INTRODUCTION

Top Previous Next References

The development and use of antibiotics has been one of the most important steps towards controlling infectious diseases in the 20th century. However, the subsequent appearance and spread of antibiotic resistance in pathogenic organisms have made many currently available antibiotics ineffective (160, 169). As a consequence, tuberculosis and other infectious diseases are reemerging, causing a serious public health problem (32, 38). To successfully fight the increasing numbers of drug-resistant and multidrug-resistant (MDR) bacteria, extensive knowledge of the molecular mechanisms underlying microbial antibiotic resistance is required.

Microorganisms have developed various ways to resist the toxic effects of antibiotics and other drugs (Fig. 1) (83, 169). One of these mechanisms involves the production of enzymes that inactivate antibiotics by hydrolysis or the formation of inactive derivatives (Fig. 1A) (40). Well-known examples are -lactamases (29, 206) and enzymes that phosphorylate, adenylate, or acetylate aminoglycoside antibiotics (234). A second mechanism of resistance is target alteration (Fig. 1B). Cellular targets can be altered by mutation or enzymatic modification in such a way that the affinity of the antibiotic for the target is reduced (95, 239, 265, 277). A third, more general mechanism of resistance is the inhibition of drug entry into the cell (Fig. 1C). Due to the low permeability of the outer membrane of gram-negative bacteria (175, 176) and the exceptionally efficient barrier of the gram-positive mycobacteria (101), drug diffusion across the cell envelope is reduced. The permeability of the outer membrane can be further decreased by the loss of porins (77, 157, 177). These barriers, however, cannot prevent the drugs from exerting their toxic action once they have entered the cell, and the active efflux of drugs (Fig. 1D) is essential to ensure significant levels of drug resistance (122, 177).

View larger version (63K): [in this window] [in a new window]   FIG. 1.   Resistance mechanisms in bacteria comprise (A) drug inactivation, (B) target alteration, (C) prevention of drug influx, and (D) active extrusion of drug from the cell.

Some transporters, such as the tetracycline efflux proteins (209, 237), are dedicated systems which mediate the extrusion of a given drug or class of drugs. In contrast to these specific drug transporters, the so-called multidrug transporters can handle a wide variety of structurally unrelated compounds (127, 178, 190, 257). On the basis of bioenergetic and structural criteria, multidrug transporters can be divided into two major classes (Fig. 2). Secondary multidrug transporters utilize the transmembrane electrochemical gradient of protons or sodium ions to drive the extrusion of drugs from the cell. ATP-binding cassette (ABC)-type multidrug transporters use the free energy of ATP hydrolysis to pump drugs out of the cell.

View larger version (43K): [in this window] [in a new window]   FIG. 2.   Schematic representation of the two major classes of multidrug transporters. (A) ABC-type multidrug transporters utilize the free energy of ATP hydrolysis to pump drugs out of the cell. (B) Secondary multidrug transporters mediate the extrusion of structurally unrelated drugs in a coupled exchange with protons or sodium ions.

Here we present a comprehensive review describing the currently known multidrug transporters in bacteria. Important issues addressed are (i) the functional reconstitution of purified multidrug transporters into proteoliposomes, which allows detailed molecular characterization of the proteins, (ii) the transcriptional regulation of genes encoding multidrug efflux systems, (iii) the structure-function relationships that dictate drug recognition and transport, (iv) the presence of multiple drug-binding sites, and (v) the mechanism of transport of both ATP-driven and secondary multidrug transporters.

Most bacterial multidrug efflux systems known to date are sensitive to agents that dissipate the proton motive force (PMF), indicating that they mediate the extrusion of toxic compounds from the cells in a coupled exchange with protons (15, 42, 84, 133, 190). On the basis of size and similarities in the primary and secondary structure, these secondary multidrug transporters can be subdivided into distinct families of transport proteins: the major facilitator superfamily (MFS) (146), the small multidrug resistance (SMR) family (191), the resistance-nodulation-cell division (RND) family (216), and the multidrug and toxic compound extrusion (MATE) family (28). These families are not solely associated with multidrug export but include proteins involved in other PMF-dependent transport processes or other functions.

The MFS consists of membrane transport proteins which are found from bacteria to higher eukaryotes and are involved in the symport, antiport, or uniport of various substrates, such as sugars, Krebs cycle intermediates, phosphate esters, oligosaccharides, and antibiotics (146). Hydropathy analysis and alignment of conserved motifs of the resistance-conferring drug efflux proteins revealed that these proteins can be divided into two separate clusters, with either 12 or 14 transmembrane segments (TMS) (Fig. 3 and 4) (188). Table 1 summarizes the multidrug transporters from both the 12-TMS and 14-TMS clusters found in gram-positive and gram-negative bacteria.

View larger version (30K): [in this window] [in a new window]   FIG. 3.   Structural model for the 12-TMS multidrug transporters of the MFS. The residues constituting the conserved sequence motifs are shaded.

View larger version (29K): [in this window] [in a new window]   FIG. 4.   Structural model for the 14-TMS multidrug transporters of the MFS. The residues constituting the conserved sequence motifs are shaded.

The 12-TMS cluster of multidrug transporters. The Staphylococcus aureus NorA protein was first discovered in a quinolone- and methicillin-resistant clinical isolate (251). Sequencing and characterization of the norA gene revealed that NorA is a 388-amino-acid membrane protein that confers resistance to hydrophilic compounds and only low or no resistance to hydrophobic drugs (271, 278). Studies on the substrate specificity demonstrated that NorA is a true multidrug transporter, mediating resistance to a range of structurally dissimilar drugs (171, 172).

The 14-TMS cluster of multidrug transporters. Resistance to antiseptic and disinfectant compounds, including intercalating dyes, quarternary ammonium compounds, and diamines, is widespread among multidrug-resistant strains of Staphylococcus aureus. One of the energy-dependent export mechanisms responsible for this resistance is encoded by the qacA gene (248). Hydropathy analysis (213) and topological analysis using alkaline phosphate and -galactosidase fusions (192) indicated that the QacA protein contains 14 TMS. Whereas QacA confers resistance to monovalent and divalent organic cations, the closely related QacB only confers resistance to monovalent organic cations (192). It has been speculated that the extensive use in hospital environments of bacteriostatic divalent cations, such as chlorhexidine, resulted in the evolution of qacA from qacB (193).

