No evidence for a relationship between farm or transformation process locations and antibiotic resistance patterns of Pseudomonas population associated with rainbow trout (Oncorhynchus mykiss).

AIMS: Study the relationship between antibiotic resistance patterns of Pseudomonas isolated from farmed rainbow trout fillets and farm or transformation process locations. METHODS AND
RESULTS: Pseudomonas strains were isolated from rainbow trout sampled in two differently located farms and filleted in laboratory or in a processing factory. One hundred and twenty-five isolates were confirmed as belonging to Pseudomonas using CFC selective media, Gram staining, oxidase test and quantitative polymerase chain reaction methods. Fifty-one isolates from separate fish fillets were further identified using MALDI-TOF mass spectrometry, and the minimal inhibitory concentrations (MIC) of 11 antibiotics were also determined by microdilution method. Most of the isolates belonged to the Pseudomonas fluorescens group (94.1%), and no relationship was established between antibiotic resistance patterns and sampling locations (farms or filleting areas). Multiple resistance isolates with high MIC values (from 64 µg ml-1 to more than 1024 µg ml-1 ) were identified.
CONCLUSIONS: Antibiotic resistance patterns found in Pseudomonas isolates were not influenced by farms or transformation process locations. Seven isolates were found highly resistant to four different antibiotic classes. SIGNIFICANCE AND IMPACT OF THE STUDY: This study does not provide evidence of a relationship between farm or transformation process locations on antibiotic resistance patterns of Pseudomonas population.

Introduction

Fish farms are interfaces between aquatic environments and food‐producing animals whose bacterial ecosystems are exposed to effluents from human activities (e.g. effluents from wastewater treatment plants, pastures) and to antibiotics used for the treatment of fish diseases (Woolhouse & Ward, 2013). Therefore, fish foodstuffs have been considered as potential reservoirs of antibiotic‐resistant bacteria. However, the contamination sources—such as farming practices (including environmental contamination associated with the farming environment and the slaughtering process)—and the filleting processing types have rarely been considered. In Europe, rainbow trout is one of the most produced farmed fish; its production must comply with strict regulations, especially on the use of antibiotics. Limited to a few classes, these molecules can only be used for the treatment of a fish disease following a veterinary prescription. In France, the six antibiotics with marketing authorizations for the sole treatment of fish diseases are florfenicol, flumequine, oxolinic acid, oxytetracycline (OTC), sulphadiazine and trimethoprim. Antibiotic resistance genes have been described in bacterial ecosystems associated with rainbow trout samples, in skin, gut (Muziasari et al., 2017) and fillets samples (Helsens, Calvez, Bouju‐Albert, et al., 2020), suggesting that these antibiotics——whether from effluents or fish treatment—play a role in the selective pressure related to the emergence of antibiotic‐resistant bacteria.

The genus Pseudomonas is considered as the most diverse and ecologically important bacterial group. It has an outstanding ability to survive in different types of habitats (soil, freshwater and marine environments) and to grow well at cold temperatures (as low as 4°C). As such, Pseudomonas are frequently found in aquatic samples (Akinbowale et al., 2007; Gordon et al., 2006; Miranda & Zemelman, 2002), and in the skin and gut microbiota of farmed rainbow trout (Lowrey et al., 2015; Wong et al., 2013; Zhang et al., 2018). As a result of their ability to persist in processing environments, they can be found in the bacterial communities of fresh fish fillets (Helsens, Calvez, Bouju‐Albert, et al., 2020; Langsrud et al., 2016; Møretrø et al., 2016). The Pseudomonas genus is composed of more than 200 ubiquitous species (Peix et al., 2009, 2018) classified into several groups, such as the P. fluorescens group which is composed of more than 50 species (Garrido‐Sanz et al., 2017). Various species can cause fish diseases, like Pseudomonas anguilliseptica (Wiklund & Bylund, 1990), P. aeruginosa (Algammal, Mabrok, et al., 2020) and Pseudomonas putida (Altinok et al., 2006), but non‐human pathogenic Pseudomonas species have also been described in samples from aquatic environments and animals (Boutin et al., 2014; Luczkiewicz et al., 2015). Among bacterial pathogens, P. aeruginosa is recognized as the main cause of septicaemia in freshwater fishes (Roberts, 2012). Its pathogenicity is often due to several virulence genes as oprL or toxA coding for outer‐membrane protein and exotoxin A respectively (Algammal, Mabrok, et al., 2020). Pseudomonas fluorescens is scarcely involved in diseases in French aquaculture but some cases have been reported in several fish species whom histopathological aspects have been described (Miyazaki et al., 1984). More recently, authors have identified several iron‐acquisition proteins involved in bacterial virulence (Sun & Sun, 2015). At least, an experimental infection of rainbow trout carried out with P. fluorescens has shown a mortality rate of 40% at an infectious dose of 108 CFU per ml (Dinçtürk & Tanrıkul, 2021).

