Are the genomes of motile Aeromonas species from the aquatic environment shaped by local geography and host species?

Abstract

Aeromonas dhakensis and Aeromonas hydrophila cause significant economic losses within the global aquaculture sector, affecting numerous farmed fish species and posing a zoonotic threat to human health. In this study, we sequenced, assembled and analysed the genomes of seven and five A. dhakensis and A. hydrophila isolates, respectively, recovered from disease outbreaks in various fish hosts across Southeast Asia using a hybrid sequencing approach with Illumina and Oxford Nanopore technologies. To assess the relatedness of our isolates with those from the global aquatic environment, with particular reference to aquaculture-relevant systems, and compare their resistome and virulome profiles, we also conducted comparative genomic analysis with an additional 57 publicly available genomes of A. dhakensis and A. hydrophila, recovered from different aquatic sources. Findings from this study revealed large genomic variability across global Aeromonas populations, with a clear distinction in the pan-genome between Aeromonas species. In silico multi-locus sequence type analysis revealed a wide distribution of sequence types (STs) 656 and 251 in Asia in A. dhakensis and A. hydrophila, respectively, although the novel STs detected in Indonesia and the Philippines suggest local adaptation of populations circulating in these countries. Analysis of mobile genetic elements revealed plasmids, insertion sequences and genomic islands to be country- or host-specific. Exploration of resistome data revealed a high prevalence of multidrug resistance across Southeast Asia, with genomes frequently carrying resistance genes against antibiotic classes commonly used across the aquaculture sector, including potentiated sulphonamides and tetracyclines. However, several antimicrobial resistance genes were found to be country- or host-specific, suggesting local adaptation to anthropogenic or environmental conditions within specific regions. A diverse repertoire of virulence genes was also detected in this study, with A. hydrophila demonstrating greater diversity in virulence genes compared with A. dhakensis. Country-specific virulence pathways were also noted for several toxins and secretion systems, including aerA/act, rtx and type III and IV secretion systems. This work provides new insights into the genomic features and relatedness of Aeromonas spp. circulating within aquaculture systems in Southeast Asia. Further, findings from comparative genomic analysis highlight the influence of geographical or host pressures on the molecular drivers of antimicrobial resistance and pathogenicity, directly impacting animal health and aquaculture production.

Impact Statement.

Motile Aeromonas species, including Aeromonas dhakensis and Aeromonas hydrophila, are economically important bacterial pathogens that continue to post a significant threat to the production of freshwater fish across the global aquaculture sector. In this study, we sequenced the genomes of seven and five A. dhakensis and A. hydrophila recovered from different fish species across Southeast Asia. We then performed comparative genomic analysis with publicly available motile Aeromonas genomes from different aquatic sources, investigating the Aeromonas pan-genome and genomic drivers behind antimicrobial resistance and pathogenicity. We showed that sequence types (STs) 656 and 251 dominated the genome collection of A. dhakensis and A. hydrophila from Vietnam, respectively, although novel STs were detected in Indonesia and the Philippines. We also demonstrated high levels of predicted antibiotic resistance across Southeast Asia, with several antibiotic resistance genes against popular antibiotic compounds detected. Further, both country- and host-specific resistome profiles were identified. A diverse virulome was detected in both A. dhakensis and A. hydrophila, albeit shaped by geography. This work has considerably expanded our understanding of the genomic features and relatedness of Aeromonas spp. circulating within global aquaculture systems. Findings from this study will contribute to ongoing surveillance programmes and help to inform disease control and management strategies applied across the Southeast Asia aquaculture sector.

Data Summary

Whole-genome sequencing data generated in the study are available in the National Center for Biotechnology Information (NCBI) Sequence Read Archive repository under the accession number PRJNA1101888, and the accession numbers for all datasets used are provided in Table 1. The authors confirm that all supporting data, code and protocols have been provided within the article or through supplementary data files.

Introduction

The genus Aeromonas (phylum, Proteobacteria; class, Gammaproteobacteria; order, Aeromonadales, and family, Aeromonadaceae) is characterized as Gram-negative, rod-shaped bacteria [1]. The genus is found ubiquitously in the aquatic environment, primarily inhabiting freshwater sources, although representative isolates have also been recovered from marine environments as well as potable and wastewater sources [2,3]. First recovered in the late nineteenth century, where isolates were identified as Bacillus hydrophilus fuscus but re-classified as Aeromonas hydrophila in 1943, the genus now comprises 34 valid species [4]. Several Aeromonas species cause disease in humans and terrestrial animals [5,9]. Likewise, members of this genus also infect a range of aquatic animals, including fish, marine reptiles and marine mammals [10,12].

In aquaculture, several Aeromonas species are known to cause high mortality and significant economic losses in production systems through disease outbreaks from furunculosis and motile Aeromonas septicaemia (MAS), caused by Aeromonas salmonicida and members of the motile Aeromonas species, respectively [13]. The close relatedness of the motile Aeromonas species associated with natural disease outbreaks in farmed fish has led to the description ‘motile Aeromonas complex’, which generally presents externally in infected fish as reddened or eroded fins, ulceration, scale loss and internally as haemorrhaging and swelling of the kidney and spleen, although host-specific differences have been noted [14]. Taxonomic assignment of motile Aeromonas species has been historically challenging, with Aeromonas dhakensis isolates often misidentified as A. hydrophila due to similarities in phenotypic profiles between the two species or the poor discriminatory power of traditional housekeeping genes and databases used in species identification [15]. However, advancement in molecular approaches, including whole-genome sequencing and average nucleotide identity (ANI), has greatly improved species-level identification, particularly in distinguishing A. dhakensis and A. hydrophila [14]. As such, whilst A. hydrophila is frequently associated with MAS outbreaks in several fish groups, including carps, tilapia and catfish, A. dhakensis is now considered an emerging bacterial pathogen of importance within the aquaculture sector [16].

Advances in next-generation sequencing have revolutionized the study of bacterial genomes, leading to deeper insights into genome structure and evolution, as well as identifying genetic drivers in pathogenicity and antimicrobial resistance (AMR). Indeed, comparative genomics has allowed for a comprehensive understanding of the genome structure and variability in several Aeromonas species, including Aeromonas caviae, A. salmonicida and Aeromonas veronii [17,20]. Comparative genomics has also revealed country- or host-specific genotypes in other notable bacterial pathogens affecting fish [21,23]. In this study, the genomes of seven A. dhakensis and five A. hydrophila isolates from various hosts in Indonesia, the Philippines and Vietnam were sequenced using Illumina and Oxford Nanopore Technology platforms. To better understand the population dynamics of A. dhakensis and A. hydrophila circulating the global aquatic environment, comparative genomic analysis was conducted among these and an additional 57 publicly available genomes associated with different aquatic hosts (including aquaculture-relevant species) and geographical locations, to determine their phylogenetic relatedness and provide deeper insights into unique genomic factors related to antimicrobial resistance, evolution and virulence.

Methods

Aeromonas isolates

A total of 69 motile Aeromonas isolates (n=26 A. dhakensis and n=43 A. hydrophila) from various aquatic hosts and different geographical locations were investigated in this study. The genome sequences from 19 and 38 A. dhakensis and A. hydrophila isolates, respectively, were retrieved from the National Center for Biotechnology Information (NCBI) RefSeq database (28 August 2024, Table 1). Genomes were selected based on the following criteria: where ‘Complete Genome’ or ‘Chromosome’ level was available, and the NCBI taxonomy check verified species identity as either A. dhakensis or A. hydrophila. The genomes of the other 12 isolates (n=7 A. dhakensis and n=5 A. hydrophila) were sequenced in this study (Table 1).

Table 1. Genomic features of the A. dhakensis and A. hydrophila isolates investigated in this study.

