Bacterial strains and plasmids used in this study are listed in Table 1. Escherichia coli strains were grown on standard trypticase soy broth (TSB, BD) at 37°C. For P. salmonis growth, plates filled with nutrient-rich medium ADL-PSA were left for 5 days at 18°C (Henríquez et al., 2015). Bacteria were subsequently used for genomic DNA purification with the GeneJet Genomic DNA Purification Kit according to the manufacturer's instructions (Thermo Fisher Scientific). DNA concentration and quality were assessed by means of a fluorometric method (Qubit, Thermo Fisher Scientific) and agarose gel electrophoresis, respectively. Plasmid miniprep purifications were carried out using a commercial kit (Quiaprep, Qiagen). To collect enough material for restriction digestion analysis, a midiprep was subsequently done (Heilig et al., 2001). Plasmids precipitated with isopropanol and pelleted by centrifugation were finally purified with a commercial kit (GeneJet). Internal procedure PGSe-03 “Management of biological and toxic waste,” which is in compliance with the Chilean directive D.S. 148 “Sanitary Regulation for Hazardous Wastes Management,” was followed for the management of biological residues.

Cloning of putative resistance markers was done using the PCR2.1 TOPO TA Cloning Kit (Invitrogen). PCR fragments comprising putative native promoter and coding sequences (CDS) were amplified with primers listed in Table 2. The PCR mix contained 2.5 μL of buffer, 1.0 μL of 50 mM MgCl₂, 0.2 μM of each primer and 1 U of Taq Platinum in a final volume of 25 μL. PCR conditions were as follows: initial denaturation at 95°C for 5 min, followed by 30 cycles of denaturation at 95°C for 15 s, annealing at 60°C for 15 s, extension at 72°C for 15 s, with a final extension at 72°C for 5 min. Primers p3PS10_F and R and conditions similar to these described above were used to verify the presence of p3PS10 in selected isolates of our P. salmonis strain collection. For restriction analysis, 0.5 μg of plasmid DNA were digested with EcoRI or BamHI for 1 h at 37°C. Samples were separated by 1% agarose gel electrophoresis at 110 V for 30 min.

qPCR-based parameters were applied for the calculation of p3PS10 plasmid copy numbers (PCN; Skulj et al., 2008). Briefly, aliquots of broth cultures of AY3800B, E. coli p3PS10, and E. coli pOTC were each collected after 24, 48, 72, 96, and 120 h of cultivation at 18°C for P. salmonis or 37°C for E. coli. Bacterial disruption was accomplished by heating the samples to 95°C for 10 min in PCR water. qPCR were done using the primer pairs tetR_F and R, BC_F and R, and the corresponding probes tetR_P and BC_P. For E. coli strains, primers and probe targeting the rpoD gene as a single-copy chromosomal marker were used. PCR conditions were those recommended for TaqMan Gene Expression Master Mix (Applied Biosystems): a holding stage at 50°C for 2 min, enzyme activation at 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Calibration curves were obtained using log₁₀ dilutions of ultramers tetR_U, rpoD_U and BC_U (Table 2). Efficiencies of PCR reactions were taken into account for PCN calculations derived from plasmid/chromosomal quotients as described by Skulj and colleagues.

The genomes of OTC-resistant AY3864B, AY6297B, and AY6532B P. salmonis LF-89-like isolates were sequenced using Pacific Bioscience (PacBio) chemistry. For AY3864B, Illumina technology (HiSeq 2000 platform, RRID:SCR_010233) was utilized additionally. Libraries were prepared using the kits recommended by the manufacturers: DNA Polymerase Binding Kit P6 v2 for PacBio and Nextera for Illumina. Sequencing was performed at Macrogen Inc., Seoul, South Korea. Long reads were de novo assembled with SMRT Analysis version 2.3.0.1 (PacBio, RRID:SCR_002942) to yield the full sequence of each replicon. Subsequently, read correction was done by mapping Illumina paired-end short reads against the PacBio assembly with CLC Genomic Workbench v6.5 software (CLC bio, RRID:SCR_011853). Annotation was accomplished with the NCBI Prokaryotic Genome Annotation Pipeline using the best-placed reference protein set as the annotation method implemented in GeneMarkS+ revision 3.0 software (RRID:SCR_011930). Sequencing statistics and accession numbers for the sequences deposited in GenBank are found in Supplementary Table 1.

