The global world fish market reached 175 million tons in 2016 (Food and Agriculture Organization (FAO), 2016). There is a growing concern in the fisheries sector about the increasing prevalence of multidrug-resistant bacterial strains in the food chain (EFSA, 2010; ECDC, 2012, 2013; EFSA-ECDC, 2013; Watts et al., 2017). Multidrug-resistant bacteria carrying extended-spectrum (ESBL) β-lactamase genes can now be found in healthy humans as well as in various animal species, food and even in environmental samples, and have become a threat in hospital settings. Carbapenemase-producing Enterobacteriaceae are particularly of concern because they tend to spread, making infection treatment difficult (Iovleva and Doi, 2017). Resistance genes for antimicrobials such as β-lactams, quinolones, and fluoroquinolones may be associated with plasmids. Plasmids of fluoroquinolone resistance in Enterobacteriaceae are widespread, and are often associated with resistance to β-lactams in strains with multidrug resistant phenotypes (Crémet et al., 2011). Some genes, such as the variant aac(6′)-Ib-cr simultaneously confer resistance to aminoglycosides, and fluoroquinolones (Kim et al., 2011). Others like the efflux pump oqxAB described in Escherichia coli, confer resistance to various antibiotics and biocides (Hansen et al., 2007; Wong and Chen, 2013).

Different types of selective pressure (such as antibiotics, biocides, or heavy metals) could play a role in the prevalence of antimicrobial resistance in the food chain. Biocides may co-select strains resistant to antibiotics of clinical use, as verified in the case of triclosan, and others (Chuanchuen et al., 2001; Braoudaki and Hilton, 2004; Ortega-Morente et al., 2013). Biocides are widely used in disinfecting equipment and facilities, including fish farming, and fish processing facilities (Directive 98/8/CE). In Gram-negative bacteria, several genes for resistance to biocides belonging to the group of quaternary ammonium compounds (QACs) as qacE, qacEΔ1, qacF, qacG, and qacH have been described. Among them, qacEΔ1 is the most widespread, as it is found in various groups of Gram-negative bacteria (Kücken et al., 2000). Association of qacEΔ1 (and less frequently also qacH and qacF) with class I integrons together with antibiotic resistance genes has been reported (Mulvey et al., 2006; De Toro et al., 2011). A study of microbial populations in environments contaminated with QACs related the increased incidence of resistance to QACs with a higher incidence of class I integrons (Gaze et al., 2005), suggesting that exposure to QACs could also selected for antibiotic resistance associated with such integrons. Other antimicrobials such as lysozyme-EDTA combinations and chemical preservatives such as sodium lactate and trisodium phosphate also deserve attention because of their wide use and potential applications in the food industry for decontamination and food preservation (Lucera et al., 2012; Bjornsdottir-Butler et al., 2015; Wang et al., 2015). There is also a growing interest in extending the use of plant essential oils or their antimicrobial compounds (such as carvacrol or thymol) for disinfection and food preservation (Lucera et al., 2012; Patel, 2015; Wang et al., 2015). One study showed that exposure to pine oil induced a decreased susceptibility to a range of antimicrobial compounds (including antibiotics and biocides; Moken et al., 1997). Yet, little is known about development of resistance/tolerance to these compounds, and co-resistance to antibiotics.

Metal salts are used for decontamination in fish farming. Metals such as copper (Cu) and zinc (Zn) are essential micronutrients in living things, but they can become toxic if they are above a certain concentration. Copper and zinc are frequently used in aquaculture and also as antifouling paints on boats (Yebra et al., 2004; Watermann et al., 2005; Guardiola et al., 2012). Copper can be released into the environment by both human activities and natural processes. Zinc is rarely found in nature in its metallic state, but many minerals contain zinc as a main component. The main anthropogenic sources of zinc are mining, zinc production facilities, iron, and steel production, corrosion of galvanized structures, coal, and fuel combustion, waste disposal and incineration, and the use of fertilizers, and pesticides containing zinc (WHO, 2001). Fish can accumulate and transmit heavy metals along the food chain. The genes that control resistance to metals may be associated with plasmids, which provide bacteria a competitive advantage over other organisms when specific metals are present (Trevors et al., 1985; Hobman and Crossman, 2014). The appearance of multidrug resistance plasmids carrying resistance to heavy metals is alarming and requires additional monitoring (Gómez-Sanz et al., 2013). The co-location of metal-resistance and antimicrobial-resistance genes can facilitate their persistence, co-selection, and dissemination (Gómez-Sanz et al., 2013; Wales and Davies, 2015).

