Piscirickettsia salmonis type strain LF-89^T (ATCC VR-1361) and the field Chilean strain CA5 were studied. These strains have been previously identified as belonging to the LF-89 and EM-90 genogroups, respectively. The field strain was isolated in 2012 from the liver of a clinically infected Atlantic salmon (Salmo salar) during a piscirickettsiosis outbreak affecting a Chilean farm in the Aysén Region. The two strains were confirmed as P. salmonis by standard phenotyping and routine internal transcribed spacer-based PCR analysis (Marshall et al., 1998). The strains were routinely cultured at 18°C for 4-5 days in Austral-TSHem agar plates (Yáñez et al., 2013) and agitated-liquid AUSTRAL-SRS medium at 120 rpm (Yáñez et al., 2012). The strains were stored in Cryobille tubes for long-term storage at -80°C, as per the manufacturer’s instructions (AES Laboratory). The purity of the strains was corroborated by Gram staining and optical microscopy prior to beginning experiments.

Once the two P. salmonis strains were routinely grown in AUSTRAL-SRS medium, they were then inoculated into 96-well microplates (flat-bottom, SPL Life Sciences Co., Ltd.) to determine the specific biofilm formation (SBF) index using the crystal violet (CV) method. Briefly, inocula were adjusted to ~0.8 McFarland and sizes were determined by counting the number of colony-forming units (CFU) on Austral-TSHem agar plates (incubated as described above) by the standard serial-dilution method. Later, microplate wells were inoculated with aliquots of 150 μL of AUSTRAL-SRS medium (8 wells per inoculum) containing 7.86 ± 6.06 × 10⁶ (i.e., 0.81 ± 0.02 McFarland) and 3.42 ± 1.61 × 10⁶ CFU mL^-1 (i.e., 0.81 ± 0.01 McFarland) of P. salmonis LF-89^T and CA5, respectively. Similarly, eight wells per microplate were inoculated with 150 μL of sterile AUSTRAL-SRS broth as negative controls. Three independent experiments were performed by inoculating thirteen 96-well microplates in each experiment. Inoculated microplates were statically incubated at 18°C until further individual processing at 24 h intervals to evaluate biofilm development over a 13-day total period. After incubation, inoculated and control wells were emptied and washed (3X) with 200 μL of sterile milli-Q water and then stained with 180 μL of CV solution (1% w/v) at room temperature for 15 min. Afterward, the CV solution was discarded, and the wells were washed with abundant tap water until no more dye was released. The microplate was turned upside down to dry at room temperature for 15 min before adding 180 μL of absolute ethanol per well for CV solubilization at room temperature for 15 min. The SBF index was determined by reading absorbance of the EtOH-solubilized CV at 585 nm using a Tecan Microplate Reader (Infinite 200 PRO, Männedorf, Switzerland) and computed in accordance with Niu and Gilbert (2004) as (B - NC)/G, where B is the amount of EtOH-solubilized CV released from biofilms; NC is the amount of EtOH-solubilized CV that adhered to the microplate surfaces in negative controls; and G is the absorbance of bacterial supernatants at 620 nm.

The physiological characteristics for 48-h-old biofilms were determined using the miniaturized API ZYM system in accordance with the manufacturer’s directions (API-BioMerieux, La Balmeles-Grottes, France), but with two modifications regarding the incubation temperature (at 18°C) and the time elapsed between inoculation and revealing (4 to 5 days). Briefly, 48-h-old biofilms were formed on the bottom of sterile glass Petri dishes by inoculating 20 mL of AUSTRAL-SRS medium containing P. salmonis LF-89^T and CA5 at ~1 × 10⁶ CFU mL^-1. Once biofilms were established, the supernatants were withdrawn to wash (3X) sessile bacteria and discard non-adherent ones with 10 mL of chilled sterile NaCl solution (2.5% w/v). A washed Petri dish of each strain was randomly selected for staining with CV (1% w/v) at room temperature for 10 min. Once the dye was discarded, CV-stained Petri dishes were washed with tap water (until no more dye was released) and dried at room temperature for 15 min ( Supplementary Figure S1A ) to confirm the formation of biofilms under an optic microscope at 1,000X magnification ( Supplementary Figure S1B ). The remaining washed Petri dishes were used to prepare 48-h-old biofilm inocula by resuspending the sessile bacteria from the bottoms by using cell scrapers and 5 mL of saline solution (2.5% NaCl). To verify the biofilm detachment procedure, some scraped Petri dishes were randomly selected for CV (1% w/v) staining ( Supplementary Figure S1C ) and optical microscopy ( Supplementary Figure S1D ). Before assays, biofilm inocula of P. salmonis LF-89^T and CA5 were adjusted to 6.0 McFarland equivalent to 1.40 ± 0.87 × 10⁸ CFU mL^-1. Planktonic inocula of both strains were harvested by centrifugation at 9,000 rpm and 4°C for 2.5 min from culture supernatants, and the resulting pellets were washed (3X) with a sterile saline solution and resuspended in the same saline solution at 1.65 ± 0.65 × 10⁸ CFU mL^-1 for subsequent API ZYM assays for comparison purposes.

