Extended antibiotic treatment in salmon farms select multiresistant gut bacteria with a high prevalence of antibiotic resistance genes

The high use of antibiotics for the treatment of bacterial diseases is one of the main problems in the mass production of animal protein. Salmon farming in Chile is a clear example of the above statement, where more than 5,500 tonnes of antibiotics have been used over the last 10 years. This has caused a great impact both at the production level and on the environment; however, there are still few works in relation to it. In order to demonstrate the impact of the high use of antibiotics on fish gut microbiota, we have selected four salmon farms presenting a similar amount of fish of the Atlantic salmon species (Salmo salar), ranging from 4,500 to 6,000 tonnes. All of these farms used treatments with high doses of antibiotics. Thus, 15 healthy fish were selected and euthanised in order to isolate the bacteria resistant to the antibiotics oxytetracycline and florfenicol from the gut microbiota. In total, 47 bacterial isolates resistant to florfenicol and 44 resistant to oxytetracycline were isolated, among which isolates with Minimum Inhibitory Concentrations (MIC) exceeding 2048 μg/mL for florfenicol and 1024 μg/mL for oxytetracycline were found. In addition, another six different antibiotics were tested in order to demonstrate the multiresistance phenomenon. In this regard, six isolates of 91 showed elevated resistance values for the eight tested antibiotics, including florfenicol and oxytetracycline, were found. These bacteria were called “super-resistant” bacteria. This phenotypic resistance was verified at a genotypic level since most isolates showed antibiotic resistance genes (ARGs) to florfenicol and oxytetracycline. Specifically, 77% of antibiotic resistant bacteria showed at least one gene resistant to florfenicol and 89% showed at least one gene resistant to oxytetracycline. In the present study, it was demonstrated that the high use of the antibiotics florfenicol and oxytetracycline has, as a consequence, the selection of multiresistant bacteria in the gut microbiota of farmed fish of the Salmo salar species at the seawater stage. Also, the phenotypic resistance of these bacteria can be correlated with the presence of antibiotic resistance genes.

Introduction injection [26,27]. Medicated feed is not fully digested by fish and, in many cases, is in fact poorly digested and metabolized, with the consequence being a constant liberation of antibiotic substances into the environment [28]. Without doubt, medicated feed also fundamentally affects the intestinal microbiota of fish [29]. Constant exposure to antibiotic substances leads to the selection of resistant bacteria and an increase in the horizontal transfer of antibioticresistance genes (ARGs) [30]. This situation ultimately means that the feces of medicated-feed fish are rich in ARGs.
The primary aim of this study was to characterize the antibiotic-resistant bacteria present in the intestinal microbiota of farmed Atlantic salmon treated with high antibiotic doses. Four salmon farms were sampled, and different antibiotic-resistant bacteria were recorded at each. Specific genetic elements involved in resistance to florfenicol and oxytetracycline were identified, as were integron elements. This paper establishes a clear relationship between the use of antibiotics, the presence of resistant bacteria and the high prevalence of antibiotic resistance genes in the intestinal-microbiota system from the Atlantic Salmon. Additionally, which could be an important source for the dispersion and liberation of ARGs into the environment, with potential impacts for human health.

Sample collection
Four salmon farms located in the Aysén Region, in the Cupquelan Fjord (Northern Patagonia, Chile) were assessed. The farms were chosen for four fundamental characteristics: (i) all the farms contain fish of the Salmo salar species in a similar period of the productive cycle; (ii) similar mass of fish (between 4500 and 6000 tons); (iii) all the farms had more than one outbreak of SRS in the productive cycle; and (iv) the four farms had more than one treatment with medicated food at the time of sampling. ( Table 1).

