Genomics of Tenacibaculum Species in British Columbia, Canada

Tenacibaculum is a genus of Gram-negative filamentous bacteria with a cosmopolitan distribution. The research describing Tenacibaculum genomes stems primarily from Norway and Chile due to their impacts on salmon aquaculture. Canadian salmon aquaculture also experiences mortality events related to the presence of Tenacibaculum spp., yet no Canadian Tenacibaculum genomes are publicly available. Ribosomal DNA sequencing of 16S and four species-specific 16S quantitative-PCR assays were used to select isolates cultured from Atlantic salmon with mouthrot in British Columbia (BC), Canada. Ten isolates representing four known and two unknown species of Tenacibaculum were selected for shotgun whole genome sequencing using the Oxford Nanopore’s MinION platform. The genome assemblies achieved closed circular chromosomes for seven isolates and long contigs for the remaining three isolates. Average nucleotide identity analysis identified T. ovolyticum, T. maritimum, T. dicentrarchi, two genomovars of T. finnmarkense, and two proposed novel species T. pacificus sp. nov. type strain 18-2881-AT and T. retecalamus sp. nov. type strain 18-3228-7BT. Annotation in most of the isolates predicted putative virulence and antimicrobial resistance genes, most-notably toxins (i.e., hemolysins), type-IX secretion systems, and oxytetracycline resistance. Comparative analysis with the T. maritimum type-strain predicted additional toxins and numerous C-terminal secretion proteins, including an M12B family metalloprotease in the T. maritimum isolates from BC. The genomic prediction of virulence-associated genes provides important targets for studies of mouthrot disease, and the annotation of the antimicrobial resistance genes provides targets for surveillance and diagnosis in veterinary medicine.


Introduction
Tenacibaculum is a genus of Gram-negative bacteria that are ubiquitous in marine environments and have beneficial, neutral, or negative interactions with marine organisms [1]. Experimental exposures and circumstantial evidence indicates multiple species of Tenacibaculum as the causative agents of disease in fishes of economic or cultural significance (e.g., salmonids [2][3][4], temperate basses [5], and flatfishes [6]). Fishes affected by 'tenacibaculosis' often display epidermal lesions that can be accompanied by the development of yellow plaques and abnormal behaviours [1]. The signs and severity of tenacibaculosis appear to depend on the species or strain of Tenacibaculum [2][3][4]; the host species, health-status, and life-stage; and the environmental conditions. In British Columbia (BC), Canada, a regional presentation of tenacibaculosis called 'mouthrot' presents as oral plaques and ulcerations on Atlantic salmon (Salmo salar L.). Mouthrot is treated based on

DNA Extractions
The selected isolates were grown in a FMM+K broth to an absorbance (A 600 ) of 0.49 ± 0.11, followed by DNA extractions (gBAC Mini DNA Bacteria Kit, IBI Scientific, Iowa, USA), according to the manufacturer's instructions. Concentrations of the extracted products were quantified using spectrophotometry (NanoVue, GE Healthcare, Illinois, USA), where A 260/280 , and A 260/230 were also recorded. Concentrations of the extracted samples were also quantified using fluorescence (Qubit TM dsDNA BR Assay kit, Invitrogen, Massachusetts, USA), according to the manufacturer's instructions, in a BioSpectrometer ® fluorescence (Eppendorf, Hamburg, Germany).

MinION Sequencing
The isolates were subjected to Oxford Nanopore long-read MinION sequencing at the BC Center of Aquatic Health Sciences (CAHS) in Campbell River (BC, Canada). The samples underwent DNA repair and end-prep, native barcode ligation, adapter ligation and clean-up, priming and loading the flow cell using Oxford Nanopore protocol with the Native Barcoding Expansion 1-12 kit (EXP-NBD104, Oxford, England) and Ligation Sequencing kit (SQK LSK 109, Oxford, England). During the adaptor ligation step, the long fragment buffer was utilized instead of the short fragment buffer. The pooled reactions from the adapter ligation and clean-up were loaded into MinION flow-cell (FLO MIN 106D v.R9, Oxford, England) according to the Oxford Nanopore protocol.

1.
Subsample each barcode independently into 12 maximally independent read sets.

3.
Cluster the contigs produced by each assembly per barcode (resulting in a hierarchical cluster dendrogram, from which subjective decisions were made).