Comparative analyses were carried out with CLC software after the p3PS10 sequence was retrieved from GenBank under the accession number CP013819. Standard algorithms BLASTN and BLASTX at NCBI were also used for sequence analyses. Refinement of antibiotic resistance marker prediction was done with tools to be encountered in the Comprehensive Antibiotic Research Database, CARD¹ (McArthur et al., 2013). This analysis was extended to the entire genome of AY3800B and the reference LF-89 strains (Pulgar et al., 2015; Bohle et al., 2017). Primer design was done with the Primer3 webtool²

Due to the lack of standardized procedures for the determination of MIC for aquatic microorganisms like P. salmonis, we carried out a protocol previously established in our laboratory to this end (Henríquez et al., 2015). This assay relies on the use of PSB medium at 18°C, which supports the growth of P. salmonis. This medium is equivalent to the one recommended by the Clinical Laboratory Standards Institute (CLSI) to perform MIC assays for P. salmonis (Yañez et al., 2012). Stock solutions of chloramphenicol (CFC), FFC, OTC, streptomycin (STR), and sulfamethoxazole/trimethoprim 19:1 (SXT) (all purchased from Sigma-Aldrich) were prepared following the procedure described in the guidelines (CLSI, 2014a). Solvents for the dilution of antibiotics corresponded to those described in the CLSI document VET03/VET04_S2 (CLSI, 2014b). 96-well microplates were filled with twofold serial dilutions of 0.06–1,024 μg/mL of each antibiotic, then inoculated with approximately 5 × 10⁵ colony forming units/well. MIC were determined with a microplate reader after 96–120 h, considering the lowest concentration of antibiotics that impeded bacterial growth. For the assessment of MIC for E. coli, Mueller-Hinton broth with cation-adjusted medium (BD) at 22°C was used along with E. coli ATCC 25922 as control strain.

Donors P. salmonis AY3800B and p3PS10-lacking P. salmonis (PM15972A1, PM32597B1, PM63907), as well as E. coli TOP10 recipient strains were grown to the early exponential phase of their respective growth curves. Aliquots of 1.0 mL (OD₆₀₀ = 1.0) were mixed, centrifuged at 5,000 rpm for 5 min, and resuspended in 0.1 mL of saline before being spread onto a non-selective ADL-PSA plate in the absence of antibiotic. Plates were incubated at 18°C for 24 h. Bacteria were collected by scraping plates and were transferred to 1.0 mL of saline solution, spinned down for resuspension in 0.1 mL of saline and finally spread onto ADL-PSA or TSA plates under selective conditions depending on the recipient. P. salmonis transconjugants were recovered after 7 days at 18 or 22°C, and E. coli transconjugants after 24 h at 22°C. In the case of P. salmonis mating, transconjugants were selected on PSA agar plates plus OTC (50 μg/mL), in some instances after their additional supplementation with oxolinic acid (5 μg/mL). In all experiments performed, donor and recipient cells streaked on plates selective for transconjugants were used as controls. Transconjugant genotypes were confirmed by PCR. Transfer frequencies were measured as the ratio of transconjugants to recipient cells assuming similar numbers of donor and recipient cells.

Plasmid p3PS10 is predicted to encode 52 CDS, which are listed in Table 3. According to their hypothetical functions, the entire sequence can be split into three sections: replication/maintenance, transmissibility and antibiotic resistance determinants (Figure 1A). Interestingly, a pair of tet(31) tetracycline (TET) resistance genes encoding a transcriptional regulator and an efflux pump protein were found, resembling a canonical array for TET-resistance cassettes (Thaker et al., 2009). In close proximity to the tet element we found a gene encoding dihydropteroate synthase (sul2), an enzyme associated with reduced susceptibility to sulfonamides, and an acetyltransferase encoded by cat2, which likely confers resistance to CFC (Sköld, 2000; Schwarz et al., 2004). Moreover, integron-associated acetyltransferase sat1, along with an adenylyltransferase (aadA1) related to aminoglycoside resistance were identified (Ramírez et al., 2010). All these CDS were detected within an 11-kb locus (Figure 1B). This region has a complex organization including integron and transposon arrangements. Restriction digestion of p3PS10 with EcoRI or BamHI yielded fragments of 6.2 and 2.4 kb, suggesting that the size of the antibiotic resistance locus was accurately estimated from the sequence assembly (Figure 1C). Comparative analysis performed with CARD matched with all the mentioned markers, but also identified five additional chromosomal targets for CFC resistance in sequences of both the wild-type LF-89 and AY3800B strains (Supplementary Table 2).

