Repeated Occurrence of Mobile Colistin Resistance Gene-Carrying Plasmids in Pathogenic Escherichia coli from German Pig Farms

The global spread of plasmid-mediated mobile colistin resistance (mcr) genes threatens the vital role of colistin as a drug of last resort. We investigated whether the recurrent occurrence of specific E. coli pathotypes and plasmids in individual pig farms resulted from the continued presence or repeated reintroduction of distinct E. coli strains. E. coli isolates (n = 154) obtained from three pig farms with at least four consecutive years of mcr detection positive for virulence-associated genes (VAGs) predicting an intestinal pathogenic pathotype via polymerase chain reaction were analyzed. Detailed investigation of VAGs, antimicrobial resistance genes and plasmid Inc types was conducted using whole genome sequencing for 87 selected isolates. Sixty-one E. coli isolates harbored mcr-1, and one isolate carried mcr-4. On Farm 1, mcr-positive isolates were either edema disease E. coli (EDEC; 77.3%) or enterotoxigenic E. coli (ETEC; 22.7%). On Farm 2, all mcr-positive strains were ETEC, while mcr-positive isolates from Farm 3 showed a wider range of pathotypes. The mcr-1.1 gene was located on IncHI2 (Farm 1), IncX4 (Farm 2) or IncX4 and IncI2 plasmids (Farm 3). These findings suggest that various pathogenic E. coli strains play an important role in maintaining plasmid-encoded colistin resistance genes in the pig environment over time.


Introduction
The emergence and spread of multidrug-resistant bacteria is a rising problem that threatens the effective treatment of infectious diseases in humans and animals [1].Escherichia (E.) coli is a ubiquitous Gram-negative bacterium that is regularly found in the intestinal tract of humans and animals.Intestinal pathogenic E. coli (InPEC) strains can cause a number of different diseases, including diarrhea.
In pigs, neonatal and post-weaning diarrhea is a widespread and often severe disease, resulting in significant economic losses in the swine industry worldwide [2].Certain E. coli pathotypes are associated with causing enteric diseases in piglets, e.g., enterotoxigenic E. coli (ETEC) and atypical enteropathogenic E. coli (aEPEC) [3].E. coli isolates producing the Shiga toxin subtype Stx2e, encoded by the stx2e gene, are the causative agent of edema disease in weaned piglets [4].Colistin has been widely used to treat intestinal infections in swine caused by InPEC [5].In 2015, the use of colistin came under scrutiny due to the emergence of colistin-resistant bacteria due to a transferable colistin resistance gene [6].A recent study from 2022 reported that 51.9% of 662 veterinarians surveyed stopped the use of colistin and 33.4% reduced their usage.The main indication for the use of colistin was gastrointestinal disease in pigs [6].The increasing occurrence of antimicrobial resistance (AMR) against this last resort antibiotic incites new debates for additional regulations of its usage, especially in veterinary medicine [7].
This study investigated the occurrence and genomic location of mcr genes in pathogenic E. coli isolates obtained from three German pig farms over at least four years, which were chosen from a comprehensive in-house database.The genomes of a representative set of isolates were sequenced to determine the presence of virulence-associated genes (VAGs) characteristic for intestinal pathogenic E. coli (InPEC) pathotypes and AMR genes as well as their plasmid location.Based on core genome and plasmid comparisons, the repeated occurrence of distinct E. coli clones, as well as distinct resistance and virulence plasmids on different farms, was examined.

