By using CEMOVIS, it is possible to visualize a “granular layer” in between the plasma membrane and the peptidoglycan layer of Mycobacterium bovis BCG, Mycobacterium smegmatis, and Corynebacterium glutamicum (Zuber et al., 2008). A previous study made on other Gram-positive bacteria has shown that the granular layer is possibly linked to the plasma membrane and composed of penicillin-binding proteins, lipoproteins, and lipoteichoic acids (or teichuronic acid in some species; for review see Tul’skaya et al., 2011) (Zuber et al., 2006). Although, it seems this structure is common to several Gram-positive bacteria, the function and its precise composition are poorly characterized.

The cell envelope of Actinobacteria is composed of a layer of peptidoglycan that provides essential functions such as rigidity and helps to maintain an optimal osmotic stability (Vollmer et al., 2008; Jankute et al., 2015). Although peptidoglycan is common amongst bacteria, there are many subtle differences in its composition that have been used in the past as a way to identify distinct species in the context of chemotaxonomy (Lechevalier and Lechevalier, 1970). The standard peptidoglycan consists of short peptides and glycan strands that are composed of N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc) residues linked byβ-1→4 bonds. The third amino acid in the peptide stem is usually meso-diaminopimelic acid (meso-DAP) for Gram-negative and L-lysine for Gram-positive bacteria (Egan et al., 2017). In Actinobacteria, this amino acid is variable with numerous species (including the Corynebacteriales order) that harbor meso-DAP. This property is potentially due to the species-specific affinity of the ligase MurE (UDP-N-acetylmuramoylalanyl-D-glutamate-2,6-diaminopimelate ligase) for meso-DAP (Hammes et al., 1977; Basavannacharya et al., 2010). Furthermore, differences in these third amino-acids were used as a marker for Actinobacteria cell wall types II, III, and IV (Lechevalier and Lechevalier, 1970).

In addition to having meso-DAP, the peptidoglycan of members of the Corynebacteriales order has another major distinction: a portion of MurNAc molecules is oxidized to become N-glycolylmuramic acid (MurNGlyc). This characteristic, in addition to playing an important role in the structure of peptidoglycan, has been suggested to increase resistance to lysozyme and β-lactam antibiotics (Raymond et al., 2005). The hydroxylase responsible for this modification is encoded by the gene namH (Rv3818; as per the reference M. tuberculosis H37Rv genome) (Raymond et al., 2005; Coulombe et al., 2009; Hansen et al., 2014). As expected, this gene is found in several members of the Corynebacteriales order (Figure 2). However, it appears to be absent from the genomes of genera such as Hoyosella and in those that are the most “basal” of the order: Corynebacterium, Turicella, Lawsonella, Dietzia, and Segniliparus. Interestingly, a few other genera in the Actinobacteria phylum, but not in the Corynebacteriales, also have a putative namH homolog (Figure 2). This is the case for Stackebrandtia (Glycomycetales), Micromonospora (Micromonosporales), and Salinispora (Micromonosporales) that form a discrete sub-lineage and have been described to harbor a cell wall with MurNGlyc (Parte et al., 2012). As an exception, a few scattered species (such as Catenulispora acidiphila and Rubrobacter radiotolerans) possess a homologous copy of NamH, although to the best of our knowledge there is no evidence for the presence of MurNGlyc in their peptidoglycan (Parte et al., 2012).

In addition to its distinct composition, the peptidoglycan of Corynebacteriales is also the point of attachment of arabinogalactan, a highly branched heteropolysaccharide composed of galactose and arabinose in a furanoid form (Figure 1; McNeil et al., 1987). It is estimated that approximately 10% of the MurNAc residues of the peptidoglycan are covalently linked to arabinogalactan [_α-_L-rhamnopyranose-(1→3-α-D-GlycNAc-(1→P))] (McNeil et al., 1990). Although this linkage is unique, it resembles the one that connects cell wall teichoic acids to peptidoglycan in other Gram-positive species. This similarity led to the discovery of a phosphotransferase of the LytR-CpsA-Psr family, named Lcp1, which is encoded by the Rv3267 gene and plays a leading role in linking arabinogalactan to peptidoglycan in M. tuberculosis (Baumgart et al., 2016; Grzegorzewicz et al., 2016; Harrison et al., 2016).

In terms of its distribution within the Actinobacteria phylum, arabinogalactan is produced exclusively by the members of the Corynebacteriales order and in some species from related orders such as Pseudonocardiales and Actinopolysporales (Goodfellow et al., 1984) [with the exception of Actinosynnema mirum (Hasegawa et al., 1978)]. These latter two orders possess a typical type IV cell wall (Lechevalier and Lechevalier, 1970), although they lack mycolic acids (Ludwig et al., 2015). The global distribution of genes implicated in the synthesis and transport of arabinogalactan is well correlated with the fact that these species produce arabinogalactan (see Figure 2). Nevertheless, a deeper investigation of these latter two orders will be required to exclude possible exceptions or misinterpretation.

The inner leaflet of the external mycomembrane is homogeneous and is mainly composed of mycolic acids (MAs) – long-chain fatty acids that are exclusive to the order Corynebacteriales (Bansal-Mutalik and Nikaido, 2014; Marrakchi et al., 2014). On one hand, MAs form a barrier to hydrophilic molecules including several antibiotics (Liu et al., 1995), while on the other hand, they are also the target of some first- and second-line anti-TB drugs including isoniazid (Jackson et al., 2013; Marrakchi et al., 2014). In an evolutionary context, MAs play a prominent role in taxonomic differentiation within the Corynebacteriales order. In this sense, there is a great diversity in the number of carbons and functional groups comprising the MAs, thus making it possible to use them as a chemotaxonomic marker to differentiate between genera and also at the species level (Minnikin et al., 1984; Barry et al., 1998; Song et al., 2009; Marrakchi et al., 2014).

