Vibrio campbellii LMB29 was isolated from skin ulcer sample of cage-cultured red drum in Shenzhen, China by 2012. Briefly, diseased tissue mixtures were plated onto thiosulphate citrate bile salts sucrose (TCBS) agar for incubation under 30°C. 24–36 h later, morphologically uniform colonies of Vibrio were passaged onto fresh media to obtain pure colonies. The strain identity of individual isolate was confirmed with optimized multilocus sequence analysis (MLSA) (Gabriel et al., 2014). The predominant strain, V. campbellii LMB29, was grown in TSB media at 30°C and its genomic DNA was extracted by using the Cetyltrimethyl Ammonium Bromide (CTAB) method with minor modification (Healey et al., 2014). The quantity and quality of genomic DNA were assessed by using a Qubit 2.0 Fluorometer (Thermo Scientific, USA) and a NanoDrop 2000 Spectrophotometer (Thermo Scientific, USA), respectively. The whole genome sequencing was performed by a commercial vendor (Personalbio, Shanghai, China) by using Illumina MiSeq (400 bp inserts library with 251 bp paired-end sequencing; Illumina, San Diego, CA, USA) and PacBio RS II platform (10 kb inserts library; Pacific Biosciences, Menlo Park, CA, USA).

The trimmed sequences from Illumina MiSeq were assembled using Newbler version 2.8 (454 Life Sciences, Branford, CT, USA) and PacBio sequencing reads were de novo assembled with hierarchical genome-assembly process (HGAP) (Chin et al., 2013). Sequencing data were combined, and further polished with Pion (Walker et al., 2014) to obtain the complete sequences. The prediction of open reading frames (ORFs) and their annotations were performed using Glimmer 3.0 (Delcher et al., 2007). The genome annotation was further corrected and confirmed with the NCBI Prokaryotic Genome Automatic Annotation Pipeline (PGAAP) (Angiuoli et al., 2008) and DOE-JGI Microbial Genome Annotation Pipeline (Huntemann et al., 2015). The putative virulence factors of the V. campbellii LMB29 strain were predicted using the basic local alignment search tool (BLAST) in the Virulence Factor Database (VFDB; http://www.mgc.ac.cn/VFs/main.htm) (Chen et al., 2016). Contigs were subsequently queried with Resfinder (Zankari et al., 2012) and CARD (McArthur et al., 2013) to perform antimicrobial resistance genes analysis.

The strain was originally identified as V. campbellii based on 16S rRNA sequencing and its identity was further confirmed with comparative phylogenetic tree analysis using MEGA7 (Kumar et al., 2016) with various 16S rRNA sequences of the genus Vibrio. The phylogenetic tree was constructed from the aligned sequences using Kimura 2-parameter model with the neighbor-joining method, bootstrapped 1,000 times via MEGA7 software. The V. campbellii LMB29 assembled contigs were also submitted to JGI IMG/MER database (Chen et al., 2017) (https://img.jgi.doe.gov/cgi-bin/mer/main.cgi) to perform the Average Nucleotide Identity (ANI) calculations (IMG Genome ID: 2718217691). In addition, 1,045 Vibrio complete genomes (Supplementary Table 1) were retrieved from JGI (all available Vibrio genome in JGI) to perform genome comparisons (Figure 9) and functional profile analysis (Figure 10) based on the Clusters of Orthologous Groups (COGs) database (Tatusov et al., 2000). Bacterial genome, plasmids and ICEs sequences alignments were performed using Mauve with progressive Mauve option (Darling et al., 2004). IslandViewer 4 was applied to predict genomic islands in V. campbellii LMB29 with default settings (Bertelli et al., 2017). DEG (Database of Essential Genes) database (version 14.7) was used to assess gene essentiality of ORFs predicted in LMB29 with BLAST search (Luo et al., 2014).

Fathead minnow (FHM) epithelial cells were routinely cultured in M199 medium supplemented with 10% (v/v) fetal bovine serum (FBS, Gibco) at 28°C. For LDH assays, FHM cells were seeded into 96-well plates and grown overnight to 90% confluence of cell monolayers. Prior to infection, growth media was replaced by 110 μL of serum-free TC199 medium per well, and cells were infected with V. campbellii LMB29 at a multiplicity of infection (MOI) of 5 over a 6 h time course. At the indicated time point, the 96-well plates were centrifuged at 3,200 × g for 2 min, and a 100 μl aliquot of the supernatant was removed for measuring LDH release with the Cytotoxicity Detection kit^PLUS per manufacturer's instructions (Roche).

