A total of 415 bacterial strains were obtained from the retrospective collections (2012–2021) of two partner laboratories specializing in the diagnostics of fish diseases, GDSAA in the New Aquitaine region (south-west France) and LABOCEA in the Brittany region (north-west France). Only bacteria identified as Aeromonas salmonicida salmonicida (ASS) by the partner laboratories were collected. All strains were isolated from samples obtained from fish exhibiting clinical signs of furunculosis, by fish farming veterinarians in various regions of France. This isolates were identified at the species level in the partner laboratories using bacterial culture characteristics [morphology and growth rate of colonies in Petri dishes, dissociation of the culture with a mixture of small and large colonies from a perfectly isolated clone, specific brown pigmentation of ASS in TSA (Tryptic Soy Agar) medium], and a MALDI-TOF (Bruker mass spectrometer) score above 1.9. The geographical origin of our strains reflects the location of the partner laboratories in the Brittany and New Aquitaine regions: 58% of our strains originated from the Brittany region and 24% from the New Aquitaine region.

The strains, stored at −80°C, were streaked on TSA at the two partner laboratories and transported to our laboratory at room temperature. To ensure the purity of the strains, they were streaked again on receipt. Identification of isolates was then confirmed in our laboratory by the PCR method of Byers et al. (2002) and by culture characteristics, in particular brown pigmentation at late stationary phase in TSA medium. All confirmed isolates were then stored in Tryptic Soy Broth (TSB) with 40% glycerol at −80°C.

The broth micro-dilution method was used to determine the MICs of 8 antimicrobial agents in accordance with CLSI guidelines (CLSI, 2020a). The antibiotics selected correspond to the molecules used in fish farming for the treatment of furunculosis. All tests were performed in 96-well microplates. The plates were formatted in 8 identical series of 10 two-fold dilutions for florfenicol (32 to 0.06 mg/L), oxolinic acid (32 to 0.06 mg/L), flumequine (32 to 0.06 mg/L), enrofloxacin (2 to 0. 004 mg/L), oxytetracycline (64 to 0.125 mg/L), doxycycline (32 to 0.06 mg/L), sulfadiazine (256 to 0.5 mg/L) and the combination of sulfadiazine + trimethoprim (154/8 to 0.3/0.016 mg/L). The antibiotics (Sigma-Aldrich France, Saint Quentin Fallavier) were certified for purity and dissolved in solvents specific to each molecule (concentration of stock solutions equal to 5120 mg/L for all antibiotics except the combination sulfadiazine + trimethoprim concentrated at 6160 mg/L).

On removal from storage at −80°C, ASS strains were streaked onto TSA agar and incubated at 22°C ± 1°C for 48–72 h. Colonies of ASS were then inoculated onto Mueller Hinton agar and incubated at 22°C ± 1°C for 12 h. From this stationary phase broth, a new culture was initiated approximately 3 h before inoculation of the 96-well microplates, in order to obtain an exponential growth phase. The suspension was adjusted to approximately 10⁶ CFU/mL, using both the optical density and appropriate dilutions. Fifty microliters of the suspension were added to each well of the microplate. Four wells were used as positive controls (bacterial suspension only), and eight as negative controls (uninoculated MH broth medium only, using the medium used for antibiotic dilutions and the medium used for strain dilution for plating). The microplates were incubated at 22°C ± 1°C for 48 h. A. salmonicida subsp salmonicida ATCC 33658 was used as a quality control. For all ASS isolates tested, the inoculum density was verified by inoculation of MH agar through a Steers replicator (Sigma-Aldrich France, Saint Quentin Fallavier) with 1 μL of the suspension from the positive control well before incubation of the microplate.

