Bacterial strains and plasmids used throughout this study are listed in Table 1 . P. salmonis strains were cultivated in P. salmonis broth (PSB) or on its solid equivalent PSA at 18°C (Henriquez et al., 2016). E. coli strains were propagated in Luria-Bertani broth (LB, Difco) or on tryptic soy agar plates (TSA, Difco) at 37°C. Plasmids were introduced into E. coli strains by transformation and into P. salmonis strains through mating with the donor strain E. coli S17-1λpir (Simon et al., 1983). Where necessary, antibiotics (Sigma) were used at the following concentrations: kanamycin at 50 μg/ml and polymyxin B at 10 μg/ml. Any potentially infectious residue was inactivated by autoclaving before disposal according to the biosafety protocols of ADL Diagnostic Chile. Plasmids pCR2.1 TOPO (Invitrogen) and pJQK (Quandt and Hynes, 1993) were manipulated according to standard cloning procedures.

The P. salmonis PM15972A1 genome sequence was retrieved from the GenBank database (accession n° CP012413). The annotation was examined with CLC Genomics Workbench software (Qiagen), looking for genes encoding functions related to the biosynthesis of precursors of LPS and EPS. Sequence alignments were accomplished with BLAST. For comparative purposes, sequences from identified gene clusters were compared with those annotated in the KEGG database (https://www.genome.jp/kegg).

All primers used in this study are listed in Table 2 . The sequences of the wzx and wcaJ genes of P. salmonis PM15972A1 were obtained from the NCBI (National Center for Biotechnology Information, http://www.ncbi.nlm.nih.gov/) (Bohle et al., 2014). P. salmonis wzx or wcaJ in-frame deletion mutants were engineered by overlapping PCR following a protocol previously described for Brucella (Conde-Alvarez et al., 2006). Briefly, two DNA fragments obtained with the primer pairs wzx_F1/wzx_R2 and wzx_F3/wzx_R4 were ligated by PCR extension, and the resulting fragment was cloned into pCR2.1-TOPO supported by the TOP10 E. coli strain ( Figure S1 ). Subcloning into the pJQK plasmid was achieved by restriction digestion of the corresponding pCR2.1 TOPO derivative and transformation into the S17-1λpir E. coli donor strain. The resulting pJQK derivative was introduced into P. salmonis by conjugation. Transconjugants carrying the pJQwzx plasmid integrated into the chromosome by a single crossover event were then selected on PSA containing kanamycin and polymyxin B. Allelic exchange between the chromosomal gene and the mutagenized plasmid copy was achieved by a second crossover event and was counter-selected on PSA containing 5% sucrose ( Figure S2 ). An identical procedure but different primers were used for the construction of the pJQwcaJ mutator plasmid and the corresponding P. salmonis ΔwcaJ mutant strain. The presence of the respective mutated genes in each strain was confirmed by the analysis of PCR amplicons generated with primers F1-R4. Mutated genes were also sequenced by Sanger chemistry to confirm the interchange with the wild-type allele in the resulting mutant clones. Bacterial strains were stored in glycerol stocks at -70°C.

100 µL aliquots of deep-frozen P. salmonis wild-type and mutant strains were streaked onto PSA plates and incubated for 2 days at 18°C. By this time, a homogeneous bacterial lawn had formed and we proceeded to suspend a loopful of bacteria in 1 mL of buffered saline solution (BSS: 8.5 g/L NaCl, 2.0 g/L K₂HPO₄, 2.0 g/L KH₂PO₄, pH 6.4), transferred them onto new PSA plates, and incubated as mentioned above. Bacterial suspensions (OD₆₀₀ = 1.0) were prepared by mixing bacteria with 5 mL of cold BSS in a conical tube kept at 10°C. For growth kinetics, 96-well microplates were inoculated with 150 µL/well of 1:100 bacterial suspensions in PSB and incubated during 70 h. Hourly reads were realized with an EPOCH2 spectrophotometer (Biotek, USA). By contrast, self-agglutination tests lasted 8 h, but used bacterial suspensions prepared in PSB that were held in plastic cuvettes at 18°C and pH 7.2. Reads were collected each hour with a densitometer (Ultrospec 10, GE LifeSciences). In order to determine bacterial counts, aliquots of log₁₀ dilutions in PSB were seeded on PSA plates in triplicates and incubated at 18°C for one week. Acriflavine tests were carried out using fresh bacterial suspension in BSS mixed with identical volumes (10 µL) of 0.1% acriflavine on a glass slide.