4.
Reconcile (circularize and align to a consistent start position) all of the contigs included in the cluster from the previous step per barcode.

5.
Compute the multiple sequence alignments between all of the reconciled contigs per cluster. 6.
Partition the initial read files (complete sets per barcode) into their appropriate clusters (i.e., chromosome reads to chromosome cluster, plasmid reads to plasmid cluster, etc.). 7.
Compute the consensus for each cluster, for each barcode. Use the reconciled contigs (step 4), alignments (step 5), and raw reads for the given cluster (step 6). 8.
Following 'step 2 in the Trycycler pipeline, barcodes 2, 9, and 10 were in numerous contigs that differed between the read subsets, preventing effective clustering in 'step 3 . As a result, a single-assembler approach was used to assemble these barcodes. Two programs (Flye v.2.9, Raven v.1.6.1) were initially used, and that which produced the most contiguous assemblies was used. Post-assembly and polishing, Ori-Finder 2022 (http://tubic.tju.edu. cn/Ori-Finder2022/public/index.php/index, accessed on 4 November 2022) was used to find the predicted origins of the circular chromosomes [31].

atpA and fusA
Similar to 16S rDNA, each nucleotide FASTA file was mined by text-searching for atpA and fusA genes, except that alignment to interpret single nucleotide polymorphisms were not applied as there was generally one copy for each isolate. Each gene was based on sequences described for atpA [37] and fusA [38]. Due to the limited sequence availability for fusA, select sequences on NCBI that identified as either gene for Tenacibaculum species were used in the alignments. For the generated phylogenies, Kordia algicida strain OT-1 T (NZ DS544873.1) was used instead of Kordia algicida strain OT-1 T (AB681152) as both genes were not described in this sequence. Phylogenetic comparisons were then conducted in a similar manner to the '16S rDNA' methodology.

Multilocus Sequence Analysis (MLSA)
MLSA was accomplished for each annotated barcode by text-searching for atpA, dnaK, glyA, gyrB, infB, rlmN, and tgt. Each sequence was then aligned by gene in MEGAX using MUSCLE and the aligned sequences were trimmed [37]. Post-trimming, the isolated sequences for each barcode were then joined in the order described by the Tenacibaculum PUBMLST database (https://pubmlst.org/organisms/tenacibaculum-spp (accessed on 8 November 2022)). The Tenacibaculum type strain MLSA profiles on the PUBMLST database were exported and included in the alignment. The NCBI sequences for each gene from the Kordia algicida strain OT-1 T (NZ DS544873.1) and the Flavobacterium johnsoniae strain UW101 T (CP000685.1) were collected, and the MLSA profiles were made in a similar manner and included in the alignment. Phylogenetic comparisons were then conducted in a similar manner to the 16S rDNA methodology.

Average Nucleotide Identity (ANI)
The complete genomes of each barcode were compared to each other, as well as other NCBI genomes of Tenacibaculum, using FASTANI (v.1.33, https://github.com/ParBLiSS/ FastANI, accessed on 28 November 2022, [39]). For this comparison, putative plasmid sequences were also included due to their genomic similarities to the contigs in the published assemblies [40]. The included NCBI Tenacibaculum genomes are presented in Table 2.  [42,43]); where Listeria, S. aureus, E. coli, and Enterococcus were designated as 'Select species', the 'Select threshold for %ID' was 90% and the 'Select minimum length' was 60%. A BLAST search also occurred for the putative virulence factors described as toxins in T. maritimum NCIMB 2154 T [7] for barcodes 1-10. Subsequently, a text-search occurred in the Bakta annotated files for the toxin categories described in T. maritimum NCIMB 2154 T [7].

MinION Post-Processing and Quality Control
All of the isolates were assigned a unique barcode for sequencing ( Table 3). The sequencing statistics post-sequencing and -processing indicated a basecalling accuracy greater or equal to 96.8% (Mean Q-score ≥ 15), and over 84 x coverage of the estimated genome size (Table 4).

