Tissues collected from suspected furunculosis cases were obtained from diagnostic routine, ADL laboratory, Puerto Montt, Chile. Samples, kept at -20°C in preserving solution, were thawed to run a commercial RNA purification protocol (E.Z.N.A. Total RNA kit, Omega Bio-tek, GA, USA). Bacterial isolates used in this study were from the ADL strain collection, kept in 20% glycerin (v/v) at -80°C, and cultured on trypticase soy agar (TSA; BD, MD, USA) plates at 18°C for 48 h. Some isolates were cultured in trypticase soy broth (TSB, BD). Dilutions prepared with saline were placed on TSA plates with Congo red added at 30 µg/ml before autoclaving the medium (Ishiguro et al., 1985). For subsequent molecular analysis, a loop of each isolate derived from a single colony was processed for DNA purification employing a GeneJET genomic DNA purification kit (Thermo Fisher Scientific, OR, USA). Epidemiological data of bacterial isolates are shown in Table 1 .

Selected isolates, representative of the different results obtained in vapA PCR assays ( Table 1 ), were grown on TSA plates at 18°C for 48 h. Then, a loop was used to prepare bacterial suspensions adjusted to optical density (OD₆₀₀) = 1.00. Subsequently, 96-well microplates were inoculated with the bacterial suspensions diluted in TSB at 1:100. Bacterial growth kinetics of four A. salmonicida isolates were observed in an automatic spectrophotometer EPOCH (Biotek, VT, USA) at incubation temperatures of 10, 23, and 37°C. Reads were taken up to 48 h. To determine the optimal salinity for growth, TSB was supplemented with 0, 0.5, and 2% of NaCl. Aliquots of bacterial suspensions were log10 serially diluted in saline, plated in triplicates onto TSA, incubated at 18°C for 48 h, and counted. For serum killing assays, a saline suspension of freshly cultured bacteria was prepared and adjusted to OD₆₀₀ = 1.00. Samples were exposed to naïve serum collected from juvenile S. salar for 3 h at 18°C. Bacterial counts were performed on control and treatment samples to compare the bactericidal action of the complement.

A. salmonicida was identified in tissue samples and isolates using specific TaqMan qPCR assays targeting well-known virulence factors (Gulla et al., 2016a; Chapela et al., 2018). Samples were not treated with DNAse, hence, a combination of cDNA and gDNA was used as template. The sequences of primers and probes are listed in Supplementary Table S1 . PCR conditions were the following: 300 nM of each primer along with 200 nM of probe for each primer/probe set were mixed with 3 µl of nucleic acid template, enzyme and master mix using AgPath-ID™ One-Step RT-PCR and nuclease-free water according to the manufacturer’s instructions (Applied Biosystems™, TX, USA). To proceed with the amplification, 10 min at 45°C for reverse transcription and 10 min at 95°C for reverse transcriptase inactivation, then 45 cycles of 5 s at 95°C and 30 s at 60°C for annealing-extension were considered.

Whole genome sequencing was conducted at Codebreaker Bioscience facilities (Santiago, Chile). We first created a set of libraries using the standard protocol of Nextera Flex (Illumina™, CA, USA). Each standard Flex library was constructed using all standard kit reagents from the Nextera DNA Flex library prep kit, following the manufacturer’s protocol. The concentration of eluted libraries and the library size were measured using a Qubit high-sensitivity (HS) dsDNA kit (Thermo Fisher Scientific) and the S2 Cartridge from BIOtic with the Qsep1 Machine. Sequencing was performed on an Illumina™ Nextseq1000™ platform using a 600-cycle kit configured for 2×300 bp paired-end reads, incorporating a 5% PhiX mix as a diversity control.