The complement killing activity of naïve serum was evaluated following a previously described protocol (Saavedra et al., 2017). Atlantic salmon (Salmo salar) smolts with an average body weight of 96 g were caudally punctured, and naïve serum was obtained after blood clotting for 2 h at 4°C and centrifugation at 2,000 × g for 5 min. Samples from 5 fish were pooled. Bacterial inoculum was set at ~10⁷ colony forming units (cfu)/mL and was treated with undiluted fish serum for 3 h at 18°C. Heat-inactivated (44°C, 20 min) serum aliquots were used as negative controls.

LPS was extracted from P. salmonis wild-type and mutants following a previously described method (Yi and Hackett, 2000). 15 mL of a fresh culture with OD₆₀₀ = 1.9 were harvested, which corresponded to roughly 10 mg of freeze-dried bacteria. LPS was isolated with a commercial RNA isolation reagent (Trizol) and subsequently resuspended in 200 µL of PBS. Discontinuous SDS-PAGE electrophoresis was run on 15% and 20% polyacrylamide stacking/solving gels. Gels were silver-stained according to a well-known protocol (Hitchcock and Brown, 1983). Alternatively, LPS was extracted without organic solvent following a protocol described before (Renzi et al., 2016). Briefly, 1 mL of bacterial suspension in PBS, adjusted to OD₆₀₀ = 1.0, was pelleted by centrifugation and subsequently lysed with Laemmli buffer. Samples were boiled and treated twice with proteinase K (50 µg/mL final protease concentration, overnight at 37°C). Crude protein samples and LPS prepared following the recommendations by Renzi and colleagues were used for independent Western blot assays. These samples were eletrophoresed in 15% polyacrylamide gels. Nitrocellulose membranes were blotted according to a conventional protocol, using semi-dry transfer equipment (Bio-Rad) for 35 min at 20 V. These blots were incubated with polyclonal serum extracted from a rabbit that had been immunized with heat-inactivated P. salmonis PM15972A1 cells. Signals were then revealed with an HRP-conjugated secondary antibody and a suitable chemical substrate.

Biofilm formation assays were performed following a previously described method (Pal et al., 2019), with few modifications. Briefly, the wells of a sterile 24-well polystyrene microtiter plate were inoculated with 1 mL of P. salmonis wild-type and mutant strain suspensions adjusted in PSB to obtain ~7.0 × 10⁶ cfu/well. Sterile PBS was used as a negative control. Plates were incubated at 18°C for 30 days under static conditions. Then, wells were washed twice with BSS to remove the planktonic cells and stained with 0.1% crystal violet (CV) for 15 min. Excess CV was removed by renewed washing with BSS, and bound dye was solubilized in 1 mL of 33% acetic acid. Solubilized CV was subsequently transferred to a new microplate and its absorbance was spectrophotometrically determined at 620 nm. This same measurement was done on negative-control wells containing bacteria-free medium in order to account for non-specific and abiotic factors. Total biofilm formation was calculated by subtracting the average OD₆₂₀ value of the stained control wells from the OD₆₂₀ value of the stained biofilm (Naves et al., 2008). Calculated values were finally normalized to those obtained for the P. salmonis PM15972A1 positive control.

