The mycobacterial cell envelope consists of three main structural components: (a) a network of peptidoglycan (PG), (b) the arabinogalactan (AG) polysaccharide, and (c) the long-chain mycolic acids (MA) (Alderwick et al., 2015; Daffé, 2015; Jankute et al., 2015). This cell wall-based permeability barrier shields the organism from environmental stress and contributes to resistance of mycobacteria to many antibiotics. Thus, defects in these lipids would damage the function of the cell wall as a barrier and increase the sensitivity to various anti-mycobacterial drugs (Jackson et al., 2013). The enzymes involved in cell wall integrity play an important role in the development of drug resistance in Mycobacterium (Jackson et al., 2013; Xu et al., 2014). Table 1 shows some important enzymes require for integrity of mycobacterial cell wall.

MurA and MurB are the key biosynthetic enzymes involved in the formation of UDP-MurNAc, an important step in biosynthesis of PG (Alderwick et al., 2015). The naturally occurring broad-spectrum antibiotic fosfomycin is a well-known inhibitor of MurA (Alderwick et al., 2015; Moraes et al., 2015). It specifically inhibits MurA by forming a covalent adduct with a cysteine residue in the active site (Moraes et al., 2015). However, MTB are intrinsically resistant to fosfomycin, as their corresponding cysteine residue is changed into aspartic acid (Alderwick et al., 2015). Conversely, mutation of the wild-type aspartate residue in the MurA active site to a cysteine renders an enzyme sensitive to fosfomycin in MTB (Castañeda-García et al., 2013).

In most bacteria, β-lactam antibiotics inactivate the essential D,D-transpeptidase activity of classical penicillin-binding proteins (PBPs). These enzymes cross-link glycan chains by forming 4 → 3 peptide bonds connecting residues at the fourth and third positions of stem peptides (Gupta R. et al., 2010). However, MTB has been recently found to have a second class of transpeptidases, the L,D-transpeptidases (Ldt) (Gupta R. et al., 2010; Schoonmaker et al., 2014; Basta et al., 2015; Kieser et al., 2015), which are responsible for resistance to β-lactam antibiotics such as amoxicillin and carbapenems (Gupta R. et al., 2010; Alderwick et al., 2015). Till date, five Ldts (Ldt_Mt1 to Ldt_Mt5) enzymes which form the non-classical 3 → 3 linkages between opposing stem peptides in PG have been identified in MTB (Schoonmaker et al., 2014). MTB strain that lacks both ldt_Mt1 and ldt_Mt2 displays enhanced susceptibility to amoxicillin and a glycopeptide drug, vancomycin (Schoonmaker et al., 2014).

Moreover, two proteins, PonA1 and PonA2, are bifunctional penicillin-binding proteins that contribute to the biosynthesis and homeostasis of the cell wall components (Jankute et al., 2015). Recent studies indicated that ponA1 mutant MTB had the same minimum inhibitory concentration (MIC) for b-lactams as wild-type MTB; however, ponA2 mutant cells were four- to eightfold more susceptible to b-lactams (Shi et al., 2011, 2014; Farhat et al., 2013).

Likewise, any defects in enzymes and proteins that are involved in the cell wall integrity can result in increased susceptibility to multiple drugs. Among the various enzymes in the cell envelope biogenesis, one can find the well-known proteins of the antigen 85 (Ag85) complex (Jankute et al., 2015). These proteins, which show mycolyltransferase activity, are involved in the biogenesis of trehalose dimycolates (TDM), a dominant structure necessary for maintaining cell wall integrity (Jankute et al., 2015). It was shown that inactivation of the Ag85 gene profoundly affects the mycolate content and alters the permeability of the MTB cell envelope (Jackson et al., 1999; Ronning et al., 2000). This change in the permeability of the cell envelope can also result in changes in drug resistance phenotype (Jackson et al., 1999; Nguyen et al., 2005). For instance, individual knockout studies on the Ag85 genes (fbpA, fbpB, and fbpC) showed that the mutant strains had 40% less mycolate moieties on the cell wall (Torres et al., 2015; Zhang et al., 2017). Nguyen et al. also indicated that the Ag85 mutants display increased sensitivity to both first-line TB drugs as well as other broad-spectrum antibiotics widely used for antibacterial chemotherapy. In their study, they showed that mutant strains of Mycobacterium smegmatis were 24-, 85-, 8-, and 12-fold more susceptible to erythromycin, imipenem, RIF, and vancomycin, respectively (Lingaraju et al., 2016). Furthermore, recent studies have shown that the production of TDM by Ag85 is essential for the intrinsic antibiotic resistance of some mycobacteria and support the concept that Ag85-specific inhibitors, alone or in combination with other antibiotics, could provide an effective treatment for TB and other mycobacterial diseases (Nguyen et al., 2005).

