Introduction

Aeromonas hydrophila is a gram-negative facultative anaerobic bacterium. It has been isolated from diverse environmental habitats (e.g., drinking water, wastewater, sewage, food, milk, soil) and has a broad host range from fish to humans [1,2,3,4]. A. hydrophila causes infectious disease with severe clinical signs such as gastroenteritis, necrotizing fasciitis, exophthalmia, fin rot, dropsy, hemorrhage, and septicemia as opportunistic bacterial pathogens in aquatic to terrestrial species [5,6,7]. In terms of fish diseases, A. hydrophila can infect various ranges of species [8,9,10]. Among all the genomes publicly available on GenBank until 2022, A. hydrophila has been reported to have been isolated from dozens of fish species, such as goldfish, carp, snakehead, eel, tilapia, catfish, bluegill, large yellow croaker, sturgeon, and salmon as summarized in Table S1. A. hydrophila can be both a primary and secondary bacterial pathogen in fish, depending on the fish’s health conditions [8, 11]. It is well known as an opportunistic pathogen that can infect individuals under poor and stressed conditions. Due to these characteristics, outbreaks of A. hydrophila are used as indicators of poor water quality management and polluted environments. In addition, A. hydrophila can survive in a wide range of water temperatures (0–45 °C) [8, 12, 13]. A. hydrophila naturally confronts a variety of environments that require adaptation. These attributes have the potential to act as evolutionary drivers, allowing bacteria to endure prolonged exposure to various challenging environments and adapt accordingly. A. hydrophila harbor numerous extracellular proteins which are known to be associated with environmental adaptability [14]. In response to varying environmental conditions, bacteria undergo adaptations by modulating gene expression and composition, leading to the emergence of diverse phenotypic characteristics, including virulence [4, 15, 16]. Not only Aeromonas but there are many cases that the genotypic differences affected significantly phenotypic characteristics such as pathogenicity and pH resistance in other fish bacterial pathogens [17, 18]. Alterations in the bacterial genome can exert profound effects on the intricate transcriptional network, subsequently influencing the associated phenotypic traits [19]. This suggests an increasing importance in acquiring comprehensive insights into the genotypic characteristics of A. hydrophila that are associated with its transcriptomic and/or phenotypic traits. Despite its significance, the integration of genomic and transcriptomic data in A. hydrophila across diverse evolutionary contexts has remained insufficiently addressed up until now.

In recent years, massive genomic and transcriptomic results have been released, and all these works contribute to the understanding of the physiological characteristics of fish bacterial pathogens [20,21,22,23,24,25,26,27]. The significance of interpreting and analyzing massive and deep sequencing data is increasingly recognized in the exploration of novel insights into bacterial evolution and the dynamics of genetic content alteration [28,29,30]. The integration of omics data derived from independent studies holds significant potential in providing novel insights and generating substantial impact through the application of big data analysis in microbiological research [31]. However, each study employs different bioinformatics pipelines, strategies, and tools for individual specific purposes, thus posing a challenge in intuitively comparing results across different datasets.

Flagella are widely recognized for their crucial roles in motility, adhesion, and biofilm formation, which are directly associated with the growth and adaptation of bacteria in viscous environments or spaces [32,33,34]. The diversity of flagella subunits can influence the assembly, shape, size, and movement of flagella, which results in variations in functionality [35,36,37]. Given the fact that a single polar flagellum allows A. hydrophila to swim efficiently and move in diverse environments, facilitating its invasion of host cells, the understanding of genotypic and transcriptomic characteristics in the flagellin subunits can reveal the microbiological significance of flagellin in A. hydrophila [38, 39]. During the pan-genome and comparative genomic study, we have noticed that strains from terrestrial and aquatic environments tend to harbor different compositional patterns of flagellin subunits, known to possess unique and typical/atypical expression patterns and/or functionalities. Accordingly, this study manually aligned and processed A. hydrophila genome and/or transcriptome data using a standardized in-house pipeline. The main objective was to elucidate the genotypic and transcriptomic characteristics of A. hydrophila flagellin subunits that might greatly impact their phenotypic traits.

Materials and Methods

Strains and Genomic Data

In this study, A. hydrophila genomes with available sample histories, including information on isolation site, date, and source from the BioProject database, were utilized. Table S1 provides comprehensive information about the strains used in this study, including isolated year, country, host, and accession numbers in GenBank, RefSeq, BioProject, and BioSample. In total, 151 A. hydrophila strains from 18 countries between 1980 and 2022 were analyzed in this study. Average nucleotide identity (ANI) in all A. hydrophila genomes and six other Aeromonas species (Aeromonas caviae, Aeromonas salmonicida, Aeromonas sobria, Aeromonas veronii, Aeromonas jandaei, Aeromonas dhakensis) were calculated using FastANI (version 1.34) [40]. A heatmap and dendrogram were created using the “gplots” package in R (version 4.2.3), based on ANI results among strains [41]. The dendrogram analysis employed the “Euclidean” method for distance calculation and clustered using the “average” method. Based on the isolation source, all strains were categorized into either aquatic or terrestrial environments. Among the 151 strains, 108 were classified as originating from aquatic environments, while the remaining 43 were categorized as terrestrial strains (Table S1). Flagellin homologs in A. hydrophila were examined, and three subunits of flagellin genes, annotated as flagellin A (flaA), flagellin B (flaB), and flagellin (fla), were identified using the NCBI Prokaryotic Genome Annotation Pipeline [42]. The number and type of flagella genes varied among strains [43].

