Efflux Pumps and Nosocomial Antibiotic Resistance: A Primer for Hospital Epidemiologists

Bacterial resistance to antimicrobial agents, including multidrug resistance, is an increasing problem in health care in both community and hospital settings. Bacterial adaptability in circumventing antimicrobial action stems from an array of mutational resistance mechanisms and from the acquisition of resistance genes that are present on plasmids that can transfer between bacteria within and among species. Resistance is mediated by 1 or more mechanisms from 3 general categories: alteration or protection of the antimicrobial target, alteration of drug access to the target, and/or inactivation of the antimicrobial. Although there are multiple, often interesting variations of these mechanisms, resistance results from 1 or a combination of mechanisms from these 3 categories. Understanding of the resistance caused by alteration of drug access to its target, the subject of this review, has recently progressed [1–4]. The eubacterial cell envelope is composed of a phospholipid bilayer constituting the cytoplasmic membrane, the peptidoglycan cell wall, and, in gram-negative bacteria, an additional outer membrane. Peptidoglycan is thought to provide little barrier to penetration of small molecules, such as most antimicrobial agents. These agents, with a few exceptions, appear to cross the cytoplasmic membrane by simple diffusion [5]. In the case of β-lactams, glycopeptides, and lipopeptides, the targets of which are either external to or at the outer-surface of the cytoplasmic membrane, drug action does not require diffusion across the membrane, but most other agents in clinical use must cross the membrane to reach their intracellular drug targets. The outer membrane in gram-negative bacteria is composed of an asymmetric bilayer of phospholipid and lipopolysaccharide that is poorly permeable to both hydrophobic and hydrophilic molecules. Porin proteins located in the outer membrane serve as general diffusion channels allowing permeation of hydrophilic small molecules, including many antimicrobial agents [6]. Reduced uptake by diffusion. Bacterial resistance to antimicrobials could, in theory, occur by slower diffusion from reduced numbers of porin channels, but seldom is such reduced diffusion itself rate-limiting in relation to bacterial doubling times [7]. Thus, susceptibility and steady-state levels of drug accumulation appear to be little affected by reduced numbers of porin channels alone. Reduced diffusion becomes important, however, when combined with additional mechanisms of resistance. For example, cefepime, in contrast to ceftriaxone or cefotaxime, is sufficiently stable with respect to the action of the chromosomally encoded AmpC β-lactamase ofEnterobacter species that AmpC overexpression, which causes resistance to ceftriaxone and cefotaxime, generally fails to cause clinically important resistance to cefepime. Cefepime-resistant clinical isolates ofEnterobacter have been found, however, that both overexpress AmpC and have reduced numbers of porin channels [8]. Thus, reduced diffusion of cefepime allows the weak action of AmpC to reduce steady-state levels of cefepime enough to cause clinical resistance. Similarly, reduced diffusion can also act in concert with endogenous efflux pumps, which actively remove antimicrobials from the cell, to reduce steady-state levels of antimicrobials and increase resistance. As with some β-lactamases, some efflux pumps are also sufficiently active on their own to cause resistance when overexpressed. Active specific drug efflux. The first efflux pumps to be recognized as causing antimicrobial resistance were the Tet pumps, which specifically pump tetracyclines and are encoded by genes that can be found on the chromosome or plasmids in a broad array of both gram-negative and gram-positive bacteria [9]. Members of the tetracycline class differ in the extent to which they are substrates of Tet pumps, and the related glycylcycline derivatives were developed for their ability to avoid the actions of Tet pumps and other specific tetracycline-resistance mechanisms [10, 11]. Specific efflux of macrolides has also been recognized inStreptococcus pneumoniae and other streptococci. The macrolide eflux (Mef) pumps, which are encoded by chromosomal genes in some strains of pneumococci, generally do not affect susceptibility to clindamycin or streptogramins, drugs that are affected by the other major mechanism of macrolide resistance caused by modification of the ribosomal target by methylase enzymes (erythromycin methylase; Erm) [12]. Mef-related resistance is limited to the 14-membered ring macrolides (e.g., erythromycin, clarithromycin, and dirithromycin) and the 15-membered ring macrolides (e.g., azithromycin), and is at a lower level than Erm-mediated resistance. The relative prevalence of the 2 mechanisms differs between the United States, where Mef-mediated resistance predominates, and other areas of the world, where Erm mechanisms dominate [13, 14]. The reasons for these differences are not known. In staphylococci, resistance to macrolides (only those with 14- and 15-membered rings) and to type B streptogramins can be caused by the MsrA pump encoded on plasmids [15]. Active multidrug efflux. Most if not all free-living bacteria produce an array of efflux pumps, many of which have broad substrate profiles, including antimicrobials of different structural classes. The normal functions of these pumps are uncertain