During 2011 and early 2012, chronic disease episodes characterized by largely non-specific clinical signs and mortalities of up to 60% were experienced in farmed red and wild type Nile tilapia fingerlings at two farms in Lincolnshire UK. In July 2012, five fish from each farm were randomly collected with a net, including apparently healthy and fish showing clinical sings from different sections of the farm and different sizes. The fish were euthanized with a lethal overdose of Tricaine methanesulfonate 1,000 mg/g (TPQ) (Pharmaq, Hampshire, UK), and necropsied. Samples of gills, heart, kidney, liver, and spleen were aseptically collected and fixed in 10% (v/v) neutral buffered formalin, 2.5% (v/v) glutaraldehyde in 100 mM sodium cacodylate buffer (pH 7.2) and 96% ethanol for histology, electron microscopy, and molecular diagnosis, respectively. The samples were sent for diagnosis to the Aquatic Vaccine Unit at Institute of Aquaculture (IoA), University of Stirling (UoS).

Formalin fixed tissues were processed using standard histological methods, stained with haematoxylin and eosin (H&E) and examined at 40x and 200x magnification on a Olympus BX51 light microscope (Olympus, Tokyo, Japan) equipped with a AxioCam MRc digital camera (Carl Zeiss, Göttingen, Germany). Glutaraldehyde fixed spleen and head kidney tissues were processed using standard methods for TEM and SEM. The sections were observed under an FEI Tecnai Spirit G2 Bio Twin TEM and a Jeol JSM6460LV SEM.

A Francisella genus specific PCR (Forsman et al., 1994) was performed using total genomic DNA (gDNA) from the fish sampled in July 2012 as a template. Previous studies have validated the use of this PCR as an inexpensive and practical tool to identify Francisella spp. in fresh and archived fish tissues (Hsieh et al., 2006; Soto et al., 2009).

The gDNA was extracted from the ethanol fixed spleens using a Nucleo Spin Tissue® kit (Macherey & Nagel, Düren, Germany) according to the manufacturer's instructions. A negative control was included using gDNA from tilapia reared at the Tropical Aquarium (TA) IoA-UoS. The TA was confirmed to be a source free of francisellosis after sampling 25 fish for bacteriological (culture in plates) and molecular (PCR) diagnosis. The PCR was performed using Illustra PuReTaq Ready-To-Go Beads™ (200 μM each dNTP in 10 mM Tris-HCl, pH 9.0, 50 mM KCl, and 1.5 mM MgCl2) (GE Healthcare, Chalfont St. Giles, UK) reconstituted to a final volume of 25 μl with: 5 μl of DNA template (~80 ng/μl), 2.5 μl of each primer (F11 5′-TAC CAG TTG GAA ACG ACT GT-3′ and F5 5′-CCT TTT TGA GTT TCG CTC C-3′) at a concentration of 10 pmol/ml (10 μM) and 15 μl of ultrapure water. Cycling conditions consisted of an initial denaturation step of 2 min at 93°C, followed by 35 cycles of: 1 min at 94°C, 1 min at 65°C, and 1 min at 72°C, and a final extension step of 5 min at 72°C in a Biometra TGradient Thermocycler (Biometra, Göttingen, Germany). Amplification products were visualized on ethidium bromide stained 1% agarose gel after electrophoresis for 35 min at 90 V.

During a follow up visit to the farms in November 2012, 10 red Nile tilapia from Farm 1 and 8 red and 2 wild type from Farm 2, were randomly sampled including apparently healthy fish as well as individuals showing clinical sings from different sections of the farms. The fish spleens were aseptically collected and half of them homogenized in 1 ml of 1x sterile phosphate buffered saline (0.02 M phosphate, 0.15 M Na Cl, and pH adjusted at 7.2) (PBS) using a Cordless Motor Pellet Pestle (Sigma-Aldrich, Dorset, UK). The other half of the spleens and all the kidneys were fixed in 96% ethanol and screened with the genus specific PCR previously described. To achieve primary isolation 5 different media were evaluated. Thus, ~20 μl of the spleen homogenates were streaked onto: cystine heart agar complemented in a 50% solution (v/v) with 2% bovine hemoglobin (CHAH; BD, Oxford, UK), modified Martin Lewis agar (MMLA; BD, Oxford, UK), modified Thayer-Martin Agar, (MTMA; BD, Maryland, USA), tryptone soya agar (TSA; Oxoid Ltd., Hampshire, UK), and cystine heart agar supplemented with 5% (v/v) tilapia blood (CHTB). In addition CHAH and CHTB were prepared with and without polymyxin B sulfate salt 100 units/ml (Sigma-Aldrich, Dorset, UK) and Ampicillin Ready Made Solution 50 μg/ml (Sigma-Aldrich, Dorset, UK).

