Zobellella denitrificans ZD1 (JCM 13380) were purchased from Riken BRC Microbe Division, Japan Collection of Microorganisms (JCM). FastRNA^™ Pro Blue Kit and Isopropyl-β-D-1-thiogalactopyranoside (IPTG) was purchased from MP Biomedicals (Santa Ana, CA, USA). Imperial^TM protein stain was purchased from Fisher Scientific (Waltham, MA, USA). OneStep Ahead RT-PCR Kit was obtained from Qiagen (Hilden, Germany). Power SYBR Green PCR Master Mix was purchased from Applied Biosystem (Waltham, MA, USA).

Strain ZD1 was grown aerobically in ammonium ((NH₄)₂HPO₄) and nitrate (KNO₃) modified mineral salts medium, respectively, supplied with 20 g/L glycerol and 0% or 3% NaCl (w/v) to above an optical density 600 nm (OD₆₀₀) of 1.0 (i.e., in stationary growth phase) before harvested for PHB genes and salt tolerance gene analysis. The modified mineral salts medium contained K₂HPO₄ (5 g/L), Na₂SO₄ (0.5 g/L), MgSO₄·7H₂O (0.4 g/L), and 0.1% (v/v) trace mineral solution. The trace mineral solution contained FeSO_4·7H₂O (2.78 g/L), MnCl₂·4H₂O (1.98 g/L), CoSO₄·7H₂0 (2.81 g/L), CaCl₂·2H₂O (1.47 g/L), CuCl₂·2H₂O (0.17 g/L), and ZnSO₄·7H₂O (0.29 g/L). Total RNA was extracted using FastRNA^™ Pro Blue Kit according to the manufacturer’s instruction. PHB concentration in dry cell weight (DCW) was determined using a spectrophotometric assay via conversion of PHB to crotonic acid as described previously [13–15].

The cDNA was synthesized using OneStep Ahead RT-PCR Kit according to the manufacturer’s instruction. Briefly, 1μg of total RNA was added to a 25-μL reaction mix containing Omniscript^® and Sensiscript^® reverse transcriptases, DNA polymerases, 0.5 μM primers (listed in S1 Table), 2.5 mM MgCl₂ in RNAase free water. The cDNA was obtained by the following cycling condition: 10 min of reverse transcription at 50 ^oC, 5 min of DNA polymerase activation at 95 ^oC, followed by 40 cycles of three-step amplification (95 ^oC for 20 sec, 54 ^oC for 20 sec and 72 ^oC for 20 sec, and a final extension at 72 ^oC for 2 min). The cDNA was stored at -80 ^oC before used. The cDNA was diluted 200 times before real-time quantitative PCR for PHB synthesis genes and a housekeeping gene (16S rRNA) using an IQ5 multicolor real-time PCR detection system (Bio-Rad, Hercules, CA, USA). Primer sets specific to the PHB synthesis genes and housekeeping gene were listed in S1 Table. All of the primers were synthesized by Integrated DNA Technologies (Coralville, IA, USA). Briefly, the reaction mix contained 1 μL of diluted cDNA, 10 μL of PowerSYBR Green PCR Master Mix, 0.5 μM of each of the primer sets and nuclease-free water to 20 μL. The cycling condition was conducted in 1 cycle of denaturation at 95°C for 10 min, followed by 35 cycles of three-step amplification (i.e., 95°C for 30 sec, followed by 54°C (PHB synthesis genes) or 55°C (housekeeping gene) for 45 sec, and then 72°C for 45 sec). The fluorescence intensity was measured during each three-step cycle. The Ct value, defined as the PCR cycle number that resulted in exceeding an arbitrarily signal threshold, was used as the endpoint of the real-time PCR quantification. Gene expression was calculated by Double delta Ct method as described in previous study [16]. The fold change of ZD1 in nitrate treatment was normalized to that observed in ammonium treatment.

