Between 2012 and 2016, 760 non-duplicate swabs were collected randomly from retail meat products (n = 520), porcine excrement (n = 68), and clinical samples (stools from patients with intestinal/respiratory diseases)(n = 172) from the provinces of Hebei (n = 473), Henan (n = 201), and Sichuan (n = 86) in China. Subsequently, the carbapenem-resistant strains were selected by MacConkey plates containing 2 μg/ml MEM and identified by 16S rRNA gene amplification and sequencing.

According to CLSI and EUCAST¹ recommendations, the minimal inhibitory concentrations (MICs) of the identified strains were determined for 17 antimicrobial compounds (Supplementary Table 1) conducted using the broth microdilution method. Escherichia coli ATCC™ 25922 and Klebsiella pneumoniae ATCC™ 700603 were included as quality-control strains. Susceptibility/resistance was interpreted according to the CLSI standard 2020 (Clinical and Laboratory Standards Institute [CLSI], 2020). MDR organisms were defined as acquired non-susceptibility to at least one agent in three or more antimicrobial categories. Extensively drug-resistant (XDR) organisms were defined as those that acquired non-susceptibility to all but two drug classes. Pan-drug-resistant (PDR) organisms were defined as those that acquired non-susceptibility to all drug classes.

All CRE strains were classified by ApaI nuclease-digested pulsed-field gel electrophoresis (PFGE; Ribot et al., 2006). Pulsotypes (PTs) were compared using BioNumerics version 6.6 (Applied Math, Sint-Martens-Latem, Belgium). PTs were interpreted using an 85% similarity cut-off. The XbaI nuclease-digested Salmonella enterica serotype Braenderup H9812 was used as a size marker.

Genomic DNA was sequenced using the Illumina HiSeq 2500 platform (Illumina, San Diego, CA, United States). Open reading frames (ORFs) were designated using the Rapid Annotation using Subsystem Technology (RAST) annotation pipeline². The ResFinder 4.1³ and PlasmidFinder 2.1⁴ databases from the Center for Genomic Epidemiology were also used to identify genomic features of interest. All the E. coli and K. pneumoniae isolates were subsequently classified for bacterial source tracking analysis by implementing a core genome SNP (cgSNP) strategy that was integrated with the BacWGSTdb 2.0 server⁵. An E. coli ST10 strain MG1655 (NCBI reference sequence: U00096) and a K. pneumoniae ST11 strain HS11286 (NCBI reference sequence: CP003200) were used as the reference genomes for all E. coli and K. pneumoniae sequence analyses, respectively. The resolution power of cgSNP analyses was assessed using Simpson’s diversity index. A phylogenetic tree was constructed using the resulting SNPs with recombination regions removed using the maximum parsimony algorithm. Annotation for each isolate and tree embellishment were visualized using iTOL⁶. The Global Optimal eBURST (goeBURST) full MST algorithm was used to compare against the PubMLST database, to highlight the potential relationships of different strain isolates and sequence types (STs) (Lu et al., 2020). The genetic information and evolutionary characteristics of all strains were considered. Nucleotide sequence analyses of 16S rRNA genes of all strains were conducted with MEGA (Lu et al., 2020). Virulence determinants were investigated by aligning the reads to the Virulence Factor Database (VFDB)^{7, 8}. In addition, MLST was defined as previously described (Francisco et al., 2009).

After sequencing, it was found that all strains carried plasmids, and carbapenem resistance genes were located on their plasmids. Then, all isolates were selected and analyzed individually for their ability to transfer the IPI/cefotaxime (CTX) resistance phenotype to a rifampicin-resistant, plasmid-free E. coli recipient (26R 793 rif^R), which was performed as previously described (Milton et al., 1996). In brief, donor and recipient cells were mixed at a ratio of 10:1 and spotted on LB at 37°C overnight. Transconjugants were selected on Mueller-Hinton (MH) agar plates containing rifampicin (100 mg/L) and IPI (1 mg/L) or CTX (4 mg/L). Susceptibility tests were performed on the transconjugants to confirm the resistance profile, followed by S1-PFGE to confirm plasmid transfer.