The deletion of arr-9 gene was performed by allelic exchange. Briefly, primers pSW7848-F/R were used to linearize the suicide vector pSW7848, and the primers Arr9-up-F/R and Arr9-down-F/R (Supplementary Table 2) were used to amplify two fragments flanking the coding sequence of arr-9. The two fragments, respectively incorporated 18 and 17 bp overlapping sequences with vector pSW7848, which also contained 17 bp overlapping sequences between them. The linearized vector and flanking fragments were assembled according to the instruction of the ClonExpress MultiS One Step Cloning Kit (Vazyme Biotech, China), generating pSW7848_Δarr-9. The resultant plasmid was transformed into chemical competent cells of E. coli GEB883 and then transferred into V. campbellii LMB29 by conjugation. The arr-9 deletion strain (LMB29Δarr-9) was selected on TCBS plates containing 20 μg/mL chloramphenicol and 0.2% glucose followed by a 0.2% arabinose selection process. Gene deletion was then confirmed by PCR and sequencing with primers Arr9-outer-F/R.

To complement the arr-9 gene, primers pMMB207-F/R were used to linearize the expression vector pMMB207 and primers Arr9-F/R were used to amplify the complete arr-9, including the ORF, the promoter and the terminator. The resulting arr-9 amplicon was assembled into the linearized expression vector pMMB207 by using the ClonExpress II One Step Cloning Kit (Vazyme Biotech, China). The recombinant pMMB207_arr-9 plasmid was respectively introduced by conjugation into the arr-9 deletion strain and wild type LMB29, resulting in LMB29Δarr-9:pMMB207_arr-9 and LMB29:pMMB207_arr-9. The empty pMMB207 plasmid was included as a control.

Rifampicin susceptibility of the wild type LMB29, the arr-9 deletion mutant and the complementation strain were tested with the E-test (BioMerieux SA, France) according to the manufacturer's guidance. Briefly, overnight cultures were diluted 1:25 into fresh TSB, and each 20 μL of dilution was then plated on TSB plates using sterile swab. The antibiotic strips were individually placed on the dried plates and the Minimum Inhibitory Concentration (MIC) was determined after an incubation of 48 h at 30°C. The MIC was interpreted the value at which the elliptical inhibition zone intercepted the scale on the E-test strip.

A total of 2,689,618 high-quality paired reads of 634,307,780 clean bases were generated through the Illumina MiSeq pipeline which represents a 102X sequencing coverage. The PacBio SMRT sequencing, with a N50 of 7,484 bp, was also introduced to close the gaps among assembled contigs. Combined sequencing analysis revealed that the complete genome of V. campbellii LMB29 includes two circular DNA chromosomes and four megaplasmids. Chromosome I consists of 3,486,048 bp with a GC content of 45.54% containing 3,072 predicted ORFs, 110 tRNAs and 34 rRNAs. Chromosome II consists of 2,214,480 bp with a GC content of 45.23% containing 2,004 predicted ORFs, 15 tRNAs and 3 rRNAs. Plasmid pLMB157 consists of 157,074 bp with a GC content of 40.37% containing 160 predicted ORFs. Plasmid pLMB143 consists of 143,114 bp with a GC content of 42.30% containing 165 predicted ORFs, and 2 tRNAs. Plasmid pLMB99 consists of 99,317 bp with a GC content of 39.57% containing 146 predicted ORFs. Plasmid pLMB96 consists of 95,945 bp with a GC content of 40.76% containing 148 predicted ORFs. Circular genome maps were generated for individual chromosome and plasmid based on bioinformatics analysis (Figure 1).

The predicted ORFs are further classified into COGs functional groups (2,154 ORFs in total; Figure 2A). According to the COG categorization analysis, based on gene counts, the abundant groups include class of R (486 ORFs, general function prediction only), class of E (363 ORFs, amino acid transport and metabolism), class of S (346 ORFs, function unknown), class of K (339 ORFs, transcription), and class of T (283 ORFs, signal transduction mechanisms; Figure 2A). The genomic sequences were used to against the DEG database to perform gene essentiality analysis. A total of 2,871 ORFs were predicted as essential genes in V. campbellii LMB29 with multiple genes were predicted on genomic islands with horizontal transferability (Supplementary Table 3).

