High-Level Ciprofloxacin Resistance from Point Mutations in gyrA and parC Confined to Global Hospital-Adapted Clonal Lineage CC17 of Enterococcus faecium

Abstract

To substantiate a common genetic background of ciprofloxacin-resistant Enterococcus faecium, 32 ciprofloxacin-resistant (Cip^r) and 31 ciprofloxacin-susceptible (Cip^s) isolates from outbreaks, clinical infections, surveillances, and animals from 10 different countries were genotyped by multilocus sequence typing. Additionally, susceptibilities to ampicillin and vancomycin and the presence of esp were determined and the quinolone resistance-determining regions of parC, gyrA, parB, and gyrE were sequenced. High-level Cip^r (MIC ≥ 64 μg/ml) due to point mutations in the quinolone resistance-determining region was unique to a distinct hospital-adapted genetic complex in E. faecium, previously designated CC17. Low-level Cip^r (MIC = 4 μg/ml) in non-CC17 strains is not attributable to point mutations in any subunit of the topoisomerase genes, and the mechanism of resistance remains unclear. Acquisition of mutations in parC and gyrA, leading to high-level Cip^r, is, in addition to ampicillin resistance and the presence of a putative pathogenicity island, another cumulative step in hospital adaptation of CC17.

Over the last two decades, enterococci have become increasingly important as nosocomial pathogens (3). These organisms are intrinsically resistant to a large number of antimicrobials and have the ability to easily acquire new resistance traits (40, 50). The emergence of nosocomial infections caused by ampicillin-, high-level aminoglycoside-, and glycopeptide-resistant Enterococcus faecium has caused clinical concern due to intra- and interhospital spread and limited therapeutic options (40, 50). Glycopeptide-resistant enterococci (vancomycin-resistant enterococci [VRE]) are nowadays endemic to the United States, with ∼30% of enterococcal infections caused by VRE (38, 43); in Europe, the epidemiology of VRE is now changing from a near absence of VRE in hospital-acquired infections at the turn of the century to a situation in which nosocomial epidemics and infections are increasingly reported (3, 13, 17). We recently described a hospital-adapted genetic subtype of E. faecium associated with epidemics and clinical infections, which has spread globally (20, 32, 62, 65). This subpopulation belongs to a distinct genetic lineage labeled clonal complex 17 (CC17) (62) and is associated with the presence of the variant esp gene as part of a pathogenicity island (31) and resistance to ampicillin. Recently, resistance to ciprofloxacin appeared to be associated with ampicillin resistance in genotypically related E. faecium isolates from Norway (25, 55, 56) and Spain (7).

To substantiate a common genetic background of quinolone-resistant E. faecium and association with the hospital-adapted CC17, we studied the genetic relatedness of 32 ciprofloxacin-resistant and 31 ciprofloxacin-susceptible E. faecium isolates from various human and animal origins by multilocus sequence typing (MLST) and determined susceptibility to vancomycin and ampicillin and the presence of esp. Finally, we sequenced the quinolone resistance-determining regions (QRDR) of parC and gyrA to identify mutations involved in ciprofloxacin resistance.

(This study was presented in part at the 44th Interscience Conference on Antimicrobial Agents and Chemotherapy, Washington, D.C., November 2004.)

MATERIALS AND METHODS

Bacterial isolates and growth conditions.

Sixty-three isolates of E. faecium (32 ciprofloxacin resistant and 31 ciprofloxacin susceptible) were collected from 10 different countries from nosocomial epidemics (n = 13; United Kingdom, n = 1; The Netherlands, n = 5; United States, n = 7), clinical infections (n = 27; Austria, n = 2; Germany, n = 4; Spain, n = 3; France, n = 2; United Kingdom, n = 2; Israel, n = 1; Italy, n = 3; The Netherlands, n = 6; Portugal, n = 3; United States, n = 7) (e.g., from blood, urine, or wounds), surveillance for colonization among hospitalized patients and in the community (France, n = 3; The Netherlands, n = 8) (all fecal samples), and animals (The Netherlands, n = 12) (2, 4, 10, 24, 32, 54, 65). Strains were considered epidemic as defined before (65). Epidemic isolates were recovered from clinical sites, blood, and urine, as well as from feces. Only one representative isolate from each outbreak was used for analysis. Bacterial strains were grown according to standard growth conditions on tryptic soy agar sheep blood plates and Todd Hewitt broth. DNA was extracted according to previous defined methods (64).

Identification and susceptibility testing.

