Bacterial Diversity and Antibiotic Susceptibility of Sparus aurata from Aquaculture

In a world where the population continues to increase and the volume of fishing catches stagnates or even falls, the aquaculture sector has great growth potential. This study aimed to contribute to the depth of knowledge of the diversity of bacterial species found in Sparus aurata collected from a fish farm and to understand which profiles of diminished susceptibility to antibiotics would be found in these bacteria that might be disseminated in the environment. One hundred thirty-six bacterial strains were recovered from the S. aurata samples. These strains belonged to Bacillaceae, Bacillales Family XII. Incertae Sedis, Comamonadaceae, Enterobacteriaceae, Enterococcaceae, Erwiniaceae, Micrococcaceae, Pseudomonadaceae and Staphylococcaceae families. Enterobacter sp. was more frequently found in gills, intestine and skin groups than in muscle groups (p ≤ 0.01). Antibiotic susceptibility tests found that non-susceptibility to phenicols was significantly higher in gills, intestine and skin samples (45%) than in muscle samples (24%) (p ≤ 0.01) and was the most frequently found non-susceptibility in both groups of samples. The group of Enterobacteriaceae from muscles presented less decreased susceptibility to florfenicol (44%) than in the group of gills, intestine and skin samples (76%). We found decreased susceptibilities to β-lactams and glycopeptides in the Bacillaceae family, to quinolones and mupirocin in the Staphylococcaceae family, and mostly to β-lactams, phenicols and quinolones in the Enterobacteriaceae and Pseudomonadaceae families. Seven Enterobacter spp. and five Pseudomonas spp. strains showed non-susceptibility to ertapenem and meropenem, respectively, which is of concern because they are antibiotics used as a last resort in serious clinical infections. To our knowledge, this is the first description of species Exiguobacterium acetylicum, Klebsiella michiganensis, Lelliottia sp. and Pantoea vagans associated with S. aurata (excluding cases where these bacteria are used as probiotics) and of plasmid-mediated quinolone resistance qnrB19-producing Leclercia adecarboxylata strain. The non-synonymous G385T and C402A mutations at parC gene (within quinolone resistance-determining regions) were also identified in a Klebsiella pneumoniae, revealing decreased susceptibility to ciprofloxacin. In this study, we found not only bacteria from the natural microbiota of fish but also pathogenic bacteria associated with fish and humans. Several antibiotics for which decreased susceptibility was found here are integrated into the World Health Organization list of “critically important antimicrobials” and “highly important antimicrobials” for human medicine.


Introduction
In the last 30 years, global production of the aquaculture sector has increased and today represents nearly half of the fish consumed worldwide, with China as the main producer. This was a consequence of higher demand due to the reduction or stagnation of fishing catches and an increasing world population [1].
Portugal is the third major consumer of fish, in Europe [2]. Although aquaculture represents a small portion of this consumption, this sector has grown in the last decades and is expected to increase in the coming years. This is a country with favorable conditions for aquaculture, where the main production is of bivalve mollusks, and extensive and intensive systems are predominant. Tanks for fish production represent about 5% of all Portuguese aquaculture infrastructures (in 2012), and turbot (Scophthalmus maximus), European seabass (Dicentrarchus labrax) and gilthead seabream (Sparus aurata) are main species produced [3]. S. aurata is a species belonging to the Sparidae family and Perciformes order [4] and is often found in shallow waters and is sensitive to water temperature. The gilthead seabream has great commercial importance in Europe, representing one of the main species cultivated in this continent [5].
However, despite all the advantages, aquaculture production can have a negative impact, specifically in degradation of natural resources and rising of antibiotic use [1,6]. A higher density of fish in a specific area is usually related to an increase in stress conditions, leading to a predisposition for infectious diseases and a higher antibiotic consumption. Therefore, these environments can function as a reservoir of antibiotic resistance and/or antibiotic-resistant genes. In Europe, antibiotics for growth promotion are not authorized, and only eight antibiotics are allowed for prophylaxis and therapeutics: ampicillin, oxytetracycline, florfenicol, flumequine, oxolinic acid, sarafloxacin, erythromycin and sulphonamides associated with trimethoprim or ormetoprim. These antibiotics are mostly administered by oral-medicated feed or bath, which are the easiest methods to apply but expose both sick and healthy individuals to antibiotics and allow the accumulation of these substances in sediments and water. This contributes to selective pressure in bacteria from these environments [7,8].
In this contexts, this study aimed to contribute to a deeper knowledge of the diversity of bacterial species found in S. aurata from aquaculture and the respective antibiotic susceptibilities that can be disseminated in their environment. The correlation of antibiotic resistance with the World Health Organization (WHO) list of "critically important antimicrobials" and "highly important antimicrobials" for human medicine will also be established.

