The in vivo furunculosis challenge was performed at the Veterinary Medical Center Aquatic Facility, Cornell University (Ithaca, NY, United States). Fish were maintained and euthanized prior to measurements and samplings in compliance with standard operating procedures of Cornell University’s Institutional Animal Care and Use Committee (IACUC, protocol number 2017-0033).

Brook charr fingerlings (weight ∼5 g, length ∼5 cm) were obtained from the Aquaculture Center of the State University of New York in Morrisville, NY, United States. Fish were acclimated in 40 L tanks with recirculation system for a period of 30 days before being assigned to an experimental group (Infected, Baseline). For either the “Infected” or “Baseline” group, three independent recirculating aquariums were set up, with each one containing 21 fish in 40 L dechlorinated freshwater maintained at 13.6 ± 0.2°C, pH 7.2, throughout the experiment (i.e., three biological replicates per treatment for a total of six independent aquariums). Dissolved oxygen was maintained above 8 mg/L with air bubbling stones running at maximum flow (1 per tank). Fish were fed 1% body weight daily with 1.5 mm Bio-Vita pellets (Bio-Oregon).

Aeromonas salmonicida subsp. salmonicida strain 890054 was selected from a collection of seven A. salmonicida subsp. salmonicida strains isolated in fish from the New York state to assess a set of genomic and proteomic factors associated with virulence.

Strain 890054 met all three virulence criteria in addition of having been isolated from brook charr. Then, its virulence was verified in a preliminary experiment using adult brook charr ranging from 160 to 260 g body weight. Doses of 100–10,000 CFU were injected intraperitoneally in anesthetized fish. Fish receiving a dose of ≈1,000 and 2,000 CFU showed signs of infection 4–10 days post-infection (dpi).

For each replicate tank, three randomly selected fish (average 7–8 g body weight) received 2,000 CFU of A. salmonicida subsp. salmonicida 890054 suspended in 0.1 mL 1× PBS, pH 7.4 (Infected group) or sterile PBS (Baseline group) by IP injection. A small segment of the caudal fin was clipped for identifying injected fish (Axelsson et al., 2020). Fish were anesthetized in 80 mg/L tricaine methanesulfonate (TMS) during the procedure. Fish were returned to their respective units after recovery from anesthesia.

Survival probability was assessed throughout the 37 days of experiment using R (R Core Team, 2019) and RStudio (RStudio Team, 2016). Survival fits were calculated from mortality counts with package survival (Therneau, 2015). Package survminer (Kassambara and Kosinski, 2018) was used for plotting Kaplan–Meier survival curves. To verify that each death event was caused by furunculosis, kidney stabs from dead or moribund individuals were streaked on Brain–Heart Infusion (BHI) agar (BD) and incubated at 18°C for 3 days. When no colonies producing dark brown pigments [a key phenotype of A. salmonicida subsp. salmonicida (Austin and Austin, 2012)] were found on agar, death events were censored (i.e., not caused by furunculosis). Only one such event occurred out of all 91 recorded mortality events over the whole experiment (event #24, see Supplementary Table 1).

First, the 890054 strain was grown in pure culture (in LB broth) for 24 h at 18°C and then adjusted to an OD₆₀₀ of ∼1.0. A volume of 1 mL of this culture was used for DNA extraction using the DNeasy Blood and Tissue Kit (QIAGEN) with the recommended protocol for Gram-negative bacteria. Genomic DNA was sequenced in short, paired-end reads (2 × 300 bp) with the Illumina MiSeq technology at the IBIS Genomic Analysis Platform (Université Laval, Quebec, QC, Canada). Long reads were obtained with MinION sequencing (Oxford Nanopore); input DNA was prepared using the Ligation Sequencing Kit (SQK-LSK109). The resulting DNA library was analyzed on a R9.5 (FLO-MIN106) flow cell and sequenced on a MinION Mk1B apparatus.

The sequencing effort resulted in 1,177,819 raw short read pairs and 346,938 raw long reads. Short reads with mean Phred quality score greater than 20 were selected with seqtk¹ and adapters were trimmed with Trimmomatic v0.36,² resulting in 1,070,840 filtered short reads. Long reads were basecalled and adapters were trimmed with Guppy v4.2.2-GPU.³ Long reads with mean Phred quality score over 10 were selected with NanoFilt v2.3.0.⁴ An additional filtering step was applied in which Nanopore reads matching with MiSeq reads were enriched with Filtlong v0.2.0,⁵ with the –keep_percent option set to 90 and –min_length set to 1,000. A total of 500,000,000 bases from long reads were retained.

