Isolation and culture

Strain MaLMAid0302^T was isolated in July 2021 from the coral species Scleronephthya gracillimum collected near Jeju Island, Republic of Korea (33°13′40″N, 126°34′07″E). Strain SPO723^T was isolated in November 2008 from a leaf of Zostera marina found on Weno Island, Chuuk State, Micronesia (50°03′N, 73°03′E). For isolation, the substratum of the coral sample or a piece of the leaf sample was homogenized and diluted with sterilized seawater. The diluted solution was spread onto Reasoner’s 2A (R2A) agar supplemented with 1.5% (w/v) sodium chloride (MR2A) for coral samples and incubated at 25°C for 7 days. For plant leaf samples, the solution was spread onto marine agar 2216 (MA; Difco) and incubated at 30°C for 2–3 days. Individual colonies were picked from MR2A or MA, and the morphologically distinct strains MaLMAid0302^T and SPO723^T were selected for further characterization. Purified strains were cultivated on MA at 25°C or 30°C and stored at –80°C in a medium supplemented with 20% (v/v) glycerol. Strain MaLMAid0302^T has been deposited in the Korean Collection for Type Cultures (KCTC 92348^T) and the Japan Collection of Microorganisms (JCM 35474^T). Strain SPO723^T is deposited in the Korean Culture Center of Microorganisms (KCCM 42324^T) and the Japan Collection of Microorganisms (JCM 14066^T). For comparative phenotypic and biochemical analyses, type strains of the Pseudovibrio species were obtained from the Leibniz Institute DSMZ and KCTC. These strains were cultivated on MA at 28°C. As the bacterial strains were isolated from publicly accessible marine environments and the study did not involve human or animal subjects, no ethical approval was required.

Phylogenetic analysis based on 16S rRNA gene sequences

Genomic DNA was extracted from the isolated strains using a Qiagen genomic DNA extraction kit following the manufacturer’s protocol. The 16S rRNA gene was amplified using universal bacterial primers 27F and 1492R, and the PCR product was purified using a MinElute PCR Purification Kit (Qiagen). Sequencing was performed by Macrogen (Seoul, Republic of Korea) on a 3730XL DNA Analyzer (Applied Biosystems), and the sequences were assembled using MEGA version X software. To determine phylogenetic relationships, 16S rRNA gene sequences of closely related taxa were retrieved from the GenBank [23] and EzBioCloud databases [24]. Sequence alignment was carried out using CLUSTAL_W in MEGA X [25]. Phylogenetic trees were constructed with 1,000 bootstrap replicates using three methods: the neighbor-joining method (NJ) [26] with the Kimura 2 parameter model [27], the maximum-likelihood (ML) method [28], and the maximum-parsimony (MP) method [29]. The partial deletion option (<80%) was applied during tree construction. Martellela mediterranea MACL11^T (AY649762) and Hyphomicrobium vulgare ATCC 27500^T (Y14302) were used as outgroup taxa. The 16S rRNA gene sequences of strains MaLMAid0302^T and SPO723^T have been deposited in GenBank under the accession numbers OP606461 and DQ660386, respectively.

Phenotypic and chemotaxonomic characteristics of strains

All strains used for phenotypic and chemotaxonomic testing were routinely cultivated on marine agar at 28°C for 3 days. The morphological characteristics of strains MaLMAid0302^T and SPO723^T were observed using transmission electron microscopy (HT-7800, Hitachi Ltd.) after negative staining with 2% (w/v) uranyl acetate. Anaerobic growth was assessed using the BD GasPak EZ pouch system on marine agar (MA; BD Difco) at 28°C for 7 days. Temperature growth ranges (4–40°C, at intervals of 5°C), pH tolerance (4.0–11.0, at increments of 1.0), and NaCl tolerance (0–8.0%, w/v, in 0.5% increments) were evaluated in marine broth (MB; BD) or modified ZoBell 2216 broth over a 7-day period. pH adjustments were made using specific buffers: MES (pH 4–6), HEPES (pH 6–8), AMPSO (pH 8–10), or sodium bicarbonate/sodium carbonate (pH 11). Growth was monitored by measuring absorbance (OD600) every 12 hours using a spectrophotometer. Catalase activity was tested by bubble formation in 3% (v/v) hydrogen peroxide, and oxidase activity by oxidation of 1% (w/v) tetramethyl-p-phenylenediamine. Enzyme activity, carbon source utilization, and other biochemical properties were analyzed using API 20E, API 20NE, and API ZYM kits with sodium chloride concentration adjusted to 3%. Results were recorded after cultivation at 28°C for several days. For chemotaxonomic analysis, strains were grown on marine agar for 3 days at 25°C. Cellular fatty acid methyl esters were prepared following the Sherlock Microbial Identification System (MIDI) version 6.1 protocol. Fatty acid profiles were separated and identified using an Agilent 6890N network GC system (Agilent Technologies) with the TSBA6 library.

