The Role of European Starlings (Sturnus vulgaris) in the Dissemination of Multidrug-Resistant Escherichia coli among Concentrated Animal Feeding Operations.

Antimicrobial use in livestock production is a driver for the development and proliferation of antimicrobial resistance (AMR). Wildlife interactions with livestock, acquiring associated AMR bacteria and genes, and wildlife's subsequent dispersal across the landscape are hypothesized to play an important role in the ecology of AMR. Here, we examined priority AMR phenotypes and genotypes of Escherichia coli isolated from the gastrointestinal tracts of European starlings (Sturnus vulgaris) found on concentrated animal feeding operations (CAFOs). European starlings may be present in high numbers on CAFOs (>100,000 birds), interact with urban environments, and can migrate distances exceeding 1,500 km in North America. In this study, 1,477 European starlings from 31 feedlots in five U.S. states were sampled for E. coli resistant to third generation cephalosporins (3G-C) and fluoroquinolones. The prevalence of 3G-C and fluoroquinolone-resistant E. coli was 4% and 10%, respectively. Multidrug resistance in the E. coli isolates collected (n = 236) was common, with the majority of isolates displaying resistance to six or more classes of antibiotics. Genetic analyses of a subset of these isolates identified 94 genes putatively contributing to AMR, including seven class A and C β-lactamases as well as mutations in gyrA and parC recognized to confer resistance to quinolones. Phylogenetic and subtyping assessments showed that highly similar isolates (≥99.4% shared core genome, ≥99.6% shared coding sequence) with priority AMR were found in birds on feedlots separated by distances exceeding 150 km, suggesting that European starlings could be involved in the interstate dissemination of priority AMR bacteria.

antimicrobials included in the testing panel were further typed by conventional PCR to screen for several major class A β-lactamase genes. β-lactamase genes identified included bla AmpC-CIT (n = 56 isolates), bla CTX-M (n = 35 isolates), and bla TEM (n = 90 isolates) (Supplementary Table 2).
The presence of bla AmpC-CIT was identified in isolates exhibiting a broad spectrum of resistance, often conferring resistance to all β-lactam antimicrobials tested (except IPM). Such relationships between the presence of bla AmpC-CIT and the β-lactam resistance phenotypes observed is highlighted in isolates belonging to antibiograms # 8, 20, and 22. The presence of bla TEM was determined in the isolates which were more frequently comprised in antibiograms # 3, 4, 5, 6, 34, 35, 40, and 41, and primarily conferred resistance to penicillins (AMP, piperacillin; PIP) or intermediate resistance to cefazolin (CFZ) and amoxicillin-clavulanate (AMC). bla CTX-M was detected primarily in isolates grouped within antibiograms # 11, 16, 17, 25, and 27, and was mostly linked to AMR involving increased resistance to first-generation (1G-C) and second-generation (2G-C) cephalosporins. The association of TEM-1 and TEM-2 with increased hydrolysis of penicillins and 1G-C, and that of bla CTX-M (bla CTX-M- 15 ) with increased hydrolysis of other β-lactams and monobactams is well established 30 .

Genetic context of AMR in E. coli collected from European starlings. A prioritized subset of 66
isolates was selected for whole genome sequencing (WGS). In silico analyses of the partial genomes identified a total of 3,435 AMR genetic determinants (94 unique putative AMR genes), with between 35 and 61 unique genes identified per isolate. The majority of the putative AMR genes were associated with drug efflux, and numerous genes involved in antibiotic inactivation and target alteration were identified (Table 1).
Analyses of the predicted protein structures indicated that bla CMY-2 , bla CTX-M-1 , and bla CTX-M-32 were conserved among the isolates tested here. Two distinct sequence variants of bla TEM-1 were present, with three isolates having an N-terminal truncation, which removed the first 28 amino acids compared to the other isolates (sequences otherwise identical except for V29 to M29 in isolates with the truncated determinant). Structural variability in bla amp genes was linked to both a conserved N-terminal truncation as well as amino acid substitutions at 35 different sites.
Phylogenetic relationships and molecular epidemiology of AMR E. coli collected from european starlings. Phylogenetic analysis of the 66 sequenced E. coli genomes showed that the core genomes of the isolates did not always align with isolates within respective collection sites, rather several isolates showed higher similarity with disparate sites across the four states represented in sequenced samples (Fig. 4). Geographically separate isolate groups displayed up to 99.9% similarity at the nucleotide level and up to 100% in shared coding sequences (CDS; Supplementary Table 4) across their assembled genomes and exhibited identical or nearly identical AMR phenotypes and molecular features predicted in silico (Supplementary Tables 2 and 3). Four of the isolate groups were collected on farms separated by 21-183 km (Supplementary Table 4).