MIC₅₀ and MIC₉₀ were calculated from the distribution of MICs obtained. MIC₅₀ is the MIC value at which ≥50% of the isolates in a test population are inhibited. MIC₉₀ is the MIC value at which ≥90% of the strains within a test population are inhibited (Schwarz et al., 2010). The abbreviations ECV and ECOFF refer to consensus-based epidemiological cut-off values from CLSI and EUCAST, respectively. To date, no ECOFF has been defined by EUCAST for this bacterium. Oxolinic acid, florfenicol, oxytetracycline and the combination of sulfadimethoxine + ormetoprim are the only antibiotics for which ECVs have been determined by CLSI in A. salmonicida (including diverse subpopulations, i.e., typical as well as some atypical strains), based on data from different laboratories and different countries (Miller and Reimschuessel, 2006; Smith et al., 2024). Our data were analyzed in relation to the ECVs reported by CLSI for these antibiotics. In table 1, white fields represent the tested dilution range, while gray fields represent untested concentrations. Strains with MIC values below the tested concentration range were assigned to the prior gray two-fold dilution box. Strains with MIC values above the tested concentration range were assigned to the next gray two-fold dilution box.

For the sake of clarity, the abbreviation CO_Epid is used throughout the study to designate the provisional epidemiological cut-off values obtained as part of this study. Our provisional CO_Epid were statistically determined according to two methods. The first method is designated the “NRI method” (Normalized Resistance Interpretation) as proposed by Kronvall (2010). The epidemiological cut-off value was obtained through the automatic excel programme available by courtesy of P. Smith, W. Finnegan and G. Kronvall, used with permission from the patent holder, Bioscand AB, TÄBY, Sweden (European patent No 1383913, US Patent No. 7,465,559). The second method is designated “ECOFFinder method” as proposed by Turnidge et al. (2006). The estimation was obtained using nonlinear regression on expanding subsets, through the 2020 updated version of the automatic excel programme available by courtesy of J. Turnidge, used with permission from EUCAST. For this method, the provisional ECOFF defined with lower percentages provide better sensitivity for non-wild type selection, while higher percentages provide better specificity for the wild-type population (see Table 2).

The susceptibility of the strains to different families of antibiotics, i.e., with different modes of action, was also assessed. To compare the susceptibility of each strain to different families of antibiotics, we proceeded as follows: a strain was considered to belong to the NWT category to a particular family of antibiotics if it fulfilled the following conditions: (1) for phenicols, a MIC value above the ECV for florfenicol (4 mg/L); (2) for tetracyclines, a MIC value above both the ECV for oxytetracycline (1 mg/L) and the CO_Epid for doxycycline (1 mg/L); (3) for quinolones, a MIC value above the corresponding CO_Epid for oxolinic acid (4 mg/L) [be careful, not the ECV defined by CLSI] and flumequine (4 mg/L), and enrofloxacin (0.5 mg/L); (4) for the combination trimethoprim + sulfadiazine, a MIC value above the CO_Epid (0.25/4.8 mg/L).

The strains were collected over several years, and originate from different regions of France. The evolution of the proportion of strains with MIC above CO_Epid was observed as a function of year of sampling and region of origin by statistical regression. The Odds Ratios and their Confidence Intervals were calculated using R Studio software version 2023.12.1.402 (R Studio: integrated development for R, PBC, Boston) for R (R Core Team, 2021, R Foundation for Statistical Computing, Vienna, Austria).

