The strains, plasmid and primers used in this study are listed in Table 1. Escherichia coli strains S17-1, BW19851 (Metcalf et al., 1994) or MFDpir (Rochat et al., 2017b) were used to transfer DNA into F. psychrophilum THCO2-90 by conjugation. E. coli strains were grown at 37°C in Luria Bertani (LB) with 15 g of agar per liter added for solid medium. F. psychrophilum THCO2-90 was grown at 18°C in tryptone yeast extract salts (TYES) broth [0.4% (w/v) tryptone, 0.04% yeast extract, 0.05% (w/v) MgSO₄ 7H₂O, 0.02% (w/v) CaCl₂ 2H₂O, 0.05% (w/v) D-glucose, pH 7.2] or in modified Bushnell-Haas broth (BH; Sigma-Aldrich Co.) supplemented with a vitamin cocktail at pH 7 (5 mg L⁻¹ pyridoxamine, 1 mg L⁻¹ nicotinic acid, 1 mg L⁻¹ thiamine, 1 mg L⁻¹ riboflavine, 1 mg L⁻¹ D,L-panthotenic acid, 10 mg L⁻¹ 4-aminobenzoic acid, 1 mg L⁻¹ D-biotine, 1 mg L⁻¹ folic acid, 1 mg L⁻¹ vitamin B12, 5 mg L⁻¹ orotic acid anhydrous, 5 mg L⁻¹ thymidine, 5 mg L⁻¹ inosine and 2.5 mg L⁻¹ thioctic acid), 100 μM FeCl₃, 50 μM CaCl₂ and 0.5% (w/v) casein or gelatin when needed. Growth in liquid culture was carried out at 200 rpm and 18°C and evaluated by measuring OD_{600 nm} at different times. Stock cultures were preserved in TYES broth containing 20% (v/v) glycerol at −80°C. To observe colony spreading, F. psychrophilum strains were grown on 1/5 TYES with 15 g L⁻¹ of agar (Pérez-Pascual et al., 2009). Extracellular proteolytic activity on solid medium was visualized by using TYES containing 15 g L⁻¹ of agar and supplemented with 0.75% (w/v) gelatin or casein (Álvarez et al., 2006). For selective growth of E. coli strains carrying pEP4351 and pCP-derivative plasmids, transformants were selected with 20 μg mL⁻¹ chloramphenicol and 100 μg mL⁻¹ ampicillin, respectively. Cultures of E. coli MFDpir were supplemented with 0.3 mM diaminopimelic acid (Sigma-Aldrich Co.). Selection of F. psychrophilum transconjugants was carried out with 10 μg mL⁻¹ of gentamycin or erythromycin.

Genomic DNA extraction was performed with the Gen Elute Bacterial DNA (Sigma-Aldrich Co.) extraction kit. Plasmid DNA was purified with the NucleoSpin® Plasmid (Machery-Nagel) kit. PCR amplification products were separated on 1% agarose gels and bands were purified with the Illustra^TM GFX, PCR DNA and the Gel Band Purification Kit gel extraction system.

E. coli BW19851 was used for conjugative transfer of pEP4351 plasmid carrying Tn4351 into F. psychrophilum strain THCO2-90 as previously described (Álvarez et al., 2004). Briefly, the donor E. coli strain was grown to mid-log phase in LB broth and 10 mL were centrifuged to harvest cells. Cells were washed twice with TYES broth, and suspended in 50 μL of TYES broth. The recipient F. psychrophilum strain was grown to mid-log phase in TYES broth, 10 mL of culture was centrifuged, and the cell pellet was washed twice with TM buffer, consisting of 20 mM Tris-HCl and 20 mM MgSO₄ pH 7.2, and suspended in 50 μL of TM buffer. Cell suspensions of F. psychrophilum and E. coli were mixed together, spotted onto TYES agar, and incubated at 20°C for 48 h. After conjugation, cells were scraped off the plates, diluted in 1 mL of TYES broth, and plated on TYES agar containing 10 μg mL⁻¹ erythromycin. Erythromycin-resistant colonies of F. psychrophilum appeared on TYES agar after 5–7 days of incubation at 18°C.

