A total of 1,354 individual E. coli isolates were collected from Guangzhou, Guangdong province, China between July 2011 and May 2013, including 509 animal strains (372 pig and 137 chicken strains) isolated from fecal samples of healthy animals from one hog market and one live poultry market, 346 strains from retail meat (247 pork and 99 chicken meat samples) recovered from fresh or chilled pork and chicken samples purchased from supermarkets and farmers' markets, and 498 strains from patients (71 E. coli strains from urine samples from inpatients with urinary tract infection, 427 E. coli strains from fecal samples of inpatients and outpatients) from four hospitals. The samples were inoculated onto MacConkey agar, and suspected E. coli colonies (one isolate per sample) were selected and identified by standard biochemical testing. All the isolates were tested for the presence of oqxAB by PCR, and 100 amplified products were randomly selected for sequencing (Table S1).

All oqxAB-positive E. coli isolates were tested for their MICs of ampicillin, cefotaxime, amikacin, gentamycin, neomycin, apramycin, tetracycline, florfenicol, ciprofloxacin, olaquindox, colistin, and sulfamethoxazole/trimethoprim using the agar dilution or broth microdilution method (limited to colistin). Antimicrobial susceptibility tests were performed and evaluated according to the protocols recommended by M100-S25 of the Clinical and Laboratory Standards Institute (Wayne, PA, USA) (Clinical and Laboratory Standards Institute, 2015). Colistin (>2 mg/L), neomycin (>8 mg/L), and florfenicol (>16 mg/L) were interpreted according to the clinical breakpoints or epidemiological cutoff values of the European Committee on Antimicrobial Susceptibility Testing (EUCAST; https://mic.eucast.org/Eucast2/); apramycin (>16 mg/L) and olaquindox (>64 mg/L) were analyzed according to DANMAP 2012 and DANMAP98, respectively (http://www.danmap.org/Downloads/Reports.aspx). The E. coli reference strain ATCC 25922 was used as the quality control. Epi Info version 7.2 (CDC) was used to perform statistical analysis. Comparison of prevalence of oqxAB and antimicrobials resistance rates was conducted by the χ²-test. P < 0.05 were considered statistically significant.

The genetic diversity of 50 randomly selected oqxAB-positive E. coli isolates from animals (n = 21), food products (n = 9), and patients (n = 20) was characterized by multilocus sequence typing (MLST; http://mlst.warwick.ac.uk/mlst/dbs/Ecoli) and PFGE (Gautom, 1997). MLST data were aligned using ClustalW, and a phylogenetic tree was constructed by the neighbor joining algorithm using MEGA 7.0 (Kumar et al., 2008). PFGE patterns were compared and analyzed using BioNumerics (Applied Maths, Sint-Martens-Latem, Belgium) using a cut-off value of 90% to define PFGE clusters.

The 50 oqxAB-positive E. coli strains subjected to PFGE were further analyzed by performing conjugation experiments with streptomycin-resistant E. coli C600 as the recipient strain following a previously described protocol (Chen et al., 2007). Transconjugants were selected using 32 μg/mL olaquindox and 3,000 μg/mL streptomycin.

Transformation was conducted via heat-shock or electroporation using E. coli strain DH5α as the recipient (Hanahan, 1983; Dower et al., 1988). Plasmid DNA was extracted using the E.Z.N.A. Plasmid DNA Midi Kit (Omega, Norcross, GA, USA). Transformants were selected using MacConkey agar plates containing 16 μg/mL olaquindox.

The presence of oqxAB in the transconjugants/transformants was confirmed via PCR and sequencing. Other resistance genes, including floR, bla_CTX−M, rmtB, and fosA3 were also screened using the primers listed in Table S1. Antimicrobial susceptibility of all the transformants and E. coli DH5α recipient strain were determined using the agar dilution or broth microdilution method (limited to colistin). All the transformants were characterized by PCR-based replicon typing and were screened IncX plasmids as described previously (Carattoli et al., 2005; Johnson et al., 2012). Replicon sequence typing and plasmid double locus sequence typing were performed to further characterize IncFII and IncHI2 plasmids according to previously described protocols (García-Fernández and Carattoli, 2010; Villa et al., 2010). S1-PFGE (Barton et al., 1995), and Southern blot hybridization were performed to determine the number of plasmids and the sizes of oqxAB-carrying plasmids in all the transformants, using a non-radioactively labeled oqxAB-specific probe. The genetic context of oqxAB was determined by PCR mapping (Table S2). Ten transformants with a single oqxAB-carrying plasmid were further analyzed by restriction fragment length polymorphism (RFLP) using the endonuclease ApaLI, TZC215-1, and AHH13-1 were excluded since their sizes were significantly different from other oqxAB-carrying plasmids.

