A total of 323 pig nasal swab samples and 442 fecal samples were collected in 2013 from six swine farms located in four geographically distinct provinces in China: one farm in Jilin (DBNA), two farms in Guangdong (GDEP and GDYD), one farm in Jiangxi (JXTC), and two farms in Henan (HNYM and HNYC). The usage of antibiotic agents at these farms over the past year (except farm DBNA) was supplied by their veterinarians. All animal studies were performed with the approval of the Animal Care and Use Committee of South China Agriculture University. Nasal swab samples were incubated in 7.5% Sodium Chloride Broth for 16–18 h at 37°C and then streaked onto Mannitol Salt agar plates and grown for 16–18 h at 37°C. Next, one presumptive staphylococcal isolate per sample was randomly selected for further analysis. Fecal samples were used to isolate E. coli as previously described (Chen et al., 2007). Species identification of all cfr-positive strains was determined by the ID32 STAPH system (bioMérieux, Craponne, France) and was further confirmed by sequencing the 16S rRNA using universal prokaryotic primers (Weisburg et al., 1991).

All Staphylococcus spp. and E. coli isolates were screened for the presence of cfr by polymerase chain reaction (PCR) as previously described (Kehrenberg and Schwarz, 2006). The cfr-positive staphylococci were also PCR-screened for the presence of the fexA, fexB, mecA, and mecC genes (Murakami et al., 1991; Kehrenberg and Schwarz, 2006; Liu et al., 2012; Paterson et al., 2012). Additionally, the cfr-positive E. coli strains were screened for floR (Chen et al., 2004).

The genetic context surrounding cfr was examined by PCR mapping and sequencing in eight randomly selected cfr-positive Staphylococcus spp. and two cfr-positive E. coli isolates. The primers used to determine the regions upstream and downstream of cfr are listed in Supplementary Table S1.

All cfr-positive Staphylococcus spp. and E. coli isolates were tested for their antimicrobial susceptibility using the agar dilution method. Antimicrobial susceptibility tests were conducted and evaluated according to the recommendations specified in CLSI documents VET01-S2 (CLSI, 2013) and M100-S25 (CLSI, 2015). Minimal inhibitory concentrations (MICs) of ≥16 mg/L were tentatively considered as criteria of florfenicol resistance (Kehrenberg and Schwarz, 2006). S. aureus ATCC^®29213 and E. coli ATCC^®25922 served as quality control strains.

Seven housekeeping genes, including adk, fumC, gyrB, icd, mdh, purA, and recA, were used to determine the sequence types (STs) of all the cfr-positive E. coli (Lau et al., 2008). The MLST databases were used for BLAST analysis¹.

Pulsed-field gel electrophoresis (PFGE) was used to determine the clonality of the cfr-positive S. sciuri isolates. Whole-cells were packed with an equal volume of 2% (wt/vol) low melting point agarose (Bio-Rad, Laboratories, Hercules, CA, USA) dissolved in EC buffer (Liu et al., 2014) and poured into plug molds. The frozen plugs were addressed according to the previously described with minor modification. The digested fragments were separated using a CHEF-DRIII system (Bio-Rad) with a clamped homogeneous electric field of 6 V/cm, using a 120° switch angle for 24 h at 14°C, with the pulse time linearly ramped from 3 to 40 s. PFGE was performed for all cfr-positive E. coli isolates using the CHEF Mapper System (Bio-Rad), according to a previously described protocol (Chen et al., 2007). Comparison of PFGE patterns was performed using the BioNumerics software (Applied Maths, Sint-Martens-Latem, Belgium) using the Dice coefficient (1.5% optimization and 1.0% tolerance); a similarity cutoff of 80% was used to identify a PFGE cluster.

