Genomic, physiologic, and proteomic insights into metabolic versatility in Roseobacter clade bacteria isolated from deep-sea water

Roseobacter clade bacteria are ubiquitous in marine environments and now thought to be significant contributors to carbon and sulfur cycling. However, only a few strains of roseobacters have been isolated from the deep-sea water column and have not been thoroughly investigated. Here, we present the complete genomes of phylogentically closed related Thiobacimonas profunda JLT2016 and Pelagibaca abyssi JLT2014 isolated from deep-sea water of the Southeastern Pacific. The genome sequences showed that the two deep-sea roseobacters carry genes for versatile metabolisms with functional capabilities such as ribulose bisphosphate carboxylase-mediated carbon fixation and inorganic sulfur oxidation. Physiological and biochemical analysis showed that T. profunda JLT2016 was capable of autotrophy, heterotrophy, and mixotrophy accompanied by the production of exopolysaccharide. Heterotrophic carbon fixation via anaplerotic reactions contributed minimally to bacterial biomass. Comparative proteomics experiments showed a significantly up-regulated carbon fixation and inorganic sulfur oxidation associated proteins under chemolithotrophic conditions compared to heterotrophic conditions. Collectively, rosebacters show a high metabolic flexibility, suggesting a considerable capacity for adaptation to the marine environment.


Results and Discussion
General genome organization and content. The general features of the T. profunda JLT2016 and P.
abyssi JLT2014 genome are summarized in Table 1 and Fig. 1. Both genomes are comprised of a single circular chromosome and 8 plasmids. The T. profunda JLT2016 and P. abyssi JLT2014 chromosome size is 4.61 Mbp and 4.25 Mbp long, harboring a total of 4,539 and 4,204 predicted protein-coding sequences (CDSs) with a GC content of 67% and 66%, respectively. The plasmid size accounts for 15% and 19% of the total genome size, respectively. This is a moderate amount in known complete roseobacters genomes with plasmids, which possess between 2 and 11 plasmids in addition to their chromosome and the 2-33% plasmid borne in their genomes 32 . The average nucleotide sequence similarity between T. profunda JLT2016 and P. abyssi JLT2014 is 81%. Bidirectional BLAST analyses showed that approximately 3,000 genes exhibited > 60% sequence identity among T. profunda JLT2016, P. abyssi JLT2014 and P. bermudensis HTCC2601, indicating identical or equivalent function. A total of 665 and 537 genes in T. profunda JLT2016 and P. abyssi JLT2014, respectively, shared a pangenome with other non-Roseobacter clade bacteria.

Major functions shared by two deep-sea roseobacters.
The presence of gene sets in their chromosomes and plasmids assumed to be important for inhabiting marine environments within the genomes of the two strains were selected and summarized in Fig. 1 and Supplementary Table S1. T. profunda JLT2016 and P. abyssi JLT2014 shared major metabolic modules (Fig. 2).
Carbon fixation and sulfur oxidation. A striking genomic feature of the two strains is that they have the potential for CO 2 fixation via the CBB cycle and inorganic sulfur oxidation, suggesting potential chemoautotrophic ability (Fig. 2). The complete CBB operons in T. profunda JLT2016 and P. abyssi JLT2014 showed high sequence identity and conserved gene structure ( Supplementary Fig. S1). Phylogenetic analysis showed that RuBisCO in T. profunda JLT2016 and P. abyssi JLT2014 belong to the IC class RuBisCO (Supplementary Fig. S2). The Sox enzyme complex (soxABXYZ) for oxidation of thiosulfate (S 2 O 3 2− ) to elemental sulfur was found in their genomes ( Supplementary Fig. S1). Sox is proposed to produce energy and reducing power for carbon fixation via a reverse and forward electron transfer of sulfur oxidation (Fig. 2) 32 . The components of forward and reverse electron transport chains involved in sulfur oxidation include cytochrome c proteins, cytochrome bc1 complex and NADH (or NADPH)-quinone oxidoreductases ( Fig. 2) 33 . Both genomes include terminal cytochrome oxidase genes for low (types aa3 and bo3) and high (types cbb3 and bd) affinity cytochrome c oxidases 34 , suggesting adaptation to growth in a wide range of oxygen concentrations. Other genes encode enzymes for the oxidation of reduced sulfur compounds, including sulfide quinone oxidoreductase and reverse dissimilatory sulfite reductase complex, which are responsible for the oxidation of sulfide and sulfite, respectively (Supplementary Table S1).
