Multiple mechanisms contributing to ciprofloxacin resistance among Gram negative bacteria causing infections to cancer patients

Fluoroquinolones have been used for prophylaxis against infections in cancer patients but their impact on the resistance mechanisms still require further investigation. To elucidate mechanisms underlying ciprofloxacin (CIP) resistance in Gram-negative pathogens causing infections to cancer patients, 169 isolates were investigated. Broth microdilution assays showed high-level CIP resistance in 89.3% of the isolates. Target site mutations were analyzed using PCR and DNA sequencing in 15 selected isolates. Of them, all had gyrA mutations (codons 83 and 87) with parC mutations (codons 80 and 84) in 93.3%. All isolates were screened for plasmid-mediated quinolone resistance (PMQR) genes and 56.8% of them were positive in this respect. Among PMQR genes, aac(6′)-Ib-cr predominated (42.6%) while qnr genes were harbored by 32.5%. This comprised qnrS in 26.6% and qnrB in 6.5%. Clonality of the qnr-positive isolates using ERIC-PCR revealed that most of them were not clonal. CIP MIC reduction by CCCP, an efflux pump inhibitor, was studied and the results revealed that contribution of efflux activity was observed in 18.3% of the isolates. Furthermore, most fluoroquinolone resistance mechanisms were detected among Gram-negative isolates recovered from cancer patients. Target site mutations had the highest impact on CIP resistance as compared to PMQRs and efflux activity.

resistance 8 , while OqxAB is a novel transmissible resistance-nodulation-division (RND) multidrug efflux pump that was found to reduce susceptibility to CIP and nalidixic acid 9 .
Some studies worldwide have previously investigated the effect of CIP prophylaxis on the emergence of resistant bacteria in cancer patients [10][11][12] , but its impact on various mechanisms of resistance was not deliberately addressed. Accordingly, the aim of the current study was to elucidate the contribution of different mechanisms of CIP resistance among CIP non-susceptible Gram negative bacteria causing infections to cancer patients attending a tertiary cancer hospital in Cairo, Egypt during the study period.
Target-affecting resistance mechanisms. DNA sequencing of the QRDRs of gyrA and parC genes from 15 isolates revealed at least one missense mutation, including a gyrA mutation, in all isolates. Codon 83 of the gyrA gene had the highest incidence of such mutations (15/15, 100%) with the resulting substitutions: Ser/Leu, Ile, Tyr or Phe in all species except P. aeruginosa where threonine was substituted by isoleucine. This was accompanied by another mutation at codon 87 in 40% (6/15) of the isolates (Asp/Asn, Tyr or Ala). Amino acid substitutions in ParC were evident in 93.3% (14/15) of the isolates at codon 80 (13/15, 86.7%) including: Ser to Ile or Leu substitutions and codon 84 where Lys was substituted for Glu in one isolate (6.7%). Detailed information on the isolates tested for target site mutations are shown in Table 2. Double mutations (1 gyrA/1 parC) were the most frequently encountered among the tested isolates (8/15, 53.3%) followed by three mutations (2 gyrA/1 parC) found in 40% (6/15), while only 6.7% (1/15) showed a single gyrA mutation.
The Kruskal-Wallis test showed that the distribution of CIP MIC across the groups of isolates carrying different numbers of target site mutations is significantly different (P-value = 0.013). The highest MIC 50 (128 mg/L) was   Of all tested PMQR determinants, only qnrS and qnrB genes were found in non-fermenter isolates. One A. baumannii isolate (4.3%) harbored the qnrS gene and the qnrB gene was carried by one P. aeruginosa isolate (20.0%). The prevalence of different PMQR determinants with respect to bacterial species is shown in Table 3.
Although higher CIP MIC 50 was observed among PMQR-positive isolates than PMQR-negative isolates (128 versus 64 mg/L, respectively), the distribution of CIP MIC across the two groups was not significantly different (Mann Whitney U test, P-value = 0.140).
Clonal Relatedness of qnr-positive isolates. ERIC-PCR was performed on 33 qnr-positive K. pneumoniae isolates and showed 32 banding patterns classified into 22 ERIC types (K1-K22) based on a similarity percentage of 85% or more (see Supplementary Fig. S1). On the other hand, greater diversity was evident among 19 qnr-positive E. coli isolates that showed 19 amplification patterns comprising 14 ERIC types (E1-E14) (see Supplementary Fig. S1). Significant association between ERIC types and PMQR determinants was observed in K. pneumoniae isolates (Fishers Exact Test, P = 0.005). In contrast such association was not evident in E. coli isolates (Fishers Exact Test, P = 0.674).
