Dr S.B. Levy Centre for Adaptation Genetics and Drug Resistance and the Departments of Molecular Biology and Microbiology and of Medicine, Tufts University School of Medicine, Boston, MA 02111, USA (e-mail: email@example.com).
Energy-driven drug efflux systems are increasingly recognized as mechanisms of antibiotic resistance. Chromosomally located or acquired by bacteria, they can either be activated by environmental signals or by a mutation in a regulatory gene. Two major categories exist: those systems energized by proton motive force and those dependent on ATP. The pumps may have limited or broad substrates, the so-called multiple drug resistance pumps, which themselves form a number of related families. The multiple antibiotic resistance (mar) locus and mar regulon in Escherichia coli and other members of the Enterobacteriaceae is a paradigm for a generalized response locus leading to increased expression of efflux pumps. One such pump, the AcrAB pump extrudes biocides such as triclosan, chlorhexidine and quaternary ammonium compounds as well as multiple antibiotics. In Pseudomonas aeruginosa, a number of multidrug efflux pumps export a broad range of substrates. Since bacteria expressing these pumps thwart the efficacy of both kinds of therapeutic agents which control infectious diseases – biocides which prevent transmission of infectious disease agents and antibiotics which treat and cure infectious diseases – they are of particular concern. The prudent use of antibiotics and biocides will guard against the selection and propagation of drug – resistant mutants and preserve the efficacy of these valuable anti-infective agents.
Bacteria respond to the widespread use of growth inhibitory agents, such as antibiotics, by emerging with progeny resistant to these substances. While some bacteria acquire resistance traits from other bacteria, many become resistant following mutations in the chromosomal gene coding for the target of the growth inhibitory compound. Increasingly recognized are energy-driven drug efflux systems intrinsic to bacteria which are either activated in response to environmental signals or by a mutation in a gene which regulates their expression. Regulatory loci, such as the multiple antibiotic resistance (mar) locus in E. coli and other members of the Enterobacteriaceae, are also subject to activation or mutation and affect the expression of genes which mediate multidrug resistance. When MarA of the locus is activated, either through mutation in the regulator MarR or in response to external stimuli, the expression of over 60 genes is altered, including the upregulation of the AcrAB multidrug efflux pump. The latter exports biocides (e.g. triclosan, chlorhexidine and QACs) as well as multiple antibiotics as do endogenous multidrug efflux pumps in Ps. aeruginosa. These kinds of pumps raise special public health concern since bacteria expressing them will resist biocides designed to prevent transmission of infectious disease agents and antibiotics aimed at curing infectious diseases.
Efflux as a mechanism of drug resistance has clearly come of age. Initially considered a curiosity when first described for the tetracyclines (McMurry et al. 1980), today a large number of integral membrane and membrane-associated proteins are involved in pumping antibiotics, biocides and other substances out of the microbial cell. Efflux pumps come in a variety of structures (Nikaido 1996; Paulsen et al. 1996). A single protein may act alone to perform efflux. Some of these single protein systems use ATP as an energy source, e.g. the Lmr protein in Lactobacillus (van Veen 1999). Other forms of a single polypeptide use proton motive force to energize transport of the drug. Tet proteins, which deal with tetracyclines, are prototype examples of proton motive force-dependent single polypeptide efflux pumps (Levy 1992; McMurry and Levy 2000). Alternatively, three-component systems, as exemplified by the MexAB/OmpM protein complex in Pseudomonas (Li et al. 1995) and AcrAB-TolC in E. coli (Fralich 1996), involve an integral membrane protein and an outer membrane porin connected by a cytoplasmic fusion protein.
