Expression of a Shiga-Like Toxin during Plastic Colonization by Two Multidrug-Resistant Bacteria, Aeromonas hydrophila RIT668 and Citrobacter freundii RIT669, Isolated from Endangered Turtles (Clemmys guttata)

Aeromonas hydrophila RIT668 and Citrobacter freundii RIT669 were isolated from endangered spotted turtles (Clemmys guttata). Whole-genome sequencing, annotation and phylogenetic analyses of the genomes revealed that the closest relative of RIT668 is A. hydrophila ATCC 7966 and Citrobacter portucalensis A60 for RIT669. Resistome analysis showed that A. hydrophila and C. freundii harbor six and 19 different antibiotic resistance genes, respectively. Both bacteria colonize polyethylene and polypropylene, which are common plastics, found in the environment and are used to fabricate medical devices. The expression of six biofilm-related genes—biofilm peroxide resistance protein (bsmA), biofilm formation regulatory protein subunit R (bssR), biofilm formation regulatory protein subunit S (bssS), biofilm formation regulator (hmsP), toxin-antitoxin biofilm protein (tabA) and transcriptional activator of curli operon (csgD)—and two virulence factors—Vi antigen-related gene (viaB) and Shiga-like toxin (slt-II)—was investigated by RT-PCR. A. hydrophila displayed a > 2-fold increase in slt-II expression in cells adhering to both polymers, C. freundii adhering on polyethylene displayed a > 2-fold, and on polypropylene a > 6-fold upregulation of slt-II. Thus, the two new isolates are potential pathogens owing to their drug resistance, surface colonization and upregulation of a slt-II-type diarrheal toxin on polymer surfaces.


Introduction
Antibiotic resistance is an increasing crisis as both the range of resistance in clinical settings expands and the pipeline for development of new antibiotics contracts [1]. This problem is compounded by the global genomic scope of the antibiotic resistome, so that antibiotic resistance spans a continuum from genes in clinical pathogens to those of benign environmental microbes along with their proto-resistance gene progenitors [2,3]. Further, increased resistance to antimicrobial agents occurs in biofilms [4]. Biofilm-associated cells differ from their suspended counterparts by generation of an extracellular polymeric substance (EPS) matrix, reduced growth rates, and the up/downregulation of specific genes [5]. Plasmid conjugation occurs at a greater rate between cells in biofilms than

Bacterial Isolation
Microbial samples were isolated from 12 adult rescued infected spotted turtles (Clemmys guttata) seized by the United States Fish and Wildlife Service from an illegal reptile trade operation (chain of custody ID number-ST#032797). The spotted turtle is a small, semi-aquatic, North American species commonly targeted and illegally harvested for sale in the pet trade and overseas for other uses [99,100]. The eyes, nostrils and limbs of turtles were swabbed on to agar plates and the samples were initially subjected to biochemical assays. For subsequent analysis, the two strains identified as Aeromonas hydrophila RIT 668 and Citrobacter freundii RIT 669 (based on 16S rDNA sequences) were routinely cultured on blood agar plates (5% sheep blood) and MacConkey plates respectively (BD BBL™, prepared media, 100 mm × 15 mm, San Diego, CA, USA). Hemolysis on the blood plates was examined by observing the presence of complete lysis around the colonies and a clearing on the medium. All twelve turtles (Clemmys guttata) were infected, lethargic and had reduced/stopped food intake for around a week. The eyes, nostrils and feet of the spotted turtles were surrounded by a slimy substance, and the eyes were inflamed. All the turtles were infected by A. hydrophila and C. freundii.

Characterization and Identification: Biochemical Assay and 16S rDNA Amplification
Primary identification was achieved via Gram staining. Oxidase and catalase tests were performed, followed by five groups of biochemical assays for microbial identification: Group 1-glucose, gas, and lysine; Group 2-ornithine, H 2 S, and indole; Group 3-adonitol, lactose, and arabinose; Group 4-sorbitol, Voges Proskauer, and dulcitol; and Group 5-phenylalanine (PA), urea, and citrate. Of the four different microbes identified, two strains were further analyzed and identified via 16S rDNA PCR amplification. For the 16S rDNA amplification, the microbial DNA was isolated using the 'UltraClean Microbial DNA Isolation Kit' (MO BIO Laboratories Inc., San Diego, CA, USA). Colony PCR was performed to obtain~500 base pair amplicons of the 16S V3/V4 regions of the rRNA gene using the following conditions: 1 cycle at 95 • C for 2 min, followed by 30 cycles at 95 • C for 30 s, 52 • C for 30 s, and 72 • C for 3 min. The forward and reverse primers used for this was 5 -CCTACGGGNGGCWCGAG-3 and 5 -GACTACHVGGGTATCTAATCC-3 . The amplicons were separated by gel electrophoresis, followed by commercial Sanger nucleotide sequencing (GeneWiz LLC, South Plainfield, NJ, USA) using the V3/V4 forward primer.

