Introduction

Bacterial diseases in pigs reared on production farms are responsible for high morbidity and mortality rates and subsequently also for increased economic losses [1, 2]. The main indications for antibiotic therapy in pigs in Switzerland are gastrointestinal and respiratory diseases [3]. Therapy may include (fluoro)quinolones, which are categorized by the World Health Organization (WHO) as critically important antimicrobial agents (CIAs) [4]. Despite the worldwide growth of the pig-farming industry, monitoring systems on the use of antibiotics in veterinary medicine have led to reduced antibiotic use in many European countries, including Germany, Denmark, The Netherlands, and Sweden [5].

In pig husbandry, the usual way of drug application is oral, via the feed mixture. This is a practical way of drug application from a farmer’s point of view, but the drawback is the release of antimicrobial substances into the farm environment [6]. Data from several studies suggest that there is a link between the agricultural use of antibiotics and the prevalence of antibiotic-resistant bacteria in the pig farm environment, including (fluoro)quinolone resistance [7,8,9,10,11].

The pathways of antibiotic residues into the environment after animal treatment are numerous. Moreover, certain antibiotics such as fluoroquinolones and tetracyclines are not fully metabolized in pigs and their residues may be detected in dust, manure, sewage, soil, ground- and surface water and crops [6, 12,13,14,15]. These various antibiotic residue reservoirs are perfect breeding grounds for resistant bacteria, including (fluoro)quinolone resistant E. coli [16]. During a previous study aimed at analysing the use of fluoroquinolones in Swiss pig farms (von Ah, et al., manuscript under review), quinolone non-susceptible E. coli were isolated from environmental samples (dust, liquid manure and wipe samples of bay walls) of the farms. The goal of this study was to characterize these isolates with regard to the two main mechanisms of (fluoro)quinolone resistance in Enterobacteriaceae, i.e., the accumulation of mutations in the Quinolone Resistance Determining Region (QRDR) of gyrA encoding DNA gyrase, and the acquisition of plasmid-mediated quinolone resistance (PMQR) genes [11]. A further aim was to characterize the strains by phylogenetic grouping and multilocus sequence typing (MLST).

Material and methods

Strains and data collection

In this study, we analysed 174 E. coli isolates collected during 2016 by the Division of Swine Medicine of the Vetsuisse Faculty Zurich (von Ah, et al., manuscript under review). The collection consisted of strains obtained using both qualitative and semi quantitative methods to isolate quinolone resistant E. coli from dust, wipe and slurry samples from farm environments. For the qualitative method, an average of 1.91 g of sample was diluted 1:10 in Enterobacteriaceae Enrichment (EE) broth (Becton, Dickinson, Heidelberg, Germany) and incubated at 37 °C overnight. The enrichment was then spread on RAPID’E. coli agar (Biorad, Munich, Germany), supplemented with 8 μg/ml nalidixic acid, and incubated overnight at 37 °C. Using the semi-quantitative approach, an average of 1.97 g of sample, were diluted in a ratio of 1:10 in 0.85% saline solution and homogenized in a Stomacher sample blender (Seward Medical Ltd., London, UK). The homogenate was spread in dilution steps of 1:100 and 1:1000, respectively, on RAPID’E. coli agar plates supplemented with 8 μg/ml nalidixic acid. One E. coli isolate randomly selected from each positive sample was collected for further analysis. Isolates with questionable identity on RAPID’E. coli agar were confirmed by matrix-assisted laser desorption/ionization time-of- flight mass spectrometry (MALDI-TOF–MS, Bruker Daltronics, Bremen, Germany).

In total, the strains originated from dust (n = 48), wipes (n = 56), and slurry (n = 70), collected from 55 different farms (24 farrowing and rearing farms, 23 fattening farms and 8 mating and gestation farms) located in central and north-eastern Switzerland. All farms were part of a sow pool system. Of the 55 farms, 23 (41.8%) reported use of fluoroquinolones during the study period of 2016, including 16 (66.7%) of the farrowing and rearing, four (17.4%) of the fattening, and three (37.5%) of the mating and gestation farms.