S1-pulsed-field gel electrophoresis analysis of successfully transferred CRE strains and transconjugants was performed as previously described (Bret et al., 1995). Briefly, plugs were made using a mid-log bacterial cell suspension in Tris-EDTA (TE) buffer and further treated with S1 nuclease (Thermo Fisher Scientific, Waltham, MA, United States). Switch times were ramped between 3 and 36 s for 18.5 h at 6 V/cm. The gels were stained with 1 mg/ml ethidium bromide (Sigma-Aldrich, St. Louis, MO, United States) in 0.5× TBE for 30 min in the dark and then washed with distilled water for 90 min to remove excess stains. Salmonella enterica serotype Braenderup H9812 was used as a size marker.

The nucleotide sequences of all 53 CRE strains have been submitted to the NCBI BioProject database⁹ under accession numbers SRR11517132–SRR11517184.

In this study, a total of 53 CRE strains were isolated from 760 samples including retail meat products, porcine excrement, and patients. HK and SK in strain names mean they were identified from samples collected from Zhengzhou in Henan province and Mianyang in Szechwan province, respectively. And SJ and SJE strains were collected from Shijiazhuang in Hebei province. Thirty-six isolates were cultured from retail meat products from Shijiazhuang and Zhengzhou, including 30 E. coli, 5 K. pneumoniae, and 1 Enterobacter cloacae. Six E. coli isolates were from the porcine excrement only from Mianyang. Twelve isolates were from respiratory secretions of patients with respiratory infection, including seven E. coli and four K. pneumoniae strains (Supplementary Table 2) from Shijiazhuang and Zhengzhou.

Susceptibility testing confirmed that 53 CRE strains showed resistance to both IPI and MEM. Few isolates showed resistance to polymyxin B (n = 3, 5.7%) and amikacin (n = 14, 26.4%), while the majority was resistant to ampicillin (n = 53, 100%), cefepime (n = 53, 100%), cefotaxime (n = 53, 100%), ceftazidime (n = 52, 98.1%), tetracycline (n = 50, 94.3%), and nalidixic (n = 50, 94.3%) (Table 1). Of the 53 strains, 46 (86.79%) were XDR Enterobacteriaceae (resistant to all we tested but polymyxins). And the other seven strains were PDR Enterobacteriaceae.

There are 14 different STs in the 43 E. coli isolates, with 30 retail meat product isolates clustered in 13 STs (ST10, ST48, ST101, ST156, SST167, ST354, ST457, ST746, ST1011, T2973, ST7109, ST7122, and ST7507), 7 E. coli isolates from patients clustered in 4 STs (ST101, ST167, ST354, and ST617), and 6 isolates from porcine excrement clustered in 1 ST (ST48) (Figure 1). Of 43 E. coli strains, 60.47% (Thanassi et al., 2012) was identified as belonging to ST156 (7/43), ST354 (7/43), ST48 (7/43), and ST167 (5/43). ST156 strains were mainly isolated from samples of retail meat products, while ST167 strains were isolated principally from samples of patients. Five of seven E. coli strains’ STs from patients were compatible with E. coli from retail meat products, particularly ST167 and ST354. In contrast, ST167 E. coli was mainly distributed in patients, and ST354 E. coli was mostly distributed in retail meat products. A total of 141,114 SNPs (minimum: 106; maximum: 70,504) were identified among the core genomes of the 43 isolates, on the basis of which the isolates could be allocated into least six major clades (Supplementary Figures 1, 2). The first clade includes 12 E. coli. Seven of them from retail meat products and five from patients. Clade 2 comprised six ST167 E. coli isolates, of which one was isolated from retail meat products and five isolated from patients. Clade 3 includes eight isolates, of which six isolates were from retail meats. Just one ST1011 E. coli isolate from retail meat products was identified as clade 4. Six unclassifiable isolates were included in clade 5.

