Whole Genome Sequencing Differentiates Presumptive Extended Spectrum Beta-Lactamase Producing Escherichia coli along Segments of the One Health Continuum

Antimicrobial resistance (AMR) has important implications for the continued use of antibiotics to control infectious diseases in both beef cattle and humans. AMR along the One Health continuum of the beef production system is largely unknown. Here, whole genomes of presumptive extended-spectrum β-lactamase E. coli (ESBL-EC) from cattle feces (n = 40), feedlot catch basins (n = 42), surrounding streams (n = 21), a beef processing plant (n = 4), municipal sewage (n = 30), and clinical patients (n = 25) are described. ESBL-EC were isolated from ceftriaxone selective plates and subcultured on ampicillin selective plates. Agreement of genotype-phenotype prediction of AMR ranged from 93.2% for ampicillin to 100% for neomycin, trimethoprim/sulfamethoxazole, and enrofloxacin resistance. Overall, β-lactam (100%; blaEC, blaTEM-1, blaSHV, blaOXA, blaCTX-M-), tetracycline (90.1%; tet(A), tet(B)) and folate synthesis (sul2) antimicrobial resistance genes (ARGs) were most prevalent. The ARGs tet(C), tet(M), tet(32), blaCTX-M-1, blaCTX-M-14, blaOXA-1, dfrA18, dfrA19, catB3, and catB4 were exclusive to human sources, while blaTEM-150, blaSHV-11–12, dfrA12, cmlA1, and cmlA5 were exclusive to beef cattle sources. Frequently encountered virulence factors across all sources included adhesion and type II and III secretion systems, while IncFIB(AP001918) and IncFII plasmids were also common. Specificity and prevalence of ARGs between cattle-sourced and human-sourced presumptive ESBL-EC likely reflect differences in antimicrobial use in cattle and humans. Comparative genomics revealed phylogenetically distinct clusters for isolates from human vs. cattle sources, implying that human infections caused by ESBL-EC in this region might not originate from beef production sources.


Whole Genome Sequence Analyses
Pan genome analyses was conducted using Roary [43], followed by AMR mechanism identification using the National Center for Biotechnology Information (NCBI) and ResFinder databases [44]. ARGs were designated as present if sequence identity was ≥90%. Agreement between ARGs and AMR phenotypes were examined as described by McDermont et al. [24], where resistance or susceptibility to a given antimicrobial was compared with the presence or absence of a known corresponding resistance gene(s) and/or structural gene mutation(s). WGS sensitivity was computed as the number of isolates that harbored resistant determinants divided by the total number of isolates exhibiting clinical resistance, with the phenotypic resistance as the standard outcome. Specificity was calculated as the number of isolates that did not harbor genetic determinants divided by the total number of phenotypic susceptible isolates [24,45]. Separate binary logistic regression [46] was used to examine a subset of cattle versus human isolates (CFeces and CBasins vs. CHumans and MSewage) for differences in ARG prevalence. The few isolates obtained from the BProcessing environment did not allow for them to be included in more extensive data analyses.
Based on single nucleotide polymorphisms (SNPs) in the core genes, phylotyping was performed using SNVPhyl [47] to generate a maximum likelihood tree with the NCBI reference strain E. coli str. K-12 substr. MG1655. The maximum likelihood tree was generated for the entire 162 presumptive ESBL E. coli isolates as well as a subset of 108 isolates which harbored bla TEM , bla CTX-M , bla OXA , or bla SHV and were designated as true ESBLs. The reference strain E. coli MG1655 was used as it has been widely used by others [48]. The tree was subsequently visualized using iTOL [49]. Whole genome multi-locus sequence typing (wgMLST) and a minimum span tree was also generated using BioNumerics (BN v7.6) (BioNumerics, Applied Maths, Keistraat, Belgium) with default parameters for categorical data. Multi-locus sequence typing (MLST) was achieved with PubMLST E. coli#1 (Achtman) scheme [50], while virulence factors were identified from the Virulence Factors Data Base (VFDB) [51]. To investigate the prevalence of ICEs in E. coli, BLAST homology searches were conducted against 1032 ICE sequences obtained from ICEberg database 2.0 [15]. In silico E. coli subtyping was performed based on outer membrane lipopolysaccharide (O) and flagellar (H) surface antigens using an in-house ECTyper and EcOH [52,53] to designate isolates as either O or H groups.
A set of presumptive ESBL-ECs randomly selected from each source of the beef production system that identified as non-ESBL-EC by WGS (total n = 23 not carrying any bla TEM , bla CTX-M , bla OXA , and bla SHV ) and those identified as true ESBL-EC (n = 24 carrying bla TEM , bla CTX-M , bla OXA , and bla SHV ) were compared using a GView BLAST atlas [54] to elucidate possible similarities or differences in genomes of the two groups using E. coli MG1655 as reference. The draft genome sequences of the 162 E. coli isolates have been deposited to GenBank under Bio project PRJNA556083.

