Introduction

The phylum Bacteroidetes is among the most abundant phyla in the ocean, accounting for an average of approximately 10% of marine surface bacterioplankton cells1,2. Marine Bacteroidetes specializes in the utilization of particulate and high molecular weight organic matter predominately from micro- or macroalgae3, and many studies have shown that Bacteroidetes increases in abundance during phytoplankton blooms in coastal waters, occupying more than 50% of the bacterial community at certain times4,5. Of the many groups within the phylum Bacteroidetes, the family Flavobacteriaceae is the most abundant in marine environments4, and has been subjected to several cultivation and genome-based studies6,7,8,9. Currently, more than 100 genera with validly published names have been described within the family Flavobacteriaceae.

The genus Kordia belonging to the family Flavobacteriaceae was proposed by Sohn et al.10 and currently comprises eight species: K. ulvae11, K. algicida10, K. aquimaris12, K. zhangzhouensis13, K. jejudonensis14, K. periserrulae15, K. zosterae16, and K. antarctica17. The type strains of these Kordia species have been isolated from a range of marine habitats, including surface seawater, a connection between the ocean and a freshwater spring, seaweed surfaces16, and the digestive tract of a marine polychaete. Despite enriched taxonomic diversity and wide distribution, physiological, ecological, and genomic studies on the genus Kordia have yet to be performed in-depth beyond those on the algicidal activity of K. algicida strains10,18,19,20 and a recent report detailing microdiversity among uncultured Kordia species identified by single-amplified genomes from a seawater sample of the Indian Ocean21.

Currently, whole-genome sequences of four strains of Kordia species are publicly available: K. algicida, K. zhangzhouensis, K. jejudonensis, and K. periserrulae. However, all of these genome sequences remain in the draft stage, and the only published genome sequence is that of K. algicida via a brief single-page genome announcement22. Recently, a high-quality, metagenome-assembled genome belonging to the genus Kordia was obtained by sequencing a non-axenic culture of the marine diatom Skeletonema marinoi23. This genome sequence has been described as being acquired from the Kordia sp. strain SMS9, but according to reports, attempts to obtain a pure culture of strain SMS9 have been unsuccessful.

This report provides the first complete genome sequence of the genus Kordia obtained from K. antarctica strain IMCC3317T, as well as comparative genome analyses with the four existing genomes of other Kordia species. Strain IMCC3317T is a Gram-negative, chemoheterotrophic, yellow-pigmented, non-motile, flexirubin-negative, facultative anaerobic bacterium that was isolated from a coastal seawater sample at the Antarctic Peninsula17. To our knowledge, strain IMCC3317T is the only Kordia strain isolated from the polar environment thus far. The complete genome sequence of strain IMCC3317T confirmed a chemoheterotrophic lifestyle and showed the potential to utilize polysaccharides and synthesize secondary metabolites. Comparative analyses also indicated that the IMCC3317T genome contains a nosZ gene encoding nitrous oxide (N2O), unique to this strain among the Kordia species.

Results and discussion

General genome features

Strain IMCC3317T was isolated from a coastal seawater sample at King George Island in western Antarctica (62°14′ S 58°47′ W) using a standard dilution-plating method and was established as K. antarctica based on phylogenetic, biochemical, and physiological characterization17. General features of strain IMCC3317T are summarized in Baek et al.17 and are also presented in Table 1. Comparison of 16S rRNA gene sequences with other type strains of the genus Kordia showed that strain IMCC3317T is most closely related to K. zosterae ZO2-23T (97.7%, sequence similarity), followed by K. ulvae SC2T (97.0%), K. jejudonensis SSK3-3T (96.5%), K. algicida OT-1T (96.6%), and K. periserrulae IMCC1412T (96.1%). Since a complete genome of an axenic strain has yet to be reported for the genus Kordia, we derived the complete genome of IMCC3317T using PacBio sequencing.

Table 1 General features and genome sequencing information of Kordia antarctica strain IMCC3317T according to the MIGS recommendations.
Full size table

The complete genome of strain IMCC3317T was obtained by de novo assembly followed by five rounds of polishing and comprised a single circular contig. The total length of the genome was 5,500,985 bp and the DNA G+C content was 33.23 mol% (Fig. 1). A total of 4,761 genes were predicted for the genome via the IMG-ER pipeline, including 4,697 protein-coding genes, nine rRNA genes, and 49 tRNA genes. Among the 4,697 protein-coding genes, 62.7% (2,944) could be assigned a putative function. The major COG categories were translation, ribosomal structure and biogenesis (7.74%), coenzyme transport and metabolism (5.97%), and inorganic ion transport and metabolism (5.92%). Overall information on the sequencing and annotation of the IMCC3317T genome is presented in Table 1.