Enterococcal species were confirmed by multiplex PCR, as described by Dutka-Malen et al. (11). Ciprofloxacin, vancomycin, and ampicillin susceptibilities were determined by standard agar dilution methods, according to the CLSI (formerly NCCLS) guidelines (42), and according to CLSI guidelines, MICs of ≥4 μg/ml of ciprofloxacin, ≥16 μg/ml of ampicillin, and ≥8 μg/ml of vancomycin were considered resistant. High-level ciprofloxacin resistance was defined in this study by a MIC of ≥64 μg/ml, low-level ciprofloxacin resistance was defined in this study by a MIC of 4 μg/ml.

esp PCR.

Seventy strains were screened for esp by PCR, with two different primer sets (esp 11 [5′-TTGCTAATGCTAGTCCACGACC-3′] and esp 12 [5′-GCG TCAA CAC TTG CAT TGC CGA A-3′] and the combination 14F [5′-AGA TTT CAT CTT TGA TTC TTG G-3′] and 12R [5′-AAT TGA TTC TTT AGC ATC TGG-3′]). PCR conditions included an initial denaturation at 95°C for 15 min for activation of the HotStarTaq DNA polymerase (QIAGEN GmbH, Hilden, Germany), followed by 30 cycles of 94°C for 30 s, 52°C for 30 s, and 72°C for 1 min, followed by an extension at 72°C for 7 min. Reactions were performed in 25-μl volumes by using the HotStarTaq master mix (QIAGEN GmbH). Strains negative by PCR were checked for the presence of the esp gene by Southern hybridization, as described previously (65). For this check, we generated an esp-specific probe (956 bp) using primers esp 11 and esp 12 (see previous explanation).

Sequencing the QRDRs of parC, gyrA, parE, and gyrB.

The QRDRs of the E. faecium parC and gyrA genes were amplified and sequenced (12). The primers used were previously designed by el Amin et al. (12) for gyrA (gyrA-1F [5′-CGG GAT GAA CGA ATT GGG TGT GA-3′] and gyrA-1R [5′-AAT TTT ACT CAT ACG TGC TTC GG-3′]) and Brisse et al. (5) for parC (parC-A [5′-TTC CCG TGC ATT TCG ATC AGT ACT TC-3′] and parC-C [5′-CGT ATG ACA AAG GAT TCG GTA AAT C-3′]). The QRDRs of parE and gyrB were determined by aligning the Escherichia coli ParE and GyrB QRDR sequence to the E. faecium DO genome in GenBank. With forward and reverse primers (parE-F [5′-GTC CGT AAA GCA ATC AAA G-3′] and parE-R [5′-CTT TAT ATA AAG GCG GTA ACG-3′], gyrB-F [5′-TGA AAT TCT TGC TGG AAA A C-3′] and gyrB-R [5′-CAA CAA TAG GAC GCA TGT AAC-3′]), the parE and gyrB genes of 44/63 and 63/63 isolates, respectively, were amplified. All PCR conditions were similar to those for esp (see above). PCR products were purified with a PCR purification kit (QIAGEN, Hilden, Germany) and sequenced with PCR forward and reverse primers, an ABI PRISM Big Dye cycle sequencing ready reaction kit (Perkin-Elmer, Applied Biosystems, Foster City, Calif.), and an ABI PRISM 3700 DNA sequencer (Perkin-Elmer). Alignments of amino acids were made with Megalign software (DNAstar). Sequences with no mutations were defined as being identical to the reference partial EMBL sequence (gyrA, accession no. AF060881; parC, accession no. AB017811) or the E. faecium DO sequence determined by E. coli QRDR sequence alignment (gyrB and parE).

MLST.

MLST and computer analysis of allelic profiles of E. faecium isolates to determine the genetic relatedness of isolates was performed as described previously (20) with the use of updated primer sequences (available at http://efaecium.mlst.net/misc/info.asp). Different sequences of a given locus were assigned allele numbers, and different allelic profiles were assigned sequence types (STs) by interrogating the E. faecium MLST database, which is available on the MLST website (http://www.mlst.net). Cluster analysis of allelic profiles was performed by using a categorical coefficient and a graphic method called a minimum spanning tree with Bionumerics software (version 4.0; Applied Maths, Sint-Martens-Latem, Belgium), as previously described by Schouls et al. (49). The criteria to first link types that have the highest number of single-locus variants, taken from the eBURST criteria (14), were chosen. The stringent criterion of a maximum one-allele difference was used to create complexes.

RESULTS

Genotyping and population modeling.

