Mycobacteria protect themselves with an outer lipid bilayer, which is the thickest biological membrane hitherto known and has an exceptionally low permeability rendering mycobacteria intrinsically resistant to many antibiotics. Pore proteins spanning the outer membrane mediate the diffusion of hydrophilic nutrients. Mycobacterium tuberculosis possesses at least two porins in addition to the low activity channel protein OmpATb. OmpATb is essential for adaptation of M. tuberculosis to low pH and survival in macrophages and mice. The channel activity of OmpATb is likely to play a major role in the defence of M. tuberculosis against acidification within the phagosome of macrophages. MspA is the main porin of Mycobacterium smegmatis. It forms a tetrameric complex with a single central pore of 10 nm length and a cone-like structure. This structure differs clearly from that of the trimeric porins of Gram-negative bacteria, which form one 4 nm long pore per monomer. The 45-fold lower number of porins compared to Gram-negative bacteria and the exceptional length of the pores are two major determinants of the low permeability of the outer membrane of M. smegmatis for hydrophilic solutes. The importance of the synergism between slow transport through the porins and drug efflux or inactivation for the development of drugs against M. tuberculosis is discussed.
The main interest in the genus Mycobacterium is because of Mycobacterium tuberculosis, which has infected about two billion people and causes about two million deaths from tuberculosis (TB) per year, more than any other single infectious agent (Bleed et al., 2001). Mycobacterium tuberculosis remains one of the most deadly pathogens of humans even though infections with drug-sensitive strains can be cured. In addition to the paramount medical importance of M. tuberculosis, mycobacteria are of great interest biologically because they have specialized in fortifying themselves with a thick and lipid-rich cell envelope. Mycobacterium tuberculosis synthesizes a vast amount of exceptionally long fatty acids, the mycolic acids, which account for 30% to 40% of the dry weight of the cell envelope (Rastogi et al., 2001). Mycolic acids are α-branched β-hydroxy fatty acids and consist of up to 90 carbon atoms (Barry et al., 1998). The maximal lengths of the hydrocarbon chains of the α-branch and the meromycolate branch are 26 and 60 carbon atoms, respectively, for M. tuberculosis H37Rv (Fig. 1A) (Watanabe et al., 2001). To date, the mycolic acids produced by mycobacteria are the longest fatty acids identified in nature. Minnikin originally proposed that the mycolic acids are part of an outer membrane (OM) in addition to the cytoplasmic membrane (Minnikin, 1982; 1991). Experimental evidence for this model was provided by X-ray diffraction studies, which showed a quasi-crystalline packing of lipids in purified cell walls of Mycobacterium chelonae. It appears very likely that non-covalently bound lipids complement the ordered arrangement of mycolic acids to an asymmetric bilayer (Nikaido et al., 1993). Hence, the current model of the mycobacterial cell envelope includes the presence of an OM (Fig. 1B), although mycobacteria are classified as Gram-positive bacteria based on the phylogenetic analysis of 16S rRNA sequences (Pitulle et al., 1992). Interestingly, recent genome-based phylogenetic trees suggest a closer relationship of mycobacteria to Gram-negative bacteria (Fu and Fu-Liu, 2002). In summary, mycobacteria have perfected the protective function of their cell envelope by combining several unique features: (i) because of the length of the mycolic acids, the OM of mycobacteria is thicker than any other known membrane. The thickness was estimated to be 9–10 nm (Brennan and Nikaido, 1995) based on electron microscopic examinations of stained thin-sections of mycobacterial cells. (ii) The covalent attachment of the polar head groups to the arabinogalactan-peptidoglycan co-polymer of the cell wall results in the absence of any lateral mobility of the mycolic acids of the inner leaflet (Barry et al., 1998). Both the length of the mycolic acids and the attachment to a single polymeric head group contribute to the very low fluidity of the mycobacterial OM. This membrane does not melt at temperatures as high as 70°C (Liu et al., 1996), in contrast to membranes of mesophilic organisms, which begin to melt at − 10°C (Melchior and Steim, 1976). Consequently, lateral diffusion of proteins and other molecules within the mycobacterial OM must be very restricted. (iii) OM fluidity decreases towards its periplasmic face, in contrast to the OM of Gram-negative bacteria (Liu et al., 1995). For further details of the physical organization and the chemical composition of the mycobacterial cell wall, the reader is referred to other reviews (Brennan and Nikaido, 1995; Daffé and Draper, 1998; Draper, 1998; Brennan, 2003).
