Resistance in Escherichia coli: variable contribution of efflux pumps with respect to different fluoroquinolones

Authors


Correspondence

Antoine Huguet, Anses, Fougères laboratory, La Haute Marche BP 90203, 35302 Fougères Cedex, France. E-mail: antoine.huguet@anses.fr

Abstract

Aims

Resistance to fluoroquinolones is partially the result of a decrease in drug accumulation in Escherichia coli through different mechanisms. However, the variable contribution of these mechanisms with respect to different fluoroquinolones is poorly investigated. Therefore, the current study aimed to compare the contribution of resistance attributed to efflux-mediated mechanisms for different fluoroquinolones.

Methods and Results

Susceptibility of enrofloxacin, marbofloxacin and ciprofloxacin were compared after treatment with an efflux pump inhibitor in 17 ciprofloxacin-resistant E. coli isolates, and also the expression profile of the genes encoding the porins and efflux pumps involved in this resistance was evaluated. After treatment with the efflux pump inhibitor Phe-Arg-β-naphthylamide (PAβN), susceptibilities differed significantly between antimicrobial agents, the decrease for MIC being higher for enrofloxacin than for marbofloxacin or ciprofloxacin. AcrB expression level increased significantly (+26%) in ciprofloxacin-resistant E. coli isolates compared with ciprofloxacin-susceptible isolates, whereas the expression level decreased for ompF (−50%) and ompC (−30%).

Conclusions

There was a higher contribution of resistance nodulation division (RND) efflux pumps to resistance to hydrophobic fluoroquinolones.

Significance and Impact of the Study

Comparison between expression profile of efflux pumps and hydrophobicity of the antimicrobial agents could result in variable resistance for different fluoroquinolones.

Introduction

In Escherichia coli, resistance to fluoroquinolones is the result of different mechanisms: mutations in genes encoding the drug target enzymes, DNA gyrase and topoisomerase IV, a plasmid-mediated quinolone resistance, but also a decrease in drug accumulation through an increase in bacterial impermeability and/or an active drug efflux (Hopkins et al. 2005). Indeed, an absence of outer membrane porins (OMPs) associated or not with an overexpression of efflux pumps leads to a decreased susceptibility of E. coli to fluoroquinolones (Hooper 2001; Webber and Piddock 2001). Different chromosome-dependent efflux systems responsible for fluoroquinolone resistance have been reported, and they include the RND family, the major facilitator superfamily (MFS) and the multidrug and toxic compound extrusion (MATE) family (Poole 2000). OMPs and efflux pumps involved in fluoroquinolone resistance are OmpC and OmpF of the porin family (Cohen et al. 1988; Tavio et al. 1999), AcrB, AcrF and YhiV of the RND family, NorE of the MATE family, and MdfA of the MFS family (Poole 2000; Cattoir 2004; Bohnert et al. 2007). While other studies have evaluated the contribution of efflux-mediated mechanisms to fluoroquinolone resistance (Mazzariol et al. 2000; Saenz et al. 2004; Kern et al. 2006), direct comparisons between fluoroquinolones are limited.

In the current study, we investigated the effect of an efflux pump inhibitor on the ciprofloxacin, enrofloxacin and marbofloxacin susceptibility of ciprofloxacin-resistant (CIP-R) E. coli isolates, and the expression level of ompC, ompF, acrB, acrF, yhiV, norE and mdfA in ciprofloxacin-susceptible (CIP-S) and CIP-R E. coli isolates.

Materials and methods

Bacterial isolates

During the French monitoring programme on antimicrobial resistance in bacteria of animal origin, on average 100 caecal samples for poultry production, and 100 faecal samples for pig and bovine production were collected each year from 10 slaughter houses between 1999 and 2006. Isolation of E. coli and antimicrobial susceptibility testing were performed according to Clinical and Laboratory Standards Institute (CLSI) documents (2008). From all isolates, in the present study 34 bacterial isolates were selected according to their susceptibility to ciprofloxacin, as well as their animal origin: 17 E. coli isolates with MIC value ranging from 0·008 to 0·03 μg ml−1 were qualified as CIP-S, and 17 E. coli isolates with a MIC value ranging from 4 to 16 μg ml−1 were qualified as CIP-R. For both CIP-S and CIP-R isolates, 7 isolates were collected from bovine faecal samples, four from pig faecal samples and six from poultry caecal samples.

