|Home | About | Journals | Submit | Contact Us | Français|
Summary: Although plasmid-mediated quinolone resistance (PMQR) was thought not to exist before its discovery in 1998, the past decade has seen an explosion of research characterizing this phenomenon. The best-described form of PMQR is determined by the qnr group of genes. These genes, likely originating in aquatic organisms, code for pentapeptide repeat proteins. These proteins reduce susceptibility to quinolones by protecting the complex of DNA and DNA gyrase or topoisomerase IV enzymes from the inhibitory effect of quinolones. Two additional PMQR mechanisms were recently described. aac(6′)-Ib-cr encodes a variant aminoglycoside acetyltransferase with two amino acid alterations allowing it to inactivate ciprofloxacin through the acetylation of its piperazinyl substituent. oqxAB and qepA encode efflux pumps that extrude quinolones. All of these genes determine relatively small increases in the MICs of quinolones, but these changes are sufficient to facilitate the selection of mutants with higher levels of resistance. The contribution of these genes to the emergence of quinolone resistance is being actively investigated. Several factors suggest their importance in this process, including their increasing ubiquity, their association with other resistance elements, and their emergence simultaneous with the expansion of clinical quinolone resistance. Of concern, these genes are not yet being taken into account in resistance screening by clinical microbiology laboratories.
The first quinolone, nalidixic acid (possessing a naphthyridone core), was introduced into clinical use in 1962 (10). In the mid-1980s, ciprofloxacin, a fluoroquinolone (with a quinolone core) that had a wider spectrum of in vitro antibacterial activity, particularly against gram-negative bacteria, first became available clinically (146). Since then, newer agents with increased antimicrobial activity against gram-positive pathogens have been developed, but the activity of ciprofloxacin against gram-negative pathogens has been largely unsurpassed (80, 177). In the decades that have elapsed since the introduction of fluoroquinolones, resistance of the Enterobacteriaceae to these agents has become common, widespread, and generally nonclonal. This implies that fluoroquinolone resistance has emerged independently many times (44, 66, 114, 132).
The main mechanism of quinolone resistance is the accumulation of mutations in the bacterial enzymes targeted by fluoroquinolones: DNA gyrase and DNA topoisomerase IV (79). Multiple mutations are generally required for clinically important resistance to result, as wild-type organisms are highly susceptible (79). Because spontaneous double mutations are rare genetic events (occurring at a frequency of 10−14 to 10−16 for fluoroquinolones) (227), the traditional understanding of quinolone resistance as a mutational phenomenon has not provided a fully satisfying explanation for the frequency with which this resistance has arisen. Such a phenomenon could perhaps be better accounted for if horizontally transferable elements could provide some degree of reduced quinolone susceptibility, enough so that organisms could survive in the face of quinolones, while resistance mutations occurred sequentially rather than simultaneously. The presence of such horizontally transferable elements might also help account for the strong association between resistance to quinolones and resistance to other agents. In the past, the recessive nature of resistant gyrA alleles in merodiploids (187) was used to account for the lack of naturally occurring plasmid-mediated quinolone resistance (PMQR) (48).
The discovery in 1998 (123) of qnr, a plasmid-mediated horizontally transferable gene encoding quinolone resistance, has shed light on these phenomena. Qnr proteins are capable of protecting DNA gyrase from quinolones and have been in circulation for at least 20 years (90). During this time they have achieved global distribution in a variety of plasmid environments and bacterial genera. Two additional mechanisms of resistance that were predicted to occur were subsequently found (48). AAC(6′)-Ib-cr, a variant aminoglycoside acetyltransferase capable of reducing ciprofloxacin activity, is also carried on plasmids and may be even more prevalent than Qnr proteins (165). Quinolone extrusion, a prevalent chromosomally encoded mechanism of resistance, has also been found to be plasmid borne (70, 143, 215). These mechanisms provide the low-level quinolone resistance shown in vitro to facilitate the emergence of higher-level resistance in the presence of quinolones at therapeutic levels.
Several reviews of the topic of PMQR have been published (22, 121, 133, 163). Research on PMQR is, however, rapidly expanding. Two of the five known qnr gene families (qnrC and qnrD) were first reported within the past year (32, 208). The recent solution of the structure of aac(6′)-Ib importantly advanced our understanding of the aac(6′)-Ib-cr mechanism of action (124, 204), and various new plasmids carrying PMQR genes have been reported. Ten years have passed since the first report of PMQR, but knowledge is still expanding quickly. Remarkably, a PubMed search using the keywords PMQR, qnr, and aac(6′)-Ib-cr revealed 212 available original research publications, and 129 (61%) were published since January 2008. In this review we address new information on PMQR in the context of previously established data.
The discovery of PMQR in the late 1990s was made serendipitously by Martínez-Martínez and colleagues (123). That group was studying pMG252, a plasmid from a multiresistant strain of Klebsiella pneumoniae that was isolated from a urine specimen from a patient at the University of Alabama at Birmingham in 1994. A quinolone was included as a control in a study of the ability of pMG252 to increase resistance to β-lactam antibiotics in porin-deficient strains of Klebsiella pneumoniae. Unexpectedly, an increase in the quinolone MIC was found. The effect of the plasmid was increased 4- to 16-fold in this porin-deficient isolate, but even in an Escherichia coli strain with intact porins, pMG252 increased the quinolone MICs between 8- and 64-fold. Although this increase from baseline was not to the level designated to represent clinical resistance (the resistance breakpoint), the plasmid also facilitated the selection of higher-level quinolone resistance. A wild-type E. coli strain carrying pMG252 plated onto agar containing nalidixic acid or ciprofloxacin was 100 times more likely to give rise to spontaneous resistant mutants than was a plasmid-free strain (123). The plasmid did not have a general mutator effect since the frequency of other genetically defined mutations was equivalent to that of plasmid-free E. coli (123). Subsequent cloning of the gene responsible for this phenotype revealed it to be a 657-bp open reading frame, and the protein which it encoded was named Qnr, for quinolone resistance (GenBank accession number AY070235) (197). (Note that the term “resistance” in the setting of PMQR is used to refer to any increase in MIC—a biological definition—rather than to an increase above a susceptibility breakpoint—a clinical definition.) More recently, this protein has been renamed QnrA1, since related proteins have been identified.
A Klebsiella oxytoca isolate from Anhui Province, China, was reported to carry a variant of qnrA differing from the originally detected gene by four amino acids. This variant was designated qnrA2 (GenBank accession number AY675584) (133). While searching for a chromosomal orthologue of qnrA in the genome sequences of environmental organisms, Poirel et al. identified three additional variants (qnrA3, qnrA4, and qnrA5) of this gene in Shewanella algae, varying from qnrA1 in two to four codons (GenBank accession numbers DQ058661, DQ058662, and DQ058663, respectively) (155). At about the same time, qnrA3 was also detected in clinical salmonella isolates (GenBank accession number AY906856) (38). Subsequently, a group searching for qnrA determinants among isolates of the Enterobacteriaceae with reduced susceptibility to quinolones found qnrA6 in a Proteus mirabilis isolate (GenBank accession number DQ151889) (17).
A subsequent search for plasmids exhibiting transferable resistance to quinolones led to the discovery of four additional similar proteins, QnrS (75), QnrB (91), QnrC (208), and QnrD (32). In October 2003, a single clone of Shigella flexneri 2b caused a food-borne outbreak of enterocolitis in Aichi Prefecture, Japan. One of eight strains of this clone was resistant to ciprofloxacin. This strain was found to harbor a unique conjugative plasmid that transferred quinolone resistance. Cloning identified an open reading frame encoding a 218-amino-acid protein of the pentapeptide repeat family. This protein shares only 59% amino acid identity with QnrA1 and was named QnrS (GenBank accession number AB187515) (75).
In the course of searching for qnr genes among clinical non-Typhi Salmonella isolates from the United States, a qnrS variant (qnrS2) (GenBank accession number DQ485530) was detected on a plasmid from Salmonella enterica serovar Anatum that codes for a protein that is 92.2% identical in amino acid sequence to QnrS1 (70). qnrS2 was found on a plasmid isolated from the activated sludge basin of a wastewater treatment plant in Germany at about the same time (14, 99). qnrS from veterinary clinical E. coli isolates in Guangdong, China, that was deposited in GenBank as qnrS1 (GenBank accession number ABU52984) differed from qnrS1 in one codon and has thus been renamed qnrS3. To avoid such confusion and to bring order into qnr numbering, a database of qnr allele designations has been established at http://www.lahey.org/qnrStudies (87). A fourth qnrS variant in a salmonella isolate from Denmark was recently described (195).
While investigating strains of K. pneumoniae from India, some of which contained qnrA, Jacoby and colleagues found that several could transfer low-level quinolone resistance but were negative by PCR for qnrA (91). The PMQR gene responsible for this phenotype coded for a 214- or 226-amino-acid protein (depending on which potential initiation codon was taken as the start) and was termed qnrB1 (GenBank accession number DQ351241). A recent determination of the transcription start site supports a peptide length of 214 amino acids, and homology between different qnrB variants bears this out (23, 209). The QnrB1 protein shares 43% and 44% amino acid identities with QnrA and QnrS, respectively (91).