Nine K. pneumoniae isolates covered five STs. The most prevalent ST was ST11 (5/9, 55.6%), which was all from patients (Figure 2). The distribution of STs in retail meat products showed diversity including ST307, ST340, and ST1311, and the ST of SJ13 was not defined. A total of 49,975 SNPs (minimum: 51; maximum: 24,889) were identified among the core genomes of the nine isolates, on the basis of which the isolates could be allocated into least three major lineages (Supplementary Figures 3, 4). Clades 1 and 2 comprised isolates from retail meat products, and clade 3 from patients.

We aimed to discover if there was a trend in the dissemination of CR-producing E. coli, so that the risks of transmission of different strains could be estimated. The goeBURST full MST algorithm used in the present study of E. coli categorized all STs into four clone complexes (CC), CC10, CC101, CC156, and CC354 (Figure 3). CC10 was the biggest group which harbored 5 STs and 15 E. coli strains. Next is CC354, which harbored ST354 with only seven E. coli strains, and CC156, which included six ST156 E. coli strains, all of which were collected from retail meat products. SJ16 of patients and SJ22 of retail meat products are ST101 E. coli belonging to CC101.

As for PTs, there were 26 PTs observed for 43 E. coli, among which 30 were E. coli isolates from retail meat products and clustered in 20 PTs, 7 E. coli isolates from patients and clustered in 5 PTs, and 6 isolates from porcine excrement and clustered in 5 PTs. A crossover between patients and retail meat products was observed in PT2 and PT3; PT2 E. coli (SJ25, SJ26, and SJ28) from patients belonged to ST167, while PT2 E. coli from retail meat products belonged to ST167 (SJ36) and ST354 (SJ29); similarly, PT3 E. coli from patients and retail meat products respectively belonged to ST354 (SJ31) as well as ST354 (HK11, HK13, and HK16) and ST746 (HK1). To some extent, strains of ST167 and ST354 had a certain similarity and relationship, both of them being associated with PT2. Strains from both porcine excrement (SK4) and retail meat products (SJ3, SJ12, SJ7, SJ15, and SJ20) were associated with PT17, but their STs were not definite (Figure 1). Nine K. pneumoniae isolates were divided into six PTs: three PT5 strains from patients (HK19, SJ17, and SJ24) and two PT6 strains from patients (SJ11 and SJ18) belonged to ST11 K. pneumoniae; the other four isolates from retail meat products of different STs each had four different PTs (Figure 2).

bla_NDM was the most common carbapenemase gene type among the CRE strains, including bla_NDM–5 (42/53, 79.25%), bla_NDM–9 (4/53, 7.55%), bla_NDM–1 (1/53, 1.89%), and bla_NDM–7 (1/53, 1.89%), suggesting bla_NDM as the key gene mediating the development of the carbapenem resistance phenotype in this study. The bla_NDM–5 gene was found in 39 E. coli (26 strains from retail meat products, 6 from porcine excrement, and 7 from patients) and 3 K. pneumoniae [2 strains from retail meat products and 1 strain (SJ24) from patients]. Three bla_NDM–9 from E. coli and one bla_NDM–9 from K. pneumoniae were all from retail meat products, as well as bla_NDM–1 from E. coli and bla_NDM–7 from K. pneumoniae.

The gene encoding K. pneumoniae carbapenemase (KPC) has been detected in five ST11 K. pneumoniae isolates collected in the hospital, with each being bla_KPC–2. Since its identification in North Carolina, United States, in 1996, in a clinical K. pneumoniae isolate, the epidemiology of bla_KPC has expanded globally, especially the clonal group (CG) 258, which includes the lineages ST258 and ST11. SJ24, carrying bla_NDM–5 and bla_KPC–2 simultaneously, had nine different plasmid replicon types, which was the largest number of plasmid replicon types in this study.

All isolates harbored beta-lactamase gene(s), aminoglycosides gene(s), and sulfonamide resistance gene(s). Most isolates harbored fluoroquinolone resistance genes(s), fosfomycin resistance gene(s), macrolide resistance gene(s), phenicol resistance gene(s), tetracycline resistance gene(s), and trimethoprim resistance gene(s) (Supplementary Table 3). Additionally, a few isolates harbored streptogramin B, polymyxin B, and/or rifampicin resistance genes. One out of the four K. pneumoniae from meat samples harbored the AmpC beta-lactamase bla_CMY–2. Four E. coli harbored a macrolide–lincosamide–streptogramin B resistance gene [erm(42) or erm(B)].