Information on genomic sequences

Strains MaLMAid0302^T and SPO723^T were sequenced using PacBio 10K (Pacific Biosciences) and Illumina TruSeq technology, respectively, by CJ Bioscience and Macrogen in Seoul, Republic of Korea. For the strain MaLMAid0302^T, the raw reads were assembled into a 4.96 Mb genome with 967X coverage. For the strain SPO723^T, the assembly was performed using the SPAdes v3.15.0 assembly tool [30], resulting in a 4.29 Mb genome with 145X coverage. Both genome sequences have been deposited in the NCBI GenBank under accession numbers JARRCW00000000.1 and JAWXVZ000000000.1. To facilitate a comparative taxogenomic analysis, genomes from other species within the family Stappiaceae were selected as reference targets. Genome assembly statistics, including the number of contigs, total bases, N50, and G + C content, were gathered using the GAAS-toolkit. The quality of the genome, completeness, and contamination were evaluated using CheckM v1.2.3 [31]. Genome annotations were performed using Prokka version 1.14.6 [32], which identified key features such as coding DNA sequences (CDS), rRNA, CRISPR repeat regions, tRNA, and tmRNA.

Genome-based phylogenetic analysis

For the genome-based phylogenetic analysis, a core gene phylogeny was constructed for each genome based on Prokka annotations [32]. To provide a comprehensive comparison, two additional species from the order Rhizobiales—Martelella mediterranea DSM17316^T and Hyphomicrobium nitrativorans NL23^T—were included as outgroups. The genome-based phylogenetic analysis was performed using PhyloPhlAn v3.1.68 [33], and the detailed steps were carried out using its built-in tools as follows. The core gene phylogeny was established using 400 universal protein markers extracted from the genomes, which were then sorted using USEARCH v5.2.32 [34]. The selected universal proteins underwent multiple sequence alignment using MUSCLE v3.6 [35]. A phylogenetic tree was constructed from this alignment using FastTree v2.1.10 [36], with 1,000 bootstrap replicates to ensure robust statistical support for the tree’s topology [37]. The constructed genome-based phylogenetic tree was evaluated using clustering trees based on ANI and AAI distance matrices. The ANI and AAI score matrices were obtained using the genomic index calculation methods described later. Tree-building was performed from the distance matrices using the R package ape. A tree was constructed using the balanced minimum evolution algorithm implemented in the FastMe [38] function of ape [39], with the nearest neighbor interchange and subtree pruning and regrafting parameters enabled.

Genus and species delineation based on genomic indices

In genus and species delineation, genomic indices such as Average Amino Acid Identity (AAI), Percentage of Conserved Proteins (POCP), digital DNA-DNA Hybridization (dDDH), and Average Nucleotide Identity (ANI) are widely used standards in taxogenomics. In this study, ANI and dDDH were employed to delineate species, with ANI calculated using the OrthoANI-usearch tool [40] and dDDH computed with the Genome-to-Genome Distance Calculator (GGDC) v3.0 [41]. For genus delineation, AAI and POCP were utilized. POCP was assessed by comparing genomes using BLASTP, where proteins with an E-value below 10^-5 and an identity of 50% or higher were considered conserved proteins, and the ratio of conserved proteins was calculated [42]. AAI was calculated using the AAI calculator from the Kostas Lab [43], with repetitive clustering to determine a clear reference value for genus delineation. These metrics collectively provided a robust approach for evaluating genetic relatedness and facilitating accurate taxonomic classification at both the genus and species levels [22].