The aim of this study was to provide insights on resistance to clinically relevant antibiotics in bacterial strains isolated from seafood sold at supermarkets and fishmarket. Since exposure to other types of antimicrobials such as biocides and metals could co-select for antibiotic resistance, we hypothesized that antibiotic-resistant strains from seafoods could also be phenotypically resistant, and/or carry genetic determinants of resistance to other antimicrobials, thus increasing the risk of transmission of antibiotic resistance through the seafood production chain. Therefore, susceptibility of bacterial strains isolated from different seafood samples against various biocides, antibiotics, and metals was evaluated, and the presence of resistance genes in multiresistant strains was determined. A special emphasis was made on extended-spectrum β-lactamase genes because of the risks that ESBL-producing bacteria may pose to human health.

In the present study, biocide tolerance and antibiotic resistance were detected among bacteria isolated from seafoods. Biocides are used for many different purposes, including health care products and in disinfection processes in the food industry. For example, benzalkonium chloride is used for water treatment, general site disinfection, fish parasite removal, and prevention of infectious disease in fish and shellfish. As a result, large amounts of biocides arrive to waters. The impact of triclosan on aquatic bacterial communities has been described (Dann and Hontela, 2011; McNamara et al., 2014). Previous contact with biocides as well as natural background resistance could explain the biocide tolerances observed in the present study. It is also worth noting that there were positive correlations not only for tolerance to biocides of the same chemical group but also between biocides from different groups. However, there were differences between polyguanides, since poly-(hexamethylen guanidinium) hydrochloride showed positive correlation with several other biocides while chlorhexidine did not. These results could be explained by differences in chemical formula, mechanisms of adaptation (including intrinsic resistance), and also by the development of specific mechanisms of tolerance upon exposure to multiple biocides.

A relatively high percentage of isolates were resistant to at least one biocide and at least one antibiotic, and there were significant (P < 0.05) positive correlations between biocide tolerance and antibiotic resistance. These results reinforce the general concern that the use of biocides may co-select for antibiotic resistance (SCENIHR, 2009; Ortega-Morente et al., 2013; Wales and Davies, 2015). Cross-resistance between antibiotics and biocides and between different biocides has been reported for different bacteria, like for example Pseudomonas aeruginosa (Lambert et al., 2001; Lavilla Lerma et al., 2015). Furthermore, previous studies have shown that adaptation to biocides by repeated exposure results in an increased resistance to antibiotics (Gadea et al., 2016, 2017a). It is worth noting that there were also some positive correlations between tolerance/resistance to biocides, antibiotics, and other antimicrobials. For example, the phenolic biocide triclosan showed positive correlation with a lower sensitivity to the phenolic compound thymol (but not with carvacrol), and the antibiotic ampicillin also showed positive correlation with thymol. Previous studies have shown that exposure to plant essential oils (which are rich in phenolic compounds) such as pine oil resulted in the selection of mutants with deregulated mar operon that had decreased susceptibility to a range of antimicrobial compounds (including antibiotics and biocides) as a consequence of reduced cell permeability and increased efflux pump activity (Moken et al., 1997; Ortega-Morente et al., 2013). It is worth mentioning that the chemical preservatives sodium lactate and trisodium phosphate only showed moderate positive correlations with the antibiotics ampicillin and imipenem, but not with biocides. A recent study indicated that bacteria adapted to quaternary ammonium compounds under laboratory conditions showed a generalized increased tolerance to preservatives (such as 4-hydroxybenzoic acid, thyme, and clove oil, sodium, and potassium nitrates, potassium sorbate), while the opposite was observed in the case of triclosan (Gadea et al., 2017b).

Sulfonamides potentiated with trimethoprim or ormethoprim and florfenicol are some of the antibiotics commonly used in aquaculture (Hernández Serrano, 2005). In the present study, the sulfonamide resistance sul1 gene was the genetic determinant detected most frequently. Sulfonamide resistance is most often linked to Class I integrons. These mobile genetic elements tend to accumulate different antibiotic resistance genes and also biocide tolerance genes like qacEΔ1. However, qacEΔ1 was detected in combination with sul1 only in the A. calcoaceticus isolates, and PCR amplification with a forward primer for qacEΔ1 and a reverse primer for sul1 did not suggest a close physical location of the two genes unless they were in opposite direction. Class 1 resistance integrons are located on mobile elements like transposons and plasmids and are widely distributed among clinical strains and also in environmental isolates, and play an important role as reservoirs of antimicrobial resistance genes (L'Abée-Lund and Sørum, 2001; Stokes and Gillings, 2011; Koczura et al., 2013, 2014). Remarkably, one study carried out at aquaculture facilities in the northern Baltic Sea (Finland) reported that antibiotic resistance genes for sulfonamides (sul1 and sul2) and trimethoprim (dfrA1) and an integrase gene for a class 1 integron (intI1) persisted in sediments below fish farms at very low antibiotic concentrations during a 6-year observation period (Muziasari et al., 2014). Presumably, antimicrobial resistance genes could spread in marine sediments to other bacteria that colonize non-aquaculture fish and from these to seafood processing environments. This could explain the finding of sul1 in bacteria isolated in the present study from fish like sardines, anchovies, Athlantic horse mackerel, sea bass, gilthead seabream, and salmon, or from squid. Furthermore, in a number of cases, bacterial isolates carrying sul1 also tested positive for the florfenicol resistance gene floR, which can also be associated with Class 1 integrons (Toleman et al., 2007; Lin et al., 2016).