Multidrug transporters of the SMR family (Table 2, Fig. 5), the smallest secondary drug efflux proteins known, are typically about 107 amino acid residues in length. Hydropathy analysis (191), gene fusions with alkaline phosphatase and -galactosidase (179), -periodicity analysis (51), cysteine accessibility scanning (242), and Fourier transform infrared spectroscopy (11) support a model of a tightly packed four-helix antiparallel bundle. Due to the small size of the multidrug transporters of the SMR family, it has been proposed that they may function as homooligomeric complexes (187, 189).

View larger version (34K): [in this window] [in a new window]   FIG. 5.   Structural model for multidrug transporters of the SMR family. The residues constituting the conserved sequence motifs are shaded.

The first gene encoding a multidrug transporter of the SMR family was detected on both conjugative and nonconjugative plasmids from clinical isolates of S. aureus and other staphylococci (120, 121, 132). This gene, first described as qacC and also known as qacD (132) and ebr (221), has been renamed smr (71). Grinius and Goldberg (72) demonstrated that the purified and reconstituted Smr protein mediates electrogenic drug/proton antiport. In staphylococci isolated from the food industry, plasmids pST94 and p2H6 were found. Sequence analysis of these plasmids revealed two new members of the SMR family, designated QacG and QacH (84, 85). QacG and QacH have 69 and 76% identity to Smr, respectively. Both transporters confer resistance to ethidium bromide and quaternary ammonium compounds.

The E. coli multidrug transporter EmrE (also known as MvrC) has been identified and cloned on the basis of its ability to confer resistance to ethidium and methyl viologen (162, 202). Yerushalmi et al. (275) demonstrated that resistance is caused by PMF-dependent extrusion of drugs. Negative dominance studies indicated that EmrE functions as a homooligomer, probably composed of three monomers (276). Studies using single cysteine mutants of EmrE showed that none of the residues in the putative transmembrane helices react with N-ethyl-maleimide (NEM), a small membrane-permeant sulfhydryl reagent. As the reaction with NEM requires water, the results imply very tight packing of the protein without a continuous aqueous domain. Based on these findings, it was proposed that EmrE translocates its substrates through a hydrophobic pathway (242).

The antiseptic resistance genes qacE and qacE1 are located on an integron, a potentially mobile element. As a consequence, these genes are widely spread among gram-negative bacteria (108). The qacE1 gene but not the qacE gene was detected in clinical isolates of gram-positive bacteria (109). The qacE gene was originally found on the Klebsiella aerogenes plasmid R751 (187). The qacE1 gene probably represents a disrupted form of qacE that evolved as a result of the insertion of a DNA segment near the 3' end of the qacE gene. Ethidium transport experiments indicated that the mechanism of resistance mediated by QacE is active export driven by the PMF. Neither QacE nor QacE1 is inhibited by reserpine, a potent inhibitor of multidrug transporters such as Bmr, NorA, and LmrP (187).

The multidrug transporter Mmr of M. tuberculosis confers resistance to tetraphenyl phosphonium (TPP⁺), ethidium bromide, erythromycin, acriflavine, safarin O, and pyronin Y (42). TPP⁺ accumulation experiments showed that Mmr actively extrudes TPP⁺, using the PMF as the driving force. The presence of mmr-like genes in other Mycobacterium species (M. simiae, M. gordonae, M. marinum, and M. bovis) was demonstrated by Southern hybridization (42).

The Bacillus subtilis genome encodes seven SMR-type proteins (194). Surprisingly, six of these proteins are encoded from gene pairs in three distinct operons, ebrAB, ykkCD, and yvdRS. Masaoka et al. (154) demonstrated that neither EbrA nor EbrB is sufficient for conferring drug resistance. However, coexpression of the two proteins causes an MDR phenotype in both E. coli and B. subtilis. A similar result was obtained for the YkkCD pair. Only when both the ykkC and ykkD genes were expressed together was resistance observed against a broad range of drugs, including cationic dyes and neutral and anionic antibiotics (97). In contrast to the SMR-type multidrug transporters, which appear to function as homooligomers, EbrAB and YkkCD are composed of two dissimilar but homologous subunits. One member of each pair is short (105 to 106 amino acids), while the other is longer (111 to 117 amino acids). This difference is due to a C-terminal hydrophilic extension (97).

The tehA gene of E. coli, which was initially identified as part of an operon associated with resistance to potassium tellurite, encodes a 36-kDa protein with 10 TMS (247). Surprisingly, a region of TehA which includes TMS 2 through 5 was found to be homologous to members of the SMR family (250). Expression of TehA increased the resistance to monovalent cations, such as tetraphenylarsonium and ethidium bromide, but decreased the resistance to the divalent cations dequalinium and methyl viologen (250). Thus, like MdfA of E. coli, TehA confers resistance to one group of toxic compounds and hypersensitivity to another group. Since deletion of the C-terminal region of TehA (TMS 8 to 10) did not decrease transport significantly, TMS 2 to 5 may be primarily responsible for the activity of TehA (250). Although TehA contains only one of the three SMR signature sequences (191) and is larger than the reported SMR-type multidrug transporters, it may represent a distantly related member of the SMR family.

Multidrug transporters belonging to the RND family (Table 3) interact with a membrane fusion protein (MFP) and an outer membrane protein to allow drug transport across both the inner and outer membrane of gram-negative bacteria. The secondary structure of RND-type efflux proteins (Fig. 6) was proposed to consists of 12 TMS, with two large loops between TMS 1 and 2 and TMS 7 and 8 (190, 216). Membrane topology analysis of the Pseudomonas aeruginosa RND protein MexB recently verified this model (75). The MFP proteins, which contain a single N-terminal TMS and a large C-terminal periplasmic domain, are thought to induce fusion of the inner and outer membrane (45, 216) or form a channel-like structure that spans the periplasmic space (281).

View larger version (61K): [in this window] [in a new window]   FIG. 6.   Structural model for multidrug transporters of the RND family. The residues constituting the conserved sequence motifs are shaded.

