Background

Coastal ecosystems are subject to various natural perturbations (e.g., variations of physical, chemical and biological conditions) and increasing anthropogenic pressures (e.g., overpopulation of coastal areas, mariculture, agriculture, maritime traffic). This creates conditions in which allochthonous human pathogens, e.g., introduced via wastewater, ballast water or coastal runoff, and indigenous marine animal pathogens are likely to thrive [1]. As coastal waters are used for recreation and food production, the occurrence of pathogens can have a direct high economic and social impact [2]. Fast and accurate surveillance of potential pathogens is therefore crucial to predict the risk of disease outbreaks and to understand disease-promoting environmental conditions.

Advanced molecular approaches and next-generation sequencing (NGS) led to the widespread use of culture-independent monitoring methods, such as high throughput sequencing of marker genes (i.e., amplicon sequencing) [3]. However, in the case of many bacterial pathogens, these approaches are not sufficient for their accurate identification. The decreasing costs of whole genome sequencing (WGS) and the development of new bioinformatics tools for genomic analyses provide new opportunities not only to accurately detect pathogens, but also to gain valuable insights into their functional potential [4,5,6]. Whole genome analyses were successfully applied in epidemiological studies, revealing sources, means of transmissions, and outbreak dynamics of non-marine bacterial pathogens [7, 8]. Detecting pathogens at different spatial-temporal scales in different ecosystems and analyzing their functional potential using their complete genomes can provide answers to important ecological questions, such as adaptation to different ecological niches, pathogen-host interactions and dispersion of functional genes between different strains [9, 10].

The long-established approach of obtaining a pure (axenic) culture of the strain of interest, followed by DNA extraction and high-throughput sequencing, is still probably the best way to obtain a high-quality bacterial genome [11]. However, obtaining an axenic bacterial culture from environmental samples is often challenging since contamination can occur during any of these steps, even when strict microbiological standards and aseptic techniques are applied [12, 13]. Therefore, non-axenic cultures represent a practical challenge to obtain a high-quality genome of a specific bacterium.

One of the globally monitored marine bacterial lineages, which includes strains associated with human diseases and connected with mass mortality events of economically and ecologically important marine organisms, is the genus Vibrio [14,15,16,17]. This genetically diverse lineage is part of the ambient microbiome in estuaries, coastal seawater, deep sea, and even marine sediments [17, 18]. Although Vibrio spp. usually comprises a minor fraction of the bacterial community (< 1%) [19, 20], it can become abundant under specific environmental conditions [21, 22]. For example, the increase in abundance of Vibrio spp. was related to the rise of seawater temperature and the decrease in seawater salinity [20]. Higher seawater temperatures were also associated with higher expression of its virulence genes in Vibrio harveyi [23]. This relationship is important in the context of projected future changes of coastal habitats (e.g., increase of seawater temperatures, droughts, sea level rise) [17, 24,25,26,27]. However, due to high genomic and phenotypic similarity, conventional analyses relying on marker genes or phenotypes frequently encounter challenges in distinguishing between closely related pathogenic and non-pathogenic Vibrio lineages [28,29,30], making it challenging to monitor and control Vibrio-associated disease outbreaks [31, 32]. In addition, as Vibrio-associated infections have become more frequent in recent years [25], it is crucial to improve our understanding of the functional and ecological traits of this bacterial lineage.

Previous microbial monitoring, performed by diversity analysis using 16S rRNA gene amplicon sequencing, revealed that Vibrio spp. are members of the core ambient microbiome of the coastal ecosystem in the northern Adriatic Sea [33,34,35], specifically in the shallow, semi-enclosed Gulf of Trieste, characterized by high salinity and temperature fluctuations. However, the resolution of these analyses was too low to accurately determine the taxonomy of the detected Vibrio spp. and to determine whether they are pathogenic. Therefore, our objective was to perform WGS of Vibrio spp. isolates from coastal waters of the Northern Adriatic Sea to acquire their accurate taxonomic identification and to elucidate their functional and pathogenic potential. Genomic analysis of two selected isolates revealed a cross-contamination event between them, where one Vibrio isolate was introduced into the culture of the second isolate during laboratory processing. Having sequencing libraries from both an axenic and non-axenic culture of the same Vibrio isolate allowed us to test the potential for recovering similar high-quality genomes from both cultures. We report here the result of our thorough bioinformatic analysis, which we believe will be useful to our peers dealing with this common analytical challenge.