Short and long filtered reads were then assembled with Unicycler v0.4.8⁶ using the default parameters for hybrid genome assembly. Briefly, Unicycler assembled short reads into contigs and bridged those contigs with long reads, and then circularized the complete chromosome and or plasmid sequences when possible.

The taxonomy of 890054 was confirmed with an average nucleotide identity (ANI) analysis using pyani v0.2 (Pritchard et al., 2016). In addition to 890054, a total of 88 representative A. salmonicida genomes spanning across all five known subspecies (achromogenes, masoucida, pectinolytica, salmonicida, and smithia) was included in this analysis. The representative genomes formed clusters that correlated with their subspecies above 99.99% ANI. Therefore, this threshold was used to verify 890054’s belonging to a subspecies.

The final assembly was then annotated with RAST v2.0⁷ using the “RASTtk” algorithm with default parameters. AMR genes were specifically annotated with Resistance Gene Identifier v5.1.1⁸ using the CARD v3.1.1 database. VF subsystems were retrieved from RAST annotations and compared with the ones of reference strain A449 (also annotated for this purpose with RAST to limit annotation bias). Secondary metabolite production was predicted from genome data using antiSMASH v6.0 alpha 1.⁹ Plasmid maps were drawn with SnapGene v5.2¹⁰ and pairwise alignments were drawn with EasyFig v2.2.5.¹¹

The presence of three small cryptic plasmids (pAsa1, pAsa2, and pAsa3) as well as and pAsal1 in 890054 was confirmed by PCR and plasmid profiling using agarose gel electrophoresis as previously described (Boyd et al., 2003; Attéré et al., 2015). For the needs of this study, an agarose gel electrophoresis was performed to visualize small cryptic plasmids (4–6 kb range) in the 890054 strain as well as in three control strains: 01-B526, which is known to bear all four small plasmids, and strains SHY13-2317 and SHY13-3795 which lack pAsa3 and pAsal1, respectively (Attéré et al., 2015). For each strain, plasmid extractions were made by resuspended one quarter of a TSB petri dish culture directly in the first buffer from the QIAprep Spin Miniprep Kit (QIAGEN). After DNA quantitation with a NanoDrop 2000 spectrophotometer (Thermo Fisher), 1 μg of each plasmid DNA extraction was digested for 2 h at 37°C with restriction enzyme EcoRI-HF from New England Biolabs. The DNA fragments were separated on a 0.7%-agarose gel for 90 min at 90 V in 1× tris-borate-EDTA (TBE) buffer. The Quick-load 2-log DNA ladder from New England Biolabs was used as a molecular weight marker. The agarose gel was stained in a bath of ethidium bromide (0.5 μg/mL) after migration to visualize DNA under UV illumination.

We obtained a complete genome assembly that includes one 4.7 Mbp circular chromosome and eight circular plasmids from ∼5.2 to 157 Kbp (Supplementary Table 2). Seven of these plasmids are typical for a strain isolated in North America (pAsa1, pAsa2, pAsa3, pAsal1, pRAS3.3, pAsa9-like, and pAsa5) (Vincent et al., 2014; Attéré et al., 2015; Tanaka et al., 2017). The genome of 890054 contains the typical prophages 1 and 2 found in virtually all A. salmonicida subsp. salmonicida strains as well as AsaGEI1a, a genomic island, identical to the one found in 01-B526 (Reith et al., 2008; Emond-Rheault et al., 2015).

A novel 46.7 kb plasmid (contig 4) with no known equivalent identified in the A. salmonicida plasmidome (Vincent et al., 2020) was compared to sequences from the chromosome and the pAsa5 and pAsa9-like plasmids to rule out potential misassembly from other genetic elements. No significant alignments between contig 4, the chromosome or other plasmids were found (Supplementary Figure 1). This data indicated that contig 4 is a novel plasmid in A. salmonicida subsp. salmonicida that we hereby refer to as p890054-1.