Study Inclusion Criteria for Farms
For this retrospective survey on the repeated occurrence of mcr-positive porcine pathogenic E. coli, suitable pig farms were selected based on specific criteria.From our database of more than 3000 registered pig farms in Germany that had sent samples for molecular typing of E. coli isolates in the past, we selected only farms from which at least 25 pathogenic E. coli isolates had been obtained and preserved over the years (n = 23 farms).This was based on our recently published collection of 10,573 E. coli isolates, each harboring at least one of ten VAGs, which were tested using PCR for the presence of mcr-1 to mcr-10 genes [30].Briefly, E. coli isolates were obtained from feces or mucosal swabs (rectum or small intestine).Upon arrival at the laboratory, the samples were streaked for single bacterial colonies on blood agar plates (blood agar base, Merck Chemicals, Darmstadt, Germany) containing 5% sheep blood and on Gassner agar (sifin diagnostics GmbH, Berlin, Germany).After approximately 18 h of incubation at 37 • C, up to six morphologically different, putative E. coli colonies were picked per sample and tested for the presence of VAGs.More details of sample collection and processing were published recently [30].
Subsequently, pig farms with less than 25 pigs sampled (n = 11) or no detection of mcr genes (n = 5) were excluded from the study, as well as farms with no isolates for six consecutive years (n = 3).Finally, only farms were selected where mcr-positive E. coli had been isolated from pig fecal samples for at least four consecutive years.

Whole Genome Sequencing
The genomic DNA of E. coli bacteria was extracted using the Master Pure™ DNA Purification Kit (Biozym Scientific GmbH, Hessisch Oldendorf, Germany).Bacterial genomes were sequenced using an Illumina MiSeq sequencer (MiSeq Reagent Kit V.3; Illumina Inc., San Diego, CA, USA) via multiplexing of 30 samples per flow cell using 2 × 150 bp paired-end reads to achieve an average coverage of 90-fold.Quality control, including contamination removal and adapter trimming, was carried out using an in-house pipeline.De novo assemblies were generated using the SPAdes Genome Assembler (v3.15.5) with the "-isolate" flag [31].The Bakta pipeline (v1.8.2) was employed using species-specific databases for genomic annotation of the bacterial genomes [32].
Multilocus sequence types (STs) were determined by applying MLST 2.0 (https:// cge.food.dtu.dk/services/MLST/,accessed on 22 November 2023), which is based on the Achtman seven-gene MLST scheme that includes seven housekeeping genes (adk, fumC, gyrB, icd, mdh, purA and recA).E. coli isolates were defined as clonal when they displayed the same sequence type in addition to identical virulence and resistance gene profiles.
The core genome was calculated using Roary [37].Phylogenetic analysis was performed by using the web-based tool ClermontTyper (IAME, http://clermontyping.iameresearch.center/index.php,accessed 18 December 2023), allowing us to assign tested strains to E. albertii, E. fergusonii, Escherichia cryptic clades I-V, E. coli sensu strico as well as to the main phylogroups A, B1, B2, C, D, E, F and G [38,39].The remaining non-sequenced E. coli isolates were tested in a quadruplex PCR, allowing for the assignment to the eight main phylogroups A, B1, B2, C, D, E, F, G and Escherichia cryptic clades I-V [39,40].

Genomic Location of Virulence-Associated Genes and mcr Genes, Plasmid Analysis
To determine the location of mcr genes, two methods were used.Whole genome sequence data were analyzed for the location of AMR and virulence genes using the tool "Chromosome & Plasmid Overview" implemented in the software package Ridom SeqSphere + (http:// www3.ridom.de/seqsphere,accessed 22 January 2024).To identify plasmid replicon types, PlasmidFinder 2.1 (https://cge.food.dtu.dk/services/PlasmidFinder/,accessed 22 January 2024) and Ridom SeqSphere+ were employed.The BacWGST database (http://bacdb.cn/BacWGSTdb/Tools.php,accessed on 17 December 2023) and RidomSeqSphere+ were applied to detect closely related plasmids via sequence comparison.To illustrate circular comparisons between the plasmids, we used the blast ring image generator software BRIG Version 0.95 [41].
To confirm the results of sequence analysis regarding the genomic location of mcr-1 genes, an additional method was employed.The location of mcr-1 genes in 38 E. coli isolates was determined via the S1 nuclease digestion of genomic DNA and pulsed-field gel electrophoresis (PFGE), followed by Southern blot hybridization (SBH).For this method, digoxigenin-labeled DNA probes ("DIG luminescent detection Kit" Boehringer Mannheim GmbH, Mannheim, Germany) were generated, targeting the PCR fragments specific for the mcr-1 gene [10].