Raw fastq files were processed using a custom-developed Nextflow pipeline (https://github.com/gene2dis/mgap), which included quality control and cleaning of the raw reads with FastP (v.0.23.2) (Chen, 2023), followed by assembly using SPAdes (v.3.15.5) (Bankevich et al., 2012). We employed Bakta (v.1.7.0) for contig annotation (Schwengers et al., 2021) and BLASTN v2.15.0 to look for homologous sequences of contigs of interest in public databases (Altschul et al., 1990). For the visualization of groups of homologous biosynthesis gene clusters, the clinker tool was applied (Gilchrist and Chooi, 2021). Average nucleotide identities (ANI) were calculated using the ANI calculator tool hosted at https://www.ezbiocloud.net/ (Yoon et al., 2017).

A phylogenomic tree was generated by downloading all available A. salmonicida genomes that have been isolated from fish (as of January 2025), and filtering by genome completeness (>95%) and contamination (<5%) using CheckM2 (Chklovski et al., 2023). Filtered genomes were annotated with Bakta (Schwengers et al., 2021), and a pangenome analysis was performed using Roary (Page et al., 2015). The core genome, defined as genes shared by more than 95% of all genomes, was aligned in MAFFTT (Katoh and Standley, 2013), and a phylogenetic tree was constructed with IQtree2 (Minh et al., 2020).

A modified, Sarkosyl-based extraction strategy was applied to obtain the outer membrane protein (OMP) fraction (Maiti et al., 2011). Briefly, a loopful of isolates cultured on TSA was suspended in saline and centrifuged at 10,000 g for 10 min. The pellet was suspended in lysis buffer (300 mM NaCl, 10 mM HEPES, 2 mM PMSF, 8 M urea, pH 8.0), then sonicated on ice for 1 min at 70% potency (300 V/T, Biologics Inc, VA, USA). The crude extract was centrifuged at 10,000 g for 10 min at 4°C, resuspended in 1 ml of 0.2% Sarkosyl and agitated over night at room temperature (RT). Samples were centrifuged at 10,000 g for 1 h at 4°C, followed by pellet washing twice with 10 mM Tris-HCl pH 7.4 buffer and recovering the pellet by centrifugation at 10,000 g for 10 min at 4°C. Finally, the OMP fraction was suspended in buffer containing 10 mM Tris-HCl pH 7.4 plus 8 M urea. Proteins were kept at -20°C before use.

For LPS extraction and purification, we adapted a method described by others (Yi and Hackett, 2000). A loop of bacteria cultivated on TSA was suspended in 300 µl of Trizol reagent (Thermo Fisher Scientific) and incubated for 15 min at RT. Then, 100 µl of chloroform were added mixing thoroughly, followed by another incubation period of 10 min at RT. The water phase was centrifuged at 12,000 g for 10 min, then treated with 500 µl of 0.375 M MgCl₂ in cold ethanol (-20°C). The pellet was finally recovered by centrifugation at 12,000 g for 15 min at 4°C. Membrane antigens were analyzed with discontinuous SDS-PAGE electrophoresis on 5% and 15% polyacrylamide stacking/solving gels. Gels were stained with Coomassie blue for protein analysis and silver-stained for LPS assessment according to a protocol previously described (Hitchcock and Brown, 1983).

For Western blots, a semi-dry protocol was followed (Bio-Rad, CA, USA). Signals in the blots were revealed using a polyclonal antibody prepared from serum of a rabbit immunized with a VapA+ isolate (Aquit, https://aquit.net/) and the corresponding HRP-conjugated secondary antibody against salmon Ig heavy chain (LSBio, https://www.lsbio.com/).

Juvenile Atlantic salmon, Salmo salar (mean weight 28 g) were intraperitoneally challenged with either one of three infectious doses of representative A. salmonicida vapA-absent PM-118565 field isolate (~10⁶, ~10⁷, or ~10⁸ colony forming units [cfu]/fish) in independent tanks. Two tanks containing 500 L of freshwater, filtered and UV treated with a turnover rate of 1 h, were used for the assay. Each tank contained 120 fish in total: 30 healthy control fish (25%) that received an intraperitoneal injection (ip) of 0.1 ml of saline, and 90 fish (75%) which were injected with bacterial inoculum. This group of 90 fish was divided into three groups of 30 fish, as marked by Visible Implant Elastomer (VIE tagging), for one to receive 10⁸ cfu/fish and the other ones to be challenged with 10⁷ and 10⁶ cfu/fish, respectively. Biomass was adjusted to a density of ~15 kg/m³ in one tank. In the second tank, we simulated overcrowding as a stress condition using 2× biomass density by diminishing the water level.