Genotyping of the 63 E. faecium isolates by MLST revealed 32 different STs, of which 10 STs (18 isolates) were found exclusively among Cip^r strains and 18 STs (22 isolates) were found only among Cip^s strains. Only 4 STs (23 isolates) were found among both Cip^r and Cip^s strains, strongly suggesting that the majority of Cip^r and Cip^s strains are genetically unrelated. A minimum spanning tree based on allelic profiles, using the most stringent definition of complexes of ≥6 identical alleles, resolved 2 dominant clonal complexes (CC), CC17 and CC22, 2 minor complexes, and 11 singletons (Fig. 1). Almost all isolates (28/30, 93%) belonging to the previously described hospital-adapted CC17 were Cip^r. In the other dominant complex, CC22 (n = 18), 7 isolates (39%) were Cip^r, whereas all isolates belonging to minor complexes and singletons, defined in this study as “other genotype” (n = 15), were Cip^s (Table 1). According to the source of isolation, all epidemic isolates (100%), 61% of clinical infectious isolates, 17% of animal isolates, 9% of clinical surveillance isolates, and none of the community isolates were Cip^r (data not shown). Ampicillin resistance (Amp^r) was almost exclusively confined (28/30 isolates, 93%) to CC17, and 20 CC17 isolates (67%) carried the variant esp gene. Vancomycin resistance (Van^r), though, was distributed among all major and minor complexes (Table 1; Fig. 2). Within CC17, Cip^r was associated with the presence of esp and Amp^r; 56% of Cip^r isolates were esp-positive Amp^r, compared to 3% of Cip^s isolates (chi-square test, P < 0.01) (Table 1). Amp^r was correlated to high-level Cip^r (chi-square test, P < 0.01) and with a concomitant presence of esp (chi-square test, P < 0.01) (Table 2).

FIG. 1.

Population structure of 63 Cip^r and Cip^s E. faecium isolates in CC. Clusters of STs are displayed in a minimum spanning tree. Each circle in the tree represents a different ST, and the ST is indicated by the number in the circle. One ST can contain multiple strains; circle size corresponds to the number of strains. Two CCs are identified (17 and 22) as well as 2 minor complexes and a couple of singletons. The CC and minor complexes are surrounded by gray shading. Heavy lines connecting two STs denote STs differing at a single locus, thin lines connect double-locus variants, and dotted lines connect STs differing at more than 2 loci. The colors indicate susceptibility to ciprofloxacin. If different resistance types occur simultaneously in one ST, pie charts are used to indicate distribution.

TABLE 1.

Origin, percentage of isolates resistant to shown antibiotics and positive for esp, and detailed Cip^r range among 63 E. faecium isolates of CC17, CC22 and other genotypes

CC (n)	Sourcea				% of isolates				Ciprofloxacin MIC (μg/ml)
	Out	Inf	Surv	Anim	Vanr	Ampr	esp+	Cipr	MIC50	MIC90	Range
17 (30)	13	15	2	0	77	93	67	83	128	256	1-256
22 (18)	0	6	6	6	83	0	0	39	2	4	0.5-4
Other (15)	0	6	3	6	93	13	0	0	1	2	0.25-2

^aOut, outbreak; Inf, infectious; Surv, surveillance; Anim, animal.

FIG. 2.

Cip^r in association with the presence and absence of esp and Amp^r among E. faecium isolates (n = 63) of CC17, CC22, and others (genotypes not belonging to CC17 or CC22). Percentages of Amp^r, presence of the esp gene, and Amp^r in combination with the esp gene are shown by bars for Cip^r and Cip^s isolates in CC17, Cip^r and Cip^s isolates in CC22, and Cip^s isolates in other genotypes (no Cip^r isolates found). For Cip^r and Cip^s in each CC, the number of isolates is given. R, resistant, S, susceptible.

TABLE 2.

MIC of ciprofloxacin in association with presence and absence of the esp gene and ampicillin resistance

^aVertical dotted line divides resistant and nonresistant E. faecium isolates.

Sequence analysis of the QRDRs of parC, gyrA, parE, and gyrB.

