In 2010, it was estimated that there were 8.8 million incident cases of tuberculosis (TB), 1.1 million deaths from TB among HIV-negative people and an additional 0.35 million deaths from HIV-associated TB [1]. Despite the availability of effective treatment options since the 1950s, and the implementation of well-structured treatment programs, the current TB epidemic is not being controlled. Frontline anti-tuberculous drugs have gradually become ineffective because of the increasing incidence of resistance. Multidrug-resistant TB (MDR-TB) is a difficult-to-treat form of M. tuberculosis that fails to respond to the two most effective first-line anti-tuberculous drugs, rifampicin and isoniazid. The World Health Organization (WHO) estimated that in 2009, around 5% of all new tuberculosis cases involved MDR-TB [2]. Strains that combine MDR with additional resistance to fluoroquinolones and at least one injectable drug have been appropriately named extensively drug-resistant tuberculosis (XDR-TB). The burden of tuberculosis on global health has pushed the research community into focusing efforts on the development of new vaccines, diagnostics and chemotherapy against Mycobacterium tuberculosis, the causative agent.

The TB pathology is diverse, generating different types of lesions, containing several micro-environments each harboring metabolically distinct bacterial sub-populations, some of which are not effectively killed by most existing drugs [3]. This drug tolerance phenomenon typical of tuberculosis has been coined “phenotypic drug resistance” [4], and is partly attributed to the pathogen’s ability to remain sequestered in macrophages and other stress-inducing micro-environments in a non-replicating state of persistence (see Figure 1) [5]. These dormant bacilli are primarily responsible for the persistent and latent forms of the disease, but retain the potential to resume growth and produce an active infection, making them a critical target population of antimycobacterial agents [5,6,7].

The development of new antimycobacterials active against dormant cells and resistant strains is in need of novel drug targets. The failure of existing chemotherapeutic options to control the TB epidemic can be attributed in part to sub-therapeutic concentrations at the site of action [8]. The longer a pool of bacteria is exposed to sub-inhibitory levels of an antimicrobial agent, the more likely the emergence and selection of resistant clones becomes [9]. This has prompted researchers and drug discovery experts to turn to strategies which would potentiate existing therapeutics by increasing their intracellular levels through the use of small molecule inhibitors against efflux pumps [10].

The cell envelope of mycobacteria is notorious for being several-fold less permeable to chemotherapeutic agents when compared to functionally similar cell walls of other bacteria [11]. The knowledge of drug transport pathways could assist in the successful design of novel chemotherapeutic combinations against M. tuberculosis. Here, we review the current understanding of the various influx and efflux pathways in mycobacteria while focusing our attention on details specific to M. tuberculosis. The function and expression of transport proteins such as porins, drug importers and efflux pumps are summarized and their respective influence on the drug-resistant and non-replicating persistent states is highlighted. Collectively, the literature data compiled here show that M. tuberculosis and other mycobacteria have evolved several intrinsic and adaptive mechanisms to increase their level of tolerance towards xenobiotic substances, by preventing or minimizing their entry: (i) natural or intrinsic resistance mediated by the thickened highly hydrophobic and waxy envelope; (ii) reduced permeability resulting from physiological adaptations under unfavorable environmental conditions; (iii) drug-induced resistance acquired via increased expression of various classes of efflux pumps; and (iv) genetically encoded resistance conferred by mutations in efflux complexes.

The cell envelope of mycobacteria is structurally distinct from that of both Gram-positive and Gram-negative bacteria. The entire mycobacterial cell envelope can be broken down into two main structural components: cell membrane and cell wall. The outer leaflet of the cell wall is composed of mycolic acids which are covalently linked to the arabinogalactan-peptidoglycan complex of the inner leaflet. Mycobacteria are capable of producing a multitude of mycolic acids with varying lengths and modifications depending on species, strain and growth conditions [12,13,14]. It is widely believed that the unusually high mycolic acid content, combined with a variety of other intercalated lipids, contributes to the wall’s limited permeability [15]. The mycobacterial cell wall is also composed of phosphotidyl-myo-inositol derived glycolipids such as lipomannan and lipoarabinomannan which have potent immunomodulatory activities [16].