Farm Selection
Farms in Hesse (n = 2) and North Rhine Westphalia (n = 1), Germany, met the chosen study inclusion criteria.Farm 1 sent samples (41 × feces, 1 × intestine + feces, and 1 × E. coli isolate) from May 2002 to October 2019.Samples were obtained from 43 pigs, most of them suffering from diarrheal disease.From this sample material, 50 E. coli isolates were cultivated that were positive for at least one VAG associated with certain pathotypes of InPEC as determined by PCR [30].Twenty-two (44%) of these E. coli isolates, obtained from July 2009 to September 2012, tested positive for mcr-1 genes (Table 1).
Farm 2 submitted samples from 69 pigs between June 2004 and February 2021.Samples were mainly feces (n = 43), followed by directly submitted E. coli isolates (n = 22) and by isolates obtained from the intestine and feces (n = 4).From these samples, 78 VAG-positive E. coli isolates were stored.Twenty-four (30.8%) isolates, obtained from July 2013 to February 2018, proved mcr-1-positive (Table 1).From October 2014 to April 2019, 26 VAG-positive E. coli were isolated from 25 sampled pigs of Farm 3. Most isolates (n = 23) were received through submissions from other veterinary diagnostic laboratories for further molecular typing in our institute, supplemented by two fecal samples.In total, 16 (61.5%)mcr-positive pathogenic E. coli were retrieved over more than four years (Table 1).
In total, 154 intestinal pathogenic E. coli isolates were obtained from 137 pigs housed on the selected three farms from May 2002 to February 2021.The total number of mcr-positive isolates was 62 (40.3%).

E. coli Pathotypes
A prediction of pathotypes was conducted for all 154 E. coli isolates included in this study based on the presence of certain VAGs obtained from PCR analysis, as described previously [30]: adhesive fimbriae E. coli (in the following termed AdhF-Ec), positive only for at least one adhesive fimbriae gene (faeG, fanA, fasA, fedA and fimF41a); AEEC (often also referred to as atypical EPEC), positive for eae; EDEC, positive for fedA and stx2; ETEC, positive for at least one adhesive fimbriae gene (faeG, fanA, fasA, fedA and fimF41a) and at least one enterotoxin gene (eltB-Ip, estap and estb); ETEC-like, positive only for at least one enterotoxin gene (eltB-Ip, estap and estb); ETEC/STEC hybrid (in the following simply termed ETEC/STEC), positive for at least one adhesive fimbriae gene (faeG, fanA, fasA, fedA and fimF41a) and at least one enterotoxin gene (eltB-Ip, estap and estb) and stx2; STEC, positive only for stx2.