Occurrence of Phenotypic Antimicrobial Resistance
Of the phenotypic resistances tested, oxytetracycline and ampicillin resistance was most common, while neomycin resistance was the least (Figure 1; Table S1). Although the CLSI minimum inhibitory concentration (MIC) ampicillin break-point was used for isolating presumptive ESBL-EC, not all E. coli were resistant to ampicillin based on subsequent disk diffusion susceptibility tests. Differences may have resulted from resistance gene silencing, loss of plasmids, or perhaps differences between MIC and disk diffusion assays. Among the isolates collected, 1.9% were not resistant to any of the antimicrobials, 5.6% were resistant to only one class, while 69.1% were resistant to five or more classes. Overall, a total of 92.6% of E. coli strains showed MDR in terms of resistance to three or more classes of antimicrobials. It is worthy to note that, despite enrichment, a limited number of presumptive ESBL-EC isolates were obtained from BProcessing environments (n = 4), an attestation to the sanitary procedures within the plant.  MDR referred to resistance to two or more classes of antibiotics.

Antimicrobial Resistance Genes Prevalence
Tetracyclines: A total of 77 unique ARGs encoding diverse antimicrobial resistance were detected. The tetracycline determinants tet(A) and tet(B) were detected in both cattle and human sources, whereas tet(C) and tet(M) were detected in cattle sources, but not MSewage nor CHumans. In contrast, tet (32) was exclusive to isolates associated with human sources (Figure 2; Table S2).

Antimicrobial Resistance Genes Prevalence
Tetracyclines: A total of 77 unique ARGs encoding diverse antimicrobial resistance were detected. The tetracycline determinants tet(A) and tet(B) were detected in both cattle and human sources, whereas tet(C) and tet(M) were detected in cattle sources, but not MSewage nor CHumans. In contrast, tet (32) was exclusive to isolates associated with human sources (Figure 2; Table S2).
Folate pathway inhibitors: For folate synthesis inhibitors, sul2 was ubiquitous but sul1 was not detected in BProcessing isolates, whereas sul3 was not found in either BProcessing or SStreams isolates. dfrA1 was present in MSewage, CBasins, and CFeces, while dfrA12 was unique to CBasins, CFeces, and SStreams. dfrA7 was unique to CBasins and dfrA18, and dfrA19 was unique to CHumans isolates. DfrA14, dfrA17, and dfrA27 were present in all CBasins, CFeces, SStreams, CHumans, and MSewage isolates. Heat map of resistance genes associated with various classes of antimicrobials as detected by the whole genome sequences of 162 presumptive ESBLproducing E. coli across a One Health continuum: Isolates originated from multiple segments of the One Health continuum including human clinical, municipal sewage, beef processing plant, cattle feces, catch basin, and surrounding streams. aac-(6′)-Ib-cr classified under aminoglycoside also encodes quinolone resistance. The overall prevalence of each determinant is displayed as a numeric percentage at the bottom of each determinant.     Sensitivity (%) equals number of isolates that harbored resistant determinants divided by the total number of isolates exhibiting clinical resistance phenotypes, while specificity (%) was calculated as the number of isolates that did not harbor genetic determinants divided by the total number of phenotypically susceptible isolates. Whole genome sequence of positive predictive value (PPV%) and negative predictive value (NPV%).

