Antimicrobial resistance and population genomics of multidrug-resistant Escherichia coli in pig farms in mainland China

The expanding use of antimicrobials in livestock is an important contributor to the worldwide rapid increase in antimicrobial resistance (AMR). However, large-scale studies on AMR in livestock remain scarce. Here, we report findings from surveillance of E. coli AMR in pig farms in China in 2018–2019. We isolated E. coli in 1,871 samples from pigs and their breeding environments, and found AMR in E. coli in all provinces in mainland China. We detected multidrug-resistance in 91% isolates and found resistance to last-resort drugs including colistin, carbapenems and tigecycline. We also identified a heterogeneous group of O-serogroups and sequence types among the multidrug-resistant isolates. These isolates harbored multiple resistance genes, virulence factor-encoding genes, and putative plasmids. Our data will help to understand the current AMR profiles of pigs and provide a reference for AMR control policy formulation for livestock in China.

A ntimicrobial resistance (AMR) is one of the most urgent threats to global public health. Recently, the emergence of and rapid increases in multidrug-resistant (MDR), extensively drug-resistant (XDR), and even pandrug-resistant (PDR) bacteria, particularly bacteria resistant to last-resort drugs (carbapenems, colistin, and tigecycline), have catalyzed serious concern 1-4 . Several drug-resistant bacteria are of great concern worldwide, among which Escherichia coli is particularly notable 1,5,6 . This bacterial species is not only a leading cause of foodborne infections but also represents a major reservoir of antimicrobial resistance genes (ARGs) due to its high capacity to accumulate ARGs, mostly through horizontal gene transfer 7 . A recent study showed that the total economic cost of AMR in Staphylococcus aureus, E. coli, Klebsiella pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa has reached $2.8 billion per year in the United States 8 . In addition, E. coli has been found to play important roles in the dissemination of bla NDM-1 , mcr, and/or tet(X3)/(X4); these ARGs mediate resistance to lastresort drugs (carbapenems, colistin, and tigecycline) in gramnegative bacteria, which may lead to unavailability of antimicrobials in both human and veterinary medicine 3,4,9 . In addition, extended-spectrum beta-lactamase (ESBL)-producing E. coli is also a major global health problem 1,5 . Therefore, E. coli has been commonly used as a biomarker to monitor AMR on livestock farms and/or in hospitals [10][11][12] .
The widespread and expanding use of antimicrobials in food animals is considered one of the main reasons for the worldwide rapid increase in AMR 13 . In the past two decades, meat production has plateaued in high-income countries but has greatly increased in low-and middle-income countries, particularly in Asia 14 . Against this background, antimicrobials are widely used in livestock for growth promotion and/or health maintenance. Globally, 73% of all antimicrobials sold are used in animals raised for food 14 . The widespread and expanding use of antimicrobials can lead to the development of drug-resistant bacteria in animal guts, and these resistant bacteria can spread to humans 1 . To date, AMR is poorly documented in animals in low-income and middle-income countries, and this may be partly due to the absence of systematic surveillance systems and large epidemiological studies 14 .
The domestic pig (Sus scrofa domesticus) is one of the most important human food-producing animals, as over 40% of all meat consumed on earth comes from pigs 15 . Geographically, pigs are farmed in many regions around the world, but more than 60% of the world's pigs are found in Asia, particularly eastern Asia and southeastern Asia 16 . Recently, the analysis of E. coli isolates from pigs and other animals by antimicrobial susceptibility testing (AST) in combination with whole-genome sequencing (WGS) has facilitated a better understanding of AMR development and dissemination [17][18][19][20] . For example, a recent study showed that plasmid-borne mcr-1 is responsible for the rapid dissemination of polymyxin resistance in Enterobacteriaceae in animals and humans 4 . Another recent study revealed that IncX4, IncI2, and IncHI2 were the common plasmid types carried by mcr-1-positive E. coli on pig farms in China after the cessation of colistin use as an animal feed additive, and significant increases in IncI2-associated mcr-1 and a distinct lineage of mcr-1associated IncHI2 were observed after this agent was banned 21 . Based on WGS data, many studies have also shown that E. coli sequence type (ST) 10 and its related STs are important AMR-associated genotypes that predominate in swine, and a higher prevalence of plasmid-borne ARGs is found in these types than in others 7,[21][22][23][24][25] .
