The bacterial genome is composed of a core genome containing the genetic information required for essential functions, and a flexible gene pool, which encodes additional traits that can be beneficial under certain circumstances. The flexible gene pool represents variable chromosomal regions and includes mobile and accessory genetic elements, such as bacteriophages, plasmids, insertion sequences, and genomic islands (GIs) (Dobrindt et al., 2004).

Genomic islands are syntenic blocks of accessory genes acquired by horizontal gene transfer (HGT) which contribute to the diversification and adaptation of microorganisms and offer a selective advantage for host bacteria. According to their gene content, GIs are described as pathogenicity, resistance, symbiosis, metabolic or fitness islands (Weinstock et al., 2000; Juhas et al., 2009; Rao et al., 2020). Most GIs are relatively large segments of DNA, usually between 10 and 200 kb, detected by comparisons among closely related strains; however, regions with sizes < 10 kb have also been identified and named genomic islets (Hacker et al., 1990; Juhas et al., 2009; Li and Wang, 2021). GIs usually differ from chromosomes in terms of GC content, tetranucleotide frequency, and codon usage. They are often flanked by small directly repeated (DR) sequences and are often inserted into tRNA genes. They can carry functional or cryptic genes encoding integrases, factors related to plasmid conjugation, phages involved in GI transfer, and transposons or insertion sequence (IS) elements, which may be implicated in mobilizing genetic material onto or deleting DNA from GI (Buchrieser et al., 1998; Juhas et al., 2009; Partridge et al., 2018). Currently, bioinformatics tools developed for GI prediction (Vernikos and Parkhill, 2006; Pundhir et al., 2008; Che et al., 2014; Soares et al., 2016; Wei et al., 2016; Bertelli et al., 2017) are available for the study of these mobile elements.

Aeromonas are Gram-negative bacteria ubiquitous in aquatic environments and can be isolated from virtually every environmental niche where bacterial ecosystems exist (Janda and Abbott, 2010). There are 36 species recognized of Aeromonas (Fernández-Bravo and Figueras, 2020), of which A. hydrophila is the most studied and the first species of the genus to have a sequenced genome (Seshadri et al., 2006; Grim et al., 2014; Rasmussen-Ivey et al., 2016b). A. hydrophila has been isolated from fresh and marine waters, diseased fish, poikilothermic aquatic animals, and warm-blooded animals (Martin-Carnahan and Joseph, 2005). In fish, A. hydrophila can cause a variety of diseases, including septicemia, red sore disease, and ulcerative infections, and has been linked to fish death worldwide, resulting in great economic losses (Janda and Abbott, 2010; Hossain et al., 2013, 2014; Rasmussen-Ivey et al., 2016a). In humans, A. hydrophila is associated with both intestinal and extraintestinal diseases (Martin-Carnahan and Joseph, 2005), such as gastroenteritis, wound infection, septicemia, pneumonia, soft tissue infections, peritonitis, hepatobiliary tract infections, and necrotizing fasciitis (Janda and Abbott, 1998, 2010; Grim et al., 2014).

Genome sequencing and comparative genomic analysis have allowed major advances in the study of Aeromonas virulence. Highly virulent strains recovered from humans or fish have been identified, several genes have been associated with pathogenicity and some are related to GIs in some strains (Seshadri et al., 2006; Hossain et al., 2013; Grim et al., 2014; Pang et al., 2015; Rasmussen-Ivey et al., 2016a). The objective of this study was to analyze A. hydrophila genomes using a GI predictor, determine the distribution of predicted GIs, their characteristics, functions encoded, core genome content, and sharing among strains.

The presence of genes from the core genome was detected in several of the predicted GIs from all strains (Supplementary Table 3), which reached 13% (strain AH10) to 27% (strain AL06-06), with an average of 18% among the A. hydrophila isolates.

However, in a few islands, a high number of genes from the core genome were detected. This was observed mainly in the predicted GIs containing the gene cluster dhaKLM gldA, which were identified as part of the core genome. In addition, genes encoding ribosomal proteins and transcription-related factors, rRNAs, RNA polymerase subunit ββ′, and some tRNAs, among others, are present in the predicted GIs. The presence of repeat regions and transposases between these genes was also observed, which may have contributed to its identification as a GI. The predicted GI containing the cluster scsABCD, associated with copper resistance, also contains a high proportion of genes from the core genome.

