Assessing the Load, Virulence and Antibiotic-Resistant Traits of ESBL/Ampc E. coli from Broilers Raised on Conventional, Antibiotic-Free, and Organic Farms

Poultry is the most likely source of livestock-associated Extended Spectrum Beta-Lactamase (ESBL) and plasmid-mediated AmpC (pAmpC)-producing E. coli (EC) for humans. We tested the hypothesis that farming methods have an impact on the load of ESBL/pAmpC-EC in the gut of broilers at slaughter. Isolates (n = 156) of antibiotic-free (AF), organic (O), and conventional (C) animals were characterized for antibiotic susceptibility and antibiotic resistance genes. Thirteen isolates were whole-genome sequenced. The average loads of ESBL/pAmpC-EC in cecal contents were 4.17 Log CFU/g for AF; 2.85 Log CFU/g for O; and 3.88 Log CFU/g for C type (p < 0.001). ESBL/pAmpC-EC isolates showed resistance to antibiotic classes historically used in poultry, including penicillins, tetracyclines, quinolones, and sulfonamides. Isolates from O and AF farms harbored a lower proportion of resistance to antibiotics than isolates from C farms. Among the determinants for ESBL/pAmpC, CTX-M-1 prevailed (42.7%), followed by TEM-type (29%) and SHV (19.8%). Avian pathogenic E. coli (APEC), belonging to ST117 and ST349, were identified in the collection. These data confirm the possible role of a broiler as an ESBL/AmpC EC and APEC reservoir for humans. Overall, our study suggests that antibiotic-free and organic production may contribute to a reduced exposure to ESBL/AmpC EC for the consumer.


Introduction
Extended-spectrum β-lactamase (ESBL) and AmpC β-lactamase (AmpC)-producing Escherichia coli (ESBL/AmpC-EC) isolates are resistant to extended-spectrum cephalosporins (ESC), a class of antibiotics classified as highest-priority critically important antimicrobials (HPCIA) by the WHO [1,2]. Resistance to ESCs is linked to the presence of genetic determinants belonging to the TEM, SHV, and CTX-M families. OXA-CMY and other families are responsible for resistance to ESBL and AmpC enzymes [3,4]. Such determinants are mostly localised on mobile genetic elements that can be transferred among different lineages of the same bacterial species or between different species [4]. Therefore, the presence of ESBL/AmpC-EC in livestock is a major public health concern, not only with regards to the direct transmission of bacteria from animals to humans, but also with regards to the possible transfer of resistance determinants from commensal to zoonotic bacteria in the gut flora [3]. Among livestock production chains, poultry is the most likely livestock-associated ESBL/AmpC-EC reservoir relevant to human health [5].
According to EUCAST guidelines, the isolates were considered as ESBL producing Enterobacteriaceae and/or AmpC β-lactamase-producing Enterobacteriaceae using the double disk synergy test (DDST) with a disc of cefotaxime (30 µg) and a disc of amoxicillinclavulanate (containing 10 µg of clavulanate), positioned at a distance of 30 mm (center to center), according to EUCAST guidelines [27].
Finally, susceptibility to colistin was assessed using the broth microdilution method in order to determine the minimal inhibitory concentration (MIC), using polystyrene microtiter plates (LP Italiana SpA, Milan, Italy) and sulfate salt of colistin (Sigma Aldrech SRL, Milan, Italy), according to the EUCAST recommendations [28]. E. coli strains ATCC 25922 and ZTA14/0097EC [a kind gift from Professor Lucas Dominguez Rodriguez, Centro de Vigilancia Sanitaria Veterinaria (VISAVET), Universidad Complutense, Madrid, Spain] served as the quality-control strains. The classification of isolates as resistant was based on MIC values using the EUCAST criteria (resistant: MIC > 2 mg/L).

DNA Extraction
For the DNA extraction of bacterial isolates, the QIAamp DNA Mini Kit (Qiagen, Hombrechtikon, Switzerland) was used following the manufacturer's instructions.

Determination of Phylogenetic Groups
E. coli isolates were tested by PCR for characterization of the phylogenetic groups A, B1, B2, C, D, E, and F, according to Clermont et al. [29].

Determination of ESBL-and AmpC-Associated Genes
The presence of AR genes was determined on isolates classified as ESBL and/or AmpC. For ESBL-enzymes detection, two multiplex PCRs and one simplex PCR were performed in this study: a bla TEM /bla SHV /bla OXA-1-like multiplex PCR; a bla CTX-M multiplex PCR including phylogenetic groups 1, 2 and 9; a bla CTX-M-8/-25 simplex PCR [30]. For AmpC-enzymes ACC, FOX, MOX, DHA, CIT and EBC one multiplex PCR was performed as previously described [31]. The presence of the allelic variant bla CTXM-15 was then investigated in all isolates positive for bla CTX-M group 1 [32] and all the isolates positive for bla CIT -group were further investigated for bla CMY-2 group using a previously described method [33].