Central carbon metabolism. Genes for a complete tricarboxylic acid cycle (TCA), Embden-Meyerhof-Parnas (EMP, glycolysis), Entner-Doudoroff (ED) and pentose phosphate pathway (PP) were identified in both deep-sea roseobacter genomes (Fig. 2). Additionally, both genomes contained genes encoding enzymes in anaplerosis (pyruvate carboxylase) and cataplerosis (malic enzyme and phosphoenolpyruvate carboxykinase) that work together to ensure the appropriate balance of carbon flow into and out of the TCA cycle 35 . Moreover, malic enzyme may act as a major anaplerotic factor in Roseobacter denitrificans as well, which catalyzes the reversible oxidative decarboxylation of malate to pyruvate and CO 2 17 .
Transporter systems. RCB contain the genes for a relatively high number of ATP-binding cassette systems (ABC transporters) and tripartite ATP-independent periplasmic (TRAP) systems compared to other bacterial groups 36 , which allow them to efficiently import organic matter. Consequently, they are adapted for dilute, heterogeneous growth substrates 12 Table S2). The type IV secretion systems are present in chromosomes of T. profunda JLT2016 or P. abyssi JLT2014 ( Supplementary Fig. S3), which are commonly used for both DNA and protein transfer between bacteria and between bacteria and host 37 .
Exopolysaccharide (EPS) biosynthesis. A number of genes associated with putative production and biosynthesis of exopolysaccharides present in P. abyssi JLT2014 and T. profunda JLT2016 (Supplementary Table S1) showed high sequence identity to those in an exopolysaccharide-producing RCB strain, Salipiger mucosus DSM16094, found in soil 38 . Two roseobacter bacteria share other genetic potentials for likely an adaption to the marine environments characterized by low-nutrient ("Nitrogen and phosphorus acquisition", "Motility and chemotaxis" in the Supplementary Material), low temperature and high-pressure conditions ("Responses to stress" in the Supplementary Material).
Comparative genomics of marine roseobacters. T. profunda JLT2016 and P. abyssi JLT2014 show substantial genomic dissimilarity. For instance, T. profunda JLT2016 contains gene clusters associated with hydrogen oxidation and pectin utilization, whereas P. abyssi JLT2014 contains a CRISPR-Cas system (Fig. 1), indicating a high potential for individual adaptations ("Distinct functional gene clusters for two deep-sea roseobacters" in the Supplementary Material). The CRISPR-Cas system is also found in other RCB species, Dinoroseobacter shibae DFL-12. (Supplementary Table S3).
Similar to other many roseobacter genomes, two roseobacter genomes exhibit high plasticity with the rearrangement of large portions of its genome and gene acquisition, as evidenced by abundant transposases, and the presence of the conjugative plasmid, prophage-like elements and genomic islands (Supplementary Table S3, "Mobile genetic elements" and "Genomic islands" in the Supplementary Material). The genomic plasiticity might shape diverse metabolic adaptations to a wide diversity of ecological niches, as has been observed for RCB.