The acrA locus of E. coli has long been known to be involved in resistance to basic dyes, detergents, and antibiotics (166, 167). Characterization of this locus revealed the presence of two genes, acrA and acrB (formerly acrE), which are located within a single operon (141). AcrA is a highly asymmetric protein that belongs to the MFP family, while AcrB is a member of the RND family (45, 281). Insertion mutations showed that both AcrA and AcrB are required for drug resistance (143). A study by Fralick (58) suggested that the outer membrane protein TolC is the third component of the AcrAB efflux system. Recently, the 2.1-Å crystal structure of TolC has been determined (118). Three TolC promoters assemble to form a continuous, solvent-accessible "tunnel" that spans the outer membrane and periplasmic space. This tunnel may transiently open up upon the interaction with an energized inner membrane transporter and establish a conduit between the cytosol and the external environment. E. coli contains two AcrAB homologs, AcrEF (formerly known as EnvCD [110]) and YhiUV (formerly called OrfAB), as well as an AcrB homolog, AcrD, which is not associated with a gene for an MFP (142). Inactivation of AcrEF decreased the resistance to a broad range of drugs, including dyes, detergents, and various classes of clinically important antibiotics, indicating that AcrEF functions as a multidrug transporter (141). Mutations in the yhiUV genes result in drug hypersusceptibility, suggesting that YhiUV also plays a role in antibiotic resistance (142, 143). Direct multidrug transport, however, has not yet been demonstrated.

P. aeruginosa is a clinically important opportunistic pathogen characterized by relatively high intrinsic resistance to a variety of antimicrobial agents. This property is now recognized to result from the synergy between low outer membrane permeability and at least four RND efflux systems. The mexAB-oprM operon was first thought to play a role in the export of the siderophore pyoverdine (198). Further characterization of the operon, however, showed that overexpression of MexAB-OprM increased the resistance of P. aeruginosa to various antibiotics (199). A direct assay of drug accumulation showed that MexAB-OprM indeed pumps out various antibacterial agents in an energy-dependent fashion (129). MexA and MexB are members of the MFP and RND families, respectively. The third gene of the operon was originally identified as oprK, but Gotoh et al. (64) demonstrated that the outer membrane component of this efflux system is OprM rather than OprK. Results obtained by Evans et al. (52) suggested that the quorum-sensing-related homoserine lactone PAI-1 is a substrate for MexAB-OprM and that exclusion of PAI-1 limits the production of virulence factors, such as the blue-green pigment pyocyanin. Examination of P. aeruginosa strains showing widely different levels of intrinsic resistance suggested the presence of more than one endogenous multidrug efflux system (128). Indeed, Poole and coworkers (200) identified a second efflux operon in P. aeruginosa, designated MexCD-OprJ. The third antibiotic efflux system of P. aeruginosa is composed of the MFP MexE, the RND protein MexF, and the outer membrane protein OprN (115). Similar to MexAB-OprM, MexEF-OprN expression correlates inversely with the production of the virulence factor pyocyanin (115). In contrast to MexAB-OprM, MexCD-OprJ and MexEF-OprN do not confer resistance to -lactam antibiotics (65, 115). Subunit swapping experiments demonstrated that the inner membrane efflux components of the MexAB-OprM transporter are responsible for the -lactam specificity of the multidrug efflux pumps in P. aeruginosa (66, 155, 240). Recently, two new genes mediating resistance to quinolones, aminoglycosides, and macrolide antibiotics were cloned from the chromosome of P. aeruginosa (6, 158). These genes, designated mexX and mexY, show significant similarity to mexAB, mexCD, and mexEF. No open reading frame corresponding to an outer membrane protein was found downstream of mexXY. Mine et al. (158) demonstrated that TolC or OprM is required for MexXY activity in E. coli. It has been reported that OprM contributes to antibiotic resistance in P. aeruginosa independent of MexAB (283). The substrate specificity of this OprM-dependent and MexAB-independent system corresponds to that of the MexXY system, suggesting that OprM functions as the outer membrane protein of the MexXY drug efflux pump in P. aeruginosa (158).

The resistance of Neisseria gonorrhoeae to hydrophobic agents, including detergent-like fatty acids and bile salts, is conferred by the MtrRCDE efflux system. The mtrRCDE region of the gonococcal chromosome contains a transcriptional regulator gene (mtrR) and three tandemly linked genes (mtrC, mtrD, and mtrE), encoding proteins of the MFP, RND, and outer membrane protein families, respectively (80). Insertional inactivation of either the mtrC (80) or mtrE (41) gene resulted in a significant decrease in resistance to erythromycin, penicillin, Triton X-100, and fatty acids, suggesting that both proteins are required in the efflux process.

Genome sequencing of Haemophilus influenzae has identified a three-gene cluster, consisting of HI0893, HI0894, and HI0895, which is homologous to the acrRAB cluster of E. coli (57). Disruption of either HI0894 (an acrA homolog) or HI0895 (an acrB homolog) caused hypersusceptibility to erythromycin, rifampin, novobiocin, sodium dodecyl sulfate, and cationic dyes. Erythromycin accumulation experiments indicated that the HI0894 and HI0895 proteins contribute to the resistance of H. influenzae through PMF-dependent drug efflux (220).

Burkholderia pseudomallei, a motile gram-negative rod, is the causative agent of melioidosis. Successful treatment of melioidosis patients is difficult because B. pseudomallei is intrinsically resistant to a variety of antibiotics, including -lactams, aminoglycosides, and macrolides. Recently, the amrAB-OprA operon was identified, which encodes a multidrug efflux system in B. pseudomallei which confers resistance to both aminoglycoside and macrolide antibiotics. AmrA and AmrB show strong homology to MexC and MexB, respectively, of P. aeruginosa (161).

Norfloxacin accumulation experiments revealed that Vibrio parahaemolyticus contains an energy-dependent efflux system, designated NorM (163). NorM and its E. coli homolog YdhE mediate resistance to dyes, hydrophilic fluoroquinolones, and aminoglycosides. NorM contains 12 putative TMS, and on this basis it has been suggested to be a member of the MFS (163). NorM and YdhE, however, do not have sequence similarity with any members of the MFS and do not exhibit any of the signature sequences (see below). Brown et al. (28) found that NorM and YdhE are members of a previously unidentified family which contains more than 30 proteins, termed the multidrug and toxic compound extrusion (MATE) family. Phylogenetic analysis of the MATE family revealed the presence of three distinct clusters, of which the first includes NorM, YdhE, and hypothetical multidrug efflux proteins from H. influenzae and B. subtilis (28).

