Molecular typing of fluoroquinolone-resistant and fluoroquinolone-susceptible Escherichia coli isolated from blood of neutropenic cancer patients in a single center

INTRODUCTION

Fluoroquinolones are potent antimicrobial agents used in the treatment and prophylaxis of community-acquired and nosocomial infections [1]. When the new fluorinated quinolones were introduced more than 10 years ago, optimism concerning their usefulness in the prophylaxis of febrile neutropenia was justified by their broad antibacterial spectrum, high intraluminal concentration, systemic bactericidal activity, good tolerability, and apparent lack of potential for development of resistance [1]. All these characteristics contribute to their widespread use for antibacterial chemoprophylaxis in neutropenic patients [2, 3]. Although the risk of development of resistance was considered to be extremely low [4], after their introduction to clinical practice an increasing emergence of quinolone-resistant bacteria was documented [5–7], particularly in Staphylococcus aureus and Pseudomonas aeruginosa [8, 9]. The emergence of high-level fluoroquinolone-resistant (FQ-R) strains (MIC >128 mg/L) has also been reported for various members of the Enterobacteriaceae family, including Escherichia coli [10–12]. A clear association was observed between increased use of quinolones and emergence of fluoroquinolone resistance. Neutropenic patients with cancer who receive prophylaxis with fluoroquinolones may be at risk of developing E. coli bacteremia due to FQ-R strains [13], and resistant E. coli isolates from the hematologic ward in Rome, Italy appeared after the introduction of fluoroquinolone prophylaxis for neutropenia, increasing from 5% to 78% between 1990 and 1996. The increasing rate of breakthrough bacteremias with resistant E. coli seems to be due to the independent emergence of several new clones [14]. However, nosocomial transmission may contribute to these resistant strains becoming a cause of concern [15].

Phenotypic characterization of E. coli isolates has proved to be inadequate for identifying strain diversity or similarity, which thus requires genotypic analysis. Various different typing methods have been used to study the genotypes of bacteria involved in nosocomial infections. Pulsed-field gel electrophoresis (PFGE) is the method of reference, although it is costly, time-consuming, and requires expensive instrumentation and trained personnel [16]. PCR typing methods, e.g. random amplified polymorphic DNA (RAPD) analysis [17–19], are rapid and less expensive, but are not standardized because there is no consensus regarding the conditions, instruments, enzymes and primers to be used [20, 21].

The aim of this study was to investigate the molecular epidemiology of FQ-R and fluoroquinolone-susceptible (FQ-S) isolates of E. coli causing bacteremia in neutropenic patients with two different typing methods, i.e. PFGE versus RAPD analysis.

Materials and Methods

PFGE

E. coli isolates were grown overnight in 10 mL of brain–heart infusion broth (BHI, Oxoid). After incubation, bacteria were harvested, washed twice in PIV buffer (1 M NaCl. 10 mM Tris-HCl. pH 7.6) and resuspended in PIV buffer to achieve an extinction coefficient of 0.9 at a wavelength of 550 nm. A 0.5-mL aliquot of this bacterial suspension was quickly mixed with 0.5 mL of 1.5% (w/v) Incert agarose (FMC, Rockland, ME, USA). After agarose plug casting, samples were transferred to 5 mL of lysing solution (6 mM Tris-HCl (pH 7.6), 1 M NaCl, 100 mM EDTA (pH 7.5), 0.2% deoxycholate. 0.5% sodium lauroyl sarcosine, 20 μg/mL of RNase. 1 mg/mL lysozyme) and incubated overnight for 24 h. The plugs were transferred to ESP solution (0.5 M EDTA (pH 9), 1% sodium lauroyl sarcosine, 100 μg/mL of proteinase K) and incubated for 48 h at 50°C. The plugs were washed twice with phenylmethylsulfonyl fluoride (PMSF) 1 mM in TE buffer (10 mM Tris-HCl, 1 mM EDTA (pH 7.5)) and three times with TE buffer. Xbal (New England BioLabs, Beverly, MA, USA) was used for cleavage. Plugs were equilibrated in 250 μL of restriction buffer plus spermidine (1 mM), and cleavage was performed overnight with 40 U of enzyme.

Electrophoresis was performed in 1% (w/v) FastLane agarose (FMC), at 14°C in TBE buffer (89 mM Tris-borate, 2 mM EDTA (pH 8.5)). PFGE was performed in a Chef Mapper apparatus (Bio-Rad Laboratories, Hercules, CA, USA). The pulse time was ramped from 1 to 50 s over 24 h at 200 V. Lambda concatemers (Bio-Rad) were used as DNA size markers. Gels were then stained with ethidium bromide and acquired digitally (Gel Doc 1000, Bio-Rad). Restriction fragment patterns were analyzed with MA Fingerprinting software (Bio-Rad). Similarity coefficient (S_AB) values were calculated directly by the software. Since band position alone was used, S_AB values ranged from 0 to 1.0, where 0 indicates that the fingerprints had no bands in common and 1.0 indicates that the fingerprints were identical. The unwarping of gels requires subjective input from the investigator, so pairs of strains with S_AB values ≥0.9 were regarded as highly similar and not distinguishable by PFGE. The cut-off value of 0.9 was based on previous experience of using the software to compare profiles of the same run on different gels with various degrees of gel warping [24, 25].