A total of 406 isolates of ASS were collected from two diagnostic laboratories specializing in bacterial diseases of fish. These laboratories are located in two of the main trout producing regions of France, Brittany and New Aquitaine. The geographical origin of our strains is therefore not homogeneous across France, reflecting the location of our partner laboratories. Around 58% of our strains originated from the Brittany region, which accounts for 20% of French trout production, and 24% from the New Aquitaine region, which accounts for 30% of French trout production (Agreste-French Minitry of Agriculture, 2024). Our selection of strains was highly targeted to ASS and aimed to limit the possibility of misidentifications. Firstly, all strains had been obtained from samples collected by expert veterinarians from fish showing clinical signs of furunculosis. Aeromonas salmonicida is one of the oldest known pathogens of fish, particularly salmonids (Roberts, 2012). Historically, their classification has been based on the source of isolation: from salmonids (such as trout) they are classified as “typical,” from non-salmonid hosts or environmental samples they are classified as “atypical” strains (Vasquez et al., 2022). As all our strains were obtained from diseased farmed trout, it is reasonable to assume that the vast majority, if not all, of our strains are typical A. salmonicida (also known as ASS) (Roberts, 2012). However, only sequencing of each strain could guarantee this. Within our partner laboratories, strains that do not originate from diseased fish (strains of food origin, environmental origin, other animal species, or even unspecified) have an identification that is limited to the genus level (Aeromonas sp). Our study did not collect these strains. Secondly, the bacteria were identified as ASS by laboratories specializing in aquaculture bacteriology. In current practice in these veterinary laboratories, identification to the species level is achieved by a combination of bacterial culture characteristics, MALDI-TOF scoring, and clinical signs reported by veterinarians. The following species or subspecies are at risk of misidentification: Aeromonas salmonicida pectinolytica, Aeromonas salmonicida masoucida, Aeromonas eucrenophila, Aeromonas bestiarum, and, to a lesser extent, Aeromonas popoffi, Aeromonas encheleia. As MALDI-TOF alone has recently been reported to be insufficient for accurate species-level identification of Aeromonas isolates (Pérez-Sancho et al., 2018; Kitagawa et al., 2022), this combination reduces the risk of misidentifying strains, at least until sequencing becomes routinely available for diagnostic purposes. Thirdly, each isolate included in this study, obtained from the ASS collections of the partner laboratories, was streaked on receipt to ensure isolate purity, and its identity was confirmed by culture characteristics and the PCR method of Byers et al. (2002). None of the strains received showed characteristics that would question their species identification. We are therefore very confident in the quality of the identification of ASS in this study.

Our 406 strains come from 96 different farm companies, with strains collected over a 10-years period. Thus, some of the strains in our set originate from the same farm company and may be epidemiologically related. However, a given company may operate several ponds, and several separate operations (hatcheries, nurseries, grow-out), sometimes in very distant locations, making it difficult to assess the epidemiological link, as all strains are registered by the veterinary laboratory under the same company name. Although most farm companies submitted samples to the partner laboratory less than five times during the study period (2012–2021), a few farms provided between 20 and 40 samples each. The specific antibiotic susceptibility pattern of ASS on these farms could skew the observed global distribution of MICs in ASS strains, particularly if a high prevalence of furunculosis has led to extensive antibiotic use on these farms. However, a comparative study of strains from the same company showed no obvious association. The large number of strains tested in our study may limit the potential bias of testing isolates that may be epidemiologically related. Furthermore, this potential bias should not affect the determination of CO_Epid. Detailed information on the origin of each strain can be found in the Supplementary Data.

To our knowledge, this is the largest antibiotic susceptibility study of ASS ever conducted. However, the proportion of NWT stains in this study does not necessarily reflect the actual prevalence encountered in commercial fish farms, as there was no random sampling. Strains were collected by the laboratories on an event-driven basis, by isolating and identifying bacteria from samples taken from fish by veterinarians. It could be argued that the proportion of NWT strains in our database may be overestimated, because veterinarians are more likely to send bacteriological samples to the laboratories when treatment fails, which may lead to an over-representation of resistant strains in particular. However, in France, when furunculosis is observed in a trout farm, an antibiotic susceptibility test is requested in more than 80% of cases before antibiotics are used as first-line treatment, and in 100% of cases when antibiotics are used as second-line treatment (Carabin, 2022). Without claiming to reflect the true prevalence of antibiotic susceptibility in ASS strains, the large number of strains collected in this study provides a credible approximation of the distribution of antibiotic susceptibility in the trout industry in Western France. Considering that the farms are generally flow-through systems, with water being collected and discharged directly into watercourses without any specific treatment, these data provide an original picture of the epidemiological situation of antibiotic susceptibility in ASS, both from the point of view of improving aquaculture production practices (particularly with regard to the use of antibiotics) and in support of public health policies.

