Genomic Diversity and Virulence Potential of ESBL- and AmpC-β-Lactamase-Producing Escherichia coli Strains From Healthy Food Animals Across Europe

The role of livestock animals as a putative source of ESBL/pAmpC E. coli for humans is a central issue of research. In a large-scale pan-European surveillance, 2,993 commensal Escherichia spp. isolates were recovered from randomly collected fecal samples of healthy cattle, pigs and chickens in various abattoirs. One-hundred Escherichia spp. isolates (0.5% from cattle, 1.3% pigs, 8.0% chickens) fulfilled the criteria for cefotaxime and ceftazidime non-wildtype (EUCAST). In silico screening of WGS data of 99 isolates (98 E. coli and 1 E. fergusonii) revealed blaSHV–12 (32.3%), blaCTX–M–1 (24.2%), and blaCMY–2 (22.2%) as predominant ESBL/pAmpC types. Other types were blaSHV–2 (1.0%), blaCTX–M–2/–14/–15 (1.0/6.1/1.0%), and blaTEM–52 (5.1%). Six isolates revealed AmpC-promoter mutations (position −42 (C > T) and one carried mcr-1. The majority (91.3%) of ESBL/pAmpC genes were located on plasmids. SHV-12 was mainly (50%) encoded on IncI1α plasmids (pST-3/-26/-95), followed by IncX3 (12.5%) and IncK2 (3.1%). The blaTEM–52 genes were located on IncI1α-pST-36 (60%) and IncX1 plasmids (20%). The dominant plasmid lineage among CTX-M-1 isolates was IncI1α (pST-3/-295/-317) (87.5%), followed by IncN-pST-1 (8.3%). CMY-2 was mostly identified on IncI1α (pST-12/-2) (54.5%) and IncK2 (31.8%) plasmids. Several plasmids revealed high similarity to published plasmids from human and animal Enterobacteriaceae. The isolates were assigned to phylogroups A/C (34.7/7.1%), B1 (27.6%), B2 (3.1%), D/F (9.2/10.2%), E (5.1%), and to E. clades (3.0%). With 51 known and 2 novel MLST types, a wide variety of STs was found, including STs previously observed in human isolates (ST10/38/117/131/648). ESBL/AmpC types or STs were rarely correlated with the geographic origin of the isolates or animal species. Virulence gene typing identified extraintestinal pathogenic E. coli (ExPEC; 2.0%), avian pathogenic E. coli (APEC; 51.5%), and atypical enteropathogenic E. coli (EPEC; 6.1%). In conclusion, the high diversity of STs and phylogenetic groups provides hardly any hint for clonal spread of single lineages but hints toward the dissemination of cephalosporin resistance genes in livestock via distinct, globally successful plasmid lineages. Even though a number of isolates could not be assigned to a distinct pathotype, our finding of combined multidrug-resistance and virulence in this facultative pathogen should be considered an additional threat to public health.

In conclusion, the high diversity of STs and phylogenetic groups provides hardly any hint for clonal spread of single lineages but hints toward the dissemination of cephalosporin

INTRODUCTION
Since the turn of the century, the prevalence of infections due to extended-spectrum cephalosporin-resistant (ESC-R) Enterobacteriaceae increased globally both in hospitals and in the community, entailing a major public health concern. At the same time, the prevalence of ESC-R Escherichia coli has been increasingly reported in livestock, in the food chain, and in companion animals (Ewers et al., 2012;Kaesbohrer et al., 2019). Consequently, there has been a controversial discussion about the role of animals as a putative source of ESC-R E. coli for humans either by direct contact or consumption of contaminated food for many years (Carattoli, 2008;Ewers et al., 2012).
Resistance to third and fourth generation cephalosporins, which are rated highest priority critically important antimicrobials as defined by the World Health Organisation (WHO, 2019), is frequently conferred by the hydrolytic activity of extended-spectrum β-lactamases (ESBL) and plasmidmediated AmpC-β-lactamases (pAmpC). The most clinically significant ESBL variants belong to the CTX-M, TEM and SHV families, pAmpCs are mostly represented by the CMY family (Bush and Fisher, 2011;Ewers et al., 2012;Carattoli, 2013). The frequent localization of bla ESBL and bla AmpC genes on plasmids and their common association with mobile genetic elements, like transposons and insertions sequences, contributes to their successful spread. Horizontal transfer facilitates easy transmission of ESBL/pAmpCs between bacteria of the same or closely related species, including commensals and pathogens, and including the intestinal microbiota of animals and humans (Ewers et al., 2012;Carattoli, 2013;Liebana et al., 2013;Madec et al., 2017;Rozwandowicz et al., 2018). ESBL-/pAmpC-encoding plasmids determined in the European Union (EU) frequently belong to the incompatibility groups (Inc) F, A/C, N, HI2, I1, and K, which differ in their host range and self-transmissibility (EFSA Panel on Biological Hazards (BIOHAZ), 2011; Liakopoulos et al., 2016). To understand the spread of ESC-R E. coli it is not only essential to consider the genetic context of ESBL/pAmpC plasmids but also the phylogenetic epidemiology and pathogenic potential of the bacterium.
E. coli comprises non-pathogenic commensals and strains that cause a variety of diseases in humans and animals. E. coli strains capable of causing extraintestinal infections, such as urinary tract infection, blood stream infection and soft tissue damage, are termed extraintestinal pathogenic E. coli (ExPEC), whereas strains that lead to intestinal diseases are designated as intestinal pathogenic E. coli (InPEC). While the classification of InPEC into different pathovars follows defined genetic criteria, i.e., concrete virulence-associated genes (VAGs) and/or phenotypic features, a wide range of VAGs have been associated with ExPEC and its various pathovars. It is assumed that ExPEC reside in the normal gut microbiota of healthy mammals and birds (Ewers et al., 2009;Koehler and Dobrindt, 2011;Starcic Erjavec and Zgur-Bertok, 2015). The combination of antimicrobial resistance (AMR) and virulence in E. coli strains constitutes a significant public health risk as exemplified by the successful global dissemination of ESBL-ExPEC lineages ST131 und ST648 (Ewers et al., 2014a;Schaufler et al., 2019), or by the CTX-M-15 producing enteroaggregative and Shigatoxin producing E. coli O104:H4 strains that caused the large German EHEC outbreak in 2011 (Karch et al., 2012).
In this study, we determined the prevalence of ESC-nonsusceptible (ESC-non-S) and ESC-R isolates in a European collection of presumptive ESBL/pAmpC-producing E. coli and characterized ESBL/pAmpC plasmids and AMR genes as well as the clonal diversity and serotype of the strains based on whole genome sequence (WGS) analysis. Virulence gene profiles were determined to facilitate classification into ExPEC and InPEC pathovars in order to assess the pathogenic potential of antimicrobial resistant strains.

