-containing lipoglycans it is possible that this single Emb protein possesses broader acceptor specificity than its mycobacterial counterparts and can direct the arabinan deposition in both molecules but is unable to produce the more elaborate structures that gene duplications and the accumulation of point mutations may have allowed in the mycobacteria.
Mycolic acid biosynthesis and deposition
Fatty acid synthase systems
Fatty acid biosynthesis occurs via the repetition of a cycle of four reactions, each cycle culminating in the extension of the alkyl chain by a two-carbon unit. The first reaction is a Claisen-type condensation reaction in which a malonate residue is decarboxylated and undergoes a condensation reaction with the thioester of the growing fatty acid and an active site thiol of the appropriate enzyme . The extension of the acyl chain is achieved thus and the remaining three reactions progressively remove the resulting β-keto-group of the product and return the chain to an alkane form. Firstly, the β-keto product is reduced to form a β-hydroxy-acyl chain and then dehydrated to form an enoyl intermediate that, after further reduction, yields the aliphatic substrate for the next round of biosynthesis.
Two types of fatty acid synthase (FAS) are known, both systems containing similar enzymic functions but differing in their organisation. A single gene encodes FAS-I, the system commonly found in vertebrates. The active protein comprises a homo-dimer and contains all the necessary functions to achieve de novo fatty acid synthesis . Most bacteria, however, utilise the FAS-II system in which the growing fatty acyl chain is shuttled between the active sites of the dissociated component enzymes as an acyl thioester of a small and highly acidic acyl carrier protein (ACP). Mycobacteria are unusual in that they contain both systems. The FAS-I system, encoded by fas (Rv2524c), is responsible for de novo fatty acid synthesis, producing a bimodal distribution of C16–C26 fatty acids. The FAS-II system of M. tuberculosis is similar to other bacterial FAS-II systems except for its primer specificity. This system is widely believed to extend relatively long acyl chains as thioesters of AcpM, a C-terminally-extended homologue of bacterial ACPs, to presumably provide the long meromycolate chains of the mature mycolic acids [152, 153]. Other scenarios are possible, however and have recently been reviewed elsewhere .
The pivotal link between the FAS-I and FAS-II is the β-ketoacyl-ACP synthase III, mtFabH (Rv0533c) . This enzyme elongates the acyl CoA primers derived from FAS-I forming β-keto-ACP thioester products through their condensation with malonyl-AcpM. In the proposed reaction mechanism the enzyme is firstly transacylated, the incoming acyl chain forming a thioester with an active site cysteine residue and its CoA moiety being released. The docking of malonyl-AcpM is followed by the decarboxylation of the malonyl residue, the resulting carbanion attacking the fatty acyl-enzyme thioester linkage thus affecting the condensation reaction [150, 155]. The carbonyl oxygen of the primer thus contributes the characteristic β-keto group of the product. The reduction of this group to a hydroxyl function is then carried out by the β-keto-acyl-ACP reductase, MabA. This gene has been cloned and expressed in E. coli and its NADPH-dependent β-keto-acyl reductase activity demonstrated using acetoacetyl-CoA . The preference of the enzyme for long-chain substrates has subsequently been demonstrated  and the recent determination of its crystal structure has led to the recognition of structural differences between this enzyme and its relatives . In M. tuberculosis, and all other mycobacteria tested, mabA lies adjacent to inhA. This gene has been shown to be the target of the anti-tubercular drugs isoniazid (INH) and ethionamide (ETH)  and to encode the NADH-specific reduction of 2-trans-enoyl-ACP reductase involved in FAS-II . By comparing the crystal structures of the wild-type enzyme and an INH-resistant mutant, this study also demonstrated that the perturbations in a hydrogen-bonding network that stabilizes NADH binding led to INH-resistance . A detailed biochemical study demonstrated the enzyme's preference for long chain substrates consistent with its involvement in long chain fatty acid synthesis . The gene encoding the dehydratase function necessary to convert the β-OH product of MabA into the enoyl substrate of InhA has yet to be identified. The subsequent rounds of acyl extension in FAS-II are thought to be initiated by the highly similar β-keto-acyl-ACP synthases, KasA and KasB [43, 162]. Both enzymes have been expressed in E. coli, purified and characterised. Both enzymes extend acyl-ACP thioesters rather than acyl CoAs condensing them with malonyl-ACP [163, 164]. Their substrate specificity, with both enzymes preferring acyl-ACPs of at least 16 carbons, is consistent with a role in FAS-II extending FAS-I products towards the biosynthesis of long chain fatty acids . Kremer et al. also showed that over-expression of kasA from M. tuberculosis in both M. smegmatis and M. chelonae led to a decrease in the amounts of the shorter α′-mycolates in their cell walls with a concomitant increase in α′-mycolates . The KasAB enzymes of M. smegmatis are very similar to those of M. tuberculosis and over-expression of M. smegmatis’ own KasA produced the same effect suggesting that the balance of α′- and α-mycolates is probably achieved purely by regulation of the amounts of the Kas and associated enzymes rather than via any qualitative differences between them and their M. tuberculosis orthologues . In M. tuberculosis, kasAB are clustered with a series of genes implicated in fatty acid synthesis, which presumably form an operon. These include acpM and fabD, the latter encoding a malonyl-CoA acyltransferase responsible for the formation of the malonyl-AcpM used as an extension substrate by mtFabH, KasA and KasB .
