Comparison of the inoculum size effects of antibiotics on IMP-6 β-lactamase-producing Enterobacteriaceae co-harboring plasmid-mediated quinolone resistance genes

Almost all cases of carbapenemase-producing Enterobacteriaceae infections in Japan are caused by blaIMP-positive Enterobacteriaceae (especially blaIMP-6) and infections caused by other types of carbapenemase-producing Enterobacteriaceae are quite rare. We examined drug resistance genes co-harboring with blaIMP-6 and their inoculum size effects. We screened β-lactamase genes, plasmid-mediated quinolone resistance (PMQR) genes, and aminoglycoside-modifying enzyme genes by PCR and performed sequencing for 14 blaIMP-6-positive Enterobacteriaceae. Further, all PMQR-positive isolates were submitted to conjugation and inoculum effect evaluation. Our data showed that 13 of the 14 isolates harbored CTX-M-2 and one co-harbored CTX-M-2 and CTX-M-1 as extended-spectrum β-lactamases. All isolates carried one or more PMQRs; aac(6’)-Ib-cr was the most prevalent (92.8%), and was followed by oqxA (64.3%), qnrS (50%), oqxAB (21.4%), and qnrB (14.3%). However, Klebsiella pneumoniae contains chromosomal OqxAB. Inoculum size effects were significant in all strains for meropenem, 13 strains for imipenem, 7 for levofloxacin, and 3 for amikacin. We observed that 11 of the experimental strains (100%), 8 strains (72.7%), and 1 strain showed inoculum size effects for meropenem, imipenem, and amikacin, respectively. However, four strains harbored qnr genes and two strains harbored qnr genes and QRDR mutations concurrently; no inoculum size effect was seen for levofloxacin. The blaIMP-6-positive Enterobacteriaceae that we studied was found to harbor at least one plasmid-mediated drug resistance gene. The inoculum size effect for carbapenems was thought to be mainly due to IMP-6-type metallo-β-lactamase; however qnrB and qnrS also had a minimal impact on the inoculum size effect for levofloxacin.


Introduction
Antimicrobial resistance in Gram-negative bacteria is an emerging and serious global threat to public health. Most importantly, carbapenemase-producing Enterobacteriaceae (CPE) confers a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 broad resistance to most β-lactam antibiotics including the carbapenems [1][2][3]. Although the number of CPE infection cases is increasing, the optimal treatment paradigm for CPE infections has not been well defined. Furthermore, there are numerous different types of carbapenemase enzymes, such as the IMP, VIM, KPC, OXA-48-like, and NDM enzymes, each conferring varying spectrums of resistance.
Much of the existing knowledge arises from reviews of case series and retrospective studies on the VIM-or KPC-producing CPEs which are now widespread in Europe and the United States. Based on published data, combination therapy is recommended for CPE infections with reference to the results of drug susceptibility tests, even when the organisms are susceptible to carbapenems. Almost all cases of CPE infection in Japan have been caused by bla IMPpositive Gram-negative bacteria (especially bla IMP-6 ) and cases of infection caused by other types of CPE are quite rare [4]. Unlike other types of CPE, bla IMP -positive Enterobacteriaceae usually show susceptibility to imipenem [5]. Furthermore, in many cases, bla IMP-6 -positive CPEs were susceptible to levofloxacin and amikacin [6]. Thus, based on the results of drug susceptibility tests, some infectious cases caused by IMP-type CPEs have been treated with quinolone monotherapy [7].
Quinolone resistance in Gram-negative bacteria is mostly mediated by point mutations that arise in the quinolone resistance-determining regions (QRDRs) of the gyrase and topoisomerase IV genes, leading to the modification of the target [8]. However, a previously unidentified resistance to quinolones mediated by the plasmid-mediated quinolone resistance genes (PMQRs) was recognized as a drug resistance mechanism [8][9][10]. These genes were first identified in 1998 and included the qnr proteins, aminoglycoside acetyltransferase, and the plasmid-mediated efflux pumps QepA and OqxAB [10]. Klebsiella pneumoniae contains chromosomal OqxAB. Aac(6 0 )-Ib-cr is one variant of aminoglycoside acetyltransferase which was found to reduce the activity of ciprofloxacin and aminoglycosides, including amikacin [8].
The inoculum size effect is a phenomenon whereby the measured value of the minimum inhibitory concentration (MIC) changes depending on the number of bacteria and its effect on β-lactamases, including metallo-β-lactamases (MBLs), has been described previously [11,12]. More recently, the inoculum size effect for the action of quinolone on bacteria expressing PMQR genes was described [9,13]. However, in one of these studies, it was reported that only half of the transconjugants of PMQRs (2/4 strains) showed an inoculum size effect for quinolone [9]. Furthermore, there has been no study describing the inoculum size effect for strains expressing both MBL and PMQRs.
The main aim of this study was to examine the drug resistance genes found in clinical isolates which produce IMP-6-type carbapenemase and PMQRs. In addition, we assessed the inoculum size effect on the MICs of carbapenems and quinolones.

