Plasmid-mediated quinolone resistance determinants in quinolone-resistant Escherichia coli isolated from patients with bacteremia in a university hospital in Taiwan, 2001–2015

The aim of this study was to characterize fluoroquinolone (FQ)-resistant Escherichia coli isolates from bacteremia in Taiwan in 2001–2015. During the study period, 248 (21.2%) of 1171 isolates were identified as levofloxacin-resistant. The results of phylogenetic group analysis showed that 38.7% of the FQ-resistant isolates belonged to phylogenetic group B2, 23.4% to group B1, 22.6% to groupA, 14.9% to group D, and 0.4% belonged to group F. FQ-resistant isolates were highly susceptible to cefepime (91.5%), imipenem (96.0%), meropenem (98.8%), amikacin (98.0%), and fosfomycin (99.6%), as determined by the agar dilution method. β-lactamases, including blaTEM (66.1%), blaCMY-2 (16.5%), blaCTX-M (5.2%), blaDHA-1 (1.6%), and blaSHV-12 (1.6%), were found in FQ-resistant isolates. The results of PCR and direct sequencing showed that 37 isolates (14.9%) harbored plasmid-mediated quinolone resistance (PMQR) genes. qnrB2, qnrB4, qnrS1, coexistence of qnrB4 and qnrS1, oqxAB, and aac(6′)-Ib-cr were found in 1, 4, 4, 1, 15, and 14 isolates, respectively. PMQR genes were successfully transfered for 11 (29.7%) of the 37 PMQR-harboring isolates by conjugation to E. coli C600. These findings indicate that qnr genes remained rare in E. coli but demonstrate the potential spread of oqxAB and aac(6′)-Ib-c in Taiwan.


Results
Long-term surveillance and antimicrobial susceptibility of FQ-resistant E. coli. During the study period, 2001-2015, we randomly selected 1,171 E. coli isolates from patients with bacteremia, of which 248 (21.2%) were identified as levofloxacin-resistant by using the disk diffusion method ( Table 1). The trend in the prevalence of FQ-resistant invasive isolates remained stable during the 15-year surveillance (19.2-24.3%) ( Table 1). The phylogenetic analysis revealed five groups (A, B1, B2, D, and F) in 248 FQ-resistant isolates. Nintysix (38.7%) of the FQ-resistant isolates belonged to phylogenetic group B2. Phylogenetic group B1 was the second most common, representing in 23.4% of the isolates, followed by group A (22.6%), group D (14.9%), and group F (0.4%) ( Table 1). The dramatically increasing ratio of phylogenetic group B2 among FQ-resistant isolates was revealed during the study period (Table 1).    The susceptibilities of the 248 FQ-resistant isolates to 15 antimicrobial agents are shown in Table 2. All isolates were resistant to levofloxacin and ciprofloxacin, as determined by the agar dilution method. However, the entire collection was highly susceptible to cefepime (91.5%), imipenem (96.0%), meropenem (98.8%), amikacin (98.0%), and fosfomycin (99.6%) ( Table 2). One isolate showed resistance to tigecycline, and all isolates were susceptible to colistin. Moreover, a total of 89 (35.9%) and 223 (89.9%) isolates were defined to be ESBL-producers and multidrug resistant (MDR) strains, respectively. The trends of resistance of FQ-resistant invasive isolates to 11 selected antimicrobial agents were generally stable during this 15-year surveillance (Fig. 1). The prevalence of antimicrobial resistance to tetracycline decreased from 86.7% to 55.6% during this period (Fig. 1).