Ten different single and double amino acid changes were detected in ParC and GyrA of 19 of 32 Cip^r isolates of E. faecium (Table 3). No amino acid changes in ParC or GyrA were found in the remaining 13 Cip^r isolates or in the 31 Cip^s isolates. Nineteen isolates had mutations in parC, leading to an amino acid change at codon 80 (Ser to Ile [n = 12] or Ser to Arg [n = 6]) or codon 84 (Glu to Lys [n = 1]). In 15 of the 19 isolates with a mutation in parC, a secondary mutation was found in gyrA leading to an amino acid change at codon 83 (Ser to Arg [n = 7]) or codon 87 (Glu to Gly [n = 6] or Glu to Lys [n = 2]). In conclusion, among Cip^r isolates, 4 contained amino acid changes only in ParC and 15 had mutations in both parC codon 80 and gyrA. Subsequently, the QRDRs of the gyrB and parE genes were sequenced of 11 Cip^r isolates that did not contain mutations in gyrA and parC, 10 Cip^r isolates containing QRDR mutations, and 23 Cip^s isolates. No amino acid substitutions were found in the topoisomerases (data not shown).

TABLE 3.

parC and gyrA mutations in 63 CIP^r and CIP^s E. faecium isolates corresponding to ciprofloxacin MIC and CC^a

^aS, serine; E, glutamic acid; R, arginine; K, lysine; I, isoleucine; G, glutamine. Boldface type indicates wild-type amino acids; mutations are shown below. Dashed vertical line divides resistant and nonresistant E. faecium isolates.

Isolates with amino acid changes in ParC and GyrA had higher MICs (MIC₅₀ = 128 μg/ml; MIC₉₀ > 128) than isolates without mutations (MIC₅₀ = 2 μg/ml; MIC₉₀ = 8) (Student's t test, P < 0.01). No correlation was found between the number of substitutions in ParC and GyrA and the level of ciprofloxacin resistance (Table 3). Although some silent mutations were identified in ParE and GyrB, these were not associated with higher MICs (data not shown). Mutations in parC and gyrA associated with high-level Cip^r were only found in isolates confined to CC17.

DISCUSSION

High-level ciprofloxacin resistance (MIC ≥ 64 μg/ml) in E. faecium is associated with amino acid changes in topoisomerase IV and DNA gyrase and confined to a single clonal complex, previously identified as a global hospital-adapted subpopulation of E. faecium (CC17). CC17 is further characterized by ampicillin resistance and a pathogenicity island, including the variant esp gene (20, 31, 32, 65) (Fig. 1). Cip^r isolates belonging to other genetic backgrounds are only low-level resistant (MIC = 4 μg/ml) and do not carry these point mutations. These findings further disclose the cumulative adaptive mechanisms that have resulted in the evolution of a hospital-adapted genetic complex in E. faecium.

We have previously described the global spread of CC17 E. faecium, which is characterized by ampicillin resistance, although pandemic spread of VRE was instrumental for its identification. The strong association with quinolone resistance strongly suggests that this clonal complex has previously emerged as ampicillin-resistant but vancomycin-sensitive in Scandinavian countries. Torell et al. (55) described a clonal subset of Swedish vancomycin-susceptible Amp^r E. faecium strains with high-level ciprofloxacin resistance and mutations in the QRDR genotyped by biochemical fingerprinting (PhenePlate). In addition, Jureen et al. (25), showed a correlation between ampicillin and ciprofloxacin resistance among genetically highly similar Norwegian isolates that also carried the purK-1 allele, indicative for CC17. More specifically, Mohn et al. (38) identified fluoroquinolone prescription as a risk factor for fecal carriage with an outbreak of ampicillin-resistant E. faecium in a Norwegian hospital. However, Fortún et al. (15) found that Spanish patients with bacteremia from ampicillin-resistant E. faecium were only likely to have received quinolones, but there is no significant association. van der Steen et al. identified previously administered ciprofloxacin as a risk factor for colonization with an outbreak-associated, vancomycin-resistant, ampicillin-resistant E. faecium isolate in a Dutch hospital (57). A mouse model nevertheless showed that ciprofloxacin administration does not promote high-level vancomycin-resistant, ampicillin-resistant E. faecium colonization (9).

Linkage of ampicillin and quinolone resistance in CC17 illustrates that specific clones have different likelihoods of acquiring resistance traits, which depends on the availability of resistance genes in the local gene pool and intrinsic capacities of bacterial strains to acquire foreign DNA or to accumulate mutations. It is interesting, in this respect, that E. faecium isolates that harbored amino acid substitutions in the DNA mismatch repair proteins MutS and MutL belong to CC17 (61), which may suggest that CC17 isolates are more prone to undergo mutations in, e.g., parC or gyrA, conferring high-level resistance to quinolones. Clonal dissemination of quinolone resistance was also reported among methicillin-resistant Staphylococcus aureus, Neisseria gonorrhoeae, Salmonella enterica serotype Typhimurium strain DT 104, Streptococcus pneumoniae, and Yersinia enterocolitica (28, 46, 48).