Two IncHI2 plasmids, namely, pHNLDH400 from a patient with E. coli colonization and pHNYJC8 from chicken meat, were selected for sequencing. Plasmid DNA was purified from the transformants using QIAGEN® Plasmid Midi Kit according to the manufacturer's instructions (Qiagen, Hilden, Germany), and was completely sequenced by PacBio single-molecule real-time sequencing (RSII platform) (Pacific Biosciences, Menlo Park, CA, USA). Raw reads were introduced into the non-hybrid Hierarchical Genome Assembly Process. Analysis and annotation of the resulting sequences were performed using RAC (http://rac.aihi.mq.edu.au/rac/), ISfinder (https://www-is.biotoul.fr//), BLAST (http://blast.ncbi.nlm.nih.gov/Blast.cgi), RAST server (Aziz et al., 2008), and the Gene Construction Kit 4.0 (Textco BioSoftware, Inc., Raleigh, NC, USA). Plasmids pHNLDF400 and pHNYJC8 were compared with similar oqxAB-bearing IncHI2 plasmids pHXY0908 and pHK0653 obtained from S. Typhimurium, as well as our previously reported plasmids pHNSHP45-2 (Zhi et al., 2016), and pHNAH67 from E. coli isolate AHC67 (Yang X. et al., 2014) using BLASTn and BRIG (Alikhan et al., 2011). The IncHI2 plasmid R478 (GenBank accession number BX664015) served as the reference plasmid for annotation.

The nucleotide sequences of plasmids pHNLDF400 and pHNYJC8 have been deposited in the GenBank database under the accession numbers KY019258 and KY019259, respectively.

Of 1,354 E. coli isolates examined in this study, oqxAB was detected in 322 (23.78%) E. coli isolates. Consistent with the results of our previous study (Yang T. et al., 2014), E. coli isolates from animal sources showed the highest oqxAB prevalence rate (172/509, 33.79%). However, oqxAB prevalence among the isolates of human origins (90/498, 18.07%) was slightly higher than those of food origins (60/346, 17.34%), contradictory to previous results, which was probably because we detected clinical E. coli isolates instead of strains from healthy volunteers. The oqxAB-positive isolates from patients were possibly selected under pressure exerted by antimicrobials, such as fluoroquinolones. However, oqxAB was previously reported to have low prevalence among clinical E. coli isolates in Europe (0.46%; Ciesielczuk et al., 2013), Korea (0.4%; Kim et al., 2009), Taiwan (6.05%; Kao et al., 2016), and mainland China (6.6%; Yuan et al., 2012; Zhao et al., 2015). Interestingly, the chicken isolates (51/137, 37.23%) showed a slightly higher oqxAB prevalence rate than the porcine isolates (121/372, 32.53%); similarly, the chicken meat isolates (22/99, 22.22%) showed higher oqxAB prevalence than the pork isolates (38/247, 15.38%).

As shown in Figure 1, the isolates from food-producing animals showed the highest resistance rates against multiple antimicrobials, including neomycin (57.6%), apramycin (60.5%), florfenicol (74.4%), olaquindox (98.3%), and colistin (18.0%), which are used only in veterinary medicine in China (colistin was banned from animal feed in April 2017 in China), as well as tetracycline (97.1%) and sulfamethoxazole/trimethoprim (97.7%). It was followed by the isolates from food (56.7, 51.7, 61.7, 55.0, 16.7, 81.7, and 93.3%, respectively), and humans (26.7, 17.8, 41.1, 44.4, 1.1, 58.9, and 83.3%, respectively). However, the food isolates showed the highest resistance rate of 51.7% against gentamycin, which was considerably higher than the resistance rates of animal and human isolates (37.2 and 30%, respectively). In addition, the resistance rates against cefotaxime were significantly higher in the oqxAB-carrying isolates from humans (50%) than from animals (19.8%; P < 0.01) and food products (30%; P < 0.05), which could be explained by the relatively more frequent use of cephalosporins in clinical medicine.