The unique mycolic acid bilayer is an extremely efficient permeability barrier protecting the cell from toxic compounds and is generally thought to be the major determinant of the intrinsic resistance of mycobacteria to most common antibiotics, chemotherapeutic agents and chemical disinfectants (Brennan and Nikaido, 1995). The mycolic acid bilayer is functionally analogous to the OM of Gram-negative bacteria, a similarity that merits a more detailed comparison. Three general and several specific pathways for transport across the OM of Gram-negative bacteria exist: (i) hydrophobic compounds penetrate the membrane by temporarily dissolving in the lipid bilayer. (ii) Small and hydrophilic compounds diffuse through water-filled protein channels, the porins (Nikaido, 1994). Some of the diffusion channels show specificity toward certain classes of compounds such as maltodextrin (Dumas et al., 2000) or nucleosides (Maier et al., 1988). (iii) Polycationic compounds are thought to disorganize the OM thereby mediating their own uptake in a process, which was termed ‘self-promoted uptake’ (Hancock et al., 1991). (iv) Certain compounds are specifically taken up by transporter proteins such as FhuA and FepA, which transport iron-loaded siderophores in an energy-dependent process across the OM of E. coli (Braun and Killmann, 1999). For further consideration of the permeability properties of the mycobacterial cell wall, it is important to distinguish between these different pathways which are likely to exist also in mycobacteria. The mycobacterial porins are the focus of this review because they are essential for nutrient uptake and are assumed to be the major pathway for influx of hydrophilic drugs. Knowledge of the structure and the transport properties of porins and how expression of porin genes is regulated will provide clues about nutrient supply and protection of mycobacteria under different physiological conditions. This information might be exploited to improve the transport efficiencies of currently used and new antibiotics directed against pathogenic mycobacteria.
Observation of porins of M. tuberculosis and M. bovis BCG
The porins of M. tuberculosis are medically the most relevant mycobacterial porins, because they are thought to be the key proteins for the uptake of hydrophilic drugs. This view is based on the fact that three out of the four current first line TB drugs, namely isoniazid, ethambutol and pyrazinamide, are small and hydrophilic molecules (Lambert, 2002). In apparent contrast to this assumption both isoniazid and pyrazinamide penetrated lipid bilayers in liposome swelling experiments indicating that these compounds might also be able to cross the mycobacterial OM directly without the help of porins (Jackson et al., 1999; Raynaud et al., 1999). However, it should be emphasized that the pathway of any TB drug across the the OM of M. tuberculosis is not known. Transport experiments with purified porins and analysis of porin mutants are necessary to dissect OM diffusion pathways for TB drugs.
The existence of porins in M. tuberculosis was demonstrated in two laboratories (Senaratne et al., 1998; Kartmann et al., 1999). Channels with conductances of 0.7 nS and of approximately 3 nS in 1M KCL were observed in detergent extracts of M. tuberculosis (Table 1). Apparently consistent with these results was the detection of porins that produced channels of 0.8 nS and 4 nS in detergent extracts of M. bovis BCG (Lichtinger et al., 1998), a species that is genetically nearly identical to M. tuberculosis (Garnier et al., 2003). However, the low-conductance channel of M. bovis BCG was found to be anion-selective (Lichtinger et al., 1998) in contrast to the low-conductance channel of M. tuberculosis (Kartmann et al., 1999) indicating that the proteins producing these channels are likely not identical. These studies showed that M. tuberculosis and M. bovis BCG contained at least two different types of porins. Unfortunately, in all cases the amount of porins in detergent extracts was too low to allow purification and identification of these proteins.
determined with planar lipid bilayer experiments in the presence of 1 M KCl.
b. Apparent molecular weights of the porin oligomer MWO and monomer MWM as determined by gel electrophoresis.
