The morphology of bacterial colonies was assayed on 0.5xN, 1x and 2xN Shieh plates, as these nutrient levels have previously been shown to be useful for exemplifying gliding motility in F. columnare (Laanto et al., 2012). The bacteria originating from the same liquid culture were spread on agar plates and grown for 2 days after which the colony morphology was imaged under a light microscope (Figure 1). Rz colonies grown on a 0.5xN Shieh plate were spreading with increased mean colony size (3.95 mm, SE ±0.42) and the production of root-like protrusions typical for Rz morphology (Figures 1, 2A). Rz colonies grown on 2xN Shieh plates had a smaller mean colony size (0.73 mm, SE ±0.04), and root-like structures, if seen, were only moderate (Figures 1, 2A). Type S responded to changing nutrient availability comparably to Rz (mean colony size 2.4 mm, SE ±0.23 in 0.5xN and 0.75 mm, SE ±0.06 in 2xN). However, when grown at lower nutrient conditions (0.5xN Shieh), root-like structures were also observed in the S type. These colonies were, nevertheless, distinguishable from Rz colonies by their non-adherent, opaque, and moist appearance (Figure 1). R type did not remarkably alter the colony size in response to changing nutrient availability (mean colony size 0.7 mm, SE ±0.03 in 0.5xN and 0.5 mm, SE ±0.00 in 2xN) and root-like structures were only occasionally seen under low nutrient conditions (Figures 1, 2A).

The viability of Rz, R, and S colony types in 0.5xN and 2xN Shieh media was measured as the maximum optical density (OD_max) reached during a 65-h cultivation period (Figure 2B). All colony types reached the highest OD_max at higher nutrient level (2xN Shieh). The biofilm forming ability was remarkably higher under low nutrient conditions (0.5xN Shieh) in the Rz type compared to 2xN Shieh as well as to the R and S types, which were weaker biofilm producers at both nutrient levels (Figure 2C).

The movements of individual Rz cells grown on 0.5xN and 2xN Shieh agar plates were recorded with a confocal microscope. The movements of Rz cells were comparable with the previously described gliding motility of F. johnsoniae; the cells glided over the surface in a straightforward manner, occasionally attaching to the surface with one end of the cell, rotating, and changing the moving direction (Supplementary Videos 1, 2). The gliding speed was slower than that seen in F. johnsoniae, which was used as a reference (data not shown).

The movements of individual bacteria growing as a part of forming biofilm were recorded between the agar layer and microslide chamber bottom. The colony types Rz and S formed a monolayer on the edges of spreading bacterial biofilm, where the cells were organized side by side and glided along the adjacent cells (Figure 3 and Supplementary Videos 3, 4, 7, 8). The cells formed branching rhizoid-like structures a few cells wide (here referred to as microrhizoids), which involved both motile and non-motile cells. In both Rz and S types, more active gliding motility was seen under low nutrient concentrations (0.5xN). Colony type R expressed only occasional movements regardless of the nutrient level, and cellular organization as a spreading biofilm was not observed (Supplementary Videos 5, 6). In order to visualize F. columnare colony formation on a longer timescale, the growth of the Rz colony type on 1x Shieh was recorded over the course of 8 h (Supplementary Video 9). In the front of the biofilm, the bacteria were characteristically organized in microrhizoids, which moved cooperatively toward the spreading direction of the biofilm and seemed also to serve as routes along which other cells were able to glide further and support biofilm expansion.

Genes putatively involved in gliding motility were sequenced from Rz, R, and S types of strain F. columnare B067. No genetic differences were detected between the colony types in the gene sequences of operons gldFG and gldKLMN, genes gldH, sprA, sprC, sprD, sprE, sprF, sprT, porV, porX, and porY, or in the predicted regulative regions upstream from the genes gldH, sprA, sprE, and porV or operons gldFG, gldKLMN, and sprCDBF. Expression of genes gldG, gldH, gldL, sprA, sprB, sprE, sprF, sprT, and porV, which are putatively involved in F. columnare gliding motility, was measured in B067 Rz, R, and S colony types that had been grown on 0.5x and 2x Shieh agar plates. Of these genes, gldL, sprA, sprE, sprT, and porV are associated with the T9SS. Gene expression results were normalized with reference genes gapdh and glyA, which are stably expressed in the current dataset (the M value was 0.5834 for both genes, with a variance coefficient of 0.2114 for gapdh and 0.2007 for glyA). Relative expressions are presented in Figure 4 (for statistics, see Tables 1, 2). Significant differences between colony types were observed in the expression of genes gldG, gldH, gldL, and sprE (Table 1 and Figure 4; for pairwise comparisons, see Table 2). Nutrient levels had a significant effect on gldL, sprA, sprB, and sprF expression. The pairwise comparisons revealed that sprA was expressed at significantly higher levels in low-nutrient conditions in Rz and R types, and the same pattern was detected in sprF expression in the R type (Table 2 and Figure 4). However, significant interaction between the colony type and nutrient was detected in gldH, gldL, sprA, and sprE (Table 1), indicating that gene expression of colony types may differ between nutrient conditions. Indeed, direct associations between spreading behavior and gliding motility gene expression were challenging to form as different colony types seemed to respond differently to the nutrient level with motility gene expression. Even though a significant effect of colony type was not observed in either sprT or porV, a significant difference between Rz and R was observed in sprT and between Rz and S in porV expression (Table 2).

