Whole genome sequencing and characteristics of Escherichia coli with co-existence of ESBL and mcr genes from pigs

This study aimed to analyze three ESBL-producing E. coli co-harboring mcr and ESBL genes from a healthy fattening pig (E. 431) and two sick pigs (ECP.81 and ECP.82) in Thailand using Whole Genome Sequencing (WGS) using either Illumina MiSeq or HiSeq PE150 platforms to determine their genome and transmissible plasmids. E. 431 carrying mcr-2.1 and mcr-3.1 belonged to serotype O142:H31 with ST29 sequence type. ECP.81 and ECP.82 from sick pigs harboring mcr-1.1 and mcr-3.1 were serotype O9:H9 with ST10. Two mcr-1.1 gene cassettes from ECP.81 and ECP.82 were located on IncI2 plasmid with 98% identity to plasmid pHNSHP45. The mcr-2.1-carrying contig in E. 431 showed 100% identity to plasmid pKP37-BE with the upstream flanking sequence of IS1595. All three mcr-3.1-carrying contigs contained the ΔTnAs2-mcr-3.1-dgkA core segment and had high nucleotide similarity (85–100%) to mcr-3.1-carrying plasmid, pWJ1. The mobile elements i.e. IS4321, ΔTnAs2, ISKpn40 and IS3 were identified in the flanking regions of mcr-3. Several genes conferring resistance to aminoglycosides (aac(3)-IIa, aadA1, aadA2b, aph(3’’)-Ib, aph(3’)-IIa and aph(6)-Id), macrolides (mdf(A)), phenicols (cmlA1), sulphonamide (sul3) and tetracycline (tet(A) and tet(M)) were located on plasmids, of which their presence was well corresponded to the host’s resistance phenotype. Amino acid substitutions S83L and D87G in GyrA and S80I and E62K in ParC were observed. The blaCTX-M-14 and blaCTX-M-55 genes were identified among these isolates additionally harbored blaTEM-1B. Co-transfer of mcr-1.1/blaTEM-1B and mcr-3.1/blaCTX-M-55 was observed in ECP.81 and ECP.82 but not located on the same plasmid. The results highlighted that application of advanced innovation technology of WGS in AMR monitoring and surveillance provide comprehensive information of AMR genotype that could yield invaluable benefits to development of control and prevention strategic actions plan for AMR.


Introduction
Antimicrobial resistance (AMR) in bacteria is one the serious public health threats worldwide and has been complicated by emerging and spread of bacterial pathogens that are resistant to

DNA preparation and Whole Genome Sequencing (WGS)
Genomic DNA of E. 431, ECP.81 and ECP.82 were extracted by using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany). Plasmid DNA of ECP.81T and ECP.82T, the transconjugants obtained from ECP. 81 and ECP.82, were extracted by using Qiagen Plasmid Maxi Kit (Qiagen). The amount of all DNA samples was quantified according to Illumina sequencing sample requirements. The DNA samples were dissolved in 10mM Tris buffer to obtained at least 10 nM in 10 μl of minimum volume. The purity of DNA was determined at A260/280 and A260/ 230 using a Nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Delaware, USA).
The quantified DNA were subjected to Whole Genome Sequencing (WGS) by using Illumina platform MiSeq or HiSeq PE150 (Illumina, San Diego, CA, US). The libraries were prepared by using Nextera XT sample preparation kit and sequenced with 2×250 or 350 pairedend reads protocol on an Illumina platform (MiSeq or HiSeq PE150) at Omics Sciences and Bioinformatics Center (OSBC), Faculty of Science, Chulalongkorn University and Singapore Joint Venture & Sequencing Center Novogen AIT.

Analysis of DNA sequence data
Trimming of raw sequence reads was performed using the CLC Genomics Workbench software version 11.0.0 (CLC bio, Aarhus, Denmark) with default settings. De novo assembly of the trimmed reads was conducted using CLC Genomics Workbench or SPAdes 3.5.0 software [32]. Identification of Open Reading Frames (ORFs) and genome annotation of the assembled genetic elements was performed by using Prokka [33] and/or PATRIC 3.6.5 [34] with default settings.
WGS data sets were analyzed using Open-access bioinformatic webtool available at the Center for Genomic Epidemiology (http://www.genomicepidemiology.org/). In silico typing based on WGS of assembled genomes/contigs in FASTA format was carried out by using Serotype Finder 2.0 [35] with selected threshold of 90% identity and 60% total serotype gene length. Multi-Locus Sequence Typing (MLST 2.0) was applied for molecular typing of E. coli [36]. The MLST allele sequences and profile data used in MLST 2.0 were obtained from https://www.PubMLST.org and determined at 90% identity and 60% minimum length. The core genome MLST (cgMLST) profile of these isolates were investigated using cgMLSTFinder 1.1 from CGE databases [37]. Res Finder 4.1 [38] and Point Finder [39] were used to predict AMR genes and chromosomal point mutations from genomic sequences based on 90% identity. Plasmid replicon sequence analysis and identification of virulence genes were performed using Plasmid Finder 2.1 [40] and Virulence Finder 2.0 [41], respectively with the threshold of 90% identity and 60% minimum length. Genome assemblies of three E. coli isolates were further analyzed for genomic relatedness using the E. coli cgMLST scheme available at the BacWGSTdb 2.0 (http://bacdb.cn/BacWGSTdb/) [42].