Genome Assembly and Annotation
The de novo assembly of the Nanopore reads generated closed singular chromosomes for seven of the ten barcodes, with sizes ranging between 2.7-4.2 Mb. Barcodes 2 and 9 had 2-3 chromosomal contigs between 0.7-1.9 Mb (Table 5), based on alignments to the T. dicentrarchi AY7486TD (now T. finnmarkense) and T. maritimum NCIMB 2154 T circular chromosomes ( Figure S1). Barcode 10 assembled into 45 contigs, preventing the confident classification of the chromosomal versus extrachromosomal contigs. Orifinder predicted origins of replication adjacent to the rpiB gene in six of the seven circular chromosomes, and near rpiB in the remaining assembly. As such, the rpiB locus was used to orient all of the circular chromosomes ( Figure 1). Barcodes 2, 3, 4, 8, and 9 also contained smaller, circular contigs between 2-154 kb ( Table 5). The genomic images from barcodes 1 and 3-8 can be found in Figure 1, with high-resolution images in Figure S2. The Bakta annotated nucleotide locus tags are provided in Table S1.

Phylogenetic Resolution Varies with Tenacibaculum Identification Methods
To establish the taxonomic identity of the genomic assemblies, several tools were applied (i.e., ANI, MLSA, fusA, atpA, 16S) that are used to identify Tenacibaculum species. The results are described in the order of highest to lowest species resolution. For downstream figures, colours denoting the species level predictions for each barcode are based

Phylogenetic Resolution Varies with Tenacibaculum Identification Methods
To establish the taxonomic identity of the genomic assemblies, several tools were applied (i.e., ANI, MLSA, fusA, atpA, 16S) that are used to identify Tenacibaculum species. The results are described in the order of highest to lowest species resolution. For downstream figures, colours denoting the species level predictions for each barcode are based on the highest resolution comparison (i.e., ANI analysis). The locations for each used genomic sequence are available in Table S2.

Average Nucleotide Identity (ANI)
Using ANI analysis with a 95% threshold to determine the species based on the previous Tenacibaculum research [11,12], barcode 1 had the highest percentage of nucleotide identity with T. ovolyticum, barcodes 2 and 7 had the highest percentage of nucleotide identity with T. dicentrarchi, barcodes 3, 4, and 8 had the highest percentage of nucleotide identity with T. finnmarkense, and barcodes 9 and 10 had the highest percentage of nucleotide identity with T. maritimum ( Table 6, Table S3). Barcodes 5 and 6 were unlike any of the compared assemblies and were below the 95% threshold required to determine a specieslevel identity ( Table 6, Table S3).

Multilocus Sequence Analysis (MLSA)
Similar to ANI, an MLSA phylogeny using concatenated sequences of atpA, dnaK, glyA, gyrB, infB, rlmN, and tgt resulted in unambiguous species level predictions for barcodes 1-4 and 7-10 ( Figure 2). These primarily mirrored the species assignments made using ANI; however, T. finnmarkense was grouped in a paraphyletic clade. Barcodes 5 and 6 could not be confidently classified but resolved closest to T. piscium and T. ovolyticum, respectively.

atpA and fusA
AtpA was annotated in nine out of ten isolates and was identified using text-search A BLAST comparison was needed to identify the gene in barcode 4. FusA was annota for eight out of ten isolates using text-searches, but a BLAST comparison was needed  Table 6. Average nucleotide identity comparison (%) summary comparing barcodes 1-10 against putative pathogenic Tenacibaculum species. Green highlighted cells represent T. ovolyticum, yellow indicates T. dicentrarchi, blue indicates T. finnmarkense, purple indicates T. maritimum, and grey indicates unclassified barcodes. Bolded cells represent over 95% nucleotide identity between intersecting query and reference sequences. All other comparisons can be found in Table S3.

atpA and fusA
AtpA was annotated in nine out of ten isolates and was identified using text-searches. A BLAST comparison was needed to identify the gene in barcode 4. FusA was annotated for eight out of ten isolates using text-searches, but a BLAST comparison was needed to identify the gene in barcode 4 and 5. Similarly to when using ANI and MLSA, the same species-level predictions could be made for barcodes 1-4 and 7-10 using both atpA and fusA. Based on the branch length, using atpA and a named Tenacibaculum species, barcodes 5 and 6 were positioned closest to T. dicentrarchi and T. ovolyticum, respectively. In contrast, when using fusA and a named Tenacibaculum species, barcodes 5 and 6 were positioned closest to T. finnmarkense and T. ovolyticum. A Flavobacterium sequence was clustered within the Tenacibaculum species for fusA; however, the prediction was not confident and the branch length was greater by an order of two magnitudes, relative to the sister branch ( Figure 3).