Bacterial isolates
From April 2015 to December 2016, sequentially collected clinical isolates of bla IMP-6 -positive Enterobacteriaceae from Japan were studied. They included Escherichia coli (n = 2) and Klebsiella pneumoniae (n = 12). These isolates were non-duplicated and were obtained from 12 different hospitals around West Japan; eight isolates were from urine, four from respiratory specimens, one from bile, and one from a surgical wound.
The presence of aminoglycoside-modifying enzymes (aph(3 0 )-VI and ant(4 00 )-IIa), which were reported to reduce susceptibility to amikacin and were detected in Enterobacteriaceae, was also assessed by PCR using specific primers [21,22]. The QRDRs in gyrA and parC were amplified as previously described [23,24] and were sequenced to assess any co-existing chromosomal mutations [25].

Antimicrobial susceptibility testing
The MICs were evaluated using the agar dilution method for piperacillin, cefotaxime, cefmetazole, ceftazidime, aztreonam, cefzopran, meropenem, imipenem, levofloxacin, amikacin, and colistin at the standard inoculum size according to the Clinical and Laboratory Standards Institute (CLSI) guidelines [26]. We also evaluated the MICs of levofloxacin, amikacin, imipenem, meropenem, and colistin for strains using 10-and 100-fold of the colony forming units (CFU) in the standard inoculum. All results were interpreted according to the CLSI criteria describing in vitro susceptibility [27]. We defined that an inoculum size effect was significant if the MICs of the antibiotics showed a � 4-fold increase at 100-fold of the inoculum size compared to the standard inoculum size. We performed these tests at least twice for each strain to confirm the accuracy of the results.
Sequencing of the PCR products derived from the QRDRs in gyrA and parC showed a substitution in the QRDR of GyrA in one E. coli isolate (NR379) and four K. pneumoniae isolates (NR411, 417, 462, 490), and in the QRDR of ParC in the K. pneumoniae isolate NR417.

Conjugation experiment results
Conjugation experiments were performed successfully with 8 isolates, and 11 strains were obtained ( Table 2). The recipient strain (E. coli J53, NR3500) did not show an inoculum size effect and was susceptible to all the antibiotics we tested with or without the presence of QRDR mutation. Eleven isolates (100%), eight strains (72.7%), and one strain showed inoculum size effects for meropenem, imipenem, and amikacin, respectively. In addition, based on our definition, no inoculum size effect for levofloxacin was seen in the conjugant strains. Of these strains, four harbored qnr genes and two harbored both qnr genes and a QRDR mutation concurrently. When NR379/NR3500 was compared with NR 379/J53-2, NR379/NR3500 harbored the aac(6')-Ib-cr gene but did not harbor the gyrA mutation. There was no difference in the inoculum size effect for amikacin. Thus, our experiments showed that the inoculum size effects for carbapenems were apparent with IMP-6-type carbapenemase and that these effects were greater for meropenem than for imipenem. In addition, the qnrS and qnrB genes did not provide an apparent inoculum size effect for levofloxacin between the standard inoculum size and 100-fold the standard inoculum size.