in many cases, but it appears that some of these pumps allow survival in environments that may contain toxic substances and remove hydrophobic molecules that can accumulate in the cytoplasmic membrane. For example, the AcrAB-TolC pump ofEscherichia coli allows growth and survival in the presence of bile salts, and bile salts induce expression of the pump [16]. Thus, AcrAB-TolC is likely important for the survival ofE. coli in the gastrointestinal tract. That antimicrobials are substrates of these pumps likely reflects an evolutionarily developed protection against antimicrobials present in natural environments. Interestingly, however, synthetic classes of antimicrobials can also be within the broad substrate spectrum of these pumps. The evolution of the large number of efflux pumps is incompletely understood, but drug-resistance pumps appear to have emerged relatively few times in evolutionary history, with drug specificities and diversification suggesting that drug-specific and multidrug-resistance (MDR) efflux pumps may have evolved from one another in both directions over time [17]. A common but not exclusive property of MDR pump substrates is molecules that are both lipophilic and cationic. But even this generalization has some notable exceptions; for example, hydrophilic, amphipathic quinolones, such as norfloxacin and ciprofloxacin, and hydrophilic aminoglycosides have been found to be substrates for some MDR pumps. Although acquired resistance by active efflux (discussed in the next section) generally results from increased expression of MDR pumps, some MDR pumps are expressed at sufficient levels in wild-type, nonmutant bacteria to affect their intrinsic susceptibility. Two clinically relevant examples include the following:P. aeruginosa, which generates cells with a susceptibility profile similar to that of wild-typeE. coli when its MexAB-OprM pump is inactivated [18]; andE. coli, which produces cells that are susceptible to linezolid, a drug ordinarily active only against gram-positive bacteria, when its AcrAB-TolC pump is inactivated [19]. MDR efflux pumps are located in the cytoplasmic membrane in both gram-positive and gram-negative bacteria, and drug substrates appear to be taken up from the inner leaflet of the membrane and possibly also from the cytoplasm and pumped outward [20]. In the case of gram-negative bacteria, the membrane pump is also associated with an outer membrane protein and a membrane fusion protein, which current structural models suggest serves to link the pump itself with the outer membrane protein, thereby forming a pathway to the exterior of the cell (figure 1). MDR pumps are energized by either the proton gradient across the membrane(s) or, in some cases, by ATP hydrolysis that is usually mediated by an ATPase subunit linked to the pump itself [21]. Schematic diagram structures of efflux pumps. Qac and Smr are representative of pumps found in gram-positive bacteria. EmrAB and MexAB-OprM represent pumps found in gram-negative bacteria. P-glycoprotein (Pgp), which is found in mammalian cells, represents an ATP-binding cassette superfamily of pumps. ADP, adenosine diphosphate; Pi, inorganic phosphate. Adapted from [65] with permission. Mechanisms of acquisition of resistance. In many cases, the expression of the genes encoding efflux pumps is regulated and often inducible by pump substrates. Genes encoding efflux pumps located on plasmids can be acquired by any bacteria in which the plasmid can replicate, and are not present in all members of a bacterial species. Plasmid-encoded inducible efflux is a common mechanism for tetracycline resistance in both gram-negative and gram-positive bacteria [11]. As another example, in staphylococci, plasmid-encodedqac genes mediate resistance to the commonly used quarternary ammonium and related cationic disinfectants and other compounds [22]. Expression of Qac pumps can be induced by cationic disinfectants [23]. Chromosomally encoded efflux pump genes are often present in most, if not all, members of a species, although there are exceptions (e.g.,mefA of streptococci). In both cases, increases in gene expression and the amounts of the efflux pump are affected either by substrate induction, environmental conditions, or mutations in regulatory genes. Often, multiple chromosomally encoded efflux pumps occur in a given bacterial species (table 1). Their expression is often regulated, although in many cases the regulatory genes and conditions affecting expression are incompletely understood. In gram-negative bacteria, 2 of the best studied MDR-pump systems are AcrAB-TolC ofE. coli and the several Mex-Opr pumps ofP. aeruginosa [2]. AcrAB-TolC is expressed in wild-type bacteria, and its expression underlies the resistance to quinolones, tetracycline, and chloramphenicol that is found in strains with the multiply antibiotic resistant (Mar) mutation [24]. Efflux pumps mediating antimicrobial resistance identified in common human pathogens. The best studied chromosomally encoded pump in pathogenic gram-positive bacteria is NorA inStaphylococcus aureus. NorA overexpression in some resistant mutants reduces susceptibility to chloramphenicol and to quinolones, such as norfloxacin and ciprofloxacin, but not to moxifloxacin [25, 26]. NorA expression has a complex regulation involving a 2-component sensor regulator and MgrA, which regulates NorA and other efflux pumps and expression of capsular polysaccharides, autolysins, and virulence factors [26–28]. Thus, efflux-pump