Sample Collection and Preparation
Five commercial-size S. aurata (500-1500 g) were collected in March 2018 in a land tank from a fish farming pilot station in the south of Portugal by the Portuguese Institute of Sea and Atmosphere. This station is located in the Ria Formosa Natural Park and is an integrated multi-trophic aquaculture with a semi-intensive system. The weight, furcal length, total length, and condition index of the 5-gilthead seabream were measured (Table 1). Each fish was divided into 4 samples (gills, intestine, muscle, and skin), that were frozen and transported on ice to the National Institute of Health Dr. Ricardo Jorge, where they were analyzed. This study includes the results of the testing for 5 samples of muscle (from fish 1, 2, 3, 4 and 5) and the gills, intestine and skin samples from fish 1 that were treated separately (excepting for some results presentation) ( Figure S1).

Bacterial Isolation and Identification
Ten grams of each sample was homogenized in peptone water (Stomacher 80 Biomaster ® , Seward, UK), incubated for 12 to 18 h at 37 • C and further diluted [10,11]. Each dilution was plated in selective media (MacConkey agar, Mannitol salt agar and UriSelect™4 chromogenic agar) and incubated for 18 to 20 h at 37 • C. Colonies with different morphology (to avoid duplications) were selected and DNA extracted, according to manufacturer's instructions (MagNA Pure 96 Instrument, Roche, Manheim, Germany). Strains were identified by VITEK2 and amplification of the 16S rRNA gene, as already described [12].

Antibiotic Susceptibility Testing
Antibiotic susceptibility testing was performed by disk diffusion (Bio-Rad, Marnes-la-Coquette, France) and minimum inhibitory concentration (MIC) by in-house broth microdilution and E-test®(bioMérieux, Marcy l'Etoile, France). Different bacterial families were tested for different antibiotics (Table 2).
For Gram-negative bacteria, the antibiogram was completed with (1) disc combination test (DCT or combined disk test, CDT) [13], which is based on the comparison between zone diameters of one disc of antibiotic alone and another with an inhibitor, here using cefotaxime (30 µg) and cefotaxime/clavulanic acid (30 µg + 10 µg) to search for the presence of extended-spectrum β-lactamase (ESBL-positive if there is a difference of ≥ 5 mm); (2) DCT to compare zone diameters of meropenem and meropenem/dipicolinic acid (1000 µg), to search for the presence of Metallo-β-lactamase (MBL-positive if ≥ 5 mm) [14]; (3) DCT with amoxicillin and amoxicillin/clavulanic acid plus cloxacillin (500 µg), to search for the presence of AmpC (positive for cephalosporinases if ≥ 5 mm) [13]; (4) double disc synergy test (DDST) with two discs containing predefined amounts of the β-lactam and the inhibitor, placed close to each other, here using boronic acid (300 µg) and carbapenems to search for any class A carbapenemase (positive if a synergy is observed between carbapenemes discs) [14]; (5) DDST to search for MBL when a synergy is observed between dipicolinic acid and carbapenemes [14]; (6) DDST to detect AmpC when a synergy is observed between boronic acid and third-generation cephalosporins and/or cloxacillin and cefoxitin/ceftazidime [13]; (7) temocillin disc to indicate the presence of an OXA-48 carbapenemase; and (8) faropenem disc to indicate the presence of carbapenemases, confirming the results from (4), (5) and (7) tests.  MIC 50 and MIC 90 were calculated for Enterobacteriaceae since it was the most represented family, as reported elsewhere [15]. MIC 50 represents the MIC value that inhibits 50% of the strains tested (and is equivalent to the median MIC value), whereas MIC 90 represents the MIC value that inhibits 90% of the strains tested. Escherichia coli strain ATCC 25922 was used as quality control for Gram-negative bacteria, whereas Staphylococcus aureus strain ATCC 25923 and Enterococcus faecalis ATCC 29212 were used for Gram-positive bacteria.