Microbiological methods

Antimicrobial susceptibility testing

The antimicrobial resistance profiles of the isolates were determined using the disk diffusion (Kirby Bauer) method according to the Clinical and Laboratory Standards Institute (CLSI) performance standards and breakpoints for human clinical isolates [17]. Mueller-Hinton agar culture medium (Becton Dickinson, Allschwil, Switzerland) was inoculated with a saline suspension of isolated colonies adjusted to 0.5 McFarland turbidity standard. Antibiotic disks (Becton Dickinson and company, Sparks, MD USA) were used, containing 30 μg nalidixic acid (NA), 5 μg ciprofloxacin (CIP), 10 μg ampicillin (AM), 20 μg/10 μg amoxicillin/clavulanic acid (AMC), 30 μg cefazolin (CZ), 30 μg cefotaxime (CTX), 30 μg cefepime (FEP), 23.75 μg /1.25 μg sulfamethoxazole/trimethoprim (SXT), 30 μg chloramphenicol (C), 15 μg azithromycin (AZM), 30 μg tetracycline (TE), 10 μg streptomycin (STR), 30 μg kanamycin (K), 10 μg gentamicin (GM), 300 μg nitrofurantoin (FM), and 200 μg fosfomycin (FOS). After 18 h of incubation at 35 °C ± 2 °C, results were interpreted as either sensitive (S), intermediate (IR), or resistant (R) according to the zone diameters around the disks using CLSI breakpoints [17]. Results were confirmed to be within the quality control ranges described by CLSI for E. coli ATCC25922 [17].

Isolates displaying resistance to three or more classes of antimicrobials (counting β-lactams as one class) were defined as multidrug-resistant (MDR).

Molecular methods

Analysis of the quinolone resistance-determining region (QRDR) in gyrA

All strains were examined for mutations in the quinolone resistance-determining regions (QRDRs) of gyrA, using PCR amplification and sequencing primers as described previously [18]. Synthesis of primers and DNA custom sequencing was carried out by Microsynth (Balgach, Switzerland). Nucleotide sequences were analyzed with CLC Main Workbench 8.0.1 and aligned with the sequence gyrA reference strain E. coli K-12, substrain MG1655 (GenBank: U00096). For database searches the BLASTN program of the National Centre for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/blast/) was used.

Detection of plasmid mediated quinolone resistance genes (PMQR)

The plasmid-mediated fluoroquinolone resistance genes aac(6′)-Ib-cr, qnrA, qnrB, qnrC, qnrD, qnrS, and qepA were detected by PCR as described elsewhere [19,20,21,22,23,24,25]. Synthesis of primers and DNA custom sequencing was carried out by Microsynth (Balgach, Switzerland) and nucleotide sequences were analyzed with CLC Main Workbench 8.0.1. In addition, qnrB and qnrS genes were sequenced.

The presence of qnrB was confirmed by PCR as described by Abgottspon et al., using strain N05–2379 as a positive control [26]. Purified amplicons were custom sequenced (Microsynth, Balgach, CH) using the forward primer (qnrB_Seq_F) [26] . Sequences were analysed using the Basic Local Alignment Search Tool (BLAST) of the NCBI (https://blast.ncbi.nlm.nih.gov/Blast.cgi) (Bethesda, USA).

The presence of qnrS confirmed and sequenced as described by Zurfluh et al., using E. coli OW95E1 as a positive control and the forward primer (qnrS_orf_F) [27]. Sequences were analysed as described for qnrB.

Phylogenetic characterization and multilocus sequence typing

Phylogenetic classification of the E. coli isolates into one of the eight groups A, B1, B2, C, D, E, F (E. coli sensu stricto), or Escherichia clade I, was performed as described by Clermont et al. [28].

Sequence type (ST) determination of the E. coli isolates was carried out as described by Wirth et al. [29]. Sequences were imported into the E. coli multilocus sequence type (MLST) database (http://enterobase.warwick.ac.uk) to determine MLST types and clonal complexes (CC).

Serotyping of E. coli ST301

Strains belonging to CC165 and ST301 frequently belong to the unusual O80 serogroup [30]. To test this possibility, all isolates belonging to ST301 were serotyped. The O80 serogroup was identified by O80-specific PCR using primers and conditions described previously [30]. The H2 type was determined by PCR targeting the flicH2 gene with primers described elsewhere [31]. The presence of the intestinal virulence genes stx, eae and of extra-intestinal virulence genes associated with plasmid pS88 [32] was evaluated as described previously [33].

Results

Antimicrobial resistance phenotypes

Antimicrobial susceptibility testing by the disc diffusion method showed that 81% (n = 141) of the strains were resistant and 19% (n = 33) were intermediately resistant to nalidixic acid (Table 1). Furthermore, 36.2% (n = 63) of the isolates were also resistant to ciprofloxacin (Table 1).