Isolation of bacteria from intestinal fish microbiota
In order to obtain the largest number of bacterial isolates, two sources were used: (i) fecal matter; and (ii) the intestines. For this, 15 apparently healthy (i.e., no clinical signs of SRS) Atlantic salmon were randomly selected from each salmon farm. Fecal matter was obtained by applying perianal stimulation to the fish. Prior to fecal extraction procedure fish were anesthetized with benzocaine (25 μg/mL). The collected fecal samples from three individuals were pooled, stored in a 5 mL saline solution (0.85% NaCl), and kept on ice until laboratory analyses 24 h later. To collect the intestine samples, the animals were euthanized by immersion in a solution of 50 mg/L of benzocaine. The intestines of three fish were also pooled. This study was carried out in accordance with law 20,380 regarding animal welfare, as set out by the Chilean Health Ministry in the use of wild or protected animal species in biomedical research and approved by the National Fisheries Service (SERNAPESCA) and the Pontificia Universidad Católica de Valparaíso Bioethical Committee. Table 1. Number of florfenicol (FLO) and oxytetracycline (OTC) oral treatments, sum of the amount of medicated feed with antibiotics and total fish weight in each Atlantic salmon farm at the time that sampling.

Farm
No. treatments and types Amount of medicated feed with antibiotics (in kg) To obtain the bacterial isolates, homogenates were obtained from the pooled fecal matter and intestines. A dilution series was applied, where 10 −4 and 10 −5 dilutions (100 μL) were seeded in sextuplet on tryptic soy agar plates (TSA) (Difco, USA). Plates were incubated at 25˚C, 30˚C, and 37˚C for 48 h and at 15˚C for 72 h, the aim of which being to assess the widest temperature range possible. TSA medium and incubation temperature have been commonly used to isolate bacteria from fish microbiota [31][32][33]. To prevent fungus and yeast growth, a 50 μg/mL concentration of cycloheximide was added as an antifungal agent.

Construction of a bank of bacterial isolates
Colonies grown in TSA medium were classified according to standard patterns: shape, color, texture and shape of the colony-border. Each of the differentiated colonies was seeded in 96-well plates in TSB medium and grown at the temperature at which it was isolated, regardless of whether it came from intestine or fecal material. In order to ensure the purity of the isolates, each of the colonies grown in TSB medium were Gram-characterized and the mixed cultures or cultures contaminated with yeasts were discarded.

Determinations of minimum inhibitory concentration (MIC)
The MICs for all the bacterial isolates were determined following the agar double-dilution protocol established by the Clinical and Laboratory Standards Institute [34,35]. Briefly, each MIC was estimated by inoculating square plates (120 mm 2 ) with the Müller-Hinton medium (Sigma-Aldrich, USA) and increasing concentrations of florfenicol and oxytetracycline (0-2,048 μg/mL), using a 96 pin replicator (Boekel 140500). As a control was used the strain E. coli K12. Once the MIC of each isolate was calculated, those isolated showing a MIC ! 128 μg/ mL for florfenicol and ! 64 μg/mL for oxytetracycline were considered to be resistant bacteria, according to EUCAST clinical standard [36]. Finally, two banks of resistant bacteria were created-a florfenicol-resistant bank (FB) and an oxytetracycline-resistant bank (OB)

Molecular identification of resistant bacteria
To identify the resistant isolates, 16S rRNA gene sequencing analysis was used for taxonomic classification. Genomic DNA extracts were obtained with Chelex-100 (Bio-rad, USA) according to manufacturer instructions. PCR amplification of the 16S gene was carried out using the universal primers 27F (5'-AGAGTTTGATCMTGGCTCAG-3') and 1492R (5'-GGTTACCT TGTTACGACTT-3'). The amplification conditions were as follows: initial denaturalization at 95˚C for 3 min, followed by 35 cycles at 95˚C for 45 s, 55˚C for 30 s, and 72˚C for 1 min, with final extension at 72˚C for 5 min. The obtained sequences were analyzed using the BLAST tool [37], with comparisons conducted against sequences available in GenBank (NCBI). Five rounds of amplification and sequenciation were used to verify the sequences. The obtained 16S rRNA sequences for each resistant bacterial isolate were submitted to the GenBank database (S1 Table).