Efflux activity. Using the efflux pump inhibitor, four fold or more CIP MIC reduction (MDF value of more ≥4) was observed in 18.3% of the isolates, most commonly in non-fermenters (50.0%). On the other hand, 66 isolates (39.1%) exhibited a MDF value of 2, while CIP MIC of 72 isolates (42.6%) did not show any difference by combining CCCP. The distribution of MDF values among various species as well as the prevalence of isolates showing active efflux activity are shown in Fig. 1.
Although efflux activity was associated with higher CIP MIC 50 (128 versus 64 mg/L), the distribution of CIP MIC was found to be the same across the groups of isolates showing positive efflux activity and others (Mann Whitney U test, P-value = 0.406).

Discussion
The current study was conducted with the aim of elucidating CIP resistance acquisition mechanisms among Gram negative pathogens causing infections to cancer patients, a population to whom FQs are widely prescribed for prophylaxis of bacterial infections. For this purpose, 169 Gram-negative isolates recovered from different clinical specimens from infected cancer patients in Egypt were investigated. Only CIP-resistant isolates were included in the current study. Broth microdilution assays revealed high-level CIP resistance (CIP MIC ≥ 32 mg/L) in the majority of isolates (89.3%). Such a noticeable prevalence of high-level resistance was also reported among Enterobacteriaceae isolates recovered from a tertiary hospital in Poland 13 .
The acquisition of FQs resistance is known to be most attributable to mutations in the target enzymes-coding genes. A genetic analysis based on the QRDR sequences of gyrA and parC genes of selected isolates of various Gram negative genera revealed that all strains carried point (missense) mutations in gyrA codon 83 with simultaneous mutations at codon 87 in 40.0% of the isolates. Such mutation sites are the most commonly encountered worldwide 5,14 . Their high mutability is likely due to being located near the active site of the DNA gyrase 15 . On Nonfermenters (n = 28) All tested non-fermenter isolates showed two target site mutations (1 gyrA/1 parC). Such mutation pattern was reported by previous studies as the most common among CIP-resistant A. baumannii [16][17][18] and P. aeruginosa isolates [19][20][21] . Nevertheless, single gyrA mutations were sufficient to cause clinically significant levels of resistance in both species according to some studies [19][20][21][22][23][24] . Triple mutations (2gyrA/1parC) were also previously reported 16 .
Consistent with the literature 5,14 , our results showed a substantial increase in CIP resistance levels with accumulating mutations in one or both target enzymes. Notably, a wide variability in CIP MIC was shown by the isolates carrying two target site mutations, likely due to the contribution of other resistance mechanisms. Of particular interest was the finding that double mutations (1gyrA/1parC) were sufficient to confer high-level CIP resistance among non-fermenter isolates. In contrast, three mutations were carried by 75.0% of Enterobacteriaceae isolates showing high-level CIP resistance. This is probably due to the intrinsic resistance of non-fermenters to antibacterial agents compared to members of Enterobacteriaceae as a result of low permeability and/or over expression of some efflux pumps 5,25 .
Despite their inability to confer FQs resistance, PMQRs play an important role in the acquisition of clinical resistance to FQs. The low-level quinolone-resistance phenotypes conferred by PMQR genes may allow some low-fitness mutants below the resistance breakpoint to evolve clinical resistance with just one or two mutations 26 . Most importantly, PMQRs can spread horizontally among different Gram negative species 5 . PCR screening of PMQR genes in the current study revealed their carriage by 66.7% of Enterobacteriaceae isolates. Such a high prevalence reported here and by other studies performed worldwide among FQ-resistant Enterobacteriaceae isolates, reflects their crucial role in acquisition of FQs resistance. In contrast, lower prevalence of PMQRs was reported in CIP-resistant Enterobacteriaceae isolates from Poland (22.8%) 13 and FQ-resistant E. coli from Taiwan (14.9%) 27 . Consistent with other studies 13,28 , PMQR genes were more frequently detected among the isolates of K. pneumoniae (70.4%) than E. coli (67.5%) and other Enterobacteriaceae species (28.6%).
Although higher frequencies of PMQRs are expected from isolates preselected for extended-spectrum β-lactamases (ESBLs) production 29 , lower PMQRs prevalence was reported among ESBL-producer K. pneumoniae from Kuwait (30.6%) 30 , E. coli from Tunisia (50.0%) 28 and Argentina (62.5%) 31 . Whether the high prevalence of PMQR-producers among our collection is linked to their recovery from cancer patients or due to their high regional prevalence, this was further analyzed by reviewing other Egyptian studies. In a recent study by El-badawy et al., qnr genes (qnrB and qnrS) were detected in 91.2% of FQ-resistant K. pneumoniae isolates. This was much higher than their prevalence among the same species in our collection (61.1%). On the other hand, the prevalence of qnr genes in E. coli isolates in the current study was comparable (23.8%) to that reported, in ESBL-producer E. coli, by another Egyptian study (26.6%), while aac(6′)-Ib-cr was more frequently detected in our collection (61.2% versus 23.3%) 32 .
