We collected samples from 10 rabbit farms across Sichuan province for a total of 187 samples. These included swabs from healthy rabbits as well as environmental samples, including anal swabs, nasal swabs, water, feed, cages, feces and sewage. The specific number of samples of each type collected and the corresponding cities are detailed in Table 1 . To collect the rabbit anal samples, a sterile cotton swab was gently inserted into the rabbit’s anus and carefully rubbed against the rectal mucosa. Subsequently, the swab was placed into a sterilized container containing brain heart infusion medium. To collect environmental samples, a swab was moistened in brain heart infusion medium in a tube. Next, the environmental surfaces were systematically swabbed, including cages, feed troughs, and floors, ensuring thorough coverage. Finally, the swab was returned to the tube.

The collected samples were pre-cultured in an incubator at 37°C, after which a single loop of each bacterial solution was streaked on a MacConkey agar plate. The plates were cultured in an incubator at 37°C for 24 h. Then, single pink clones were selected and incubated on trypticase soy agar plates. A pure and single colony was carefully aspirated with a sterile pipette tip and gently deposited onto a clean MALDI target plate to form a thin even layer. Subsequently, 1 μL of α-cyano-4-hydroxycinnamic acid matrix solution was accurately dispensed onto each spotted sample and allowed to dry naturally at room temperature. The MALDI target plate was then immediately transferred to a MALDI-TOF mass spectrometer (Bruker) for isolate identification. To avoid strain duplication, a single representative strain was retained from each sample.

The minimum inhibitory concentration (MIC) was determined according to the Clinical and Laboratory Standards Institute (CLSI) M100-33rd guidelines using 96-well plates to test the 17 antimicrobials: nalidixic acid, ciprofloxacin, colistin, tigecycline, tetracycline, chloramphenicol, azithromycin, trimethoprim-sulfamethoxazole, amikacin, streptomycin, ampicillin, cefotaxime, ceftazidime, ceftazidime-avibactam, ampicillin-sulbactam, meropenem, and ertapenem. Single and pure isolated colonies were picked to prepare a 0.5 McFarland standard bacterial suspension. Subsequently, the bacterial suspension was diluted 100-fold using Müller–Hinton broth. Then, the bacterial suspension was added to a 96-well plate manufactured by Meihua Company (China) that had been preloaded with concentration gradients of antimicrobial drugs. The Escherichia coli ATCC^® 25922 reference strain served as the quality control strain throughout the study. MIC determinations were conducted in triplicate for each clinical isolate. The plates were incubated at 37°C for 18–20 h. The resistant phenotype of the isolates was determined according to the MIC breakpoint criteria outlined in the CLSI M100-33rd guidelines.

In the conjugation assay, Ec-A21, Ec-A24, Ec-A29, Ec-JB2, and Ec-CD45 isolates that exhibited colistin MIC values exceeding 4 mg/L served as donor strains, while E. coli J53 functioned as the recipient strain. Donor and recipient strains were cultured to logarithmic growth and subsequently mixed at a 1:1 volumetric ratio (0.4 mL of the donor and 0.4 mL of the recipient). Following static incubation for 10 min, 80 μL of the bacterial suspension (40 μL of the donor and 40 μL of the recipient) was aseptically transferred onto sterile 0.22-μm nitrocellulose membranes placed on trypticase soy agar plates and incubated for 12 h at 37 °C. Post-incubation, all cultures (donors, recipients, and conjugation mixtures) were washed with phosphate-buffered saline. Transconjugant selection was performed using Müller–Hinton agar supplemented with 100 mg/L sodium azide and 4 mg/L colistin. Donor viability was quantified by plating serial dilutions on sodium azide-containing agar (100 mg/L). PCR amplification of mcr-1 (primer sequences, F: 5'-CGG TCA GTC CGT TTG TTC-3' and R: 5'-CTT GGT CGG TCT GTA GGG-3') was performed (Liu et al., 2016). The conjugation transfer frequency was calculated by dividing the number of transconjugants by the number of recipients.