Table 1 Escherichia coli isolated from dust, wipes and slurry samples from pig farms
Full size table

Additional antimicrobial resistance was most frequently observed for streptomycin (72.4% /n = 126), tetracycline (60.9% /n = 106), sulfamethoxazole/trimethoprim (50% /n = 87), ampicillin (46.6% /n = 81), kanamycin (19.5% /n = 34), chloramphenicol (15.5% /n = 27), and gentamicin (14.4% /n = 25), respectively (Table 1). Resistance to all other tested antibiotics was detected for at least one isolate, except to nitrofurantoin (Table 1).

Of the 174 isolates analysed in this study, 68.4% (n = 119) were resistant to three or more classes of antibiotics and therefore categorised as MDR. The most frequent MDR combinations detected were SXT-TE-STR (n = 15), AM-SXT-TE-STR (n = 10) and AM-SXT-STR-K (n = 8) (Table 1). E. coli strains resistant to four and five antibiotics were the most prevalent (21.3 and 19.0%, respectively).

Molecular properties

Of 141 isolates with a nalidixic acid resistant phenotype, 98.6% (n = 139) possessed at least one nucleotide mutation in the QRDR of gyrA. Thereof, 49.6% (n = 70) showed single amino acid substitution at codon Ser83, namely Ser83 to Leu (n = 67), or Asp87 to Tyr (n = 2), or Asp87 to Gly (n = 1). Further, 48.9% (n = 69) possessed double substitutions at Ser83 to Leu and Asp87 to Asn (n = 68) or Tyr (n = 1). Two isolates (isolates no. 65 and 106, respectively) tested negative for mutations in the QRDR of gyrA (Table 1).

A total of 38 strains possessed one or more PMQR genes, representing 21.8% of the 174 analysed strains (Table 1). Among the 19.5% (n = 34) of the isolates with one PMQR gene, twenty (11.5%) possessed qnrB, thirteen (7.5%) qnrS and one isolate (0.6%) possessed aac(6′)-Ib, respectively (Table 1). Four isolates (2.3%) possessed a combination of qnrB and qnrS genes. No isolates tested positive for qnrA, qnrC, qnrD or qepA. The occurrence of PMQR positive isolates was remarkably higher in strains exhibiting intermediate resistance to nalidixic acid (90.9% /n = 30), than in nalidixic acid resistant strains (5.7% /n = 8). Moreover, all qnrB/qnrS combinations were detected in intermediately resistant isolates (Table 1). Isolates possessing PMQR were found in 11 (22.9%) of the dust samples 16 (28.6%) of the wipe samples. and 11 (15.7%) of the slurry samples (Table 1).

Of the 23 farms with reported use of fluoroquinolones, 12 (52.2%) yielded environmental E. coli containing PMQR genes. Thereof, the majority (7 farms/58.3%) were farrowing and rearing farms, three (25%) were fattening farms and two (16.7%) were mating and gestation farms (Table 1).

By contrast, of the 32 farms without a history of fluoroquinolone use during the study period, nine (28.1%) tested positive for E. coli harbouring PMQR genes. Thereof, five (55.6%) were fattening farms, four (44.4%) were mating and gestation farms, and none (0%) were farrowing and rearing farms (Table 1).

Phylogenetic grouping

The majority of the isolates were assigned to phylogenetic groups A (48.3%/n = 84) and group B1 (33.3% /n = 58). The remaining strains were classified into group C (9.8% /n = 17), E (6.9%/n = 12), F (1.1%/n = 2) and D (0.6%/n = 1), respectively (Table 1). None of the isolates belonged to phylogenetic group B2.

MLST

Overall, a total of 50 STs were found. The most common sequence types were ST10 (n = 20), ST297 (n = 20), ST453 (n = 10), ST88 (n = 9), ST898 (n = 8), ST93 (n = 6), ST2197 (n = 6), ST737 (n = 5), and ST2509 (n = 5) (Table 1). Seven isolates could not be assigned to any known ST, because the allele combinations were new (Table 1). The allele combinations are listed in Additional file 1: Table A1.

Characteristics of E. coli ST301 isolates

Three isolates belonged to ST301 (Table 1). Thereof, two (isolates 30 and 31, respectively) belonged to serotype O80:H2 and possessed the eae-ξ variant. None of the isolates harboured stx or any genes related to pS88. Both isolates were therefore classified as enteropathogenic E.coli (EPEC).