Bacterial Isolates
The European Antimicrobial Susceptibility Surveillance in Animals (EASSA) monitors the antimicrobial susceptibility of zoonotic and commensal bacteria in healthy food-producing animals at slaughter across Europe (de Jong et al., 2013). The EASSA project includes the major countries of production of beef cattle, slaughter pigs and broiler chickens in the EU. Five or six countries were selected per animal species with ≥4 slaughterhouses in each country providing samples (usually about 200 per animal species). Colon and cecal samples were randomly collected from each of the major foodproducing animal species. At most one isolate of each bacterial species was retained from each animal, which was randomly selected as being representative of a whole herd or flock. Our study comprised 2,993 non-repetitive E. coli isolates from broiler chicken (n = 1016), cattle (n = 841), pig (n = 1136) isolated in Belgium (n = 167), Denmark (n = 208), France (n = 588), Germany (n = 426), Hungary (n = 328), Poland (n = 182), Spain (n = 410), the Netherlands (n = 387), and the United Kingdom (n = 297) isolated from February 2013 to March 2015March (20132014, 1,588 isolates;2015, 224 isolates).

Antimicrobial Susceptibility Testing and Verification of Bacterial Species
Of the total Escherichia spp. collection 100 isolates fulfilled the criteria for cefotaxime and ceftazidime non-wild type according to EUCAST criteria, i.e., reduced susceptible or resistant to one or both cephalosporins [isolates with MICs of cefotaxime and/or ceftazidime ≥ 1 mg/L; EUCAST (The European Committee on Antimicrobial Susceptibility Testing), 2015; Table 1]. MALDI-TOF MS analysis (Bruker Daltonics, Bremen, Germany) confirmed 99 of these isolates as E. coli and one isolate as E. fergusonii.

Whole Genome Sequencing
Ninety-nine cefotaxime and ceftazidime non-wild type Escherichia spp. isolates (one isolate was excluded due to repeated DNA degradation) were whole genome sequenced. DNA was extracted from the isolates using the DNA Blood and Tissue Kit according to the manufacturer's instruction (Qiagen, Hilden, Germany), followed by library preparation, using Nextera XT library (Illumina, San Diego, United States). DNA was sequenced by using an Illumina NextSeq 550 with multiplexing of 70 samples per flow cell using 150 bp paired end reads and a minimum of 70-fold coverage. Raw reads were adapter-trimmed by Flexbar v.3.0.3 (Resource Identification Portal RRID: SCR_013001), corrected using BayesHammer and assembled de novo using SPAdes v3.12.1 (RRID: SCR_000131). Assembled draft genomes were annotated using Prodigal (Prodigal, RRID: SCR_011936).
To investigate the distribution of class 1, class 2, and class 3 integrons with prevalent arrays in the genomes of our Escherichia spp. isolates, we performed BLAST search against the available databases GenBank and INTEGRALL 1.2, using filter parameters of more than 90% nucleotide identity and at least 80% query coverage.
To display circular comparisons between plasmids we used the blast ring image generator software BRIG Version 0.95 (RRID: SCR_007802).
To test whether ESBL and AmpC-β-lactamase genes were transferable, conjugation was performed by the broth filter mating method at 37 • C on 40 isolates which represented different ESBL/pAmpC genes and plasmid replicon types using plasmid-free sodium azide resistant E. coli J53 (J53 Azi R ) as recipient. Transconjugants were selected on Mueller Hinton agar plates supplemented with 100 mg/L sodium azide and 4 mg/L cefotaxime (Sigma-Aldrich, Germany). To confirm successful plasmid transfer, antimicrobial susceptibility testing of transconjugants and PCR for ESBL/pAmpC genes were performed. In addition to WGS analysis, plasmid sizes and location of ESBL/pAmpC genes were determined for all 99 isolates by S1 nuclease-restriction followed by pulsed-field gel electrophoresis and subsequent southern blotting. To verify the genomic location of ESBL and pAmpC genes we further performed an in silico search of whole genome sequences with mlplasmids v. 1.0.0 (Arredondo- Alonso et al., 2018).