Genome tree analysis was performed on 15 Vibrio species with 16S rRNA sequences alignment and strain LMB29 shows very close phylogenetic distance to V. campbellii ATCC BAA-1116 (Figure 2B). ANI values indicate that strain LMB29 shares a 96.66% similarity with the ATCC strain BAA-1116 based on a 75% AF (average fraction) calculations, but multiple gene rearrangements also happened in the chromosomes (Figure 3). Compared to the single plasmid (~89 kb; NC_022271.1) in V. campbellii ATCC BAA-1116, the V. campbellii LMB29 has a unique plasmid profile with four megaplasmids (Figure 3) which indicates the bacterial evolution of V. campbellii.

The putative virulence factors in V. campbellii LMB29 were analyzed by inquiring the VFDB database (Chen et al., 2016; Table 1). T3SS, which is usually involved in Vibrio pathogenesis, was found on chromosome I. Importantly, the recently identified T6SS (Coulthurst, 2013; Ho et al., 2014) was predicted in strain LMB 29 with T6SS1 (1,649,571–1,672,337 region) on chromosome I, T6SS2 (850,171–881,172 region) and T6SS3 (936,359–959,932 region) on chromosome II (Table 1). T6SSs have been implicated in eukaryotic cell targeting and virulence (Coulthurst, 2013; Ho et al., 2014), and is also believed to have anti-bacterial properties in Vibrio (Church et al., 2016). In addition, the tlh gene encoding a thermolabile hemolysin (TLH) of V. parahaemolyticus was found on chromosome II in strain LMB29. TLH is expressed by all clinical and environmental strains of V. parahaemolyticus (Bej et al., 1999) and it may have similar biological functions similar to these of the TDH (thermostable direct hemolysin) and TRH (TDH-related hemolysin) toxins, playing a key role in Vibrio infections (Wang R. et al., 2015). We also performed LDH release assay using FHM cells (Liu et al., 2016) to test the cytotoxicity activity (Figure 4). Cell death was observed after 3 h post-infection and up to 80% of the cells are lysed within 6 h (Figure 4). Nuclear condensation and fragmentation, based on one DNA staining assay, was confirmed to contribute to the cell death after the infection with V. campbellii LMB29 (Figure 5). Taken together, we conclude here that V. campbellii strain LMB29 is pathogenic which is consistent with our clinical observations in Shenzhen, China.

Thirty-seven genes in V. campbellii LMB29 were observed showing >45% identity to well-characterized antimicrobial resistance genes in reference databases Resfinder and CARD (Table 2). Multiple genes encoding aminoglycoside phosphotransferase [strB, strA and APH(3″)-Ib], tetracycline resistance genes [tet(A), tet35 and tet34] and blaACARB-18 (a beta-lactamase coding gene) were predicted in the LMB29 genome. These results are consistent with our antibiotic susceptibility assay data for strain LMB29 by showing its resistance to ampicillin, amoxicillin, rifampicin, tetracycline, streptomycin, oxacillin, lincomycin, teicoplanin, and vancomycin, some of which are classified as human medically-important antibiotics.

Moreover, in plasmid pLMB157, we named a gene arr-9 which shows 99% identity to a previously described rifampin ADP ribosyltransferase (Arr) (Rodriguez-Blanco et al., 2012). We demonstrated that the arr-9 is responsible for rifampin-resistance phenotype of LMB29 since the deletion mutant was showing susceptibility to rifampin (Figure 6B), and complementation of this gene successfully restored the resistance phenotype (Figure 6). We also constructed the phylogenetic tree for arr-9 with all previously reported rifampin ADP-ribosyltransferase and arr-9 shows a <10% identity to all other genes (Figure 7). This gene was originally reported in integrating conjugative elements (ICEs) in Vibrio splendidus and Vibrio alginolyticus (Rodriguez-Blanco et al., 2012), and the current study is the first report showing its presence in a mobilized plasmid (Figure 8B).

Genomic islands (GIs), regions of probable horizontal origin, were predicted in V. campbellii LMB29, and multiple GIs were identified in this strain with transferable potentials (Figure 8A; Supplementary Table 3). arr-9 was predicted sits on a 6,844 bp genomic island in pLMB157 which highlights its potential of horizontal transmission (Figure 8A). In addition to arr-9, two aminoglycoside O-phosphotransferase (strA and strB), one sulfonamide-resistant dihydropteroate synthase (sul2) and one tetracycline resistance MFS efflux pump gene [tet(A)] were predicted on another genomic island (7,797 bp) in plasmid pLMB157. These findings are consistent with bacterial adaptations to various environments by acquiring antimicrobial resistance.

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