Multidrug Resistance Dissemination in Escherichia coli Isolated from Wild Animals: Bacterial Clones and Plasmid Complicity

: Objectives. Epidemiological data concerning third-generation cephalosporin (3GC) resistance in wild fauna are scarce. The aim of this study was to characterize the resistance genes, their genetic context, and clonal relatedness in 17 Escherichia coli resistant to 3GC isolated from wild animals. Methods. The isolates were characterized by short-read whole genome sequencing, and long-read sequencing was used for the hybrid assembly of plasmid sequences. Results. The 3GC resistance gene most identiﬁed in the isolates was the extended-spectrum β -lactamases (ESBL)-encoding gene bla CTX-M-1 (82.3%), followed by bla CTX-M-32 (5.9%), bla CTX-M-14 (5.9%), and bla SHV-12 (5.9%). E. coli isolates mainly belonged to the sequence types (STs) rarely reported from humans. The single nucleotide polymorphism (SNP)-based typing showed that most E. coli genomes from wild animals (wild boars, birds of prey, and buzzards) formed clonal clusters (<5 SNPs), showing a clonal dissemination crossing species boundaries. bla CTX-M-1 -harboring IncI1-ST3 plasmid was the predominant ESBL-encoding plasmid (76.4%) in wild animal isolates. Plasmid comparison revealed a 110-kb self-transferable plasmid consisting of a conserved backbone and two variable regions involved in antimicrobial resistance and in interaction with recipient cells during conjugation. Conclusion. Our results highlighted the unexpected clonal dissemination of bla CTX-M-1 -encoding clones and the complicity of IncI1-ST3 plasmid in the spread of bla CTX-M-1 within wild fauna. ESBL-producing E. coli strains for 10 out of 17 isolates in wild fauna.


Introduction
Antibiotics overuse is a key factor in the emergence of antimicrobial resistance (AMR) mechanisms, especially extended-spectrum β-lactamases (ESBL) [1]. ESBLs are enzymes that confer resistance to penicillins and cephalosporins, including third-generation cephalosporins (3GC), classified by the World Health Organization (WHO) as critically important antimicrobial agents in human medicine [2].
Antimicrobial resistance is also a complex and multifaceted threat to humans, animals, and the environment. Advanced knowledge on AMR has demonstrated that it not only affects the human sector but is also widely distributed across animals and the environment [3]. A major cause of the AMR burden is the capability of AMR to transmit within and
a The association of ESBL genes with replicons was determined from Southern blot experiments and in silico analysis of WGS; b deduced from in silico NGS data analysis.

Whole-Genome Sequencing
Whole-genome sequencing (WGS) was performed using the next-generation sequencing platform of the teaching hospital of Clermont-Ferrand, France. DNA was extracted with a DNeasy UltraClean Microbial kit (Qiagen). The libraries were prepared with a Nextera XT Kit (Illumina, San Diego, CA, USA), and sequenced by the Illumina MiSeq system generating 2 × 301-base pair (bp) paired-end reads. The Fastp v0.19.10 [31] was used for quality filtering of Illumina reads and SPAdes [32] for short reads assembly. The average sequencing depth was ≥75×; the number of assembled contigs ranged between 69 and 154, and genome sizes between 4,567,913 and 5,210,322 nucleotides. The raw reads have been deposited in the European Nucleotide Archive (ENA) under the project accession number PRJEB36175.
Genomic Data Analysis E. coli phylogroups and multi-locus sequence typing (MLST) were determined in silico according to ClermonTyping method [33] and Achtman's MLST scheme [34]. The molecular typing of isolates was performed by core genome SNP-based typing (cgSNP) and core genome MLST (cgMLST). BactSNP was used to perform cgSNP using the E. coli EC958 genome [35] as reference (genome size: 5,162,892 bases) as previously described [36,37]. After the filtration of recombination zones detected by Gubbins [38], a phylogenetic tree was inferred from the resulting alignment by maximum likelihood using RAxML [39]. The cgMLST was performed according to the Enterobase cgMLST scheme (https://enterobase.warwick. ac.uk (accessed on 4 February 2021)) and allele calling with package chewBBACA [40]. The antibiotic resistant genes were identified by alignment against a database including the online databases CARD [41], Resfinder [42], and the NCBI National Database of Antibiotic Resistant Organisms (https://www.ncbi.nlm.nih.gov/pathogens/antimicrobialresistance/ (accessed on 1 December 2020)) using a 95% minimum threshold for the breadth of coverage and identity percentage, as previously described [43].

