ABSTRACT
Quinolone antibiotics constitute a clinically successful and widely used class of broad-spectrum antibiotics; however, the emergence and spread of resistance increasingly limits the use of fluoroquinolones in the treatment and management of microbial disease. In this study, we evaluated the quantitative contributions of quinolone target alteration and efflux pump expression to fluoroquinolone resistance in Pseudomonas aeruginosa. We generated isogenic mutations in hot spots of the quinolone resistance-determining regions (QRDRs) of gyrA, gyrB, and parC and inactivated the efflux regulator genes so as to overexpress the corresponding multidrug resistance (MDR) efflux pumps. We then introduced the respective mutations into the reference strain PA14 singly and in various combinations. Whereas the combined inactivation of two efflux regulator-encoding genes did not lead to resistance levels higher than those obtained by inactivation of only one efflux regulator-encoding gene, the combination of mutations leading to increased efflux and target alteration clearly exhibited an additive effect. This combination of target alteration and overexpression of efflux pumps was commonly observed in clinical P. aeruginosa isolates; however, these two mechanisms were frequently found not to be sufficient to explain the level of fluoroquinolone resistance. Our results suggest that there are additional mechanisms, independent of the expression of the MexAB-OprM, MexCD-OprJ, MexEF-OprN, and/or MexXY-OprM efflux pump, that increase ciprofloxacin resistance in isolates with mutations in the QRDRs.
INTRODUCTION
Fluoroquinolones are very potent antimicrobial agents with excellent oral bioavailability, reaching concentrations in serum equivalent to those for intravenous administration. They are broad-spectrum antibiotics with antibacterial activity against Gram-positive as well as Gram-negative bacteria (1, 2). As a consequence, fluoroquinolones are widely and increasingly used for the treatment of bacterial infections not only in the hospital setting but also for outpatients. The broad, frequent, and worldwide use of the fluoroquinolones, as well as the frequently inappropriate application of these antibiotics, is an important factor driving resistance, which has reached clinically relevant levels in the last decade (3, 4, 5).
The fluoroquinolones act by directly inhibiting DNA replication via an interaction of the drug with complexes composed of DNA and either of the two target enzymes, DNA gyrase and topoisomerase IV (1, 6). The molecular mechanisms of fluoroquinolone resistance include two dominant mechanistic categories for all bacterial species studied so far (7, 8). The activity of multidrug resistance (MDR) efflux pumps decreases intracellular fluoroquinolone concentrations (9), and alterations of the drug target by mutations at key sites in the so called quinolone resistance-determining regions (QRDRs) of the genes encoding DNA gyrase (gyrA and gyrB) and/or topoisomerase IV (parC and parE) lead to decreased binding affinity of the quinolones for their respective drug targets (10, 11). More recently, mobile genetic elements carrying the qnr (12), qepA (13), or aac(6′)-Ib-cr (14) gene, which confer reduced susceptibility to quinolones on members of the Enterobacteriaceae family, have also been described.
In this study, we analyzed a panel of 100 clinical Pseudomonas aeruginosa isolates with respect to the presence of mutations in the QRDRs and the expression of four major efflux pumps in a subset of these isolates. We furthermore introduced the most frequent mutations in the QRDRs into the susceptible P. aeruginosa reference strain PA14 and inactivated the efflux regulator-encoding genes nfxB, mexR, and mexZ, as well as the oxidoreductase-encoding gene mexS, with the aim of generating mutants that overexpress the MexCD-OprJ, MexAB-OprM, MexXY, and MexEF-OprN efflux pumps, respectively. The results of this study suggest that in addition to mutations in the QRDRs and overexpression of the MexAB-OprM, MexCD-OprJ, MexEF-OprN, and/or MexXY-OprM efflux pump, there are other, yet unknown mechanisms contributing to increased ciprofloxacin MIC levels for clinical P. aeruginosa isolates.
MATERIALS AND METHODS
Bacterial isolates and antibiotic resistance profile.A collection of 100 clinical P. aeruginosa isolates collected at the Hannover Medical School (MHH) between 2005 and 2007 was used in this study. These isolates were obtained from 90 individuals, 24 of whom were cystic fibrosis (CF) patients, with clinical infections at various sites (see Table S1 in the supplemental material). Two isolates per patient were analyzed in this study when the isolates clearly differed in their antibiotic resistance profiles; otherwise, one isolate per patient was analyzed. Antibiotic resistance profiles (see Table S1) were determined using a Vitek 2 system (bioMérieux).