The insertion of Tn4351 into the genome of 439 transconjugants was firmly established using inverse-PCR as previously described (Álvarez et al., 2006). Briefly, genomic DNA of the mutant strains was digested with HindIII followed by a re-ligation process. The resulting circular molecules were used as a template to amplify by inverse PCR the sequences adjacent to the Tn4351 insertion site using a specific pair of primers TN-1/IS4351-F (Table 1) and the GoTaq DNA polymerase (Promega, France). Sanger sequencing of the PCR amplified products was performed on an ABI PRISM 3100 (Applied Biosystems, CA, USA) and sequences were used to locate the transposon insertion site on the THCO2-90 genome (Rochat et al., 2017a).

The coding sequence of tetQ was replaced by aac(6′)-aph(2′) gene encoding a gentamycin resistance marker, while keeping expression signals unchanged (plasmid map in Figure S1). pCPGm^R was constructed as follows: the two DNA fragments were amplified by PCR using Phusion High-Fidelity DNA polymerase (Thermo Fisher), the vector fragment using pCP23 DNA as matrix and primers TRO300/TRO301, aac(6′)-aph(2′) using pZXL5 DNA (Zhang et al., 2012) and primers TRO302/TRO303. The resulting PCR products were assembled by the method developed by Gibson (Gibson et al., 2009) using the Gibson Assembly Master Mix (New England Biolabs). Engineered plasmids were constructed in E. coli S17-1. Correct replacement of tetQ by aac(6′)-aph(2′) was verified by PCR and DNA sequencing with primers TRO308 and TRO319. pCPGm^R was then transferred by electrotransformation to E. coli MFDpir, the donor strain used subsequently to introduce plasmids into F. psychrophilum by conjugation. Transconjugants were selected on TYES agar supplemented with 10 μg mL⁻¹ gentamycin and appeared between 3 and 5 days of incubation at 18°C. The presence of plasmid was checked by plasmid DNA extraction.

Two pCPGm^R derivative plasmids containing the coding sequence of gldD and gldG were constructed using Gibson's method (plasmid map available in Figure S2). Briefly, DNA sequences of interest were amplified from THCO2-90 genomic DNA by PCR using primers TRO370/TRO371 and TRO372/TRO373 for gldD and gldG, respectively. The vector was amplified by PCR using primers TRO350/TRO351 and pCPGm^R as DNA matrix. The DNA assembly results in the insertion of gld gene upstream of the expression signals of ORF1 of pCP1, a cryptic plasmid isolated from a F. psychrophilum isolate (McBride and Kempf, 1996). Resulting plasmids pCPGm^R–gldD and pCPGm^R–gldG were checked by PCR and DNA sequencing with primers TRO137 and TRO138. Plasmids were transferred into relevant F. psychrophilum strains by conjugation with E. coli MFDpir.

Adhesion ability of each strain to 96-well microtiter polystyrene plates with flat bottom (Nunclon^TM Delta surface, Nunc) was evaluated as previously described (Högfors-Rönnholm et al., 2015) with few modifications. Briefly, the wild-type strain THCO2-90 and the two mutant strains were grown in TYES broth to OD_600nm = 0.5. One milliliter of each bacterial culture was centrifuged at 11,093 g for 5 min, the supernatant was removed and cells pellet was resuspended in sterile distilled water. One hundred of microliters of each bacterial suspensions were added in quadruplicate to the microplate and incubated at 18°C for 3 h without shaking. Then, wells were washed twice with sterile distilled water. The adherent cells were stained with 100 μL 1% (w/v) crystal violet solution for 30 min at room temperature. Excess stain was removed by washing the wells four times with sterile distilled water, and stain bound to the adherent cells was released with 100 μL absolute ethanol for crystal violet solubilization. The adhesion ability of the bacterial cells was determined by measuring the OD_595nm using a Tecan Microplate Reader (Infinite 200 PRO). As a negative control non-inoculated sterile milliQ water was used. Percentages of binding refer to the level of adhesion observed for each strain, compared with adhesion of the wild-type strain which higher OD_595nm value was set to 100%. All assays were performed in quadruplicate and repeated at least two times for reproducibility.

