ESBL and AmpC β-Lactamase Encoding Genes in E. coli From Pig and Pig Farm Workers in Vietnam and Their Association With Mobile Genetic Elements

Animals are considered important sources of ESBL/AmpC-producing bacteria in humans. We analyzed indications of transfer of ESBL/AmpC genes between pigs and pig farmers in Vietnam by analyzing whole genome sequences of 114 ESBL/AmpC-producing E. coli isolated from the two hosts, and performed conjugation experiments and plasmid profiling to confirm that such transfer could have happened. ESBL-encoding genes detected in pigs and pig farmers included blaCTX–M-55, blaCTX-M-27, blaCTX-M-65, blaCTX-M-15, blaCTX-M-14, blaCTX-M-3, blaCTX-M-24, and blaCARB-2, and AmpC β-lactamases included blaCMY-2, blaDHA-1, and blaCMY-42. The most frequent ESBL gene, blaCTX-M-55, was carried on plasmid with replicons types IncF, IncX, IncH, IncN, IncR, and IncP. The insertion transposases downstream of the blaCTX-M-55 gene were different in plasmids carried by different strains. The second most detected gene, blaCTX-M-27, is found in a stable genetic arrangement with the same flanking transposons seen across strains, and the gene was located on similar conjugal IncF plasmid types, suggesting a horizontal spread of these plasmids. In three strains, we observed a novel blaCTX-M-27 harboring IncF type of plasmid which had not been reported before. Its closest reference in NCBI was the non-ESBL Salmonella Typhimurium plasmid pB71 that might have experienced an insertion of blaCTX-M-27. Our data also point to an emergence of plasmids co-carrying ESBL genes, mcr genes, quinolones and other antimicrobials resistance determinants, and such plasmids require special attention. Plasmids phylogeny confirmed that the blaCTX-M-55 encoding plasmids varied considerably, while those encoding blaCTX-M-27 were closely related. Plasmids harboring both ESBL genes were confirmed to be conjugative and not to differ in transfer efficacy. The isolates carrying the plasmids, even those with plasmids of similar types, showed wide genetic variation with high number of SNPs, suggesting horizontal spread of plasmids into different clonal lines. Their virulence profiles did not confirm to known pathotypes, suggesting that unrelated commensals are a main reservoir for ESBL and AmpC β-lactamases in both humans and pigs. Overall, despite evidence of transferability of plasmids in the analyzed strains, our findings do not support that ESBL-producing E. coli from pigs or their ESBL/AmpC encoding plasmids are commonly spread to workers in close contact with the animals.

Animals are considered important sources of ESBL/AmpC-producing bacteria in humans. We analyzed indications of transfer of ESBL/AmpC genes between pigs and pig farmers in Vietnam by analyzing whole genome sequences of 114 ESBL/AmpCproducing E. coli isolated from the two hosts, and performed conjugation experiments and plasmid profiling to confirm that such transfer could have happened. ESBLencoding genes detected in pigs and pig farmers included bla CTX-M-55 , bla CTX-M-27 , bla CTX-M-65 , bla CTX-M-15 , bla CTX-M-14 , bla CTX-M-3 , bla CTX-M-24 , and bla CARB-2 , and AmpC β-lactamases included bla CMY-2 , bla DHA-1 , and bla  . The most frequent ESBL gene, bla CTX-M-55 , was carried on plasmid with replicons types IncF, IncX, IncH, IncN, IncR, and IncP. The insertion transposases downstream of the bla CTX-M-55 gene were different in plasmids carried by different strains. The second most detected gene, bla CTX-M-27 , is found in a stable genetic arrangement with the same flanking transposons seen across strains, and the gene was located on similar conjugal IncF plasmid types, suggesting a horizontal spread of these plasmids. In three strains, we observed a novel bla CTX-M-27 harboring IncF type of plasmid which had not been reported before. Its closest reference in NCBI was the non-ESBL Salmonella Typhimurium plasmid pB71 that might have experienced an insertion of bla CTX-M-27 . Our data also point to an emergence of plasmids co-carrying ESBL genes, mcr genes, quinolones and other antimicrobials resistance determinants, and such plasmids require special attention. Plasmids phylogeny confirmed that the bla CTX-M-55 encoding plasmids varied considerably, while those encoding bla CTX-M-27 were closely related. Plasmids harboring both ESBL genes were confirmed to be conjugative and not to differ in transfer efficacy. The isolates carrying the plasmids, even those with plasmids of similar types, showed wide genetic variation with high number of SNPs, suggesting horizontal spread of plasmids into different clonal lines. Their virulence profiles did not confirm to known pathotypes, suggesting that unrelated commensals are a main reservoir for