RESULTS

The ciprofloxacin MICs for the E. coli isolates included in the study are shown in Table 2. PFGE typed 17 of the 19 isolates of FQ-R E. coli examined. Two isolates (L444 and L62) were not typable, showing only a diffuse DNA band of between 150 and 350 kb in size. Fifteen of the 17 typable isolates were unrelated. Two isolates (L454 and L456), obtained from different patients in the same period (February 1996), were indistinguishable (S_AB=0.9). Two other isolates (L177 and L347), obtained from the same patient during two episodes of bacteremia occurring 1 year apart, were different (S_AB=0.25).

All 27 FQ-S isolates were typable by PFGE and were unrelated. No relationship was found between FQ-S and FQ-R isolates. All FQ-R E. coli isolates were typable by RAPD analysis using the primer 1247, and five pairs and one triad of indistinguishable isolates were identified (L449–L177, L412–L457, L444–L445, L281–L428, L454–L456, L62–L98–L194).

All FQ-R strains were also typable with primer 1283, which was able to identify four pairs of indistinguishable isolates (L454–L456, L449–L177, L445–L444, L281–L428). These four pairs were identical to four of the five pairs identified with primer 1247. One of these pairs (L454–L456) was also found to be indistinguishable by PFGE analysis. The two isolates that were not typable with PFGE (L444 and L62) were successfully typed by RAPD analysis and were unrelated.

RAPD analysis with primer 1247 identified only one indistinguishable pair of FQ-S isolates (L68–L61), while primer 1283 identified three pairs and one triad of indistinguishable isolates (L426–L433, L59–L61, L200–L198, L79–L76–L132). No pairs or triads of indistinguishable FQ-S isolates were common to both primers.

The discriminatory index was 0.999 for PFGE, 0.992 for RAPD with primer 1247, and 0.991 for RAPD with primer 1283.

DISCUSSION

According to the PFGE results, horizontal spread of a single clone of FQ-R E. coli in the hematologic ward in Rome appears to be a rare event and does not represent a cause of concern in terms of infection control. Furthermore, no genotype correlation was found among FQ-S isolates and between FQ-R and FQ-S isolates. Horizontal transmission has, however, been described previously for several multiresistant enteric bacteria [28, 29]. It is noteworthy that horizontal spread of resistant clones of E. coli was reported in the Cancer Center of Ulm [15]. However, reports suggest that resistance to quinolones in E. coli is mainly of multiclonal origin [14]. Oethinger et al evaluated FQ-R isolates of E. coli from several centers in Europe and the Middle East and found no correlation between strains provided by different hospitals, supporting the lack of horizontal transmission [14]. In the present study, PFGE analysis was able to distinguish all FQ-S and most (17/19, 89%) FQ-R E. coli isolates. Genotyping by PFGE is a powerful tool for studying the genetic diversity of E. coli strains and, as in the Ulm experience, a similar percentage of FQ-R strains were typable (13/16, 81%) [14].

RAPD analysis typed two FQ-R isolates that were untypable by PFGE, and so was a more sensitive technique. This could be explained by the fact that, although degraded, DNA from these two isolates was still available as a template for the primers used in RAPD analysis. The results obtained with RAPD using different primers yielded different results (four pairs of isolates found to be indistinguishable, but seven pairs or triads identified as different with the two primers). Many factors can affect the reproducibility of RAPD analysis, including Mg²⁺ concentration [32], batch-to-batch variation in primer synthesis [31], ratio of DNA template concentration to primer concentration, model of thermocycler used [20], and the supplier and concentration of Taq DNA polymerase [32], Furthermore, RAPD is a combination of artifactual variations due to the low annealing temperature and true polymorphism [20]. In our experience. RAPD analysis was less discriminatory than PFGE for typing both FQ-R and FQ-S E. coli isolates. RAPD analysis was unable to distinguish between isolates found to be different by PFGE. In a previous report, some FQ-R isolates with identical RAPD genotypes had different PFGE genotypes [14].

A possible explanation for the different discriminatory power of the two genotyping methods may be that PFGE examines about 80% of the genomic DNA. while RAPD is able to analyze only 0.1–0.2% of DNA, as calculated by the ratio between the sum of the sizes of all bands generated by Xbal and the size of the E. coli chromosome for PFGE, and the sum of sizes of bands generated by RAPD and the size of the E. coli chromosome (G. Cardinali, personal communication).

As observed in this study, the discriminatory index (D) of Hunter and Gaston [29] is unable to discriminate between genotyping methods that group few strains in many types, such as the new genotyping methods. This characteristic reduces the capability of this index to evaluate the best method for typing a given species. However, with the limited number of isolates examined in this study, RAPD analysis had a more effective typability, but was less discriminative than PFGE for characterizing bacteremic E. coli isolates.