Isolate	Accession	Species	Country	Host	Completeness (%)	Contamination (%)	Size (Mbp)	GC %	CDS (total)	rRNA gene	No. plasmid (accession)
Newly sequenced genomes
AD-92HK1	SAMN41005112	A. dhakensis	Indonesia	Oreochromis niloticus	100	0	4.83	61.64	4258	31	–
AD-69	SAMN41005113	A. dhakensis	The Philippines	Oreochromis niloticus	100	0	4.89	61.56	4367	31	–
AD-VN1	SAMN41005116	A. dhakensis	Vietnam	Channa striata	100	0	4.97	61.1	4411	31	4 (CP152260, CP152261, CP152262,CP152263)
AD-VN2	SAMN41005117	A. dhakensis	Vietnam	Channa striata	100	0	4.96	61.11	4403	31	4 (CP152255,CP152256, CP152257, CP152258)
AD-27	SAMN41005114	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	100	0	4.94	61.22	4401	31	4 (CP152270, CP152271, CP152272, CP152273)
AD-1	SAMN41005115	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	100	0	4.97	60.97	4431	31	4 (CP152265, CP152266, CP152267, CP152268)
AD-9	SAMN41005118	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	100	0	5.03	61.01	4497	31	4 (CP152250, CP152251, CP152252, CP152253)
AH-40	SAMN41005119	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	100	0	4.84	61.01	4294	31	1 (CP152248)
AH-32	SAMN41005121	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	100	0	5.03	60.91	4478	31	2 (CP152242, CP152243)
AH-34	SAMN41005120	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	100	0	5.07	60.83	4521	31	2 (CP152245, CP152246)
AH-12	SAMN41005122	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	100	0	5.06	60.85	4501	31	2 (CP152239, CP152240)
AH-7	SAMN41005123	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	100	0	5.1	60.81	4536	31	1 (CP152237)
RefSeq genomes from NCBI
Aer_OnIF1	GCF_022703095.1	A. dhakensis	Brazil	Oreochromis niloticus	99.91	0.23	4.83	61.65	4290	31	–
Aer_On24M	GCF_017310095.1	A. dhakensis	Brazil	Oreochromis niloticus	99.73	0.29	4.93	61.37	4410	31	–
Aer_Pi12.1HTAS	GCF_025266835.1	A. dhakensis	Brazil	Phractocephalus hemioliopterus	99.95	0.61	4.82	61.35	4303	25	1 (CP050696)
202108B1	GCF_034143565.1	A. dhakensis	China	Ancherythroculter nigrocauda	99.96	0.23	5.04	60.97	4457	31	4 (CP133574, CP133575, CP133576, CP133577)
202108 C2	GCF_035658375.1	A. dhakensis	China	Ctenopharyngodon idella	99.96	0.23	5	61.03	4409	31	–
b2-100	GCF_023920205.1	A. dhakensis	China	Danio rerio	99.92	0.56	5.02	61.46	4537	31	–
VA54	GCF_037076385.1	A. dhakensis	Vietnam	Anabas testudineus	99.96	0.51	4.85	61.55	4274	31	–
VA6	GCF_037101205.1	A. dhakensis	Vietnam	Mastacembelus	99.93	0.58	4.97	61.43	4379	31	1 (AP027932)
TN45	GCF_905132775.1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	99.96	0.2	4.77	61.4	4318	2	4 (LR963080, LR963081, LR963082, LR963083)
TN49	GCF_905132875.1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	99.96	0.31	4.83	61.32	4382	2	4 (LR963108, LR963109, LR963110, LR963111)
TN50	GCF_905132865.1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	99.96	0.31	4.83	61.31	4386	3	4 (LR963095, LR963096, LR963097, LR963098)
TN59	GCF_905132855.1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	99.96	0.31	4.83	61.31	4386	3	4 (LR963085, LR963086, LR963087, LR963088)
TN10	GCF_905132905.1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	99.96	0.31	4.83	61.32	4383	2	4 (LR963100, LR963101, LR963102, LR963103)
TN11	GCF_905132845.1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	99.96	0.31	4.83	61.32	4384	2	4 (LR963090, LR963091, LR963092, LR963093)
TN4	GCF_905132895.1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	99.96	0.31	4.83	61.32	4381	2	4 (LR963123, LR963124, LR963125, LR963126)
TN14	GCF_905132925.1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	99.96	0.31	4.82	61.34	4367	1	2 (LR963105, LR963106)
TN3	GCF_905132935.1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	99.96	0.31	4.83	61.32	4381	3	4 (LR963113, LR963114, LR963115, LR963116)
TN5	GCF_905132915.1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	99.96	0.31	5.83	61.32	4383	3	4 (LR963118, LR963119, LR963120, LR963121)
VA84	GCF_037076415.1	A. dhakensis	Vietnam	Pangasius bocourti	99.96	0.51	4.85	61.55	4275	31	–
T4	GCF_905132965.1	A. hydrophila	Bangladesh	Labeo rohita	98.14	1.19	4.82	61.21	4431	1	–
Aer_Brac14A	GCF_022700855.1	A. hydrophila	Brazil	Dendrocephalus brasiliensis	98.25	0.96	4.76	61.47	4224	28	–
Aer_Brac66	GCF_022700815.1	A. hydrophila	Brazil	Dendrocephalus brasiliensis	98.25	0.96	4.75	61.47	4220	27	–
Brac6	GCF_017310195.1	A. hydrophila	Brazil	Dendrocephalus brasiliensis	98.25	0.96	4.77	61.46	4225	31	–
Aer_LaG33	GCF_017310135.1	A. hydrophila	Brazil	Lophiosilurus alexandri	98.25	0.96	4.75	61.47	4225	28	–
Aer_LaG34	GCF_017310075.1	A. hydrophila	Brazil	Lophiosilurus alexandri	97.93	1.28	4.72	61.54	4224	9	–
Aer_OnP2.2	GCF_017310155.1	A. hydrophila	Brazil	Oreochromis niloticus	98.25	1.28	4.75	61.49	4224	22	–
Aer_OnP4.2	GCF_017310115.1	A. hydrophila	Brazil	Oreochromis niloticus	97.93	1.28	4.75	61.49	4237	24	–
OnP3.1	GCF_017310215.1	A. hydrophila	Brazil	Oreochromis niloticus	98.25	0.96	4.77	61.46	4224	31	–
Aer_Pi25.1HTAS	GCF_022700835.1	A. hydrophila	Brazil	Phractocephalus hemioliopterus	98.58	1.64	4.81	61.23	4302	28	–
516 (Sample ID B3)	GCF_028621795.1	A. hydrophila	China	Aquatic animal	97.71	3.3	5.02	60.79	4472	31	–
516 (Sample ID HG13)	GCF_029320915.1	A. hydrophila	China	Aquatic animal	96.93	3.97	5.02	60.76	4521	31	1 (CP119765)
ZYAH72	GCF_003491225.1	A. hydrophila	China	Carassius carassius	98.82	0.82	5.16	60.68	4618	31	–
NJ-35	GCF_001019645.1	A. hydrophila	China	Carassius carassius	98.19	1.47	5.28	60.51	4776	31	–
JBN2301	GCF_001455365.1	A. hydrophila	China	Carassius carassius	98.81	1.22	5.13	60.41	4640	31	3 (CP013179, CP013180, CP013181)
J-1	GCF_000819505.1	A. hydrophila	China	Carassius carassius	97.55	1.14	5	60.89	4481	31	–
AH10	GCF_000963645.1	A. hydrophila	China	Ctenopharyngodon idella	97.28	0.83	4.91	61.15	4386	33	–
D4	GCF_001518775.1	A. hydrophila	China	Fish	98.33	1.98	5.1	60.46	4770	31	4 (CP013966, CP013967, CP013968, CP013969)
AH230115	GCF_037765365.1	A. hydrophila	China	Hypophthalmichthys nobilis	98.93	1.43	5.27	60.48	4688	31	2 (CP148750, CP148751)
Ah27	GCF_023823085.1	A. hydrophila	China	Ictalurus punctatus	97.81	0.82	5.1	60.76	4576	31	–
HX-3	GCF_009791455.1	A. hydrophila	China	Larimichthys crocea	98.03	1.03	4.94	61	4376	31	–
LHW39	GCF_011602425.1	A. hydrophila	China	Megalobrama amblycephala	98	1.22	5.1	60.8	4583	31	–
MaAh001	GCF_035319865.1	A. hydrophila	China	Monopterus albus	98.38	0.81	4.89	61.2	4318	31	–
GYK1	GCF_001683535.1	A. hydrophila	China	Siniperca chuatsi	98.11	1.14	4.95	60.82	4391	31	–
AC185	GCF_022631175.1	A. hydrophila	South Korea	Anguilla rostrata	99.3	1.78	4.96	61.34	4422	31	–
AC133	GCF_022631195.1	A. hydrophila	South Korea	Carassius carassius	97.51	0.82	5.04	60.9	4492	31	–
LP0103	GCF_022557195.1	A. hydrophila	Taiwan	Hypostomus plecostomus	97.63	1.26	5.02	60.91	4437	31	–
AL06-06	GCF_000940915.1	A. hydrophila	USA	Carassius auratus	98.01	2.34	4.88	61.36	4450	31	3 (CP010948, CP01949, CP01950)
PC-104A	GCF_000635955.1	A. hydrophila	USA	Catfish	99.07	0.82	5.02	60.82	4515	31	–
3019	GCF_018802385.1	A. hydrophila	USA	Fish	96.73	3.94	5.13	60.95	4687	31	–
AL09-71	GCF_000633175.1	A. hydrophila	USA	Ictalurus punctatus	99.11	0.82	5.02	60.82	4511	31	–
ML09-119	GCF_000401555.1	A. hydrophila	USA	Ictalurus punctatus	99.42	0.82	5.02	60.82	4504	31	–
AG-2013-AG1	GCF_029027865.1	A. hydrophila	Vietnam	Catfish	98.51	1.09	4.94	60.93	4440	2	–
TN42	GCF_905132945.1	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	97.98	1.33	4.95	60.95	4478	3	2 (LR963130, LR963131)
TN46	GCF_905132995.1	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	97.98	1.33	4.95	60.95	4472	3	2 (LR963134. LR963133)
TN22	GCF_905132975.1	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	97.98	1.33	4.96	60.93	4480	4	2 (LR963136, LR963137)
TN27	GCF_905132955.1	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	97.98	1.33	4.95	60.97	4471	3	1 (LR963128)
TN1	GCF_905132985.1	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	98.51	1.14	4.94	60.91	4456	3	2 (LR963139, LR963140)

CDS, coding sequence.