Scientific RepoRts | 6:35528 | DOI: 10.1038/srep35528 cluster (Fig. 3). Most genes in the Sox, CBB and H 2 gene clusters had high sequence identity (> 70%) and conserved structures (Supplementary Figs S1 and S4). T. profunda JLT2016 and P. abyssi JLT2014 have CO dehydrogenase for aerobic CO oxidation but lack AAPB genes and DMSP utilization genes frequently found in roseobacters isolated from the surface water and alga-associated samples. Heterotrophy is essential for RCB success in the costal ocean whereas photoheterotrophy provide a survival advantage for AAPB in RCB in surface waters of the oligotrophic open ocean 12,16 . Pervious studies demonstrated that the Gammaproteobacteria (SUP05 clade and Oceanospirillales) and the Deltaproteobacteria (SAR324 clade) lineages that are ubiquitous in the dark oxygenated ocean are likely mixotrophs and have the potential for autotrophic CO2 fixation, coupled to the oxidation of reduced sulfur compounds 23 . Similarly, RCB bacteria within the Alphabacteria have the versatile metabolic potentials (heterotrophy, lithotrophy and authotrophy) that are beneficial for thriving at lower nutrients in the deep-sea than those in the surface water 39 . Growth experiments and physiological characteristics of T. profunda JLT2016. T. profunda JLT2016 could grow autotrophically (sodium bicarbonate as carbon source) and hetotropically (glucose as carbon source) by generating energy from chemolitotrophic sulfur oxidation (thiosulfate as an energy resource) ( Fig. 4A-C). Moreover, T. profunda JLT2016 is able to grow with thiosulfate as an energy resource and using both sodium bicarbonate and glucose as carbon sources for growth (termed mixotrophy) ( Fig. 4A-C). The new generations of sulfate concentrations under chemolithotrophic conditions are approximately 1.9-fold change relative to the added thiosulfate concentrations after exhaustion of thiosulfate, suggesting that bacteria could produce minimal amounts of elemental sulfur during thiosulfate oxidation, in contrast to the aerobic sulfur oxidizer Gammaproteobacteria SUP05 clade 40 . There was a significant difference in bacterial abudance under five culture conditions (one-way ANOVA, p < 0.05). The maximum final cell density during chemolithoheterotrophic growth (mean ± SD, 9.26 ± 0.31 cells × 10 7 /mL) was higher than that in heterotrophic growth with glucose (4.53 ± 0.08 cells × 10 7 /mL) (Fig. 4A), suggesting that heterotrophic cells could regulate the energy metabolism and biosynthesis after being supplied with thiosulfate as an alternative energy resource. R. denitrificans Och 114 has been reported to use anaplerotic pathways mainly via malic enzyme to fix 10% to 15% of protein carbon 17 . In contrast, 13 C content in T. profunda JLT2016 when grown with glucose and NaH 13 CO 3 is signifcantly less than that in autotrophic and mixotrophic cells (paired t-test, p < 0.001) (Fig. 4C), accounting for only 0.5-1.2% of the total carbon of cells. This finding suggests that the level of anaplerotic carbon fixation is low and bacterial growth relies primarily on heterotrophic metabolism (herein referred as heterotrophy). The 13 C content in mixotrophic cells at 24 h was close to that in other growth phase, when cluture medium contained more than half of glucose added (Fig. 4B,C), indicating that autotrophic carbon fixation and heterotrophic metabolic pathways occur simultaneously. The average 13 C content during the mixotrophic growth is only approximately 1/3 that during autotrophic growth (Fig. 4C), indicating that glucose was assimilated at the expense of chemolithotrophic energy, resulting in less carbon fixation.
The average cellular carbon (C) and nitrogen (N) ratio (4.09) of T. profunda JLT2016 is lower than the previously estimated for marine heterotrophic bacteria (4.31) 41 , but close to that estimated for marine dissolved organic carbon (4.13) 41 . Moreover, cellular C/N ratios varied very little among different trophic growth cells ( Supplementary Fig. S5), suggesting that the trophic strategy does not have a great impact on cellular biochemical composition. Autotrophic growth cells yield the lowest amount of poly-3-hydroxybutyrate (PHB), which is a cellular carbon storage material (Fig. 4D). In contrast, autotrophic cells yield the highest amount of EPS, followed by cells grown on glucose and sodium bicarbonate. Specifically, their EPS accounted for 25.9% ± 2.9% and 25.5% ± 4.5% of dry cell weight, respectively, which was significantly higher than that observed for cells cultivated under other growth conditions (paired t-test, p < 0.05) (Fig. 4D). EPS biosynthesis consumed a great deal of energy and carbon, which may have been responsible for the relatively low cell densities in the presence of glucose and sodium bicarbonate mixture compared to those grown in the presence of glucose (Fig. 4A). EPS helps deep-sea bacteria endure extremes of temperature, salinity and nutrient availability 42 . Moreover, EPS can speed the rate of substance uptake and trap dissolved organic matter in the marine environment 42 . During T. profunda JLT2016 growth, microbial aggregates were detected in the liquid media and their aggregation ability was observed in by electron micrographs (Supplementary Fig. S6). Furthermore, aggregated bacterial cells surrounded by a continuous film of extracellular substances were observed by confocal laser scanning microscopy ( Supplementary Fig. S7), indicating that T. profunda JLT2016 has biofilm-forming ability. Bacterial aggregates and biofilms enhance their competition and adaptation to the environment 43 . EPS may be involved in cell aggregation and biofim formation as previously suggested 43 .
