Paraphotobacterium marinum NSCS20N07D^T (=KCTC 52126^T = MCCC 1A01886^T = CIP 111031^T) was cultured with Marine Broth 2216 (BD, Difco). The cell biomass of 100 ml was obtained by centrifugation at 6,000 × g for 10 min. The genomic DNA was extracted using the conventional phenol/chloroform/isoamyl alcohol (25:24:1) extraction method. The completeness of the extracted DNA was tested using 1.0% agarose gel electrophoresis. The concentration and purity were determined using TBS-380 and NanoDrop 2000 spectrophotometer.

A 10 kb DNA fragment library was constructed according to the manufacturer’s instruction and sequenced on the PacBio RSII sequencing platform with one SMART cell (MajorBio Co., Shanghai, China). Sequencing reads were assembled using the hierarchical genome-assembly process (HGAP 3.0) (Chin et al., 2013). A complete genome with two chromosomes and one plasmid of P. marinum NSCS20N07D^T was obtained.

The prediction of the protein coding genes (opening reading frames, ORFs) was carried out using Glimmer 3.0 (Delcher et al., 2007). The ribosomal RNA (rRNA) genes were predicted using Barrnap¹. The tRNA genes were predicted using tRNAscan-SE (Lowe and Eddy, 1997). Gene annotation was carried out by BLASTP search against non-redundant protein sequences (nr database), Swissprot database, the clusters of orthologous groups of proteins (COG) database, and KEGG database (Kanehisa et al., 2016) with e-value cutoff of 1e-5 (Tatusov et al., 2001). Further annotation was performed using the RAST server online (Aziz et al., 2008) and MicroScope platform (Vallenet et al., 2009). The circular genome was drawn using Circos (Krzywinski et al., 2009).

The in situ surface seawater of two sites (NSCS1 [119°26′24.62′′E, 21°58′35.71′′N] and NSCS2 [117°36′11.76′′E, 19°19′17.30′′N]) in the South China Sea was collected on May 14–15, 2015. P. marinum NSCS20N07D^T was isolated from site NSCS2 at a depth of 3,048 m (Huang et al., 2016). Triplicate 2 l seawater of each sampling site was filtered on board using 0.22 μm filter membrane (GF, Millipore, United States). The membrane was put into a 2.0 ml EP tube and store at -20°C until use.

The total DNA of filtered surface seawater was extracted using the traditional proteinase K/SDS lysis followed by phenol/chloroform/isoamyl alcohol (25:24:1) extraction method (Huang et al., 2014). The quality of total DNA was evaluated by 1.0% agarose gel electrophoresis. The 16S rRNA gene V4 region was amplified with the specific primers 515F (GTGCCAGCMGCCGCGGTAA) and 806R (GGACTACHVGGGTWTCTAAT) with dual barcode (Kozich et al., 2013). The PCR amplification mixture was made in 50 μl reactions with total DNA of 10 ng. The PCR reaction was conducted with TransStart Fastpfu DNA Polymerase as follows: initial denaturation at 98°C for 1 min, followed by 30 cycles of denaturation at 98°C for 10 s, annealing at 55°C for 30 s and elongation at 72°C for 30 s, and extension at 72°C for 5 min. For each sample, three parallel PCR reactions were carried out and pooled together before purification of the PCR product with AxyPrepDNA gel extraction Kit (Axygen). The high-throughput sequencing of the 16S rRNA gene V4 region was performed using Illumina Miseq PE250 platform (Illumina, San Diego, CA, United States).

The pair-end reads of each sample were merged into contigs using make.contigs command implemented in mothur v.1.34.4 (Schloss et al., 2009). The contigs were then trimmed using Qiime pipeline (Caporaso et al., 2010) as follows: (1) allowing no mismatch with dual barcode and primers; (2) allowing no ambiguous bases; and (3) removal of contigs containing the homopolymers over 6. Then, the chimeras were checked using usearch61 algorithm against the latest Greengenes database of gg_13_8_97_otus (DeSantis et al., 2006). The operation taxonomic unit (OTU) was picked with a 97% similarity threshold using the de novo picking method.

The contigs of V4 region of each sample affiliated with P. marinum were identified by mapping to the full-length 16S rRNA gene of P. marinum NSCS20N07D^T (1,539 bp, accession number: KU140414) using BLASTN with e-value of 1e-5 and sequence similarity of >97%.

The closely related sequences of P. marinum were searched in GenBank using Nucleotide BLAST using the full-length 16S rRNA gene of P. marinum NSCS20N07D^T. The sequence similarity above 97% with P. marinum NSCS20N07D^T was determined as the same species, and the isolation source of the closely related sequences was summarized.

A total of 131,203 clean reads were obtained with the largest of 40,999 bp retrieved from one SMART cell of PacBio RSII sequencing. The complete genome of P. marinum NSCS20N07D^T was obtained, containing two circular chromosomes with a total size of 2,593,992 bp with G+C content of 31.2 mol%, and one plasmid with a size of 5,539 bp. The larger chromosome (Chr. I) had a genome size of 1,426,504 bp with G+C content of 31.6 mol%, and the smaller one (Chr. II) had a genome size of 1,161,949 bp with G+C content of 30.8 mol% (Figure 1 and Table 1).