Other genes from the core genome in predicted GIs encode the enzyme IMP dehydrogenase, enzymes associated with biosynthesis of tryptophan or threonine, and the enzyme adenylylsulfate kinase. These and other genes related to the core genome are highlighted in blue in Supplementary Table 3.

Several strains share predicted GIs, as indicated by comparative analysis performed considering at least 75% amino acid sequence identity and coverage as criteria (Figure 3; Supplementary Table 6). Some of the predicted GIs were shared among most of the strains, while others were shared only among those of phylogenetic group 1 or group 2. Strains from group 1 shared all or most of their predicted GIs, while those from group 2 shared only a few (Figure 3; Supplementary Table 6).

The main predicted GIs with identified functions shared among two or more strains from both groups included those encoding ScsABCD, DhaKLM/GldA, and enzymes associated with tryptophan biosynthesis/Ivy. Of these, the first is the most widespread and is present in all strains except KN-Mc-1R2. This strain also contains the cluster ScsABCD; however, it is encoded in a larger GI that did not meet the criteria used in this study to define a shared island with other strains. The predicted GIs with characterized functions shared only among strains from group 1 included those encoding inositol and sialic acid/L-fucose metabolism; QueDEC, the O antigen cluster for putative OX1 or OX6 serogroups, and Fic PA0574 (Supplementary Tables 3,6). Additionally, predicted GIs encoding toxin/antitoxin systems, phages or hypothetical proteins are shared among them. In group 2, predicted GIs encoding the zona occludens toxin/accessory cholera enterotoxin and flagellar regions 1 and 2 were the most common shared GIs (Supplementary Tables 3,6).

In this study, complete genomes of A. hydrophila strains were analyzed using the Island Viewer 4 bioinformatics tool for GI prediction in order to determine the distribution of GIs and their contributions in such strains, however, while some of the features encoded have functions defined, others are putative and have to be experimentally demonstrated in Aeromonas. A phylogenetic analysis performed with SweeP software showed that two strains identified as A. hydrophila (4AK4 and YL17) were misidentified at species level. These findings are in agreement with the studies by Beaz-Hidalgo et al. (2015) and Moriel et al. (2021) which indicated several problems with the taxonomic affiliation of Aeromonas genomes deposited at the NCBI database. These findings show the relevance of using complete genomes for taxonomy of Aeromonas. The A. hydrophila strains were classified into two distinct phylogenetic groups (Figure 1). Group 1 contained strains isolated from diseased fish, except pc104A, which was recovered from the soil of a catfish pond that experienced an epidemic outbreak of septicemia caused by Aeromonas (Pang et al., 2015). Most of the strains were related to epidemic outbreaks of motile Aeromonas septicemia in catfish that occurred in the United States (ML09-119, AL09-71, and pc104A) or carp in China (NJ-35, JBN2301, and J1), and share a recent common ancestor, as shown by comparative genomic analyses (Hossain et al., 2014). These strains were previously classified as hypervirulent A. hydrophila (vAh) based on phenotypic and genotypic tests or in silico analysis for pathotype-specific PCR of vAh isolates (Pang et al., 2015; Rasmussen-Ivey et al., 2016a).

Strains ZYAH72, D4 and GYK1, isolated from diseased fish in China, are also part of phylogenetic group 1. These United States and Chinese strains were separated into distinct subclades of group 1 (Figure 1), what is in agreement with the data of Pang et al. (2015).

Strains from phylogenetic group 2 were recovered from different sources and were also separated into two subclades based on their geographical origins (Figure 1), and the South Korean strain grouped together with the United States strains. Three of these strains (AH10, ATCC 7966, and AL06-06) were previously studied and considered to be non-virulent A. hydrophila (non-vAh) in relation to the hypervirulent vAh strains recovered from fish disease outbreaks (Pang et al., 2015; Rasmussen-Ivey et al., 2016a). Our data confirmed the results of other phylogenetic studies indicating the separation of vAh from non-epidemical strains (Pang et al., 2015; Rasmussen-Ivey et al., 2016a).

Strains from groups 1 and 2 differed in several characteristics, including genome size, number of predicted GIs, GI average size, and ratio of GIs relative to genome size (Table 2), in which strains from group 1 generally present greater numbers. Genomic GC content also varied among these strains, from 60.50% to 60.90% in group 1 and from 61% to 61.60% in group 2. Strains from groups 1 and 2 also differed in terms of the functions encoded on the predicted GIs, although some features were common among them (Table 3).