Characterizing genetic determinants of resistance
The incidences of genetic determinants implicated in resistance to florfenicol and oxytetracycline were evaluated through PCR and sequencing of the most relevant genes described in relation to antibiotic resistance. The partial sequences of the genes (three or four for fish farm) were submitted to the GenBanK database (S2 Table). For florfenicol, these genes included floR [38,39], fexA [40,41], and cfr [42]. For oxytetracycline, these genes included tetA, tetB, tetE, tetL, tetH, tetM, tet34, and tet35 [43]. The primers used for each gene are given in Table 2. All primers were validated with the NCBI primer-Blast tool before their use.

Statistical data analysis
Statistical analyses were performed using SPSS 22.0 for Windows (Chicago, IL). The prevalence of antibiotic resistance bacteria and antibiotic resistance genes was compared among isolates from different origins or taxonomic groups using the chi-square test at a significance level of 0.05.

Characterization of the bacterial isolates bank
More than 12,000 colonies were obtained in the TSA plates without antibiotics from the four farms sampled. Of these colonies, 28% was obtained at 15˚C, while 63% of the isolates came from the growth at 25˚C. Finally only 9% of the colonies came from the growth at 30˚C. In none of the fish farms were bacteria found that could grow at 37˚C. After the classification of the colonies according their pattern of size, shape, color, texture and the shape of the border, and after discarding the mixed cultures or those contaminated with yeast, the number of colonies was considerably reduced: 2,628 coming from the intestine and 2,390 colonies from the feces, showing no differences between this two different sources. The individual analysis per farm showed that 1,248 colonies were obtained from the farm I; while 1,035 are from farm II; 1,472 from farm III; and 1263 from farm IV. The amount and diversity of the patters of size, shape, texture, color and shape of the border from the colonies were similar for the four farms.

Determinations of MICs for all isolates of the bacterial bank
The MICs to florfenicol or oxytetracycline were determined using microdilutions in Müller-Hinton agar, adding increasing concentrations (0-2,048 μg/mL) of each antibiotic for all isolates of the bacterial bank. Once MICs were determined, those isolates evidencing resistance to !64 μg/mL for oxytetracycline or !128 μg/mL for florfenicol were classified as resistant. A total of 47 (23 from feces and 24 from intestine) florfenicol-resistant and 44 (21 from feces and 23 from intestine) oxytetracycline-resistant isolates were classified (Figs 1 and 2). Of the bacteria resistant to florfenicol, 93.6% presented a MIC >256 μg/mL, and 38.2% of these isolates had a MIC > 1,024 μg/mL. Of the bacteria resistant to oxytetracycline, 56.8% presented an MIC >256 μg/mL, and 34.1% of this group had an MIC of 1,024 μg/mL. These findings suggest a high degree of resistance in bacteria from the gut microbiota of fish fed with medicated feed containing these antibiotics. Even at the three farms where medicated food did not contain oxytetracycline, high MICs against this antibiotic were found.
With this data the Antibiotic Resistance Rate (ARR) (resistant bacteria/total culturable bacteria) was calculated, and for the florfenicol resistance samples this rate was estimated in 9.4x10 -3 and 8.7x10 -4 for the oxytetracycline resistance samples. In other words, less than 1% of the total isolates were classified as resistant bacteria. If this value is analyzed by farm and by resistant bank, it can be seen that farm I shows a value of 8x10 -3 , while farm II shows a value of 1.2x10 -2 , and farms III and IV show values of 7.4x10 3 and 7.9x10 -3 respectively flor florfenicol. In this case the farm II shows an antibiotic resistance rate almost two times higher compared with the others three farms. This results could be explained by the amount of florfenicol treatment used in the farm II. In the case of the OB the values are the following: Farm I: 1.4 x10 -2 ; farm II 9.6x10 -3 ; farm III 4,7x10 -3 ; and farm IV 7.1x10 -3 . In this case it is the farm II that has a value almost two times higher compared to the rest of the farms, which could be explained as it is the only farm where oxytetracycline is used as a treatment in seawater.