INTRODUCTION
Extended-spectrum β-lactamases (ESBLs) and AmpC β-lactamases are the most widespread mechanisms of cephalosporin resistance in E. coli in animals and humans worldwide, including the Asia Pacific region (Jean and Hsueh, 2017;Naas et al., 2017). Some of the ESBLs and AmpC β-lactamases described in E. coli have spread globally in the last two decades, including CMY-2, which is the most common AmpC enzyme (Naas et al., 2017), and CTX-M, which is the most widespread ESBL type (Jean and Hsueh, 2017); however, there are marked differences in the CTX-M variants occurring across sources and geographical areas. In Vietnam, CTX-M-1 group enzymes including CTX-M-15 and CTX-M-55, and CTX-M-9 group enzymes including CTX-M-14, CTX-M-27, and CTX-M-65 have been described in E. coli from diseased and healthy humans, chickens, pigs, meat, and vegetables (Bui et al., 2015;Zurfluh et al., 2015;Nguyen et al., 2016;Hoang et al., 2017;Jean and Hsueh, 2017).
In most cases, ESBL/AmpC-encoding genes are located on mobile genetic element, and many are plasmid-mediated and transferable between bacteria of different species. Studies from Europe have shown that farmers share highly related and even identical ESBL/AmpC E. coli and/or ESBL/AmpC plasmids with their livestock (Madec et al., 2017). It is likely that this risk is enhanced within countries with higher level of animal to human contact due to low hygienic standards in their livestock production, such as in pig production in Vietnam (Hoang et al., 2017), but this remains to be further investigated.
Knowledge and control of ESBL and AmpC in E. coli from animals is crucial to preserve the effect of medically important antimicrobials such as cephalosporins. Therefore, ESBL/AmpC monitoring in E. coli from animals and food has been mandatory in EU countries since 2014 (Commission implementing Decision 2013/652/EU) and such monitoring is a suggested focus area by the World Health Organization in low-and middle-income countries (The ESBL Ec Tricycle AMR surveillance project) 1 .
In Vietnam, knowledge on the ESBL/AmpC variants in E. coli from pigs and pig farmers is limited, and thorough characterization of the genetic location of the ESBL/AmpCencoding genes in E. coli is lacking. Moreover, most genomic studies that address transmission of resistance genes between animals and humans rely on tracking clonally evolving lineages excluding extrachromosomal plasmids despite plasmids being the primary carriers of antibiotic resistance genes across the pathogens (David et al., 2020).
The objective of this study was therefore to characterize ESBL/AmpC genes in E. coli in pigs and pig farmers in northern Vietnam and determine their location and context on mobile genetic elements to estimate the level of transfer between pigs and humans. Information from this study illustrates the role of different genetic elements in the spread of ESBL/AmpC genes between human and animal hosts.

Isolates and Genome Collections
In a previous study, cefotaxime-resistant E. coli collected in 2015 from pigs and farm workers in Thai Binh and Hanoi provinces in Vietnam were characterized (Dang et al., 2018). Forty of these cefotaxime-resistant E. coli was selected for whole genome sequencing by stratified random selection of isolates. At least one isolate per cephalosporin resistance profile compatible with ESBL/AmpC production was included. When the same resistance profile was present in more than one isolate, a number of isolates proportional to the frequency of that phenotype within the overall collection of isolates was selected. Choice of isolates within specific resistance profiles was random, in that, each isolate was assigned a consecutive number and random numbers were generated by using online freely available software 2 . Since we aimed at having pig and human isolates equally represented within each profile, the numbers were randomly generated until this criterion was satisfied. Only in one case, four isolates from humans and two isolates from pigs (instead of 3 and 3) were selected since the third isolate originating from pigs was no longer available.
Additionally, 74 genomes of ESBL/AmpC producing E. coli isolated in 2018 from pigs and pig farm workers in Bac Ninh province, Vietnam were retracted from a sequence collection submitted to the European Nucleotide Archives under the project accession number PRJEB37980. Thus in total 114 genomes were included in the study.
All 114 strains were previously subjected to antimicrobial susceptibility testing, and results are available from Dang et al. (2018) and a recently submitted manuscript (Duong et al., under review).
Whole Genome Sequencing, Genome Assembly, and Genome-Characterization Genomic DNA was extracted from the 40 E. coli isolates from 2015 using an Invitrogen Easy-DNA Kit (Invitrogen, Carlsbad, CA, United States) and DNA concentration was determined using the Qubit dsDNA BR assay kit (Invitrogen). The genomic DNA was prepared for Illumina pair-end sequencing using the Illumina (Illumina, Inc., San Diego, CA) NexteraXT R Guide 14 150319425031942 following the protocol revision C. The libraries were sequenced using an Illumina MiSeq platform. Raw sequence data were submitted to the European Nucleotide Archive (ENA) under study accession no: PRJEB30991. Along with the 2018 genomes retracted from ENA under the project accession PRJEB37980., the raw reads of all the 114 strains were de novo assembled using SPAdes algorithm for de novo short reads assembly (Bankevich et al., 2012). Assembled sequences were analyzed using the CGE tools 3 including MLST finder to determine E. coli multi-locus sequence types (MLST), ResFinder 3.1 for detection of genes and chromosomal mutations mediating AMR, VirulenceFinder 2.0 with default settings for detection of virulence genes, PlasmidFinder for detection of plasmid replicons, and pMLST for further subtyping of specific plasmids.