The MIC distribution appeared to be bimodal for quinolones, with two distinct populations identified for oxolinic acid (1st generation quinolone), flumequine (2nd generation quinolone) or enrofloxacin (3rd generation quinolone). Enrofloxacin has been classified as a critical molecule under French regulation since 2016. As a result, its use in fish farming has been very limited, unlike the other quinolones. Oxolinic acid and flumequine have been used to treat furunculosis in salmonids in France since the 1980s. These molecules, together with oxytetracycline, are still the only ones with a marketing authorization (MA) for use in fish in France. This means that they have to be used preferentially. However, to date in France, the antibiotics used as first-line treatment for furonculosis in trout farms are mainly florfenicol (38% of veterinary drugs prescribed) and the combination of sulfadiazine + trimethoprim (31% of veterinary drugs prescribed), drugs without a MA for use in fish (Carabin, 2022). The NWT strains identified by our CO_epid for a given quinolone were practically all classified as NWT for the other quinolones. This finding is consistent with current knowledge on quinolones, particularly fluoroquinolones, where resistance to one agent often leads to resistance to all agents of this class (Goñi-Urriza et al., 2000; Naviner et al., 2011; Giguère and Dowling, 2013). Thus, the use of first- or second-generation quinolones in fish farming may contribute to the selection of ASS resistant to the latest generation fluoroquinolones. Indeed, the European Medicines Agency has placed this entire family into “category B – restricted” (EMA, 2019). This category is intended for antibiotics that are of critical importance in human medicine and whose use in animals should be restricted in order to reduce the risk to public health by giving preference, when possible, to antibiotics in “category C – caution” or in “category D – prudence”, and by testing for antimicrobial susceptibility whenever possible. To date, only enrofloxacin has been classified under the French veterinary legislation as being subject to specific restrictions on prescription by veterinarians.

The CLSI established an ECV for oxolinic acid in Aeromonas salmonicida of 0.125 mg/L in VET04Ed3 (CLSI, 2020b), which is much lower than the value measured in our study (4 mg/L). In 2024, by aggregating the data used by CLSI to define the ECV for oxolinic acid (Miller and Reimschuessel, 2006) with new data from various laboratories, a new provisional epidemiological cut-off lowered to 0.0625 was proposed (Smith et al., 2024). The MIC₅₀ can be used as an estimate of the susceptibility of a particular population of bacteria to a particular antibiotic. Thus, it is important to note that the MIC₅₀ of oxolinic acid calculated in our study (1 mg/L) was close to that previously reported (2 mg/L) on 74 isolates collected from infected fish in France during the period 1985–2020 (Baron et al., 2021). However, it was much higher than that reported in other studies on other fish species, or in water samples from the environment (Tsoumas et al., 1989; Barnes et al., 1990, 1991; Miller and Reimschuessel, 2006; Baron et al., 2017; Scarano et al., 2018) – see Table 3. The disappearance of the WT population of ASS from French trout farms, as already suggested by the MIC₅₀ data presented, is the most likely explanation for this discrepancy between the international epidemiological cut-off values (ECV) and the provisional cut-off calculated in this study (CO_Epid), with a rightward shift of MIC values in the French ASS subpopulation associated with farmed trout.