Phylogenetic Analysis
Phylogenetic groups were determined by using the ClermonTyping method and its associated web-interface ClermonTyper, that allows a given strain sequence to be assigned to E. albertii, E. fergusonii, Escherichia clades I-V, E. coli sensu stricto as well as to the seven main E. coli phylogroups A, B1, C, E, D, F, and B2 (Clermont et al., 2013;Beghain et al., 2018). MLST 2.0 3 (Larsen et al., 2012) was applied to identify the multilocus sequence type (RRID: SCR_010245) of E. coli isolates following the Achtman scheme 4 , which represents a 7-gene-scheme including genes adk, fumC, gyrB, icd, mdh, purA, and recA.

Detection of Serotype and Virulence-Associated Genes
The sero(geno)type of E. coli strains was determined using the web-based serotyping tool SerotypeFinder 2.0 5 (Joensen et al., 2015). Screening for VAGs was carried out by NCBI BLASTn (RRID: SCR_004870) analysis against homologous genes present in an in-house database of 800 VAGs, gene variants or genomic islands from a subset of the VirulenceFinder database and inhouse created and manually curated VAG reference sequences. Coverage length and sequence identity thresholds were 80 and 90%. We searched for genes that were previously linked with intestinal and extraintestinal pathogenic E. coli pathovars and that belonged to different categories (adhesin, toxin, iron uptake system, capsule synthesis, auto-transporter, and secretions system). A VAG score was definded as the total number of VAGs within an isolate. In case a VAG was detected multiple times within a single isolate, it was only counted once. The kpsM, afa/dra, and sfa/foc operons were considered present if any of the corresponding genes or allelic variants were identified. All E. coli isolates were further analyzed for the presence of fimH gene and allele type by aligning to a FimH database using FimTyper 1.0 6 .

Determination of Associations Between
E. coli Host, Country of Origin, Phylogroups, VAGs, and AMR Genes The clustering of binary matrix indicating the presence and absence of antibiotic and virulence classes was performed by using package Rtsne version 0.15 (RRID: SCR_016900) with 5000 iteration and perplexity 15 in R version 3.6.1 (R Project for Statistical Computing, RRID: SCR_001905). Microreact version 5.93.0 was applied to visualize the clustering pattern in association with the metadata of the strains (Argimon et al., 2016). Kruskal-Wallis test as well as χ 2 and Fisher's exact test (IBM SPSS Statistics 27) were used to determine whether proportions for one variable (AMR gene, VAGs) were different between phylogroups and isolate origin (host and country). P-values < 0.001 were considered as statistical significant.

Core Genome Analysis
Phylogenetic and population genetic relationships were determined by applying a gene-by-gene approach on the dataset to generate a core genome alignment and subsequently a phylogenetic tree. The core genome alignment was assembled by a gene-wise alignment with Mafft v7.407 (RRID: SCR_011811), of 1,366 core genes that were present in at least 99% of the strains (sequence similarity min. 70%, sequence coverage min. 90%) and were concatenated afterward. The resulting alignment was used to infer a phylogeny with 100 bootstrap replicates using RAxML v.8.2.10 (RAxML, RRID: SCR_006086) with a General Time Reversible model and gamma correction for among site rate variation. iTOL v5 (RRID: SCR_018174) was used to visualize the population structure in the context of available metadata.
Almost all isolates (97%) revealed MDR. The most frequent MDR phenotype was combined resistance to penicillins, cephalosporins, tetracycline, sulfisoxazole, and either phenicols or folate pathway inhibitors (n = 8 each). Different proportions of isolates showed resistance to two (3%), three (18%), four (27%), five (23%), six (19%), seven (7%), and eight (3%) antibiotic classes. On average, isolates from cattle, chickens and pigs revealed a mean number of resistance to 4.3, 3.5, and 3.0 antibiotic classes, respectively. Thirty-four phenotypic AMR patterns (based on resistance of isolates to 12 antibiotic classes) were distinguished (Supplementary Table S2). Eleven of these patterns were unique; the remaining patterns were dispersed among a number of 2-8 isolates. AMR patterns appeared to be unrelated to host and country origin of the E. coli isolates.