Figure 1
figure 1

Circular map of the Kordia antarctica IMCC3317T genome. From outside to center: Genes on forward strand (colored by COG categories), genes on reverse strand (colored by COG categories). RNA genes: tRNAs, green; rRNAs, red; other RNAs, black. GC content, black; GC skew, purple/olive. Color codes for COG function categories shown below genome map. Position of the ‘giant gene’ indicated by red arrow.

Full size image

ANI values among the five Kordia genomes (including the IMCC3317T genome) ranged from 76 to 80%. These ANI values are lower than 95–96%, a widely accepted threshold for prokaryotic species demarcation24,25, and therefore are consistent with the fact that these five genomes are from five different species of the genus Kordia. When compared using BLASTn, the IMCC3317T genome showed similarities to the other four Kordia genomes over the entire length of the genome (Fig. 2). However, many genomic regions not similar to the four other Kordia genomes were also found for IMCC3317T. Metabolic genes that were predicted to be unique to the IMCC3317T genome, such as cysC and nosZ (see below), were found within these regions (Fig. 2). Comparison with other Kordia genomes based on the inference of orthologous protein clusters showed that the IMCC3317T genome shared around 55% of its protein clusters (2,324 among 4,221) with all other four Kordia genomes, while 900 protein clusters were not found in any other Kordia genome (Fig. 3).

Figure 2
figure 2

Genomic comparisons among K. antarctica IMCC3317T and other Kordia strains K. algicida, K. jejudonensis, K. zhangzhounensis and K. periserrulae. Innermost rings show G+C content (black) and GC skew (purple/green) of IMCC3317T genome. Other rings show BLASTn-based similarities of IMCC3317T genome to other Kordia genomes. Relative shading density within each circle represents level of nucleotide identity. Red boxes indicate the approximate locations of genes unique to the IMCC3317T genome. Abbreviations: gdhA, glutamate dehydrogenase (NAD(P)); mqo, malate dehydrogenase (quinone); cysC, adenylylsulfate kinase; ldhA, d-lactate dehydrogenase; nosZ, nitrous oxide reductase.

Full size image
Figure 3
figure 3

Venn diagram showing number of shared and unique orthologous protein clusters among five Kordia strains: K. antarctica, K. periserrulae, K. algicida, K. jejudonensis, and K. zhangzhounensis. Total number of protein clusters of each genome indicated in parentheses below strain names.

Full size image

A long, 50,835 bp gene was predicted in the IMCC3317T genome. This ‘giant gene’ (IMG gene ID: 2713618241; hereafter, ten-digit numbers in the parentheses following gene or protein names are their IMG gene IDs) had a G+C content of ~ 42 mol% that deviated considerably from the average26 (Fig. 1). This gene encoded a protein of 16,944 amino acids in length with more than 100 tandemly-repeated, single Pfam domains (PF03160; Calx-beta). Though a specific function could not be assigned to this protein, the prediction of a signal peptide and a TIGR04131 (gliding motility-associated C-terminal) motif indicated the protein might be involved in cell surface function. BLASTp analysis against the NCBI nr database (accessed at Jul, 2020) showed that many similar proteins are found mostly (but not exclusively) in the family Flavobacteriaceae, including other Kordia species. The top 10 best hit proteins were also composed of a very large number of amino acids (3,864–12,288), suggesting a wide distribution of long-length proteins similar to the one predicted in the IMCC3317T genome.

Carbon metabolism

The metabolic capacity of strain IMCC3317T was analyzed using KEGG pathway maps obtained from BlastKOALA and the IMG-ER annotation. The other four Kordia genomes available in IMG were also analyzed by BlastKOALA and used for comparative analyses. Since the completeness of other Kordia genomes (99.48–99.59%; calculated by CheckM) was only slightly lower than that of the IMCC3317T genome (99.92%), these genome sequences were confident to be used for comparative genome analyses.