Changes in biofilm cell viability over time were determined with the LIVE/DEAD BacLight Bacterial Viability Kit (Invitrogen, Molecular Probes, Oregon, USA); this kit was evaluated prior to use in samples. To do so, 1 mL-aliquots of fresh P. salmonis cultures (prepared in AUSTRAL-SRS medium) were fixed overnight in formaldehyde (10% v/v) to obtain dead cells. Afterward, fixed cells were harvested by centrifugation at 4°C and 10,000 rpm for 2.5 min, washed (3X) with sterile NaCl solution (2.5% w/v), and then resuspended in the same saline solution. To obtain living cells, 1 mL aliquots of fresh cultures were directly harvested by centrifugation at 4°C and 9,000 rpm for 2.5 min, washed (2X) in sterile NaCl solution, and resuspended in the same saline solution. Afterward, bacterial suspensions with different live-to-dead cell ratios (0: 100, 10: 90, 50: 50, 90:10, and 100: 0; hereafter, expected ratios) were prepared by mixing live and dead bacterial suspensions. The expected ratios of live-to-dead bacteria in bacterial suspensions were correlated with the empiric ratios of live-to-dead bacteria in the same bacterial suspensions by using the Pearson’s correlation coefficient. Empiric ratios of live-to-dead bacteria in bacterial suspensions (with different expected ratios of live-to-dead bacteria) were determined by direct cell counting on an Olympus BX41 microscope at 1,000X magnification (Olympus Corporation, Tokyo, Japan). Pearson’s correlation coefficients between expected and empiric ratios of live-to-dead bacteria indicated that the LIVE/DEAD Viability kit detected physiologically contrasting states of P. salmonis LF-89^T (r = 0.9995, P < 0.05) and CA5 (r = 0.9975, P < 0.05).

Three independent experiments were conducted to determine the changes in both the surface coverage and bacterial viability of biofilms formed at 18°C and at 24 h intervals over a 360 h period. The two P. salmonis strains were grown in AUSTRAL-SRS medium and, prior to conducting the experiments, inoculum sizes were adjusted to 5.09 ± 0.59 × 10⁶ CFU mL^-1 (0.79 ± 0.01 McFarland) and 3.52 ± 3.15 × 10⁶ CFU mL^-1 (0.80 ± 0.03 McFarland) for P. salmonis LF-89^T and CA5, respectively. Ninety-six-well microplates were inoculated in triplicate (at 24 h intervals) with 150 µL of McFarland-adjusted inocula to monitor different stages of biofilm development (i.e., from early to advanced states of biofilm formation, including the mature stage) on a single microplate. Furthermore, triplicate 150 µL aliquots of sterile AUSTRAL-SRS medium were included as negative controls in each experiment. After incubation, inoculated and negative-control wells of microplates were emptied and washed (3X) with 200 μL of a sterile NaCl solution (2.5% w/v). Later, washed biofilms were stained with the LIVE/DEAD Kit as per the manufacturer’s instructions, except that dye salts were dissolved in a sterile NaCl solution (2.5% w/v).