Analysis of Contigs Harboring ESBL/AmpC-Encoding Genes
In all genomes, contigs harboring ESBL/AmpC-encoding genes were analyzed in detail to gain insights into the genetic context of the ESBL/AmpC-encoding genes. First, nucleotide BLAST at NCBI was used to determine the level of identity of the contig with publicly available sequences. Then, Open Reading Frames (ORFs) were predicted and annotated using Artemis software version 8 (Carver et al., 2012), and each predicted protein was compared against the all-protein database at NCBI using BlastP. ISfinder was used for identification of insertion sequences (IS) (Siguier et al., 2006). Comparisons of contigs harboring identical ESBL/AmpCencoding genes were performed using CLC genomics workbench v.8. Furthermore, the Artemis comparative tool (Carver et al., 2005) was used to perform in-depth comparative genomics between ESBL-hosting contigs of our samples and their closest references from NCBI.

Plasmids Reconstruction and Characterization
To retract plasmid components from the short illumina reads for downstream analyses, the raw reads of all genomes with and without plasmid replicons were analyzed with PlasmidSpades (Bankevich et al., 2012). The predicted plasmids were then used as input files in PlasmidFinder to detect their replicon types and analyzed in ResFinder to detect antimicrobial resistance genes found on these plasmids including ESBL/AmpC encoding genes as well as co-occurrence of the plasmidmediated colistin resistance mcr genes and more. Additionally, the concatenated plasmid component files were analyzed in Bacmet (Pal et al., 2014) for experimentally confirmed metal and biocide tolerance genes co-occurring on the ESBL/AmpC carrying plasmids. With a two-two-table from STATCALC in EPIINFO v.7.2, we assessed the association of metal and biocide resistance genes co-occurring with β-lactamases on the same predicted plasmids. The reconstructed plasmids were also annotated in RAST (Overbeek et al., 2014) and visualized with CLC Genomics workbench v8 and displayed as graphics for selected samples.

Plasmid Profiling
To confirm the presence and the size of predicted plasmids in the strains, a plasmid profiling was carried out where plasmid DNA was extracted according to Kado and Liu (1981). The purified plasmid DNA was separated on a 0.8% (W/V) agarose gel for 3.5 h at 150 V and stained with 0.5 µg ml −1 ethidium bromide. The approximate molecular weight of each plasmid was determined by comparison with two reference E. coli strains, 39R861 (Threlfall et al., 1986) and V517 (Macrina et al., 1978), containing multiple reference plasmids.

Diversity of Plasmids Harboring the Predominant ESBL Genes bla CTX-M-55 and bla CTX-M-27
Due to the fact that the ESBL/AmpC contigs were of highly variable sizes and also do not represent the whole plasmid, the predicted plasmid components were annotated using Prokka v1.13.3 (Seemann, 2014) and used for comparative analyses. Two pan-genome analyses were performed using roary v3.12 (Page et al., 2015) across the annotated gff files of predicted plasmids harboring bla CTX-M-55 and for those harboring bla CTX-M-27 to identify the number of shared genes among the plasmids. For better representation of the gene contents of the plasmids, only plasmid components with the size above 40 kb were included in these analyses making 18 plasmid components in each pangenome.
A single nucleotide polymorphism analysis was not appropriate due to the small fraction of nucleotides representing the core-genome in the analyzed plasmids; therefore, the phylogenetic analysis was performed using nucleotides alignment method. This analysis included all predicted plasmids harboring the two bla CTX-M-55 and bla CTX-M-27 ), and DNA sequences were aligned using the Clustal Omega multiple sequence alignment tool 4 . The aligned files were converted to MEGA file format then used as input to construct a neighbor-joining phylogenetic tree in MEGAx (Kumar et al., 2018) using the bootstrap method with 500 replicates. The final trees were visualized in iTOL (Letunic and Bork, 2016) and rooted with the shortest plasmids in each group.

Genetic Diversity of the E. coli Isolates
For the overall population analysis of the E. coli genomes, the whole genome sequences were analyzed in CSIPhylogeny for core-genome phylogenetic analysis using the genome of E. coli K12 as reference and the resulting tree was visualized in iTOL (Letunic and Bork, 2016). In this analysis, the pig isolate 51A3 obtained in 2015 containing only the narrow-spectrum β-lactamase bla TEM −1A was not included.

