Multidrug-resistant enterobacteriaceae in coastal water: an emerging threat

Background The environmental role of carbapenemase-producing Enterobacteriaceae (CPE) acquisition and infection in human disease has been described but not thoroughly investigated. We aimed to assess the occurrence of CPE in nearshore aquatic bodies. Methods Enterobacteriaceae were cultured from coastal and estuary water near Netanya, Israel in June and July of 2018. Bacteria were identified by VITEK2® and their antimicrobial susceptibility was tested according to the CLSI guidelines. Enterobacteriaceae genomes were sequenced to elucidate their resistome and carbapenemase types. Results Among other clinically relevant bacteria, four CPE (three Enterobacter spp and one Escherichia coli isolate) were isolated from two river estuaries (Poleg and Alexander Rivers) and coastal water at a popular recreational beach (Beit Yanai). Molecular analysis and genome sequencing revealed the persistent presence of rare beta-lactamase resistance genes, including blaIMI-2 and a previously unknown blaIMI-20 allele, which were not found among the local epidemiological strains. Genome comparisons revealed the high identity of riverine and marine CPE that were cultivated one month apart. Conclusions We show that CPE contamination was widespread in nearshore marine and riverine habitats. The high genome-level similarity of riverine and marine CPEs, isolated one month apart, hints at the common source of infection. We discuss the clinical implications of these findings and stress the urgent need to assess the role of the aquatic environment in CPE epidemiology.


Background
Carbapenem-resistant Enterobacteriaceae (CRE), in particular, carbapenemase-and extended-spectrum β-lactamase-producing Enterobacteriaceae (CPE and ESBL-PE) endanger global health, as they have spread worldwide during the last two decades. The risk factors for the acquisition of these multidrug-resistant organisms (MDROs) are usually associated with healthcare [1] and overseas travel [2]. Potential modes of environmental transmission of these bacteria to humans outside of healthcare facilities following exposure to wildlife, livestock and pet animals have been reported [3]. The environmental routes of CRE infection may play a role in causing human disease and spreading it globally, however, they have not been thoroughly investigated yet. In particular, little is known about the epidemiology of CRE in the aquatic environment.
CRE may reach aquatic bodies as a result of organic contamination from multiple sources [21], including hospital effluents [22][23][24], wastewater treatment plants (WWTPs) [24][25][26], discharge of livestock farms and agriculture [3,27], seepage water [28] as well as others [21]. Once mixed with the aquatic body, these effluents may introduce not only foreign MDROs, but also high doses of antibiotics, which likely trigger resistance propagation [29]. In Israel, the main sources of aquatic contamination comprise of regional councils that are not connected to WWTPs, controlled discharges from fish farms and WWTPs [30] and untreated sewage from the Palestinian Authority [31]. Israeli governmental programs that monitor the quality of coastal seawater and rivers report coliform concentrations, but not the presence of MDROs. MDRO infection, however, poses a real threat to those engaged in recreational water activities.
A young patient who suffered a near-drowning experience in one of Netanya's beaches was admitted to the Laniado medical center, Netanya, Israel, in June 2014. Bacterial screening upon his admission to the intensive care unit found he was a carrier of two CPE species: Enterobacter cloacae and Klebsiella oxytoca, both carrying bla KPC . We assumed that seawater ingestion and aspiration led to this infection, as no other risk factors could explain the high rates of CPE colonization. We hypothesized that these CPEs originated from rivers supplying water to the shoreline in the vicinity of Netanya. To test this hypothesis, we aimed to identify CPE in coastal waters and two river estuaries in the Sharon district.