Endpoint images (single-point reading) from LIVE/DEAD-stained biofilms were automatically captured on a Cytation 5 Imaging Multi-Mode Reader (BioTek Instruments Inc., Winooski, VT, USA) by using the 20X objective with the correction collar set to 0.5 mm (for the usual bottom thickness of 96-well microplates). Image capturing was carried out in two fluorescence channels; green (GFP: 469, 525) and acridine orange (AO: 469, 647). Optimal image exposure was achieved using LED intensities of 3 and 5 (for GFP and OA channels, respectively), an integration time of 10 and 35 ms (for GFP and OA channels, respectively), and a camera gain equal to zero (both channels). The laser autofocus method allowed us to find the optimal focus position using a laser beam that corrects for any irregularities in plastic. Once the trained position was found, the instrument used the described channels and settings to automatically capture the biofilms formed on the bottom of each well. In this regard, the bottom elevation (a parameter that refers to the distance in microns between the top of the microplate carrier and the bottom of the well where the sample is visible) was 3,250 microns. Additionally, endpoint fluorescence intensities were quantified for each well using the cell imaging multi-mode reader with a monochromator-based system, with excitation wavelength and bandwidth set to 485/20 nm and emission wavelengths and bandwidths set to 530/20 nm (green emission for live bacteria) and 630/20 nm (red emission for dead or damaged bacteria). The “wavelength switch per well” feature was set to “on” for sequential signal detection of both emissions before moving to the next well. Reads were performed using top fluorescence optics, a normal read speed (i.e., delay after plate movement of 100 ms), 10 measurements per data point, a height of 7 mm, and an extended dynamic range for automatic gain adjustment. Fluorescence data were corrected by the background fluorescence emitted by negative-control wells and expressed as the ratio of the green-to-red fluorescence; this approach has previously been used as a proxy to evaluate the changes in bacterial viability of biofilms formed by other strains of P. salmonis (Levipan et al., 2020).

For scan laser confocal microscopy, sterile glass slides were dipped into 20 mL of P. salmonis LF-89^T and CA5 cultures (prepared in AUSTRAL-SRS medium at ~1 × 10⁶ CFU mL^-1) inside sterile glass Petri dishes. After static incubations, 48-h-old biofilms formed on glass slides were washed (3X) with a chilled sterile NaCl solution (2.5% w/v) and stained with the LIVE/DEAD Viability kit (following the manufacturer’s instructions) to determine the contribution of living and dead (or inactive) bacteria to the biofilm and the distribution of sessile bacteria on glass slides. A drop of mounting antifade oil (Invitrogen) was placed on LIVE/DEAD-stained slides and covered with a coverslip for subsequent microscopic observation on a Leica TCS SP5 II spectral confocal microscope (Leica Microsystems Inc., Jena, Germany). The images were collected using the HCX PL APO 100x/1.44 Oil CORR CS objective and the Leica Confocal software version 2.6 (Leica Inc.). In addition, 48-h-old biofilms formed on other set of glass slides were fixed in chilled methanol for 15 min, dipped once in sterile NaCl solution (2.5% w/v), air-dried at room temperature, and stained as per the manufacturer’s specifications with Alexa fluor 488-conjugated wheat germ agglutinin (Invitrogen) (5 μg mL-1) and 4’,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich). Wheat germ agglutinin- and DAPI-stained slides were mounted using a drop of Dako mounting medium (Invitrogen) and coverslips for subsequent observation and imaging of the exopolysaccharide matrix of biofilms with embedded bacteria. The images were collected as previously described, and at least 10 microscopic fields were analyzed per sample.

RNA extraction and subsequent RNA-based experiments were carried out using P. salmonis strain LF-89^T considering three criteria. First, strain LF-89^T is the type of the species, the most studied worldwide, and is available in bacterial culture collections. This allows researchers to replicate published results as P. salmonis LF-89^T can be included for comparison when testing other field isolates. Second, this strain formed biofilms with higher levels of fluorescence of live bacteria compared with P. salmonis CA5, especially in the mature biofilm state. Finally, the increase in cytotoxicity induced by planktonic and biofilm-derived bacteria of P. salmonis LF-89^T was faster and sustained over time (see in the Results section). Briefly, total RNA was extracted with the TRIzol™ Max Bacterial RNA Isolation Kit (Thermo Fisher Scientific, NY, USA) from 24-h and 48-h-old biofilms of P. salmonis LF-89^T. Biofilms were induced to form on the bottom of sterile glass Petri dishes under standard and static incubation conditions (as described before). After incubation, the resulting biofilms were washed (3X) with 10 mL of chilled sterile NaCl solution (2.5% w/v) to discard non-adherent cells and then scraped off with a cell scraper in 1 mL of TRIzol™ reagent. Three independent experiments were conducted, and, in each, four Petri dishes with biofilms were scraped using the same milliliter of TRIzol™ reagent. This 1 mL-suspension was later used for RNA extraction according to the manufacturer’s specifications. Complete cell detachment was confirmed by microscopic observation of TRIzol-scraped Petri dishes that were randomly chosen for CV (1% w/v) staining ( Supplementary Figure S1E ) and optical microscopy ( Supplementary Figure S1F ). For the sake of comparison, 1 mL aliquots of culture supernatant (containing planktonic bacteria) were collected from randomly selected glass Petri dishes in each independent experiment. Planktonic bacteria were harvested by centrifugation at 10,000 rpm and 4°C for 2.5 min, and each bacterial pellet was resuspended in 1 mL of TRIzol™ reagent for RNA extraction per the manufacturer’s instructions.