Genomic DNA Isolation
Genomic DNA was extracted from mid-log phase of Aeromonas hydrophila RIT668 and Citrobacter fruendii RIT669 using the GenElute Bacterial Genomic DNA kit, as per the manufacturer's protocol (Sigma-Aldrich, St. Louis, MO, USA), and quantified using a Nanodrop One spectrophotometer.

Agarose Gel Electrophoreses
The DNA fragments from PCR and RT-PCR experiments were run on 1% agarose gels submerged in 1X TAE buffer, and stained with ethidium bromide, followed by visualization with a UV light source. A 1 kb ladder was used as the DNA standard (New England Biolabs, Ipswich, MA, USA).

Whole-Genome Sequencing, Assembly and Annotation
Using the Nextera XT library prep kit (Illumina) and Nextera XT index kit (Illumina), 1 ng of genomic DNA was processed to generate a sequencing-ready library, as per the manufacturer's protocol. After library prep, 5 µL tagmented and indexed library DNA was quantified via the PicoGreen broad range assay on a Biotek Synergy H1 plate reader. To determine the fragment size distribution of DNA in the tagmented and indexed library, 1 µL of DNA was run using the DNA 1000 chip kit on an Agilent 2100 bioanalyzer. Based on the DNA concentration and size distribution, the library was manually normalized to a concentration of 4 nM in molecular-grade water. Of the resulting 4 nM library, 5 µL was then pooled with other tagmented, indexed, and normalized libraries. The pooled libraries were denatured and diluted to a loading concentration of 12 pM following the manufacturer's protocol for manually normalized libraries. The pooled, denatured and diluted 12 pM library was sequenced using the MiSeq Reagent Kit V3 on the Illumina MiSeq for 2 × 151 cycles at the Rochester Institute of Technology Genomics Facility. Adapter trimming was performed automatically on the Illumina MiSeq during FASTQ generation. Trimmed reads were uploaded to the Galaxy web platform, and assembled de novo at the public server at usegalaxy.org [101], using Unicycler version 0.4.6.0 [102], with a minimum contig length of 200 bp.

Phylogenetic Analysis
The assembled FASTA contig files for A. hydrophila RIT668 and C. freundii RIT669 were uploaded to the Type Strain Genome Server (TYGS), found at https://tygs.dsmz.de. TYGS is a freely available tool for creating taxonomic assignments based on whole-genome sequence data [103]. The ten closest strains in the TYGS database to the query assembly were determined using Mash, a whole-genome clustering method [104]. Additionally, another ten closest relatives to the query assembly were selected by BLAST comparison of the 16S rDNA sequences extracted from the query using RNAmmer [105] against all 10,342 type strains in the TYGS database [106], selecting the top 50 hits, and calculating the Genome BLAST Distance Phylogeny distance (GBDP) with the query genome to determine the closest 10 type strains [107].

Resistance Gene Identifier (RGI)
The antibiotic resistance genes in both the model organisms were identified using the Resistance Gene Identifier (RGI) tool for understanding the mechanisms of antimicrobial resistance. The AMR gene family was identified along with the drug class, percentage identify of matching region and length of reference sequence. The genomes of A. hydrophila and C. freundii were analyzed using the bioinformatics platform RGI CARD. FASTA files were uploaded to RGI CARD software server RGI 5.1.0, CARD 3.0.4 [108] using the following parameters: Perfect, Strict and Loose hits, and complete genes only. The Perfect algorithm is most useful for clinical surveillance as it detects perfect matches to the curated reference sequences and mutations in the CARD. The Strict algorithm is aimed at unearthing previously unknown variants of known antibiotic resistance genes, including secondary screen for key mutations.

Predictions of Secondary Metabolite Production
The assembled genome sequence of A. hydrophila RIT668 and C. freundii RIT669 were analyzed using the Antibiotics and Secondary Metabolite Analysis Shell (antiSMASH5.0) webserver [109]. This tool identifies the biosynthetic loci covering the whole range of known secondary metabolite compound classes. It aligns the identified regions at the gene cluster levels to the closest match from a database, which contains all other known gene clusters and integrates the previously available secondary metabolite-related genes.

