Broiler Chickens as Source of Human Fluoroquinolone-Resistant Escherichia coli, Iceland

To investigate feed as a source for fluoroquinolone-resistant Escherichia coli in broiler chickens, we compared antimicrobial drug–resistant E. coli from broiler feed and broilers with ciprofloxacin-resistant human clinical isolates by using pulsed-field gel electrophoresis. Feed was implicated as a source for ciprofloxacin-resistant broiler-derived E. coli and broilers as a source for ciprofloxacin-resistant human-derived E. coli.

I n a previous study, we found a relatively high prevalence of antimicrobial and especially quinolone resistance among Escherichia coli isolates from broiler chickens and broiler meat (1), despite no known antimicrobial drug selection pressure in chicken farming in Iceland and biosecurity measures to prevent transmission of infectious agents into farms. Broiler houses are cleaned and disinfected after broiler fl ocks are transported to slaughter. Therefore, resistant bacteria are unlikely to persist in the broiler houses. However, animal feed can be contaminated with antimicrobial drug-resistant E. coli (2).
The high prevalence of quinolone-resistant E. coli isolates obtained from broilers and broiler meat coincides with an increasing prevalence of fl uoroquinolone resistance among human clinical E. coli isolates in Iceland. This increase correlated with increased use of fl uoroquinolones in clinical settings (3).
We examined whether the prevalence of resistant E. coli strains in broilers had changed since our previous study and whether broiler feed could be a source for the resistant strains. Furthermore, we compared the genotypes of ciprofl oxacin-resistant broiler, broiler meat, and broiler feed E.
coli isolates with ciprofl oxacin-resistant human clinical E. coli isolates.

The Study
The sampling period for this study was May through November 2008. Pooled cecal samples (20 ceca from each fl ock) were taken from the 30 fl ocks slaughtered at all 3 broiler slaughterhouses in Iceland in June 2008. Ceca were stomachered in phosphate-buffered saline, spread on Mac-Conkey agar with and without enrofl oxacin (0.25 mg/L), and incubated overnight. Feed was sampled from feed stalls at 18 farms (of which 14 had participated in the previous study) and from 2 feed mills; the feed was suspended in buffered peptone water, mixed, incubated overnight, and spread on MacConkey agar as described above. One colony from each agar plate was selected for susceptibility testing as described in our previous study (1). Because isolates were collected from the 18 largest broiler farms (of the 27 farms operating in Iceland), all the broiler slaughterhouses, and the only 2 feed mills operating in Iceland, we believe this study provides a representative sample.
We selected all 34 available human E. coli isolates recovered from routine clinical specimens (mostly urine and blood) at the main clinical microbiology/reference laboratory in Iceland (Landspitali University Hospital) during 2006-2007, which had similar susceptibility patterns to the strains previously isolated from broilers (1) (ampicillintetracycline-sulfamethoxazole/trimethoprim-ciprofl oxacin or ciprofl oxacin alone). Only 1 isolate was chosen from each patient.
We compared all E. coli broiler ceca and feed isolates resistant to >1 antimicrobial agents and the 34 ciprofl oxacin-resistant human E. coli isolates with resistant E. coli isolates from the previous study (2005)(2006)(2007) (1) by pulsed-fi eld gel electrophoresis (PFGE) using a slightly modifi ed method of Ribot et al. (6). Comparison of PFGE patterns was made by visual inspection and BioNumerics software (Applied Maths, Sint-Martens-Latem, Belgium). For cluster analysis, the Dice coeffi cient for band matching (band-position tolerance 1.5%) was used to generate an unweighted pair group method with arithmetic averages dendrogram. Isolates from the previous and present study that did not yield a satisfactory banding pattern by PFGE were genotyped by randomly amplifi ed polymorphic DNA (RAPD) analysis as described (7). Reaction products were analyzed by electrophoresis on 1.5% agarose gels stained with ethidium bromide. Patterns were considered different when the profi les differed by at least 1 band. Similarity among RAPD patterns was compared as described for PFGE. Clusters for PFGE and composite RAPD profi les were defi ned as >2 isolates with >80% similarity. Prevalence values were compared by using the Fisher exact test.
Of the 40 broiler isolates, 20 were resistant to >1 of the antimicrobial drugs tested (Table); only 1 was multidrug resistant (resistant to streptomycin, tetracyclin, sulfamethoxazole, and trimethoprim). Ciprofl oxacin and nalidixic acid were always cross-resistant. Compared with the previous sampling, the prevalence of resistance increased significantly for ciprofl oxacin and nalidixic acid (from 18.2% to 42.5%; p<0.0001) but decreased signifi cantly for ampicillin (from 18.2% to 0.0%; p = 0.002) and sulfamethoxazole (from 19.1% to 5.0%; p = 0.0398). This fi nding suggests that quinolone resistance was not transferred with resistance to the other antimicrobial agents and that it was selected for by other factors. Of the 22 E. coli isolates obtained from feed, 7 (32%) were resistant to ciprofl oxacin and nalidixic acid, and all were susceptible to the other antimicrobial agents tested. Although no E. coli were isolated from the 2 feed mill samples, other Enterobacteriaceae grew on the agar plates, which could have overgrown existing E. coli strains, if any, demonstrating that the feed was not sterile.
The 27 resistant broiler and feed isolates were compared with 76 resistant isolates analyzed in our previous study (1) along with the 34 ciprofl oxacin-resistant human E. coli isolates. Of 137 broiler, broiler meat, feed and human isolates, 110 (80%) yielded interpretable, reproducible PFGE patterns. We detected 92 profi les, of which 81 (88%) were represented by a single isolate. Isolates of different origin were intermixed forming 26 clusters, of which 12 were seen in the previous study. Of the 14 new clusters, 10 contained isolates of different origins (Figure). Human isolates clustered with broiler (2005)(2006), broiler meat, broiler (2008), and feed isolates in 6 clusters (Figure). This supports previous fi ndings of chickens and their products as a possible source of antimicrobial drug-resistant E. coli in humans (8,9). With the extensive genomic diversity of E. coli and the discriminative power of PFGE typing, fi nding indistinguishable isolates of different origin collected over several years is unlikely, except from a large collection (8,10). Therefore, fi nding human isolates closely related (>80 similarity) to broiler, broiler meat, and feed isolates suggests an epidemiologic link between the populations. Additionally, we found closely related isolates from feed and broiler (samples from 2008 and 2005-2006) from geographically distant farms, supporting previous fi ndings that antimicrobial drug-resistant E. coli could be introduced into the farm environment through broiler feed (2).

Conclusions
Prevalence of fl uoroquinolone-resistant E. coli remains moderately high in broilers, but resistance to other antimicrobial drugs is decreasing. Fluoroquinolone-resistant E. coli isolated from broiler feed implicates feed as the source  of resistant strains into farms. Resistant isolates from feed, broilers, broiler meat, and humans were closely related, demonstrating that poultry and their food products can be a source of resistant E. coli in humans.