Comparative and functional genomics analysis

GFF file outputs generated by Prokka were used as inputs for the pan-genome pipeline Roary software [44] to determine coding sequence (CDS) clusters. A clustering criterion of > 71.92% average amino acid identity (AAI) was applied, as the genomes were relatively distantly related and classified into several distinct genera. The distribution of CDS clusters across bacterial genomes was visualized using Phandango [45]. To identify genes specific to bacteria inhabiting particular ecological niches, canonical correspondence analysis (CCA) was conducted using the vegan package in R. CDS clusters served as species data, while habitat types, such as sponge, coral, flatworm, eelgrass, and sediment, were used as environmental variables. CCA identified CDS associated with specific habitat types. Additionally, CDS exclusively present in a taxon from a specific habitat and absent in all related taxa were considered habitat-specific genes. Functions of CDS were mapped to KEGG metabolic pathways using BlastKOALA [46] with a criterion of allowing a maximum of one missing component. For CDS associated with specific niches by CCA, gene-set enrichment analysis was performed using ShinyGO [47] to infer metabolic pathways responsive to ecological niches.

Isolation and general genomic features of the strains MaLMAid0302^T and SPO723^T

The coral species Scleronephthya gracillimum from which strain MaLMAid0302^T was isolated and the eelgrass species Zostera marina from which strain SPO723^T was isolated represent different hosts, yet both strains share the commonality of originating from marine organisms. When cultured on marine agar, strain MaLMAid0302^T formed circular, smooth, convex, non-transparent, creamy-white colonies, while strain SPO723^T produced light pink to cream-colored circular, convex, and erose margin colonies with an average diameter of 1.0 mm. The genomes of strains MaLMAid0302^T and SPO723^T, sequenced using PacBio 10K (Pacific Biosciences) and Illumina TruSeq technology, respectively, revealed that strain MaLMAid0302^T has a 4.96 Mb genome, while strain SPO723^T has a 4.29 Mb genome. Both genomes showed high quality, with completeness exceeding 95% and contamination below 1%. Additionally, strain MaLMAid0302^T contains seven rRNA sets and has a G + C content of 49.4%, whereas strain SPO723^T has a G + C content of 55.5%. The genomic data confirmed that these strains belong to the family Stappiaceae, prompting further in-depth analysis to explore their novelty, characteristics, and taxonomic significance.

16S rRNA gene sequence-based phylogenetic analysis

Further investigation using 16S rRNA gene sequence-based phylogenetic analysis of strains MaLMAid0302^T and SPO723^T confirmed their affiliation within the genus Pseudovibrio. Both strains exhibited high sequence similarity to various Pseudovibrio species, with strain MaLMAid0302^T showing 97.79% similarity to Pseudovibrio flavus RKSG542^T and strain SPO723^T showing 97.21% similarity to the same strain. Both strains also exhibited relatively high similarity to Roseibium species, indicating potential classification challenges within these genera. Phylogenetic analysis revealed that both strains clustered with members of Pseudovibrio, though strain MaLMAid0302^T did not form a strong clade with P. flavus, and strain SPO723^T instead clustered with P. exalbescens (Fig 1). Additionally, Roseibium species did not form a monophyletic group, and certain Stappia species shared a clade with Hongsoonwoonella zoysiae. These findings suggest that the classification of the family Stappiaceae may need further revision to account for these phylogenetic nuances. The analysis emphasized the complexity of taxonomic classification within the family Stappiaceae, highlighting the need for refined taxonomic approaches, particularly when species exhibit high levels of similarity yet do not always form distinct clades.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Phylogenetic relationships based on 16S rRNA gene sequences of members in the family Stappiaceae.

The tree was reconstructed using the neighbor-joining method inferred by the Kimura2 parameter distance model with the bootstrap values of 1000 replicates. Nodes recovered by three (N-J, MP, and ML) methods with >90% (★), > 70% (●), and > 50% (ϒ), at least one lower than 50% (▲), and recovered by two methods (■) are designated by symbols. Bar, 0.01 substitutions per nucleotide position.