Chemicals

Ciprofloxacin and enrofloxacin were purchased from Fluka (Lyon, France), marbofloxacin was obtained from Vetoquinol (Paris, France), and the Phe-Arg-β-naphthylamide (PAβN) was supplied by Sigma-Aldrich (Lyon, France).

Antimicrobial susceptibility testing

The susceptibilities to ciprofloxacin, enrofloxacin and marbofloxacin of CIP-R E. coli isolates were tested using the broth microdilution method according to CLSI documents (2008). Assays were performed in sterile 96-well polypropylene microtitre plates (Greiner, Courtaboeuf, France) containing serial twofold dilutions of each antibiotic at final concentrations ranging from 128 to 0·25 μg ml−1. Each CIP-R E. coli isolate was tested for MIC in the presence or absence of 100 μg ml−1 of the efflux pump inhibitor PAβN. Escherichia coli ATCC 25922 was used as quality control for antimicrobial susceptibility testing. In addition, for the effect of the PAβN on antimicrobial susceptibility, a MIC decrease factor (MDF) was determined. For each CIP-R E. coli isolate, the MDF was calculated as follows: MDF = MICwithout PAβN/MICwith PAβN.

RNA extraction, cDNA synthesis, primer design and quantitative PCR

Total RNA of CIP-R and CIP-S E. coli isolates was extracted with the RNeasy® Protect Bacteria Mini Kit (Qiagen, Courtaboeuf, France) according to the manufacturer's instructions and with the following minor modifications. Bacterial cells were disrupted using lysozyme (Sigma-Aldrich) and proteinase K (Qiagen), and residual amounts of DNA were removed during the RNA purification using the RNase-Free DNase Set Kit (Qiagen) according to the manufacturer's instructions. Reverse transcription (RT) was performed for one μg of total RNA with the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions, and ribonucleases were inhibited with two IU of RNaseOUT (Invitrogen, Cergy Pontoise, France). From the National Center for Biotechnology Information (NCBI), complete sequence of each gene, forward and reverse specific complementary primers were designed using the Molecular Beacons Designer software (version 6·0; PREMIER Biosoft, Palo Alto, CA, USA). The gene gapA was chosen as a reference gene as it did not exhibit any significant variation in expression among the samples. All primers were purchased from Sigma-Aldrich and are listed in Table 1. Quantitative PCR was carried out on a Chromo4 Real-Time Detector (Bio-Rad, Marnes-la-Coquette, France). For each gene (the efflux pumps AcrB, AcrF, YhiV, MdfA and NorE; the porins OmpC and OmpF; the reference GapA), reactions were performed in duplicate in a total volume of 20 μl containing 10 μl Power SYBR® GREEN PCR Master Mix (Applied Biosystems, Foster City, CA), 5 μl cDNA, 1 μl of each primer (0·3 μmol l−1 final concentration) and 3 μl sterile water. After an initial step (94°C for 7 min), the thermal cycling protocol was as follows: 40 cycles of PCR for 15 s at 95°C for denaturation, for 15 s at 52 or 55°C for annealing, and for 15 s at 72°C for extension. Data were analysed with the Opticon Monitor software (version 3·0; Bio-Rad). Melting curve analysis was performed for each gene to check the specificity and identity of the quantitative PCR products. For each tested sample, the relative amount of the target transcript, called the relative quantity (RQ), was determined by comparison with the corresponding standard curve. The cDNA samples were diluted sufficiently to obtain Ct values in the middle of the standard curve, and the quantitative PCR was performed for this dilution. The RQ values were calculated by applying the following equation: RQ = 10([Ct-intercept]/slope). For each sample, the RQ was then normalized to that of the reference transcript gapA leading to a normalized relative quantity (NRQ) calculated as follows: NRQ = RQtarget transcript/RQgapA.

