Assessment of a phenotypic algorithm to detect plasmid-mediated quinolone resistance in Enterobacteriaceae

Sir, Quinolone resistance in Gram-negative bacteria is the result of mutations in the quinolone resistance-determining region (QRDR) in chromosomally located genes encoding type II topoisomerases and, to a lesser extent, altered permeability.1,2 Plasmid-mediated quinolone resistance (PMQR) has also been reported as encoding different proteins: Qnr proteins, acetyltransferase AAC(6′)-Ib-cr variant and the QepA and OqxAB active efflux pumps.2 In Enterobacteriaceae, a single mutation in the QRDR of GyrA (frequently at Ser83) results in high-level resistance to nalidixic acid (MIC > 32 mg/L with absence of inhibition zone) and is easily detected. Clinical resistance to fluoroquinolones (ciprofloxacin MIC >2 mg/L according to CLSI)3 is attained with additional chromosomal mutations or the acquisition of PMQR. In the case of isolates harbouring PMQR, but lacking QRDR mutations, these determinants confer low-level resistance to quinolones and/or fluoroquinolones and molecular tests are needed to detect it. The detection of PMQR usually takes up to six PCRs. There are to date no specific phenotypic tests for the detection of PMQR mechanisms. The current prevalence of PMQR seems to be increasing.2 The aim of this study was to determine the capacity of phenotypic detection of PMQR in Enterobacteriaceae with low-level quinolone resistance (LLQR) without modifications at the QRDR. This phenotype has been reported mainly in Salmonella.4 To assess this issue, a sample size of 49 positive cases and 49 negative cases was estimated necessary for a 95% CI, proportion expected of 85%, and estimation accuracy of 0.1, using NQuery Advisor. In our study, 99 (57 Klebsiella pneumoniae, 27 Escherichia coli, 7 Enterobacter cloacae, 2 Klebsiella oxytoca, 1 Enterobacter aerogenes, 1 Citrobacter koseri, 1 Citrobacter freundii, 1 Morganella morganii, 1 Proteus mirabilis and 1 Serratia marcescens) isolates (of 10 874 clinical isolates of Enterobacteriaceae) were selected during October 2012 to June 2014. They were susceptible to nalidixic acid, MIC ≤16 mg/L, and showed reduced susceptibility to ciprofloxacin, MIC = 1–2 mg/L, on a Wider C095-31WREV1 panel (Soria Melguizo). The panel included the following quinolones and concentrations: nalidixic acid 16 mg/L and ciprofloxacin 0.12–1–2 mg/L. Ten additional enterobacterial isolates (3 K. pneumoniae and 7 E. coli) fully susceptible to quinolones were included as controls. Antibiotic susceptibility testing was performed by both microdilution and disc diffusion tests as follows: susceptibility to ciprofloxacin and nalidixic acid was determined by disc diffusion (http://www.eucast.org/) and microdilution;3 and susceptibility to norfloxacin, ofloxacin, moxifloxacin and levofloxacin was determined by microdilution. After microdilution (Table S1, available as Supplementary data at JAC Online), MIC values ranged between 2 and 64 mg/L for nalidixic acid and between 0.03 and 4 mg/L for ciprofloxacin, so that 93% of isolates could be defined as included in this low-level ciprofloxacin-resistant phenotype. To investigate this LLQR phenotype with no mutations in the genes coding for type II topoisomerases (data not shown), the isolates were screened by PCR for qnr (qnrA, qnrB, qnrC, qnrD and qnrS), aac(6′)-Ib-cr, qepA and oqxAB genes.5 For K. pneumoniae isolates, the presence of oqxAB genes in plasmids was confirmed by blot hybridization since that species naturally harbours those genes.6 No PMQR genes were identified for the 10 fully susceptible isolates. For the LLQR phenotype, 51 isolates (51%) harboured at least one PMQR mechanism. qnrS-like genes were the most prevalent (25 isolates, 49% of all PMQR-positive isolates) followed by qnrB-like (20, 39%), aac(6′)-Ib-cr (9, 18%), oqxAB (6, 12%), qnrA-like (2, 4%) and qnrD-like (2, 4%). The presence of at least two PMQR genes in the same isolate was observed in 13% of the collection. Using IBM® SPSS® Statistics 20.0, receiver operating characteristic (ROC) curves were constructed for the 99 isolates with an LLQR phenotype to assess the sensitivity and specificity for PMQR detection according to susceptibility data. Ciprofloxacin MIC AUC was compared with the other quinolones using the Delong's test (P < 0.05 indicating significance). ROC curve analysis of MIC values showed that the maximum sensitivity/specificity combinations for detecting a PMQR determinant were ≥0.25 mg/L for ciprofloxacin (94%/85%, AUC 0.932), ≥0.5 mg/L for ofloxacin (94%/79%, AUC 0.867, P = 0.013), ≥0.5 mg/L for norfloxacin (94%/79%, AUC 0.878, P = 0.008) ≥0.5 mg/L for levofloxacin (86%/89%, AUC 0.918, P = 0.098) and ≥1 mg/L for moxifloxacin (90%/92%, AUC 0.947, P = 0.195) (Figure 1). Disc diffusion showed the maximum sensitivity/specificity combinations for detection of the PMQR determinant were ≤27 mm for ciprofloxacin (82%/83%, AUC 0.882) and ≤19 mm for nalidixic acid (88%/88%, AUC 0.909), and only one isolate yielded a nalidixic acid inhibition zone <13 mm (Figure 1). The combination of inhibition zone information for ciprofloxacin and nalidixic acid showed a sensitivity of 90% and specificity of 92% (AUC 0.933, P > 0.05) for isolates with inhibition zones of 13–21 mm for nalidixic acid and ≤28 mm for ciprofloxacin. ROC curves for several fluoroquinolones (a, ofloxacin; b, norfloxacin; c, levofloxacin; d, moxifloxacin), compared with ciprofloxacin, according to susceptibility values obtained by microdilution (mg/L), and for nalidixic acid and ciprofloxacin, according to disc diffusion (mm) susceptibility testing (e). In conclusion, in the absence of QRDR modifications, the presence of PMQR genes can easily be suspected using combinations of specific quinolone inhibition zones to infer its molecular basis. Nalidixic acid alone cannot be used as an indicator of fluoroquinolone susceptibility7,8 because hidden PMQR mechanisms are frequently associated with the phenotype being studied here. In daily practice, however, PMQR genes may be suspected using the combination of inhibition zone information for nalidixic acid and ciprofloxacin to rule out the presence of PMQR. Detection of these mechanisms would play a role for epidemiological purposes, to know the trend and to assess the impact in fluoroquinolone resistance prevalence. This work was supported by the Ministerio de Sanidad y Consumo, Instituto de Salud Carlos III (projects PI11-00934 and PI14/00940) and the Consejería de Innovación Ciencia y Empresa, Junta de Andalucía (P11-CTS-7730), Spain, the Plan Nacional de I+D+i 2008-2011 and the Instituto de Salud Carlos III, Subdirección General de Redes y Centros de Investigación Cooperativa, Ministerio de Economía y Competitividad, the Spanish Network for Research in Infectious Diseases (REIPI RD12/0015)—co-financed by the European Development Regional Fund ‘A Way to Achieve Europe’. None to declare.