https://doi.org/10.1371/journal.pone.0322500.g001

Phenotypic and chemotaxonomic characteristics

Strains MaLMAid0302^T and SPO723^T are Gram-stain negative bacteria, with strain MaLMAid0302^T being a strict aerobe and strain SPO723^T being a facultative anaerobe. Under transmission electron microscopy (TEM), strain MaLMAid0302^T displayed peritrichous flagella (S2 FigA), whereas strain SPO723^T lacked flagella under scanning electron microscopy (SEM) (S2 FigB), emphasizing their structural differences. Both strains thrived in mesophilic and neutrophilic conditions and required NaCl for growth, consistent with other members of the family Stappiaceae. The polar lipid composition of both strains included PE, PG, DPG, AL, PL, and L, aligning with that of their closest relatives (S3 Fig). Phenotypic and chemotaxonomic tests were conducted and compared with eight closely related species (Pseudovibrio ascidiaceicola, Pseudovibrio denitrificans, Pseudovibrio japonicus, Pseudovibrio axinellae, Polycladidibacter stylochi, Polycladidibacter hongkongensis, Pseudovibrio flavus, and Pseudovibrio exalbescens) (S1 Table), except for the type strain of Pseudovibrio brasiliensis, which was unavailable. All strains, including strains MaLMAid0302^T and SPO723^T, showed no activity for urease, tryptophan deaminase, α-galactosidase, β-glucuronidase, α-mannosidase, and α-fucosidase but exhibited activity for alkaline phosphatase, esterase (C4), leucine arylamidase, and naphthol-AS-BI-phosphohydrolase. None of the strains metabolized L-arabinose, D-mannitol, capric acid, adipic acid, trisodium citrate, L-lysine, L-ornithine, sodium thiosulfate, inositol, D-sorbitol, L-rhamnose, or amygdalin, but all hydrolyzed esculin ferric citrate. Strain MaLMAid0302^T demonstrated clear distinctions from its closest relative, Pseudovibrio flavus, particularly in its inability to produce indole, reduce nitrate to nitrite, or exhibit N-acetyl-β-glucosaminidase activity. It standed out by displaying unique activities of esterase lipase (C8), trypsin, and acid phosphatase, differentiating it functionally from P. flavus. When compared to its closest relative Pseudovibrio exalbescens, strain SPO723^T showed notable differences in its metabolic profile. Strain SPO723^T metabolized L-arginine, gelatin, D-glucose, 4-nitrophenyl-β-D-galactopyranoside, D-mannose, N-acetyl-glucosamine, D-maltose, potassium gluconate, malic acid, phenylacetic acid, 2-nitrophenyl-β-D-galactopyranoside, sodium pyruvate, and gelatin, while notably failing to metabolize D-sucrose. Furthermore, it is distinguished by its unique enzymatic activities of esterase lipase (C8), lipase (C14), valine arylamidase, cysteine arylamidase, α-chymotrypsin, acid phosphatase, and β-galactosidase, setting it apart from P. exalbescens. These results suggest the novelty of strains MaLMAid0302^T and SPO723^T beyond the species level. In the previous 16S rRNA sequence-based phylogenetic analyses, a cluster comprising Pseudovibrio denitrificans, Pseudovibrio japonicus, Pseudovibrio ascidiaceicola, and Pseudovibrio axinellae consistently grouped with a cluster of Polycladidibacter hongkongensis and Polycladidibacter stylochi with over 90% bootstrap recovery across three methods (ML, MP, and NJ). However, no consistent phenotypic differences were observed between these clusters. This highlighted the need for a more refined and precise classification system beyond the current 16S rRNA-based framework. Chemotaxonomically, the fatty acid composition of the two strains was dominated by summed feature 8 (C18:1ω6c and/or C18:1ω7c), a characteristic common in the family Stappiaceae (S2 Table). Strain MaLMAid0302^T showed a higher diversity of fatty acids compared to Pseudovibrio flavus, with a notable difference in the higher proportion of C16:0 and a lower proportion of cyclo C19:0ω8c, C17:0. Some differences were also observed between strain SPO723^T and other Pseudovibrio species, particularly strain SPO723^T had the higher proportion of C17:0, C18:0, but lower proportion of cyclo C19:0ω8c than P. exalbescens. The distinctive fatty acid profile of Pseudovibrio axinellae further differentiated it from other Pseudovibrio species. These findings emphasized the need for continued taxonomic revisions, particularly for the family Stappiaceae, based on both phenotypic and chemotaxonomic data.