All inoculated plates were incubated at 28°C for 10 days. For purification, single smooth, convex, circular colonies with a greenish-grayish color were subcultured twice on CHAH at the same temperature as for isolation. Gram staining, catalase and oxidase production, oxidation/fermentation of glucose (O/F test), and motility tests were carried out using standard methods.

Suspected Francisella like colonies (smooth, convex, with a greenish-grayish color) were grown in Modified Mueller-Hinton II cation adjusted broth supplemented with 2% IsoVitaleX (BD, Oxford, UK) and 0.1% D-(+)-glucose ACS reagent (Sigma-Aldrich, Dorset, UK) (MMHB). The broth cultures were grown overnight at 28°C in a shaker incubator at 175 rpm and stored at −80°C with 20% sterilized glycerol (BDH Prolabo-VWR International Eurolab, Leuven, Belgium). To confirm the identity of the colonies, their genomic DNA was obtained using the boiling technique outlined by Seward et al. (1997) with modifications. Briefly, five bacterial colonies were re-suspended in a 1.5 ml Eppendorf tube containing 100 μl of Tris-EDTA (TE) buffer and heated for 10 min at 99°C. The suspensions were then cooled on ice for 5 min and centrifuged at 15,800 g in a benchtop IEC Microlite Centrifuge (Thermo Electron Corporation, Massachusetts, USA) for 1 min. The supernatant containing the crude gDNA was then transferred into a fresh 0.5 ml tube and used as template in the genus specific PCR previously described.

Of the isolates recovered from the diseased fish, two isolates i.e., STIR-MATT-F1-f6 (from Farm 1) and STIR-GUS-F2f7 (from Farm 2), confirmed as Francisella spp. by PCR, were selected as representative of the outbreaks for further phenotypic analyses. The Fno type strain Ehime-1 (DSM 21254), Fno PQ1104 (from different geographical origin) and Escherichia coli ATCC 25922 were included as controls. All Fno isolates were grown on CHAH and MMHB, whereas E. coli was cultured on TSA and in tryptone soya broth (TSB; Oxoid Ltd., Hampshire, UK). Fno was cultured on solid media for 96 h, and E. coli for 24 h. Broth cultures were incubated in 15 ml aliquots for 18–21 h (i.e., mid log phase of the growth curve) in a shaker incubator at 150 rpm. All isolates were grown at 28°C. The Fno isolates used and recovered in this study are presented in Table 1.

The optimal in vitro growth temperature of each Fno isolate was investigated on agar by plating in triplicate six 20 μl drops containing the same bacterial concentration (dilution 10⁻⁶) with an optical density of 0.4 at a wavelength of 600 nm (OD₆₀₀ 0.4) and incubating them at 5, 15, 18, 21, 22, 24, 26, 28, 29, 30, 32, 33, and 37°C for 10 days.

Growth curves were established for STIR-GUS-F2f7 and Ehime-1 by inoculating and incubating triplicate flasks containing 99 ml of MMHB with 1 ml of starting culture (OD₆₀₀ 1.0) for 72 h at 28°C, on an orbital shaker incubator at 150 rpm. To monitor growth, a 1 ml sample was taken every 3 h and its optical density recorded. The bacterial growth curve was produced by plotting the absorbance of the culture at 600 nm against time (h).

Enzymatic activity and fermentation of carbohydrates was assessed for each isolate in triplicate using API 20E and API ZYM kits (BioMerieux, Marcy l'Etoile, France). The kits were used according to the manufacturer's instructions with the following modifications: bacteria were precultured in CHAH at 28°C, the API ZYM kit-strips were visually read at 4, 8, and 24 h post inoculation (hpi) and the API20E strips after 24 hpi.