Conjugation Experiment
To experimentally investigate the transferability of the plasmids harboring the bla CTX-M-27 and bla CTX-M-55 genes, a conjugation experiment was conducted. We randomly selected four strains to serve as donors including EC224 (IncFII) and EC170 (IncFIB) representing bla CTX-M-27 , while bla CTX-M-55 was represented by strains EC297 (IncFIB) and EC116 (IncN). The recipient strain in all experiments was E. coli J53-1 encoding a chromosomally located rifampicin resistance gene. The conjugation experiment was performed as previously described (Møller et al., 2017). Briefly, the donor and recipient strains were grown in Luria-Bertani (LB) broth (Sigma, Copenhagen, Denmark) with shaking (180 rpm) at 37 • C to exponential phase (OD600 = 0.5). Conjugation was performed by mixing donor and recipient strains in a 1:1 ratio on filter papers (0.22 µM, Millipore, Copenhagen, Denmark) placed on LB agar plates (Becton, Dickinson, Albertslund, Denmark) at 37 • C for 1 and 6 h. The bacterial material was washed from the filters using isotonic NaCl and plated on LB agar plates containing either 2 mg L −1 cefotaxime (to quantify donor + transconjugants), 50 mg L −1 rifampicin (to quantify doner and transconjugants), or 50 mg L −1 rifampicin and 2 mg L −1 cefotaxime (to quantify transconjugants) and incubated overnight at 37 • C. The conjugation experiments were performed with two biological replicates and two technical replicates each. Selected colonies of the transconjugants were subjected to plasmid profiling using the method described above to confirm the transfer of the ESBL-gene encoding plasmids from the donors. The conjugation frequency of each strain was calculated as the number of transconjugants divided by the number of donors. The results are presented as means cfu. mL −1 ± SD.

Genetic Context of ESBL/AmpC Genes bla CTX-M-55
The gene bla CTX-M-55 was the most abundant ESBL-encoding gene observed (n = 40 strains). Twenty-eight strains originated from pigs and 12 were from farm workers. In all cases, the bla CTX-M-55 gene was flanked by an ISEcp1 elements upstream and a partial orf477 downstream (Figure 1). However, genetic variations between bla CTX-M-55 encoding elements were frequent due to different length of the ISEcp1 elements, different length of the region between ISEcp1 and bla CTX-M-55 , and/or different Frontiers in Microbiology | www.frontiersin.org length of the partial orf477 (Figure 1). In seven of the 40 strains containing bla CTX-M-55 , BLAST in NCBI suggested a chromosomal location (Table 1 and Supplementary Table S1), as the bla CTX-M-55 encoding elements were flanked by genes which are normally chromosomally located. In the remaining 33 strains the hosting contigs not only corresponded to plasmid sequences in NCBI (Supplementary Table S1) but also included ORFs which clearly indicated plasmid location (e.g., plasmid DNA primase, DNA ligase, and conserved plasmid hypothetical proteins). The predicted plasmid components carrying bla CTX-M-55 were of variable replicon types including IncF (IncFIA, IncFIB(K), IncFII), IncX, IncH, IncN, IncR, and IncP (Supplementary File 1). Overall, bla CTX-M-55 was predominantly plasmid mediated, surrounded by variable genetic contexts and carried on plasmids with multiple different replicon-types. Comparison of isolates obtained from humans and pigs did not suggest transmission of plasmids carrying this gene between pigs and farmers.
bla CTX-M-27 The gene bla CTX-M-27 was detected in 36 E. coli strains including 24 strains from pig farm workers and 12 from pigs. bla CTX-M-27 was flanked in all strains by a partial ISEcp1 upstream and partial IS903B downstream (Figure 2). In all 36 strains, the contigs where bla CTX-M-27 was located produced exclusively plasmid hits by BLAST in NCBI, mostly with 99-100% coverage and percent  identity match to published plasmid sequences (Table 1 and  Supplementary Table S1). The gene bla CTX-M-27 was associated with IncF conjugative plasmids of variable sizes as confirmed by plasmid profiling (Supplementary File 1). Strikingly, in three strains, we observed that bla CTX-M-27 was inserted on a novel IncF type of plasmid that had not been reported before (Figure 3). Its closest reference in NCBI was the non-ESBLencoding Salmonella Typhimurium plasmid pB71, Accession KP899806 (70% coverage and 99% identity). bla CTX-M-27 in ESBL-producing E. coli from pigs and pig farm workers were all carried by the conjugal IncF types of plasmid with a consistent genetic context suggesting horizontal transmission of the mobile elements.