The RNA integrity number for each RNA extract was determined on an Agilent 2100 Bioanalyzer instrument using the RNA 6000 Nano kit in accordance with the manufacturer’s instructions (Agilent Technologies, CA, USA). Four total RNA extracts were obtained in triplicate with an RNA integrity number ≥ 7.0; the RNA extracts were quantified on a QuantiFluor RNA System using the instructions provided by the manufacturer (Promega, WI, USA). The RNA concentration was equal to 127 ± 23 μg μL^-1 and 116 ± 26 μg μL^-1 for 24-h and 48-h-old biofilm samples, respectively. In turn, measured RNA concentrations for 24-h and 48-h-old planktonic samples were 87 ± 30 μg μL^-1 and 81 ± 36 μg μL^-1, respectively. All RNA extracts were stored at -20°C until further use in both RNA sequencing and validation via RT-qPCR of the resulting transcriptomes.

For the sake of validating the transcriptomic data, 10 DEGs (differentially expressed genes as determined by RNA sequencing) were randomly chosen for design of qPCR primers ( Supplementary Table S1 ) at Genoma Mayor Spa (http://www.genomamayor.com/) by using the Primer3 software (https://primer3.org/). To confirm specificity of the primers, a resulting amplicon from each primer pair was randomly selected for cloning with the pGEM-T Easy Vector System (Promega, Madison, CA, USA) and sequencing at Macrogen Inc. (https://dna.macrogen.com/). Two-step reverse transcription-qPCR (RT-qPCR) was performed from complementary DNA (cDNA) synthetized using the ImProm-II Reverse Transcription System (Promega), 20 ng of DNase-treated RNA (TURBO DNA-free kit, Applied Biosystems, Austin, TX, USA), and random primers from the ImProm-II System. All RT-qPCRs were performed on a Stratagene Mx3000P PCR device (Agilent Technologies, CA, USA) in a total volume of 20 μL with 1 μL of cDNA, 2X Brilliant II SYBR Green QPCR Master Mix (1X, final concentration; Agilent Technologies), forward and reverse primers (1 μM, final concentration), and ROX as a passive reference dye (18 nM, final concentration; Agilent Technologies). The description of target genes, amplicon lengths, primer sequences, and annealing temperatures are listed in Supplementary Table S1 . The qPCR program consisted of initial denaturation for 3 min at 95°C, followed by 40 amplification cycles that consisted of denaturation at 95°C for 30 s, primer annealing for 45 s at different melting temperatures ( Supplementary Table S1 ), and 45 s of extension at 72°C. Amplicons were within the expected size, as assessed by analyzing melting curves and standard electrophoresis in agarose gels.

Absolute qPCR data were processed with the MXPro software version 4.10 (Agilent Technologies). For quantification of the gene transcripts, the sequenced clones were used to construct 7-point standard curves (in triplicate) by 10-fold dilution series starting from 4 × 10⁷ copies. To do so, the concentration of each clone was determined with the QuantiFluor dsDNA System per the manufacturer’s instructions (Promega, WI, USA). Afterward, the copy number of clones was calculated by dividing the DNA concentration (in ng μL^-1) of a given clone by its mass (in ng) estimated with the following formula: mass = [(n) × (M/NA)] × 10⁹, where n is the clone size in bp (vector plus insert); M, is the mean molecular weight of a nitrogenous bp (660 g mol^-1); NA, is the Avogadro’s number (6.0221 × 10²³ molecules mol^-1); and 10⁹ is the factor that converts grams to nanogram. The efficiencies (E%) and correlation coefficients (r²) of the standard curves used for quantification of DEGs were 102.17 ± 7.14% and 0.9930 ± 0.0077, respectively.

CHSE-214 embryo cells (ATCC 1681) from Chinook salmon were grown at 18°C on 24-well microplates (flat-bottomed; SPL Life Sciences Co., Ltd.) at 1 × 10⁵ cells per well in the Leibovitz’s L-15 medium (HyClone Laboratories Inc., Logan, Utah, USA) supplemented with 10% fetal bovine serum (Gibco, Invitrogen Laboratories), 6 mM L-glutamine (Gibco), 15 mM HEPES pH 7.3 (Gibco), and 100 μg mL^-1/100 IU mL^-1 streptomycin/penicillin (Gibco). CHSE-214 cells were washed in phosphate-buffered saline (1X PBS, pH 7.0) and then refreshed with fresh antibiotic-free medium (2 mL per well) with 2% fetal bovine serum (Gibco) prior to bacterial infections. Preliminary experiments indicated that the supplemented Leibovitz’s L-15 medium alone was unable to support the growth of P. salmonis.