Statistical Analyses
Statistical analyses of the results were performed to detect positive or negative associations between fish samples (muscle vs. gills, intestine and skin) and each bacterial family/species and non-susceptibility to different antibiotic's class (only factors identified as statistically significant are shown). Fisher exact test was used to assess differences in bacterial families/species/non-susceptibility to different antibiotic's class between fish samples, and one-tailed p-values of ≤ 0.05 were considered to be statistically significant. Associations were established by calculation of odds ratios with 95% confidence intervals. The null hypothesis was rejected for p-values of ≤ 0.05. All statistical analyses were calculated using OpenEpi software, v. 3.01 [16].

Detection of Antibiotic Resistance Genes
The interpretation of the antibiotic susceptibility testing results guided the research of resistance genes by PCR (Polymerase Chain Reaction), and all the positive results were sequenced, as described elsewhere [12].
Apart from the antibiotic susceptibility testing results, all Staphylococcus spp. were tested for the presence of mecA, mecC, vanA, vanB and vanD genes ( Table 3) [33]. On the other hand, all Enterococcus spp. and three Bacillus spp. resistant to vancomycin were investigated for the presence of vanA, vanB and vanD genes. The cycling conditions for the PCR multiplex for detection of vanA, vanB and vanD genes were as follows: 1 cycle of denaturation at 94 • C for 10 min, followed by 30 cycles of denaturation at 94 • C for 30 s, annealing at 55 • C for 1 min and elongation at 72 • C for 1 min, and a final cycle of elongation at 72 • C for 10 min.

Bacterial Diversity in S. aurata Samples
One hundred thirty-six bacterial strains were recovered from the total of S. aurata samples. Eighty-eight were Gram-negative bacteria, and 48 were Gram-positive bacteria. The results of VITEK2 and amplification of the 16S rRNA gene revealed that the majority of strains belonged to the Enterobacteriaceae family (55% in muscle samples and 60% in gills, intestine and skin samples), followed by Staphylococcaceae in muscle samples (16%) and Bacillaceae in gills, intestine and skin samples (17%) ( Table 4). Bacillales Family XII. Incertae Sedis, Comamonadaceae and Micrococcaceae families were only found in muscle samples, whereas Erwiniaceae family was only found in gills samples. Nevertheless, these results do not represent statistically significant differences. Finally, Erwiniaceae family included Pantoea vagans species. Despite appearing in both groups of samples, Enterobacter sp. are more frequently found in gills, intestine and skin group than in muscle group (p ≤ 0.01; Table 5, Table S1). Statistically significant associations were not found for the other species analysed. Table 5. Odds ratio (OR) and 95% confidence intervals (CI) (p ≤ 0.05) from the analysis of negative and positive correlations between fish samples (muscle vs. gills, intestine and skin) and each bacterial species and non-susceptibility to different antibiotics classes.