. It is not known whether these porins are monomers or oligomers.
. The indicated channel properties are most likely those of MspA, the main porin of M. smegmatis. It is not known, which porin genes are expressed in addition to mspA and whether the other porins MspB, C and D have different channel properties than MspA.
e. Very low channel activity; active only in the presence of the putative signal peptide; important role in adaptation of M. tuberculosis to low pH (Raynaud et al., 2002); not considered as a classical porin.
In a different approach to identify porins, a protein of M. tuberculosis with significant homology to the OmpA protein family was produced in E. coli (Senaratne et al., 1998). Whereas purified recombinant OmpATb without the putative N-terminal signal peptide did not show channel activity, the recombinant protein with signal peptide increased the permeability towards various sugars in liposome swelling experiments and produced channels with conductances of 0.7 nS to 3.5 nS in lipid bilayer experiments (Table 1). These results were interpreted as proof that OmpATb was a porin of M. tuberculosis (Senaratne et al., 1998). However, the recombinant OmpATb protein used in this study was probably not a good mimic of the native protein, because it still contained the putative signal peptide. It is expected that signal peptides are cleaved off by signal peptidases upon translocation of the protein across the inner membrane of M. tuberculosis (Braunstein et al., 2001). Thus, it is not clear to what extent the channel properties of recombinant OmpATb were influenced by the presence of the signal peptide. Planar lipid bilayer experiments indicated that other porins than OmpATb gave rise to the 0.7 nS channels observed in detergent extracts of M. tuberculosis (Kartmann et al., 1999). Thus, the in vitro data regarding porin function of OmpATb are not conclusive.
To demonstrate the porin function of OmpATb in vivo, the rate of uptake of hydrophilic solutes by an ompATb mutant of M. tuberculosis was analysed. Uptake of serine was reduced in the ompATb mutant compared to wild-type M. tuberculosis, but the uptake rates differed only marginally within the first 15 min at pH 7.2 (Raynaud et al., 2002a). Moreover, uptake of glycine was even faster in the mutant. These results are not consistent with OmpATb being a major general porin of M. tuberculosis. Indeed, the overall permeability of the OM of M. tuberculosis was reduced at pH 5.5 compared to pH 7.2, although the levels of OmpATb in the OM were strongly increased. This also suggests that other pore-forming proteins are the main determinants of the cell wall permeability of M. tuberculosis. These porins await discovery.
OmpATb is required for adaptation to low pH and for survival in macrophages and mice
The ompATb mutant of M. tuberculosis revealed a new and unexpected phenotype. Whereas growth of M. tuberculosis was delayed for approximately two days after reinoculation into a medium with a pH of 5.5 compared to medium with a pH of 7.2, growth of the ompATb mutant was delayed by between 8 and 15 days. Furthermore, multiplication of the ompATb mutant was hampered in macrophages and in intravenously infected mice (Raynaud et al., 2002a). The failure of the ompATb mutant to grow at low pH is likely the cause of the virulence defect in mice, because activated macrophages are able to override the block of phagosome acidification exerted by M. tuberculosis and to lower the pH inside phagocytic vacuoles (Schaible et al., 1998). The strong induction of ompATb transcription at low pH (30-fold) and in macrophages (fivefold) (Raynaud et al., 2002a) suggests that acidification of the phagosome is the signal which triggers an OmpATb-depending defence mechanism of M. tuberculosis in macrophages to cope with growth-limiting proton concentrations. Thus, OmpATb is one of the few protein virulence factors associated with the OM of M. tuberculosis in addition to the exported repeated protein (Berthet et al., 1998), heparin-binding haemagglutinin (Pethe et al., 2001) and phospholipases C (Raynaud et al., 2002b).