The mycolyl-arabinogalactan-peptidogalactan complex is acknowledged as being a more efficient permeability barrier than cell walls of any other class of bacteria [11]. Jarlier and Nikaido attempted to clearly define the mycobacterial permeability barrier to hydrophilic molecules by studying the uptake kinetics of small nutrient molecules (glucose, glycine, leucine and glycerol) in M. chelonae [17]. The permeability coefficients (P) for these nutrients were found ranging from 1.4 to 62 nm/s; specifically 2.8 nm/s for glucose. K_m values of the overall transport of glucose and glycerol were 1,000 µM and 200 µM, respectively, as measured in the same study. In comparison, a different study had measured a permeability coefficient of glucose for E. coli (1.4 × 10⁵ nm/s) that was about five orders of magnitude higher [18]. It should be noted that the precise values of permeability differ among different species of mycobacteria. M. chelonae, being one of the most drug-resistant species, has a cell wall that is about one to two orders of magnitude less permeable than M. tuberculosis, M. smegmatis and M. phlei [11,18]. This intra-species difference in cell wall permeability may be attributed to variability in its content and organization. Detailed structural and quantitative analysis has revealed a higher mycolate-to-peptidoglycan ratio in M. leprae than M. tuberculosis; peptidoglycan coverage by mycolate was estimated at 80% and 63% for M. leprae and M. tuberculosis, respectively [19].

This unique cell wall composition and organization is believed to render mycobacteria less susceptible than other bacterial pathogens to various antibiotic classes [11,20]. Several pathways exist for compounds to cross this permeability barrier. It is assumed that hydrophobic compounds should be able to penetrate cell walls by simply dissolving into and through the lipophilic cell wall unassisted, whereas the influx of hydrophilic compounds is largely facilitated by porins, which are water-filled open channels that span the cell wall [21]. It appears that the mycobacterial plasma membrane plays a limited role in pathogenicity and maintenance of the influx-efflux equilibrium [13,22].

In principle, antimicrobial agents of the more lipophilic classes such as the rifamycins, macrolides and fluoroquinolones are more likely to diffuse into and through the lipid-rich environment of the mycobacterial cell wall in order to transverse its depth [20]. This passive transport has been coined “hydrophobic (or lipid) pathway”, characterized by the nature of the interactions between structural lipids and small molecules [23]. However, lipophilic agents are presumably slowed down by the low fluidity and unusual thickness of the cell wall [24]. It has been demonstrated that lipophilic derivatives within single drug classes are more active against mycobacteria when compared to their hydrophilic counterparts [20]. This was more recently supported by evidence from a comparison of Minimum Inhibitory Concentrations (MIC) between hydrophilic and hydrophobic fluoroquinolone analogs. Moxifloxacin (cLogP 0.6) was 32-fold more effective than norfloxacin (cLogP −0.1) at inhibiting the growth of M. smegmatis [25], though the relative affinity of fluoroquinolones for the gyrase and differential susceptibility to efflux pumps also contribute to the net difference in MIC. Brennan et al. postulated that an increase in the rate of drug penetration resulting from an increase in incubation temperature is also evidence of the predominant role of the hydrophobic pathway or passive diffusion in drug penetration [20].

Porins are large water-filled channels allowing the penetration of small hydrophilic molecules without requiring energy consumption. This uptake pathway caters to a limited range of compounds since channel diameters at the narrowest point define the exclusion limit, and parameters such as channel length and the number of open pores determine the velocity of transport [26]. As demonstrated by studies in E. coli, diffusion rates through porins are further affected by the charge, hydrophobicity, and size of the solute [21,27,28]. Several types of porins have been identified and studied in Gram-negative and some Gram-positive bacteria. To date, two putative classes of porins have been identified and characterized in mycobacteria; they are MspA-like and OmpA-like porins in M. smegmatis and M. tuberculosis respectively [25].