Classical Biofilm Forms
The colonization of polyethylene (PE) and polypropylene (PP) plastics was studied using the two model organisms. High-density 1.6 mm PE and PP sheets were laser cut to 1 inch squares and sterilized with alcohol followed by UV radiation for 3 h. Sterilized squares were placed aseptically in blood agar plates and MacConkey plates, and inoculated with 5 mL each of overnight culture of A. hydrophila and C. freundii, respectively, for biofilm formation. The plates were incubated at 37 • C for 6 weeks with a steady supply of 3 mL of liquid bacterial cultures and analyzed for plastic biofilm colonization via scanning electron microscopy (SEM). This set up in solid agar media is referred to as 'biofilms' throughout the study and was used only for SEM comparisons.

Planktonic and Adherent Forms
A single plastic PE or PP square was placed in each well in a six-well dish with 5 mL of C. freundii or A. hydrophila culture with an OD 600 of 0.5. Planktonic cells were aspirated out every other day in order to replace the well with fresh TSB depending on the dish. Dishes were incubated at 37 • C. Controls were initially placed alongside cultures in a six-well dish. To allow for maximum possible plastic colonization, except for aspiration of old TSB media and addition of new media, the six-well dish was left undisturbed for the duration of 6 weeks. This set up in liquid media is referred to as 'planktonic' throughout the study and the adherent cells as 'adherent'.

RNA Isolation
RNA was isolated from the A. hydrophila and C. freundii samples using the Omega E.Z.N.A. bacterial RNA isolation kit as per the manufacturers' protocol (Omega Bio-tek Inc., Norcross, GA, USA). The RNA integrity was analyzed NanoDrop followed by running the samples in a 1.5% agarose gel alongside a 1 Kb ladder (Invitrogen, Carlsbad, CA, USA). An array of nine different annealing temperature were tested on eight different primers in order to elucidate and optimize the temperatures for each of the samples.

Detection of Biofilm-Related Genes and Virulence Factors by RT-PCR
RT-PCR was performed based on the primer sequences and melting temperatures listed in Table 4. The genes encoding the following proteins were analyzed (a) Biofilm-related BsmA (biofilm peroxide resistance protein), BssR (biofilm formation regulatory protein), BssS (biofilm formation regulatory protein), HmsP (biofilm formation regulator), TabA (toxin-antitoxin biofilm protein) and CsgD (transcriptional activator curli operon). (b) Virulence-related: ViaB (primers VIAB-1 and VIAB-2, expected product size 516 bp; [110]), and Shiga-like toxin II (SLT-II) using the published primers GK1 and GK4, designed to amplify slt-II genes with an expected product size of 1260 bp [81,111]. A reverse transcriptase system (2 step) with cDNA synthesis (Promega Corp., Madison, WI, USA) was used for RT-PCR followed by the thermal cycle program. Since nine different annealing temperatures were tested in order to optimize the temperatures for each of the samples and each primer, the thermal cycle program varied for the primers and samples used in this study. A sample thermal cycle program involved initial denaturation at 95 • C for 3 min ×denaturation, annealing and extension at varying temperatures for 1 min (35 cycles) and 72 • C for 1 min respectively, followed by a final extension at 72 • C for 5 min and a final hold at 4 • C.

Estimation of Gene Expression
The change in gene expression was analyzed by running electrophoresis gels with sample concentrations normalized. Then, each set of bands corresponding to amplified fragments was analyzed for intensity and peak area with background subtraction, using the scientific image analysis open platform ImageJ [112]. The values were then plotted in Microsoft Excel as bar graphs showing fold change normalized to the 16S gene values, in the form of bar graphs. The 16S gene was used as the housekeeping standard during the RT-PCR and its values were used in this analysis.

Scanning Electron Microscopy Analysis
The biofilm-covered PE and PP squares were gently rinsed with phosphate-buffered saline (PBS) buffer at pH 7.4 and 2% glutaraldehyde to fix the cells and then incubated at 25 • C for 90 min. The samples were dehydrated with ethanol, incubated, and rinsed successively (50% for ten minutes, 70% for ten minutes, 80% for ten minutes, 95% twice for ten minutes, and 100% three times for fifteen minutes) inside a Petri dish. All the liquid was removed and the Petri dish containing the samples was covered in parafilm and stored at 25 • C until further analysis. Samples were covered with gold-palladium for two minutes with an SPI sputter coater to mitigate charging in the electron beam. The SEM was performed at 5 kV using a Mira3Tescan field emission SEM at the Rochester Institute of Technology (RIT) Nano-Imaging Lab.