Strain IMCC3317T and other four Kordia type strains were all predicted to possess complete pathways for central carbon metabolism, such as the Embden–Meyerhof–Parnas (EMP) glycolysis pathway, the tricarboxylic acid (TCA) cycle, and the non-oxidative branch of the pentose phosphate pathway (Table 2). However, strain IMCC3317T was distinct from other strains in the number of enzymes it possessed for several steps in these pathways. While the other four strains encoded but a single copy of 6-phosphofructokinase, one of the most important regulatory enzymes of the EMP pathway, IMCC3317T encoded two copies thereof, with one copy (2713615308) showing more than 90% similarity to homologs from other strains and the other (2713616029) having with less than 50% similarity to these homologs. In addition, while the other four strains each had one class II fructose-bisphosphate aldolase, IMCC3317T encoded the class I fructose-bisphosphate aldolase (2713616027) as well as the class II enzyme. All five strains contained NAD-dependent malate dehydrogenase (EC 1.1.1.37), an enzyme that converts malate to oxaloacetate in the TCA cycle, but strain IMCC3317T additionally harbored a malate dehydrogenase (quinone) (EC 1.1.5.4; 2713614246), a membrane-associated enzyme that can donate electrons to quinone (Table 2).

Table 2 Distribution of major metabolic pathways and genes among K. antarctica strain IMCC3317T and four other Kordia strains.
Full size table

Several differences were found amongst the Kordia genomes regarding accessory carbon metabolism. Strain IMCC3317T was predicted to lack a glyoxylate cycle, while the other four strains were equipped with isocitrate lyase (EC 4.1.3.1) and malate synthase (EC 2.3.3.9), two key enzymes for this cycle (Table 2). Strains IMCC3317T and K. periserrulae IMCC1412T encoded genes for the Leloir galactose utilization pathway, including galactokinase (EC 2.7.1.6) and galactose-1-phosphate uridylyltransferase (EC 2.7.7.12). Only strain IMCC3317T possessed a lactate dehydrogenase (EC 1.1.1.28; 2713616031), which may facilitate fermentation under anaerobic conditions. All five Kordia genomes encoded phosphoenolpyruvate carboxylase (EC 4.1.1.31), an anaplerotic enzyme that converts pyruvate to oxaloacetate via the incorporation of CO2. However, pyruvate carboxylase (EC 6.4.1.1), another representative anaplerotic enzyme, was not found in any Kordia genomes.

Nitrogen related metabolism

All five Kordia genomes encoded genes for glutamine synthetase and glutamate synthase for ammonia assimilation (Table 2). All genomes possessed two glutamine synthetase genes of different lengths located next to each other in opposite directions. One group of glutamine synthetase genes encoded a type III enzyme of ~ 730 amino acids and the other encoded enzymes of ~ 340 amino acids. In the IMCC3317T genome, genes for an ammonia transporter and nitrogen regulatory protein P-II were located just upstream of glutamate synthase. The five Kordia genomes also contained genes for glutamate dehydrogenase (GdhA) responsible for another ammonia assimilation pathway that synthesizes glutamate directly from ammonia. There are three types of GdhA, each showing different cofactor specificities: NAD-dependent (EC 1.4.1.2), NADP-dependent (EC 1.4.1.4), and NAD(P)-dependent (EC 1.4.1.3). All Kordia genomes possessed the NADP-dependent type (EC 1.4.1.4), but only the IMCC3317T genome encoded an additional GdhA annotated as a dual cofactor-specific type27 (EC 1.4.1.3; 2713618789). Interestingly, a nosZ gene was found in the IMCC3317T genome, but not in other Kordia genomes. nosZ encodes a nitrous oxide reductase that reduces nitrous oxide to nitrogen, the final step in the denitrification pathway. No other genes involved in denitrification were found in the IMCC3317T genome. A more detailed analysis of the nosZ gene found in strain IMCC3317T is presented below.