Comparative proteomics of T. profunda JLT2016. In total, 2,656 iTRAQ-labeled proteins, including 283 proteins on plasmids, were identified. Of these, 1,621 displayed significant differential expression between at least two growth conditions at early stationary phase (fold change > 1.2 or < 0.8 adopted in most iTRAQ studies based on accuracy and resolution of iTRAQ quantitation 44,45 . Overall, compared to heterotrophic growth, most of proteins in the CBB cycle and the Sox enzymes during autotrophic growth showed significantly higher abundance (Fig. 5). For example, the cbbL (ribulose-bisphosphate carboxylase large subunit) and soxB (sulfur-oxidizing protein soxB) proteins significantly increased by 3.8-and 4.2-fold, and 3.1-and 3.3-fold changes, respectively, in autotrophic cells relative to those in two heterotrophic cells grown with glucose, and glucose and sodium bicarbonate, whereas the mRNA abundance of the two genes relative to that of heterotrophic cells showed 8.2-and 5.5-fold changes, and 5.4-and 11.5-fold changes upon qRT-PCR, respectively ( Supplementary Fig. S8). Moreover, cytochrome bc1 complexes, cytochrome c proteins, and a NAD(P)H-ubiquinone oxidoreductase (Fig. 5), which are predicted to be components of a reverse and forward electron transfer of sulfur oxidation, were up-regulated in autotrophic cells relative to those in heterotrophic growth. Cytochrome c oxidase type cbb3 in an electron transport was the only detected terminal oxidase under different conditions. Furthermore, most of proteins in the CBB cycle and the Sox enzymes in chemolithoheterotrophic and mixotrophic growth were present at higher levels than those in heterotrophic growth (Fig. 5). These findings combined with physiological evidence confirmed the existence of autotrophic and chemolithotrophic metabolic pathways in T. profunda JLT2016, as shown in Fig. 2.
When compared to other growth, key enzymes of central metabolisms were significantly up-regulated in heterotrophic growth including edd (phosphogluconate dehydratase) for ED pathways, and zwf (glucose-6-phosphate dehydrogenase) and pgl (6-phosphogluconolactonase) shared by ED pathway and PP pathway (Fig. 5). Most of enzymes in the TCA cycle in heterotrophic bacteria were also significantly upregulated than other growth (Fig. 5). ATP, NADPH, and NADH concentrations were higher in hetrotrophic cells than that in chemolithoheterotrophic and mixotrophic cells at early stationary phase ( Supplementary Fig. S9, paried t-test, p < 0.01), thereby indicating an enhanced energy and reducing power production from central metabolism in heterotrophic cells. Bacterial growth is crucially dependent on protein synthesis and thus on the number of ribosomes 46 . Specifically, upregulated ribosomal protein expression was observed during chemolithoheterotrophic and mixotrophic growth ( Supplementary Fig. S10), thereby indicating bacteria enhanced protein biosynthesis and futher increased cell abundance. Although chemolithoheterotrophic and mixotrophic growth required more energy for biosynthesis than heterotrophic growth, sulfur oxidation provides another ATP resource and reduces ATP production through oxidative phosphorylation, as indicated by the upregulation of the Sox enzymes and downregulation of the TCA cycle proteins (Fig. 5). These then gave rise to increased carbon flux toward biomolecules biosynthesis for cells under chemolithoheterotrophic and mixotrophic growth compared to heterotrophic growth. The expression of malic enzyme and pyruvate carboxylase was not affected by the switch from heterotrophic growth on glucose to heterotrophic growth on glucose and sodium carbonate (Fig. 5), indicating that anaplerotic carbon fixation in T. profunda JLT2016 did not appear to be important for bacterial growth. When compared to other types of growth, autotrophic growth upregulated phosphoenolpyruvate carboxykinase and the gluconeogenic enzyme fructose-1,6-bisphosphatase (fbp), which are required to bypass the irreversible step in glycolysis (Fig. 5). The increasing gluconeogenesis might lead to increased carbon flux to hexoses that are used in polysaccharides production. Most EPS biosynthesis proteins have relatively high abundance in autotrophic growth and heterotrophic growth in the presence of glucose and sodium bicarbonate mixture (Supplementary Fig. S10).