Drug transport by multidrug transporters of the MFS, SMR, RND, and MATE families is driven by the transmembrane electrochemical gradient of protons or possibly sodium ions. The PMF is composed of a chemical proton gradient (pH, inside alkaline) and an electrical potential (, inside negative). The ionophores nigericin and valinomycin, which can selectively dissipate the pH and , respectively, are valuable tools for the study of the energetics of secondary multidrug transporters. Ng et al. (174) demonstrated that NorA-mediated norfloxacin uptake in inside-out membrane vesicles was completely abolished by nigericin. In contrast, addition of valinomycin resulted in a slight increase in the uptake of norfloxacin. These results indicate that the pH is the major driving force for NorA. Grinius et al. (73) reported that NorA-mediated drug transport in proteoliposomes is an electrogenic drug/proton antiport process. Transport experiments in the presence of valinomycin and/or nigericin showed that both the pH and the are driving forces of LmrP-mediated transport of trimethyl ammonium-diphenyl hexatriene (TMA-DPH) (22) and Hoechst 33342 (203), consistent with an electrogenic exchange of one lipophilic cation for two or more protons. Mitchell et al. (159) investigated which component of the PMF is involved in energizing QacA-mediated drug efflux. It was found that both the pH and the drive the electrogenic transport of monovalent and divalent cations by QacA. The involvement of both components of the PMF as the driving force in Smr-mediated drug transport was demonstrated by studies using purified Smr reconstituted into proteoliposomes. A (positive inside) superimposed on pH (acidic inside) resulted in additional ethidium uptake via Smr (72). An electrogenic drug/proton antiport reaction by Smr was further supported by the observation that Smr increased the resistance of E. coli cells to ethidium in alkaline medium, when the PMF predominantly consists of a (72). The electrogenic drug/nH⁺ (n  2) antiport mechanism observed for LmrP, NorA, QacA, and Smr appears to be a general feature of multidrug transporters of the MFS and SMR families and may also be valid for the RND- and MATE-type multidrug transporters. The substrates used in the studies on the energetics of secondary multidrug transporters are monovalent cations. However, several multidrug transporters, such as QacA, MdfA, and YkkCD, also mediate resistance to divalent cations, neutral drugs, and anionic compounds. An analysis of the relative contribution of the and the pH to the driving force for transport of these cations can provide further insight into the drug/proton ratio of the export process.

Multiple-sequence analysis of transporters within the various clusters of solute and drug transporters of the MFS revealed a substantially larger sequence similarity in the N-terminal halves of these proteins than in their C-terminal halves (70, 146, 188, 213). It has been hypothesized that the N-terminal domain is involved in proton translocation, while the C-terminal domain of the transporters is primarily involved in determining the substrate specificity (70, 213). On the other hand, the significant sequence similarity observed between the N- and C-terminal halves of transporters of the MFS suggests that they have evolved from a duplication of a common ancestor with six TMS (122, 145, 215, 274). The 14-TMS cluster of the MFS may have evolved via a similar gene duplication event and the acquisition of two additional TMS (70, 188, 217). The homology between the first and second halves of RND proteins suggests that these proteins also may have arisen via an intragenic duplication event during evolution (75, 216).

The multiple alignments of amino acid sequences further revealed a number of strongly conserved amino acid sequence motifs (Table 4 and Fig. 3, 4, 5, and 6) throughout the multidrug transporters of the MFS, SMR, and RND families (188, 190, 191, 213, 216). The conservation of these motifs suggests that they play an important structural or functional role in the transporters. Motif A in the cytoplasmic loop between TMS 2 and 3 in transporters of the MFS, which is predicted to contain a -turn structure, may be involved in the reversible conformational change required for the opening and closing of a transport channel (102, 188, 272, 273). The absolute conservation of the basic arginine residue suggests that motif B in TMS 4 of MFS transporters may be involved in proton transfer (188). Interestingly, motif C in TMS 5 is found in multidrug transporters and specific drug efflux pumps, but not in symporters of the MFS. This suggests that motif C dictates the direction of transport (70, 188). The GP dipeptide of this motif causes a bend in helix 5, and the repeating pattern of glycine residues forms a pocket devoid of side chains. These structures may determine the orientation of the unoccupied substrate-binding site(s) and, consequently, the direction of transport (261). The C-terminal motifs F and G represent partial duplications of motif C and may have a similar function. Site-directed mutagenesis studies can provide valuable information on the structure and mechanism of multidrug transporters of the MFS, RND, and SMR families. The conserved motifs may be directly involved in drug or proton translocation. The charged residue in motif B (Arg) and in motif D1 (Asp) in members of the MFS, in motif A (Glu) in SMR-type multidrug transporters, and in motif C (two Arg) in drug efflux systems of the RND family are predicted to be located in TMS. The presence of positively or negatively charged amino acid residues in the membrane is energetically unfavorable, suggesting that these motifs play a functional role in the binding and transport of charged drugs and/or translocation of protons. Alternatively, they may play a structural role in the interaction between the transporters of the RND family (motif A) and proteins of the MFP family, or in the formation of homo- and heterooligomers of SMR-type multidrug transporters. The conserved sequences are characteristic of proteins of the MFS, SMR, and RND families and can be used to identify new members of these families in bacterial genomes.

The secondary multidrug transporters identified in bacteria (Tables 1 to 3) belong to one of four distinct families of proteins, MFS, SMR, RND, and MATE. Whereas multidrug transporters of the MFS and SMR families have been found in both gram-positive and gram-negative bacteria, RND- and MATE-type multidrug transporters have so far only been identified in gram-negative organisms. Table 5 indicates that bacteria can contain several multidrug transporters from different families. The presence of a variety of multidrug transporters with an overlapping substrate spectrum would seem redundant. However, the availability of an array of multidrug transporters, which are controlled by both global and specific regulatory proteins, enables the bacterium to fine tune the response to different environmental signals.

The most striking difference between multidrug transporters of the SMR family and drug efflux systems from the other families is their size. Whereas members of the MFS, RND, and MATE families consist of 12 or 14 TMS, SMR-type multidrug transporters only contain 4 TMS. These small multidrug efflux pumps are thought to function as homooligomers, presumably composed of three monomers. However, two new SMR-type multidrug transporters were recently identified in B. subtilis which require two homologous but dissimilar subunits for drug efflux. This observation raises the question of what molecular determinants dictate whether an SMR transporter functions as a homo- or heterooligomer.