The repertoire of qnrB variants is broader than that of qnrA and qnrS. The first variant, qnrB2, was found in the first survey among several isolates of Enterobacteriaceae from the United States. The qnrB2 gene codes for a 214-amino-acid protein (GenBank accession number DQ351242) that differs from qnrB1 in five codons (91). Subsequent surveys in the United States have identified qnrB3 and qnrB4 (GenBank accession numbers DQ303920 and DQ303921, respectively) among the Enterobacteriaceae (166) and qnrB5 in non-Typhi salmonella isolates (GenBank accession number DQ303919) (70); these differ from qnrB1 in 2, 14, and 6 codons, respectively. More recently, using the same multiplex PCR method, qnrB13, qnrB14, and qnrB15 (GenBank accession numbers ABX72042, ABX72044, and ABX72227, respectively), differing from qnrB1 in four, five, and five amino acids, respectively, were identified among Citrobacter freundii strains from South Korea (190). Using degenerate primers for the detection of qnrB in a collection of isolates of Enterobacteriaceae from Kuwait City, Kuwait, Cattoir et al. identified qnrB7 in Enterobacter cloacae and qnrB8 in C. freundii (23). These genes differ from qnrB1 by 4 and 11 amino acids, respectively (GenBank accession numbers EU043311 and EU043312, respectively). qnrB19, which differs from qnrB1 by six amino acids, was also found in an E. coli isolate from Colombia by the same group (GenBank accession number EU432277) (23). qnrB10 (161), discovered in C. freundii, differs from qnrB1 by five amino acids, and qnrB12 (104), identified in Citrobacter werkmanii, differs from qnrB1 by eight amino acids in addition to the five differences that characterize qnrB10. Sequences of six more qnrB alleles have been deposited in GenBank. qnrB6, which differs from qnrB1 by two amino acids, was found among several cephalosporin-resistant isolates of Enterobacteriaceae in western China (GenBank accession number EF520349). qnrB9 (GenBank accession number EF526508), qnrB11 (GenBank accession number EF653270), qnrB16 (GenBank accession number EU136183), and qnrB17 and qnrB18 (GenBank accession numbers AM919398 and AM919399, respectively) were all found among different isolates of C. freundii. Before agreement on qnrB allele numbering (87), sequences were submitted to GenBank with arbitrary, overlapping, and frequently changing allele numbers. Consequently, qnrB allele designations in GenBank files should be accepted with caution.
A clinical strain of Proteus mirabilis from Shanghai, China, which transferred low-level quinolone resistance, was negative by PCR for the known qnr genes. Plasmid pHS9, which, upon conjugation, increased the MIC of ciprofloxacin, carried a 666-bp gene, designated qnrC1, coding for a 221-amino-acid protein. QnrC shared 64%, 41%, 59%, and 43% amino acid identities with QnrA1, QnrB1, QnrS1, and QnrD, respectively (208).
Four Salmonella enterica isolates obtained from humans in the Henan Province of China showed reduced susceptibility to ciprofloxacin that was transferable on a small plasmid of about 4.3 kb, which in E. coli conferred a 32-fold increase in the MIC of ciprofloxacin. This plasmid was negative for qnrA, qnrB, qnrS, aac(6′)-Ib-cr, and qepA (other PMQR genes [see below]) (32). The plasmid encodes a 214-amino-acid pentapeptide repeat protein designated QnrD. qnrD (GenBank accession number EU692908) showed 48% similarity to qnrA1, 61% similarity to qnrB1, and 32% similarity to qnrS1 (32).
In silico analyses of deposited complete genomes led to the discovery of qnr-like genes in the genomes of several gram-positive bacteria including Enterococcus faecalis, Enterococcus faecium, Listeria monocytogenes, Clostridium perfringens, Clostridium difficile, Bacillus cereus, and Bacillus subtilis (171) as well as in gram-negative, mainly waterborne, bacteria (153, 178, 179).
A recently proposed consensus for qnr nomenclature defined qnr as a naturally occurring allele encoding a pentapeptide repeat protein that confers reduced susceptibility to nalidixic acid or a fluoroquinolone. qnr families (such as qnrA, qnrB, or qnrC) are defined by a 30% or more difference in nucleotides or derived amino acids. Within each family, qnr alleles differ in one or more amino acids. qnr genes found on a bacterial chromosome are named after the host organism or assigned to a family if the gene is at least 70% identical to an established qnr family (e.g., SaqnrA3 from the chromosome of Shewanella algae) (87).
The Qnr proteins belong to the pentapeptide repeat family, which is defined by a series of tandem 5-amino-acid repeats. In the pentapeptide repeats, no position is completely conserved, but each of the residues of an individual pentapeptide exhibits a propensity for a restricted number of amino acids with the recurrent general motif approximately represented by the sequence A(D/N)LXX and more precisely represented by [Ser, Thr, Ala, or Val][Asp or Asn][Leu or Phe][Ser, Thr, or Arg][Gly] (12, 203). Qnr proteins, however, often have a cysteine at position i−2 (with position i representing the central amino acid of each repeat) (77). A characteristic feature of the Qnr proteins is that they are formed by two domains of pentapeptide repeats separated by a single amino acid, usually glycine. The primary structures of QnrA, QnrB, and QnrS are similar, with nine pentapeptide repeat units connected by a single glycine, followed by a cysteine, with variable numbers of units (22 in QnrS, 28 in QnrA, and 29 in QnrB, QnrC, and QnrD) (Fig. (Fig.1).1). These differ from MfpA, a pentapeptide repeat protein found in Mycobacterium smegmatis that is also capable of gyrase protection from quinolone and for which the three-dimensional structure has also been solved (see below). In that protein, there is a kink in the helical axis midway through the β-helix but no putative glycine “hinge” (77).
As the three-dimensional structure of Qnr has not yet been solved, the contribution of the glycine moiety to function is unknown. Likewise, it is impossible to predict increases in activity from changes in primary structure among the many Qnr variants. For example, when the Vibrio parahaemolyticus qnr homologue, VPA0095, was cloned onto a plasmid, it did not confer fluoroquinolone resistance unless cysteine 115 was mutated to a tyrosine (178). When the same amino acid substitution was created in QnrA1 or QnrS1, however, the ability to protect against quinolones declined rather than increased (26). Mutations in amino acids that are conserved among Qnr proteins had a similar effect of a decrease in quinolone resistance, but the change in the MIC varied among QnrA1, QnrB1, and QnrS1 with the same substitution (167). A phylogenetic analysis of Qnr alleles suggests that recombination has played an important part in their evolution but that a specific role in protection against quinolones is not evident (11).
Genes for pentapeptide repeat proteins with sequence similarity to plasmid-borne Qnr proteins have been found on the chromosomes of both gram-positive and gram-negative bacteria. Some of these proteins also have the primary structure of two pentapeptide strings of variable length connected by a single amino acid. For example, Efsqnr, found on the chromosome of E. faecalis strains V583 and JH2-2, encodes a 211-amino-acid pentapeptide repeat protein that shares 25% identity and 40% similarity with QnrA. The protein is divided into two pentapeptide repeat domains of 9 and 33 units, each connected by a single asparagine (9). In silico analysis of the pentapeptide domains of Listeria monocytogenes (GenBank accession number EAL07413), Clostridium perfringens (accession number NP_561876), Clostridium difficile (accession number CAJ69589), Bacillus cereus (accession number NP_831602), and Bacillus subtilis (accession number CAB12929) using the Pfam database of the Wellcome Trust Sanger Institute (64) as a search platform failed to identify two separate pentapeptide domains in any of these sequences (J. Strahilevitz, unpublished data).