Table 1. Sequence of the forward (F) and the reverse (R) primers used in quantitative PCR
GenesNCBI accession numberSequence 5′-3′Annealing T (°C)
acrB NC_000913.2 F: GAAGAGCACGCACCACTACAC55
R: GCAGACGCACGAACAGATAGG
acrF NC_000913.2 F: CGACTGGAACGCTACAAC55
R: CGGCACTACCAACATAACC
yhiV NC_007946.1 F: GGCTATCATCCTCGTCTTCC54
R: GCGTCATCCACCAGTAACC
norE NC_004431.1 F: TCGCAGGACATCAGATTG55
R: CAGACACCCACCATAAGC
mdfA NC_010498.1 F: TTTATGCTTTCGGTATTGGT52
R: GAGATTAAACAGTCCGTTGC
ompC NC_000913.2 F: GTTGATGTTGGTGCTACCTACTAC51
R: ATGCCAGCGTCACGAGTG
ompF NC_000913.2 F: TCTTCGTTGGTCGTGTTG53
R: AAGCCTTCGTATTCGTAGC
gapA NC_000913.2 F: GGACGAAGTTGGTGTTGAC54
R: TTCTGAGTAGCGGTAGTAGC

Statistical analyses

Statistical analyses were performed using GraphPad Prism software (version 5·0; GraphPad Software Inc., La Jolla, CA, USA). The MIC (in the presence or absence of the PAβN) and the MDF data were analysed for the effect of antimicrobial agent using the Friedman test. When the effect was significant (< 0·05), MICs50 (or medians for MDF data) were compared using Dunn's multiple comparison test. The MDF data for each antibiotic were also analysed using the Wilcoxon signed-ranked test with ‘1’ as theoretical median. For NRQ data, first, the Gaussian distribution of values for each susceptibility group (CIP-R and CIP-S) was evaluated using the Shapiro–Wilk test. Second, the variances were compared using the Fisher test. For the Gaussian distribution and variance comparison, when > 0·1, the NRQ data were analysed for the effect of susceptibility using the Student's t-test with the one-tailed option.

Results

Results of the antimicrobial susceptibility are summarized in Table 2. Isolates had MICs against ciprofloxacin ranging from 4 to 32 μg ml−1, against enrofloxacin ranging from 8 to 64 μg ml−1 and against marbofloxacin ranging from 4 to 32 μg ml−1. However, susceptibility differed significantly between antimicrobial agents. Compared with those for enrofloxacin, MIC50 values were lower for ciprofloxacin (< 0·01) and marbofloxacin (< 0·01). Treatment with the efflux pump inhibitor PAβN modified susceptibility. Indeed, in the presence of PAβN, the 17 CIP-R E. coli isolates showed MICs against ciprofloxacin ranging from 2 to 8 μg ml−1, MICs against enrofloxacin ranging from 1 to 4 μg ml−1, and MICs against marbofloxacin ranging from 1 to 4 μg ml−1. In CIP-R E. coli isolates treated with the PAβN, susceptibilities differed significantly between antimicrobial agents. MIC50 values with PAβN treatment decreased for enrofloxacin and marbofloxacin (< 0·01), compared with ciprofloxacin.

Table 2. Effects of PAβN on the fluoroquinolone MIC of 17 ciprofloxacin-resistant Escherichia coli isolatesa
Antimicrobial agentsMIC (μg ml−1)MDF
PAβN (100 μg ml−1)
+
  1. MDF, MIC decrease factor; CIP, ciprofloxacin; ENR, enrofloxacin; MAR, marbofloxacin.

  2. a

    Values presented are medians with minimum and maximum values in parenthesis.

  3. †,‡ Within a column, medians without a common superscript symbol differ (< 0·01); statistical tests used were the Friedman test followed by the Dunn's multiple comparison test.

CIP8 (4–32)4 (2–8)2 (1–8)
ENR16 (8–64)2 (1–4)16 (4–32)
MAR8 (4–32)2 (1–4)4 (2–16)

For each CIP-R E. coli isolate, an MDF was calculated for each fluoroquinolone. This value represented a quantitative variation of the susceptibility for fluoroquinolone after treatment with the PAβN in CIP-R E. coli. Compared with the theoretical median, the MDF increased significantly (< 0·05) irrespective of the antimicrobial agent (Table 2). Results indicated median values of 2 for ciprofloxacin (MDF ranging from 1 to 8, only one bacterial isolate having a value of 1), 16 for enrofloxacin (MDF ranging from 4 to 32) and 4 for marbofloxacin (MDF ranging from 2 to 16). Moreover, there were significant differences in MDF data between antimicrobial agents. MDF medians were lower for ciprofloxacin (< 0·001) and marbofloxacin (< 0·01) as compared to that of enrofloxacin.