Bacterial multidrug resistance increase globally and is considered as public health threat (https://www.who.int/news‐room/fact‐sheets/detail/antimicrobial‐resistance). Antibiotic‐resistant Pseudomonas isolates have been found in samples from diseased farmed fish such as catfish (Nguyen et al., 2014), Oreochromis niloticus and Clarias gariepinus (Algammal, Mabrok, et al., 2020), in influents and effluents of wastewater treatment plants (Luczkiewicz et al., 2015), in river water (Gordon et al., 2006) and in sediment from rainbow trout farms (Akinbowale et al., 2007; Ishida et al., 2010). Furthermore, they possess mobile genetic elements and can transfer antibiotic resistance determinants to various non‐pathogenic and pathogenic Pseudomonas species (Cazares et al., 2020; Kottara et al., 2018; Molina et al., 2014; Yomoda et al., 2003) and to pathogenic species from other genera like Escherichia coli (Dominguez et al., 2019). Several previous studies revealed the emergence of multidrug‐resistant bacterial pathogens from different origins especially fish, warranting proper use of the antimicrobial agents in both health and veterinary sectors (Abolghait et al., 2020; Algammal, Mohamed, et al., 2020; El‐Sayed et al., 2019; Yang et al., 2018).

In a context admittedly characterized by the presence of antibiotic‐resistant Pseudomonas species in aquatic samples, this work aimed at determining whether the location of farm on a same river and the transformation process area (laboratory vs. factory facility) has an effect on the antibiotic resistance patterns of Pseudomonas isolates sampled from rainbow trout fillets. The isolates were tested against a selection of antibiotics including those used for the treatment of fish diseases in France, and various other antibiotics of clinical interest in human therapeutics. The following results constitute a first approach toward the characterisation of antibiotic‐resistant Pseudomonas isolates according to the sampling location of farmed rainbow trout.

MATERIALS AND METHODS

Fish and fillet sampling

Rainbow trouts were sampled from two farms (A and B) located on the same river in Brittany. Farm A was located at the upstream of an urban area and the other one at the downstream. The farm owners were recruited on a voluntary basis according to the experimental design of Helsens, Calvez, Prévost, et al. (2020). In this previous study, a questionnaire filled by farmers identified that only Farm B had used an antibiotic treatment, either one OTC treatment 27 weeks before the sampling date. Twenty‐eight fish were randomly collected from the raceway environment of each farm; 14 were filleted in the laboratory, and the other 14 were filleted in a processing factory. All fish and fillets were individually packed in plastic bags and stored on melted ice (0°C) prior to processing in laboratory (6 h of handling). In laboratory, each fillet was portioned under sterile conditions to obtain at least 30 g of tissue with the most uniform possible flesh‐to‐skin ratio. Thirty grams from one fillet from one rainbow trout were considered as one sample. Each fish sample (1–28) was encoded with two letters indicating the original farm (A or B), and the filleting location (R for raceway/laboratory; F for factory) The 56 samples were vacuum packed individually in plastic bags and stored at −20°C until analysis.

Total mesophilic bacteria count and Pseudomonas isolation

Each 30‐g sample was transferred to a 63‐µm stomacher bag (Interscience) with 50 ml of phosphate buffer saline containing 5% TWEEN® 80 (Sigma‐Aldrich). Bacterial suspensions were obtained by the rinsing method according to Helsens, Calvez, Bouju‐Albert, et al. (2020). The filtered rinse solution was collected from the stomacher bag for enumeration of total viable mesophilic aerobic count (TVMC) as an indicator of fish fillet quality and good manufacturing processing (Özyurt et al., 2015).

Enumeration was performed using a 10‐fold serial dilution of the rinse solution (10−1, 10−2) spread onto brain heart infusion (BHI) agar plates (Difco™, VMR USA) and incubated at 20°C for 48 h. Colonies were initially quantified using eCOUNT™ (Heathrow Scientific).

Presumptive Pseudomonas strains were isolated according to the ISO 13720 standard (International Organization for Standardization, 2010). For each sample, 400 µl of the fillet rinse solution were collected and spread onto two plates of selective cephalothin sodium fusidate cetrimide (CFC) agar (Merck Millipore), and then incubated at 25°C for 48 h. A maximum of three yellow and other light‐coloured colonies were selected per sample and tested for their Gram and oxidase properties. The bacteria that appeared as Gram‐negative rods and showed a positive reaction for oxidase were considered as presumptive Pseudomonas spp. Each selected isolate was grown on BHI agar (BHI, VWR Chemicals Prolabo) to obtain pure cultures. For each isolate, one colony was inoculated in 10 ml of BHI broth and incubated overnight at 25°C. Then, each bacterial suspension was cryopreserved by adding 20% of glycerol, and stored at −80°C.

