Carbohydrate Metabolism and Carbon Fixation in Roseobacter denitrificans OCh114

The Roseobacter clade of aerobic marine proteobacteria, which compose 10–25% of the total marine bacterial community, has been reported to fix CO2, although it has not been determined what pathway is involved. In this study, we report the first metabolic studies on carbohydrate utilization, CO2 assimilation, and amino acid biosynthesis in the phototrophic Roseobacter clade bacterium Roseobacter denitrificans OCh114. We develop a new minimal medium containing defined carbon source(s), in which the requirements of yeast extract reported previously for the growth of R. denitrificans can be replaced by vitamin B12 (cyanocobalamin). Tracer experiments were carried out in R. denitrificans grown in a newly developed minimal medium containing isotopically labeled pyruvate, glucose or bicarbonate as a single carbon source or in combination. Through measurements of 13C-isotopomer labeling patterns in protein-derived amino acids, gene expression profiles, and enzymatic activity assays, we report that: (1) R. denitrificans uses the anaplerotic pathways mainly via the malic enzyme to fix 10–15% of protein carbon from CO2; (2) R. denitrificans employs the Entner-Doudoroff (ED) pathway for carbohydrate metabolism and the non-oxidative pentose phosphate pathway for the biosynthesis of histidine, ATP, and coenzymes; (3) the Embden-Meyerhof-Parnas (EMP, glycolysis) pathway is not active and the enzymatic activity of 6-phosphofructokinase (PFK) cannot be detected in R. denitrificans; and (4) isoleucine can be synthesized from both threonine-dependent (20% total flux) and citramalate-dependent (80% total flux) pathways using pyruvate as the sole carbon source.


Introduction
Two of the most important sources of carbon sinks known in nature are absorption of CO 2 by the oceans and photosynthesis by photosynthetic organisms. The marine Roseobacter clade are potentially major contributors to global CO 2 fixation as they make up 10-25% of the total microbial community in some surface ocean ecosystems [1][2][3][4]. Some members of the Roseobacter clade belong to a group known as Aerobic Anoxygenic Phototrophs (AAPs), the only known organisms performing photosynthesis requiring oxygen but not producing oxygen, while other members are non-phototrophic. CO oxidation was confirmed experimentally for the non-phototrophic Roseobacter clade bacterium Silicibacter pomeroyi [5] and other Roseobacter clade [6], and CO 2 fixation was suggested in several marine AAPs [7], while bioinformatic analysis in Roseobacter clade with completed genome sequence indicated that the genes encoding ribulose bisphosphate carboxylase/oxygenase (RUBISCO) and phosphoribulokinase required in the Calvin cycle for carbon fixation, as well as genes for other autotrophic CO 2 fixation pathways, are missing in these bacteria [5,[8][9][10]. Thus, it has been of great interest to determine how Roseobacter clade bacteria can fix CO 2 , if they indeed fix CO 2 .
Anaplerotic pathways have been proposed as an alternative mechanism for CO 2 fixation in Roseobacter clade [5,8,11], but have not been verified experimentally. Given that organisms in the Roseobacter clade are known to require organic carbon sources for growth [12,13], understanding how Roseobacter clade bacteria utilize organic carbon and assimilate CO 2 will help us understand the bio-energy metabolism, production of bioactive metabolites, and roles of global carbon cycle in these wide-spread marine bacteria.
Here, we report metabolic and biochemical studies of Roseobacter denitrificans OCh114 [14], which can denitrify, as its name indicates [13], and produce bacteriochlorophyll a (BChl a) aerobically. The genomic sequence of R. denitrificans [8] suggested that the tricarboxylic acid (TCA) cycle and anaplerotic pathways are complete, and that most of the genes for carbohydrate metabolism in the Embden-Meyerhof-Parnas (EMP, glycolysis), Entner-Doudoroff (ED), and pentose phosphate (PP) pathways are present and annotated. To understand the contributions of these enzymes and reaction pathways in the metabolism of R. denitrificans, we developed a minimal growth medium containing only defined carbon source(s), optimized the growth conditions of R. denitrificans in different defined carbon sources, and used isotopomer assisted metabolite analyses, biochemical approaches, and gene expression profiles to investigate the carbon assimilation, carbohydrate utilization, and amino acid biosynthesis pathways in R. denitrificans.