Pyrosequencing.To extract DNA for pyrosequencing, 500 μl of an overnight culture was harvested and was lysed for 15 min at 95°C in 100 μl lysis buffer (0.25% [mass/vol] sodium dodecyl sulfate [SDS], 50 mM NaOH). After the addition of 900 μl dH2O, 2 μl was used as a PCR template.
To identify mutations at amino acid positions 83 and 87 in gyrA and position 87 in parC, a pyrosequencing assay was established. Amplification was performed as described by Doostzadeh et al. (15) using a 24-mer universal biotinylated primer (UBP) adopted from the work of Royo et al. (16). All primers used in this study are shown in Table S2 in the supplemental material. Sequencing primers were designed with Primer3 (17) to anneal 3 bp upstream (gyrA) or 5 bp upstream (parC) from the single nucleotide polymorphism (SNP). Pyrosequencing was performed on a PSQ 96MA pyrosequencer (Pyrosequencing AB) with PyroMark Gold chemistry (Qiagen) as described by Royo et al. (16).
Sanger sequencing.To identify mutations in gyrB and parE, the QRDRs of both genes were amplified with primers gyrBfp4, gyrBrp4, parEfp3, and parErp3 (see Table S2 in the supplemental material). PCR products were sequenced by the Sanger method using the same sets of primers on a 3730xl DNA analyzer (Applied Biosystems).
qRT-PCR.The expression levels of mexA, mexC, mexE, mexX, and the housekeeping gene rpoD were determined by quantitative real-time reverse transcription-PCR (qRT-PCR). RNA was isolated at the late-logarithmic growth phase (optical density at 600 nm [OD600], 1.5 to 2.0) from 3 ml Mueller-Hinton liquid culture by using the RNeasy kit (Qiagen), as described by the manufacturer. RNA was eluted from the RNeasy columns in a volume of 50 μl water and was treated with a DNA-free kit (Ambion). cDNA was synthesized by using random hexamer primers (Invitrogen) and SuperScript II reverse transcriptase (Invitrogen) according to the manufacturer's instructions. qRT-PCRs were performed in duplicate in a 20-μl volume with 25 ng cDNA and a primer concentration of 500 nmol/liter on a LightCycler 480 system (Roche) using the SYBR green I master mix (Roche). The primers were adopted from the work of Tomás et al. (18); sequences are listed in Table S2 in the supplemental material. Gene expression was calculated using the ΔΔCT method and a standard curve to measure PCR efficiency. All results were normalized to the expression of the housekeeping gene rpoD of the same clinical isolate and were calibrated relative to expression in P. aeruginosa PA14. According to Cabot et al. (19), isolates with ≥3-fold mexA overexpression were regarded as positive, whereas values between 2- and 3-fold were regarded as borderline expression. For mexC, mexE, and mexX, ≥10-fold overexpression was regarded as positive and 5- to 10-fold overexpression was regarded as borderline expression.
Genetic manipulations.To generate mutations of gyrA, gyrB, and parC, as well as knockouts of the nfxB, mexR, mexS, and mexZ genes in an isogenic background, the P. aeruginosa strain PA14 was used. Mutagenesis was carried out by homologous recombination using the pEX18Ap plasmid (20) and approximately 1,000-bp mutagenic fragments created by overlap extension PCR as described previously (21). To generate knockout mutants, the 500-bp regions upstream and downstream of the gene of interest were amplified using overlapping primers. All primers are listed in Table S2 in the supplemental material. Mutant candidates were identified by replica plating on LB agar plates with appropriate antibiotic concentrations and were further analyzed by PCR and Sanger sequencing to verify the mutation. The antibiotic resistance profiles of all mutants were determined in Mueller-Hinton broth as described previously (22).