The expression of all tetracycline resistance efflux systems described to date is regulated (Levy 1989; McMurry and Levy 2000). In fact, all variants of the TetA type Gram-negative efflux pump have a similar genetic organization, consisting of a repressor gene upstream and transcribed in the opposite direction to the gene for the structural efflux protein. The repressor acts in a classic negative fashion to prevent transcription of the efflux gene. Tet proteins appear to have evolved specifically to handle tetracyclines or very related molecules since the affinity of tetracyclines for the protein is high (Km 6–20 μmol l−1; McMurry et al. 1980; Yamaguchi 1990). In contrast, the multidrug efflux pumps, e.g. MexAB and AcrAB, are not saturable by the known substrates; the affinity for any of the substrates is very low.
3.1 Tetracycline-resistant bacteria
In the author's initial studies of tetracycline-resistant bacteria, tetracycline accumulation in tetracycline-resistant and -susceptible isogenic strains with and without energy was compared (McMurry et al. 1980). The drug was actively accumulated in susceptible cells only when energy (e.g. lactate) was supplied. When the cells were de-energized, in the presence of the ionophore dinitrophenol (DNP), accumulation was greatly reduced and represented that occurring by passive diffusion. Provided with lactate, the same bacterium, bearing a plasmid which specified resistance to tetracycline by an efflux pump, accumulated tetracycline at a level below that observed in the de-energized cell, i.e. that achieved by passive diffusion. This finding was the first indication that cells were actively keeping the drug out of the cell by some means. Before it could be called efflux, it had to be demonstrated that tetracycline was being actively removed from the cell. A system initially described by Barry Rosen to demonstrate calcium efflux was adapted (Tsuchiya and Rosen 1976). After the outer membrane was removed, E. coli were ruptured by passing them through a French pressure cell. In the process, the inner membrane was flipped inside out to produce everted inner membrane vesicles (McMurry et al. 1980). Radioactive tetracycline entered the inverted vesicles in the presence of lactate as an energy source and was assayed when the vesicles were trapped on filters. In this assay, the tetracycline being pumped into the vesicles actually represented what would normally be pumped out of whole cells. This technique demonstrated very clearly that, in the presence of an energy source, tetracycline accumulated in the everted membrane vesicles (to 300 μmol l−1), significantly above the external concentration (5 μmol l−1). When an energy inhibitor (such as DNP or carbonyl cyanide m-chlorophenol hydrazone (CCCP)) was added, the accumulated tetracycline came streaming out of the vesicles. This finding showed that the drug was not accumulating in the everted membrane vesicles because of precipitation, but was clearly being pumped into the vesicle; in the absence of energy, this accumulation did not occur. This transport could then be described as an `active efflux' system, the first described for an antibiotic (McMurry et al. 1980; Levy 1992). Further studies by the author's group (McMurry et al. 1980) and also by Yamaguchi and his group (Kaneko et al. 1985) demonstrated that this efflux by Tet proteins involved an antiport system exporting a tetracycline : cation complex in exchange for a proton in an electroneutral exchange (Fig. 1).
3.2 Tet proteins
Most Tet proteins in Gram-negative bacteria consist of a six-plus-six organization of 12 transmembrane regions. The author and colleagues demonstrated some time ago that the two halves of the protein were highly homologous and appeared to have arisen by gene duplication (Rubin et al. 1990). In fact, by looking at other members of the major facilitator family, a strong similarity can be seen between the alpha (N-terminal) and beta (C-terminal) halves suggesting, as for Tet protein, a common evolution by duplication of one domain and linkage of the two by a cytoplasmic string of amino acids. If either the alpha or beta domain of the Tet protein is deleted or mutated, resistance is not mediated, both domains being needed for resistance (Curiale and Levy 1982; Curiale et al. 1984). If inactive Tet proteins with a single mutation in either the alpha or beta domain are placed in the same bacterial cell, resistance is restored (Curiale et al. 1984). If an alpha domain from a Class C determinant (there are now over 30 classes of tetracycline resistance determinants) is fused with the beta domain from a Class B determinant, resistance is not achieved (Rubin and Levy 1990). There is a necessary `cross-talk' between the alpha and beta domains of the same class in order to provide resistance, i.e. the alpha and beta halves have evolved together to be able to interact. However, suppressor mutations in one or the other half of a TetC/B hybrid allow the domains to `talk' again, providing tetracycline resistance (Saraceni-Richards and Levy 2000a, 2000b). Such studies have suggested a model for the Tet protein. These findings complement much more comprehensive site-directed mutational studies by Yamaguchi and his group (Tamura et al. 2001).