The controversy over the transport function of OmpA-like proteins takes a new twist with OmpATb of M. tuberculosis
OmpA-like proteins exist in all Gram-negative bacteria examined, possess a similar sequence at their carboxy-terminal ends, and share many biochemical properties suggesting common functions (Beher et al., 1980). Whereas the functions of E. coli OmpA in conjugation (Schweizer and Henning, 1977), in maintaining the structural integrity of the OM (Sonntag et al., 1978) and in resistance to environmental stresses (Wang, 2002) have been clearly established, there is an ongoing debate about whether OmpA forms a pore. Numerous reports describe the pore-forming activity of OmpA of E. coli in artificial membranes (Sugawara and Nikaido, 1992; Arora et al., 2000; Saint et al., 2000). However, the structure of the N-terminal transmembrane β-barrel domain of E. coli OmpA both in crystaIs (Pautsch and Schulz, 1998; Bond et al., 2002) and in solution (Arora et al., 2001) did not show an open water-filled channel. Furthermore, deletion of ompA in Salmonella and E. coli did not affect the OM permeability (Bavoil et al., 1977; Nikaido et al., 1977) in contrast to many mutants lacking functional porins (Nikaido et al., 1977; Harder et al., 1981; Sugawara and Nikaido, 1994). Nikaido and co-workers noted early that the rate of penetration of solutes mediated by OmpA is at least 50-fold lower than the rate produced by E. coli OmpF porin (Sugawara and Nikaido, 1992). This difference was attributed to a small fraction of 2–3% of the OmpA molecules displaying open channels (Sugawara and Nikaido, 1994). Two conclusions can be drawn from these experiments. First, OmpA clearly forms pores when reconstituted in planar lipid bilayers or liposomes. Second, OmpA does not contribute significantly to the permeability properties of the OM of E. coli and Salmonella under most experimental conditions.
The observations of Colston and colleagues that OmpATb plays a significant role for the OM permeability of M. tuberculosis at low pH and is necessary for adaptation of M. tuberculosis to low pH (Raynaud et al., 2002) implies a mechanism which might solve the apparent contradiction about the pore activity of OmpA proteins in vitro and in vivo. Low pH might trigger a switch of OmpATb to an open-pore conformation which allows export of a basic compound, e.g. ammonia, to neutralize excess protons in the environment and, thereby, might increase OM permeability for hydrophilic solutes. This mechanism requires the existence of at least two conformations of OmpATb with significantly different permeability properties, which have not been observed so far. However, two conformations of E. coli OmpA with different channel conductances were observed in planar lipid bilayer experiments and it was suggested that the small and large channels conformations were interconvertible (Arora et al., 2000). The OmpA-like outer membrane protein OprF of Pseudomonas fluorescence was isolated in a low and high conductance state when the cells were grown at 8°C and 28°C, respectively (De et al., 1997) suggesting that interconversion of the two conformers occurs in vivo. Switching of porins between an open and a closed conformation was observed by Delcour and co-workers using patch clamp measurements of live bacterial cells, spheroplasts and reconstituted liposomes containing membrane fragments (Delcour, 1997). The closure frequency of porins is modulated by polyamines, external pH and voltage, and this regulated closure enables E. coli to quickly respond to environmental signals by altering the OM permeability (Dela Vega and Delcour, 1995; Liu and Delcour, 1998; Iyer et al., 2000). This showed that porins are not static, permanently open pores, but dynamic proteins. In a similar but opposite manner, OmpA proteins might be able to switch from a closed to an open-channel conformation. Certainly, more experiments are needed to prove this hypothesis. In this regard, it would be interesting to analyse whether the function of OmpATb in adaptation to low pH is specific for M. tuberculosis or is also a property of other OmpA proteins.
Identification of porins of M. smegmatis
In contrast to other fast-growing mycobacteria, M. smegmatis can be transformed with high efficiency (Snapper et al., 1990) and as a consequence has become widely used as a model organism (Reyrat and Kahn, 2001). However, the numerous physiological differences between the saprophyte M. smegmatis and the human pathogen M. tuberculosis do not allow direct extrapolations (Barry, 2001). Nevertheless, M. smegmatis is certainly useful to study common aspects of the biology of mycobacteria. For example, the principal structure of the cell wall was found to be identical for many mycobacteria (Paul and Beveridge, 1992), although there is substantial structural variation of cell wall bound mycolic acids (Barry et al., 1998). Thus, there may be common mechanisms of solute transport across the mycobacterial cell wall, which can be studied in M. smegmatis, even if there are different proteins involved than in M. tuberculosis.