Flagellin Homologs

The number of flaA, flaB, and fla gene copies from each strain was used for the dendrogram analysis using “ape” and “phytools” packages in R (version 4.2.3) [44,45,46]. Five distinct groups were identified based on the composition and number of flagellin homologs. This study named each strain named into fla(+ 1) to fla(+ 4) groups based on the number of flagellin copies. These groups encompassed strains with two to four copies of fla (abbreviated as fla(+ 2), fla(+ 3), or fla(+ 4)), strains with a combination of a single copy of flaA and flaB (abbreviated as flaA(+ 1)/flaB(+ 1)), and strains with a single copy each of the three flagellin subunits (abbreviated as fla(+ 1)/flaA(+ 1)/flaB(+ 1)). The relative proportion of the aquatic strains was determined using the following formula:

$$\mathrm R\mathrm e\mathrm l\mathrm a\mathrm t\mathrm i\mathrm v\mathrm e\;\mathrm p\mathrm r\mathrm o\mathrm p\mathrm o\mathrm r\mathrm t\mathrm i\mathrm o\mathrm n\;\mathrm o\mathrm f\;\mathrm s\mathrm t\mathrm r\mathrm a\mathrm i\mathrm n\mathrm s\;\mathrm f\mathrm r\mathrm o\mathrm m\;\mathrm a\mathrm q\mathrm u\mathrm a\mathrm t\mathrm i\mathrm c\;\mathrm e\mathrm n\mathrm v\mathrm i\mathrm r\mathrm o\mathrm n\mathrm m\mathrm e\mathrm n\mathrm t\;\left(\text{AE}\right)=\frac{\mathrm N\mathrm u\mathrm m\mathrm b\mathrm e\mathrm r\;\mathrm o\mathrm f\;\mathrm s\mathrm t\mathrm r\mathrm a\mathrm i\mathrm n\mathrm s\;\mathrm f\mathrm r\mathrm o\mathrm m\;\mathrm o\mathrm f\;\mathrm a\mathrm q\mathrm u\mathrm a\mathrm t\mathrm i\mathrm c\;\mathrm e\mathrm n\mathrm v\mathrm i\mathrm r\mathrm o\mathrm m\mathrm e\mathrm n\mathrm t\mathrm s\;\mathrm i\mathrm n\;\mathrm e\mathrm a\mathrm c\mathrm h\;\mathrm f\mathrm l\mathrm a\mathrm g\mathrm e\mathrm l\mathrm l\mathrm i\mathrm n\;\mathrm g\mathrm r\mathrm o\mathrm u\mathrm p}{\text{Total}\;\mathrm{nu}\mathrm m\mathrm b\mathrm e\mathrm r\;\mathrm o\mathrm f\;\mathrm s\mathrm t\mathrm r\mathrm a\mathrm i\mathrm n\mathrm s\;\mathrm f\mathrm r\mathrm o\mathrm m\;\mathrm a\mathrm q\mathrm u\mathrm a\mathrm t\mathrm i\mathrm c\;\mathrm e\mathrm n\mathrm v\mathrm i\mathrm r\mathrm o\mathrm n\mathrm m\mathrm e\mathrm n\mathrm t\;\left(n=108\right)}$$
(1)
$$\mathrm R\mathrm e\mathrm l\mathrm a\mathrm t\mathrm i\mathrm v\mathrm e\;\mathrm p\mathrm r\mathrm o\mathrm p\mathrm o\mathrm r\mathrm t\mathrm i\mathrm o\mathrm n\;\mathrm o\mathrm f\;\mathrm s\mathrm t\mathrm r\mathrm a\mathrm i\mathrm n\mathrm s\;\mathrm f\mathrm r\mathrm o\mathrm m\;\mathrm t\mathrm e\mathrm r\mathrm r\mathrm e\mathrm s\mathrm t\mathrm r\mathrm i\mathrm a\mathrm l\;\mathrm e\mathrm n\mathrm v\mathrm i\mathrm r\mathrm o\mathrm n\mathrm m\mathrm e\mathrm n\mathrm t\;\left(\text{TE}\right)=\frac{\mathrm N\mathrm u\mathrm m\mathrm b\mathrm e\mathrm r\;\mathrm o\mathrm f\;\mathrm s\mathrm t\mathrm r\mathrm a\mathrm i\mathrm n\mathrm s\;\mathrm f\mathrm r\mathrm o\mathrm m\;\mathrm o\mathrm f\;\mathrm t\mathrm e\mathrm r\mathrm r\mathrm e\mathrm s\mathrm t\mathrm r\mathrm i\mathrm a\mathrm l\;\mathrm e\mathrm n\mathrm v\mathrm i\mathrm r\mathrm o\mathrm m\mathrm e\mathrm n\mathrm t\mathrm s\;\mathrm i\mathrm n\;\mathrm e\mathrm a\mathrm c\mathrm h\;\mathrm f\mathrm l\mathrm a\mathrm g\mathrm e\mathrm l\mathrm l\mathrm i\mathrm n\;\mathrm g\mathrm r\mathrm o\mathrm u\mathrm p}{\text{Total}\;\mathrm{nu}\mathrm m\mathrm b\mathrm e\mathrm r\;\mathrm o\mathrm f\;\mathrm s\mathrm t\mathrm r\mathrm a\mathrm i\mathrm n\mathrm s\;\mathrm f\mathrm r\mathrm o\mathrm m\;\mathrm t\mathrm e\mathrm r\mathrm r\mathrm e\mathrm s\mathrm t\mathrm r\mathrm i\mathrm a\mathrm l\;\mathrm e\mathrm n\mathrm v\mathrm i\mathrm r\mathrm o\mathrm n\mathrm m\mathrm e\mathrm n\mathrm t\;\left(n=43\right)}$$
(2)
$$\begin{array}{c}\mathrm{Relative\;proportion\;of\;aquatic\;strains\;in\;each\;flagellin\;group}=\\\frac{\text{AE}}{(\text{AE}+\text{TE})}\end{array}$$
(3)