Sampling and bacterial isolation
On two occasions (June, July 2018), water samples were collected with sterile-glass containers (pre-combusted at 500 °C) and delivered to the laboratory at room temperature. 200-1300 ml of freshwater or seawater was filtered, using a vacuum pump, through a 0.2 µm sterile cellulose acetate membrane (Sartorius Stedim ® ). Distinct volumes of water were filtered at each site (Table 1), due to differences in turbidity and membrane clogging. Control membranes were obtained by filtering room air for the same duration. The membranes were soaked in an enrichment medium (thioglycolate broth (Novamed ® )), that served as the source of cultures: first after overnight incubation at room temperature, and after 24 h of incubation at 37 °C. The media were cultured on TSBA + Blood/Chocolate agar, MacConkey agar, CNA agar, Chromagar Orientation, Chromagar MSupercarba and Streptococci Select medium plates (all from HyLabs ® ). Colonies were picked, isolated and then identified using VITEK2 ® (bioMérieux). Antimicrobial susceptibility testing was performed by VITEK2 ® (bioMérieux) and Etest (bioMérieux) on Mueller-Hinton agar according to the Clinical & Laboratory Standards Institute (CLSI) guidelines. Enterobacteriaceae growing on CRE plates were subjected to the modified Hodge test or CARBA NP hydrolysis. Following species isolation and identification, the carbapenemase type was identified using Cepheid ® GeneXpert Carba-R for bla KPC , bla NDM , bla VIM , bla IMP , bla OXA-48 genes. The bla IMI was identified using the polymerase chain reaction (PCR) [32].
We repeated the sampling from the same sites 3 weeks later. During the second sampling session, we mainly attempted to identify CREs, and used 60 ml of sterile media as a negative control.

Genome sequencing, assembly and bioinformatics
DNA was extracted from three Enterobacter isolates: nEC133, nEC134 and nEC135 using the Presto Mini gDNA Bacteria kit (GeneAid). DNA libraries were prepared using the Illumina Nextera XT kit (Illumina). Sequencing was performed using the Illumina MiSeq platform with 2 × 250 bp paired-end reads aiming at > 100 × coverage, MiSeq Reagent Kit v2.

Bioinformatic analyses
Sequencing was performed using the Illumina MiSeq platform with 2 × 250 bp paired-end reads aiming at > 100 × coverage, MiSeq Reagent Kit v2. Genomes were de-novo assembled with SPAdes V3.12 [33] and their quality was estimated with checkM [34]. The whole-genome multilocus sequence typing (wgMLST [35]) for Enterobacter cloacae complex was performed using the BioNumerics version 7.6.3 cloud-based calculation engine. wgMLST alleles were determined using assembly-free and assembly-based approaches. The resistome was identified using a standalone Resistance Gene Identifier (RGI) application with the Comprehensive Antibiotic Resistance Database (CARD [36]). Plasmid sequences were predicted with plasFlow [37] because we were unable to assemble complete plasmid sequences with PlasmidSPAdes [38]. Average nucleotide identity (ANI) was calculated as orthoANIu [39] using EzBioCloud [40]. The maximum likelihood tree of BlaIMI amino acid sequences was constructed with MEGA7 [41], based on the LG model [42]. Genomic sequences are available under the NCBI Bio-Project accession number PRJNA578038. The bla IMI-20 sequence was submitted to GeneBank with the accession number MN619794.

Water-borne bacterial isolates
Enterobacteriaceae, non-fermenter bacilli and Grampositive cocci were isolated from the coastal aquatic environment (Table 1). Four strains of CPE were found in the two sampling sessions: three isolates of CP-Enterobacter spp and CP-Escherichia coli bla OXA-48 . The three Enterobacter isolates, which were isolated from seawater and freshwater, had remarkably similar antibiogram phenotype, different from that of E. coli bla OXA-48 ( Table 2).