Virulence Associated Genes and Virulence Plasmids
The genomes of 87 E. coli isolates were sequenced to conduct further molecular analysis, specifically regarding VAGs, as well as mcr detection and genomic localization.The isolates included 38 mcr-1.1-positive, 1 mcr-4.8-positiveand 48 mcr-negative E. coli isolates (Table 2).At least one isolate per year and farm was selected for genome sequencing.If there were several mcr-positive isolates of different pathotypes per year and farm, at least one representative isolate per pathotype was selected.The mcr-negative isolates (one per pathotype) were additionally sequenced from the years with detected mcr occurrence.
The majority of the VAGs identified via PCR (Section 3.1) were also identified in the genomic data (Table 2).The adhesive fimbriae F18 (encoded by the fedA gene) were further classified into F18 fimbrial subtypes F18ab (fedAab) and F18ac (fedAac).In total, 19 isolates encoded for the F18ab subtype, while 21 isolates encoded for F18ac.One isolate carrying the fedA gene was not further typeable (Farm 1, IHIT52950).EDEC, defined as isolates positive for fedA and stx2e, was predicted in 94.7% of all F18ab-positive bacteria.Isolates encoding for F18ac mostly harbored enterotoxin genes (ETEC; 61.9%) or enterotoxin and Shiga toxin genes (ETEC/STEC; 33.3%).All of the isolates that tested positive for F4 fimbriae (encoded by faeG) in the PCR were identified as the F4ac subtype (faeGac) via genome analysis.This subtype is known to be the most prevalent in piglets with post-weaning diarrhea [2].Genomic data from 30 isolates positive for the stx2 gene via PCR revealed that they encoded for Shiga toxin subtype Stx2e (stx2e), which plays a pivotal role in the pathogenesis of edema disease in swine [4].
Over the years, E. coli isolates from all three farms repeatedly harbored similar virulence plasmids that either carried fimbrial genes, enterotoxin genes or both.Two reference plasmids were identified in the NCBI database, which showed high similarities to these virulence plasmids.To illustrate similarities between the study and reference plasmids, we selected one representative isolate per year and farm for BRIG analysis (Figure 1).Table 2 provides details on the E. coli strains used as representative isolates, including isolation year, pathotype, and ST.
Reference plasmid p15ODTXV (NCBI Reference Sequence: MG904998.1)was identified in an ETEC/STEC strain isolated from a diarrheic pig in Switzerland in 2014/2015.It carried estap, estb, fedA and hlyDBAC virulence genes.Representative plasmids identified in four isolates from Farm 3 were highly similar to the IncFII/IncX1 multivirulence reference plasmid used in Figure 1A.Virulence plasmids detected in Farm 1 (n = 19) and Farm 2 (n = 10) resembling p15ODTXV showed varying similarities over the years and mostly carried either fedAab (n = 15) or fedAac (n = 11; Table S1) genes.
Plasmids with high similarity to the IncFII reference plasmid pUMNK88_K88 (NCBI Reference Sequence: CP002730.1)were found in all three farms.In 2007, one ETEC strain from a pig with neonatal diarrhea in the USA was found to contain the pUMNK88_K88 plasmid, which carried faeG as a virulence gene.BLAST analysis revealed two very similar IncFII plasmids harbored by two strains positive for faeGac from Farm 1 and Farm 3, respectively (Table S1).In Farm 2, IncFII plasmids encoding for faeGac were found in ten ETEC strains (all ST100) isolated in the years 2005 to 2018 (Figure 1B).A comparison of the plasmids from this study with reference plasmids demonstrated structural resemblance of virulence plasmids over several years within and, in part, also across farms (e.g., plasmid pUMNK88_K88).plasmids were identified in the NCBI database, which showed high similarities to these virulence plasmids.To illustrate similarities between the study and reference plasmids, we selected one representative isolate per year and farm for BRIG analysis (Figure 1).Table 2 provides details on the E. coli strains used as representative isolates, including isolation year, pathotype, and ST.Reference plasmid p15ODTXV (NCBI Reference Sequence: MG904998.1)was identified in an ETEC/STEC strain isolated from a diarrheic pig in Switzerland in 2014/2015.It carried estap, estb, fedA and hlyDBAC virulence genes.Representative plasmids identified in four isolates from Farm 3 were highly similar to the IncFII/IncX1 multivirulence reference plasmid used in Figure 1A.Virulence plasmids detected in Farm 1 (n = 19) and Farm 2 (n = 10) resembling p15ODTXV showed varying similarities over the years and mostly carried either fedAab (n = 15) or fedAac (n = 11; Table S1) genes.
Plasmids with high similarity to the IncFII reference plasmid pUMNK88_K88 (NCBI Reference Sequence: CP002730.1)were found in all three farms.In 2007, one ETEC strain from a pig with neonatal diarrhea in the USA was found to contain the pUMNK88_K88 plasmid, which carried faeG as a virulence gene.BLAST analysis revealed two very similar IncFII plasmids harbored by two strains positive for faeGac from Farm 1 and Farm 3, respectively (Table S1).In Farm 2, IncFII plasmids encoding for faeGac were found in ten ETEC strains (all ST100) isolated in the years 2005 to 2018 (Figure 1B).A comparison of the plasmids from this study with reference plasmids demonstrated structural resemblance of virulence plasmids over several years within and, in part, also across farms (e.g., plasmid pUMNK88_K88).
Table S2 provides a detailed distribution of all additionally detected VAGs among whole genome-sequenced isolates, sorted by farms and the categories: adhesion, tissue Table S2 provides a detailed distribution of all additionally detected VAGs among whole genome-sequenced isolates, sorted by farms and the categories: adhesion, tissue damage (hemolysin/toxin genes), invasion and protection, iron acquisition and secretion system.