Genome information and phylogenomic analysis

S3 Table provides a comprehensive list of 38 genomes collected for future analysis, including strains MaLMAid0302^T and SPO723^T. In terms of genome characteristics, the size of the genomes ranged from 3.68 to 6.90 Mb, displaying a wide spectrum of genetic diversity. Additionally, the G + C content ratio exhibited considerable variation, spanning from 47.0 to 67.1%. The number of coding DNA sequences (CDSs) possessed by each genome also displayed considerable variability, ranging from a minimum of 3,349 to a maximum of 6,402. Although no unique genomic features were observed specific to each genus under the current classification system, the phylogenomic analysis revealed new trends worth exploring.

The phylogenomic analysis conducted using the acquired genomes revealed the potential for dividing the data into at least 17 distinct groups (Fig 2). This grouping was further supported by the clustering tree based on distance matrices (S4 Fig). Notably, the clustering of Stappia albiluteola with Hongsoonwoonella zoysiae, rather than with other species of the genus Stappia, highlighted the necessity for reclassification and further investigation. Additionally, Pannonibacter and Polymorphum showed close proximity on the phylogenomic tree. To ascertain their accurate classification, a thorough comparison of index values was imperative. If Polymorphum and Pannonibacter are indeed determined to be distinct genera through further analysis, it could potentially result in the reorganization of numerous species within Roseibium and Pseudovibrio into new genera, as species within Roseibium and Pseudovibrio exhibited greater phylogenetic distances between each other than those observed between Polymorphum and Pannonibacter. Consequently, this reorganization could profoundly impact our understanding of diversity and relationships within the family Stappiaceae. Furthermore, under the current classification system, the genera Pseudovibrio and Polycladidibacter were intermixed, with strain MaLMAid0302^T forming a deep branch. Since these strains form distinct clusters, further in-depth analyses were necessary. Moreover, regarding the strains MaLMAid0302^T and SPO723^T, while they appeared to belong to the existing Pseudovibrio group, a more in-depth examination was required due to the classification of the genus Pseudovibrio itself into at least four distinct clusters. If the classification is conducted based on the phylogenomic results, Group 1 exhibits a genome size ranging from 4.96 to 6.18 Mb, with a G + C content of 50.3 to 52.7% and 4,551–5,769 CDSs. In contrast, Group 2 displays a genome size of 3.68 Mb, a G + C content of 47.0%, and 3,349 CDSs, while Group 3 possesses a genome size of 3.75 Mb, a G + C content of 53.3%, and 3,534 CDSs. These distinct characteristics are consistently observed across all 17 groups delineated by phylogenomic analysis, demonstrating clear trends in genome features.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The mirror image of the phylogenomic tree of Stappiaceae members.

The left tree represents the previous classification, while the right tree shows the revised classification. The tree was generated using PhyloPhlAn, with groups divided based on AAI values ranging from 71.92 to 72.88. In the right tree, species highlighted in purple represent those reclassified into new genera, while species highlighted in blue indicate newly classified species. Nodes recovered by both distance clustering trees (ANI and AAI scores) are marked with ●, while those recovered by a distance clustering tree are marked with ○.