Phenotypic Characterization of the Bacterial Strains
In this study, non-susceptibility to phenicols was the most frequently found in both groups of samples, being significantly higher in gills, intestine and skin samples (45%; Table 6) than in muscle samples (24%) (p ≤ 0.01; Table 5). Non-susceptibility to β-lactams is the second most prevalent in both groups of samples (13% in muscles and 10% in gills, intestine and skin samples; Table 6). Decreased susceptibility to quinolones was also found in this study (7% in muscle and 5% in gills, intestine and skin samples). On the other hand, decreased susceptibility to glycopeptides and to mupirocin was only found in muscle samples.
Among Enterobacteriaceae strains, no differences were observed in MIC 50 and MIC 90 between the two groups of samples for the four antibiotics tested (Table 7). For each antibiotic, slight differences were registered between MIC 50 and MIC 90 (just 1-fold dilution). When comparing the decreased susceptibilities between the two groups of samples (muscle vs. gills, intestine and skin), major difference were found only in florfenicol, with the group of Enterobacteriaceae from muscles presenting 44% of nonsusceptible strains, while in the group of gills, intestine and skin samples, the values were higher, with 76% (Table 7). Table 6. Antibiotic susceptibility testing results of the 136 strains found in this study (these results do not include known intrinsic non-susceptibilities).  In Table 8 is registered the decreased susceptibility profile for the 61 strains that had a non-susceptibility result for at least one antibiotic (including intrinsic non-susceptibilities). In Gram-positive bacteria, we found decreased susceptibilities to β-lactams and glycopeptides in the Bacillaceae family and to quinolones and mupirocin in the Staphylococcaceae family. In Gram-negative bacteria, Enterobacteriaceae family showed several decreased susceptibility profiles, with non-susceptibilities to β-lactams, phenicols and quinolones; the same non-susceptibilities to these antibiotic classes were found in strains from the Pseudomonadaceae family. The non-susceptibility to ertapenem of seven Enterobacter spp. strains and to the antipseudomonal carbapenem (meropenem) of five Pseudomonas spp. was found.

Genotypic Characterization
The search for resistance genes by PCR revealed a qnrB19 gene in a L. adecarboxylata strain, with decreased susceptibility to ciprofloxacin (zone diameter = 24 mm). The analyses of the mutations present in gyrA, gyrB, parC and parE genes of a K. pneumoniae strain with decreased susceptibility to ciprofloxacin (zone diameter = 25 mm) showed the presence of non-synonymous mutations G385T (Ala129Ser in protein) and C402A (Ser134Arg in protein) in parC gene. No other resistance genes were identified in the studied bacteria among all the searched genes.