Plasmid and phylogenetic analysis
Plasmid reconstruction from WGS data was conducted by using highly homologous complete plasmid sequence references available in NCBI. The alignment and assembly of sequences was performed using CLC Genomics Workbench or PATRIC 3.6.5 [34]. Plasmid sequences were annotated by Prokka [33] or PATRIC 3.6.5 [34] and manually edited. Then, annotated plasmid sequences were analyzed. Circular comparison between IncI2 plasmid carrying mcr-1 and most identical reference plasmids available at NCBI database was generated by CG view server [43]. All reference plasmids used for comparison in this study and obtained from NCBI database are listed in (S1 and S2 Tables).
The phylogenetic three of the mcr-1.1-carrying IncI2 plasmids was analyzed by comparing the two mcr-1.1-carrying IncI2 plasmids in ECP81 and ECP82 with 20 of mcr-1-carrying IncI2 plasmids from E. coli deposited in the GenBank database. The sequences were aligned using MUSCLE and phylogenetic interferences were obtained using the neighbor-joining method within the MEGA 10 software at set up of 1000 times bootstrap values to generate a majority consensus tree. The phylogenetic relationship of the core genome segment of mcr-3 carrying contigs was conducted by MEGA 10.0 program with maximum likelihood method (1,000 bootstrap replicates). All mcr-3 harboring contig sequences used to produce the phylogenetic trees were generated in this study and the mcr-3 reference genome sequences were obtained from the GenBank database (S2 Table).

Nucleotide sequence accession number
The complete sequences of mcr-1 carrying IncI2 plasmids from ECP81 and ECP82 in this study have been deposited at in GenBank under the accession numbers OK323956 and OK323955, respectively. Partial sequences of mcr-3-bearing plasmids from E431, ECP81 and ECP82 and partial mcr-2 sequence from E431 were deposited in GenBank database with the accession of OK323954, OK323952, OK323953 and OK323951, respectively. The cgMLST scheme based on the presence or absence of 2,513 genes in E431, ECP81 and ECP82 were assigned into cg13958, cg40289 and cg 136327, respectively. Based on the cgMLST  Table 2).

Presence and horizontal transfer of AMR determinants
WGS data analysis using Resfinder revealed the presence of AMR genes that were consistent with their resistance phenotype ( Table 2) In addition to ESBL and mcr genes, all three isolates concurrently contained various AMR genes conferring (but not limited) resistance to aminoglycoside (aac(3)-IIa, aadA1, aadA2b, aph(3'')-Ib, aph(3')-IIa and aph(6)-Id), macrolide (mdf(A)), phenicol (cmlA1), sulphonamide (sul3) and tetracycline (tet(A) and tet(M)) that were consistent with the AMR phenotypes. The single point mutations in quinolone resistant-determining regions (QRDR) of gyrA and parC resulting in amino acid substitution in GyrA and ParC were identified including S83L and  Based on in silico phylogenetic analysis, phylogenetic of mcr-1.1-harboring IncI2 plasmids revealed 8 distinct clades (Fig 1), of which each clade consisted of DNA sequences from different origins. Both plasmid from ECP.81 and ECP.82 were found on the same clonal linages and clonally related to pHNSHP45. The mcr-1.1-harboring IncI2 plasmids identified in ECP. 81 and ECP. 82 were also similar to the mcr-1 harboring IncI2 plasmids isolated from a cattle in Japan (pMRY16-002, Accession No. AP017622), a wild boar in China (pMRY15-131, Accession No. AP017614) and chicken in China (pGD16-131, Accession No. MN232187) (Fig 2).