Transporter was the major functional category of detected proteins in all samples, accounting for 8.2% of the total. The most frequently observed transporters were ABC transporters (154/219) such as sugar, amino acid, and dipeptide/oligopeptide. The next most frequently observed transporter is TRAP (36/219). A SulP family transporter predicted to take up sodium bicarbonate was significantly up-regulated during autotrophic and mixotrophic growth, whereas glucose transporter was down-regulated in autotrophic growth (Supplementary Fig. S11). The transporters for ammonia, phosphate and sulfate were also detected ( Supplementary Fig. S11). Bacteria expressed not only transporters for essential nutrients, but also transporters for other originally non-existent nutrients in media. For example, autotrophic growth cells have higher expression of a glycerol ABC transporter system in a plasmid (Supplementary Fig. S11). Even though, hydrogen gas, hydrogen sulfide and urea were not selected as growth substrates in this study, hydrogenase, sulfide quinone oxidoreductase and urease were detected in all samples (Supplementary Fig. S11). A previous study showed that approximately 50% of proteomes in several roseobacter strains represent an opportunistrophic gene pool, allowing rapid bacterial response to specific environmental changes 11 . Similarly, T. profunda JLT2016 not only has the bare minimum proteome for survival under different trophic conditions, but also follows an opportunitrophic lifestyle of maintaining broad functional potential to exploit spatially and temporally variable substrates as carbon and energy resources.

Methods
Genome sequencing. T. profunda JLT2016 and P. abyssi JLT2014 DNA were extracted using a TIANamp Bacteria DNA Kit (Tiangen, Beijing, China). The genome sequencing was performed at the Chinese National Human Genome Center at Shanghai. The genome of T. profunda JLT2016 was sequenced using a massively parallel pyrosequencing technology (GS FLX+ , Roche Diagnostics, Indianapolis, IN, USA). Briefly, 5 μ g of sample DNA was sheared to 500-1000 bp by Covaris S2 (Covaris, Woburn, MA, USA) and purified using AMPure Beads (Beckman-Coulter, Fullerton, CA, USA). A DNA sequencing library was generated using a GS DNA Library Preparation Kit (Roche Diagnostics, USA) and a GS emPCR Kit (Roche Diagnostics, USA) for emPCR. A total of 248,283 reads counting up to 133.79 Mbp were obtained, covering 25.3-folds of the genome. Assembly was performed using Newbler (v2.7) and produced 136 contigs ranging from 500 bp to 383,719 bp. Relationships of the contigs were determined by multiplex PCR. Gaps were then filled in by sequencing PCR products using ABI 3730xl capillary sequencers (Applied Biosystems, Foster city, CA, USA). Finally, sequences were assembled using Phred, Phrap and Consed software packages (http://www.phrap.org/), and low quality regions of the genome were resequenced.
Whole genome sequencing of P. abyssi JLT2014 was accomplished using a hybrid approach, combining Illumina short read data with PacBio long read data 47  Flow cytometry measurements. Cells were stained with SYBR Green Ι (Invitrogen, Eugene, OR, USA) for 15 min, after which the abundance was monitored by EPICS ALTRA II flow cytometry (Beckman-Coulter, USA).