The MLST analysis identified 21 different sequence types (STs) among 39 oqxAB-positive E. coli isolates (Table S3). The most commonly identified genotypes were ST10 (n = 10), followed by ST410 (n = 4) and ST744 (n = 3), whereas the isolates belonging to ST93 (n = 2), ST165 (n = 2), ST542 (n = 2), ST602 (n = 2), ST48, ST57, ST58, ST178, ST206, ST224, ST301, ST359, ST453, ST746, ST847, ST1421, ST3339, and ST6697 were also identified (Table S3). The remaining 11 oqxAB-positive E. coli were assigned to 11 novel STs (Table S3). In our study, the dominant ST10 clone, which has been frequently detected in animals, humans, food products, and the environment, and is responsible for the spread of multiple antibiotic resistance genes (Belmar Campos et al., 2014; Müller et al., 2016; Röderova et al., 2017), was detected in the isolates from animals, chicken meat, and patients. Additionally, other STs described in this study, such as ST410, ST744, ST93, ST48, and ST58, have been detected in various sources (Loncaric et al., 2013; Belmar Campos et al., 2014; Abraham et al., 2015; Falgenhauer et al., 2016; Röderova et al., 2017). The results of the phylogenetic analysis revealed the genetic relatedness of oqxAB-positive isolates (Figure S1).

Among the 50 oqxAB-positive E. coli isolates from different sources analyzed using XbaI-PFGE, 44 oqxAB-carrying E. coli isolates from animals (n = 19), animal-derived food products (n = 6), and patients (n = 19) exhibited 43 distinct PFGE patterns (Figure 2). Notably, the isolates belonging to the same STs identified in our study showed different PFGE patterns; however, the ST410 E. coli isolates BYMP20 and YZHF29, which were obtained from pork and human sources from different districts in Guangzhou within 12 km, showed similar PFGE patterns (E1 and E2; Figure 2).

MLST and PFGE demonstrated the molecular diversity of the oqxAB-positive E. coli isolates, and suggested that clonal transfer might not be the main mechanism underlying oqxAB dissemination among E. coli isolates in Guangzhou, China.

Surprisingly, none of the 50 randomly selected oqxAB-positive E. coli isolates transferred oqxAB to E. coli C600 via conjugation, which is consistent with the results reported for S. Typhimurium (Wong et al., 2014). Therefore, transformation assays were performed, and 21 transformants were obtained successfully. All the transformants were subjected to S1-PFGE and Southern blotting. The results demonstrated that the transformants carried one to four plasmids of various sizes. oqxAB was detected in plasmids with sizes ranging from ~33.3 to ~500 kb. Four transformants (SNJ23-1, SNJ41-1, ZYTM154-1, and LDHF159-1) showed two hybridization signals, whereas no hybridization signal was observed in SNX19-2 (Table 1). Thirteen transformants contained IncFII (n = 4), IncFIA (n = 2), IncFIB (n = 3), IncHI2 (n = 3), and IncN (n = 5) replicons. Notably, three transformants carried both IncN and IncFII replicons, which were classified as N1-F33: A-: B- (Table 1). Interestingly, the three transformants were obtained from three original E. coli isolates exhibiting different PFGE patterns from the same hospital, of which two carried only one oqxAB-bearing plasmid with similar sizes (Table 1 and Figure S2). Three IncHI2 plasmids from food product and patients were assigned to ST3 via plasmid double locus sequence typing (Table 1). PCR-based replicon typing could not determine the replicon types of the remaining eight transformants. Although rare, an untypable oqxAB-carrying plasmid was described previously (Liu et al., 2013). Ten transformants with single plasmid were selected for RFLP with ApaLI. Interestingly, two N1-F33:A-:B- plasmids from the patients of the same hospital (ZYTM118-1 and ZYTF32-1) showed identical patterns, as well as two F-:A18:B- plasmids from pigs (TZC152-6 and TZC212-1). Additionally, three ~76.8 oqxAB-bearing plasmids obtained from pig, chicken meat, and human, exhibited identical patterns (Figure S3).

Sequences of the regions surrounding oqxAB were determined via PCR mapping and sequencing. oqxAB was flanked by two IS26 elements in the same orientation, and formed a composite transposon Tn6010 (IS26-oqxA-oqxB-oqxR-IS26) in 19 transformants. Additionally, one IS26 element was present in upstream or downstream of oqxAB in the remaining transformants TZC338-4 and SNX19-2. Since its first identification in the plasmid pOLA52 from porcine E. coli in 2003 (Sørensen et al., 2003), Tn6010 is believed to play a vital role in oqxAB transmission among Enterobacteriaceae isolates (Liu et al., 2013; He T. et al., 2015). Our results further support that horizontal transfer mediated by mobile elements seems to be the main mechanism for oqxAB transmission in E. coli from different sources.