Sulfur metabolism

Only strain IMCC3317T had a complete assimilatory sulfate reduction pathway wherein sulfate is reduced to sulfide and sequentially incorporated into the biomass, usually via sulfur-containing amino acids such as methionine and cysteine (Table 2). In many bacteria, assimilatory sulfate reduction is accomplished through four steps sequentially mediated by sulfate adenylyltransferase (CysND; EC 2.7.7.4), adenylylsulfate kinase (CysC; EC 2.7.1.25), phosphoadenylylsulfate reductase (thioredoxin) (CysH; EC 1.8.4.8), and sulfite reductase (ferredoxin) (Sir; EC 1.8.7.1). In the genomes of the strains K. algicida, K. zhangzhouensis, and K. antarctica, a homologous gene cluster containing cysND, cysH, and sir genes, but not the cysC gene, was found in the vicinity of methionine biosynthesis-related genes. The genomes of K. algicida and K. zhangzhouensis did not encode a cysC gene, suggesting an incomplete assimilatory sulfate reduction pathway in these strains. In contrast, the IMCC3317T genome encoded a cysC gene (2713614888) in another gene cluster that contained an additional copy of the cysND gene (2713617813, 2713617814), located at the 5´ region of a non-ribosomal peptide synthetase (NRPS) biosynthetic gene cluster. This finding indicates that only IMCC3317T possesses a complete pathway for assimilatory sulfate reduction, which may confer a competitive advantage when reduced sulfur compounds are scarce. Genes for the Sox pathway of sulfur oxidation that enables sulfur compound utilization as an energy source were not predicted for any of the Kordia genomes. However, a sulfide:quinone oxidoreductase that mediates the electron transfer from sulfide to the quinone pool was found in the K. zhangzhouensis (2628586580) and K. periserrulae genomes (2735934522) (Table 2).

Phosphorus metabolism

Regarding phosphorus metabolism, strain IMCC3317T and other Kordia strains appear to specialize in the utilization of organophosphate but not inorganic phosphate. Many representative marine bacterial groups abundant in pelagic ocean, such as the SAR11 clade and Prochlorococcus, possess a high-affinity phosphate ABC transporter (EC 7.3.2.1; PstSCAB), which may be an adaption to phosphate-depleted oligotrophic pelagic waters. However, all Kordia genomes lacked this ABC transporter, implying that the genus Kordia has not adapted to phosphate-depleted oligotrophic condition. Instead, all Kordia genomes encoded genes for alkaline phosphatase D (PhoD) and PhoPR, a two-component signal transduction system. Alkaline phosphatases (EC 3.1.3.1) can liberate phosphate from organic phosphorus compounds via hydrolysis of phosphoester bonds. Among the three prokaryotic alkaline phosphatase groups known to date, PhoA, PhoX, and PhoD, PhoD was more abundant than the other two groups among global ocean sampling (GOS) metagenomes28. Given that all Kordia PhoD proteins contain signal peptides, it is likely that Kordia strains secrete PhoD into the periplasm or extracellular milieu and subsequently uptake the phosphate released from dissolved organic phosphorus compounds. The possession of secretable alkaline phosphatase suggests that the genus Kordia may be adapted to coastal waters replete with organophosphate compounds. Whether and how the expression of PhoD is regulated by the PhoPR system in the Kordia strains, however, remains unclear.

Secondary metabolites

Analyses by antiSMASH showed that the IMCC3317T genome contained nine putative biosynthetic gene clusters (BGCs) with lengths ranging from 21 to 225 kb. These BGCs included a range of clusters, such as NRPS, T1PKS [Type I polyketide synthase (PKS)], T3PKS (Type III PKS), transAT-PKS (Trans-AT PKS), CDPS (tRNA-dependent cyclodipeptide synthase), arylpolyene, lanthipeptide, and terpene. Considering that BGCs are found rarely in marine bacterial groups that have streamlined genomes and are abundant in pelagic waters (e.g. the SAR11 clade), the possession of multiple BGCs that can produce secondary metabolites may be an adaptation to coastal waters where the interactions with other organisms are expected to be more active than pelagic ocean.

The longest BGC, spanning ~ 225 kb (802,427–1,027,059 bp), was identified as a polyketide-NRPS hybrid gene cluster and predicted to synthesize a polymer consisting of at least 30 amino acids (Suppl. Fig. S1). The antiSMASH results showed that gene clusters similar to this BGC were found in certain flavobacterial strains, such as Flavobacterium spartansii MSU, Flavobacterium chilense DSM 24724, Flavobacterium johnsoniae UW101, Flavobacterium sp. WG21, and Kordia zhangzhouensis MCCC 1A00726, suggesting wide distribution among the family Flavobacteriaceae. Of note, however, none of these gene clusters have been fully characterized, leading to a low similarity between this K. antarctica BGC and those curated in the MIBiG, a database of BGCs and their products. This K. antarctica BGC shared only 3% to 4% of its genes with the most similar BGCs found in the MIBiG database, and there was no homology with the core biosynthetic genes of this BGC, indicating the K. antarctica BGC and similar BGCs found in other flavobacterial genomes might synthesize novel secondary metabolites.