Multilocus Sequence Types, Clonotypes, Phylogenetic Groups
In the phylogenetic tree of all whole genome-sequenced E. coli isolates in this study, the majority of isolates from each farm clustered together, irrespective of the year of isolation (Figure 2).Clustering was also observed for mcr-positive isolates on each farm.
Overall, 20 known multilocus sequence types (STs) were determined (Table 2).ST1 was the predominant ST (60.9% of the isolates) on Farm 1, while the most prevalent STs on farms 2 and 3 were ST100 (n = 21; 48.8%) and ST86 (n = 7; 43.8%), respectively (Figure 2).Three mcr-1-positive ETEC isolates from Farm 2 (isolated between June 2010 and June 2011) belonged to ST131.This sequence type has previously been detected in ETEC strains isolated from pigs with diarrhea in Spain [3].It is also known as the predominant E. coli lineage among extraintestinal pathogenic E. coli (ExPEC) worldwide, frequently associated with an ESBL phenotype and multidrug resistance [42].
Phylogroups A (n = 58; 37.7%) and D (n = 53; 34.4%) were found to be the most common phylogroups among all 154 E. coli isolates (Figure 3).While the majority of E. coli isolates of Farm 1 were assigned to phylogroup D (n = 34; 68%), phylogroups A (n = 45; 57.7%) and B1 (n = 16; 61.5%), which were most frequent in isolates from Farms 2 and 3, respectively.Overall, only four, three and nine isolates were assigned to phylogenetic groups B2, C and E. No isolate belonged to phylogroup F or G.

Genomic Location of mcr Genes and Plasmid Analysis
The genomic location of mcr-1.1 in 38 whole genome-sequenced E. coli isolates was determined via Southern blot hybridization (SBH) of S1 nuclease-digested whole-cell DNA and analysis of genomic sequence data.We identified mcr-1.1 on four different plasmids in this study.S1 nuclease PFGE and SBH initially revealed mcr genes on plasmids with approximate sizes of 250 kbp (n = 13), 35 kbp (n = 19) and 65 kbp (n = 3).Three isolates (IHIT34315, IHIT34769, and IHIT36427) did not reveal the genomic location of mcr genes through SBH despite repeated attempts.
The BacWGST database was used to identify closely related MCR-1 plasmids, which were then used as reference plasmids to illustrate circular comparisons with MCR-1 plasmids of representative isolates of this study (Figure 4).mcr-1.1 on IncX4 plasmids.E. coli isolates obtained from Farm 3 carried the mcr-1.1 gene on IncX4 (n = 4), IncI2 (n = 3) and IncI2 (delta) (n = 1) plasmids.The mcr-4.8 gene was located on a ColE10 plasmid, as previously reported [30].
The BacWGST database was used to identify closely related MCR-1 plasmids, which were then used as reference plasmids to illustrate circular comparisons with MCR-1 plasmids of representative isolates of this study (Figure 4).For comparison, we selected one mcr-positive representative E. coli isolate per year and farm.When two different pathotypes tested positive for mcr in the same year, both positive pathotypes were included.Six representative plasmids were detected in Farm 1 and identified as similar to two IncHI2 reference plasmids.The plasmids were found in IHIT45339 (EDEC; 2009), IHIT32406 (EDEC; 2010), IHIT46534 (ETEC; 2010), IHIT45341 (ETEC; 2011), IHIT45342 (EDEC; 2011) and IHIT46538 (EDEC; 2012), with nucleotide sequence identities ranging from 98.8% to 99.99% and coverage ranging from 90% to 97% compared to the references.The reference plasmids harboring mcr-1.1 had been identified in E. coli strains isolated from swine in China (pDJB-3; NCBI Reference Sequence: MK574666.1;used as a reference plasmid in Figure 4A) and broiler chicken in China (pGD27-70; NCBI Reference Sequence: MN232195.1).
An E. coli isolate obtained from influent municipal wastewater in Japan harbored mcr-1.1-carryingIncI2 plasmid pC2 (NCBI Reference Sequence: LC473131.1),which was used as a reference in Figure 4D.This plasmid was highly similar to plasmids from Farm 3 (isolates IHIT47047, IHIT34769, IHIT48353), with identities ranging from 92.66% to 99.99% and coverage greater than 99%.However, IHIT48352 harbored one IncI2 (delta), which was similar to one mcr-1.1-carryingplasmid from a Salmonella enterica isolate obtained from a child in Ecuador (p778; 98.4% identity, coverage 100%; NCBI Reference Sequence: MN746292.1).None of the selected reference plasmids contained additional resistance genes besides mcr-1.1.