SDS-PAGE analysis was performed to examine the expression of genes (ectA, ectB, and ectC) responsible for ectoine synthesis by strain ZD1 and compared to over-expression of each gene in the plasmid by E coli. Over-expression of these genes were induced by adding 0.2 mM IPTG to E. coli BL21 (DE3) that contains a plasmid of each of three genes constructed previously [12]. Briefly, ectA, ectB, and ectC in strain ZD1 were cloned into three pET11a vectors, using traditional cloning with T4 DNA ligase (New England Biolabs, Ipswich, MA, USA). The constructed plasmids were transformed into E. coli competent cells (NEB 5-alpha), followed by screening on LB agar plates containing 100 mg/L ampicillin. The plasmids of the colonies on the plates were then extracted by QIAprep Spin Miniprep Kit (QIAGEN, Hilden, Germany), followed by sequence confirmation. The correct plasmids were then transformed into E. coli BL21 for protein expression. The E. coli BL21 and ZD1 cultures were concentrated to OD₆₀₀ around 16.7 in sodium phosphate buffer (MP Biomedicals, Santa Ana, CA, USA) and lysed by sonication. Sonication was conducted in 6 rounds of 20 sec sonication (50% amplification) and 20 sec on ice intervals. A 15 μL supernatant of lysed cultures was mixed with 15 μL of 2X SDS buffer in a 1.5-mL microcentrifuge tube and then boiled for 10 min. SDS-PAGE was conducted on Novex^™ wedge well 14% tris-glycine gel (Invitrogen, Carlsbad, CA, USA) in XCellSureLock^™ Mini-Cell equipment (Novel Experimental Technology, San Diego, CA, USA). 1X Tris-Glycine (MP Biomedicals, Santa Ana, CA, USA) running buffer and 10 μL of the Mark12 Unstained Standard (Invitrogen, Carlsbad, CA, USA) were used. The electrophoresis was run at 90 V for 2.5 hrs. The gel was stained with Imperial^TM protein stain for 1 hr, and then destained for 2 hrs with deionized water.

The draft genome of strain ZD1 [7] was used for a phylogenetic analysis. A phylogenetic tree was constructed using ezTree [17] (https://github.com/yuwwu/ezTree), which was capable of mining single copy marker genes and generating phylogenetic trees from the concatenated alignment of the marker genes, and was visualized using MEGA5 [18]. Genes were predicted using the NCBI Prokaryotic Genome Annotation Pipeline (PGAP) [19] and used in a downstream functional analysis. Sequences for building the phylogenetic tree of the norB genes that code nitric oxide reductase [20] were downloaded from the NCBI database (https://www.ncbi.nlm.nih.gov/) according to their accession numbers and aligned using MUSCLE [21]. The phylogenetic tree was then built using FastTree [22] with JTT model.

To probe the existence of gene clusters among the Bacteria kingdom, 36,682 bacterial genomes were downloaded from the NCBI ftp site (ftp://ftp.ncbi.nlm.nih.gov/genomes/) in April 2018. Only the best-assembled strain (i.e., with the highest N50 length) of each species was retained for analysis. A total of 8,482 genomes were kept in this process. Genes were predicted from the selected genomes using Prodigal [23], and the presence of the PHB gene cluster was checked using BLASTP [24]. The phylogenetic tree of PHB gene clusters was made by separately aligning individual genes using MUSCLE [21], concatenating the alignments, and then building the tree using FastTree [22] with JTT model. The tree was then plotted and colored using the Evolview web service [25]. The conserved functional domains of phasin proteins were obtained by searching against NCBI Conserved Domain Database (CDD) [26] using the NCBI CD-Search service [27]. The phylogenetic tree of phasin proteins was constructed by first aligning genes using MUSCLE, followed by building the tree using FastTree with the JTT model, and then visualized using MEGA5 [18].

Clustered regularly interspaced short palindromic repeat (CRISPR)-Cas systems were checked using the CRISPRone web service [28]. The taxonomic distribution of CRISPR spacers was obtained as described by Shmakove et al [29]. Briefly, a total of 154,924 viral sequences was downloaded from the NCBI genome website (https://www.ncbi.nlm.nih.gov/genome/viruses/) in June 2018. The CRISPR spacers were then searched against the downloaded viral sequences using BLASTN with the following parameters: “-dust no -word_size 8 -max_target_seqs 1 -outfmt 6.” Only hits with at least 95% identity were parsed to identify the taxonomic profiles of the most closely related phages.