3.4.4.16S rDNA
Each barcode had six to nine copies of 16S rDNA (Table 7). There were also variable SNPs (≤ 34) and INDELs (≤14) among the 16S loci within each genome ( Table 7). The multiple partial 16S sequences prevented SNP and INDEL calling for barcode 10. All of the 16S rDNA sequences most closely matched the Tenacibaculum species. Depending on the 16S rDNA copy used in the BLAST comparison, the closest species match could change

16S rDNA
Each barcode had six to nine copies of 16S rDNA (Table 7). There were also variable SNPs (≤34) and INDELs (≤14) among the 16S loci within each genome ( Table 7). The multiple partial 16S sequences prevented SNP and INDEL calling for barcode 10. All of the 16S rDNA sequences most closely matched the Tenacibaculum species. Depending on the 16S rDNA copy used in the BLAST comparison, the closest species match could change for a given barcode (Table 6), which was also supported by the maximum-likelihood phylogenetic comparisons (Figure 4). When using both BLAST comparisons (Table 6) and 16S phylogenies (Figure 4): barcode 1 was most similar to T. ovolyticum; barcodes 9 and 10 were generally most similar to T. maritimum; however, the species level designations for barcodes 2-4 and 7-8 could not be confidently determined, but were clustered in a paraphylogeny around T. dicentrarchi, T. finnmarkense, and other Tenacibaculum species. Based on branch length, barcodes 5 and 6 were positioned closest to T. dicentrarchi and T. haliotis, respectively.

Virulence Related Genes
FeGenie predicted between 40 and 77 iron-related genes per barcode that passed the bitscore cutoff and fell into one of seven categories (i.e., Iron Storage, Iron Gene Regulation, and Iron Acquisition (Siderophore synthesis, Iron Transport, Siderophore Transport, Siderophore Transport potential, and Heme Transport)) ( Table S4). The Virulence Finder

Virulence Related Genes
FeGenie predicted between 40 and 77 iron-related genes per barcode that passed the bitscore cutoff and fell into one of seven categories (i.e., Iron Storage, Iron Gene Regulation, and Iron Acquisition (Siderophore synthesis, Iron Transport, Siderophore Transport, Siderophore Transport potential, and Heme Transport)) ( Table S4). The Virulence Finder tool did not find any notable matches within barcodes 1-10. Each category of putative virulence factors previously described as toxins in T. maritimum NCIMB2154 T [7] were identified by a BLAST search in barcodes 9 and 10 (Table S5). A manual search among the annotated genes for the putative virulence factors described as toxins in T. maritimum NCIMB2154 T [7] identified hemolysins in all of the barcodes, barcodes 1-8 had variable matches to several categories of the described toxins, and barcodes 9 and 10 matched all of the categories (Table S5).

Antimicrobial Resistance Genes
CARD-RGI identified between 1-4 strict matches and 142-244 loose matches for barcodes 1-10 (Table S6). The strict matches included a tetT, tetQ, or tetB like gene in barcodes 1-8, and a vanT, vanX, or vanY like gene in barcodes 1-3 and 5-10. Manual searches of the annotated genes in the barcodes indicated the presence of tetR and tetQ in barcodes 1-8 and 1-10, respectively. No ARG or resistance phenotypes were predicted using ResFinder and ARG-ANNOT.

Genomic Islands
IslandViewer4 and GYPSy are tools that utilize codon-usage, sequence composition, mobile element presence, and genomic comparisons to infer the locations of genomic islands (GI). IslandViewer4 predicted between 3-9 GI within barcodes 1-8, and 18 and 16 GI for barcodes 9 and 10, respectively (Table S7). In contrast to IslandViewer4, GYPSy predicted between 5-26 GI within barcodes 1-10 (Table S7). GYPSy often identified more tRNA and transposases, with the exception of barcode 3. Beyond determining the GI, GYPSy also predicted that 1-10 islands in each barcode were related to pathogenicity, 0-10 islands were related to antimicrobial resistance, 0-6 islands were related to metabolism, and 0-3 islands were related to symbiosis (Table S7).