The 12 Aeromonas isolates sequenced in this study were recovered from storage on Protect Beads (SLS, UK) at −80 °C and grown on tryptone soya agar (E and O Laboratories Ltd., UK) at 28 °C for 24 h. All bacterial isolates were grown in 10 ml tryptone soya broth (Oxoid, UK) for 18 h at 28 °C, under agitation (140 r.p.m.). High molecular weight DNA was extracted from 1 ml of each bacterial culture using the Wizard^® HMW DNA Extraction kit (Promega, UK) following the manufacturer’s instructions. A minor exception was the addition of 5 µl of RNase A solution to the cell lysate, and DNA was eluted in 85 µl of EB buffer (Qiagen, UK). Before purification, extracted DNA was treated with RNAse A/T1 (20 µl ml⁻¹) (Fisher Scientific, UK) and incubated at 37 °C for 30 min. Following additional RNAse treatment, DNA was further purified using the DNeasy PowerClean Pro kit (QIAGEN, UK) following the manufacturer’s instructions with minor modifications. Briefly, an additional dry spin at 21,000 g for 2 min was performed after washing DNA with Solution CB, and spin columns were incubated at 60 °C for 5 min to remove any residual ethanol. Purified DNA was eluted in 60 µl preheated Solution EB and incubated at room temperature for 5 min. Purified DNA was quantified using the Qubit dsDNA BR Kit (Fisher Scientific, UK) and Qubit machine (Fisher Scientific, UK), before being stored at −20 °C until required.

Whole-genome sequencing, assembly and annotation

The genomes of 12 Aeromonas isolates were sequenced using the NovaSeq 6000 (Illumina) and MinION Mk1B (Oxford Nanopore Technology) platforms at the University of Exeter Sequencing Facility. The Nanopore reads were default filtered, and Illumina reads were adapter and quality trimmed using Fastp [24]. Trimmed reads were uploaded to the Bacterial and Viral Bioinformatics Resource Center (BV-BRC) version 3.35.5 [25], where they were used in hybrid assembly by Unicycler version 0.4.8 [26] with default settings. The genomic sequences were submitted to GenBank and annotated using the Prokaryotic Genome Annotation Pipeline [27]. Genome completeness and contamination were checked by checkM [28]. The genomic sequence and annotation information for isolates investigated in this study are publicly available via their accession numbers (Table 1).

Species identification

The newly sequenced isolates were identified through the KmerFinder 3.2 database [29,31], hosted on the CGE platform (https://cge.food.dtu.dk/services/KmerFinder/). The assembled genomes in FASTA format were uploaded to the KmerFinder tool, and the ‘extended output’ was used to identify isolates, with the species assigned to the template with the highest score considered a positive match with the query genome.

Core- and pan-genome analysis

The Roary pipeline [32] was employed to determine the core and pan-genomes (core plus accessory genome) of A. dhakensis and A. hydrophila isolates, using the GFF3 files generated by prokka 1.14.6 [33]. Phylogenetic trees were constructed from the output core genome alignment file using FastTree version 2.1.10 [34] on the Galaxy server (https://usegalaxy.org/), and pan-genomes were visualized using the roary_plots.py script, employing the absence and presence matrix of genes. A genome-wide association study was performed on the gene_presence_absence.csv file generated by Roary using Scoary [35]. A Fisher’s exact test and Benjamini–Hochberg false discovery rate adjustment were performed in Scoary to score each gene in the pan-genome against its potential correlation with either Aeromonas species.

Genome-based phylogeny

In silico multi-locus sequence type (MLST) analysis was performed using assembled genomes in FASTA format, which were uploaded to the PubMLST database [36]. The ‘Aeromonas’ database was selected, which consists of alleles from six loci: gltA, groL, gyrB, metG, ppsA and recA [37]. Allele sequences and profile data were obtained by aligning loci using default settings on PubMLST. Single nucleotide polymorphisms (SNPs) were called against the core genome (generated with Roary) using snp-sites v2.5.1 [38]. The SNP-based alignment was used to infer a maximum likelihood (ML) phylogeny using RaxML v8.2.12 with the Gamma+GTR rate model [39]. The ML phylogeny was visualized in Figtree v1.4.4 and was mid-rooted.

Accessory genome analysis

Genome assemblies were inspected for putative plasmids by screening contigs (<1 Mb) against a database of plasmid sequences from A. dhakensis or A. hydrophila recovered from Vietnam (six sequences). Contigs that returned no hits were then screened against the full plasmid database available on the NCBI (86,009 sequences). A similarity threshold of >97 % and an E-value of 0 was set to determine closely related plasmids. Plasmid contigs were visually checked for structure using the assembly graphs generated in BV-BRC and Bandage [40]. Further, contigs were screened for plasmid replicon genes using the PlasmidFinder database via ABRicate (https://github.com/tseemann/abricate) and applying a threshold of 60 % sequence coverage and 80 % sequence identity [41]. Features, including GC content and coding sequences (CDSs) of novel plasmids, were mapped using Proksee [42]. To investigate transmission of plasmids across the global Aeromonas genome collection, the ANI was calculated for each plasmid using FastANI (https://github.com/ParBLiSS/FastANI). Here, rows and columns represent each plasmid, and the ANI is calculated based on orthologous mapping of plasmid sequences. A heatmap of ANI distances was generated in RStudio v1.4.1717 using the ‘pheatmap’ package. Insertion sequences (ISs) were detected in assembled genomes using ISFinder (http://www-is.biotoul.fr) [43] and blastn, with an E-value threshold set to 1.00E-06 [17]. Genomic islands (GIs) were identified in assembled genomes and compared using IslandCompare [44]. Antimicrobial resistance genes (ARGs) were identified in chromosome and plasmid contigs by blast analysis against the Comprehensive Antibiotic Resistance Database [45] via ABRicate (https://github.com/tseemann/abricate). Virulence genes were identified by blast analysis against the Virulence Factor Database (VFDB) [46] via ABRicate. The presence of ARGs and virulence genes was determined based on a 60 % sequence coverage and 90 % sequence identity [47]. Protein secretion systems and related appendages were detected in genomes using MacSyFinder v2.1.4 [48] and annotated protein files. Systems were detected using default settings on ordered replicons and the model database TXSScan v 1.1.3.

Statistical analysis

Statistical differences between the means of variables were determined by non-parametric Wilcoxon rank-sum or Kruskal–Wallis tests. A logistic principal component analysis (PCA) was conducted on a binary (0 and 1) matrix of genome components and accessory genes. A two-tailed Fisher’s exact test was performed to determine the statistical significance of distributions between categorical variables. All statistical analysis was performed on JMP Pro software v19.0.0, with significance determined when P<0.05.

Results

Genome characteristics

The general genomic features of 26 A. dhakensis and 43 A. hydrophila isolates from various aquatic hosts in different geographical locations are presented in Table 1. There was considerable overlap in genomic features between the two species investigated, with inter-individual variability also noted (Fig. 1a). Moreover, genome size was positively correlated with the number of CDS regions, whilst GC content was negatively correlated with both genome size and number of CDS regions (Fig. 1). The genomic features of isolates sequenced in this study fell within the expected range for the two species (Fig. 1). Mean genome size was not found to vary significantly between species (t-test, P>0.05), ranging from 4.93±0.03 and 4.96±0.03 Mb in A. dhakensis and A. hydrophila, respectively. Whilst mean GC content was found to be significantly higher in A. dhakensis (61.3±0.05 %) compared with A. hydrophila (61.00±0.04; t-test, P<0.0001), the total number of CDS regions was found to be significantly higher in A. hydrophila (4,448±19) genomes compared with those from A. dhakensis (4,379±25; t-test, P=0.03). All genomes sequenced in this study had a completeness of 100% and 0 % contamination, as revealed by CheckM.