Plasmid Analysis
The plasmid content of bacteria and the size of plasmids were determined using plasmid DNA extracted using the method of Kado and Liu [44]. The PCR-based replicon typing (PBRT) scheme (Diatheva, Fano, Italy) and Southern blots using probes specific for the IncI1, IncN, and IncK replicons and CTX-M-1, CTX-M-14, and SHV-12-encoding genes were used to localize the ESBL-encoding genes in plasmids, as previously described [45]. The PCR-generated probes derived from bla CTX-M -harboring sequenced strains and PBRT kit controls (Diatheva, Fano, Italy). The experimental results were confirmed in silico from WGS using plasmidFinder and the plasmid MLST typing, as previously described [46]. Long reads were obtained from E. coli isolates 2-101 (birds of prey) and J31 (wild boar) using the ligation sequencing kit 1D SQK-LSK109 with the barcoding extension kit EXP-NBD 104 according to the Oxford Nanopore Technologies protocols (Oxford, United Kingdom). The same DNA extract was used for sequencing across Illumina and Nanopore platforms to minimize biological variability introduced in bacterial culture. A flowcell FLO-MIN106 connected to the MinION sequencing device was used to sequence the library during 48 h. Real-time base calling of MinION reads was performed with the MinIT and integrated Guppy software to produce fastQ files. The reads size from Nanopore varied between 120 and 69,617 bases for the J31 dataset (39,647 reads with a median of average quality score per read of Q15-base call accuracy: 97%-and an average size of 11,691 bases corresponding to a total of 463,529,496 bases) and between 107 and 87,164 bases for 2-101 dataset (85,034 reads with a median of average quality score per read of Q15-base call accuracy: 97%-and an average size of 10,090 bases corresponding to a total of 857,958,899 bases). These long reads were de novo assembled with the Illumina short reads using Unicycler v. 0.4.7 [47]. The average sequencing depth of the hybrid assembly was 573× (minimum 139×), and 169× (minimum 30×) for J31 isolate and 370× (minimuFm 89×) and 110× (minimum 20×) for 2-101 isolate from Nanopore and Illumina reads, respectively. Plasmids annotation was performed by the National Center for Biotechnology Information prokaryotic genome annotation pipeline (https://www.ncbi.nlm.nih.gov/genome/annotation_prok/ (accessed on 4 May 2020), project accession number PRJEB36175).

Resistance Phenotype and Resistome
The E. coli isolates exhibited a positive synergy test and resistance to penicillin and cephalosporins, including the oxyimino-cephalosporins cefotaxime or ceftazidime, in agreement with ESBL production. A total of 41% of the isolates were resistant to all quinolones, 76.4% to trimethoprim, and 82.3% to sulphonamides. All the tested isolates were susceptible to amikacin, and only one isolate was resistant to gentamicin. The antimicrobial susceptibility profiles are listed in Supplementary Table S1. WGS analysis detected ESBL-encoding genes in all isolates ( Table 1). The predominant ESBL gene was bla CTX-M-1 (82.3%), followed by the point variants bla CTX-M-32 (5.9%), bla CTX-M-14 (5.9%), and bla SHV-12 (5.9%). Analysis of gyrA and parC mutations responsible for quinolone resistance showed the predominance of the double mutations gyrA (S83L and D87N) associated with parC mutation (S80I; 35.3%). The gene qnrS1 that decreases quinolone susceptibility was identified in 23.5% of the isolates. The most prevalent aminoglycoside resistance genes were the genes aadA (aadA1, aadA2, and aadA5), strA, and strB that affect susceptibility to streptomycin. Only one gene known to affect the susceptibility to gentamicin was detected (aac(3)-VIa; 5.9%), and none affecting amikacin were detected. Overall, the content in resistance genes agreed with the resistance phenotype.
The core genome alignment consisted of 3,499,887 bp, with the isolates differing from each other by 0 to 77,550 SNPs. Three clonal clusters comprising 2 to 4 isolates differed from each other by less than 5 SNPs, while the others diverged by 1831 to 77,550 SNPs (Figure 1a). To provide an unbiased phylogeny, recombination zones were removed. The resulting alignment consisted of 568,547 nucleotides with 0 or 1 SNP for the clonal isolates (≤5 SNPs) and of 211 to 32,758 SNPs for the other isolates. The clonality relationships were also inferred by a cgMLST scheme including 2513 conserved loci (Figure 1b). In agreement with the cgSNP approach, the isolates differed from each other by 0 or 3 alleles for the clonal isolates (≤5 alleles) and by 258 to 2375 alleles for the other isolates (Figure 1b). The clonal clusters belonged to ST48, ST1800, and ST1198, and corresponding isolates were isolated from carnivore birds and a wild boar. These results reveal the clonal dissemination of ESBL-producing E. coli strains for 10 out of 17 isolates in wild fauna.
( Figure 1a). To provide an unbiased phylogeny, recombination zones were removed. The resulting alignment consisted of 568,547 nucleotides with 0 or 1 SNP for the clonal isolates (≤5 SNPs) and of 211 to 32,758 SNPs for the other isolates. The clonality relationships were also inferred by a cgMLST scheme including 2513 conserved loci (Figure 1b). In agreement with the cgSNP approach, the isolates differed from each other by 0 or 3 alleles for the clonal isolates (≤5 alleles) and by 258 to 2375 alleles for the other isolates (Figure 1b). The clonal clusters belonged to ST48, ST1800, and ST1198, and corresponding isolates were isolated from carnivore birds and a wild boar. These results reveal the clonal dissemination of ESBL-producing E. coli strains for 10 out of 17 isolates in wild fauna.