Cloning of gyrA and complementation of clinical isolates.To complement gyrA mutations in clinical isolates with the wild-type gene, gyrA was amplified from the PA14 chromosome with primers gyrAFPSacI and gyrARPSacI and was cloned into the SacI restriction site of plasmid pME6032 (23), yielding plasmid pME::gyrA. The correct insertion and sequence were verified by Sanger sequencing using primers gyrAseqF, gyrAseqR, pMEseqF, and pMEseqR. Complementation was performed as follows. Each isolate was grown overnight at 37°C and 180 rpm in 4 ml LB broth. Portions (1.5 ml) of these cultures were centrifuged, washed three times with 1 ml 0.3 M sucrose, and resuspended in 100 μl 0.3 M sucrose. Five hundred nanograms of plasmid pME::gyrA and 50 μl of the cell suspension were used for electroporation. The cells were plated on LB agar plates supplemented with 100 μg/ml tetracycline. The ciprofloxacin MICs for isolates containing pME::gyrA with and without the addition of 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) were determined using Etest strips (bioMérieux) on LB agar plates supplemented with 100 μg/ml tetracycline. Results were recorded after 24 h of growth at 37°C.
RESULTS AND DISCUSSION
Frequency and nature of mutations in the QRDRs of gyrA, gyrB, parC, and parE in clinical P. aeruginosa isolates.In this study, we used Sanger sequencing and pyrosequencing to determine the nature and frequency of hot spot SNP mutations in the quinolone resistance-determining regions (QRDRs) of the gyrA and gyrB genes, encoding DNA gyrase, as well as in those of parC and parE, encoding topoisomerase IV. A panel of 100 clinical P. aeruginosa isolates isolated from patients of the Hannover Medical School over a period of 2 years (2005 to 2007) was analyzed in this study. The isolates were recovered from various clinical sites, and most of them exhibited resistance not only to fluoroquinolones but also to various other antimicrobial compounds (see Table S1 in the supplemental material). Pyrosequencing technology has been proven to be time and cost competitive and to allow efficient detection of SNPs in localized regions where the nucleotide variants are known (24). We have designed two different pyrosequencing assays for sequencing analysis of the most prominent mutation hot spots in the QRDR of the A subunit of the DNA gyrase, encoded by gyrA, which spans amino acid positions 83 to 87, and in the QRDR of the A subunit of topoisomerase IV, encoded by parC, which spans amino acid positions 82 to 84.
The QRDRs of the B subunits of DNA gyrase (encoded by gyrB) and topoisomerase IV (encoded by parE) are larger, spanning amino acid positions 429 to 585 in the GyrB protein and 357 to 503 in the ParE protein. Therefore, Sanger sequencing was performed for the identification of relevant mutations in the QRDRs of gyrB and parE.
Sequencing confirmed the presence of mutations in the QRDRs in most of the clinical isolates. The relative frequencies of the specific mutations are shown in Fig. 1A. In accordance with the findings of several previous studies (25, 26, 27, 28, 29, 30, 31), the most frequently observed mutation, T83I, was encoded in the QRDR of gyrA, whereas mutations in gyrB were less frequent (30, 32, 33). Here the majority of mutations were found at amino acid positions 466 to 468; however, we also found two isolates with an I529V mutation, which, to our knowledge, has not been described previously. Two mutations within the QRDR of parC were detected in our panel of 100 clinical isolates (S87W and S87L), and only three mutations overall were present in parE (one M437I and two A473V mutations). The majority of clinical isolates harbored either a single mutation in gyrA or gyrB or a combination of mutations in gyrA and parC (Fig. 1B). Fewer isolates exhibited mutations in gyrB in combination with parE or in gyrA in combination with gyrB, with or without additional mutations in the QRDR of parC. As in previous studies (28, 33), no single parC mutations were found in our panel of clinical P. aeruginosa isolates. Two of the isolates harbored a single mutation in parE, and for 14 isolates, no mutations in the QRDRs were detected.
Mutations identified in the gyrA, gyrB, parC, and parE genes of 100 clinical isolates. (A) Frequency and nature of mutations found in the QRDRs of gyrA, gyrB, parC, and parE in 100 clinical Pseudomonas aeruginosa isolates. Each mutation (given as the wild-type allele, amino acid position, and mutant allele) is followed by a semicolon and the number of isolates harboring the mutation. Δ, deletion at the specified position; wt, wild-type allele or silent mutation. (B) Co-occurrence of mutations in the QRDRs of gyrA, gyrB, parC, and parE in individual clinical P. aeruginosa isolates. The proportions of isolates with a single (light shaded sectors), double (dark shaded sectors), or triple (filled sectors) mutation, or with no identified QRDR mutation (open sectors), are shown.