DNA extraction and confirmation of the Pseudomonas genus by quantitative polymerase chain reaction

The isolates were recovered on BHI agar plates and incubated at 30°C for 24 h. One colony was sub‐cultured in BHI broth overnight at 25°C under stirring at 180 rpm. One millilitre of this culture was centrifuged at 6000 g for 5 min in a 1.5‐ml microtube. The supernatant was removed, and the cell pellet was suspended in 100 µl of sterile water and boiled at 95°C for 3 min in a water bath. The boiled cell suspension was centrifuged at 6000 g for 5 min, and the supernatant containing DNA was transferred into a new microtube. These DNA samples were stored at −20°C until use. Each DNA sample was diluted 100‐fold in a total water volume of 500 µl and used as a template for quantitative polymerase chain reaction (qPCR) amplification.

qPCR amplification was performed on 143 isolates to confirm their Pseudomonas affiliations. Primers Pse 435F (5′‐ACTTTAAGTTGGGAGGAAGGG‐3′) and Pse 449R (5′‐ACACAGGAAATTCCACCACCC‐3′) targeting the 16S rRNA V3–V4 region were used to amplify a specific 251‐bp fragment (Bergmark et al., 2012). The reaction was carried out in a final volume of 20 µl containing 10 µl of 2 X Sso Advanced universal SYBR Green Supermix (Bio‐Rad), 0.2 µl of each primer at 50 µM, 4.6 µl of water, and 5 µl of 1/100 diluted DNA sample. A P. fluorescens CIP 69.13T DNA sample was used as positive control, and water was used as a no‐template control. The amplification reaction was conducted using a Bio‐Rad CFX connect real‐time PCR detection system (Bio‐Rad). The cycling parameters were as follows: initial denaturation step at 98°C for 3 min, 39 cycles of 98°C for 10 s and 60°C for 30 s. A melting curve from 55 to 95°C was determined after the last amplification cycle and at a temperature transition rate of 0.5°C/5s. Quantification cycles (Cq) values were automatically obtained from the Biorad CFX Manager software program.

Identification of the Pseudomonas group

Among all the Pseudomonas isolates confirmed by qPCR, 51 isolates were randomly chosen to represent each fish fillet positive for this genus. The isolates were recovered on MH agar plates and sent at ambient temperature to LABOCEA lab (https://www.labocea.fr/) for MALDI‐TOF analysis. Each isolate was deposited in a 96‐well plate and recovered with 1 µl of 70% formic acid for fast extraction. One microlitre of co‐crystallizing matrix solution containing α‐cyano‐4‐hydroxycinnamic acid (HCCA) solubilized in acetonitrile (50%, v/v), ultra‐pure water (47.5%, v/v) and trifluoroacetic acid (2.5%) was added to each deposit. Once dry, each MALDI target was introduced in the mass spectrophotometer for analysis. The spectra were generated and analysed by a MALDI‐TOF Microflex™ system (BrukerDaltonics, Germany) using the flexControl and MBT Compass software programs. The criteria for successful identification were within a confidence score of ≥2.0 for species confirmation and ≥1.7 for genus confirmation.

Antimicrobial susceptibility test

Minimum inhibitory concentrations (MICs) were determined by the microdilution method in cation‐adjusted Mueller‐Hinton broth (CAMHB) (Oxoid) according to the MIC guidelines of the Clinical and Laboratory Standards Institute (CLSI M07‐A10 [2015], M100S‐26th [2016], VET03/VET04‐S2 [2014]). The 51 Pseudomonas isolates identified by MALDI‐TOF mass spectrometry were tested against 11 antibiotics purchased from Sigma‐Aldrich (Merck) and listed in Table 1. Six molecules subjected to the European marketing authorisation for the treatment of fish bacterial infections were tested: florfenicol (1–128 µg ml−1), flumequine (2–256 µg ml−1), oxolinic acid (0.0625–8 µg ml−1), OTC (0.5–64 µg ml−1), and the sulfadiazine/trimethoprim association (0.5/9.5 to 64/1216 µg ml−1). In the absence of interpretive criteria for sulfadiazine/trimethoprim, sulfamethoxazole was used to detect phenotypic resistance to sulfonamides. Four antibiotics used in human treatment were tested according to the CLSI M100 referential: colistin (1–128 µg ml−1), piperacillin (8–1024 µg ml−1), gentamicin (2–256 µg ml−1) and ceftazidime (4–512 µg ml−1). Azithromycin was chosen to represent macrolides and was used from 8 to 1024 µg ml−1. The concentrations corresponded to the CLSI recommendations (2016) for non‐Enterobacteriaceae bacteria including Pseudomonas spp. non‐aeruginosa for piperacillin, ceftazidime, gentamicin, colistin and the trimethoprim‐sulfamethoxazole association (CLSI M100). For oxolinic acid, OTC and florfenicol, the recommended concentrations for Aeromonas salmonicida were tested according to the CLSI VET 03 reference database. No indication was available for azithromycin and flumequine, so the concentrations were chosen according to Boss et al. (2016) with azithromycin on P. aeruginosa and to Scarano et al. (2018) with flumequine on Aeromonas spp. Eight isolates presenting highly resistant profiles for some antibiotics were tested at higher concentrations (Table 1).