Antimicrobial Resistance Genes and Integrons
The genomes of 99 out of 100 ESC-non-S/ESC-R isolates were sequenced successfully. One isolate from a chicken in Spain turned out to belong to the species E. fergusonii. Ninety out of these 99 isolates (90.9%) possessed one or more ESBL/pAmpC genes ( Table 1). Another six isolates that lacked ESBL/pAmpC genes revealed mutations in the promoter of the chromosomally encoded ampC gene when compared with the wildtype E. coli K12 sequence (ATCC 25922; NZ_CP009072.1). Mutations were located at positions −42 (alternate −35 box, C > T), −18 (alternate −10 box, G > A), + 81 (AmpC coding region, A > G), and outside functional promtoter elements and AmpC coding region (−1, C > T; + 58, C > T). In three ESBLand AmpC negative isolates, namely IHIT31981, IHIT31985 (both from chicken in the Netherlands) and IHIT32069 (pig, Netherlands), the genetic background of the ESC-R phenotype remained unclear.
Furthermore, we identified the plasmid-mediated colistin resistance gene mcr-1 in one colistin resistant strain isolated from a broiler chicken sample in the Netherlands in 2013. Other mcr variants (mcr-2-mcr-10) or alleles were not present.
The majority of strains (83/91; 91.2%) carried their ESBL/pAmpC/MCR-1 genes on plasmids. For 30 isolates, there was a link between ESBL/pAmpC genes and replicon sequences on the same de novo assembled contig. In the remaining isolates, WGS-based reconstruction, i.e., a mapping of contigs to previously published plasmid sequences together with plasmid prediction of contigs by using mlplasmids was performed to deduce the respective plasmid contigs. In some cases, we could not reconstruct the entire plasmid backbones. Consequently, Inc types of ESBL/pAmpC encoding plasmids could either not (n = 3 isolates Inc not typable) or only partially (n = 7 isolates non-IncI1-type) be assigned Supplementary Document S2. Transconjugation assays with 40 selected strains confirmed the transferability of ESBL/pAmpC/MCR-1 plasmids and phenotypic resistance (beta-lactam or colistin) for all strains and transconjugants, respectively (data not shown).
Twenty of these plasmids showed the genomic backbone of pC60-108, a 108.6 kb plasmid determined in an E. coli strain from chicken in Switzerland in 2013  Supplementary Document S3 and Supplementary Figure S3).
Another two isolates from pigs in France and Germany revealed bla CTX−M−1 on IncN/pST-1 plasmids that were 42.4 kb in size and showed a similarity of 98.7% to IncN/pST-1 plasmid pL2-43 (Supplementary Document S3 and Supplementary Figure S4). Like pL2-43 (E. coli, lamb, Switzerland, 2013), our plasmids harbored macrolide resistance gene mph(A) in addition to bla CTX−M−1 .
In four avian E. coli isolates from Spain, we located the bla CTX−M−14 gene on a plasmid with unknown replicon type. The CTX-M-14 plasmids were about 110 kb in size and showed partial similarity with 55.1 kb CMY-2 plasmid 2016C-3936C1-unnamed2 (GenBank: CP018772) (Supplementary Document S3 and Supplementary Figure S5).
The only ESBL-producing strain identified among E. coli isolates from Poland (cattle, ST617, phylogroup C) was at the same time the only strain that harbored CTX-M-15. The bla CTX−M−15 gene was located on the chromosome and the genetic environment shows high similarity to that of human intra-abdominal infection ST4981 E. coli strain CH611_eco (CP017980.1) from China (data not shown). Although we could verify plasmid location of the bla CTX−M−2 gene of a bovine isolate from Belgium, a comparison with published plasmid sequences revealed no significant results.

MCR-1
The mcr-1 gene in avian E. coli isolate IHIT31981 was encoded on a 33.3 kb IncX4 plasmid that was almost identical (99.9% nucleotide sequence identity) with other globally distributed MCR-1 IncX4 plasmids, such as pMCR-1-CT (GenBank: CP018773.2), determined in an E. coli isolate from a human patient in the United States.

Phylogenetic Grouping
Using the ClermonTyping method and its associated webinterface ClermonTyper, 98 isolates were verified as E. coli and one, as expected, as E. fergusonii. The majority of the 98 E. coli isolates were assigned to phylogenetic groups A (34.7%) and B1 (27.6%). Group C, a newly defined phylogroup within the formerly highly diverse group A (Clermont et al., 2013), was identified in 7.1% of the isolates. Phylogenetic groups B2 (3.1%), D (9.2%), and F (10.2%), the latter one representing a novel group D-related cluster, were less prevalent. Three isolates (3.0%) from chickens in Spain belonged to clade I. Five isolates (5.1%) were assigned to phylogenetic group E (Figure 2A). All phylogroups B2 and D and 90% of group F isolates, often referred to as extraintestinal pathogenic E. coli (ExPEC) groups, were from chickens. Those phylogroups regarded as commensals (A, B1, C) were detected in 62.5% of chicken, 100% of cattle, and in 93.3% of pig isolates. The distribution of ESBL/pAmpC types revealed no significant association to the phylogenetic group of a strain ( Figure 2B). Among 12 antibiotic classes tested, phylogroup B1 strains showed the highest mean number of resistance (resistant to 5.19 ± 1.36 classes), followed by isolates belonging to group E (4.8 ± 0.87), A (4.79 ± 1.28), clade I (4.67 ± 1.33), F (4.4 ± 1.05), and D (4.11 ± 1.04). Group C (3.71 ± 0.89) and B2 (3.67 ± 0.33) isolates revealed the lowest antimicrobial resistance with regard to different antibiotic classes.