It has been observed before that highly resistant P. aeruginosa isolates harboring a double gyrA and parC mutation are isolated almost exclusively from non-CF patients, whereas in isolates from CF patients, single mutations within the QRDRs dominate (34, 35, 36). Interestingly, in accordance with the previous reports, 27 of the 29 gyrA parC mutants in this study were isolated from non-CF patients. It has been suggested that higher ciprofloxacin levels in non-CF patients might account for this phenomenon, since levels of the drug in the sputa of CF patients were found to be significantly lower than those in blood (37). Although the lower drug concentrations might select for strains with intermediate resistance in distinct niches, it might also indicate coselection of single mutations in QRDRs with other phenotypic traits that provide the strains with a selective advantage. Thereby, the unique environment of the lung in CF patients might play a significant role in the process of mutation and selection (35).
Correlation of the presence of SNPs in the QRDRs of gyrA, gyrB, parC, and parE with the ciprofloxacin resistance phenotype in clinical P. aeruginosa isolates.The presence of SNPs within the QRDRs of gyrA, gyrB, parC, and parE was correlated with phenotypic resistance to fluoroquinolones in the clinical P. aeruginosa isolates. Figure 2 shows the relationship of the ciprofloxacin MIC values for all 100 clinical P. aeruginosa isolates to the presence of mutations in the QRDRs. The majority of clinical isolates harbored single mutations in parE, gyrA, or gyrB; however, those mutations did not necessarily lead to ciprofloxacin MIC values exceeding 2 μg/ml (noteworthy, resistance according to the Clinical and Laboratory Standards Institute [CLSI] breakpoints is categorized by MIC values exceeding 2 μg/ml). In contrast, combinations of mutations in the QRDRs of gyrA and parC always resulted in a ciprofloxacin-resistant phenotype, with MIC values of ≥8 μg/ml. The 2 isolates that harbored single parE mutations and the 14 isolates without mutations in the QRDRs exhibited MIC values that did not exceed 2 μg/ml and thus were categorized as susceptible or intermediate according to CLSI breakpoints.
Correlation of ciprofloxacin MIC values with the presence of mutations in the QRDRs of gyrA, gyrB, parC, and parE (and combinations thereof) for 100 P. aeruginosa clinical isolates. The number of isolates with the same combination of MIC and genotype is given inside each circle. Light, medium, and dark shaded circles represent sensitive, intermediate, and resistant isolates, respectively, according to the CLSI clinical breakpoints.
Introduction of dominant SNPs in the QRDRs of gyrA, gyrB, and parC into the susceptible P. aeruginosa reference strain PA14.In order to pinpoint the contributions of the most frequent mutations in the QRDRs of gyrA, gyrB, and parC to fluoroquinolone resistance, we introduced the respective SNPs into the fluoroquinolone-susceptible reference strain PA14 and measured the resistance profile. Plasmid constructs for allelic exchange were generated for two SNPs in gyrA (resulting in T83I and D87N), three SNPs in gyrB (S466F, S466Y, and E468D), and two SNPs in parC (S87L and S87W). Those SNPs were introduced into the reference strain singly and in various combinations. As shown in Table 1, the introduction of parC mutations alone had no impact on susceptibility to ciprofloxacin, whereas mutations in the QRDR of gyrB or gyrA increased the MIC of ciprofloxacin 8- to 16-fold. Similarly, the introduction of a single parC mutation did not alter the susceptibility of Escherichia coli to fluoroquinolone (38). The simultaneous introduction of two SNPs into the QRDR of gyrA (T83I and D87N) did not increase ciprofloxacin resistance over that with T83I alone. However, the simultaneous introduction of SNPs in gyrA (T83I) and parC (either S87L or S87W) increased the ciprofloxacin MIC 256-fold over that for the reference strain. None of the mutations in the QRDRs had any impact on the resistance of the parental strain to beta-lactam antibiotics, carbapenems, or aminoglycosides (data not shown).