Two multilocus sequence typing schemes were used for further subtyping assessment of the E. coli isolates (Supplementary Table 5), with 23 and 18 clonal groups detected with high confidence, respectively. According to the Enterobase database, the majority of the E. coli sequence types (ST) identified here are recognized to occur in livestock, food, wildlife, as well as the environment and were previously shown to have priority AMR phenotypes/ genotypes. E. coli isolates 1567 (ST4542) and 2966 (ST4380) had exceedingly rare STs, with less than 10 entries for these STs described in the Enterobase database. These STs were previously observed in poultry-associated isolates from China, and are primarily recognized in Asia and the Philippines, although ST4380 was also detected in a bla CTX-M-55 -carrying E. coli isolate collected from a monkey in France in 2011 31 .

Discussion
In this study, we examined the AMR phenotypes and molecular epidemiology of CTX and CIP-resistant E. coli at the interface of livestock and European starlings. Importantly, we provide evidence for genetically conserved AMR isolates in European starlings separated by distances exceeding 150 km on cattle production facilities in the Central United States. While directionality of the exchange of these bacteria could not be specifically inferred, our data strongly suggest that European starlings are involved in interstate dissemination of specific lineages of bacteria with priority AMR phenotypes. Our future research efforts will determine if AMR strains harbored by these birds can become established in cattle and CAFO environments.
The identification of CIP-and CTX-resistant E. coli in European starlings from CAFOs was not altogether unexpected, as AMR in the microbial communities associated with livestock production is widespread and is linked to a diverse array of phenotypic and genetic determinants 32 . The observation that CTX-selected isolates were resistant to a greater number of antibiotics was also not surprising considering the types of antibiotics included in the testing panel. Specifically, nine different β-lactams were employed in susceptibility testing, more than any other type of antimicrobial. Further, AMR in CTX-selected isolates was mediated by genetic determinants which typically conferred resistance to a broad range of antimicrobials.
The dynamics of AMR dissemination and introduction into agricultural systems are difficult to interpret considering the number of events and sources in which AMR bacteria could be introduced and disseminated.  www.nature.com/scientificreports www.nature.com/scientificreports/ Nonetheless, based on current research, a logical assumption is that competent wildlife hosts that are found in high numbers on livestock facilities and can travel over large distances would be particularly important contributors to AMR in these settings. High resolution analyses of the genetic backgrounds of such a diversity of E. coli isolates obtained from European starlings had not been investigated previously. This information is critical for assessing the potential role of European starlings as contributors to AMR in livestock production. Molecular typing revealed striking similarities between serotypes and STs from European starlings and isolates from food animals, food, humans, and the environment. This information, in conjunction with the high degree of genomic conservation between certain isolates found in European starlings, provides evidence that clonal spread of AMR E. coli may be mediated by these birds. Similarly, a role of European starlings in the clonal dissemination of the foodborne bacterium Campylobacter jejuni in cattle operations has been previously suggested 33 , even though the analysis of evolutionary relationships of E. coli using molecular typing more closely aligns with reticulate evolution as opposed to clonality 34 . Further, evidence points to presence of multi-resistant E. coli clones (i.e. Figure 2. In silico assessments of AMR genes found within 66 E. coli isolates collected from European starlings that were predicted to confer resistance to β-lactams. CARD-RGI analyses of isolates identified a total of 8 genes which putatively confer resistance to β-lactams. An additional 12 genes putatively involved in efflux of both β-lactams and quinolones were identified (see Supplementary Table 3). The column labeled ' AMR' represents phenotypic resistance characterization, and it is matched by row to the genotypic data presented in the left side columns. Cells shaded in white indicate that the gene was not present or the isolate was susceptible to the particular drug. (2020) 10:8093 | https://doi.org/10.1038/s41598-020-64544-w www.nature.com/scientificreports www.nature.com/scientificreports/ ST131, ST69, ST23) in animals (companion and food animals), foods, and the environment, indicating complex transmission patterns and wide distribution 35 . In our work, we observed no preponderance of specific STs, although certain patterns have emerged, as for example the identification of ST10 in several isolates associated with Inc-type plasmids, with this ST linked by others to both production of ESBL enzymes 36 and hyperexpression of AmpCs 37 .