ExPEC-Related Genes
More than half of the 99 isolates fulfilled the criteria to be classified as avian ExPEC, also termed avian pathogenic E. coli (APEC). In detail, 51.5% of our isolates harbored at least three of the five VAGs iroN, iss, iutA, ompT, and hlyF that have been suggested to define APEC ( Table 3; Johnson et al., 2008). A few years earlier, Johnson and Russo (2005) defined the group of ExPEC on the basis of the presence of ≥ 2 VAGs or gene combinations, including genes for P fimbrial genes (papAH and/or papC), S and FIC fimbriae (sfa/focDE), Dr. antigenbinding adhesin (afa/draBC), group II capsule polysaccharides (kpsMTII), and ferric aerobactin receptor (iutA) (Johnson and Russo, 2005). This definition includes all pathotypes that were primarily "ill-defined, " including uropathogenic E. coli (UPEC), neonatal meningitis-associated E. coli (NMEC), sepsis-causing E. coli (SEPEC), and APEC. Following this definition, only two of our isolates qualified as ExPEC, namely a CMY-2-positive avian isolate from Spain (phylogroup D, ST38) and a CTX-M-1-producing porcine isolate from Germany (phylogroup B1, ST453).

Clonal Diversity
With 51 known and two novel types, we found a wide variety of multilocus sequence types (STs) among our strains. For isolates from broilers (n = 80), pigs (n = 15) and cattle (n = 4) we determined 42, 14, and 3 different STs, respectively ( Table 4). Thirty-two of these STs have been assigned to 18 clonal complexes (CCs), while 21 STs did not cluster within one of the CCs defined in the Enterobase E. coli database (Supplementary Document S2) 7 . The most often identified STs were ST117 and ST10 (n = 6 each), ST665 (n = 5) ST23, ST38, ST155 (n = 4 each), ST101, ST354, ST4980, ST752, ST770, and ST88 (n = 3 each). ST648 and ST131, two globally distributed lineages that are frequently associated with ESBL/AmpC production were identified in a CTX-M-1-producing chicken isolate from Spain (ST648) and in two CMY-2 positive chicken ST131 isolates from different farms in Hungary. A correlation between STs and animal species or country of origin was not evident. In addition, ESBL and AmpC β-lactamase types, as well as plasmid replicon types showed a random distribution among the different STs.
We further determined the relatedness of strains from different animal species and countries by WGS comparison. Although the application of comparative core genome analysis of 98 E. coli isolates, based on 1,330 orthologous genes, achieved a much higher resolution regarding the genetic relatedness of the ESBL/pAmpC/MCR-1 strains (SNP-based comparison of 1,330 genes vs. allele-based comparison of seven MLST genes), it basically reflected the high genetic diversity observed by MLST analysis (Figure 3). As expected, isolates forming one ST and/or CC by based on MLST data as well as isolates that shared the same phylogenetic groups almost always clustered together in the cgMLST scheme. In contrast, neither ESBL/AmpC-types nor plasmid Inc types followed the phylogenetic distribution of the isolates but were broadly distributed among the different cgMLST clusters.
The same analysis was done using VAGs grouped into different categories as the clustering pattern

Chicken
Chicken ( *In case an ESBL/AmpC type appears more than once, the number is indicated in brackets. **Escherichia fergusonii (IHIT32041). Figure S2). Significant associations were predominantly observed between VAGs and the phylogenetic group. In total 36 adhesion-related, 11 autotransporter and toxin genes, 22 invasion and protectin genes, 14 iron acquisition genes and 33 secretion systems genes revealed several P-values < 0.001. In contrast, only 14, 21, and 7 VAGs showed significant associations with the variables host, country, and ESBL/AmpC type.