Resistome of aquatic Enterobacter revealed by genome analysis
We sequenced and assembled three high-quality genomes (completeness > 99%, contamination < 1%) of the aquatic Enterobacter isolates. nEC134 and nEC135 genomes were highly similar to each other, based both on the 99.9% ANI and 5 out of 15,612 allele differences on wgMLST analysis. These genomes were classified as E. bugandensis (98.7% ANI against the GenBank genomes). nEC133 genome was more diverged when compared to nEC134 and nEC135 genomes (91.5-91.6% ANI, 2600 distinct alleles), and was classified as E. asburiae (97.5% ANI against the GenBank genomes). The genomederived resistomes of all the three isolates were similar and included IMI, ACT and ampC-type beta-lactamases, as well as numerous genes that encode components of antibiotic efflux pumps, and other potential antibiotic resistance mechanisms (Table 3). E. asburiae nEC133 bla IMI was classified as IMI-2, based on 100% sequence identity (Fig. 2, Additional file 1: Fig. 1). E. bugandensis nEC134 and nEC135 carried a previously undescribed IMI-20 allele, whose sequence was distinguished from that of IMI-2 by a single nucleotide polymorphism (SNP), resulting in cysteine to phenylalanine substitution (Fig. 2, Additional file 1: Fig. 1). As in other Enterobacteriaceae, imiR gene that encodes a LysR-type regulator was found upstream of the bla IMI gene [8,43]. We were not able to assemble complete plasmids using short reads, and only 3943-6219 bp long scaffolds included the bla IMI and imiR genes. However, plasFlow analysis suggested these scaffolds may be plasmid-related. The total length of scaffolds that were assigned to proteobacterial plasmids in E. asburiae nEC133 genome was 74,645 bp, which is comparable to that of the plasmid p3442-IMI-2 (78,374 bp). In E. bugandensis nEC134 and nEC135 genomes, the cumulative length of plasmid sequences was 142,368-176,174 bp, hinting at the presence of larger, or more than one plasmid. Genome assembly graphs of these isolates revealed a close linkage between the bla IMI-20 and plasmid scaffolds that carry the traXIDTHBFNCU-WICVBKEL genes of the IncF conjugal transfer system, providing further evidence that bla IMI-20 is encoded on an IncF plasmid (Additional file 2: Fig. 2).