Gram-negative species were initially screened for qnr-like genes in a search for the reservoir of Qnr determinants (see Origins of qnr Genes below) (155). The conclusion that qnrA3 has a chromosomal location in Shewanella algae was based upon whole-genome restriction with the ribosomal endonuclease I-CeuI followed by double hybridization with rRNA and qnrA DNA probes to the same high-molecular-weight band. This received further support from the identical G+C content of qnrA3 and the genome of S. algae and the lack of common region 1 that is part of the sul1-type integron, which was previously associated with the mobilization of qnrA1 onto plasmids (155). Subsequent in silico analyses identified qnr-like genes in the genomes of other members of the Shewanellaceae (153, 179), including Vibrio vulnificus (GenBank accession number AA007889), Vibrio parahaemolyticus (accession number BAC61438), Photobacterium profundum (accession number CAG22829), and other Vibrio and Shewanella spp. Similarly to S. algae, they are not flanked by insertion sequences or genetic structures known to mobilize resistance genes (153). The encoded proteins have at most 67% identity with QnrA1 and are made of two domains of 11 and 32 units connected by a single glycine. Upon expression in a heterologous E. coli host system, they conferred an increased MIC of quinolones similar to that obtained with the recombinant plasmid that expressed the qnrA3 determinant from S. algae (153). Similar observations were made for a qnr-like sequence in Vibrio parahaemolyticus (178). Although a qnr-like gene was not identified in Vibrio cholerae in the above-mentioned in silico analysis of other Vibrio spp. (153), another study examining a ciprofloxacin-resistant V. cholerae O1 clone from a cholera epidemic in Brazil identified a qnr-like gene encoding a protein with 69% sequence identity to that of the P. profundum Qnr-like peptide (65). The gene was designated qnrVC1, but according to recommended nomenclature (87), it might better be termed Vcqnr1. Vcqnr1 had a G+C content of 36.8%, which is considerably different from that of the V. cholerae genome (47.5%), supporting introduction through horizontal gene transfer. It also has an affiliated aatC recombination site, unlike other qnr genes, and is incorporated as a cassette into a chromosomal class 1 integron (65).
qnrB12 was found in three epidemiologically and clonally unrelated Citrobacter werkmanii isolates of poultry origin from Germany (104). A chromosomal location of this gene appeared most likely for a number of reasons: repeated plasmid transformation and conjugation experiments failed, Southern blot hybridization studies of I-CeuI-digested genomic DNA gave a signal only with the largest I-CeuI fragment (approximately 800 kb) in each strain, and further S1 nuclease digestion followed by pulsed-field gel electrophoresis did not identify any large qnrB-carrying plasmid which might comigrate with this 800-kb I-CeuI fragment (104).
Smqnr, identified in the genome of Stenotrophomonas maltophilia strain R551-3, codes for a 219-amino-acid protein that shares about 60% amino acid identity with QnrB. SmQnr has two domains of 5 and 28 pentapeptide repeats separated by a glycine (179, 182). An even closer homologue of qnrB has been found in DNA sequences in a marine metagenome, but its location on a bacterial chromosome or on a mobile genetic element is unknown (179).
In silico analysis of the genome sequence of Vibrio splendidus identified an open reading frame encoding a 218-amino-acid protein sharing 84% and 87% amino acid identities with QnrS1 and QnrS2, respectively. When overexpressed in E. coli cells, recombinant plasmids with the V. splendidus qnr genes conferred an eightfold increase in the MIC of nalidixic acid and 4- to 16-fold increases in the MICs of fluoroquinolones (25).
The vast majority of the pentapeptide repeat-containing proteins currently listed in the Pfam database of the Wellcome Trust Sanger Institute (PF00805) (www.sanger.ac.uk/cgi-bin/Pfam) are found in prokaryotes. However, the function of nearly all of these proteins is unknown. Three pentapeptide repeat proteins are of particular interest because they confer some level of quinolone resistance. The detailed mechanism of action of the PMQR protein Qnr is still unknown. Some lessons have been learned from studies of analogous pentapeptide repeat proteins that also confer fluoroquinolone resistance: McbG and MfpA.
McbG is a pentapeptide repeat protein sharing 19.6% amino acid identity with Qnr. It protects DNA gyrase against the effect of a microcin (69). Microcins are a class of small inhibitory proteins (less than 10 kDa) that vary in their mechanisms of action. One of these, microcin B17 (MccB17), is a bacterial poison that, like the quinolones, inhibits DNA gyrase (76) (though at a different site than that of the quinolones ). Organisms producing MccB17 also produce McbG, which protects them from the effect of this toxin, and mcbG has been found on resistance plasmids in clinical isolates (89). In host E. coli J53 cells, plasmid-mediated mcbG produced a slight increase in the MIC of sparfloxacin. Additionally, in the presence of mcbG, mutants with resistance to sparfloxacin arose at a concentration of sparfloxacin that was twofold higher than that for J53 R−. Thus, the pentapeptide repeat protein McbG, like Qnr proteins, appears to provide some antiquinolone protection. The effect, however, was much smaller than that of the Qnr proteins (89). Also, McbG did not appear to affect the susceptibility of E. coli J53 to ciprofloxacin or nalidixic acid.
MfpA, a pentapeptide repeat protein having 18.9% amino acid similarity to QnrA, has been more thoroughly studied. The mfpA gene was first identified on the chromosome of Mycobacterium smegmatis (127). When expressed on a multicopy plasmid, this gene resulted in an increase in the MIC of ciprofloxacin for this organism of between four- and eightfold, and the inactivation of the gene on the M. smegmatis chromosome resulted in increased ciprofloxacin susceptibility. The three-dimensional structure of this gene variant in Mycobacterium tuberculosis, MfpAMt, demonstrated that the pentapeptide repeat sequence encodes a right-handed quadrilateral β-helix (77). MfpAMt is notable among pentapeptide repeat proteins in that (like qnrA) the first residue in the pentapeptide (residue i−2) is often a cysteine (203).
It has been shown that MfpAMt inhibits ATP-dependent DNA supercoiling and ATP-independent relaxation reactions catalyzed by E. coli DNA gyrase. The apparent median inhibitory concentration (IC50) values were calculated to be ~1.2 μM (based on an active dimer) for both reactions in an assay containing 3 units of DNA gyrase.
Surface plasmon resonance is a technique of profiling the binding and dissociation of molecules. In brief, molecules immobilized on a sensor surface alter the refraction of polarized light. When a sample is passed over the sensor surface, and molecules interact with the immobilized molecules, the degree of light emission changes in proportion to the mass of bound material. Experiments employing this methodology indicated that MfpA interacts directly with DNA gyrase. Both the three-dimensional structure of this MfpA homologue and its charge distribution closely resemble those of B-form DNA. Based on these data, it was proposed that MfpA binds DNA gyrase in place of DNA. This displacement may generate some resistance to fluoroquinolones, as DNA gyrase bound to MfpA will not participate in the deleterious quinolone-gyrase-cleaved DNA complex that is the basis for quinolone cell killing (77).
Transconjugants of E. coli with qnrA do not demonstrate changes in quinolone accumulation, outer membrane porins, or drug inactivation (197); none of these are the mechanism of Qnr activity. Instead, the mechanism by which QnrA protects DNA gyrase from quinolones appears similar, but not identical, to that of MfpA.
The two quinolone targets, the type II topoisomerase enzymes DNA gyrase and topoisomerase IV, regulate conformational changes in DNA topology by catalyzing the breakage and rejoining of DNA strands during normal cellular growth (53, 207). The main function of DNA gyrase is to catalyze the ATP-dependent negative supercoiling of DNA, a unique property of this enzyme that is necessary for the initiation of DNA synthesis, which can be measured in a DNA supercoiling assay (84). In contrast, the main role of topoisomerase IV is to decatenate daughter replicons, measured in the decatenation of kinetoplast DNA (54). Both enzymes are heterodimers; DNA gyrase is composed of two A subunits and two B subunits, and topoisomerase IV is also an A2B2 enzyme composed of two ParC and two ParE subunits. The ParC subunit is homologous to GyrA (29% amino acid identity in E. coli), and ParE is homologous to GyrB (52).
Quinolones inhibit gyrase-mediated DNA supercoiling and topoisomerase IV-mediated DNA decatenation (71, 106). Purified N- or C-terminal histidine-tagged QnrA reversed this quinolone inhibition of DNA gyrase and topoisomerase IV activities in a dose-dependent manner. The median concentration of QnrA required to protect against the DNA gyrase-inhibitory action of 1.5 μM (0.5 μg/ml) ciprofloxacin was ~0.32 μM (197-199) (0.93 nM DNA gyrase, which corresponded to ~2 units). QnrA-His6 alone did not itself effect DNA supercoiling (197), nor did it inhibit DNA supercoiling even at a high concentration, unlike MfpA (77). QnrB1-His6 or QnrB4-His6 could also protect DNA gyrase in vitro, although at high concentrations, an inhibitory effect was seen (91, 125).
The mechanism of the Qnr protective effect is not completely understood. It has been shown through gel retardation assays that QnrA can bind to the DNA gyrase holoenzyme as well as to its respective subunits, GyrA and GyrB. This binding occurred in the absence of relaxed DNA, ciprofloxacin, or ATP, indicating that the binding of QnrA to gyrase did not require the presence of the ternary complex of enzyme, DNA, and quinolone (198). Similar findings were also reported for QnrA and E. coli topoisomerase IV (199). DNA filter binding assays have shown that DNA binding of gyrase decreased when the gyrase was exposed to Qnr (198). Thus, it was proposed that QnrA protection from quinolones is the result of QnrA binding to gyrase or topoisomerase IV at a site overlapping the DNA binding site. However, it is not clear how, in contrast to MfpA, QnrA might compete with DNA for gyrase binding without functionally inhibiting gyrase activity in vitro.