China is the largest pig-rearing country in the world and many provinces include low-income and middle-income regions. China is also the top antimicrobial-producing, and pig-producing, and pork-consuming country worldwide 26,27 . Therefore, understanding the current profile of AMR on pig farms in China is of great significance for global AMR surveillance. However, pig production in China is very complex, involving different sizes of pig farms and production models in various regions. There remains a lack of systemic data reflecting the condition of AMR in the pork production chain in China. Here, we identified AMR on pig farms in most regions of China by investigating E. coli, which is a commonly used biomarker of AMR on pig farms 10 . To the best of our knowledge, this is the first time that pig farms in all regions of mainland China have been included in an epidemiological investigation. This study will contribute to understanding the current profile of AMR in pigs and provide a reference for the formulation of AMR control policies for livestock in China.

Results
Antimicrobial-resistant phenotypes. Altogether, a total of 1871 E. coli isolates were recovered from pigs or their breeding environments in 31 Chinese provinces between 1 October 2018 and 30 September 2019 ( Fig. 1a and Supplementary Fig. 1). The most represented provinces by isolate count were Henan (n = 191) and Hubei (n = 250), which are the two largest pigfarming provinces in China. Initially, we attempted to include more farms in each of the provinces, but the sudden outbreak of African swine fever (ASF) in late 2018 and coronavirus disease  in late 2019 made it extremely difficult to obtain more samples from additional farms.
The AST results also revealed that resistance to tetracycline, chloramphenicol, moxifloxacin, and trimethoprim/sulfamethoxazole was a common phenotype of the isolates from the pig farms in all 31 provinces of mainland China (Fig. 1c). Carbapenemresistant isolates were found in seven provinces, but the majority came from Henan Province; isolates resistant to colistin were found in 12 provinces, and most of these isolates were also recovered from Henan Province (Fig. 1c). Tigecycline-resistant E. coli were found in 28 provinces, including Tibet, but highly tigecycline-resistant strains (MIC value ≥4 µg/ml) were only identified in Anhui, Hunan, Guizhou, Hebei, and Hubei (Fig. 1c). E. coli strains resistant to broad-spectrum cephalosporins (ceftriaxone and cefepime) were isolated from pig farms in 30 and 24 provinces, respectively (Fig. 1c). Tibet was the only region where no strains from pig farms with the abovementioned resistant phenotypes were detected (Fig. 1c). Notably, MDR isolates were identified on pig farms in 31 Chinese provinces, but a relatively high proportion of the MDR isolates were identified on farms in Sichuan (69.44%) and Tibet (42.86%)  Among the 118 identified STs, ST10 (n = 52), ST101 (n = 39), and ST48 (n = 22) were the most prevalent (Fig. 2b).
Resistance phenotypes generally corresponded to carriage of one or more genes known to confer that phenotype (Fig. 3). For example, ESBL genes were associated with resistance to broadspectrum cephalosporins. In addition, only carbapenem-resistant isolates contained the three carbapenem-resistance genes, and only colistin-resistant isolates carried the three mcr genes. Although different tetracycline resistance genes (tetA, tetB, tetC, and tetM) were found in tetracycline-resistant isolates, tetX4 was only identified in isolates with a high level of resistance to tigecycline (MIC value ≥4 µg/ml). Notably, over 15% of NDMproducing isolates carried mcr-1, and only one isolate contained mcr-1 and tetX4 (Fig. 3).
Class I integrons and plasmids. To understand the genetic basis of ARG dissemination, we first detected the presence of the class 1 integrase gene intI1 in the genomes of the isolates, and 491 isolates were found to harbor intI1 (Supplementary Data 5). More than half of the isolates (61.71%, n = 303) possessed a single intI1 gene, and 52 isolates harbored three intI1 genes. Interestingly, more than half (71.15%, n = 37) of the ST10 isolates harbored only one intI1 gene, while only three ST10 isolates possessed three intI1 genes. We next investigated the presence of plasmids in the wholegenome sequences of these drug-resistant isolates, and a total of 53 groups of plasmid replicons were detected (Supplementary We then generated the complete genome sequences of the XD33 isolate producing both NDM and MCR (GenBank accession no. JAENDM000000000) through Oxford Nanopore Sequencing (ONT). This approach identified an 85.9 kb IncFII plasmid (designated pXD33-05) carrying the bla NDM-1 gene and a 33.3 kb IncX4 plasmid (designated pXD33-06) carrying the mcr-1 gene (Fig. 4a, b). Strikingly, other isolates that co-produced NDM and MCR (n = 6) displayed high sequence homology across the vast majority of these plasmid backbones, strongly suggesting carriage of highly similar plasmids ( Supplementary Fig. 3). BLAST analysis of the pXD33-05 sequence showed that the backbone of pXD33-05 (plasmid excluding the MDR element) displayed 99% homology to that of pHNEC55 (GenBank accession no. KT879914), a plasmid carried by E. coli HNEC55 ( Fig. 4d and Supplementary Fig. 2b), a carbapenem-resistant strain also obtained from pig farms in Henan Province 29 . Interestingly, plasmids with similar backbones carrying different MDR cassettes have also been reported previously, including plasmid pHNHNC02 (GenBank accession no. MG197497), which was recovered from an E. coli strain from chickens, and plasmid pKPC-L388 (GenBank accession no. CP029225), which was carried by a clinical K. pneumoniae isolate from humans (Fig. 4d) elements of pXD33-06 harbored only the mcr-1 gene ( Fig. 4e and Supplementary Fig. 2c).