Other CTX-M Genes
The gene bla CTX-M-14 was detected in 15 strains including five strains from humans and ten strains from pigs. bla CTX-M-14 was flanked by partial IS903B interrupted by an ISEcp1 upstream and partial IS903B downstream (Supplementary Figure S1). This ESBL gene occurred either in the chromosome or on plasmids, i.e., in nine out of the 15 contigs, BLAST showed best hits for E coli chromosomal fragments (Table 1 and  Supplementary Table S1). In the remaining six strains, the contigs harboring bla CTX-M-14 showed highly similar nucleotide sequences and perfect matches to plasmid sequences including mostly the E. coli plasmid with accession number #MT318677 (Supplementary Table S1). Interestingly, the NCBI BLAST of the bla CTX-M-14 contig of the human isolate EC514 yielded a perfect match to the Escherichia coli plasmid pESBL57, Accession MT230319 initially reported with bla CTM−M-97 (Figure 4). The predicted plasmids harboring bla CTX-M-14 were of varying sizes as confirmed by profiling (Supplementary File 1) and showed different replicon types including IncB/O/K/Z, IncFIB, and IncX4 (Table 1 and Supplementary Table S1). In summary, bla CTX-M-14 was predominantly chromosomally located with insertion in variable genetic contexts, and there was little evidence that the gene was transferred between E. coli in pigs and farmers. The gene bla CTX-M-15 was detected in 10 strains including seven strains from humans and three strains of pig origin. The bla CTX-M-15 was flanked by an ISEcp1 gene upstream and partial orf477 downstream, which was followed by partial Tn2 sequence (Supplementary Figure S2). In the human isolate 29C1, bla CTX-M-15 was predicted to be located in the chromosome ( Table 1 and Supplementary Table S1), and annotation of the contig showed that bla CTX-M-15 and its associated transposon elements were inserted downstream of a 3-dehydroquinate dehydratase annotated as RR31_09005 in E. coli strain 6409 (GenBank #CP010371). In the remaining nine strains, the bla CTX-M-15 contig generated plasmid hits in NCBI to which perfect matches (100% coverage and 100%ID) were determined (Supplementary Table S1). The reconstructed plasmids varied in size (Supplementary Table S1 and Supplementary File 1) and were mostly of the IncF conjugal type. Two of them were however multi-replicon plasmids such as   Figure S4A). By BLAST at NCBI, the contig showed the highest similarity (88% coverage and 99.8% ID) with the E coli plasmid with the accession number MF136778 (Supplementary Table S1). bla CTX-M-24 was confirmed on an IncP1 plasmid.
Strain EC488 isolated from a farm worker in 2018 was the only isolate carrying bla CTX-M-3 on a short 1,289 bp contig that yielded a perfect sequence match by BLAST in NCBI to the Klebsiella pneumonia plasmid of accession number LC556222 (Supplementary Table S1). The only annotated ORF on the contig hosting bla CTX-M-3 was a tryptophan synthase beta chain located downstream (Supplementary Figure S4B). In addition, a 200 kb plasmid of the IncF(II) replicon type was predicted in the isolate and confirmed the plasmid location of bla CTX-M-3 .

AmpC Genes
The gene bla CMY-2 was detected in seven strains. Five of these strains contained the gene in the chromosome (Table 1 and Supplementary Table S1) located together with genes encoding for beta-ketoacyl-ACP synthase, aspartate racemase and transcriptional activator protein LysR. The other two strains carried bla CMY-2 gene plasmid contig corresponding to E coli plasmids #CP034399 and #AP023192 (100% coverage and 99.9%ID). Reconstruction put the gene on IncF conjugal plasmids. In all the seven isolates, bla CMY-2 was flanked by a complete ISEcp1 upstream, whereas the region downstream varied among the strains (Figure 5).
The gene bla CMY-42 was detected in a single strain from a farm worker on a 33,046 bp contig with IncI1 plasmid elements including repI1, ardA, and trbA. The bla CMY-42 gene was detected downstream of a partial ISEcp1 and had a hypothetical mobile element downstream (Supplementary Figure S5A). By BLAST at NCBI, this contig showed a perfect match (100% identity and 100% coverage) with E. coli plasmid pCMY-42 (KY463221). The reconstructed plasmid from this strain was also of IncI1 replicon type which confirmed the presence of bla CMY-42 as a plasmidmediated gene.
The bla DHA-1 gene was detected in one human (EC495) and one pig strain (EC67). The genetic arrangement of annotated ORFs around bla DHA-1 in the two strains showed that it was flanked by a transposase InsE for insertion sequence IS3E (upstream) and an undefined mobile element protein downstream (Supplementary Figure S5B). The human isolate harbored bla DHA-1 on a 145 kb IncFIA plasmid and the pig strain contained the gene on a 170 kb IncF(IB/II) replicon plasmid.