Chi-square analysis was performed using the relative proportion of strains from the aquatic environment (AE) and relative proportion of strains from terrestrial environment (TE) using the “gmodels” package in R [47]. p-value less than 0.05 was considered a statistical significance.

Phylogenetic Analysis of Flagellin Homologs

The sequences of all flagellin homologs were obtained from genomic coding sequences and GFF/GTF files. Approximately, 400 flagellin nucleotide sequences were utilized for phylogenetic analysis using Nextstrain [48]. Briefly, all flagellin sequences are grouped by type of flagellin composition, isolation year, and isolation site. Also, “sequences-per-group,” which specifies the maximum number of sequences per group, was assigned to 50. All sequences were aligned by the option of “–file-gap” without reference sequence for constructing the phylogenetic tree. Time-resolved tree was generated through the option of “data-confidence”, “coalescent = opt”, and “data-inference = marginal”. The trait, ancestral sequence inference, and mutation sites were annotated and profiled using the “traits”, “ancestral”, and “translate” commands. The phylogenetic results were then visualized in auspice using the “nextstrain view” command line.

Flagellin Expression Level and Global Transcriptome

The investigation into the transcriptomic characteristics of various flagellin genes involved the selection of five strains (23-C-23, ATCC7966, NJ-35, SCV-1, and WCX23). These strains were chosen given their composition of different flagellin genes in their genome and availability in open databases. At least one of selected five strains has been confirmed to harbor the fla, flaA, and flaB genes. The flagellin gene and global bacterial transcriptomic expressions under ordinary cultural environments currently available in SRA (Sequence Read Archive) were profiled from five A. hydrophila strains using raw transcriptomic files (FastQ) [22,23,24]. Three to four biologically replicated FastQ files derived from the transcriptome of each A. hydrophila strain cultured under ordinary culturable conditions (e.g., tryptone soy or Luria–Bertani medium at 28 °C for ~ 1 day) were downloaded from SRA [20,21,22,23,24]. The strain WCX23 and 23-C-23 belong to flaA(+ 1)/flaB(+ 1) flagellin group, and ATCC7966 harbored single flaA, flaB, and fla gene in the genome (flaA(+ 1), flaB(+ 1), and fla(+ 1) group). On the other hand, NJ-35 and SCV-1 strain have three fla genes belonging to a flagellin fla(+ 3) group (Table 1).

Table 1 The information of strains used for transcriptomic study
Full size table

In this study, two distinct bioinformatics pipelines were utilized to assess the diverse levels of flagellin expressions across each strain, as well as to conduct a comparative analysis of the global transcriptome among the strains. To quantify the expression of various subunits of flagellin present in individual strains, transcriptomic reads were aligned to the genome of each strain (Fig. 1). In detail, FastQ files after trimming bad quality sequence reads below Q20 scores using Trimmomatic (Galaxy version 0.38.1) [49] were mapped onto each reference genome (WCX23; NZ_CP038463, 23-C-23; NZ_CP038465, ATCC7966; NZ_JAGDEM010000001, NJ-35; NZ_CP006870, and SCV-1; DANIKG010000001) using Bowtie2 with very sensitive end-to-end option (–very-sensitive) (Galaxy version 2.5.0 + galaxy0) [50, 51]. The transcriptome assembly, merge, and quantification were carried out using cufflinks (Galaxy version 2.2.1.3), cuffmerge (Galaxy version 0.0.5), and cuffdiff (Galaxy version 2.2.1.6) [52]. The FPKM (fragments per kilobase per million mapped fragments) value was validated to TPM (transcripts per million) value using below formula. TPM values of flagellin homologs from each sample were utilized for the level of flagellin gene expression as sample traits in this study. Welch’s t-test and Duncan’s multiple range test (analysis of variance; ANOVA) were carried out for the transcriptomic levels of each type of flagellin using an “aricolae” package in R (version 4.2.3) [53].