Carbohydrate-active enzymes (CAZymes)

Many CAZymes and SusCD proteins were found in the IMCC3317T genome, suggesting a metabolic potential for polysaccharide utilization. Members of the marine Flavobacteriaceae specialize in degradation and utilization of polysaccharides produced primarily by algae and phytoplankton3,29, making this heterotrophic bacterial group a major contributor to the turnover of dissolved or particulate high-molecular weight organic matter30. Recent studies on genomics, transcriptomics, and proteomics of representative marine flavobacterial strains showed that a diverse set of proteins, including transporters, CAZymes, and sulfatases, are critical to polysaccharide metabolism; moreover, the genes encoding these proteins are usually gathered into clusters termed polysaccharide utilization loci (PULs)8,31.

As such, we screened and analyzed PULs from the IMCC3317T genome, focusing on CAZymes and SusCD, the two major constituents of PULs, to reveal the genomic potential of marine flavobacterial strain IMCC3317T for polysaccharide metabolism. The susC and susD genes, located adjacently in many PULs, encode a TonB-dependent transporter and carbohydrate-binding lipoprotein, respectively. SusC and SusD form a complex in the outer membrane to bind and uptake carbohydrates32. CAZymes found in PULs include glycoside hydrolases (GH), glycosyltransferases (GT), carbohydrate esterases (CE), and polysaccharide lyase (PL), and are involved in the further breakdown and metabolism of various carbohydrates. Analysis of the IMCC3317T genome using dbCAN showed a total of 194 putative CAZyme proteins, including 53 GH (20 families), 57 GT (12 families), 36 CE (8 families), nine PL (5 families), and 45 carbohydrate-binding modules (14 families). PFAM and TIGRFAM-based searches for SusC and SusD revealed that the IMCC3317T genome possessed 10 SusCD pairs. An inspection of genomic regions around these pairs showed that eight SusCD pairs constituted seven PULs together with neighboring CAZymes (Fig. 4). GntR family transcriptional regulators or sulfatases were also found within many PULs. One of the PULs (PUL1, Fig. 4) contained GH families 16, 17, and 30, suggesting the utilization of laminarin8. Collectively, the IMCC3317T genome has many genes necessary for polysaccharide utilization, which may be regarded as an adaptation to coastal seas where phytoplankton and algae may produce a large amount of polysaccharides.

Figure 4
figure 4

Seven polysaccharide utilization loci (PUL) of K. antarctica IMCC3317T predicted based on eight SusCD pairs and adjacent CAZymes. Genes are colored according to function (color key at bottom). Families of CAZyme genes indicated by numbers within arrows. Numbers on either side of each PUL map indicate start and end positions in the genome. 'Others' refers to annotated genes with functions other than CAZyme, SusCD, and sulfatase.

Full size image

Nitrous oxide reductase gene (nosZ)

A nosZ gene (2713616369) encoding nitrous oxide reductase was found in the IMCC3317T genome, but not in the other Kordia genomes. The NosZ enzyme localizes to the periplasmic space and is involved in the final step of denitrification; that is, the reduction of nitrous oxide (N2O) to nitrogen gas (N2)33. According to a recent phylogenetic analysis37, there are two distinct clades of NosZ, each with different signal peptides related to different translocation pathways. While the clade I NosZ localizes to the periplasm via the twin-arginine translocation pathway, the clade II NosZ is transported via the Sec-dependent translocation pathway. An analysis of signal peptides showed that the NosZ of strain IMCC3317T belongs to clade II. Since no other enzymes involved in the denitrification pathway (such as nitrate reductase and nitrite reductase) were found in the IMCC3317T genome, like many other prokaryotes with clade II NosZ, IMCC3317T was designated as a non-denitrifying N2O reducer. It has been suggested that phylogenetically diverse non-denitrifying N2O reducers with a clade II NosZ may be important N2O sinks since these organisms can consume, but not produce, N2O in the environment34. From a climate change perspective, given that N2O is a major greenhouse gas, at around 300 times more warming potential than CO2 and implicated in ozone depletion, the existence of non-denitrifying N2O reducers, such as IMCC3317T, is critical for the conversion of N2O to harmless N235.

