MexXY multidrug efflux system of Pseudomonas aeruginosa

1Pseudomonas aeruginosa has been recognized as an increasingly important and worrisome species in health care-associated infections. This bacterium possesses intrinsic resistance to many antimicrobials because of the low permeability of its outer membrane barrier and the presence of multidrug efflux transporters. Although fluoroquinolones (e.g., ciprofloxacin), broad-spectrum β-lactams (e.g., imipenem), and anti-pseudomonas aminoglycosides (e.g., amikacin) are often available for treatment, P. aeruginosa readily acquires resistance to these anti-pseudomonas agents via chromosomal mutations and lateral gene transfer. The emergence and spread of multidrug-, extensive drug-, and pandrug-resistant P. aeruginosa infections is of great concern as very few agents are effective against these strains. The problem of increasing antimicrobial resistance is even more threatening when considering the very limited number of new antimicrobial agents in development. In particular, the prospects of finding new antibiotics for Gram-negative pathogens are poor because of the low permeability of their outer membrane barriers and the presence of multidrug efflux transporters. To combat these bacteria efficiently, it is necessary to understand the molecular basis of the efflux mechanisms involved in limiting the intracellular (or periplasmic) concentration of many antibiotics.

2The most clinically relevant multidrug efflux systems in Gram-negative bacteria are those of the resistance-nodulation-cell division (RND) family. P. aeruginosa expresses several RND-type multidrug efflux systems, of which MexAB-OprM, MexCD-OprJ, MexEF-OprN, and MexXY are significant determinants of multidrug resistance in laboratory and clinical isolates. These pumps are three-component systems comprising antiporters of the RND family driven by proton motive force (MexB, MexD, MexF, and MexY), outer membrane factors (OMF; OprM, OprJ, and OprN), and periplasmic membrane fusion proteins (MFP; MexA, MexC, MexF, and MexX;; ). These pumps probably function in a similar manner with AcrAB-TolC, which is the best-studied RND-type multidrug pump of Escherichia coli.

3Among them, the MexXY system is intriguing in that it is a significant determinant of aminoglycoside resistance only in P. aeruginosa, with numerous reports of clinical isolates during the last decade. Relatively few bacterial drug efflux systems are known to accommodate aminoglycosides. In addition to MexXY, AmrAB-OprA of B. pseudomallei is noteworthy for its contribution to this organism’s intrinsic aminoglycoside resistance, while AdeABC of Acinetobacter baumannii is implicated in acquired aminoglycoside resistance. The upregulation of the MexXY pump is considered the most common mechanism of resistance and appears to be the major determinant of aminoglycoside resistance in cystic fibrosis (CF) lung isolates of P. aeruginosa. Here, we summarize the properties of these pumps and discuss how to combat efflux-mediated aminoglycoside resistance.

4In 1993, the first RND-type multidrug efflux system of P. aeruginosa, MexAB-OprM (OprM was called OprK at that time), was discovered at approximately the same time as the AcrAB (AcrB was called AcrE at that time) system of E. coli. It was the first genetic evidence that an efflux operon was involved in multiple antibiotic resistance in P. aeruginosa. The following year, the efflux activity of tetracycline, chloramphenicol, norfloxacin, and benzylpenicillin was shown using antibiotic accumulation assays in intact cells, which was the first biochemical evidence of the role of efflux in intrinsic multidrug resistance in P. aeruginosa. Taken together, the MexAB-OprM system was shown to contribute to the intrinsic resistance of P. aeruginosa to a wide range of antimicrobial compounds including fluoroquinolones, tetracycline, chloramphenicol, and β-lactams such as carbenicillin. Its homologs (MexCD-OprJ and MexEF-OprN) were then discovered as determinants of nfxB and nfxC fluoroquinolone-resistant (e.g., norfloxacin) mutants from P. aeruginosa. Incidentally, these two pumps are not expressed in normal laboratory growth conditions, but are induced under some conditions in wild-type P. aeruginosa (e.g.,; ). An unidentified efflux system that requires OprM was shown to contribute to resistance to quinolones and cephalosporins, such as cefpirome, erythromycin, and tetracycline, but not β-lactams, such as cefoperazone and carbenicillin, in the P. aeruginosa PAO1 background. More details on these three pumps can be found in recent reviews (e.g.,,; ).