In our study, the polymyxin B resistance gene mcr-1 was detected in 17/43 (40.5%) E. coli isolates, coexisting with bla_NDM (bla_NDM–5 and mcr-1 14/17, bla_NDM–1 and mcr-1 1/17, and bla_NDM–9 and mcr-1 2/17). Only one came from porcine excrement (harboring bla_NDM–5 and mcr-1); the rest of the 17 strains were all collected from retail meat products (Figure 1).

Virulence genes have also been tallied and collated (Supplementary Table 4). Forty-six, 27, and 15 virulence genes were detected from the E. coli strains of retail meat products, patients, and porcine excrement, respectively. Among the 16 virulence genes which represent the main categories of virulence determinants (sfa, papC, sepA, etrA, aer, feaG, fsaA, eaeA, rfc, cnf1, hlyA, eltA, estA, exhA, stx1, and stx2), we only detected the papC gene in 10 E. coli, which encode the protein of pilus associated with pyelonephritis and belong to ExPEC. In addition, irp2 (iron-repressible protein gene) was detected from 6 E. coli, cva (colicin V plasmid operon gene) was detected from 17 E. coli, astA (enteroaggregative toxin gene) was detected from 12 E. coli, tsh (temperature-sensitive hemagglutinin gene) was detected from 1 E. coli, iucD (ferric aerobactin receptor gene) was detected from 22 E. coli, and iss (increased serum survival protein gene) was detected from 27 E. coli.

Virulence factor profiling of the nine K. pneumoniae isolates revealed the presence of 98 virulence factors. The fim gene cluster (fimABC, etc.), mrk genes (A, B, C, D, F, H, I, and J), fep genes (A, B, C, D, and G), ent genes (A, B, C, D, E, F, and S), tss genes (A, B, C, D, F, G, H, I, J, K, L, and M), iro genes (D, E, and N), ybt genes (A, E, P, O, S, T, U, and X), iuc genes (A and C), iutA gene, and rmpA gene were highly represented.

None of the nine K. pneumoniae plasmids and one E. cloacae plasmid could be transferred by conjugation in three attempts.

Forty-three E. coli strains (seven strains from patients, six from porcine excrement, and the rest from retail meat products) were tested for their ability to transfer IPI and CTX resistance phenotypes by conjugation. A total of 34 transconjugants were obtained. All the transconjugants were resistant to AMP and CTX. Of 34 transconjugants, 30 were resistant to IMI and MEM. After PCR identification, there were 30 transconjugants harboring the bla_NDM gene; there were 8 transferred donor strains without bla_CTX–M genes but whose transconjugants were resistant to CTX, and their resistance to CTX may be mediated by other ESBLs. mcr-1 was detected in four transconjugants, of which two were recovered from the same donor (SJ22 carries mcr-1, bla_NDM–5, and bla_TEM–1B) (Figure 4 and Table 2).

S1-nuclease plasmid analysis revealed that all 27 transferred donor strains contained large detectable plasmids: mostly possessing three plasmids (n = 15, 55.56%), two plasmids (n = 6, 22.22%), four plasmids (n = 4, 14.81%), and five plasmids (n = 2, 7.21%). Compared with those found in the S1-nuclease plasmid analysis in donors, the numbers of plasmids carried by 29/34 transconjugants were reduced (Figure 4).

In this study, the most prevalent plasmid replicon types in the 43 E. coli were IncFIB (35/53, 81.40%, 27 strains were from retail meat product samples, 3 strains were from patient samples, and 5 strains were from porcine excrement), IncX3 (34/43, 79.07%, 25 strains were from retail meat product samples, 3 strains were from patient samples, and 6 strains were from porcine excrement), and IncFIC (19/43, 44.18%, 17 strains were from retail meat product samples, 1 strain was from patient samples, and 1 strain was from porcine excrement). IncFIB (7/9, 5 from patients and 2 from retail meat products), IncFII (as well as IncFIB), IncR (5/9, 3 from patients and 2 from retail meat products), ColRNAI (5/9, all of them collected from patients), and IncX3 (4/9, 1 strain collected from patients and others from retail meat products) were the dominant plasmid replicon types of K. pneumoniae. Except for IncFIB and IncX3, the plasmid types of E. coli and K. pneumoniae were distinct (Figure 5).