Plasmid pHNEP28 from porcine E. coli isolate EP28 was purified from a transconjugant using a Qiagen plasmid midi kit (Qiagen, Hilden, Germany) and 3 μg DNA was quantified by NanoDrop^TM 2000/2000c Spectrophotometers from Thermo Scientific. Sequence was performed by the Roche 454 GS-FLX system. Contigs were assembled with the 454 GS de novo assembler (Newbler) v2.8. Gaps between contigs were closed by PCR and sequencing. Annotation of pHNEP28 was performed using the RAST server (Aziz et al., 2008), ISfinder², and BLAST³.

Statistical significance for the comparison of prevalence data was determined by the x²-test. p-values less than 0.05 were regarded as statistically significant.

The nucleotide sequence of the plasmid pHNEP28 has been deposited with the GenBank nucleotide sequence database under accession number KT845955.

Two-hundred and fifty four Staphylococcus spp. strains and 398 E. coli strains were recovered from the 323 nasal swab and 442 fecal samples, respectively (Table 1). Among them, cfr was detected in 40 (15.7%) Staphylococcus spp. and two (0.5%) E. coli isolates. Farm GDEP in the Guangdong province showed the highest prevalence of cfr-positive (37.5%) Staphylococcus spp. strains and it was also the only farm with cfr-positive E. coli strains (Table 1). The predominant cfr-positive Staphylococcus spp. species was S. sciuri (n = 38); additionally, one S. chromogenes isolate was obtained from farm HNYC and one S. lentus isolate was found at farm GDYD.

Antimicrobial susceptibility assays showed that all 40 cfr-positive Staphylococcus spp. isolates exhibited resistance to erythromycin, tetracycline, clindamycin, and trimethoprim-sulfamethoxazole. In addition, 39 (97.5%), 37 (92.5%), 36 (90%), 36 (90%), and 34 (85%) cfr-positive Staphylococcus spp. isolates demonstrated resistance to oxacillin, tiamulin, valnemulin, florfenicol, and gentamicin, respectively. Of the 40 cfr-positive staphylococci, 19 (47.5%) and 11 (27.5%) were resistant to ciprofloxacin and rifamycin, while only three (7.5%) isolates were linezolid-resistant (Table 2 and Supplementary Table S2). The two cfr-positive E. coli strains also exhibited a multiresistance phenotype, namely, ampicillin, tetracycline, streptomycin, florfenicol, and trimethoprim-sulfamethoxazole resistance. Strain EP28 was also ciprofloxacin-resistant (Table 3).

Of the 40 cfr-positive staphylococcal isolates, 25 strains carried fexA, while fexB was not detected. Interestingly, 32 of the 39 oxacillin-resistant staphylococci harbored mecA; mecC was not detected in the remaining seven mecA-negative strains. In addition, mecA, fexA, and cfr coexisted in 22 staphylococci (Table 2). The cfr-positive E. coli strain EP28 also carried floR (Table 3).

The 38 cfr-positive S. sciuri strains were analyzed by PFGE and 28 major SmaI patterns were observed, whereas seven strains were nontypeable by PFGE using SmaI (Table 2 and Supplementary Figure S1). Although obtained from the same farm, the two cfr-carrying E. coli isolates were not related (Table 3). The cfr-positive E. coli strain EP25 was shown to belong to ST1602, whereas E. coli EP28 represented a novel ST, ST4241 (Table 3). Interestingly, five cfr-positive S. sciuri strains isolated from the same farm located in the Jiangxi province were grouped in two clusters (Dice similarity index was ≥80%): strains with similar PFGE patterns A1, A2, and A3, and two strains showing related PFGE patterns B1 and B2.

The regions surrounding cfr were determined by PCR mapping and sequencing in eight randomly selected Staphylococcus strains. In six isolates, IS21-558 was identified 99% identity with upstream of cfr and truncated transposase gene tnpB was found downstream of cfr (Table 2). This arrangement (IS21-558-cfr-ΔtnpB) is identical 99% identity to the corresponding region of plasmids pSS-02 (accession no. JF834910, S. saprophyticus, swine, China), pHNTLD18 (accession no. KF751702, S. equorum, pork, China), pSA737 (accession no. KC206006, S. aureus, patient, USA), pHK01 (accession no. KC820816, S. cohnii, patient, China), and pHNCR35 (accession no. KF861983, S. simulans, hog market worker, China) (Wang et al., 2015).

