Transferable Extended-Spectrum β-Lactamase (ESBL) Plasmids in Enterobacteriaceae from Irrigation Water

Extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae are classified as serious threats to human health by the U.S. Centers for Disease Control and Prevention. Water used for irrigation of fresh produce can transmit such resistant bacteria directly to edible plant parts. We screened ESBL-producing Escherichia coli, Enterobacter cloacae, and Citrobacter freundii isolated from irrigation water for their potential to transmit resistance to antibiotic-susceptible E. coli. All strains were genome-sequenced and tested in vitro for transmission of resistance to third-generation cephalosporins on solid agar as well as in liquid culture. Of the 19 screened isolates, five ESBL-producing E. coli were able to transfer resistance with different efficiency to susceptible recipient E. coli. Transconjugant strains were sequenced for detection of transferred antibiotic resistance genes (ARGs) and compared to the known ARG pattern of their respective donors. Additionally, phenotypic resistance patterns were obtained for both transconjugant and corresponding donor strains, confirming ESBL-producing phenotypes of all obtained transconjugants.


Introduction
The discovery and subsequent broad use of antibiotics in the first half of the last century has revolutionized human and veterinarian medicine. However, a setback promptly followed, as upon discovery of each novel antibiotic, emergence of resistant bacteria soon followed [1]. Antibiotic resistance is conferred by three mechanisms: (1) bacterial target alteration, e.g., by mutation, (2) drug inactivation, e.g., by hydrolyzing enzymes, and (3) decreased intracellular drug concentration through low cell permeability or drug efflux pumps [2]. The genes conferring such antibiotic resistance (ARGs) are of special concern to human health when carried by mobile elements such as phages or plasmids, as these can spread rapidly within bacterial communities and, thus, from commensal or environmental bacteria to human pathogens [3]. The extent and speed of this so-called horizontal gene transfer (HGT) depends on various factors such as bacterial density (biofilms vs. planktonic cells), conditions supporting bacterial growth, or plasmid type (broad-or narrow-host-range plasmids). Barriers to HGT exist, such as recipient helicase compatibility or host restriction system susceptibility, which can reduce but not completely prevent the transfer and establishment of a transferable plasmid in the bacterial host cell [4]. Of note, antibiotic resistance (AR) plasmids usually carry several ARGs, often conferring resistance to multiple antibiotic classes. Antibiotic-resistant bacteria (ARB) with acquired resistance to at least one agent in three or more antibiotic classes are referred to as multidrug-resistant (MDR) [5].

MALDI Biotyping
MALDI biotyping was performed by direct smearing as described previously [32] using a microflex LT MALDI-TOF mass spectrometer (Bruker Daltonics, Bremen, Germany) and the associated MALDI biotyper RTC Software (Version 3.1; Bruker Daltonics). Notably, species belonging to either E. cloacae complex (such as E. ludwigii or E. asburiae) or C. freundii complex (such as C. braakii or C. gillenii) could not be distinguished according to the manufacturer and were grouped therefore into the corresponding complex.

Antibiotic Susceptibility Testing
For all 19 ESBL-producing Enterobacteriaceae from irrigation water, resistance to 32 clinically relevant antibiotics was determined in disk diffusion assays as described previously [15]. Additionally, all ESBL-producing donor and transconjugant strains obtained from filter conjugation experiments were screened for resistance against 31 microbial agents by broth dilution using the MicroScan autoSCAN-4 System (Beckman Coulter Life Sciences, Indianapolis, IN, USA) with the Neg MIC 44 panel. Susceptibility testing was performed according to the manufacturer's guidelines.