Discussion
The aquaculture sector has experienced a strong growth in recent decades and with it the number of studies to answer some concerns about the quality and safety of its products. Some studies have focused on the search of antibiotic residues in waters and/or sediments from fish farms, while others have focused on the search for specific pathogens, such as E. coli, not only in water and/or sediments but also in fish and shellfish [34][35][36][37][38][39]. This study aims to provide data, not only on the bacterial diversity in S. aurata from aquaculture, but also on antibiotic resistant genes that circulate in these environments. This information is crucial in the construction of science-based policies for the suitable use of antibiotics in aquaculture.
In this study, we found a very diverse bacterial population with 31 species that belonged to nine different families. Microbiota present in fish depends not only on the fish genetics and diet but is also determined by microbiota present in their environment, such as water and sediments. Microbiome composition is normally different between individual fish belonging to the same species but also varies between healthy and sick individuals, wherein the healthy individuals seem to have a higher diversity of bacterial population [40]. The condition index of fish belonging to this study varied between 1.56 and 2.04 (Table 1), meaning that these S. aurata were healthy (values higher than 1), possibly explaining the diversity found.
Of the species found in this study, only Pseudomonas spp. and K. rhizophila appeared to frequently cause diseases in fish, K. rhizophila being an emerging pathogen [41,42]. Pseudomonas spp. are ubiquitous in nature, and P. fluorescens is the most important species in fish infections (not found in this study), although P. putida had already been found in internal organs of fish and P. stutzeri in sediments of marine waters [43]. This genus is responsible for strawberry disease and septicaemia in some fish species [42]. P. putida and P. stutzeri are also opportunistic pathogens in humans and were already associated with bacteraemia, endocarditis, keratitis, meningitis, pneumonia, skin and soft tissue infections and urinary tract infections [44][45][46][47][48]. Kocuria spp. are Gram-positive and coccoid bacteria isolated from numerous environments: skin of mammals, marine sediments, soil and food [49][50][51][52][53]. Specifically, K. rhizophila causes a variety of lesions in fish and is responsible for a 50% mortality rate: exophthalmia, skin petechiae, increased skin melanization, liver congestion, inflammation of the intestine and hemorrhages [42]. Reports revealed that this species was already responsible for some infections in humans, mainly catheter-related bacteremia [54].
There is a single report about the association of L. adecarboxylata with fish, more specifically in the oral cavity of sharks [70]. However, this species is frequently found in water environments. It can cause endocarditis, bacteremia and peritonitis in human hosts [71,72]. Some studies indicate that E. hirae could be a part of the natural microbiota of fish. There is no literature involving this bacterium in fish infections, but this species was well characterized in human infections, such as endocarditis, pyelonephritis, acute pancreatitis and septic shock, representing 1 to 3% of infections caused by Enterococcus spp. in clinical practice [73].
Together with E. coli, E. faecalis is also an indicator of fecal contamination [87,88]. Species like Bacillus spp., E. acetylicum and Enterococcus spp. can be used as probiotics in aquaculture to reduce antibiotic consumption and avoid the spread of antibiotic resistance genes [66,89,90]. We could not obtain information if this was the case of this fish farming pilot station, in Portugal, possibly justifying the presence of these bacteria in our samples.
β-Lactam antibiotics are used in aquaculture in many countries, especially amoxicillin [7]. Several studies revealed a high prevalence of non-susceptibility to this class of antibiotic in fish, shrimps and water samples from aquaculture farms. These studies reported the discovery of bla SHV and bla TEM genes, associated with β-lactam non-susceptibility [91][92][93][94]. In our study, β-lactam resistance genes were not detected. The non-susceptibilities found here were probably related to genes or other resistance mechanisms not studied (e.g., efflux pumps). Likewise, some of the non-susceptibilities found to β-lactams are intrinsic, such as the resistance to ERT in Pseudomonas spp. [95]. Non-susceptibility to AMC, FOX, AZT, FEP, CTX, CAZ and PTZ in Enterobacter spp., E. coli and Citrobacter spp. could be explained by chromosomal AmpC β-lactamase or AmpC hyperproduction [96]. Worryingly, our study revealed decreased susceptibilities to carbapenems of Enterobacter spp. (ertapenem), usually associated with acquired resistance, as well as in Pseudomonas spp. (meropenem), which are not used in aquaculture and are considered last-resort antibiotics to treat serious infections caused by multidrug-resistant bacteria in humans [97].
Interestingly, although banned for use in food-producing animals in many countries (since 1994 in Europe; [98]), decreased chloramphenicol susceptibility continues to be described, not only in the present study (with MIC of 32 to > 64 mg/l), but also in others, with the presence of cat genes [87,99]. Chloramphenicol can occur naturally, produced by soil organisms such as Streptomyces venezuelae. The persistence of decreased susceptibility to this antibiotic may be due to intrinsic mechanisms of resistance as possibly in P. putida and P. stutzeri from our study or co-selection with other antibiotics and/or heavy metals [98,100,101]. On the other hand, florfenicol is widely used in aquaculture [8]. In a review article, Miranda et al. [102] compiled several studies in Chilean salmon farms that revealed a high frequency of decreased susceptibility to florfenicol, such as in our samples. However, there are further studies that described lower rates [103,104].
Another antibiotic class commonly used in aquaculture is quinolones, specifically flumequine and oxolinic acid [6,7]. In this study, we found low frequency of decreased susceptibility to quinolones (7% in muscles and 5% in gills, intestine and skin group). These results are confirmed by other works in S. aurata samples [38] and other fish species [105], while others revealed higher frequencies, also in fish samples [106]. Of the nine strains with decreased susceptibility to quinolones in this study, we found genes that justify that resistance in only two; the others may have untested resistance mechanisms. qnrB genes are frequently found in aquaculture environments [107,108], but to our knowledge, this is the first description of qnrB19 gene in an L. adecarboxylata strain. This strain had decreased susceptibility to ciprofloxacin and was susceptible to flumequine (MIC = 4 mg/L). The qnrB19 gene had already been reported in food-producing animals, such as pigs, poultry and veal calves, sometimes associated with mobile genetic elements, like plasmids and insertion sequences, demonstrating a potential for spreading [26,109]. Moreover, qnrB19 gene was also found in Enterobacteriaceae from human clinical samples [110][111][112]. K. pneumoniae strain with decreased susceptibility to ciprofloxacin and susceptibility to flumequine (MIC = 1 mg/L) revealed two non-synonymous mutations in parC gene within the quinolone-resistance-determining regions (QRDR), one of them (G385T) already described in a K. pneumoniae strain from nosocomial origin [113]. It is known that mutations in QRDR of parC gene can result in structural changes in topoisomerase IV, reducing the affinity of this enzyme to fluoroquinolones [114].
Vancomycin and mupirocin are not used in aquaculture [7]. Decreased susceptibilities to these antibiotics were found only in muscle samples in this study. Bacillus spp. resistance to vancomycin probably does not represent an intrinsic resistance, since several studies indicated the susceptibility of these bacteria to this glycopeptide [67]. The search for vanA, vanB and vanD genes was negative, so this non-susceptibility could be explained by the presence of other genes not studied, like vanC, vanE and vanG, or mutations in walKR, vraRS and graRS genes involved in cell wall metabolism and cellular response to cell wall damage [115]. To our knowledge, this is the first description of Bacillus sp. strains resistant to vancomycin in S. aurata from aquaculture. This resistance was already found in this environment but in Enterococcus species [87]. Acquired reduced susceptibility to mupirocin in S. petrasii could be related to mutations in the ileS gene and, more rarely, to the acquisition of plasmid-mediated mupA gene, like in S. aureus [116]. There are no data available about the frequency of detection of decreased susceptibility to mupirocin in aquaculture.
As we demonstrated, some antibiotics used in food-producing animals are the same, or belong to the same class, as antibiotics used in humans [117]. Several antibiotics for which decreased susceptibility was found in this study are integrated into the WHO list of "critically important antimicrobials" and "highly important antimicrobials" for human medicine. This means that for some bacterial infections, these antibiotics are the only therapy available, and that they are also used in humans to treat infections caused by bacteria/resistance genes that may be transmitted from non-human sources (namely food-producing animals) [118].