https://doi.org/10.1371/journal.pone.0322500.g002

Comparative analysis of genomic indices

ANI and dDDH comparisons among Pseudovibrio ascidiaceicola strains Ad26 and AU243 revealed that, despite belonging to the same species, their ANI values did not exceed the species delineation threshold of 95–96%, nor did their dDDH values surpass 70% (S4 and S5 Tables). Similarly, Pseudovibrio denitrificans strains FO-BEG1, DSM17465^T, and JE062 also did not exceed these species classification thresholds. In contrast, the three strains of Pseudovibrio exalbescens (strains CDO22, WB1–6, and DSM16456^T) exceeded the species-level criteria, with ANI values close to 100% and dDDH values above 80%, confirming that they belong to the same species. The ANI comparison of strain MaLMAid0302^T with Pseudovibrio flavus RKSG542^T revealed an ANI value of 71.31, markedly lower than the standard value for species classification. Similarly, the strain SPO723^T showed its highest ANI value with Pseudovibrio exalbescens strains, with a value of 76.13 ± 0.08, which is also much lower than the species delineation criteria. Having confirmed the novelty of strains MaLMAid0302^T and SPO723^T at the species level through the comparison of ANI and dDDH values, we further investigated whether they exhibited marked differences at the genus level by comparing AAI and POCP values (S6 and S7 Tables). To establish the optimal threshold for AAI, we performed repeated calculations based on the AAI results from Park et al., 2022. The possible thresholds obtained were as follows: [60.68–60.78], [71.92–72.88], [75.90–76.69], [76.70–77.32], and [78.42–79.99]. Applying the lowest value [60.68–60.78] led to the grouping of 22 species from Roseibium aggregatum to Hongsoonwoonella zoysiae into one genus. However, this criterion encompassed five genera, including Roseibium, Pannonibacter, Polymorphum, Stappia, and Hongsoonwoonella, rendering it incorrect. On the other hand, applying the threshold [71.92–72.88] resulted in the division of a total of 17 groups. Interestingly, these 17 groups exactly matched the groups delineated in the previous phylogenomic analysis (Fig 2). Notably, Stappia albiluteola and Hongsoonwoonella zoysiae formed one group, indicating a need for internal reorganization within the genera Stappia and Hongsoonwoonella. The remaining species of the genus Stappia belonged to Group 16, those of the genus Pannonibacter to Group 14, and Polymorphum gilvum to Group 15, demonstrating results consistent with the existing classification system. Group 7 included some species, including the type species of Roseibium. Roseibium hamelinense was located in Group 8, Roseibium aquae in Group 9, and Roseibium sediminis in Group 10. These species formed a slightly distant cluster from the one formed by Roseibium, the close species of the type species, further supporting their reasonable division based on the phylogenomic tree. Members of the genus Pseudovibrio and Polycladidibacter, including strains MaLMAid0302^T and SPO723^T, were classified into six distinct groups. Pseudovibrio flavus was assigned to Group 4, while strain MaLMAid0302^T formed Group 5. Pseudovibrio ascidiaceicola, Pseudovibrio denitrificans, Pseudovibrio brasiliensis, Pseudovibrio japonicus, and Pseudovibrio axinellae clustered together in another group. Meanwhile, strain SPO723^T grouped with Pseudovibrio exalbescens, and Polycladidibacter stylochi and Polycladidibacter hongkongensis were placed in separate clusters. Given that AAI thresholds higher than [71.92–72.88] resulted in excessive single-member groups in Stappiaceae, this threshold was considered optimal as it aligned with the results of phylogenomic analysis and provided a reasonable subdivision. We extended our investigation to assess whether the patterns observed in AAI comparisons were consistent when using POCP as another indicator for genus delineation. POCP employs a clear genus classification standard value of 50. However, we found that in most cases, the comparative values exceeded 50, and the individual genera were not distinctly separated. This inconsistency and lack of clear differentiation made POCP unsuitable as a reliable classification standard index for our study. This trend has also been observed in other studies, suggesting that setting thresholds using AAI could be the optimal standard for genus delineation. Ultimately, we established [71.92–72.88]% AAI as the genus classification threshold and reorganized Stappiaceae into 17 genera (Pseudovibrio, Parapolycladidibacter, Polycladidibacter, Astericibacter, Flexibacterium, Nesiotobacter, Roseibium, Aliiroseibium, Laciiroseibium, Soliroseibium, Novilabrenzia, Litoriroseibium, Algilabrenzia, Pannonibacter, Polymorphum, Stappia, and Hongsoonwoonella). The characteristics of these newly organized genera are described in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. General characteristics of the genera in the family Stappiaceae.