To assess the metabolic activity (carbon utilization) of Fno, three Biolog-GN2 microplates (Biolog Inc., California, USA) were used per each strain. The plates were set up and analyzed according to the following protocol: bacteria grown on CHAH, as previously described. For each microplate, three 15 ml aliquots of MMHB were prepared in 50 ml centrifuge tubes. After incubation at 28°C for 18–21 h, the cultures were centrifuged in a Sigma 4K15C refrigerated benchtop centrifuge (Sigma Laborzentrifugen GmbH, Osterode am Harz, Germany) at 4°C for 15 min at 2,602 × g. The bacterial pellets were then washed by re-suspending them in 20 ml of sterile PBS (a sterile disposable 5 μl loop was required to break resuspend the pellets) and vortexing for 5 min. The tubes were centrifuged as described before and the PBS discarded. Each pellet was then suspended in 3 ml of GN inoculating fluid (Biolog Inc., California, USA) in a 50 ml centrifuge tube. In order to optimize the protocol i.e., identify the optimal inoculation density, a gradient of bacterial concentrations was prepared for the type strain Ehime-1 and STIR-GUS-F2f7 using the GN inoculating fluid. The OD₆₀₀ tested were 0.36, 0.46, 0.56, 0.66, 0.76, 0.86, 0.96, 1.06, and 1.5.The plates were inoculated with 150 μl of the adjusted bacterial suspension per well, using a multichannel pipette and a sterile reservoir (Biolog Inc., California, USA). After inoculation, the Biolog GN2 microplates were incubated at 28°C for 24 h and the color change recorded every 3 h by visual inspection. Following preliminary results from the density gradient, the remaining Fno isolates were subsequently tested at an OD₆₀₀ 0.85. The criteria chosen for the selection of this value was: the higher density at which the negative control remains as negative.

The cellular fatty acid methyl esters (FAME) composition of Fno was analyzed by gas chromatography (GC) according to the protocol established by Tocher and Harvie (1988). The two representative Fno isolates and the type strain Ehime-1 were grown in MMHB as previously described, using three 50 ml centrifuge tubes containing 20 ml of MMHB per isolate, and incubated for 43 h. After incubation, the absorbance of the cultures at 600 nm was determined and the bacterial suspension centrifuged at 4°C for 15 min at 2,602 × g. The resulting bacterial pellets were then washed by re-suspending them in 5 ml of sterile PBS, vortexing for 5 min and centrifuging at 4°C for 15 min at 2,602 × g.

The lipid content was extracted by suspending the pellets in 5 ml of ice cold chloroform/methanol (2:1 v/v) using a disposable glass Pasteur pipette and quantified gravimetrically. FAME were prepared by acid catalyzed trans-esterification at 50°C for 17 h. FAME were extracted (not purified) from the total lipid content and separated and quantified by GC using a Fisons GC-8160 (Thermo Scientific, Milan, Italy) equipped with a 30 m × 0.32 mm × 0.25 mm ZB-wax column (Phenomenex, Cheshire, UK) “on column” injection and flame ionization detection. Hydrogen was used as carrier gas with an initial oven thermal gradient from 50 to 150°C at 40°C per min to a final temperature of 230°C at 2°C per min. Individual FAME were identified by comparison to known standards i.e., Supelco™ 37-FAME mix (Sigma-Aldrich, Dorset, UK). Data were collected and processed using Chromcard version 1.19 (Thermoquest Italia SpA., Milan, Italy).

The minimal inhibitory concentration (MIC) of 39 different antimicrobial compounds was investigated using GN2F and AVIAN1F Sensititre® Plates (Trek Diagnostic System, West Sussex, UK). This procedure was performed in duplicate following the manufacturer's instructions and previously published protocols for Fno and Francisella tularensis (Baker et al., 1985; Brown et al., 2004; García del Blanco et al., 2004; Urich and Petersen, 2008; Soto et al., 2012). The media preparation, inoculation densities, incubation temperature, quality control organism, and interpretation of results were performed in compliance with the standards of the Clinical and (Clinical and Laboratory Standards Institute, 2014a). Briefly, the Fno isolates and E. coli ATCC 25922 were grown on agar as previously described and colonies suspended in sterile PBS to McFarland standard 0.5. This suspension was diluted 100-fold (Fno) or 1,000-fold (E.coli ATCC25922) in MMH and 50 μl of these added with a multichannel pipette to each well of the Sensititre® Plates. The plates were then incubated at 28°C and bacterial growth visually checked at 48 (Fno isolates) or 24 (E. coli ATCC 25922) hpi. The MIC value was defined as the lowest concentration with no visible growth.

The susceptibility or resistance of Fno to 16 different antibiotics was investigated using the disc diffusion method on agar plates following the protocol established by the Clinical and Laboratory Standards Institute (2006) and (Soto et al., 2012) Briefly, bacteria were harvested after incubation in CHAH as previously described and suspended in PBS to achieve a turbidity equivalent to McFarland standard 0.5. Fresh CHAH plates were inoculated with 100 μl of the suspension using sterile disposable L shaped spreader. After 60 min when the plates had dried, antibiotic discs (Oxoid) were dispensed using a self-tamping antimicrobial susceptibility disc dispenser (Oxoid). Plates were incubated at 28°C for 96 h and the diameter of inhibition zones measured after 72 h.