The channel-forming properties of a crude cell wall extract of M. smegmatis were characterized in lipid bilayer experiments (Table 1) (Trias and Benz, 1994). In an attempt to identify the porins responsible for channel activity, Mukhopadhyay et al. (1997) observed low channel activity in liposome swelling experiments using crude cell wall fractions of M. smegmatis enriched in a 40 kDa protein. However, they did not purify the 40 kDa protein or demonstrate by other means that this protein is a porin. In addition, the N-terminal sequence determined for the cell wall fraction with channel activity is identical to the N-terminus of the DNA-binding protein HU of M. smegmatis (Lee et al., 1998), which excludes porin function for this protein.
A porin with an apparent molecular mass of 100 kDa protein was purified from CHCl3/MeOH extracts of whole cells of M. smegmatis (Niederweis et al., 1999). This protein is detectable down to a concentration of a few pg ml−1 in lipid bilayer experiments indicating high channel activity and had a single channel conductance similar to that of whole cell extracts of M. smegmatis (Table 1). The N-terminal and C-terminal sequences of this protein were entirely encoded by the mspA gene and indicated that a signal peptide of 27 amino acids was cleaved off of the mature protein. Expression of the mspA gene encoding the mature MspA protein in E. coli yielded a folded, but inactive 20 kDa monomer (Heinz et al., 2003a) and an oligomer with an apparent molecular mass of 100 kDa, which formed channels similar to those of the native protein (Niederweis et al., 1999). These results demonstrated that the mspA gene indeed encodes a porin of M. smegmatis and that oligomerization is necessary for channel activity of MspA.
MspA is the major porin of M. smegmatis and is responsible for the uptake of hydrophilic solutes
The chromosome of M. smegmatis encodes four very similar porins designated MspA, B, C and D. The mature MspB, C and D proteins differ only at 2, 4 and 18 positions, respectively, from MspA (Stahl et al., 2001). Detergent extracts of a ΔmspA mutant exhibited significantly lower channel activity in lipid bilayer experiments and contained less porins than extracts of wild-type M. smegmatis. The lower porin content reduced the cell wall permeability of the ΔmspA mutant towards cephaloridine and glucose nine- and fourfold respectively (Stahl et al., 2001). These results showed that MspA is the main general diffusion pathway for hydrophilic molecules across the OM of M. smegmatis and was only partially replaced by fewer porins in the cell wall of the ΔmspA mutant. Furthermore, these studies demonstrated for the first time that porins limit the rate of uptake of hydrophilic nutrients in M. smegmatis (Stahl et al., 2001). One problem in assessing the physiological function of MspA quantitatively is that it is not known whether the other Msp porins have identical transport properties compared to MspA. One out of two, two out of four and seven out of 18 of the amino acids differing in MspB, MspC and MspD, respectively, compared to MspA alter the charge of the protein and might affect the transport of ionic compounds. The importance of charged amino acids for the transport properties of porins has been conclusively shown for porins of E. coli (Bauer et al., 1989; Phale et al., 2001). Additionally, the interpretation of the properties of the ΔmspA mutant is complicated because we do not know whether all msp genes are expressed in an identical manner in wild-type M. smegmatis and the ΔmspA mutant.