Whole Genome Sequencing
In order to investigate sequence type, serotype, virulence profile, and antimicrobial resistant genes, 13 E. coli isolates were WG sequenced. The subset isolates were selected among the F and D phylogenetic group isolates, which are known to be pathogen for humans.
Genomic DNAs of the pure E. coli cultures were extracted from 1 mL of logarithmic phase broth cultures using QIAamp DNA Mini Kit (Qiagen Inc., Hilden, Germany) following the manufacturer's protocol for Gram-negative bacteria organisms. Each sample was then quantified with the Qubit fluorometer (QubitTM DNA HS Assay, Life Technologies, Thermo Fisher Scientific Inc., Milan, Italy). Library preparation was obtained using the Nextera XT Library Prep kit (Illumina Inc., San Diego, CA, USA) according to the manufacturer's manual. The prepared libraries were loaded onto NextSeq 500/550 Mid Output Reagent Cartridge v2, 300 cycles kit (Illumina Inc., San Diego, CA, USA) and then sequenced on an Illumina NextSeq 500 platform to generate 150 bp paired-end reads.
The bioinformatics analysis was carried out using the services of the Centre for Genomic Epidemiology (CGE), Technical University of Denmark (DTU, https://cge.cbs.dtu. dk/services/ accessed on 9 November 2021) [38][39][40][41][42]. Briefly, the fasta files were analyzed using the following CGE databases: MLST, SeroTypeFinder, ResFinder for the acquired antibiotic resistance genes, and VirulenceFinder for identifying the putative virulence factors of isolates. Based on the virulence genes described, the 13 isolates were screened for the presence of APEC virulence genes. Strains were classified as APEC, a subtype of ExPEC pathotype, when at least four among iroN, iutA, iss, ompT, and hlyF genes were present [43]. Moreover, we tested on our isolates the refined definition of APEC according to Johnson T. et al. [44]. Furthermore, we tested our fasta files using the Clermon-Typer [45], a user-friendly tool for Escherichia species/phylogenetic group identification (http://clermontyping.iame-research.center/ accessed on 9 November 2021).

Statistical Analyses
The bacterial counts were converted to Log CFU/g of cecal content for statistical analysis. The difference in bacterial loads from animals from the three production types were evaluated by using the Kruskal-Wallis test, after assessing the normality of data by using the Shapiro-Wilk method. The analyses were performed by using Stata 11.2 (StataCorp, College Station, TX, USA). The difference in the proportion of isolates between animals from the three production lines was evaluated by using Pearson's χ2 or Fisher's test, with a significance threshold of p ≤ 0.05. The strength of the association was evaluated by using the odds ratio (OR), with the conventional production type as a reference (OR = 1). The analysis was performed using R (version 4.0.2, access date: 14 July 2020) package epiR [46]. To show the distribution of the putative virulence genes across the pathotypes, we performed Kruskal-Wallis rank sum testing. To investigate the distribution of genes encoding putative virulence factors, we constructed a heat map based on the distance metric "euclidean" and complete linkage method.

E. coli loads in the Three Management Systems
A total of 809 fecal samples from broiler cecal content at a slaughterhouse in Umbria, Italy, were collected. Samples were from three production types of breeding: conventional (C; N = 10), organic (O; N = 9) and antibiotic-free (AF; N = 11).
A total of 156 isolates were cefotaxime-resistant after culturing on MacConkey cef and they were further classified as E. coli. As showed in Table S1, fifty-five isolates belonged to AF production type (35%), 47 to organic production type (30%) and 54 to conventional type (35%).

Statistical Analyses
The bacterial counts were converted to Log CFU/g of cecal content for statistical analysis. The difference in bacterial loads from animals from the three production types were evaluated by using the Kruskal-Wallis test, after assessing the normality of data by using the Shapiro-Wilk method. The analyses were performed by using Stata 11.2 (StataCorp, College Station, TX, USA).
The difference in the proportion of isolates between animals from the three production lines was evaluated by using Pearson's χ2 or Fisher's test, with a significance threshold of p ≤ 0.05. The strength of the association was evaluated by using the odds ratio (OR), with the conventional production type as a reference (OR = 1). The analysis was performed using R (version 4.0.2, access date: 14 July 2020) package epiR [46]. To show the distribution of the putative virulence genes across the pathotypes, we performed Kruskal-Wallis rank sum testing. To investigate the distribution of genes encoding putative virulence factors, we constructed a heat map based on the distance metric "euclidean" and complete linkage method.