CTX-M-1-Encoding Plasmids
Hybridization experiments revealed that ~110-kb IncI1 plasmids were mainly associated with blaCTX-M-1 (n = 13) and, to a lesser extent, with blaSHV-12 (n = 1). Furthermore, blaCTX-M-1 and blaCTX-M-32 were also localized on ~50-kb IncN plasmid in isolates GV23 and W151, respectively. The probe specific for the blaCTX-M-14 gene hybridized with a ~90-kb IncK plasmid in L16 isolate. The in silico analysis of WGS performed with plasmidFinder identified seven replicons among the isolates, with IncF and IncI1-ST3 replicons among the predominant ones (88.2%). The experimental and in silico plasmid analysis results are listed in the Supplementary Table S2.
The genetic context of the predominant ESBL-encoding gene blaCTX-M-1 was further investigated. Based on short-read mapping and WGS assembly, blaCTX-M-1 was localized in large contigs (size ranged between 38,271-88,238 bp), including the replicon origin IncI1 (60%) or at least a fragment of the region previously called shufflon that is specific to IncI1 plasmids [65,66]. These data agreed with the hybridization experiments and the detection of plasmid IncI1-ST3 in the 13 blaCTX-M-1 isolates. The mapping of corresponding reads against IncI1 reference plasmid R64 confirmed the results and showed a highly similar structure of IncI1-ST3 plasmids except in two variable regions, including the shufflon.
While the ESBL-producing plasmid IncI1-ST3 has already been sequenced in strains isolated from human and human-influenced habitats [67,68], data that have been collected from wild fauna are scarce [23]. The isolates J31 (wild boar) and 2-101 (bird of prey) were, Figure 1. The core genome of extended-spectrum β-lactamases (ESBL)-encoding E. coli isolates analyzed by the cgSNP (a) and cgMLST (b) approaches. The dendrogram from cgSNP was inferred from recombination-free core genome SNPs by the maximum likelihood method (scale: Number of mutations per site). The dendrogram from cgMLST was inferred by the neighbor-joining method (scale: Number of differences in allele). The bootstrap values indicated for the main nodes were calculated from 100 replicates. The clusters corresponding to E. coli phylogroups A, B1, B2, and D are indicated. The number of differences in allele or SNP are indicated for the clonal isolates according to the cgMLST and cgSNP approaches ( 1 before recombination filtration, and 2 after recombination filtration). The clonal relatedness of the strains was determined using less than five variant sites as clonality criteria.