Conclusions
This study unraveled some information about bacterial diversity and antibiotic resistance genes that circulate in S. aurata raised in aquaculture.
It is noteworthy that the difficulty of characterizing the strains as susceptible or non-susceptible due to be the absence of breakpoints to some of the species found in these aquatic environments and to antibiotics commonly used in aquaculture. Further research is needed in this area, also in relation to intrinsic resistances since opportunistic environmental bacteria can represent a health threat and a reservoir of antibiotic-resistant bacteria/resistance genes.
In the current study, we can observe protective associations in muscle samples, while positive correlations were found in gills, intestine and skin group of samples. Indeed, being part of the gills, intestine and skin group represents a risk factor for the presence of Enterobacter sp. and non-susceptibility to phenicols.
Antibiotics are used in humans and aquaculture to treat infections caused by bacteria. Furthermore, it was already described that antibiotic residues can remain in fish tissues for long periods of time, increasing exposure of commensal bacteria and/or fish pathogens, along with aquatic bacteria, to these antibiotics, enhancing the development of resistance [8]. Therefore, it is fundamentally a "One Health" approach, combining the efforts of human and veterinary medicine, along with agriculture sector, to reduce the spread of bacterial resistance and/or bacterial resistance genes.
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-2607/8/9/1343/s1, Figure S1: Diagram summarizing the experimental design; Table S1: Odds ratio (OR) and 95% confidence intervals (CI) (p ≤ 0.05) from the analysis of negative and positive correlations between fish samples (muscle versus gills, intestine and skin) and each bacterial species and non-susceptibility to different antibiotic's class (detailed results).

Conflicts of Interest:
The authors declare no conflict of interest.