From May to December 2021, 2,103 E. coli strains were isolated from 11 cities in Zhejiang, including Hangzhou, Jinhua, Jiaxing, Qvzhou, Ningbo, Taizhou, Shaoxing, Zhoushan, Lishui, Wenzhou and Huzhou, including 1,186 strains from pigs, 904 from strains in chickens and 13 strains from ducks. Thirty-six E. coli strains were isolated from Jiangxi Province, 25 E. coli strains were isolated from Hunan Province, and one was isolated from Anhui Province (Table 1). PCR screening of isolated strains was performed to obtain the prevalence of the cfr gene in the above E. coli isolates with primer sequences (F: GTGAAGCTCTAGCCAACCGTC; R: GCAGCGTCAATATCAATCCC), as described previously (Osman et al., 2019).

Escherichia coli was inoculated on Luria-Bertani (LB) agar medium for pure culture, according to the micro-dilution method recommended in the M100-S31 document of the American Committee for Clinical Laboratory Standardization (CLSI) (Humphries et al., 2021; Tang et al., 2022b). The antimicrobial susceptibility of E. coli to 13 tested antibiotics were, ampicillin (2–128 μg/ml), amoxicillin-clavulanate acid (4/2–128/64 μg/ml), cefotaxime (0.06–64 μg/ml), meropenem (0.5–16 μg/ml), amikacin (2–64 μg/ml), gentamicin (0.25–32 μg/ml), colistin (0.125–8 μg/ml), ceftiofur (0.25–32 μg/ml), ciprofloxacin (0.06–8 μg/ml), trimethoprim-sulfamethoxazole (0.5/9.5–16/304 μg/ml), tetracycline (0.25–64 μg/ml), tigecycline (0.25–32 μg/ml), and florfenicol (2–128 μg/ml). E. coli ATCC 25922 served as quality control bacteria.

To further understand the genetic background of themultiple AMR gene cfr in E. coli, a genomic DNA extraction kit (Generay, Shanghai, China) was used to extract bacterial genomic DNA from all cfr positive strains for whole-genome sequencing (WGS). An Illumina sequencing library was generated using the NEXTflex DNA sequencing kit (Bioo Scientific, Austin, United States). Illumina paired-end sequencing was performed using the HiSeq-PE150 strategy, and the readings were filtered using fastp v0.12. Clean data were reconstructed using CLC Genomic Workbench 12.0. Prototypical strains were simultaneously whole-genome sequenced on the Oxford Nanopore GridION platform (Oxford, United Kingdom). The above genomic DNA library was prepared using the SQKLSK109 kit (Oxford Nanopore Technologies, Oxford, United Kingdom). Guppy v3.2.4 was used for base invocation and removal of adapter sequences. Sequences were assembled from scratch using a mixture of short and long reads from the Unicycler v0.4.4 pipeline (Wick et al., 2017). The reconstruction of plasmids from next generation sequence pair-end datasets was performed by PLACNETw (Vielva et al., 2017).

Acquired AMR genes and chromosomal mutations were predicted using ResFinder 4.1¹ with a percentage identification threshold of 90% and a minimum coverage length of 60%. The virulence genes were predicted using VirulenceFinder 2.0.² Plasmid replicon type identification using PlasmidFinder 2.1³ with a percentage identification threshold of 95% and percentage coverage length of 60%. Multilocus sequence typing (MLST) was performed using MLST 2.0.⁴ Phylogenetic analysis of genomes and plasmids based on maximum likelihood was performed using kSNP3 (Gardner et al., 2015). Easyfig 2.2.3 was used to compare the gene–environment (Sullivan et al., 2011). BRIG was used to plot circles of multiple plasmids for comparison (Alikhan et al., 2011).

The E. coli strain J53 was selected as the recipient strain, and cfr-positive E. coli was selected as the donor strain. Florfenicol and sodium azide were added for the selection. First, we determined that cfr-positive E. coli could not be grown on LB plates containing 100 mg/l sodium azide, and J53 could no longer be grown on LB plates containing10 mg/l florfenicol. The method of conjugation transfer was mentioned in previous reports (Xu et al., 2021; Tang et al., 2022a). The donor bacteria and recipient bacteria were inoculated into LB broth and cultured on a shaker for 4–6 h. One milliliter of the bacterial solution was taken for centrifugation, and the donor and recipient bacteria were added to the LB plate overnight at 37°C. After gradient dilution with PBS, they were inoculated onto LB square plates containing 10 mg/l florfenicol and 100 mg/l sodium azide. Finally, single colonies that grew after mating were identified via PCR to exclude false-positive cases.

To verify the circularization potential of the IS26 flanking fragments in a plasmid, a pair of primers were designed and amplified by PCR to observe whether they could form the circular intermediate of cfr-IS26. The primers used to identify the cfr-IS26 circular intermediate are shown (F: GTTGCCTGGTGTAAATGATTC; R: CTGCTAAGAGCTTGATATTC). The size of the cfr-IS26 circular intermediate was determined by Sanger sequencing.

Eight cfr-positive isolates were identified from 2,165 E. coli isolates (1,187 from pigs, 929 from chickens, and 49 from ducks), and the prevalence was 0.37% (Table 2). Seven of the cfr-positive strains were isolated from pigs, and one strain was isolated from chicken. The AST results of eight positive E. coli isolates showed that all strains were resistant to ampicillin, amoxicillin-clavulanic acid, tetracycline, and florfenicol (Figure 1; Supplementary Table 1). All the strains were sensitive to colistin, meropenem, tigecycline, and amikacin.