Fig. 1
figure 1

Schematic for profiling the expression of each flagellin gene and global gene expressions among the strains used in this study. The figure was drawn using BioRender (https://biorender.com/)

Full size image
$$TPM= \frac{({\text{FPKM}}\;\mathrm{value\;at\;each\;gene\;}\times\;1000000)}{\mathrm{Sum\;of\;all\;transcriptome\;FPKM\;values}}$$
(4)

On the other hand, as for the global gene expression for the comparison among strains, all transcription reads were mapped onto one specific reference genome (FDAARGOS_916; NZ_CP065651). Except for the reference genome, the same pipeline was used as described above for obtaining TPM values (Fig. 1).

Bioinformatics

Differentially Expressed Genes (DEGs)

Based on global transcriptome results, the differentially expressed genes between flaA(+ 1)/flaB(+ 1) (WCX23 and 23-C-23 strains) and fla(+ 3) (NJ-35 and SCV-1 strains) groups were observed. The p-value of each gene between flagellin groups was calculated using edgeR in R. The gene lower than p-value 1.14 × × 10 −5 from the Bonferroni adjustment (0.05/total number of genes observed in this study) was regarded as DEGs in this study. The DEGs were annotated and mapped onto Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways [54].

Weighted Correlation Network Analysis (WGCNA)

The transcriptomic results from the strains that harbored flaA and flaB genes (WCX23, 23-C-23, and ATCC7966) were utilized for WGCNA. FlaA and flaB TPM results were used as sample traits. The signed network with the option of verbose = 3, merge cut height = 0.25, minimum module size = 30, and soft threshold power = 20 was generated to identify the gene cluster shared similar pattern of expression. Each module eigengene was used for correlation analysis with the expression of flaA and flaB to identify modules significantly correlated with their pattern of expression (p-value < 0.05). All genes belonging to significant modules were annotated using KEGG. Similarly, WGCNA was conducted under the same conditions as described above, utilizing NJ-35 and SCV-1 strains, which possess three fla genes (fla-2, fla-3, and fla-4) belonging to the fla(+ 3) group. All analyses and data visualization were carried out using “matrixStats”, “impute”, and “WGCNA” packages in R (version 4.2.3) [55].

Correlation Analysis

Correlation coefficients between sample traits (flaA and flaB) and all gene expressions were used for the clustering analysis. The heatmap with dendrogram was visualized using “RColorBrewer”, “gplot”, “pheatmap”, and “d3heatmap” in R (version 4.2.3). Based on the pattern of correlation coefficients between all genes and flaA and flaB, eight gene clusters were divided. Based on the pattern of overall correlation coefficients with flaA and flaB in a cluster, positive coefficients were denoted with “ +” , while negative coefficients were denoted with “ −” . Gene clusters harbored both positive correlation coefficients in flaA and flaB (flaA + /flaB +), opposite direction of correlation coefficients between homologs (flaA + /flaB- or flaA-/flaB +), or both negative correlation coefficients (flaA-/flaB-) were annotated through KEGG pathways.

KEGG Pathway Enrichment

Genes in significant WGCNA modules and gene clusters were mapped onto KEGG pathways. Regarding the DEG results, genes showing significantly higher expression in the flaA(+ 1)/flaB(+ 1) groups compared to the fla(+ 3) groups were denoted as upregulated, while the reverse scenario was labeled as downregulated. The significance of up- and downregulated DEG numbers in each pathway was evaluated using the Chi-square analysis, and the pathway with a p-value less than 0.05 was visualized. Likewise, top five pathways enriched by genes from significant modules through WGCNA and gene clusters with the different patterns of flaA and flaB correlation coefficients were drawn by “EnrichmentMap” in Cytoscape 3.10.0 [56, 57]. The KEGG pathways relevant to human diseases were not used for the analyses in this study. As for the WGCNA and gene cluster analyses, the KEGG two-component system, which includes many genes and has relatively broad biological meaning and implications, was not included in data visualization.

Results

Average Nucleotide Identity (ANI)

In this study, ANI results of 151 genomes were in a range of 95.3555–99.9997%, with one exception in B11 strain. The ANI between the label A. hydrophila strain B11 and other A. hydrophila strains was between 93.2555 and 94.1305%. However, it exhibits a higher ANI with A. dhakensis (97.1336%) (Figure S1; Table S2). The ANI among five strains (WCX23, 23-C-23, ATCC7966, NJ-35, and SCV-1) utilized for transcriptomic analysis in this study ranged from 96.6255 to 99.9980% (Table S2). Among them, a high ANI sharing of over 99.9% was observed between the 23-C-23 and WCX23 strains, as well as between the NJ-35 and SCV-1 strains. Interestingly, among these five strains, those sharing high ANI values correspondingly belong to the same flagellin group (flaA(+ 1)/flaB(+ 1) and fla(+ 3)).