Recent data cast further doubt on the functional similarity between MfpAMt and Qnr proteins. The effect of purified MfpAMt and QnrB4 was investigated using various catalytic and noncatalytic type II topoisomerase enzyme assays (125). Histidine-tagged MfpAMt inhibited the catalytic activity of M. tuberculosis gyrase; IC50s for supercoiling, relaxation, and decatenation were 1.75 μM, 2 μM, and 2 μM, respectively, similar to previously reported results (77). MfpAMt also inhibited DNA supercoiling mediated by the E. coli gyrase, with an IC50 of 3 μM. In contrast, histidine-tagged QnrB4 did not inhibit the catalytic activity of E. coli or M. tuberculosis gyrase (unless concentrations were over 30 μM). In addition to suggesting that MfpAMt and QnrB4 interact with DNA gyrases in a different manner, this work also indicates that the mechanism of these proteins may be gyrase specific: QnrB4 protected E. coli but not M. tuberculosis gyrase from the inhibitory effect of fluoroquinolones; MfpAMt protected M. tuberculosis but not E. coli gyrase. Similarly, QnrA does not protect against the inhibition of gyrase by the GyrB-targeting agent novobiocin or MccB17 (G. A. Jacoby and D. C. Hooper, unpublished data).
In general, the acquisition of a qnrA-bearing plasmid will not render a wild-type organism fluoroquinolone insusceptible according to CLSI clinical breakpoints. The extent to which QnrA protects isolates of Enterobacteriaceae against fluoroquinolones has usually been examined by measuring the difference in quinolone MICs for an E. coli strain with and without a qnrA-bearing plasmid. The first report of a qnrA plasmid found that the MIC of ciprofloxacin increased from 0.008 μg/ml to 0.25 μg/ml in an E. coli J53 transconjugant, with a range from 0.125 μg/ml (212) to 2.0 μg/ml (211) for other qnr plasmid transconjugants of this strain. One study assessed the quinolone resistance conferred by 17 clinical qnrA-bearing plasmids (Table (Table1).1). Donor bacteria originally harboring these plasmids all had exhibited higher levels of resistance to quinolones than the transconjugants, suggesting that additional mechanisms of quinolone resistance frequently coexist with qnrA. There were also differences among transconjugants in the qnrA effect on fluoroquinolone MICs. Although for most agents, the presence of a qnr plasmid increased their MIC by between 16-fold and 125-fold, this increase was less (16-fold to 32-fold) for sitafloxacin. The agent for which the loss of activity was least pronounced was nalidixic acid (twofold to eightfold increases in MIC) (211). Illustrating this phenomenon, Hopkins et al. found that in non-Typhi Salmonella isolates, a phenotype of reduced susceptibility to ciprofloxacin (MIC > 0.06 μg/ml) but preserved susceptibility to nalidixic acid (MIC ≤ 16 μg/ml) identified qnr-positive strains (82, 83). Also noteworthy is the finding that some qnrA-carrying plasmids from U.S. K. pneumoniae isolates yielded transconjugants with very similar quinolone susceptibilities (211), whereas other qnrA-carrying plasmids from U.S. and Chinese isolates of Enterobacteriaceae varied in ciprofloxacin susceptibilities by 16-fold (212). There are several reasons for this phenomenon. In some cases these differences resulted from the presence of an additional resistance determinant, aac(6′)-Ib-cr, on some plasmids (165, 220). For other strains, the copy number and especially the transcriptional level of the qnr genes affected quinolone resistance (174, 220).
qnrS- and qnrB-carrying plasmids confer quinolone resistance that is similar to that conferred by qnrA1 (Table (Table1).1). When cloned into a derivative of E. coli DH10B, qnrS1 increased the MICs of nalidixic acid, ciprofloxacin, and ofloxacin 8-, 83-, and 24-fold, respectively. These changes led only to nalidixic acid resistance by CLSI breakpoints (26). The impact of some qnr allele variants on quinolone MICs has additionally been examined. Overall, the patterns of resistance have been similar, with 2- to 8-fold and 8- to 32-fold increases in MICs of nalidixic acid and ciprofloxacin, respectively (17, 23, 70, 91, 126, 161, 166).
MIC studies assess the effect of a resistance gene on growth inhibition by an antimicrobial agent. There are other indices by which the effect of a resistance gene can be assessed. A time-kill study has examined the bactericidal activities of ciprofloxacin and ofloxacin in the presence of QnrA. Despite the fact that QnrA protects against quinolone growth inhibition, it did not block the bactericidal activity of these quinolones at concentrations of twice the MIC or greater (119, 169).
Another measure of resistance gene effect is a change in the mutant prevention concentration (MPC). The MPC is the lowest concentration of quinolone required to prevent the growth of quinolone-resistant mutants from a starting inoculum of about 1010 bacteria (a large inoculum is used to ensure the presence of single-step mutants occurring at a low frequency). When the quinolone concentration remains above the MPC, single-step resistant mutants are unlikely to arise. The early finding that QnrA facilitated the recovery of mutants with higher levels of quinolone resistance prompted an assessment of its effect on the MPC of ciprofloxacin. The MPC for wild-type E. coli J53 is 0.125 μg/ml; E coli J53 carrying a qnrA plasmid has an MPC that is more than 10-fold greater (88) (Fig. (Fig.2).2). Thus, the low-level resistance conferred by these mechanisms might not allow a population of bacteria to survive in the presence of a quinolone. However, it substantially enhances the number of strains with resistance mutations that can be selected from the population, as also occurs with chromosomal quinolone resistance mutations. In the case of QnrA, this phenomenon has been experimentally shown both for E. coli (123, 173) and for Enterobacter spp. (164) and probably holds true for other genera as well. Indeed, data from a pharmacodynamic model suggested that Providencia stuartii with qnrA (but not without it) is insufficiently killed by a large single ciprofloxacin dose and rapidly acquires resistance (215).
By increasing the MIC and widening the mutant selection window (55), Qnr may thus lead to decreased therapeutic efficacy. Rodriguez-Martinez et al. recently demonstrated this effect in a murine model of K. pneumoniae pneumonia (169). The K. pneumoniae strain used in that study lacked the porins OmpK35 and OmpK36, had an active efflux system for quinolones, and carried a resistance mutation in gyrA and therefore magnified the protective effect of qnrA1 to the maximum. All 30 animals infected with the qnrA mutant organism and treated with a fluoroquinolone survived. In contrast, the survival rate among 30 animals infected by a pMG252 (qnrA+) transconjugant of that strain was 53% (P < 0.03). Similarly, the clearance of bacteria from the lungs was affected by qnrA; the log10 CFU/g of lung in untreated, qnrA+ strain-infected, and qnrA-negative strain-infected mice were ca 9.16, 7.74, and 3.53, respectively (P < 0.001) (169). Whether qnr would have a similar effect on a strain without enabling mutations is not yet known.
qnrA has frequently been observed in the company of other resistance mechanisms in clinical strains and is capable of acting additively with these mechanisms. pMG252, the plasmid on which qnrA1 was originally identified, was introduced into E. coli strains containing a variety of chromosomal mutations that enhanced or diminished resistance through alterations in DNA gyrase, topoisomerase IV, efflux, or outer membrane porin channels (92, 122, 173). The presence of qnrA was found to supplement all types of mutation-based resistance. Interestingly, the presence of a qnr gene may affect the distribution of chromosomal resistance mutations that are selected in the presence of fluoroquinolones. In a recent study, quinolone-resistant mutants were selected from wild-type E. coli strains and corresponding transconjugants harboring qnrA1, qnrA3, qnrB2, or qnrS1 (34). Although the proportion of resistant mutants selected was the same, fewer qnr-positive transconjugants yielded clones with a quinolone resistance-determining region mutation. The cause of this change in mutant distribution is not known but might be explained if Qnr binding results in alterations in the positioning of quinolones in relation to the gyrase-DNA complex (34).
A few studies have found bacteria harboring more than one qnr gene. This occurrence has been usually but not exclusively qnrS with either qnrB or qnrA (27, 85, 110, 217). Whether multiple Qnr proteins have an additive effect on the MIC is unclear. In one example of the cooccurrence of qnrB4 and qnrS1, transconjugation experiments showed that the effect of both genes on ciprofloxacin MIC was the same as that of one alone, suggesting that the two gene products may compete for binding to gyrase (27, 85).
Genes for quinolone resistance have been found on plasmids varying in size and incompatibility specificity (Table (Table2),2), indicating that the spread of multiple plasmids has been responsible for the dissemination of this resistance around the world. The immediate genetic environment of each gene type, however, is similar enough to suggest a limited number of acquisition events followed by transposition, recombination, replicon fusion and resolution, and deletion and insertion of DNA to generate the diversity of plasmid structures seen today. Sometimes the plasmids are so similar that a broad dissemination of the same plasmid does appear likely.