MspA is the prototype of a new class of tetrameric porins
The Msp porins do not show significant sequence similarity to other proteins in the database. Electron microscopy and cross-linking experiments revealed that MspA is a tetrameric protein forming a central pore of 10 nm length (Fig. 1C, Engelhardt et al., 2002). This arrangement is different than that of the trimeric porins of Gram-negative bacteria, which have one pore per monomer and are approximately 4 nm long (Koebnik et al., 2000). These differences indicate that MspA represents a new type of pore protein. Analysis of the secondary structure by infrared spectroscopy revealed that MspA consists of approximately 40–45% antiparallel β-strands and 10% α-helices (Heinz et al., 2003b). Denaturation of the β-structure occurred between 102 and 110°C, indicating that the β-strands of MspA are organized in an extraordinarily stable coherent domain (Heinz et al., 2003b). It was proposed that the β-sheets of MspA form an open β-barrel. Indeed, a β-barrel as a common structural element of outer membrane proteins (OMP′s) of both Gram-negative (Koebnik et al., 2000) and Gram-positive bacteria is reasonable. Hydrophobic α-helices would be alternative membrane-spanning structural elements of proteins, but would stop the transfer of the polypeptide to the OM by preventing its release from the inner membrane (Pugsley, 1993). A mixture of α-helices and β-sheets or an open β-sheet structure would leave the amide hydrogen bond donors and acceptors at the β-sheet edges exposed (Schulz, 2002) and would therefore represent energetically unfavoured structures in lipid environments. At present, transmembrane β-barrel proteins have been found exclusively in the OM of Gram-negative bacteria and these membranes lack α-helical proteins (Schulz, 2002). Thus, the sole solution to the architectural problems of OMP′s appears to be the association of both edges of a β-sheet into a barrel. The unique mycobacterial OM provides an interesting test case for this idea.
Why is the permeability of mycobacterial outer membranes for hydrophilic solutes so low?
The first report about mycobacterial porins noted the relatively low yield of purified porin from M. chelonae (Trias et al., 1992). It was tempting to explain this result as being due to a low number of porins in the outer membrane, which would have provided an obvious reason for the low permeability of the outer membrane. However, other explanations such as poor solubilization of the porin could not be excluded. Direct counting of uranyl acetate stained pores by electron microscopy analysis revealed approximately 1000 MspA-like pores per µm2 cell wall of M. smegmatis (Engelhardt et al., 2002). Thus, the density of porins is about 15-fold less than that in the OM of Gram-negative bacteria, in which more than 15 000 porin trimers per µm2 in a two-dimensionally crystalline arrangement were observed (Kessel et al., 1988). This is equivalent to a 45-fold lower number of OM pores, because MspA contains only one channel per porin molecule in contrast to the three channels per porin of E. coli (Cowan et al., 1992).
Diffusion rates through channels are reduced compared to those in bulk solutions due to interactions with the channel wall or with internal loops (Im and Roux, 2002) and decrease with the size of the substrate (Nikaido and Rosenberg, 1983). Solute interactions with the channel interior are likely to be more pronounced in the 2.5-fold longer MspA pore than in E. coli porins. Thus, the low number of porins and the exceptional length of the pores are two determinants of the low permeability of the OM of M. smegmatis for hydrophilic solutes. Because low porin activity and a low yield of porins after purification have been noted in several other mycobacterial species (Trias et al., 1992; Kartmann et al., 1999; Lichtinger et al., 1999) and all mycobacterial porins have to span an OM with a similar thickness (Paul and Beveridge, 1992; Paul and Beveridge, 1993), similar causes for low permeability are likely to exist for all mycobacteria. Considering that mycobacteria have designed their cell wall more like a fortress than any other known bacteria, it would not be surprising to find more mechanisms by which mycobacteria prevent intrusion of toxic compounds through the porins. Such mechanisms may involve the so-called eyelet loop, which folds back into the channel interior of porins and drastically restricts the area accessible to diffusion (Cowan et al., 1992), or the directed closing of the porins as occurs in Gram-negative bacteria (Delcour, 1997).
How do mycobacterial porins integrate into membranes?