Table 1

Antibiotic concentrations (µg ml−1) tested on the 51 isolates of Pseudomonas spp

	Pseudomonas spp.	Highly resistant profile Pseudomonas spp.	Escherichia coli ATCC 25922
Oxytetracycline	0.5–64	32–512	0.25–4
Oxolinic acid	0.062–8	4–128	0.016–0.25
Florfenicol	1–128	64–1024	0.5–8
Trimethoprim/sulfamethoxazole	0.5/9.5–64/1216	32/608–512/9728	0.031/0.594–0.5/9.5
Flumequine	2–256		0.062–1
Colistin	1–128	64–1024	0.125–2
Piperacilline	8–1,024		0.250–4
Ceftazidime	4–512		0.031–0.5
Gentamicin	2–256		0.062–1
Azythromycin	8–1024

Stock solutions of antibiotics corresponding to 20× concentrations of the maximum concentrations tested in the MIC experiments were prepared in their appropriate diluent solutions and kept at 4°C. They were extemporaneously diluted to 2× in CAMHB. One hundred microlitres of the 2× solutions were placed in the first well of a 96‐U‐well plate. Fifty microlitres were sampled and serially diluted in 50 µl of CAMBH in the following 7 wells in order to test eight concentrations for each antibiotic. Bacterial suspensions were prepared from colonies cultured on Mueller Hinton agar plates at 30°C for 24 h for the Pseudomonas spp. isolates and at 37°C for the quality control (QC) E. coli ATCC 25922. The suspensions were adjusted at an optical density at 600 nm (OD600) of 0.13–0.14 for the Pseudomonas spp. isolates and at 0.1 OD600 for the QC to obtain suspensions containing 108 CFU per ml. The suspensions were diluted to 106 CFU per ml in CAMBH, then 50 µl were distributed in each well containing the antibiotic to be tested. The positive control was a mix of 50 µl of CAMBH and 50 µl of inoculum suspension. The negative control contained 100 µl of CAMBH.

The 96‐U‐well plates were sealed with Parafilm®. Pseudomonas spp. plates were incubated at 28°C for 24–28 h, and the QC plate was incubated at 35°C for 16–20 h, both under stirring at 180 rpm. The bacteria in the inoculum suspension were counted on BHI agar plates. MICs were determined as the lowest concentrations with no bacterial or hardly any bacterial development compared to the positive control.

No MIC breakpoints were available for OTC, florfenicol, flumequine and oxolinic acid. For this reason, we calculated ECOFF (Epidemiological Cut‐Off) values using ECOFF Finder software (Turnidge et al., 2006). These values determined within the Pseudomonas population isolated in our study were used to check for the presence of isolates possessing atypical (or non‐wild‐type) phenotypes compared to the wild‐type population.

Statistical analyses

Nonparametric data were analysed using a Wilcoxon Mann–Whitney test in the R environment (version 4.0.3). The e1071 package (Meyer et al., 2021) was used for statistical analyses. The ggplot2 (Wickham, 2016) and ggpubr (Kassambara, 2020) packages were used for the graphic design of the statistical test. A multiple correspondence analysis (MCA) was performed using the FactoMineR 2.4 and factoextra 1.0.7 packages to visualise and summarise the dataset built with Pseudomonas isolates as individuals and the active variables that described them (e.g. antibiotic resistance phenotypes). The aim was to assess the relationship between the antibiotic resistance patterns and the sampling locations of Pseudomonas isolates (farm and filleting locations). Within categorical variables, phenotypes for antibiotics with available clinical breakpoints were defined as follows: (i) susceptible: strains susceptible at the standard dose, and also “intermediate” strains newly proposed as susceptible when exposed to increased doses of antibiotics (European Committee on Antimicrobial Susceptibility Testing [EUCAST], 2019), (ii) resistant and (iii) highly resistant strains that we defined with MICs at least 32‐fold the ones of the clinical breakpoint. For antibiotics without a clinical breakpoint, two phenotypes (wild type; atypical/non wild type) were defined according to the calculated ECOFF values of the total population. The number of resistances to different antibiotics were used as quantitative variables, and two illustrative variables (farm and filleting location) were also considered in the MCA. Confidence ellipses (95%) were built to exhibit significant differences between the farming/filleting locations and the antibiotic resistance patterns of Pseudomonas isolates.

RESULTS

Total aerobic bacteria count and Pseudomonas isolation

The measurement of total viable mesophilic aerobic bacteria is mainly used as an indicator of fish fillet quality and is very helpful to determine the presence of spoiler bacteria when values are beyond 7 log CFU per g. In this study, TVMCs ranged from 2.19 log CFU per g (farm A) to 3.33 log CFU per g (farm B) in rainbow trout fillet samples (Figure 1). A significant difference (p < 0.0001) was highlighted between the two farms whatever the filleting location (laboratory or factory). On CFC medium, 154 isolates with yellow and other light‐coloured colonies phenotypes, Gram‐negative aspects and oxidase activities were considered as presumptive Pseudomonas (80 isolates from farm A; 74 from farm B). Then, 54 isolates (farm A) and 71 isolates (farm B) were confirmed as belonging to the Pseudomonas genus by qPCR. After these confirmations, 51 out of 56 fish fillet samples were considered as Pseudomonas‐positive. We randomly chose one isolate per fish fillet, distributed as follows: 12 and 13 isolates from fish samples from farm A (and 14 and 12 from farm B) filleted in the laboratory or in the factory respectively. These 51 isolates were further analysed for identification and assessment of the MICs of different antibiotics.