Fig. 1. Genomic features of 26 A. dhakensis and 43 A. hydrophila isolates, recovered from various aquatic hosts and from different geographical locations. A three-dimensional plot of genomic features (genome size, GC content and CDS regions (a). Principal component analysis and canonical plot of genomic features in motile Aeromonas isolates (b). Comparison of genome size, GC content and CDS regions (c). In (a) and (b), dots coloured by species (red, A. dhakensis; blue, A. hydrophila). In Fig. 1c, dots were coloured based on genome assembly origin (green, RefSeq genome assemblies; yellow, genomes sequenced in this study). Significant differences (P < 0.05) in genomic features between species are denoted by *.

Core- and pan-genome analysis

Pan-genome analysis revealed a total of 13,977 genes detected across the 69 Aeromonas genomes investigated (Fig. 2). Moreover, the core genome shared by both A. dhakensis and A. hydrophila represented 19 % (n=2,775 genes) of the total genes detected, whereas the cloud genome represented 58 % (n=8,038 genes) and was the largest component of the global gene pool amongst these two species. Accessory genomes showed distinct clustering based on species. Further, a genome-wide association study (GWAS) with Scoary identified a diverse repertoire of species-specific genomic markers associated with A. dhakensis (n=1,119 genes) and A. hydrophila (n=1,311 genes) (Table S1, available in the online Supplementary Material). Exploration of GWAS data revealed 108 genes to be virulence related and involved in adhesion (n=40), capsule/biofilm (n=1), iron acquisition (n=1), motility (n=15), secretion systems (n=39), toxins (n=9) and ‘other’ (n=3) (Fig. 3). Of these genes, A. dhakensis harboured a greater number of unique genes related to secretion systems (n=26) and toxins (n=7), compared with A. hydrophila (n=13 and 2, respectively). Further, unique A. dhakensis secretion system genes were primarily related to type VI secretion system (T6SS) (e.g. tss-family). In addition, the toxin–antitoxin (TA) system RelBE was also identified to be exclusive to A. dhakensis. In contrast, the thermostable cytotonic enterotoxin ast was found to be unique to A. hydrophila. A. hydrophila (n=32) also harboured a larger number of genes associated with adhesion compared with A. dhakensis (n=8). These genes were primarily related to pilus and fimbrial biogenesis, including tapNM and msh-family, respectively. Species-specific genes related to motility were evenly distributed between bacterial species, although pathways differed, with fliSD and flhF being exclusive to A. dhakensis, whereas the genes flgM and flgJ were unique to A. hydrophila.

Fig. 2. Pan-genome of A. dhakensis (n=26) and A. hydrophila (n=43) genomes from various aquatic sources. The figure shows the gene presence/absence matrix generated from the Roary pan-genome analysis pipeline. The left side illustrates the evolutionary relationship among isolates based on their core genomes. The matrix on the right side shows where genes were either present (dark blue) or absent. p denotes genomes sequenced in this study.

Fig. 3. Gene-level enrichment of virulence factors between A. dhakensis and A. hydrophila from various aquatic sources. The lollipop plot shows virulence genes significantly associated with each species based on Scoary’s genome-wide association analysis. Stick length represents the magnitude of enrichment (based on the log₂ odds ratio), point size corresponds to statistical significance (based on -log₁₀ P values, adjusted using Benjamini–Hochberg method), and point colour indicates the species in which the gene was enriched. Genes are grouped by virulence category.

Further exploration of the pan-genomes at the species level found accessory genome composition of A. dhakensis to be more influenced by geographical location than host origin, confirmed by PCA (Fig. 4a and b). The majority of A. dhakensis isolates from China and Vietnam were found to share similar accessory genomes and clustered to the left of the PCA plot, whereas those recovered from Brazil clustered to the right. The A. hydrophila genomes showed more diversity in their accessory genomes (Fig. 4c and d), although isolates recovered from catfish species (e.g. Ictalurus punctatus and Pangasianodon hypophthalmus) had closely related accessory genomes (Fig. 4d).

Fig. 4. PCA of country and host influence on the accessory genome of A. dhakensis (n=26; a and b) and A. hydrophila (n=43; c and d) isolates recovered from various aquatic sources. The accessory genome of each isolate is coloured according to country (a and c) or host (b and d).

Genome-based phylogeny

The 69 Aeromonas isolates were resolved into 21 sequence types (STs) and 3 novel STs, with a clear distinction between A. dhakensis and A. hydrophila (Fig. 5). Sequence type 656 dominated the A. dhakensis genome collection (65 %; n=17), where it was assigned to isolates from various host species in Vietnam (83 %; n=15) and China (67 %; n=2). In contrast, A. dhakensis isolates recovered from Brazil (ST 812, 2518, 2519), Indonesia (novel ST) and the Philippines (novel ST) were assigned to distinct, country-specific STs. Amongst the A. hydrophila genomes investigated, ST 251 dominated the collection (49 %; n=21) and was assigned to isolates recovered from China, South Korea, Vietnam and the USA. In contrast, isolates recovered from Bangladesh (ST 1826), Brazil (ST 2521, 2522) and Taiwan (ST 2528) were assigned to distinct STs that were country-specific, similar to findings from the A. dhakensis collection. On closer inspection, A. hydrophila isolate AH-40, sequenced in this study, was identified as a novel ST.

Fig. 5. Population structure of 69 Aeromonas isolates recovered from various aquatic sources. ML phylogenetic tree of A. dhakensis (n=26) and A. hydrophila (n=43) genomes based on 403,996 high-quality single-nucleotide polymorphism positions called against the core genome. Species, country, host, lineage and ST for each isolate are provided at the same horizontal position. p denotes genomes sequenced in this study.

A total of 403,996 SNPs were detected across the core genome of the 69 Aeromonas isolates investigated. An ML tree based on the core genome SNPs is presented in Fig. 5. The core genome SNP phylogeny revealed a clear distinction between A. dhakensis and A. hydrophila isolates with five larger clades, consistent with in silico MLST results. A. dhakensis was dominated by a large clade, represented by isolates recovered from various host species in China and Vietnam, assigned to ST 656, and lineages A2 and A3. Three large clades were detected across A. hydrophila genomes. The largest clade was represented by isolates from various host species in Asia (China, Taiwan and Vietnam) and assigned to a single lineage (A6) and ST 251. The remaining ST 251 isolates were resolved into a closely related clade assigned to lineage A7, which also included the ST 2524 isolate JBN2301. This clade was found to be more globally distributed, represented by isolates from China as well as the USA. The third clade was revealed to be Brazil-specific, with isolates assigned to the lineage A10 and ST 2521.

Mobilome

When all genomes were explored, 48 % (n=33) of isolates carried at least one plasmid type in their genome, whilst 39 % (n=27) of isolates carried at least two plasmid types. A significantly higher prevalence of isolates carrying at least one plasmid type was found for A. dhakensis (69 %; n=18) compared with A. hydrophila (35 %; n=15) (Fisher’s exact, two-tailed; P<0.01). Further, when geographical origin was explored, a significant association was detected between sampling country and plasmid-carrying isolates (Fisher’s exact, two-tailed; P<0.0001). Correspondence analysis showed a clear separation of Vietnam toward the plasmid-positive axis, where 90 % (n=26) of isolates from Vietnam carried at least one plasmid type. In contrast, all other countries clustered near the plasmid-negative axis, consistent with substantially lower plasmid prevalence (≤30 %).

The genomes of A. dhakensis sequenced in this study were found to carry between 0 and 4 plasmids (Table 2). No plasmids were detected in isolates AD-69 and AD-92HK1, recovered from the Philippines and Indonesia, respectively. Most A. dhakensis plasmids sequenced in this study were uniform in size at 4.7 Kb, 6.2 Kb, 6.3 Kb and 6.4 Kb and had high similarity (100 % sequence coverage and >99 % sequence identity) to the A. dhakensis plasmids 4.7-HicB, 6.2-Stb, 6.3-Rel and 6.4-Qnr. When compared with the global isolate collection, plasmids 4.7-HicB, 6.3-Rel and 6.4-Qnr were exclusive to A. dhakensis in Vietnam from both striped catfish (Pangasianodon hypophthalmus) and striped snakehead (Channa striata), where they shared >98 % ANI (Fig. 6). Whilst the plasmid 6.2-Stb was more frequently detected in A. dhakensis from Vietnam, a closely related (>98 % ANI) 6.2 Kb plasmid was also found in the genomes of two A. hydrophila isolates recovered from China. Isolate AD-9 harboured a novel, larger plasmid at 79.9 Kb, which comprised 78 CDS regions, including several genes from the type II TA system and a type IV secretion system (T4SS) (Fig. 7a). Between one and two plasmids were detected in the genomes of A. hydrophila sequenced in this study. Two plasmids were found to be identical (100 % sequence coverage and sequence identity) to other A. hydrophila plasmids (3.6-parDE and 5.0-YhdJ plasmids) previously detected and associated with striped catfish in Vietnam. Analysis by ANI revealed plasmid 3.6-parDE to be exclusive to A. hydrophila from striped catfish in Vietnam, where plasmids shared >98 % similarity, although a closely related (>91 % similarity) 6 Kb plasmid was also detected in A. hydrophila from non-catfish hosts in China. Isolate AH-12 harboured a 4.0 Kb plasmid, which was closely related (100 % sequence coverage and 99 % sequence identity) to a plasmid in Edwardsiella ictaluri from the USA and Vietnam. This was confirmed by ANI, which found the plasmid to be distinct (0 % similarity) from any other Aeromonas plasmid investigated. Further, a novel 9.2 Kb plasmid was detected in isolate AH-40, which encoded 12 CDS regions including several type II TA system and tetracycline resistance genes (Fig. 7b).