MspA was the first porin of its class identified in a mycobacterial species, with proven oligomerization and channel-forming activity in vitro and when cloned in E. coli [29]. Subsequent sequencing of the M. smegmatis genome revealed three more porin genes with homology to mspA, namely mspB, mspC and mspD [30]. Numerous studies have documented MspA-enabled transport of hydrophilic solutes and drug molecules across the cell wall of M. smegmatis. Table 1 summarizes drug transport specificities of various M. smegmatis porins and their impact on drug uptake and MIC. The studies show that porin deletion is clearly linked to increases in MICs of various antibiotics. In several instances, this increase in MIC has been associated with reduction in drug uptake. Furthermore, heterologous expression of M smegmatis mspA accelerated the growth and increased the susceptibility of M. tuberculosis and M. bovis BCG to various classes of antibiotics [23]. This establishes the relationship between porin function, small molecule and nutrient uptake, and drug susceptibility in M. smegmatis.

Porins are only minor proteins in the mycobacterial cell wall unlike enterobacterial porins which are the most abundant proteins in the cell [23,26]. Direct counting of stained pores by electron microscopic analysis revealed a 45-fold lower number of MspA pores on the outer membranes of M. smegmatis when compared to pore counts of the outer membranes of Gram-negative bacteria [32]. The octameric MspA porin consists of two consecutive hydrophobic β-barrels, a more hydrophilic globular rim domain, and a single central channel of 9.6nm in length. This porin is largely embedded into the cell membrane of M. smegmatis with the embedded region including a portion of the hydrophilic rim domain [33]. Crystal structures revealed that the constriction zone of MspA is rich in aspartate residues. Together with the high number of negative charges in the vestibule and channel interior, this could explain the cation preference of MspA [34].

OmpATb was the first suggested porin-like molecule identified in M. tuberculosis. Encoded by the rv0899 gene, the name OmpATb was coined because of its homology to the E.coli porin OmpA [35].A study by Teriete et al. showed that Rv0899 does not form a transmembrane β-barrel, but a mixed α/β globular structure encompassing two independently folded modules which correspond to the B and C domains of the protein. The core of this B domain appears hydrophobic while its exterior is both polar and acidic [36]. Altogether, this proposed structure for OmpATb makes it unlikely for it to function as a porin. More recent structural elucidation by Yang et al. suggests that OmpATb forms a heptameric ring complex, driven by interactions between the α/β structured monomers, and is hypothetically capable of inserting itself into a biological membrane and form channels [37], allowing ion diffusion as observed in vitro. This model is based on NMR data of minor oligomeric populations of OmpATb in solution, and lies in contrast with available data suggesting the lack of functional porin assembly in vitro [38].

OmpATb plays a key role in conferring M. tuberculosis the ability to survive under acidic conditions. Deletion mutants in ompATb exhibit a significant reduction in permeability to several hydrophilic molecules and impaired ability to grow at reduced pH. The role of OmpA in acid resistance was reinforced by the observation of increased ompATb transcription levels in M. tuberculosis growing within macrophages, given that vacuole acidification is known to occur in infected phagocytes [39]. More recent functional studies revealed no obvious porin activity of OmpATb. Rather, chemical analysis of low-pH M. tuberculosis culture filtrates showed that OmpATb is involved in rapid ammonia secretion capable of neutralizing medium pH and restoring exponential bacterial growth. This is further substantiated by the discovery that Rv0899-like proteins are present predominantly in bacteria with functions in nitrogen fixation and metabolism [40]. OmpATb-mediated ammonia extrusion may be one of the multiple adaptations of M. tuberculosis to acidic environments and, on its own, is not critical for virulence in mice [38]. The porin function of OmpATb in M. tuberculosis clearly remains a controversial issue.

A recent attempt to predict outer membrane proteins in M. tuberculosis via a bioinformatics approach has led to the identification of two novel Omp-like proteins (Rv1698 and Rv1973) in M. tuberculosis. Both proteins have been proven to localize to the outer membrane [41]. Since then, the channel-forming activity of Rv1698 has been successfully characterized; Rv1698 expression has proven to restore sensitivity of MspA-deletion mutants of M. smegmatis to ampicillin and chloramphenicol, and complement the permeability defect of the mutant for glucose. Single homologues of Rv1698 are found only in mycolic-acid containing bacteria belonging to the suborder Corynebacterineae of the Actinomycetales, which includes mycobacteria. It therefore represents the first protein identified as specific for this suborder [42]. Orthologous porins PorM1 and PorM2 have since been characterized in M. fortuitum [43].