Biofilm formation was evaluated using the standard assay with crystal violet staining as previously described for F. psychrophilum (Álvarez et al., 2006; Levipan and Avendaño-Herrera, 2017) with some modifications. The wild-type strain THCO2-90 and the two mutant strains were grown in half-strength TYES broth to the mid-exponential phase. The cultures were diluted 1:100 in half-strength TYES broth, and 150 μL were deposited in wells of 96-well microtiter polystyrene plates with flat bottom (Nunclon^TM Delta surface, Nunc). Wells containing non-inoculated medium were used as negative controls. The plate was incubated at 18°C under static condition for 120 h. Every 24 h, biofilm development was evaluated in four wells by strain. The supernatants were discarded, the wells were washed twice with 200 μL of sterile distilled water then 150 μL of 1% (w/v) crystal violet was added to each well. After 30 min, excess stain was removed by washing the wells four times with 200 μL of sterile distilled water and the stain bound to adherent cells was subsequently released by adding 100 μL of absolute ethanol. The biofilm formation was determined by measuring the OD_{595 nm} using a Tecan Microplate Reader and quantified as the specific biofilm formation (SBF) index proposed by (Niu and Gilbert, 2004): SBF = (B − NC)/G, where B is the amount of ethanol-solubilized crystal violet released from biofilm cells, NC is the amount of ethanol-solubilized crystal violet adherent to wells of negative controls, and G is the absorbance of the cell supernatant.

Cultures of F. psychrophilum strains THCO2-90, gldD::Tn and gldG::Tn were grown in 50 mL of TYES broth at 200 rpm and 18°C until reaching late exponential phase (OD₆₀₀ = 0.8). For secretome analysis, the supernatants were recovered by centrifugation at 6,000 g for 10 min at 4°C and concentrated by ultrafiltration using Amicon Ultra-centrifugal filters (Millipore, MW cut off 10 kDa) at 4,000 g for 30 min at 4°C. Then, 10 μg of each protein suspension was separated using one-dimensional short migration in SDS-PAGE. In-gel digestion of the proteins was performed on bands excised from one-dimensional SDS-PAGE. Each lane of short migration was cut and washed for 15 min with an acetonitrile/100 mM ammonium bicarbonate mixture (1:1). Digestion was performed in 50 mM ammonium bicarbonate pH 8.0 and the quantity of modified trypsin (Promega, sequencing grade) was 0.1 μg per sample. Digestion was carried out for 6 h at 37°C. The supernatant was reserved. Peptides were extracted by 5% formic acid in water/acetonitrile (v/v). Supernatant and extracted tryptic peptides were dried and resuspended in 50 μL of 0.1% (v/v) formic acid and 2% (v/v) acetonitrile. For shaving, cells pellets obtained from the culture centrifugation were washed twice in PBS and resuspended in 0.8 mL of PBS containing 1.2 mM sucrose and 1 mM CaCl₂ pH 7.4. Samples were digested in-solution for 10 min at 37°C by adding 2 μg mL⁻¹ of sequencing-grade modified trypsin. Enzymatic reaction was quenched by reducing the pH of peptide mixtures with 0.1% formic acid. Samples were filtered with Millex GV (0.25 μm, Millipore, ref SLCG004SL) and the resulting peptide mixtures were pre-cleaned with a Strata-X column (Phenomenex, ref. 8B-S100-TAK). Columns were washed with 1.5 mL of washing buffer [3% acetonitrile (ACN) and 0.1% trifluoroacetic acid (TFA)]. The peptide mixtures were charged into the columns, followed by three washing steps of 500 μL. Elution of peptides was achieved using 600 μL of elution buffer (40% ACN and 0.1% TFA). The resulting samples were concentrated under vacuum to dryness and resuspended in 50 μL of 0.1% TFA and 2% ACN. Before analysis in a high-resolution mass spectrometer, samples were diluted 1/100.