CTX-M-1-Encoding Plasmids
Hybridization experiments revealed that~110-kb IncI1 plasmids were mainly associated with bla CTX-M-1 (n = 13) and, to a lesser extent, with bla SHV-12 (n = 1). Furthermore, bla CTX-M-1 and bla CTX-M-32 were also localized on~50-kb IncN plasmid in isolates GV23 and W151, respectively. The probe specific for the bla CTX-M-14 gene hybridized with a~90-kb IncK plasmid in L16 isolate. The in silico analysis of WGS performed with plasmidFinder identified seven replicons among the isolates, with IncF and IncI1-ST3 replicons among the predominant ones (88.2%). The experimental and in silico plasmid analysis results are listed in the Supplementary Table S2.
The genetic context of the predominant ESBL-encoding gene bla CTX-M-1 was further investigated. Based on short-read mapping and WGS assembly, bla CTX-M-1 was localized in large contigs (size ranged between 38,271-88,238 bp), including the replicon origin IncI1 (60%) or at least a fragment of the region previously called shufflon that is specific to IncI1 plasmids [65,66]. These data agreed with the hybridization experiments and the detection of plasmid IncI1-ST3 in the 13 bla CTX-M-1 isolates. The mapping of corresponding reads against IncI1 reference plasmid R64 confirmed the results and showed a highly similar structure of IncI1-ST3 plasmids except in two variable regions, including the shufflon.
These data confirmed the presence of two variable regions. The first variable region was located downstream of the gene repZ and flanked by an ISSbo1 and ISVsa3, two IS91like elements in plasmids p2-101 and pJ31. This region (12.725 kb) harbored a class 1 integrase gene int1, two resistance gene cassettes (dfrA17 and aadA5), IS26, and truncated IS5075 insertion sequences and sul2 (Figure 2a,b), as previously reported IncI1-ST3 plasmids (e.g., KJ484638, MH846978, and LT985235), but was absent in plasmid of wolf's strain W4 ( Table 1).
The second variable region, the shufflon, was located between the conserved genes rci and pilVA and usually consisted of four segments A to D randomly rearranged by the activity of the rci-encoded recombinase [69]. In contrast to IncI1 reference plasmid R64, shufflon segment D was absent in the InI1-ST3 plasmids, and the remaining segments exhibited different organizations in the circularized plasmids p2-101 and pJ31 (Figure 2b). An additional level of rearrangement was provided by the presence of insertion sequence ISEcp1 associated with bla CTX-M-1 . The element ISEcp1-bla CTX-M-1 , which is well-known to be mobile, was flanked by direct repeats (TTTTTA) and was inserted in the same site within the shufflon segment B. However, this block ISEcp1-bla CTX-M-1 -B exhibited two different orientations in plasmids p2-101 and pJ31 (Figure 2b). Since the shufflon segments are involved in the synthesis of PilV adhesin of the type IV conjugative pilus, the insertion of ISEcp1-bla CTX-M-1 associated with the shuffling of segments could provide diversity in PilV adhesin and, therefore, modify the recognition and the binding of recipient cell during IncI1-ST3 conjugation [65]. The resistance element ISEcp1-bla CTX-M-1 could thereby modulate the transfer efficiency of IncI1-ST3 bla CTX-M-1 plasmids and may be, therefore, involved their successful spread in a broad diversity of E. coli lineages.