Plasmid p890054-1 is roughly divided into three main regions: a conjugal transfer system, a Vir-like type IV secretion system gene cluster and a putative composite transposon containing several AMR genes (Figure 2A). It is highly similar in structure with pRA3 (Figure 2B), a IncU plasmid from A. hydrophila, a mesophilic Aeromonas species also known for causing infections in fish, amphibians, and even humans (Janda and Abbott, 2010). Globally, p890054-1 and pRA3 share a similar backbone, but a different transposon. It is the second IncU plasmid recently identified in A. salmonicida subsp. salmonicida strains. The first one, pRAS1b (Vaillancourt et al., 2021), shares backbone similarity with p890054-1 except major differences in their transposon (Supplementary Figure 2A). However, unlike pRA3 and pRAS1b, an IS5195 family transposase, flanked by 23 bp inverted repeats, is inserted in a DNA topoisomerase I (topA) gene within the IncU backbone of p890054-1 (Supplementary Figure 2B). This insertion appears to truncate topA in two open reading frames, potentially leading to a non-functional TopA protein. The exact function of this topA gene in p890054-1 is unknown as well as the impact of the inserted transposase gene within this sequence.

Despite being undetected in the Unicycler hybrid assembly, the sequence of plasmids pAsa3 and pAsal1 could be retrieved, each in a single contig, with a reference-based approach, either with Illumina short DNA reads or with Oxford Nanopore long DNA reads (Supplementary Figure 3). Both plasmids were also visible in agarose gels within expected sizes as in other strains known to possess those plasmids (Supplementary Figure 4).

The CARD AMR database identified a series of AMR genes in the chromosome, in the pRAS3.3 plasmid and in the novel p890054-1 plasmid (Supplementary Table 3). More specifically, p890054-1 encodes the following genes: gcn5, sul1, qacEdelta1, dfrB3, and tet(A) (Supplementary Table 3 and Supplementary Figure 5). Based on this analysis, 890054 contains AMR genes against several families of antibiotics: fluoroquinolone, tetracycline, carbapenems, cephalosporin, elfamycin, sulfonamides, acridine, polymyxin, and diaminopyrimidine.

An antiSMASH analysis detected five antimicrobial and secondary metabolite biosynthetic gene clusters in 890054’s chromosome (Supplementary Figure 6A), but none in its plasmids. Among those five, a biosynthesis cluster for pseudomonine, a siderophore usually found in Pseudomonas species. This cluster had 100% similarity with the pseudomonine siderophore of Pseudomonas fluorescens WCS374 (Supplementary Figure 6B), which is a well-known strain used in agriculture against plant pathogens such as pear and apple blight (Erwinia spp.) (Mercado-Blanco et al., 2001). Therefore, using P. fluorescens WCS374 as a candidate to control an outbreak of furunculosis might prove unsuccessful, considering that A. salmonicida subsp. salmonicida strain 890054 possesses its own pseudomonine iron uptake system.

The VF profile of 890054 was similar to reference strain A449 with a few exceptions (Supplementary Table 4). Notably, a Vir-like type IV secretion system (located on p890054-1) was absent in A449, whereas the type VI secretion system (T6SS) genes located on the pAsa4 plasmid in A449 was absent in 890054. However, the A449 T6SS is unlikely to be functional because of several truncated T6SS genes (Reith et al., 2008). Whether this Vir-like system (if active) is involved in conjugation of secretion of effector protein remains to be determined. Also notable was the absence of mannose-sensitive fimbriae and colicin-like bacteriocin production cluster genes, all present in A449 (Supplementary Table 4). Note that the presence of one or more genes does not necessarily imply that a given cluster is active.

Strain 890054 was initially selected as a representative virulent A. salmonicida subsp. salmonicida strain because it possesses its hallmarks of virulence, namely T3SS, protease activity and A-layer (see section “Materials and Methods” “Pathogenic Strain Selection”). We have verified our taxonomic assessment for 890054 by analyzing its complete genome sequence. First, an ANI analysis has clearly situated 890054 within subspecies salmonicida (summarized in Supplementary Table 5; see full table with 88 representative genomes in Supplementary Data Sheet 2). Indeed, 890054 and other genomes from subspecies salmonicida form a subspecies group with less than 0.01% dissimilarity, which is 10 times less dissimilarity than with genomes from other subspecies (i.e., achromogenes, masoucida, pectinolytica, and smithia). In addition, 890054 possesses all four small cryptic plasmids (Supplementary Table 6) from the core A. salmonicida subsp. salmonicida plasmidome proposed by Vincent et al. (2020) as well as AsaGEI1a which is typically found in about 1/3 of the North American A. salmonicida subsp. salmonicida strains (Emond-Rheault et al., 2015). Therefore, we conclude that 890054 is indeed a representative virulent A. salmonicida subsp. salmonicida strain.