LC-MS/MS analysis was performed using an Ultimate 3000 RSLC system (Dionex, Voisins-le-Bretonneux, France) connected to a LTQ Orbitrap mass spectrometer (Thermo Fisher) by a nanoelectrospray ion source. Samples were resuspended in 50 μL of nano HPLC buffer (2% ACN/ 0.1% formic acid) and a dilution 1/100 was achieved for shaving experiments.

Mass spectrometry was performed using an Orbitrap Fusion™ Lumos™ Tribrid™ (Thermo Fisher Scientific) coupled to an UltiMate™ 3000 RSLCnano System (Thermo Fisher Scientific). Four microliters of each sample were loaded at 20 μL min⁻¹ on a precolumn (μ-Precolumn, 300 μm i.d × 5 mm, C18 PepMap100, 5 μm, 100 Å, Thermo Fisher) and washed with loading buffer. After 3 min, the precolumn cartridge was connected to the separating column (Acclaim PepMap®, 75 μm × 500 mm, C18, 3 μm, 100 Å, Thermo Fisher). Buffer A consisted of 0.1% formic acid in 2% acetonitrile and buffer B of 0.1% formic acid in 80% acetonitrile.

The peptide separation analysis was achieved at 300 nl min⁻¹ with a linear gradient from 1 to 35% buffer B for 50 min and 35 to 45% for 5 min. One run took 65 min, including the regeneration step at 98% buffer B. Ionization (1.6 kV ionization potential) and capillary transfer (275°C) were performed with a liquid junction and a capillary probe (SilicaTip™ Emitter, 10 μm, New Objective). Peptide ions were analyzed using Xcalibur 3.1.66.10. The machine settings were as follows: 1) full MS scan in Orbitrap (scan range [m/z] = 400–1,500) and 2) MS/MS using CID (35% collision energy) in Orbitrap (AGC target = 4.0 × 10², max. injection time = 50 ms, data type = profile). Analyzed charge states were set to 2–5, the dynamic exclusion to 60 s and the intensity threshold was fixed at 5.10³.

The genome of F. psychrophilum THCO2-90 (Rochat et al., 2017a) was searched by the X!TandemPipeline (open source software developed by PAPPSO, version 3.4.3, http://pappso.inra.fr/bioinfo/xtandempipeline/). Protein identification was run with a precursor mass tolerance of 10 ppm and a fragment mass tolerance of 0.5 Da. Enzymatic cleavage rules were set to trypsin digestion (“after Arg and Lys, unless Pro follows directly after”) and no semi-enzymatic cleavage rules were allowed. The fix modification was set to cysteine carboxyamidomethylation and methionine oxidation was considered as a potential modification. In a second pass, N-terminal acetylation was added as another potential modification, whereas all other previous settings were retained. The identified proteins were filtered as follows: 1) peptide E < 0.01 with a minimum of 2 peptides per protein and 2) a protein E < 10⁻⁴.

Peptide quantities of the proteome were analyzed by spectral counting (SC). SC takes into account the number of assigned spectra for each protein and is correlated to relative protein abundance. The P-values obtained from both ANOVA for the SC were considered significant below a value of 0.01.

The rainbow trout (O. mykiss) homozygous line A36 was used (Quillet et al., 2007). The uniformity of genetic background in isogenic lines and the high susceptibility of line A36 to F. psychrophilum infection makes this line highly relevant to test for changes in bacterial virulence. Fish were reared at 10°C in dechlorinated recirculated water until they reached 3–4 g, and were then transferred to continuous flow aquaria for infection experiments. Bacteria used for infections were prepared as follows: strains THCO2-90, gldD::Tn and gldG::Tn were grown in TYES broth at 200 rpm and 18°C until late-exponential phase (DO_{600 nm} = 1). This culture density corresponds to 10⁹ colony-forming units (CFU) mL⁻¹, determined by serial dilutions and plate counting on TYES agar. Two experimental infection models differing by the infection route were tested: intramuscular injection and immersion. Two independent experiments were performed for each experimental infection model.

For injection challenge, 50 μL of serial dilutions performed in TYES broth by diluting bacterial cultures to obtain 10⁶, 10⁷, and 10⁸ CFU mL⁻¹ were used. These doses correspond to theoretical 0.1, 1, and 10 LD₅₀ previously determined for the wild-type strain. Groups of 10 fish were challenged with each dose by intramuscular injection after anesthesia. As a negative control, a group of 10 fish were injected with 50 μL of sterile TYES broth.