Table 2. blastn results of plasmids detected in A. dhakensis and A. hydrophila genomes sequenced in this study.

Isolate	Species	Country	Host	Plasmid	Size (Kb)	Match	Species	Accession	E-value	Coverage	Match
AD-92HK1	A. dhakensis	Indonesia	Oreochromis niloticus	None detected	–	–	–	–	–	–	–
AD-69	A. dhakensis	The Philippines	Oreochromis niloticus	None detected	–	–	–	–	–	–	–
AD-VN1	A. dhakensis	Vietnam	Channa striata	pAD-VN1-6.4	6.4	6.4-Qnr	A. dhakensis	NZ_LR963083.1	0	100%	100%
				pAD-VN1-6.3	6.3	6.3-Rel	A. dhakensis	NZ_LR963080.1	0	100%	100%
				pAD-VN1-6.4	6.2	6.2-Stb	A. dhakensis	NZ_LR963081.1	0	100%	100%
				pAD-VN1-4.7	4.7	4.7-HicB	A. dhakensis	NZ_LR963082.1	0	100%	100%
AD-VN2	A. dhakensis	Vietnam	Channa striata	pAD-VN2-6.4	6.4	6.4-Qnr	A. dhakensis	NZ_LR963083.1	0	100%	100%
				pAD-VN2-6.3	6.3	6.3-Rel	A. dhakensis	NZ_LR963080.1	0	100%	100%
				pAD-VN2-6.2	6.2	6.2-Stb	A. dhakensis	NZ_LR963081.1	0	100%	100%
				pAD-VN2-4.7	4.7	4.7-HicB	A. dhakensis	NZ_LR963082.1	0	100%	100%
AD-27	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	pAD-27-6.4	6.4	6.4-Qnr	A. dhakensis	NZ_LR963083.1	0	100%	100%
				pAD-27-6.3	6.3	6.3-Rel	A. dhakensis	NZ_LR963080.1	0	100%	99%
				pAD-27-6.2	6.2	6.2-Stb	A. dhakensis	NZ_LR963081.1	0	100%	100%
				pAD-27-4.7	4.7	4.7-HicB	A. dhakensis	NZ_LR963082.1	0	100%	100%
AD-9	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	pAD-9-79.9	79.9	Novel	–	–	–	–	–
				pAD-9-6.4	6.4	6.4-Qnr	A. dhakensis	NZ_LR963083.1	0	100%	100%
				pAD-9-6.2	6.2	6.2-Stb	A. dhakensis	NZ_LR963081.1	0	100%	100%
				pAD-9-4.7	4.7	4.7-HicB	A. dhakensis	NZ_LR963082.1	0	100%	100%
AD-1	A. dhakensis	Vietnam	Pangasianodon hypophthalmus	pAD-1-6.4	6.4	6.4-Qnr	A. dhakensis	NZ_LR963083.1	0	100%	100%
				pAD-1-6.3	6.3	6.3-Rel	A. dhakensis	NZ_LR963080.1	0	100%	99%
				pAD-1-6.2	6.2	6.2-Stb	A. dhakensis	NZ_LR963081.1	0	100%	100%
				pAD-1-4.7	4.7	4.7-HicB	A. dhakensis	NZ_LR963082.1	0	100%	99%
AH-40	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	pAH-40-9.2	9.2	Novel	–	–	–	–	–
AH-34	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	pAH-34-5.6	5.6	5.0-YhdJ	A. hydrophila	NZ_LR963130.1	0	100%	100%
				pAH-34-3.7	3.7	3.6-parDE	A. hydrophila	NZ_LR963134.1	0	100%	100%
AH-32	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	pAH-32-5.6	5.6	5.0-YhdJ	A. hydrophila	NZ_LR963130.1	0	100%	100%
				pAH-32-3.7	3.7	3.6-parDE	A. hydrophila	NZ_LR963134.1	0	100%	100%
AH-12	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	pAH-12-4.0	4	19-pEI1	Edwardsiella ictaluri	NC_020281.1	0	100%	99%
				pAH-12-3.7	3.7	3.6-parDE	A. hydrophila	NZ_LR963134.1	0	100%	100%
AH-7	A. hydrophila	Vietnam	Pangasianodon hypophthalmus	pAH-7-3.7	3.7	3.6-parDE	A. hydrophila	NZ_LR963134.1	0	100%	100%

Fig. 6. A dendrogram of average nucleotide identity for plasmid sequences (n=94) detected in A. dhakensis and A. hydrophila genomes. The heatmap shows a similarity matrix, which is coloured according to percentage nucleotide identity. On the left side, dendrogram nodes are coloured according to country, host and bacterial species origin for distinction (see legend).

Fig. 7. Map of novel plasmids pAD-9-79.9 (a) and pAH-40-9.2 (b) detected by whole-genome sequencing in this study. CDS regions, GC content and skew visualized using Proksee. Genomic features highlighted include type II TA systems, T4SS and tetracycline resistance gene Tet(E).

The number of ISs per genome in A. dhakensis (57±5 IS per genome) was significantly lower than in A. hydrophila (80±6 IS per genome) (Wilcoxon two-sample, P<0.001) (Table S2). When country-level association was further explored for A. dhakensis, isolates from China were found on average to carry a greater number of ISs (69±21 IS per genome), compared with isolates from other countries (8–63 IS per genome), although no significant difference was found (Kruskal–Wallis; X²=7.4, DF=4, P>0.05). The ISs detected in A. dhakensis genomes fell into 18 IS families, which were all detected in the chromosome. Further, seven of the IS families were also detected on various plasmids, including IS110, IS1202, IS3, IS481, IS5, IS66 and Tn3. The IS family Tn3 was conserved across all A. dhakensis genomes. Further, the families IS66, IS481 and TN3 were globally distributed and detected in at least one genome from each country investigated. Several IS families were also found to be country- or host-specific (Fig. 8a and b). The IS family ISL3 was only detected in genomes associated with China, whereas IS30, IS1202 and ISCR were exclusive to genomes from Vietnam, and IS1202 was exclusive to striped catfish-associated isolates only. Insertion sequences were detected in every A. hydrophila genome, with the number detected per genome ranging from 11 to 122 (Table S2). At the country level, a Kruskal–Wallis test showed significant differences in the average number of IS per genome between sampling countries (X²=27.15, DF=6, P=0.0001). Further, Dunn’s post hoc analysis revealed isolates from Vietnam (105±6 IS per genome) and China (98±3.7 IS per genome) to carry a significantly higher number of ISs average per genome compared with those from Brazil (14 IS per genome), whilst all other pairwise comparisons were not significant. A total of 21 IS families were detected across the A. hydrophila genomes. All families were detected in the chromosome of isolates; however, the families IS21 and Tn3 were also detected in the plasmids of some isolates. The IS families IS3, IS66 and Tn3 were conserved across A. hydrophila genomes. Three IS families (IS3, IS66 and Tn3) were globally distributed across all countries and associated with all host species investigated. Further exploration also revealed several IS families to be country- or host-specific (Fig. 8c, d). For example, IS1182 was exclusive to AC185, recovered from American eel (Anguilla rostrata) in South Korea, ISL3 was detected in isolate 3019 recovered from the USA and the family ISNCY was found only in the genomes of isolates recovered from striped catfish in Vietnam.

Fig. 8. Chord diagram showing the distribution of IS families across the genomes of A. dhakensis (n=26; a and b) and A. hydrophila (n=43; c and d) recovered from various aquatic sources. For each species, chord diagrams illustrate the relationships between IS families and genome origin. Panels (a) and (c) show IS–country associations, and panels (b) and (d) show IS–host associations. Ribbon width is proportional to the number of genomes carrying a given IS family. Colours denote either country or host origin.