Whole-genome sequencing of the isolated strains was conducted by Majorbio Bio-pharm Technology (China) using the Illumina platform. The sequencing procedure was as follows: Initially, total genomic DNA was extracted from the isolated strains using a bacterial genomic DNA extraction kit. The extracted genomic DNA was then fragmented using Covaris technology, and a genomic sequencing library was constructed. Draft genomes were generated on the Illumina sequencing platform. Sequencing libraries with insert sizes of approximately 400 bp were constructed using only DNA samples that met stringent quality control standards. The libraries were subsequently subjected to paired-end sequencing with a read length of 150 bp in each direction. This process generated raw sequencing data with a minimum coverage depth of 100× across the genome. SOAPdenovo 2.04 software was used for genome assembly, leading to the construction of multiple scaffolds.

Three E. coli complete genomes were obtained using an Oxford Nanopore Technologies (ONT) system in combination with Illumina genome data. This was performed by Biomaker Technology Company (China). The experimental procedure was conducted in accordance with the standard protocol provided by ONT, which includes sample quality assessment, library preparation, library quality evaluation, and sequencing. The main steps were as follows: high-quality genomic DNA was extracted using bacterial genome extraction kits and subsequently assessed for purity, concentration, and integrity using Nanodrop, Qubit, and 0.35% agarose gel electrophoresis; large DNA fragments were size-selected and recovered using the BluePippin fully automated nucleic acid recovery system; library construction was carried out using the SQK-LSK109 ligation kit, followed by sequencing. To assemble the genome, the filtered reads were first assembled using Canu v1.5 software, followed by circularization of the assembled genome using Circlator v1.5.5. For functional annotation, the predicted proteins were compared against the Nr, Swiss-Prot, TrEMBL, KEGG, and eggNOG databases using BLAST with an e-value threshold of 1e−5. The Escherichia coli strains Ec-JB2 and Ec-CD45 were subjected to whole-genome sequencing using ONT sequencing system. Both strains contained five plasmids, with plasmids pEc-JB2-5 (GenBank accession: CP182207) and pEc-CD45-5 (GenBank accession: CP182224) of particular interest as they carried the mcr-1 colistin resistance gene.

Isolate identification was validated using conserved housekeeping genes via the Majorbio cloud platform (Ren et al., 2022). Subsequently, the genomes were uploaded to KmerFinder 3.2 (https://cge.food.dtu.dk/services/KmerFinder/) and subjected to BLAST analysis to identify isolates. ResFinder 4.7.2 (http://genepi.food.dtu.dk/resfinder) and the Comprehensive Antibiotic Resistance Database (https://card.mcmaster.ca/) was used to predict ARGs, and the Virulence Factor Database (Liu et al., 2022) was used to predict virulence factor genes (VFG). Both analyses used a BLAST nucleotide identity threshold of ≥90% and length coverage ≥90%. The Center of Genome Epidemiology MLST 2.0 tool (https://cge.food.dtu.dk/services/MLST/) was used to predict the sequence types (ST) of the isolates. Pathogenwatch (https://pathogen.watch/) was used for Klebsiella pneumoniae and Salmonella enterica serotype prediction, and ClermonTyping (http://clermontyping.iame-research.center/) was used to predict E. coli phylogroups. Proksee (https://proksee.ca/) was used to annotate resistant plasmids, the genome sequences were annotated using Prokka, and mobileOG-db (beatrix-1.6) was used to find mobile genetic elements (MGE) (Seemann, 2014; Brown et al., 2022). BacWGSTdb (http://bacdb.cn/BacWGSTdb/index.php) was utilized for phylogenetic analysis of E. coli based on core single nucleotide polymorphisms (SNP) (Feng et al., 2021), with E. coli MG1655 selected as the reference genome.