By comparison to 134 species affiliated with seven genera in the family Vibrionaceae, P. marinum NSCS20N07D^T possessed the smallest genome size and lowest G+C content (Figure 2). As far as we know, P. marinum NSCS20N07D^T was the smallest genome in Vibrionaceae species reported so far.

To compare P. marinum with other Vibrionaceae species in more details on each chromosomes, we selected seven representative species with complete genome, Photobacterium profundum SS9, Photobacterium gaetbulicola Gung47, Aliivibrio fischeri ES114, Aliivibrio salmonicida LFI1238, V. cholerae O1 biovar El Tor str. N16961, V. parahaemolyticus RIMD 2210633, and Grimontia hollisae ATCC 33564 (Table 1). Obviously, Chr. I of P. marinum NSCS20N07D^T was the smallest chromosome among those of the above representative species, the size of which ranged from 2,897,536 bp of A. fischeri ES114 to 4,085,304 bp of P. profundum SS9. Strangely, Chr. II of P. marinum NSCS20N07D^T was not the smallest. The size of the smaller chromosome showed considerably variable pattern among the compared members of the family Vibrionaceae, ranging from 783,080 to 2,237,943 bp (Table 1). The variable size of Chr. II was recognized as more flexible genes therein (Okada et al., 2005). G+C content of the two circular chromosomes of P. marinum NSCS20N07D^T had less than 1% difference. This was in congruence with the genome characteristics of Vibrionaceae species, which was regarded as a long evolutionary co-existence for the divided chromosomes (Okada et al., 2005). Furthermore, on the aspect of genome size, the difference between Chr. I and Chr. II was much smaller in P. marinum NSCS20N07D^T than the corresponding other species within this family (Table 1).

Through gene annotation, we found that Chr. I encoded 1,481 ORFs, and had 8 copies of rrn operon (16S–23S–5S rRNA genes), an extra 5S rRNA gene, and 69 tRNA genes; while Chr. II encoded 1,148 ORFs and had 1 copy of rrn operon and 7 tRNAs. It was noteworthy that all these 16S rRNA genes shared 100% sequence identity (Figure 1 and Table 1).

By means of BLASTP search against nr database, Chr. II had a larger proportion of unknown proteins and hypothetical proteins, accounting for 29.4 and 29.8%, respectively, indicating there was a larger proportion of genes not yet elucidated (Supplementary Table S1).

By COG functional annotating, we found that 42.47% in Chr. I and 13.50% in Chr. II can be annotated. Surprisingly, a huge amount of ORFs (86.50%) in Chr. II could not be classified (No assigned COG in Figure 1). This was in congruence with the result of BLASTP against nr database, indicating that the biological function of most genes in Chr. II was still unknown. On the other hand, the vast majority of genes for essential cell function were located in Chr. I determined from COG classes, such as Class J of translation, ribosomal structure, and biogenesis; Class C of energy production and conversion; Classes of D, E, F, G, and P of transport; and Class L of replication, recombination, and repair. Cell motility and secretion (Class N) were undertaken mainly by Chr. I other than Chr. II. COG functional categories revealed that the two chromosomes have different distributions of functional classes, indicating they take different cellular functions.

The study on chromosome replication of the two chromosomes of V. cholerae attracted much more attention in recent years (Baek and Chattoraj, 2014; Val et al., 2016). DnaA was reported as the main controller in Chr. I in V. cholerae (Baek and Chattoraj, 2014). Similarly, in P. marinum NSCS20N07D^T, the replication initiation region of Chr. I was identified by the gene configuration of dnaA, dnaN, and recF, which encoded chromosomal replication initiation protein DnaA, DNA polymerase III subunit beta, and DNA replication/repair protein RecF, respectively (Supplementary Table S1). CrtS, the RctB binding locus, was reported to mediate the replication initiation of Chr. II (Baek and Chattoraj, 2014). However, in P. marinum, we did not find CrtS, whereas the replication initiation region of Chr. II was identified by chromosome partitioning protein ParB (orf00002_1) and protein ParA (orf00003_1). With respect to the plasmid replication, there was a RepA protein found in the plasmid of P. marinum, which can control the replication of plasmid.

Complete genes taking part in the central carbohydrate metabolism could be found in Chr. I (Figure 3), including the Embden–Meyerhof–Parnas pathway (EMP or glycolysis), pyruvate oxidation to acetyl-CoA, tricarboxylic acid (TCA) cycle, and glyoxylate bypass (also called glyoxylate cycle). Chr. I also contained an Entner–Doudoroff pathway, which could replace the glycolysis to catabolize glucose to pyruvate. Most genes involved in pentose phosphate pathway were encoded in Chr. I, however, the transaldolase catalyzing the step of sedoheptulose 7-phosphate and glyceraldehyde 3-phosphate to form fructose 6-phosphate and erythrose 4-phosphate was encoded in Chr. II. This indicated the pentose metabolism could be regulated by the smaller chromosome Chr. II.