Examples of other important proteins include: GlmU, MurX, Alr, Ddl, RmlC, accD6, and Pks12 that were reported to be involved in cell wall integrity of mycobacteria (Table 1; Mills et al., 2004; Philalay et al., 2004; Farhat et al., 2013; Calvanese et al., 2014; Schoonmaker et al., 2014; Alderwick et al., 2015; Jankute et al., 2015). These enzymes were shown to be essential and were attractive targets for the development of new antibacterial agents (Alderwick et al., 2015).

Three terms, latency, dormancy and persistence, will be defied here as follow:

Latency refers to a state of asymptomatic infection without apparent disease. Dormancy refers to a non-replicating state with low-to-absent metabolic activity. A subpopulation of dormant cells may be responsible for latency and persistence despite host defense and drug treatment. Persisters, are bacteria that survive antibiotic treatment by remaining dormant even though they are genetically susceptible to these antibiotics (Gomez and McKinney, 2004; Chao and Rubin, 2010).

MTB enters a dormant state during latent infection which is characterized by a shutdown of most of its metabolism causing increased tolerance to antimicrobial agents that are lethal to replicating bacilli a phenomenon referred to as phenotypic drug resistance (Gomez and McKinney, 2004; Gengenbacher and Kaufmann, 2012). In contrast to genotypic drug resistance, this form of drug tolerance is due to reduced metabolic activities or cell division rather than chromosomal resistance mutations (Garton et al., 2008). This is reasoned by the fact that the low metabolic activity in slowly growing or non-replicating dormant MTB is associated with reduced production of antibiotic target proteins or machineries by the bacteria. In fact these bacteria were found to be resensitized upon reentering active growth phase due to resumed production of antibiotic cellular targets (Smith et al., 2012). It's been demonstrated that, in vitro induction of dormant state in MTB by oxygen deprivation is associated with shutdown of DNA, RNA and protein synthesis and subsequently increased resistance to anti-TB drugs such as INH and RIF but susceptibility to the anaerobic bactericidal action of metronidazole, which has no effect on aerobic cultures of MTB (Wayne and Hayes, 1996; Hu et al., 1998).

Another system has used nutrient starvation for induced transition of MTB from active growth into a dormant state and revealed decreased respiration and transcription rate, energy metabolism, lipid biosynthesis, cell division as well as concomitant increased drug tolerance to antibiotics targeting functions required for active growth (Betts et al., 2002; Xie et al., 2005). Therefore, the ability of MTB to enter a metabolically inactive dormant phase can be classified as an intrinsic drug tolerance mechanism. Identifying the factors inducing the dormant state and promoting long term survival or persistence of latent MTB might provide novel targets for development of compounds that can treat TB infections by inhibition of these persistence factors. The Ldts enzymes are among the most important factors that plays a vital role in MTB adaptation to stationary phase (Lavollay et al., 2008). As mentioned above, PG of MTB contains a complex network of classical (4 → 3) and non-classical (3 → 3) transpeptide bonds. However, recent studies have shown that upon MTB entry in to the dormancy phase the number of 3 → 3 cross-linkages significantly increases through PG rearrangement by Ldts (Lavollay et al., 2008; Jankute et al., 2015). The rearrangement of PG may have at least two selective advantages. First, Ldts are the only enzymes able to catalyze the formation of new cross-links in the absence of de novo synthesis of precursors (Lavollay et al., 2008). Second, modification of the cross-links may render the peptidoglycan resistant to the hydrolytic activity of endopeptidases (Lavollay et al., 2008). Thus, the development of Ltd inhibitors is a novel and promising approach to obtain drugs for treatment of TB.