Discussion
The global dissemination of colistin resistance genes in bacteria from humans, animals and the environment has been reported over the last years [24,44].The occurrence of mcr genes has been investigated in various studies, especially those conducted concerning pig and poultry production [45][46][47].However, data on recurrent mcr-mediated resistance in livestock farms are scarce [48,49].We were not able to identify clones of pathogenic E. coli over the years on the specific farms.However, our study described three German pig farms with the repeated occurrence of highly similar mcr-carrying plasmids in pathogenic E. coli isolates over two to more than four years, respectively.
In our study, ETEC was the prevalent pathotype for all collected isolates (78/154; 50.7%) as well as for mcr-1-positive isolates (30/61; 49.2%).The most common pathotypes for each farm also coincided with the prevalent pathotypes of mcr-1-harboring E.coli.ETEC has been reported to be the most common cause of post-weaning diarrhea (PWD) in swine, but is also known as a pathogen in humans [5,50].An epidemiological study from Spain reported a significant association of ETEC (67%) with the occurrence of PWD in swine [3].That study investigated 481 E. coli isolates from diarrheic pigs, of which 123 (25.6%) strains carried the mcr-1 gene, with 57.7% belonging to the ETEC pathotype.This study also investigated the prevalence of the Shiga toxin gene stx2, which was found to be lower than in our study (10% vs. 35%) [3].
We identified IncFII/IncX1 and IncFII plasmids as the most frequent virulence plasmid types among the isolates from all three farms over the years.VAGs like eltB-Ip, estap, estb, and fedABCEF have previously been reported to be encoded on IncF or IncFII/IncX1 plasmids of E. coli strains isolated from diarrheic pigs [51,52].Virulence plasmids have occasionally been described in ETEC and ETEC/STEC strains [52,53], but to the best of our knowledge, not in individual pig farms over the years.The striking similarity of certain virulence plasmids in single farms, such as in Farm 3 for four consecutive years (Figure 1A), suggests a local clonal distribution of virulence plasmids.However, it should be taken into account that specific virulence plasmids can be distributed across farms (Figure 1B) and countries.In summary, our data do not support clones (according to the strict definition used in our study design) of E. coli strains carrying specific virulence plasmids over the years.
Effelsberg et al. investigated the occurrence of mcr-1 to mcr-5 genes in 318 porcine fecal samples from 81 pig farms in northwest Germany collected from March 2018 to September 2020 [54].Overall, ten farms (12.3%) provided fecal samples that contained E. coli harboring mcr-1.Two farms showed repeated presence of mcr-1-positive strains from two different dates of sampling.A retrospective study examined 436 boot swab and pooled fecal samples from 58 German pig-fattening farms for the presence of mcr-1 and mcr-2 [29].The mcr-1 gene was detected in 43 (9.9%)E. coli isolates obtained from 15 (25.9%) farms, which were almost evenly distributed in northern/western (25%), southern (25%), middle (21%) and eastern (36%) Germany in 2011 to 2012.Considering the presence of mcr-positive pathogenic E. coli strains on the farms examined in our study, the prevalence was high, with 42%, 30.8% and 61.5% for farms 1, 2 and 3, respectively.A longitudinal study from Thailand investigated the occurrence of mcr genes in one representative pig farm from 2017 to 2020 after the cessation of prophylactic colistin usage [48].Samples were taken from pigs (n = 70), farm workers (n = 50) and wastewater (n = 50).While mcr-1-positive isolates (n = 4; 8%) were