Phylogenetics, MLST Cluster, and Serogroups Analysis
Evaluation of the genetic relatedness of the beef/cattle-associated and human-associated isolates using a maximum likelihood phylogenetic tree showed that 71.64% of the core genome of isolates from these sources was shared. Limited phylogenetic intermixing among isolates from segments along the One Health continuum was observed (Figure 3). Single nucleotide variant phylogenomics revealed that, in general, human-sourced ESBL-EC comprising CHumans and MSewage formed one major cluster within which a number of distinct clades and sub-clades occurred. The majority of CHumans and MSewage isolates were closely related, although about 7.5% of all beef/cattle-associated isolates (n = 107) were in this cluster. Interestingly, 92.5% of the isolates from CFeces, CBasins, and SStreams formed a separate cluster of mixed clades with only a few MSewage (n = 9) and CHumans (n = 2) isolates in this clade. Isolates from CBasins and CFeces often clustered together. Although the few BProcessing isolates formed a distinct sub-clade from CBasins and SStreams, they also clustered separately from isolates obtained from other points within the beef production chain. Apart from an isolate from CFeces, that was within 18 SNP of MSewage isolates (=2) (Figure S3), the lowest SNP pairwise differences between CHumans and MSewage vs. all cattle-sourced isolates were 552 and 718, respectively. Minimum span tree (MST) generated with Bionumerics and whole genome MLST (wgMLST) corroborated our phylogenetic analysis and showed distinct clusters of cattle-versus human-sourced isolates (Figure 4; Figure S4). Subsequently, a separate phylogenetic tree which featured only true ESBL isolates also showed limited intermixing between cattle and human isolates ( Figure S5). BLAST atlas of isolates identified as non-ESBL versus true ESBL-EC by WGS did not reveal much diversity among non-ESBL and ESBL-EC sequences ( Figure 5), although the genetic elements that appeared to be commonly absent in both E. coli groups were associated with prophage elements when compared to the reference genome (data not shown).
In silico typing using the ECTyper and EcOH databases yielded 67/162 and 37/162 unknown and novel O-antigen loci, respectively. About half of the clinical isolates (52%; n = 13) were O25:ST131, while three MSewage isolates were of the wzx-Onovel31:ST131 subtype. The majority of cattle-sourced isolates were found to be O8, O89, O9, O15, and O99. Two BProcessing O group isolates remained unclassified, while the other two isolates both classified as O128:H2. All presumptive ESBL-EC identified by MLST as ST224 (n = 21) showed an unknown O-antigen with a H23 flagella antigen. Figure 3. Phylogenetic tree generated on the basis of single-nucleotide polymorphisms (SNPs) of the core genes of 162 E. coli isolates obtained from multiple segments of the beef production system and human-associated isolates with reference genome E. coli str. K-12 substr. MG1655 (GenBank accession # GI: 545778205/U00096.3). Inner, middle ring, and outer rings are representative of isolate sources, phenotypic antimicrobial, and multidrug resistance, respectively.