Discussion
This work aimed to analyze at the genomic level ESBL-producing E. coli collected from wild animals with different eating habits and living in natural parks and reserves or other rural conservation areas in distinct regions of Portugal [24][25][26][27][28][29][30]. All ESBL-producing E. coli strains were resistant to penicillin, cephalosporin 1st generation (cephalexin), and cefuroxime and 94% were resistant to cefotaxime. The isolates also exhibited associated resistances to quinolones, fluoroquinolones, sulfonamides, and trimethoprim, showing that multidrug-resistant bacteria have emerged in wild birds, as previously reported [6], and in a broad range of other wildlife species.
The bla CTX-M-1 gene was the predominant ESBL-encoding gene (82.3%), followed by bla CTX-M-14 (5.9%), and bla CTX-M-32 (5.9%). In addition to bla CTX-M genes, we detected the bla SHV-12 gene, previously reported in isolates from humans, livestock, companion animals, and wild birds [6,[70][71][72][73]. We also detected genes bla OXA-1 and bla TEM-1 , which have been reported in waterbirds on the Baltic sea coast of Poland [20] and from wild boars and Barbary macaques [74], respectively. Except for the absence of bla CTX-M-15 , these results were also consistent with those previously reported from wild birds [6], in which the bla CTX-M-1 and bla CTX-M-15 genes were the most widespread ESBL genes, followed by bla CTX-M-14 and bla CTX-M-32 . The bla CTX-M-15 gene is the predominant ESBL in humans [6] . The absence of bla CTX-M-15 in our study may reflect a weak anthropization of the wild animals included in this study. Previous reports identified the impact of the human-influenced habitat in the diffusion of the antimicrobial resistance determinant in the wildlife, especially in urban areas where wild birds can live and feed [6,10,75]. Our isolates were collected from natural parks and reserves or other rural conservation areas [24][25][26][27][28][29][30], which are less influenced by human activity than urban areas.
E. coli ST131 and ST405 are the major STs associated with the rapid worldwide rise in bla CTX-M genes, including bla CTX-M-15 [76] and other CTX-M genes, such as bla CTX-M-14 and bla CTX-M-1 [77]. None of our bla CTX-M -positive isolates belonged to ST131 or ST405, showing that the dissemination of bla CTX-M genes in our isolates is not related to wellknown clonal strains, in contrast to previous reports about wildlife [6,10]. Furthermore, E. coli isolates from this study essentially belonged to novel STs (ST1998, ST1800, ST10430), STs rarely reported in human and human-influenced habitats (ST361, ST155, ST117, ST710) and/or STs reported in wild animals (ST117 and ST155) [10]. A striking finding from this study is the clonal dissemination of ESBL-producing E. coli in wild fauna. The isolates belonged to ST48, ST1800, and ST1198 formed clusters that diverged by less than 5 SNPs, suggesting direct and/or indirect transmissions between wild animals. This contrasts with data obtained from human isolates, in which ESBLs are often observed in distantly related E. coli isolates [48][49][50][51][52][53][54][55][56][57][58][59][60][61][62][63][64]. Overall, the genetic background of isolates suggests that ESBL spread in the wild animals of this study is not directly linked to contacts with humans or human-influenced habitats and is marked by direct or indirect transmission within wild fauna despite a probable low antibiotic selection pressure.
Consequently, anthropization is probably not the main spreading factor of bla CTX-M-1, which was observed in five different species (bird of prey, bat, wolf, buzzard, wild boar) out of the seven wild animal species included in this study. 92.8% of the gene bla CTX-M-1 was encoded by the Incl1-ST3 plasmid. Incl1 is one of the most common plasmid families in Enterobacteriaceae, in both animals and humans [6,10,78], and one of the major plasmids contributing to the dissemination of the bla CTX-M-1 gene in animals, the environment, and humans [79]. On the Baltic coast of Poland, ESBL-producing bacteria isolated from waterbirds predominantly had the bla CTX-M-1 gene located on plasmid Incl1 [20]. Similarly, in the Netherlands, the bla CTX-M-1 gene was carried by Incl1 plasmids among E. coli isolates from several wild birds [21]. This association was also more recently detected in ESBL-producing E. coli from various wild animals isolated in Guadeloupe [23]. The identification of the epidemic plasmids IncI1-ST3 encoding bla CTX-M-1 in samples of different sources, including humans, animals, and the environment, further underlined the role of horizontal transfer in the dissemination of resistance genes [80]. Since the IncI1 shufflon segments are involved in the synthesis of PilV adhesin of the type IV conjugative pilus, the insertion of ISEcp1bla CTX-M-1 associated with the shuffling of segments could modify the recognition and the binding of recipient cells during IncI1-ST3 conjugation [65], modulate the transfer efficiency of IncI1-ST3 bla CTX-M-1 plasmids, and may be, therefore, involved in their successful spread in a broad diversity of E. coli lineages even in an antibiotic-free ecosystem.

Conclusions
It emerges from this study that the clonal dissemination of ESBL-producing E. coli strains in wild fauna crosses the species barriers. It is worth noting the moderate diversity of ESBLs detected in our E. coli isolates. The bla CTX-M-1 gene, associated with tetracyclines, fluoroquinolones, aminoglycosides, sulphonamides, and trimethoprim-encoding resistance genes, was the predominant resistance mechanism observed in our isolates, which belonged to unusual E. coli lineages collected from wild animals. Our results also underline the key role of the epidemic plasmid IncI1-ST3 in the diffusion of CTX-M-1 in areas with low or no antibiotic exposure and weak anthropization. The present study is a step forward to understanding the dissemination of antibiotic resistance in the wild ecosystem. More surveillance programs focusing on the spread of antibiotic resistance mechanisms at the interface of ecosystems, including the wildlife, are needed to better assess the emergence of antibiotic resistance in the context of the "one health" approach.