MICs for in vitro-generated PA14 mutants
Most clinical P. aeruginosa isolates harboring mutations in the QRDR additionally express efflux pumps.Mutations in genes encoding the two subunits of DNA gyrase raised the ciprofloxacin MIC 8- to 16-fold over that for the P. aeruginosa reference strain. Those gyrA mutants, as well as the majority of the clinical isolates harboring relevant mutations in the QRDRs of gyrA and/or gyrB, exhibited MIC values of ≤2 μg/ml. However, we identified clinical isolates with a single mutation in gyrA for which MIC values reached 8 μg/ml. The broad MIC range for clinical P. aeruginosa gyrA mutants has been observed in several studies previously (34, 39), and although it is tempting to speculate that this can be explained by differential expression of efflux pumps (32, 40), no clear association between increased MICs for the gyrA mutants and increased expression of efflux pumps could be demonstrated in previous studies (39, 41, 42). Along the same lines, it has been demonstrated for some individual clinical P. aeruginosa isolates that elevated meropenem MIC levels could not be explained by decreased levels of OprD and/or overexpression of the MexAB-OprM and MexEF-OprN efflux pumps (43, 44), and it was thought that other resistance mechanisms yet to be identified might account for the resistance phenotype. Therefore, to test whether differential expression of efflux pumps in those isolates could account for the high MIC values, we monitored the expression of four efflux pumps—MexAB-OprM, MexCD-OprJ, MexEF-OprN and MexXY-OprM—in 29 selected clinical isolates. Nine of these isolates harbored no mutation in any of the QRDRs; 10 isolates harbored a single gyrA mutation; and 10 isolates had mutations in gyrA and parC. The transcription of the genes encoding the membrane fusion proteins of the pumps (mexA, mexC, mexE, and mexX) was quantified using qRT-PCR, and the results are shown in Table 2. According to work done by Cabot et al. (19), ≥3-fold overexpression of mexA was regarded as positive, whereas values between 2- and 3-fold were regarded as borderline expression. For mexC, mexE, and mexX, ≥10-fold overexpression was regarded as positive, and 5- to 10-fold overexpression was regarded as borderline expression. The majority of clinical isolates exhibited increased expression of at least one efflux pump; MexXY-OprM expression was increased the most. This result has also been observed previously (36). However, no clear association between the expression of efflux pumps and increased fluoroquinolone MIC values for isolates harboring mutations in the QRDRs could be observed (Fig. 3).
Expression of mexA, mexC, mexE, and mexX in 29 clinical isolates and 4 in vitro-generated PA14 knockout mutants compared to that in the wild-type strain PA14
Influence of multidrug efflux pump overexpression and QRDR mutation on the ciprofloxacin MIC. The differences in MIC values for a particular genotype cannot be explained by the additional expression of multidrug efflux pumps. The expression of four multidrug efflux pumps in 29 selected clinical P. aeruginosa isolates is shown. Each circle represents one clinical isolate, while each quarter represents one efflux pump (top left, MexAB-OprM; top right, MexCD-OprJ; bottom left, MexEF-OprN; bottom right, MexXY-OprM). Filled quarters, overexpression of a pump; shaded quarters, borderline expression; open quarters, wild-type expression levels (according to Cabot et al. [19]). The isolates are arranged according to their QRDR genotypes (wild type, single mutation in gyrA, or simultaneous mutations in gyrA and parC) and their ciprofloxacin MICs.
Inactivation of the efflux regulator-encoding genes mexR, nfxB, mexS, and mexZ in the susceptible P. aeruginosa reference strain.In order to pinpoint the contributions of overexpression of the MexAB-OprM, MexCD-OprJ, MexEF-OprN, and MexXY-OprM efflux pumps to fluoroquinolone resistance, we inactivated the respective efflux regulator-encoding genes in the fluoroquinolone-susceptible reference strain PA14 and measured the resistance profile. Deletion of the efflux regulator-encoding gene mexR, nfxB, or mexZ or of the oxidoreductase-encoding gene mexS led to overexpression of the efflux pump MexAB-OprM by 1.6-fold, MexCD-OprJ by 16-fold, MexXY-OprM by 6-fold, or MexEF-OprN by 320-fold, respectively (Table 2). We also inactivated the efflux regulator-encoding genes in the PA14 strain background in various combinations (ΔnfxB-mexZ, ΔmexR-mexS, ΔmexR-mexZ, and ΔmexS-mexZ) and also combined the deletion of the four efflux regulator-encoding genes with mutations in the QRDRs (gyrA, gyrB, and gyrA parC). As shown in Table 1, overexpression of the efflux pumps clearly increased the ciprofloxacin MIC values for the susceptible P. aeruginosa reference strain 2- to 16-fold. Inactivation of nfxB or mexS had the most pronounced phenotype. The combined inactivation of various efflux regulator-encoding genes (ΔnfxB-mexZ, ΔmexR-mexS, ΔmexR-mexZ, and ΔmexS-mexZ) did not lead to further increases in MIC levels. This absence of an additive effect might be explained by antagonistic interactions of efflux pumps during planktonic growth, which have been found to occur in nfxB mutants (45). However, the inactivation of efflux regulator-encoding genes in the PA14 gyrA, gyrB, and gyrA parC mutant backgrounds clearly enhanced the fluoroquinolone resistance level further in an additive manner. In agreement with our results, the deletion of efflux pumps in resistant P. aeruginosa strains with multiple target alterations has been demonstrated previously to lead to a reduced fluoroquinolone MIC (46). It might thus be surprising that we did not find a clear correlation between increased fluoroquinolone MIC values for clinical isolates harboring a particular QRDR genotype and the expression of major efflux pumps.