Data were sought to understand how the Tet protein existed in the membrane. The amino acid sequence would suggest that there are 12 alpha-helical transmembrane structures. Tet protein was purified using a method that had previously been successful for cytoplasmic proteins. A polyhistidine tail was placed on the Tet protein and then purified on a nickel column. Greater than 90% pure Tet-his protein was obtained after washing the column with low concentrations of imidazole and then elution with high-dose imidazole. Circular dichroism (CD) spectral analysis in vitro was performed with the Tet protein preparation in dodecylmaltoside. The analysis at wavelengths of 180–260 and 200–260 nm provided information on the alpha-helical content of the purified protein (Aldema et al. 1996). These data correlated well with the theoretical value of alpha-helical structure and made it clear that the protein contained about 65% alpha-helical structure.
In other studies, binding to a nickel column was used to investigate the interaction between the alpha and beta domains as suggested by genetic studies (Curiale and Levy 1982; Curiale et al. 1984). The results obtained were not those expected. Purified alpha or beta domain polypeptides or full-length Tet protein were passed over nickel columns bearing either the alpha or beta domain. No Tet protein or half Tet protein bound to the column bearing the beta domain. In contrast, the column to which the alpha domain was attached retained alpha domain polypeptides or full-length Tet proteins but no beta domain polypeptide (McMurry and Levy 1995). This result allowed the design of a model for Tet as a dimer in the membrane based on alpha–alpha domain interaction (McMurry and Levy 1995). The structure agreed with prior genetic complementation data that implied a multimer structure (Hickman and Levy 1988). It was concluded that the most likely structure for the protein in the cell was a multimer, at least a dimer.
Current data suggest that the structure is more complicated. With collaborators in the UK, the author's group has obtained two-dimensional crystals of Tet protein in lipid membranes. Both multiple and single lipid-embedded crystals were analysed. A trimer picture emerged at 17 Å resolution, involving three alpha domains forming a central core with beta domains in the periphery (Yin et al. 2000). There are a number of related membrane proteins which confer tetracycline resistance bearing 12 or 14 transmembrane alpha-helical structures (Levy 1992). Besides Tet proteins, there are also those which mediate the efflux of chloramphenicol, ethidium bromide, norfloxacin, as well as 14 trans-membrane proteins which efflux tetracycline and others which efflux the QACs (Levy 1992). Each is energized by proton motive force. The Tet protein can serve as a model for understanding how these other related proteins function.
3.3 Efflux blocking
Work that began in the author's laboratory at Tufts University and is now being continued by Paratek Pharmaceuticals (Boston, MA, USA) aims to identify new tetracyclines which either block the efflux pump or get around it. If the ability of Tet protein to pump tetracyclines out of the cells could somehow be interrupted, the classic tetracyclines would be returned to power. Mark Nelson, in the author's group, began to design and synthesize new tetracyclines. In the initial work, very simple changes made on the molecule provided effective efflux blockers (Nelson and Levy 1999). One contained a 13-cyclopentyl on the sixth carbon of the tetracycline molecule. The studies used the everted vesicle assay to identify the blocking tetracycline derivatives. When the potential inhibitor was isotetracycline, there was no effect on tetracycline uptake (Fig. 2) but, with 13-cyclopentyl tetracycline (CPTC), there was significant inhibition of the tetracycline uptake at less than 1 μmol l−1. Further studies showed that CPTC competitively inhibited tetracycline uptake in everted vehicles (Levy and Nelson 1998). If everted vesicles were pretreated with 13-cyclopentyl, the efflux of tetracycline was inhibited producing a synergistic drug combination (Nelson and Levy 1999).