Biochemical Characterization and Taxonomy
When the research was begun, there was 50% mortality, in spite of following appropriate animal care. By the end of the study period, however, all twelve turtles died (mortality rate 100%). Microbial isolates from rescued spotted turtles (Clemmys guttata) were identified as A. hydrophila and C. freundii, and were Gram-negative, beta hemolytic, lactose fermenting, and potential opportunistic pathogens. The two strains were initially identified through the 16S rDNA sequence coupled with NCBI-BLAST searches. Owing to their possible pathogenicity, the genomes were sequenced and annotated; a summary of the genome characteristics is shown in Table 1. Taxonomy based on whole-genome comparisons of the ten nearest relatives using the TYGS tool [103] confirmed the identification of RIT668 as A. hydrophila and RIT669 as Citrobacter. The nearest relative of A. hydrophila RIT668 is A. hydrophila ATCC 7966 ( Figure 1). C. freundii RIT669 is the closest relative of Citrobacter portucalensis A60 ( Figure 2). palladium for two minutes with an SPI sputter coater to mitigate charging in the electron beam. The SEM was performed at 5 kV using a Mira3Tescan field emission SEM at the Rochester Institute of Technology (RIT) Nano-Imaging Lab.

Biochemical Characterization and Taxonomy
When the research was begun, there was 50% mortality, in spite of following appropriate animal care. By the end of the study period, however, all twelve turtles died (mortality rate 100%). Microbial isolates from rescued spotted turtles (Clemmys guttata) were identified as A. hydrophila and C. freundii, and were Gram-negative, beta hemolytic, lactose fermenting, and potential opportunistic pathogens. The two strains were initially identified through the 16S rDNA sequence coupled with NCBI-BLAST searches. Owing to their possible pathogenicity, the genomes were sequenced and annotated; a summary of the genome characteristics is shown in Table 1. Taxonomy based on whole-genome comparisons of the ten nearest relatives using the TYGS tool [103]confirmed the identification of RIT668 as A. hydrophila and RIT669 as Citrobacter. The nearest relative of A. hydrophila RIT668 is A. hydrophila ATCC 7966 ( Figure 1). C. freundii RIT669 is the closest relative of Citrobacter portucalensis A60 ( Figure 2).

Resistome Analysis and Secondary Metabolite Analysis
According to the Resistance Gene Identifier (RGI) analysis of its genome, A. hydrophila RIT668 is potentially resistant to the carbapenem, penem, cephalosporin, fluroquinolone, tetracycline and elfamycin antibiotic classes ( Table 2). It has homologs with 43-99.7% sequence identities to known examples of multidrug, unclassified, MLS (macrolide, lincosamide, streptogramin), aminoglycoside, β-lactam, bacitracin and glycopeptide resistance genes, with putative efflux pumps present. C. freundii RIT669 potentially resists up to 19 different classes of antibiotics, including carbapenem, penem, cephalosporin, fluroquinolone, tetracycline, peptides and elfamycin; mutilple efflux pumps are present (Table 3). Notably, it may also be resistant to classes such as fosfomycin, rhodamine, triclosan and benzalkonium chloride (the last two are widely used in consumer and sanitizer products). The closest relative C. portucalensis A60 is also aquatic and harbors resistance to β-lactams and quinolones [113].

Resistome Analysis and Secondary Metabolite Analysis
According to the Resistance Gene Identifier (RGI) analysis of its genome, A. hydrophila RIT668 is potentially resistant to the carbapenem, penem, cephalosporin, fluroquinolone, tetracycline and elfamycin antibiotic classes ( Table 2). It has homologs with 43-99.7% sequence identities to known examples of multidrug, unclassified, MLS (macrolide, lincosamide, streptogramin), aminoglycoside, β-lactam, bacitracin and glycopeptide resistance genes, with putative efflux pumps present. C. freundii RIT669 potentially resists up to 19 different classes of antibiotics, including carbapenem, penem, cephalosporin, fluroquinolone, tetracycline, peptides and elfamycin; mutilple efflux pumps are present (Table 3). Notably, it may also be resistant to classes such as fosfomycin, rhodamine, triclosan and benzalkonium chloride (the last two are widely used in consumer and sanitizer products). The closest relative C. portucalensis A60 is also aquatic and harbors resistance to β-lactams and quinolones [113].