expression is affected by global regulators of cellular physiology. InStreptococcus pneumoniae, the PmrA pump also contributes to reduced susceptibility to quinolones [29]. Clinical importance of efflux-mediated resistance. Defining the extent to which efflux pumps contribute to antimicrobial resistances in clinical isolates can be difficult. In the case of Tet, Qac, and other plasmid-encoded pumps, the presence of the pump-encoding genes themselves may allow tracking of efflux-mediated resistance in an epidemiologically interpretable manner, as has been done with outbreaks of infection with tetracycline-resistant strains ofE. coli O157 H7 [30] and resistantShigella species [31]. Thus, an outbreak of infection associated with efflux-mediated resistance can be ascribed to the spread of plasmid-containing strains that can be specifically distinguished from their susceptible counterparts. Tracking of tetracycline resistance alone, however, is insufficient to ascribe resistance to an efflux mechanism, because other acquiredtet genes that can mediate resistance by ribosomal protection and many native MDR pumps also cause tetracycline resistance when overexpressed [9]. Nevertheless,tet genes encoding Tet pumps are widespread in bacterial isolates and have been the dominant tetracycline-resistance determinants in clinical isolates of resistantE. coli and other enteric bacteria [11]. In the case of native, chromosomally encoded MDR efflux pumps, attribution of resistance to the increased expression of these pumps alone can be difficult in clinical isolates, in which other mechanisms may contribute to resistance. Even multidrug resistance is insufficient to define MDR efflux-mediated resistance, because resistant clinical isolates can have mutational-resistance mechanisms unrelated to efflux or have acquired plasmids containing multiple resistance determinants. For this reason, assessment of efflux-mediated resistance has used comparisons of drug susceptibilities in the presence and absence of pump inhibitors that lack intrinsic antibacterial activity. To infer on the basis of these analyses that acquired resistance results from efflux requires that the difference in susceptibilities in the presence and absence of the pump inhibitor be greater in resistant strains than in susceptible wild-type strains, which may also exhibit lower levels of pump expression. Other assessments have compared the expression of pump genes or proteins between resistant and susceptible isolates. Levels of pump expression, however, are not discrete, and classification of the amount of increase that is sufficient to cause resistance can be ambiguous and can vary between studies. Interpretation is also limited for putative efflux pumps whose ability to cause resistance has not been validated in the laboratory. In the case ofP. aeruginosa, most clinical isolates, including those susceptible and resistant to fluoroquinolones, show substantial changes in susceptibility to levofloxacin in the presence of MC-207110, an inhibitor of all 5 known MDR-efflux pumps in this organism: the MIC50 of levofloxacin was reduced from 0.5 to 0.03 g/mL, and the MIC90 was reduced from 8 to 0.5 g/mL [32]. These findings indicate that the active efflux likely contributes to resistance in most isolates. Increased resistance due to the increased expression of the MexAB-OprM pump, which is also expressed in wild-type cells, and to the upregulation of other pumps, such as MexCD-OprJ and MexEF-OprN, which are not normally expressed, have been found to contribute to resistance in clinical isolates, particularly those from respiratory specimens obtained from patients with cystic fibrosis [33]. Cross-resistance among fluoroquinolones and other antimicrobials has been associated with an increased expression of MexAB (conferring resistance to ciprofloxacin, nalidixic acid, tetracycline, and chloramphenicol), MexCD (conferring resistance to fluoroquinolones, erythromycin, trimethoprim, and triclosan), MexEF (conferring resistance to fluoroquinolones, chloramphenicol, trimethoprim, imipenem, and triclosan), and MexXY (conferring resistance to fluoroquinolones, erythromycin, and aminoglycosides). Thus, MDR phenotypes could potentially be selected by exposure to a broad range of antimicrobials ofP. aeruginosa. Particularly noteworthy is a study of 21 pairs of clinical isolates collected from patients before and after therapy with antipseudomonal β-lactams. Among isolates obtained after therapy, but not among those obtained before therapy, about half of the isolates exhibited increased expression of the MexAB-OprM pump [34], suggesting that selection of strains ofP. aeruginosa with efflux-pump mutations can occur readily in patients. InE. coli and, similarly, in several other enteric gram-negative bacteria [24, 35], there is a principal, chromosomally encoded MDR pump, AcrAB-TolC. The magnitude of efflux-mediated resistance inE. coli is less than that inP. aeruginosa; in part, this is because of the greater permeability of the outer membrane ofE. coli, compared with that ofP. aeruginosa, that thereby reduces the net effect of efflux. Central regulatory mechanisms involving themar, sox, androb regulons mediate resistance by mutations that coordinately increase expression ofacrAB and reduce the number of porin channels [36]. Strains with a mar mutation that overexpress AcrA have been found among fluoroquinolone-resistant clinical isolates ofE. coli, but often also have other quinolone-resistance mutations