The Pseudovibrio strain was isolated as previously described (Rua et al., 2014). Pseudovibrio sp. Ab134^T has been deposited in the Collection of Environmental and Health (CBAS) at the Oswaldo Cruz Institute (IOC), FIOCRUZ (Rio de Janeiro, Brazil) (http://cbas.fiocruz.br/) and assigned the accession number CBAS 623^T. The strain was also deposited in the Collection of Aquatic Microorganisms (CAIM) in Mazátlan, Sinaloa, Mexico (http://www.ciad.mx/caim/CAIM.html) and assigned the accession number CAIM 1924^T.

Genomic DNA was extracted using a previously described method (Pitcher et al., 1989). DNA libraries were built using the Nextera DNA Sample Preparation Kit (Illumina, San Diego, CA, USA). The DNA was purified using AMPure XP beads and quantified using the fluorometric Qubit dsDNA HS Assay Kit (Life Technologies). Quantification of libraries was performed with the 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) and the KAPA Library Quantification Kit (Kapa Biosystems, Wilmington, MA, USA). Library size distribution was determined using the Agilent 2100 Bioanalyzer. Genome sequencing was performed using the Illumina MiSeq platform (paired-end sequencing, 2 × 300 base pairs). The sequences obtained were preprocessed using PrinSeq software to remove small reads (<35 bp) and low-score sequences (Phred score <30) (Schmieder and Edwards, 2011). Two programs assembled high-quality reads: A5-miseq (Coil et al., 2015) assembled the sequence data, and then the generated contigs and the singletons were used as input for the CAP3 sequence assembly program (Huang and Madan, 1999). Functional annotation was carried out by the Rapid Annotations using Subsystem Technology (RAST) platform (Aziz et al., 2008).

The 16S rRNA gene of strain Ab134^T was obtained as previously described (Moreira et al., 2014; Rua et al., 2014) and compared to known sequences in the NCBI GenBank database using the Basic Local Alignment Search Tool (BLAST). The closest matches were included in the phylogenetic analysis. The 16S rRNA gene identities were calculated using Jalview V.2 (Waterhouse et al., 2009). Pairwise and multiple alignments were performed using ClustalW (Larkin et al., 2007). Evolutionary history was inferred using the neighbor-joining method (Saitou and Nei, 1987). The bootstrap consensus tree inferred from 1,000 replicates was taken to represent the evolutionary history of the taxa analyzed (Felsenstein, 1985). Evolutionary distances were computed using the p-distance method (Nei and Kumar, 2000), and evolutionary analyses were conducted in MEGA6 (Tamura et al., 2013).

The recA gene sequences for Pseudovibrio genomes available on GenBank (P. denitrificans JCM 12308^T, P. sp. FO-BEG1, P. axinellae Ad2^T, and P. hongkongensis UST20140214-015B^T) were analyzed using the BLASTN algorithm (Altschul et al., 1990). Multiple alignment, phylogenetic reconstruction, and bootstrap consensus were conducted as described above for the 16S rRNA genes (Felsenstein, 1985; Saitou and Nei, 1987; Larkin et al., 2007; Tamura et al., 2013). Accession numbers are listed in Table S1.

In silico DNA-DNA hybridization analysis was carried out by genome-to-genome comparison (Auch et al., 2010). Average nucleotide identity (ANI) was calculated as previously described (Thompson et al., 2013). The intraspecies genomic relatedness ranged from 95 to 100% ANI. Genome distance was calculated using a genome-to-genome distance (GGD) calculator (Meier-kolthoff et al., 2016), with intraspecies genomic similarity ranging from 70 to 100%.

Phenotype analysis was based on genome sequences (Amaral et al., 2014), focusing on eight biochemical characteristics that have been used to identify Pseudovibrio species (Shieh et al., 2004; Hosoya and Yokota, 2007). For each characteristic, we searched for the corresponding genes. If one or more genes involved in a phenotype was present in the genome, the organism was considered positive for this phenotype. Genes encoding proteins involved in those characteristics were detected by using the RAST program (Overbeek et al., 2014). Genes associated with related biochemical pathways were identified with the BLASTP algorithm (Altschul et al., 1990). We used the antiSMASH 2.0 software pipeline (Weber et al., 2015) to identify secondary metabolite biosynthesis clusters in the whole genome sequences of P. brasiliensis Ab134^T and P. denitrificans JCM 12308^T. The metabolic features of P. denitrificans JCM 12308^T and the Ab134^T strain were compared using the RAST platform (Overbeek et al., 2014).