5In those days, RND multidrug efflux systems such as MexAB-OprM and AcrAB-TolC, which can handle a wide variety of drugs that appear to contain hydrophobic domains of significant sizes, were considered to be similar to the P-glycoprotein multidrug efflux pump of mammalian cells, which extrudes not only basic compounds but also neutral and weakly acidic compounds. However, there was no evidence for the efflux of aminoglycosides, which are very hydrophilic compounds, among the antibiotics used to treat P. aeruginosa infections.

6The MexXY system was discovered in 1999 in Japan as the fourth RND-type multidrug efflux system of P. aeruginosa PAO1. This system was functionally expressed and conferred resistance to fluoroquinolones, tetracycline, erythromycin, etc. in the E. coli KAM3 mutant lacking the acrB gene, which is an RND transporter component of the major multidrug efflux pump (AcrAB-TolC) in E. coli. Interestingly, unlike the other three already known systems, no open-reading frame encoding the outer membrane component, such as OprM, was found in the region downstream from the mexY gene. However, this system was found to function cooperatively with OprM of P. aeruginosa and TolC of E. coli.

7Nine months after the discovery described above, a French group showed that MexXY was involved in the natural resistance of P. aeruginosa PAO1 to aminoglycosides as well as tetracycline and erythromycin. Although the overexpression of MexXY increased the level of resistance to fluoroquinolones in P. aeruginosa PAO1 cells, disruption of mexXY from PAO1 had no detectable effect on susceptibility to these agents. mexZ, which is located upstream of but transcribed separately from mexXY, was identified. Its product, MexZ, contains a helix-turn-helix motif, which is characteristic of DNA-binding proteins, at its N-terminus, similar to the repressors of RND-type multidrug efflux genes (e.g., AcrR, a repressor of acrAB in E. coli), supporting the notion that mexZ negatively controls the expression of the operon.

8The following month (10 months after the first discovery), a group in the USA showed that MexXY, which they called AmrAB, was an aminoglycoside impermeability factor in spontaneous aminoglycoside-resistant mutants of the impermeability phenotype from P. aeruginosa PAO1. Interestingly, a dramatic decrease in the amount of OprM was observed in the mutants compared to wild-type PAO1, indicating that OprM is unlikely to be the outer membrane component associated with this efflux system in the mutants. In addition, MexXY was shown to be upregulated in clinical P. aeruginosa isolates displaying aminoglycoside impermeability, suggesting that the pump is a clinically relevant mechanism of aminoglycoside resistance in P. aeruginosa.

9The following year, the complete genomic sequence of P. aeruginosa strain PAO1 (PAO1-UW) was published in Nature. Although the locus IDs PA2019-18 of the PAO1-UW genome sequence correspond to the mexXY genes, the nucleotide sequences of PA2019-18 were not identical with those of previously published mexXY findings. This is probably because the DNA sequencing technology at that time was unable to analyze GC-rich bacteria such as P. aeruginosa (66–67% GC content; ). Therefore, we analyzed MexXY using the nucleotide sequences from the PAO1-UW complete genome because we live in the post-genome era.

10The RND components of RND-type tripartite multidrug efflux pumps determine substrate specificity (e.g.,; ); therefore, we focused on the structure and function of MexY rather than MexX or OprM. Very recently, the crystal structure of the RND-type multidrug efflux pump AcrB of E. coli revealed the presence of two discrete, high-volume multisite binding pockets that contribute to the remarkably broad substrate recognition of AcrB and its homologs. Although we basically assume that MexY pumps out antimicrobials in a similar manner as AcrB, it will be intriguing to uncover the molecular basis of how MexY accommodates aminoglycosides because they are strongly hydrophilic molecules that are completely different from the relatively hydrophobic compounds (e.g., minocycline, doxorubicin, rifampicin, and erythromycin) used as substrates of AcrB.

11Generally speaking, the function of a transporter (e.g., substrate specificity and energy coupling) should be determined by its time course efflux assay and evaluated using kinetic constants (e.g.,;;; ). However, it is difficult to conduct such an assessment in a small bacteriology laboratory. The reconstitution of proteoliposomes revealed that AcrB, AcrD, and MdtBC of E. coli were H+/drug antiporters, and we assume that MexY pumps out antimicrobials coupled with the same energy. In addition, five charged and polar amino acid residues that are involved in the proton translocation pathway are conserved between MexY and AcrB of E. coli. Unfortunately, the purification, reconstitution, and characterization of the MexXY pump remain to be established, and the energy coupling and substrate specificity of MexXY has not been examined through its efflux activity. On the other hand, the MexXY-mediated energy-dependent efflux activity of ethidium, aminoglycosides, tetracycline, Ala-Nap (MC-005,556), and fluorescein-di-β-D-galactopyranoside has been measured in whole cells.

