Persistence and Dissemination Capacities of a blaNDM-5-Harboring IncX-3 Plasmid in Escherichia coli Isolated from an Urban River in Montpellier, France

To investigate the capacities of persistence and dissemination of blaNDM-5 within Escherichia coli and in aquatic environment, we characterized E. coli (sequence type 636) strains B26 and B28 isolated one month apart from the same urban river in Montpellier, France. The two isolates carried a pTsB26 plasmid, which sized 45,495 Kb, harbored blaNDM-5 gene and belonged to IncX-3 incompatibility group. pTsB26 was conjugative in vitro at high frequency, it was highly stable after 400 generations and it exerted no fitness cost on its host. blaNDM-5harboring plasmids are widely dispersed in E. coli all around the world, with no lineage specialization. The genomic comparison between B26 and B28 stated that the two isolates probably originated from the same clone, suggesting the persistence of pTsB26 in an E. coli host in aquatic environment.


Introduction
Antimicrobial resistance (AMR) occurs worldwide and the World Health Organization has identified it as one of the three main threats to human health [1,2]. Due to the overuse and misuse of antimicrobials, selective pressure exerted on bacteria significantly enhances AMR, with consequences for antimicrobial treatments failures [3]. At first neglected, the environment is now considered as a main player in the emergence and diffusion of AMR. Water plays a major role in interconnecting different ecosystems such as humans, animals, soils and hydrosystems [4][5][6][7][8]. Urban and rural surface waters constitute hotspots for exchanges among microorganisms of human and environmental origin, which are all subject to strong selection pressures due to diverse pollutions [3,5]. β-lactams are by far the most widely consumed antibiotics worldwide [9], and among β-lactams, carbapenems are last resort treatment for multidrug-resistant bacterial infections [10][11][12]. Several studies reported the occurrence of Carbapenemase Encoding Genes (CEGs) in aquatic environments [6][7][8]. However, there is a lack of studies on the persistence and dissemination capacities of CEGs and bacteria carrying CEGs in waters.

Characteristics of the Two NDM-5 Producing E. coli Isolates B26 and B28
B26 and B28 E. coli were isolated in the urban river of Font d'Aurelle, in August and September 2015, respectively. They both displayed high level of carbapenem resistance, with minimum inhibitory concentrations (MIC) over 15 mg/L for ertapenem, meropenem and imipenem.
The B26 genome is 4927 Mb and includes four plasmids belonging to IncFIA/FIB, IncFII, IncQ1 and IncX-3 incompatibility groups. The inventory of antimicrobial resistance genes identified bla NDM-5 , bla TEM-1B and bla SHV-12 (previously identified as bla SHV-5 by multiplex PCR GeneXpert Cepheid). Genome of B28 is 4858 Mb, including a unique IncX-3 plasmid and bla NDM-5, was the only antimicrobial resistance gene detected.
Both isolates were affiliated to the sequence type (ST) 636 (B2 phylogroup) by in silico multilocus sequence typing. Whole-genome alignment and comparison showed no large indels. Only 21 variations between B26 and B28 genomes were identified, including 19 single nucleotide polymorphisms (SNPs) and two multi-nucleotide polymorphisms (four nucleotides). Among the 21 variations, six variations were related to hypothetic encoding sequences, four of them corresponding to synonymous mutations and two to missense mutations.