ONT sequencing of two highly tigecycline-resistant isolates, HB50 and SY36, revealed that tetX4 was carried by an IncX1 plasmid in both isolates (Fig. 4c). BLAST analyses revealed that the sequence of this IncX1 plasmid displayed 99% homology to plasmid pYY76-1-2 (GenBank accession no. CP040929), which is also a tetX4-bearing plasmid carried by an E. coli isolate from the feces of a cow 30 . Plasmids with similar backbones were also reported previously, including the pEC931_tetX plasmid (Gen-Bank accession no. CP049121), which was recovered from a clinical E. coli isolate from humans. In all identified tetX4-carrying elements, tetX4 was adjacent to an ISCR2 element ( Supplementary Fig. 2d)  isolates from pig farms in China to 287 publicly available draft genomes of human commensal E. coli (Bioproject no. PRJNA400107) was analyzed 31 . The phylogenetic analysis of the 802 E. coli isolates showed that the 515 MDR E. coli isolates from pig farms were closely related to the 287 human E. coli strains (Fig. 5a). Single nucleotide polymorphism (SNP) analyses revealed that a large proportion of the pig farm isolates showed high genetic similarity (228/515, as few as three SNPs) to the human isolates ( Fig. 5 and Supplementary Data 8). MLST of the 287 human commensal isolates also revealed significant diversity of STs, where the isolates were assigned to 132 distinct STs, with the exception of 11 isolates with novel STs (Fig. 5b and Supplementary Data 9). Interestingly, ST10 was the most common ST and was carried by 44 human isolates (15.33%). Among the 132 distinct STs, 45 STs were also shared by the 515 drug-resistant isolates from pig farms (Fig. 5b).

Discussion
In this study, we present a nationwide surveillance report encompassing isolates from pig farms in all of China's mainland provinces. While a large number of similar studies have been conducted on different Chinese pig farms [32][33][34] , these studies have been limited at a spatial level. Most of them were conducted in regions (e.g., Henan, Sichuan, Hubei, Hunan, and Shandong) with large pig populations, where the Chinese pig-farming industry has been developed. In regions, where small pig populations are reared (e.g., Qinghai and Tibet), small amounts of associated data have been reported 14 . To the best of our knowledge, this is the first study to investigate the AMR phenotypes of E. coli isolates from pig farms in all provinces of mainland China during the same period.
Our AST results suggest a worrisome AMR situation on pig farms, as evidenced by the common recovery of MDR E. coli isolates from both pigs and their breeding environments on farms in different provinces, including Tibet, Xinjiang, and Qinghai. This worrisome situation is widely accepted to be the result of antibiotic overuse and misuse in the Chinese pig industry 26,35 . During recent decades, with the rapid increase in the economy, the production of meat, eggs, and milk has rapidly increased in China; this is especially true for pork, the main source of animal protein for most Chinese people 26,36 . To meet the increasing demand for pork, both the number and the size of pig farms have increased markedly, and massive amounts of antibiotics are used in the country to support the rapid increase in pig production 26 .