For immersion challenge, bacterial cultures performed in TYES broth were diluted directly into the water of aquaria (15 L) at a final concentration of 5 × 10⁶ CFU mL⁻¹. Bacteria were maintained in contact with fish for 24 h by stopping the water flow then subsequently removed by restoring the water flow. Sterile TYES broth was used for the control group. Bacterial counts were determined at the beginning and at the end of the immersion challenge by plating serial dilutions of water samples on TYES agar. Water was maintained at 10°C and under continuous oxygenation for the duration of the immersion. Groups were composed of 46 fish. Virulence was evaluated according to fish mortality 14 days post-infection. Six fish of each group were randomly chosen and sacrificed 6 h after the end of immersion challenge to evaluate the bacterial load from spleen, gill, and skin mucus. Organs were mechanically disrupted in Lysing Matrix tubes containing 500 μL of 1% peptone water and 1 mm ceramic beads (Mineralex). Samples were homogenized at 6.0 m s⁻¹ for 45 s on a FastPrep-24 instrument (Thermo Fisher). Serial dilutions of the homogenized solution were plated on TYES agar.

Animal experiments were performed in accordance with the European Directive 2010/2063/UE and approved by the institutional review ethics committee, COMETHEA, of the INRA Center in Jouy-en-Josas, France. Authorizations were approved by the Direction of the Veterinary Services of Versailles (authorization number 15-58).

To develop a functional genomic approach aiming to understand the role and the relative importance of genes of F. psychrophilum, a Tn4351-mutant library was constructed in strain THCO2-90 according to the previously developed strategy (Álvarez et al., 2004). About 2,000 erythromycin-resistant transconjugants were obtained. The specific localization of Tn4351 was performed by inverse PCR (Álvarez et al., 2006) on a subset of this library, formally identifying the insertion site for 439 transconjugants. Among them, two strains carrying a transposon insertion into gld homologous genes were selected for further characterization (Figure 1). In the first one, hereafter named gldD::Tn, the Tn4351 transposon is located after position 1,222,107 of the THCO2-90 chromosome and interrupts the THC0290_1046 gene, disrupting the protein after amino acid residue 36 (out of a total of 187). THC0290_1046 encodes the gliding motility lipoprotein precursor GldD (protein_id = SHH90844.1). In the other mutant, hereafter named gldG::Tn, the Tn4351 transposon is located after position 2,148,164 and disrupts the THC0290_1849 gene that encodes the gliding motility transmembrane protein GldG (protein_id = “SHI04365.1”), leading to a truncated protein after residue 307 (out of a total of 559). GldG likely forms along with GldA and GldF an ATP-dependent ABC transporter, found to be required for gliding in F. johnsoniae (Hunnicutt and McBride, 2001). In silico analysis of their genetic context suggests that gldD could be transcribed from the promoters of mutY and ssb genes (Figure 1). gldG lies downstream of gldF, a genetic organization highly conserved among genomes of the class Flavobacteriia carrying gldG and gldF homologous genes (Figure S3; McBride and Zhu, 2013) and upstream of dnaN, an essential gene likely transcribed from its own promoter (Figure 1).

The two mutant strains showed a lack of colony spreading when grown on 1/5 TYES agar, while the wild-type strain displayed the characteristic wide spreading phenotype (Figure 2A). In addition, strains gldD::Tn and gldG::Tn displayed a significant diminution of their extracellular proteolytic activity compared to the wild-type strain using gelatin as substrate (Figure 2B). Similar defects were observed using casein as substrate (data not shown). Interestingly, whereas no differences were observed in their growth kinetics when grown in the tryptone-rich TYES broth, with a doubling time of about 3 h for strains THCO2-90, gldD::Tn and gldG::Tn, both mutant strains were unable to grow in BH minimal broth supplemented with 0.5% casein or gelatin as C-source, contrary to the wild-type strain (Figure 2C).