Characteristics of the bla NDM-5 -Harboring IncX-3 Plasmids
Transformation assays in E. coli TOP10 with plasmids extracted from B26 and B28 were successful and gave two transformants, TsB26 and TsB28. These two transformants displayed high level of resistance to carbapenems (MIC ≥ 7.5 mg/L whatever the carbapenem). They were positive to specific PCRs anchored in pir gene (i.e., specific of IncX-3 plasmid) and bla NDM . PCR tests indicated that both transconjugants did not contain any other plasmid.
Genomic sequences of TsB26 and TsB28 were aligned and compared with those of E. coli TOP10. TsB26 presented a unique additional sequence onto a single contig corresponding to IncX-3 plasmid. The extremities of this contig were bla NDM-5 and ISAba125, corresponding to a plasmid region already sequenced [29]. Complete identity was observed between pTsB28 and pTsB26 encoding bla NDM-5 plasmids. So, the plasmid was called pTsB26 from then on. pTsB26 sizes 45,495 kb with a GC content of 46.5%. It belongs to IncX-3 plasmid incompatibility group. Nucleotide sequence analysis revealed 57 predicted open reading frames corresponding to 57 encoding genes ( Figure 1). Alignment of pTsB26 with IncX3plasmid conserved backbone, described by Liakopoulos [30], showed that pTsB26 backbone was typical of IncX-3 group. It is approximately 25 kb and includes encoding genes for replication (pir and bis), entry exclusion (eex), plasmid stability (parAB, topB and hns) and conjugative transfer (pilX1-11 and taxA-C) [30]. The accessory module of about 20 kb contains bla NDM-5, which is preceded by IS3000 and ISAba125 and followed by ble MBL (bleomycin resistance gene), trpF (N-5 phosphoribosylanthranilate isomerase), dsbD (disulfide oxidoreductase) and umuD (encoding a protein implicated in the SOS system).

Conjugative Transfer Success of pTsB26 In Vitro
Conjugative transfer rate of pTsB26 was studied by mating assays, with B26 TsB26 and TsB28 as donor strains and XL1-Blue E. coli as receptor. All strains succes transferred pTsB26, with high transfer rates (Table 1), and all transconjugant strain played carbapenem resistance (MIC ≥ 7.5 mg/L).

In Silico Population Study of BlaNDM-5-Harboring IncX-3 Plasmids among E. Coli Spec
A genomes dataset was constructed with 28 complete genomes of E. coli-car blaNDM-5 on an IncX-3 plasmid and the genomes of B26 and B28 strains. In silico seq types, phylotypes and metadata associated with the genomes are presented in Ta More than 96% of the genomes belong to A, B1 and C phylogroups. B26 and B28 we only genomes of the B2 phylogroup in the dataset. The 31 genomes correspond strains isolated from diverse origins: humans (n = 20), environment (n = 6) and anim = 4).  BLASTn analysis showed that pTsB26 displayed nearly the same sequence (45494/45495 bp) as the plasmid pEC7-NDM-5 (accession number: MH347484) found in a E. coli strain isolated from dog in South Korea. A high homology (45475/45495 bp) was also observed with the well characterized IncX-3-blaNDM5 pEC463-NDM5 plasmid (accession number: MG545911) in an E. coli clinical strain from China [31].

Conjugative Transfer Success of pTsB26 In Vitro
Conjugative transfer rate of pTsB26 was studied by mating assays, with B26, B28, TsB26 and TsB28 as donor strains and XL1-Blue E. coli as receptor. All strains successfully transferred pTsB26, with high transfer rates (Table 1), and all transconjugant strains displayed carbapenem resistance (MIC ≥ 7.5 mg/L).