DNA Extractions, Sequencing, and Bioinformatics
Genomic DNA was extracted from all selected ESBL-producing Enterobacteriaceae and transconjugants using the commercial GenElute TM Bacterial Genomic DNA Kit (Sigma-Aldrich) according to the manufacturer's instructions. Genomic DNA was sequenced at Eurofins Genomics (Konstanz, Germany) on an Illumina HiSeq4000 instrument (paired-end, 2 × 150 bp). Raw data were assembled to contigs using CLC Genomics Workbench 10.0 (Qiagen, Venlo, Netherlands) for further analysis. The Illumina raw reads (ERR4065331-ERR4065341; ERR4065362-ERR4065369; and ERR4065372-ERR4065376) were submitted to the European Nucleotide Archive (ENA) under BioProject PRJEB37967. Additionally, plasmid DNA from ESBL-producing E. coli and their transconjugants was enriched using a commercial PureYield Plasmid Maxiprep System (Promega, Fitchburg, WI, USA). Enriched plasmid DNA was sequenced on a Pacific Biosciences RSII instrument and used for hybrid assemblies as described previously [15]. The PacBio raw reads (ERR4235917-ERR4235921; ERR4234695; ERR4235916) were also submitted to ENA under PRJEB37967.
In order to estimate the diversity of the tested strains, multi-locus sequence typing (MLST) was performed and a phylogenetic tree constructed from the assembled contigs, using the Center for Genomic Epidemiology (CGE) online tools MLST 2.0 and CSI Phylogeny 1.4 [33,34]. Phylogenetic trees were illustrated using FigTree v1.4.4 [35]. Antibiotic resistance genes and plasmid replicons were identified using ResFinder 3.2 and PlasmidFinder 2.1 [36,37]. All CGE tools were run with default settings for the respective tool version. One type strain per ESBL-producing species was included in the analysis (C. freundii ATCC8090, E. cloacae ATCC13047, and E. coli ATCC11775).

Plasmid Transfer Assays
To examine the frequency of ESBL resistance genes and phenotype transfer to non-ESBL-producers, ESBL-producing irrigation water isolates were used as donors in filter and broth conjugation assays. The donor strain collection comprised 11 ESBL E. coli, 5 ESBL E. cloacae complex, and 3 ESBL C. freundii complex strains. Escherichia coli CV601 served as a recipient strain in all experiments.
The filter conjugation assay was adapted from a previously described method [38]. Donor and recipient strains were grown overnight in Tryptic Soy Broth (TSB, Merck, Darmstadt, Germany) supplemented with CAZ (8 or 0.8 mg/L), RIF (50 mg/L), and KAN (50 mg/L), respectively. Absorbance was measured and adjusted to OD 600 = 4.0 ± 0.5 for all cultures. Recipient and donor cultures were centrifuged (3100 × g, 5 min), the supernatants discarded, the pellets washed twice with 1 mL of 0.1 × TSB, and finally re-suspended in 1 mL of 0.1 × TSB. Then, 500 µL of each donor culture was added to 500 µL of recipient culture and mixed by pipetting. The resulting conjugation mixture was centrifuged (3100 × g, 5 min), the supernatant discarded, and the pellet re-suspended in 200 µL of 0.1 × TSB. For conjugation, Durapore ® membrane filters (Merck Millipore, Burlington, MA, USA) were placed on Plate Count Agar (PCA; Thermo Fisher Scientific, Waltham, MA, USA), and the entire conjugation mixture was transferred to the center of the filter. As controls, after washing, 1 mL of each donor culture only or recipient culture only was centrifuged, re-suspended in 200 µL of 0.1 × TSB, and transferred to a filter. After overnight incubation at 28 • C, filters were placed in 2 mL of sterile 0.9% NaCl solution and vortexed for 1 min. Conjugation suspensions and controls were serially diluted and appropriate dilutions plated in duplicates on PCA plates containing RIF (50 mg/L), KAN (50 mg/L), and CAZ (8 or 0.8 mg/L). To determine the number of recipient cells, spot plating was performed for each conjugation mixture by spotting three 20 µL drops per dilution (10 −5 to 10 −8 ) on PCA supplemented with RIF (50 mg/L) and KAN (50 mg/L). To determine transfer frequencies, the total number of transconjugants was divided by the number of recipients as determined by spot plating. All plates were incubated at 28 • C for up to 72 h.
The broth conjugation assay was performed as described previously [38]. In short, donor and recipient strains were grown in LB overnight (37 • C, 180 rpm). In 2 mL Eppendorf tubes containing 1 mL of LB, 500 µL of each donor strain were mixed with 500 µL of recipient strain and incubated for 24 h at 37 • C. Conjugation mixtures were serially diluted, and 100 µL of appropriate dilutions were plated in duplicate onto LB agar plates containing KAN (50 mg/L), RIF (50 mg/L), and CAZ (8 or 0.8 mg/L) followed by incubation at 37 • C for up to 48 h.
For both, filter and broth conjugation assays, green fluorescence was confirmed under UV light for putative transconjugants. Per replicate, three transconjugants were picked and re-streaked. For confirmation of E. coli CV601 background in transconjugants, repetitive sequence-based PCR fingerprinting (REP-PCR) was performed using the (GTG) 5 primer (5 -GTGGTGGTGGTGGTG-3 ) [39]. Template DNA of putative transconjugants and recipient was obtained by suspending a single colony in 50 µL of sterile water followed by boiling for 10 min at 99 • C. For PCR conditions see Table A1.