The patterns of Flagellin Homolog Compositions and Phylogenetics in A. hydrophila

The number and type of flagellin genes in each strain were examined, revealing that 37 strains belonged to the fla(+ 3) group, 8 strains to the fla(+ 4) group, 56 strains to the flaA(+ 1)/flaB(+ 1) group, 12 strains to the fla(+ 1)/flaA(+ 1)/flaB(+ 1) group, 32 strains to the fla(+ 2) group, and the rest 6 strains that did not belong the majority. The strains harboring flaA and flaB were found to originate predominantly from terrestrial environments, whereas the fla(+ 3) and fla(+ 4) flagellin groups predominantly consisted of strains from aquatic environments. Specifically, the relative proportion of aquatic strains within the fla(+ 3) group was significantly higher compared to strains from terrestrial environments (Fig. 2). Based on phylogenetic analysis, the flagellin genes were classified into six subunits: four fla groups (named as fla-1, fla-2, fla-3, and fla-4, which were annotated same term but phylogenetically far distance flagellin), flaA, and flaB (Fig. 3). Interestingly, a notable observation was that the majority of strains did not possess a single type of flagellin gene, but instead exhibited a combination of different flagellin subunits. For instance, in the fla(+ 2) and fla(+ 3) flagellin groups, it was found that most strains harbored fla-1, fla-2, fla-3, and/or fla-4 genes, rather than multiple copies of the same flagellin type (Fig. 3A). The majority of strains belonging to the fla(+ 2), fla(+ 3), and fla(+ 4) groups possess at least one fla-4 gene. However, there were slight phylogenetic differences observed between fla-4 genes from the fla(+ 2) and fla(+ 3) groups (Fig. 3A, B). Additionally, it was noted that fla(+ 2) strains tended to be isolated more recently compared to fla(+ 3) strains (Fig. 3B). In terms of gene expression levels, flaB and flaA exhibited the highest and second-highest expression levels among the seven flagellin homologs, respectively. In contrast, fla-2 to fla-4 displayed relatively similar lower levels of expression. Conversely, the fla-1 gene showed very low or no expression (Fig. 3C). The phylogeographic results revealed that a significant number of strains positive for flaA and flaB originated from terrestrial sources, particularly strains from China and the USA (Fig. 4). On the other hand, most fla-2 to fla-4 genes were derived from aquatic strains, and notably, no terrestrial strain was observed in fla-2. In contrast, fla-1, which displayed the most distant phylogenetic relationship with other flagellin genes, exhibited a higher proportion of terrestrial strains compared to fla-2 to fla-4 (Fig. 4).

Fig. 2
figure 2

Dendrogram based on the type and number of flagellin genes by Aeromonas hydrophila strains. The blue and red circles indicate the isolated site of each strain (either aquatic or terrestrial environment). Bar color and length denote the type and number of flagellin genes. The relative proportion of strains originating from aquatic environment was described by the group sharing the same number of type of flagellin genes (fla(+3), fla(+4), fla(+1)/flaB(+1), fla(+1)/flaA(+1)/flaB(+1), and fla(+2)). Asterisks (***) indicates the significantly different number of aquatic strains compared to terrestrial strains in same flagellin group using Chi-square analysis (p-value < 0.001)

Full size image
Fig. 3
figure 3

Phylogenetic tree of all flagellin genes in Aeromonas hydrophila observed in this study (A). Zoom in on the fla-4 clade with the label of isolation years (B). The transcription level of flagellin paralogs (fla-1, flaB, flaA, fla-2, fla-3, and fla-4) in 23-C-23, ATCC7966, NJ-35, SCV-1, and WCX23 strain based on transcripts per kilobase million (TPM) value (C). There was no statistically different expression between or among the strain within a same type of flagellin based on Welch’s t-test or one-way analysis of variance test using Duncan’s multiple comparison test

Full size image
Fig. 4
figure 4

Phylogeographic analysis and distribution of aquatic and terrestrial strains in flaA (A), flaB (B), fla-1 (C), fla-2 (D), fla-3 (E), and fla-4 (F) genes

Full size image

DEGs and Featured Pathways in the Flagellin Group

In total, 486 DEGs were identified between flaA(+ 1)/flaB(+ 1) and fla(+ 3) group, and 459 DEGs showed higher expression in flaA(+ 1)/flaB(+ 1) group. Five featured KEGG pathways, namely two-component system (ko02020), ABC transporters (ko02010), amino sugar and nucleotide sugar metabolism (ko00520), galactose metabolism (ko00052), and phosphotransferase system (PPTS) (ko02060), were identified as the most significantly different numbers of up- and downregulated DEGs across different flagellin groups. Notably, the flaA(+ 1)/flaB(+ 1) group exhibited a high number of upregulated DEGs in all five identified pathways (Fig. 5A). In particular, the two-component system pathway exhibited the highest number of DEGs among the identified pathways. Within this pathway, the flagellar motor switch adaptation system played a significant role, involving several DEGs such as methyl-accepting chemotaxis (mcp); aerotaxis receptor (aer); purine-binding chemotaxis protein (cheW); two-component system, chemotaxis family, sensor kinase (cheA); and two-component system, chemotaxis family, chemotaxis protein (cheY). In the A. hydrophila ASM162068 genome, five copies of mcp and two copies of aer were identified. The average gene expression levels of these genes were significantly higher in the flaA(+ 1)/flaB(+ 1) flagellin group. Similarly, the flaA(+ 1)/flaB(+ 1) group exhibited significantly higher TPM values in the genes cheW, cheA, and cheY compared to the fla(+ 3) group (Fig. 5B).