FIGURE 1

Enumeration of total viable mesophilic aerobic bacteria (log CFU per g) on fish fillets sampled in two farms (A and B) and according to the filleting location (laboratory versus factory). Significant TVMC differences between the farms were tested using a non‐parametric Wilcoxon test (α = 0.05)

Identification of Pseudomonas isolates using MALDI‐TOF mass spectrometry

Conventional approaches (phenotypical and biochemical characterization, selective media and PCR) did not allow us to determine the affiliation of the Pseudomonas isolates within the different phylogenetic groups or many species. However, molecular approaches (as MLST or whole genome sequencing) for group or species identification of Pseudomonas isolates are available. In our study, we have chosen to use a faster approach based on the identification of proteins profiles of each isolates compared to a database of known micro‐organisms. The 51 Pseudomonas isolates were identified using MALDI‐TOF mass spectrometry associated with the Bruker database (Table 2). All isolates were confirmed as belonging to the Pseudomonas genus with a confidence score greater than 1.7, and 33 isolates (64.7%) were confirmed at the species level with a score ≥2.0. Most of the isolates belonged to the P. fluorescens group (n = 48/51), and to a far lesser extent to the Pseudomonas lutea (n = 2/51) and Pseudomonas syringae (n = 1/51) groups.

Table 2

Distribution of Pseudomonas isolates from two different farms and filleting locations

Group affiliation	FARM A (N)		FARM B (N)		Total (%)
	Laboratory	Factory	Laboratory	Factory
P. fluorescens	11	12	13	12	94.1
P. lutea	1	0	1	0	3.9
P. syringae	0	1	0	0	2.0

N, number of isolates.

Antibiotic resistance of the Pseudomonas isolates

The MICs of 10 antibiotic compounds were assessed on all 51 Pseudomonas isolates (Table 3). As regards antibiotics commonly used in aquaculture, the trimethoprim/sulfamethoxazole association was the only one with a clinical breakpoint. Based on this, 21 Pseudomonas isolates out of 51 were considered resistant. The MICs of other compounds without a clinical breakpoint available (OTC, florfenicol, flumequine and oxolonic acid) were highly variable. The different antibiotic susceptibility phenotypes highlighted variable proportions of isolates resistant to compounds of critical interest in human medicine, such as ceftazidime (n = 20/51) or colistin (n = 11/51). Conversely, all Pseudomonas isolates were highly sensitive (<2 µg ml−1) to gentamicin.

Table 3

Distribution of the MICs of 11 antimicrobial drugs for Pseudomonas isolates from two rainbow trout farms and filleting locations

Vertical dark lines: breakpoints used with the following sequence of interpretation: S = susceptible; I = susceptible, increased exposure; R = resistant/NS = non‐susceptible; Vertical white lines: calculated ECOFF values using the ECOFF Finder program with 95% endpoint. Grey areas, range of antimicrobial concentrations used during the tests.

Among all isolates, 16 (31.4%) were sensitive to four different antibiotic classes (β‐lactamines, aminoglycosides, sulphonamides and polymixins) with available clinical breakpoints. Twenty‐four (47.1%), nine (17.6%) and one (2.0%) isolates were resistant to one, two and three antibiotic classes respectively. The ECOFF values of our Pseudomonas population varied for antibiotics without available breakpoints (Figure S1). For this population, calculated ECOFF values were: 8 µg ml−1 for OTC, 32 µg ml−1 for flumequine, 256 µg ml−1 for florfenicol and 512 µg ml−1 for azithromycin. Five isolates displayed atypical (or non‐wild‐type) resistance phenotypes to OTC and florfenicol. The respective MICs of OTC and florfenicol for this atypical resistant sub‐population found in both farms ranged from 64 to 256 µg ml−1 and from 512 µg ml−1 to more than 1024 µg ml−1. These isolates were also resistant to the trimethoprim/sulfamethoxazole association, with MICs ranging from 128/2432 µg ml−1 to more than 512/9728 µg ml−1. Two more isolates were highly resistant to colistin at a MIC of 512 µg ml−1 in farm B, compared to 3 isolates at 32 µg ml−1 maximum in farm A. Thirteen isolates were resistant to ceftazidime in farm B with MICs of 16–128 µg ml−1, and 7 in farm A with MICs of 16–64 µg ml−1. Finally, the analysis of antibiotic resistance patterns based on both breakpoint and ECOFF values revealed that eight (15.7%), five (9.8%) and one (2.0%) were resistant to two, three and four antibiotic classes respectively (Table 4). Among these subpopulations of fourteen isolates, the main patterns were composed by resistance to (i) trimethoprim/sulfamethoxazole association and ceftazidime (6 out of 14 multiple resistant isolates), (ii) colistin associated with ceftazidime or trimethoprim/sulfamethoxazole (3 out of 14 isolates) and (iii) those described above with high MIC values (OTC, florfenicol, trimethoprim/sulfamethoxazole).