For example, qnrA1 plasmids can vary in size from 20 to 320 kb (17, 112) and belong to at least three plasmid incompatibility (Inc) groups (157). qnrA1 is usually associated with ISCR1 (formerly orf513) (193), although 6.2% of qnrA1-positive strains in a study from South Korea were negative for ISCR1 by PCR (142). Usually, a single copy of ISCR1 is found downstream from qnrA1, but in pMG252 and related plasmids, the qnrA1 gene is bracketed by two copies of ISCR1 (163, 172). The qnrA1 ISCR1 complex is inserted in turn into a sul1-type integron containing several other resistance gene cassettes. Some examples are shown in Fig. Fig.3.3. Genes for extended-spectrum and AmpC β-lactamases are often found on the same plasmid. qnrA1 was discovered in an isolate from Alabama on plasmid pMG252 also expressing the uncommon FOX-5 β-lactamase. Plasmids carrying qnrA1 and FOX-5 have subsequently been found in specimens from Delaware, Kentucky, New York, North Carolina, and Tennessee (168, 210) and from as far away as Brazil (21). In isolates from Canada, France, Thailand, and Turkey, qnrA1 is associated with VEB-1 β-lactamase (120, 131, 154, 156), while in samples from other countries, qnrA1 and extended-spectrum β-lactamase (ESBL) SHV-12 and various CTX-M enzymes are linked (17, 30, 47, 56, 95, 111, 112, 136, 138, 152, 170, 189, 190, 212, 217, 223).
qnrB1 was discovered on a 340-kb multiresistance plasmid from India (91) and was subsequently reported for isolates from Algeria (86), Denmark (195), South Korea (142), Nigeria (185), and Scotland (130). It is associated not with ISCR1 but with another putative transposase, Orf1005 (91, 185). The qnrB2 allele has an even wider distribution, having been found originally in isolates from the United States (91) and subsequently from China (223), Brazil (126), Hungary (189), Israel (40), South Korea (142), The Netherlands (200), Portugal (158), Scotland (130), Senegal (68), Switzerland (113), Taiwan (218), and the United Kingdom (82). qnrB2 is linked to a single copy of ISCR1 in some plasmids (60), while in other plasmids, duplicate copies of ISCR1 surround the gene (68) (Fig. (Fig.3).3). As with qnrA1 plasmids, the ISCR1 complex is inserted into a sul1-type integron. β-Lactamase genes associated with qnrB2 include CMY-1 (138), CTX-M-14 (223), CTX-M-15 (113, 158), IMP-4 (60), IMP-8 (218), KPC-2 (40), and SHV-12 (91); qnrB3 has been reported only from the United States (166), but qnrB4 has a worldwide distribution and a strong linkage with the gene for the AmpC-type β-lactamase DHA-1, with plasmids in strains from China (223), France (202), South Korea (138, 190), Switzerland (113), Taiwan (218), and the United States (G. A. Jacoby, unpublished data) showing this association. The qnrB4 allele has also been associated with, or linked to, the ESBLs SHV-12 (27, 138), TEM-52 (138), CTX-M-3 (95, 113), CTX-M-9 (113), CTX-M-14 (95, 138), and CTX-M-15 (113). In addition, qnrB4 alleles are closely linked to eight genes or pseudogenes unrelated to antibiotic resistance: sapA and sapB (both peptide transport system permeases); sapC, pspA, pspB, pspC, and pspD (all phage shock proteins); and pspF (a transcriptional activator for the psp operon) (Fig. (Fig.3).3). sapA-like and pspF-like genes surround qnrB1 (91), qnrB2 (60, 68, 91), qnrB10 (161), and probably other qnrB alleles, while the other six sap and psp genes appear to be missing.
qnrB5 has been found in K. pneumoniae (20, 138) but seems especially common in serovars of S. enterica (70, 82, 195). With the exception of qnrB19, the remaining qnrB alleles have been too recently described for much to be known about their plasmid locations. The qnrB19 allele was found on a 40-kb plasmid from an E. coli strain isolated in Colombia, South America, in a transposon comprising ISEcp1 and qnrB19 inserted into another transposon carrying a gene for tetracycline resistance (23). Within the year of its publication, qnrB19 was reported for an S. enterica serovar Typhimurium strain from The Netherlands (67), commensal enterobacteria from children living in Bolivia and Peru (140), and a K. pneumoniae isolate on an 80-kb plasmid from the United States (58). In the U.S. isolate, a transposon composed of ISEcp1 and qnrB19 and another one containing blaKPC-3 were inserted into a third transposon related to Tn3, forming what has been termed the KQ element (162).
Unlike qnrA and qnrB, qnrS genes are not associated with ISCR1, nor are they part of complex integrons. More qnrS plasmids are small and nonconjugative (14, 67, 105, 216, 217) than are qnrA or qnrB plasmids, and they are also more likely to be found in S. enterica serovars and not to carry ESBL or AmpC β-lactamase genes. In several qnrS1 plasmids (103), including pAH0376, in which qnrS was discovered (75), the gene is found near Tn3 encoding TEM-1 β-lactamase. In other plasmids of similar structure, the neighboring gene is blaLAP-1 (151), blaLAP-2 (67), or mobABC (105). An insertion sequence variably termed ΔIS2 (105) or ISEc12 (151) lies downstream from qnrS1, and elements of IS26 also surround it in plasmid pINF5 (105). The qnrS2 allele has been found in two different genetic environments. In small plasmids in isolates from the United States (70) and Germany (14), open reading frames related to replication genes are located immediately downstream from qnrS2, while in plasmids from Aeromonas spp., qnrS2 is part of a transposon-like structure inserted into a putative metalloprotease gene (24, 149).
It would seem likely from the impressive number, variety, and geographical and bacteriological penetration of qnr variants that these or similar genes existed for a considerable time before the report of their discovery in 1998 (105, 123). This raises questions about from where these genes originated and what purpose they served before being recruited to protect bacteria from antimicrobial agents.
Postulating that qnr genes originated on the chromosome of an organism occupying a human, veterinary, or environmental reservoir, Poirel and colleagues (155) screened the genome sequences of 48 gram-negative species from a wide range of genera for qnrA. Variants of qnrA (qnrA3 to qnrA5) were located on the chromosome of Shewanella algae. The quinolone MICs of this organism were four- to eightfold higher than those of Shewanella putrefaciens, a closely related organism lacking a chromosomal qnrA gene (155). These data suggest that S. algae is a possible reservoir of qnrA. Shewanella spp. are water dwellers present in both marine and freshwater environments. Recently, qnrA3-positive Klebsiella pneumoniae and Kluyvera ascorbata strains isolated from the feces of an immunocompromised outpatient in Paris were analyzed (109). The sequence immediately downstream from qnrA3 was identical to that found downstream from qnrA3 in the S. algae chromosome, supporting the notion that qnrA3 has been excised from chromosomal DNA of S. algae or similar organisms (109).
As noted previously, pentapeptide repeat proteins showing 40 to 67% amino acid identity to qnrA were present in other waterborne Shewanella spp. (153, 178). Additional studies found qnrB-like and qnrS2 genes in water and other environmental isolates.
Sánchez et al. and Venter et al. found a gene encoding a pentapeptide repeat in a sequence database of microbial populations collected en masse from seawater samples collected from the Sargasso Sea (179, 201). The protein, designated marine metagene Qnr (GenBank accession number AACY020347520), was 88% similar to QnrB5 and QnrB19, but its functionality was not examined. We note that sequence analysis shows that metagene qnr is 98% similar to qnrB8, and the first 214 amino acids are 99% identical to QnrB8. In addition, BLAST analysis revealed the same sapA sequence that is closely linked to qnrB genes on plasmids (see above) downstream of metagene qnr, further supporting the aquatic origin of this PMQR determinant.
qnrS2 genes have also been found on plasmids carried by environmental organisms. In 2006, Cattoir et al. sampled water from urban sites in the Seine river and found Aeromonas punctata subsp. punctata and Aeromonas media strains carrying plasmids that transferred quinolone resistance (24). These plasmids were shown to carry qnrS2 (24). The same gene has also been found in plasmids from Aeromonas allosaccharophila found in Lake Lugano (a Swiss-Italian vacation area) in 2005 (149) and, together with a Tn1721-like transposon, on plasmid pGNB2 in sludge basin bacteria (from a German wastewater treatment plant) in 2004 (14). This plasmid also conferred decreased susceptibility only to quinolones (14).
Collectively, these findings suggest that some qnr genes in circulation likely originated in the chromosomes of water-dwelling environmental organisms. The recent detection of qnr-bearing plasmids in water organisms suggests that freshwater in inhabited areas may be a reservoir in which pathogens acquire these elements. Quinolones are excreted unmetabolized by mammals into wastewater. The carboxylic acids of quinolones are degradable by sunlight in aqueous solution (194), but when these compounds enter the aquatic environment via sewage water, photodegradation may be of only minor importance. Elimination can also occur via adsorption to sediments, degradation by terrestrial fungus (145), or environmental Mycobacterium spp. (3), but a substantial quantity of active drug may remain in the environment (5). For example, ciprofloxacin and other fluoroquinolones have been found at concentrations of up to 0.005 μg/ml in water sources, including water downstream from a wastewater treatment plant in the United States and in the Seine River in France (13, 191). Thus, it is possible that environmental quinolone accumulation has contributed to the success of these genes, perhaps by helping to maintain a reservoir of aquatic organisms for which a low-level quinolone resistance gene provided a survival advantage.