Multiple studies have suggested that interactions between CAFO cattle and wildlife, including European starlings, contribute to increased cattle infection/fecal shedding of AMR bacteria, and environmental dissemination of AMR bacteria 22,24,[38][39][40] . This risk is further highlighted in a study by Medhanie et al. which suggested that increased prevalence of CTX and CIP-resistant E. coli in bovine feces correlated with increased populations of European starlings, and that the distance between livestock facilities and European starling roosts was significantly associated with the presence of these AMR indicators 41 . A similar result was achieved in the companion study to this work, which indicated that shedding of CIP-resistant E. coli in cattle is positively correlated with a greater abundance of European starlings utilizing the associated production environments 28 .
In assessments of β-lactam resistance among the isolates collected in this study, the presence of bla AmpC-CIT was often found in conjunction with broad-spectrum resistance to β-lactams, including 3G-C antimicrobials such as CTX and CAZ, as well as monobactams represented by ATM. Resistance phenotypes involving reduced susceptibility to 3G-C, and monobactams are commonly mediated by extended-spectrum and AmpC β-lactamases 42,43 . CMY-2 is considered to be the most common type of plasmid-mediated AmpC β-lactamase, frequently isolated from patients in hospitals, livestock, and ground meat [44][45][46] . It should be noted that the majority of isolates characterized in our study, including those containing presumptive plasmid-mediated AmpCs, were determined to belong to the phylogenetic group B1. This phylogroup comprises primarily commensal bacteria 7 , which have been shown to resist stress and persist better in the environment 47 . Additionally, E. coli phylogenetic group B1 was previously found in European starlings sampled in Ireland 24 , providing an indication of host preferences for Figure 3. In silico assessments of AMR genes found within 66 E. coli isolates collected from European starlings that were predicted to confer resistance to quinolones. CARD-RGI analyses of isolates identified 7 genes that putatively confer resistance to quinolones. An additional 12 genes putatively involved in efflux of both β-lactams and quinolones were identified (see Supplementary Table 3). The column labeled ' AMR' represents phenotypic resistance characterization, and it is matched by row to the genotypic data presented in the left side columns. Cells shaded in white indicate that the gene was not present or the isolate was susceptible to the particular drug.

Scientific RepoRtS |
(2020) 10:8093 | https://doi.org/10.1038/s41598-020-64544-w www.nature.com/scientificreports www.nature.com/scientificreports/ this phylogroup. PCR, which we utilized as a screening tool for select antimicrobial resistance genotypes, was generally in excellent agreement with outcomes of WGS characterization. When discrepancies occurred, these were primarily due to incomplete coverage of the primer sets utilized, or a limitation associated with short-read sequencing, which generally results in partial genomes.
A large percentage of the isolates in our study were resistant to TET and CAM. To date, TET resistance remains one of the most commonly encountered resistances in E. coli 48 . TET resistance has been associated with decreased susceptibility to SXT and AMP in E. coli from food animals in multiple European countries. The resistance genetic determinants to these antimicrobials are often found on the same mobile genetic elements, potentially reflecting the history of antimicrobial use in these countries 49 . The high percentages of resistance to CAM in our isolates were somewhat unusual. Only two of the isolates sequenced here encoded specific exporters recognized to confer resistance to phenicols (60 of the 66 isolates sequenced had clinical-resistance to CAM). However, several mechanisms are associated with CAM resistance in bacteria 50 , including target site mutations or modifications 51 , acetylation via acetyltransferases or chloramphenicol phosphotransferases 52,53 , efflux pumps 54 , and decreased outer membrane permeability 55 . Additionally, we note that CAM is banned for use in veterinary medicine (including food animals) in the United States due to issues related to its toxicity 56,57 . Nevertheless, florfenicol (fluorinated synthetic analog of thiamphenicol, a CAM analog) is used to treat respiratory infections in cattle and resistance to this antibiotic is documented in livestock and may be co-selective 58,59 .
Here, we assessed the phenotypic and genetic diversity of AMR E. coli isolates collected from European starlings found on CAFOs in order to improve our understanding of the role of these birds in the transmission of AMR in livestock production systems. We utilized an array of methods, ranging from culture-based to molecular screening to select for priority phenotypes and genotypes for advanced molecular typing and characterization. We selected priority phenotypes and genotypes for downstream WGS analysis in order to facilitate phylogenetic and AMR determinant characterization; however, this approach would not be considered optimal should the goals of the study benefit from determination of genomic diversity in all isolates.