TBtools v2.210 was used to generate heatmaps of the ARGs and VFGs (Chen et al., 2023). Office 2021 Excel was used to collect and process data in tables. BLAST Ring Image Generator (BRIG) V0.95 was used to analyze the resistant plasmid homology. NCBI BLAST v2.16.0 was used for the local alignment of plasmid sequences. GraphPad Prism 8.0 was used to create column charts. Evolview 2.0 was used to modify the phylogenetic tree (He et al., 2016). Proksee was used to visualize the map of resistant plasmids (Grant et al., 2023).

All genome sequences were uploaded to NCBI and whole-genome shotgun data was deposited in GenBank under at Bioproject PRJNA1223317. Data will be made available on request.

A total of 187 samples were collected from four cities in Sichuan province, China (Chengdu, Leshan, Rongxian, and Zigong), originating from ten rabbit farms. The samples included anal swabs, nasal swabs, feces, water, feed, floor, cages, and sewage. A total of 73 Enterobacteriaceae strains were isolated using MacConkey selective culture and MALDI-TOF mass spectrometry, comprising 3 Klebsiella pneumoniae strains, 3 Salmonella enterica strains, 6 Enterobacter hormaechei strains, and 61 Escherichia coli strains. Detailed information on the collected samples and isolates is provided in Table 1 .

Antimicrobial susceptibility profiles of the isolates were determined against 17 antimicrobials spanning seven therapeutic classes. As depicted in Figure 1 , five antimicrobials demonstrated resistance rates exceeding 60%: tetracycline (78%, 57/73), ciprofloxacin (74%, 54/73), nalidixic acid (68.5%, 50/73), sulfamethoxazole-trimethoprim (68.5%, 50/73), and ampicillin (60.3%, 44/73). Moderate resistance was observed for chloramphenicol (46.6%, 34/73) and streptomycin (37%, 27/73). Notably, emerging resistance to last-line antibiotics was detected, with 17.8% (13/73) of isolates demonstrating colistin resistance and 6.9% (5/73) showing reduced tigecycline susceptibility. In addition, the resistance rates to cefotaxime, ceftazidime, and ampicillin-sulbactam were 11% (8/73), 4.1% (3/73), and 11% (8/73), respectively, suggesting that these isolates may include Extended-spectrum β-lactamase (ESBL)-producing strains. Azithromycin resistance occurred in 12.3% (9/73) of strains, while two isolates exhibited resistance to amikacin. Importantly, all isolates were susceptible to carbapenems (meropenem and ertapenem) and the novel β-lactamase inhibitor combination ceftazidime-avibactam. Multidrug resistance (MDR), defined as resistance to ≥3 antimicrobials, was observed in 86.3% (63/73) of isolates ( Supplementary Figure S1 ). The antimicrobial resistance profiles of the isolates are summarized in Supplementary Table S1 , which indicates that certain individual isolates demonstrate resistance to as many as 11 antimicrobial agents. One isolate exhibited pan-susceptibility, while the majority of isolates (75.3%, 55/73) displayed resistance to 3–7 antimicrobial classes. Supplementary Table S1 shows the MICs and resistance profiles of all isolates.