Our AST results also demonstrated that E. coli isolates from Chinese pig farms were commonly resistant to several specific antimicrobial classes, including sulfonamides, tetracyclines, fluoroquinolones, macrolides, and β-lactams (excluding carbapenems). In addition, ARGs associated with resistance to these antimicrobial classes were also commonly identified in isolates from the farms. This may be because all of these antimicrobial classes have been extensively used on Chinese pig farms, particularly fluoroquinolones and β-lactams, which account for more than half of the application of these agents 27,37,38 . A previous study reported that total tetracycline concentrations in manure and soil samples from several Chinese pig farms were as high as 15.2 and 0.78 mg kg −1 , while the highest mean concentrations of ofloxacin and enrofloxacin observed in samples collected from these farms were as high as 335 µg kg −1 and 96.0 µg kg −1 , respectively 39 . In the present study, although farmers on the majority of farms included in our investigation were not willing to disclose the antibiotics used in their facilities, several disclosed that antibiotics belonging to the sulfonamide (e.g., sulfamonomethoxine), tetracycline (e.g., chlortetracycline, doxycycline, and oxycycline), fluoroquinolone (e.g., enrofloxacin), macrolide (e.g., ivermectin and kitasamycin tartrate), and β-lactam (e.g., cefotaxime, penicillin, and amoxycillin) classes were used on their farms for growth promotion or disease therapy.
Although the use of chloramphenicol in animal-producing facilities was banned in China in 2002 40 , chloramphenicolresistant isolates were still frequently recovered from pig farms in different provinces in this study. This might be because florfenicol, a member of the same class as chloramphenicol, is still in use 26 and may be responsible for the maintenance of chloramphenicol resistance in E. coli from Chinese pig farms. Another noteworthy finding was that colistin-resistant isolates were recovered from pig farms in 12 provinces in low proportions (3.79%). This relatively low proportion of recovery is attributed to China's policy banning the use of colistin in agriculture in 2017, which has had a significant effect on reducing colistin resistance in both animals and humans 21,41 . Although many E. coli isolates from pig farms in China showed resistance to tigecycline (37.31%, n = 698), most of them displayed a low level of resistance (MIC values ranging from 0.5 to 1 µg/ml; 92.98%, n = 649). This may be because the breakpoint used for the interpretation of tigecycline resistance recommended by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) is low (0.5 µg/ml, EUCAST document v 8.1). A phenotype of low tigecycline resistance may be conferred by tetracycline resistance genes (e.g., tetA, tetB, tetC, and/or tetM 42 ) since these genes are commonly carried by E. coli from animals.
Our investigation also showed the worrisome situation of the presence carbapenem-resistant isolates on pig farms in seven provinces, all of which were NDM-producing isolates. The recovery of these isolates was very unusual and unexpected because carbapenem antibiotics are not approved for use in livestock in China according to the official policy. A possible explanation is that some pig farms might use these antibiotics as a therapy without receiving approval because infections caused by MDR bacteria are very common on livestock farms 2,32,33,43 , and carbapenem antibiotics are considered last-resort effective options for treating MDR bacterial infections 44 . However, we did not obtain the required information since the farmers on these farms were not willing to disclose the antibiotics used at their facilities. Another possible reason might be contamination in the in-house environment, as a recent study showed a high detection rate (26.8-31.4%) of bla NDM in environmental samples, except in air, after standard cleaning and disinfection during the vacancy period on a Chinese poultry farm 43 . Interestingly, the reported rates of carbapenem-resistant E. coli in human health care facilities in provinces from which farm-originating carbapenemresistant isolates have been recovered are higher than those in provinces, where farm-originating carbapenem-resistant isolates have not been recovered, according to the China Antimicrobial Resistance Surveillance System (CARSS) 45,46 . Therefore, it can be hypothesized that carbapenem-resistant bacteria or bla NDM from human health care facilities may lead to contamination by these hazardous agents in environments (e.g., water, air, etc.); this contamination may in turn spread to livestock farms, resulting in the emergence of carbapenem-resistant bacteria in animals. In addition, a third possibility worth considering is that the plasmid and/or isolate was co-selected by the use of another antimicrobial that may be present in the plasmid and/or isolate. However, further investigations are necessary to verify the above three assumptions.