In parallel, we performed a lethal dose 50 (LD50) assay in a third tank containing 100 fish. These fish were inoculated with the vapA+ PM-118328 isolate at doses ranging from 10⁴-10⁷ cfu/fish, with each group comprising 20 animals. Accordingly, we distinguished five groups in this tank: the negative control group and fish that received 10⁴, 10⁵, 10⁶, and 10⁷ cfu, respectively. All experiments were carried out at 10.5 ± 0.1°C and terminated 30 days post-inoculation (dpi). Mortality was monitored and registered daily. Anterior kidney, liver and spleen pooled samples from each dead fish were obtained, placed in tubes with RNAlater (Thermo Fisher Scientific), and stored at -80°C until further analysis. At 30 dpi, all survivor fish, including control fish, were sacrificed to take internal organ samples for PCR analysis. The experimental design was approved by ADL’s Bioethics Committee and performed in TEKBios fish trial center (study code 000-017).

Immune responses to vaccination against A. salmonicida were evaluated in serum field samples. During vaccination activities that took place in farms 5, 7 and 10 in winter 2023 ( Table 1 ), five unvaccinated S. salar pre-smolts were euthanized with an overdose of benzocaine, before being bled to obtain serum samples. Further serum samples were obtained from identical numbers of fish per farm during the freshwater stage at ~300 and ~600 thermal units (TU) time points post vaccination with a pentavalent product, including A. salmonicida¸ P. salmonis, Vibrio anguillarum, Infectious Pancreatic Necrosis virus (IPNV), and Infectious Salmon Anemia virus (ISAV) antigens. All serum samples were kept at -80°C until processing. Briefly, 96-well microplates were coated using 5 µg/well of bacterial proteins purified from representative A. salmonicida vapA+ or vapA-absent isolates (PM-118328 and PM-118565, respectively; see Table 1 ). Enzyme-linked immunosorbent assays (ELISA) were performed on protein-coated microplates containing 5% skim milk as blocking agent, washed with phosphate buffer saline (PBS)-0.05% Tween (PBS-T). Fish serum was added 1:100 in skim milk, followed by incubation for 1 h at RT. Microplates were subsequently washed with PBS-T, then incubated with a monoclonal HRP-conjugated antibody anti-salmon IgM, and washed again. Finally, the signal was revealed with the chromogenic substrate 3,3′,5,5′-tetramethylbenzidine (TMB; Thermo Fisher Scientific). Absorbance was captured at 450 nm with a spectrophotometer.

Plots were constructed using the ggplot package on R (version 4.0.5). Differences in serum killing assays were assessed by t-test. For the in vivo assay, the percentages of cumulative mortalities were analyzed using the Kaplan-Meier method, and the differences were evaluated using the log-rank. For LD50 calculations, we used the BioRssay R package. Statistical tests were also performed with R.

ADL Diagnostic’s routine database was interrogated for A. salmonicida testing and detection between January 2021 and June 2024. As shown in Figure 1A , a dramatic increase in the demand for A. salmonicida PCR diagnostics took place during 2022 in comparison with 2021. Since tissue samples from 2023 and earlier were analyzed in a single-marker qPCR assay, all positive cases were considered as caused by the A. salmonicida vapA+ strain. During 2024, however, after the modification of the diagnostic routine and the introduction of the two-marker strategy, several cases of furunculosis tested positive only for fstA but not for vapA. Data revealed A. salmonicida vapA-absent types to account for 88% of all tissue samples tested this year ( Figure 1B ). Typical clinical signs observed in field cases of vapA-absent furunculosis included skin lesions on the back, abdomen and fins, which were often coinfected with Flavobacterium psychrophilum. Signs of inflammation of internal organs, as usually described in vapA+ furunculosis, were rare in cases caused by vapA-absent types, but vapA-absent bacterial colonies were isolated from those samples ( Figure 1C ; Table 1 ).