DISCUSSION
This study focused on the identification and investigation of ESBL and pAmpC genes in ESC-non-susceptible/ESC-resistant commensal E. coli isolates. Special attention was paid to the antimicrobial resistance phenotype of the isolates and their plasmids with regard to replicon type and genetic plasmid structure. We further investigated the molecular epidemiology of E. coli isolates with respect to their genome-based phylogeny and assignment to intestinal and extraintestinal E. coli pathotypes. The presence of potentially virulent and antimicrobial resistant ESBL/pAmpC-producing E. coli originating from healthy production animals could be a significant issue for human and animal health. One hundred ESC-non-susceptible/ESC-resistant isolates were identified among 2,993 commensal E. coli isolates collected over a time period of 25 months (February 2013-March 2015) from different livestock species and European countries. The highest rate of ESC-non-susceptible/ESC-resistant isolates was detected in poultry (8.0%), whereas pigs (1.3%) and cattle (0.5%) revealed comparable low numbers of such isolates. These findings are fairly in agreement with the numbers reported by the European Food Safety Authority (EFSA) and the European Centre for Disease Prevention and Control (ECDC). Here the number of presumptive ESBL and AmpC phenotypes identified in indicator E. coli isolates rated to 5.4% for broiler flocks in 2014 [EFSA (European Food Safety Authority) and ECDC (European Centre for Disease Prevention and Control), 2016], and to 1.5 and 2.0% for fattening pigs and calves under 1 year of age, respectively, in 2015 [EFSA (European Food Safety Authority) and ECDC (European Centre for Disease Prevention and Control), 2017]. The lower number observed for cattle in our study might be due to the mixed origin of isolates according to animal age. Different studies could show that veal calves harbor higher proportions of ESBL/AmpCproducing bacteria than beef cattle [Haenni et al., 2014; EFSA (European Food Safety Authority) and ECDC (European Centre for Disease Prevention and Control), 2017]. Recent data from 2017 to 2018 indicated that the occurrence of presumptive ESBL-and/or AmpC-producing indicator E. coli isolates was still generally low in reporting European countries, ranging from 0.6 to 6.3% in isolates from pigs, from 1.2 to 5.3% in isolates from calves and from 0.6 to 7.9% in isolates from broilers [EFSA (European Food Safety Authority) and ECDC (European Centre for Disease Prevention and Control), 2020]. Carbapenemase producing (CP) E. coli were not detected in our study, which may be due to the absence of such isolates and/or the FIGURE 3 | Maximum-likelihood tree of 98 E. coli genomes (E. fergusonii isolate IHIT32041 was not included). The phylogeny is based on 1336 orthologous genes and was calculated by RAxML. Graphical visualization was achieved by iTOL v5. Colored rings represent (from inner to outer ring the (i) phylogroup, (ii) ESC-non-S/ESC-R resistance determinant, (iii) ESBL/AmpC/MCR-1 plasmid replicon type, (iv) country of isolation, and (v) host.
cultivation of the samples on non-selective media. The presence of CP bacteria on livestock farms was first reported in 2012 in Germany (Fischer et al., 2013). Since then, increasingly emerged in different livestock species and countries (Anderson and Boerlin, 2020). Despite of these findings, the overall prevalence of CPs in livestock is still low as shown by the results from the specific carbapenemase producers monitoring in European countries between 2016 and 2018 [EFSA (European Food Safety Authority) and ECDC (European Centre for Disease Prevention and Control), 2020].
Almost all isolates (97.0%) from our study revealed an MDR phenotype, i.e., resistance to antimicrobial agents of at least three classes, and 29.0% of the isolates showed resistance to ≥ six classes of antimicrobial agents. This is consistent with previous studies from different countries. Ninety percent of cefotaxime and/or ciprofloxacin resistant, commensal E. coli isolated from food-producing animals in Belgium were resistant to several other antibiotics (Lambrecht et al., 2018). MDR was also observed in 88.1% of ESBL-producing isolates obtained between 2008 and 2014 from diseased food-producing animals in Germany . In this study, 39.4% of the isolates, predominantly from cattle, were resistant to six antibiotic classes. Several authors assumed a correlation between antimicrobial usage in food-producing animals and the prevalence of AMR bacteria in those animals. According to publicly available national or international reports from seven European countries, the level of veterinary use of specific antimicrobials strongly correlates to the level of resistance toward these agents in commensal E. coli isolates in pigs, poultry and cattle (Chantziaras et al., 2014). A significant correlation between antimicrobial use and resistance in commensal E. coli was also demonstrated for livestock animals in Belgium (Callens et al., 2018). In our study, isolates were frequently (32-100%) resistant to ampicillin, sulfisoxazole, tetracycline, trimethoprim/sulfamethoxazole, and ciprofloxacin. Antimicrobial agents of such classes are among the most sold agents to veterinarians in different European countries (Lekagul et al., 2018;Roth et al., 2019), indicating that AMR and frequent antibiotic usage could coincide to an extent that remains to be determined.
ESBL genes bla CTX−M−14 (6.1%, chicken and pig) and bla TEM−52 (5.1%, chicken and pig) were present in lower numbers among our isolates, while bla SHV−2 (chicken), bla CTX−M−2 (cattle) and bla CTX−M−15 (cattle) occurred only once. Together with CTX-M-1 and CTX-M-15, CTX-M-14 represents one of the most common ESBL type in E. coli isolates from cattle and its largely moderate occurrence in E. coli isolates derived from chickens and pigs has also been demonstrated (Ewers et al., 2012;Palmeira et al., 2020).
AmpC activity due to chromosomal ampC promoter mutations was found in 6.1% (6/99) of our isolates. All seven isolates revealed a mutation pattern previously described as ampC type 03 (Mulvey et al., 2005). This type was reported as one of the most prevalent variants among cefoxitin-resistant clinical E. coli isolates in different studies (Mulvey et al., 2005;Peter-Getzlaff et al., 2011). The ampC promoter alterations created an alternate displaced promoter whose mutation at position −42 (C > T) is thought to have strong effect on promoter strength and consequently on hyperproduction of ampC (Caroff et al., 2000).
CTX-M-1-producing isolates belonged largely (75.0%) to IncI1α/pST-3 plasmids, which represent a globally distributed, genetically highly similar plasmid lineage that is disseminated in livestock animals and humans in Europe (Carattoli, 2008;Zurfluh et al., 2014;Irrgang et al., 2018;Touzain et al., 2018). 8 IncI1 plasmids are highly conjugative, and transmission from commensal to pathogenic bacteria poses a risk for animal and human health (Carattoli, 2013). One of our CTX-M-1 plasmids was assigned as pST-295, a plasmid type that was first identified in a commensal CTX-M-1 ST1604 E. coli isolate from chicken in Denmark in 2015 (Liu et al., 2019). In that study, four novel CTX-M-1 IncIα plasmid lineages (pST-293-pST-296) were determined among livestock isolates. Together with our finding of novel type pST-317, this could reflect a possible diversifying evolutionary process in these otherwise highly conserved plasmids. Noteworthy, the plasmid transfer of several plasmid types, including IncI1-pST3-CTX-M-1 and IncI1-pST295-CTX-M-1 could be significantly increased in vitro in a strain-dependent manner following exposure to the antibiotics cefotaxime, ampicillin, and ciprofloxacin (Liu et al., 2019).