Using additional chemistry at the C13 position, Nelson was able to derive structure–activity relationships, providing a length and width relationship of a proposed hydrophobic pocket in the Tet protein (Levy and Nelson 1998). This information has helped to define potential efflux blockers as well as new tetracyclines which are not subject to efflux. Fortunately, these same tetracycline derivatives were not subject to the second mechanism of tetracycline resistance, ribosomal protection.
4. REGULATION OF MULTIDRUG EFFLUX
Some time ago, the author discovered a multiple antibiotic resistance (mar) operon was discovered in E. coli which consisted of genes for three proteins: a regulator MarR, a transcriptional activator MarA and a small protein MarB, for which a function is not yet known. A second transcript in the other direction which overlaps the operator-promoter site (marO) specifies a six transmembrane integral membrane protein designated MarC, the function of which is unknown (Alekshun and Levy 1997). The mar locus controls the cell's response to multiple different toxic substances. While named the `multiple antibiotic resistance' locus, it is probably best described as a `multiple adaptational response' locus because it deals not only with antibiotic resistance but also cell metabolism, DNA repair, superoxide stress and other physiological systems linked to cell danger (Barbosa and Levy 2000; Okusu et al. 1996). For antibiotic and biocide resistance, mar upregulates expression of the multidrug efflux pump AcrAB and its outer membrane TolC component, a functional homologue of the Mex pumps in Pseudomonas (Li et al. 1995). The mar locus, through MarA, increases expression of micF, an RNA molecule which causes reduced production of the principal outer membrane porin OmpF (Cohen et al. 1988). The MarR protein is affected by a number of structurally unrelated compounds (Seoane and Levy 1995). By examining the fold increase in LacZ activity when fused to marO in the presence of MarR, different inducers of the mar operon were identified, including uncouplers, such as CCCP and DNP, and redux-cycling agents, such as menadione. In collaboration with Lee Rosner at the National Institutes of Health, salicylate and acetometaphen were found to be inducers (Cohen et al. 1993). Many of the inducers can directly affect MarR in vitro (Alekshun and Levy 1999). When the cell goes into a so-called `Mar state', numerous changes occur. Quinolones lose their bacteriocidal activity and become bacteriostatic agents at relatively low levels of drug resistance: two- to four-fold resistance to multiple drugs (Goldman et al. 1996). Mar does not initially provide high-level resistance; it is often the first step to higher levels of resistance, secondary to drug target mutations (Oethinger et al. 1998). In collaboration with colleagues at Boston University, the author's group resolved the crystal structure of a dimer of MarR at 2·3 Å (Alekshun et al. 2001). It has a winged helix form with two likely DNA-binding domains. The protein is now being crystallized with different substrates as well as with DNA to determine its substrate-binding site and to confirm the putative DNA-binding sites. There are over 50 known homologues of MarR, one of which is MexR which controls the MexAB efflux system in Ps. aeruginosa (Li et al. 1995; Nikaido 1996).
The author's group recently published an examination of the mar regulon using macroassays (Barbosa and Levy 2000). Overexpression of MarA affected the expression of more than 60 genes. In addition to genes whose functions were known, the work identified genes for which no function has been assigned. Some are conserved among other bacterial genomes. While of unknown function, they presumably have something to do with stress because they are affected by this stress locus. Through the MarA activity, it was further noted that combining decreased outer membrane permeability (by decreasing porins) with indigenous efflux systems produced resistance much higher than expected from diffusion or efflux alone (Levy 1990). The combination of outer membrane impermeability and inner membrane efflux enhanced the effect of the two changes in drug transport. If resistance is linked to a cytoplasmic target mutation, such as mutated topoisomerase for fluoroquinolone resistance, efflux amplifies the effect of the mutation (Kern et al. 2000).