Phylogenetic Diversity of Donor Strains
The phylogenetic diversity of the ESBL-producing strains was analyzed by MLST typing and by calculating single-nucleotide polymorphism (SNP)-based phylogenetic trees ( Figure 1). strains (H10 and H30) belonged to ST68 and consequently clustered together in the tree ( Figure 1C). Strain H38, belonging to the global pandemic clone ST131, closely clustered with strain H6 (ST1193) and the reference strain. Taken together, the ESBL-producing E. coli used in this study represented a broad diversity of STs, with the reference type strain lying in between them ( Figure 1C). Interestingly, three strains originating from the same water source (H2, H6, and H17) did not cluster together in the tree, demonstrating that irrigation water can contain multiple and distantly related ESBLproducing E. coli. On the other hand, H10 and H25, also originating from one irrigation water sample, clustered closely together, although they belonged to different STs.
Finally, SNP-distance matrices were calculated for each species group, featuring the number of SNPs detected by pairwise comparison. Based on the minimum and maximum SNP differences detected, E. coli showed the largest range from nearest to farthest strain (min. 25; max. 46,239), followed by E. cloacae (min. 6913; max. 52,453) and C. freundii (min. 13,962; max. 66,505). In terms of maximum SNP difference, the C. freundii strains were the most dispersed.

Antibiotic Resistance and Plasmid Profiles of Donor Strains
For the ESBL-producing E. coli from different irrigation waters, phenotypic resistance determined by disk diffusion as well as ARGs and plasmid replicons detected by sequencing have been published previously [15]. Briefly, the ARGs responsible for the observed ESBL phenotype all belonged to the blaCTX-M gene family. The most frequently detected were blaCTX-M-1 (4 of 11) and blaCTX- For C. freundii complex strains, 78.5% of the reference type strain genome was covered by all three isolates, with the individual strains sharing at least 80.6% with the reference. Two strains (H12 and H41) clustered together in one clade, while the third strain (H42) was separated from them by the reference strain ( Figure 1A). All four strains belonged to different sequence types (STs).
In the E. cloacae complex, 64.7% of the reference strain genome was covered by all five irrigation water E. cloacae strains and at least 71.1% by the single strains. Strains H16 and H24 clustered closely in one clade, while the others were separated clearly, with the reference strain placed in between them. They all belonged to different STs, while three belonged to unknown sequence types for which the nearest known ST is indicated ( Figure 1B).
All together, the analyzed ESBL-producing E. coli covered 72.6% of the reference strain genome with a minimum of 80.2% shared genome when comparing individual strains to the reference. Two strains (H10 and H30) belonged to ST68 and consequently clustered together in the tree ( Figure 1C). Strain H38, belonging to the global pandemic clone ST131, closely clustered with strain H6 (ST1193) and the reference strain. Taken together, the ESBL-producing E. coli used in this study represented a broad diversity of STs, with the reference type strain lying in between them ( Figure 1C). Interestingly, three strains originating from the same water source (H2, H6, and H17) did not cluster together in the tree, demonstrating that irrigation water can contain multiple and distantly related ESBL-producing E. coli. On the other hand, H10 and H25, also originating from one irrigation water sample, clustered closely together, although they belonged to different STs.
Finally, SNP-distance matrices were calculated for each species group, featuring the number of SNPs detected by pairwise comparison. Based on the minimum and maximum SNP differences detected, E. coli showed the largest range from nearest to farthest strain (min. 25; max. 46,239), followed by E. cloacae (min. 6913; max. 52,453) and C. freundii (min. 13,962; max. 66,505). In terms of maximum SNP difference, the C. freundii strains were the most dispersed.