The diverse population of E. coli isolates carrying CTX-M-1 IncI1 plasmids (21 IncI1 plasmids among 18 different MLST types) suggests that the dissemination of this ESBL gene is not due to the spread of single clonal lineages but is more likely the result of vertical transmission of plasmids by horizontal transfer. Also Irrgang and co-workers (2018) reported 51 different STs among 89 CTX-M-1 producing E. coli isolates from German food samples.
Our SHV-12 IncX3 plasmids were almost identical in plasmid backbone and bla SHV−12 flanking region to pEC-244, which was the first completely sequenced IncX3 plasmid of animal origin. In this study, IncX3 plasmids from humans and animals exhibited remarkable synteny in their backbone and differed only in their bla SHV−12 -flanking region (Liakopoulos et al., 2018). IncX3 plasmids are conjugative and highly stable, while they exert no fitness cost on their bacterial host, thus highlighting the epidemic potential of these plasmids.
In eight E. coli isolates from chickens, bla CMY−2 (n = 7) and bla SHV−12 (n = 1) has been linked to plasmids of the recently defined incompatibility group IncK2 (Seiffert et al., 2017). IncK2 plasmids are mainly associated with the spread of bla CMY−2 and bla CTX−M−14 genes in Europe, particularly in Spain and the United Kingdom, and are frequently reported from E. coli from livestock sources (Dhanji et al., 2012;Rozwandowicz et al., 2018;Apostolakos et al., 2020). de Been et al. (2014) identified bla CMY−2 carrying IncK2 plasmids in human and poultry isolates belonging to evolutionary distinct backgrounds, suggesting that these plasmids efficiently spread through E. coli populations of different reservoirs. We also identified IncK2 plasmids in isolates belonging to different phylogenetic groups (B2, D, E, and F) and to six different MLST types, including ST38, ST117, and ST131. The most frequent replicon type identified among bla CMY−2 carrying plasmids in our study was IncIα (n = 12 isolates from Netherlands, Hungary, and Spain). IncI1α and IncK2 were also found as major replicon types of bla CMY−2 carrying plasmids from E. coli isolates from humans, animals and food in Germany (Pietsch et al., 2018). Due to high sequence identity of plasmids, also shown for several plasmids in our study, the authors suggested that plasmidmediated, rather than clonal spread likely plays an important role for emergence and transmission of bla CMY−2 between aminals and humans.
The virulence potential of ESBL-and AmpC-producing E. coli originating from healthy livestock animals could be a significant issue for public health. Nevertheless, data on extensive virulence gene typing of AMR E. coli from livestock animals, including a rigorous screening for a set of > 800 VAGs related with intestinal and extraintestinal pathotypes, are scarce (Hussain et al., 2019;Apostolakos et al., 2020). We identified the intimin gene eae, which is a surrogate for atypical EPEC and several other InPECrelated genes, including the translocated intimin receptor gene tir in five ESBL-(SHV-12 and TEM-52) and one CMY-2-producing chicken E. coli isolates belonging to four different STs. Mueller et al. determined the eae gene in 3.6 and 2.9% of ESBL-producing E. coli isolates from livestock (n = 28) and healthy humans (n = 34), respectively. Consistent with our findings, they did not identify genes indicating the presence of EIEC (ipaH), ETEC (sta, stb, est), and STEC (stx1, stx2), suggesting a low prevalence of intestinal pathotypes among MDR bacteria from livestock (Mueller et al., 2016). Although bovines are the primary reservoir of STEC, previous reports confirm a low prevalence of ESBLproducing STEC among cattle and other livestock animals (Ewers et al., 2014b;Pietsch et al., 2018). Apostolakos et al. (2020) could assign the majority of 100 ESBL/pAmpC-E. coli isolated from the broiler production pyramid in Italy to defined ExPEC or InPEC pathotypes by virulence gene analysis. They identified 56% of their isolates as atypical EAEC, based on the presence of aatA and the absence of aggR. Both typical and atypical EAEC, that have been predominantly associated with pediatric diarrhea in developing countries, were not present in our isolates as they all lacked the pathotype-specific genes aatA, aggR, and aaiC (Bamidele et al., 2019).
Whether an E. coli isolate falls into the category of ExPEC, likely depends on the definition proposed by different authors. More than half of our 99 isolates fulfilled the criteria to be classified as avian ExPEC (APEC), including 19.6% isolates from nonavian sources. Apostolakos et al. (2020) identified 39 ESBL-E. coli broiler isolates (39%) belonging to 13 different STs as APEC. While they APEC group mainly (38.5%) consisted of ST744 and ST429/ST9298, our APEC strains were even more diverse (31 STs among 51 isolates) and only few STs were not unique, including ST117 (n = 6), ST23 (n = 4), ST38 (n = 3), ST88 (n = 3), and ST101 (n = 3). While we had no indication for a co-location of APEC-VAGs and ESBL-/AmpC genes on plasmids, others identified several virulence genes, including sitA-D, iucA-D, iutA, hlyF, ompT, iss, iroN, cvaA-C, and cvi on bla CMY−2 IncF plasmid obtained from a diseased French broiler, suggesting a putative threat for the easy dissemination of ExPEC virulence factors and resistance determinants (Touzain et al., 2018).
Only two of our isolates (CMY-2-ST38 and CTX-M-1-ST453) could be classified as ExPEC, whereas 67.7% of the isolates were considered as ExPEC-like (>10 ExPEC genes). Using the same criteria we applied, Apostolakos et al. (2020) identified 51% of their isolates belonging to 12 STs from different production chains of an integrated broiler company in Italy as ExPEC. In a study from Brazil, 58% of ESBL E. coli isolates from chicken carcasses harbored 3-5 of the ExPEC VAGs iutA, hlyF, iss, iroN, and ompT (Cyoia et al., 2018). The authors suggested that chicken meat is a potential reservoir of MDR E. coli strains harboring resistance and virulence genes that could pose a serious threat to human public health. This is further corroborated by a study dealing with ESBL-E. coli from poultry and human in India. Among 15 ExPEC-associated genes (papA/C/E/F/G, fimH, pic, sat, tsh, vat, iutA, ireA, iroN, fyuA, and usp) Hussain et al. (2019) identified comparable numbers in broiler (median 4) and human isolates (median 3) and overlaps in the AMR and VAG profile and phylogenetic background of a subset of strains.
Interestingly, several isolates harbored both InPECand ExPEC-related genes (Supplementary document S1), resembling what has previously been reported as hybrid pathotype, which is probably best exemplified by EHEC as a long-standing EPEC/STEC hetero-pathogen (Santos et al., 2020). Our aEPEC strains additionally harbored 9-12 InPEC related genes, such as ColV plasmid gene cvi, iron acquisition gene fyuA, iucA-D, irp2, and sitA-D, increased serum survival protein gene iss, and haem acquisition protein gene hma, which plays a critical role in the colonization of the urinary tract. The ExPEC and APEC strains carried up to four InPECrelated genes, including those encoding for AidA-I adhesin-like protein invasion-associated protein Tia, long polar fimbriae Stg, serine protease Pic, iron-related haem receptor Iha, and ETEC F4 and F5 fimbriae. Apostolakos et al. (2020) identified 30 ESBL/AmpC broiler isolates that displayed an aEAEC/ExPEC pathotype. They suggested that the virulence gene repertoire of ESBL/pAmpC E. coli may explain their adaptation to and persistence in different niches. As the capacity for genome interrogations is constantly rising, it seems obvious that various VAGs, previously either linked with InPEC or ExPEC pathotypes, will be found among different E. coli pathotypes to which they have not traditionally been associated (Santos et al., 2020). This, together with the availability of extensive AMR data based on an ever growing number of genome sequences may help to predict the emergence of novel multidrug resistant and virulent strains.