E. coli loads in the Three Management Systems
A total of 809 fecal samples from broiler cecal content at a slaughterhouse in Umbria, Italy, were collected. Samples were from three production types of breeding: conventional (C; N = 10), organic (O; N = 9) and antibiotic-free (AF; N = 11).
A total of 156 isolates were cefotaxime-resistant after culturing on MacConkey cef and they were further classified as E. coli. As showed in Table S1, fifty-five isolates belonged to AF production type (35%), 47 to organic production type (30%) and 54 to conventional type (35%).

Antimicrobial Susceptibility Testing Results
Fifteen antimicrobial molecules were tested. All isolates were susceptible to colistin and imipenem. The proportion of isolates resistant to the other tested antibiotics is shown in Table 1. Antibiotic resistance equal to or higher than 50% was found for ampicillin, first generation cephalosporins (cefazolin), sulfonamides and tetracycline, independently from the production type. Resistance equal to or higher than 50% for amoxicillin/clavulanic acid and cefotaxime was found in organic and conventional production lines, for nalidixic acid in antibiotic-free and organic lines and for sulphametoxazole + trimethoprim only for conventional the production line. A resistance to colistin was not detected. The percentages of isolates resistant to chloramphenicol, ciprofloxacin, streptomycin, sulphonamides and trimethoprim + sulfamethoxazole were different among the three groups (Pearson χ2: p < 0.05). The organic production type was a protective factor compared to the conventional production type for chloramphenicol and sulphonamides (OR 0.20; 95%CI 0.07-0.54), while belonging to the antibiotic-free type represented a protective factor for ciprofloxacin, sulphonamides and sulfamethoxazole +trimethoprim resistances ( Table 2). For the other antibiotic molecules, there were no differences in the percentages of resistance among the three production types (Pearson χ2: p > 0.05).   All isolates were classified as MDR, because of resistance to three or more antimicrobial classes.
ESBLs and/or AmpC β-lactamases were detected by PCR in 156 isolates. The results are summarized in Table 3. According to the distribution of resistance genes, 131 (84%) isolates were phenotypically ESBL-producing E. coli, 18 (11%) were AmpC producers, 3 (2%) were both ESBL and AmpC producers, while 4 (3%) were classified as negative, according to EUCAST, 2018 (EUCAST, 2018).  ESBL-producing isolates belonged to CTX-M group 1 (42.7%), even in combination with other enzymes, in particular with TEM-type (29%), followed by SHV alone (19.8%). All CTX-M group 1 positive isolates, except three, belonged to the CTX-M-15 allelic variant. The CTX-M-1 + isolates were not randomly distributed among the groups (Pearson χ2: p = 0.0013): belonging to an antibiotic-free system was a protective factor (OR 0.23, 95%CI 0.10-0.54), while isolates from organic farms were not different from the conventional ones (OR 0.6, 95%CI 0.24-1.50). None of the isolates belonged to the other investigated CTX-M groups. The other most prevalent group, bla SHV , was found mainly in antibiotic-free (47%), followed by conventional type (16.7%) and organic type (8.5%). The proportion of SHV-1 isolates was different among the three groups (Pearson χ2: p < 0.001), with isolates from antibiotic-free farms having 4.48 OR (95%CI 1.84-10.92) of being SHV + compared to conventional farms. The CIT and FOX enzymes were the most frequently observed plasmidmediated AmpC-lactamases, but always in association with other enzymes belonging to ESBL-type. One out of twenty positive isolates for the CIT group was also positive for the CMY-2 group.
Neither the proportion of isolates with an ESBL phenotype, AmpC phenotype, or being positive for TEM and CIT varied among the three categories (Pearson χ2: p > 0.05). Using the in silico method of ClermonTyping we identified three phylogroups: D, F and G. Seven isolates out of thirteen (53.8%) belonged to the phylogenetic group D, 4/13 (30.8%) to the G and two isolates were phylogenetic group F (15.4%).