https://doi.org/10.1371/journal.pone.0322500.t001

Genome contents and comparative genomics of novel taxa

To conduct an in-depth analysis of the two novel microbial species we isolated, a pangenome analysis was performed using genomes from 16 closely related strains, belonging to the genus Pseudovibrio, Parapolycladidibacter, Polycladidibacter, Astericibacter, Flexibacterium, and Nesiotobacter. The purpose of this analysis is to determine whether the newly restructured taxa exhibit any distinct gene content patterns (Fig 3). A total of 19,139 coding domain sequence (CDS) clusters were identified through Roary analysis, using a clustering criterion of AAI > 71.92%. Among these, 1,239 clusters were shared across all 16 genomes, while the majority (12,083 clusters) consisted of singletons unique to individual genomes of the type strains. Additionally, 5,817 CDS clusters were variably present or absent depending on the taxa. The genus Pseudovibrio appeared to be taxonomically well-defined as the largest number of CDS clusters were conserved among its species-level type stains. The genus Nesiotobacter also showed a substantial number of genus-specific clusters. In addition, type strain genomes of the novel genera Asteriscibacter and Flexibacterium hardly shared CDS among themselves and with other genera. The two flat-worm dweller strains of Polycaldidibacter and Parapolycladidibacter also showed relatively fewer CDS clusters shared with other genera. The four strains belonging to Asteriscibacter, Flexibacterium, Polycaldidibacter, and Parapolycladidibacter carried more than 1,500 CDS clusters not found among their phylogenetic relatives and exhibited a long stretch of clusters unique only in the specific genomes (Fig 3). To identify genes specific for each of the novel taxa, sets of CDS clusters associated with the formation of terminal branches corresponding to Asteriscibacter flavus gen. nov., Flevibacterium corallicola gen. nov., sp. nov., Nesiotobacter zosterae sp. nov. were further annotated with KEGG modules (Table 2). Only a small fraction of more than 1,500 clusters could be identified as KEGG-annotated proteins. While genes related to carbohydrate and amino acid metabolism were notable for the Asteriscibacter genome, genes unique to the strain MaLMAid0302^T of Flexibacterium were diverse as encompassing nitrate metabolism (both assimilatory and dissimilatory), histidine degradation, and virulence factor (dTDP-L-rhamnose and aerobactin) biosynthesis. In the case of the strain SPO723^T, which is Nesiotobacter zosterae sp. nov., capacities of producing a virulence factor and coenzyme F420 were noted. The taxon-specific or genome-specific diversification of those taxa can be interpreted as a result of natural selection specific to their habitats.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Distribution of CDS in genomes of strain MaLMAid0302^T, strain SPO723^T, and 14 related type strains.

The top panel is the frequency of CDS among 16 genomes; the left panel is an AAI-based genomic tree with the number of CDS common in each branch; the bottom panel is the presence of CDS in each genome.

https://doi.org/10.1371/journal.pone.0322500.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. KEGG gene modules annotated among CDS appearing specifically in the genomes of novel taxa.

https://doi.org/10.1371/journal.pone.0322500.t002

Genome variability and habitat-driven functional analysis

Notably, the majority of the family Stappiaceae originated from marine environments. The 38 strains derived from marine environments encompass sediment, water, and a diverse range of hosts, such as corals, marine sponges, flatworms, oysters, dinoflagellates, unicellular protists, and halophytes. Given the potential influence of habitat identified in the previous gene content analysis, we conducted canonical correspondence analysis (CCA) of the presence/absence of genes in the pangenome of the 16 genomes of closely related species, including strains MaLMAid0302^T and SPO723^T. During this stage, we classified genes distinctively make their presence in a habitat-specific manner, (i.e., the presence of a gene in response to environmental forcing). The isolation sources of the 16 genomes could be classified into 6 kinds: sponge, coral, flatworm, seawater, sea sediment, and seagrass (Table 3). With CCA analysis, five axes were identified to explain 36% of the variability in the presence/absence of CDS at each of the six habitat types. This estimate implies that 64% of genome contents in the 16 bacteria were determined by their phylogeny, while 36% of genome contents were present due to environmental forcing. The first CCA axis was the strongest environmental forcing and explained 15% of the total CDS variability, and it separated the deep-sea sediment of the Indian Ocean from other habitats (Fig 4A). The second axis separated the flatworm habitat, which was represented by the two strains from marine flatworms in Hong Kong coastal waters, from the other five kinds of habitats, and it explains 8% of genome variance. (Fig 4A and B). Since genomes of the two flatworm strains diverged enough to form separate genera, genes present in both strains imply supra-generic phylogenetic forcing or the environmental forcing from the flatworm habitat and the geographical location of the Hong Kong Sea. Therefore, environmental forcing from deep-sea sediment of the Indian Ocean and flatworms in the Hongkong Sea appeared to determine 23% of CDS presence/absence. CDS differentially distributed between coral strain genomes and sponge strain genomes constituted the third axis factor in CCA, which explained 5.3% of total CDS variability. It should be noted that the estimate also includes the contribution of genome contents of the eelgrass strain Nst. zosterae SPO723^T, which behaved the same way as those of sponge strains (Fig 4B). The genome contents of strains from the eelgrass habitat and sponge habitat were similar in spite that they are affiliated with 3 different genera or 6 different species. On the contrary, the genome contents of the two coral strains, which belong to two different genera, included some common features not found in those of eelgrass and sponge strains. A possible explanation for the similarity in genome contents of the eelgrass strains to those of sponge strains is that many sponges tend to have algal symbionts in their bodies [73–75]. While coral provides an animal host biosphere, the sponges might provide an algal host biosphere by which environmental forcing from plant physiology might contribute to the genomes of their bacterial commensals.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Characteristics of strains whose genome contents were analyzed for environmental forcing.