In Silico Population Study of bla NDM-5 -Harboring IncX-3 Plasmids among E. coli Species
A genomes dataset was constructed with 28 complete genomes of E. coli-carrying bla NDM-5 on an IncX-3 plasmid and the genomes of B26 and B28 strains. In silico sequence types, phylotypes and metadata associated with the genomes are presented in Table 2. More than 96% of the genomes belong to A, B1 and C phylogroups. B26 and B28 were the only genomes of the B2 phylogroup in the dataset. The 31 genomes corresponded to strains isolated from diverse origins: humans (n = 20), environment (n = 6) and animals (n = 4). In order to study the distribution of E. coli encoding bla NDM-5 on IncX-3 plasmids within the whole E. coli populations, we reconstructed genetic links by goeBURST analysis. The dataset of 30 genomes was matched with 178 776 available genomes of E. coli (11 058 STs) in the dataset EnteroBase (13 July 2021) ( Figure 2). The genomes of E. coli-carrying bla NDM-5 IncX-3 plasmid spread out in 16 STs with no obvious lineage specialization. However, about half of the genomes (41.9%) belong to the CC10. This CC is a major sub-population in E. coli because it gathers 11.85% of the STs available in EnteroBase. The other genomes are scattered in the overall E. coli population structure. B26 and B28 belonged to ST636, which forms the CC636 together with 4 related STs. CC636 was relatively isolated in the E. coli population and gathers 451 strains (0.25%) of the 178,776 strains of EnteroBase. Each node corresponds to a Sequence Type (ST). The size of the node is scaled to the number of genomes of that ST. Nodes linked between them present one allele in common among the 7 genes considered in the MultiLocus Sequence Type scheme. Red nodes correspond to STs for which genomes with a blaNDM-5 encoding IncX-3 plasmid was identified; the proportion in the ST of genomes containing the plasmid is noted in parentheses.

Stability and Fitness Cost of pTsB26 on B26 and B28 E. Coli
To evaluate the stability of the pTsB26 plasmid in B26 and B28, strains were passaged daily for 40 days without antibiotic selection. pTsB26 is highly stable, with more than 96% of plasmid containing cells after approximately 400 generations ( Figure 3). Each node corresponds to a Sequence Type (ST). The size of the node is scaled to the number of genomes of that ST. Nodes linked between them present one allele in common among the 7 genes considered in the MultiLocus Sequence Type scheme. Red nodes correspond to STs for which genomes with a bla NDM-5 encoding IncX-3 plasmid was identified; the proportion in the ST of genomes containing the plasmid is noted in parentheses.

Stability and Fitness Cost of pTsB26 on B26 and B28 E. coli
To evaluate the stability of the pTsB26 plasmid in B26 and B28, strains were passaged daily for 40 days without antibiotic selection. pTsB26 is highly stable, with more than 96% of plasmid containing cells after approximately 400 generations (Figure 3).
These assays allowed one to isolate strains without pTsB26: B26∆pJ40 and B28∆pJ19, respectively, isolated at day 40 and day 19 of the experiment. In parallel, two strains containing pTsB26 plasmid (B26J40a and B28J19a) were isolated the same day as B26∆pJ40 and B28∆pJ19. These strains, used as control, were thereafter called "LB-adapted" strains, with the letter "a" at the end of the strain name.  These assays allowed one to isolate strains without pTsB26: B26ΔpJ40 and B28ΔpJ19, respectively, isolated at day 40 and day 19 of the experiment. In parallel, two strains containing pTsB26 plasmid (B26J40a and B28J19a) were isolated the same day as B26ΔpJ40 and B28ΔpJ19. These strains, used as control, were thereafter called "LB-adapted" strains, with the letter "a" at the end of the strain name.
Growth kinetics assays were performed for the strains with (ancestral strains and LBadapted strains) and without pTsB26 plasmid (Table 3). Growth rates of B26, B26ΔpJ40 and B26J40a did not vary significantly (p > 0.05), suggesting that the carriage of pTsB26 did not produce fitness lost for B26. Of note, B26ΔpJ19 and B26J19a contained the other plasmids IncFIA/FIB, IncFII and IncQ1 as the ancestral B26. On the other hand, growth rates of ancestral B28 or LB-adapted B28 (B28J19a) were equivalent and significantly better than observed for B28 cured for pTsB26 (p < 0.05). This last observation suggests that pTsB26 could provide fitness advantage even to its host, without selective pressure.