In S. sciuri strain GDEP27, cfr was bracketed by two copies of ISEnfa4 in the same orientation, which was identical 98% identity to the corresponding region of E. faecalis pW9-2 (accession no. JQ911741, sewage at pig farm, China) and pHOU-cfr (accession no. JQ660368, patient, Thailand), E. thailandicus pW3 (accession no. JQ911739, sewage at pig farm, China), and p3-38 (accession no. JQ911740, pig, China).

In S. sciuri strain JXTC28, cfr is bracketed by two copies of IS256 in the same orientation, a partial fragment of IS1216 and the aminoglycoside resistance gene aacA-aphD were located downstream of IS256. This ΔIS1216-aacA-aphD-IS256-cfr-IS256 module was 98% identity to that found on plasmid pSS-04 (accession no. KF129410) carried by porcine S. sciuri GN5-1 in Sichuan province, China.

In both cfr-positive E. coli strains, the cfr gene was flanked by two copies of IS26 (Supplementary Figure S2). Strain EP25 had a similar genetic environment to plasmids pEC-01 (accession no. JN982327) identified in a porcine E. coli strain from the Shandong province, China, and pSD7 (accession no. KJ453116) found in a porcine E. coli isolate from the Guangdong province, China; 2 copies of IS26 located in the same orientation, but opposite to cfr gene (Supplementary Figure S2B). In EP28, the 2 copies of IS26 are in opposite orientation relative to one another, with cfr in the same orientation than IS26 right-copy (Supplementary Figure S2A).

S1-PFGE and Southern blotting showed that the cfr gene in isolates EP25 and EP28 were located on a plasmid of approximately 95 and 105 kb, respectively. The cfr-carrying plasmids from isolate EP28 and EP25, designated as pHNEP28 and pHNEP25, were successfully transferred to E. coli C600 and DH5α, respectively. Plasmid pHNEP28 was identified as F43:A-:B-, while the replicon type of pHNEP25 could not be determined using the PBRT method. Both the transconjugant EP28-2J and the transformant EP25-1Z exhibited resistance against ampicillin, florfenicol, and elevated MICs of cefotaxime and ciprofloxacin, compared with the recipient E. coli strain (C600 or DH5α; Table 3). In addition, EP25-1Z also showed resistance to tetracycline, streptomycin, and sulfamethoxazole-trimethoprim, whereas the acquisition of pHNEP28 increased tetracycline MIC by only onefold.

Sequence analysis revealed that pHNEP28 is 108,837 bp in size with an average GC content of 49.85% (contigN50: 108837 bp). Plasmid pHNEP28 consisted a typical IncFII-type backbone (83,398 bp) encoding plasmid replication, horizontal transfer, maintenance and stability functions, and two multi-resistance regions (MRR) (Figure 1).

The replication region (position 36,852–38,269 bp), consisting of repA2 and repA1, was classified as F43:A-:B:- by RST. This region shared 97% identity to two F2:A-:B- NDM-producing plasmids from E. coli isolates: pMC-NDM (accession no. HG003695) and pGUE-NDM (accession no. JQ364967). The transfer-leading region comprised genes related to plasmid maintenance and stability, including parA, parB, and ssb, as well as many putative genes. Additionally, genes involved in induced mutagenesis and DNA damage tolerance (umuC and umuD) and in the inhibition of the bacterial SOS response (psiA and psiB) were also identified. A section of the plasmid transfer-leading region (position 4,906–15,340 bp) shared 99% nucleotide identity with Enterobacter cloacae plasmid pIMP-HB623 (accession no. KM877517). Another section (position 77,101–88,852 bp) showed 97% identity to enteropathogenic E. coli plasmid p026_2 (accession no. AP010955).