Co-occurrence of ESBL/AmpC Encoding Genes With Metal/Biocide, Colistin, and Other Antimicrobial Resistance Genes on Same Plasmids
The plasmid-mediated colistin resistance gene mcr-1 was detected in three of the 40 (7.5%) strains from 2015, while 21 of the 74 (28.4%) strains isolated in 2018 carried mcr-1 (in combination with mcr-3 in four strains). Moreover, multiple co-occurrences were observed, where a chi-square test show a significant association (p < 0.018) between the presence of metal/biocide resistance genes and β-lactam genes on the same plasmids (Supplementary Table S1).
The bla CTX-M-27 encoding plasmids mostly harbored additional antimicrobial genes including narrow spectrum β-lactamases such as bla TEM −1B and bla LAP-2 as well as quinolone, aminoglycoside, sulfonamide, and trimethoprim resistance genes (Supplementary Table S1). Moreover, bla CTX-M-27 plasmids co-carried biocide and metal resistance genes such as qacEdelta1, emrE/mvrC, and qacF for quaternary ammonium compounds, zraS and zinT encoding for zinc resistance, hydG encoding for lead resistance, merR for mercury and the cadmium resistance gene yodA (Supplementary Table S1).

Diversity and Transferability of Plasmids Harboring the Two Predominant ESBL Genes (bla CTX-M-55 and bla CTX-M-27 )
Pan-genome analysis of the plasmids harboring bla CTM−M-27 showed that 48 genes out of 1,532 core genes were shared across 80% of the analyzed plasmids. In contrast, only nine genes out of 1,831 constituted the core-genome of plasmids carrying bla CTM−M-55 . In both cases, the main core gene was the bla gene followed by antimicrobial resistance genes and the plasmid replicon proteins common to the analyzed plasmids.
The neighbor-joining phylogeny of the alignments of all reconstructed plasmids carrying bla CTX-M-27 and bla CTX-M-55 confirmed that plasmids carrying bla CTX-M-55 , showed wider genetic variations with four main clades and longer branch lengths within clades ( Figure 6A) compared to the more homogenous IncF bla CTX-M-27 plasmids that was found in pigs  Table S1). and farmers (Figure 6B). Although some bla CTX-M-55 plasmids from pigs and farmers clustered in the same clades (Figure 6A), these plasmids were not of the same replicon types and therefore did not provide evidence of transmission between hosts, but rather supported that this gene is carried by different plasmids with different transposons.
The plasmids carrying bla CTX-M-27 and bla CTX-M-55 were confirmed as conjugative with transfers frequencies varying from 6.55 × 10 -6 to 3.06 × 10 -4 with a slight increase when the conjugation time increased (Table 2). Moreover, in the strains with multiple plasmids subjected to the conjugation experiment, only plasmids encoding the bla CTX-M gene were transferred to the transconjugants (Supplementary Figure S6).

E. coli Diversity
Of the 114 E. coli strains analyzed, 82 different sequence types were detected (Supplementary Table S2, sheet 2) but only three sequence types were shared between human and pig isolates. These include ST10 reported in 10 pig isolates and in 1 human isolate, while ST48 and ST2170 were found in one pig and 1 animal isolate, respectively. None of these few-shared STs was from the same farms. The core genome phylogenetic analysis of the 113 isolates against the reference E. coli K12 confirmed the wide genetic variation among the isolates regardless of host, the farm and years of isolation with up to 39,306 SNPs difference between strains grouped in nine clades (Figure 7). None of the STs was predominantly driving a specific plasmid or ESBL encoding gene. In all of these clades, there were isolates from both pig and farm workers, however the isolates were often of different sequence types with wide SNPs differences (Supplementary Table S2). The strains were predicted to be of commensal types, since their virulence profiles did not correspond to main E. coli pathotypes (Supplementary Table S1, sheet 2).