Growth of R. denitrificans OCh114 in different growth conditions
Similar growth curves and spectral features were obtained in the R. denitrificans OCh114 cultures grown aerobically in the light (20 W/m 2 ), dark, and day-night cycles ( Figure S1a and S1b), and slightly higher OD 600 were reached in the dark and day-light cycles compared to in the continuous light growth conditions ( Figure S1c). It is consistent with the reports that light has a negative effect on pigment formation in some AAPs [15]. No growth was observed anaerobically regardless of the light intensity (data not shown). Multiple organic carbon sources have been tested for the growth of R. denitrificans OCh114 [13]. Here, we report studies of three carbon sources: pyruvate, D-glucose, and CO 2 (or HCO 3 2 ). Spectral features of the photosynthesis system and light-harvesting complexes for the cultures grown in the minimal medium containing pyruvate or glucose are similar to the cultures grown in the rich medium (Figure 1a and 1b) as well as the results reported previously [13,16]. The image of R. denitrificans cells grown in the minimal medium with pyruvate was examined by OLYMPUS FV1000/BX61 highresolution confocal microscope, and the morphology and average cell size is consistent with previous reports [13]. Uptake of pyruvate, 2.5610 22 65610 24 mmole per hour (Figure 1d and S1d), is approximately 2 to 3-fold faster than uptake of glucose by R. denitrificans (Figure 1e), consistent with the faster growth in pyruvate than in D-glucose ( Figure 1c). Higher OD 600 and better cell growth can be obtained using 0.2% pyruvate in the minimal medium (data not shown). No differences in the 13 C-isotopomer abundances of the protein-derived amino acids for cultures grown in the minimal medium containing pyruvate were observed under either dark or illuminated conditions (Table S1a and S1b). Similar results were also observed in the cultures grown in the minimal medium supplied with glucose.

Vitamin B 12 is required for the growth of R. denitrificans
The previous studies of the growth of R. denitirficans OCh114 or other AAPs have been performed in either a rich medium or a medium containing either 0.02 g/liter [13] or 0.1 g/liter [17] of yeast extract (undefined carbon sources). We confirmed the necessity of yeast extract, as poor growth of the R. denitrificans cultures without yeast extract was observed (data not shown). Although it has been recognized that yeast extract contains rich vitamin mixtures, it also includes amino acids and other undefined carbon sources. Ideally, the 13 C-isotopic labeling studies require a minimal medium containing only defined carbon source(s), thus it is desirable to optimize the growth conditions by eliminating the yeast extract. We found that vitamin B 12 (cyanocobalamin) can serve as an alternative to yeast extract for growing R. denitrificans in defined carbon sources, as cultures with OD 600 $3 can be reached in our minimal growth medium with glucose as the sole carbon source (Figure 1e). Vitamin B 12 and different forms of cobalamin are required for methionine and protein synthesis, deoxyribonucleotide triphosphate synthesis, amino acid metabolisms, and CO 2 fixation (in methanogens, although there is no such pathway identified in R. denitrificans), and are included in the growth media of many photoautotrophic, photoheterotrophic and chemoheterotrophic bacteria. Requirement for vitamin B 12 in the growth of R. denitrificans may partially explain why yeast extract was necessary to be included in the minimal medium containing either pyruvate or glucose.