12It is conventional to use differences in minimum inhibitory concentrations (MICs) between bacterial cells with and without a multidrug efflux transporter to estimate substrate specificity (e.g.,; ). Although the comparison of MICs can sometimes be a misleading indicator of pump function (e.g., ), it can still indicate the possible clinical relevance of a pump (e.g.,,; ). Mutant strains lacking major multidrug efflux pump(s) have been used to determine substrate specificity (e.g.,; ). The substrate specificity of MexXY-OprM was determined using a mutant from PAO1 that overproduced MexXY-OprM, but not MexAB (and AmpC in the case of β-lactams), and were compared with a mutant lacking MexXY/MexAB-OprM (and AmpC in the case of β-lactams; ). MexXY-OprM-mediated resistance was then observed for quinolones, macrolides, tetracyclines, aminoglycosides, chloramphenicol, lincomycin, and most β-lactams, but not for novobiocin, polymyxin B, and some β-lactams (carbenicillin, sulbenicillin, cefsulodin, ceftazidime, oxacephem, imipenem, and aztreonam) among a wide variety of antimicrobial agents. In conclusion, MexXY-OprM is a multidrug efflux transporter whose specificity is extraordinary broad, but different compared with MexAB-OprM, MexCD-OprJ, MexEF-OprN, and other RND efflux transporters in P. aeruginosa. In addition, MexXY-OprM was the only pump to mediate aminoglycoside resistance and was thus considered to recognize aminoglycosides as substrates.

13Basic local alignment search tool (BLAST) analysis showed that MexY was highly conserved in P. aeruginosa strains: more than 99% (99%) identity (positive) for most strains and 97% (98%) identity (positive) for PA7. There was no functional difference between the MexYs of PAO1 and PA7 when they were expressed in either E. coli or P. aeruginosa. MexY was more similar [70–73% (83–86%) identity (positive)] to orthologs of B. pseudomallei and various B. cepacia complexes than other RND pumps of P. aeruginosa and other Pseudomonas species. These Burkholderia species, except for B. mallei, are intrinsically resistant to aminoglycosides (e.g.,;;; ). B. gladioli is also known to be involved in human infections; however, no MexY (AmrB) ortholog exists in B. gladioli BSR3, consistent with the fact that all isolates tested were susceptible to aminoglycosides. Interestingly, the most similar functional ortholog to MexY exists in Achromobacter xylosoxidans and has a 74% identity (86% positive); this pump was named AxyY in strain AXX-A. A. xylosoxidans is also an opportunistic human pathogen capable of causing a wide range of infections. Most A. xylosoxidans clinical isolates were resistant to the tested aminoglycosides, including amikacin. The AxyY pump contributes to aminoglycoside resistance in a similar manner to MexY and AmrBs.

14COBALT analysis is a multiple sequence alignment tool for finding a collection of pairwise constraints. Such constraints are derived from data of the conserved domain database, protein motif database, and sequence similarity of RND pumps, including all pumps from P. aeruginosa PAO1-UW and E. coli K12 (MG1655). The exception is heavy metal efflux pumps, which are characterized by their relationships. Therefore, we focused on the four branches containing the four Mex pumps in P. aeruginosa (Figure 1). The MexY branch is located next to the MexD branch and includes the AmrBs of Burkholderia species (e.g., ) and AxyY of A. xylosoxidans. The MexD branch includes the AdeB pump of A. baumannii, MtrD of Neisseria gonorrhoeae,and BdeB of Bradyrhizobium japonicum. The SmeZ pump of S. maltophilia, which can mediate aminoglycoside resistance, also belongs to the MexD branch. Many pumps in the MexY/MexD branches can mediate aminoglycoside resistance (e.g.,;;; ), which hints at the structure-function relationship of pumps involved in aminoglycoside resistance. MexB is located in the branch that contains the AcrB/D/F and MdtF pumps of E. coli, AdeJ of A. baumannii, BpeB of B. pseudomallei, AxyB of A. xylosoxidans, and VmeB of Vibrio parahaemolyticus. Among them, some pumps (e.g., AcrD and MexB) were reported to be involved in aminoglycoside resistance under some conditions. The MexF branch includes AdeG of A. baumannii and MdsB of Salmonella enterica.