So as to evaluate possible differences in gene expression triggered by the treatment with P. salmonis wild-type and mutant strains, RNA extraction and cDNA synthesis were performed as described by Mancilla et al. (2018). RNA was purified using the EZNA Total RNA Kit I (Omega Bio-Tek, Norcross, GA, USA). After DNAse I digestion, purified RNA was stored at -70°C. Reverse transcription was performed using the M-MLV RT Kit (Promega, USA) and synthesized cDNA was stored at -20°C. Gene expression was analyzed by qPCR using the Maxima SYBR^® Green qPCR Master Mix (Thermo Scientific) in a StepOne PCR machine (Applied Biosystems, Waltham, MA, USA), whereby each reaction was performed using 250 nM of primers ( Table 2 ) and 1.0 µL of 1:2 diluted cDNA as a template. The thermal profile included an initial step of 10 min at 95°C, followed by 40 cycles of 15 s at 95°C and 1 min at 60°C. The fold change of gene expression was calculated according to the 2^-ΔΔCt method (Pfaffl, 2001), using the expression of the elongation factor 1α (efl1α) as a normalizer and the expression of all markers under control conditions as a calibrator.

SHK-1 cells (passage 72) were cultured without antibiotics in Leibovitz’s L-15 medium supplemented with 10% fetal bovine serum (FBS) at 20°C in 75-cm² flasks. Cells were seeded into 24-well plates at a concentration of ~1.0 × 10⁵ cell/well, incubated for 24 h at 20°C and then infected with the selected P. salmonis strain at a multiplicity of infection (MOI) = 10. Purified nucleic acid samples were collected from the cell culture at 5 and 10 days post infection (dpi). Cell damage induced by P. salmonis in SHK-1 cell monolayers was determined by the release of the cytosolic enzyme lactate dehydrogenase (LDH) into the medium using a commercial Cytotoxicity LDH Detection Kit (Thermo Scientific). LDH activity levels were analyzed in 50 µL aliquots of cell-free supernatant obtained from each well. Additionally, the supernatant of cells lysed with L-15 medium containing 1% Triton X-100 was used as a total lysis control (maximum LDH liberation) and the supernatant of uninfected cells was used as a negative control (basal LDH liberation). Optical densities measured at 490 nm and 680 provided the basis for calculations as recommended by the manufacturer of the detection kit.

The in vitro infectivity of P. salmonis was assessed using an endpoint dilution assay (TCID₅₀) as described early, with few modifications (House et al., 1999). CHSE-214 cells were cultured in 96-well flat-bottom microplates in antibiotic-free L-15 Medium supplemented with 10% FBS and incubated at 20°C until they reached confluence (~5.0 × 10⁴ cells/well). As inoculum, bacterial suspensions (~5.0 × 10⁷ cfu/mL) were prepared in L-15 supplemented with 2% FBS, and log₁₀ dilutions were made up to the 10^th dilution. Monolayers were infected (8 wells/dilution), incubated for 15 days at 18°C, and then visually inspected for the presence of cytopathic effects. TCID₅₀ were calculated according to the Spearman–Kärber algorithm (Hierholzer and Killington, 1996).

Atlantic salmon smolts were confirmed as “pathogen free” using RT-PCR routine diagnostic methods for IPNV, ISAV, PRV, P. salmonis and Renibacterium salmoninarum. After their transfer to the experimental center, fish were tagged using Passive Integrated Transponder tags (PIT tags). The animals were allowed to acclimatize to seawater for 21 days and fed ad libitum; they were maintained at 14 ± 1°C, with a water change rate of 1.2 - 1.5/hour, pH of 7.0 - 8.0, a photoperiod of 24 light hours, and a specific feeding rate of 1.5 - 2%. Prior to any handling and marking, fish were anesthetized with benzocaine. Finally, euthanasia was performed using an overdose of anesthetic. All efforts were made to provide best growth conditions and minimize suffering.