The distribution of predicted GIs containing the main functions identified, and some features whose function have yet to be demonstrated in A. hydrophila, varied among the strains (Table 3), while some were common and other were encoded in a few strains. Among the identified characteristics are clusters for the synthesis of O Ag. O Ag, the most surface-exposed part of the lipopolysaccharide, a component of the outer membrane of Gram-negative bacteria, mediates pathogenicity by protecting infecting bacteria from serum complement killing and phagocytosis. It consists of repeating oligosaccharide subunits whose variability confers immunological specificity (Tomás, 2012). Several O Ag gene cluster types have been described in A. hydrophila (Hossain et al., 2013; Pang et al., 2015; Cao et al., 2018) and previously associated with GIs (Hossain et al., 2013; Pang et al., 2015). Gene clusters for O Ag were found in the predicted GIs of all strains; however, they differed in their gene content and were related to 10 distinct serogroups (Figure 2).

Thirteen of the strains analyzed in this study had their O Ag gene clusters previously determined (Hossain et al., 2013; Pang et al., 2015; Cao et al., 2018).

The putative serogroup OX1 was proposed for strains ML09-119, AL09-71, pc104a, JBN2301, and D4 strains (Cao et al., 2018). Strain ZYAH72, which had no previously analyzed O Ag cluster, shared the same cluster with the strains above (Figure 2). Thus, it should also be included in the putative OX1 serogroup. Strains NJ-35 and GYK1 were classified as putative serogroups OX6 (Cao et al., 2018). According to our data, strain J-1 should also be included in this serogroup because it shares the same O Ag biosynthesis cluster. This is in agreement with Pang et al. (2015), who showed that J-1 and NJ-35 have identical O Ag gene clusters.

All other strains had unique O Ag clusters. In contrast to the vAh fish pathogens that presented putative OX1 or OX6 O Ag clusters, strains AL06-06 and AH10 were also recovered from diseased fish but were considered as non-vAh (Rasmussen-Ivey et al., 2016a) and presented clusters OX4 and OX5, respectively (Cao et al., 2018). Strain ATCC 7966 belongs to the O1 serogroup (Seshadri et al., 2006), AHNIH1 to O25, WCHAH045096 to OX7 (Cao et al., 2018). The ZYAHA75 O cluster is similar to O18, and those of MX16A and KN-MC-1R2 were not determined (Figure 2).

Thus, no O Ag gene cluster was shared among strains from phylogenetic groups 1 and 2. All strains from group 2 exhibited unique O Ag clusters. Of the 10 distinct serogroups detected, only OX1 and OX6 were shared, and the identification of these clusters might be an alert for potential virulent strains related to outbreaks of epidemic diseases in fish.

Several genes associated with polar flagella biogenesis and chemotaxis were identified in the predicted GIs of most strains (Table 3; Supplementary Table 3). Mesophilic Aeromonas have a polar flagellum for swimming in liquid, and some strains can also express multiple inducible lateral flagella for swarming over viscous environments or surfaces (Janda and Abbott, 2010). The polar flagellum genes are clustered in five distinct regions, whereas lateral flagella genes are organized in only one cluster (Tomás, 2012). Beyond their role in motility, in Aeromonas flagella act as adhesins for human intestinal enterocytes and in biofilm formation (Kirov et al., 2004), and are essential for adhesion and the ability to invade fish cell lines (Beaz-Hidalgo and Figueras, 2013). In the current study, GIs containing gene clusters for polar flagella region 2 (flaA flaBGHJ maf-1) (Canals et al., 2006) were predicted. However, the data indicated that gene content downstream of polar flagella region 2 is heterogeneous in different strains (Supplementary Table 3) and that genes associated with flagella glycosylation, such as flmD and neuB (Tabei et al., 2009; Wilhelms et al., 2012; Fulton et al., 2015) are adjacent to region 2 in certain A. hydrophila isolates. These data are in agreement with a recent study (Forn-Cuní et al., 2021) that showed that pse/flm genes (flmA flmB neuA flmD neuB) related to polar flagellin glycosylation are clustered in highly polymorphic GIs with three main genomic glycosylation islands identified.

Groups I and III contain almost exclusively the pse genes (flmA flmB neuA flmD neuB). Aeromonas belonging to these groups have flagellins modified with a single monosaccharide pseudaminic acid derivative (Forn-Cuní et al., 2021). Strains from phylogenetic group 1 and strain AH-10 belong to the glycosylation genomic island group I, while MX16A and WCHA045096 were included in group III according to Forn-Cuní et al. (2021).