In order to identify genes related to symbiosis in P. brasiliensis Ab134^T, secretion system types III, IV, and VI and toxin-like proteins were predicted using two different approaches. The search for secretion systems was carried out using T346Hunter, a web-based tool for the prediction of type III, type IV, and type VI secretion systems (Martínez-García et al., 2015). Detection of toxin-like proteins was carried out using BLASTP and the Virulence Factors Database (VFDB; Chen et al., 2016). For the BLASTP analysis, amino acid sequences from P. brasiliensis Ab134^T were predicted using Prodigal software v2.6.2 (Hyatt et al., 2010) with default parameters and used as query sequences. Results annotated as “toxin” were used in the subsequent analysis. Genes associated with the secretion systems of P. brasiliensis Ab134^T and P. denitrificans JCM 12308^T were detected out using the same approach.

Growth of Pseudovibrio sp. Ab134^T at different NaCl concentrations (0% and 5.0% [w/v]), temperatures (4, 10, and 13–44°C), and pH values (5–10, adjusted with NaOH or HCl) were tested on marine agar medium. Plates were incubated at 27°C (or the test temperature) for up to 7 days in triplicate. Motility was evaluated on semisolid marine agar and stab inoculation into tubes and confirmed by phase contrast microscopy, which was also used to evaluate Gram staining, shape, and diameter. Features that differed between Ab134^T and the reference strains of the three Pseudovibrio species are listed in Table S2.

A total of 1,967,588 paired-ends reads for the Ab134^Tstrain were generated. The genome assembly resulted in 39 contigs, and the coverage was 164-fold. Estimated genome size was 5,975,631 bp, and the number of coding sequences estimated by Prodigal was 5476. Of the 66 RNA sequences, 61 were tRNAs, and 5 were rRNAs. Functional annotation revealed that most of the 2,260 genes identified were assigned to carbohydrates (301), amino acid and derivatives (281), and cofactors, vitamins, prosthetic groups, and pigments (234) (Figure S1).

Our results showed that P. brasiliensis Ab134^T is phylogenetically and genetically related to P. denitrificans (Figure S2 and Table S3). A functional comparison between the P. denitrificans JCM 12308^T and P. brasiliensis Ab134^Tgenomes revealed 51 genes unique to P. denitrificans JCM 12308^T, including those related to urea decomposition, acetyl-CoA conversion to butyrate, menaquinone and phylloquinone biosynthesis, curli production, dihydroxyacetone kinases, and the G3E family of P-loop GTPases (metallocenter biosynthesis, urease accessory). The 34 genes unique to the P. brasiliensis Ab134^T genome included those related to phages and prophages, maltose and maltodextrin utilization, transport of nickel and cobalt, toxin-antitoxin replicon stabilization systems, and thioredoxin-disulfide reductase (Table S4). Genes shared by P. brasiliensis Ab134^Tand P. denitrificans JCM 12308^T included those related to nitrogen metabolism (dissimilatory nitrite reductase, nitrate and nitrite ammonification) (Table S5) and fermentation processes (butanol biosynthesis, mixed acid, lactate, acetolactate synthase subunits, and acetyl-CoA fermentation to butyrate) (Table S6), suggesting the potential of P. brasiliensis Ab134^T to carry out denitrification and fermentation in the sponge holobiont.

Using the antiSMASH program we identified clusters related to secondary metabolite biosynthesis (Figures 1,2 and Figures S3,S4) and found differences in gene cluster abundance and diversity between P. denitrificans JCM 12308^T and P. brasiliensis Ab134^T. The nine clusters identified in P. denitrificans JCM 12308^T are involved in the biosynthesis of bacteriocins (2), terpene (1), non-ribosomal peptide synthases (NRPSs) (2), hybrid NRPS-polyketide synthase (PKS)-T1 (1), hybrid PKS T1-bacteriocin (1), homoserine-lactone (1), and T3PKS-T1PKS (1). The five clusters identified in P. brasiliensis Ab134^T were assigned to the biosynthesis of terpene (1), bacteriocins (2), PKS (1), and NRPS (1). Some gene clusters showed no similarity to known Pseudovibrio gene clusters.

We identified four gene clusters encoding non-flagellar type III secretion systems (T3SSs) in the P. brasiliensis Ab134^T genome (Figure S5). All essential genes were present in three of these clusters. Genes in the four clusters showed high similarity to those from P. denitrificans JCM12308^T (Figure S5). The main difference was the presence of three genes from other secretion systems. T3SS cluster 1 (T3SS-1) contained the gene tfc4 (155 amino acids), which is characteristic of T4SS. T3SS cluster 3 (T3SS-3) and cluster 4 (T3SS-4) contained vasH genes (854 and 421 amino acids, respectively), which are characteristic of T6SS. We also identified one cluster that appears to encode 12 proteins of a T4SS, containing mainly virB genes and trb (Figure S6).