Discussion
Recently, bla IMP -positive Enterobacteriaceae has become a serious problem throughout Asian countries, including Japan [4]. However, very few studies have reported other drug resistance genes besides β-lactamases in Japan. In this study, we found that bla IMP-6 -positive Enterobacteriaceae harbored multiple drug resistant genes including ESBLs and PMQRs. To the best of our knowledge, this is the first report evaluating drug resistance genes other than β-lactamases and the inoculum size effects for isolates producing IMP-6-type MBLs and PMQRs.
The resistance mechanisms for antibiotics are very complicated because organisms can harbor multiple resistant genes in their plasmids or chromosomes, and some of these genes act against multiple drug classes. In our study, we focused on IMP-6-type carbapenemase and PMQRs and the impact of the inoculum size effects of these genes. Margaritis et al. reported that half of the transconjugants of PMQRs (2/4 strains) showed inoculum size effects for levofloxacin (1-, 2-, 4-, and 16-fold: MIC of 10 7 CFU/mL per MIC of 10 5 CFU/mL) [9]. However, they also showed that the inoculum size effects for quinolone were significant for laboratory strains which did not harbor acquired quinolone resistance genes. Rice reported that resistance to quinolone arises as a result of a combination of various resistance mechanisms [29]. We conclude that qnr genes had only a small impact on the inoculum size effects for levofloxacin at 100-fold of the standard inoculum size. The change might be significant at higher concentrations than the standard inoculum size. Further studies should be performed to identify other factors, such as efflux pumps or unknown mechanisms that could explain this difference. One experimental strain (NR 379 and NR 3500), which harbored the qnrB gene and gyrA mutation, showed susceptibility to levofloxacin based on the CLSI breakpoint (M100-S25) but was resistant based on the EUCAST clinical breakpoint (ver. 8.1) at 100-fold of the inoculum size [28]. Therefore, qnr genes should not be completely ignored if the strain shows a decreased susceptibility to levofloxacin in the setting of a huge level of organismal infection (e.g. an abscess or bacteremia).
Presently, a combination therapeutic strategy is recommended for carbapenemase-producing Enterobacteriaceae infections [1,2,30]. These recommendations are based on clinical experiences with VIM-or KPC-type carbapenemase-producing Enterobacteriaceae infections where the MICs for carbapenem were high. In contrast, the reported susceptibility to meropenem and imipenem among bla IMP-6 -positive E. coli was approximately 70% and 100%, respectively [4]. Pang et al. have reported that most pathogens were confirmed to produce IMP-type carbapenemases and some cases were successfully treated with quinolone monotherapy as a definitive therapy based on drug susceptibility tests [7]. However, in the present study, the isolates did not show resistance to levofloxacin, amikacin, and carbapenems based on the CLSI definition. the MICs of some isolates turned out to be high enough to be resistant to these drugs, especially to meropenem (the MIC for all isolates was � 16 μg/mL), at 100-fold of the standard inoculum size. In fact, it has been reported that 55.6% of infectious cases caused by KPC-type carbapenemase-producing K. pneumoniae showed susceptibility in automated drug susceptibility tests, but failed to be treated by imipenem or meropenem [31]. Thus, it might be necessary for successful antibiotic treatment to introduce aggressive interventional procedures to reduce the quantity of the organisms which produce drug resistant genes such as IMP-6 MBLs.

Conclusions
The bla IMP-6 -positive Enterobacteriaceae we studied harbored at least one plasmid-mediated drug resistance gene, other than ESBLs and carbapenemase, at the same time. Furthermore, these isolates showed inoculum size effects for levofloxacin, amikacin, and carbapenems (especially for meropenem compared with imipenem). The inoculum size effect for carbapenems was thought to be mainly due to IMP-6-type MBLs; however qnrB and qnrS also had a small impact on the inoculum size effect for levofloxacin between the standard inoculum size and 100-fold of the standard inoculum size.