15Substrate specificity is determined almost entirely by the periplasmic domain (i.e., two large extramembrane loops that largely protrude toward the periplasmic space) of MexY (and MexB; ). MexYF1018L (the F1018L mutation is located in TMS-12 of MexY base on the structure of AcrB; ), enhanced the function of MexY, presumably by increasing the efflux of aminoglycosides, cefepime, and fluoroquinolones, which was the first example of an improved efflux pump in vivo.

16MexXY was shown to be induced by sub-inhibitory concentrations of tetracycline, erythromycin, aminoglycosides, tigecycline, and LMB415 (a peptide deformylase inhibitor), but not ofloxacin in P. aeruginosa PAO1. Moreover, ofloxacin and cefpirome were also shown to be inducers, but only in a PAO1 mutant lacking MexAB (and AmpC in the case of cefpirome;,). MexZ was shown to bind an inverted repeat region located in the mexZ-mexX intergenic region directly as a homodimer, which encompasses the putative mexXY promoter, but the inducers failed to alter the MexZ-operator interactions. The crystal structure of MexZ has since been solved. Induction of the MexXY efflux pump in P. aeruginosa PAO1 was shown to be dependent on drug–ribosome interactions, and the pump remained inducible, but to a lesser degree, by ribosomal inhibitors, even in the mexZ mutant. These data demonstrate the physiological interactions between MexXY and the ribosome and are suggestive of an alternative function for MexXY beyond the efflux of antibiotics. Microarray analysis showed that mexXY were the most highly upregulated genes in P. aeruginosa PAO1 after 4 h of interaction with primary normal human airway epithelial cells and in response to sub-inhibitory concentrations of tobramycin under normal aerobic conditions, but not under lethal aerobic conditions or anaerobic conditions.

17The antibiotic inducibility of the MexXY multidrug efflux system of P. aeruginosa was shown to be involved in the modulation of MexZ activity by the antibiotic-inducible PA5471 gene product (; Table 1). PA5471 encodes a predicted product of 43.5 kDa, which was identified as a hypothetical protein conserved between bacteria and archaea, and is a representative of the uncharacterized protein family UPF0027 in the Pfam protein families database or the PRK09588 cluster in ProtClustDB (NCBI Protein Clusters Database; ). Recently, it was demonstrated that RctB of E. coli, which is related to members of this family, is a novel RNA ligase and functions as a bona fide RNA repair protein in vivo. PA5471 is found upstream of and in a possible operon with an open-reading frame dubbed PA5470; RT-PCR confirmed both the drug inducibility of PA5470 and its expression from a polycistronic message that also contains PA5471. PA5470 is predicted to encode a peptide chain release factor of 22.3 kDa. A homolog of PA5471 from E. coli K12, ykfJ (b0235), which was, however, C-terminally truncated (approximately 1 kb; ), was also shown to be inducible by 4-azaleucine, which is known to interfere with translation, and it too is linked to a putative peptide release factor gene. P. aeruginosa senses antibiotic-mediated ribosomal disruption and links it to PA5471 gene expression by monitoring the translation of a 13-amino-acid-leader peptide region (PA5471.1) found ~250 bp upstream of the PA5471 coding sequence on PA5471 mRNA. The antimicrobial-inducible PA5471 gene product has been shown to interact with the repressor MexZ and interfere with its DNA binding activity in vitro, and this finding contributed to elucidating the molecular mechanisms of the MexXY induction. However, PA5471 is not sufficient for MexXY recruitment in response to antibiotic exposure, and additional antibiotic-dependent effects are needed in P. aeruginosa. Exposure to reactive oxygen species (ROS; e.g., peroxide) induces the expression of the PA5471 gene, leading to MexXY-dependent aminoglycoside resistance. Moreover, long-term (8-day) exposure of P. aeruginosa to peroxide (mimicking chronic in vivo ROS exposure) increased the frequency of PA5471- and mexXY-dependentaminoglycoside resistance. Recently, reduced (approximately twofold) expression of the rplU-rpmA operon (encoding the 50S ribosomal proteins L21 and L27) was shown to promote mexXY expression via the PA5471 gene in pan-aminoglycoside resistant mutants from PAO1 and a CF clinical isolate. Such expression was in the form of ribosomal protein mutations that influence mexXY expression, including rplY (encoding ribosomal protein L25; ) and rplA (encoding ribosomal protein L1; ). Transcriptome profiling revealed that significantly increased expression was observed for the mexXY and PA5471 genes in both the PA2572 and PA2573 mutants compared with the wild-type PAO1 strain during exponential growth in Luria–Bertani media. PA2572 encodes a putative response regulator of a two-component system required for full virulence to Galleria mellonella (Wax moth) and PA2573 also encodes an ophan chemotaxis sensor which seems to function in part through signal transduction involving PA2572.