Discussion
Our results suggest that CPE, as well as other MDR pathogens, occur in the coastal waters of a popular public recreational beach offshore Netanya, Israel and nearby river estuaries. These CPE appear to be abundant and well-acclimated to freshwater and seawater, as they were easy to isolate and at least on one occasion, the same E. bugandensis was found twice, in distant sampling sites, one month apart. It is important to state that at the time of our sampling (June and July 2018) no considerable fluctuations in seawater quality were reported [44]. Residents and tourists visiting popular recreation and leisure sites may be exposed to these MDROs. The fact that isolates with highly similar genomes were isolated from both the riverine and marine environments, implies a mutual source, however, its origin is still unclear. Rivers are putative reservoirs and sources of MDROs, which were shown to be virulent [16,45] and MDROs are less frequently encountered in patients drowning in seawater [46]. The pollution sources of Poleg River include discharge of urban sewage [47,48], authorized dumping of treated sewage water [49] and cattle herding as seen on site. One of the main sources of Alexander River is the Nablus River, running from Nablus city through Tul-Karem located in the West Bank (Fig. 1). Other possible contamination sources include effluents from the adjacent WWTPs, a sea-turtles rescue center, an algae plant, an agricultural catchment basin and reservoir waters [30,50]. Becker et al. (2013) reported that since 1995 several polluting sources were successfully treated, yet Wadi Zeimar, a major contamination source concentrating pollutants from Tul-Karem to Nablus River, remained [31]. The polluted rivers are the likely sources of marine contamination. Alternatively, coastal water can migrate up to e few kilometers inland, as shown by the daily salinity profiles measured by Ruppin Estuarine and Coastal Observatory [50], potentially reaching the possible contamination sources. Despite the limited mixing of water masses, bacterial cross-contamination may be substantial, hence bidirectional contamination is feasible.
Gut microbes may be adapted to the aquatic environment, in which their survival rates are poorly understood. Yet, cultivation-based studies demonstrate that common gut bacteria, such as Enterobacteriaceae, are frequently detected in freshwater [8,10,11,15,51,52] and seawater [5][6][7]. These bacteria may be able to cope with different salinities because the human gut environment is characterized by spatial and temporal heterogeneity in osmolarity, based on the kinds of meals consumed [53]. The fitness of microbes and their growth rates depends not only on osmolarity but also on taxon-specific physiology and additional external factors such as nutrient abundance, pH and oxygen levels [54]. The three sequenced CP Eneterobacter spp. genomes encoded proteins that are involved in halotolerance, including the osmosensitive K + channel histidine kinase KdpD, NhaA type Na + / H + antiporter DNA-binding protein H-NS [55]. These sequences were found in scaffolds that were > 400 kbp in length and were defined as chromosomal by plasFlow. This indicates that at least some of the salt tolerancerelated traits in these strains are not linked to plasmids as in other bacteria, and therefore not to bla IMI genes, which most likely are encoded on plasmids. Salinity has been recently shown to be the most important factor modulating the distribution patterns of antibiotic resistance genes in oceans and river beach soils [56], although other studies of aquatic ecosystems failed to show this [57]. Thus, it is feasible that less-studied traits that mediate salinity tolerance may be linked to antibiotic resistance genes. Most importantly, these aquatic isolates are only remotely associated with the local clinical epidemiology. CP Enterobacter spp., are uncommon in our hospital's clinical settings: between August 2013 and February 2019, we identified 129 Enterobacter spp. out of 798 CRE rectal screening isolates (16%). Carbapenemase-producing Enterobacter spp. were detected in 65 isolates. bla IMI genes were found only in three of these CPE isolates, while bla KPC (50 isolates) and bla NDM (10 isolates) were more common. However, the IMI mechanism can often go undetected, because only the five major enzymes OXA-48 has also not been frequently encountered during this period in our facility (19 isolates out of 798 CREs, mostly seen in E. coli spp). Nevertheless, IMI carbapenemases appear to be emerging in clinical practice [58] as well as causing nosocomial outbreaks [59].

Conclusions
The widespread occurrence of CPE contamination in popular recreational beaches of the Sharon district is alarming and has major ramifications for environmental and public health, as well as to the public perceptions and awareness. However, occupation-or recreation-related seawater or river exposure is currently not considered a risk factor for CPE contamination, as it should probably be. It is important to consider MDROs (including CPE) when providing empiric antibiotic therapy for aspiration pneumonia during near-drowning as well as other injuries occurring during water-related activities, such as surfing and fishing.
Future large scale studies that will investigate the roles of the contamination sources along the river route to the sea, as well as resistome-related monitoring programs based on the frequent sampling of water and sediments, are crucial to mitigate aquatic CPE infection risks.
Additional file 1. Figure 1: Alignment of IMI type carbapenem-hydrolyzing class A beta-lactamase amino acid sequences (see Figure 2 in the main text for accession numbers). Names of novel IMI alleles are marked in blue, cysteine to phenylalanine substitution is shown in the frame.
Additional file 2. Figure 2: The link between bla MI-20 and traXIDTHBFN-CUWICVBKEL genes of the IncF plasmid, based on the simplified de Bruijn assembly graph of the E. bugandensis nEC134 genome, generated by the SPAdes assembler and viewed in Bandage (the whole graph in the top panel and zoom in into the paths that contain the target genes in the bottom panel). The target genes are shown (bla IMI-20 -red, lysR-yellow, traXIDTHBFNCUWICVBKEL-shades of blue).

Fig. 2
Phylogenetic tree of IMI type carbapenem-hydrolyzing class A beta-lactamases (alignment of 21 sequences of 292 amino acids). The tree is drawn to scale, with branch lengths representing the number of substitutions per site. The percentage of trees in which the associated taxa clustered together was determined based on 100 bootstrap resamples. Sequences from this study are marked in blue