Antibiotic Resistance and Plasmid Profiles of Donor Strains
For the ESBL-producing E. coli from different irrigation waters, phenotypic resistance determined by disk diffusion as well as ARGs and plasmid replicons detected by sequencing have been published previously [15]. Briefly, the ARGs responsible for the observed ESBL phenotype all belonged to the bla CTX-M gene family. The most frequently detected were bla CTX-M-1 (4 of 11) and bla CTX-M-15 (3 of 11). Five of the 11 detected bla CTX-M genes were linked to a known plasmid replicon, 3 of which were identified as IncI1.
The recipient strain E. coli CV601 showed intermediate resistance towards amikacin (AMK) and tobramycin (TOB), as expected due to presence of a gfp-aph(3′)-III cassette. These two resistances were therefore detected in most transconjugants as well (Table 2). Most importantly, the ESBL phenotype was detected phenotypically in all transconjugants. Resistance phenotypes of H6 and H6-TC were identical with the exception of fluoroquinolone resistance (ciprofloxacin (CIP) and levofloxacin (LVX)) observed in the donor strain only. Strains H22 and H22-TC shared all phenotypic resistances, with the exception of recipient-associated aminoglycoside resistance (Table 2). Strain H38 and the respective transconjugant strain were identical in their resistances towards β-lactams. They differed in H38 having fluoroquinolone resistances (CIP and LVX; Table 2). Comparing H40 to its transconjugant, we again observed a common β-lactam resistance profile. In terms of differences, H40 showed resistance to trimethoprim-sulfamethoxazole (SXT), minocycline (MI), and tetracycline (TE), while H40-TC to TE only (and one in three replicates resistant towards MI). Finally, H45 and H45-TC shared all β-lactam resistances.

Strain ID AMK AMC SAM AM ATM FEP CTX CAZ CXM CIP LVX MI PIP TE TOB SXT
The recipient strain E. coli CV601 showed intermediate resistance towards amikacin (AMK) and tobramycin (TOB), as expected due to presence of a gfp-aph(3 )-III cassette. These two resistances were therefore detected in most transconjugants as well (Table 2). Most importantly, the ESBL phenotype was detected phenotypically in all transconjugants. Resistance phenotypes of H6 and H6-TC were identical with the exception of fluoroquinolone resistance (ciprofloxacin (CIP) and levofloxacin (LVX)) observed in the donor strain only. Strains H22 and H22-TC shared all phenotypic resistances, with the exception of recipient-associated aminoglycoside resistance (Table 2). Strain H38 and the respective transconjugant strain were identical in their resistances towards β-lactams. They differed in H38 having fluoroquinolone resistances (CIP and LVX; Table 2). Comparing H40 to its transconjugant, we again observed a common β-lactam resistance profile. In terms of differences, H40 showed resistance to trimethoprim-sulfamethoxazole (SXT), minocycline (MI), and tetracycline (TE), while H40-TC to TE only (and one in three replicates resistant towards MI). Finally, H45 and H45-TC shared all β-lactam resistances.