Whole Genome Sequencing (WGS)
We identified nine STs and the most prevalent were ST117, counting for 30.8% (n = 4) of isolates, and ST349, counting for 15.4% (n = 2) of isolates. Other STs, with one isolate for each ST, are reported in Table 4. Nine isolates (69.3%) showed the APEC pathotype (Table 4). Among those, all isolates belonging to the ST117 and phylogenetic group G were APEC and ESBL, except one, which showed an APEC and AmpC phenotype (Figure 2). The assembled contigs were typed in silico using the CGE databases and SerotypeFinder predictions corroborated O and H antigens, shown in Table 4.
When we investigated the distribution of putative virulence genes in a subset of 13 isolates, we found 49 putative virulence genes and included them in the heat map ( Figure 3). The most frequent virulence genes were chuA, gad, iss, terC, present in all isolates, while traT was detected in 12 out of 13 isolates (92.3%), ompT and sitA in 11/13 (84.6%), and hlyF, iucC, iutA in 10/13 (76.9%). Genes encoding for the heat-lable (LT) and heat-stable (ST) toxin, which characterize the Enterotoxigenic E. coli (ETEC) pathotype, were not detected in our isolates [47]. Putative virulence genes were not randomly distributed across the pathotypes (Kruskal-Wallis rank sum test; p = 0.01). The APEC isolates showed a higher number of virulence genes than the AFEC isolates ( Figure 3). We did not observe differences in the presence of virulence factors between ESBL and AmpC phenotypes (Kruskal-Wallis rank sum test; p > 0.05).

Discussion
Resistance to extended-spectrum cephalosporins, molecules classified as HPCIAs by the WHO (WHO, 2019), in E. coli in the poultry sector has raised serious concerns regarding the transfer of clones to humans or the exchange of resistant genes between poultry and human flora. In the present study, the presence of ESBL/AmpC-EC in broilers at a slaughterhouse was observed in animals that were not subjected to antibiotic therapy during the production cycle. This confirms what has been described in the literature, where there was an expansion of ESBL/AmpC-EC clones along the broiler production cycle, even in the absence of selective pressure, despite low starting loads [48,49]. Resistance can be maintained, in the absence of antibiotic selection, through resistance mutations that may incur no fitness costs and compensate for the costs of resistance via second-site mutations, which restore organismal fitness [50].
The load of commensal E. coli was not even in the three types of production, with broilers from conventional farms having higher E. coli loads than O and AF broilers, although the difference among the three medians was small (approximately 0.2 Log CFU). This difference is likely due to the younger age at slaughter for C broilers compared to that of O and AF broilers [13], since a young age is associated with higher E. coli loads in the gut [51].