ADL’s strain collection encompasses atypical A. salmonicida isolates that have been collected since 2013. Those old isolates stem from seawater cases, but the newer additions originate from freshwater outbreaks ( Table 1 ). Bacterial specimens were isolated from diseased fish belonging to 11 different salmon farming companies located across five administrative regions. While they have their own farms, some companies utilize third-party facilities for rearing fish up to smolt size (e.g., Farm 5). Although some isolates did not amplify in vapA qPCR, they had been stored for further investigation due to their appearance and colony morphology. All isolates were genotyped for the presence of vapA and fstA. As shown in Table 1 , the majority were vapA+/fstA+, with the vapA-/fstA+ genotype accounting for only 17% of the total number of isolates, all of which date to 2022 and later. This share is in disagreement with the prevalence detected in tissue samples, presumably because the number of processed samples and the success rate of isolation are not comparable to the PCR detection.

The A-layer phenotype was investigated using a classic culture method, taking advantage of the retention capacity of vapA+ colonies for Congo red (Ishiguro et al., 1985). As illustrated in Figure 2A , A. salmonicida vapA+ (represented by isolates AY-2485 and PM-118328) can adsorb the dye, forming deep red colonies, while those typed as vapA-/fstA+ were unable to retain the dye and remained gray (isolates PM-118565 and PM-119448). Since the retention of Congo red depends on the expression of the A-layer protein, these observations allow us to infer that vapA-absent colonies are devoid of this molecule. Another important phenotypic feature we assessed was the growth at different temperatures and salinities. Growth curves obtained demonstrate optimal growth for all isolates at mesophilic temperatures (23°C) at different salinity levels, except for vapA+ AY-2485 at high salinity. All A. salmonicida isolates showed the capacity to adapt to psychrophilic temperatures (10°C) and mild salinity, with the exception of AY-2485, which did not replicate. Attempts to define conditions that enable better growth of AY-2485 failed, and this isolate behaved persistently as a slow-growing, fastidious bacterium ( Supplementary Figure S1 ). Growth at 37°C, albeit suboptimal, could be observed for vapA-absent genotype isolates only. In this context, high salinity proved to be a detrimental factor for growth for PM-118565 but not PM-119448 ( Figure 2B ). Regarding these isolates’ susceptibility to naïve fish serum, vapA-absent isolates – which are devoid of the A-layer – resulted more susceptible to being killed than those carrying an intact copy of vapA ( Figure 2C ).

The sequences of all draft genomes were deposited in GenBank under the BioProject PRJNA1168199. The corresponding BioSample accession numbers along with data of draft genomes are listed in Table 2 . Genome-based phylogeny placed sequences of vapA+ PM-118328 and AY-2485 isolates within a cluster of A. salmonicida sequences derived from diseased fish, mostly from salmonids ( Figure 3 ; Supplementary Table S2 ). Interestingly, PM-118328 grouped together with sequences derived from reared salmon in Canada, while AY-2485 forms a single-unit subcluster. In contrast, vapA-absent types clustered adjacent to the sequences derived from non-salmonid fish and environmental samples, suggesting a different lineage compared to Chilean vapA+ types.