Secondary multidrug transporters extrude drugs in exchange with protons (or sodium ions). Both components of the PMF can drive the extrusion of drugs by multidrug transporters from the MFS, SMR, and RND families, indicating that an electrogenic antiport mechanism is a general feature of secondary multidrug transporters.

Multiple sequence alignments of transporters within one family revealed the presence of three or more conserved motifs in multidrug transporters of the MFS, SMR, and RND families (Table 4). Hydropathy analysis, the recognition of family-specific conserved motifs, and the availability of the complete genomic sequences of 28 bacteria and 8 archaea (http://www.tigr.org/tdb/mdb) enable the identification of putative multidrug transporters. However, as multidrug transporters are frequently more homologous to substrate-specific transporters than to other multidrug transporters, the activity of newly identified proteins needs to be characterized.

Although most bacterial multidrug drug transporters utilize the PMF (or sodium) for the extrusion of cytotoxic compounds, some drug efflux systems are driven by the free energy of ATP hydrolysis (Fig. 7). All ATP-dependent drug efflux proteins known to date are members of the ABC superfamily (87), also referred to as traffic ATPases (8). In general, ABC transporters require four distinct domains: two highly hydrophobic membrane domains, which usually consist of six putative transmembrane -helices each, and two hydrophilic nucleotide-binding domains (NBDs), containing the Walker A and B motifs (263) and the ABC signature (96). The individual domains can be expressed as separate proteins or may be fused into multidomain polypeptides in a variety of ways (55, 87, 96).

View larger version (37K): [in this window] [in a new window]   FIG. 7.   Structural model for multidrug transporters of the ABC superfamily. Shown is the multidrug transporter LmrA, with six transmembrane helices and an NBD, containing the Walker A and B motifs and the ABC signature sequence. In view of the general four-domain organization, it may function as a homodimer. The Walker A and B motifs and the ABC signature sequence are shaded.

Most bacterial ABC drug transporters mediate the export of specific antibiotics (17, 55, 76, 130, 183, 197, 210, 212). The first true bacterial ABC multidrug transporter was found in Lactococcus lactis (20). The gene encoding this transporter, designated lmrA, encodes a 590-amino-acid membrane protein with an N-terminal hydrophobic domain with six putative transmembrane helices and a C-terminal hydrophilic domain, containing the ATP-binding cassette (256). LmrA is homologous to each of the two halves of the human multidrug transporter P-glycoprotein (256), suggesting that LmrA is functional as a homodimer. This suggestion was confirmed by studies on covalently linked dimers composed of wild-type LmrA units (designated KK), mutant LmrA units containing an inactivated NBD (designated MM), and combinations of wild-type and mutant units (designated KM and MK) (258). The ethidium transport activity and substrate-stimulated ATPase activity of the KK dimeric form of LmrA was comparable to that of wild-type LmrA. However, the KM, MK, and MM mutant forms of dimeric LmrA had lost all LmrA-associated ATPase activity and the ability to transport ethidium from the cell. These results are supported by reconstitution experiments which demonstrate the negative dominance of M units over K units in proteoliposomes and suggest that homodimeric LmrA is the minimal functional unit. Hence, functional crosstalk between two LmrA monomers is essential for activity (258).

When expressed in human lung fibroblast cells, LmrA is targeted to the membrane and confers multidrug resistance on these human cells (257). The pharmacological characteristics of LmrA and the human multidrug transporter P-glycoprotein (67) and the affinities of both proteins for vinblastine and Mg²⁺-ATP are very similar. Blockers of P-glycoprotein also inhibited LmrA-mediated resistance. These results demonstrate that LmrA and the human multidrug transporter P-glycoprotein are functionally interchangeable, indicating that this type of multidrug transporter is conserved from bacteria to humans (258).

Kinetic analysis of vinblastine dissociation from LmrA and characterization of transport of Hoechst 33342 or vinblastine in the presence of low concentrations of a second substrate revealed that LmrA contains at least two allosterically linked drug-binding sites with overlapping substrate specificities (see below) (257, 258). Purified LmrA reconstituted in proteoliposomes mediates ATP-dependent transport of Hoechst 33342 (147). Certain ABC transporters, including the LmrA homolog MsbA of E. coli (285), are able to translocate lipids. The specificity of reconstituted LmrA for C₆-NBD-labeled phosphatidylethanolamine and phosphatidylcholine was investigated. Interestingly, LmrA catalyzes the transport of C₆-NBD-labeled phosphatidylethanolamine but not of C₆-NBD-labeled phosphatidylcholine. These results indicate that LmrA exhibits specificity for phospholipid headgroups and that LmrA may be involved in the translocation of specific lipid or lipid-linked precursors in L. lactis (147).

Hop-resistant lactobacilli can cause serious spoilage of beer. Sami et al. (218, 219) demonstrated that plasmid pRH45 of Lactobacillus brevis is involved in the high resistance to hop compounds (iso--acids). pRH45 contains the horA gene, which encodes a protein of the ABC superfamily. HorA is a structural homolog of LmrA, containing six putative TMS and a C-terminal ATP-binding domain. In addition to conferring hop resistance, HorA confers resistance to the structurally unrelated drugs novobiocin and ethidium (219). HorA has been heterologously expressed in Lactococcus lactis, purified, and functionally reconstituted in proteoliposomes. HorA-mediated Hoechst 33342 transport in proteoliposomes is inhibited by hop compounds (K. Sakamoto et al., submitted for publication).

Kaidoh et al. (105) described an ATP-dependent drug efflux system in the archaeon Haloferax volcanii. This efflux system mediates the transport of doxorubicin, vinblastine, vincristine, ethidium, and monensin. Expression of the transporter is induced in media containing a large excess of essential amino acids, glucose, or fructose. These observations suggest that this efflux system plays a role in the active extrusion of cytotoxic compounds or toxic metabolites generated in the cell. The gene encoding this ABC-type multidrug transporters remains to be cloned and sequenced (105). Genome sequencing revealed the presence of putative LmrA homologs in E. coli, B. subtilis, and the pathogens Helicobacter pylori, Mycobacterium genitalium, H. influenzae, and S. aureus (259). Functional characterization of these proteins is required to demonstrate their involvement in the ATP-dependent transport of structurally unrelated drugs.