The gross pathology displayed by the diseased fish included fin erosion, scale loss, pale skin, white gills, and emaciation. At necropsy, most internal organs were enlarged and either haemorrhagic or pale. In some fish, the anterior kidneys were hyperaemic and enlarged, resembling a raspberry in appearance. Most of the fish presented white nodules in the spleen, posterior kidney, and liver, in some severe cases covering over 60% of the organ surface (Figure 1A).

Histological observations revealed the presence of extensive diffuse granulomatous inflammation. The affected tissues showed necrotising vasculitis and infiltration of mononuclear cells and neutrophils. The granuloma content was dominated by hypertrophied macrophages, fibroblasts, and leukocytes. Most granulomas exhibited necrotic cores. The most severely affected tissues were the spleen and the anterior kidney, with granuloma structures comprising up to 60% of the parenchyma. Granuloma formation was also seen in the heart, liver, and gills (Figure 1B).

By TEM, pleomorphic coccobacillary bodies ranging in size from 0.2–0.4 μm (width) to 0.4–1.7 μm (length) were observed (Figure 1C). These structures could be observed inside phagocytic cells, free in the cytoplasm and most frequently within vacuoles surrounded by an electron lucent membrane. In SEM micrographs of spleen and head kidney the 3 dimensional structure of the bacterial cells resembled the shape of a corn grain (Figure 1D). Additionally, vesicles with electron dense membranes detaching from the bacterial cells were also detected by TEM. The size of these structures was 60–80 nm (width) 90–100 nm (length) (Figure 2).

The Francisella genus specific PCR yielded a product of ~1.2 Kbp in five of the 10 samples tested from July 2012 confirming the presence of the bacteria within the fish tissues. In the samples recovered during the follow up visit in November 2012, the PCR analyses confirmed as positive all the fish from which it was possible to isolate the bacteria and also detected as positive four fish from where the isolation was not achieved. Interestingly the three biggest individuals, two of them the wild type Nile tilapia, were negative for both PCR and primary isolation (Table 3). In some fish, the results were not always the same for both tissues, in total four weak bands were observed in the agarose gel but were considered as positive after comparison with the controls (Supplementary Figure 1).

All the MICS for antibiotics which reference values at 28°C are available were within the range considered as acceptable by the Clinical and Laboratory Standards Institute (2014b), the lowest MIC values observed in the AVIAN1F Sensititre® plates across the replicates of the Fno tested were enrofloxacin (<0.12 μg/ml), gentamicin (<0.5 μg/ml), neomycin (<2 μg/ml), and streptomycin (<8 μg/ml) while the highest were ceftiofur (2 to >4 μg/ml), erythromycin (>4 μg/ml) sulphadimethoxine (128–256 μg/ml), trimethoprim/sulfamethoxazole (>2/38 μg/ml), penicillin (4 to >8 μg/ml), tylosin tartrate (20 to >20 μg/ml), and clindamycin (>4 μg/ml).

In the GN2F Sensititre® plates the lowest MIC values were amikacin (<8 μg/ml), ciprofloxacin (<0.5 μg/ml), gatifloxacin (<1 μg/ml), nitrofurantoin (<16 μg/ml), and tobramycin (<4 μg/ml) whereas the highest were aztreonam (16–32 μg/ml), cefazolin (32 to >32 μg/ml), cefotetan (>32 μg/ml), cefuroxime (32 to >32 μg/ml), and cefoxitin (>32 μg/ml). The range of variability observed in MIC values in both the AVIAN1F and the GN2F Sensititre® plates are summarized in Supplementary Table 2.

In the disc diffusion method, 8 of the 16 antimicrobials developed clear and reproducible zones of inhibition among the Fno strains i.e., enrofloxacin (5 μg/disc), kanamycin (30 μg/disc), gentamicin (2 μg/disc), tetracycline (30 μg/disc), oxytetracycline (30 μg/disc), florfenicol (30 μg/disc), oxolinic acid (2 μg/disc), and streptomycin (10 μg/disc). No substantial differences were found amongst Fno isolates. The list of all the compounds tested and inhibition zone sizes (means and standard deviations) are presented in Supplementary Table 3.