Discussion
The spread of NDM-5 variant has been extensively described in hospital environments, including hospital sewage water [32,33] and in wastewater [34]. In France, in 2020, about 20% of carbapenemase-producing enterobacteria isolated from clinical samples were NDM-producers [35]. Among them, more than 35% were NDM-5 variant. Its emergence and successful spread are worth emphasizing as this variant represented only 5% Growth kinetics assays were performed for the strains with (ancestral strains and LB-adapted strains) and without pTsB26 plasmid (Table 3). Growth rates of B26, B26∆pJ40 and B26J40a did not vary significantly (p > 0.05), suggesting that the carriage of pTsB26 did not produce fitness lost for B26. Of note, B26∆pJ19 and B26J19a contained the other plasmids IncFIA/FIB, IncFII and IncQ1 as the ancestral B26. On the other hand, growth rates of ancestral B28 or LB-adapted B28 (B28J19a) were equivalent and significantly better than observed for B28 cured for pTsB26 (p < 0.05). This last observation suggests that pTsB26 could provide fitness advantage even to its host, without selective pressure.

Discussion
The spread of NDM-5 variant has been extensively described in hospital environments, including hospital sewage water [32,33] and in wastewater [34]. In France, in 2020, about 20% of carbapenemase-producing enterobacteria isolated from clinical samples were NDMproducers [35]. Among them, more than 35% were NDM-5 variant. Its emergence and successful spread are worth emphasizing as this variant represented only 5% of NDMs reported in 2013, 15% in 2016 and more than 30% in 2017, with stabilization since this date [35].
Only 21 variations (19 SNPs and 2 multi-nucleotide polymorphisms) were detected between B26 and B28 isolated in the same urban aquatic environment a month apart. The scarce genomic differences strongly suggest their clonal origin. Clonal strains have certainly persisted for at least one month in the river, with iterative input being highly unlikely but not excluded. This hypothesis is strongly supported by the fact that the two isolates carried the same plasmid pTsB26 (100% nucleotide identity). Interestingly, the IncFIA/FIB, IncFII and IncQ1 plasmids carried by B26 (isolated in August 2015) were absent from the genome of B28 (isolated in September 2015). Only pTsB26 has persisted in the E. coli strain isolated in September, showing that pTsB26 is stable in B26 and B28 isolates, relative to other plasmids in the same strains. This in situ observation was verified by in vitro evolution experiments, demonstrating the longtime persistence of the plasmid after 400 generations in B26 and B28 isolates ( Figure 3). Several factors could explain the stability of pTsB26 in the cell lineage. First, like the other IncX-3 plasmids, pTsB26 harbors the widespread partitioning system ParAB. This system limits the number of segregant cells during cell division, ensuring the correct inheritance of the plasmid to the daughter cells [51]. Moreover, the conjugative traits of pTsB26 allow for infection of segregant cells and thus limit their number. Thus, harboring pTsB26 does not reduce the fitness of B26 and B28 (Table 3) and enhances the growth of B28. It is generally admitted that plasmids cause a fitness burden on their bacterial host [52][53][54][55] and that the plasmid could be quickly eliminated from its host. High stability of IncX-3 plasmids has already been reported in Enterobacterales transconjugant strains [28,56]. The absence of fitness cost of IncX-3 plasmids could be explained by their small size [25] and by the presence of transcriptional regulator H-NS like protein [28,30,[57][58][59][60]. The observed high stability of pTsB26 has potential due to its high conjugation frequency, to a low rate of segregational loss and due to the fact it does not pass a fitness cost onto the bacterial host.
The isolates B26 and B28 are the unique representants of ST636 in the studied dataset of E. coli-carrying bla NDM-5-encoding IncX-3 plasmid (Table 2). However, in their study, Kumwenda et al. reported two E. coli ST636 clinical isolates that carried bla NDM-5 onto a plasmid not affiliated to an incompatibility group [61]. EnteroBase reported 118 isolates belonging to ST636 (Figure 3). They were isolated in different countries (all continents are represented) and in environmental and clinical samples. Other studies reported the occurrence of ST636-producing ESBLs in clinical and environmental samples [62][63][64], suggesting a generalist trait for this ST.
Persistence of strains or STs carrying CEGs on self-transmissible plasmid such as pTsB26 in aquatic environment is of concern. Aquatic environment contains diverse autochtonous bacteria, including Enterobacterales (e.g., Enterobacter sp. [65] or Raoultella sp. [66]), which can exchange and receive IncX-3 plasmids [27,28,41]. These autochtonous bacteria can constitute an environmental reservoir and shuttles for bla NDM genes. Aquatic environments are strongly linked with anthropic activities, and during recreational activities, after flood episodes or by alimentation, humans can become exposed to bacteria from aquatic environments [7,[67][68][69][70]. Thus, if water contains carbapenemase-producingbacteria such as B26 or B28 or other autochtonous bacteria, and has acquired the plasmid by horizontal gene transfer, it represents a risk for human health (i) directly by causing antimicrobial-resistant bacterial infections [71] and (ii) indirectly by participating in the dissemination of bla NDM-5 on the occasion of gut colonization [71,72], or transit. These resistant bacteria can transfer the plasmid to host microbiota bacteria, making a "shuttle" between aquatic environment and humans [7,71].
Antibiotics 2022, 11, 196 8 of 14 We described for the first time the in situ persistence of a bla NDM-5 harboring IncX-3 self-conjugative plasmid in an E. coli lineage in an aquatic environment. This study underlines, once again, the importance of investigations into environmentally emerging, resistant bacteria. Beside genomics, testing genetic transfer and resistance stability by in vitro evolution is a proxy for diffusion and persistence in natural environment. In addition to the strategy proposed in this study, experiments of resistance genes transfer to waterborne autochthonous bacteria in microcosm would be interesting to conduct for a better description of the resistance reservoirs and the conditions influencing these reservoirs. On another hand, rapid alerts on environmentally emerging antimicrobial resistance are needed for rapid responses to resistance with public health concern. For this, efficient surveillance of AMR in environment should be undertaken. This is one of the challenges of the current national and international projects aiming to limit global AMR outbreak.