A complete pathway for converting glycerol to PHB was identified in the ZD1 genome (Fig 2). As shown in Fig 2A, series of reactions and enzymes are involved in the uptake and conversion of glycerol to PHB. Briefly, glpF encoding glycerol facilitator is responsible for uptake of glycerol into cells and glpK encoding glycerol kinase converts glycerol to glycerol-3-phosphate. Glycerol-3-phosphate is then converted to dihydroxyactonephosphate (DHAP) and eventually to pyruvate. Pyruvate is then oxidized to acetyl-CoA, which is converted into PHB. A PHB synthesis gene cluster, consisting of four genes: phaB, phaA, PFP (a phasin-family protein gene), and phaC, was shown in Fig 2B. The three genes (phaA,-B, and -C) encoding enzymes that are known to convert acetyl-CoA into PHB, in which PhaA served as the first enzyme to catalyze acetyl-CoA to acetoacetyl-CoA, PhaB transforms acetoacetyl-CoA to (R)-3-hydroxybutyryl-CoA, and PhaC catalyzes the polymerization and formation of PHB granule [30]. The PFP may also play a role in the formation of the PHB granule [31]. Other phasin-encoding genes, including phaP and phaF [32], along with a phaR gene (a phasin transcription factor known to bind to the phaP promoter region and block phaP expression [33]) were also found in the genome.

The PHB synthesis gene cluster consists of four genes: phaB, phaA, PFP (a gene encoding a phasin-family protein), and phaC. The order of the PHB synthesis gene cluster (i.e., phaB, -A, and -C genes) is similar to that of Azotobacter sp. FA8 [35]. However, to the best of our knowledge, the PFP between phaC and phaA in this gene cluster has not been previously documented. After searching for this four-gene-cluster in genomic sequences in the NCBI database, 117 species containing this gene cluster were identified (Fig 4). The majority of these species were Gammaproteobacteria, belonging to the Vibrionaceae family, in which the Vibrio genus accounts for 55.6% of species, followed by Photobacterium (18.8%). Only eight species (four Oceanimonas, three Oceanisphaera, and the Z. denitrificans species) were in the Aeromonadaceae family to which strain ZD1 belongs to. Almost all of these species were marine microorganisms and thus halophilic, suggesting that this gene cluster may specifically belong to ocean-dwelling bacterial organisms.

To explore evolutionary relationships among these 117 species, a phylogenetic tree was constructed from the concatenated amino acid sequences of the four genes in the PHB synthesis gene cluster (Fig 5). The Aeromonadales tree topology (including Oceanimonas, Oceanisphaera, and Zobellella) along with the order Alteromonadales generally agrees with the phylogenetic tree based on 16S rRNA gene (Fig 1B). However, several horizontal gene transfers seem to have occurred in the order Vibrioles. The most visible one might be a transfer event from Vibrio to the common ancestor of Salinivibrio, Grimontia, and Enterovibrio. A bacterial symbiont of the triplewart seadevil (Cryptopsarascouesii) and a Photobacterium species (Photobacterium damselae subsp. damselae CIP 102761) may have acquired the cluster from Enterovibrio and Vibrio, respectively. The observation suggests that this PHB synthesis gene cluster may have occasionally been horizontally transferred among different marine microorganisms.

The PFP in ZD1 was also investigated. A phaP gene was identified in the ZD1 genome and the phaP is similar to those encoding phasin proteins. The PHB synthesis gene cluster encoding proteins were assigned to the conserved protein domain family phasin_2_superfamily (cl11491), which is designated as “associated with polyhydroxyalkanoate (PHA) inclusions, commonly consisting of polyhydroxybutyrate (PHB) (https://www.ncbi.nlm.nih.gov/Structure/cdd/cddsrv.cgi?uid=325059)” (S2A Fig). The phylogenetic tree demonstrates that in their evolutionary history, the PFP identified from the ZD1 PHB gene cluster may have diverged from the phaP clade (S2B Fig).