Although a fas orthologue (DIP1846) is evident in the C. diphtheriae genome, homologues of the FAS-II components are not present. The absence of long-chain mycolic acids in corynebacteria is therefore, a consequence of these bacteria representing ‘natural mutants’ in FAS-II. As the corynomycolic acids from C. diphtheriae contain 32 carbons , it is entirely possible that these are formed via the Claisen-type condensation of C16–CoA esters.
Mycolic acid maturation
The relative simple corynomycolates are much shorter than those of the mycobacteria and, unless the bacteria encounter exogenous desaturated fatty acids, contain no functionality other than the β-keto group introduced by the mycolic condensation that ligates the two acyl chains. Several of the genes associated with the introduction of functionality to meromycolates have been identified in M. tuberculosis and have been extensively reviewed elsewhere [43, 152, 153]. As these modifications are only found in the mycobacteria they will only be discussed here briefly to highlight their importance to mycobacterial permeability, persistence and pathogenicity.
Takayama et al.  documented that exposure to isoniazid inhibited the synthesis of a series of non-mycolate long chain fatty acids from M. tuberculosis that they considered precursors to mycolic acids. Extraction and characterisation of these fatty acids revealed that they contained cyclopropanyl and methoxy functions and a series of unsaturated fatty acids which all appeared to be structurally related to the α- and methoxymycolates . Similarly, a range of monounsaturated and diunsaturated long chain fatty acids were extracted from M. smegmatis that appeared to be structurally related to its α-mycolates . These meromycolate-like C48–C56 fatty acids could be synthesised using a palmitate primer in an M. tuberculosis-derived cell free system and were also labelled using [14C]S-adenosyl-methionine (SAM) suggesting the involvement of methyltransferases in meromycolate derivatisation . Recently Barry and colleagues have provided strong evidence supporting the hypothesis that the meromycolate chain is modified during elongation prior to its condensation with the α-alkyl chain . They observed that a range of long chain fatty acid residues derived from heat-inactivated extracts of M. smegmatis act as acceptors of radiolabelled methyl groups from SAM. Labelling was strongly inhibited by an antiserum that specifically recognises AcpM implicating the protein as the acyl chain carrier throughout meromycolate elongation and modification.
Two putative long-chain fatty acid desaturase genes are present in the genome of M. tuberculosis H37Rv  occupying loci at Rv0824c and Rv1094. The former, desA1, encodes a product (Des) that was originally detected as an exported component of an M. tuberculosis PhoA fusion library processed in M. smegmatis. The amino acid sequence of this protein shows significant homology to plant acyl-ACP desaturases and contains two copies of the characteristic (D/E)ENXH motifs of the class II diiron-oxo proteins of which acyl-ACP desaturases are members. The sequence of the product of desA2, however does not contain these motifs although it does possess homology throughout with stearoyl-ACP desaturases. Neither of the gene products have yet been characterised in terms of desaturase activity. The significance of Des secretion is, at this point, unknown. Consistent with its elaboration of the shorter unmodified corynomycolic acids, we have not identified any desA homologues within the genome of C. diphtheriae.