Escherichia coli Strains
E. coli strains B26 and B28 were isolated from water sampled at the same site of the urban river Font d'Aurelle in the city of Montpellier (N43.62711 E003.85316), France. They were isolated in August and September 2015, respectively [29].

Transformation and Conjugation Assays
Plasmid DNA extraction was done using the NucleoSpin Plasmid Kit (Macherey-Nagel, Allentown, PA, USA). Plasmid extracts were used for transformation assays using One Shot TOP10 chemically competent E. coli (Invitrogen, ThermoFisher Scientific, Paisley, UK) as recipient cell.
Conjugation experiments were performed using non-competent XL1-Blue E. coli MRF', a recipient strain resistant to tetracycline and sensitive to meropenem. Briefly, donor (B26, B28 and transformants TsB26 and TsB28) and recipient strains were grown overnight at 37 • C in Luria Bertani (LB) broth supplemented (donor strains) or not (recipient strain) with ertapenem (4 mg/L). Cells were washed from antibiotic and resuspended in LB broth, and each donor strain suspension was mixed (1:1 ratio) with the receptor strain. 200 µL of each mix was deposited onto nitrocellulose membrane, itself stuck on LB agar media and incubated at 37 • C during 24 h. Transconjugants were selected by plating the bacteria from the nitrocellulose membrane onto LB agar plates supplemented with ertapenem (4 mg/L) and tetracycline (12 mg/L). The conjugative frequencies were determined by calculating the transfer rate (ratio transconjugant/donor).
The presence of bla NDM and pir (encoding the IncX-3 plasmid-specific Pir protein) in selected transconjugants and transformants was assessed by specific PCRs [14,73].