Mutant analysis and photoaffinity labeling experiments have been used to identify essential residues and putative substrate-binding sites in the human multidrug transporter P-glycoprotein. The results demonstrated that, although some mutations leading to changes in specificity are located in other regions of the protein, TMS 5, 6, 11, and 12 and the loop between TMS 2 and 3 and TMS 11 and 12 are most important in drug binding and/or transport (39, 43, 68, 69, 82, 106, 137, 139, 164, 282). Interestingly, sequence conservation between the lactococcal multidrug transporter LmrA and P-glycoprotein includes parts of TMS 5 and 6 and the loop between TMS 2 and 3 (256), indicating that these particular regions may be important for substrate binding in both human and bacterial ABC-type multidrug transporters.

Several studies demonstrated that the expression levels of multidrug transporters such as Bmr, NorA, QacA, and P-glycoprotein can be enhanced by the addition of structurally diverse compounds which are exported by these pumps (3, 74, 104, 107). These observations indicate that the expression of multidrug transporters is controlled by regulatory proteins which, like the transporters, are capable of recognizing structurally diverse compounds. The activator and repressor proteins involved in the transcriptional regulation of multidrug transporter genes are summarized in Table 6.

The best-studied regulator of multidrug transport is BmrR, the transcriptional activator of the B. subtilis multidrug transporter Bmr. BmrR, which is encoded by a gene immediately downstream of bmr, belongs to the MerR family of regulators and has strong homology with other members of the family only in the N-terminal helix-turn-helix DNA-binding domain. Binding of one rhodamine 6G molecule to a dimer of BmrR increases the affinity of BmrR for the bmr promoter, resulting in enhanced transcription of the bmr gene (3). The individually expressed C-terminal region of BmrR forms dimers and binds a broad range of drugs with the same affinity as full-length BmrR (148, 150). High-resolution crystal structures of the BmrR C-terminal region and its complex with TPP⁺ revealed drug-induced unfolding and relocation of an -helix, which exposes an internal drug-binding pocket. Residue Glu-134, which is completely buried in the unliganded structure, is the key to cation selectivity, while a number of nonpolar and aromatic side chains impose specific requirements on the size and shape of the substrate (284). The transcription of the blt gene is regulated by another member of the MerR family, BltR. The C-terminal domains of BmrR and BltR have no sequence similarity, which is consistent with the observation that Bmr and Blt are expressed in response to different inducers. This difference in inducer specificity may explain why Bmr and not Blt is normally expressed in wild-type B. subtilis (4).

The expression of the S. aureus multidrug efflux gene qacA is regulated by the divergently transcribed repressor protein QacR. QacR binds specifically to the qacA promoter, preventing transcription of the gene. Substrates of QacA, such as ethidium and rhodamine 6G, interact with QacR directly and inhibit the binding of QacR to the qacA promoter, resulting in derepression of qacA expression (74).

The EmrAB multidrug pump of E. coli is induced in the presence of CCCP, the weak acid salicylate, and a number of other structurally unrelated drugs. The derepression is controlled by the EmrR, a MarR type of repressor protein (136). EmrR was found to bind putative substrates of the EmrAB pump, such as 2,4-dinitrophenol, CCCP, and carbonyl cyanide p-(trifluoro-methoxy)-phenylhydrazone (FCCP), with micromolar affinity. Equilibrium dialysis experiments suggested one bound ligand per EmrR dimer (27).

Analysis of the sequence upstream of the mtrCDE multidrug efflux system of N. gonorrhoeae revealed the presence of the divergently transcribed repressor gene mtrR. MtrR binds to the 13-bp inverted repeat sequence between the mtrR and mtrC promoters, thereby inhibiting expression of the mtrRCDE gene complex (80, 185). Studies by Hagman and Shafer (79), however, showed that the MtrCDE efflux system is subject to both MtrR-dependent and MtrR-independent regulation. Recently, the transcriptional activator MtrA was shown to be involved in the modulation of mtrCDE gene expression (214).

The transcription of the acrAB operon in E. coli is regulated by the repressor AcrR (144). Gel mobility shift assays and lacZ transcriptional fusions suggested that the general-stress-enhanced transcription of acrAB (144) is primarily mediated by global regulatory pathways such as the mar regulon (see below) and that a major function of AcrR is that of a specific secondary modulator, which fine tunes the level of acrAB transcription (144).

The transcriptional repressor AmrR of Burkholderia pseudomallei shows strong similarity to the regulator MtrR. The expression of the AmrAB-OprA efflux system was not affected by the substrate streptomycin or by stress conditions (4% ethanol or 0.5% NaCl), and further studies are required to identify the inducers of this system (161).

The expression of the P. aeruginosa multidrug transporter MexAB-OprM is regulated by MexR, a MarR-type regulator (53, 201). The transcriptional repressor NfxB regulates a second MDR in P. aeruginosa, MexCD-OprJ, as well as its own expression (200). The mexEF-oprN efflux operon differs from the other P. aeruginosa efflux systems in that it is positively regulated by MexT, a protein belonging to the LysR family of transcriptional activators (115, 116). A third MDR in P. aeruginosa, MexXY, may be regulated by MexZ, which is homologous to the repressor proteins MtrR and AmrR (6).

In addition to the specific regulatory mechanisms that affect the expression of single multidrug transporters, induction of multidrug efflux systems can also result from global regulatory mechanisms. Multiple-antibiotic resistance (Mar) mutants of E. coli express elevated levels of resistance to a wide range of antibiotics. The alleles that affect the Mar phenotype are located in the marRAB operon (34). Expression of this operon is normally repressed by the transcriptional repressor MarR (9, 225) but can be derepressed by various compounds, such as tetracycline, chloramphenicol, and salicylates (35, 78). The marA gene encodes an AraC-type activator that is structurally and functionally similar to two other global regulators of E. coli, SoxS and Rob (60, 235, 270). The MarA, SoxS, and Rob proteins are transcriptional activators of at least a dozen promoters, the mar and soxRS regulons, that confer resistance to a variety of antibiotics and superoxide-generating agents (10, 62, 99, 100, 226). The observed resistance is partly due to increased micF transcription and the consequent decrease in expression of the outer membrane porin OmpF (33, 60). The RND-type multidrug transporter AcrAB-TolC plays a major role in the antibiotic and organic solvent resistance induced by MarA, SoxS, or Rob (182, 266). The mar operon is widespread among enteric bacteria, including the genera Salmonella, Shigella, Klebsiella, Citrobacter, Hafnia, and Enterobacter (36). Expression of E. coli MarA in Mycobacterium smegmatis resulted in increased resistance to multiple antimicrobial agents, suggesting that MarA functions in M. smegmatis and that a mar-like regulatory system exists in mycobacteria (156). Recently, Baranova et al. (16) identified a global MerR-type regulator in B. subtilis, termed Mta. The individually expressed N-terminal DNA-binding domain of Mta mimics the inducer-bound form and interacts directly with the promoters of bmr and blt, thereby inducing transcription of these genes. Additionally, this domain stimulates the expression of the mta gene and at least one more gene, ydfK, which encodes a hypothetical membrane protein (16).