Discussion
In this study, we present the characteristics of 248 FQ-resistant bacteremia isolates of E. coli from Taiwan, 2001-2015. Among them, 37 isolates harbored at least one PMQR gene. oqxAB and aac(6′ )-Ib-cr genes were most prevalent among PMQR-producers. In addition, horizontal transmission of PMQR genes is often accompanied by transmission of genes conferring resistance to other antimicrobial agents. Antimicrobial resistance in Gram-negative bacteria is on the rise worldwide, particularly in E. coli, which constitutes a majority of invasive Gram-negative isolates. Wong et al. showed that ciprofloxacin resistance in E. coli isolated from bacteremia in Canada peaked in 2006 at 40% and subsequently stabilized at 29% in 2011, corresponding to decreasing ciprofloxacin usage after 2007 10 . In this study, we showed the prevalence of FQ-resistant invasive E. coli isolates is lower compared with Canada (Table 1). In addition, the prevalence of FQ resistance in bacteremia-causing E. coli was lower than urinary-tract-related E. coli in Taiwan (21.2% vs. 32%) 11 . Moreover, the entire collection was highly susceptible to cefepime, imipenem, meropenem, amikacin, and fosfomycin ( Table 2). Fosfomycin is found active against Enterobacteriaceae, particularly E. coli, regardless of source (urinary tract infections or bacteremia), ciprofloxacin resistance, and ESBL production [12][13][14] . In addition, fosfomycin is recommended as one of the first-line agents for treatment of urinary tract infections (UTIs) in the latest guidelines endorsed by the Infectious Diseases Society of America and the European Society for Clinical Microbiology and Infectious Diseases 15 . As a result, the clinical usefulness of fosfomycin, as a first-line treatment agents of bacteremia E. coli infections, should be evaluated further, especially in regions where ciprofloxacin resistance rates are high.
The phylogenetic group B2 was the most common pathogenic E. coli in many countries, and group A and group B1 were usually isolated as commensals 16,17 . Massot et al. showed a parallel and linked increase in the frequency of the B2 group strains (from 9.4% in 1980 to 22.7% in 2000 and 34.0% in 2010) and of virulence factors 18 . Here, we showed 38.7% of the FQ-resistant bacteremia E. coli isolates belonged to phylogenetic group B2, followed by group B1 (23.4%), group A (22.6%), group D (14.9%), and group F (0.4%) ( Table 1). Moreover, based on the 15-year epidemiologic analysis, we further showed that the increasing trend of group B2 among bacteremia E. coli isolates (Table 1). Phylogenetic group B2 dominates the bacteremia E. coli isolates during the period 2007-2015, but group B1 was most prevalent among bacteremia E. coli isolates during the period 2001-2006 ( Table 1). As a result, the longitudinal collection of clinical isolates provides the opportunity to characterize the dynamics of the epidemiologic trend and evolution in infectious pathogens over long periods.
Zhao et al. showed that qnr, aac(6′ )-Ib-cr, qepA, and oqxAB were found in 2.7%, 24.5%, 11.9% and 6.3% of ciprofloxacin-resistant E. coli isolates in China, respectively 19 . Yang et al. showed that PMQR genes were detected in 59 of 80 (73.8%) ciprofloxacin-nonsusceptible bacteremia E coli isolates from Korea 20 . In this study, we revealed the prevalence of PMQR genes among FQ-resistant E. coli in Taiwan (14.9%) was relatively lower than in China (37.3%) 19 or in Korea (73.8%) 20 . In addition, the dominant PMQR genes among FQ-resistant E. coli in Taiwan is oqxAB (40.5%), followed by aac(6′ )-Ib-cr (37.8%), and qnr alleles (27.0%). No qepA-producer was found in this study. Although PMQR genes provide a low level of FQ resistance, they have been reported to favor the selection of additional chromosome-encoded resistance mechanisms 21 . Moreover, all of the PMQR-positive isolates had QRDR mutations (Table 3). These results suggest that along with high-level resistance mediated by QRDR mutations, selection pressure from FQs was absent, and in this case PMQR genes may be lost 21 . It is possible that evolution by natural selection may explain the higher level of FQ resistance and the relatively lower prevalence of PMQR genes in FQ-resistant invasive E. coli from Taiwan. As a result, continual epidemiologic surveillance of PMQR genes is necessary to evaluate whether there are specific plasmids disseminated in Taiwan.
Previous studies showed the most common point mutations in ciprofloxacin-resistant E. coli isolates from China were GyrA S83L/D87N (263 isolates, 87.1%) and S83L (21 isolates, 7.0%), and those in ParC were S80I (233 isolates, 77.2%) and S80I-E84V (35 isolates, 11.6%) 19 . Our results regarding the distribution of QRDR mutations among FQ-resistant isolates were consisted with previous studies (Table 3). Isolate 1019 showed low-level FQ resistance presented S129A/S134G/A141V/L151M substitutions in ParC in the absence of GyrA substitutions raised the possibility that these mutations were not associated with FQ resistance. However, the direct evidence to demonstrate the association of specific QRDR mutations with FQs susceptibility is still limited and thus worth investigating.
In summary, plasmid profiling of E. coli isolates exhibiting the co-existence of both PMQR genes and other antimicrobial resistance genes on a single plasmid shows how they contribute to the rapid spread and increase in bacterial resistance, which is important to public health. The plasmid backgrounds of the PMQR genes were variable, ruling out the hypothesis for the spread of specific plasmids in Taiwan, however, continual epidemiologic surveillance and monitoring antimicrobial prescriptions and consumption would decrease the prevalence of FQ-resistant organisms and PMQR spread.