Polymer Adhesion and Gene Expression
This study examined the expression of six biofilm-related genes as well as slt-II and viaB, based on the primers listed in Table 4, which were derived from earlier studies [81,85,114,115]. The genes bsmA, bssR, bssS and csgD were reported to contribute to planktonic and biofilm growth in Citrobacter werkmanii BF6 [114]. The genes tabA and hmsP were identified as biofilm related in a subsequent study on the same organism [115]. The gene bsmA (biofilm stress and motility) or biofilm peroxide resistance protein, known as yjfO in E. coli, is upregulated during biofilm growth [116] Mutant analysis revealed roles in microcolony formation, flagellar motility as well as resistance to acid and peroxide stresses [116]. It has been suggested that bsmA is involved in controlling cell aggregation at specific points in biofilm development [116,117]. The genes bssR and bssS (regulator of biofilm through signal secretion), also called yliH and yceP in E. coli K-12, were shown to repress motility by mediating cell signaling [118]. Both bssR and bssS were postulated to be global regulators of several genes involved in catabolite repression, stress response, quorum sensing and the putative stationary-phase signal [118]. A. hydrophila samples did not amplify PCR products corresponding to bsmA and bssR. However, planktonic cells showed the amplification of bssS, while cells adhering on PE did not express bssS and adherent cells on PP appeared to repress bssS expression ( Figure 3). C. freundii expressed bsmA, bssR and bssS; most were slightly upregulated in adherent cells on both polymers, but bssR was upregulated over 3.5-fold in the PP-adherent cells (Figure 4).

Polymer Adhesion and Gene Expression
This study examined the expression of six biofilm-related genes as well as slt-II and viaB, based on the primers listed in Table 4, which were derived from earlier studies [81,85,114,115]. The genes bsmA, bssR, bssS and csgD were reported to contribute to planktonic and biofilm growth in Citrobacter werkmanii BF6 [114]. The genes tabA and hmsP were identified as biofilm related in a subsequent study on the same organism [115]. The gene bsmA (biofilm stress and motility) or biofilm peroxide resistance protein, known as yjfO in E. coli, is upregulated during biofilm growth [116] Mutant analysis revealed roles in microcolony formation, flagellar motility as well as resistance to acid and peroxide stresses [116]. It has been suggested that bsmA is involved in controlling cell aggregation at specific points in biofilm development [116,117]. The genes bssR and bssS (regulator of biofilm through signal secretion), also called yliH and yceP in E. coli K-12, were shown to repress motility by mediating cell signaling [118]. Both bssR and bssS were postulated to be global regulators of several genes involved in catabolite repression, stress response, quorum sensing and the putative stationaryphase signal [118]. A. hydrophila samples did not amplify PCR products corresponding to bsmA and bssR. However, planktonic cells showed the amplification of bssS, while cells adhering on PE did not express bssS and adherent cells on PP appeared to repress bssS expression ( Figure 3). C. freundii expressed bsmA, bssR and bssS; most were slightly upregulated in adherent cells on both polymers, but bssR was upregulated over 3.5-fold in the PP-adherent cells (Figure 4).   The gene hmsP (biofilm formation regulator) is one of two genes controlling the amount of biofilm produced in the pathogen Yersinia pestis via possible phosphodiesterase activity to modulate the levels of cyclic nucleotides [119]. It was shown that the EAL domain is essential for the biofilm inhibition activity of hmsP [120]. For A. hydrophila, hmsP was 4-fold upregulated in adherent cells on PE, but downregulated upon adherence to PP (Figure 3). C. freundii upregulated hmsP about 2-fold upon adherence to either PE or PP (Figure 4). The toxin-antitoxin gene tabA, known as yjgK in E. coli, influences biofilm formation in a time-dependent manner [121]. During the early phase tabA expression suppresses biofilm formation, while in later phases, it increases biofilm formation; it also suppresses biofilm dispersal [121]. PE-adherent A. hydrophila did not amplify tabA, but PP-adherent samples showed a slight downregulation (Figure 3). C. freundii showed a slight upregulation of tabA upon attachment to either polymer (Figure 4).
The curli operon-related gene csgD is a master regulator of biofilm formation and activates the synthesis of curli fimbriae and EPS in E. coli; csgD suppresses cell motility, triggering biofilm formation [122]. A homolog of csgD is expressed in the irreversible step of biofilm formation in Actinobacillus pleuropneumoniae [123]. Bacterial invasion of host cells may be controlled by c-di-GMP via csgD [124]. Cellulose biosynthesis is promoted by csgD, which is the transcriptional activator for curli production [125]. For both polymers, adherent A. hydrophila showed slightly upregulated expression of csgD (Figure 3), whereas adherent C. freundii downregulated csgD (Figure 4). The Vi antigen is commonly found in strains of Salmonella enterica serovars Typhi and Paratyphi, where it contributes to bacterial virulence and pathogenesis [126][127][128]. However, viaB-related genes for parts of the Vi antigen, which may contribute to invasiveness, have also been reported from Salmonella The gene hmsP (biofilm formation regulator) is one of two genes controlling the amount of biofilm produced in the pathogen Yersinia pestis via possible phosphodiesterase activity to modulate the levels of cyclic nucleotides [119]. It was shown that the EAL domain is essential for the biofilm inhibition activity of hmsP [120]. For A. hydrophila, hmsP was 4-fold upregulated in adherent cells on PE, but downregulated upon adherence to PP (Figure 3). C. freundii upregulated hmsP about 2-fold upon adherence to either PE or PP (Figure 4). The toxin-antitoxin gene tabA, known as yjgK in E. coli, influences biofilm formation in a time-dependent manner [121]. During the early phase tabA expression suppresses biofilm formation, while in later phases, it increases biofilm formation; it also suppresses biofilm dispersal [121]. PE-adherent A. hydrophila did not amplify tabA, but PP-adherent samples showed a slight downregulation (Figure 3). C. freundii showed a slight upregulation of tabA upon attachment to either polymer (Figure 4).
The curli operon-related gene csgD is a master regulator of biofilm formation and activates the synthesis of curli fimbriae and EPS in E. coli; csgD suppresses cell motility, triggering biofilm formation [122]. A homolog of csgD is expressed in the irreversible step of biofilm formation in Actinobacillus pleuropneumoniae [123]. Bacterial invasion of host cells may be controlled by c-di-GMP via csgD [124]. Cellulose biosynthesis is promoted by csgD, which is the transcriptional activator for curli production [125]. For both polymers, adherent A. hydrophila showed slightly upregulated expression of csgD (Figure 3), whereas adherent C. freundii downregulated csgD (Figure 4). The Vi antigen is commonly found in strains of Salmonella enterica serovars Typhi and Paratyphi, where it contributes to bacterial virulence and pathogenesis [126][127][128]. However, viaB-related genes for parts of the Vi antigen, which may contribute to invasiveness, have also been reported from Salmonella enterica serotype Dublin, E. coli and Citrobacter strains [129,130]. A. hydrophila samples did not amplify viaB (Figure 3), but slight upregulation is seen in PE-adhering C. freundii cells, whereas the PP-adhering cells did not amplify this gene (Figure 4). A. hydophila showed more than 2-fold upregulation of slt-II upon attachment to PE or PP (Figure 3). PE-adhering C. freundii upregulated slt-II expression more than 2-fold, while PP-adhering cells upregulated the expression of the same gene more than 6-fold (Figure 4). Biofilm-related genes were not upregulated and the lack of correlation between the expression of biofilm-related genes and the toxin genes remains unexplained. In Escherichia coli O104:H4 expressing the stx2 gene, the correlation of the expression of other genes such as pgaA and aggR with stx2 expression is strain dependent [131].