Fig. 5
figure 5

The KEGG enrichment analysis revealed a significantly different number of DEGs (differentially expressed genes) based on Chi-square analysis (p-value < 0.05) between flaA(+1)/flaB(+1) and fla(+3) strains (A). The two-component system relevant to flagellar motor switch adaptation based on the KEGG pathway for the prokaryote (B). The TPM level of relevant genes (mcp, aer, cheY, cheA, and cheW) was shown in raincloud plots. The expression of five homologous or paralogous mcp (Gene id: ①I6G73_RS10070, ②I6G73_RS19490, ③I6G73_RS04110, ④ I6G73_RS08330, and ⑤ I6G73_RS16810) and two aer (Gene id: ①I6G73_RS19930 and ②I6G73_RS05190) was shown as dot plots in accordance with different strains and groups. Asterisks (*p < 0.05, **p < 0.01, and ***p < 0.001) showed statistical difference between flaA(+1)/flaB(+1) and fla(+3) groups based on Welch two sample t-test

Full size image

Module Identification Correlated with Flagellin Expressions Through WGCNA

The dendrogram showed overall transcriptomic expression of ATCC7966, WCX23, and 23-C-23, along with the relative expression levels of flaA and flaB utilized in the WGCNA (Fig. 6A). In total, eight modules were identified, and the eigengene of yellow and turquoise modules was positively and negatively correlated with the expression of flaA (Figure S2; Fig. 6B). On the other hand, flaB showed only a negatively significant correlation with brown eigengene (Fig. 6B). In total, 2125, 366, and 188 genes were labeled and identified as the turquoise, brown, and yellow modules, respectively. Also, since two-component system was the pathway composed of the highest numbers belonging to each module, regardless of module’s significance, it was not visualized in this study. ABC transporters and quorum sensing are the pathways with high number of genes in turquoise modules. But, featured pathways in brown module were mainly linked to biomolecule degradation (e.g., fatty acid, valine, leucine, and isoleucine) and metabolism (e.g., pyrimidine and nitrogen metabolism). Although the yellow module consisted of a relatively small number of genes compared to others, these genes were found to be closely associated with flagellar assembly, biofilm formation, and bacterial chemotaxis. On the other hand, fla-2, fla-3, and fla-4 expression patterns were almost identical among samples, and their correlation coefficient values with six detected modules were also similar (Figure S3).

Fig. 6
figure 6

Clustering dendrogram with the flaA and flaB expression in each strain (A). The thicker red color denotes the higher flaA or flaB expression. The correlation between module eigengene and flaA or flaB expression (B). The highest enriched KEGG pathways in a turquoise (C-a), brown (C-b), and yellow module (C-c)

Full size image

Correlation Analysis Between All Gene and flaA/flaB Expressions

The correlation coefficients between all gene expressions and flagellin genes (flaA and flaB) are shown in Fig. 7. Eight clusters (cluster 1–cluster 8) were identified based on correlation coefficient values between flaA and flaB. The top five pathways, each consisting of more than three genes from each cluster, were selected. Cluster 8, with both high expressions in flaA and flaB (flaA + /flaB +), comprised the genes that positively correlated with the expression of both flaA and flaB and were relevant to the propanoate and butanoate metabolism, valine, and leucine and isoleucine degradation. On the other hand, all genes that have both flaA and flaB negative correlation coefficients belonging to cluster 3, with both downregulations in flaA and flaB, (flaA-/flaB-), were mainly mapped onto the pyrimidine and purine metabolism, bacterial secretion system, etc. Clusters 1 and 2 showed a positive flaA but negative flaB correlation coefficient (flaA + /flaB-), and ribosome was the only pathway shared between clusters. Although many KEGG pathways share the same number of genes, relatively, many pathways were visualized in cluster 2, O-antigen nucleotide sugar biosynthesis, nitrogen metabolism, and ABC transporters were the highest pathways (Fig. 7). On the other hand, quorum sensing and ABC transporters were one of the featured pathways in cluster 6, composed of genes with negative flaA but positive flaB correlation coefficients (Fig. 7).