detected in humans only in 2017, the prevalence of colistin-resistant strains was 28.6% in pigs, with a declining trend over the years, and 18% in wastewater.Another study obtained mixed bacterial cultures (n = 35) from the environment of three German pig farms from 2011 to 2012 [55].Seven E. coli isolates tested positive for mcr-1-harboring IncX4 plasmids.The positive samples were obtained from boot swabs, barn dog feces, stable flies and manure.In our study, we did not expect the recurrent identification of highly similar mcr-positive pathogenic E. coli isolates in three German pig farms over more than four years.This is a concerning observation because it suggests that the occurrence of resistant strains over the years may be largely undetected due to missing systematic observational data.
The predominant mcr-1.1-harboringplasmid types in our study were IncHI2 for farm 1, IncX4 for Farm 2 and both IncX4 and IncI2 plasmids for Farm 3. A systematic review by Matamoros et al. reported a total of 13 plasmid incompatibility types for mcr-1carrying plasmids for 217 Enterobacteriaceae isolated from human, animal and environmental samples [56].The majority of plasmid types were IncX4 (35.2%),IncI2 (34.7%) and IncHI2 (20.5%), with a significant geographical clustering of IncHI2 plasmids in Europe and a regional spread of IncI2 plasmids in Asia.A recent study from France investigated the occurrence of mcr-1-mcr-5 and mcr-9 genes in colistin-resistant E. coli isolates obtained from over 1500 goats of 80 breeding and five fattening goat farms [57].In total, 149 mcr-1-positive E. coli were identified, with 146 mcr-1 genes located on either IncX4 (38.9%) or IncHI2 (26.8%) plasmids and on the chromosome (32.2%).The mcr-1-carrying plasmids of types IncX4 and IncHI2 were never detected on the same farm in that study.
Bok et al. analyzed 274 E. coli strains isolated from healthy post-weaning piglets and sows in Poland with the phylogenetic assignment of isolates mainly to phylogroups A (37.6%) and B1 (33.2%) [60].This is in line with our findings of phylogroups A and B1 being the most prevalent phylogroups for Farm 2 and Farm 3, respectively.Another study from Thailand reported phylogroups A (44.3%), B1 (34.4%) and D (14.8%) being the predominant phylogroups for mcr-harboring E. coli from slaughtered pigs in 2014-2015 [61].Here, we observed similar prevalences for collected mcr-positive E. coli isolates.
Our study has some limitations.This retrospective survey was based on the receipt of samples for diagnostic purposes and allows for bias due to the random nature of sample submissions.Only porcine E. coli isolates that proved positive for certain VAGs were documented and stored, leaving out non-pathogenic E. coli.In addition, viable but nonculturable (VBNC) E. coli were not studied.Bacteria in the VBNC state are viable but unable to grow on nutrient culture media.Pathogenic E. coli entering the VBNC state as a survival strategy has been reported [62], and would have been missed in our study design.Furthermore, comprehensive data were not available for all three farms, including information on farm characteristics such as potential external contamination, management practices related to animal health, and previous veterinary treatments, including the use of colistin or other antimicrobials.
The collection of a large pool of pathogenic E. coli isolates over a period of more than twenty years is a major advantage for the selection and evaluation of sampled pig farms in Germany.To the best of our knowledge, this study is the first to present phenotypic and molecular data on mcr-mediated resistance and intestinal pathogenic E. coli isolates from individual pig farms over an extended period of time.