Results and discussion

Sequencing and genome assembly

To identify Vibrio candidates for WGS, we carried out taxonomic classification of a collection of bacterial isolates from the Gulf of Trieste using Sanger sequencing of ~ 1400 bp of 16S rRNA gene (27F – 1492R). The two selected isolates were affiliated with the Vibrionaceae family (Table 1). However, the 16S rRNA gene did not allow accurate classification at a lower taxonomic rank (e.g., genus), a common problem with marker gene-based analyses of Vibrio lineages [31, 32].

Table 1 Bacterial cultures analyzed in this study and their closest relatives according to 16S rRNA
Full size table

Genomic DNA from cultures of both isolates was sequenced in parallel using long (MinION, Oxford Nanopore Technology) and short-read (Illumina) techniques (Table 2). To assemble bacterial genomes, we implemented the Trycycler workflow, which produces a consensus assembly based on manually selected contig clusters from multiple long-read-only assemblers (methodology described elsewhere [36]). In our case we combined genome assemblies from three different assembly tools (Flye [37], Miniasm+Minipolish [38, 39] and Raven [40]), followed by post-assembly long- and short-read polishing (described in detail in Methods). Genomic sequences assembled from the BF5_0283 culture formed three distinct contig clusters that resulted in a consensus sequence of three circular DNA molecules with a total length of 6.03 Mb (Additional file 2: Fig. S1 A, Fig. 1a). In contrast, a similar approach on sequences from the Mt009 culture did not produce clear clusters (Additional file 2: Fig. S1 B) and the resulting three consensus DNA sequences had a total length of 7.66 Mb. In an attempt to improve the genomic assembly from the Mt009 culture, we implemented two other approaches: (1) short read-first hybrid assembly tool Unicycler - specifically designed for the assembly of bacterial genomes [41] and (2) long-read metagenome assembler metaFlye [42]. Both tools, the Unicycler and the metaFlye, resulted in even larger assemblies (17.48 and 19.92 Mb, respectively) and a higher number of contigs (21 and 188, respectively), compared to Trycycler (Table 2). The Trycycler consensus contigs of both cultures, as well as the contigs in other Mt009 assembly attempts covered approximately 71% of the V. campbellii ATCC BAA 1116 genome (Table 2, metaQUAST calculation). However, the assembly from the Mt009 culture also covered a large fraction of Enterovibrio norvegicus Alg239-V16 and Klebsiella pneumoniae KCTC 2242 genomes, indicating that Mt009 culture was either non-axenic or contaminated.

Table 2 Sequencing information and assembly statistics for BF5_0283 Trycycler and Mt009 Trycycler, Unicycler and metaFlye assembly
Full size table
Fig. 1
figure 1

Refinement of the assembled genomes. Graphical representation of the BF5_0283 Trycycler assembly (a) and the Mt009 Unicycler assembly (b) along with associated data: GC-content, mapping of Illumina reads, 16S rRNA and genes taxonomy. The bins in (b) were manually refined based on differences in mean coverage (mapping of Illumina short reads), differences in GC content and gene taxonomy