The in vivo challenge was carried out on 420 fish with an approximate weight of 90 - 100 g. These fish were uniformly allocated to three 0.5 m³ tanks, yielding initial densities of 25 - 30 kg/m³ and 140 fish/tank. 35 fish of each tank comprised a single treatment group that received 0.1 mL of bacterial suspension (either P. salmonis wild-type, P. salmonis Δwzx, or P. salmonis ΔwcaJ) or vehicle (BSS). Fresh bacterial suspensions (OD₆₀₀ = 0.7) were used to prepare the inocula that were administered to the fish via intraperitoneal injection. In parallel, a sample of each inoculum was utilized to perform bacterial counts at the time of injection.

Fish were monitored daily for 30 days, and mortalities were removed from the tanks. As for tanks 1 and 2, mortalities were recorded and subjected to necropsy. Additionally, PCR of internal organ samples were done to assess for the presence of P. salmonis. The third tank was intervened on days 5, 12 and 20 post infection, removing three live fish/group to collect head kidney tissue samples for gene expression analysis.

Data obtained from the bactericidal serum effect studies were tested for normal distribution using the Shapiro–Wilk test and were subsequently analyzed with the non-parametric Mann–Whitney rank-sum test. For biofilm quantification and cytotoxicity assays, the mean and standard deviation values were compared with the Kruskal-Wallis test and Dunn’s multiple comparison test as a post-hoc analysis (p < 0.05). These same tests were applied to the mean and standard deviation values of the fold change of gene expression, which are given in terms of relative expression. PCR efficiencies were determined by linear regression analysis of sample data using LinReg-PCR. With regard to in vivo testing, the percentages of cumulative mortalities were analyzed using the Kaplan-Meier method, and the differences were evaluated using the log-rank test performed with GraphPad Prism v6.0.

BLAST searches and annotation analyses allowed for the identification of a ~18 kb cluster carrying 15 genes putatively involved in the biosynthesis of LPS ( Figure 1 ). Those genes presumably encode enzymes involved in furnishing (glycosyltransferases), O-chain transport across the membrane, and precursor biosynthesis ( Table 3 ). An O-antigen flippase (wzx, KW89_2967, 462 amino acids), which catalyzes the transport of the O-antigen unit across the inner membrane, was predicted. We also found genes encoding proteins possibly participating in the polymerization of the O-antigen (wzy, KW89_2964; wzz, KW89_2972). Surprisingly, homologs of the waaL O-antigen ligase, which facilitates the transference of polymerized O-antigen to the core-lipid A, could not be identified. A putative undecaprenyl-phosphate glucose phosphotransferase or wcaJ (KW89_2959, 466 amino acids) was found to be located in that same cluster, but upstream of wzx and in close proximity to another gene encoding an S-layer protein. The enzyme encoded by wcaJ is allegedly responsible for the initiation of the synthesis of colanic acid, which is part of the EPS and the capsule of the Enterobacteriaceae family members. In detail, undecaprenyl-phosphate glucose phosphotransferase accounts for the transfer of glucose-1-phosphate onto the carrier lipid undecaprenyl phosphate. Interestingly, a second cluster of ~8.5 kb containing genes for the biosynthesis of a capsular polysaccharide was found ( Figure S3 and Table S1 ). Some genes of this cluster show similarities with orthologs in Francisella philomiragia and F. noatunensis subsp. orientalis.

According to the literature, a knock-out mutation of wzx or wcaJ would carry a detrimental effect on the biosynthesis of LPS and EPS, respectively. Therefore, altered virulence phenotypes of the corresponding mutants could be expected if those structures play a role in the pathogenesis of piscirickettsiosis.