Biochemical measurements. Glucose concentration in growth media was determined using a Glucose
Colorimetric/Fluorometric Assay Kit (Biovision, Milpitas, CA, USA). Concentrations of NADH, NADPH and ATP were chemically tested with NAD + /NADH and NADP + /NADPH quantification kits (Biovision, USA) and an ATP Assay Kit (Abcam, Cambridge, MA, USA) according to the manufacturer's instructions. Sulphate and thiosulfate in the culture medium measurements were carried out using a Dionex ICS-2500 ion chromatography system (Dionex, Sunnyvale, CA, USA). To analyze the elemental composition (C and N) samples were evaluated using a Vario EL Cube (Elementar Analysen Syetem GmbH, Germany). The PHB content in cells was determined by acid hydrolysis 52 with HPLC using a Dionex ASI-100 autosampler injector (Dionex, USA) equipped with an Aminex HPX-87 H ion exchange organic acids column (300 × 7.8 mm) (BioRad, Hercules, CA, USA). Exopolysaccharides were determined by the phenol-sulfuric acid method 53 .
Carbon isotope measurements. For analysis, 13 C-labeled NaHCO 3 (99% 13 C, Cambridge Isotope Laboratories, MA, USA) was added into the culture medium as described above. Each sample was collected at different culture times using 0.3-μ m glass fiber filters (GF-75, Advantec, Japan) that had been weighed and burned at 450 °C for 4 h. Sample filters were then freeze dried for about 2 days, after which they were weighed and packed into tin cans. The 13 C content in cells was determined using a Flash EA 1112 Series elemental analyzer coupled with a ConfloIII interface to a Delta V Advantage isotope ratio mass spectrometer (Thermo Electron, San Jose, CA, USA). All defined samples were collected and analyzed in triplicate.
Scientific RepoRts | 6:35528 | DOI: 10.1038/srep35528 iTRAQ-based quantitative proteomic analysis. Isobaric tags for relative and absolute quantitative (iTRAQ) MS/MS were used to analyze the proteome of T. profunda JLT2016 under different growth conditions as described above. The iTRAQ-MS/MS was performed at the Rearch Center for Proteome Analysis, Shanghai Institute for Biological Sciences, Chinese Academy of Sciences. Total proteins were extracted before iTRAQ analysis. Briefly, cell suspensions at early stationary phase (120 h for autotrophic cells, 48 h for mixotrophic cells and chemolithoheterotrophic cells, and 62 h for heterotrophic cells) were pelleted by centrifugation (7,000g, 4 °C, 15 min). After washing pellets with 0.2 M potassium phosphate buffer (pH 7.4), cells were added to STD buffer (4% SDS, 100 mM DTT, 150 mM Tris-HCl pH 8.0) and incubated in a water bath at 99 °C for 5 min. To improve cell lysis, samples were sonicated for 10 s, interval 15 s, and lasting for 2-3 min. After centrifugation (16,000g, 4 °C, 30 min), the supernatants were transferred into new tubes and prepared for subsequent analysis.
Protein digestion was performed according to a previously described FASP procedure 54 , after which the resulting peptide mixture was labeled using the 8-plex iTRAQ reagent according to the manufacturer's instructions (Applied Biosystems, USA). iTRAQ labeled peptides were fractionated by strong cation exchange chromatography (SCX) using the AKTA Purifier system (GE Healthcare Amersham Biosciences, USA) as previously described 55 . The collected fractions were finally combined into 10 pools and desalted on SPE C18 Cartridges (Empore ™ standard density bed I.D. 7 mm, volume 3 ml, Sigma-Aldrich, USA). Each fraction was concentrated by vacuum centrifugation and reconstituted in 40 μ l of 0.1% (v/v) formic acid. Experiments were performed on a Q Exactive mass spectrometer (Thermo Electron, USA) coupled to an Easy nLC (Thermo Fisher Scientific, Waltham, MA, USA) according to a previous method 55 .
MS/MS spectra were searched using the MASCOT engine (Matrix Science, London, UK; v2.2) embedded into Proteome Discoverer 1.3 (Thermo Electron, USA) against T. profunda JLT2016 protein sequences and the decoy database. The MASCOT parameters were set as follows: peptide mass tolerance, 20 ppm, MS/MS tolerance, 0.1 Da, trypsin enzyme with up to 2 missed cleavages, fixed modification of iTRAQ 8-plex (K), iTRAQ 8-plex (N-term), variable modification of oxidation (M), and decoy database pattern of Reverse. A protein quantified with at least three peptides in experimental replicates was kept for further analysis (p < 0.05). Final ratios of protein quantification were then normalized by the median average protein quantification ratio for unequally mixed differently labeled samples.