The CHSE-214 cell line was infected with biofilm-derived and planktonic bacteria of P. salmonis LF-89^T and CA5. The inocula were prepared by growing both P. salmonis strains in AUSTRAL-SRS medium under routine conditions until achieving a concentration of ~1 × 10⁶ CFU mL^-1 (determined by standard plate counting). Afterward, 20 mL of each culture were added to sterile glass Petri dishes and statically incubated in quadruplicate per strain to induce biofilm formation at 18°C for 48 h. After incubation, bacterial supernatants were withdrawn from the Petri dishes, while biofilms formed in the bottom were washed (3X) with 10 mL of sterile AUSTRAL-SRS medium by gentle manual shaking of the plates for 5 min. Later, washed Petri dishes were scraped with cell scrappers to collect biofilm-derived bacteria from the bottom by resuspending in 5 mL of sterile AUSTRAL-SRS medium to obtain a single 5 mL biofilm inoculum per strain in each independent experiment. The biofilm inocula of P. salmonis LF-89^T and CA5 were adjusted to 3.10 ± 0.42 × 10⁶ CFU mL^-1 (i.e., 0.77 ± 0.04 McFarland) and 6.17 ± 0.76 × 10⁶ CFU mL^-1 (0.78 ± 0.05 McFarland), respectively. Later, bacterial supernatants that were previously removed from the Petri dishes were adjusted to 6.93 ± 0.40 × 10⁶ CFU mL^-1 (i.e., 0.80 ± 0.01 McFarland) and 2.53 ± 0.41 × 10⁶ CFU mL^-1 (i.e., 0.78 ± 0.04 McFarland) to be used as planktonic inocula of P. salmonis LF-89^T and CA5, respectively. These planktonic inocula were included in the cytotoxicity trials for comparison purposes.

CHSE-214 cells were infected with biofilm-derived or planktonic bacteria of P. salmonis LF-89^T and CA5 by adding 100 µL of each inoculum per well. Considering the inocula sizes, each infection was carried out using a multiplicity of infection between 3 and 7. Three infection experiments were independently conducted with each strain (grown under biofilm and planktonic conditions) using triplicate infections at 18°C over a 72 h period. Afterward, aliquots of 100 μL were taken from each well at six different hours post-infection (i.e., at 0, 6, 12, 24, 48, and 72 hpi) to measure the LDH-based cytotoxicity with the Takara Kit (Bio Inc., Otsu, Japan) by spectrophotometric reading at 500 nm (Tecan microplate reader, Infinite 200 PRO, Männedorf, Switzerland). The absorbance measurements were corrected with the background absorbance measured in low controls (non-infected CHSE-214 cells) by using the equation supplied by the manufacturer. Resulting measurements were expressed as a percentage of the LDH-based cytotoxicity measured in high controls that consisted of CHSE-214 cells incubated without bacteria with 1% Triton X-100.

Next-generation sequencing libraries and bioinformatic analyses were performed at Genoma Mayor Spa starting from total RNA extracts from biofilm-derived and planktonic samples of P. salmonis strain LF-89^T. Total RNA extracts were subjected to digestion with DNase I to avoid contamination with genomic DNA. Later, the concentration and integrity of the RNA extracts were determined using the Quant-iT™ RiboGreen RNA Assay Kit (Thermo Fisher Scientific) and a Bioanalyzer 2100 (Agilent Technologies, CA, USA), respectively. The RNA libraries were prepared with the TruSeq™ RNA sample preparation kit according to the manufacturer’s protocol (Illumina Inc., CA, USA). Briefly, 28S, 18S, 5.8S, and 5S rRNAs were depleted starting from 500 ng of total RNA from each sample using the Ribo-Zero rRNA Removal Kit (Illumina Inc.); the depletion was confirmed by Bioanalyzer profiling. The rRNA-depleted RNA extracts were fragmented with divalent cations at an elevated temperature. The first-strand cDNA synthesis produced (by reverse transcription) single-stranded DNA copies from the fragmented RNA. After second-strand cDNA synthesis, the double-stranded DNA was end repaired, and the 3’ ends were adenylated. Universal adapters were then ligated to the cDNA fragments, and the final sequencing library was generated by PCR. After validation of the library by using the DNA 1000 chip and a Bioanalyzer (Agilent Technologies), the samples were pooled together in equal concentrations and run on an Illumina HiSeq System for 150 cycles of paired-end sequencing. The RNA sequencing data reported herein were deposited in the Sequence Read Archive (SRA) database under the BioProject number PRJNA854674 for biofilm-derived (SAMN29445255 to SAMN29445260) and planktonic (SAMN29445261 to SAMN29445266) BioSamples.