Conclusions
We describe three German pig farms in which pathotypes of E. coli, including ETEC and EDEC, and various MCR plasmids have been found repeatedly over the years.Isolates obtained on each farm over 17.4, 26.6, and 4.5 years, respectively, differed from each other in their multilocus sequence types and/or VAG and AMR gene profiles and were therefore not considered clonally related.However, a comparison of the plasmid sequence data with reference plasmids demonstrated the structural resemblance of virulence plasmids over several years within and, in part, also across farms.
The data suggest that the repeated occurrence of mcr-carrying plasmids in pathogenic E. coli isolates may be due to the local long-term occurrence of mcr-carrying plasmids rather than reinfection with different plasmids.In affected farms, specific eradication programs, such as rigorous hygiene management and prevention of pig trafficking, may be at least as equally important and effective in reducing the burden of colistin resistance as general recommendations to reduce colistin usage.

Figure 1 .
Figure 1.Schematic circular representation of E. coli virulence plasmids detected in all three farms over the years (details given in Table 2).For comparison, one representative isolate per year and farm was selected.The reference plasmids used were p15ODTXV (A); NCBI Reference Sequence: MG904998.1)and pUMNK88_K88 (B); NCBI Reference Sequence: CP002730.1),shown in the inner rings.The plasmids detected in isolates from the three farms are arranged according to farm and year of isolation.Isolates from the same farm are color-coded (A): Farm 1 = red, Farm 2 = green, Farm 3 = blue; (B): Farm 1 = orange, Farm 2 = turquoise/blue, Farm 3 = violet).The color of the earliest isolation is displayed as the lightest, while the color of the latest isolation is displayed as the darkest.

Figure 1 .
Figure 1.Schematic circular representation of E. coli virulence plasmids detected in all three farms over the years (details given in Table 2).For comparison, one representative isolate per year and farm was selected.The reference plasmids used were p15ODTXV (A); NCBI Reference Sequence: MG904998.1)and pUMNK88_K88 (B); NCBI Reference Sequence: CP002730.1),shown in the inner rings.The plasmids detected in isolates from the three farms are arranged according to farm and year of isolation.Isolates from the same farm are color-coded (A): Farm 1 = red, Farm 2 = green, Farm 3 = blue; (B): Farm 1 = orange, Farm 2 = turquoise/blue, Farm 3 = violet).The color of the earliest isolation is displayed as the lightest, while the color of the latest isolation is displayed as the darkest.

Figure 2 .
Figure 2. Neighbor-joining tree based on the comparison of 2670 core genome genes of 87 E. coli isolates from Farm 1, Farm 2 and Farm 3, with the respective multilocus sequence types (MLST) and isolation years color-coded.The occurrence of mcr genes and their location on different plasmids are shown.Farm specific clustering of isolates from different years can be observed for Farm 1, Farm 2 and Farm 3.

Figure 2 .
Figure 2. Neighbor-joining tree based on the comparison of 2670 core genome genes of 87 E. coli isolates from Farm 1, Farm 2 and Farm 3, with the respective multilocus sequence types (MLST) and isolation years color-coded.The occurrence of mcr genes and their location on different plasmids are shown.Farm specific clustering of isolates from different years can be observed for Farm 1, Farm 2 and Farm 3.

Figure 3 .
Figure 3. Overview of the phylogenetic grouping of mcr-positive and mcr-negative E. coli isolates sorted by Farms 1 to 3 and in total.

Figure 3 .
Figure 3. Overview of the phylogenetic grouping of mcr-positive and mcr-negative E. coli isolates sorted by Farms 1 to 3 and in total.

Figure 4 . 4 .
Figure 4. Schematic circular representation of detected MCR-1 plasmids from three farms in our study compared to most similar reference plasmids predicted using the BacWGST database.One Figure 4. Schematic circular representation of detected MCR-1 plasmids from three farms in our study compared to most similar reference plasmids predicted using the BacWGST database.One representative mcr-positive E. coli isolate per year and farm was selected for comparison (details given in Table 2).If two different mcr-positive pathotypes were detected per year and farm, we selected both pathotypes.The plasmids from our study are arranged in order of isolation year, with the earliest (light color) on the third innermost ring and the latest (dark color) on the outermost ring in (A-D).The reference plasmids are shown in the two innermost rings.Reference plasmids (NCBI Reference Sequence in brackets) included are as follows: pDJB-3 (MK574666.1),pGD27-70 (MN232195.1)for Farm 1 (A); pMCR-1-IHIT35346 (KX894453.1),pGMI17-004_2 (NZ_CP028167.1)for Farm 2 (B); pMCR-1-IHIT35346 (KX894453.1),pCFSAN061769_01 (CP042970.1)for Farm 3 (C); pC2 (LC473131.1)and p778 (MN746292.1)for Farm 3 (D).

Table 1 .
Sample collection from three farms and predicted pathotypes from all E. coli isolates and mcr-positive E. coli isolates per farm. E.