18A recent study identified a gene, parR, encoding the response regulator of a two-component system, ParRS, which promotes either induced or constitutive mexXY upregulation, thereby activating the MexXY efflux system as well as OprD porin loss and lipopolysaccharide modification in a MexZ-independent manner. Overexpression of PaeIII, a small non-coding RNA between PA3505 and PA3536 in the genome of P. aeruginosa PAO1, in the stationary phase increased the expression of the mexXY and mexZ genes as well as type III secretion genes, while reducing the expression of genes for arginine metabolism.

19Pseudomonas aeruginosa shows intrinsic resistance against many antimicrobials because of the low permeability of its outer membrane and the presence of efflux systems. MexXY was shown to be involved in natural resistance to aminoglycosides, tetracycline, tigecycline, erythromycin, and LBM415 in P. aeruginosa PAO1. MexXY was also shown to be necessary for the adaptive resistance of P. aeruginosa PAO1 to aminoglycosides. It is of note that MexXY is the only pump of the 12 identified RND systems that mediates aminoglycoside resistance in P. aeruginosa PAO1. The antagonism of aminoglycosides by the divalent cations Mg2+ and Ca2+ is well documented, and culture in cation-adjusted Mueller–Hinton broth is recommended as a susceptibility test to ensure acceptable results when P. aeruginosa isolates are tested. MexXY was shown to be required for the antagonism of aminoglycosides by divalent cations in P. aeruginosa PAO1. Although Phe-Arg-β-naphthylamide (PAβN, MC-207,110) is known as a non-specific inhibitor against RND-type multidrug efflux pumps, this inhibitor, as observed for divalent cations, antagonized the activity of aminoglycosides (amikacin and netilmicin) in a MexXY-dependent manner, even though it also inhibited MexXY-dependent fluoroquinolone (levofloxacin) resistance. Conversely, PAβN inhibited MexXY-mediated aminoglycoside (gentamicin) resistance. The reason for the discrepancy between these two results remains unknown. Increased susceptibility to aminoglycosides in nfxB mutants, which upregulate mexCD-oprJ expression, was correlated with increased resistance to fluoroquinolones and some β-lactams, such as cefepime, concomitant with a higher susceptibility to aminoglycosides and some β-lactams, such as ticarcillin, aztreonam, and imipenem. This was shown to be partly due to the impaired activity of MexXY-OprM because of major changes in cell physiology, but not the expression/production of mexY/MexY and oprM/OprM (; Mulet et al., 2011). The increased susceptibility to aminoglycosides in MexEF-OprN-overproducing nfxC mutants was also observed, apparently owing to impairment of the MexXY system. mexXY expression (and so MexXY-mediated resistance) was independent of the AmgRS two-component system in which mutations enhanced aminoglycoside action to control an adaptive response to membrane stress.

20Multidrug resistant P. aeruginosa clinical isolates have often been reported to be MexXY overproducers (e.g.,,;;;;;;;;;; ). Time series analysis (January 1999 to January 2005) revealed a significant relationship between antibiotic use (aminoglycosides, fluoroquinolones, and cefepime, but not carbapenems) and the incidence of MexXY-overproducing P. aeruginosa in a French hospital. MexXY (n = 39) and MexAB (n = 31) were the most frequently overproduced pumps in 85 non-CF P. aeruginosa strains with low-level ciprofloxacin resistance (MICs ranging from 0.25 to 2 µg/mL, which are still susceptible or intermediate according to the CLSI breakpoints; ). A large proportion of the strains were MexXY overproducers in genotypically distinct P. aeruginosa clinical isolates that were less susceptible to cefepime than to ceftazidime, and these were identified in Europe and the USA. In contrast, both cefepime and ceftazidime are potent β-lactam antibiotics with similar MICs (1–2 µg/mL) for wild-type P. aeruginosa strains. Moreover, ceftobiprole, similar to cefepime, selected MexXY overproducers in clinical studies. Actually, a single step MexXY overproducer was selected in vitro by cefepime and ceftobiprole, but not ceftazidime. MexXY contributed very significantly to the development of high-level (100–1000 µg/mL MIC) aminoglycoside resistance via a combination of aminoglycoside-modifying enzymes (AMEs) in multidrug resistant P. aeruginosa non-CF clinical isolates. However, AMEs are common determinants of aminoglycoside resistance in P. aeruginosa,except for CF isolates. In clinical CF isolates, MexXY has been primarily implicated in pan-aminoglycoside resistance (e.g.,;;, ). MexXY was also shown to be necessary in subpopulations of P. aeruginosa CF isolates that are hypersensitive to ticarcillin (called Tichs; ). mexZ was shown to be one of the most frequently mutated genes during chronic infection by P. aeruginosa in CF patients. However, a number of studies highlighted the absence of mutations in mexZ or the mexXY promoter region in MexXY-overproducing P. aeruginosa CF isolates. To date, three kinds of mutants (agrZ, agrW1, and agrW2) have been recognized as MexXY overproducers as a result of genetic mechanisms: mutants with impaired binding or unbinding of MexZ due to alterations in the mexZ or mexZ-mexX intergenic region (type agrZ); mutants with impaired protein synthesis (type agrW1); and mutants with alterations in parRS (type agrW2; ). Oxidative stress, a component of the host’s immune system in the CF lung, induced mexXY expression via PA5471 and promoted aminoglycoside resistance. Under conditions of oxidative stress, P. aeruginosa can develop aminoglycoside resistance, even in the absence of aminoglycosides. It is also very plausible that the routine use of aminoglycosides (e.g., tobramycin) might simply select for MexXY-overproducing P. aeruginosa in theCF lung.