Full size image

Tracking the contamination

To investigate whether the contamination of Mt009 occurred already during isolation, we used dedicated polymerase chain reaction (PCR) primers (Vca-hly-5 / Vca-hly-3 and KP878-F / KP878-R) to test for the presence of V. campbellii and K. pneumoniae in the cryo-preserved stock of the initial Mt009 and BF_0283 isolate. The PCR results did not confirm the presence of V. campbellii in the initial cryo-preserved culture stock of Mt009 but did show a weak signal of K. pneumonia (Additional file 2: Fig. S3 A, Fig. S3 B). The presence of E. norvegicus was not tested, due to the lack of published taxa-specific PCR primers. Contamination may have also occurred during the sequencing process (i.e., cross-barcode contamination). However, usually in such cases the contaminated contigs show lower than expected read depth [43], which was not the case in Mt009, as revealed by our further analysis (Fig. 1b, Additional file 1: Table S1). Taken together, these results suggested that the initial Mt009 isolate most likely contained a co-culture of K. pneumoniae and E. norvegicus, while V. campbellii was introduced in the laboratory during secondary cultivation.

Retrieving the Vibrio campbellii genome from the non-axenic culture

To retrieve the genome of interest from the non-axenic culture, we addressed the Mt009 sequencing dataset as a metagenome and performed binning of the assembled contigs. Through a combination of Illumina short-read coverage and G + C content, we were able to manually refine three genomic bins from the Mt009 assembly (Fig. 1b, Additional file 1: Table S1). Based on single copy gene taxonomy, as well as BLASTn search of the 16S rRNA genes, the bins were assigned to V. campbellii (Mt009_b1), E. norvegicus (Mt009_b2), and K. pneumoniae (Mt009_b3) (Additional file 1: Table S2). Unicycler has been previously suggested to retrieve metagenome assembled genomes (MAGs) from metagenomics samples with a combination of short- and long-reads [44]. Indeed, out of the three tested tools, the binned contigs assembled using Unicycler gave the most complete genome and were therefore chosen as the consensus for further genomic analyses of the Mt009 dataset (Additional file 2: Fig. S2).

Comparison of assembled genomes from axenic and non-axenic cultures

Our WGS study resulted in two V. campbellii genomes, the first assembled from the axenic culture (BF5_0283) and the second acquired from a non-axenic culture (Mt009). In accordance with the known structure of the V. campbellii genome, both assembled genomes had two circular chromosomes of 3.7 and 2.1 Mbp (Table 3). The number and length of plasmids varies between different V. campbellii strains [45, 46], and in the case of the assembled genomes both likely contain a putative plasmid of 150 Kbp (Table 3). The particularly high Average Nucleotide Identity (ANI) of 99.98% between the assembled V. campbellii genomes (Additional file 1: Table S6) strongly indicates that most likely there was a cross-contamination event between the two cultures and that we generated the genome of the same V. campbellii strain (BF5_0283), once from an axenic culture and once “salvaged” from a contaminated one.

Table 3 Comparison of genomic features between BF5_0283 and Mt009_b1 assemblies
Full size table

The unexpected cross-contamination allowed us to compare the two assemblies (BF5_0283 and Mt009_b1) to assess the extent of genomic information loss when performing WGS from a non-axenic culture. BLASTn was used for bidirectional best hit analysis (i.e., identification of the pairs of genes in two different genomes that are more similar to each other than to any other gene). We found that 5394 genes (the vast majority of the genes) were similarly represented in both assemblies (Additional file 1: Table 3, Table 4). A total of 24 genes from the BF5_0283 assembly, mostly with unknown functions, was missing in the Mt009_b1 assembly (Additional file 1: Table 3). However, there were 50 genes in the Mt009_b1 assembly not present in the BF5_0283 (Additional file 1: Table 4). The mean coverage of these 50 additional genes was slightly higher than the mean coverage of all genes in the Mt009_b1 assembly (86.31 vs. 83.46, respectively), potentially suggesting that they could be an artifact introduced from the other binned genomes in the non-axenic culture. Nonetheless, this comparison confirmed that we successfully assembled an almost identical genome of a V. campbellii isolated from both, the axenic and contaminated culture.