Multiresistant bacterial isolates to florfenicol and oxytetracycline
To estimate the multiresistance capacity against other antibiotics in strains classified as resistant to florfenicol and/or oxytetracycline (i.e., MIC !64 μg/mL for oxytetracycline or !128 μg/mL for florfenicol), MIC values were determined against the following antibiotics:  chloramphenicol, tetracycline, erythromycin, ampicillin, ciprofloxacin, and kanamycin. Among the FB resistant isolates, 4.3% showed resistance phenotype to the eight assessed antibiotics, and 57.3% showed resistance to at least four antibiotics (Table 3). Different results were found for OB resistant bacteria, with a 8.5% of the bacterial isolates showed resistance to the eight antibiotics. Nevertheless, 100% of OB-classified isolates were resistant to at least four of the tested antibiotics (Table 4). These findings evidence a clear selection tendency for multiresistant bacteria when florfenicol and oxytetracycline are used as treatments.
Particularly interesting results were found for resistance tests against ciprofloxacin and kanamycin, because only two of the florfenicol-resistant isolates showed resistance for ciprofloxacin (Table 3). But in the case of the OB a 54.5% of the bacterial isolates showed resistance to ciprofloxacin (Table 4). Similar results were found when testing against the antibiotic kanamycin-only nine florfenicol-resistant bacteria evidenced a resistance phenotype against kanamycin (Table 3). In turn, 45.4% of those bacteria resistants to oxytetracycline were also resistant to kanamycin, and nearly all of these bacteria originated for farm I (Table 4).
It is important to mention that only six bacterial isolates, two from the FB and four from the OB showed resistance to the eight antibiotics evaluated. The two isolates of the FB correspond to members of the Pseudomonas genus, while of the OB two of them were classified as member of the Serratia genus and two of the Carnobacterium genus. Among the six identified multiresistant bacteria, a Pseudomonas sp. from the FB is particularly concerning since this bacterium presented a MIC ! 2,048 μg/mL to six of the eight evaluated antibiotics. Considering this finding, the assessed Psudomonas sp. isolate was termed "super-resistant" and is currently being sequenced for continued research.