The glycosylation GI group II contains other genes between the pse/flm genes, and are present in strains that modify polar flagellins with heterogeneous glycan moieties. Strains ATCC 7966, AL06-06, and AHNIHI are part of this group together with KN-Mc-1R2 and ZYAHA75 (Forn-Cuní et al., 2021). Flagellin glycosylation is essential for flagellar function (Tabei et al., 2009) and polar flagellar glycan composition can differ between A. hydrophila strains (Wilhelms et al., 2012; Fulton et al., 2015; Forn-Cuní et al., 2021). Because Aeromonas flagella have several roles in pathogenesis (Kirov et al., 2004; Beaz-Hidalgo and Figueras, 2013), based on data showing that strains ML09-119 and AL09-71 (pathogens of catfish), AH10, J-1, JBN2301, and NJ 35 (pathogen from carp) share the same predicted GIs related to flagellar region 2, and results from Forn-Cuní et al. (2021), showing that they belong to the glycosylation GI group I, we speculate that flagellin glycosylation patterns may favor interactions with determined hosts.

VgrG and Hcp play roles in Aeromonas virulence, influencing bacterial motility, protease production, and biofilm formation, inducing host cell toxicity, apoptosis, and activation of macrophages (Suarez et al., 2008, 2010a,b; Sha et al., 2013). They are components of the T6SS which in addition to virulence plays roles in inter-bacterial competition. T6SS is encoded within gene clusters containing the core genes of the secretion machinery, however, additional copies of hcp and vgrG can be found outside the main cluster (Cianfanelli et al., 2016; Moriel et al., 2021), which explains the presence of these genes in two distinct predicted GIs in certain strains (Supplementary Table 3). However, a search in the genome showed that strain AL06-06 does not have the complete T6SS cluster, differently from strains AH10 and KN-Mc-1R2. This is in agreement with Tekedar et al. (2019) which showed that this strain have only three components of T6SS, hcp, tssH and vgrG. In their study they also showed that virulent strains including ML09-119, AL09-71 and pc104A also have only these T6SS components, and that hcp1 and vgrG1 contribute for virulence of the ML09-119 in catfish.

Some other genes with functions identified in predicted GIs (Supplementary Table 3) are associated with virulence in other bacteria, but their roles in Aeromonas have yet to be determined. Among them are the zonula occludens toxin and the accessory cholera enterotoxin, which have been described in Vibrio cholerae and have the ability to affect intestinal function in rabbits (Trucksis et al., 1993; Fasano et al., 1997). Another is the ankyrin protein, AnkB. In plant-pathogens Xanthomonas oryzae pv. oryzae and X. oryzae pv. oryzicola, AnkB participates in biofilm formation, swimming ability, exopolysaccharide production, and defense against oxidative stress. In these bacteria, the ankB gene is located 58 bp downstream of catB, which encodes the enzyme catalase. AnkB affects catB gene expression, catalase activity, and sensitivity to H₂O₂ (Pan et al., 2018). In A. hydrophila, the ankB gene is in tandem with katE, which encodes catalase, and in most strains with a gene encoding a superoxide dismutase precursor. Although A. hydrophila is an opportunistic pathogen in animals, this led us to speculate regarding its roles in defense against oxidative stress, as shown in X. oryzae.

Fic proteins are widespread in bacteria and are often found encoded in GIs. They may play a role as toxins secreted from pathogens, or as the toxin component of toxin/antitoxin modules. However, the great majority of Fic proteins have unknown functions (Veyron et al., 2018).

Several genes from the core genome were found in the predicted GIs (Supplementary Table 3), which should be further considered. Except for the scsABCD locus associated with copper resistance, which was present in the predicted GIs of all strains analyzed, the distribution of GIs encoding resistance-associated features was limited to four strains in phylogenetic group 2 (Table 3). The full scs operon appears to be limited to some Enterobacteriaceae (Salmonella enterica, Serratia proteamaculans, Citrobacter koseri, and Klebsiella pneumoniae), Aeromonas hydrophila, and Photobacterium profundum (Subedi et al., 2019). The scsABCD locus was originally identified in S. enterica Typhimurium, and is required to deal with copper and H₂O₂ stress, helping to increase Salmonella survival under severe copper and oxidative stress, hostile conditions encountered by the pathogen during its intracellular survival (López et al., 2018) thus, contributing to virulence. However, copper may also be found in water (Domek et al., 1984). Considering that Aeromonas are ubiquitous in aquatic habitats and are considered as opportunistic pathogens (Janda and Abbott, 2010), potential roles of scsABCD may contribute to fitness in environment and virulence in the host.