Colony types Rz, R, and S were cultivated on 0.5xN and 2xN Shieh plates containing 1.5% skim milk. Proteolytic activity was observed in each colony variant, seen as the formation of a clear degradation zone peripheral to the bacterial growth, but no differences between the colony types were observed, and the nutrient concentration did not affect the proteolytic activity (Supplementary Figure S1). The effect of colony type and nutrient availability on the contents of extracellularly secreted products (ECPs) was analyzed further. Generally, Rz, R, and S grown in 0.5xN and 2xN liquid Shieh cultures shared a common overall ECP profile, with some moderate changes in individual protein bands between colony types (Supplementary Figure S2). However, a strong protein band, approximately 13 kD in molecular weight, was detected in Rz type grown in both 0.5xN and 2xN Shieh media. The corresponding band was absent or barely detectable in R or S types.

In all experiments in this study, we used three different colony types (rhizoid [Rz], rough [R] and soft [S]) of F. columnare strain B067 (Laanto et al., 2011). The bacterial strain was originally isolated from a trout that was killed during a columnaris disease outbreak at Finnish fish farm (Laanto et al., 2011). The R colony type was obtained after exposing the original Rz isolate to bacteriophages (Laanto et al., 2012). The S type appeared spontaneously during laboratory culture of the Rz type (Laanto et al., 2014).

Bacterial stocks were stored at -80°C in 10% fetal calf serum and 10% glycerol and revived from the freezer in fresh modified Shieh medium according to Song et al. (1988), which is referred to as Shieh medium in this study and is used as a base of nutrient-modified media. In nutrient-modified media (0.5x or 2x Shieh), all the ingredients were either halved or doubled. In 0.5xN and 2xN Shieh media, only the concentration of peptone and yeast extract were halved or doubled, respectively (for detailed compositions of the media, see Penttinen et al., 2016). After revival from the freezer, liquid bacterial cultures were cultivated at RT/26°C, with agitation of 115/150 rpm for 24–48 h to obtain dense cultures. Cultures were refreshed with fresh 1x Shieh medium, and cultivation was continued for 16–24 h. For plate cultures, dense liquid culture was streaked on a Shieh agar plate, which was incubated at RT for 2 days.

Liquid bacterial cultures of Rz, R, and S grown in 1x Shieh were streaked on 0.5xN, 1x, and 2xN Shieh plates. Plates were incubated for 2 days at RT, after which the colony morphology was determined.

In order to evaluate the bacterial viability in different nutrient concentrations, 1.18 × 10⁷–1.28 × 10⁷ colony forming units of B067 Rz, R, and S in a total volume of 400 μl containing 0.5xN, 1x, or 2xN Shieh (N = 8) was cultivated on Honeycomb 2^® microplate (Growth Curves Ltd.) in a Bioscreen C^TM spectrophotometer (Growth Curves Ltd.) at 26°C. The absorbance (600 nm) was measured every 5 min for 65 h. The viability was estimated as the maximum absorbance recorded during the cultivation.

The biofilm formation capacity under various nutrient conditions was determined by cultivating 1.7 × 10⁶-1.83 × 10⁶ colony forming units of B067 Rz, R, and S in a total volume of 100 μl in 0.5xN and 2xN Shieh media on a Maxisorp plate. After a 44-h -incubation at RT, the emptied wells were rinsed twice with 200 μl of phosphate buffer solution (PBS). The biofilm-forming bacteria were stained with 125 μl of 0.1% crystal violet solution for 10 min and rinsed three times with 200 μl of PBS and the plate was dried at RT overnight. To solubilize the crystal violet, 125 μl of 96% ethanol was added. Finally, 100 μl of the solution was transferred to a fresh microplate, and absorbance was determined at the wavelength of 595 nm with a Multiskan FC spectrophotometer (Thermo Scientific).