Genetic determinants involved in antimicrobial resistance
After an exhaustive analysis of phenotypic resistance, a search was conducted for genetic elements directly involved in resistance to florfenicol and oxytetracycline. Integron elements were also evaluated since these elements can collect antibiotic-resistance genes (ARGs), thus facilitating expression [44]. These elements can not only incorporate and actively express ARGs, because if they are incorporated in mobile genetics elements like plasmids, they can be easily transferred also, increasing considerably the dispersion of those genes [45,46]. As expected, FB resistant bacteria showed a high incidence of floR and fexA, genes that code for efflux pumps from the major facilitator superfamily group MFS, that expulse a significant number of molecules, including antibiotics, thereby conferring resistance to florfenicol. Of the bacterial isolates with the FB, 70.2% carried floR, 46.8% carried fexA, and 34% evidenced both genes. Only 17% did not carry either floR or fexA (Table 5). Regarding bacteria from the OB, none carried floR. However, 52.2% did present fexA (Table 6). These results support that determinants of florfenicol resistance are widely distributed across the gut microbiota of farmed fish as a result of high antibiotics administration.
After assessing the presence of florfenicol-resistance genes among the FB and OB isolates, analyses were conducted for genes connected to oxytetracycline resistance. Eight resistance genes were included in evaluations, all of which confer resistance to the tetracycline drug Isolation and characterization of multiresistant bacteria from farmed salmon gut    Isolation and characterization of multiresistant bacteria from farmed salmon gut family. These apply named tet genes included tetA, tetB, tetE, tetH, tetL, tet34, and tet 35, all of which were previously described in sediments associated with salmon farms [43], but not with intestinal microbiota of the farmed fish. The presence of tetM was also included considering prior reports of being a widely dispersed and prevalent resistance mechanism against tetracyclines [47,48]. For OB resistance bacteria, 100% of isoaltes presented at least one of the evaluated tet genes, the more prevalent of which was, not surprisingly, tetM (72.7%). Following in prevalence was tetL (68.1%). Among this group of bacterial isolates, 4.5% of the bacterial isolates carried six tet genes, 9,1% five and 15.9% four. In other words, 29.5% of oxytetracyclineresistant bacteria carried at least four different resistance genes to this antibiotic (Table 6).
Regarding those bacteria included in the FB, 78.7% of isolates had at least one of the evaluate tet genes, 48.9% of which were carries of tetH (i.e., the most abundant in florfenicol-resistant bacteria; Table 5). Likewise, an estimated 12.8% of the FB isolates carried at least three of the evaluated tet genes. The presence of specific ARGs against florfenicol and oxytetracycline corroborate the phenotypic resistances observed for each of the resistant isolates. Further research is needed regarding those resistant isolates for which no ARGs were detected. To understand these results, it is necessary to take into account the limitations of detection using the PCR technique, since only eight out of 28 tet genes were evaluated [49], the most frequently described in aquaculture [50]. The same happens with the genes of resistance to phenicoles since only those that are detected more frequently at environmental level were evaluated [38,51]. Therefore a metagenomic analysis would be very useful to detect all the possible genes involved in the resistance against these antibiotics.
Once resistance elements to the most widely used antibiotics were determined, a search for integron-like elements was conducted. The presence of type I integrase was established through PCR analyses and posterior sequencing. In the case of OB isolates, 100% were positive for type I integrase, while 96% of FB isolates presented type I integrase (Tables 5 and 6). These results suggest that this type of element might be widely distributed among resistant bacteria residing within the gut microbiota of fish. These are highly dangerous elements that play a fundamental role in the distribution and phenotypic expression of ARGs. A dedicated study on the detected integrons is currently underway in our lab.

Taxonomic differences among bacteria resistant to florfenicol and oxytetracycline
In contrast to previously published reports, the present study simultaneously provides information related to the phenomena of resistance to florfenicol and oxytetracycline, with the different tested isolates having been directly obtained from the intestinal microbiota of farmed salmon exposed to high concentrations of antibiotics. Of note, the conducted analyses detected differences in species richness for resistant bacteria. More specifically, the FB isolates evidenced a high predominance of Pseudomonas species, a phenomenon that was found across the sampled farming centers and among all the evaluated temperatures. To a lesser degree, members of the Gammaprotebacteria class were also isolated, including Aeromonas, Serratia, and Hafnia species. Regarding OB isolates, a high incidence of Gammaprotebacteria were also observed, as represented by the genera Serratia, Hafnia, Rouxiella, Rahnella, Kluyvera, Shewanella, psychrobacter, Aeromonas, and Pseudomonas. Furthermore, Gram-positive bacteria were represented by the Firmicutes phylum, notable among which were Carnobacterium and Brochothrix, among others.