Conventionally, the determination of the plasmid incompatibility (Inc) group is based on the amino acid sequence of the replication protein Rep, a key component of plasmid replication systems (Shintani et al., 2015). Comparative analyses of the rep sequence of p3PS10 (WP_012850373) showed a perfect match with the corresponding rep gene of pEIB202, a plasmid that has been classified as a member of the IncP group (Wang et al., 2009). Since components of the partition system of p3PS10 are also related to the IncP group (Siddique and Figurski, 2002), we postulate that p3PS10 belongs to this same Inc group.

Concerning the PCN of p3PS10, we predict the value to be low and rest our appraisal on the yield obtained by midipreps (Supplementary Figure 1). Indeed, qPCR-based calculations indicated an average PCN of 3-5 in P. salmonis cells (Figure 2). This low PCN was also observed in the heterologous host E. coli p3PS10, but not in the control strain E. coli pOTC, which carries a high-copy-number pCR2.1 derivative. In this strain, PCN of pOTC exceeded 1,000, as can be expected for this type of plasmid (Invitrogen). Strikingly, PCN of p3PS10 remained constant even after 120 h (~12 generations) of culture in the absence of selective pressure, irrespective of the host. The same observation was made after 60 generations of cultivation (not shown). In contrast, PCN of pOTC started to decline after 48 h in medium without antibiotic.

Bioinformatic analyses disclosed that p3PS10 encodes a conjugative T4SS gene cluster, which may permit its transmission to different hosts. This possibility was tested in mating assays. These experiments revealed that the donor strain AY3800B is able to transfer the plasmid into homologous hosts. In fact, our results show that the plasmid was mobilized between P. salmonis isolates displaying different antibiotic susceptibility patterns (Table 4). Remarkably, p3PS10 was able to replicate within an EM-90-like isolate, representative of a genogroup that has only shown wild-type susceptibility for the major classes of antibiotics utilized for SRS treatment (Saavedra et al., 2016). We calculated the frequencies of plasmid transference and confirmed that such events are rare when donor and recipient derive from different genetic backgrounds, e.g., when LF-89-like AY3800B was paired with EM-90-like PM15972A1. In contrast, transfer frequencies were one or even two orders of magnitude higher for mating pairs with similar genetic backgrounds: in AY3800B-PM63907 and AY3800B-PM32597B1 mating assays, donors and recipients belonged to the LF-89 genogroup (Otterlei et al., 2016). The plasmid was also transferred to the heterologous recipient E. coli. In this species, the formation of transconjugants occurred at a frequency similar to that observed in P. salmonis mating pairs with the LF-89 background. Antibiotic-resistant colonies were considered as true transconjugants since no spontaneous mutants with reduced susceptibility to either CFC, OTC, SXT, or STR were detected in recipient backgrounds.