Antimicrobial Resistance Determinants
The overall ubiquity of β-lactam and tetracycline genes among cattle-and human-sourced isolates across the One Health continuum was expected as both physicians and veterinarians use these classes of antimicrobials to treat humans and beef cattle, respectively [17,57,58]. Tetracycline resistance genes, especially the efflux pumps encoded tet(A), have been detected in high frequencies in generic E. coli and ESBL-EC from bovines and humans from United States and China [58][59][60] and from Canadian feedlot environments using metagenomics [61]. However, the ribosomal protection mechanisms conferred by tet(M) exclusive to cattle isolates has rarely been reported in fecal samples from beef cattle. Metagenomics studies of manure, soil, and wastewater from dairy and beef production systems [61] within the same region in Canada did not identify tet(M). Since the present study investigated only presumptive ESBL-EC, it is possible that tet(M) may be present in the selected ESBL-EC subpopulation of the feedlot environment, likely acquired through horizontal gene transfer (HGT) as proposed by Bryan et al. [58,62]. The predominance of sul2, aph(3")-Ib, aph(6)-Id, and floR has also been previously observed in cattle-and human-sources [63][64][65]. These resistances seem to arise with AMU and likely reflect the use of aminoglycosides, sulfonamides, quinolones, and/or phenicols for therapy in humans and cattle.
The exclusivity of bleomycin anticancer antimicrobial inactivator ble [66] in MSewage ESBL-EC likely reflect the sole use of this antimicrobial in humans and not cattle, as sewage would receive human excrement [67,68]. Although this appears to be the first report of the ble gene in municipal sewage in Canada, emerging carbapenemase NDM-1 has been associated with bleomycin resistance in clinical Enterobacteriaceae and attributed to selective pressure of bleomycin or bleomycin-like molecules [69].
Quaternary ammonia compounds are often used as disinfectants in processing plants and clinical settings. A number of qac subgroup members of small multidrug proteins have been previously identified in Enterobactericaeae including qacE, qacE∆1, qacF, qacG, and qacH. These genes are usually associated with plasmid-mediated class 1 integrons, which carry a variety of ARGs [70,71]. The qac genes are frequently found in combination with genes coding for β-lactams, aminoglycoside, sulfonamide, chloramphenicol, and trimethoprim resistance [4,72]. The regular association of qac genes with human isolates in this study is probably due to frequent use of hydrogen peroxide, quaternary ammonium compounds, and sodium hypochlorite-based chemical disinfectants in clinical and household settings [73,74]. Considering that only four isolates from BProcessing were evaluated in this study, the likelihood of detecting qac genes was limited even though these disinfectants are often used in meat processing plants. Overall, the generally low occurrence of QACs in CFeces, CBasins, and SStreams in contrast to occurrence in human isolates may reflect the rare use of these sanitizers in feedlot environments.
Enrichment was required to isolate ESBL-EC from feedlot-related samples but not for samples from human sources. It may be deduced that ESBL-EC in human-sourced samples were levels of magnitude higher, possibly selected on the basis of clinical illness and/or resistant infections. Variants of bla CTX-M-1,-14,-27 were only associated with ESBL isolates from humans, a finding which is compatible with other reports of bla CTX-M-14,-15,-27 association with the global spread of bla CTX-M in human clinical isolates [1,75,76]. Although bla CTX-M-15 was more frequently detected in human-than cattle-related isolates, their occurrence in both cattle-and human-sources may possibly imply transmission between humans and animals [75,77] as ESBL-EC can be zoonotic [78]. In this study, bla SHV genes were rather rare in contrast to other Canadian investigations where bla SHV2 variants have been frequently reported [75,79]. An interesting finding of the current work is that both bla SHV-11 and 12 were detected in a single isolate from CFeces. E. coli bla SHV -12 was reported in one of the first clinical ESBL-EC cases in a dog with recurrent urinary tract infection in Spain in 2000 [80,81]. These findings implicate that it is possible that even minimal interactions may cause resistance gene transfer within a One Health continuum [81].
We also report what is believed to be the first detection of the recently described bla LAP-2 class A β-lactamase gene [82] from E. coli in Canada. The LAP-2 gene (originally described in 2007) encodes narrow spectrum β-lactamase resistance. All the isolates from CFeces (n = 2) and CBasins (n = 1) with LAP-2 also harbored qnrS1, bla TEM -1, and dfrA14, together with several other ARGs. Co-carriage of LAP-2 and plasmid mediated qnrS1 has been previously reported, and LAP has been found to be associated with the same gene cluster together with insertion elements and transposons in clinical E. coli [83] and Enterobacter cloacae [82,84]. The present study found that isolates carrying LAPwere resistant to 8-10 antimicrobial agents and carried at least five distinct plasmid types, implying a high potential for horizontal gene transfer of MDR.

Genotype-Phenotype Antimicrobial Resistance Concordance
There was a strong agreement (≥93.2%) between clinical resistance phenotypes and genetic determinants across the One Health continuum for all antimicrobials tested. Notably, sensitivity and specificity for quinolone (based on qnr and aac(6')Ib-cr genes) were initially low (data not shown). Upon examination of QRDR mutations, sensitivity catapulted to 100%, with each quinolone resistant isolate harboring a resistant gene and/or mutation. In contrast, WGS specificity reduced to 1.1%, with a large proportion of quinolone susceptible isolates also possessing a gene and/or point mutations. These differences between WGS sensitivity and specificity based only on resistance genes versus a combination of genes and mutations may be attributed to the possibility that not all QRDR mutations confer clinical resistance to quinolones or have yet to be confirmed as causing quinolone resistance. A recent study by Varughese et al. [85] found that the commonly reported S83L QRDR mutations did not yield quinolone resistance; although Varughese et al. [85] also noted that coupled with other mutations, S83L conferred quinolone resistance in uropathogenic E. coli.
To assess and facilitate adoption of WGS as the gold standard for AMR surveillance, a number of studies investigating the sensitivity and specificity of WGS in E. coli [86,87] have reported high agreement between a variety of gene determinants and associated phenotypes. In this study, tetracycline, ampicillin, streptomycin, trimethoprim/sulfamethoxazole, and florfenicol resistance exhibited a high concordance for both sensitivity and specificity. An important aspect of WGS for gene detection is the utilization of already identified sequences as references. It is possible that discrepancies between WGS and phenotypic resistance may result from the presence and expression of novel genes which encode for associated resistance phenotypes. Indeed, previous studies have identified phenotypic resistance which were unaccounted for by known resistance genes (e.g., chloramphenicol, gentamicin, streptomycin, and cefoxitin [24]