Fig. 7
figure 7

Dendrogram with heatmap based on the correlation coefficient between flagellin gene (flaA or flaB) and other gene expressions among A. hydrophila strains. Enriched KEGG pathways from the gene cluster 8 (A), cluster 6 (B), cluster 3 (C), cluster 1 (D), and cluster 2 (E)

Full size image

Discussion

In this study, we maximized the use of genomic and transcriptomic data to investigate the characteristics of different subunits of flagellin in A. hydrophila. This study incorporated comparative genomics analyses and conducted comprehensive profiling of transcriptomic responses across five representative strains of A. hydrophila. These strains harbored various copies of flagellin genes with characteristics that have yet to be clearly elucidated in their genomes. To assess the expression levels of individual flagellin genes, we mapped the transcriptomic data to their respective reference genomes. On the other hand, in examining the overall and global transcriptomic responses across strains, we analyzed the transcriptomic data by aligning it to a single reference genome. This approach facilitates a thorough investigation of the global transcriptomic expression across strains, enabling the preservation of diverse flagellin expression profiles within each genome. Additionally, it helps to the identification of correlated genes associated with flagellin expression.

Most of the A. hydrophila genomes exhibited a high ANI of over 95%; however, the B11 strain showed ANI values in the range of 93 to 94%, which are lower than the ANI comparison with A. dhakensis (~  97%). The mislabeling issue of Aeromonas species in the GenBank database has been raised previously. Beaz-Hidalgo et al. [58] found that mislabeled A. hydrophila genomes in the GenBank, which had ANI values ranging from 93.70 to 93.84% with the A. hydrophila ML09-119 strain, actually belonged to A. dhakensis. Taken together, although the B11 strain has been labeled and utilized as A. hydrophila, it actually belongs to A. dhakensis. Considering that A. dhakensis has been documented as a highly virulent species capable of causing mortality in both humans and fish [59,60,61], further studies on the classification and understanding of A. dhakensis are necessary. However, the B11 strain did not harbor any distinctiveness in the pattern of flagellin subunit composition compared to other A. hydrophila strains. Among the five strains used for the transcriptomics study, the 23-C-23 and WCX23 strains, as well as the NJ-35 and SCV-1 strains, shared a high ANI of over 99.9%. Given that former studies [62, 63] reported that the isolation of clonal related A. hydrophila strains sharing high ANI (>  99%) in the fish, the strains over 99% ANI in this study may be genotypically clonal-related strains.

Flagella play a crucial role not only in the motility and virulence of bacteria but also in their evolutionary adaptation to diverse environmental conditions [64, 65]. Six flagellin clusters were phylogenetically identified among the flagellin sequences obtained from 151 A. hydrophila genomes, including flaA, flaB, and four additional flagellin. Notably, this study revealed significant phylogenetic variations among four fla genes annotated with the same term. Consequently, these flagellins were arbitrarily designated as fla-1 to fla-4. In general, the different subunits of flagellin composition can affect the swimming behaviors of bacteria [66]. Kühn et al. [67] investigated the phenotypic characteristics of the different assemblies of flaA and flaB in Shewanella putrefaciens, and flaA mutant strain (ΔflaA) and flaB mutant strains (ΔflaB) showed completely different patterns of swimming and colony structure. An impaired swimming capacity with significantly reduced filament length was observed in the ΔflaB strain, while ΔflaA strain did not exhibit substantial differences compared to the wild-type strains. Conversely, in the case of Helicobacter pylori, the ΔflaA strains showed a loss of motility, whereas the ΔflaB mutants retained their motility [68]. This indicates that the significance of flaA and flaB concerning bacterial motility may vary depending on the species and could be directly associated with bacterial motility. Interestingly, this study revealed that the majority of fla(+ 3) strains were derived from aquatic sources rather than terrestrial environments. In contrast, A. hydrophila strains in the flaA(+ 1)/flaB(+ 1) and fla(+ 1)/flaA(+ 1)/flaB(+ 1) groups tend to be more frequently isolated from terrestrial environments and animals. In Shewanella putrefaciens, diverse fractions of cells demonstrating screw motion in a high-viscosity environment were found to be influenced by the specific flagellin type [67]. Similarly, in the case of Aeromonas hydrophila, the observed variations in flagellin subunits harbored by different strains are likely to be a consequence of adaptive responses to distinct environmental challenges. Likewise, Nakamura and Minamino [36] suggested that flaA can be the principal component in soil bacteria such as Sinorhizobium meliloti and Rhizobium lupine. These results highlight the potential impact of flagellin diversity on the motility behavior of microorganisms and the adaptive significance of such variations in coping with varying environmental conditions.

In Aeromonas species, the sequence and genetic diversity in flagellin genes have greatly influenced the activity of biofilm formation, pathogenicity, and swimming motility [69]. Miyagi et al. [69] investigated genetic diversity in flaA in A. hydrophila, and the phylogenetic clusters were clearly divided in accordance with the type of origins, either clinical or environmental basis. The different level of motility and biofilm activity observed by flaA clusters indicated the genetic changes in the flagellin gene could drive different phenotypic characteristics. Given that flagella-relevant proteins, which directly exposed to outer environments, they can exhibit high variability due to external conditions. Based on the massive sequence information in Vibrio harveyi, the group incubated in seawater at 30 °C revealed a possibility that changes the genetic mutation in the flagellar relevant gene (putative flagellar motor switch protein) [29]. This study also verified some flagellin genes, for example fla-4, were clustered with the time of isolation (isolation year). This implied that the status of flagellin genes in A. hydrophila is not static but highly mutable and changeable, which can lead to environmental adaptation. However, the composition or nucleotide sequence of flagellin subunits did not show a significantly different pattern by country, not like subunits of environments such as aquatic and terrestrial traits.