The architecture of a membrane is an important determinant of the structure and function of integral membrane proteins. For example, the thickness of the hydrophobic region is known to influence properties such as fluidity, ion permeability, capacitance of a membrane and structure and function of transmembrane proteins (Killian, 1998). A striking example is the strict dependence on membrane thickness of the formation of β-sheet secondary structure, closure of the β-barrel and membrane insertion of OmpA of E. coli (Kleinschmidt and Tamm, 2002). Examinations of stained thin-sections of mycobacterial cells revealed a 5–12 nm thick electron transparent zone (ETR) outside the cytoplasmic membrane (Koike and Takeya, 1961; Paul and Beveridge, 1992; 1993), which is thought to be the hydrocarbon core of the OM (Barry et al., 1998). Therefore, reconstitution efficiency and/or activity of mycobacterial OMP′s might be reduced when analysed in thinner membranes produced from conventional lipids. Surprisingly, porins of M. chelonae and M. smegmatis reconstituted very efficiently into liposomes (Trias et al., 1992) and into planar lipid membranes (Trias et al., 1992; Trias and Benz, 1994; Niederweis et al., 1999) and showed a high channel activity, although these membranes were only approximately 4 nm thick. This should leave a considerable stretch of the M. smegmatis porin MspA, which is presumably covered by lipids of the mycobacterial OM, exposed to water in those experiments. Indeed, electron micrographs showed that approximately half of the MspA pore sticks out of dimyristoyl phosphatidylcholine (C14) liposomes (Engelhardt et al., 2002). Such a large hydrophobic mismatch would be energetically very unfavourable, but apparently only ∼5 nm of the MspA channel are very hydrophobic, whereas a large domain of ∼2 nm length is thought to be surface-exposed in M. smegmatis (Engelhardt et al., 2002). Such a hydrophobicity gradient along the 10 nm long surface of the MspA channel would reduce the number of hydrophobic residues exposed to a hydrophilic environment in artificial lipid bilayers with acyl chains considerably shorter than the mycolic acids of M. smegmatis. The strongly asymmetric insertion into planar lipid bilayers is consistent with the presence of a very hydrophobic end of MspA (Engelhardt et al., 2002).
It is obvious that mycobacterial porins must span the OM completely to fulfil their transport function. Therefore, these results also suggest a maximal thickness of ∼8 nm for the OM of M. smegmatis consistent with the electron microscopy pictures of the mycobacterial cell envelope (Paul and Beveridge, 1992) and molecular dynamic calculations, which indicate a maximal thickness of ∼9 nm for the OM of M. smegmatis (unpublished results). In conclusion, mycobacterial porins do not appear to be strongly affected by a hydrophobic mismatch with the membrane in contrast to many other membrane proteins (Killian, 1998; Kleinschmidt and Tamm, 2002).
Evidence for porins in other mycobacteria
Porins in the genus Mycobacterium were discovered when Trias and colleagues observed the diffusion of sugars and amino acids into lipid vesicles after reconstitution of a cell wall extract from M. chelonae (Trias et al., 1992). They purified a 59 kDa protein from the isolated cell wall of M. chelonae and demonstrated its channel activity by both liposome swelling experiments and lipid bilayer measurements. Although the specific activity for arabinose diffusion in liposome swelling experiments was more than 10-fold lower for the M. chelonae porin compared to OmpF of E. coli, it showed a high channel activity at a rather low protein concentration of 2 ng ml−1 in lipid bilayer measurements. In subsequent experiments, the channel properties of this putative M. chelonae porin were determined (Table 1) (Trias and Benz, 1993). Unfortunately, the identity of the 59 kDa protein is unknown and final evidence that this protein gave rise to the observed channel activity, is lacking. Mycobacterium phlei is another fast growing species of the genus Mycobacterium, is closely related to M. smegmatis (Roth et al., 1998) and contains a set of porin genes very similar to the msp genes of M. smegmatis. Not surprisingly, a porin with similar properties as MspA was isolated from M. phlei (Table 1) (Riess et al., 2001).
Why do slow-growing mycobacteria have different porins?