Technical sequences removal (adapters) and filter quality of the next-generation sequencing reads were performed using Trim Galore (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/), discarding sequences with Phred quality scores (on average) < 18. The pre- and post-filtering quality assessment of reads was performed with the FastQC online tool (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). The high-quality reads were aligned to the reference genome P. salmonis LF-89^T (ATCC VR-1361; NZ_CP011849) using the bowtie2 aligner (version 2.3.4.3) (Langmead and Salzberg, 2012), which creates Sequence Alignment/Map output files. These files were then changed to Binary Alignment Map files using the “sort” tool from SAMtools (Li et al., 2009). The number of mapped reads to a single gene for each “.bam” file was computed using the HTSeq-count tool (Anders et al., 2015). By using the “bedtools getfasta” function in BEDTools (Quinlan, 2014), the genes of the reference genome were extracted to generate a transcript FASTA file with the coordinates of all genes in the genome. Starting from these genes, proteins were predicted with the Prokka tool (Seemann, 2014), and functional annotation at the GO level was performed using the EggNOG-Mapper method (Huerta-Cepas et al., 2017). Starting from the found GO terms, functional profiles of gene expression in biofilm-derived and planktonic samples were obtained with the Web Gene Ontology Annotation Plot (WEGO) tool (Ye et al., 2018), which assigns gene transcripts into the three broader GO functional categories (i.e., Cellular Component, Molecular Function, and Biological Process), as well as into subsets.

DEGs were identified from RNA sequencing data using the DESeq2 package (Love et al., 2014) for the following four comparisons between two conditions: (1) 24-h versus 48-h-old biofilm conditions (BC 24h vs. 48h), (2) 24-h versus 48-h-old planktonic conditions (PC 24h vs. 48h), (3) biofilm versus planktonic conditions at 24 h (BC vs. PC 24h), and (4) biofilm versus planktonic conditions at 48 h (BC vs. PC 48h). Fold Changes (FC) of gene expression were determined using threshold values of 4-fold (i.e., log₂ FC ≥ 2) and 0.25-fold (i.e., log₂ FC ≤ -2) for up- and down-regulated genes, respectively. Ten DEGs identified from RNA sequencing data ( Supplementary Table S1 ) were randomly selected to validate the respective dataset by absolute DEG quantification using RT-qPCR; absolute DEG counts were normalized by the concentration of the total RNA extracts. Normalized DEG counts were log₂-transformed for computing FC ratios that were then compared with log₂ FC values directly computed from RNA sequencing data ( Supplementary Figure S2 ).

Beta regression analysis (Cribari-Neto and Zeileis, 2010) was performed to model beta-distributed dependent variables (e.g., rates and proportions) to describe and evaluate how the percentage of cytotoxicity induced on CHSE-214 cells varied through five-time intervals (0, 6, 12, 24, 48, and 72 hpi) considering two modes of growth (i.e., biofilm and planktonic states) and two infecting P. salmonis strains (LF-89^T and CA5). Similarly, beta regression analysis was implemented to model the fluorescence ratio of live-to-dead bacteria in biofilms through time (24-360 h) considering the factors strain (i.e., LF-89^T and CA5) and mode of growth (i.e., planktonic vs biofilm states). As for the percentage of cytotoxicity and fluorescence ratio, a completely orthogonal model was built accounting for the effect of each factor on each response variable. All beta regression analyses were run using the “betareg” package (Cribari-Neto and Zeileis, 2010), while the “emmeans” package (Lenth et al., 2018) was used to implement the posteriori Tukey’s test to identify significant differences between treatment combinations. Additionally, a generalized linear model was implemented by assuming a Gaussian distribution with an identity function to test how the fluorescence signal (in RFU) of live bacteria in biofilms and specific biofilm formation (SBF) index varied through time intervals (i.e., between 24-360 h [RFU] and 24-312 h [SBF index]) considering P. salmonis strains LF-89^T and CA5. Here, a Gaussian distribution was assumed by the Akaike information criterion using the ‘propagate’ package (Spiess, 2013). For all analyses, an orthogonal model was implemented considering two factors (i.e., P. salmonis strain and time). Later, significant differences between treatment combinations were evaluated by the Tukey’s HSD post hoc test by using the “emmeans” package. All analyses considered a critical value of 0.05. Finally, a lognormal distribution of four parameters was fitted across time-intervals (between 24-360 h) to predict the temporal trajectory of the fluorescence signal of live sessile bacteria of P. salmonis LF-89^T and CA5 (see equation in Supplementary Figure S3 ). All analyses were performed using R version 4.2.0 (R Core Team, 2013).