Cyclopropanation is a common modification in the mycolic acids of the slow-growing pathogenic mycobacteria. The fast-growing species, such as M. smegmatis do not produce cyclopropananted mycolates but produce large amounts of unsaturated mycolic acids . Yuan et al.  isolated a cosmid containing DNA from M. tuberculosis that mediated the transformation of the distal cis-double bond in the major mycolic acid of M. smegmatis to a cis-cyclopropane ring. The activity was traced to a single gene cma1 that encoded a protein with significant homology to the cyclopropane fatty acid synthase from E. coli. Interestingly, M. smegmatis overexpressing cma1 significantly resisted killing by hydrogen peroxide suggesting that distal cyclopropanation may represent an important adaptation of the pathogenic mycobacteria to oxidative stress.
A second gene, cma2, was subsequently discovered whose product CMAS-2 shared significant homology (52% identity) with that of cma1 (CMAS-1) but appeared more closely related to a putative cyclopropane synthase from M. leprae. Expression of cma2 in M. smegmatis resulted in the cis-cyclopropanation of the proximal double bond of α1 mycolates. Expression of both cma1 and cma2 was required to cis-cyclopropanate both double bonds producing a molecule similar to the α-mycolates of M. tuberculosis. Differential scanning calorimetry (DSC) studies using purified mycolic acids and cell wall preparations from M. smegmatis overexpressing these genes individually and in combination revealed that expression of cma2 mediated a significant increase in phase transition temperatures suggesting a decrease in the fluidity of the cell wall mycolate domain and thus an important role for proximal cyclopropanation in cell wall integrity .
More recently Glickman et al.  have identified and characterised a novel mycolic acid cyclopropane synthase in M. tuberculosis. An M. bovis BCG transposon library was screened for mutants deficient in cord formation, a morphological property that can be correlated to the virulence or persistence of a particular strain of M. tuberculosis or M. bovis BCG for mice [174–176]. Disruption of pcaA, a homologue of cma1, led to the loss of the cording phenotype. The involvement of pcaA in cording in M. tuberculosis was confirmed by its deletion by allelic exchange. Careful analyses of mycolic acids from the mutant revealed that PcaA is necessary to synthesise the proximal cis-cyclopropane ring of M. tuberculosisα-mycolates. When the M. tuberculosis pcaA mutant was used to infect mice, the mutant grew more rapidly than the parent strain, reaching 3-fold higher levels after the first three weeks. However, the mutant was eliminated from the animals much more quickly and after 135 days the numbers of mutant bacteria isolated from the mice was significantly lower than for the parent strain despite its initial numerical superiority. Significantly after 219 days of infection, all of the mice infected with the parent strain had died whereas the mutant infected mice had all survived and were healthy.
The apparent redundancy in proximal cyclopropanation was resolved by the construction of a chromosomal cma2 deletion mutation in M. tuberculosis. Unsaturated derivatives of both the methoxy- and ketomycolic acids accumulated in the mutant strain. Analysis of mycolic acids derived from this strain by NMR revealed that they were normal in terms of cyclopropane and methyl branch content but lacked trans-cyclopropane rings; the proximal cyclopropane rings of the oxygenated mycolates of the pathogenic mycobacteria can occur in both conformations, whereas the α-mycolates are decorated by two cis rings. The product of cma2, therefore appears to be a mycolic acid proximal trans-cyclopropane synthase. It is plausible that the substrate specificity of these methyltransferases may be partly defined by interaction with other meromycolate modifying enzymes . Although there is no supporting experimental evidence, the coordination of meromycolate elongation and modification in multi-enzyme complexes seems a likely route to efficient meromycolate biosynthesis and ultimately to the regulation of the classes of mycolic acid produced. It is, therefore, possible that the production of cis-cyclopropane rings in M. smegmatis overexpressing cma2 may reflect a misincorporation of CMAS-2 in such a complex.