Since the target genes are likely to be part of operon units, the mutagenesis approach was engineered to yield in-frame deletion mutants. Mutator plasmids were designed using the pJQK genetic tool, a suicide plasmid already tested in P. salmonis (Mancilla et al., 2018). We have chosen a strategy similar to that used in other bacterial models, replacing the wild-type gene by a short, afunctional copy of the same gene. This way, we could avoid polar effects and remove antibiotic resistance markers ( Figure 2A and Figure S4 ). The interchanged alleles remaining after the mutagenesis process were detected and confirmed by conventional PCR ( Figure 2B ). It shall be mentioned that the mutant strains have lost their kanamycin resistance markers and displayed minimal inhibitory concentration (MIC) values at control, wild-type levels.

Cell envelope mutants frequently present colony phenotypes that are easily distinguished on agar plates. However, isolated colonies from the P. salmonis mutants described in this study looked macroscopically identical to the wild-type strain (not shown). Growth kinetics in liquid medium showed no differences between the mutants and the parent strain ( Figure S5 ). Considering prior reports on the self-agglutination ability of mutants with cell envelope defects, we studied our P. salmonis strains in this regard. On the one hand, we measured the optical density of bacterial suspensions in liquid medium, and we conducted acriflavine tests. In liquid medium, mutant strains displayed a higher self-agglutination capacity than their parent strain, and this characteristic was particularly pronounced in ΔwcaJ ( Figure 3A ). On the contrary, the acriflavine test was positive only for the Δwzx strain, where cell clumps became evident ( Figure 3B ). Since LPS is the main target of some antibiotics, such as polymyxin B, we assessed the mutant strains’ MIC to this antibiotic. Unexpectedly, similar MIC values were determined for the parent and mutant phenotypes (8 - 16 µg/mL).

In a Δwzx genetic background, it is expected to find O-chain-deficient LPS species with a lower molecular mass when compared with the parent strain. That feature can be explored by electrophoretic profiling, so we examined the profiles of LPS samples from the parent and mutant strains in a 15% conventional acrylamide gel. Consistent with previous reports (Kuzyk et al., 1996; Vadovic et al., 2007), we visualized a single-band profile with an apparent molecular mass of 11 kDa, but did not find any differences among tested strains (not shown). Upon increasing the resolution of polyacrylamide gels, we observed subtly augmented electrophoretic mobility for the Δwzx LPS fraction, when compared with the parent and ΔwcaJ mutant strains ( Figure 4A ). This pattern resembled the canonical core-lipid A profile. In a next step, we used crude protein samples that would be expected to carry LPS, performed Western blots and finally incubated with immunoreactive rabbit serum against P. salmonis wild-type. While samples corresponding to the parent and ΔwcaJ mutant strains yielded similar profiles, the Δwzx mutant lacked a band ( Figure 4B ). An identical pattern was observed on a Western blot of LPS samples extracted without organic solvent and extensively treated with proteinase K ( Figure 4C ).

Some bacterial pathogens are able to resist the killing activity of naïve serum. While this applies to P. salmonis, the resistance of this pathogen seems to depend on its genetic background (Saavedra et al., 2017). In this regard, we found the parent and the Δwzx mutant strains to display partial susceptibility to salmon naïve serum. The ΔwcaJ strain, however, proved to be completely susceptible to the serum treatment and was unable to survive ( Figure 5 ).

Available experimental evidence does not unequivocally confirm the role of the wcaJ gene in the biosynthesis of EPS. While a knock-out mutation of this gene in the fish pathogen Edwarsiella tarda caused an impairment of biofilm production (Yu et al., 2015), enhanced biofilm formation was described for Klebsiella pneumoniae carrying the same mutation (Pal et al., 2019). As for our P. salmonis strains, the evaluation of biofilm production after 30 days of incubation showed that the disruption of wcaJ is associated with enhanced biofilm formation. The corresponding mutant strain formed up to 50% more biofilm than the wild-type strain. In contrast, no differences were observed between the Δwzx and wild-type strains ( Figure 6 ).