As reviewed elsewhere (Marrakchi et al., 2014), the raw materials of MAs, fatty acids, are produced by the combination of two fatty acid synthases (FAS-I and FAS-II). The first, FAS-I, is a eukaryotic-like protein narrowly distributed in bacteria (Cabruja et al., 2017) and encoded by the fas gene (Rv2524c). FAS-I produces acyl-CoAs with a bimodal distribution of C₁₆-C₁₈ and C₂₄-C₂₆. FAS-II, on the other hand, is far more common across bacterial species (Cabruja et al., 2017) and begins with the formation of β-ketoacyl-ACP by Claisen condensation of malonyl-ACP [produced from malonyl-CoA by MtFabD (Rv2243)], with acyl-CoA (the product of FAS-I), through the action of MtFabH (Rv0533). Other enzymes such as HadA (Rv0635), HadB (Rv0636), HadC (Rv0637), InhA (Rv1484), KasA (Rv2245), and KasB (Rv2246) are involved in the subsequent steps of elongation and maturation of fatty acids by FAS-II. The final MA structure is produced though several crucial steps that include, among others, activation (FadD32 and Rv3801c), condensation (Pks13 and Rv3800c) and reduction (CmrA and Rv2509).

Although the presence of MAs is well-conserved throughout the Corynebacteriales, there are some exceptions. For example, basal species of Corynebacteriales (such as Corynebacterium, Dietzia, Lawsonella, and Turicella) that form a monophyletic group, lack multiple genes confirming the well described diversity in length and composition of MA. It has already been reported that some species of Corynebacterium possess two FAS-I genes (such as C. glutamicum), while some do not have the genes encoding for the typical FAS-II machinery (Radmacher et al., 2005). Interestingly, Corynebacterium cannot elongate MAs (Radmacher et al., 2005; Burkovski, 2013), which is consistent with the lack of more than 50% of genes implicated in this pathway.

Turicella otitidis possesses almost none of the genes involved in MA biosynthesis (see “^∗” in Figure 2). This species is clearly in the order Corynebacteriales based on its genome sequence (Baek et al., 2018). However, a recent study suggests that the genome of T. otitidis has lost the key genes involved in MA synthesis (Baek et al., 2018). This result corroborates the fact that this bacterium does not produce MAs which is exceptional for a bacterium of the order Corynebacteriales and this microevolution has made its positioning within this order rather difficult (Funke et al., 1994). Interestingly, this species still produces arabinogalactan (Renaud et al., 1996).

While the composition of the inner leaflet of the mycomembrane is homogenous, the outer leaflet is highly heterogeneous and consists of lipids, lipoglycans, and proteins (Chiaradia et al., 2017). Due to its complexity and species-specific constitution, it remains poorly characterized with the exception of the pathogen M. tuberculosis and closely related slow-growing mycobacteria. In these bacteria, several lipids are associated with the external leaflet including – but not limited to – phthiocerol dimycocerosate (PDIM), phenolic glycolipid (PGL, phenolphthiocerol-based glycolipids that share a similar long-chain fatty acid backbone with PDIM), and lipooligosaccharides (LOS). These components seem to be important virulence factors (Reed et al., 2004; Forrellad et al., 2013) and have been shown to be involved in the infection of macrophages as well as in escape from the immune system (Cambier et al., 2014). The PDIM and PGL-associated genes (Rv2928 to Rv2963), are all co-located within the same cluster of the chromosome (Goude and Parish, 2008; Supplementary Table S2). LOS are common to several slow growing mycobacteria (including “M. canettii” from the M. tuberculosis complex) although they are not produced by M. tuberculosis due to the loss of Pks5.1 and truncation of PapA4. The remaining genes implicated in this pathway are present (Figure 2) which suggests micro-evolution in this species has led to the loss of LOS production (Boritsch et al., 2016; Brennan, 2016). This apparent evolutionary change correlates with a marked increase in whole cell hydrophobicity and enhanced aerosol transmission of current TB strains (Jankute et al., 2017).

The external layer of the mycomembrane also contains trehalose monomycolate – TMM, and trehalose dimycolate – TDM (also called “cord factor”) that also have crucial functions in the regulation of the host-symbiont relationship (Hunter et al., 2006). These molecules consist of glucose disaccharides (α-D-glucopyranosyl-α-D-glucopyranoside) esterified with MAs. TMM is generated in the cytoplasm, whereupon MmpL3 (Rv0206c) subsequently transports this molecule to the periplasm (Grzegorzewicz et al., 2016). Finally, the accepted model (for a review see Marrakchi et al., 2014) is that TMM in the periplasm serves as the MA donor for the mycolylation of arabinogalactan, and is also processed to give TDM. Both reactions involve the Ag85 enzyme complex (Belisle et al., 1997). Three genes (Rv3803c, Rv1886c, and Rv0129c) encode for the Ag85 complex and they are only present in the MA-positive species as seen in Figure 2. As TMM and TDM rely on the presence of MAs and Ag85, their distribution is therefore restricted to MA positive species. Indeed, they have been isolated from Mycobacterium, Corynebacterium, Nocardia, and Rhodococcus (Noll et al., 1956; Ioneda et al., 1970; Jean-Claude et al., 1976; Mompon et al., 1978; Lopes Silva et al., 1979; Pommier and Michel, 1979; Rapp et al., 1979; Batrakov et al., 1981). However, further characterization is necessary for other MA-positive species (such as Segniliparus).