In A. hydrophila, scsSBCD is located in the predicted GIs distinct from those encoding features associated with resistance (Supplementary Table 3).

Aeromonas have intrinsic resistance to some β-lactams due to the production of one to three inducible β-lactamases. The resistance phenotypes vary in relation to the species according the production of one to three inducible β-lactamases, with five basic phenotypes described (Fosse, 2010). In A. hydrophila three classes of β-lactamases, a class B metallo-β-lactamase, a class C cephalosporinase, and a class D oxacillinase were described (Fosse, 2010; Janda and Abbott, 2010), conferring to the species the penicillinase-cephalosporinase-carbapenemase phenotype exhibiting resistance to ampicillin, amoxicillin, ticarcillin, cephalothin, and profile susceptible or intermediate to imipenem (Fosse, 2010). We looked for genes encoding these beta lactamases and subclass B2 metallo-β-lactamase (CphA family, or ImiH or ImiS), cephalosporin-hydrolyzing class C β-lactamase (CepS or CepH) and OXA-12 family class D β-lactamase (AmpH/OXA-724 or AmpS/OXA-725 or OXA-726) were identified in the chromosome of all 17 strains. Additionally, a tetracycline efflux MFS transporter TetE was identified in the chromosome of 2 strains.

Aeromonas are usually susceptible to aminoglycosides, tetracyclines, chloramphenicol, sulfonamides, trimethoprim, nitrofurans, nalidixic acid, fluoroquinolones (Fosse, 2010), carbapenems, third- and fourth-generation cephalosporins, macrolides, monobactams and extended spectrum penicillins (Janda and Abbott, 2010; Fernández-Bravo and Figueras, 2020). However, strains presenting resistance to one or more of these antibiotics are common in clinical and environmental strains (Aravena-Román et al., 2012; Esteve et al., 2015). Studies on molecular antibiotic resistance have shown that class 1 integrons, involved in intraspecific and interspecific dissemination of resistance, and large plasmids are associated to resistance to several antibiotics in Aeromonas (Nguyen et al., 2014; Hughes et al., 2016).

Aeromonas hydrophila multidrug-resistant strains are frequent, mainly among clinical isolates (Esteve et al., 2015; Hughes et al., 2016). Among the strains analyzed, the following contains plasmids, as informed at the NCBI database: JBN2301 (3 plasmids), D4 (4 plasmids), AHNIH1 (1 plasmid), AL06-06 (3 plasmids) and WCHAH045096 (6 plasmids). Plasmids from strains JBN2301and AL06-06 do not contain resistance genes. In strain D4, only one of the plasmids contain a gene encoding a quinolone resistance pentapeptide repeat protein QnrS2. The plasmid of strain AHNIH1, contains several genes encoding for resistance, as described (Hughes et al., 2016). Includes three class A beta-lactamases (KPC-2, SHV-12 and CARB-12) and genes encoding for resistance to aminoglycosides, chloramphenicol, fluoroquinolones, macrolides, sulfonamide and trimethoprim. Three of the plasmids of WCHAH045096 carry resistance genes. They are associated to the resistance to several classes of antibiotics including aminoglycosides, sulfonamide, trimethoprim, quinolone, tetracycline and carbapenem.