Discussion
Antibiotic use in the production of animal proteins is a practice that should be controlled on a global scale, specifically since the application of these medicines for veterinary ends many times lacks the degree of strict control imposed on human medicines. One of the primary methods for administering antibiotics to industrially farmed animals is medicated feed [52][53][54]. This delivery method is the most popular among farmed fish due to the notable technical difficulties associated with administering antibiotics through other means, such as injection [55,56]. The ingestion of medicated feed implicates that the intestinal microbiota of these animals is strongly affected. Antibiotics substantially change the bacterial diversity of gut microbiota in farmed animals [53], which, often, can result in functional and growth problems [57]. Another effect caused by the high use of the medicated food on the intestinal microbiota is that the feces could contain resistant bacteria, which could be easily dispersed in the environment [58,59]. In this work it was possible to demonstrate that the feces contain multiresistant bacteria, which are constantly being released into the environment. Additionally, it was observed that the diversity of resistant isolates is similar to that found in the intestine, reaffirming in this way that eventually any resistant bacteria selected in the intestine carrying ARGs could be disseminated through the feces.
Various studies have extensively characterized the gut-microbiota composition of Atlantic salmon, reporting total dominance by the Proteobacteria phylum. More specific still, Pseudomonas species of the Gammaproteobacteria class are the most represented [29,60]. These prior descriptions were supported by the present study, with most of the resistant isolates being of the Pseudomonas genus. Prior research has similarly characterized the effect that oxytetracycline use has on the microbiota composition of Atlantic salmon. For example, Navarrete et al. (2008) [29] exposed Atlantic salmon to high oxytetracycline concentrations under experimental conditions (i.e., not in the field) and found a notable enrichment of Aeromonas salmonicida, a bacterial pathogen. In the present study, only one Aeromonas isolate was found in the OB group, thus corroborating the possible selection of this bacterium under high oxytetracycline concentrations.
A fundamental finding of the present study is that oxytetracycline-resistant isolates were found to be more diverse than those resistant to florfenicol. This could have direct implications on treatments and the strategies applied to the use of these antibiotics. Since the early report of P. salmonis infection by the end of the 80s, oxytetracycline and quinolones have been used to tackle the infection. Also, oxytetracycline is the most widely used antibiotic during freshwater rearing, with florfenicol use nearly null during this stage [25]. Consequently, the large majority of farmed fish with a freshwater stage have been subjected to oxytetracycline treatments [25]. This could, in turn, mean that the gut microbiota of these fish is already rich in bacteria resistant to oxytetracycline. Little research has reported on substantial changes to the microbiota of freshwater-versus seawater-stage salmon. However, some authors speculate that a large part of the microbiota would be conserved between systems [61]. In contrast to freshwater farming, florfenicol is the most widely used antibiotic in seawater farming. Supporting this claim, all of the currently assessed fish farms had applied three to four oral applications of florfenicol, and only farm I had used an additional treatment of oxytetracycline at the time of sampling. This might explain the lesser diversity of bacteria resistant to florfenicol; i.e., the constant florfenicol treatments may have selected and maintained only those bacteria with the most efficient mechanisms against this antibiotic. This is the perfect scenario for Pseudomonas members, the genomes of which are equipped with multiple resistance elements that make these bacteria competitive in environments saturated with antibiotics [62,63].
In addition to the well-known antibiotic resistance capacity of Pseudomonas, the present study detected most of the resistance genes included in assessments, and the majority of these genes, excepting tetM, were drug efflux pumps. These findings corroborate that these genetic elements are a principal trait of Pseudomonas bacteria [64]. Considering the phenotypic versatility of Pseudomonas, it was unsurprising that the majority of the obtained florfenicol-resistant isolates were of this genus. Regarding oxytetracycline-resistant isolates, the detected genera, as with Pseudomonas, are highly ubiquitous in different systems. For example, Hafnia alvei has been isolated from the intestinal microbiota of various organisms, including pigs, cows, and fish [65,66]. This bacterium is an important contaminant of packaged food [67], meaning that H. alvei is an ideal resistant-isolate candidate in any system. These traits can also be found in species of Shewanella [68,69], Serratia [70,71], Carnobacterium [72], Brochotrhix [73,74], and Kluyvera [75,76]. All of these bacteria can colonize multiple environments, and the ability to acquire resistance in any given environment represents a significant risk for human health. Only Aeromonas molluscorum [77,78] and Psychrobacter nivimaris [79] are exclusive to the aquatic environment, which translates into a highly probable exchange of ARGs between bacteria directly related to the human system and environmental bacteria from the gut microbiota of fish.