The transfer region of pHNEP28 included 24 tra genes and finO and is less highly related to other IncFII plasmids, with ∼94%-97% identity to this region in E. coli plasmid pEB3 (accession no. CP006001) from Vietnam, which was designed to F30:A-:B:- by RST. Furthermore, the partial tra region of pHNEP28 (position 69475-77025 bp) showed high homology (98% nucleotide identity) with plasmids pMC-NDM and pGUE-NDM.

The pHNEP28 contained two mosaic MRR regions (Figure 2). The first 9,677 bp MRR, located downstream of the replication region, contains two resistance genes (tet(M) and bla_TEM-1) and complete or truncated insertion sequences and transposons (IS1B, ΔintI4, ΔTn2, ISAplI, and IS26). The tet(M) gene found in pHNEP28 shared 96–98% nucleotide identity with sequences in S. pneumoniae strain 11930 (accession no. FR671416), Tn916-like transposon Tn6085 from E. faecium (accession no. HM243623), Tn916/Tn1545-like transposon from E. faecalis (accession no. DQ223241), and S. aureus strain TW20 (accession no. NG_048252); whereas it was identical to the tet(M) gene from porcine E. coli isolate CICYT-332 (accession no. KJ755873) found in Spain, although only 1802/1935 bps of CICYT-332 tet(M) were sequenced. While tet(M) demonstrated only 96% identity to that in plasmid pTCY4 (KJ772289, duck E. coli, China), the 1,204-bp module (IS26-orf12-orf13) upstream of tet(M) was identical to the corresponding region of pTCY4. Transposon Tn2 carrying resistance gene bla_TEM-1 was interrupted by ISApl1, which generated 2-bp direct repeats (DR) (5′-CT-3′); tnpR of Tn2 was truncated by IS26, which interrupted another Tn2 with an opposite orientation (Figure 2A).

Another MRR module (14,370 bp), including the multiresistance gene cfr, florfenicol efflux pump gene floR, and quinolone resistance gene qnrS1, was inserted into the plasmid stability region. As shown in Figure 2B, the cfr gene was flanked by two IS26 located in the opposite orientation, in agreement with the result obtained by PCR mapping. In addition, the 2,513-bp module (Δrep-IS26) was identical to the corresponding region in plasmids pHNFP671 (accession no. KP324830), pSD6 (accession no. KJ453115), and pFSEC-01 (accession no. KR779901) from porcine E. coli in China. In plasmid pHNEP28, the Δrep-IS26 region was found in the opposite orientation with a second copy of IS26 present in the downstream region. Furthermore, another IS26 element is present downstream of cfr on plasmids pHNFP671 and pSD6, though the 2,414-bp segment containing cfr and IS26 was identical to plasmid pHNEP28 in our study (Figure 2B).

Moreover, plasmid-mediated quinolone resistance gene qnrS1 was flanked by two copies of ISKpn19 located in the opposite orientation, although the first one was interrupted by IS26 (Figure 2B). The 4615-bp module (qnrS1-ISKpn19) was present in multiple plasmids, such as Klebsiella pneumoniae plasmids pNDM-1fa (CP014757, human, USA) and pK1HV (HF545434, human, Vietnam), and E. coli plasmids pPGRT46 (KM023153, human, Nigeria), pEQ2 (KF362122, equine, Czech Republic), and pKT58A (JX065631, water fowl, Slovakia) (Dolejska et al., 2013). The IS26-ΔISKpn19-ΔqnrS1 module was identical 99% to that in E. coli isolate E80 (KF921555, raw wastewater, Portugal), containing a disrupted qnrS1 gene (Figure 2B). The 4,292-bp segment (ΔvirD2-floR-ΔISCR2) was identical 99% identity to the corresponding region of plasmids pHNSHP45-2 (KU34138, porcine E. coli, China), pHXY0908 (KM877269, chicken Salmonella Typhimurium, China), and pK1HV (HF545434, human K. pneumoniae, Vietnam); however, both virD2 and ISCR2 were truncated by IS26 in plasmid pHNEP28 (Figure 2B).