Interestingly, M. tuberculosis does not express the Msp-like porins that are found in faster-growing M. smegmatis [32]. This and other significant differences between the pathogenic and saprophytic mycobacterial species call into question the appropriateness of M. smegmatis as a model organism for anti-tuberculosis drug discovery and virulence studies of M. tuberculosis [15].

Table 2 summarizes several biophysical characteristics of mycobacterial and other bacterial porins. Single-channel conductance often provides an estimation of channel diameters of porins, and gives an indication of the relative mobilities of solutes through them. It can be observed from Table 2 that some proportionality exists between channel width, channel conductance and size exclusion limits. If OmpATb of M. tuberculosis does indeed form functional porin units, this trend would place its limit in the approximate range of 600–800 Da. This implies that the rifamycin and macrolide classes are too large to utilize OmpATb to transverse the cell wall. Danilchanka et al attempted to illustrate the fit of a drug when oriented along their longest axes within the MspA porin constriction zone by using 3D structure visualization and surface representations of structural models of antibiotics. They predicted that ampicillin, chloramphenicol and norfloxacin are able to utilize this porin molecule, as opposed to antibiotics such as erythromycin, kanamycin and vancomycin [25].

The specific role played by porins in intracellular drug accumulation within other bacterial species has been well studied. The outer membrane porin protein OprD of Pseudomonas aeruginosa has been directly implicated in the influx of imipenem; a staggering 98% of imipenem- and meropenem-resistant P. aeruginosa clinical isolates have been identified as being negative for OprD porin production [52]. Similarly, studies on expression levels of the porin protein OmpF in clinical isolates of E. coli have linked the decreased expression levels of this porin with resistance to quinolones [53,54]. In these enterobacterial pathogens, the reduction in the number of functional porins per cell is mostly due to a decrease or complete shutdown of synthesis, or the expression of an altered porin. These changes bring about decreased susceptibility to antimicrobials and favor the acquisition of additional mechanisms of bacterial resistance [55]. Changes in the expression levels of functional porins should therefore be viewed as potential contributing factors in the development of resistance in mycobacteria.

In conclusion, porins appear to be less varied and less abundant in mycobacteria than in other bacterial families such as the enterobacteriaceae, though the possibility remains that we have only detected a small fraction of the total mycobacterial porin panel. The wide range of metabolic and physiologic adaptations seen in M. tuberculosis, combined with the generally complex regulation of porin expression in other species, suggest that M. tuberculosis has likewise exploited porin modulation as a strategy to fence itself off from harmful small molecules. Some of these putative schemes are discussed in a later section.

In the hunt for novel anti-tuberculosis drugs, the physico-chemical properties driving cell wall permeation by chemotherapeutic agents remain a critical but largely unsolved question. Table 4 lists biological and molecular properties of selected anti-TB agents, compiled in an attempt to detect correlation trends between intracellular accumulation and physico-chemical properties. The level of drug accumulation within M. tuberculosis cells varies significantly between drug classes but drawing comparisons between them is difficult due to differences in experimental methods used to measure accumulation. Intracellular concentrations of the fluoroquinolones listed were determined by measuring their fluorescence in cell lysates [120]. Pyrazinamide, isoniazid and rifampicin intracellular concentrations were calculated from scintillation counts of radio-labeled compounds in whole-cell preparations [121,122,123], which includes cell wall-associated drug content. This results in a possible overestimation of intracellular drug concentration. In the study of efficacy of drugs targeting cytosol-localized proteins, only the drug content of the cytosol compartment is of concern.

Molecular weight is often speculated to be an important determinant of the rate of diffusion and cell wall permeation; the smaller the compound the higher the rate of passive diffusion [124], while ClogP, a measure of the lipophilicity of a compound, reflects partitioning into the hydrophobic phase of the cell wall [125]. Attempts to plot these parameters as a function of intra-bacillary accumulation failed to reveal any correlations (not shown), consistent with the growing realization that many anti-TB agents lie outside the drug-like chemical space in which the Lipinski rule-of-five prevails [124,126]. The observation that the hydrophilic fluoroquinolones efficiently accumulate within M. tuberculosis supports the hypothesis that diffusion through porins serves as a route of transport across the mycobacterial cell wall for this class of compounds. Altogether, weak correlations are very likely due to the fact that many of these anti-tuberculous agents are transported via a combination of pathways. This analysis is also limited by the small number of drugs included and the fact that we attempted to compare physico-chemical properties across different compound classes. The intracellular concentrations of pro-drugs that require enzymatic activation are not always reflective of their potency; the intracellular concentration of active metabolites may be more relevant in the cases of isoniazid and pyrazinamide.