Tetracyclines are commonly used in veterinary medicine around the world, and, specifically, oxytetracycline is the most widely used antibiotic during the freshwater stage of salmon growth in Chile [25]. Origin records for the Atlantic salmon sampled in the present study were not obtained, but it is very likely that the sampled fish were exposed to high doses of oxytetracycline during the freshwater stage. This would explain the high instance of oxytetracyclineresistant bacteria, even when oxytetracycline was not applied during the seawater stage. Similarly, the high prevalence of tet genes found for the FB and OB could be explained by the high frequency at which these genes localize in mobile genetic elements, such as plasmids or transposons [49]. The genes tetA, tetB, tetE, tetL, tetH, tet34, and tet35 have been recorded in the sediments and water column of Chilean salmon farms [43,80]. An interisting case is the gene tetM, this globally distributed gene presents a clinically confirmed resistance mechanism to tetracyclines [81,82]. One trait of this gene is that it is frequently associated with transposons and plasmids in human pathogenic bacteria [83,84]; this could be a clear sign of the impacts this gene is having from the clinical to the environmental spheres. Consequently, antibiotics used in industrially farmed animals could substantially contribute to the maintenance and dispersion of ARGs in natural environments.
As with the tet genes, the presence of floR, a drug efflux pump of the major facilitator superfamily (MFS), has previously been reported in bacteria collected from marine sediments associated with salmon farming in Chile [39]. This gene has even been detected in a plasmid that also transported tet and qnr genes, the latter of which are implicated in resistance to quinolones [85]. However, the present study is the first to report on the existence of floR in the gut microbiota of salmon. Furthermore, this investigation is the first to find and report on fexA, an efflux pump of the major facilitator superfamily also, in association with Chilean salmon farming, with no previously published study having detected this gene in relation to the Chilean aquaculture industry. This fenicol-resistance gene was first described in the bacterium Staphylococcus lentus [40], and has since almost been exclusively described in members of the Staphylococcus genus [41,42,51,[86][87][88], both for humans and animals, but never in the marine environment. Apart from the Staphylococcus genus, fexA has been described in members of the Enterococcus genus, specifically in association with industrially farmed terrestrial animals [89,90]. No Staphylococcus or Enterococcus bacteria were found among any of the obtained resistant isolates, but the presented findings do support the ability of fexA to disperse across a wide variety of both Gram-positive and-negative bacteria.
One curious outcome of the conducted study was that none of the oxytetracycline-resistant bacteria presented floR, but, by contrast, 52,2% did present fexA. As mentioned, the most widely used antibiotic used in Chilean salmon farming is oxytetracycline, and there is a nearly null use of florfenicol in the freshwater stage. Prior reports have found fexA in plasmids containing a great number of resistance genes to other antibiotics [87]. Therefore, this gene has probably been selected during the freshwater stage in plasmids that also carry tetracylcin-resistance genes. This co-selection is likely since the water used in the freshwater rearing stage is collected from the lakes and rivers most exposed to human activities, including livestock farms, and, as stated, fexA has almost exclusively been isolated from industrially farmed terrestrial animals. This is another compelling case for ARGs primarily originating from human activities and not directly from salmon farming.