Methods
Sampling and isolation of E. coli. Bacteremia E. coli isolates were recovered in National Cheng Kung University hospital, 2001 to 2015. The Ethics Committee approved that no formal ethical approval was needed to use these clinically obtained materials, because the isolates were remnants from patient samples, and the data were analyzed anonymously. A total of 1,171 non-duplicate clinical isolates were randomly selected and stored at − 80 °C in Luria-Bertani (LB) broth containing 20% glycerol (v/v) until used. E. coli was identified in the clinical laboratory by colony morphology, Gram stain, biochemical tests, and the Vitek system (bioMérieux, Marcy l′ Etoile, France) according to the manufacturer's recommendations. Susceptibility to levofloxacin for E. coli isolates was determined by the disk diffusion method (5 μ g/disc, BD BBL ™ Sensi-Disc ™ , Sparks, MD, USA) on Mueller-Hinton (MH) agar (Bio-Rad, Marne la Coquette, France) based on the CLSI guidelines 24 . A total of 248 levofloxacin-nonsusceptible bacteremia E. coli isolates were identified for further analysis.
Antimicrobial susceptibility testing. Antimicrobial susceptibilities to ampicillin, ampicillin-sulbactam, gentamicin, colistin, and tigecycline (BD BBL ™ Sensi-Disc ™ ) were determined by the disk diffusion method on Mueller-Hinton agar 24 . MICs of selected antimicrobial agents (from Sigma-Aldrich: amikacin, cefepime, cefotaxime, ceftazidime, ciprofloxacin, fosfomycin, kanamycin, levofloxacin; from USP Standards: cefoxitin, imipenem, meropenem) were determined by the agar dilution method in accordance with CLSI guidelines 24 . E. coli ATCC 25922 and Pseudomonas aeruginosa ATCC 27853 were used as quality control strains. The interpretation of resistance to these antimicrobial agents was determined according to the recommendations of the CLSI 25 . Tigecycline and colistin susceptibilities were interpretated according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) 26 and previous study 27 , respectively. MDR E. coli was defined as isolates that were resistant to at least 3 classes of the tested antimicrobial agents 28 .
Characterization of antimicrobial resistance genes. All 248 FQ-resistant E. coli isolates were further screened for selected β -lactamases (bla TEM , bla SHV , bla CTX-M , bla DHA , and bla CMY ) and PMQR genes (qnr alleles, oxqAB, qepA, and aac(6′)Ib-cr) by PCR amplification with specific primers (Supplmentary Table S1). DNA sequencing was further carried out on β -lactamases (except bla TEM ) and PMQR genes, and the DNA sequences and deduced amino acid sequences were compared with genes in the GenBank database (http://www.ncbi.nlm. nih.gov/genbank/) to confirm the subtypes of antimicrobial resistance genes.
Screening for mutations in quinolone resistance-determining regions. GyrA and ParC QRDRs of 37 isolates harboring PMQR genes were examined by amplifying and sequencing gyrA (490 bp) and parC (470 bp) genes using primers (Supplmentary Table S1) described by Zhao et al. 19 . Amplimers were sequenced and amino acid mutations were determined using the control strain E. coli K-12 (NZ_AKBV01000001.1) as a reference.
Determination of the phylogenetic origin of E. coli isolates. Phylogenetic grouping of E. coli isolates was performed using a previously published method 29 . Primers used are described in Supplmentary Table S1. The PCR-amplified products were separated by electrophoresis on 1.8% agarose gels, stained with ethidium bromide, and assigned to one of the seven phylogenetic groups A, B1, B2, C, D, E and F.

Conjugation experiments and plasmid analysis.
The liquid mating-out assay was carried out to transfer PMQR genes from 37 FQ-resistant E. coli isolates to rifampicin-resistant E. coli C600 as described previously 30 . Transconjugants were selected on LB plates containing 256 μ g/mL rifampicin (Sigma) and 0.06 μ g/mL ciprofloxacin. The plasmids were extracted as described previously 9 , followed by electrophoresis in a 0.6% agarose gel at 50 V for 3 h and compared by co-electrophoresis with plasmids of known sizes from Salmonella OU7526 and a GeneRular TM DNA ladder (Fermentas, Burlington, ON, Canada) to predict the plasmid sizes 30 . Plasmids were typed by PCR-based replicon typing according to the previous study 31 .