Secondary Metabolite Production via antiSMASH
Genome mining for secondary metabolite-producing gene clusters using the antiSMASH tool (108) showed that A. hydrophila has the potential to produce non-ribosomal peptides (NRP), arylpolyene, homeserine lactone and bacteriocin type of molecules (Table 5). C. freundii, on the other hand, possesses two biosynthetic gene clusters predicted to produce arylpolyene and NRP-type molecules ( Table 5). The A. hydrophila gene clusters have high similarity to those of other aeromonads. Interestingly, homeserine lactone signals are known to coordinate quorum sensing and biofilm formation in many Gram-negative bacteria [132][133][134].

Electron Microscopy Analysis
Scanning electron microscopy (SEM) was performed to compare the planktonic phases, adherent cells and blood/MacConkey media-induced classical biofilms. SEM analysis of A. hydrophila in the presence of PE showed that the long rod-shaped clumps of cells ( Figure 5B) were transformed into long rows of cells adhered to the surface with some cell curling and the formation of early biofilm-like structures ( Figure 5C). The mature biofilms induced by blood agar, on the other hand, had layers of EPS and only tiny remnants of cells visible ( Figure 5D). For PP, the adherent cells had already formed EPS layers and only tiny cell fragments were visible ( Figure 6C), but the blood agar induced the formation of irregularly shaped aerial structures ( Figure 6D). For C. freundii, planktonic cells had a round morphology with clumping ( Figure 7B), but the adherent cells already formed linked cells in a branching pattern and three-dimensional structures ( Figure 7C). Mature biofilms induced by MacConkey medium appeared similar to the adherent cells, but were more densely networked ( Figure 7D). C. freundii on adhered on PP differed significantly from earlier samples; here the cocci clumped together or linked by fibers ( Figure 8B), are transformed into biofilms containing curled cells on top of cells trapped in EPS ( Figure 8C). The MacConkey-induced mature biofilm was largely flat with cocci trapped in EPS on top ( Figure 8D).