21Although it is obvious thatMexXY is one of the determinants of antimicrobial resistance in P. aeruginosa in the clinical setting, only a few reports have assessed the in vivo impact of the MexXY system on antibiotic therapy for P. aeruginosa infections (e.g., ).

22The mexXY operon lacks a gene coding for the outer membrane protein in P. aeruginosa PAO1. OprM is necessary for the function of MexXY and MexAB in P. aeruginosa PAO1, although overproduced OpmB (PA2525) can function as an outer membrane component of MexXY, MexAB, and MexCD. Intriguingly, the multidrug resistant taxonomic outlier P. aeruginosa PA7 possesses a unique gene (oprA) downstream of mexXY encoding an outer membrane channel that is absent in most P. aeruginosa strains. MexXY in this strain utilizes either the OprA or OprM outer membrane channel (; Table 1). While OprM is functional with both MexXY and MexAB, OprA did not associate as strongly with MexAB as it did with MexXY. We compared the OprA of P. aeruginosa PA7 with the OprM family from P. aeruginosa PAO1 as well as TolC of E. coli K12, OprA of B. pseudomallei 1026b, OprZ of A. xylosoxidans AXX-A,and AdeC of A. baumannii AYE (; Figure 2). COBALT analysis showed that OprA of P. aeruginosa PA7 and its close orthologs (OprA of B. pseudomallei 1026b and OprZ of A. xylosoxidans AXX-A) is located close to OprJ and is followed by OprM of the OprM outer membrane family of P. aeruginosa PAO1 (Figure 2).

23Interestingly, a small portion of the oprA gene immediately downstream of the mexY gene in PAO1 was identified, suggesting that non-PA7 P. aeruginosa strains might have possessed, but lost, the intact mexXY-oprA efflux pump locus (; Table 1). Consistent with this, the majority of a panel of serotype strains possessed the truncated oprA, but the serotype O12 isolate had an intact mexXY-oprA locus, similar to PA7 and the related strain DSM 1128. O12 is a predominant serotype associated with multidrug resistance to a number of antibiotic classes, including aminoglycosides and β-lactams, although it represents a minor serotype in the environment. O12 might be more dominant due, in part, to the presence of oprA in hospitals in which antimicrobials promoting MexXY-OprA-mediated multidrug resistance, such as aminoglycosides, were used. P. aeruginosa PA7 isolated before 1984 from a wound infection in Buenos Aires, Argentina, might also have acquired multidrug resistance via the heavy use of antibiotics, including gentamicin or tobramycin, to treat wounds at that time. Apparently, a slightly increased resistance (two- to fourfold) to amikacin, ciprofloxacin, and cefpirome was shown in the presence and absence of oprA. Such a small difference might be significant during antibiotic treatment or in the presence of sub-inhibitory concentrations of antibiotics.