The MICs of ciprofloxacin and olaquindox against the transformants were 2- to 32-fold (0.015–0.25 μg/mL) and 2- to > 32-fold (32 to >256 μg/mL) higher than the recipient E. coli DH5α strain (0.008 and 16 μg/mL), respectively. Furthermore, the co-transfer of resistance to ampicillin (n = 12), gentamycin (n = 10), tetracycline (n = 9), florfenicol (n = 9), sulfamethoxazole/trimethoprim (n = 11), cefotaxime (n = 3), and amikacin (n = 3) was observed in 12 transformants containing one plasmid. Multiple resistance genes were observed to coexist with oqxAB within the same plasmid accounting for the antimicrobial resistance; such as bla_CTX−M−55, rmtB, fosA3, and floR were detected in two N1-F33:A-:B- plasmids from patients; aac(6′)-Ib-cr and/or qnrS were identified in both F-:A18:B- plasmids from pigs (Table 1). This allows for the co-selection of other genes conferring resistance to antibiotics that are routinely used in clinics, such as cephalosporins, aminoglycosides, and fosfomycin, or subject to selective pressure by quinolones, as well as quinoxalines used in food-producing animals. In turn, the co-selection of multiple drug resistance genes might facilitate further oqxAB transfer among Enterobacteriaceae isolates.

Two oqxAB-bearing IncHI2 plasmids in this study, namely pHNYJC8 from YJMC8 of food origin and pHNLDF400 from LDHF400 of human origin, were selected for sequencing. Plasmids pHNLDF400 and pHNYJC8 have sizes of 249,152 and 249,153 bp, respectively, and both have a GC content of 46.51%. They contain a typical IncHI2-type backbone, which encodes genes for plasmid replication, conjugative transfer, maintenance, and stability. Similar to other IncHI2 plasmids, such as pEC5207 (KT347600, porcine E. coli, China), pHNLDF400 and pHNYJC8 also carry a set of tellurite resistance determinants (terZABCDEF).

Sequence comparisons demonstrated that pHNYJC8 and pHNLDF400, obtained from the E. coli clones ST93 and ST57, respectively, were identical with only 2-bp nucleotide difference, thereby confirming that the same oqxAB-carrying IncHI2 plasmid was transferred horizontally among E. coli isolates in human and food samples. Furthermore, both plasmids showed 99% identity to the previously sequenced oqxAB-bearing IncHI2 plasmids pHXY0908 from S. Typhimurium in chicken from China and pHK0653 from clinical S. Typhimurium in Hong Kong (Wong et al., 2016), except that a 4,284-bp fragment (floR-ISCR2) was absent in pHK0653 (Figures 2, 3), further suggesting that the same or similar oqxAB-carrying IncHI2 plasmids are transmitted among Enterobacteriaceae isolates from various sources in different geographic regions of China. Additionally, the backbones of pHNLDF400 and pHNYJC8 showed 99% identity to those of our previously sequenced colistin-resistant mcr-1-bearing plasmids pHNSHP45-2 (KU341381) obtained from porcine E. coli isolate SHP45 in Shanghai, China (Zhi et al., 2016) and pHNAH67 (KX246266) obtained from an E. coli isolate from chicken in Anhui province, China. However, the abovementioned plasmids contain distinct variable regions. pHNSHP45-2 contains a 3,376-bp fragment harboring fosA3 and bla_CTX−M−14. Contrastingly, in our study, both pHNLDF400 and pHNYJC8 contain a 4,713-bp segment harboring the |aac(6′)-Ib-cr|bla_OXA−1|catB3 |arr3|qacEΔ1|sul| cassette array, which is flanked by the 3′-conserved segment (3′-CS) and interrupted by a Tn21-like transposon Tn_chrA, itself truncated by IS26. Most importantly, pHNSHP45-2 also harbors the colistin resistance gene, mcr-1, which is associated with ISApl1, and is inserted in the plasmid backbone (Figure 3). Similarly, a 6,760-bp fragment containing fosA3 and bla_CTX−M−65 is identified in pHNAH67, whereas pHNLDF400 and pHNYJC8 contain a ~28.3-kb fragment with multiple resistance genes oqxAB, sul1, ΔaadA2, dfrA12, aphA1, sul3, aadA1a, cmlA1, and aadA2, which is absent in pHNAH67 (Figure 2). The absence/presence of these regions might be explained by IS26-mediated homologous recombination or replicative transposition (He S. et al., 2015). Furthermore, the oqxAB-associated composite transposon Tn6010 and the upstream fragment (3,678-bp) containing ΔTn2, IS26 (which interrupts the Tn2 segment), and the bleomycin resistance genes blms, orf63, and ΔISEnca1 (which is truncated by IS26 of Tn6010), were identical to the corresponding regions in the originally identified IncX1 oqxAB-carrying plasmid pOLA52, as well as pHXY0908, pHK0653, and pHNSHP45-2. The above mentioned results suggest that these plasmids may have originated from the same IncHI2-type plasmid, acquiring or losing various regions containing resistance genes, and spread among Enterobacteriaceae species in livestock, food products, and humans in different regions in China.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.