An ectoine synthesis gene cluster (ectA, ectB, and ectC) was identified in the ZD1 genome. The structure of the ectoine gene cluster in strain ZD1 was identical to those of some Firmicutes and Proteobacteria species (Halomonas sp. QHL1 [36], Marinococcus halophilus [37], and Bacillus alcalophilus DTY1 [11]. Expression of proteins (EctA, EctB, and EctC) in strain ZD1 grown under saline conditions was confirmed using SDS-PAGE analysis. As shown in Fig 6, the enzymes produced from the recombinant E. coli BL21 were used as positive controls; a band for EctA, 19 kilo-dalton (kDa) (lane 4), a band for EctB, 45 kDa (lane 5), and a band for EctC, 15 kDa (lane 6). Comparing to the positive control (lanes 4, 5, and 6) and 0% NaCl samples (lanes 2 and 7), ammonium sample with 3% NaCl showed EctB and EctA bands clearly (lane 3) and a clear band for EctB was shown in nitrate sample with 3% NaCl (lane 8). However, the faint band for EctC was only observed in ammonia sample with 3% NaCl (lane 3).

Strain ZD1 has been shown to perform complete denitrification, i.e., converting nitrate to nitrogen gas and its ability is supported by a complete set of denitrifying genes identified in the ZD1 genome (Fig 7A). Fig 7B shows genes and gene clusters associated with denitrification process in ZD1. They are i) a gene cluster, napFDAGHBC, encoding a periplasmic nitrate reductase responsible for conversion of nitrate to nitrite; ii) a nirK gene encoding a copper-containing nitrite reductase which can convert nitrite to nitric oxide; iii) a norB gene encoding nitric oxide reductase subunit B that is involved in conversion of nitric oxide to nitrous oxide; and iv) a nosZ gene encoding nitrous-oxide reductase that is involved in conversion of nitrous oxide to nitrogen gas.

The structure of the periplasmic nitrate reductase nap gene cluster is very similar to those in E. coli [41], Haemophilus influenza [38], Rhodoacter sphaeroides [42], and Magnetospirillum gryphiswaldense [40]. NapF and NapD involve in post-translation of naturalization of NapA, while NapG, NapH, NapB and NapC assist electron transport to NapA (as demonstrated by Sparacino-Watkins et al. [43]). Two nitrite reductase genes (nirK and nirS) are functional biomarkers that are commonly used to investigate denitrifying bacteria [44]. In strain ZD1, only nirK was identified. The lack of a norC gene (which is usually found near or next to the norB gene) suggests that ZD1 may belong to single-component type of norB gene (quinol-oxidizing single-subunit class; qNorB) which is similar to that of Ralstonia eutropha H16 [45]. The phylogenetic tree of qnorB and cnorB (encoding cytochrome bc-type complex, cNorB) (S3 Fig) indicates that the ZD1 norB gene may be a single-component qNorB type.

An assimilatory nitrate reductase gene cluster (nasFEDRACB) was also identified in the ZD1 genome (Fig 7C and 7D). The cluster encodes enzymes involved in transport of nitrate or nitrite into the cytoplasm, followed by the conversion of nitrate to ammonia using nasA-encoded NADH-dependent nitrate and nitrite reductases [46]. This assimilatory nitrate reductase gene cluster is similar to that in Klebsiella oxytoca M5a1 (an assimilatory nitrate reductase gene cluster of K. oxytoca M5a1 is nasRFEDCBA) [46]. In Klebsiella, nasFED genes encode a typical ABC transporter for both nitrate and nitrite [47]. Thus, the presence of nasFED genes suggests that strain ZD1 might be able to uptake nitrate through cellular membranes. Differences in the structures of the assimilatory nitrate reductase gene clusters of ZD1 and K. oxytoca were observed. For example, the nasR gene, encoding an anti-terminator, is located in the downstream of the nasFED gene cluster, not the upstream. Additionally, nasACB genes are five genes away from the nasRFED gene cluster.