Flavobacterium columnare B067 Rz cells were scratched from 0.5xN and 2xN Shieh agar plates and suspended in 0.5xN and 2xN Shieh liquid media, respectively. The cell suspension was pipetted and imaged on an eight-chambered ibidi^® ibiTreat μ-Slide (ibidi GmbH) covered with CID lid for μ-dishes (ibidi GmbH). The bacterial cells were imaged with a Nikon AR1 laser scanning confocal microscope using a 488 nm Argon laser and CFI Apo VC 60x water immersion objective (numerical aperture 1.2).

In order to image the spreading behavior of the different colony types, 3–5 μl of overnight culture of F. columnare B067 Rz, R, or S was pipetted onto an eight-chambered ibidi^® ibiTreat μ-Slide (ibidi GmbH) between the bottom of the chamber and a 0.5xN, 1x, or 2xN Shieh agar layer. The bacteria were cultivated overnight at RT, after which the motility on the edge of the spreading colony was imaged as described above. In order to make slow bacterial movements detectable to the human eye, the videos were sped up as follows: Supplementary Videos 1–8: 4×; Supplementary Video 9: 1,800×.

Several dilutions (with Shieh media) were made from liquid bacterial cultures, which were then spread on 0.5x and 2x Shieh agar plates in order to obtain plates with separate colonies and on which the colony types were recognizable. Plate cultures were incubated at RT for 2 days, after which each plate was inspected to ensure it contained only the appropriate colony type. By diluting the bacterial cultures, close to round-shaped colonies were observed and their size was measured. A colony was considered to be the area covered with bacterial cells, including both the denser area in the middle of the colony (if present) and the more transparent area around it. Following the manufacturer’s instructions, the bacterial colonies were suspended in RNA Protect^TM Bacteria Reagent (QIAGEN), which protects RNA from degradation. Total RNA was extracted from Rz, R, and S colonies grown on 0.5x and 2x Shieh agar plate cultures with an RNeasy^® Mini Kit (QIAGEN). If there was any remaining genomic DNA, DNAse treatment with DNA-free^TM (Ambion by Life Technologies) was carried out. RNA quality was verified by running the samples on an Agilent RNA 6000 Nano Chip (Agilent Technologies) in an Agilent 2100 Bioanalyzer (Agilent Technologies) and determining the RNA integrity number (RIN) for each sample. Only qualified samples (RIN above 8.7) proceeded to cDNA synthesis which was performed immediately after RNA validation. RNA was reverse-transcribed into cDNA in triplicate reactions with iScript^TM cDNA Synthesis Kit (Bio-Rad) according to the manufacturer’s instructions. cDNA reactions with a volume of 20 μl contained 1x iScript^TM reaction mix, 1 μl iScript^TM reverse transcriptase, and 40 ng of template RNA. Replicate reactions were pooled and used as a template in qPCR.

Each 20 μl qPCR reaction, run in triplicates, contained 40 ng (gapdh, glyA, gldG, gldH, gldL and, sprB) or 80 ng (sprA, sprE, sprF, sprT and porV) of cDNA template, 0.5 μM of both forward and reverse primers and 1X iQ^TM SYBR Green Supermix (Bio-Rad) that contained iTaq DNA polymerase (25 U/ml). qPCR reactions conditions were as follows: 95°C for 3 min, followed by 40 cycles of 95°C for 10 s, Tm °C for 20 s and 72°C for 20 s, [melting temperature (Tm) was chosen according to the primer pair (Table 3)]. CFX96^TM Real-Time System C1000^TM and C1000^TM Touch Thermal Cyclers (Bio-Rad) were used in qPCR plate runs on 96-well Hard-Shell^® PCR plates (Bio-Rad). On each plate, two interplate calibrator samples in triplicates were run to normalize interplate variation.

Primer-BLAST¹ was used to design a primer pair for each target gene. Primers used in this study are presented in Table 3. The specific binding of each primer pair was tested by checking the amplicon length on agarose gel and with melt curve produced by CFX Manager^TM Software v3.0. glyA and gapdh have been qualified as valid and stably expressed in each F. columnare colony type in various nutrient conditions (Penttinen et al., 2016). They were, thus, used as reference genes to normalize the gene expressions of gliding motility-associated genes. M-value (Vandesompele et al., 2002), which indicates gene expression stability, was measured for the current dataset for reference genes glyA and gapdh using CFX Manager version 3.0 (Bio-Rad).

For the following data prehandling, GenEx version 6.0 (MultiD Analyses) was utilized. Any missing quantification cycle (Cq) value was replaced with the average Cq of its two qPCR replicates. IPC samples run on each plate were used to minimize variation between different plate runs. Efficiency for each primer pair was calculated from a standard curve (with CFX Manager version 3.0) and Cq values were corrected with the efficiency within each gene. The averaged Cq values were normalized with reference genes and transformed into relative gene expression with GenEx version 6.0 (MultiD Analyses).