Discussion
Resistance to extended-spectrum cephalosporins, molecules classified as HPCIAs by the WHO (WHO, 2019), in E. coli in the poultry sector has raised serious concerns regarding the transfer of clones to humans or the exchange of resistant genes between poultry and human flora. In the present study, the presence of ESBL/AmpC-EC in broilers at a slaughterhouse was observed in animals that were not subjected to antibiotic therapy during the production cycle. This confirms what has been described in the literature, where there was an expansion of ESBL/AmpC-EC clones along the broiler production cycle, even in the absence of selective pressure, despite low starting loads [48,49]. Resistance can be maintained, in the absence of antibiotic selection, through resistance mutations that may incur no fitness costs and compensate for the costs of resistance via second-site mutations, which restore organismal fitness [50].
The load of commensal E. coli was not even in the three types of production, with broilers from conventional farms having higher E. coli loads than O and AF broilers, although the difference among the three medians was small (approximately 0.2 Log CFU). This difference is likely due to the younger age at slaughter for C broilers compared to that of O and AF broilers [13], since a young age is associated with higher E. coli loads in the gut [51].
In our study, we found that the ESBL/AmpC-EC load in the caecum of broilers from conventional and antibiotic-free farms was approximately 10 4 /g, which was similar to that described in the literature for broilers of the same age [48,52].
In contrast, in O broilers, ESBL/AmpC-EC loads were more than ten times lower than those recorded in C broilers. Our data were obtained from samples collected from the caecum, and not from the meat. The final contamination of meat is generally lower than that of the gut. The contamination of meat depends on several factors apart from the presence of ESBL/AmpC-EC in the gut, including cross-contamination among carcasses, contamination by operators or from the premises of the food chain, and reduction of contamination after chilling [53]. According to the findings of the European official monitoring for AMR, the prevalence of ESBL/AmpC-EC in meat is lower, but still comparable to that in the gut, suggesting a strong connection between the contamination of the gut and that of meat in the broiler production chain [53]. Therefore, the low ESBL/AmpC-EC loads in the organic production system observed in this study may represent a reduced risk to the consumer [52].
ESBL/AmpC-EC isolates also showed resistance to multiple antibiotic classes historically used in the poultry sector, including penicillins, tetracyclines, quinolones, and sulfonamides, in agreement with what was observed in previous studies [16,54]. ESBL/AmpC-EC isolates from organic and antibiotic-free farms were less resistant to some antimicrobial classes. Among these, we observed minor resistance to antibiotics historically used in livestock, such as sulfonamides. Interestingly, the odds of being resistant to fluoroquinolones were halved in ESBL/AmpC-EC isolates from antibiotic-free farms compared to those from the conventional farms. The combined resistance to ESC and fluoroquinolones, two classes classified as HPCIA, is worrisome [13,55]. Our data suggest that isolates from antibiotic-free and organic broilers are less resistant than those from conventional farms.
Biomolecular analyses have also revealed the presence of genes encoding TEM, SHV, and CTX-M in ESBL/AmpC-EC, confirming their involvement in the poultry sector [55]. In particular, CTX-M-1 is generally found on conjugative plasmids, is associated with other resistance genes, and is prevalent as the CTX-M-15 variant. CTX-M-15 has undergone global dissemination over the last 30 years [55,56]. This has been attributed to several factors, including the dispersion of mobile genetic elements and/or the dissemination of successful bacterial clones, favoured by the low fitness cost associated with the presence of CTX-M-1 genes [55]. In the veterinary sector, selective pressure generated by antibiotics is likely to have contributed to the spread of CTX-M-1 ESBL [55]. In our study, we observed a reduced proportion of CTX-M-1 in ESBL E. coli isolated from antibiotic-free broilers. This result might be explained by the presence of different bacterial clones on the antibiotic-free farms. Alternatively, the lack of selective pressure generated by antibiotic use may have limited the spread of CTX-M-1 among animals.
Taken together, our findings suggest that broilers from non-conventional farms, characterised by higher biosecurity standards, older age at slaughter, and no antibiotic use, are less risky in terms of ESBL/AmpC-EC contamination as compared to broilers from conventional farms.
The A phylogenetic group, associated with commensal bacteria in humans, is the most represented among ESBL/AmpC-EC of avian origin [57,58]. Numerous isolates from human extra-intestinal infections belong to the F and D phylogroups described in our study, together with commensal phylogroups [58]. Using WGS, we analysed the presence of E. coli isolates belonging to the G phylogroup, a new phylogenetic group generally misidentified as the F phylogenetic group through PCR analysis [59]. In humans, phylogenetic group G strains represent approximately 1% of the E. coli isolates and are found in both commensal and extra-intestinal pathogenic conditions, including septicaemia [60]. A limitation of our study is that only a small fraction of isolates underwent WGS. Nevertheless, we were able to describe the APEC pathotypes in our dataset independent of the production systems. APEC strains utilize different virulence and pathogenesis factors to cause disease in broilers, primarily adhesins, invasins, protectins, iron acquisition systems, and toxins [9], determinants that we found in our isolates. These factors facilitate the adhesion, invasion, evasion from host immune responses, colonisation, proliferation, and systemic dissemination of APEC, thereby allowing the establishment of infection in broilers [61]. Despite this classification, a considerable overlap of virulence determinants with human ExPEC can be found among a subset of APEC, indicating the non-host specificity of these strains and underscoring their zoonotic potential [9]. ExPEC subtypes, such as APEC isolates, can asymptomatically colonise the gut of a fraction of healthy animal population and survive in extra-intestinal environments, causing diseases in animals and humans throughout the food chain [4].
Consistently, phylogroup G strains belonged to ST117, which is the most prevalent phylogroup G lineage in broilers and poultry meat products from Northern Europe and Canada [62,63]. This sequence type has spread throughout the Nordic broiler production and has been implicated in large outbreaks of colibacillosis [63]. In our study, four ST117 strains were classified as APEC, confirming their pathogenic potential in poultry [64,65]. As part of ST117, we found another relevant ST associated with extra-intestinal infections, namely the ST362, with increased antibiotic resistance and enhanced virulence [66]. Thus, our data confirmed that ESBL/AmpC-EC in poultry could be APEC pathogens. These data confirm the possible role of the broilers as an ESBL/AmpC-EC reservoir associated with extra-intestinal forms in humans. In this study, broilers from non-conventional farms showed decreased ESBL/AmpC-EC loads at the slaughterhouse. As previously reported, ESBL/AmpC-EC isolates from non-conventional farms showed a more favourable antibiotic resistance profile than those from conventional farms. Overall, our study confirmed that antibiotic-free and organic production may contribute to a reduction in consumer exposure to ESBL/AmpC-EC.