mcr-harboring plasmids
The mcr-2.1 gene was identified in E.431, of which only short assembled contig of 2,142 bp was generated. This mcr-2.1 cassette contained a hypothetical protein at downstream and its All three porcine E. coli isolates in this study harbored mcr-3.1. The size of assembled contigs ranged from 8,412 to 11,172 bp. The WGS data showed that mcr-3.1 carrying plasmid in all the isolates had a backbone similar to mcr-3.1-carrying plasmid pWJ1 (Accession No. KY924928) (Fig 3). Five insertion sequences (IS) flanking the mcr-3.1 cassette were identified. The upstream IS were IS4321 (ECP.81 and ECP.82) and ΔTnAs2 (E.431, ECP.81 and ECP.82) and those at downstream were ISKpn40 and IS3 family transposase (ECP.81 and ECP.82) and IS26 (ECP.82). A core segment ΔTnAs2-NimC/NimA-mcr-3.1-dgkA and conjugation transfer genes (i.e. trb and traO, and traG) were presented in all three mcr-3.1-carrying plasmid. The additional insertion sequence, IS4321, was immediately upstream of the ΔTnAs2-NimC/ NimA-mcr-3.1-dgkA segment in mcr-3 carrying contigs of ECP.81 and ECP.82. None of the contigs contained other AMR genes and plasmid replicon type sequence identified in the Plas-midFinder database.
Phylogenetic tree of the core genome sequences was analyzed in all three mcr-3.1-carrying isolates and 12 mcr-3.1 harboring-plasmids deposited in the GenBank database (Fig 4). The members in this phylogenetic tree can be grouped into three clades. The mcr.3.1 carrying contigs in ECP. 81 and EC.P.82 had a core segment that were similar to pWJ1 (Accession No. KY924928) isolated from in E. coli from pigs in China and in Klebsiella pneumoniae from pigs in Nakhon Pathom, Thailand (Accession No. CO041095 and CP041104). However, the mcr-

PLOS ONE
Whole genome sequencing and characteristic of E. coli with co-existence of ESBL and mcr genes 3.1 carrying contig from fattening pig (E.431) was different from those in sick pigs and reference plasmids.

Discussion
The emergence and spread of colistin resistance mediated by plasmid-borne mcr genes has occurred globally, of which the majority was associated with bacterial pathogens including E. coli., Moraxella spp., Klebsiella spp., Salmonella spp., Enterobacter spp. and Citrobacter spp [44]. The mcr-carrying plasmids frequently co-harbor other resistance determinants e.g. bla-VIM-1 [17] and bla CTX-M1 [45]. These raise a particular challenge for treatment of MDR Gramnegative bacterial infection due to limited availability of effective antibiotics. E. coli, particularly from pig, has been a major species among the mcr-carrying bacterial strains [10,11,46].