Several MTB protein chaperones may also possibly function as a stress responder and consequently contribute to persistence of MTB (Vaubourgeix et al., 2015; Lupoli et al., 2016; Botella et al., 2017). Many of the stresses that MTB encounters, such as reactive oxygen species, changes in pH and mistranslation caused by antibiotics, can cause reversible or irreversible damage to bacterial proteins (Lupoli et al., 2016). In stationary phase, exposure to an aminoglycoside antibiotic like streptomycin, corrupts ribosomal fidelity and induces mistranslation (Dukan et al., 2000; Ballesteros et al., 2001; Vaubourgeix et al., 2015). Recent works suggest that ClpB/DnaK chaperones play unique roles in mycobacteria (Vaubourgeix et al., 2015; Lupoli et al., 2016). These ATP-powered chaperones have been shown to unfold and refold aggregated proteins to their native structures to restore the function of protein (Lupoli et al., 2016). Recent study indicated that ClpB-deficient MTB had a marked recovery defect from stationary phase or antibiotic exposure and survived poorly in mice (Vaubourgeix et al., 2015). In this regard, ClpB and DnaK chaperones in mycobacteria may serve as a novel class of targets for the development of drugs that can sensitize MTB to other drugs (Vaubourgeix et al., 2015).

Another important mechanism which allows MTB to temporarily survive lethal concentration of antibiotics is stochastic variation in the expression of resistance genes. During this phenomenon, MTB persister cells stochastically enter a dormant state, allowing the population to hedge against the appearance of an antibiotic. It is important to know that transient resistance is not caused by genetic changes, rather MTB induces gene expression to generate a resistant phenotype (Wakamoto et al., 2013; El Meouche et al., 2016). For example, stochastic expression of genes affecting antimicrobial action such as catalase-peroxidase katG, which activates the prodrug INH accounts for differential antibiotic susceptibility of MTB to INH in a subset of cells (El Meouche et al., 2016; Wakamoto et al., 2013).

The permeability of lipid membranes of mycobacteria for hydrophilic solutes is extremely low compared to Gram-negative bacteria. Therefore, hydrophilic agents often utilize channel proteins, such as porins, to cross the outer membrane (Faller et al., 2004; Mailaender et al., 2004; Danilchanka et al., 2008). Porin-like proteins can be detected by channel-forming activities in detergent extracts of MTB and M. smegmatis (Heinz and Niederweis, 2000; Raynaud et al., 2002; Niederweis, 2003; Faller et al., 2004; Nguyen and Thompson, 2006; Danilchanka et al., 2008; Song et al., 2008; Niederweis et al., 2010). One of these proteins, MspA has been well characterized and constitutes the major porin of M. smegmatis (Niederweis et al., 1999). Deletion of the MspA, drastically increased the resistance of M. smegmatis to several agents, indicating that MspA plays an important role in the uptake of hydrophilic antibiotics (Stahl et al., 2001; Stephan et al., 2004). Heterologous expression of the M. smegmatis porin gene mspA in MTB increasesits susceptibility to hydrophilic antibiotics such as β-lactams as well as anti-TB drugs INH, EMB and streptomycin (Stephan et al., 2004). An understanding of the pathways that enable passage of these drugs through the envelope of MTB is essential to the successful design of new therapeutic agents.

The intrinsic resistance of mycobacterial species to most antibiotics is generally attributed to the low permeability of the mycobacterial cell wall (Brennan, 2003). Along with cell wall permeability, active efflux systems also provide resistance by expelling the drug molecules that enter the cell. Several mycobacterial drug efflux pumps have been identified and can be grouped into five different structural families: the major facilitator superfamily (MFS) (Siddiqi et al., 2004; Balganesh et al., 2012; Xu et al., 2015), the small multidrug resistance (SMR) family (De Rossi et al., 2002; Garima et al., 2015), the resistance-nodulation-cell division (RND) superfamily (De Rossi et al., 2006; Murakami et al., 2006), the adenosine triphosphate (ATP)-binding cassette (ABC) superfamily (Bhatt et al., 2000; Choudhuri et al., 2002; Pasca et al., 2004; Lubelski et al., 2007; Caleffi-Ferracioli et al., 2016) and the multidrug and toxic compound extrusion (MATE) family (Table 2) (Mishra and Daniels, 2013).