Multidrug transporters have predominantly been studied in whole cells and membrane vesicles, in which other proteins and cellular components or processes may complicate the interpretation of the data. In order to elucidate the structure, function, and mechanism of multidrug transporters at the molecular level, methods have been developed for overexpression, purification, and functional reconstitution into liposomes (72, 73, 147, 203, 227, 275, 280). Although the overexpression of membrane proteins is often deleterious for the cells (48, 119, 203, 243), drug transporters could be overexpressed up to 5 to 30% of total membrane protein using homologous expression systems and inducible promoters, such as the T7 polymerase promoter (275), the lac promoter (236), and the nisA promoter (147, 203). The first step in the purification of membrane proteins involves extraction of the protein from the membrane. The E. coli multidrug transporter EmrE can be quantitatively and functionally extracted with a mixture of organic solvents, such as chloroform-methanol (275). However, solubilization of most other membrane proteins requires the use of detergents (115). The choice of detergent is crucial, since some detergents can irreversibly inactivate the protein (227; J. Knol, R. H. E. Friesen, and B. Poolman, submitted for publication). In addition, detergents are potential substrates for multidrug efflux proteins and can affect the transport activity (46, 203, 268). Retention of detergent during the reconstitution procedure is a serious problem which further restricts the choice of detergent for the solubilization and reconstitution of multidrug transporters (115). Affinity chromatography, a highly specific and efficient technique, allows the purification of several multidrug transporters to a high degree of purity (>90%) in a single step (72, 147, 203, 227, 280). After reconstitution in liposomes, Smr, P-glycoprotein, EmrE, NorA, LmrA, LmrP, AcrB, and HorA mediated the transport of multiple drugs in response to the imposition of an artificial proton gradient or the addition of ATP (72, 73, 147, 203, 227, 275, 280; K. Sakamoto, H. W. van Veen, and W. N. Konings, unpublished data). Taken together, these results demonstrate that multidrug transporters of the MFS, SMR, RND, and ABC families are capable of mediating drug transport independent of accessory proteins or cytosolic components.

Both secondary and ATP-dependent multidrug transporters extrude a wide variety of toxic compounds. Although the substrates have very different structures, they share physical characteristics, such as high hydrophobicity, an amphiphilic nature, and a positive or neutral charge. In the absence of three-dimensional structural information, the identification of structure-function relationships in multidrug transporters has relied on amino acid sequence alignments and site-directed mutagenesis. The localization of a charged residue in the membrane-embedded part of the protein is normally considered energetically unfavorable, yet many multidrug transporters of the MFS and SMR families, such as MdfA, QacA, EmrE, and Smr, contain an acidic amino acid residue in the middle of TMS 1 (50). Furthermore, Guan et al. (75) identified three highly conserved negatively charged amino acid residues in transmembrane domains of efflux proteins of the RND family. Since the majority of the MDR substrates are positively charged, the negative charge of an intramembranous acidic residue may be involved in substrate binding. The role of the conserved Glu in TMS 1 of Smr has been investigated using site-directed mutagenesis. Even conservative substitutions, such as the replacement of Glu with Asp, effectively abolished the activity of the efflux system, indicating that this residue may be directly involved in the multidrug efflux process, potentially in substrate binding and/or the exchange of drug molecules for protons (72). Edgar and Bibi (50) demonstrated that a single membrane-embedded negative charge is critical for the recognition of positively charged drugs by the multidrug transporter MdfA of E. coli but not for transport activity. Replacement of the Glu residue in TMS 1 of EmrE abolished binding activity, but replacement with a negatively charged Asp created a mutant protein that retained its ability to bind substrate (165). Whereas QacA transports both monovalent and divalent cations, the closely related multidrug transporter QacB only transports monovalent cations. Sequence analysis and mutagenesis revealed that the presence of an acidic residue in TMS 10 (position 322 or 323) of these transporters is crucial for the recognition of divalent cations (192). Altogether, these results suggest that charge-charge interactions play an important role in the recognition and export of lipophilic cations by multidrug transporters.

Based on the observation that aromatic amino acid residues in transmembrane regions of the human multidrug transporter P-glycoprotein are highly conserved, Pawagi et al. (195) hypothesized that the side chains of these residues can participate in the initial binding and subsequent transport of a variety of ring-containing compounds. Indeed, Dougherty (47) indicated that cation- interactions between cations, such as quarternary ammonium compounds, and aromatic side chains from the amino acids Phe, Tyr, and Trp are important in a variety of proteins that bind cationic substrates. The potential role of the conserved aromatic residues Tyr⁵⁹ and Trp⁶² in the staphylococcal multidrug transporter Smr has been investigated by site-directed mutagenesis (189). Substitutions at these positions abolished the multidrug efflux capacities of the protein, indicating that these residues play a critical structural or functional role. Mutations of residues Phe¹⁴³, Val²⁸⁶, and Phe³⁰⁶ of Bmr altered the sensitivity to reserpine and changed the substrate specificity, indicating that these residues are likely involved in substrate recognition (2, 111). Furthermore, mutations of Phe, Trp, and Tyr residues in TMS 6, 11, and 12 drastically altered the drug resistance profile of P-glycoprotein (82, 137). These results illustrate the importance of aromatic and nonpolar amino acid residues in drug binding.