PLOS ONE
Whole genome sequencing and characteristic of E. coli with co-existence of ESBL and mcr genes In the present study, we performed WGS analysis for molecular characterization of three ESBL producing E. coli additionally carrying mcr originated from pigs.
The E. coli isolates from healthy fattening pig, E.431, carrying mcr-2/mcr-3 belonged to serotype O142:H31 and had sequence type, ST29. The other two isolates originated from clinically sick pigs, ECP.81 and ECP.82, with coexistence of mcr-1 and mcr-3 appeared to be serotype O9:H9 with sequence type ST10. This supports previous studies demonstrating that the ST10 E. coli served as an important reservoir of mcr-1 [46,47]. The clonal group ST10 of mcrcarrying E. coli was reported previously in the clinical isolates from humans [48] and food-producing animals (i.e. poultry and swine) in Thailand [49] and other Asian countries [50,51] as well as some European countries [16,52]. It was shown that the ST10 strains harboring mcr commonly carried additional-multiple AMR genes e.g. bla CTX-M1 and bla SHV-12 [16], in agreement with the isolates in this study.
The conjugation experiments and plasmid analysis confirmed that mcr-1.1 of ECP.81 and ECP. 82 was located on IncI2 plasmid, in agreement with a previous study [10]. The mcr-1 gene and its variants have been identified to be associated with four major plasmid incompatibility groups i.e. IncX4, IncI2, IncHI2 and ColE10-like [46,47]. In Thailand, a previous study in clinical CRE isolates carrying mcr-1 revealed that IncX4 was predominant replicon type, followed by IncI2 [54]. Taken together, these results suggest the circulation of IncI2 plasmid harboring mcr-1 among bacterial species of animal and human origins. It was previously shown that IncI2 plasmid can migrate between different bacterial species and E. coli serves as a potential carrier of this plasmid replicon [55].
ISApl1 has been the main driver of mobilized colistin resistance gene, mcr-1, via horizontal gene transfer [56]. Four genetic contexts surrounding mcr-1 cassette were identified including the composite transposon Tn6330 (ISApl1-mcr-1-orf-ISApl1) and a single-ISApl1 located upstream (ISApl1-mcr-1-orf); a single-ISApl1 (mcr-1-orf-ISApl1) located downstream and a structure lacking both copies of ISApl1 (mcr-1-orf) [56]. In present study, the two mcr-1.1 carrying IncI2 plasmids from ECP.81 and ECP. 82 had the structure of ISApl1-mcr-1-orf. The results from the conjugation experiment and plasmid analysis in donor (ECP. 82) and transconjugant (ECP. 82T) confirmed that ISApl1 is associated with the mobilization of mcr-1. This transconjugant, ECP. 82T, carried mcr-1.1 with translocation of ISApl1 to its upstream or ISApl1-mcr-1-orf as observed in its donor strain. A recent study hypothesized that mcr-1 translocation could be mediated through a circular intermediate that mediates the insertion of the mcr-1 gene cassette into other bacterial plasmids or genome [19].
The mcr-2 gene was first described on IncX4 plasmid in E. coli isolated from calves and piglets in Belgium [26]. However, the prevalence of mcr-2 was low and limited to E. coli, Moraxella and K. pneumoniae from pigs, claves, bird, and chicken in some countries e.g. China [57], Belgium [26], USA and Great Britain [58]. Up to date, six mcr-2 variants (Accession no. LT598652, MF176239, NG065452, MT757845, MT757842 and MT757844) were deposited on to GenBank database but only one plasmid sequence, pBK37-BE (Accession no. LT598652), was characterized. In our study, the mcr-2.1 gene in E.431 showed 100% identity to pBK37-BE, with IS1595 located upstream. However, analysis of genetic contexts of the genomic location mcr-2 was limited due to the limited size of the contig and non-transferability in the conjugation experiment.
ESBL genes are frequently located on conjugative plasmids and has been previously shown to be associated with several plasmid incompatibility groups e.g. IncF, IncI, IncH and IncA/C plasmids [62]. IncF is a commonly described plasmid type identified in humans and animals and E. coli is considered a major reservoir of this plasmid [62]. A previous study in Korea demonstrated the dissemination of bla CTX-M14 gene driven by IncF plasmid [63]. A study in China reported that bla CTX-M55 in E. coli from pets and food animals were linked to IncI2 plasmid [64]. Both IncHI1 and IncHI2 plasmids are frequently associated with resistance to multidrugs e.g. sulphonamides, aminoglycosides, tetracyclines and streptomycin in additions to cephalosporins. The bla CTX-M2 and bla TEM-1 genes were previously reported to be on a IncHI2 plasmid [65]. In this study, many plasmid replicon types were detected in the three E. coli strains, of which the most frequent plasmid replicons were IncFIB, IncFII and IncHI2. The bla CTX-M14 gene was detected in all E. coli isolates. The bla CTX-M55 gene was found only in ECP.81 and bla-TEM-1B was identified in both isolates from ill pigs. Four mobile elements (i.e. Tn3 transposase, Tn903, IS1 and IS26) were identified in flanking regions of ESBL genes. This supports that, in addition to conjugative plasmid, the dissemination of ESBL genes may occur through transposons.
Moreover, the co-transfer of mcr-1.1 and bla TEM-1B was observed in ECP. 82 and the cotransfer of mcr-3.1 and bla CTX-M55 was observed in ECP. 81. Using data from WGS analysis of the donor and plasmid analysis of transconjugants, plasmid reconstruction did not show the presence of other AMR genes. This implies that the genes encoding colistin resistance and ESBL production (i.e. mcr-1.1/ bla TEM-1B and mcr-3.1/ bla CTX-M55 ) were not colocalized on the same plasmid with other AMR genes. These observations are in agreement with a previous study in China demonstrating the co-transfer of mcr-1.1, mcr-3.5, bla NDM-5 and rmtB could occur even though they were not located on the same plasmid [66]. The co-transfer of many clinically important AMR genes at the same time is a big challenge for clinical treatment and disease controlling in both human and veterinary medicine. Such co-transfer of multiple resistance genes could provide the fitness advantage to the host strains [67]. Therefore, using colistin alone may not be the only reason of mcr dissemination. This is additionally supported by the use of ampicillin as selective pressure in conjugative experiment in this study, where colistin resistance gene(s) was co-selected.
In conclusion, the pandemic spread of mcr genes is attributed to several factors e.g. MDR profile in colistin-resistant strains, host fitness adaptation, and co-selection by other antimicrobials. Co-transfer of mcr and ESBL genes has generated a challenging for clinical treatment and a serious concern to global health. Judicious use of antibiotics in livestock and human medicine should be encouraged. Characterization of the coexistence ESBL and mcr genes in three E. coli isolates by using WGS approach provided large database for rapid analysis of genetic context of resistance determinants and supported better understanding on the dissemination of these resistance markers among bacteria from different sources. However, a shortread sequencing platform was used in this study and deemed insufficient to elucidate complete sequence and genetic organization of plasmid. Therefore, long read sequencing approach (e.g. Pacific Biosciences and Oxford Nanopore Technologies) producing long reads with higher-quality genome assemblies and thus improving de novo assembly is suggested for further study.
Supporting information S1