A comparative analysis of draft genomes at a large scale is presented in Table 3 . We performed a pair-wise genome comparison using an improved algorithm to calculate ANI. Even though vapA+ isolates were collected with a decade of difference, their sequences exhibited the highest identity value (99.59%). The identity between mixed group-pairs resulted to be over 97%. On the contrary, the comparison of vapA-absent genomes yielded even higher identity scores, reaching 99.96% of identity, despite them not being epidemiologically related. Comparative analysis performed on the vapA locus and its genomic context disclosed a striking result: As expected, we could not find homologous sequences for vapA in PM-118565 and PM-119448 genomes, which is consistent with the PCR results. Notably, further contig and annotation analysis allowed us to unveil a fragment encoding a capsule biosynthesis gene cluster flanked by some O-antigen biosynthesis wb _salmo genes in vapA-absent types ( Figure 4 ). This region is quite different not only in vapA+ types, in which the locus harbors genes of wb _salmo, vapA and its protein-exporting machinery, but also in the vapA-absent SRW-OG1 genome. The genomic organization found in vapA+ isolates resembles that described for well-known A. salmonicida strains (Merino et al., 2015). According to the synteny analysis, it seems that a complex sequence rearrangement comprising rfb genes took place in vapA-absent types, since that cluster appears to be split into rfbBA and rfbDC, which are flanking the capsule biosynthesis cluster instead of presenting as a continuous sequence of elements. Evidence on the presence of the same locus organization was collected by means of BLAST searches, which identified an 14-kb orthologous region in Aeromonas sp. O23A, an environmental strain related to A. salmonicida (Uhrynowski et al., 2017).

Outer membrane protein profiles disclosed specific patterns for both vapA-absent and vapA+ A. salmonicida isolates ( Figure 5A ). Further analysis concerning the protein fraction reactive to rabbit immunized serum revealed some common bands but most notably confirmed the absence of a ~49-kDa band compatible with the VapA protein in vapA-absent isolates ( Figure 5B ). Isolates could also be grouped according to their LPS profiles: vapA-absent types presented a single-band pattern which is structurally consistent with the lipid A-core oligosaccharide. By contrast, vapA+ isolates exhibited two bands. One of them probably corresponds to the lipid A-core (low molecular size), while the other likely represents the lipid A-core plus O-antigen (high molecular size) ( Figure 5C ). LPS Western blots evidenced a high-molecular size single-band pattern in vapA+ isolates ( Figure 5D ). This finding is in accordance with the above description of the LPS profiles.

Remarkable differences in virulence between vapA-absent and vapA+ isolates became evident in the intraperitoneal challenge model of infection. The vapA+ strain killed fish in a few days: Mortality rates of 100% were reached as fast as five dpi in S. salar inoculated with the highest bacterial dose (3.17 × 10⁷ cfu/fish, Figure 6A ). The same level of mortality was recorded at six dpi in fish challenged with the lower dose of 2.67 × 10⁶ cfu, while 90 and 65% mortality were accumulated with 1.50 × 10⁵ and 1.80 × 10⁴ cfu/fish, respectively. In contrast, mortality observed in fish infected with the vapA-absent PM-118328 strain scaled up to 80 and 6.7% after the inoculation of 4.0 × 10⁸ and 3.67 × 10⁷ cfu/fish, respectively ( Figure 6B , left panel). Interestingly, the same bacterial doses induced 90 and 23% mortality in fish that experienced overcrowding ( Figure 6B , right panel). The relative bacterial loads in kidney, liver, and spleen pooled samples were determined by PCR for all fish, mortality and survivors. Results show that both vapA+ and vapA-absent isolates did infect internal organs, producing the death of fish most probably from a septicemic disease ( Figure 6C ). According to calculated LD50 values, virulence ranked as vapA+ > vapA-absent+stress ≈ vapA-absent ( Supplementary Figure S2 ).

Relative IgM serum levels against A. salmonicida were determined by ELISA. As expected, vaccinated fish exhibited an increasing level of antibodies during the development of immunity in fresh water (up to ~600 TU), which reflects the homologous character of the antigen present in the vaccine formulation (VapA+-coated microplates, Figure 7 ). A different situation was observed in farm 7, where fish presented a high level of anti-A. salmonicida IgM prior to the administration of the vaccine. These fish’ specific IgM serum levels dropped to basal values at ~600 TU. When the same samples were tested on microplates coated with VapA- extract, very low antibody levels were detected, with the exception of fish from farm 10, which disclosed a behavior similar to that obtained with VapA+-coated microplates, albeit at a lower relative level ( Figure 7 ).