5. RESISTANCE TO BIOCIDES
The recognition by the author's group of the relationship between household products and antibiotic resistance initially emerged from work with Merri Moken, a New Jersey high school student. She selected E. coli mutants resistant to pine oils and noted that they were also resistant to antibiotics. She shared her unexpected findings and it was determined that the mutants constitutively expressed MarA (Moken et al. 1997). The mutations were in MarR which allowed MarA expression and activation of the multidrug efflux pump AcrAB, which was able to efflux pine oils as well as antibiotics.
While removing the mar locus did not affect triclosan resistance, overexpression of the MarA protein via a MarR mutation led to a three- to four-fold increased triclosan resistance – exactly as occurs with antibiotics (McMurry et al. 1998a). The same phenomenon is seen with a SoxR mutation leading to overexpression of the related SoxS protein. Similarly, overexpression of the AcrAB efflux pump itself will provide resistance (Wang et al. 2001). Mar is able to turn on efflux systems, and perhaps other mechanisms, for drug resistance.
Given the two types of effect of triclosan on bacterial cells, bacteriostatic at low and bacteriocidal at high levels, the effect of the AcrAB pump on these two activities was investigated (Levy 2001). At low levels, triclosan inhibited growth in the wild type cell (50% at 0·15 μg ml−1; up to 90% at 0·6 μg ml−1); the amount causing lysis (bactericidal) was 8 μg ml−1. When the acrAB gene complex, and thus the constitutive expression of the AcrAB pump, was removed the growth inhibitory level of triclosan was reduced 10-fold in the wild type cell. Deletion of the pump also affected cell lysis; the amount of triclosan needed to cause lysis dropped from 8 to 3–4 μg ml−1. This result was unexpected since lysis had been considered a chemically-modified membrane disruption.
In one of the high level triclosan E. coli mutants with a mutation in the FabI protein (the target for triclosan (McMurry et al. 1998b)) 50% growth inhibition was at 13 μg ml−1, a 100-fold increase over the wild type cell. It was not possible to find sufficiently soluble triclosan, above 32 μg ml−1, to give 90% inhibition. Similarly, lysis did not occur within the solubility limits of triclosan. When the AcrAB pump was eliminated, growth inhibition dropped by 10-fold with the MIC 90% at 2·1 μg ml−1. Remarkably, the amount needed for cell lysis became the same as that for the wild type cell, i.e. 3–4 μg ml−1. These data strongly imply that triclosan does not cause lysis by a non-specific chemical reaction; it has to get into the cell and its action is impeded by the AcrAB efflux pump. The pump somehow protects the cell by keeping the drug from another target independent of FabI. If the multidrug resistance pumps can be inhibited, the activity of triclosan and other drugs against E. coli and other bacteria can be greatly increased.
In summary, the MarA protein of the mar operon upregulates AcrAB and the TolC outer membrane protein to produce resistance to antibiotics, organic solvents (e.g. cyclohexane) pine oils, bile salts and antiseptics and disinfectants, such as triclosan, QACs and chlorhexidine (Fig. 3). This regulatory protein protects the cell from biocides and antibiotics through at least one known mechanism – antibiotic efflux pumps.
As with antibiotics, given enough time and dose, biocide resistance will emerge. The continued effect of the biocides and the increasing number of resistance determinants (target gene or otherwise) will produce widespread biocide resistance. If biocides are spread all over households, diluted down to less than their growth inhibitory concentrations, bacteria with resistance mutations will emerge (McMurry et al. 1998b). The same phenomenon has occurred for antibiotics, as illustrated by the newly emerging resistance to vancomycin among Staphylococcus aureus and to fluoroquinolones among E. coli. Some of the resistance mechanisms, such as efflux, can provide cross-resistance to other drugs. As urged in the use of antibiotics, prudent use of antibacterial agents, namely those which leave residues, should be followed in order to curb resistance and preserve efficacy of these products when needed to protect the vulnerable patient.