CONCLUSION
Our data suggest that cephalosporin resistance genes are mainly disseminated in livestock animals via distinct plasmids. Plasmid backbones within different plasmid lineages were often almost identical and were shared by phylogenetically unrelated isolates from chickens, cattle, and swine. In addition, they revealed significant similarity to plasmids from human isolates. We could demonstrate that E. coli of various phylogenetic groups and STs can combine antimicrobial resistance and virulence, even though a number of isolates could not be assigned to a distinct pathotype. In summary, this work significantly contributes to the understanding of the epidemiology and virulence potential of cephalosporin-resistant E. coli from livestock animals.

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 Supplementary Material.

ETHICS STATEMENT
Ethical review and approval was not required for the study in accordance with the local legislation and institutional requirements.

AUTHOR CONTRIBUTIONS
CE supervised the entire project. CE and AJ drafted the manuscript. CE and FE designed the study. EP-B, CE, and AJ have provided raw data and analyzed the data. UL conducted the laboratory experiments. ST, CE, and TS conducted the analyses of the sequencing data. All authors critically reviewed the manuscript.

FUNDING
This study was supported in part by the Executive Animal Health Study Center (CEESA). ST and TS were funded by the Federal Ministry of Education and Research (BMBF) under project number 01KI1703B as part of the ERA-NET COFUND antimicrobial resistances (JPI-EC-AMR) "HECTOR." CE was funded by the Federal Ministry of Education and Research (BMBF) under project number 01KI1908B as part of the JPI-AMR project "OASIS."