Genomic comparison to other V. campbellii isolates

To confirm the taxonomic affiliation of assembled genome, we collected all currently available complete representative genomes of Vibrio spp. from NCBI (National Center for Biotechnology Information). In total, 32 representative genomes were collected, and three additional complete genomes representing Vibrio species commonly found in coastal marine environments (Vibrio coralliilyticus, V. mediterranei, V. splendidus) were added (Additional file 1: Table S5). The phylogenetic tree, constructed based on concatenated alignment of 1027 single copy amino acid sequences of orthologous genes, showed that both assemblies are consistently affiliated with V. campbellii (Fig. 2). Further analysis of the BF5_0283 and Mt009_b1 assemblies and 10 complete V. campbellii genomes from NCBI (Additional file 1: Table S7) revealed that the isolated strain clustered with 6 V. campbellii strains in Group 1. In accordance with previous studies, V. cambpellii contains two clusters, Group 1 - isolates originating from aquatic animals and biofilms [47, 48], and Group 2 – represents isolates of oceanic origin [49, 50].

Fig. 2
figure 2

Phylogeny of Vibrio genomes based on single-copy core orthologues. An approximately-maximum-likelihood phylogenetic tree representing 32 Vibrio genomes from NCBI (Additional file 1: Supplementary Table S5) and two genomes assembled in this study (BF5_0283 and Mt009_b1)

Full size image

Pangenome analysis performed with both, BF5_0283 and Mt009_b1 assemblies and 10 reference genomes of V. campbellii resulted in 9318 functional gene clusters (GCs) (Fig. 3). The GCs could be divided into three collections: ‘core genome’ – GCs shared among all strains (39.0% of all GCs), ‘accessory’ - GCs specific to a subset of the genomes clustering into Group 1 and Group 2 (1.9 and 0.6% of all GCs for Group 1 and Group 2, respectively), as well as ‘unique’ - GCs found in individual strains (4% of all GCs for BF5_0283). The ‘core genome’ contained the majority of chromosomal genes of BF5_0283 (~ 70 and 66% of genes on Chr I and Chr II, respectively), indicating their high conservation among V. campbellii (Fig. 4). The core genome of V. campbellii comprised a set of conserved genomic functions, with the most abundant COG categories being signal transduction mechanisms (8% of core genome GCs) and amino acid transport and metabolism (7.6%) (Additional file 2: Fig. S4, Fig. S5 A), which suggests involvement of this lineage in protein turnover in the seawater. The accessory GCs of Group 1, to which our isolate belongs, contained mainly genes connected with intracellular trafficking, secretion, and vesicular transport (14.1% of accessory GCs in Group 1), as well as signal transduction mechanisms (9.1%) (Additional file 2: Fig. S5 B), which may imply the specialization of these strains for intercellular interactions (e.g., with their host). On the KOfam level, we found type VI secretion systems (T6SS) (e.g., K11902, K11899, K11898), accessory colonization factors acfA and acfD (i.e., K10939, K10936), toxin-antitoxin systems genes ccdA and ccdB (K19163, K19164), and the toxin gene hipA (K07154) associated with Group 1 (Additional file 1: Table S9). All these markers are involved in the pathogenesis of Vibrio spp. [30, 51, 52]. It was previously reported that T6SS and HipA might contribute fitness advantages to the AHPND-causing V. parahaemolyticus over competing bacteria and in this way facilitating shrimp infection [53, 54]. T6SS systems are complex systems that inject so-called ‘anti-bacterial’ and ‘anti-eukaryotic’ effector proteins into target cells, targeting both, eukaryotic hosts and bacterial competitors [55, 56], while the serine/threonine protein kinase HipA is a toxin that causes inhibition of cell growth [57]. In contrast to previous reports, our results did not reveal functions related to antibiotic transport and galactose metabolic process associated with Group 1 [29]. The accessory GCs collection for Group 2 mainly contained genes related to transcription (8.1% of the genes in the accessory collection of Group 2), inorganic ion transport and metabolism (6.5%), and general function (6.1%), representing the three most abundant COG categories (Additional file 2: Fig. S5 B). On the KOfam level, we found sensory rhodopsin (i.e., K04643) (Additional file 1: Table S9), which suggests mixotrophy of Group 2. Although the presence of this gene has been previously described in V. campbellii BAA-1116 [58], we found that it is specific to all genomes in Group 2. We hypothesize that since these isolates originate from ocean waters, they probably undergo adaptation to survive in nutrient-poor environments. Interestingly, ‘unique’ GCs of our isolate accounted for ca. 40% of all genes present on its putative plasmid, with only a small portion of the plasmid-associated genes being part of the V. campbellii core genome and none associated with accessory genes of the Group 1 cluster (Fig. 4).