The interaction between time and strain had a highly significant effect on the variability of the specific biofilm formation (SBF) index, yet the second factor alone did not have a significant effect ( Table 1 ). Crystal violet staining revealed variations associated with changes in the bulk fraction of the biofilm structure ( Figure 1A ). These changes were coherent with those in the SBF index ( Figure 1 ). While no statistically significant interstrain differences in the SBF index were detected at any time, excepting higher indexes for strain CA5 in the first 48 h ( Figure 1 ), the two strains showed opposite general trends in variability of these indexes. In addition, there were some significant intra and interstrain differences in the SBF index between different incubation times ( Supplementary Table S2 ). For instance, differences were detected between the first 24 h and subsequent sampling times (i.e., 72, 120, and 216-312 h) for P. salmonis CA5, as well as between P. salmonis CA5 at 24 h and P. salmonis LF-89^T at subsequent times (i.e., 48-144, 240, and 312 h) ( Figure 1 and Supplementary Table S2 ).

The surface coverage of live and dead bacteria in biofilms tended to increase over time, but with a higher contribution of dead (or inactive) bacteria towards the end of the incubation period ( Figure 2 ). The strain LF-89^T formed biofilms as very thin layers at 48 h ( Figure 2 ), with an exopolysaccharide matrix revealed by wheat-germ-agglutinin lectin targeting N-acetyl sugar matrix components ( Figure 2 ). Similarly, 48-h-old biofilms of P. salmonis CA5 were structured frequently as thin layers ( Figure 2 ) supported by an exopolysaccharide matrix ( Figure 2 ). In addition, P. salmonis LF-89^T biofilms and planktonic counterparts at 48 h shared an API ZYM profile composed of four enzymatic activities (i.e., esterase, esterase lipase, leucine arylamidase, and acid phosphatase) ( Supplementary Figures S4A, B ). Four enzymatic activities (i.e., alkaline phosphatase, leucine arylamidase, valine arylamidase, and acid phosphatase) also made up the API ZYM profile of P. salmonis CA5 biofilms and planktonic counterparts at 48 h ( Supplementary Figures S4C, D ), but in addition, 48-h-old planktonic bacteria of P. salmonis CA5 exhibited beta-glucosidase activity ( Supplementary Figure S4D ).

The interaction between strain and time, as well as each factor separately, exerted a highly significant effect (P < 0.05) on the variables “fluorescence signal of live bacteria in biofilms” and “fluorescence ratio of live-to-dead bacteria in biofilms” ( Table 1 ). The two P. salmonis strains emitted maximum levels of fluorescence from live bacteria during the first 48 h of biofilm formation; levels that, in general, were statistically significantly higher than subsequent times ( Figure 3A and Supplementary Table S3 ). The two strains showed similar kinetic patterns of biofilm formation, and a four-parameter lognormal distribution provided the best fit to the fluorescence datasets ( Figures 3 ). Maximum fluorescence peaks of live bacteria in biofilms formed by P. salmonis LF-89^T were significantly higher than those from live sessile P. salmonis CA5 during the first 96 h of biofilm development ( Figure 3A ). In addition, there was a general trend towards decreasing fluorescence ratios for live-to-dead bacteria in biofilms (i.e., viability ratios) towards the end of the incubation period ( Figure 4 ). Significant differences in the viability ratio of sessile bacteria were detected between biofilms formed by P. salmonis CA5 and LF-89^T during the first 72 h, with higher viability ratios for biofilms formed by strain CA5 ( Figure 4 ). Orthogonal comparison analysis further revealed additional significant intra and interstrain differences ( Supplementary Table S4 ).