Gene Content Investigation
When comparing all 10 barcodes, thus considering several Tenacibaculum species, 191 and 973 core-genes (99-100% of barcodes) were predicted at the 95% and 80% gene cluster similarity thresholds, respectively (Table 8). Larger core-gene sets were identified in the comparisons within a single species and when comparing T. dicentrarchi to T. finnmarkense (Table 8).

Genomic Assembly Provides Novel Circular Genomes
Using Oxford Nanopore long-read sequencing technology, seven of the ten Tenacibaculum genomes were assembled into singular, circular chromosomes. This increases the number of complete Tenacibaculum assemblies available for future comparisons and provides novel genomes isolated from marine waters off the coast of BC, Canada. Several studies have suggested a hybrid approach to genome construction where long-read sequencing establishes the genome structure and short-read sequencing offers high perbase accuracy [51][52][53]. For example, a hybrid approach was used on select Tenacibaculum isolates (T. maritimum NCIMB 2154 T (GCA_900119795.1) [7], T. mesophilum DSM 13764 T (GCA_009362255.1) [23]) and provided high-quality circular chromosomes. Future research will focus on using hybrid approaches to sequence and assemble BC Tenacibaculum genomes. A limitation of this study includes the fact that three chromosomes could not be completely assembled, while an unreconciled result of this study includes that the estimated number of nucleotides for barcodes 9 and 10 were~4.2 Mb, which is greater than previously described (i.e.,~3.4 Mb) in T. maritimum [54,55]. This limitation and unreconciled result could be due to the quality of the template DNA, the extraction methodology, the sequencing platforms used, the genome assembly criteria, and potential contamination.

Variable Genomic Resolution and Novel Species
The genomic comparisons between barcodes 1-10 and the type strains of Tenacibaculum yielded phylogenetic insights similar to other published work. It is well established that 16S rDNA sequences can have limited use when comparing Tenacibaculum sequences [24,37,38,56]; however, 16S rDNA qPCR can distinguish genetically distinct species such as T. maritimum [25] and T. ovolyticum [26]. While other genes (i.e., fusA and atpA) can distinguish Tenacibaculum species [38,56], given the fewer sequences available in the databanks, it is unknown if these genes will experience the same limitations as 16S rRNA as more sequences are produced and deposited. For fusA, an outgroup was clustered within the Tenacibaculum genus. Increasing the number of outgroups often provides more robust phylogenies [57,58], and comparing more Flavobacterium and Kordia isolates could cluster outgroups away from Tenacibaculum.
The species-level identification and confidence generally improved as the amount of genome that were compared increased. ANI provided unambiguous, confident specieslevel predictions for all of the barcodes that belong to a known Tenacibaculum. The phylogenetic placement of barcodes 3 and 4 with T. finnmarkense gm. finnmarkense TNO006 and barcode 8 closer to T. finnmarkense gm. ulcerans TNO010 suggest that atpA, fusA, and MLSA may also be able to predict previously established genomovars (11); however, more work is needed to verify this inference. The ANI analysis supported the MLSA results: barcodes 3 and 4 had~98.2% and~96.85% nucleotide identity to T. finnmarkense gm. finnmarkense TNO006 and T. finnmarkense gm. ulcerans TNO010, while barcode 8 had~96.8% and~98.6% nucleotide identity to the aforementioned T. finnmarkense genomovars, respectively. The percent of nucleotide identities distinguishing T. finnmarkense genomovars were the same as Olsen et al., 2020 [11]. A limitation of using MLSA and ANI analysis for diagnostics includes the cost to complete each technique per isolate. In the field, dozens of isolates could be cultured from one collection and can cost thousands of dollars to process, depending on the number of isolates. For cost-effective diagnostic-based research, genomic investigations could establish the fewest genetic targets to reliably determine species identity, which could include sequencing fusA or atpA paired with other genes, or using species-specific genes for a multiplex PCR. Using MLSA and ANI analysis, barcode 1 was T. ovolyticum; barcodes 2 and 7 were T. dicentrarchi; barcodes 3 and 4 were T. finnmarkense gm. finnmarkense; barcode 8 was T. finnmarkense gm. ulcerans; and barcodes 5 and 6 had no similar matches but were within the genus Tenacibaculum, representing a novel species; and barcodes 9 and 10 were T. maritimum.
Barcodes 5 and 6 were unique to any of the defined and compared Tenacibaculum species using ANI. Both of these barcodes had less than 95% nucleotide identity to any of the compared Tenacibaculum species using ANI, which passes the threshold used to determine species [11,12,39], and has been used to establish novel Tenacibaculum species, such as T. piscium [11]. Previous research has used a single locus, often a 16S sequence, to identify Tenacibaculum species, while other biological, chemical, and biochemical traits are supplementary. As a result, it is proposed that barcode 5 (Tenacibaculum sp. 18-2881-A) be denoted as Tenacibaculum pacificus (pacificus; L. neut. adj. pacificum, peaceful; named after the Pacific Ocean (L. Mare Pacificum)) strain 18-2881-A T and barcode 6 (Tenacibaculum sp 18-3228-7B) be denoted as Tenacibaculum retecalamus (retecalamus; L. noun rete-calamum, net-pen) strain 18-3228-7B T .