The NRQs of different genes encoding efflux pumps and porins were evaluated for 17 CIP-S and 17 CIP-R E. coli isolates. Results are summarized in Table 3. Firstly, acrB expression levels increased significantly in CIP-R E. coli isolates. Compared with CIP-S E. coli isolates, acrB expression levels were on average 26% (< 0·05) greater. In contrast, the mdfA expression level was 27% (< 0·05) lower in CIP-R E. coli isolates when compared to that of CIP-S E. coli isolates. Secondly, porin expression levels were affected by the susceptibility profile. Indeed, compared with CIP-S E. coli isolates, the expression level of ompF was 50% (< 0·05) lower in CIP-R E. coli isolates, whereas that of ompC tended to be 30% (< 0·1) lower.

Table 3. Normalized relative quantities of the acrB, acrF, yhiV, mdfA, norE, ompC and ompF transcripts in ciprofloxacin-susceptible and resistant Escherichia coli isolatesa
GeneSusceptibleResistantP-value
  1. For each strain, relative quantity of the target transcript was normalized to that of the reference transcript gapA.

  2. a

    Relative values presented are means ± SE (= 17).

  3. †, ‡ Within a row, means with a superscript symbol is (; < 0·05) or tend to be (, < 0·1) different from the susceptible group; the statistical test used was the Student's t-test with the one-tailed option.

acrB 0·65 ± 0·040·82 ± 0·070·019
acrF 0·81 ± 0·100·92 ± 0·11>0·1
yhiV 0·51 ± 0·040·58 ± 0·06>0·1
mdfA 4·91 ± 0·423·60 ± 0·290·011
norE 5·99 ± 0·547·14 ± 0·76>0·1
ompC 1·68 ± 0·311·17 ± 0·220·098
ompF 2·60 ± 0·421·29 ± 0·320·014

Discussion

Concerning CIP-R E. coli isolates, the MIC for enrofloxacin was higher than those for ciprofloxacin and marbofloxacin. Several studies have previously described different susceptibilities in E. coli depending on the fluoroquinolone (Tavio et al. 1999, 2010; Morgan-Linnell et al. 2009; Gibson et al. 2010). For a given strain, this variability could be explained by the differential affinity of fluoroquinolones to the drug target enzymes or to the OMPs and efflux pumps involved. Therefore, our results suggest a higher affinity of hydrophobic fluoroquinolones, such as enrofloxacin, for proteins involved in the decreased susceptibility of the bacterial isolates in this study. In the current study, after treatment with PAβN, an inhibitor active against RND efflux pumps (Poole 2005), MIC decreased significantly irrespective of the fluoroquinolone. This contribution of RND efflux pumps in the susceptibility against fluoroquinolones has also been described previously (Mazzariol et al. 2000; Saenz et al. 2004; Kern et al. 2006). However, susceptibilities in E. coli isolates after PAβN treatment differed between antimicrobial agents, resulting in a higher MDF for enrofloxacin when compared to those for marbofloxacin and ciprofloxacin. Therefore, this result suggested a greater ability of the RND efflux pumps to expel hydrophobic fluoroquinolones, such enrofloxacin, than hydrophilic fluoroquinolones, such as marbofloxacin or ciprofloxacin (Teresa Tejedor et al. 2003). A previous study also reported a higher MDF for enrofloxacin than pradofloxacin after PAβN treatment in resistant E. coli (Gibson et al. 2010) with the hydrophobicity of enrofloxacin being higher than pradofloxacin. Regarding these data, we suggest that the contribution of RND efflux pumps to the decreased fluoroquinolone susceptibility was greater for hydrophobic fluoroquinolones such as enrofloxacin due to a higher affinity for these proteins. This differential affinity between antimicrobial agents was also described for NorE which is limited to the more hydrophilic fluoroquinolones such as ciprofloxacin and norfloxacin (Morita et al. 1998; Poole 2000; Yang et al. 2003).