A three-factor interaction between strain, time, and growth condition (i.e., biofilm versus planktonic states) significantly affected the variability of LDH-based cytotoxicity ( Table 2 ). Remarkably, the factor “growth condition” alone was not a meaningful factor to explain cytotoxicity patterns ( Table 2 ). There were no statistically significant differences in cytotoxicity due to P. salmonis LF-89^T or CA5 on Chinook salmon (Oncorhynchus tshawytscha) embryo (CHSE-214) cells when intrastrain comparisons between biofilm-derived and planktonic bacteria were performed at any given hpi ( Figure 5 ). For biofilm-derived bacteria of P. salmonis LF-89^T (hereafter defined as biofilm-detached bacteria as single-cells and/or cell aggregates), no statistically significant differences in cytotoxicity between 0 hpi and later infection times were detected; however, the cytotoxic effect of planktonic bacteria of P. salmonis LF-89^T triggered a significant increment of the LDH-based measurements between 0 hpi and following infection times ( Figure 5 ). Biofilm-derived and planktonic bacteria of P. salmonis CA5 induced an initial increase in cytotoxicity in CHSE-214 cells during the first 6 hpi. This was followed by a decrease in cytotoxicity between 12 and 24 hpi and, then, an important increase from 48 hpi onwards ( Figure 5 ). Supplemental material is provided about significant intra and interstrain differences in cytotoxicity level induced on CHSE-214 cells by biofilm-derived bacteria of P. salmonis LF-89^T and CA5 ( Supplementary Table S5 ).

A total of 2,310 genes were expressed in the biofilm and/or planktonic modes of growth at 24 h and/or 48 h of incubation; indeed, the elapsed time to achieve mature biofilms in P. salmonis strain LF-89^T was 48 h. A small fraction of all expressed genes was exclusively transcribed in 24-h and 48-h-old biofilms (i.e., n = 35 and 78 genes, respectively). This fraction included eight genes encoding transposases (three of which were IS6 family transposases), in addition to three proteins (i.e., a LysR family transcriptional regulator; the type VI secretion system contractile sheath large subunit and a HigB family toxin). Of the transcribed genes (n = 2,310), only 157 were significantly differentially expressed genes (DEGs) (at Padj-values < 0.05) after four pairs of different conditions were compared, with different degrees of overlap in DEGs between comparisons ( Supplementary Figure S5 ). Of the significant DEGs, 27 encoded virulence-related proteins, namely “stress response” (n = 12), “iron uptake” (n = 3), “endotoxins” (n = 2), and “other virulence-related genes” (n = 10) ( Supplementary Tables S6-S9 ).

The total number of significant DEGs between 24-h and 48-h-old biofilms ( Supplementary Table S6 ), or between biofilm and planktonic states at 48 h ( Supplementary Table S9 ), was almost twice the total number of DEGs detected for the other two comparisons of conditions ( Supplementary Tables S7, S8 ). DEGs that were upregulated between two specific conditions were found to be normally downregulated under another comparison of conditions. For example, approximately 67% of all upregulated DEGs in 24-h-old biofilms versus 48-h-old biofilms ( Supplementary Table S6 ) were downregulated in 48-h-old biofilms versus planktonic counterparts ( Supplementary Table S9 ). Mainly ribosomal proteins were involved, but DEGs encoding other proteins were detected.

Distribution analysis of the functional categories indicated that significant DEGs between 24-h and 48-h-old biofilms had a functional profile similar to profiling based on DEGs between biofilm and planktonic states at 48 h. Following in similarity was the functional profile obtained from DEGs between biofilm and planktonic states at 24 h, and finally by the functional profile from DEGs between planktonic bacteria at 24 and 48 h ( Figure 6 ).

Significant DEGs in P. salmonis LF-89^T were grouped at the “cellular process” gene ontology (GO) term ( Figure 6 ). In general, the results showed that some ribosomal proteins downregulated in 48-h-old biofilms (compared to planktonic bacteria at 48 h) were upregulated regarding the comparison 24-h versus 48-h-old biofilms ( Figure 6 ). The same was true for the phoH gene, while the sspA gene (encoding the stringent starvation protein A) was downregulated in 48-h-old biofilms (compared to planktonic bacteria at the same timepoint) ( Figure 6 ). Differential gene-expression analyses between 24-h and 48-h-old biofilms (BC 24 vs. 48 h in Figure 6 ) and/or between biofilms and planktonic bacteria at 24 h (BC vs. PC 24 h in Figure 6 ) indicated the upregulation of genes involved in protease activity, regulation of ATP synthesis, and efflux transporter activity, among others ( Figure 6 ). Genes involved in cell adhesion (cadherin-like proteins), environmental sensing (response regulators), and RNA modifications or translation (traM gene) were upregulated in 48-h-old biofilms (compared to planktonic bacteria) ( Figure 6 ). Additionally, a gene encoding an undecaprenyl-diphosphate phosphatase (involved in polysaccharide biosynthesis) was downregulated when 24-h and 48-h-old biofilms were compared; in other words, the respective gene was upregulated in mature biofilms ( Figure 6 ).