We identified two clusters that encode T6SS genes, both including vgrG and hcp genes (Figure S7A). The distribution of effector protein-coding genes was similar to that of the two T6SS clusters identified in the P. denitrificans JCM 12304^T genome (Figure S7B), but the number and arrangement of genes differed. Both Ab134^T and P. denitrificans JCM 12304^T genomes contained 16 core component genes in the first cluster (T6SS-1). Whereas in the second cluster (T6SS-2), Ab134^T and P. denitrificans JCM 12304^T genomes showed five and 19 core component genes, respectively (Figures S7A,B). Taken together, these results suggest specific interactions with eukaryotes and the potential ability to target host cell machinery.

A wide range of potential toxin-like protein-coding genes was identified in the P. brasiliensis Ab134^T genome. Of the 25 hits retrieved from a BLASTP search against the VFDB, 15 genes were unique. The genes argK, frpC, rtxA, and syrE were present in multiple copies, suggesting the presence of paralogs (Table S7).

Ab134^T clustered tightly with other Pseudovibrio spp. based on the 16S rRNA gene sequences analysis (Figure S2A), showing identity values to P. denitrificans strains of 99.4% (JE062, NW001) and 99.5% (JCM12308^T, FO-BEG1, MBIC3368). Strains JE062 and FO-BEG1 shared almost identical 16S rDNA gene sequences (99.9%) with P. denitrificans DN34^T. Based on the phylogenetic analysis of recA, Ab134^T shared only 93% sequence identity with P. denitrificans JCM 12308^T and FO-BEG1 (Figure S2B), suggesting that Ab134^T represents a new species of the genus Pseudovibrio. A formal description is provided in the Supplementary Material.

Pairwise genomic comparisons between P. brasiliensis Ab134^T and P. denitrificans JCM 12038^T (Shieh et al., 2004) showed that they share only 93% ANI and 54.3% (±3) GGD similarity (in silico DNA-DNA hybridization) (Table S3). A bacterial species is defined as a group of strains that share ≥98.7% 16S rRNA gene sequence similarity, >95% ANI and >70% GGD similarity (Thompson et al., 2015). Based on phylogenetic analysis, in silico DNA-DNA hybridization, ANI, and differential phenotypic characteristics, strain Ab134^T is proposed as the type strain of a novel species, for which the name Pseudovibrio brasiliensis sp. nov. is proposed.

Secondary metabolites and toxins may complement each other to promote holobiont homeostasis. Secondary metabolites (e.g., bacteriocins, terpenes, and NRPS) are encoded by gene clusters, whereas toxins are encoded by a few genes or a single gene (e.g., argK, frpC, rtxA) (Sertan-De Guzman et al., 2007; Penesyan et al., 2011; Vizcaino, 2011; O'Halloran et al., 2013).

Bacteriocins are ribosomally synthesized antimicrobial peptides that are lethal to closely related bacteria. Bacteriocin producers are protected from the effects of these peptides by a specific immunity protein(s) (Cotter et al., 2005). Bacteriocins have been used extensively as preservatives in the food industry (Deegan et al., 2006) and have been identified as potential alternatives to antibiotics (Piper et al., 2009). Bacteriocins may also serve as anti-competitor toxins, enabling a strain or species to invade an established microbiome (Riley and Gordon, 1999; Lenski and Riley, 2002; Riley and Wertz, 2002). In sponge holobionts, bacteriocins protect the host against pathogenic bacteria, and bacteriocin-producing bacteria may prevent the dissemination of pathogens by occupying the same ecological niche (Desriac et al., 2010).

Kennedy et al. (2008) reported that Pseudovibrio cultures from the marine sponge Haliclona simulans contain both putative PKS and NRPS genes, suggesting a potential for secondary metabolite production. These strains exhibited antimicrobial activity against methicillin-resistant Staphylococcus aureus, Bacillus cereus, Escherichia coli, and B. subtilis. P. brasiliensis Ab134^T harbored PKS and NRPS gene cluster and exhibited antimicrobial activity against B. subtilis (Rua et al., 2014) and the pathogens V. coralliilyticus P1 and V. parahaemolyticus (this study).