The nosZ gene was found only in IMCC3317T, and analyses were thereby performed to infer the phylogenetic position of the IMCC3317T NosZ. In a BLASTp search against the NCBI nr database, top hits for IMCC3317T NosZ were predominantly from marine flavobacterial genera phylogenetically related to Kordia. The IMCC3317T NosZ sequence and several best BLASTp hits were used for the construction of maximum-likelihood trees together with representative NosZ sequences36,37 and IMCC3317T NosZ, and its best BLASTp hits formed a robust branch within the Bacteroidetes group of clade II NosZ (Fig. 5). These results suggest that IMCC3317T NosZ is similar to that of other phylogenetically-related flavobacterial genera, notwithstanding the absence of NosZ in other species of the same genus. This finding is consistent with previous research indicating that NosZ phylogeny was predominantly shaped by vertical transfer, and gene gain or loss has led to a variable distribution of NosZ even among closely related organisms37,38. Further analyses of the family Flavobacteriaceae genomes (available in the IMG database) revealed that the variable distribution of NosZ among the same genera was not limited to Kordia, but widespread throughout the family Flavobacteriaceae. Many genera, including Tenacibaculum, Maribacter, Muricauda, Aquimarina, and Winogradskyella, had both NosZ-possessing genomes and NosZ-lacking genomes (Suppl. Table S1). More elaborate analyses of the phylogenomics of flavobacterial strains and phylogeny of their NosZ proteins will be necessary to reconstruct the evolutionary history of NosZ in the family Flavobacteriaceae and provide a plausible explanation for the distribution of NosZ in Kordia among the following two scenarios: all Kordia strains, except for K. antarctica IMCC3317T, have lost the nosZ gene inherited from the common ancestor of the genus Kordia; or, the common Kordia ancestor lacked the nosZ gene and strain IMCC3317T acquired it via horizontal gene transfer from closely related organisms.

Figure 5
figure 5

Maximum-likelihood phylogenetic tree of NosZ proteins from different bacterial groups. NosZ of IMCC3317T is bolded in red. Designation of NosZ clades I and II based on Hallin et al.37. Phylogenetic affiliation at phylum/class level indicated after brackets to the right of taxa names. For visualization purposes, certain NosZ genes from the same phylum/class were grouped together and described as wedges, with the number of grouped proteins indicated in parentheses following phylum/class name. Bootstrap values (≥ 70; 100 replicates) indicated above the branches.

Full size image

In many nosZ-harboring genomes, nosZ is clustered with other genes involved in protein assembly and transport of copper, a cofactor of NosZ. The nosZ gene of IMCC3317T is located within a gene cluster arranged as nosZ, nosL (copper chaperone; 2713616367), nosD (accessory protein; 2713616366), nosF (Cu-processing system ATP-binding protein; 2713616365), and nosY (Cu-processing system permease protein; 2713616364). This gene order is highly conserved among nosZ-possessing flavobacterial strains closely related to the genus Kordia.

Conclusion

The genome sequence of K. antarctica IMCC3317T reported in this study represents the first complete, published genome of the genus Kordia. Analyses of this complete genome showed that strain IMCC3317T shared certain chemoheterotroph metabolic features with other Kordia strains. Also, like the two Kordia genomes (K. algicida and K. sp. SMS9) listed in PULDB39, the IMCC3317T genome possessed many PULs, suggesting the utilization of polysaccharides originated by marine phytoplankton or algae. Based on the prediction of a nosZ gene in the IMCC3317T genome in the absence of other denitrification genes, strain IMCC3317T is considered a non-denitrifying N2O-reducing bacteria. Further studies will be required to explore whether strain IMCC3317T can actually reduce N2O, a significant greenhouse gas in climate change.