Based on our results of sequence analyses and the detection of multiple resistance traits, we decided to characterize the susceptibility of p3PS10-carrying P. salmonis AY3800B to OTC, FFC, CFC, STR, and SXT by means of MIC assays (Table 5). The comparison of MIC values for AY3800B with those obtained for the wild-type P. salmonis LF-89 reference strain showed that the AY3800B strain was indeed less susceptible to OTC, CFC, and STR, while MIC for SXT were high in both strains. Interestingly, the E. coli TOP10 derivative resulting from the transference of p3PS10 exhibited MIC values comparable to those measured for the donor strain AY3800B, which indicates the functionality of resistance markers carried by the plasmid. To confirm this hypothesis, resistance markers for OTC and CFC were cloned into E. coli TOP10. For both antibiotics tested, the phenotypic traits of the modified strains were consistent with the bioinformatically predicted function of the corresponding CDS found in p3PS10. In fact, MIC of OTC and CFC were 1,024 and 512 μg/mL in E. coli pOTC and E. coli pCFC, respectively, which are values similar to those obtained in the E. coli p3PS10 strain. The fact that E. coli pCFC did not lose susceptibility to FFC, a halogenated CFC derivative, after the transfer of p3PS10, suggests that the reduced susceptibility to CFC is mediated by an acetyltransferase selective for this drug. Due to the intrinsically high in vitro MIC of STR for E. coli TOP10, we chose to clone the predicted resistance marker for STR into the STR-susceptible E. coli DH5α. This way, we were able to show the functionality of this marker, too. With regards to susceptibility to SXT, P. salmonis isolates proved to be resistant to this compound irrespective of the presence of p3PS10. Nevertheless, cloning of sulfonamide resistance determinant sul2 into E. coli TOP10 yielded E. coli pSUL2, which displayed a significantly decreased susceptibility to SXT.

The sequencing of P. salmonis AY3864B and AY6297B showed that both isolates harbor an identical p3PS10 plasmid that has been given different names during automatic annotation (p3PS11 and p4PS8, respectively), whereas P. salmonis AY6532B carries p4PS9, a smaller plasmid encompassing 20.4 kb and lacking transfer genes, sulfonamide and TET resistance markers (Supplementary Figure 2, Supplementary Table 3). BLASTN analysis disclosed that 35 kb of the backbone of p3PS10 matched with an identity of more than 99% to the sequence of the 43.7-kb plasmid pEIB202 found in the Gram-negative bacterium Edwarsiella tarda strain EIB202, the causative agent of the fish disease edwardsiellosis (Wang et al., 2009). Both p3PS10 and pEIB202 show similar patterns of genetic organization and comprise sequences encoding multidrug resistance markers, genes required for replication, and a set of T4SS genes. Nevertheless, the 11-kb locus of p3PS10 that harbors resistance genes clearly distinguishes this plasmid from pEIB202. The sequence of this locus showed to be partially homologous to a segment of 3.5 kb found in the 48-kb plasmid pRAS2 (accession number AJ250203). pRAS2 was isolated from Aeromonas salmonicida subsp. salmonicida 1682/92, and constitutes an episome conferring STR, sulfonamide, and TET resistances (L'Abee-Lund and Sorum, 2000).

To study the similarity between p3PS10 and known plasmids at the level of amino acid sequences, BLASTX searches were carried out using the sequence of the 11-kb locus of p3PS10. We could identify several components conserved across different bacterial species, e.g., those genes encoding a class 2 integron-associated integrase, aminoglycoside resistance proteins, and a pair of IS1182 transposases flanking the cat2 gene (Figure 1B). The class 2 integron sequence conferred resistance to STR and was almost identical to the one detected in the cholera agent Vibrio cholerae VC97 (accession number GU570569; da Fonseca et al., 2011). Furthermore, the transposon-like structure in p3PS10 proved to be identical to a locus found in a plasmid carried by a CFC-resistant Photobacterium damselae subsp. piscicida strain (accession number AB453229), causative agent of pseudotuberculosis in fish (del Castillo et al., 2013). The 2-kb sequence encoding the tet-resistance determinant carried by p3PS10 was found in several aquatic pathogens, such as in A. salmonicida and Vibrio splendidus. Somewhat surprisingly, this sequence was also identified in the genome of Gallibacterium anatis strain 12656-12, a bacterium causing septicemia in chicken (Kudirkiene et al., 2013). The tet-resistance element was highly prevalent in a collection of G. anatis field isolates resistant to tetracycline, which have been gathered in Mexico and Denmark (Bojesen et al., 2011). We investigated the prevalence of the tet element in a group of representative isolates from our P. salmonis collection (Figure 3). As expected, those isolates carrying tet(31) were associated with OTC treatment failures observed in the field (Supplementary Table 4). These isolates were recovered from SRS outbreaks that took place in the south of Región de Aysén, in the period 2013-2015 (Supplementary Figure 3). This restricted dissemination pattern contrasts with the wide distribution reported for non-wild-type isolates for quinolones and FFC (Henríquez et al., 2015).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.