The contigs carrying cfr gene assembled by the second generation sequence are between 1 and 3 Kb in length (Supplementary Figure 1). The cfr-harboring E. coli strains isolated from chicken and pig belonged to different branches. Among the strains ECJHZ21-040, ECJHZ21-049, and ECNBZ21-038 were clustered together. Additionally, ECQZJ21-074 belonged to independent lineages, and there were differences between them and in the seven strains mentioned above (Figure 1). The eight E. coli isolates had distinct sequence types (STs) with ST641, ST2179, ST4434, ST88, ST349, ST10562, ST48, and ST209, indicating that cfr was widely distributed in E. coli with different genetic backgrounds.

A total of 49 types of AMR determinants within 10 classes of antibiotics were detected (Figure 2A). In addition, there were two florfenicol genes (cfr, floR), three tetracycline genes (tet(A), tet(B), and tet(M)), 10 β-lactam genes (bla_CTX-M-15, bla_TEM-150, bla_TEM-1A, bla_TEM-1B, bla_OXA-10, bla_TEM-1C, bla_OXA-20, bla_OXA-135, bla_TEM-32, bla_OXA-1), two quinolone genes (qnrS1, qnrS2), two rifamycin genes (ARR-2, ARR-3), three macrolide genes (mph(A), mdf(A), and erm(B)), one lincosamide gene (Inu(F)), six folate pathway antagonist genes (sul1, sul2, sul3, dfrA12, dfrA17, dfrA19), 14 aminoglycoside genes (aadA2b, aph(4)-Ia, aac(3)-IV, aadA2, aph(3′)-Ia, aph(3″)-Ib, aac(3)-IId, aph(6)-Id, aadA5, aph(3′)-IIa, aadA1, aac(6′)-Ib-cr, aadA22, aadA24) and some additional AMR determinants (Figure 2A). The virulence genes of the strains included terC, traT, gad, lpfA, ompT, sitA, astA, hra, etc. (Figure 2B). Among them, astA is a virulence gene encoding heat-stable enterotoxin of enteroaggregative E. coli, which may produce related toxins with the possibility of pathogenicity. Importantly, strain EC727A3 contains the virulence genes stx2A and stx2B that produce Shiga toxin, which may cause self-limiting diarrhoeal disease and sometimes bloody diarrhea as well as complications such as hemorrhagic colitis and hemolytic uremic syndrome (HUS) (Fitzpatrick, 1999; Launders et al., 2016; Mcfarland et al., 2017). Plasmid replicons include 19 types such as IncFIC(FII), IncN, IncFIA(HI1), IncFIB(K), ColE10, IncR, Col156, IncQ1, Col440II, IncFII(29), p0111, IncFII(pCoo), IncFII, IncY, IncX1, IncHI2A, IncHI2, IncX4, and IncFIB. The plasmid types of the eight isolates remained genetically diverse (Figure 2C).

The conjugation transfer assay demonstrated that all transconjugants from cfr-positive E. coli strains and E. coli J53 could grow normally on LB plates containing 100 mg/l sodium azide and 10 mg/l florfenicol. Further, PCR confirmed that the transconjugant contained the cfr gene, which indicated that the conjugative transfer experiment was successful, with a transfer frequency of 1.05 × 10⁻¹–1.01 × 10⁻⁶.

Two isolates were randomly selected from the eight cfr-positive strains for nanopore sequencing to obtain their complete genome sequences. To understand how cfr is transmitted, the genetic background of the cfr gene was further investigated. The cfr gene was located on the IncX4-type plasmid p727A3-62K-cfr (CP100066) in strain EC727A3. The length of p727A3-62K-cfr was 62,717 bp, and the GC content was 44% (Figure 3A). Moreover, p727A3-62K-cfr had high similarity with another cfr-carrying plasmid and had the highest homology with plasmid pSD11 (KM212169.1, 37,672 bp) from porcine E. coli strain 8ZG6D (65% query coverage and 99.99% identity). The collinear comparison showed that p727A3-62K-cfr and pSD11 had two different gene arrangements. The 12,647 bp region had high homology with the sequence containing the tet(M) gene in pNT1N31-93k (CP075482, 93,332 bp), and there was an insertion sequence IS1 upstream of tet(M) compared with pNT1N31-93k (Figure 3B). The other 10,831 bp region had a higher homology to a part of pSCZE4 (CP051226, 60,732 bp), and this sequence had three more IS91 insertion sequences in the same direction than pSCZE4 (Figure 3B).

The cfr gene of strain EC737A1 was located on plasmid p737A1-27K-cfr (CP100008). The length of p737A1-27K-cfr was 27,742 bp, and the GC content was 43% (Figure 4A). Plasmid p737A1-27K-cfr had a high degree of homology (100% query coverage and 100% recognition) with plasmid unnamed4 (CP037908.1, 28,519 bp). The collinear comparison showed that a 777 bp region containing the IS1 mobile element was inserted into the plasmid p737A1-27K-cfr to form unnamed4. However, the type of plasmid had not yet been determined; it was only known that the backbone of plasmid p737A1-27K-cfr was derived from pSTEC2018_607-F (CP075703.1, 24,412 bp). The 4,270 bp construct containing the IS26-cfr-IS26-higA-higB-parK was inserted into the plasmid pSTEC2018_607-F (Figure 4B).

Genome analysis found that both the upstream and downstream regions of the cfr gene in EC727A3 and EC737A1 had an IS26 element in the same direction, forming an IS26-cfr-IS26 structure (Figure 5A). However, there was a 343 bp size difference between the IS26-cfr-IS26 structures in EC727A3 and EC737A1. PCR determined that three out of eight E. coli strains could form cfr-IS26 cyclic intermediates of two different sizes. Among them, the size of the circular intermediate formed by ECJHZ21-173 and EC727A3 was the same, at 2,365 bp (Figure 5B). The size of the cfr-IS26 circular intermediate in EC737A1 was 2,022 bp (Figure 5C).