24AcrD has the highest similarity score at the amino acid level to MexY of the E. coli K12 RND pumps and was shown to be an aminoglycoside efflux pump as judged by MIC determination and the aminoglycoside efflux assay. However, differences in aminoglycoside resistance (uptake) between the parent strain JC7623 and its acrD-deletion mutant JZM320 was possibly not limited to AcrD function because JZM320 was constructed by inserting the tet gene from pBR322 into acrD. The increased aminoglycoside uptake might be due to not only AcrD deficiency but also to the production of an aberrant cytoplasmic membrane protein (the product of acrD with the inserted tet) and/or the tetracycline/H+ antiporter itself. While disruption of tolC or acrA did not increase the susceptibility of K12 to aminoglycosides, both of them were necessary for the function of acrD against various antimicrobials; however, no aminoglycosides were used in the study. We do not rule out the hypothesis of that AcrD protein can perhaps function without the participation of AcrA and TolC in the case of aminoglycoside efflux.

25It is evident that purified AcrD can function as an H+-driven aminoglycoside efflux pump. Especially, strong stimulation of proton efflux was observed when aminoglycosides (e.g., streptomycin) were added to the more acidic intra-vesicular space of reconstituted AcrD proteoliposomes containing AcrA and Mg2+, indicating that AcrD captures aminoglycosides exclusively from the periplasm in E. coli. The difference in the MICs of amikacin and gentamicin between a parent strain and its in-frame acrD-deletion mutant or between an acrBD-deletion mutant and its acrD-overexpressing complementation mutant was approximately twofold. There was no significant difference in kanamycin resistance in the case of an in-frame deletion, and a twofold difference was observed in the case of overproduction. An acrA in-frame deletion mutant also showed an approximately twofold increased susceptibility to aminoglycosides. However, similar observations were not seen for AcrB. A comparison of the entrances of the vestibules, which are found in the central cavities of AcrD (which transports aminoglycosides) and AcrB (which does not) crystal structures, shows that this area in AcrD is in line with many more acidic residues that may attract polycationic substrates. Treatment with sub-inhibitory concentrations of kanamycin induced adaptive resistance to aminoglycosides, which was dependent on acrD. Aminoglycosides are very hydrophilic and polycationic and assumingly permeate through the porin channel in E. coli, unlike P. aeruginosa in addition to so called “self-promoted” aminoglycoside uptake across the outer membrane of both of E. coli and P. aeruginosa. The MICs of aminoglycosides on E. coli, unlike P. aeruginosa,might be poor indicators of aminoglycoside efflux. There are numerous AcrD homologs in other Enterobacteriaceae. Although the AcrD of S. enterica serovar Typhimurium ATCC 14028s was studied comprehensively, no significant difference between the AcrDs of E. coli and S. enterica has been observed so far. Interestingly, AcrD pumps mediate resistance to the substrates of MexAB (e.g., carbenicillin, aztreonam, and novobiocin). However, AcrD pumps did not mediate resistance to the substrates of MexXY (e.g., cefpirome, erythromycin, and tetraphenylphosphonium) or shared substrates of both MexAB and MexXY (e.g., fluoroquinolone and tetracycline) when differences of the MICs were compared between a parent strain and its transformant overproducing the pump. MexAB-OprM was also shown to contribute to aminoglycoside resistance, presumably via active efflux in the low-ionic-strength medium used in this particular study. AcrD and the MdtABC pump were iron-regulated, induced in low-iron conditions, and export the siderophore enterobactin, which reminds us that MexAB-OprM was inducible under conditions of iron limitation and compensated for a growth defect in an iron-deficient medium in the presence of the non-metabolizable iron chelator 2,2′-dipyridyl. AcrD seems to be a functional homolog of MexB rather than MexY, as determined from substrate specificity and physiological function, consistent with the fact that phylogenetic analysis showed that AcrD is closer to MexB than to MexY (Figure 1).

26Burkholderia pseudomallei is the etiologic agent of melioidosis, a rare but serious disease endemic to South Asia, Northern Australia, and other parts of the tropics. Melioidosis is very difficult to treat because of the intrinsic resistance to many antimicrobial agents including aminoglycosides, macrolides, polymyxins, and some β-lactams. AmrAB-OprA was identified as an efflux determinant of resistance to aminoglycosides and macrolides in the B. pseudomallei 1026b clinical isolate. This pump was actually the first to be demonstrated responsible for the aminoglycoside resistance of RND pumps in Gram-negative bacteria. The gene product of amrR, which is located immediately upstream and divergently transcribed from amrAB-oprA in B. pseudomallei 1026b, showed strong homology [60% (73%) identity (positive)] to MexZ, which acts as a transcriptional repressor of the mexXY operon of P. aeruginosa PAO1 (;; Table 1).