In order to complement the Tn4351 erythromycin-resistant mutants, a new shuttle vector derived from pCP23 (Agarwal et al., 1997) was constructed by substituting the tetracycline-resistance marker tetQ by the aac(6')-aph(2') gentamycin-resistance gene. Conjugative transfer of the resulting plasmid, pCPGm^R, in F. psychrophilum led to hundreds of gentamycin-resistant clones after 3–4 days on TYES gentamycin. The presence of plasmid was confirmed by DNA extraction. No gentamycin-resistant clones were obtained using E. coli MFDpir empty of plasmid as donor. In contrast, conjugations performed using E. coli MFDpir pCP23 (tetQ) or empty of plasmid as donor strains led to hundreds of F. psychrophilum false positive clones, which appeared on TYES tetracycline after 6–7 days of incubation. To complement Tn4351 mutants, the coding sequence of gldD and gldG was cloned under the control of a F. psychrophilum promoter into pCPGm^R (see Figure S2). Introduction of pCPGm^R-gldD and pCPGm^R-gldG into gldD::Tn and gldG::Tn mutants, respectively, resulted in the complementation of each of them. Spreading on 1/5 TYES agar was comparable for the resulting colonies with those of the wild-type strain (Figure 2A). The extracellular proteolytic activity was also restored by plasmid introduction (Figure 2B). These results indicate that GldD and GldG are both required for efficient spreading and extracellular proteolytic activity in F. psychrophilum.

The ability of F. psychrophilum cells to adhere to polystyrene plates has been previously reported (Högfors-Rönnholm et al., 2015). Here, under similar assayed conditions, both strains gldD::Tn and gldG::Tn showed an impaired adhesion ability to polystyrene plates after 3 h of incubation at 18°C (Figure 3A). Biofilm formation depends on adhesion of microorganisms to each other and to biotic or abiotic surfaces and F. psychrophilum cells form biofilms on polystyrene plates (Álvarez et al., 2006; Sundell and Wiklund, 2011). Interestingly, both strains gldD::Tn and gldG::Tn showed a strongly reduced biofilm formation ability in contrast to the wild-type strain when grown in 1/2 TYES broth under static condition (Figure 3B). When grown as planktonic bacteria in liquid culture, reduced sedimentation was observed after 48 h of incubation for both mutant strains compared to the wild-type strain (Figure 3C). Altogether, these results imply that both GldD and GldG proteins are required for proper adhesion of F. psychrophilum cells to surfaces, as well as for auto-adhesion ability required for biofilm development.

As described above, inactivation of F. psychrophilum gldD or gldG genes led to pleiotropic phenotypes in vitro. The impaired bacterial surface functions such as extracellular proteolytic activity, cell sedimentation and cell adhesion of strains gldD::Tn and gldG::Tn suggest an effect of these mutations on protein secretion efficiency. To test this hypothesis, we used a label-free proteomic approach that combined SDS-PAGE electrophoresis and LC-MS/MS analyses, to compare the exoproteomes of strains THCO2-90, gldD::Tn and gldG::Tn (Table S1). Among the 414 proteins identified in total, 90 and 158 were significantly altered in abundance in the culture supernatant of strains gldD::Tn and gldG::Tn, respectively, compared to those of wild-type strain. Among them, 20 and 9 proteins were not detected at all for strains gldD::Tn (Table 2) and gldG::Tn (Table 3), respectively, including the extracellular protease Fpp1 and a probable S8 subtilisin family serine endopeptidase. Strikingly, among the proteins significantly less abundant in the mutant strains, proteins involved in T9SS, gliding and other proteolytic enzymes were identified (i.e., GldN, GldK, and PorT that belong to the core T9SS machinery, the PorU peptidase; the SprB adhesin; the Fpp2 extracellular protease and the collagenase; Tables 2, 3). Among the 39 and 9 proteins predicted to have either a TIGR04183 or a TIGR04131 CTD in the proteome of strain THCO2-90, 29, and 5, respectively, were not found or significantly less abundant in both mutants' supernatants. Finally, 11 and 83 proteins were more abundant in the culture supernatant of strains gldD::Tn and gldG::Tn, respectively, compared to the wild-type strain. Among them, most are of unknown function, and none possess a CTD domain (Table S1).