Iron-Related Genes
Iron-related proteins are necessary for basal physiological processes in prokaryotes and eukaryotes, while virulent microbes utilize similar proteins to induce disease. FeGenie predicted numerous iron-related proteins concerning storage, gene-regulation, and acquisition. Similar predicted proteins are found in Chilean T. dicentrarchi [59] and T. piscium isolates [12], and the T. maritimum type strain [7], indicating that mechanisms to utilize iron between these species, and potentially among the genus, may be similar.
All ten barcodes contained several PF00210-Ferritin-like domain proteins, a non-haem iron storage protein providing vital physiological functions [60], including protection from oxidative stress by sequestering iron and limiting oxyradical formation [60,61]. A study exposing mice to Salmonella enterica serovar (sv.) Typhimurium demonstrated low survival rates (20% survived to d28) using the wild-type, increased survival rates (60% survived until d28) using mutants without ferritin B, and no mortality using mutants without ferritin A and B, and bacterioferritin [62].
All ten barcodes had predicted iron-regulating proteins, including proteins similar to Fur, DtxR, FecR, PchR, PvdS, and Yqil. Iron regulatory proteins have roles in essential physiological processes and in regulating virulent iron-related proteins. For example, S. enterica sv. Typhimurium mutants without fur were avirulent in mice [62] and fur knockouts in Vibrio cholera experienced reduced growth in contrast to the parental strain or ectopic complemented mutants within mice [63].
Numerous predicted proteins were related to iron acquisition; all of the barcodes had proteins relating to iron, siderophore, and heme-related transport. Most notably, all of the barcodes contain an HmuY substrate-binding protein, which is a putative heme-binding lipoprotein associated with the outer membrane [64,65], is associated with virulence [66,67], and has also been described in other Tenacibaculum genomes [7,12,59]. In a study, hfpY co-transcribed with hfpR produced a protein related to HmuY in F. psychrophilum, that contributed to host colonization and disease severity [67]. Wild-type bacteria killed all of the exposed rainbow trout, hfpY and hfpR knockout mutants killed 70% and 40% of fish, respectively, and hfpY and hfpR ectopic complemented mutants killed all the fish [67].
Iron acquisition proteins concerning siderophore synthesis were only predicted in barcodes 1 (T. ovolyticum), 9 and 10 (T. maritimum), while all barcodes encoded proteins with predicted functions for siderophore utilization. Not all bacteria produce siderophores but can often utilize them; the presence of E. coli enterobactin or siderphore producing Maribacter luteus KLE1011 bacteria influenced the presence and appearance of M. polysiphoniae KLE1104 [68], a bacteria which does not produce siderophores but has the ability to utilise the small molecular weight proteins and bind insoluble Fe(III). In another study, three strains of T. maritimum produced iron-sequestering compounds and each could use the compounds secreted by the other two [69]. Potential iron-related virulence genes in Tenacibaculum species warrant investigation as no genetic knockout in-vivo research has occurred to date.