In contrast to findings from sediments related to intense salmon farming in Chile [91], the present study found a high incidence of class I integrons in resistant isolates from both the FB and OB databases. In fact, 100% of OB isolates and 96% of FB isolates were positive for type I integrase. These genetic elements are the greatest contributors to the evolution and dispersion of ARGs [92]. Integrons are true genetic platforms that acquire exogenous genes through mobile genetic cassettes, having the ability to collect up to hundreds of genes in the same unit [93]. Nevertheless, these elements did not originate in bacterial pathogens, nor is their primary function to collect antibiotic-resistance genes. Integrons frequently appear in Betaproteobacteria native to soils, freshwater, and saltwater. Genetic cassettes are incorporated alongside elements that are highly varied and much more diverse than ARGs [94]. Integrons are very successful and ubiquitous in natural environments, and it is therefore unsurprising that the majority of the isolated OB and FB bacteria would present these elements. Concern arises when these elements have a clinical origin and carry ARGs for most of the known antibiotics [95]. Integrons carrying ARGs against quinolones (qnrA and qnrB), trimethoprims (drfA12), and aminoglycosides (aad2) have recently been described in the Chilean farming system, which is in addition to some clinical Escherichia coli also carrying these plasmid-transferred integrons [96]. This appears to be credible proof that there is an interaction between the clinical environment and that of salmon farming. However, a number of questions remain-What is the origin of these integrons? Are they from aquaculture-associated bacteria? Or have contributions from the clinical environment contaminated bacteria from salmon farming? These questions need to be answered. No integron from either the OB or FB bacteria presented the traits described by Tomova et al. (2018) [90].
Another important aspect to consider in this work is the Antibiotic Resistance Rate (ARR). Overall, it was found that less than 1% of the total bacterial isolates were resistant to oxytetracycline or florfenicol (0.94% for florfenicol and 0.87% for oxytetracycline). Although there are few works in which this rate is addressed at an environmental level [97,98], values close to 1% can be considered high [97][98][99][100]. In these works it can be seen that when the concentration of antibiotics in the medium increases, the value of this range increases too. This phenomenon is corroborated in the present work since the ARR in the FB for farm II exceeds 1%, almost twice as much as that presented by the other farms and it is precisely the farm II that the major number of treatments with florfenicol presents, four. The same effect is observed in the OB in farm I where ARR again exceeds 1% and the farm I is the only one that uses oxytetracycline as a treatment in its seawater stage.
The information collected in the present study clearly indicate that using large amounts of antibiotics to treat industrially farmed animals has the consequence of selecting for multiresistant bacteria in the intestinal microbiota, which then have undeniable advantages when liberated into the environment through the feces. While the obtained results are somewhat biased towards "culturable" bacteria, metagenomics data are currently being analyzed, the results of which will reveal all of the bacteria that did not grow in a TSA medium. This information will provide a complete picture for the behaviors of the bacterial populations interacting in this complex system. Meanwhile, the currently reported data do show that the intestinal microbiota, as a system "semi-isolated" from the environment yet in direct contact with applied antibiotics, is the perfect place for interactions to occur between bacteria from different environments, including from the human clinical context. Therefore, the gut microbiota of farmed salmon could also serve as the perfect reservoir for ARGs, the dispersion of which would be fundamentally related to feces. The release of feces with resistant bacteria and ARGs is creating an environment constantly rich in resistance elements, creating an ideal situation for genetic exchanges to occur between different bacterial populations, whether in the water column or in marine sediments. These populations are precisely where fish pathogens can be found, thus notably increasing the risk that these bacteria will acquire resistance elements. If this occurs, no effective treatments will exist for the control of these pathogens. Fortunately, a number of aquaculture producers are beginning to understand the gravity of the situation. Measures aimed at decreasing antibiotic usage are increasingly being adopted, with expectations that by 2020, there will be at least a 50% decrease in the use of these substances.

Conclusion
In this study, bacterial isolates from fish gut microbiota of the Salmo salar species, showing high resistance levels to antibiotics oxytetracycline and florfenicol, were characterised. The feces, represent important vehicles of dispersion of resistant bacteria and ARGs. The analyses showed that the high use of these antibiotics selects bacteria that are multiresistant to a wide range of antibiotics. In turn, these bacteria show a high prevalence of antibiotic resistance genes, thus, verifying the phenotypic resistance with the mechanism present in the bacteria to cope with these substances. Furthermore, almost 100% of these (antibiotic) resistant isolates showed class 1 integrons, dangerous elements involved in the resistance phenomenon due to their capacity of dispersion.
Supporting information S1