that alter the topoisomerase target enzymes [37, 38]. In addition, over 60% of clinical isolates with high-level ciprofloxacin resistance had increased expression of AcrA [39, 40]. Mar homologues have also been found in other enteric bacteria, implying that mechanisms similar to those inE. coli could contribute to resistance in these species as well. MDR efflux pumps have also been characterized in drug-resistant clinical strains ofNeisseria gonorrhoeae [41, 42],Campylobacter jejuni [43, 44],Stenotrophomonas maltophilia [45], andAcinetobacter baumanii [46] (table 1). In gram-positive bacteria, the contribution of efflux pumps to antimicrobial resistance in clinical isolates is less clear. In the absence of the outer membrane present in gram-negative bacteria, which can limit drug diffusion and act in concert with efflux pumps, antibiotic resistance due to efflux in gram-positive bacteria requires pumps that are themselves sufficient to reduce cellular drug levels in order to contribute to a resistance phenotype. InS. aureus, only the NorA and Qac MDR pumps have been evaluated in clinical isolates. For NorA, the presence of mutations in thenorA promoter region and the extent to which reserpine reduced norfloxacin susceptibility correlated poorly with levels of quinolone resistance [47, 48], possibly because of the common presence of resistance mutations in the genes encoding quinolone target-enzymes and because the presence of other efflux pumps in addition to NorA could contribute to resistance [26, 49]. In the laboratory, selection of chromosomal quinolone resistance mutations occurs less readily in the presence of reserpine, suggesting that normal levels of reserpine-inhibitable pumps facilitate selection of resistant mutants [50]. In one study, a NorA variant (in which aspartate replaced glycine at position 291) was found in 14 of 28 norfloxacin-resistant isolates and in 0 of 14 norfloxacin-susceptible isolates, suggesting that this variant might contribute to resistance in clinical isolates [47]. Plasmid-encodedqacA orqacC genes encoding resistance to cationic disinfectants have been found in 10% and 20% of clinical MRSA isolates in 1 study [51]. A number of other pumps in staphylococci also encode resistance to these disinfectants (table 1), but the extent to which they compromise disinfectant action in the clinical environment is unknown. ForS. pneumoniae, the role of the PmrA pump in quinolone susceptibility was identified on the basis of increased susceptibility when thepmrA gene was disrupted [29]. Reserpine inhibits PmrA and increases quinolone susceptibility 2 times or more in 45% of clinical isolates [52]. Correlations of increasedpmrA expression and increased effects of reserpine on quinolone susceptibility with the levels of quinolone resistance in clinical isolates ofS. pneumoniae, however, have been poor [53, 54], likely for reasons similar to those causing poor correlations betweennorA expression and quinolone resistance levels inS. aureus. In bothS. pneumoniae andS. aureus, identification of the full array of relevant chromosomal efflux pumps and assessment of their levels of expression will likely provide a more complete view of the role of efflux pumps in antimicrobial resistance in these organisms. Interestingly, it appears that an efflux pump present inE nterococcus faecalis but notEnterococcus faecium may account, at least in part, for the species-specific resistance ofE. faecalis to quinupristin-dalfopristin and clindamycin [55]. The utility of regimens that combine antimicrobials with an agent that specifically blocks resistance mechanisms has been established with the successful clinical use of β-lactams in combination with β-lactamase inhibitors. Because efflux mechanisms are widespread in bacteria and have a role in establishing the levels of both susceptibility and resistance, identification and clinical development of efflux-pump inhibitors to be used in combination with existing or new antimicrobials is particularly appealing. The challenges, however, lie in the diversity of these pumps within and among bacterial species. Thus, an efflux inhibitor would require activity against many pumps if it were to have broad clinical utility [32]. For the future, it is likely that the contribution of efflux pumps to antimicrobial susceptibility and resistance will be increasingly recognized as more pumps are identified and studied in detail. With this knowledge is likely to come a better understanding of multidrug resistance, beyond what can be attributed to plasmids containing multiple resistance determinants. In particular, a better understanding of how certain antimicrobial combinations (when both components are substrates of an efflux pump) may promote rather than retard selection of resistance may be clinically important. Physiologic upregulation of efflux pumps in bacteria at the site of infection could also be important for response to therapy and needs further evaluation. Financial support. This review was supported in part by US Public Health Service grant R01 AI23988 from the the National Institutes of Health. Potential conflicts of interest. D.C.H. has received recent funding from Daiichi Pharmaceuticals, Wockhardt Pharmaceuticals, and Kyorin Pharmaceuticals; has served on scientific advisory boards for ActivBiotics and Cubist Pharmaceuticals; and has served as a consultant for Bayer, Daiichi Pharmaceuticals, Mpex Pharmaceuticals, Ortho-McNeil Pharmaceuticals, Osient Pharmaceuticals, and Pfizer.