There is limited literature on ASS susceptibility to flumequine, and no EUCAST or CLSI defined epidemiological cut-off value. However, as for oxolinic acid, our calculated MIC₅₀ for flumequine (2 mg/L) is much higher than those reported in the literature (Barnes et al., 1991; Baron et al., 2017; Scarano et al., 2018) – see Table 3. Some authors have proposed provisional epidemiological cut-off values ranging from 0.06 mg/L to 0.25 mg/L, far from our CO_Epid of 4 mg/L (Baron et al., 2017; Scarano et al., 2018). Comparison of our data with the literature and evidence of cross-resistance within this antibiotic family suggests that the WT population of ASS to flumequine has also disappeared from trout farms in western France. In light of our results, there is an urgent need to re-evaluate the current recommended dosing regimen for the use of first- and second-generation quinolones, which was established more than 30 years ago, as the dosing regimen for an antibiotic is based on the bacterial population that is sensitive to that antibiotic. Similarly, the clinical breakpoints used for antibiotic susceptibility testing should be re-evaluated, at least in France.

There is no EUCAST or CLSI defined epidemiological cut-off value for enrofloxacin. The MIC₅₀ values found in the literature for enrofloxacin, or its metabolite ciprofloxacin, are split between values around our own MIC₅₀ of 0.125 mg/L (Goñi-Urriza et al., 2000; Penders and Stobberingh, 2008; Baron et al., 2017; Viel et al., 2021; Lin et al., 2022), and values that are 10 times lower (Barnes et al., 1990; Ko et al., 1996; Kämpfer et al., 1999; Baron et al., 2017) – see Table 3. However, it is still difficult to identify ASS at the species level, and many of those studies report MIC₅₀ in sets of isolates of Aeromonas identified at the genus level only. This is an important limitation in the attempt to set epidemiological cut-off values in clinically important Aeromonas species. Nevertheless, although we are very confident in our CO_epid thanks to our high confidence in species level identification, it remains possible that the WT population has disappeared from trout farms in France, as is probably the case with first and second generation quinolones. Only 4.4% of the strains collected in our study showed an MIC of 0.03 mg/L or less, the provisional epidemiological cut-off value for Aeromonas salmonicida from river water suggested by Baron et al. (2017). Enrofloxacin displays a concentration-dependent bactericidal activity against Gram-negative bacteria. To optimize the efficacy of this antibiotic while limiting the selection of resistant bacteria, PK/PD (pharmacokinetic/pharmacodynamics) modeling was carried out on the dosing regimen for rainbow trout (Viel et al., 2021). When this model is compared with our sensitivity data for ASS, the preferred dosing regimen would be between 40 and 80 mg/kg, which is completely unrealistic in the aquaculture sector in terms of cost, production of the medicated feed, palatability (the standard dose is 10 mg/kg) or residues in tissues intended for human consumption. Therefore, the use of enrofloxacin to treat ASS in France may no longer be appropriate, even after antibiotic susceptibility testing, while under-dosing may lead to selection of resistant bacteria.

Few consensus-based epidemiological cut-off values for Aeromonas salmonicida have been defined by CLSI. The European Committee on Antimicrobial Susceptibility Testing website¹ provides MIC distributions for A. caviae, A. hydrophila and A. veronii, but no ECOFFs have been proposed due to the small number of observations, and no data are available for A. salmonicida. This work is therefore a contribution to the understanding of this bacterium.

Furunculosis is a relatively common disease in trout farms, and its prevalence is likely to increase with global warming. To control the disease, antibiotics are sometimes required in flow-through aquaculture farms, at the crossroads of animal, human and environmental health. The availability of such data on the epidemiology of ASS in France contributes to the monitoring of antibiotic resistance in a bacterial genus often considered to be a good indicator of antibiotic susceptibility in the aquatic environment (Goñi-Urriza et al., 2000; Alcaide et al., 2010; Lamy et al., 2022). In particular, Aeromonas spp. are especially able to promote interactions and gene exchange between One Health components (Lamy et al., 2022). These data are also essential for defining appropriate dosing regimens for antibiotics in fish. Indeed, in aquaculture veterinary medicine, a number of antibiotics are used off-label, with dosages defined several decades ago. A good definition of the susceptibility of ASS strains is a key element in reassessing dosage regimens for treatment. Finally, these data are essential for determining clinical breakpoints, which are crucial for interpreting antibiograms, to help prescribers make better use of veterinary medicines.