Discussion

Resistance profiles

In the present study, we determined the prevalence of point mutations within the QRDR of gyrA and the presence of PMQR genes among 174 E. coli isolated from pig farm environments. All isolates were non-susceptible to quinolones and, using the disk diffusion method, were classified as intermediate or as resistant to nalidixic acid.

All resistant isolates except two, possessed at least one mutation in gyrA. Notably, both these isolates possessed a PMQR gene, (isolate 65 possessed qnrB, and isolate 106 qnrS, respectively). Since qnr genes alone are insufficient to confer resistance [34,35,36], further mechanisms are likely associated with the resistance phenotype of these strains, such as mutations in the QRDR regions of the gyrB, parC genes, or increased efflux pump activity (both not evaluated in this study).

Of the 70 nalidixic acid resistant strains possessing one mutation in gyrA, only one (0.7%) was resistant to ciprofloxacin. By contrast, of the isolates with double mutations in gyrA, the majority (89.9%) were also resistant to ciprofloxacin. These observations correlate with previous studies that link the number of quinolone resistance-associated mutations and the resistance phenotype of an isolate [37, 38].

In this study, only a minority of the resistant strains carried PMQR genes, and, except for aac(6′)-Ib-cr, none of the PMQR genes were associated with ciprofloxacin resistance. Isolates with qnrS also possessed a Ser83 to Leu substitution in the gyrA QRDR and showed decreased susceptibility to ciprofloxacin. Two further resistant isolates harbouring qnrS or qnrB lacked mutations in the gyrA QRDR and were susceptible to ciprofloxacin. In short, these data correlate with results from previous studies on the coexistence of different resistance mechanisms in one isolate [37,38,39].

Notably, the vast majority (90.9%) of strains with intermediate resistance to nalidixic acid were associated with the presence of qnrB, qnrS or a combination thereof. The three remaining intermediately resistant isolates lacking both gyrA mutations and aac(6′)-Ib-cr, qnrA, qnrB, qnrC, qnrD, qnrS, or qepA genes, are likely to possess other resistance-mechanisms that we did not screen for in the present study, e.g., point mutations within the QRDR of gyrB or in the topoisomerase genes parC or parE [11]. Four isolates possessed a combination of qnrB and qnrS genes but remained intermediately resistant to nalidixic acid. These results are in agreement with the findings of other studies [40, 41], which demonstrate that the presence of two different qnr genes in the same strain has no additive effect on resistance levels. Further, our data correlate with previously mentioned studies concerning the coexistence of different resistance mechanisms in one isolate [37,38,39].

The predominance of PMQR among isolates that lack gyrA mutations is noteworthy, since these genes are known to facilitate the selection of resistant mutants. Data from other studies [42, 43] suggest that, depending on which mutations are already present in a strain, the acquisition of further fluoroquinolone resistance genes could increase the strain’s fitness.

In this study, the most frequent PMQR genes were qnrB and qnrS, respectively. Correlating with our data, it has been reported previously that qnrB is the most frequent PMQR gene, followed by qnrS. [11, 36, 44, 45]. The presence of qnr genes in environmental E. coli indicates that selection could occur without exposure to inhibitory concentrations of fluoroquinolones. It has been demonstrated previously for environmental Klebsiella pneumoniae isolated from wastewater that qnr genes confer a selective advantage in the presence of residual subinhibitory fluoroquinolone concentrations present in wastewater [46] Accordingly, pig farms with a history of (in-feed) application of fluoroquinolones may represent environments containing residual concentrations of antibiotics which propagate PMQR genes. Our data suggest that this may hold true in particular for farrowing and rearing farms, where such genes were detected exclusively on farms with a history of use of fluoroquinolones. Environmental pollution with residual fluoroquinolones is facilitated by their poor degradability and strong potential for binding to sediments [47]. In the absence of solar radiation some fluoroquinolones (enrofloxacin) remain stable for at least 120 days [48]. In their long-term experimental study Xu and colleagues [49] stated that the application of swine manure lead to an increase of fluoroquinolone resistances in soil, including PMQR genes, which persisted at least five months. On farms without prior treatment of animals, strains harbouring quinolone resistance genes may be introduced during transfer of pigs from other locations, comparable to inter-farm transmissions of extended-spectrum ß-lactamase (ESBL) producing E. coli in pigs [50]. Our data suggest that this may especially be the case for mating and gestation farms and for fattening farms, where, in contrast to farrowing and rearing farms, environmental contamination with E. coli harbouring PMQR genes was detected irrespective of the history of fluoroquinolone treatment. These findings suggest that animal movement to and from farrowing and rearing farms with recent histories of treatment may promote the risk of transmission of resistant bacteria and of fluoroquinolone resistance genes among farms within sow pool systems.