The native function of qnr genes is unknown. Ellington and Woodford postulated that Qnr could be an antitoxin, protecting DNA gyrase and topoisomerase IV from some naturally occurring toxins (57). There are known natural DNA gyrase toxins. These include CcdB, a toxin encoded on the F plasmid (49), ParE (distinct from the ParE subunit of topoisomerase IV), located on the broad-host-range RK2 plasmid (94), and MccB17, a posttranslationally modified peptide produced from the plasmid-borne MccB17 operon (76, 144). Each of these toxins is paired with an inhibitor (antitoxin) that protects cells from death and is encoded in the same operon: CcdA (49), ParD (94), and McbG (which is in fact a pentapeptide repeat protein) (76, 175). Qnr may thus be analogous to one of these antitoxins. However, the hypothetical toxin against which Qnr protects has not been demonstrated. Qnr may alternatively serve a function similar to that of the chromosomally encoded non-pentapeptide-repeat protein GyrI, a DNA gyrase regulator that also is capable of some antitoxin and antiquinolone effect (35, 36).
Notably, plasmid-carried qnrB alleles have been shown to have upstream LexA binding sites, leading to increased qnrB expression levels upon exposure to quinolones, mitomycin, and possibly other DNA-damaging agents as part of the SOS response (118, 209, 225). Thus, if such induction is also seen in an organism with a chromosomal qnrB progenitor, it suggests that QnrB may have a native function in protection from naturally occurring DNA-damaging agents.
Several years after the discovery of QnrA, our group was investigating the phenomenon of inequality in the level of quinolone resistance transferred with different qnr plasmids. Wild-type E. coli strains have an MIC of ciprofloxacin of about 0.008 μg/ml, and most qnr plasmids determine an MIC of ciprofloxacin of 0.25 μg/ml for E. coli. We noted that certain plasmids from clinical E. coli strains collected in Shanghai provided about fourfold-higher levels of ciprofloxacin resistance (1.0 μg/ml). We found that this high-level resistance was not caused by an increased level of expression of qnrA, as it has been with other plasmids (174, 220). By random transposon mutagenesis of plasmid DNA, we found that the gene responsible for the incremental resistance was an aminoglycoside acetyltransferase, aac(6′)-Ib, which confers resistance to tobramycin, amikacin, and kanamycin (165). Sequencing showed this allele to be unique among the approximately 30 known variants of aac(6′)-Ib in two codon changes, Trp102Arg and Asp179Tyr, which we found to be necessary and sufficient for the ciprofloxacin resistance phenotype. An acetylation assay showed the capacity of this AAC(6′)-Ib variant [which we designated AAC(6′)-Ib-cr, for ciprofloxacin resistance] to acetylate ciprofloxacin at the amino nitrogen on its piperazinyl substituent (165).
The increase in MIC conferred by AAC(6′)-Ib-cr is smaller than that conferred by Qnr proteins, and as predicted by its specific quinolone target, it was selective only for ciprofloxacin and norfloxacin, which both have piperzinyl secondary amines. Other quinolones lacking an unsubstituted piperazinyl nitrogen were unaffected. Although the increase in the MICs of ciprofloxacin and norfloxacin was modest (threefold to fourfold), the effect on the MPC was marked (Fig. (Fig.2).2). In the presence of aac(6′)-Ib-cr, resistant clones of wild-type E. coli strain J53 could still be recovered at concentrations of 1.6 μg ciprofloxacin per ml, a level approximating the peak serum concentration of free ciprofloxacin during therapy (165).
Kinetic studies of purified AAC(6′)-Ib and its cr variant indicated that the mutant enzyme had only slightly reduced efficiency (relative to that of the wild-type enzyme) for the acetylation of kanamycin (204). The acetylation of ciprofloxacin, although less efficient than that of kanamycin, was sufficient in bacterial cells to produce a reduced-susceptibility phenotype equivalent to that of cells exposed to chemically synthesized N-acetyl ciprofloxacin (165), suggesting that complete ciprofloxacin acetylation had occurred under conditions of bacterial growth. Dead-end quinolone (pefloxacin) and aminoglycoside (lividomycin) substrates, for which the target sites of drug acetylation are either blocked (pefloxacin) or absent (lividomycin), both produced a competitive inhibition of acetylation of the true substrates kanamycin and ciprofloxacin, indicating a functional overlap in the binding sites for both classes of substrate of AAC(6′)-Ib-cr (204).
The X-ray crystallographic structure of the wild-type enzyme with kanamycin and acetyl coenzyme A has been solved and allowed the construction of a molecular model of ciprofloxacin binding to the cr variant (204). In the model, the Asp179Tyr mutation had the greatest effect, resulting in a π stacking interaction with the quinolone rings to enhance drug binding. The Trp102Arg mutation was more distantly positioned, serving to stabilize the positioning of Tyr179. This model is consistent with the magnitude of the effects of the individual mutations, with Asp179Tyr having a partial resistance phenotype, Trp102Arg having little detectable resistance phenotype, and the two mutations together having the full resistance phenotype (165). An alternative model of positioning of ciprofloxacin that emphasizes plasticity in the enzyme active site proposes that Arg92 (equivalent to Arg102) has a direct interaction with the carboxyl group of ciprofloxacin (124) but does not account for the relative effects of the two individual mutations. A direct structural analysis of AAC(6′)-Ib-cr with both acetyl coenzyme A and ciprofloxacin substrates is needed.
aac(6′)-Ib-cr, like its parent aac(6′)-Ib, is in an integron cassette with an associated attC site. It is hence found in various integrons, some of which are shown in Fig. Fig.3,3, but especially on IncF11 plasmids expressing CTX-M-15 that have spread rapidly so that CTX-M-15 has become the predominant ESBL in many countries around the world (15, 46, 63, 102, 117, 139, 150, 158, 176). aac(6′)-Ib-cr has been associated with other PMQR genes including qnrA1 (95, 165, 220), qnrB2 (158, 223), qnrB4 (95, 116), qnrB6 (95, 116), qnrB10 (161), qnrS1 (95, 116), qnrS2 (149), and qepA (116) and with other β-lactamases including CTX-M-1 (185), CTX-M-14 (95), CTX-M-24 (95), DHA-1 (116), SHV-12 (116), and KPC-2 (39).
Antimicrobial efflux mechanisms can act on single or multiple agents and can be carried on both plasmids and chromosomes. Numerous such mechanisms have been described, and they are increasingly being recognized as important determinants of antimicrobial resistance in a variety of organisms (160). Efflux determinants of quinolone resistance in gram-negative bacteria are largely multidrug transporters of the resistance-nodulation-cell division (RND) family encoded by endogenous chromosomal genes (159). This family of traditionally chromosomal efflux pumps is ubiquitous in gram-negative bacteria, serving to remove unwanted compounds from the cytoplasm and membrane. Two plasmid-mediated quinolone transporters have now been found: OqxAB and, more recently, QepA.
A conjugative plasmid, pOLA52, conferring resistance to the antibiotic olaquindox (a quinoxaline derivative that is used in agriculture as a veterinary growth promoter) was found in E. coli strains isolated from swine manure (186). The resistance mechanism was identified to be a multidrug efflux pump, OqxAB, which belongs to the RND family (73). It confers resistance to other agents as well, including chloramphenicol. Upon expression in an E. coli strain lacking a native acrA gene, pOLA52 conferred 8- and 16-fold increases in the MICs of nalidixic acid and ciprofloxacin, respectively (72). In the single prevalence study performed, OqxAB was rare. Ten of 556 (1.8%) E. coli strains isolated between 1995 and 1998 in Denmark and Sweden were shown to have an MIC of olaquindox of ≥64 μg/ml; in 9 of the 10 strains, the oqxA gene was detected (74). Plasmid-mediated OqxAB was recently detected in a human clinical E. coli isolate from South Korea. oqxAB genes are also present on the chromosome of K. pneumoniae, with different levels of expression being correlated with differences in susceptibility to olaquindox (108).
The novel efflux pump QepA was found to be encoded on plasmid pHPA, discovered in an E. coli strain from a urine specimen from an inpatient in Hyogo Prefecture, Japan, in 2002 (222). This plasmid displayed a multiple-resistance profile for aminoglycosides, fluoroquinolones, and broad-spectrum β-lactams. qepA encoded a 511-amino-acid protein putatively belonging to the 14-transmembrane-segment major facilitator superfamily of transporters. The G+C content of QepA is 72%, higher than that of the chromosomes of Enterobacteriaceae (50%). Interestingly, according to phylogenetic analysis, QepA belongs to the 14-transmembrane-segment family transporters of gram-positive Actinomycetales but not those of gram-negative bacteria (222). Yamane and coworkers found that qepA cloned into pSTV28 increased the MICs of several compounds in an E. coli transconjugant (222). The MICs of nalidixic acid, ciprofloxacin, and norfloxacin increased 2-, 32-, and 64-fold, respectively; a twofold increase was observed for erythromycin, acriflavine, and ethidium bromide, whereas the MICs of other antibiotic classes and common transporter substrates did not change (222). Since its discovery, a variant of qepA possessing two amino acid substitutions has been found. This variant (named QepA2) conferred a phenotype similar to that of the QepA determinant (now renamed QepA1) (28).