An initial insight into the production of the oxygenated mycolates was gained when cma1 was used as a gene probe to aid the identification of other meromycolate modifying enzymes . This approach led to the discovery of a cluster of genes encoding four highly homologous methyltransferases designated mma 1–4. Expression of the whole cluster facilitated the production of methoxymycolates in M. smegmatis. All four gene products (MMAS1-4) and both cma products (CMAS1-2) share a high degree of homology and contain a conserved SAM binding domain. Analysis of mycolic acids derived from M. smegmatis expressing different combinations of these mma genes suggested a biosynthetic route to the methoxymycolates. Expression of mma4 led to a complete absence of α1 and α2 mycolates, an altered surface morphology and the formation of two polar mycolic acids, both containing a distal hydroxyl residue with a methyl branch on the adjacent carbon atom. MMAS-3 appears to O-methylate the nascent hydroxyl group to form the methoxy functional group. MMAS-2 apparently possessed a proximal cis-cyclopropanation activity . However, analysis of the mycolic acids from a strain of M. tuberculosis bearing a chromosomal deletion of mma2 suggested MMAS-2 as a distal cyclopropane synthase for the α-meromycolyl class, a role previously ascribed to CMAS-1 . The production of a cma1 inactivated mutant of M. tuberculosis revealed no discernible role for CMAS-1 in meromycolate synthesis. MMAS-2 does appear to play a role in methoxymycolate synthesis, however, as the null mutant accumulates unsaturated derivatives while still synthesising intact methoxymycolate. NMR analysis revealed a 2-fold reduction in cis-cyclopropanated methoxymycolates suggesting MMAS-2 also acts as the preferred proximal cis-cyclopropane synthase in methoxymycolate synthesis but can be functionally replaced, to an extent, by another enzyme such as CMAS-2. Recently the crystal structures of PcaA, CMAS-1 and CMAS-2 have been elucidated. All three methyltransferases possess a similar fold with conserved SAM and mycolic acid binding sites, the latter in the form of a 15 Å long by 10 Å wide tunnel lined exclusively by hydrophobic amino acid residues. It has been suggested that structural differences (a conformational change in a single α helix) in the region around the entrance to this tunnel may play an important role in defining the specificity of these enzymes in terms of distal and proximal modification of meromycolates .
Although the function of MMAS-1 was not obvious from studies in M. smegmatis, it became apparent when mma1 was overexpressed in M. tuberculosis. Analysis of the mycolic acids from the overproducing strain showed that the proportion of trans-olefin and trans-cyclopropanated mycolates was significantly increased suggesting that MMAS-1 regulated a branch point between cis- and trans-cyclopropane-containing oxygenated mycolates . In addition to the isomerisation of the proximal cis-olefin with the incorporation of an allylic methyl branch, the overexpression of mma1 appeared to somehow increase the keto- to methoxymycolate ratio. The permeability of the cell wall to chendeoxycholate was also increased and DSC studies demonstrated a decrease in the thermal stability of the cell wall, suggesting that the increased proportion of ketomycolates may exert a fluidising effect upon the permeability barrier .
The overexpression of mma3 in M. tuberculosis led to a total replacement of all ketomycolates with methoxymycolates. Although cell wall organisation seemed to be maintained overall, growth was impaired at temperatures below 37 °C and the cells were hypersensitive to ampicillin and rifampicin. Most interesting, however, was the effect of modulating the keto and methoxymycolate ratio in vivo. The abilities of mma3 over-producing M. tuberculosis cells both to enter and to replicate within macrophage-like THP-1 cells were severely diminished. Pulse labelling experiments were then used to demonstrate that when the bacterium was cultured in vitro it under-represented ketomycolates (5-fold) and over-represented methoxymycolates (2-fold) compared with their in vivo grown cultures.
However, a mutant strain of M. tuberculosis inactivated in mma4 was devoid of all oxygenated mycolates was able to grow in THP-1 cells at about the same rate as the parent strain H37Rv . In addition, this mutant was significantly attenuated in a mouse model of infection, growing more slowly in the lungs, spleen and liver and being cleared more rapidly from the liver than H37Rv. The attenuated phenotype was fully recoverable when mma4 was supplied on a multicopy vector . Consistent with the loss of the expected fluidising effect of the oxygen functions, the cells showed decreased permeability to both hydrophilic and hydrophobic substrates as well as resistance to hydrogen peroxide. The authors suggest a nutrient deprivation phenotype caused by this reduced permeability, pointing out that the mutant strain eventually reaches and maintains a steady-state level in the mouse lung comparable to that seen with the parent organism . Further analysis of the mycolic acids elaborated by this mutant showed that it accumulated mostly α-mycolates with about equal quantities of dicyclopropanated (α1) and distally monounsaturated, mono-cyclopropanated (α2) mycolic acids. Very small amounts of cis-epoxymycolates related to α2 mycolates were also found. When the gene was expressed in M. smegmatis the levels of diunsaturated mycolates in the bacterium was lowered, being replaced by hydroxy and ketomycolates. Together these data support the hypothesis that mycolic acid substitutions are initially affected via the double bonds, which are then processed to the cyclopropanyl, and oxygenated functions found in M. tuberculosis.