Multilocus sequence typing of the 61 Escherichia coli isolates identified 16 distinct sequence types, with three predominant clones collectively representing 52.4% of the population: ST328 (26.2%, 16/61), ST224 (13.1%, 8/61), and ST297 (13.1%, 8/61). Supplementary Table S2 shows all isolate STs. Serological profiling revealed 22 unique O:H serovars, including four strains (6.6%) with unknown O antigens. Supplementary Table S3 shows all E. coli serotype alignment results. The most prevalent serovars were O178:H7 (19.7%, 12/61), O1:H10 (9.8%, 6/61), and O172:H3 (9.8%, 6/61). Using phylogenetic grouping analysis, the isolates were classified into five phylogroups: A, B1, B2, C, and E, with the majority belonging to phylogroup B1 (78.7%, 48/61). Published reference has reported that B1 strains are predominant in domestic and wild animals (Tenaillon et al., 2010). Supplementary Table S4 shows all E. coli phylogroups alignment results. All three strains of Salmonella enterica isolate belonged to ST426, with a serotype classification of Aberdeen. The Klebsiella pneumoniae collection (n = 3) included two STs: ST1876 (n = 2) and ST294 (n = 1), with serotyping identifying one O1ab:K30 strain. The remaining two isolates displayed O13 serotype compatibility, with the H serotype not classified, indicating a possible capsular antigenic variation or genetic deletion in the K locus. The phenotypes predicted from the isolates’ genomic data are presented in detail in Table 2 .

The ARGs identified in the Enterobacteriaceae strains included the following types of resistance: polymyxin resistance (mcr-1.1), tetracycline resistance (tet(A)), fluoroquinolone resistance (qnrS1, oqxA, oqxB), sulfonamide and diaminopyrimidine resistance (sul1, sul2, dfrA12, dfrA17), extended-spectrum β-lactamase production (bla _CTX-M), chloramphenicol resistance (catB3, folR), aminoglycoside resistance (aac(3)-IV, aac(6’)-Ib-cr, aph(3’)-Ia, aph(6)-Id), macrolide resistance (mph(A)), and fosfomycin resistance (fosA). A subset of the detailed results is presented in Figure 2 , with Resfinder database annotations summarized in Supplementary Table S5 and CARD database annotations in Supplementary Table S6 . The plasmid replicon predictions found that IncFIB, IncFII, IncHI2, Incl2, and IncX1 were the primary replicon types. The annotated results are shown in Supplementary Table S7 . As shown in Figure 2 , the mcr-1.1 gene was detected in 35 isolates; among these, 29 isolates harbored the IncI2 plasmid replicon.

The VFDB database was used to annotate virulence genes encompassed: vgrG/tssl, espL1, fimC, fimD, fimH, cgsG, csgA, csgC, yagW/ecpD, ompA, clbK, clbJ, cheY, phoP, rcsB, rpoS, gndA, pic, escV and iroN, which associated with E. coli adherence, invasion, effector delivery system, exotoxin, motility, and regulation function. The carriage status of major VFGs for each isolated strain was presented in Supplementary Figure S2 , while the original annotation data are provided in Supplementary Table S8 .