https://doi.org/10.1371/journal.pone.0322500.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Canonical correspondence analysis (CCA) ordination diagram showing correlation among genomes, CDS, and habitat types.

Genoms (○), CDS (+), and habitat types (arrows; weed = eelgrass, worm = flatworm, water = seawater, sediment = deep-sea sediment, coral = coral).

https://doi.org/10.1371/journal.pone.0322500.g004

To identify CDS (coding sequences) specific to a habitat, CCA (Canonical Correspondence Analysis) scores for each CDS were analyzed. CDS positively correlated with the CCA1 axis (r = 0.99) were considered specific to the sediment habitat, while those negatively correlated with the CCA2 axis (r = -0.99) were deemed specific to the flatworm habitat. Interestingly, the CCA3 axis separated strains from corals from sponges. CDS negatively correlated with the CCA3 axis (r = -0.33) were treated as specific to the coral habitat, whereas positive values on the CCA3 (r = 0.65 to sponge, r = 0.25 to eelgrass) axis indicated specificity to the eelgrass-sponge habitat. Since CCA scores close to zero suggest neutrality to each axis, we selected CDS with deviations greater than 0.15 from zero for both positive and negative directions on each axis. CDS related to the central dogma, glycolysis, TCA cycles, and amino acid biosynthesis remained in the ± 0.15 score range for all axes. Among the collected CDS for each axis, fewer than 60% were annotated as known genes, with the remainder predicted to encode putative hypothetical proteins of unknown function. As summarized in the CCA columns of Table 4, which is based on the results shown in S8 Table, enrichment analysis identified metabolic pathways specific to sediment, flatworm, and sponge (cofounded with eelgrass) habitats. For sediment-specific CDS, pathways related to the degradation of typical substrates such as pectin and aromatic compounds were identified, along with anaerobic respiration pathways, including nitrogen reduction. In flatworm-specific collections, genes associated with nitrogen reduction and galactose metabolism were found, suggesting collaborative metabolism between the animal host and bacterial commensals under anaerobic conditions. For sponge-eelgrass CDS, degradation of aromatic compounds and galactose metabolism were noted, with flagella production emerging as a distinct genetic feature.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Presence (○or Δ) of CDS clusters specific in bacterial genomes isolated from five kinds of habitats.

https://doi.org/10.1371/journal.pone.0322500.t004

In addition to CCA, an alternative approach to identifying habitat-specific genes involved isolating CDS unique to bacteria from a given habitat. We applied this method by collecting CDS present exclusively in isolates from a specific habitat and absent in bacteria from other habitats, subsequently annotating the CDS into KEGG orthologs (Table 4 – Taxa columns). Among the habitats of deep-sea sediment, flatworm, sponge, coral, and eelgrass, only two yielded annotated CDS sets using this method. Coral strains carried genes linked to sulfur metabolism, indicating potential reliance on bacterial sulfur metabolism for inorganic material processing. Notably, a single eelgrass strain possessed genes for synthesizing dTDP-L-rhamnose, potentially enhancing biofilm formation and adhesion to eelgrass surfaces. For sponge isolates, no genes were uniquely common across all strains and absent in others. Previous studies suggest that sponge symbionts are influenced more by host species-specific factors than biogeographical variability. The absence of sponge-specific genes may result from the immense diversity of sponge strains, given their isolation from different sponge species across various global locations [76–78]. Therefore, we identified genes present in any sponge strain but absent in all other habitats (Table 4 – triangles). This approach revealed five metabolic capacities unique to sponge strains, including flagella production, drug resistance, oxygen stress response, and protein glycosylation. These findings suggest that sponge strains possess host-specific genetic adaptations.

Taxonomic conclusion