Transport and Secretion Systems
Transport systems are responsible for moving nutrients and proteins within the cell and across cell membranes, as well as moving toxins to the bacterial surface [14]. Similar to T. maritimum NCIMB 2154 T [7], the ABC-type transport, Sec-independent transport (Sec), and twin-arginine transport (Tat) proteins were present in the Bakta annotations of barcodes 1-10. Sec and Tat transport systems are universal but are often involved in virulence [14]. Type IX secretion system proteins were annotated in all ten barcodes; however, type IV and VI secretion system proteins were identified in barcodes 1, 4, 8, 9, and 10, and type II and III secretion system proteins were also present in barcode 1. Type IX secretion systems are described in other Tenacibaculum assemblies of T. maritimum NCIMB 2154 T [7], T. finnmarkense AY7486TD [10,70], T. dicentrarchi isolates [59], T. piscium isolates [12] and T. ovolyticum To-7Br [8,9], and therefore could be conserved within the genus. Previous T. maritimum NCIMB 2154 T exposure studies induced less than 30% mortality in Atlantic salmon under select conditions [2], but other Canadian T. maritimum isolates caused greater mortality, including TmarCan15-1 (100 and 75% mortality in shedders [S] and cohabitants [C]), TmarCan16-1 (100% mortality in S and C), TmarCan16-5 (>80 and 30% mortality in S and C) and T. maritimum 2.1C (barcode 9) (>70 and >60% mortality in S and C) [2,4]. The observed pathogenicity of T. maritimum 2.1C (barcode 9) could be related to the presence of type IV and VI secretion system proteins, suggesting that more work is needed to establish how virulence factors among the Tenacibaculum species are transported.

Putative Antimicrobial Resistance Determinants of BC Tenacibaculum Species
Measuring antimicrobial resistance among isolates helps understand if trends for resistance are developing in a population. The antibiotic florfenicol is commonly used to treat mouthrot among Canadian Atlantic salmon; however, no studies have identified florfenicol resistance among Canadian Tenacibaculum isolates. Studies using florfenicol minimum inhibitory concentration (MIC) on 80 Canadian T. dicentrarchi isolates determined that all of the tested isolates were within the wild-type cut-off value (i.e., 16 µg × mL −1 ) [89]. Similar work in Chile described a T. dicentrarchi isolate with greater MIC for florfenicol beyond their wild-type cut-off (i.e., 4 ug × mL −1 ) [90]. The genomic investigations for barcodes 1-10 did not identify genes or proteins known to confer 'resistance' to florfenicol (i.e., FloR, FexA, Cfr [91]). However, the proteins described to confer resistance to oxytetracycline (i.e., TetQ [92] and TetR [93,94]) were identified. Other genomic investigations have described similar tetQ or tetR genes in T. dicentrarchi [59], T. piscium [12], and T. aestuarii [95]. Other proteins annotated in all ten barcodes may help confer a basal tolerance or resistance to antimicrobials, including, but not limited to: efflux pumps (multidrug resistance protein NorM, multidrug ABC transporter, CusA/CzcA family heavy metal efflux RND transporter); peroxide stress resistance (peroxide stress resistance protein YaaA, superoxide dismutase); and transcriptional regulators (multiple antibiotic resistance protein MarR). The exact mechanism by which several Tenacibaculum species possess a basal level of tolerance or resistance is unknown.

Numerous Genomic Islands among Tenacibaculum Species
Genomic islands (GI) were identified within all ten barcodes, indicating that several Tenacibaculum species may obtain novel genetic material through horizontal transmission.
When using GYPSy, the authors advise using a genome within the same species that is also non-pathogenic [49]. However, the comparisons in this study are limited as few complete Tenacibaculum genomes are available and it is not well known if these genomes are pathogenic. It has been suggested to use multiple genomic island tools (i.e., GYPSy and IslandViewer4) to obtain the greatest accuracy in predicting genomic islands [91]. Both tools identified typical components of genomic islands, such as transposases and integrases; however, the tRNA associated with genomic islands was primarily identified using GYPSy, which is comparable to using both tools in E. coli CFT073 [96]. Few phagerelated proteins were identified in the genomic islands of barcodes 1-8, but between 12 and 24 were predicted in T. maritimum. Phages specific to T. maritimum have been identified [97], and the transmission of the genes between the two should be investigated. In barcodes 2, 3, 6, and 7, a bacteriophage abortive protein (i.e., AbiH) or an abortive infection bacteriophage resistance protein was identified. The presence of an abortive protein indicates that the bacteria could have acquired methods to defend against phage-related infection, such as in E. coli [98] or Lactococcus lactis [99]. GYPSy predicted several genomic islands to be pathogenicity islands in all ten barcodes, with barcodes 1, 2, 5, 6, 7, 9, and 10 also having antimicrobial resistance islands. The occurrence of virulence-and resistance-related genes in the GIs indicates that the horizontal transmission of these genes could contribute to reoccurring infections with several species. More research is needed to investigate how these putatively horizontally acquired genes were obtained, if they can be transferred, and if they contribute to virulence, antimicrobial resistance, or tolerance.