Plasmid pEc-JB2–5 is 64,108 bp in length and belongs to the IncI2 replicon type. Plasmid pEc-CD45–5 is 80,958 bp in length and also belongs to the IncI2 replicon type. Both plasmids harbor an extensive array of mobile genetic elements, which are organized into four functional clusters: integration and excision elements (tnp, xerC); replication, recombination, and repair systems (nikB, topB, yhcR, parA, repA-1); conjugative transfer apparatus (virB1–virB11 operon); and plasmid stability and defense mechanisms (relE toxin-antitoxin system, plasmid conjugative transfer pilus pilP–pilQ, and tcpE). Notably, the mcr-1.1 gene was positioned between a PAP2 family hydrolase gene and the nikB relaxase, a critical enzyme mediating plasmid conjugation through single-strand DNA processing. This genetic architecture, in which antibiotic resistance determinants were flanked by conjugation-associated elements, suggested that it may enhance horizontal dissemination. Plasmid pEc-JB2–5 architecture and MGE organization are schematically depicted in Figure 3A , with distinct color-coding to differentiate functional modules. A comparative genomic analysis revealed that pEc-JB2–5 exhibited similarity with clinically relevant plasmids from various host species ( Figure 3B ), nucleotide identity threshold of ≥90% and length coverage ≥90%. These plasmids include pE2865-4 (origin: cattle; geographic location: Japan; size: 62,235 bp; accession number: NZ_AP018812.1); an unnamed plasmid (origin: pig; geographic location: Henan, China; size: 142,379 bp; accession number: NZ_CP137738.1); pMCR-M19242 (origin: human; geographic location: Canada; size: 61,632 bp; accession number: NZ_KY471312.1); Sh487-m4 (origin: human; geographic location: Shanghai, China; size: 63,512 bp; accession number: NZ_KY363996.1); and an unnamed nosocomial infection-associated plasmid (origin: human; geographic location: China; size: 62,440 bp; accession number: NZ_KX580716.1). The pEc-CD45–5 plasmid was also analyzed in the same method, with the detailed results shown in Supplementary Figure S3 . pEc-CD45–5 exhibited homology with pPSS-08-2_3(origin: human; geographic location: Ecuador; size: 60,961 bp; accession number: NZ_AP027682.1), pPSS-16_2 (origin: human; geographic location: Ecuador; size: 60,960 bp; accession number: NZ_AP027715.1), pHLJ109-70 (origin: chicken; geographic location: China; size: 61,023 bp; accession number: NZ_MN232201.1), pHLJ111-18 (origin: chicken; geographic location: China; size: 60,962 bp; accession number: NZ_MN232205.1), pHLJ111-5 (origin: chicken; geographic location: China; size: 61,094 bp; accession number: NZ_MN232208.1), and pSC111 (origin: human; geographic location: China; size: 60,960 bp; accession number: NZ_MZ277864.1)The cross-species homology and transcontinental distribution of these plasmids, spanning cattle, swine, and human hosts, suggested that pEc-JB2–5 may represent a high-risk mobile genetic element. This finding again highlighted the potential for interspecies transmission of colistin resistance determinants within “One Health” ecosystems.

As shown in Figure 2 , the strains Ec-JB2, Ec-CD45, Ec-A21, Ec-A24, and Ec-A29, identified as Escherichia coli, Klebsiella pneumoniae, and Salmonella enterica, harbored the mcr-1.1 gene. These isolates harbored multiple plasmid replicons: Ec-JB2 carried IncFIB, IncFIC, IncI1-I(Alpha), IncI2, and IncX1; Ec-CD45 carried IncFIB, IncFII, IncI2; Ec-A21 carried IncI2 and IncFII; Ec-A24 carried IncI2, IncFII, and repB; and Ec-A29 carried IncI2 and IncHI2. All of them contained the IncI2 plasmid replicon. Detailed information on the specific plasmids is provided in Supplementary Material Supplementary Table S7 . A conjugation experiment demonstrated that all five strains were able to transfer the mcr-1.1 gene to a recipient strain, E. coli J53. The conjugation transfer frequency was performed in Supplementary Table S9 . The genetic environments of mcr-1.1 shown in Figure 4 . Amplification of the mcr-1 gene in donors, recipients, and transconjugants is illustrated in Supplementary Figure S4 . As depicted in Figure 4 , both copies of the mcr-1 gene were flanked by a relaxase and PAP2.

Whole-genome sequences of 48 human-derived mcr-1-carrying Escherichia coli strains were systematically retrieved from the NCBI database. These were subjected to comparative phylogenetic analysis with 32 of the rabbit-derived mcr-1-harboring E. coli strains investigated in the current study. The resulting phylogenetic reconstruction demonstrated some rabbit-origin Escherichia coli carrying mcr-1 and human-origin Escherichia coli are on the same evolutionary branch, as marked by the shaded area in Figure 5 . The phylogenetic clusters derived from human reservoirs are highlighted in pink, while strains originating from rabbit specimens in this investigation are demarcated in light blue. They did not form two distinct branches as initially hypothesized, with the rabbit-origin and human-origin strains each forming independent lineages. Notably, all the strains analyzed also carried additional antimicrobial resistance determinants, including tet(A) conferring tetracycline resistance, folR associated with sulfonamide resistance, and the multidrug efflux pump gene mdf(A).