Exploration of GIs in the genome collection revealed A. dhakensis and A. hydrophila genomes to carry a similar number of GIs (15±1 and 17±1 GIs per genome, respectively) (Wilcoxon two-sample, P>0.05) (Tables S3 and S4). A Kruskal–Wallis test found significant differences between the number of GIs per genome and origin country, irrespective of bacterial species (X²=32, DF=8, P<0.0001). Comparatively, isolates from China (19±1 GIs per genome), USA (19±3 GIs per genome) and Vietnam (17±1 GIs per genome) carried a significantly higher number of GIs per genome compared with isolates from Brazil (11±1 GIs per genome) (Dunn’s test; P<0.002) average number of GIs per genome, irrespective of bacterial species. Comparative analysis of GIs in A. dhakensis and A. hydrophila genomes revealed isolates to carry various types of GIs, including virulence and resistance islands. Virulence islands encoding type I–IV and VI secretion systems, as well as type II and IV TA systems, brkB, flagellar, haemolysins, pili, V antigen, VasX and vertebrate lysozyme inhibitor, were frequently detected in isolate genomes, irrespective of bacterial species. However, country- and host-specific differences were also noted in the resistance and virulence islands carried by A. dhakensis or A. hydrophila. Isolates AD-69 and AD-92HK1 from the Philippines and Indonesia, respectively, carried 15 and 8 unique GIs, respectively, although these were predominantly metabolic islands and none encoded for resistance. In contrast, one GI (average size 21,057 bp) detected in 94 % (n=17) of A. dhakensis genomes from Vietnam encoded resistance to potentiated sulphonamides (dfrA1 and sul1) and tetracyclines (tet(A)) (Fig. S1). Further, when host origin was explored, the isolates AD-VN1 and AD-VN2, recovered from striped snakehead, were found to carry a unique hybrid GI (average size 21,251 bp) encoding for DNA repair (radC) and the virulence factor irmA. In A. hydrophila, a resistance island detected in multiple genomes from China encoded genes against aminoglycosides (AAC(6′)-Ib-cr5), sulphonamides (sul1) and antiseptics (qacE Delta 1) (Fig. S2). Likewise, isolate AC185, recovered from South Korea, carried two islands encoding resistance to aminoglycosides (aph(3′)-lb), florfenicols (floR), potentiated sulphonamides (dfrA12 and sul1) and tetracyclines (tet(D)) (Fig. S3). Isolate AH-40 carried a unique virulence island encoding a T6SS (e.g. hcp1, tssB and vgrG) not detected in any other isolate from Vietnam (Fig. S4). Further, a GI (average size 30,697 bp) detected in eight (73 %) A. hydrophila genomes from striped catfish in Vietnam encoded resistance to potentiated sulphonamides (dfrA1 and sul1) and tetracyclines (tet(A)).

Resistome

A total of 39 different ARGs were detected across the 69 isolates (Table S5). A mean of 5±3 ARGs was detected per genome, although A. dhakensis isolates (7±3 genes per genome) carried a significantly higher number of ARGs per genome compared with A. hydrophila (4±3 genes per genome) (Wilcoxon two-sample; P<0.0001). Whilst all four resistance mechanisms were found in both Aeromonas species, ARGs encoding for efflux, target protection and target replacement were more abundant in A. dhakensis genomes compared with A. hydrophila (Fig. 9a) (Fisher’s exact test, two-tailed; P<0.0001). Although A. hydrophila displayed a higher relative proportion of drug-inactivation ARGs, Fisher’s exact test showed no significant difference in genome-level prevalence between both species (P>0.05). When ARGs were aggregated by drug class, ARGs conferring resistance to phenicol, quinolone, sulphonamide, tetracycline and trimethoprim drug classes were more abundant in A. dhakensis compared with A. hydrophila (Fisher’s exact test, two-tailed; P<0.05). In contrast, ARGs encoding resistance to aminoglycosides were more frequently detected in A. hydrophila (Fisher’s exact test, two-tailed; P<0.001). Further, resistance to glycopeptide and rifamycin was exclusive to A. dhakensis and A. hydrophila, respectively. The beta-lactamase gene OXA-724 was conserved across all genomes, irrespective of bacterial species. However, other beta-lactamase genes were identified to be species-specific. Specifically, A. dhakensis carried the families AQU and NDM, whereas A. hydrophila carried CepS (Fisher’s exact test, two-tailed; P<0.05). Other species-specific gene families included AAC (A. hydrophila; aminoglycoside), BRP (A. dhakensis; glycopeptide) and arr (A. hydrophila; rifamycins). Whilst not unique, the gene families dfrA, qnr, sul and tet, encoding resistance to trimethoprim, quinolones, sulphonamides and tetracyclines, respectively, were more abundant in A. dhakensis genomes compared with those belonging to A. hydrophila (Fisher’s exact test, two-tailed; P<0.001).

Fig. 9. Resistome profiles of A. dhakensis (n=26) and A. hydrophila (n=43) genomes recovered from various aquatic sources. Plots illustrate the relative abundance of antimicrobial resistance genes across bacterial species (a), isolation country (b) and host origin (c). Stacked bars summarize the proportion of detected genes aggregated into resistance mechanisms, drug classes and gene families.

Since different ARGs might be associated with genome origin, we also explored the influence of country and host on ARG presence and resistome profiles. When country-level data were explored, a Kruskal–Wallis test found a significant difference between the number of ARGs per genome and sampling country for A. dhakensis only (X²=17.22, DF=4, P<0.01), whilst no significant difference was observed for A. hydrophila (X²=10.71, DF=6, P>0.05). When ARGs were aggregated by resistance mechanism, the genomes from China and Vietnam showed the greatest diversity with all four resistance mechanisms detected, followed by South Korea (n=3) and USA (n=3) (Fig. 9b). Further, all four resistance mechanisms were detected at significantly higher frequencies in Vietnam compared with other countries (Fisher’s exact test, two-tailed; P<0.001). When ARGs were aggregated by drug class, ARGs encoding resistance to beta-lactam antibiotics were the most abundant, irrespective of sampling country. However, country-specific differences were also noted for several drug classes, including resistance to aminoglycosides, which was significantly associated with South Korean genomes and associated with the APH gene family (Fisher's exact test; P<0.05). Likewise, a significantly higher proportion of genomes from Vietnam carried the gene families dfrA, qnr, sul and tet, which confer resistance to quinolone, sulphonamide, tetracycline and trimethoprim classes (Fisher’s exact test, two-tailed; P<0.01). Several gene families were country-specific, including BRP (glycopeptide) and NMD (beta-lactam), which were exclusive to genomes from Vietnam. Likewise, ESBL (beta-lactams) and MPH (macrolide) ARGs were only found in genomes from South Korea. The arr family (rifampicin) was exclusive to Asia, where it was detected in genomes from China, South Korea and Vietnam, but absent in those from Brazil and the USA.

At the host level, drug inactivation was the most prevalent resistance mechanism detected across genomes (Fig. 9c). However, all four resistance mechanisms were detected in the bacterial genomes associated with three host species: Mastacembelus, striped catfish and striped snakehead. When ARGs were aggregated by drug class, ARGs encoding resistance to beta-lactam antibiotics were the most abundant irrespective of host origin. Further, the genomes from 58 % (n=15) of host species carried ARGs encoding resistance to beta-lactam antibiotics only. Macrolide resistance was exclusive to the genome of AC185 from the American eel, encoded by MPH genes. Further, when ARGs were aggregated by gene family, the genome of AC185 also displayed the greatest resistome diversity with ARGs assigned to 14 families. This was followed by Mastacembelus (n=13) and striped catfish (n=10).

Virulome

A total of 194 different virulence genes were detected across the global Aeromonas genome collection (Table S6), which comprised adherence (n=54; 28%), effector delivery systems (n=77; 40%), exotoxins (n=7; 4%) and motility (n=56; 29%) associated genes. The number of genes per genome ranged from 103 to 158, although on average, A. hydrophila genomes carried a significantly higher number of genes (n=142±14 genes) compared with A. dhakensis (n=111±2 genes) (Wilcoxon two-sample test; P<0.0001). Virulence genes were grouped into 32 gene families, and the distribution of these families across genomes is presented in Fig. 10. Sixteen gene families were conserved across the genome collection, irrespective of bacterial species. The major virulence factors identified in all genomes were associated with MSHA type IV pili (e.g. msh), polar flagella (e.g. che, flg, flh, fli, flm, flr, mot, nue, parA and pom), Tap type IV pili (e.g. tpp and tap), type I pili (e.g. fim), type II secretion system (T2SS, e.g. exe) and T6SS (e.g. hcp). Whilst no genes were unique to A. dhakensis genomes, the gene families fla and flp associated with polar flagella and MSHA type IV pili, respectively, were exclusive to A. hydrophila genomes (Fisher’s exact test, two-tailed; P<0.0001).