Taking the similar cell wall architecture into account, it is anticipated that all mycobacteria need porins to overcome the OM permeability barrier to hydrophilic solutes. However, the genomes of M. tuberculosis and M. bovis BCG do not encode for proteins with significant sequence homology to MspA (Niederweis et al., 1999) and other members of the family of Msp porins of M. smegmatis and M. phlei. This does not necessarily mean that the structure of porins of fast- and slow-growing mycobacteria is different, because porins can exhibit large sequence variations and still possess similar structures (Schiltz et al., 1991; Puntervoll et al., 2000). However, both the channel activity and the channel properties of porins in detergent extract of M. tuberculosis are different from that of M. smegmatis (Kartmann et al., 1999) indicating that the sequence divergence of the porins of these organisms may indeed reflect structural and functional differences. Porins are the first gates of the bacterial cell and as such must fulfil two major functions which at first sight seem to be mutually exclusive. Porins must allow influx of nutrients at sufficiently high rates to support growth and must protect the cells from potentially harmful concentrations of toxic compounds. Gram-negative bacteria have solved this problem by combining several strategies. Diffusion of larger solutes through the porins is restricted by a loop, which is folded inwardly into the channel (Weiss et al., 1991), and porin-mediated OM permeability is regulated on all possible levels (transcription, translation and protein activity) in a complex manner to balance these opposing requirements (Pratt et al., 1996; Ferenci, 1999; Delcour, 2003). Thus, it is conceivable that adaptation of the slow-growing pathogen M. tuberculosis to its intracellular life style also required significant changes of the porins compared to the saprophytic fast-growing mycobacteria, which are normally found in soil, water and dust (Howard and Byrd, 2000).
It has been proposed that the low permeability of the mycobacterial cell wall for hydrophilic nutrients might be the reason for the slow growth of some mycobacterial species (Jarlier and Nikaido, 1990). In apparent contrast, Nikaido and co-workers have demonstrated that the OM permeability for some penicillins and cephalosporins is similar for M. tuberculosis and M. smegmatis (Chambers et al., 1995). Thus, it is unlikely that the sequence divergence of the porins is the major determinant of the growth rate of mycobacteria. On the other hand, growth experiments with M. bovis BCG expressing mspA support the assumption that nutrient influx through the porins partially limits the growth rate of slow-growing mycobacteria (C. Mailänder and M. Niederweis, unpubl. results). Certainly, more experiments are needed to solve this puzzle.
Conclusions and prospects
During the last decade porins were discovered in many species of the genus Mycobacterium. Considerable progress has been made in unravelling the physiological functions of two of those proteins. OmpATb was originally proposed to be a porin, but turned out to have a new and essential function in the adaptation of M. tuberculosis to low pH and in the survival of M. tuberculosis in macrophages and mice. The elucidation of the molecular mechanism by which M. tuberculosis counteracts excess protons in its environment via OmpATb is a most interesting future goal. Considering their ubiquituousness in Gram-negative bacteria, it would be interesting to analyse other bacterial OmpA-like proteins for such a function, e.g. OmpA of Salmonella, which shares with M. tuberculosis the capability to survive in macrophages and also must face environments with low pH at several stages of infection. It is surprising that the often stated assumption that porins limit the uptake and efficiency of TB drugs (Brennan and Nikaido, 1995; Lambert, 2002) has not been experimentally verified. To this end, the identification of the main porins of M. tuberculosis would be helpful.
The tetrameric porin of M. smegmatis MspA displays an entirely new porin fold, supporting the assumption that the unique mycobacterial outer membrane contains a wealth of proteins with new structures. Crystal structures of mycobacterial porins are now needed to understand the mechanisms of action and biochemical properties of these proteins.
One of the most urgent problems in the TB field is to find drugs against M. tuberculosis which can shorten TB therapy. Considering the extremely low permeability of mycobacterial outer membranes, it might be wise not to focus entirely on the identification of new drugs and drug targets, but to also consider the problem of drug transport, which is likely to render ineffective many new drugs discovered with in vitro assays. In this regard, understanding the synergism between inefficient transport through porins and drug efflux or inactivation is a high priority.
I wish to thank Drs Chuck Turnbough, Mary Jackson and Harald Engelhardt for valuable comments and the members of the Mycolab for excellent contributions and stimulating discussions. Work of the Mycolab on mycobacterial porins was funded by the Deutsche Forschungsgemeinschaft (NI 412), the Volkswagenstiftung (I/77 729) and the European Union (5th framework programme, contract QLK2-2000–01761). I gratefully acknowledge continuous support of my work by Dr Wolfgang Hillen.