In summary, the mycolate profile of the M. tuberculosis appears to be highly responsive to its growth environment and proximal substitution in particular modulates the fluidity of the cell wall permeability barrier. The mycolic acid repertoire of the cell wall is also strongly implicated in the pathogenicity of the organism with the complement of oxygenated mycolates influencing intracellular growth rate and bacillary clearance in vivo.
Claisen-like condensation of meromycolate and α-alkyl chains
The mechanism and enzymology of the Claisen-like condensation required to form both mycolates and corynomycolates remains enigmatic but probably proceeds via a common route in both mycobacteria and corynebacteria. Two mechanisms are possible; one involving the decarboxylation of an acyl malonate as seen with KasAB and FabH, and a non-decarboxylating mechanism used by various biosynthetic thiolases which is, in essence, the reverse reaction to the β-oxidative degradation of fatty acids . A non-decarboxylating mechanism has been suggested for the mycolic condensation forming corynomycolates based on the retention of 2H label in the corynomycolates of C. matruchotti supplied with exogenous 2,2-[2H]palmitic acid. Both deuterium atoms were retained in the β-chain at position C-4 of the resulting corynomycolic acids and one was retained at position C-2 which might be expected to be lost in a carboxylating scheme (Fig. 2) . The data presented does not preclude a decarboxylating route to mycolic condensation as the deuterium retained at C-2 might not leave the enzyme active site. There is, however, some evidence that the mycolic condensation leading to corynomycolate synthesis from palmitate in C. diphtheriae is sensitive to the antibiotic cerulenin , an inhibitor of β-ketoacyl-ACP synthases . The association of a cerulenin-sensitive condensing enzyme similar to β-ketoacyl-ACP synthases would, therefore, suggest a decarboxylating mechanism. The corynebacteria represent the least complex system in which to define the enzymology of the mycolic condensation reaction, the lack of an ACP suggests that acyl-CoA thioesters will contribute both branches of the corynomycolates and, importantly, acyl CoAs of appropriate sizes are commercially available. The synthesis and handling of appropriate meromycolate-thioester mimics may prove difficult in the mycobacterial system.
Mycolate export and deposition in the cell wall
Almost a decade ago a novel 3-oxo-2-tetradecyloctadecanoate (dehydrocorynomycolate)-containing phospholipid was discovered in C. diphtheriae and proposed as a carrier molecule involved in the mycolic condensation reaction . Arguably the most interesting mycobacterial Pol-P carrier conjugate described to date is Myc-PL, a 6-O-mycolyl-β-d-mannopyranosyl-1-monophosphoheptaprenol from M. smegmatis. The introduction of [14C]Myc-PL into a cell wall preparation from M. smegmatis led to the transfer of [14C]mycolates into extractable TMM and TDM as well as insoluble mycolates esterified to AG. The authors also reported a similar mycolate-containing phospholipid in M. tuberculosis and hypothesised that the Myc-PL of the mycobacteria are the reduced products of the mycolic condensation and the substrate for the export and deposition of mycolates in the cell wall . However, it is possible that the mature mycolate is transferred to the mannosylated carrier lipid only after mycolic condensation is complete. As these conjugates are extremely unstable even at −20 °C, the isolation of any acyl-mannosyl-PL or acyl-malonate-mannosyl-PL substrates that may aid in the resolution of this question may prove difficult. The fact that condensation reactions involving thioesters proceed at a higher rate to an equilibrium more strongly favouring product formation compared to the condensation of esters may suggest that transfer to the carrier lipid occurs after maturation of the mycolate moiety.