Fig. 10. A heatmap illustrating the distribution of virulence gene families across A. dhakensis (n=26) and A. hydrophila (n=43) genomes from various aquatic sources. Rows represent individual genomes and columns show the presence (black) or absence (white) of 32 gene families. Row annotations correspond to genome origin (bacterial species, country and host). Column annotations indicate virulence factor (VF) classification and are coloured based on general virulence category, including adherence (blue), effector delivery system (green), exotoxin (red) and motility (yellow). p denotes genomes sequenced in this study.

As shown in Fig. 10, the genomes sequenced in this study tended to cluster together with genomes from similar geographic or host origins, broadly displaying similar virulence profiles. However, whilst origin-associated similarities in the virulome of isolates were evident, the profiles often overlapped across distinct groups. Notably, one of the newly sequenced genomes, AH-40, deviated from the general trend observed for A. hydrophila in Vietnam, supporting trends observed in phylogeny previously (Fig. 5). Instead, AH-40 displayed a similar virulence profile to isolate MaAh001, recovered from Asian swamp eel (Monopterus albus) in China. Specifically, both AH-40 and MaAh001 differed from other A. hydrophila from similar regions by lacking fla genes. When genome origin was further explored, several genes were found to be country and lineage-specific. Indeed, the flp gene associated with type IV pili was significantly associated with ST 251 (Fisher’s exact test, two-tailed; P<0.0001), specifically genomes from China, South Korea, USA and Vietnam (Fisher’s exact test, two-tailed; P<0.05). Likewise, significant lineage and country associations (Fisher’s exact test, two-tailed; P<0.0001) were also observed in the prevalence of the exotoxin aerA/act, where it was conserved in ST 251 and ST 656, whereas the prevalence of aerA/act was significantly lower in Brazil compared with other countries. At the host level, aerA/act was conserved in isolates recovered from channel catfish (Ictalurus punctatus), crucian carp (Carassius carassius), striped catfish and striped snakehead, but had significantly lower prevalence in genomes associated with fairy shrimp (Dendrocephalus brasiliensis) and Nile tilapia (Oreochromis niloticus) (Fisher’s exact test, two-tailed; P<0.0001). Genes associated with a second exotoxin, rtxA, were also detected in 90 % (n=67) of genomes; however, differences were noted in operon completeness (Table S6). Specifically, a complete exotoxin rtxA operon (rtxA/B/C/D/E/H) was significantly associated with Brazil, where it was conserved in ST 2521 (Fisher’s exact test, two-tailed; P<0.0001). A complete rtxA operon was also detected in other countries, albeit significantly lower in prevalence (Fisher’s exact test, two-tailed; P<0.0001). This included Vietnam, where it was detected in three A. dhakensis isolates (VA6, VA54, VA84), as well as several A. hydrophila isolates in China (21 %; n=3) and South Korea (50 %; n=1) carried the full rtx operon. Finally, several type III secretion system (T3SS) genes (e.g. acr, aop, ati and exs) were exclusive to two genomes recovered from China.

Macromolecular systems

A total of 11 macromolecular systems were detected across the genome collection, albeit at varied levels of completeness (Fig. 11, Table S7). Three complete systems were conserved across genomes, including the flagellum, the type I secretion system and the type II secretion system. A complete mannose-sensitive haemagglutinin (MSH) pilus system was also detected in 99 % (n=68) of genomes investigated. However, only a partial MSH system was detected in the genome of AH-32, sequenced in this study. Whilst VFDB screening detected all mandatory msh genes, including mshA in AH-32 (Table S7), MacSysFinder failed to recognize mshA and thus a complete MSH system in this isolate. All genomes carried core elements of type IVa pilus (T4aP), although they all lacked the mandatory gene pilA and thus only a partial T4aP system was conserved. A complete T3SS and T6SS were detected in 12 % (n=8) and 87 % (n=60) of genomes investigated. Further, Fisher’s exact tests found the prevalence of these systems to be significantly associated with country (two-tailed; P<0.01). Specifically, complete T3SS were more frequently detected in genomes recovered from Bangladesh, China, Indonesia and the USA. In contrast, the prevalence of complete T6SS was significantly lower in Bangladesh and the USA. The Tad (tight adherence) pilus was exclusive to A. hydrophila, where it was detected in 56 % (n=24) of genomes. Further, Tad showed significant geographic (Fisher’s exact test, two-tailed; P<0.05) and lineage (Fisher’s exact test, two-tailed; P<0.0001) associations, being significantly prevalent in China, South Korea, the USA and Vietnam, as well as being conserved across all isolates of ST 251. Complete pT5SSt and type V secretion system (T5SS) were rare, detected in 2 % (n=1), 3 % (n=2) and 4 % (n=3) of isolates, respectively. Whilst pT4SSt was exclusive to China, the T5aSS and T5bSS systems were shared across countries and hosts. The genome of AD-69, sequenced in this study, was unique in carrying complete sets of both T5SS.

Fig. 11. A heatmap illustrating the distribution of macromolecular systems in A. dhakensis (n=26) and A. hydrophila (n=43) genomes from various aquatic sources. Rows represent individual genomes, and columns show the presence or absence of 11 systems. Row annotations correspond to genome origin (bacterial species, country and host). Cells are coloured based on system completeness: complete (required number of mandatory genes detected: black), partial (≥ 1 but less than min. required mandatory genes detected: grey) and absent (no mandatory gene detected: white). Systems (n=min. required mandatory genes) include flagellum (n=11), MSH pilus (n=5), type I secretion system (T1SS; n=3), type II secretion system (T2SS; n=4), type III secretion system (T3SS; n=9), type IVa pilus (T4aP; n=6), type IV secretion system (pT4SSt; n=1), type Va secretion system (T5aSS; n=1), type Vb secretion system (T5bSS; n=1), type VI secretion system (T6SSi; n=14) and tight adherence (Tad) pilus (n= 7). p denotes genomes sequenced in this study.

Discussion

Analysis of the pan-genome in global Aeromonas isolates revealed large genomic variability with only a small set of core genes (19%) shared between all Aeromonas genomes investigated (Fig. 2). These findings agree with those reported previously for Aeromonas [49]. Whilst we recognize the cited study was based on a broader multi-species dataset, the large genomic variability observed in both studies highlights the ability of this genus to adapt and occupy diverse ecological niches, with our analysis specifically reflecting A. dhakensis and A. hydrophila. Indeed, Aeromonas can be found in various aquatic environments, as well as food (fruits, vegetables, dairy and meat) and terrestrial animals [1]. In the work presented, the presence of closely related pan-genomes in isolates from similar geographical regions and fish species (Figs2 4) suggests a divergent evolution and local adaptation of these two species, with pan-genomes shaped more by the immediate environment. Environmental niche and local geography have been shown to shape the pan-genomes of other bacterial pathogens, including E. ictaluri [23], Escherichia coli [50] and Salmonella enterica [51], and highlight the importance of local disease management strategies in controlling the spread of MAS.

Genomic comparisons of circulating Aeromonas populations revealed highly divergent clusters, dominated by ST 656 and ST 251 for A. dhakensis and A. hydrophila, respectively (Fig. 5). The ST 656 dominates the A. dhakensis recovered from the striped catfish sector in Vietnam [52]; however, findings from this study suggest that ST 656 has a wider regional distribution within Asia, with this ST also detected in China. This study also detected novel STs in the newly sequenced genomes of the A. dhakensis isolates AD-92HK1 and AD-69 from Indonesia and the Philippines, which were distinct from the STs of other isolates recovered from Nile tilapia (Oreochromis niloticus). Indeed, all four A. dhakensis genomes associated with Nile tilapia differed in every allele examined, highlighting a high level of genetic divergence within A. dhakensis, irrespective of host, and would suggest that AD-92HK1 and AD-69 may have undergone adaptation to specific environmental conditions within their respective countries; however, additional strains would be beneficial to confirm. The presence of ST 251 in China, South Korea, the USA and Vietnam suggests a global distribution of clonal A. hydrophila populations. Previous reports from Vietnam revealed the ST 251 to dominate A. hydrophila populations causing disease in striped catfish farms [15,52]. Whilst our findings would agree in principle, in our study, isolated AH-40 was confirmed as belonging to a novel ST. Further analysis of the allelic profile of AH-40 revealed this isolate to be distinct from all other isolates recovered from striped catfish in Vietnam, differing in every allele examined. A study by [53] also identified novel STs in isolates recovered from diseased catfish in Vietnam, although these isolates differed from AH-40 in the alleles of gyrB, groL, metG and ppsA. These findings would suggest that A. hydrophila populations circulating the Vietnamese striped catfish sector are more divergent than previously assumed and warrant increased microbial surveillance studies to better understand the prevalence of these novel STs associated with disease.