A two-gene cluster, including nirB and nirD, encoding a NADH-dependent nitrite reductase for converting nitrite to ammonia, was found in the ZD1 genome. This gene cluster structure is similar to those found in E. coli [48] and Staphylococcus carnosus [49]. Mycobacterium tuberculosis also possesses these two genes [50]. This nitrite reductase was reported to be mainly functional under anaerobic conditions [48, 49]; however, it was also found to be functional during aerobic growth in M. tuberculosis when nitrite was served as the sole nitrogen source [48, 49, 51]. Since ZD1 is a facultative anaerobe, we postulate that the nitrite reductase encoded by the nirBD gene cluster also functions under anaerobic conditions.

A CRISPR-Cas system, containing nine CRISPR genes along with a CRISPR array consisting of 117 spacers, was found in the ZD1 genome. The nine CRISPR genes included two universal CRISPR genes (cas1 and cas2), five subtype-I-E genes (cas6e, cas5, cas7, cse2gr11, and cas8e), one type-I gene (cas3HD), and a WYL gene (pfam13280). The presence of the CRISPR-Cas system and the vast amount of spacers indicate that the ZD1 strain may have been a host for 117 different phages and thus acquired or developed defense mechanisms against them.

As the genetic structure of this ZD1 CRISPR-Cas system is similar to that of E. coli, we searched some closely-related genomes, including the Aeromonadaceae family and Vibrio cholerae, in order to determine the distribution of this CRISPR-Cas system. Among 62 species, only 13 harbored the CRISPR-Cas system, including two Aeromonas, three Oceanimonas, two Oceanisphaera, one Tolumonas, two Vibrio cholerae species, E. coli, and Z. denitrificans strain ZD1. Fig 8 shows the distribution of the CRISIR-Cas system on the phylogenetic tree of the 62 species shown in Fig 1B. The results suggested that the CRISPR-Cas system may have been undergone extensive horizontal transfer events in the evolutionary history of the species. For example, the three Oceanimonas species do not share the same CRISPR system, in which two are subtype-I-E and one is subtype-I-F. The two closely related Vibrio cholerae species also harbor different CRISPR systems; not to mention that the three Aeromonas species contain three different CRISPR systems. Interestingly, we also observed that another Z. denitrificans strain F13-1 (referred as strain F13-1 hereafter), did not contain the CRISPR system, suggesting either that ZD1 horizontally acquired the CRISPR-Cas system from other species or that F13-1 may have lost this system.

An evolutionary tree for the cas3 gene was built, as this gene is shared by all probed species containing the CRISPR system. As shown in Fig 9, the CRISPR-Cas systems of strain ZD1 and E. coli are not very closely related despite their highly similar gene order. Instead, the cas3 gene of strain ZD1 is most closely related to those of two Oceanimonas species, suggesting that Z. denitrificans may have inherited the CRISPR-Cas system from the common ancestors, and that strain F13-1 may have recently lost the CRISPR-Cas system. The cross-comparison between the cas3 tree (Fig 9) and the CRISPR-marked phylogenetic tree (Fig 8) further suggests that CRISPR-Cas systems have been extensively exchanged among different oceanic species.

The taxonomic distribution of the ZD1 CRISPR spacers was also explored in order to determine phages that may commonly infect this species. As shown in Fig 10, 52 out of 117 spacers (i.e., 44%) belonged to double-stranded (ds) DNA viruses, and almost half of the dsDNA viruses were Caudovirales. The other 12% belongs to ssRNA-virus, ssDNA-virus, and dsRNA-virus. However, another 44% of spacers could not be identified using available viral sequences database. The diversification of viral spacers in the ZD1 CRISPR spacers suggested that strain ZD1 may have been the host target of a variety of phages. CRISP-Cas is an immune system allows a bacterium to fight against phage infection. Bacteria incorporate small fragments of phage DNA/RNA into their genome as spacers which were derived from the phages that the host bacteria previously infected. In the event of subsequent infection, these spacers work together with the Cas protein to recognize and digest the phage DNA/RNA. In this study, the high number of spacers found in the ZD1 genome would allow the strain to fight against infection of phages that they previously infected and such a strong phage immune system in the strain might explain the fail attempts in our previous efforts in isolating environmental phages to infect stain ZD1.