Molecular Detection of ARGs and Resistance Point Mutations
On the molecular side, we identified ARGs, plasmid replicons, and point mutations known to confer antibiotic resistance in transconjugant strains and their respective ESBL-producing donors. The detected ARGs grouped by contig are in complete accordance between donor and respective transconjugant strain for all five donor-transconjugant pairs ( Table 3). Each of the contigs carrying the bla CTX-M gene was transferred as an entity to the recipient strain, including the long ARG contig of strain H6 encompassing nine ARGs in total. Moreover, the plasmid replicons associated with the bla CTX-M contigs in donor strains were detected in the transconjugant strains as well, with H6-TC forming the only exception. Finally, all transconjugants contained aph(3 )-III originating from the recipient strain. Table 3. Antibiotic resistance genotype including plasmid replicons in ESBL-producing E. coli and their respective transconjugants obtained from filter conjugation. All detected plasmid replicons were full length. Strain H40 was the only donor to harbor a second, non-ESBL plasmid of the IncFIB/IncFIC replicon family with seven ARGs (Table 3). However, transfer to the recipient strain was not detected-as opposed to the IncI1-associated bla CTX-M-1 , sul2, and tet(A) genes-since the plasmid did not carry ESBL genes; thus, transconjugants were not selected.

ID
Several point mutations conferring resistance to fluoroquinolones were identified in the donors H6 and H38. These included mutations in the A-subunit of the DNA gyrase as well as in the genes parC and parE, encoding subunits of the DNA topoisomerase IV (Table 3). No point mutations were detected in any of the transconjugant strains.