This study shows that E. coli isolated from these birds have important AMR phenotypes and genes, including those which confer resistance to fluoroquinolones and β-lactam antimicrobials, including 3G-C. Our study also identified phylogenetically conserved E. coli isolates in geographically separated European starlings, highlighting a potential link between this invasive agricultural pest and interstate dissemination of AMR in food animal production.

Materials and Methods
Study sites, collection of avian fecal specimens, and culture-based isolation. Feedlots with severe European starling problems (e.g., experiencing more than 10,000 European starlings per day) were previously identified using methodology established for identifying bird damage associated with dairies 60 . Samples from 31 feedlots in Colorado (n = 8 feedlots, 400 birds), Iowa (n = 5 feedlots, 150 birds), Kansas (n = 8 feedlots, 443 birds), Missouri (n = 3 feedlots, 119 birds), and Texas (n = 7 feedlots, 365 birds) were collected between Culture-based isolation. The gastrointestinal tract, from proventriculus to cloaca, was removed from each European starling, placed into a sterile Whirl-Pak bag (Nasco, Fort Atkinson, WI), and homogenized for 120 sec at 230 rpm using a Stomacher 400 Circulator (Seward, Islandia, NY). The resulting homogenate was inoculated using a sterile cotton tipped applicator onto MacConkey agar (Acumedia, Lansing, MI) (MAC) supplemented with 2 μg/mL CTX (Calbiochem, EMD Millipore, Billerica, MA) and onto MAC supplemented with 1 μg/mL CIP (Enzo Life Sciences, Farmingdale, NY) and incubated at 37 °C for 24 hr. From each media, 1-2 presumptive E. coli colonies (occasionally lactose negative isolates were also collected) were subcultured on either MAC-CTX or MAC-CIP to yield purified isolates. Purified isolates were propagated overnight in Brain Heart Infusion Broth (BD, Franklin Lakes, NJ) at 37 °C with shaking at 200 rpm, then mixed 1:1 in 40% glycerol, and stored at −80 °C until further use. Isolate confirmation. Species-level confirmation was performed on all bacterial isolates presumptively identified as E. coli via culture-based methods by MALDI biotyping using a formic acid-acetonitrile extraction procedure 61 . Briefly, one loopful (approximate volume of 1 μl) of bacteria was suspended in a 1:3 solution of HPLC grade water (Sigma-Aldrich, St. Louis, MO) and absolute ethanol (Sigma-Aldrich). According to the manufacturer's protocol, bacteria were collected by centrifugation at 17,000 × g, pellets were allowed to air dry, and then suspended 1:1 in acetonitrile (Sigma-Aldrich, St. Louis, MO) and 70% formic acid (Sigma-Aldrich, St. Louis, MO). Insoluble material was pelleted by centrifugation as described above and 1 μl of the supernatant was applied to a polished steel target plate (Bruker, Billerica, MA). Samples were air-dried and overlaid with 1 μl of freshly prepared α-cyano-4-hydroxycinnamic acid (Bruker, Billerica, MA). MALDI biotyping for genus and species identification was accomplished using a Bruker Ultraflex II TOF/TOF or Bruker Microflex LRF (Billerica, MA) operating with Bruker Biotyper RTC software (Version 3.1) and pre-calibrated with Bruker Bacterial Test Standard. Species level identification of the isolates was accepted if a score of ≥1.7 was assigned by the MALDI Biotyper algorithm. While species-level identification is not secure with a score of ≤2.0, a lower score (≥1.7) was accepted when supported by culture-based identification. PCR-based phylotyping and detection of select AMR genes conferring β-lactam resistance. Phylotype grouping (Supplementary Table 6) was performed using a triplex PCR as previously described by Clermont et al. 63 . Briefly, all DNA was prepared via conventional boil-prep. The conventional PCR assay targeted chuA, yjaA and an anonymous DNA fragment (tspE.C2). An Applied Biosystems 2720 thermal cycler (Foster City, CA) was used for amplification. This procedure permitted classification of the isolates into phylogroups. The class A ß-lactamase genes bla CTX-M , bla SHV , bla TEM and CIT-type AmpCs were detected in a one-step multiplex PCR reaction as previously described by Roschanski  Whole genome sequencing and analysis. The epidemiological, phenotypic, and PCR data collected were used to prioritize a subset of isolates for genetic analyses by WGS. Priority was given to E. coli isolates with similar AMR phenotypes collected between multiple sampling locations and with clinically significant 3G-C resistance (n = 66).