The expression level of flagellin genes in ordinary cultural environments differed significantly depending on the flagellin type. FlaA and flaB showed relatively high expression, while fla-1 to fla-4 was not highly expressed like flaA and flaB. Interestingly, the strains belonging to the flaA(+ 1)/flaB(+ 1) group comprised many upregulated DEGs relevant to two-component system. Among many subordinated two-component systems, bacterial chemotaxis relevant to the flagellar motor switch adaptation pathway is featured and enriched subordinated systems in flaA(+ 1)/flaB(+ 1) group. In general, bacteria utilize chemotaxis to control their movement toward favorable environments by detecting the levels of chemoeffectors (e.g., attractants and repellents) through chemoreceptors [70]. The level of chemoeffectors can be detected by mcp and aer, and these signals were delivered through the cheW and cheA [71]. The level of autophosphorylated cheA is greatly influenced by signal transduction, which affects the stimulation of cheY by phosphorylation. Ultimately, phosphorylated cheY interacts with the flagellar motor and switches the direction of rotation in the flagellar [70, 72]. The high expression of flaA and flaB, along with the elevated expression of chemoreceptors and response regulator proteins (cheW, cheA, and cheY), might indicate that flaA(+ 1)/flaB(+ 1) group may have more robust flagella activity and/or movement, compared to the fla(+ 3) group. In addition, given that more than half number of strains in flaA(+ 1)/flaB(+ 1) group originated from terrestrial habitats where high variations of temperature and humidity were observed [73], it has been speculated that A. hydrophila can evolve and develop the type of flagellin fitting into the facing environment. Also, bacterial flagellins are known to the important epitope that causes the strong immuno-stimulatory response through toll-like receptor in aquatic animals [64, 74, 75]. Gao et al. [64] investigated that the level of immuno-stimulation effects varies depending on the subunits of flagellin (flaA, flaB, flaC, flaD, and flaE). This implies that the composition of flagellin genes in A. hydrophila could also have a significant impact on infection and prognosis.

In general, the construction of relationship networks based on the pattern of co-expression through pairwise measurement could be a robust approach for investigating biologically meaningful pathways. WGCNA, as the more advanced method, has received great attention to clarify the relationship between multidimensional results and featured sample traits [55, 76]. In this study, we applied the WGCNA method to identify the gene clusters that are highly correlated with flaA and flaB expression. FlaA and flaB were not typically co-expressed in A. hydrophila. For example, the yellow module, which positively correlated with the expression pattern of flaA but showed no significance with flaB, consisted of genes related to flagellar assembly, bacterial chemotaxis, and biofilm formation. On the other hand, a brown module, which showed a negative correlation with flaB expression but lacked significance with flaA, harbored the same flagellar assembly pathway. This does not necessarily imply that flaA and flaB have opposing functions in flagella assembly, but rather indicates that they do not share identical functions and characteristics. While both flaA and flaB may be associated with the flagellar assembly pathway, they may act in potentially different capacities or manners. Mohari et al. [77] investigated that Agrobacterium tumefaciens have several subunits of flagellin from flaA to flaD. FlaA in A. tumefaciens showed the highest importance that affected flagellin motility and functions, but others (flaB–flaD) had accessory functions [77]. Likewise, transcriptional expression of flaA and flaB in H. pylori is controlled by different sigma factors (ς28 and ς54), which suggests that both flagellin were regulated in a different timely manner [68, 78, 79]. Considering gene clusters or gene correlation analysis, it is a high possibility that flaA is probably associated with biofilm formation, bacterial chemotaxis, and bacterial secretion systems. Besides, flaB might be more associated with quorum sensing activities. But further studies are necessary to confirm this hypothesis.

Conclusion

This study conducted the first comprehensive investigation of hundreds of A. hydrophila genomes to profile flagellin subunits and clusters. We identified flaA, flaB, and four other phylogenetically different units of flagellin (fla-1 to fla-4). The composition of flagellin genes in A. hydrophila genomes exhibited significant diversity among strains, with a predominant presence of 3 to 4 fla subunits (fla(+ 3) and fla(+ 4) groups) originating from aquatic environments. In contrast, many terrestrial strains tended to harbor different flagellin subunits (flaA and flaB) in their genomes. Through multiple transcriptomic analyses conducted on strains with different flagellin subunits compositions under ordinary culture conditions, this study demonstrated significantly higher expression of flaA and flaB compared to the other four flagellins. Furthermore, genes associated with flagellin motor switch adaptation, controlling bacteria chemotaxis movement, showed significantly higher expression in the flaA(+ 1)/flaB( +) group compared to fla(+ 3). While further studies are necessary to elucidate the specific functions of each flagellin subunit, the pattern of flagellin composition and expression level found in this study provide valuable insights into the understanding of the role of flagellin and function and the adaptive mechanisms of microorganisms in diverse environments.