Sequence analysis revealed that the qepA gene is located in a ~10-kb region, with a flanking sequence that is in large part identical in different plasmids (28, 147, 222). However, qepA1-positive isolates from Japan (221, 222), Belgium (148), and South Korea (41, 143) were flanked by two copies of IS26 and associated with the rmtB gene, encoding an aminoglycoside ribosomal methylase, whereas qepA2 was flanked by a novel insertion sequence element (ISCR3C), and rmtB was not found (28).
The only oqxAB plasmid analyzed to date is a 52-kb IncX1 plasmid from an E. coli isolate from Denmark (186) that has been completely sequenced (134). qepA1 has been found on 113- and 168-kb IncF1 plasmids from Belgium (147, 148), while qepA2 was discovered on a 90-kb IncF1 plasmid from France.
As a rule, the resistance phenotype does not distinguish between PMQR and other resistance mechanisms. PMQR genes confer low-level quinolone resistance that is below the CLSI breakpoint for nonsusceptibility, similar to that conferred by first-step DNA gyrase mutations, transporters that extrude quinolones, and decreased levels of expression of porins. The phenotype of low-level nalidixic acid resistance and reduced ciprofloxacin susceptibility sometimes observed among qnr-positive strains is neither sensitive nor specific. Thus, screening for qnr genes is generally done by PCR amplification of the target genes. To facilitate a higher throughput, a number of groups have employed multiplex PCR for the detection of qnrA, qnrB, and qnrS (29, 166). Later, additional sets of primers were used, including multiplex PCR with degenerate primers for qnrB, to overcome the wider variability within this group (29). No method is perfect. For example, primers which we have used did not fully match all qnrB genes, and the reverse primer mismatched at the 3′ end with qnrB5, qnrB10, and qnrB19. qnrB5 was, however, detected using these primers (70). Similarly, the reverse degenerate primer for qnrB (29) did not completely align with qnrB17. Also, false-positive amplicons have been reported for multiplex PCR procedures that are not seen with monoplex PCR using the same individual primer pairs within the multiplex primer mixture, emphasizing the importance of extensive DNA sequence confirmation (107). Because a comparison of the detection methods has not been performed, the prevalence studies were potentially subject to detection bias.
Because the difference between aac(6′)-Ib-cr and aac(6′)-Ib is in only two nucleotides, screening for aac(6′)-Ib-cr has traditionally involved aac(6′)-Ib amplification followed by sequencing (141) or restriction analysis (150). To overcome this cumbersome and costly method, we recently employed the gap-ligase chain reaction for the G-to-T change in the cr variant at nucleotide 535, one of the two defining mutations of aac(6′)-Ib-cr, to screen a large database for this gene (213). In this technique, two same-directional primers, separated by a gap of several nucleotides, are chosen. These primers hybridize to complementary strands of target DNA and will be extended by a DNA polymerase and subsequently ligated into a single long oligonucleotide when the mutation of interest, which corresponds to the 3′ end of the first primer, is present. This oligonucleotide can then be amplified (1). All aac(6′)-Ib-cr-positive control strains were identified, and by optimizing the assay conditions, we were able to use unquantified extracts of whole-cell DNA and maintain specificity (213).
After the initial discovery of qnrA in a K. pneumoniae isolate obtained from a urine sample from a patient in Alabama in 1994, efforts were made to find this gene elsewhere. A survey for qnrA by PCR of more than 350 gram-negative isolates collected mainly in the 1990s and chosen to include a broad geographic range and a variety of genera of gram-negative bacteria found qnrA in only six isolates (four E. coli and two Klebsiella sp. isolates), all from the same center in Alabama where the original strain had been detected and all collected between July and December 2004 (89). Since that early study, more epidemiological surveys have been reported (Table (Table3).3). The earliest known qnr alleles are a qnrB8-like gene in a C. freundii isolate from Brooklyn, NY, and a qnrB9-like gene in a K. pneumoniae isolate from Cordoba, Argentina, both collected in 1988 (90). In general, studies have been localized to a narrow geographical region and a limited range of genera. Most studies used PCR methodologies to examine clinical isolates of Enterobacteriaceae collected in the 1990s or early 2000s, spanning periods ranging from a few months to more than a decade. Through 24 November 2008, more than 70 publications in peer-reviewed journals and conference abstracts reported over 20,960 isolates that were tested for PMQR. The average prevalences of qnrA, qnrB, qnrS, and aac(6′)-Ib-cr in this compiled database were 1.5%, 4.6%, 2.4%, and 10.8%, respectively. Early surveys looked for qnrA, and in recent studies, qnrS, qnrB, aac(6′)-Ib-cr, and qepA were also included. E. coli has been the most common species screened for PMQR. However, in the vast majority of surveys, qnr was more prevalent among Enterobacter spp. and Klebsiella spp. than in E. coli strains (17, 27, 93, 95, 98, 112, 120, 126, 141, 142, 152, 156, 166, 206). On the other hand, from surveys that included a balanced distribution of isolates of Enterobacteriaceae, it was evident that aac(6′)-Ib-cr is most common among E. coli strains (95, 141, 176, 223).
The selection criteria for isolates included in surveys could potentially bias prevalence data. Surveys generally have been performed with isolates collected over a short period or isolates that are resistant to various drugs, most commonly quinolones or ESBLs. In some studies, strains were collected only in the context of an outbreak. Hence, prevalence data in general must be interpreted with caution. It does appear to be the case, however, that the incidence of PMQR genes has increased in recent years. Several studies were conducted over sufficiently long periods to assess trends (6, 17, 27, 107, 120, 150, 152, 156, 188, 218). For example, in a survey in Paris, France, no qnr genes were found in 2002, and 10 were found in 2005 (120, 152, 156) (27). In addition, 41 of 1,147 K. pneumoniae bloodstream isolates collected in Taiwan from 1999 through 2005 were qnrB4 positive; no qnrB4 genes were found in 1999 to 2000, but in 2005, 14 isolates (7.6%) were positive (218). In a cohort of clinical Enterobacter sp. isolates from Jerusalem, Israel, that were collected from 1990 through 1993, none of 94 isolates had qnr; in isolates from 1994 through 2005, 33 out of 485 (6.8%) isolates had qnr (P < 0.01). Findings were similar for K. pneumoniae (188).
A similar picture is seen with aac(6′)-Ib-cr. The cr variant was not found among 150 aac(6′)-Ib-positive strains collected between 1981 and 1991 (90). In a study from Ljubljana, Slovenia, from 2000 through 2002, aac(6′)-Ib-cr was detected in 1 of 17 Klebsiella isolates, whereas in 2003 to 2005, the prevalence increased significantly to 24 of 57 isolates (P = 0.02) (6). In Calgary, Canada, the prevalence of aac(6′)-Ib-cr significantly increased from 5 of 121 isolates (4.1%) in 2004 to 52 of 346 E. coli strains (15%) isolated in 2007 (P = 0.001) (150). In Jerusalem, E. coli isolates bearing aac(6′)-Ib-cr emerged in 1998, and since then, aac(6′)-Ib-cr has progressively penetrated into multiple clinical E. coli clones (213). A recent survey of the qnr, aac(6′)-Ib-cr, and qepA genes among 461 unselected, consecutive bloodstream isolates collected in Seoul, South Korea, in two periods, 1998 to 2001 and 2005 to 2006, corroborate the findings from prior studies. There was a significant increase in the rate of ciprofloxacin resistance over time (P < 0.001), and the overall prevalence of PMQR genes increased significantly over time as well (P = 0.02). Although qnrB was the most prevalent PMQR gene, there was an overall increase in the prevalence of PMQR genes, representing an increasing diversity of PMQR genes rather than a dominance of a single gene (107). Also noteworthy in that study was the finding that in strains with PMQR genes, there was a significant increase in gyrA and/or parC resistance mutations over time, but in strains without PMQR genes, these mutations remained stable over time. This epidemiological association supports the role of PMQR genes in promoting higher levels of resistance by mutation in clinical settings, as has been demonstrated in the laboratory.
qnr genes have already been found in all populated continents (as well as the waters in between them) and in most clinically common Enterobacteriaceae. These species include E. coli, Klebsiella spp. (K. pneumoniae and Klebsiella oxytoca), Enterobacter spp. (Enterobacter cloacae, Enterobacter aerogenes, Enterobacter amnigenus, and Enterobacter sakazakii), Citrobacter freundii, and Providencia stuartii. Among these, qnr has been more commonly identified in Enterobacter spp. followed by K. pneumoniae and less so in E. coli, where aac(6′)-Ib-cr seems to be more prevalent. Until recently, absent from this list were Proteus spp. and clinically important nonenteric gram-negative bacteria (e.g., Pseudomonas aeruginosa and Acinetobacter spp.). All three have been included in small surveys of isolates of human origin (78, 89, 156, 181, 208), without PMQR genes being detected. qnr was also not found among multidrug-resistant Acinetobacter baumannii isolates (2, 224). Interestingly, in a recent survey of isolates from zoo animals, qnrB was detected in Pseudomonas fluorescens from a turtle and in Proteus mirabilis from feces of Bengalese finches (4). QnrA has also recently been found in an isolate of A. baumannii (196).