Aquaculture heavily depends on the use of antibiotics (Hernández Serrano, 2005), and several studies have reported on antimicrobial resistance in bacteria from aquaculture ecosystems (e.g., Shah et al., 2014; Huang et al., 2015; Xiong et al., 2015; Watts et al., 2017). Remarkably, one of the two isolates from the present study with broadest spectra of antimicrobial resistance (identified as A. oleivorans) was isolated from prawns grown in aquaculture. The A. oleivorans strain from the present study was resistant to 10 antimicrobials and tested positive for the genetic determinants bla_TEM and bla_CTX−M. Since the prawns were boiled and frozen and then sold unfrozen over the counter on ice, there is a possibility that this strain arrived to the food by cross contamination during handling. Boiled prawns are ready to eat, therefore the bacteria with multiple antibiotic resistance together with their antibiotic resistance genes could be passed from prawns to humans. Acinetobacters are inhabitants of soil and water in addition to being opportunistic pathogens for humans, and the human-pathogenic strains are known to exhibit both intrinsic and acquired resistance to a wide variety of antimicrobials (Doughari et al., 2011). It is also worth noting that the remaining acinetobacters detected in the present study (all of them identified as A. oleivorans) were isolated from raw salmon slices, also grown in aquaculture. They were resistant to a lower number of antibiotics (4 and 6), but all of them tested positive for the genetic determinants qacEΔ1, sul1 and aac(6′)-Ib, and two also were positive for the extended-spectrum β-lactamase gene bla_TEM. A. oleivorans was described as a diesel-oil and n-hexadecane-degrading bacterium isolated from a rice paddy (Kang et al., 2011). So far, antibiotic resistance of non-pathogenic Acinetobacter species has been weakly explored, but results from the present study suggest that they could be an important reservoir of antimicrobial resistance traits.

A 14.94% of the 87 isolates from the present study were resistant to imipenem. Carbapenems were the last β-lactams retaining nearuniversal anti-Gram-negative activity, but carbapenemase genes are spreading, conferring resistance (Nordmann et al., 2011). Furthermore, among the metallo-β-lactamases investigated in the present study, the imipenemase (IMP), and the New Delhi metallo-β-lactamase NDM-1 gene were detected in two isolates, both belonging to Pseudomonas. Several metallo-β-lactamases encoded by mobile DNA have emerged in important Gram-negative pathogens (Cornaglia et al., 2011). NDM has been reported mainly in Klebsiella pneumoniae and E. coli, but it has also been found in a variety of other members of the Enterobacteriaceae, including Acinetobacter spp., in Pseudomonas spp., and in Vibrio cholerae (Poirel et al., 2011; Walsh et al., 2011; Mataseje et al., 2016). In Pseudomonas aeruginosa, the chromosomal bla_NDM−1 gene has been reported to be located within a Class 1 integron bearing insertion sequence (IS) common region 1 (ISCR1) in a Tn402-like structure (Janvier et al., 2013; Jovcic et al., 2013). Remarkably, the integron also contained the QAC resistance determinant qacEΔ1 and the sulfonamide resistance gene sul1. A similar gene array has been reported for the bla_NDM−1 regions from E. coli transferable plasmid pNDM15-1078 (Mataseje et al., 2016). Furthermore, bla_NDM−1 frequently appears in association with ble_MBL gene that confers resistance to bleomycin, a glycopeptide antibiotic that is naturally produced by Streptomyces verticillus. It is possible that bleomycin-like molecules contribute to selective pressure, leading to the further spread of NDM producers in the environment (Mataseje et al., 2016).