These efflux pumps actively transport many antibiotics out of the cell and are major contributors to the intrinsic resistance of mycobacteria to many drugs. Some efflux pumps have narrow substrate specificity (such as TetV pumps), but many transport a wide range of structurally dissimilar substrates and are known as MDR efflux pumps. The MFS family of MDR efflux pumps is found in many species of mycobacteria and includes several members that are relevant to antibiotic resistance (Yamchi et al., 2015). When overexpressed, MFS pumps confer clinically relevant levels of MDR and export an extremely wide range of substrates.

For example, the increased transcription of jefA, a well-studied example of MFS pumps in MTB, leads to increased resistance to INH, EMB and streptomycin (Gupta et al., 2010b). Likewise, the best-characterized efflux pumps in NTM include the tap in M. fortuitum conferring resistance to tetracycline and aminoglycoside or lfrA efflux pumps in M. smegmatis that mediate resistance to fluoroquinolones, RIF and INH (Liu et al., 1996; Takiff et al., 1996; Aínsa et al., 1998; Li et al., 2004; Ramón-García et al., 2006; Bowman and Ghosh, 2014). In clinical strains of MTB, increased expression of efflux pumps occurs mainly as a result of induction in response to antibiotic stress (Jiang et al., 2008; Gupta et al., 2010a; Li et al., 2015). For instance, studies that investigated the relevance of active efflux in the drug resistance of clinical strains of MTB indicated that jefA, drrA, drrB, efpA, mmr, and RV1217-Rv1218 efflux pumps were overexpressed under INH and RIF stress (Wang et al., 2013; Blair et al., 2015; Li et al., 2015). Furthermore, several studies indicated that some RIF or INH-resistant strains of MTB did not have sequence alterations in the core region of the drug target encoding genes (e.g., rpoB, katG, inhA; Yamchi et al., 2015; Manson et al., 2017). Consequently, efflux pumps might play an important role in RIF- and INH- resistance in MTB, especially in those strains having no mutation in genes associated with INH and RIF resistance. Similarly, elevated level of expression in efflux pump genes are also demonstrated to be involved in drug resistance development in NTM species (Rodrigues et al., 2009; van Ingen et al., 2012). Understanding the mechanisms controlling the overexpression is important as it is a common mechanism of resistance in clinically important mycobacteria, and also provides promising target for designing novel class of anti-mycobacterial therapeutics that can treat mycobacterial infections by efflux-pump prevention.

Since efflux pumps play a major role in the development of drug resistance in MTB, many studies have focused on identifying potential efflux pump inhibitors. For example, efflux pump inhibitors reserpine, verapamil, 2,4-dinitrophenol (DNP) and pyrrole derivatives have been shown to increase the susceptibility to norfloxacin, RIF and INH by inhibiting efflux pumps (i.e., Rv1258c and MmpL3), in drug-susceptible mycobacteria (Sharma et al., 2010; World Health Organization, 2011; La Rosa et al., 2012; Machado et al., 2012; Willers et al., 2017). Another efflux pump inhibitor is a mammalian efflux pump inhibitor, timcodar (Nasiri et al., 2016). An adjuvant effect was demonstrated in combination with RIF, bedaquiline and clofazimine when MTB was cultured in host macrophages (Nasiri et al., 2016). Recent studies also demonstrate the potential role of efflux pump inhibitors in treatment regimens to improve and shorten anti-TB treatment (World Health Organization, 2010, 2011).