Enzyme modification through AAC(6′)-Ib-cr was the most prevalent PMQR mechanism among Enterobacteriaceae isolates in the current study (51.5%). A similar predominance of this mechanism was reported by other studies 13,33,34 . Higher prevalence of the aac(6′)-Ib-cr variant compared to the wild type gene was evident in our collection (87.8%) as well as previous studies [34][35][36] . This is possibly due to its wider spectrum of activity that uniquely encompasses two different classes of antimicrobial agents including quinolones and aminoglycosides 37 . Of the qnr genes, qnrB seems to be the most prevalent worldwide 29 . In our study, however, qnrS was more frequently detected than qnrB (31.2% versus 7.1%). The qnrA gene, the least frequent of the qnr genes, was described in a single P. mirabilis isolate. Nevertheless, it was absent in recent worldwide surveillance studies 28,31,36,38 .
Concerning plasmid-encoded efflux pumps, QepA-producers were not identified. Similar findings were described in several studies in Tunisia 28 , Argentina 39 , and France 36 . In contrast, qepA gene was detected among K. pneumoniae isolates from non cancer patients in Egypt 40 and E. coli isolates from Algeria 41 and Pakistan 33 . On the other hand, oqxAB genes were only detected among members of Klebsiella spp. in which the oqxAB operon is mostly located on the chromosome and has no correlation with CIP MIC, likely due to different expression levels 42 . Accordingly, oqxA and oqxB-positive Klebsiella spp. isolates were not considered among the PMQR-positive isolates in the current study.
A relatively few number of studies worldwide were concerned with detection of PMQR genes among non-fermenter isolates in which they are known to be rare. In the current study, a significantly lower prevalence of PMQRs was found among non-fermenter than Enterobacteriaceae isolates (7.1% versus 66.7%, P < 0.001) suggesting higher contribution of other resistance mechanisms 5 . The qnrS and qnrB genes were detected in 4.3% of A. baumannii and 20.0% of P. aeruginosa isolates, respectively. A similar prevalence of qnrB gene was reported from Poland among P. aeruginosa isolates 43 , other PMQR genes were also previously reported [44][45][46] . In A. baumannii, qnrS and qnrB were reported in China 47 and qnrA gene was reported in Algeria 48 .
Although their carriage did not have a significant effect on CIP MIC among the tested isolates (P = 0.140), detection of PMQR genes in almost all tested species in our collection confirms their interspecies transferability. More alarming is the previous knowledge of their existence in multi-resistance plasmids linked to other resistance determinants 14,49 . In light of this, it should be stressed that carriage of PMQR determinants are of great public health concern owing to their ability to confer multidrug resistance to other potential recipient strains. Failure of empirical antimicrobial treatment of Gram negative infections, following CIP prophylaxis, in cancer patients is then inevitably expected.
To investigate the dissemination of qnr genes among the isolates of the same species, it was of interest to examine their clonal relatedness. Molecular characterization using ERIC-PCR showed a considerable genomic diversity among the qnr-positive isolates. Twenty two ERIC types were identified among 33K. pneumoniae isolates where up to seven isolates had the same ERIC type. Moreover, ERIC types were significantly related to PMQR determinants (P = 0.005), suggesting a low-level clonal dissemination. On the other hand, among 19 E. coli isolates, 14 ERIC types were evident without significant relation to PMQRs (P = 0.674) indicating that most of them were not clonal. This provides additional evidence on the horizontal transfer of PMQR genes described elsewhere rather than their clonal dissemination 13,30,50 .
Active efflux and/or reduced influx of FQs can reduce their cytoplasmic concentrations conferring FQs resistance 5 . To elucidate the involvement of efflux pumps in CIP resistance in our collection, CIP MIC was determined in the presence of sub-inhibitory concentrations of CCCP, a protonophore that reduces ATP production and increases bacterial cell membrane permeability 51 . Based on a fourfold or greater reduction in CIP MIC (MDF ≥ 4) as a criterion for significance 52 , efflux activity was evident in 18.3% of the isolates. Consistent with the literature 5 , this was significantly more frequent among non-fermenter than Enterobacteriaceae isolates (50.0% versus 12.1%, P < 0.001). The distribution of CIP MIC was found to be the same among the isolates showing positive efflux activity and others (P = 0.406). A possible explanation is that the effect of over active efflux pumps is often limited to about four-to eightfold increase in inhibitory concentrations (except for rare cases in our collection) 5 . Such effect may be overlooked when other mechanisms, particularly target site mutations, are established. Similar findings were reported by other studies as well 24,53 . The results of our study, however, demonstrate a loss of CIP resistance in 19.3% of the isolates showing active efflux activity by combining CCCP. In this context, the use of efflux pump inhibitors appears to be a promising strategy to restore antibacterial potency.