Discussion
Bacterial MLST can be used in source tracking. As opposed to extensive E. coli isolate databases revealing associations between known STs and strain origin, information about E. cloacae and C. freundii isolates is scarcer. As described in our previous work [15], the ESBL-producing E. coli STs detected in Swiss irrigation water have been detected previously in water and sewage, but also in humans, domesticated and wild animals, and different food items. As for C. freundii, the corresponding public MLST database [42] contains one entry for a ST234 strain from a diarrheal patient in China and one record for ST345 isolated from food in China. No records exist to date for C. freundii ST402. Enterobacter cloacae ST108 seems to be among the globally most widespread STs [43]. Additionally, three major lineages have been identified among MDR E. cloacae, one of which was defined as ST108-like [44]. Finally, E. cloacae ST910 has been associated with carbapenemase-carrying E. cloacae from sewage water of different Chinese hospitals [45] as well as German surface water [46]. Overall, the diversity and geographic distance of the different STs that were isolated from these sources emphasize the global distribution of these strains.
No transfer of ESBL phenotype or genes could be observed for any of the investigated E. cloacae or C. freundii strains, although known ESBL genes were detected in all but one isolate. However, no plasmid replicons could be identified, suggesting that these genes were not plasmid-borne. The bla CMY-2 -like genes detected in all three C. freundii strains have been identified on plasmids previously [47]. However, they are also known to occur naturally in this species (that is, to confer intrinsic resistance), and could have been mobilized from the C. freundii chromosome onto plasmids by mediation of the insertion sequence ISEcp1 [47]. In E. cloacae strains bla ACT genes were detected. Transmission of such genes between bacteria has also been described [48]. However, in Enterobacter spp. they are known to be intrinsic [49], making them unable of transferring to other bacteria unless gene mobilization takes place first.
Resistance phenotype and genotype of C. freundii and E. cloacae were in good accordance ( Table 1). The ESBL phenotype relied on the presence of the above-mentioned AmpC-type ARGs. The presence of qnr genes was accompanied by phenotypic quinolone resistance. The colistin ARG mcr-9 has been documented to not always confer phenotypic resistance [40], and accordingly we did not observe any colistin resistance phenotypically. Finally, the resistance phenotype of C. freundii H42 nicely reflected all ARGs detected, including tetracycline, sulfonamide, and trimethoprim resistance. Additionally, a plasmid replicon was detected in this strain that was not, however, associated with the bla CMY gene after assembly, but with sul2 only.
Conjugation experiments were conducted on filter/solid medium as well as in liquid broth cultures, as conjugation success can vary between the two systems. We did not observe, however, a difference as the same strains conjugated under both conditions. The five E. coli strains able to transfer their ESBL phenotype by conjugation were the same five in which a known plasmid replicon was associated with the bla CTX-M gene. Plasmid-mediated transmission was further confirmed by the detection of all donor plasmid replicons associated with the bla CTX-M genes in the transconjugants, except for H6-TC. This latter displayed the lowest conjugation rates while harboring the most ARG on the bla CTX-M contig. This heavy ARG burden possibly associated with a large plasmid might be one reason for the low conjugation rates.
The phenotypic resistance profiles of donor and respective TC strains observed in broth microdilution are in good accordance with the molecular resistance profiles. Their different genetic backgrounds can explain most differences between donor and respective transconjugant strains. First, we observed intermediate to full resistance towards the tested aminoglycosides amikacin and tobramycin in most transconjugants. This resistance originates from the recipient strain E. coli CV601, which carries aph(3 )-III encoded by a gfp-cassette and conferring resistance to kanamycin and other aminoglycosides. On the other hand, the donors H6 and H38 were resistant to fluoroquinolones (CIP and LVX), which rely on point mutations in the respective strains and cannot transfer to the recipient. Finally, the H40-TC lost SXT resistance as compared to H40, since dfrA1 for trimethoprim resistance was located on the second, non-ESBL plasmid of this donor strain. Tetracycline resistance, however, transferred along with bla CTX-M-1 and was observed phenotypically in both H40 and H40-TC.
In conclusion, none of the investigated ESBL-producing E. cloacae and C. freundii strains transferred ESBL-encoding genes by conjugation, due to the intrinsic nature of these ARGs. On the other hand, we could show significant conjugative transfer of ESBL-encoding MDR plasmids from irrigation water-borne ESBL-producing E. coli to ESBL-susceptible E. coli, with conjugation in 5 out of 11 strains (up to 4.1 × 10 −5 TC per recipient bacterium). These results highlight the importance of monitoring irrigation water quality for selected ARB in order to mitigate contamination of the irrigated produce with clinically significant bacteria, whose MDR plasmids can moreover spread in bacterial communities.

Acknowledgments:
We are grateful to Khald Blau (Julius Kühn-Institut, Braunschweig) for sharing his valuable technical skills and scientific discussions. We thank Weihong Qi (Functional Genomics Center Zurich) for bioinformatics assistance.

Conflicts of Interest:
The authors declare no conflicts of interest.

Cycling Step PCR Conditions
Enzyme activation 95 • C, 5 min was located on the second, non-ESBL plasmid of this donor strain. Tetracycline resistance, however, transferred along with blaCTX-M-1 and was observed phenotypically in both H40 and H40-TC.
In conclusion, none of the investigated ESBL-producing E. cloacae and C. freundii strains transferred ESBL-encoding genes by conjugation, due to the intrinsic nature of these ARGs. On the other hand, we could show significant conjugative transfer of ESBL-encoding MDR plasmids from irrigation water-borne ESBL-producing E. coli to ESBL-susceptible E. coli, with conjugation in 5 out of 11 strains (up to 4.1 × 10 −5 TC per recipient bacterium). These results highlight the importance of monitoring irrigation water quality for selected ARB in order to mitigate contamination of the irrigated produce with clinically significant bacteria, whose MDR plasmids can moreover spread in bacterial communities.