Our WGS analysis identified several O-serogroups that are frequently associated with E. coli pathotypes 47,48 . Importantly, several marker VFGs of E. coli pathotypes (e.g., astA, eae, east1, stb, stx 2eB , and toxB 47,48 ) were also identified in E. coli isolates belonging to these O-serogroups. However, only 36.64% (66/181) of these isolates were recovered from the feces of diarrheal pigs, and the remaining isolates came from environmental samples. The widespread presence of these MDR E. coli on pig farms represents a great health risk and should receive more attention. Our WGS analysis also demonstrated that ST10 (n = 52), ST101 (n = 39), and ST48 (n = 22) were widely distributed among the MDR E. coli isolates from pig farms. It should be noted that all three of these STs, particularly ST10, have been reported to be important AMR-associated genotypes predominating in pigs 7,21-25 . Of particular note, the ST10 clone has been revealed to be a reservoir of ARGs and mobile genetic elements, such as class 1 integrons and plasmids 24,25 . It should also be noted that these three STs have been frequently identified in E. coli pathotypes (ETEC/EPEC/STEC) from pigs with diarrhea 49 . Accordingly, these three STs were commonly identified among the 207 isolates from the feces of diarrheal pigs [ST10 (n = 28), ST101 (n = 14), and ST48 (n = 8)] in this study. Notably, these three STs were also predominantly identified STs among the 287 human commensal isolates analyzed in this study. Furthermore, these findings are in agreement with those of a recent study suggesting that E. coli clonal complex 10, including ST10, may be associated with fecal carriage in humans 24 .
It has been documented that E. coli has a great capacity to accumulate ARGs, mostly through horizontal gene transfer 7 . Accordingly, a large number of acquired ARGs were detected in the genomes of the isolates recovered in this study. Along with the detection of multiple ARGs, a large number of the sequenced isolates were found to carry the class 1 integrase gene intI1. This gene is always linked to genes conferring resistance to antibiotics, disinfectants, and heavy metals, and it could be used as a proxy for anthropogenic pollution 50 . In addition to the class 1 integrase, we detected many putative plasmids in E. coli isolates from pig farms in China. Several groups of these plasmids (e.g., IncA/C, IncB/O, IncFIB, IncX1, IncFIIA, IncX2, and IncY) are capable of carrying sequences conferring transfer, MDR, and virulence functions in a broad host range 51 . The presence of those plasmids may accelerate the dissemination of ARGs. We also identified several groups of plasmids that were rarely associated with the spread of specific ARGs. For example, the IncX3 plasmid has been reported to account for the majority of bla NDM carriage in livestock farms 43,52 . However, we also identified an IncFII-typebla NDM -carrying plasmid, and this type of plasmid was observed in all E. coli isolates coproducing NDM and MCR. In addition, the mcr-carrying plasmid identified in E. coli isolates coproducing NDM and MCR in this study was different from the plasmid mediating the dissemination of mcr in China reported recently 4 . The presence of these plasmids makes the dissemination of bla NDM-1 and/or mcr more heterogeneous. It is worth noting that the elements mediating the spread of the high tigecycline resistance gene tetX4 in E. coli isolates from pig farms in this study were highly homologous to those reported previously 3,31,53 , suggesting that the dissemination of tetX4 might not be as heterogeneous as that of bla NDM and mcr. However, continuous monitoring should be undertaken in the future. Most notably, the MDR E. coli isolates from pig farms displayed very close relatedness to E. coli strains from humans in China, and most of them shared the same STs as the human isolates. Many of the E. coli isolates obtained from pig farms in this study differed by fewer than 100 SNPs from the human-originating E. coli strains. These findings suggest a very high genetic propensity of farm-sourced MDR E. coli to spread into humans.
Although this work is limited by our inability to include samples from more pig farms in different provinces of China due to the outbreaks of ASF and COVID-19, our sample collection still covers pig farms in all provinces of China. We characterized the resistance phenotypes and population genomics of E. coli on pig farms in China on a national scale for the first time. Our results revealed a worrisome situation of AMR on pig farms in China, and there is still a long way to go to reduce AMR in livestock in China. The Chinese government has implemented a series of actions to address the worrisome AMR conditions in animal husbandry. Notably, the Ministry of Agriculture and Rural Affairs (MARA) issued a policy banning the addition of antibiotics to feed to promote animal growth on July 1, 2020 (MARA Announcement No. 194, 07-10-2019). In this study, we also systematically revealed the distribution of O-serogroups, sequence types, ARGs, VFGs, and putative plasmids of MDR E. coli on pig farms in different provinces of China. These data will provide comprehensive insights to help understand AMR on pig farms, and may also be beneficial for the development of governmental policies to reduce AMR in the Chinese pig industry. Notably, we also identified many MDR E. coli with potential pathogenicity to humans, and most importantly, we found that pig farm-originating MDR E. coli showed a very high genetic propensity to spread humans. The persistence and dissemination of these isolates present important health risks and should receive more attention.