The three-dimensional structure of the multidrug-binding domain of BmrR, the transcriptional activator of Bmr, supports a model in which a negatively charged amino acid residue as well as aromatic and nonpolar residues is involved in drug binding (284). The binding of TPP⁺ to Bmr induces the displacement of a short amphipatic -helix to allow the drug to enter the drug-binding pocket. Inside the binding pocket, TPP⁺ forms a number of van der Waals' forces and stacking interactions with the surrounding hydrophobic and aromatic residues. The bottom of the binding pocket contains a buried glutamate residue, which makes an electrostatic interaction with the positively charged substrate (284). The substrate-binding site(s) of a multidrug transporter may have a similar organization. Most secondary multidrug transporters contain an essential membrane-embedded negatively charged amino acid residue. This residue could act like a magnet to attract positively charged substrates from the inner leaflet of the membrane. The surrounding hydrophobic residues would impose requirements on the size and shape of the substrate. In contrast to secondary multidrug transporters, the ABC multidrug transporters LmrA, P-glycoprotein, and others do not contain negatively charged residues in predicted TMS. It should also be noted that the binding of neutral or negatively charged substrates cannot be explained by the model described above.

There is increasing evidence for the presence of multiple drug-binding sites in both secondary and ATP-dependent multidrug transporters, rather than a single binding site, as suggested previously (126, 195). Tamai and Safa (245) indicated for the first time that the human multidrug transporter P-glycoprotein may contain at least two kinetically distinguishable drug-binding sites. Kinetic analysis of the drug-stimulated ATPase activity of P-glycoprotein showed that cyclosporin A competitively inhibits the verapamil-stimulated ATPase activity, whereas daunorubicin, gramicidin D, vinblastine, and colchicine give an allosteric inhibition. Interestingly, cooperative stimulation of the verapamil-induced ATPase activity was found with progesterone (131). Several other studies demonstrated a similar complex effect of pairs of substrates or modulators on the drug-induced ATPase activity of P-glycoprotein (30, 61, 153, 186). Noncompetitive inhibition of drug binding (44, 56, 153, 264) and drug transport by P-glycoprotein (13, 196, 230, 238) further support a model with multiple cooperating drug interaction sites.

Noncompetitive inhibition of rhodamine 6G transport by drugs such as quinidine, puromycin, and colchicine also suggested multiple cooperating drug interaction sites in the yeast multidrug transporter Pdr5p (117). Van Veen et al. (257) demonstrated that nicardipin interacts with the lactococcal multidrug transporter LmrA at a binding site distinct from but allosterically linked to the vinca alkaloid-binding site. Competition studies showed that QacA-mediated transport of ethidium is inhibited competitively by the monovalent cations benzalkonium and tetraphenylarsonium and noncompetitively by the divalent cations chlorhexidine and propamide. These observations also suggest that the S. aureus multidrug transporter QacA contains distinct binding sites for monovalent and divalent cations (159). Furthermore, kinetic analysis of LmrP-mediated Hoechst 33342 transport in inside-out membrane vesicles revealed a competitive inhibition by verapamil and quinine, a noncompetitive inhibition by nicardipin and vinblastine, and a noncompetitive inhibition by TPP⁺, implying that LmrP must contain at least two drug interaction sites (204). These results indicate that the presence of two or more interacting substrate-binding sites may be a general feature of multidrug efflux systems, including those of the SMR and RND families.

The common ability of MDR substrates to intercalate in the membrane led to the suggestion that multidrug transporters transport their substrates from the lipid bilayer (88, 207, 252) rather than from the cytoplasmic aqueous phase (7). Drug recognition within the membrane is suggested by a number of observations: (i) site-directed mutagenesis has identified specific residues implicated in substrate specificity, which are predominantly located within predicted TMS (2, 50, 65, 103, 184, 192), (ii) photoactive substrates label the human multidrug transporter P-glycoprotein in or near transmembrane helices (69, 207), and (iii) nonfluorescent acetoxymethyl esters of various fluorescent indicators, such as 2',7'-bis-(2-carboxymethyl)-5-(and 6)-carboxy-fluorescein (BCECF) and calcein, are extruded from the membrane before they can be converted into fluorescent carboxylates by nonspecific esterases in the cytoplasm (23, 89, 91).

The most convincing evidence for drug efflux from the membrane to the aqueous phase, however, is provided by transport studies using the fluorescent compounds TMA-DPH, Hoechst 33342, and 2-{4-[4-(dimethylamino)phenyl]-1,3-butadienyl}-3-ethylbenzo-thiazolium perchlorate (LDS-751) which are strongly fluorescent in a hydrophobic environment but essentially nonfluorescent in an aqueous environment. Transport of these compounds from the membrane to the aqueous phase can be monitored as a decrease in fluorescence over time (22, 23, 180, 228, 229, 231). Shapiro et al. (229) demonstrated that the initial rate of P-glycoprotein-mediated Hoechst 33342 transport is directly proportional to the amount of Hoechst 33342 in the membrane, demonstrating that P-glycoprotein removes the drug from the lipid membrane. The mechanism of transport of LmrA (23), LmrP (22), MexAB-OprM (180), and QacA (159) was studied with TMA-DPH. TMA-DPH displays a biphasic interaction with bacterial membranes; a fast partitioning of the probe in the outer leaflet of the phospholipid bilayer, followed by a slower transbilayer movement of the probe from the outer to the inner leaflet (23, 159). The initial rate of TMA-DPH extrusion by the lactococcal multidrug transporter LmrA increases with the concentration of TMA-DPH in the inner leaflet, demonstrating that LmrA transports drugs from the inner leaflet of the membrane (23). A similar relationship between the initial rate of transport and the concentration of substrate in the inner leaflet of the membrane was observed for LmrP (22) and P-glycoprotein (229, 231). Also, the MDR transporters MexAB-OprM and QacA are capable of strongly reducing the accumulation of TMA-DPH in the inner leaflet of the bilayer (159, 180). Altogether, these results demonstrate that the multidrug transporters of the MFS, RND, and ABC families extrude toxic compounds from the inner leaflet of the cytoplasmic membrane to the aqueous extracellular medium (Fig. 8). Thus, despite differences in energy coupling, ATP- and PMF-dependent multidrug efflux pumps display a general mechanism of transport for hydrophobic drugs.

View larger version (31K): [in this window] [in a new window]   FIG. 8.   Extrusion of hydrophobic drugs by multidrug transporters. Drugs are expelled from the cytoplasmic leaflet of the membrane to the external medium.

Drug transport from the inner leaflet of the cytoplasmic membrane to the extracellular aqueous medium requires a high-affinity substrate-binding site that is accessible from the inner leaflet of the phospholipid bilayer. This high-affinity site is converted to a low-affinity