Evaluation of Plasmid Stability
Strains B26 and B28 were grown overnight at 37 • C in 10 mL of LB broth supplemented with ertapenem (4 mg/L). Bacterial cells were washed from antibiotic by centrifugation, the pellet was resuspended in 1 mL of LB broth and 10 mL of fresh LB broth without antibiotic was spiked with 10 µL of the bacterial suspension and incubated 24 h at 37 • C in a shaking water bath. Serial passages of 10 µL of overnight culture to 10 mL of fresh LB broth were done daily. One passage corresponded approximatively to 10 generations of growth. Every 50 generations, samples were diluted and plated on LB agar plates. Then, 50 colonies from each lineage were screened on LB agar plates supplemented or not with ertapenem (4mg/L) to determine the fraction of plasmid-containing cells. The lack of plasmid was confirmed by the absence of bla NDM and pir genes with specific PCRs, and these strains (B26∆pJ40 and B28∆pJ19) were harvested for fitness cost assays. Parallelly, strains from the same generation, carrying pTsB26 plasmid, were harvested as controls (B26J40a and B28J19a). Experiments were done in triplicate.

Fitness Cost of Plasmid Carriage
Growth of strains carrying (B26, B28, B26J40a and B28J19a) or not (B26∆pJ40 and B28∆pJ19) pTsB26 plasmid were measured at 37 • C in LB broth without antibiotics using a CLARIOstar Plus microplate reader (BMG, Labtech). Every 15 min, the microplate was shacked at 200 rpm during 20 s, and optical density was measured at 600 nm. Growth rates were calculated according to Sandegren et al. [74].
4.6. In Silico Analysis 4.6.1. DNA Extraction and Whole-Genome Sequencing Genomic DNA of B26, B28, TsB26, TsB28 and E. coli TOP10 was extracted using the MasterPure™ purification kit (Lucigen, Middleton, WI, USA). High-throughput genome sequencing was carried out at the Plateforme de Microbiologie Mutualisée (P2M, Institut Pasteur, Paris, France). DNAs were processed for sequencing with Illumina systems (libraries using the Nextera XT DNA Library Prep kit and sequencing with the NextSeq 500 system). Paired-end reads were submitted to pre-processing using fqCleaner and to de novo assembly using SPAdes v3.12.0 [76] with k-mer lengths 21, 33, 55 and 77. The raw data and assemblies have been deposited in GenBank under the BioProject accession number PRJNA796954.

Plasmid Sequence and Annotation
The sequence of the plasmid pTsB26 was deduced from the alignment of the genomes of TsB26 and TsB28 with that of E. coli TOP10 using the ProgressiveMauve algorithm of Mauve software [79]. Identification of plasmid incompatibility group was done with PlasmidFinder [80]. Plasmid annotation was performed with Prokka v1.14.6 [81] with default parameters through the Galaxy platform (v4.6.0 + galaxy0) [82], and the plasmid sequence was manually curated; plasmid-specific genes were named according to Thomas et al. recommendations [83].

Distribution of IncX3-bla NDM Plasmids in E. coli Population
In order to select E. coli genomes carrying bla NDM-5 on an IncX3 plasmid, complete genomes available on NCBI database (6 April 2021) were investigated using BLASTn tool [84]. bla NDM-5 and pir gene of pTsB26 sequences were used, and genomes carrying the 2 genes (minimum homology 100% and 92%, respectively) on the same replicon were selected. Metadata associated with each genome were collected, and ST were determined in silico using the MLST 2.0 [89]. STs (Achtman MLST scheme) associated with selected genomes and B26 and B28 strains were compared to general E. coli population using EnteroBase database [90] (178,776 strains the 13 July 2021) and the goesBURST algorithm (single locus variant level) of PHYLOViZ 2.0 software [91]. Plasmid homologies were determined by BLASTn analysis [84] against the NCBI nr/nt database with default parameters.

Statistical Assays
Statistical analysis of growth rates was performed using Student's t-test. All statistics were made using the GraphPad Prism software V 5.03. Test results were considered as statistically significant when the associated p-value was less than 0.05.