As decreased cellular drug accumulation is a complex phenomenon involving diverse mechanisms including in part RND efflux pumps, the expression level of genes encoding the efflux pumps and porins involved in decreased fluoroquinolone accumulation was also investigated. Our results indicated a differential expression level of these genes in CIP-R E. coli isolates. Previous studies have described differential amounts of the multidrug efflux pump AcrB correlating with the susceptibility of strains, but this was demonstrated using semi-quantitative methods such as Western blot and Northern blot (Jellen-Ritter and Kern 2001; Fabrega et al. 2010; Tavio et al. 2010). In the current study, the expression level of acrB increased in CIP-R E. coli isolates in comparison with CIP-S E. coli isolates. This was in agreement with a recent study which described a strong correlation between the expression level of acrB and fluoroquinolone susceptibility(Swick et al. 2011). However, in this study, we observed no differences in the expression of acrF and yhiV, two other proteins of the RND family of multidrug efflux systems, between CIP-S and CIP-R E. coli isolates. These results support the idea that AcrB is the predominant RND efflux system leading to E. coli fluoroquinolone resistance as described previously (Poole 2000). Our results in CIP-R E. coli isolates also describe a decreased level of expression of mdfA, encoding a member of the MFS family of multidrug efflux systems. This was in contrast with previous reports suggesting a significant relationship between the overexpression of mdfA and susceptibility to sitafloxacin (Yasufuku et al. 2011), or a decreased susceptibility against norfloxacin after transfection of the MdfA encoding gene, with the susceptibility to other fluoroquinolones being unaffected (Yang et al. 2003). The limited efficiency of MdfA to enhance fluoroquinolone resistance is supported by experiments reporting this protein to be a more effective pump for nonantibiotics (Edgar and Bibi 1997; Poole 2000). In the light of the conflicting data, further investigation is required to clarify the role of these efflux systems in fluoroquinolone resistance. For NorE, a member of the MATE family multidrug efflux systems, there are no significant differences in gene expression levels in CIP-R E. coli isolates when compared to CIP-S E. coli isolates. This result is supported by a previous study that described no correlation between the level of expression of norE with the fluoroquinolone susceptibility (Swick et al. 2011). While studies have reported that NorE is capable of exporting fluoroquinolones (Morita et al. 1998; Yang et al. 2003), in these studies efflux and antibiotic resistance were assessed using genes cloned on plasmids and over-expressed in E. coli. Thus, the relevance of the chromosomal counterparts to fluoroquinolone resistance is uncertain (Poole 2000). Concerning porins, our results demonstrate decreased expression levels of ompF and ompC in CIP-R E. coli isolates when compared to CIP-S E. coli isolates, with the decrease being greater for ompF than ompC. Previous studies had reported lower amounts of OmpF in strains with greater MIC, although these data were obtained using semi-quantitative methods (Cohen et al. 1989; Fabrega et al. 2010; Tavio et al. 2010). However, much less information is available concerning OmpC. One study described a moderate decrease in the amount of OmpC in resistant E. coli, but a greater decrease in OmpF (Tavio et al. 1999). Therefore, in accordance with the previously mentioned studies, our results support findings that OmpF was the predominant porin mediating the decreased fluoroquinolone accumulation in E. coli, and OmpC was less involved.

In the current study, we demonstrate a higher contribution of RND efflux pumps to the decreased susceptibility against enrofloxacin compared with that of marbofloxacin or ciprofloxacin. This could be explained by a stronger affinity of RND pumps for hydrophobic fluoroquinolones such as enrofloxacin. Moreover, the expression profile of genes encoding the different efflux pumps revealed that only acrB was upregulated in CIP-R E. coli isolates. Therefore, among the different existing efflux pumps, only AcrB contributed to decreased susceptibility for fluoroquinolones in these CIP-R E. coli isolates. This could therefore explain the higher MIC for enrofloxacin when compared with ciprofloxacin, as this RND type efflux pump has a stronger affinity for hydrophobic fluoroquinolones. However, considering that resistance mechanisms are being more complex, the present data need further confirmation, in particular other possibilities for reduced drug accumulation need to be excluded. While the present study focused only on efflux, variable resistance for different fluoroquinolones could also be explained by a different affinity to the target proteins, combined with the expression profile of efflux pumps in resistant E. coli.

Acknowledgements

We would like to thank the National Reference Laboratory of Fougeres for providing E. coli isolates. The authors declare that there are no conflicts of interest.

Ancillary