Predicted GIs containing antibiotic resistance associated genes had heterogeneous distribution among the A. hydrophila (Supplementary Table 4), possibly reflecting horizontal gene transfer and selection with the exposure to antibiotics in their environments. These predicted GIs were identified in four strains, recovered from water, sewage, fish or human, indicating that antibiotic resistance associated with GIs is spread in clinical and environmental strains. Most of these predicted GIs also encode transposases, integrases, and mobile elements, and one of them also includes proteins related to conjugation (Supplementary Table 3). In the strain AL06-06, a predicted GI encodes for aminoglycoside, sulfonamide, chloramphenicol and tetracycline resistance. The predicted GI of WCHAH045096 encodes for resistance to the same antibiotics described above, and contains genes for several proteins associated with mercury resistance. The predicted GI of strain MX16A encodes antibiotic resistance for tetracycline, aminoglycoside, chloramphenicol, sulfonamide, macrolide and beta-lactams e proteins related to mercury resistance. Strain ZYAH75 has three predicted GIs containing resistance genes. One of them contain genes for resistance to aminoglycoside and sulfonamide together with proteins for mercury resistance. This predicted GI is a mosaic of proteins associated with mobile elements, it encodes conjugative transfer proteins TrbJ, TrbK and TrbK, IncP-type oriT binding proteins TraJ and TraK, together with integrase, transposase, IS and mobile element proteins. Genes for chloramphenicol and aminoglycoside resistance are encoded in two other islands, which contains additionally genes encoding class A extended-spectrum beta-lactamase CTM-3, class A broad-spectrum beta-lactamases TEM-1, and gene for sulfonamide resistance; or genes for resistance to macrolides and quinolone, oxacillin-hydrolyzing class D beta-lactamase OXA-10, and also proteins associated with mercury resistance. Integron integrase Intl1 are encoded in two of the predicted GI of ZYAH75, and those from strains WCHAH045096 and MX16A. In several of these predicted GIs there is more than one gene encoding for aminoglycoside modifying enzymes, and for sulfonamide resistance. Therefore, in addition to the intrinsic resistance characteristic of the species, some strains analyzed contain determinants for resistance to tetracycline in the chromosome (2 strains); plasmids carrying genes for resistance to one and up to six distinct classes of antibiotics, and even more than one plasmid related to resistance, and predicted GIs in which several genes of resistance are encoded, the most common are those related to the aminoglycoside and sulfonamide resistance. Aeromonas are ubiquitous in aquatic environments and widely distributed, being found in every environmental niche where bacterial ecosystems exist, facilitating the contact with humans (Janda and Abbott, 2010). Thus, the presence of GIs carrying resistance genes in Aeromonas may impact the fitness of the bacteria, facilitate the resistance dissemination, and spread of resistant strains in several environments.

Some of the predicted GIs shared among strains from phylogenetic group 1 encode metabolic pathways for utilization of myo-inositol and L-fucose/sialic acid, which were previously described in GIs of A. hydrophila strains NJ-35, J-1, ML09-119, AL09-71, and pc104A (Pang et al., 2015). In the current study, these features were also found in the predicted GIs of other strains of group 1. These pathways may be linked to the full virulence of fish disease epidemics A. hydrophila (Pang et al., 2015), suggesting that these GIs may be related to fitness or enhancing adaptation to hosts and competitiveness.

Other predicted GIs common to vAh contain genes encoding QueD, QueE, and QueC (Supplementary Table 3), which participate in the synthesis of PreQ₀, a precursor of queuosine- (Q) tRNA modification in bacteria. PreQ₀ is reduced to PreQ₁ and inserted in the G residue at position 34 of tRNAs with GUN anticodons by tRNA guanosine transglycosylase (bTGT); however, the enzyme can also use PreQ₀ when PreQ₁ is absent (Thiaville et al., 2016). However, predicted A. hydrophila GIs encoding QueDEC also encode TgtA5, a variant of TGT that has divergent features from bTGT, helicases, and proteins involved in DNA repair (Supplementary Table 3). Similar clusters have been described in a GI of Salmonella enterica serovar Montevideo and by comparative genomic analysis in several other bacteria, including A. hydrophila ML09-119. In S. enterica serovar Montevideo, Kineococcus radiotolerans, Comamonas testosteroni, and Sphingopyxis alaskensis, clustering of tgtA5 and preQ₀ synthesis genes is involved in inserting 7-deazapurine derivatives into DNA, suggesting a new and complex system of DNA modification (Thiaville et al., 2016). The SAM-dependent methyltransferase HI0095 (UbiE paralog), which is associated with virulence in vAh (Rasmussen-Ivey et al., 2016a) is also encoded in this predicted GI. Thus, the roles of these genes in A. hydrophila DNA modification and/or virulence have to be determined.

No distinctive characteristics of strains from group 2 were identified, and most isolates shared only a small number of predicted GIs (Supplementary Table 6), suggesting a high degree of strain diversity. Of note, several of these strains have multiple determinants of antimicrobial resistance encoded by their predicted GIs. Group 2 contains only eight strains; however, they may better reflect the diversity of GIs in A. hydrophila because previous studies have shown that most strains from group 1 are highly related and constitute a clonal group (Hossain et al., 2014; Pang et al., 2015; Rasmussen-Ivey et al., 2016a).

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

AS designed the study, collected and data analysis, and manuscript writing. JM did the funding acquisition, data evaluation, and original draft writing. RR did the conception and design of the manuscript. CD did the phylogeny analysis. DJ developed scripts and contributed to data analysis. CF-P did the data analysis and manuscript writing. GP reviewed and edited the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.