Gene Content Analysis Indicating Diversity among Tenacibaculum Species
In contrast to previous studies [10,54,55], the present study compared the gene content across four Tenacibaculum species, where only 191 and 973 loci comprised the core genome at 95% and 80% gene cluster similarity thresholds, respectively. A small core genome has been recorded in Lactobacillus spp., with 266 core genes [100]. In contrast, larger coregenomes have also been recorded for Legionella spp., Piscirickettsia spp. and Francisella spp., with 886 and 1,732 and 692 core-genes, respectively [101]. Few highly similar genes consisting of the core-genome at 95%, but more intermediately similar genes at 80%, indicate substantial genetic diversity across the genus. In contrast to previous studies [10,54,55], the different core-genome sizes also likely occurred because the genome sizes vary across the Tenacibaculum species (i.e., 2.7 to 4.2 Mb).
Previous gene content analysis predicted 2013, 1947, and 1818 CDS of the core-genomes for three T. dicentrarchi isolates, four T. finnmarkense isolates, and both groups, respectively [10]. In the present study, 2140 ± 0, 2047.5 ± 6.4, 1924.5 ± 4.9 core-genes were identified within T. dicentrarchi, T. finnmarkense, and both groups, respectively, at 95% and 80%. Large core-genomes within and between the two species in both studies suggest that T. finnmarkense and T. dicentrarchi are genetically similar and may have similar interactions with the environment.
Gene content studies comparing T. maritimum core-genomes are consistent in the literature. In one study, 2116 core-genes accounting for~75% of the genes in each genome were described in 25 T. maritimum strains [54]. In another study, 2034 core-genes were identified between 40 T. maritimum strains [55]. In this study, 3819.5 ± 2.1 CDS were similar between barcodes 9 and 10 using both 95% and 80% similarity thresholds. The large increase in the number of CDS compared and the amount of core-genes in this study, in contrast to previous work [54,55], could be attributed to factors including, but not limited to: the sample size and selected isolates (i.e., reduced sample size and geographically similar isolates for the same species would provide more similar core-genes); the sequencing platforms; how the genomes were assembled; the tools used to interpret core genes (Panaroo (this study) vs Microscope [54,55]); and potential contamination.

Conclusions
Ten Tenacibaculum isolates collected from mouthrot outbreaks in BC waters were sequenced with Oxford-Nanopore long-read sequencing technologies. Seven out of the ten isolates were assembled into circular and complete chromosomes. Larger genomic comparisons (i.e., ANI) provided improved species-level resolution in contrast to single gene comparisons. Average nucleotide identity analysis classified the isolates into four species (T. maritimum (T.mar 2.1C, T.mar ATR 174-1B) T. finnmarkense (20-4106-2, 17-2596-1, LI-C6 FM3-F), T. dicentrarchi (20-4116-9,18-3141), and T. ovolyticum (20-4135-2)), and two unknown novel species (T. pacificus sp. nov. type strain 18-2881-A T and T. retecalamus sp. nov. type strain 18-3228-7B T) . Hemolysins were predicted in all of the barcodes, but several other putative toxins were predicted in T. maritimum. Few genes related to antimicrobial resistance were predicted, most notably genes related to oxytetracycline resistance. Subsequent work will focus on identifying whether the predicted genes inform virulence and antimicrobial resistance. This study is the first to describe the genomes of several Tenacibaculum species in Canada and BC waters, which will help inform future phylogenomic, virulence, and antimicrobial resistance research for Tenacibaculum spp. in BC.