The effect of colony type and nutrient level on gliding motility gene expression was tested with ANOVA. Post hoc tests were Bonferroni-corrected. Data for sprA, sprE, and gldG were log-transformed to fulfill the assumption of normality and homoscedasticity. Statistical analyses were performed with IBM^® SPSS^® Statistics 22 (IBM Corporation), except for gldL, which could not be transformed to fulfill the assumption of ANOVA and was, thus, analyzed by ARTtool package in R (version 3.1.3) (Wobbrock et al., 2011). Pairwise comparisons were not performed for gldL.

Genomic DNA of F. columnare strain B067 colony types Rz, R and S was extracted from bacterial liquid cultures grown overnight in 1x Shieh medium at RT (115 rpm) using a GeneJET Genomic DNA Purification Kit (Thermo Scientific). The genes related to gliding motility, T9SS or their regulation (gldH, sprA, sprE, sprT, and porV), genomic regions spanning sprCDBF, gldFG and gldKLMN as well as partners of a putative two-component system, porX and porY, were first amplified with PCR using genomic DNA of B067 type Rz, R or S as a template. The 20 μl reactions were performed using Phusion Flash High-Fidelity PCR Master Mix (Thermo Fisher Scientific), with primer concentrations of 0.5 μM and a template amount of 1–10 ng per reaction. The PCR protocol described by the manufacturer was followed, taking into account the differences in primer melting temperatures and PCR product sizes.

The organization of the genes studied in RT-qPCR assay within operons was predicted with DOOR2 (Dam et al., 2007; Mao et al., 2009) according to the genome sequence of F. columnare ATCC 49512. The genomic region upstream of the predicted operon was assumed to contain the appropriate promoter region. The upstream regions of operons gldFG, gldKLMN, and sprCDBF, those operons comprising genes gldH, sprA, or porV as well as the upstream region of the gene sprE (which was predicted to be expressed alone) were sequenced in B067 Rz, R, and S types.

Prior to sequencing the PCR products were purified using QIAGEN’s QIAquick PCR Purification Kit. Primers for the sequencing reactions were designed in 500 bp intervals using VectorNTI version 11.5.1 (Invitrogen), utilizing our shotgun sequencing results as a template. A BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems) was used for sequencing the DNA fragments using the Sanger sequencing technique with an automated sequencing instrument 3130xl Genetic Analyzer (Applied Biosystems). The identity of each base was determined with at least two good quality reads. Basecalling was done using Sequence Analysis 6 (Applied Biosystems). Gene sequence assembly and the alignment of homologous sequences of different colony types were performed with Geneious 8.1.5 (Biomatters Ltd.). The assembled gene and regulative region sequences of Rz, R, and S colony types are found in GenBank (accession number in brackets): gldFG (MF278296), the upstream region of operon comprising gldH (MF278297), gldH (MF278298), gldKLMN (MF278299), the upstream region of operon comprising sprA (MF278305), sprA (MF278306), sprCD (MF278307), sprE (MF278308), sprF (MF278309), sprT (MF278300), the upstream region of operon comprising porV (MF278301), porV (MF278302), porX (MF278303), and porY (MF278304).

To study the effect of nutrient level on proteolytic activity, B067 colony types Rz, R, and S were cultivated in 0.5xN and 2xN Shieh media and 10 μl of bacterial culture (containing 1.4 × 10⁶ CFUs ±4 × 10⁴ SE on average) were spotted, respectively, on 0.5xN or 2xN Shieh agar plates containing 1.5% skim milk (Merck). Plates were incubated for 2 days at RT, after which the clear zone (indicating proteinase production) around the bacterial growth was detected.

Extracellularly secreted product samples were prepared as follows: 8 ml from F. columnare Rz, R, and S liquid cultures were added to 100 ml of fresh 0.5xN and 2xN Shieh media. The cultures were grown for 19 h. One hundred milliliters of dense bacterial culture was centrifuged at 4°C (4,500 rpm, 15 min). The supernatant was first filtered through a 0.45 μm Supor^® membrane (Pall Corporation) and then concentrated with 10 K Amicon Ultra-15 Centrifugal Filter Units (Merck Millipore) at 4°C to final volume of 2–3 ml. ECP samples were divided into 500 μl aliquots and stored at -20°C. Protein concentration of the ECP samples was determined using the Bradford method (Bradford, 1976) against a standard curve made with known amounts of bovine serum albumin. Fifty micrograms of each ECP sample (except 150 μg of Rz grown in 2xN Shieh) was loaded onto 14% Tricine-SDS-PAGE gel. The gel was run for 24 h at 90 V/30 mA and stained with Coomassie Brilliant Blue solution.