Table 4

Profiles of the 14 multiple antibiotic resistant isolates within the Pseudomonas population (N = 51) based on both breakpoint and ECOFF values

Isolate ID	Phylogenetic group affiliation	Multiple resistance patterns	Number of resistance a
Farm A—laboratory filleting
A03RPs3‐08	P. fluorescens	OTC‐FFC‐TMP/SXT‐CAZ	4
A12RPs2‐30	P. fluorescens	OTC‐FFC‐TMP/SXT	3
A14RPs3‐37	P. fluorescens	OTC‐FFC‐TMP/SXT	3
Farm A—factory filleting
A13FPs3‐75	P. fluorescens	TMP/SXT‐CAZ	2
Farm B—laboratory filleting
B03RPs3‐86	P. fluorescens	TMP/SXT‐COL‐CAZ	3
B04RPs1‐87	P. fluorescens	COL‐CAZ	2
B05RPs3‐90	P. fluorescens	TMP/SXT‐CAZ	2
B07RPs3‐98	P. fluorescens	TMP/SXT‐CAZ	2
B11RPs3‐110	P. fluorescens	TMP/SXT‐CAZ	2
Farm B—factory filleting
B03FPs1‐121	P. fluorescens	TMP/SXT‐CAZ	2
B05FPs3‐125	P. fluorescens	TMP/SXT‐COL	2
B21FPs3‐128	P. fluorescens	OTC‐FFC‐TMP/SXT	3
B26FPs3‐140	P. fluorescens	OTC‐FFC‐TMP/SXT	3
B14FPs1‐144	P. fluorescens	TMP/SXT‐CAZ	2

Abbreviations: CAZ, ceftazidime; COL, colistin; FFC, florfenicol; OTC, oxytetracycline; TMP/SXT, trimethoprim/sulfamethoxazole association.

To different antibiotic classes.

Relationship between the sampling locations of the fish fillets and the antibiotic resistance patterns of the Pseudomonas population

In order to assess the relationship between the sampling locations of the fish fillets (farm or filleting facility), we performed a MCA on 5 out of 10 active variables (Figure 2). As partly described above, all Pseudomonas isolates were susceptible or shown as having wild‐type phenotypes to five antibiotic compounds (oxolinic acid, flumequine, piperacilline, gentamicin and azithromycin). The relevance of these active variables was considered low in explaining differences between the antibiotic resistance patterns of the isolates. The MCA biplot showed that the total variance of our data was mainly explained by the first and second dimensions (62.5%, the contributions of each of the six dimensions are presented in Table S1). The variables significantly involved (p < 0.001) in the first dimension were the resistance patterns with high MICs of OTC, florfenicol and trimethoprim/sulfamethoxazole (R 2 = 0.99). For the second dimension, the variables more significantly (p = 0.002) related to its construction were the resistance phenotypes when exposed to sulfamethoxazole (R 2 = 0.62), ceftazidim (R 2 = 0.44) and colistin (R 2 = 0.28). The ellipse‐coloured biplot highlighted that the antibiotic resistance patterns of the Pseudomonas isolates from each farm (A in red, B in green) overlapped with a 95% CI, confirming a weak association between resistance phenotypes and farm location. The confidence ellipses of the antibiotic resistance patterns of the Pseudomonas isolates also overlapped when laboratory filleting (purple) and factory filleting (blue) were taken into account.

FIGURE 2

Graphical representation of the multiple correspondence analysis. The statistical data analysis by MCA did not show any relationship between the antibiotic resistance profiles of the Pseudomonas strains isolated from two farms (A and B) or according to the sampling location of fish fillets (laboratory or factory). Five active variables (resistance patterns to oxytetracycline, florfenicol, ceftazidime, colistin and trimethoprim/sulfamethoxazole) and four illustrative variables (farm A, farm B, laboratory, factory) were used. The first and second dimensions represent 62.5% of total inertia

DISCUSSION

The total viable mesophilic aerobic bacteria count (TVMC) is considered as an indicator of fish fillet quality and good manufacturing practices (Özyurt et al., 2015). Several studies performed on rainbow trout fillets in different countries have shown TVMCs ranging from 1.543 log CFU per g to 6.3 log CFU per g (Oraei et al., 2011; Özyurt et al., 2015; Pao et al., 2008; Pullela et al., 1998; Ucak et al., 2020). Ozyurt et al. (2015) defined that TVMCs ranging from 3.06 log CFU per g to 3.31 log CFU per g could indicate a high fish quality which are in agreement of our values. These results are informative of fillet freshness: they suggest that the bacterial communities are representative of those associated with rainbow trout in its environment. However, the rainbow trout fillets from farm B were significantly more contaminated than the fillets from farm A. The different TVMCs could be explained by the release of effluents upstream of farm B, in an urbanized area including two wastewater treatment plants, as previously described using tuf gene‐targeted qPCR (Helsens, Calvez, Bouju‐Albert, et al., 2020). Among aerobic mesophilic bacteria communities, several genera and species have been isolated from rainbow trout fillet, for example, Micrococcus, Corynebacterium, Aeromonas or Pseudomonas (Nedoluha et al., 2001). The present study confirms the presence of Pseudomonas associated with rainbow trout fillets, in accordance with the ubiquitous character of this genus in the environment (Pirnay et al., 2005; Sazakli et al., 2005). We identified P. fluorescens as the main phylogenetic group of the population using MALDI‐TOF mass spectrometry. This approach was recently confirmed as accurate for phylogenetic group or sub‐group identification within this genus in freshwater ecosystems and different fish species (Kačániová et al., 2019; Mulet et al., 2020). Our findings are in agreement with previous studies in which the P. fluorescens group was the most isolated one on fillets of different fish species (Duan et al., 2019; Shehata et al., 2020).