Cell surface shaving with trypsin and LC-MS/MS analysis of strains gldD::Tn and wild-type enabled the identification of 426 proteins (Table S2). Among them, 74 were significantly altered in abundance in strain gldD::Tn compared to the wild-type strain. Indeed, 39 proteins (8 with a CTD domain) were not detected or significantly less abundant, including the GldJ gliding protein and the GldN T9SS machinery subunit (Table 4). In contrast, 35 proteins were more abundant in strain gldD::Tn such as the predicted cysteine protease FcpB and the gliding motility precursor RemF. Contrary to results obtained in the supernatant fractions, the SprB surface adhesin and the PorU peptidase were not detected in the wild-type cells, whereas these proteins were identified in strain gldD::Tn (Table 4).

When analyzing the surfome of gldG::Tn compared to the wild-type strain, 582 proteins were identified in total (Table S2). Among the 69 proteins altered in abundance in strain gldG::Tn, 14 were not detected or significantly less abundant, including the above-mentioned GldJ, GldK, and GldN or the collagenase (Table 5), which is in agreement with the results obtained in the supernatant fractions (Table 3). In contrast, and as observed for strain gldD::Tn, 55 proteins were undetected or less abundant in the wild-type surfome, including SprB, PorU, and RemF (Table 5).

This proteomic analysis reveals that inactivation of gldD or gldG by Tn4351 results in a major defect in surface protein localization, especially for those proteins likely translocated through the T9SS apparatus.

The effect of gldD or gldG inactivation on the virulence of F. psychrophilum was investigated in rainbow trout using two different routes of infection, intramuscular injection and immersion. First, the LD₅₀ of strain THCO2-90 was established at 6.0 × 10³ bacteria by performing intramuscular injection challenge in rainbow trout. This value was used as a reference to compare the virulence of strains THCO2-90, gldD::Tn and gldG::Tn. Groups of 10 fish were injected with 2 x 10⁴ CFU of strains THCO2-90, gldD::Tn or gldG::Tn. This dose theoretically corresponds to 4-fold the LD₅₀ of the wild-type strain. Fish challenged with the wild-type strain quickly died and the cumulative mortality reached 100% 8 days post infection. In contrast, strains gldD::Tn and gldG::Tn showed cumulative mortalities of 20 and 0% respectively by 8 days, and 40 and 20% by 14 days (Figure 4).

We further investigated the impact of gldD or gldG inactivation on virulence using an immersion challenge model that supposedly more closely mimics the natural infection route. Fish were bathed for 24 h in water contaminated by bacterial culture of strains THCO2-90, gldD::Tn or gldG::Tn (theoretical initial concentration 5 × 10⁶ CFU mL⁻¹). Following the bath challenge, an increase in bacterial concentration in water was observed for all tanks, whatever the strain (Table 6). Four days post-infection, fish cumulative mortality reached 100% with the wild-type strain, while the mortality of fish infected with strains gldD::Tn or gldG::Tn was similar to that of the non-infected group (≤8%; Figure 5A). When bath-challenged with the wild-type strain, the bacterium was detected on the surface and in organs of all fish arbitrarily sampled 6-h after the end of immersion. The bacterial load in spleen and gills was on average 3.5 × 10⁴ and 2.8 × 10⁴ CFU, respectively. In contrast, the bacterium was only detected in 2 out of 12 fish infected with strains gldD::Tn or gldG::Tn. For those fish, the bacterial loads were 10² and 3.5 × 10² CFU in the spleen and 1.4 × 10² and 2.6 × 10² CFU in the gills, respectively (Figures 5B,C). In addition, F. psychrophilum was always detected in samples of skin mucus of fish infected with the wild-type strain, while the bacterium was systematically absent from the skin mucus of fish infected with the mutant strains (12 fish sampled for each group; data not shown).

These results revealed that the virulence of strains gldD::Tn and gldG::Tn is strongly attenuated in rainbow trout whatever the infection route, with considerably lower invasion and proliferation abilities in vivo.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.