27While the majority of B. pseudomallei clinical isolates exhibit high levels of aminoglycoside and macrolide resistance, rare isolates are susceptible to these antibiotics. While it is noted that the resistance profile of those isolates matches that of the amrAB-oprA mutants, it was shown experimentally that amrAB-oprA was missing in B. pseudomallei 708a, an aminoglycoside- and macrolide-susceptible clinical isolate, and this loss was associated with the deletion of >130 kb of genetic material. The expression of amrAB-oprA increased resistance to not only aminoglycosides and macrolides but also fluoroquinolones and tetracyclines in a BpeAB-OprA pump-deficient mutant of 1026b. Judging from the substrate specificity and sequence similarity, we have no doubt that AmrAB is a functional ortholog of MexXY in B. pseudomallei. BpeAB-OprB of B. pseudomallei also reportedly mediates aminoglycoside resistance in strain KHW, while this pump did not confer aminoglycoside resistance in 1026b. In addition, the BpeB RND transporter was also shown to be closely related to MexB of P. aeruginosa, both functionally and phylogenetically, consistent with our phylogenetic analysis (Figure 1).

28As described above, AmrB orthologs are conserved among various human pathogens belonging to Burkholderia species, but not B. gladioli. Actually, an AmrAB-OprA ortholog was shown to be a major aminoglycoside resistance contributor in B. cenocepacia, a member of the B. cepacia complex.

29Acinetobacter baumannii is the most frequently implicated species in nosocomial infections among Acinetobacter spp.. AdeABC was identified as an RND-type efflux pump involved in resistance to multiple antimicrobials including aminoglycosides, fluoroquinolones, tetracycline, erythromycin, cefotaxime, trimethoprim, and chloramphenicol in A. baumannii BM4454, a low-level pan-aminoglycoside resistant clinical isolate. The polycistronic adeABC transcript was confirmed experimentally to encode AdeA (MFP), AdeB (RND transporter), and AdeC (OMF;; ). Because the substrate specificity of AdeAB is very similar to that of MexXY, we have no doubt that AdeAB is a functional homolog of MexXY in A. baumannii. AdeC is not essential for AdeAB-mediated resistance, indicating that AdeAB recruits another outer membrane protein to form a functional tripartite complex, as observed for the MexXY pump with OprM in P. aeruginosa. We do not rule out the possibility that AdeAB is functional with AdeC, as observed for the MexXY pump with its linked outer-membrane channel OprA, which was not essential in P. aeruginosa PA7. The adeAB genes are usually present, but the adeC gene was not found in ~40% of clinical isolates. Our phylogenetic analysis showed that AdeC is more closely related to OprM and OprJ than to OprA in the OprM outer membrane family of P. aeruginosa (Figure 2).

30The adeABC operon is expressed at low levels in natural isolates of A. baumannii due to stringent control by the AdeRS two-component system, which is encoded adjacent to adeABC, but transcribed in the opposite direction (; Table 1). Mutations (e.g., AdeRPro116Leu, AdeSThr153Met, or AdeSGly30Asp) in AdeRS have been shown to be responsible for the constitutive expression of AdeABC, which reminds us that mutations (e.g., ParRMet59Ile) in the ParRS two-component system are responsible for the constitutive expression of MexXY in P. aeruginosa. Overexpression of the AdeABC system in a tigecycline non-susceptible clinical isolate was due to the transposition of a copy of ISAba1 into adeS. Very recently, a truncated AdeS kinase protein generated by an ISAba1 insertion was shown to be correlated with enhanced adeABC expression in A. baumannii. Other regulatory mechanism(s) were shown to be involved in adeABC overexpression without any previously known mutation. Recently the AdeABC ortholog was shown to be a contributor to multiple antimicrobials, including aminoglycosides, in Acinetobacter genomospecies 13TU, a non-A. baumannii species.

31MexXY is one of the potential targets for novel anti-pseudomonas agents. Its inhibitor is able to not only potentiate previously used ineffective antimicrobial agents (e.g., aminoglycosides against aminoglycoside-resistant P. aeruginosa and B. cepacia complex), but also to speed up the development of novel anti-pseudomonas agents. Because there are a significant number of potential drug targets encoded by the genome of P. aeruginosa (e.g., products of essential genes; ), it is the most promising therapeutic strategy to conquer the impermeability barriers of these bacteria. The efflux inhibitor MP 601384, which has specificity toward aminoglycoside-accommodating RND efflux systems and is not toxic to bacteria, is the only MexXY inhibitor reported so far. Uncultured bacteria and plants are predicted to be a significant reservoir of novel antimicrobial agents. Screening novel antibacterial agents, including a MexXY inhibitor, is currently in progress in our laboratory (e.g., ).

32The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.