Methods

Genome sequencing

Strain IMCC3317T was resuscitated by incubating a loopful of 10% glycerol stock on marine agar 2216 plates at 20℃ for 7 days. Genomic DNA was extracted from colonies grown on these plates with a DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol. Genome sequencing of IMCC3317T was conducted on the PacBio RS II platform (Pacific Bioscience, Menlo Park, CA, USA) with a 20 kb SMRTbell library generation. De novo assembly was conducted with the SMRT Analysis RS_HGAP_Assembly.2 protocol (v2.3.0) using ~ 38,000 filtered reads (~ 583 Mbp total; N50 read length, 20,089 bp), resulting in a single contig with approximately ~ 93 × coverage. This single contig was circularized using Circlator (v1.5.5; ‘all’ command with default options)40 and polished five times with the SMRT Analysis RS_Resequencing.1 protocol until no variants were called, generating a final, error-corrected genome sequence. The GenBank accession number of the complete IMCC3317T genome is CP019288.

Genome annotation and comparison

The complete genome sequence of IMCC3317T was submitted to the IMG-ER system41 for detailed annotation and comparative analyses. Prokka (v1.12)42 was also used for local annotation (with the options ‘--rfam’ and ‘--rnammer’). Protein coding sequences, ribosomal RNA genes, transfer RNA genes, and non-coding RNAs were predicted using the tools (implemented in Prokka) Prodigal (v2.6.3), RNAmmer (v1.2), ARAGORN (v1.2), and Infernal (v1.1), respectively. The GenBank file produced by Prokka was submitted to the RAST server for improved functional annotation43 with classic RAST annotation scheme and ‘preserve original genecalls’ option. Where necessary, annotation information from Prokka and the RAST server were used for comparison with IMG-ER annotation. HMMs from dbCAN44 were used to predict and classify CAZymes with default parameters. SusCD proteins were predicted based on the presence of PFAM (for SusD; PF07980, PF12741, PF12771, and PF14322) and TIGRFAM (for SusC; TIGR04056 and TIGR04057) domains. Biosynthetic gene clusters for secondary metabolites were predicted using antiSMASH 5.045 with default settings.

Comparative analyses with four other Kordia genomes were generally performed using the functionalities provided by the IMG-ER system. All four genomes included in comparative analyses were from the type strains of the genus Kordia available in the IMG database: K. algicida OT-1 (IMG genome ID; 641380434), K. zhangzhouensis MCCC 1A00726 (2627853697), K. jejudonensis SSK3-3 (26364156330), and K. periserrulae DSM 25731 (2734482288). The completeness of the genome sequences was calculated using CheckM (v1.1.2) with default setting46. Average nucleotide identity (ANI) values between genomes were calculated using JSpeciesWS47. Strain genome sequences were downloaded and used to draw a plot for genomic comparison between the K. antarctica IMCC3317T genome and the other four Kordia genomes using BRIG48. Protein sequences were also downloaded and used for the clustering of homologous proteins among the five Kordia genomes by OrthoFinder (v2.2.7)49 at default settings, and a Venn diagram showing the number of shared or unique protein clusters was generated using the R packages ‘limma’ and ‘gplot’. Protein sequences were also analyzed by BlastKOALA (ran on August 13, 2018)50 to assign them to KEGG orthology (KO) groups and reconstruct metabolic pathways. A comparative analysis of metabolic pathways was conducted using the KO ID, EC number, COG ID, and Pfam ID of major metabolic enzymes of five Kordia genomes in searches against multiple databases, including MetaCyc. When necessary, protein sequences were searched against the NCBI nr database.

Phylogenetic analysis of NosZ protein

Protein sequences highly similar to that of the NosZ protein found in the IMCC3317T genome were searched for and retrieved from the NCBI RefSeq database using BLASTP 2.8.1+ (ran on September 11, 2018). These sequences were used for phylogenetic analyses together with 109 representative NosZ proteins selected based on previous studies on the distribution and phylogeny of NosZ37,51. The collected protein sequences were aligned using Muscle and the aligned sequences were then used for a tree building by the Maximum Likelihood method based on the JTT model using MEGA752.

We analyzed the distribution of the nosZ gene among members of Flavobacteriaceae using genomes publicly available in the IMG database (accessed at Oct, 2018). Genomes classified as Flavobacteriaceae in IMG were analyzed using the Function Profile utility. Any genome harboring gene(s) assigned to K00376, a KEGG ortholog corresponding to nitrous oxide reductase (NosZ), was considered to contain a nosZ gene. Single-cell genomes, metagenome-assembled genomes, and genomes from bacterial isolates not classified at the genus level were excluded from the analysis.