Apart from direct transmission of resistant E. coli, PMQR genes can be transferred horizontally to other bacteria in the pig farm environment. Exposure to quinolones of bacteria containing qnr genes may increase their capacity to acquire point mutations in the gyrase and/or topoisomerase IV genes [46]. PMQR genes are often harboured on plasmids containing other resistance genes, e.g., ß-lactamases [51], thus, the use of non-fluoroquinolone antimicrobials enables their co-selection [52].

Overall, the resistance profiles of the isolates described in this study are in agreement with previous studies that have demonstrated that some of the most common antibiotic resistances (other than to quinolones) in E. coli in the pig environment are to tetracycline, ampicillin, streptomycin and sulfamethoxazole [53, 54], and that E. coli strains from the pig environment often are resistant to four or five antibiotics simultaneously [55].

Phylogenetic grouping, MLST and serotyping of ST301

The majority of the isolates belonged to phylogenetic groups A or B1. The predominance of ST10 among the E. coli belonging to group A reflect previous observations regarding isolates from pigs, chicken faeces, as well as retail chicken and pork meat [56, 57]. Furthermore, Araùjo et al. [58] observed that E. coli ST10 and ST297 were the predominant sequence types among MDR isolates isolated from irrigation water and vegetables in household farms, highlighting the wide dissemination of this sequence type and its association with MDR. Notably, E. coli ST10 (CC10), as well as E. coli CC23 also identified in this study, both including ciprofloxacin resistant and MDR strains, are also associated with urinary tract infection and sepsis in humans [59].

In addition, several other E. coli belonging to phylogenetic group A identified in this study have been associated with disease in food producing animals, such as avian pathogenic E. coli ST93 [60], E. coli ST744 isolated from diseased calves [61], and strains belonging to CC165 from food producing animals [62]. Notably, among the latter, we detected two EPEC O80:H2 isolates, which is considered an emerging pathogen among calves in Belgium [63]. This serotype has emerged among humans as a highly virulent extra-intestinal pathogenic Shiga-toxin producing E. coli (STEC) in France and Switzerland since 2015 [30, 33]. As opposed to the STEC O80:H2 found in humans, the isolates from this study lacked pS88-associated extra-intestinal virulence genes. Therefore, the relationship between EPEC O80:H2 isolated from the farm environment and human STEC isolates needs to be assessed, e.g. by whole genome sequencing, and the prevalence of EPEC O80:H2 in pigs should be established.

E. coli strains belonging to the phylogenetic group B1 are for the most part commensal, with the ability to persist in the environment [64]. Most ST of this phylogroup, such as ST58, ST162, or ST453 have been frequently detected among healthy livestock [65]. However, E. coli ST297, which was one of most frequently observed ST in this study, has been associated with disease in both poultry and humans [66]. Likewise, E. coli ST453 is known to cause extraintestinal disease in humans (urinary tract infections) and metritis in cattle [67].

A minority of strains were assigned to the extraintestinal virulent phylogroups D and F. Among these, we detected two MDR strains belonging to ST117 which is a well-recognized avian pathogenic E. coli with zoonotic potential [68].

Conclusions

Quinolone non-susceptible E. coli are widespread in the environment of Swiss pig farms. In particular, isolates showing intermediate resistance to nalidixic acid frequently possess transmissible PMQR genes. This is worrisome, since the presence of qnr genes may increase the ability of bacteria to acquire point mutations in the gyrase and topoisomerase IV genes, resulting in high level resistance to (fluoro)quinolones. Furthermore, plasmids harbouring qnr genes may contribute to the horizontal spread of antibiotic resistance in livestock and in the environment. In pig farms which are part of sow pool systems, inter-farm measures that aim to reduce the risk of spreading resistant bacteria and resistance genes from one stage of production to the next need to be assessed and promoted. Our data further show that farm environments contain commensal MDR E. coli as well as E. coli with zoonotic potential. In particular, we demonstrate for the first time the presence of EPEC O80:H2 in an environmental sample from a pig farm.

In order to preserve the usefulness of fluoroquinolones and to protect animal and human health, surveillance of antimicrobial resistance is warranted. Measures for prudent use of (fluoro)quinolones as provided by the European Union’s guidelines for use of antimicrobials in veterinary medicine are of utmost importance [69].