Electron Microscopy Analysis
Scanning electron microscopy (SEM) was performed to compare the planktonic phases, adherent cells and blood/MacConkey media-induced classical biofilms. SEM analysis of A. hydrophila in the presence of PE showed that the long rod-shaped clumps of cells ( Figure 5B) were transformed into long rows of cells adhered to the surface with some cell curling and the formation of early biofilmlike structures ( Figure 5C). The mature biofilms induced by blood agar, on the other hand, had layers of EPS and only tiny remnants of cells visible ( Figure 5D). For PP, the adherent cells had already formed EPS layers and only tiny cell fragments were visible ( Figure 6C), but the blood agar induced the formation of irregularly shaped aerial structures ( Figure 6D). For C. freundii, planktonic cells had a round morphology with clumping ( Figure 7B), but the adherent cells already formed linked cells in a branching pattern and three-dimensional structures ( Figure 7C). Mature biofilms induced by MacConkey medium appeared similar to the adherent cells, but were more densely networked ( Figure 7D). C. freundii on adhered on PP differed significantly from earlier samples; here the cocci clumped together or linked by fibers ( Figure 8B), are transformed into biofilms containing curled cells on top of cells trapped in EPS ( Figure 8C). The MacConkey-induced mature biofilm was largely flat with cocci trapped in EPS on top ( Figure 8D).

Discussion
Genome analysis of the ATCC 7966-type strain (which is closely related to RIT668) showed it to be metabolically versatile with significant virulence potential and a predicted ability to infect a variety of hosts [134]. Due to the high similarity in the genomes (99%), it is possible that A. hydrophila RIT668 shares this potential for broad metabolic capability, virulence and the ability to infect multiple hosts, including humans. The closest relative of RIT669 is a C. portucalensis strain; C. portucalensis strains can be multidrug resistant and some may be highly resistant livestock-origin pathogens or "superbugs" [135,136]. This raises the possibility that RIT669 may also be multidrug resistant.
Enterobactericeae, especially Citrobacter spp., are common among turtles, whereby immunecompromised turtles may be conducive to the expression of pathogenic potential in the gut, which should be considered during rehabilitation procedures [94]. Australian green turtles (Chelonia mydas) in captivity were previously shown to harbor antibiotic-resistant strains [90,94,137]. A. hydrophila is common to infect turtles, fish and other amphibians as well as humans, as it is widely distributed in fresh water, estuarine and marine environments [138,139]. Among turtles, A. hydrophila infections were reported in Pseudemis scripta [140] and soft-shelled turtles (Trionyx sinensis) [141][142][143]. In the 1994 outbreak of A. hydrophila in Italy, there was a 95% mortality rate and the autopsies revealed infection in visceral organs [140]. The animals in that study were apathetic, did not feed and were lethargic in their movements. Our animals were similarly sluggish and the mortality rate was 100% by the end of the study.
Previously, it was shown that healthy pet turtles from seven species (other than Clemmys guttata) and their environment harbored Citrobacter spp. that were multidrug resistant, formed biofilms and were positive for slt-II and viaB [85]. In another study, all the turtles sampled had bacteria resistant to at least two antibiotics and 24% of the isolates were resistant to seven of the eight antibiotics tested [144]. Therefore, the predicted multidrug resistance characteristics of A. hydrophila RIT668 and C.