Fig. 3
figure 3

Vibrio campbellii pangenome. The radial layers present genomes ordered according to their phylogenetic relationship based on maximum likelihood phylogenetic tree constructed from single copy orthologous genes. ANIb values are shown as heatmap with high similarity in red and lower similarity in gray color. The dark and light color of bars in radial layers show presence and absence, respectively, of 9318 gene clusters. Gene clusters (GCs) are organized based on their distribution across genomes using Euclidean distance and Ward ordination (gene tree in the center of pangenome). The “Core” collection corresponds to GCs containing genes from all genomes. The “Accessory Group 1” and “Accessory Group 2” refers to genes characteristic to Group 1 and Group 2. The “Unique” collection refers to genes unique for the new assembled genome in this study (BF5_0283 and Mt009_b1). Collections are marked in the outermost radial layer. Colored squares below the ANIb value heatmap provide additional data for each isolate

Full size image
Fig. 4
figure 4

Proportion of core genome, accessory Group 1 (G1) and Unique GCs in the chromosomes and the plasmid of BF5_0283

Full size image

Exploring the plasmid of the novel Vibrio campbellii genome

We compared the sequence of the identified putative plasmid of BF5_0283 to previously characterized plasmids in the plasmid database PLSDB (v. 2021_06_23_v2) (max_pvalue 0.1, max_distance 0.2) [59]. According to Mash distances, the most closely related plasmids were found in V. campbellii strains: plasmid pLA16–1 in strain LA16-V1 (Mash distance 0.1168), plasmid pLMB143 in strain LMB29 (0.1354), plasmid pVCGX3 in strain 20130629003S01 (0.1354), and plasmid pLA16–4 in strain LA16-V1 (0.1370) (Additional file 1: Table S11). More distant plasmids were found in V. parahaemolyticus, V. owensii, and other V. campbellii genomes. The majority of related plasmids were isolated from the host organism Penaeus vannamei (52%), and some were isolated from AHPND infected shrimps (23%) (Additional file 1: Table S11) [47, 48, 60]. Although only parts of the putative plasmid were similar to other V. campbellii plasmids, these shared genes such as the anti-restriction protein gene ardC and CRISPR Csa3 system (Fig. 5). The presence of CRISPR Csa3 system suggests that these plasmids could provide a defense function, since this system is involved in protecting the cell against foreign DNA, such as bacteriophages [61,62,63,64].

Fig. 5
figure 5

BRIG visualization of comparative sequence analysis of the plasmid. BLASTn was used to align plasmid sequences from newly assembled genomes BF5_0283 and Mt009_b1 with all V. campbellii plasmids in PLSDB. Circles from outside to inside are plasmid sequences of BF5_0283, Mt009_b1, LMB29, 20130629003S01, BoB-90, 170,502, LA16-V1, DS40M4, ATCC-1, ATCC-2, CAIM 519. In cases where the genome contained more plasmids, all plasmids were aligned. The intensity in colors indicates 100% (higher intensity), 70% (medium intensity) and 50% (low intensity) identity. The outermost circle shows protein coding genes with hits in V. campbellii plasmids from PLSDB (gray), and selected unique genes of newly assembled plasmids (red). Complete annotation of common genes is available in Additional file 1: Table S12