Our understanding of the enzymology of mycolyl transfer is now advanced in both mycobacteria and corynebacteria. Early studies by Takayama et al. [185, 186] showed that treatment of M. smegmatis with ethambutol, which is now known to inhibit cell wall arabinan polymerisation, led to the accumulation of TMM and TDM. A mycolyltransferase capable of the exchange of mycolyl residues between mycolyl-trehalose and the free disaccharide was later purified from M. smegmatis and a role in mycolyl-AGP deposition proceeding via an acyl-enzyme intermediate was postulated . Belisle et al.  demonstrated that three members of the M. tuberculosis antigen 85 complex, Ag85A, Ag85B and Ag85C2 (encoded by fbpA, fbpB and fbpC2, respectively) were able to catalyse mycolyltransferase reactions. This study also determined that 6-azido-6-deoxytrehalose could inhibit mycolyltransferase activity in vitro and also the growth of M. aurum, an established surrogate for M. tuberculosis in drug-susceptibility trials. These enzymes contain a carboxylesterase domain bearing the highly-conserved consensus sequence GXXSXXG. The introduction of a Ser → Ala substitution in this motif resulted in the complete loss of mycolyltransferase activity. Although these enzymes have a broad specificity and will catalyse the transfer of palmitate residues to trehalose, this reaction is some 40-fold slower than that involving mycolic acids (G.S. Besra, unpublished results). The crystal structures of FbpC2 and FbpB have recently been determined [189, 190] and revealed the conserved Ser residue as part of a catalytic triad completed by Glu and His residues. A hydrophobic pocket and tunnel corresponding to a probable TMM binding site and a trehalose binding site were also demonstrated. A gene encoding a probable fourth member of the antigen 85 complex, fbpC1 has been recognised. Sequence analysis revealed that three catalytic triad residues had all been replaced and, consistent with these substitutions, the purified recombinant protein did not possess a mycolyltransferase activity . This family of antigens also interact with human fibronectin thereby enhancing the uptake of mycobacteria by human macrophages via complement-mediated phagocytosis . However FbpC1 also lacks the proposed fibronectin-binding region  and its biological function remains unclear.
In order to shed light upon this apparent redundancy in mycolyltransferase and to ascertain the biological roles of the individual enzymes fbpC2, fbpA and fbpB have all been disrupted [194, 195]. The disruption of fbpC2 in M. tuberculosis by transposon mutagenesis led to the transfer of 40% fewer mycolates to the cell wall. There was no change in the types of mycolates esterified to AG or occurring as free glycolipids thus demonstrating that FbpC2 is involved directly or indirectly in the transfer of mycolates onto the cell wall of the whole bacterium and that the enzyme is probably not specific for a given type of mycolate, or at least the remaining mycolyltransferases are able to maintain the balance between the mycolate types through their own broad specificity . Homologous recombination was used to achieve the disruption of fbpA and fbpB in M. tuberculosis H37Rv. Although the fbpA mutant grew as well as the parent strain in laboratory media and macrophage-like cell lines, the fbpB mutant only grew well in laboratory media. In macrophage-like cell lines the strain grew very poorly, if at all [195, 196].
Corynebacteria possess genes with significant homology to those encoding the antigen 85 complex . Disruption of csp1 encoding the secreted Fbp-like PS1 of C. glutamicum lead to a 50% decrease in the amount of cell wall-linked corynomycolates and an alteration in the cell wall permeability . The accumulation of TMCM and a concomitant decrease in TDCM was observed in this strain , but not in the fbpC2 mutant of M. tuberculosis. The expression of fbpA, fbpB and fbpC2 from M. tuberculosis in this csp1-deficient strain restored the cell wall-linked mycolate content and the outer permeability barrier of the mutant, although the original balance between TMCM and TDCM was not rectified. It was suggested that the function of these three enzymes was to transfer mycolyl residues to AG but, given the structural differences between the corynomycolates and their mycobacterial counterparts, it was important to determine the role of each directly in the various fbp mutants of the parent bacterium. Puech et al.  analysed the fbpA and fbpB mutants of M. tuberculosis biochemically revealing that both produced comparable mycolate profiles to that of the parent strain. The persistence of the mycolyl-glycolipids in these mutants demonstrates that these Fbps are not essential individually to the production of the trehalose conjugates. As no differences were found in the cell wall-linked mycolate contents of the mutants and the parent strain it was clear that either the absence of FbpA or FbpB individually could be compensated for by the remaining cellular Fbp complement, or that FbpA and FbpB do not play a role in mycolyl-AG esterification in the intact bacterium. The ability for each Fbp to restore the AG-linked mycolylation of the fbpC mutant strain expression was assessed. The individual over-expression of all three fbp genes led to some recovery of wild-type AG-mycolylation levels. As expected over-expression of fbpC led to the most marked recovery (86% wild type from a 60% background) with fbpB (80%) proving more effective than fbpA (70%). Permeability studies and linkage analysis confirmed that both FbpA and FbpB are able to transfer mycolates to AG and that they display no preference for mycolyltransfer to terminal or 2-linked Araf residues to the terminal hexaarabinosyl motifs of AG . It appears then that M. tuberculosis exhibits a partial redundancy in mycolyltransfer with the role of FbpC being only partly fulfilled by the others.