Discussion
Genome analysis of the ATCC 7966-type strain (which is closely related to RIT668) showed it to be metabolically versatile with significant virulence potential and a predicted ability to infect a variety of hosts [134]. Due to the high similarity in the genomes (99%), it is possible that A. hydrophila RIT668 shares this potential for broad metabolic capability, virulence and the ability to infect multiple hosts, including humans. The closest relative of RIT669 is a C. portucalensis strain; C. portucalensis strains can be multidrug resistant and some may be highly resistant livestock-origin pathogens or "superbugs" [135,136]. This raises the possibility that RIT669 may also be multidrug resistant.
Enterobactericeae, especially Citrobacter spp., are common among turtles, whereby immunecompromised turtles may be conducive to the expression of pathogenic potential in the gut, which should be considered during rehabilitation procedures [94]. Australian green turtles (Chelonia mydas) in captivity were previously shown to harbor antibiotic-resistant strains [90,94,137]. A. hydrophila is common to infect turtles, fish and other amphibians as well as humans, as it is widely distributed in fresh water, estuarine and marine environments [138,139]. Among turtles, A. hydrophila infections were reported in Pseudemis scripta [140] and soft-shelled turtles (Trionyx sinensis) [141][142][143]. In the 1994 outbreak of A. hydrophila in Italy, there was a 95% mortality rate and the autopsies revealed infection in visceral organs [140]. The animals in that study were apathetic, did not feed and were lethargic in their movements. Our animals were similarly sluggish and the mortality rate was 100% by the end of the study.
Previously, it was shown that healthy pet turtles from seven species (other than Clemmys guttata) and their environment harbored Citrobacter spp. that were multidrug resistant, formed biofilms and were positive for slt-II and viaB [85]. In another study, all the turtles sampled had bacteria resistant to at least two antibiotics and 24% of the isolates were resistant to seven of the eight antibiotics tested [144]. Therefore, the predicted multidrug resistance characteristics of A. hydrophila RIT668 and C. freundii RIT669 are not surprising, and this fits into the growing number of studies which suggest that both A. hydrophila and C. freundii are emerging pathogens of concern. In addition, our strains do harbor more extensive resistomes in comparison to previous studies. Bacteria associated with reptiles could cross species barriers and infect mammals, according to a study of pet green turtles in which reptile and clinical strains for some enterobacterial genera (including Aeromonas) were identical [83]. We suggest that zoonotic transmission risks exist for these two potential pathogens and believe that their transmission needs to be monitored from a One Health perspective.
Shiga toxins (Stx) and Shiga-like toxins (Slt) are a group of bacterial toxins involved in human and other animal diseases; they are the cause of bloody diarrhea and hemolytic uremic syndrome [145,146]. Stx or Slt toxins are produced by enterohemorrhagic E. coli, Shigella dysenteriae type 1, C. freundii, Aeromonas spp. and Acinetobacter haemolyticus [38,39,81,145]. Diarrhea-associated C. freundii isolates are reported to contain several toxins, including Shiga-like toxins, heat-stable toxins and a homolog of the cholera toxin B subunit [59]. Clinical isolates of Aeromonas spp. possess Shiga toxin genes (stx1 and stx2), whose sequence is highly similar to the most virulent gene variants of E. coli strains [38]. It is notable that other studies showed specifically that microplastics in the environment may induce the overexpression of other types of virulence-related genes like the integrase int1 genes [98]. Similarly, the increased expression of the slt-II gene upon plastic colonization across all the samples suggests a possible adhesion-specific or even plastic adhesion-specific mechanism of upregulation. However, biofilm-related genes were not upregulated in this study and the lack of correlation between the expression of biofilm-related genes and the toxin genes remains unexplained. In E. coli O104:H4 expressing the stx2 gene, the correlation of the expression of other genes such as pgaA and aggR with stx2 expression is strain dependent [131]. Another study in the same E. coli strain, the only gene reported to be sufficient for plant surface adhesion is ompA, but the correlation is not strong [147]. Therefore, the co-regulated expression of biofilm and virulence-related genes is not consistent across different bacterial species, or even strains of the same species.
Electron microscopy has been used from the 1990s for the examination and characterization of biofilms on medical devices [148,149]. Scanning electron microscopy (SEM) has the level of magnification and resolution necessary to observe the overall shape of microorganisms in the biofilm, as well as their three-dimensional organization [150,151]. Aeromonas spp. were previously reported to attach to surfaces and eventually form biofilms, even if they were freely able to grow in water [52].
A recent study reported that floating biofilm-like structures (BLSs) and the attached biofilms had different metal resistance properties [152]. It has been suggested that "reversible" and "irreversible" attachment to a surface, as well as "surface-sentient" and "surface-naïve" planktonic cells are distinct [153]. The current work suggests that there could be different stages, and that adherent cells may have different properties from true biofilms. Thus, the adherent cells in a system with intermittent shear may not morphologically resemble the well-studied mature biofilm forms induced by blood/bile components, even though they may upregulate the expression of certain biofilm genes or toxins. Further, the structures formed in each case may differ based on both the bacterium and the polymer. Therefore, gene expression during different stages of attachment may be more nuanced than hitherto appreciated.

Conclusions
The isolated strains were shown to have significant resistomes, with A. hydrophila containing predicted resistance genes to six antibiotic classes and C. freundii containing resistance genes for 19 classes. The expression of many of the genes examined did not follow a specific pattern, since the adhesion to the polymers could be controlled via the complex interplay of several genes that were or were not included in this study. However, the clear exception is slt-II, whose expression is increased in response to either PE or PP for both bacteria. The toxin expression notwithstanding, electron microscopy showed that the adherent cells form structures different from well-studied biofilms growing on media with blood/bile components. Extensive antibiotic resistance repertoires, biofilm formation, colonization of common plastics and the overexpression of the slt-II-type diarrheal toxin in plastic-adherent cells, along with the origin in of the bacterial isolates from reptilian niches warrant the classification of both strains in this study as potential opportunistic zoonotic pathogens.