Methods
Sample collection, identification, and antimicrobial susceptibility testing. Between 1 October 2018 and 30 September 2019, an AMR surveillance project was established based on E. coli isolated from pig farms in China. In each of the 31 provinces of mainland China, 2-3 different pig farms were randomly selected to collect swabs of fresh feces and rectal swabs of pigs (with approximately 40 samples per farm) as well as swabs of drinking and fecal slurries, floors, and troughs (at each sampling point, at least three samples per farm were collected). All samples were immediately shipped with dry ice for E. coli isolation. The need for ethical approval has been waived by the Animal Management and Ethics Committee of Huazhong Agricultural University because the sample collection does not involve in the use of animal tissues. The reference ID number for the waiver is HZAUSW-2018-023. The swabs were incubated in Luria Bertani (LB) broth (Sigma-Aldrich, MO, USA) at 37°C and 180 rpm overnight. Thereafter, the swab cultures were streaked on MacConkey agar and incubated at 37°C for 16 h. Single colonies with similar morphological characteristics (small red/pink round colonies) were picked and confirmed by the PCR amplification of the 16S rRNA gene with primers (F: 5′-GAAGCTTGCTTCTTTGCT-3′, R: 5′-GAGCCCGGGGATTTCACAT-3′) reported previously 54 . PCR assays amplifying seven housekeeping genes of E. coli (adk, fumC, gyrB, icd, mdh, purA, and recA) 55 were set up for double confirmation.
Whole-genome sequencing and data availability. All E. coli strains with phenotypes of resistance to one of the tested carbapenems, broad-spectrum cephalosporins (ceftriaxone and cefepime), tigecycline, or colistin were selected for NGS. Bacterial genomic DNA was extracted by using a commercial DNA Kit (TIAN-GEN, Beijing, China). The quality and concentration of the bacterial genomic DNA were evaluated via electrophoresis on a 1% agarose gel and analysis on a NanoDrop2000 system (Thermo Scientific, Waltham, MA, USA) and a Qubit 4 Fluorometer (Thermo Scientific, Waltham, USA). Libraries were constructed based on the qualified DNA by using a NEBNext Ultra TM II DNA Library Prep Kit (New England BioLabs, Ipswich, USA) and sequenced on the NovaSeq 6000 platform using the paired-end 150 bp sequencing protocol (Novogene, Beijing, China). Raw reads with low quality were removed as described previously 56 . High-quality reads were de novo assembled with SPAdesv3.9.0 to generate genome contigs.
The complete genome sequences of bla NDM-1 -carrying, mcr-1-carrying, and/or tetX4-carrying plasmids were generated by ONT sequencing in combination with Illumina technology. Plasmid DNA was extracted using the phenol-chloroform protocol combined with Phase Lock Gel tubes (Qiagen GmbH), analyzed by agarose gel electrophoresis and quantified on a Qubit® 2.0 system (Thermo Scientific, Waltham, USA). Libraries for ONT and Illumina sequencing were prepared using an SQK-LSK109 kit and a NEBNext® Ultra™ DNA Library Prep kit, respectively. The prepared DNA libraries were sequenced using the Nanopore PromethION platform and an Illumina NovaSeq PE150 system at Novogene Co. LTD (Tianjin, China). ONT and Illumina short reads were finally assembled and combined using Unicycler v0.4.4 software (https://github.com/rrwick/Unicycler/ releases) with the default parameters.
Plasmid conjugation experiments. Plasmid conjugation experiments between carbapenem-resistant E. coli, colistin-resistant E. coli, and/or tigecycline-resistant E. coli (donors) and rifampin-resistant E. coli C600 (recipient) were performed as described previously 23 . Briefly, bacterial donor and recipient strains at mid-log phase (OD 600 = 0.5-0.6) were mixed at a ratio of 1:3 (v/v). The bacterial mixture was spotted on nitrocellulose membranes that were preplated on LB agar. Each of the plates were incubated at 37°C for 12 h, and the bacteria on the membrane were washed using LB broth followed by shaking at 37°C for 4 h. Transconjugants were selected on LB agar plates with rifampin (1000 mg/l) plus imipenem (20 mg/l) [to screen carbapenem-resistant transconjugants], rifampin (1000 mg/l) plus colistin (2 mg/l) [to screen colistin-resistant transconjugants], or rifampin (1000 mg/l) plus tigecycline (4 mg/l) [to screen tigecycline-resistant transconjugants]. The antimicrobial susceptibility of the transconjugants was identified using the broth microdilution method as mentioned above.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The whole-genome sequences (WGSs) of E. coli isolates generated in this study have been deposited in GenBank under accession code bioproject PRJNA688628. GenBank accession numbers are given in supplementary Data 2.