Full size image

The unique fraction of the putative plasmid of BF5_0283 comprised two complete sets of genes of the Type I restriction-modification system (Fig. 5). This is surprising, since previous studies reported that many individual genes involved in the Type I R-M system are usually present on plasmids, but only few complete systems [65]. The Type I R-M system consists of genes for methyltransferase (hsdM) that specifically methylate DNA, restriction endonuclease (hsdR) cleaving DNA that has not been properly modified (i.e., methylated), and genes for specificity (hsdS) determining the recognition sequence of restriction and modification activities [66]. The presence of this system has been previously connected with a ‘selfish-behavior’ of the plasmid carrying the R-M gene complex, since the loss of the R-M gene complex can lead to cell death, because the balance of methyltransferases and restriction endonucleases in a cell is disturbed [67, 68]. This suggests that plasmid containing R-M genes cannot be eliminated from the cell or displaced by the plasmid lacking this gene complex.

Interestingly, the ardC genes observed in the putative plasmid of BF5_0283 have anti-restriction activity against the type I R-M system, which enables the plasmid to overpass the R-M systems of the recipient cell once they are transferred by conjugation [69]. In that way, plasmids broaden their host range [70]. This, together with numerous transposases playing a role in horizontal gene transfer [71] suggests that there is potential for propagation of plasmid genes in coastal systems, as it was previously shown for Vibrio spp. [5, 72, 73].

Conclusions

Our study highlights the power of whole genome sequencing for accurate taxonomic identification and unraveling the pathogenic potential of emerging environmental pathogens. In fact, our analysis revealed that the genome of Vibrio campbellii isolated from the northern Adriatic Sea carries genes for T6SS type VI secretion systems, known for their role in pathogenesis and interbacterial antagonism, as well as novel putative Vibrio plasmid, both of which should be further explored. Besides, our approach to salvage a high-quality genome of the bacteria from a contaminated culture using state-of-the-art bioinformatics tools and a sufficient sequencing effort can be implemented when dealing with common issues of non-axenic cultures. This approach can be also applied, for example, to study bacteria that exhibit co-culture dependence (e.g., Prochlorococcus) [14, 15] or to study interspecific interactions [16] or to reduce the time and costs of analyses, such as proposed for genomic epidemiology studies [17]. Last but not least, high quality genome sequences can also serve as baseline for the development of new monitoring approaches (e.g., more specific primers for more reliable monitoring than with the 16S rRNA approach), which will allow us to track and control propagation of emerging pathogens in marine coastal ecosystems. This is crucial to constrain disease outbreaks, which will help maintaining ecosystem services in the future.

Methods

Isolation, culture condition and DNA sequencing

For bacterial isolation, a defined volume of seawater was spread on modified ZoBell solid agar media [74] and incubated in the dark at 21 °C by gently agitating for 48 h. Single colonies were clean streaked once and inoculated into ZoBell liquid medium and incubated in the dark at 21 °C for 24 h. Bacterial genomic DNA for 16S rRNA Sanger sequencing was extracted immediately, with a modified Chelex-based procedure [75], amplified with universal primers 27F and 1492R, and sent for Sanger sequencing at Macrogen Inc. (Accession number JX864957 and Additional file 3). Both isolates were stored at the culture collection of the Marine Biology Station Piran, Slovenia (in 30% glycerol at − 80 °C).

Each isolate from the cryo-preserved stock was re-grown on ZoBell agar plates (at 24 °C for 72 h in the dark). A single colony was picked from the agar plate, inoculated into 6 mL of ZoBell liquid medium, and incubated at room temperature in the dark on a shaker. For each isolate, four 1 mL replicates of the liquid culture were pelleted by centrifugation at 4000x g for 3 min. The bacterial pellets were then shipped on dry ice to the sequencing facility (Microsynth AG, Balgach, Switzerland) where high molecular weight DNA was extracted. The DNA was sequenced using the long-read MinION ONT (Oxford Nanopore Technologies, Oxford, United Kingdom) technique and complemented by short-read paired-end (2 × 75 bp) sequencing on Illumina NextSeq (Illumina, San Diego, CA, USA).