All Aeromonas isolates from the present study were resistant to chlorhexidine, ampicillin and imipenem. In addition two of them tested positive for sul1 and bla_TEM and one also was positive for bla_CTX−M. Aeromonas are widely distributed in aquatic environments (Holmes et al., 1996). The genus includes species pathogenic for fish (like A. salmonicida and others) and humans. In one study on ampicillin-resistant isolates from estuarine waters, Henriques et al. (2006) detected the presence of the integrase gene (along with other genes associated with Class I integrons) in Aeromonas strains. The authors also detected the presence of β-lactamase genes bla_TEM, bla_SHV, bla_CphA, and bla_OXA−B in Aeromonas strains. Remarkably, the bla_OXA−B detected in Aeromonas sp. and A. hydrophila was associated with Class I integrons. Another study on Aeromonas isolated from nine freshwater trout farms in Australia reported the presence of sul1 together with other antibiotic resistance genes typically associated with Class I integrons (Ndi and Barton, 2011). However, the β-lactamase genes investigated (bla_TEM and bla_SHV) were not detected. Further studies should be carried out in order to determine the possible association of the β-lactamase genes detected in the Aeromonas isolates from present study with Class I integrons.

In the present study, one isolate resistant to cefotaxime, ceftazidime, streptomycin and nalidixic acid was identified as Listeria innocua. This isolate tested positive for the β-lactamase resistance gene bla_PSE. Previous studies have reported on antimicrobial resistance in Listeria spp. isolated from raw fish and open-air fish market environment (Jamali et al., 2015). The authors reported a high resistance of L. monocytogenes to tetracycline and penicillin G in agreement with other studies (Rodas-Suárez et al., 2006; Fallah et al., 2013). Another work reported a high resistance level to ampicillin, cefotaxime (100%), and penicillin (57%) in seafood isolates of L. monocytogenes (Abdollahzadeh et al., 2016). We could speculate that the L. innocua isolated in the present study from squid originated from the seafood handling and processing environment. Although this is a non-pathogenic species, the results would suggest that different species of Listeria in seafood processing environments can act as reservoirs for antimicrobial resistance traits.

Exposure to low concentrations of antibiotics, disinfectants, chemical pollutants, and metals can act as a selective force leading to resistance processes among indigenous bacterial populations (Martinez, 2009). There is a concern that exposure to metals may co-select for antibiotic resistance. In the present study, isolates capable of growing at high concentrations of copper sulfate, and zinc chloride included representatives of Pseudomonas, Aeromonas, Acinetobacter, and Proteus. The observed metal tolerance could be due to direct exposure to these metals in the environment. Furthermore, the multicopper oxidase gene pcoA/copA was detected in several Pseudomonas strains that were also positive for different antibiotic resistance genes (bla_CTX−M, bla_TEM, bla_PSE, bla_NDM−1, bla_IMP, aadA1, aac(6′)-Ib, sul1, sul2, floR). The copper-inducible system copABCDRS was first described within the plasmid pPT23D in P. syringae (Cha and Cooksey, 1991) and is homologous to the pco system found on the conjugative plasmid pRJ1004 from E. coli. In addition to the plasmid-borne system, chromosomal P-type ATPases are responsible for conferring tolerance to metals like copper, zinc, cobalt, chromium, and cadmium (Teitzel and Parsek, 2003). As indicated by Berendonk et al. (2015), selective pressures present in natural environments such as rivers and lakes as a result of human practices are leading to the occurrence of an increasing number of multi-drug resistant environmental Pseudomonas isolates (Berendonk et al., 2015). Nevertheless, in the study of the resistome of P. aeruginosa E67, an epiphytic isolate from a metal-contaminated estuary, physical links between metal and antibiotic resistance genes were not identified, suggesting a predominance of cross-resistance associated with multidrug efflux pumps (Teixeira et al., 2016). Bacterial multidrug efflux pumps can accommodate a variety of antimicrobials, including dyes, antibiotics, and biocides (Poole, 2007). Nevertheless, a possible association of plasmidic pcoA/copA with antibiotic resistance genes in Pseudomonas deserves to be further investigated.

In conclusion results from the present study clearly indicate that bacteria from seafoods carry resistance traits against diverse antimicrobials. The possible role of these bacteria in spread of antimicrobial resistance through the food chain deserves further investigation. It is also important to find alternative ways to manage antimicrobial resistance. Biological methods such as the use of probiotics (Tan et al., 2016; Banerjee and Ray, 2017) or the application of bacteriophages (Letchumanan et al., 2016; Torres-Barceló et al., 2016; Parmar et al., 2017) could be promising alternatives to classical antimicrobials.

JR carried out seafood sampling, microbiological analysis and PCR investigation of genetic determinants of resistance. MG participated in planning and supervision of the experimental work and contributed with strain identification and preparation of PCR experiments. RP contributed with data analysis and preparation of graphical material for the manuscript. RL and AG carried out global analysis of the results and wrote the manuscript.