P. brasiliensis Ab134^T also produces bromotyrosine-derived alkaloids (Nicacio et al., 2017). Additional enzymes involved in secondary metabolism were detected in this study. Experiments aiming to elucidate the biosynthetic gene clusters responsible for these alkaloids are in progress and will be reported in due time.

Toxin genes detected in multiple copies include argK, frpC, rtxA, and syrE. argK encodes an ornithine carbamoyltransferase that confers self-resistance to phaseolotoxin, responsible for halo blight in beans (Phaseolus vulgaris L.), caused by Pseudomonas syringae pv. phaseolicola, (Mosqueda et al., 1990). This gene is known to be horizontally transferred (Sawada et al., 2002) and the product ArgK protects phaseolotoxin producers (self-resistance) in part by providing an alternative Arginine source (Chen et al., 2015). L-arginine is produced by bacterial fermentation and the main applications are in flavor and pharmaceuticals. Arginine overproducing strains has been a research target for the past decades (Lu, 2006). Another gene that was first described as a virulence determinant of P. syringae pv. syringae B301D was present in two copies. syrE is a phytotoxin gene present within the syringomycin cluster. Syringomycin is a cyclic lipodepsinonapeptide (Guenzi et al., 1998). This protein forms pores in plasma membranes, leading to electrolyte leakage (Bender et al., 1999) and might play a role in protecting the host (Table S7).

Multiple copies of frpC and rtxA were also detected. frpC encodes the iron-repressible repeat-in-toxin (RTX) protein FrpC, and rtxA encodes the structural toxin protein RtxA. Both belong to the RTX family, which consists primarily of cytotoxic pore-forming proteins (Schaller et al., 1999; Linhartová et al., 2010) acting as virulence determinants in many gram-negative pathogens. RTX proteins can also play a role in host protection as bacteriocins or by forming protective bacterial surface layers (S-layers) (Linhartová et al., 2010). RTX proteins exhibit additional biological activities as metalloproteases, lipases, pore-forming toxins, iron-regulated proteins, nodulation-related proteins and are involved both in bacterial adherence/motility and host-receptor interactions (Welch, 2001; de Souza Santos et al., 2015). RTX proteins are secreted by a T1SS via Sec-independent pathway used by gram-negative bacteria to transport proteins from the cytoplasm to the extracellular medium in a single step (Chenal et al., 2015). The presence of multiple copies of frpC and rtxA suggests a possible environmental role, including protection of the host and defense against other microorganisms and pathogens. The extracellular matrix of sponges is rich in proteoglycans, lamnin-like subunits, fibronectin, and other structural proteins (Har-el and Tanzer, 1993; Özbek et al., 2010); thus, the multiple copies of RTX proteins may help to penetrate the sponge mesohyl.

A symbiotic relationship suggested between Pseudovibrio and marine invertebrates (Taylor et al., 2007) may involve interactions with holobiont host cells via secretion systems T3SS, T4SS, and T6SS. Secretion systems were first associated with pathogenic strains but have since been widely detected in symbiotic and free-living bacteria (Dale and Moran, 2006; Persson et al., 2009). Bacteria commonly use these three secretion systems to inject effector proteins in target cells, which facilitate colonization (Costa et al., 2015). For example, the non-flagellar T3SS (injectisome) enables gram-negative bacteria to deliver effector proteins into the cytoplasm of eukaryote hosts (Büttner, 2012). T4SS (VirB system), which was first identified in Agrobacterium tumefaciens, delivers toxins into host cells, as well as DNA that integrates into the host genome (Wallden et al., 2010), contributing to the genome plasticity and virulence of the bacteria (Voth et al., 2012).

Two genes (virB and trb) in the T4SS of P. brasiliensis Ab134^T are related to conjugation, suggesting a role in genome plasticity and environmental adaptation. These genes have been identified in other Pseudovibrio strains (Romano et al., 2016). Compared to P. denitrificans JCM 12308^T, the number of genes in the T4SS cluster was the same, but the gene annotation and rearrangement differed (Figure S6).

T6SSs are thought to help bacteria conquer an ecological niche (Ma et al., 2014; Russell et al., 2014; Kapitein and Mogk, 2017), and niche-specific distribution of T6SS effectors has recently been described (Egan et al., 2015; Romano et al., 2016). T6SSs appear to be involved in biofilm formation (Aschtgen et al., 2008), quorum sensing (Weber et al., 2009), interbacterial interactions (Hood et al., 2017), and anti-pathogenesis (Chow and Mazmanian, 2010; Jani and Cotter, 2017). The T6SS gene impI was detected only in the P. brasiliensis Ab134^T genome, which contained all core components of the T6SS, as described by Boyer et al. (2009).

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.