Mutation in the QRDR of gyrA adds to preexisting isolate-specific resistance levels.Although overexpression of efflux pumps further enhanced fluoroquinolone resistance in a QRDR mutant background, we did not find a clear association between the expression of efflux pumps and increased fluoroquinolone MIC values in our set of clinical isolates harboring mutations in the QRDRs. We therefore wondered whether the contribution of a gyrA mutation to the fluoroquinolone resistance level could differ for different isolates. To address this question, we cloned the wild-type gyrA gene into the pME6032 vector, resulting in vector pME::gyrA, and introduced the gene into various clinical isolates in trans. All of those clinical isolates exhibited gyrA mutations, but the MIC values ranged from 0.125 to 2 μg/ml (Table 3). We found that complementation with the wild-type gyrA gene led to a 2- to 8-fold reduction in fluoroquinolone resistance irrespective of the original resistance level. These results indicate that mutations within the gyrA QRDR add to preexisting isolate-specific resistance levels of unknown origin. Two comprehensive screenings of a P. aeruginosa PA14 mutant library have shown that approximately 100 to 200 genes are involved in the ciprofloxacin resistome (47, 48). It thus will be an interesting task for the future to determine which of the gene inactivations identified, if any, play a role in fluoroquinolone resistance in clinical settings.
Complementation of clinical isolates with plasmid pME::gyrA
Of the 100 clinical P. aeruginosa isolates studied here, 53 harbored a single mutation within only one of the QRDRs and 40 showed MIC values of 2 μg/ml or less. Although these P. aeruginosa isolates are categorized as susceptible, it remains to be shown that it is safe to treat them with fluoroquinolones (49). The stepwise enrichment of fluoroquinolone resistance mutations has been described previously in several in vitro studies (50, 51, 52). Mutants at each step are enriched when drug concentrations fall within a specific range called the mutant selection window (1, 53). Even antibiotic concentrations below the MIC might select for resistance-conferring mutations; in a recent study, it could be demonstrated that ciprofloxacin concentrations at 1/10 the MIC were sufficient to select fluoroquinolone-resistant mutants de novo in E. coli (54).
Although low-level resistance conferred by first-step SNPs within the QRDR does not prevent bacterial eradication in the presence of sufficient levels of a quinolone, these SNPs may substantially enhance the number of (secondary) resistant mutants that can be selected from this population. In line with this, it has been demonstrated that deletion of efflux pumps significantly reduces the frequency of emergence of fluoroquinolone-resistant mutant isolates (46, 52). A key to preventing fluoroquinolone resistance in P. aeruginosa may therefore be to strictly avoid the use of low doses of fluoroquinolones and thus to preclude the emergence of first-step mutations that confer resistance to fluoroquinolones.
ACKNOWLEDGMENTS
Help from Robert Geffers (Helmholtz Centre for Infection Research) with qRT-PCR and pyrosequencing is gratefully acknowledged.
This work was supported by an ERC starter grant (RESISTOME 260276) and by the President's Initiative and Networking Funds of the Helmholtz Association of German Research Centers (HGF) under contract VH-GS-202.
FOOTNOTES
- Received 6 August 2012.
- Returned for modification 2 September 2012.
- Accepted 25 December 2012.
- Accepted manuscript posted online 28 December 2012.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/AAC.01581-12.
- Copyright © 2013, American Society for Microbiology. All Rights Reserved.