Pseudomonas in rainbow trout can cause diseases, such as infections by P. putida or the newly identified species P. tructae (Altinok et al., 2006; Oh et al., 2019). On the contrary, other species like P. fluorescens seem to be involved in the inhibition of rainbow trout colonization by other pathogenic micro‐organisms (Gonzalez‐Palacios et al., 2019; Korkea‐aho et al., 2012; Liu et al., 2015). We did not isolate any pathogenic Pseudomonas, and the health status of the farms was good based on historical data provided by the farmers (no diseased fish sampled). Even if non‐pathogenic Pseudomonas isolates can have a protective effect on fish, the isolates from the two farms showed various antibiotic resistance profiles and could be a concern in terms of animal/human health. In this study, we first identified that no relationship could be established between the farm or the factory processing locations and the Pseudomonas antibiotic resistance patterns. Indeed, similar multiple antibiotic resistance patterns were widespread on fish fillets sampled in both farms whatever the proximity to urban wastewaters. Moreover, no changes in antibiotic resistance patterns were occurred between filleting at the factory (high amount of fishes filleted per day) and at the laboratory, suggesting that factory would not be involved in potential enrichment of antibiotic‐resistant bacteria. This result is supported by previous work carried out in the same farms by Helsens, Calvez, Bouju‐Albert, et al. (2020), who hypothesized that the microbiota of fresh farmed rainbow trout was mainly influenced by the river environment. Pseudomonas is ubiquitous in many environments, and it is probably present in the river that hosts both farms. Nevertheless, to confirm these results, it would be interesting to compare the resistance patterns to the same antibiotics of Pseudomonas isolated from fillets of rainbow trout farmed in other rivers.

The data for each antibiotic confirmed results from other studies reporting a predominance of gentamicin‐susceptible Pseudomonas isolated from rainbow trout farms (Akinbowale et al., 2007; Duman et al., 2019). Populations isolated from Australian rainbow trout but also from other freshwater and marine fish showed that a great proportion of isolates were resistant to different classes of antibiotics (Akinbowale et al., 2006, 2007). Among 44 Pseudomonas isolates, Akinbowale et al. (2007) revealed that 90.9% were resistant to florfenicol, with MICs of 32–128 µg ml−1 according to the recommendations of the CLSI. Similarly, we determined that 86.3% (n = 44/51) of the Pseudomonas isolates were resistant to this range of MICs, while 5.9% (n = 3/51) had the highest MICs (>1024 µg ml−1). The resistance phenotypes related to this antibiotic appear to be widespread in different countries; the atypical resistance profiles observed in our study could be due to its use in France. This hypothesis is supported by another study in Chilean salmon farms where this antibiotic is highly used for bacterial infection treatment (Miranda & Rojas, 2007). These authors demonstrated that the MICs of isolates from two farms ranged from 256 µg ml−1 to more than 1024 µg ml−1. While we evidenced resistance to trimethoprim/sulfamethoxazole, several authors isolated Pseudomonas strains from different fish species that were fully sensitive to this antibiotic association, using dilution broth or agar diffusion methods (Akinbowale et al., 2007; Türe & Alp, 2016). A study performed on 316 Pseudomonas strains isolated from contaminated waters showed a wide, species‐dependent distribution of resistance phenotypes to trimethoprim/sulfamethoxazole association (Camiade et al., 2020). These data showed a percentage of resistant Pseudomonas (49.4%) similar to the one observed in our study (41.2%). Likewise in recent study performed in Chilean salmon farms, the authors highlighted a great proportion of Pseudomonas isolates resistant to several antibiotic classes like sulphonamides (Dominguez et al., 2019). Their MICs were similar to and even higher than those observed in our study. These authors identified several isolates belonging to the Pseudomonas genus and carrying the sul and dfr genes encoding enzymes which confer sulphonamide resistance—dihydropteroate synthase and dihydrofolate reductase enzymes. They also demonstrated that resistance to antibiotics was attributable to the acquisition, spreading and maintenance of class‐1 integrons. Among 25 isolates carrying one or two plasmids, ten were able to co‐transfer resistance to sulphonamides, trimethoprim and other classes (florfenicol, OTC) to donor bacteria (Dominguez et al., 2019).