Insights from E. coli Phylogenetics, MLST and Serogroups
Investigation of the genetic relatedness of presumptive ESBL-EC along the One Health continuum showed generally distinct clades, with evidence of genomes clustering by origin from cattle or human sources, as observed by others [25]. Genetic distinction among ESBL-EC originating from cattle versus human sources along the continuum via phylogenetics implies that isolates are separately adapted to their respective environments and that exchange of isolates between these niches maybe limited [25,89]. For example, ESBL-EC ST131 is reported to be highly human-host specific as it is frequently linked to health-care system exposures rather than environmental sources [90]. It could also be inferred that the segments of the beef production system investigated might not be directly implicated in recent human ESBL-EC infections in this region [25]. Ludden et al. [25] and Salinas et al. [91] also observed distinct clusters between E. coli isolates from livestock and humans. However, a study by Mulvey et al. [92] highlighted the possibility of linkage of MDR E. coli occurring in cattle and human infections in Canadian hospital on the basis of plasmid bla CMY A/C replicon fingerprint similarities >90%. In this study, comparative WGS discriminated between isolates originating from cattle and farm environments which were phylogenetically related, as were isolates from CHumans and MSewage. The frequent clustering of CFeces and CBasins ESBL-EC suggested that these isolates were closely related, suggesting that CBasins were catching flow from feedlot pens and reducing the dissemination of ESBL-EC to the broader environment. However, since a number of CFeces and SStreams were intermixed, it can be deduced that some ESBL-EC may be potentially released into the broader environment as water from catch basins is often utilized to irrigate surrounding crops.
Interestingly, different STs prevailed in different sources, even between CBasins and SStreams. Similar to other reports, ST244 were dominant [75] in cattle-sources, while hyperendemic ExPEC ST131 (implicated in bloodstream and urinary tract infection) [76] and, to a lesser degree, STs 38 and 648 were predominant in human-sources. E. coli ST648 exhibiting an ESBL phenotype have been reported globally in human patients [7,75]. In concordance with the frequent occurrence of ST131 human sources in this study, ESBL-EC isolated from 11 different Canadian medical centers showed a high occurrence of the clonal complex ST131 producing CTX -M-15, -14, and -27 from Brampton, Calgary, and Winnipeg, illustrating its propensity to cause nosocomial infections in Canada [93]. Additionally, the overall findings of distinct ST occurrence between sources agrees with other reports [25,75], where ST131 was only found in E. coli from dogs and not in farm animals [75]. The unique predominance of specific ST also possibly reflects the spread of clonal lineages or the outcome of selective pressures within the diverse segments of the continuum [94,95].
Serogroups in ESBL-EC from CFeces were mostly unknown O serogroups (50%), with a few O8, O89, and O99, while human-sourced ESBL-EC were mostly O25:H4-ST131. Except for a single instance of O26:H11 and wzx-/wzy-Onovel26:H28, our study did not identify any O157 nor any of the other top 6 non-O157 serogroups (O45, O103, O111, O26, O121, and O145) [96]. O157 and the top 6 non-O157 serogroups cause the majority of Shiga toxigenic E. coli (STEC)-associated foodborne infections [96,97]. The infrequent occurrence of these STEC serogroups in ESBL-EC along the continuum could be due to the unspecific nature of the ESBL-EC isolation procedure employed, unlike the specific procedures that are used to isolate STEC, such as direct PCR [98] and sequential immunomagnetic separation [99]. Notwithstanding, the O128 serogroup identified in BProcessing isolates from hides have been linked to serious human infections [97] and may serve as a risk of product contamination during processing. The majority of the O25:ST131 clones from human clinical isolates were CTX-M 15 producers, similar to findings of Aslantaş and Yilmaz [100], who characterized ESBL-EC from Canadian dogs. The specific differences between serogroups along the continuum aligns with MLST and phylogenetic differences in this study and further suggests little to no transmission between cattle and humans.