DISCUSSION
It is widely suggested that the livestock sector may be an important source of antimicrobial resistance genes, including ESBL/AmpC encoding genes, found in humans. The transfer mainly happens through horizontal spread of mobile elements (Zajac et al., 2019;Dantas Palmeira and Ferreira, 2020). To elucidate the level of transmission of ESBL/AmpC genes in E. coli from pig and pig farmers in Vietnam, we performed a detailed genetic analysis of E. coli strains, which harbored such genes, and we analyzed the location of genes in the genome. This included annotation of transposons and insertion sequences located around the resistance genes, the phylogenetic relatedness of the plasmid components reconstructed from strains, and genetic relatedness of the strains, all based on whole genome sequences. This approach provides a holistic answer to the transmission of resistance genes between hosts, either by transfer of strains or by spread of mobile genetic elements, and thus it extends the approaches used in most previous studies of spread of resistance genes and mobile genetic element between hosts (Kudirkiene et al., 2018;Zajac et al., 2019;Baniga et al., 2020). Analysis of the plasmids and the genetic contexts along with the strains diversity allow to investigate the diversity, distribution, and transmission dynamics of resistance plasmids in the strains (David et al., 2020). The ESBL/AmpC β-lactamases genes detected in the analyzed strain collection of E. coli were present across pigs and pig farmers suggesting a distribution of the same genes across hosts, however, the most commonly observed gene differed between the two hosts; bla CTX-M-55 was the most commonly observed among the pig isolates and bla CTX-M-27 among human isolates. The genes detected have previously been described from diseased and healthy humans, chickens, pigs, and food products in Vietnam in studies where the human cases were suggested to emanate from animal origins (Nguyen et al., 2016;Hoang et al., 2017;Hinenoya et al., 2018).
plasmids did not share reference plasmids in NCBI to any large extend. For instance, no more than five contigs with bla CTX-M-55 showed similarity to a common plasmid reference in the database [plasmid accessions MN823991 (5), AP023198 (5), MN158989 (5), CP034747 (4)]. Overall, these findings with regard to plasmids carrying bla CTX-M-55 , as well as plasmid carrying bla CTX-M-14 , bla CTX-M-15 , and bla CMY-2 indicated that horizontal gene transfer in a real-life situation between pigs and pig farm workers happens with so low frequency that it will only be detected by very intensive sampling strategies. However, laboratory-based conjugation experiments shows that transfer of ESBL-encoding plasmids between E. coli strains is more frequent (Liu et al., 2019). Plasmids harboring bla CTX-M-55 were confirmed to be conjugative in our investigation with similar conjugation transfer frequencies as plasmids encoding bla CTX-M-27 . Whether transferability of plasmids between strains under field conditions is as efficient as in the laboratory remains to be investigated. The presence of bla CTX-M-55 on diverse plasmid replicon types with different transposons on various E. coli lineages show the heterogeneity of the plasmids carrying this ESBL gene, and suggest multiple introductions of these genes independently in each host (David et al., 2020). Together with the high variation in genetic context around the bla CTX-M-55 gene, these results show that bla CTX-M-55 in farmers is not necessarily horizontally acquired from close contact with animals and further that there is limited transmission of this type of resistant E. coli from pigs to farmers, and vice versa. The sample size is, however, not large enough to rule out that transmission occurs on a low level, given the fact that the plasmids are transferable.
The gene bla CTX-M-27 , was the second leading ESBL gene found in pigs and pig farmers. It was located in a relatively conserved genetic context in all isolates with the same flanking transposons in all the strains and all carried by the same conjugal IncF types of plasmid. The strains carrying the gene, on the other hand did not show resemblance. This is evidence of horizontal spread of the IncF plasmids types carrying bla CTX-M-27 in the study area. The same types of conjugative IncF plasmids were found in E. coli strains from pigs and farm workers ( Supplementary Table S1), however, the direction of transfer of plasmids cannot be established from the current study. The gene bla CTX-M-27 seems to be generally associated with conjugative plasmids of the IncF replicon type including FIA, FIB, and FII (Fernandes et al., 2020;Matsuo et al., 2020). The phylogenetic analysis of the plasmids revealed that they shared a larger core-genome cluster compared to the plasmids harboring bla CTX-M-55 . In contrast to the observation with the plasmids carrying bla CTX-M-55 , many of the bla CTX-M-27 plasmids shared reference plasmids in NCBI (plasmid accession CP049168 shared by 11 strains and KX008967 by nine strains), further supporting the low genetic variation for this type of plasmids. We identified a novel plasmid species harboring bla CTX-M-27 in three strains. Its closest reference in NCBI, Salmonella Typhimurium plasmid pB71, is a non-ESBL, multi-drug resistance plasmid, and it is likely that it has experienced an insertion of bla CTX-M-27 , however, other ways including intermediate plasmid species, could also have led to the formation of this plasmid. The size of the plasmid was approximately 150 kb (Supplementary Table S1 and Supplementary File 1). It is an IncF replicon type plasmid, and encoded multi drug resistance with mcr-3, aadA1, aac(3)-Iid, aph(3 )-Ib, aph(3 )-Ia, aph(6)-Id, mef (B), cmlA1, floR, sul3, and dfrA12 genes in addition to the ESBL gene. Overall, this novel plasmid is similar to the other bla CTX-M-27 plasmid reported in this study and elsewhere (Fernandes et al., 2020;Matsuo et al., 2020).
Moreover, our comparative analyses of contigs carrying the IncF types of plasmids also revealed another plasmid that showed complete identity with E. coli plasmid pESBL57, Accession MT230319 initially harboring bla CTM−M-97 in NCBI. Our analysis indicates that bla CTX-M-14 has replaced bla CTM−M-97 .