On the aspect of utilization of carbon source, the two chromosomes hosted different genes encoding different metabolic activities. Chr. I possessed the genes encoding the utilization of D-glucose, while Chr. II mainly possessed those for the utilization of D-galactose. This demonstrated that Chr. II improved the ability of the metabolic range of carbon source. No data supported that this bacterium can utilize oligosaccharides.

Through metabolic pathway analysis, 13 kinds of amino acids can be synthesized, which were found in Chr. I. The majority of genes involved in arginine biosynthesis could be found in Chr. II except the gene argD encoding acetylornithine aminotransferase, which was encoded in Chr. I (Supplementary Table S1). Tryptophan and proline biosynthesis were also annotated in the two chromosomes (Figure 3).

Through gene annotation, Chr. I contained 19 kinds of tRNA synthetase responsible for the 19 amino acids except proline. Chr. II contained only two tRNA synthetases for proline and tryptophan, respectively.

Strain NSCS20N07D^T can produce a large amount of poly-β-hydroxybutyrate (PHB) granules inside the cell observed using transmission electron microscopy (TEM) after negative staining and Nile blue A stain (Huang et al., 2016). Gene annotation and metabolic analysis revealed that a gene cluster (atoB, phbB, and phbC) was present in Chr. II for PHB biosynthesis from two acetyl-CoA molecules to form acetoacetyl-CoA (Figure 3), whereas it seems that the degrading of PHB was controlled by Chr. I as the incomplete pathway from (R)-3-hydroxybutanoate to acetone was found, but the mechanism of polymer hydrolyze needed further study.

All genes encoding for flagellar biosynthesis proteins could be found in Chr. I (Supplementary Table S1), indicating that strain NSCS20N07D^T could be motile. However, no flagella were observed by TEM after negative stain (Huang et al., 2016), making it likely that the flagella were lost in sample preparation.

To investigate the relative abundance of P. marinum in the surface seawater where it was isolated, we carried out microbial diversity analysis based on high-throughput sequencing of 16S rRNA gene V4 region using Miseq platform. The six samples produced a total of 142,564 contigs with the minimum of 18,669 contigs for NSCS1.3 and the maximum of 28,067 contigs for NSCS1.1. The surface seawater of the sampling sites in South China Sea was dominated by Cyanobacteria, Proteobacteria, Bacteroidetes, and Actinobacteria, though the data seemed to be variable among each samples (Figure 4). Synechococcales in Cyanobacteria (3.82–50.19%), Rickettsiales in Alphaproteobacteria (9.10–38.90%), Oceanospirillales in Gammaproteobacteria (4.97–20.89%), and Acidimicrobiales in Actinobacteria (6.11–15.05%) were the dominant groups in the surface seawater (Figure 4). Flavobacteriales of the class Flavobacteriia accounted for 2.47–8.74% relative abundance (Figure 4). While sequences affiliated with the family Vibrionaceae accounted for less than 0.1% of relative abundance.

By mapping the V4 contigs to the full-length 16S rRNA gene of P. marinum NSCS20N07D^T, we found that there were four contigs (253 bp in length) in NSCS1.1 (28,067 contigs), one contig in each sample of NSCS1.2 (25,062 contigs), NSCS2.1 (24,463 contigs), and NSCS2.2 (22,064 contigs), having 99.6–100% similarity with the 16S rRNA gene of P. marinum, while other 16S rRNA gene V4 contigs showed the similarity no more than 96.0%. The seven sequences were classified into a single OTU (denovo_289). The low relative abundance of P. marinum indicated that it was a rare bacterial species which were regarded as the threshold of five sequences per OTU (Skopina et al., 2016). Thus, it is assumed that P. marinum is present as a rare species in the surface seawater.

Through 16S rRNA gene similarity search against the Nucleotide collection (nr/nt) database, we found that this bacterium P. marinum showed 99% similarity with the 21 uncultured bacterium clones detected from the Red Sea coral Pocillopora damicornis (accession number: KC668602), seawater from South China Sea (GU061990, GU061981), hydrothermal plumes of East Pacific Rise (JN874365, JN874376, JN874374, JN873939, JN874354, JN874093, JN874308, JN874318, JN874322), West Pacific (KF271041), and Gulf of Mexico oil sheen (KF786540, KF786512, KF786461), ocean water of north eastern Gulf of Mexico (KU578418), and CALCOFI Line 67-155 depth of 5 m size fraction 3.0–20 μm (KX935264, KX935265, KX935266, KX935267). These environments, to some extent, are all pelagic niches that are distinct from the coast region. Thus, we speculate that P. marinum is specific to the pelagic environment, and has a global distribution by free-living, or coral-associated, and from surface to the deep extreme hydrothermal environments.