Mobile Elements: Plasmid and Integrative Conjugative Elements
It is thought that livestock and animal-derived foods are potential sources of ESBL-EC in humans because the same ESBL genes or plasmids have been detected in livestock and farm workers [78,101]. In this study, the greatest plasmid overlaps occurring among all sources of the One Health continuum were IncFIB(AP001918) and IncFII. Johnson et al. [102] noted that, regardless of E. coli source, IncFIB was predominant in avian, human, and poultry meat isolates in the US, although plasmid replicons and colicin-related genes differed among E. coli sources. Likewise, Rodriguez-Siek et al. [103] also found differences in the actual prevalence of plasmid, virulence, phylogeny, and serogroup traits between human and avian E. coli, a finding that is compatible with the differences observed between human and cattle isolates in the present study. In Canada, there is little information regarding the occurrence and frequency of plasmid-mediated transfer of ARGs and virulence determinants among E. coli originating from different segments of the One Health continuum. Nevertheless, plasmid-mediated horizontal transfer of multiple ARGs is important for resistance circulation among bacteria in various sources of the continuum [57,92], considering that~99% of the isolates in our study bore at least one plasmid.
Surprisingly, ICE were found throughout all the genomes sourced from the One Health continuum, even though reports of cattle-related E. coli-sourced ICE are rare [15]. Apart from the characteristic set of core genes encoding functions essential for self-transmission and maintenance, ICE are known to often carry cargo genes that impart various fitness and adaptive advantages to hosts [104,105]. Consequently, besides other well-characterized MGEs including plasmids and integrons, ICE may play a crucial role in the evolution and adaptation of E. coli to various niches, including those environments subject to the selective pressure of antimicrobials. In this study, genes associated with MDR, metal resistance, and pathogenicity functions were noted for some ICE. The presence of diverse ICE families may also imply exchange of genetic material among different bacterial species. For example, the SXT/R391 family of ICE often in Vibrio spp. [106] were present in E. coli genomes. Further characterization of ARG-bearing MGEs and their roles in gene exchange via conjugation would shed more light on the extent and frequency at which these elements transfer AMR and virulence genes among E. coli.

Occurrence of Virulence Genes
Often, STEC are characterized for their stx virulence determinants due to their role in enterohemorrhagic colitis and hemolytic urea syndrome. The repertoire of virulence factors detected in this study included common genes that encode adhesins essential for biofilm formation, host cell invasion, tissue degradation, and host cell death. For example, the fimbria adhesin ecpABCDE that encodes the E. coli common pilus essential for early-stage biofilm development and host cell recognition [107] were common in most isolates (91.4%). Likewise, the type 1 fimbriae determinants fimBCDEFGHI involved in early stage biofilm formation on host mucosa and abiotic surfaces was also common. The detection of T3SS effector protein genes in >70% of isolates in this study demonstrate that all these isolates likely harbored the E. coli locus for enterocyte effacement (LEE), the most important pathogenicity island for enterohemorrhagic and enteropathogenic E. coli [108].
The exclusive occurrence of f17 fimbriae genes in pathogenic E. coli in CFeces and associated CBasins perhaps reflects point-source dissemination of virulence genes from CFeces to CBasins. The f17 fimbriae mediates binding to host intestinal microvilli and has been linked to diarrhea and septicemia outbreaks in calves and lambs [109]. Additionally, the higher occurrence of esp in cattle-sourced E. coli and the higher occurrence of pap, fim, hly, and iuc in human-sourced isolates probably reflects the importance of these factors in E. coli survival or pathogenicity in their respective hosts. All factors considered, the detection of unique virulence genes specific to niche/environmental hosts may support that virulent ESBL-EC occurring in humans did not likely originate from beef cattle. It may also be deduced that, while some virulence genes are ubiquitous due to their importance for fitness and adaptation to biotic and abiotic conditions, other virulence genes may be unique to isolates based on the source characteristic as well as isolate adaptation [26,110].