Five of the seven strains where bla CMY-2 was detected carried the gene in the chromosome. The other AmpC genes detected, bla CMY-42 and bla DHA-1 , were plasmid-mediated with consistent genetic contexts between isolates from pigs and humans, indicating that these genes are transmitted between hosts. In other studies, bla CMY-2 was reported to be chromosomally located (Harada et al., 2010), which suggests that this gene is less likely mobilizable and can only be spread together with the host strains. In many cases, the plasmids containing ESBL/Ampc β-lactamases also carried mcr-1 or mcr-3 genes encoding colistin resistance, as well as the presence of metal/biocide tolerance genes. Co-occurrence of ESBL/AmpC, colistin resistance as well as other antimicrobial resistance genes such as qnrS1, mef(B), mph(A), cmlA1, sul1, sul3, tet(A), dfrA12, aadA1/2, aac(3)-IIa was common in most isolates, and resistance genes were often occurring on the same plasmid (Supplementary Table S1). This corroborate previous resports that the majority of the strains analyses had multidrug resistant status (Dang et al., 2018). The plasmids identified therefore represent an important source of multidrug resistance, and due to their location on conjugative plasmids, they may transfer resistances of significant public health importance, and such plasmids need particular attention to track their global emergence, evolution and spread (Lupo et al., 2018;Zajac et al., 2019). Our findings further corroborate studies showing co-carriage of mcr-genes in many ESBL isolates (Lupo et al., 2018;Zajac et al., 2019) although, in those studies, the mcr-encoding plasmids are not the same as the plasmid carrying the ESBL genes.
In our study, ESBL/AmpC and other antimicrobial resistance genes were frequently co-located with heavy metal resistance genes on the same plasmids. Heavy metal exposure due to usage in livestock feed, e.g., use of zinc oxide in pig feed, can therefore co-select for antimicrobial resistance in bacteria (Ludden et al., 2019;Peng et al., 2020). In support of this, recent studies have reported that antimicrobial resistance is associated with tolerance to heavy metals existing naturally or used in food animal production, including zinc oxide and copper (Rensing et al., 2018;Cheng et al., 2019;García et al., 2020). Studies have also reported occurrence of metal tolerance genes in E. coli from post-weaning diarrhea in piglets with no documented exposure to metals, probably reflecting that over the years, metal resistance genes have been widely spread among E. coli in pigs (García et al., 2020).
The phylogenetic analysis of the isolates revealed wide genetic variation with no evidence of transmission of strains between pigs and farmers, and vice versa. All the isolates were of different MLST types, and none of them was predicted to be of known pathogenic sub-types (Supplementary Table S1). Thus, ESBL and AmpC encoding E. coli in pigs and pig farmers in Vietnam belong to unrelated commensal groups with both related and unrelated plasmids. A recent study has shown that E. coli causing bloodstream infections in the United Kingdom were not acquired from livestock, and that sharing of mobile elements between animals and humans was infrequent (Ludden et al., 2019). Moreover, a large diversity was reported in E. coli clones and plasmid types in France from food animals corroborating our findings (Lupo et al., 2018). Thus, our findings provide further evidence that plasmids carrying ESBL/AmpC β-lactamase encoding genes occur in genetically unrelated commensal E. coli. Strain sharing between the two hosts was infrequent and so was the mobile genetic element except for bla  , and this underlines that one cannot only focus on the resistant strains when analyzing spread patterns of resistance genes.
Despite that the bacterial strains from 2015 (Dang et al., 2018) and 2018 were epidemiologically related, i.e., matching fecal samples from workers and pigs collected from the same pig farms, they showed genetic variations and were predominantly commensals occurring independently in each host. Moreover, commensal E. coli are less studied than pathogenic E. coli with the former usually found genetically distinct (Ahmed et al., 2017). This may hide the occurrence of frequent transfer of mobile elements across strains as suggested by the conjugation experiments. Furthermore, as mobilizable genes, ESBL carriage is dynamic and could affect the consistency of their contexts in cross-sectional studies. However, the fact that sampling at two separate time frames did not show any significant fluctuations in the genetic context of ESBL genes in pigs and farm workers suggests that most of the ESBL genes are less frequently shared between the two hosts. Nevertheless, future longitudinal studies would provide further and improved insights.
A main limitation in this study is that only short-read technologies were used to generate the whole genome and plasmid sequences, Long and short-reads hybrid assemblies do not show significantly different outcomes (David et al., 2020), however, using combinations of the two techniques, closing of circular (plasmid) sequences is possible. To overcome the limitation, we used plasmid profiling to confirm the results of the predicted plasmids and their sizes, and conjugation experiment were further used to confirm the transferability of the plasmids between strains. We suggest that the relatedness of the plasmids encoding the resistances and the consistency of the genetic context around the ESBL/Ampc genes is considered along with the phylogenetic relationship between the strains.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.

AUTHOR CONTRIBUTIONS
AD and JO: conceptualization and supervision. YH and VB: methodology and software and data curation. YH, VB, AD, JO, SD, and DT: validation and writing-review and editing. YH: formal analysis, writing-original draft preparation and visualization. YH, SD, and DT: investigation. YH, VB, AD, and JO: resources. AD: project administration and funding acquisition. All authors contributed to the article and approved the submitted version.

FUNDING
We acknowledged the financial support from the Danish International Development Assistance (Danida) to the project "Health and Antibiotics in Vietnamese Pig Production" (grant DFC File No. 17-M06-KU).

ACKNOWLEDGMENTS
We would like to thank Jing Yuan Wang for the initial technical assistance during sequencing work on the 2015 E. coli strains.