In C. diphtheriae two genes encoding potential mycolyltransferases, csp1 and DIP2194) are present. As in M. tuberculosis, these genes occupy two adjacent loci downstream of glfT and three conserved genes encoding membrane proteins of unknown function. In M. tuberculosis fbpA lies next to the gene encoding FbpC1, which lacks the catalytic triad found in other members of the antigen 85 complex. Sequence alignments of the M. tuberculosis and C. diphtheriae mycolyltransferases suggest that both C. diphtheriae genes encode this catalytic triad and, in agreement with the 50% AG-mycolylation activity of the csp1-deficient C. glutamicum strain , that both gene products may possess corynomycolyltransferase activity. Presumably redundancy in mycolyltransferase activity is a common theme among the mycolata.
Genetic organisation of mAGP biosynthesis/assembly
Belanger and Inamine suggested the possibility of a large AG biosynthetic gene cluster comprising 31 genes and stretching from Rv3779-Rv3809c covering 48.5 kb of the chromosome (Fig. 6) . Among these genes are glf and glfT (galactan polymerisation), embCAB (arabinan deposition), fbpA (mycolyltransfer). Comparison of this region with the equivalent from C. diphtheriae shows that part of the arrangement is well conserved but with two major differences, the insertion of a probable glycine-betaine synthesis cassette between the glf and glfT homologues and the whole conserved region is split into two discontinuous segments resulting in the emb homologue of C. diphtheriae lying over 460 kB away from the glfT homologue. One cluster consists of homologues of embC-Rv3789 (DIP0159-DIP0166), which includes a three-gene insertion (Fig. 6). The other cluster encodes orthologues of Glf, GlfT, Rv3805-Rv3807, mycolyltransferases and three genes which appear to encode a polyketide synthase module devoted to the decarboxylating condensation of fatty acids. The latter genes appear to encode a fatty acid CoA ligase, a non-biotinylated CoA carboxylase subunit and a polyketide synthase (Pks13) bearing β-ketoacyl-ACP synthase and thioesterase domains. As this is the only candidate for a polyketide synthase evident in the genome of this C. diphtheriae strain, and its genetic context seems to suggest an involvement in cell wall biosynthesis it represents an attractive candidate for the mycolic condensation enzyme.
The last decade or so has seen considerable progress towards the enzymological definition of mycobacterial cell wall biosynthesis. Much of this progress has been aided by the use of the annotated genome sequence of M. tuberculosis H37Rv and its comparison with appropriate sequences of known function from other bacteria allowing the identification of mycobacterial genes encoding analogous functions. The complex cell walls of the mycolata carry several unique structures that may only be genetically and enzymologically defined by assessing the conservation of genes of ‘unknown’ function within their genomes. The proteins directing the assembly of the cell wall components are also likely to fall into this category. Here we have attempted to define the similarities and differences in cell wall structure and biosynthesis and have tentatively speculated upon the nature of some of the processes and genes involved in the as yet undefined stages of the process. The shrewd application of the burgeoning portfolio of new molecular techniques for mycobacteria combined with careful biochemical analysis of mutants and isolated gene products promises to make significant additions to our knowledge over the coming decade. Perhaps some of these studies will reveal a ‘chink in the armour’ of this resilient pathogen and lead to an effective new chemotherapeutic agent with which to combat tuberculosis.
The authors would like to acknowledge research support provided by the Lister Institute for Preventative Medicine, The Wellcome Trust and the Medical Research Council.