Contamination check using polymerase chain reaction

Cryo-preserved stocks were re-grown using the same culturing conditions as described above. Bacterial genomic DNA was isolated with a modified Chelex-based procedure [75] and amplified by PCR reaction using universal 16S rRNA bacterial primers (27F and 1492R) or species-specific primers (Vca-hly-5 / Vca-hly-3 targeting haemolysin (hly) gene of V. campbellii and KP878-F / KP878-R targeting transferase gene of K. pneumoniae) (Table 4). A total of 50 μL of PCR mixture was prepared for each isolate and each primer pair with a suitable primer concentration (0.5 μM for universal primers, 0.25 μM for species-specific primers), 1X Tris KCl-MgCl2, 1.5 mM MgCl2, 0.2 mM dNTP and 0.3 U Taq polymerase. The PCR protocol was as follows: 5 min of initial denaturation at 94 °C, 30 cycles of 30 sec denaturation at 94 °C, 30 sec of primer annealing at 54 °C, 30 sec for extension at 72 °C, followed by a final extension for 5 min at 72 °C.

Table 4 Selection of species-specific primers for PCR reaction
Full size table

Genome assembly

Raw reads were quality-filtered using the Filtlong tool for long reads (keep percent 75%) [78] and fastp (default thresholds) [79] for short reads. Assembly of isolate BF5_0283 was performed using the Trycycler tool [36] combining multiple separate long-read assemblies of the same genome. Assemblies were created by subsampling 12 long-read sets assembled using the assembling tool Flye [37], Miniasm+Minipolish [38, 39] and Raven [40]. Trycycler contigs tree was visualized using iTOL (v 6.8.1) [80]. Long-read polishing of the consensus long-read assembly was done with Medaka (v. 1.4.4) [81] and short-read polishing with Pilon tool (v. 1.24) [82].

Mt009 was assembled using three different methods. First, Trycycler was used for long-read assembly followed by long- and short-read polishing as described for BF5_0283. Second, the genome was assembled using Unicycler short-read-first hybrid assembly tool [41] which uses SPAdes for short read-assembly [83], followed with Miniasm long-read plus contig assembly and Racon polishing [84]. Third, metaFlye [42] was used for long-read assembly. Sequences shorter than 10 Kb were removed. Quality assessment of all assemblies was done with metaQUAST tool (v. 5.0.2) [85] without providing reference genomes.

Genome annotation and refinement

The assembled genomes were first annotated using Anvi’o v. 7 [86]. Briefly, for Anvi’o annotation we used ‘anvi-gen-contigs-database’ to construct the contig database for each assembly, which uses Prodigal [87] to identify ORFs in each contig. We ran HMM (Hidden Markov models) with ‘anvi-run-hmms’ and assigned functions to the genes by alignment against the COG database [88, 89] with the ‘anvi-run-ncbi-cogs’ program. We also used ‘anvi-run-kegg-kofams’, which uses hmmsearch to find hits from KOfam, database of KEGG Orthologs (KOs) [90]. Gene taxonomy was annotated with kaiju classifier [91] and ‘anvio-run-scg-taxonomy’. Short-read mapping to the assembled genome was done using bowtie2 [92]. An anvi’o profile database was generated storing coverage statistics using ‘anvi-profile’ with ‘--cluster-contigs’ option. We manually refined the bins in the Mt009 assembly to identify bacterial genomes in this sample within the ‘anvi-interactive’ interface. The taxonomy of each bin was assigned by exporting and alignment all 16S genes and by inspecting the taxonomy of single-copy genes with ‘anvi-summarize’. We used ‘anvi-split’ to split the Mt009 sample into three separated genomes (Mt009_b1, Mt009_b2, Mt009_b3).