Mycobacteria can encode several specific enzymes that have evolved to target and cleave antibiotics of different classes, including β-lactams, aminoglycosides and macrolides (Wang et al., 2006; Da Silva and Palomino, 2011; Wivagg et al., 2014). The β-lactam antibiotics are among the most important classes of antibacterial compounds and act by inhibiting the bacterial transpeptidases that are responsible for the final step of peptidoglycan cross-linking (Wright, 2005). In mycobacteria, β-lactam resistance is primarily due to the production of an Ambler class-A β-lactamase encoded by the blaC gene (Hugonnet and Blanchard, 2007). However, other factors such as low cell envelope permeability and presence of low penicillin-binding protein binding affinity for beta-lactams were also believed to contribute to the ineffectiveness of beta-lactams in mycobacteria (Wang et al., 2006; Jankute et al., 2015). Class A enzymes in MTB include the beta-lactamase encoded by the chromosomal blaC gene, which shares extensive similarity with many eubacterial b-lactamases (Wang et al., 2006; Nampoothiri et al., 2008). These enzymes are large proteins with much greater penicillinase activity than cephalosporinase activity (Wang et al., 2006). Like other Class A β-lactamases, BlaC hydrolysis the β- lactam ring via nucleophilic attack by an active site serine residue to inactive the drug (Tremblay et al., 2010). In addition, MTB BlaC displays relatively broad hydrolysis spectrum to all β-lactam classes, including the members of the carbapenem antibiotics, which are generally resistant to β-lactamases of other pathogenic bacteria (Hugonnet and Blanchard, 2007; Tremblay et al., 2010). Moreover, β-lactamase inhibitors such as clavulanic acid are less effective against BlaC compared to other class A enzymes (Hugonnet and Blanchard, 2007). Although mycobacterial beta-lactamases can hydrolysis carbapenems, reactions proceed slowly, and one member of the carbapenem family, meropenem, has been reported to be active in vitro in association with clavulanic acid against drug resistant strains (Hugonnet et al., 2009; Mainardi et al., 2011). Meropenem may also have potential applications for the treatment of susceptible TB, because meropenem–clavulanic acid is active against nonreplicating forms of bacilli, which are difficult to eradicate even with INH and RIF (Mainardi et al., 2011).

Ambler class A β-lactamases have also been described in M. fortuitum and M. smegmatis (Nguyen and Thompson, 2006; Soroka et al., 2014).

The addition of chemical groups to specific sites on the antibiotic by mycobacterial modifying enzymes causes antibiotic resistance by preventing the antibiotic from binding to its target protein.

Among mycobacterial species, aminoglycoside resistance is sometimes caused by modifying enzymes that are coded by genes on the chromosome (Zaunbrecher et al., 2009; Maurer et al., 2015). Resistance to aminoglycoside antibiotics is also conferred by target mutation, reduced uptake and/or increased efflux (Aínsa et al., 1998; Da Silva and Palomino, 2011). Till now, two main classes of aminoglycoside-modifying enzymes namely, acetyltransferase and phosphotransferase, have been demonstrated in several mycobacterial species (Zaunbrecher et al., 2009; Kim et al., 2011; Ahn, 2013). The best biochemically characterized aminoglycoside-modifying enzyme in mycobacteria is aminoglycoside N-acetyltransferase (AAC 2′) which is capable of acetylating all know aminoglycosides bearing 2′ amino group including neomycin, kanamycin, gentamycin, tobramycin and ribostamycin (Vetting et al., 2002). Distinct N-acetyltransferases have been identified in the genomes of the MTB, M. kansasii, the rapid growers M. fortuitum, M. smegmatis, and M. abscessus (Ho et al., 2000; Ripoll et al., 2009; Ramirez and Tolmasky, 2010; van Ingen et al., 2012).

Another discovery is the presence of a RIF-resistance gene in some mycobacterial species (Baysarowich et al., 2008). In the opportunistic pathogen M. smegmatis, RIF is an ineffective drug because of the presence of a chromosomally encoded RIF ADP-ribosyltransferase (Baysarowich et al., 2008). This specific enzyme, transfers the ADP-ribose unit to a hydroxyl residue at position 23 of RIF, rendering the bacterium resistant to RIFs (Baysarowich et al., 2008).