The influence of local geography in shaping the population dynamics of Aeromonas isolates was also evident in the different mobilome elements found across genomes. Indeed, ANI revealed several country- or host-specific plasmids associated with numerous genomes (Fig. 6). Likewise, statistical analysis revealed that isolates from Vietnam were significantly more likely to harbour plasmids, compared with genomes from other countries, such as Indonesia and the Philippines, where genomes lacked any plasmids. The number and identity of plasmids detected in A. dhakensis and A. hydrophila from Vietnam agree with previous reports [52]. However, the presence of a novel plasmid in isolate AD-9 (Fig. 7a) and an E. ictaluri plasmid in isolate AH-12 (Table 2) also reflects unexpected diversity in the plasmid reservoir of circulating Aeromonas populations within Vietnam. Together, these findings would suggest the potential for large-scale horizontal gene transfer within these systems, which could drive future changes in AMR and pathogenicity.

A closer inspection of the resistome of A. dhakensis and A. hydrophila revealed a high prevalence of antibiotic inactivation genes in these genomes, supporting our existing understanding of common AMR mechanisms in Gram-negative bacteria [54]. The gene OXA-724, which encodes a class D β-lactamase, was conserved across all genomes investigated (Table S5), in agreement with previous reports [52,55]. However, further analysis showed country-specific resistome profiles with varying degrees of predicted resistance levels. Specifically, the newly sequenced A. dhakensis isolates from Indonesia and the Philippines, as well as A. hydrophila from Bangladesh and Taiwan, carried three ARGs in their genomes, encoding resistance to beta-lactam compounds only (Fig. 9b). These findings also support the phylogenetic clustering and distinct ST assignment found for these isolates (Fig. 5), indicating that AMR drivers are likely clade-associated. In contrast, isolates recovered from other regions in Asia demonstrated the potential for multidrug resistance (MDR), carrying genes encoding resistance to at least three antibiotic classes [56]. In Vietnam, most isolates (86 %; n=25) carried ARGs against more than five antibiotic classes, including aminoglycoside, fluoroquinolone, potentiated sulphonamides (inc. diaminopyrimidine) and tetracycline (Fig. 9b). Likewise, several genomes associated with China (12 %; n=2) and South Korea (50 %; n=1) also presented MDR, encoding ARGs against aminoglycosides, potentiated sulphonamides and tetracyclines (Fig. 9b). These findings support existing reports across Asia of MDR development in bacterial pathogen populations associated with aquaculture, where these antibiotic classes are frequently used [15,57,60]. This contrasted with trends observed in Brazil and the USA, where low or no incidence of MDR was observed in isolates recovered from these countries (Fig. 9b). We hypothesize that these differences in MDR detected across regions reflect differences in antimicrobial stewardship and disease management practices. Whilst these will undoubtedly play a role, the factors and sources influencing AMR leading to the emergence of MDR isolates within aquatic systems, including fish farms, are complex. Surveillance and open data on antimicrobial usage are crucial to disentangle this interplay and better understand AMR drivers across the global aquaculture sector.

A. dhakensis and A. hydrophila carry an extremely diverse repertoire of genes within their virulome [61]. In this study, effector delivery systems were found to dominate the virulome of Aeromonas species investigated (Fig. 10). As these systems allow for efficient and targeted manipulation of the host [62], dominance of this pathway in Aeromonas from aquatic environments would be particularly advantageous where hosts may be more diverse or readily accessible to the bacterium. In addition, A. hydrophila was found on average to carry a significantly higher number of virulence genes within its genome compared with A. dhakensis. Further analysis also revealed fla and flp-related virulence factors to be exclusive to A. hydrophila. Whilst little is known about the role of flp pili in the virulence of motile Aeromonas [63], the flp pili facilitate host invasion and defence in other aquatic bacterial pathogens [64]. Likewise, the lateral flagella encoded by fla genes serve important functions for Aeromonas in adhesion, motility and biofilm formation [65]. In addition, findings from GWAS of the Aeromonas pan-genome in this study also identified the heat-stable cytotonic enterotoxin ast, which is involved in the pathophysiology of MAS clinical signs [57], to be specific to A. hydrophila. Together, these findings would, therefore, suggest a greater virulence potential for A. hydrophila compared with A. dhakensis, contradicting anecdotal evidence of A. dhakensis being more prevalent in MAS outbreaks in different aquatic hosts [15,52, 53, 66, 67]. However, these findings do support other studies reporting a higher prevalence of A. hydrophila associated with MAS in other fish species [68,69]. In addition, findings also demonstrated that the Aeromonas virulome was shaped by host and geographical pressures, echoing the phylogenetic topology. For example, whilst ST 2521 isolates from Brazil lacked the exotoxin aerA/act, they all carried a complete rtx operon in their genome (Table S7). In contrast, whilst aerA/act was conserved in dominant STs (ST 251 and ST 656) circulating in China, the USA and Vietnam (Fig. 10), the presence of a complete rtx operon was rare in these countries. Whilst both systems contribute to the clinical signs observed in MAS [61], these genes have distinct functions in the bacterium’s pathogenesis. Together, the findings from this study, therefore, suggest potential clade-specific virulence repertoires and reflect the complex interplay between environment, host susceptibility and pathogen virulence. Given the diverse host range of both A. dhakensis and A. hydrophila, the virulome will undoubtedly be influenced by differences in host defence mechanisms and farming conditions. Surveillance and challenge studies are, therefore, required to better understand these interactions and identify drivers behind the divergence in the virulome.

Macromolecular secretion systems are vital to the normal functioning of the bacterial cell, as well as promoting pathogenesis in bacterial pathogens through the transport of small molecules, nucleic acids and proteins, e.g. toxins [62,70]. In this study, we detected gene clusters associated with a variety of systems, including flagella, Tad pili and type secretion systems (I–VI) (Fig. 11). Whilst flagella genes were conserved across all Aeromonas genomes investigated, genes associated with Tad pili were exclusive to A. hydrophila genomes and conserved in ST 251, suggesting that the presence of this system is largely lineage-dependent. The Tad pili are an Archaea-derived pili, acquired through horizontal gene transfer, and facilitate a variety of processes, including motility, adherence and DNA uptake in the host bacterium [71]. In A. hydrophila, Tad has been shown to function in biofilm formation and virulence in channel catfish [72], thus highlighting the importance of this gene cluster in the pathogenicity of aquatic isolates affecting different farmed fish species. Local geography was also found to shape macromolecular systems in Aeromonas genomes, as the presence of a complete T3SS was isolated to a few isolates (12 %; n=8) across Bangladesh, China, Indonesia and the USA, yet all isolates from Vietnam, irrespective of bacterial species, lacked this machinery. In other bacteria, the T3SS assembles into a needle-like structure on the surface of the bacterial cell and allows the transfer of effector proteins into the target cell and allows host cell colonization [73]. The absence of such a vital secretion system in some genomes would, therefore, suggest a reliance on alternative virulence pathways in these isolates. Likewise, isolate AD-69 was unique in carrying both T5aSS and T5bSS (Fig. 11). The low prevalence of T5SS across the genomes investigated would, therefore, suggest this secretion system is of limited importance in the virulence of A. dhakensis or A. hydrophila, supporting previous reports in other Aeromonas species [19].

In conclusion, comparative genomic analysis in this study has contributed to a better understanding of A. dhakensis and A. hydrophila populations circulating in the global aquatic environment. Findings from this study suggest that whilst A. dhakensis comprises distinct clades, A. hydrophila is largely dominated by a single clade that is found across multiple continents and host species. However, local geography and host species exert significant influence on the genotypes of both A. dhakensis and A. hydrophila, likely driven by multiple factors, including anthropogenic pressure, environmental conditions within production systems and host genotypes. As evident by the high prevalence of MDR genomes in certain countries and niche-specific virulome profiles observed in this study, these pressures will likely facilitate divergence in both AMR and pathogenesis of Aeromonas populations that could lead to increases in disease burden or treatment failure within production systems. Our results, therefore, provide a critical framework to support future surveillance and comparative genomics studies, to better inform targeted and effective control strategies for MAS in aquaculture.

Abbreviations

AMR

antimicrobial resistance

ANI

average nucleotide identity

ARGs

antimicrobial resistance genes

BV-BRC

bacterial and viral bioinformatics resource center

CDSs

coding sequences

GIs

genomic islands

GWAS

genome-wide association study

ISs

insertion sequences

MAS

motile Aeromonas septicaemia

MDR

multidrug resistance

ML

maximum likelihood

MLST

multi-locus sequence type

MSH

mannose-sensitive haemagglutinin

NCBI

National Center for Biotechnology Information

PCA

principal component analysis

SNP

single nucleotide polymorphism

STs

sequence types

TA

toxin–antitoxin

T4aP

type IVa pilus

T2SS

type II secretion system

T3SS

type III secretion system

T4SS

type IV secretion system

T5SS

type V secretion system

T6SS

type VI secretion system

VFDB

virulence factor database