Comparative genomics analysis

For comparative functional analyses of the V. campbellii genomes assembled in this study (BF5_0283 and Mt009_b1) and the reference Vibrio spp. genomes, we annotated the assemblies on the RAST Server [93]. This was done by importing fasta files into the web-based annotation service, running annotation (RASTtk annotation scheme). To compare BF5_0283 and Mt009_b1 assemblies, Bidirectional Best Hits (BBH) were calculated in Seed Viewer [94]. The exported annotated genomes in GeneBank format were imported into Anvi’o with ‘anvi-script-process-genbank’ and a contig database was created using ‘anvi-gen-contigs-database’ with ‘--external-gene-calls’ flag. The annotation was completed with the COG and KOfam database as described above.

Vibrio spp. genomes were downloaded from NCBI and annotated (with RAST tool and Anvi’o) as described above. To construct the phylogenetic tree based on orthologous genes, we extracted and aligned genes from single-copy gene clusters present in all 37 genomes with ‘anvi-get-sequences-for-gene-clusters’ program. Nucleotide positions missing in more than 50% of sequences were removed (with ‘trimal’). The amino acid translated phylogenetic tree was constructed with IQ-TREE (v. 2.0.3) (options -m WAG, −bb 1000, to specify WAG substitution model and the number of bootstrap replicates to 1000 – recommended values) [95]. The resulting phylogeny was subsequently rooted and edited in FigTree (v 1.4.4) [96]. To explore similarities across genomes of Vibrio species, the average nucleotide identity (ANI) value was calculated with ‘anvi-compute-genome-similarity’ using Phyton module PyANI [97].

The pangenome was created to compare genomes assembled in this study with 10 complete genomes of V. campbellii retrieved from NCBI. FASTA files of the public genomes were downloaded and processed and annotated as described for BF5_0283 and Mt009_b1 (using RAST and Anvi’o). The pangenome was constructed following the pangenomics workflow in Anvi’o v. 7.1 [98]. Briefly, ‘anvi-gen-genomes-storage’ was used to create the genome database and the ‘anvi-pan-genome’ program that uses BLASTp for amino acid sequence similarity search, and the MCL algorithm to identify gene clusters in the amino acid sequence similarity results [99]. The inflation parameter was set to 10 to increase the sensitivity of the algorithm, suggested for closely related genomes [99]. ANI was calculated with ‘anvi-compute-ani’ using the PyANI program. Genomes in the V. campbellii pangenome were organized based on the single-copy core genes tree, constructed with IQ-TREE [95]. Gene clusters were grouped into core bin containing gene clusters present in all genomes, accessory bins with gene clusters present in genomes belonging to a specific group and unique bins with gene clusters specific to the genomes assembled in this study. Data were exported with ‘anvi-summarize’. Heatmaps of genes with COG annotations in different collections, and barplots of genes with COG annotations on chromosomes and the plasmid were plotted in R [100] using ‘tidyr’ [101], ‘dplyr’ [102], ‘ggplot2’ [103] and ‘forcats’ [104] packages.

We identified functions enriched in V. campbellii Group 1 or Group 2 in our pangenome with the program ‘anvi-compute-functional-enrichment-in-pan’. The program calculates functional enrichment scores using the Rao score test for equality of proportions. False detection rate correction is applied to the p-values to account for multiple tests.

Plasmid exploration and gene map visualization

For plasmid exploration, the sequence of the plasmid from the BF5_0283 genome was used. The similarity comparison of the novel assembled plasmid with reference plasmids was done by Mash distance search in publicly available plasmid sequences (PLSDB) [59]. The distance ranges from 0 (identical) to 1 (highly unrelated). We limited the search with a maximum p-value of 0.1 and a maximum distance of 0.2. To explore which reference plasmids contain genes similar to our plasmid, we extracted gene sequences with ‘anvi-get-sequences-for-gene-calls’ and searched with BLASTn search in PLSDB with the default parameters: minimal identity 80% and minimal query coverage/HSP 90%. Nucleotide alignment and visualization of the plasmid assembled in this study and PLSDB were performed using BRIG v 0.95 [105]. All final figures were edited using the vector graphics editor Inkscape v 1.1 [106].