A total of 3 genetic contexts (type 1 to type 3) were found in 36/43 bla_NDM–5-carrying plasmids (Figure 6). Type 1 was found in 14 bla_NDM–5-carrying E. coli from retail meats, 5 bla_NDM–5-carrying E. coli from porcine excrements, 3 bla_NDM–5-carrying E. coli from patients, and 1 bla_NDM–7-carrying E. coli from retail meat products. In type 1, ISCR1 was located downstream of bla_NDM–5, same with plasmid pYJ6 (AP023236.1) of E. coli. Type 2 was found in plasmid IncFIB of eight bla_NDM–5-carrying E. coli from retail meat products, same with plasmid pMTY18781-5 (AP023210.1) of E. coli, which was a 46,161-bp IncFII plasmid. In type 2, ISAba125 located between IS5 and IS3000 upstream of bla_NDM–5 may play an important role in recombination of plasmids by ISAba125 duplication. Type 3 was similar to type 1, except without sul1, ant(3″)-Ia, and dfrA12 genes, which were identified from one bla_NDM–5-carrying E. coli from retail meat products and four bla_NDM–5-carrying E. coli from patients. The type 1 genomic context bla_NDM–5-ble-trpF-dsbD-ISCR1-sul1-ermE-ant(3″)-Ia-orfF-dfrA12-IntI1 was the most prevalent bla_NDM–5 genetic environment in this study (63.9%, 23/36).

In six bla_NDM–1/9-carrying plasmids, there were four genetic contexts (type 1 to type 4) (Figure 7). Type 1 was found in two bla_NDM–9-carrying E. coli from retail meat products, same with plasmid p0058 (MN577015.1) of Salmonella enterica, which was mainly carried by IncFIB plasmids which obtained ISAba125 and ISCR1 elements. This genomic context was similar to type 1 of bla_NDM–5, except that there was an ISAba125 element. In type 2, they carried the same IS elements but without ermE, ant(3″)-Ia, and orfF genes. There were one bla_NDM–9-carrying K. pneumoniae and one bla_NDM–1-carrying E. cloacae from retail meat products, which belonged to type 2. Types 3 and 4 are similar; neither of them had a sul1-ermE-ant(3″)-Ia-orfF-dfrA12 genomic context, but type 4 had a truncation trpE and did not have dsbD and cutA genes. One bla_NDM–9-carrying E. coli and one bla_NDM–1-carrying E. coli from retail meat products belonged to types 3 and 4, respectively.

Three genetic contexts (type 1 to type 3) were found in five bla_KPC–2-carrying plasmids (Figure 8), all of which were ST11 K. pneumoniae from patients. Two strains were classified as type 1, which was the same with a 108,019-bp plasmid p12478-KPC (MT108212.1). Tn3 and ISKpn27 were located upstream of bla_KPC–2, and ISKpn6 was located downstream of bla_KPC–2. Type 2, including two ST11 K. pneumoniae strains, lacked tnpA and rep genes compared with type 1. And type 3 lacked a Tn3 gene compared with type 2, including one ST11 K. pneumoniae strain.

Three genetic contexts (type 1 to type 3) were found in 11/14 mcr-1-carrying plasmids (Figure 9). Type 1 was found in six mcr-1-carrying E. coli from retail meat products. In type 1, ISApl1 was located upstream of mcr-1 and an insertion sequence of ISApl1 in plasmid p0111. In type 2, it was similar to type 1, except that it had traE, which may play an important role in conjugation of plasmids. There were three mcr-1-carrying E. coli from retail meat products, which belonged to type 2. Type 3 was found in one mcr-1-carrying E. coli from retail meat products and one mcr-1-carrying E. coli from porcine excrement, same with plasmid pE105-5 (CP072316.1) of E. coli, which was mainly carried by IncX4 plasmids.