We isolated five Pseudomonas isolates resistant to OTC (from 64 to 256 µg ml−1), and also resistant to florfenicol and the trimethoprim/sulfamethoxazole combination, corresponding to 9.8% of the isolates, while all 44 Pseudomonas strains isolated from rainbow trout in Australia were sensitive to OTC (Akinbowale et al., 2007). A study on antibiotic resistance in 30 trout farms in Turkey yielded a proportion of 30% of OTC resistance out of 20 isolated Pseudomonas isolates; among them, seven carried at least one of the tetA, tetB, tetC or tetD resistance genes. However, none of them was involved in a multi‐drug resistance profile (Capkin et al., 2017). Likewise, Türe and Alp (2016) found that 20% of the species isolated from fish raised in sea‐water or freshwater farms were resistant to OTC from which only 15 Pseudomonas isolates were isolated out of 240. Among these 15 isolates, four were multi‐drug resistant, but none of them was resistant to OTC, and none carried the tetB gene (Türe & Alp, 2016). Interestingly, although OTC is commonly used in fish farming in most countries, the resistance profiles of the five isolates in our study were not in accordance with the studies of Türe and Alp (2016) and of Capkin et al. (2017). Deeper investigations of our relevant isolates would be necessary to understand the resistance profiles of these ones and to confirm the hypothesis of resistance genes carried by plasmids, integrons or phages.

Ceftazidime—a third‐generation cephalosporin—is considered as a critically important antimicrobial for human medicine by the World Health Organization (2019). Knowledge on resistance phenotypes within the Pseudomonas population is mainly supported by P. aeruginosa because of its interest in human medicine. A collection of 499 Pseudomonas isolates through an urban water cycle was recently built by Butiuc‐Keul et al. (2021). The authors showed that the percentages of ceftazidime‐resistant isolates were higher in hospital effluents and in one drinking water source (ca. 20.0%) than in wastewater treatment plant influent or effluent (ca. 12.0% and 4.0% respectively). These intriguing results suggest that ceftazidime resistance phenotypes within the Pseudomonas population are more related to found species than to proximity to a contamination source. Out of 316 Pseudomonas strains isolated from contaminated environments, only 2.5% were resistant to ceftazidime (Camiade et al., 2020). Two out of five isolates were found resistant to ceftazidime in a Pseudomonas population isolated from fin fishes (Carol et al., 2013). We identified 20 out of 51 Pseudomonas isolates (39.2%) resistant to this antibiotic in the two farms. Based on the high diversity of resistance phenotypes to ceftazidime within the Pseudomonas population probing of whether this variability could be related to bacterial species or to antibiotic use.

Polymyxins such as colistin have also been listed as critically important antimicrobials for human medicine (World Health Organization, 2019). To our knowledge, few studies have investigated resistance phenotypes to this antibiotic class within the Pseudomonas population, except for P. aeruginosa. Among Pseudomonas isolates from blood stream infections, two out of 19 isolates were resistant to colistin, with maximum MICs of 8 µg ml−1 (Varghese et al., 2020). In the same way, 40 P. aeruginosa isolates were all found susceptible to colistin, with MICs of 0.25–2 µg ml−1 (Zhang et al., 2020). In our study, most Pseudomonas isolates (76.5%) were considered susceptible to this antibiotic. Among the resistant sub‐population, eight isolates were resistant, with MICs of 4–32 µg ml−1, and two were highly resistant (MICs of 512 and >1024 µg ml−1 respectively). These two phenotypes were not associated with resistance to other antibiotic classes and did not seem to be among clinical isolates of P. aeruginosa involved in human infections. Based on the results discussed above, we could hypothesize that “environmental” Pseudomonas populations could be a reservoir of highly colistin‐resistant isolates. Considering the critical importance of this antimicrobial, it is recommended to further characterise the genetic elements involved in this phenotypic resistance in order to determine whether it would be transferable to pathogenic bacteria for fish and humans.

In conclusion, rainbow trout fillets from two French farms harboured Pseudomonas isolates mainly belonging to the P. fluorescens group. This study provided data on antibiotic resistance levels and patterns of Pseudomonas non‐aeruginosa isolated from rainbow trout fillets. No evidence for a relationship of farm or filleting processing locations has been highlighted on antibiotic resistance patterns found in Pseudomonas population. This study confirmed that Pseudomonas spp. associated with rainbow trout fillets are reservoirs of antibiotic resistance traits and revealed that these isolates were widespread in the environment independently of proximity of fish farms to human or animal inputs. Among these multiple antibiotic resistance isolates, seven were highly resistant to antibiotics used in aquaculture (OTC, florfenicol and sulphonamides) and in human medicine (colistin). To better understand their role on the spreading of antibiotic resistance to pathogenic bacteria for rainbow trouts (e.g. A. salmonicida subsp. salmonicida or Flavobacterium psychrophilum) or human (e.g. Pseudomonas aeruginosa), it would be of interest to further study this population and characterize their genetic contents by a whole genome approach.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Figure S1

Click here for additional data file.

Table S1

Click here for additional data file.