Epidemiological studies have been useful in supporting genetic data indicating a linkage of PMQR with other resistance genes, particularly ESBLs. Various investigators have demonstrated higher qnr prevalences among ESBL-positive strains (142, 188, 217, 223). For example, Strahilevitz and coworkers found that among clinical isolates of Enterobacter spp. and Klebsiella pneumoniae, the relative risk for ceftazidime resistance (a surrogate for ESBL presence) in qnr-positive K. pneumoniae isolates was 1.8 (95% confidence interval, 1.3 to 2.5), and in Enterobacter isolates, it was 3.5 (95% confidence interval, 2.7 to 4.5) (188). Further work demonstrated that the ceftazidime resistance in qnr-positive Enterobacter strains was associated with a true ESBL-mediated mechanism (188). Because fluoroquinolones remain one of the few options for treating infections caused by such organisms, it is concerning that a substantial fraction of the ceftazidime-resistant qnr-carrying isolates in this study were susceptible to ciprofloxacin according to CLSI criteria (188). Similarly, in a French survey of ESBL-positive isolates of Enterobacteriaceae, 43% of qnrA-positive isolates tested as ciprofloxacin susceptible. Whether such qnrA-positive, fluoroquinolone-“susceptible” isolates can be effectively treated with fluoroquinolones requires further investigation.
A further concern arising from epidemiological studies is the close association of aac(6′)-Ib-cr with CTX-M-15, an ESBL that has emerged worldwide in recent years, including in community settings (45, 150).
Despite continued efforts to control its spread, non-Typhi Salmonella enterica persists as the most common food-borne pathogen in the United States (33). Unlike most Enterobacteriaceae species in which qnr genes have been detected, non-Typhi serotypes of Salmonella enterica are carried largely in the intestinal tract of food animals and are transmitted to humans through the food chain (8). Therefore, quinolone use in agriculture may drive the dissemination of qnr-mediated resistance in these pathogens, and mapping of qnr in non-Typhi serotypes of S. enterica could serve as a marker of the route of infiltration of antibiotic resistance from the food animal industry to humans (16, 61). Over 2,300 non-Typhi S. enterica isolates from human cases and animals in North America, Europe, and Asia have been tested (30, 38, 50, 67, 70, 83, 104, 184, 200, 216). qnrA, qnrB, qnrS, and aac(6′)-Ib-cr were found in the majority of the surveys, at overall prevalences of 0.2%, 1.0%, 2.4%, and 6.4%, respectively, among several salmonella serotypes, thus suggesting broad host and geographic distribution. Similar to other isolates of Enterobacteriaceae, some of the PMQR-positive isolates had a relatively small increase in MICs of nalidixic acid and ciprofloxacin and were associated with an ESBL or AmpC β-lactamase phenotype.
Less information is available about the epidemiology of the newly discovered PMQR pump QepA. A survey performed in Japan found qepA in 2 (0.3%) of 751 E. coli isolates (cutoff of MIC of norfloxacin, ≥0.025 μg/ml) collected from 140 hospitals between 2002 and 2006 (221). A second large survey was done by PCR in France. A single E. coli isolate among 121 (0.8%) ESBL-positive Enterobacteriaceae strains isolated in 2007 was positive for a variant named qepA2 (28). In a study of pig farms in China, qepA was found in 28 of 48 (58.3%) rmtB-positive E. coli isolates (115). A follow-up study from the same region in China tested for qepA among ceftiofur-resistant isolates of Enterobacteriaceae. qepA was found in 16 of 101 (15.8%) isolates, including, for the first time, K. pneumoniae and E. cloacae (116). Few recently published studies indicated a broad distribution of the gene. A survey of 461 isolates of Enterobacteriaceae in South Korea found qepA in one isolate from 2005 (107). qepA has also been found in the United Kingdom. Three of 83 (3.6%) human Enterobacteriaceae clinical isolates in London were qepA positive (7). Two additional studies screened isolates from Seoul, South Korea, for qepA. Four clonally unrelated strains of 621 (0.6%) E. coli bloodstream isolates were found to be positive in one study (41), and two E. aerogenes isolates of 223 (0.9%) E. cloacae, E. aerogenes, C. freundii, and Serratia marcescens isolates with reduced susceptibility to quinolones were qepA positive in the second survey (143). qepA was not found in a large survey of non-Typhi Salmonella enterica isolates collected in the United States from 1996 to 2006 (184).
In the decade since the discovery of qnrA1, there has been an explosion of knowledge about a phenomenon previously thought not to exist. We have challenged bacteria with a class of synthetic antimicrobial agents against which mutational resistance was unlikely to develop in a clinical setting. Despite this, resistance has emerged independently, countless times, worldwide. We now understand that, concomitant with the expansion in quinolone use, gram-negative bacteria assembled an arsenal of horizontally transmissible genetic elements that facilitated the emergence of mutational quinolone resistance. These elements are preexisting tools refitted for a novel purpose. Chromosomal genes not native to the Enterobacteriaceae have been recruited to these genera to effect topoisomerase protection and quinolone efflux, and a naturally occurring resistance gene has been outfitted with a new target. These elements may have played a leading role in the drama of emerging quinolone resistance, or theirs may have been only a supporting part; we have not yet fully taken stock of their importance. However, it is clear that PMQR has made extensive inroads among organisms of clinical importance, for humans and animals, everywhere that quinolones are used. Whatever these organisms have contributed to the global rise of resistance, their current prevalence is a problem. Cooccurrence with other resistance elements allows mutual resistance promotion and the spread of organisms that are difficult to treat. Additionally, clinical breakpoints have not yet been assessed in the context of PMQR, a form of reduced susceptibility that can contribute to therapeutic failure despite going undetected by traditional phenotypic methods. With the discovery of qnr in 1998, we took an important step in the battle against resistance, but it is clear that the bacteria have had a head start.
This work was supported by grant Morasha 1833/07 from the Israel Science Foundation to J.S., by grants AI43312 (to G.A.J.) and AI57576 (to D.C.H.) from the National Institutes of Health, by the U.S. Public Health Service, and by a NorthShore University HealthSystem research career development award to A.R.
Jacob Strahilevitz received his M.D. degree from the Tel-Aviv University School of Medicine. He did residency training in internal medicine at Beilinson Hospital as well as research training at the Weizmann Institute of Science, after which he served as associate director of a medicine department at Meir Hospital. He did his fellowship in infectious diseases at Hadassah Ein-Kerem and a research fellowship at the Massachusetts General Hospital before joining the faculty at Hadassah-Hebrew University in 2005. Dr. Strahilevitz's research focuses on the mechanisms and epidemiology of antibiotic resistance in bacteria, and he has a particular interest in resistance to quinolones.
George A. Jacoby trained at Harvard Medical School, the National Institutes of Health, the National Institute for Medical Research at Mill Hill, and the Massachusetts General Hospital, where he was a consultant in the Infectious Disease Unit for 25 years before moving in 1993 to head the Infectious Disease Department at the Lahey Clinic. He retired from clinical work in 2002 at age 70 and now has a small research lab at Lahey, where he works on bacterial resistance to antimicrobial agents, especially quinolones and β-lactams. He is a past Editor and Editor-in-Chief of Antimicrobial Agents and Chemotherapy.
David C. Hooper received his Bachelor of Arts degree with a major in microbiology from the University of Texas at Austin and his M.D. degree from Washington University School of Medicine, St. Louis, MO. He did residency training in internal medicine and a clinical and research fellowship in infectious diseases at Massachusetts General Hospital (MGH) as well as additional research training at the National Institutes of Health (NIH) before joining the faculty at the MGH and Harvard Medical School (HMS) in 1983. He is currently Chief of the Infection Control Unit and Associate Chief of Infectious Diseases at MGH and Professor of Medicine at HMS. Dr. Hooper's research interests have focused on the mechanisms and epidemiology of antibiotic resistance in bacteria, with a particular interest in resistance to quinolones due to alterations in drug target enzymes, alterations in regulation of endogenous multidrug resistance efflux pumps, and acquisition and spread of plasmid-carried resistance genes.
Ari Robicsek received his medical degree from the University of Toronto, where he also trained in internal medicine. He completed his infectious disease fellowship at Harvard Medical School. He is now Hospital Epidemiologist at NorthShore University HealthSystem and is a faculty member at the University of Chicago's Pritzker School of Medicine. His research interests include the epidemiology and control of antimicrobial resistance.