The CHSE-214 cell line has been described as susceptible to infection with P. salmonis. Hence, we hypothesized that using comparable amounts of bacteria to perform TCID₅₀ studies would reveal potential differences in the virulence of mutant strains. We thus conducted the assays at a similar MOI. Notwithstanding, titers reached by the mutants and the parent strain were in the same order of magnitude and ranged from 6.0 - 6.7 × 10⁵ TCID₅₀/ml. In a second, independent experiment, calculated titers were somewhat higher, but still largely identical across P. salmonis strains, ranging from 8.0 - 9.3 × 10⁵ TCID₅₀/ml. In order to consolidate these findings, we decided to further assess the virulence of P. salmonis mutants in SHK-1 cells. The direct observation of cytopathic effects is less reliable in SHK-1 cells, which is why we opted for the quantification of LDH release and activity as an alternative, more sensitive measure of the in vitro virulence of P. salmonis strains in this cell line. To this end, monolayers of SHK-1 cells were infected at a lower MOI and incubated at 18°C for 10 days. In this model, both mutants exhibited lowered cytotoxicity when compared with their parent strain ( Figure 7 ). The Δwzx mutant was the most attenuated one.

Pathogen mutants with cell envelope defects tend to induce differential patterns of gene expression of immune response markers in host cells. We tested this hypothesis for P. salmonis strains by determining the expression profiles of a proinflammatory cytokine gene panel in SHK-1 cells, sampled at 5 or 10 dpi. All strains were able to stimulate a proinflammatory response, but clear differences became evident ( Figure 8A ). The Δwzx strain upregulated the expression of il1β and downregulated the expression of tnfα at 5 dpi. At 10 dpi, tnfα and ifnγ were downregulated. Interestingly, cells infected with the ΔwcaJ strain showed a notable upregulation of tnfα at both points in time (>20-fold change), while il1β was found to be upregulated at 5 dpi but dropped below reference values at 10 dpi. As for the expression of ifnγ after infection with the ΔwcaJ strain, a clear downregulation was registered at 5 dpi. Gene expression was also examined in samples derived from the in vivo assay. An upregulation of the expression of il1β and tnfα was recorded at 5 dpi in fish infected with all strains tested ( Figure 8B ). Consistent with the pattern observed in vitro, a downregulation of ifnγ expression was observed at 5 dpi in the group infected with the ΔwcaJ strain. Interestingly, at 12 dpi, expression profiles of all markers tended to level off at reference levels, with the exception of the ifnγ transcript. The expression pattern for this cytokine gene was characterized by persistent downregulation in fish treated with the ΔwcaJ strain. Extensive mortality was registered in groups infected with the parent and ΔwcaJ strains; only fish from the Δwzx-challenged group survived up to 20 dpi. Therefore, no comparison of gene expression could be made at this sampling point in time. We highlight the reliability of our qPCR data, which it can be deduced from the normalizer behavior across samples ( Figure S6 ).

The in vivo assay was carried out infecting S. salar smolt cohorts with similar amounts of P. salmonis strains via intraperitoneal injection. According to our experience with the challenge strain, the horizontal transmission of P. salmonis PM15972 is negligible in the period set for the experiment, and any mortality occurring within that time frame may be interpreted as an effect of the injected material. The target concentration of the inoculum was set at 1.0 × 10⁷ cfu/fish (Mancilla et al., 2018). Precisely, individual fish were inoculated with 3.0 × 10⁷ and 2.6 × 10⁷ cfu of the parent strain and Δwzx mutant, respectively, while the ΔwcaJ treatment group received 1.3 × 10⁶ cfu/fish. As shown in Figure 9 , the group injected with the parent strain accumulated almost 100% mortality within a few days. Treatment with the ΔwcaJ mutant induced a similar mortality curve, although it was slightly displaced in time. Fish infected with P. salmonis Δwzx had a significantly higher probability of survival in both tanks. Necropsies conducted on dead fish revealed pathognomonic signs of piscirickettsiosis in internal organs, regardless of the challenge strain (not shown).