In gram-negative bacteria, such as Escherichia coli, Neisseria gonorrhoeae, and Pseudomonas aeruginosa, gyrA is the primary target in mutation-mediated quinolone resistance. In these species, low-level ciprofloxacin resistance was reported in strains with mutations only in gyrA, while a higher level of resistance was discerned in strains carrying mutations in both gyrA and parC (1, 19, 30, 41). Conversely, in gram-positive bacteria, such as Staphylococcus aureus and Streptococcus pneumoniae (16, 52), parC appears to be the primary target, with mutations resulting in low-level resistance, which changes to high-level resistance with subsequent gyrA mutations. The latter phenomenon has also been proposed for Enterococcus faecalis and E. faecium from mutant analysis (5, 12, 27, 29, 53). Our finding that all ciprofloxacin-resistant isolates carried mutations in parC and only 15 isolates carried additional mutations in gyrA, confirms parC as the main target in mutation-mediated ciprofloxacin resistance in enterococci. In contrast, the only enzymatic study in E. faecalis showed that, for levofloxacin and ciprofloxacin, gyrase is the primary target, possibly explained by higher in vivo lethality of the quinolone-gyrase-DNA complex (44).

In various pathogens, including E. faecalis, a clear correlation was found between the number of mutations in gyrA and parC and the level of ciprofloxacin resistance (27). For E. faecium, Brisse et al. described higher MICs for isolates that acquired an amino acid change in gyrA in addition to an already existing mutation in parC, even though this was of only limited impact compared to S. aureus or S. pneumoniae and not consistently true for all double mutants (5). We could not confirm this correlation, since the mean MIC for isolates with only one mutation in parC was not significantly lower than that for isolates with two mutations in parC and gyrA. It is unclear whether additional mutations, without increased ciprofloxacin resistance, have other selective advantages. Selection pressure of mutants by quinolones other than ciprofloxacin could be attributable to the observed inconsistency, as already suggested for the accumulation of mutations in both subunits of both topoisomerases in levofloxacin-resistant S. pneumoniae (6).

Several other resistance mechanisms may be responsible for (low-level) ciprofloxacin resistance in CC22 and CC17 isolates that do not contain amino acid changes in the QRDRs of ParC and GyrA. In Staphylococcus aureus, Streptococcus pneumoniae, and E. coli, mutations in parE and/or gyrB have been associated with fluoroquinolone resistance. It is possible that mutations in these genes account for the low-level quinolone resistance as observed in this study, though Grohs et al. reported one high-level Cip^r E. faecalis isolate (with substitutions in ParC and GyrA) with an additional ParE substitution in codon 453 (Pro to Ser) which did not affect the MIC of ciprofloxacin (and other quinolones) significantly (18). No amino acid substitutions in parE and gyrB genes were found in the E. faecium isolates in this study. Silent mutations in parE in some highly Cip^r strains were identified, but this does not explain increased MICs and possibly reflects higher mutation frequencies in certain subtypes of E. faecium.

Resistance by active efflux of quinolones or increased expression of endogenous efflux pumps are alternative quinolone resistance mechanisms reported for S. pneumoniae (45), S. aureus (26), and E. faecalis (22, 33, 34). Davis et al. reported the presence of 34 potential multidrug resistance-encoding genes in E. faecalis by a bioinformatics approach (8). Three E. faecalis multidrug efflux pumps have been characterized, EmeA, EfrAB, and Lsa, of which only the first two confer resistance to quinolones (22, 33, 34, 51). Therefore, it seems highly likely that efflux pumps exist in E. faecium as well. So far, only active efflux of norfloxacin has been reported for E. faecium strain ATCC 19434 (35), but the efflux pump hasn't been identified yet. A novel quinolone resistance mechanism includes a Klebsiella pneumoniae multidrug resistance plasmid-mediated qnr gene that could be transferred horizontally (37) and has ever since been described in clinical isolates of K. pneumoniae in the United States (47, 58), E. coli, Citrobacter freundii, and Enterobacter sp. in Europe and China (23, 36, 59), and Providencia stuartii in Egypt (60). However, until now, the qnr locus has been described only in gram negative bacteria. Microarray hybridization failed to support the presence of a qnr probe from Klebsiella in 100 E. faecium isolates (different from the strains used in this study; data not shown). Different mechanisms leading to (low-level) ciprofloxacin resistance in E. faecium, therefore, remain to be identified.