Conclusions
This study used epidemiologically robust study design and sample collection to examine geographically and temporally related presumptive ESBL-EC isolates from beef cattle, water sources, and human clinical samples in southern Alberta. Comparison of the genotypic and phenotypic characteristics of these isolates along a One Health continuum revealed differences in prevalence of similar ARGs, MGEs, and virulence factors together with phylogenetic differences, implying that ESBL-EC originating from beef cattle may not play a significant role in ESBL-EC infections in humans in southern Alberta. The ubiquity of diverse plasmid and ICE families could be indicative of HGT, even within the same niche and warrants further investigation. This study also documents the first detection of the bla LAP-2 β-lactamase in cattle feces and catch basins and of the ble anticancer resistance genes in municipal sewage in Canada. From specificity and sensitivity outcomes, this study noted that WGS promises to be a robust method for AMR investigations, but warrants further validation and improvement in AMR phenotype prediction accuracy. Future investigation of ESBL-EC based on sound epidemiological project designs will continue to add to the body of work described here and may provide useful insights at to the significance of similarities and differences identified within the various segments of the One Health continuum.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-2607/8/3/448/s1, Figure S1: Occurrence (%) of genes that encode resistance for antibiotics in whole genome sequences of presumptive ESBL E. coli per source of isolates. Figure S2: Prevalence of genetic determinants of antimicrobial resistance in human-versus cattle-sourced E. coli isolates. Figure S3: (A) Pairwise SNP differences between clinical human isolates (n = 25) and municipal sewage isolates (n = 30), processing plants (n = 4), surrounding streams (n = 21), catch basins (n = 42), and cattle feces (n = 40). (B) Pairwise SNP differences between municipal sewage isolates (n = 30) and clinical human isolates (n = 25), processing plants (n = 4), surrounding streams (n = 21), catch basins (n = 42), and cattle feces (n = 40). Figure S4: Whole genome MLST of presumptive ESBL-producing E. coli genomes along the One Health continuum from human clinical, municipal sewage, beef processing, cattle feces, catch basins, and surface streams generated using 9580 wgMLST loci, colored based on isolate origin. Figure S5: Phylogenetic tree generated based on single-nucleotide polymorphisms (SNPs) of the core genes of 108 E. coli isolates that proved to be true ESBL producers (bla TEM , bla CTX-M , bla OXA , and bla SHV ) obtained from cattle feces, catch basin, surface streams, municipal sewage, and human clinical isolates from the One Health continuum, with reference genome E. coli str. K-12 substr. MG1655 (GenBank accession # GI: 545778205/U00096.3). Figure S6: Relative abundance of 38 plasmid types in ESBL-producing E. coli from the One Health continuum of the beef-production system. Table S1: Antimicrobial resistance phenotypes of presumptive ESBL-E. coli. Table S2: Resistance determinants identified in presumptive ESBL-E. coli. TableS3: Whole genome sequencing assembly data for 162 presumptive ESBL-E. coli isolates. Table S4: Identification of mutations in the quinolone resistance-determining regions of presumptive ESBL-E. coli. Table S5: Least square means comparing antimicrobial resistance gene prevalence in cattle versus human isolates. Table S6: Genotype and phenotype comparison of presumptive ESBL E. coli isolates from multiple sources of the One Health continuum. Table S7: Serogroup and ST type of in presumptive ESBL-E. coli. Table S8: Differences between plasmid least square means estimates of the effect of isolate from cattle versus human sources. Table S9: Plasmids identified in presumptive ESBL-E. coli. Table S10: Integrative conjugative elements identified in presumptive ESBL-E. coli. Table S11: Virulence determinants identified in presumptive ESBL-E. coli. Table S12: Differences between virulence determinants identified in presumptive ESBL-E. coli isolated from different sources across a One Health continuum.

Acknowledgments:
We thank Ruth Barbieri, Wendi Smart, and Jillian Rumore for technical assistance. We also gratefully acknowledge Francis Zvomuya, Dept. of Soil Science, University of Manitoba for his assistance with data analysis. We wish to also thank the cooperating commercial feedlots for allowing us access to collect samples.

Conflicts of Interest: CWB is part owner and managing partner of Feedlot Health Management Services and
Southern Alberta Veterinary Services. SJH is an employee at Feedlot Health Management Services, Okotoks, Alberta, Canada. Feedlot Health is a private company that provides expert consultation regarding management and production of calf grower calves and feedlot cattle, including developing veterinary protocols to support animal health. Feedlot Health also conducts in-house and contract research related to dairy calf grower and feedlot production.