Conclusion
The current study provides information on the mechanisms underlying CIP resistance among Gram negative isolates causing infections to cancer patients in Egypt. Being affected by the same resistance mechanisms, the results of the current study could be extrapolated to other FQs that are widely used for prophylaxis of bacterial infections in this population. The information provided herein showed high-level CIP resistance in the majority of isolates. Resistance was a result of a complex interplay between most of the known FQs resistance mechanisms each demonstrated in a high prevalence. The current results also provide additional evidence that chromosomal mutations in sequences encoding GyrA and ParC, detected in all tested isolates, play an essential role in CIP resistance. It is noteworthy that the dissemination of PMQR genes, with a high prevalence, was demonstrated in clonally unrelated isolates. Reduced antibiotic accumulation arising from the over expression of efflux pump systems was also evident but with low impact on CIP MIC. Failure of empirical treatments of Gram negative infections in cancer patients following CIP prophylaxis is a possible scenario and it would be attributed to the coexistence of other genes conferring resistance to different antimicrobial agents.
VITEK®2 system (bioMérieux, Marcy-l′Étoile, France) was used for species identification and antimicrobial susceptibility testing as a part of the routine laboratory work in the hospital. All CIP non-susceptible Gram negative isolates (MIC ≥ 2 mg/L) were selected for the study. The minimum inhibitory concentration (MIC) of CIP was determined using the broth microdilution method in accordance with the Clinical and Laboratory Standards Institute (CLSI) guidelines 54 . E. coli ATCC 25922 was used as a quality control strain. Approval of the study protocol was received from the Ethical Review Boards of each of Ain Shams University, October University for Modern Sciences and Arts and the National Cancer Institute, Cairo University. All methods were performed in accordance with the required guidelines and regulations.
Genotypic detection of target-affecting resistance mechanisms. Fifteen isolates of different species were selected for analysis of mutations in the QRDRs of the gyrA and parC genes. Those comprised five K. pneumoniae isolates, three E. coli isolates, three P. aeruginosa isolates, three A. baumannii isolates, and a single P. mirabilis isolate. Specific primers ( Table 4) for amplification of gene fragments encompassing the QRDR and flanking nucleotide sequences in different species were designed based on the wild-type gene sequences of corresponding standard strains published in the GenBank nucleotide sequence database (www.ncbi.nlm.nih.gov). Genomic DNA was extracted using Genomic DNA Purification Kit (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions. Amplification was performed in a final volume of 25 μL containing 1 μL of each primer at a concentration of 12.5 μM, 12.5 μl MyTaq TM Red Mix, and approximately 100 ng of chromosomal DNA. PCR reactions were performed using TAdvanced thermal cycler (Biometra, Germany) with an initial denaturing cycle at 95 °C for 5 min followed by 35 cycles of 95 °C for 30 s, annealing temperature for 30 s, and 72 °C for 1 min, with a final extension step at 72 °C for 5 min.
All PCR products were purified using DNA Clean & Concentrator TM -25 Kit (Zymo Research, Orange, CA, USA) and sequenced by the Sanger method using an ABI 373 A DNA sequencer (PE Applied Biosystems, Life technologies Inc., CA, USA). The predicted amino acid sequences were analysed for amino acid changes by comparison to the wild-type GyrA and ParC sequences of the corresponding standard strains deposited in the GenBank at the National Center for Biotechnology Information website (http://www.ncbi.nlm.nih.gov/blast) using the BLAST tool.
Screening of plasmid-mediated resistance mechanisms. Using multiplex PCR, all the isolates were screened for the PMQR genes; qnrA, qnrB, qnrS, and aac(6′)-Ib, while monoplex PCRs were used for the detection of qepA, oqxA and oqxB genes using gene specific primers listed in Table 5, as we described previously 55  Clonality analysis of qnr-positive isolates. All qnr-positive isolates of the same species were analyzed by ERIC-PCR using ERIC2 primer; 5′-AAGTAAGTGACTGGGGTGAGCG -3′ as described previously 56 . Amplification patterns were analyzed using GelCompar II software (Applied Math, Kortrijk, Belgium). Similarity clustering analyses were performed using UPGMA (unweighted pair group method with arithmetic mean) and Dice correlation coefficient with a position tolerance of 1.5%.   58 . An MDF value of 4 or more was considered as a significant effect due to efflux pump inhibition via CCCP.
Nucleotide sequence accession numbers. The nucleotide sequences of gyrA and parC genes of some clinical isolates analyzed in the current study were submitted to GenBank nucleotide sequence database. Accession numbers are shown in Table 6.
Statistical analyses. All tests of significance were two-tailed, and statistical significance was defined at   Table 6. Accession numbers submitted to the NCBI of gyrA and parC fragments amplified from different isolates of various genera.