Pyrazinamide appears to accumulate 6-fold within the bacilli, partly owing to its bioconversion to pyrazinoic acid and subsequent trapping of the charged molecule. Its transport has been determined as being ATP-dependent and reliant on the nicotinamide transport pathway [121]. However, the other drugs listed that accumulate intracellularly above the extracellular concentration have been speculated as being passively taken up by mycobacteria. In the case of isoniazid which accumulates 4–5 times within M. tuberculosis, the calculated accumulation factor accounts for both isoniazid and metabolites since the read-out measures all isoniazid-derived radioactive material. Constant conversion by KatG ensures a consistent pro-drug concentration gradient between the intracellular and extracellular compartments [122]. Fluoroquinolones appear to concentrate within the intracellular compartment despite being un-metabolized. The ability to concentrate a drug within the intracellular environment of M. tuberculosis should reflect the presence of active drug importers which have yet to be identified.

In conclusion, there is no simple formula linking physico-chemical parameters to intracellular accumulation of small molecules in M. tuberculosis. As mentioned above, the contribution of a variety of passive and active mechanisms of uptake and efflux precludes the use of a simple equation to solve this question. Further studies focusing on relatively large numbers of molecules within the same chemical scaffold should help identify the major determinants of uptake for a given class of small molecules or therapeutic agents.

Thus far, tools have not been developed to accurately predict the extent to which a drug utilizes either the un-facilitated or porin-facilitated diffusion pathway to achieve cell wall penetration in mycobacteria. Several studies have looked at the complex inhibition of E. coli porins OmpF and OmpC by polyamines. Cadaverine, putrescine, spermidine and spermine have all been shown to inhibit both chemotaxis and the flux of β-lactam antibiotics in E. coli [132]. These polyamines have been shown to increase the number and duration of channel closures, while promoting the blocked or inactivated state [133,134]. The impact of polyamine inhibition on porins has not yet been demonstrated in mycobacteria but similar studies could open up the field of porin-mediated drug uptake in these species.

A recent elegant study by Allison et al. has shed light on the ability of specific metabolites to potentiate the eradication of both E. coli and S. aureus by aminoglycosides [135]. Aminoglycoside uptake in exponentially growing bacteria requires a PMF. Metabolites entering upper glycolysis such as glucose, mannitol and fructose enhance aminoglycoside uptake and the rapid killing of persisters by inducing PMF. It remains to be shown whether PMF-stimulating metabolites are similarly able to serve as adjuvants to aminoglycoside therapy for M. tuberculosis infections.

A handful of studies have established the link between antimicrobial drug uptake and active influx mechanisms in non-mycobacterial species. Streptomycin accumulation, for example, in E. coli and P. aeruginosa is energy-dependent and saturable [136]. Similar results were attained with gentamicin accumulation in the same two bacterial species; it was shown to concentrate between 4 to 250 times over the extracellular concentration [137]. It is conceivable that similar active drug influx mechanisms exist in M. tuberculosis and other mycobacterial species.

With regards to efflux pumps, the question remains: What chance is there of devising pump inhibitors as effective clinical solutions? Thioridazine appears to be a promising therapeutic option in the fight against MDR- and XDR-TB infections [138,139]. When used in combination with second-line anti-tuberculous agents it boosts the management of such infections [140]. Other than having been approved for use on the basis of compassionate therapy for XDR-TB patients in Mumbai, India, thioridazine remains untapped as an effective anti-tuberculous agent. Fellow anti-psychotic agent chlorpromazine has also displayed efflux pump inhibition activity in M. avium and M. smegmatis [138]. It is not known which specific class of bacterial efflux pumps is inhibited by these agents.

This review attempts to provide a holistic view of drug transport mechanisms in mycobacteria, and their potential contribution to intrinsic and acquired drug resistance, phenotypic drug tolerance, and genetic resistance. Improving our understanding in this field will help to decipher the virulence and resistance mechanisms of M. tuberculosis, and other pathogenic mycobacterial species.