The carbohydrate utilization pathways in R. denitrificans
In the cultures grown in the minimal medium containing either D-[1-13 C]glucose or D-[6-13 C]glucose, the isotopomer labeling data of serine (the precursor is 3-phosphoglycerate) and alanine (the precursor is pyruvate) were different (Table S1e, S1f, S1i, and S3). Three metabolic pathways can be employed for sugar utilization by R. denitrificans and need to be considered to account for the isotopomer abundance in these protein-derived amino acids: (1) the Embden-Meyerhof-Parnas (EMP) pathway (glycolysis), by which one [1-13 C]glucose molecule is cleaved into two glyceraldehyde-3-phosphate (GAP) molecules: one is [3-13 C]labeled and the other is unlabeled using either D-[1-13 C]glucose or D-[6-13 C]glucose; (2) the Entner-Doudoroff (ED) pathway [18], by which one molecule of glucose generates one molecule of GAP, in which the third carbon is labeled using D-[6-13 C]glucose, and one molecule of pyruvate, where the first carbon is labeled using D-[1-13 C]glucose ( Figure 2a); (3) the pentose phosphate (PP) pathway, the first carbon of glucose is removed as CO 2 , if the oxidative PP pathway is active, to generate ribose-5-phosphate, which can be converted to GAP through the non-oxidative PP pathway ( Figure S2). As shown in Figure 2a, regardless of the pathways, GAP is converted into 3-phosphoglycerate (3-PGA), the precursor of glycine, serine, and cysteine, and then to pyruvate, the precursor of alanine, valine, leucine, and isoleucine.
The QRT-PCR results indicate that both eda (RD1_2878) and edd (RD1_2879) genes are expressed and the transcript level of these genes is higher in the minimal medium containing either pyruvate or glucose than in the rich medium, similar to the gene expression profiles of other genes responsible for carbon fixation and carbon metabolism examined in this report ( Figure 3). Moreover, the activity of 2-keto-3-deoxy-phosphogluconate (KDPG) aldolase and phosphogluconate dehydrase can be detected in cell-free extracts. Alternatively, the gene encoding 6phosphofructokinase (PFK, EC 2.7.1.11), an essential enzyme for the EMP pathway, is not annotated, and no PFK activity can be detected in cell free extracts.
The proposed ED pathway was further tested using D-[6-13 C]glucose, in which [6-13 C]KDPG is cleaved into unlabeled pyruvate and [3-13 C]GAP in the ED pathway ( Figure 2b). Our data showed a smaller M+0 value in serine using D-[6-13 C]glucose (0.30) than using D-[1-13 C]glucose (0.74), compared to slightly higher M+0 value in alanine using D-[1-13 C]glucose (0.56) versus D-[6-13 C] glucose (0.48) (Table S1e and S1i). The M+0 value in serine using D-[1-13 C]glucose or D-[6-13 C]glucose is not close to 1 or 0, respectively, suggesting that in addition to the ED pathway, the PP pathway is also active and generating GAP with different labeling pattern using D-[6-13 C]glucose or D-[1-13 C] glucose ( Figure 2a). Based on the isotopomer abundance of serine using D-[1-13 C]glucose and D-[6-13 C]glucose, we estimated 25,30% of GAP is produced from the (non-oxidative) PP pathway (Table S3). Figure 1. Spectra, image and organic carbon uptake of R. denitrificans OCh114. The spectra of cultures grown in the rich medium vs. in the minimal medium supplied with either 0.1% pyruvate or 0.1% glucose (a). No undefined carbon sources were included in the minimal medium reported in this paper. The normalized spectra in the 650 nm-900 nm range (b), the cell growth in the minimal medium with pyruvate, glucose, or HCO 3 2 (c), and the cell growth curve and uptake of pyruvate (d) or glucose (e) with or without the addition of 0.2% NaHCO 3 in the minimal medium containing pyruvate or glucose. More than ten biological replicates were performed in every growth conditions. doi:10.1371/journal.pone.0007233.g001 The ED pathway also leads to different predicted labeling in the aromatic amino acids. Phosphoenolpyruvate (PEP), synthesized from 3-PGA, and erythrose-4-phosphate (E4P), the intermediate in the non-oxidative PP pathway, are the precursors of phenylalanine, tyrosine and tryptophan. Labeled E4P in the non-oxidative PP pathway ( Figure S2) leads to lower M+0 value in the aromatic amino acids (0.5060.03) compared to serine (0.74) using D-[1-13 C]glucose. The ED and PP pathways can explain much higher M+0 value in phenylalanine (0.48) and tyrosine (0.53) using D-[1-13 C]glucose than the value in these amino acids (0.04) using D-[6-13 C]glucose, and more labeled carbon can be incorporated into PEP and E4P using D-[6-13 C]glucose (Figure 2a and 4).

The Pentose Phosphate pathway in R. denitrificans
A smaller M+0 value in histidine (0.28) than other amino acids (0.42-0.89) was detected using D-[1-13 C]glucose (Table S1e). Histidine is synthesized from 5-phospho-ribosyl-a-pyrophosphate (PRPP), the purine ring of ATP (the carbon from the formyl group of N 10 -formyl-tetrahydrofolate (THF)), and glutamine (the nitrogen source). The 6-phosphogluconate, an intermediate of the ED/ oxidative PP pathway, is the precursor of PRPP. In the oxidative PP pathway (Figure 4), the first (labeled) carbon in D-[1-13 C]glucose is released as CO 2 during the conversion of 6-phosphogluconate into D-ribulose-5-phosphate by 6-phosphogluconate dehydrogenase (PGD, EC 1.1.1.44), and it is unknown if the oxidative PP pathway is complete because the pgd gene in R. denitrificans is not annotated. Thus, the 13 C-isotopomer abundance in histidine using D-[1-13 C]glucose is possible from the following sources: (1) from the N 10 -formyl moiety of THF if the oxidative PP pathway is active, (2) from D-ribose-5-phosphate or D-ribulose-5-phosphate derived from the non-oxidative PP pathway, where the labeled carbon is maintained (Figure 4), or both. The N 10 -formyl moiety of THF is generated from THF and formate by transformylase, and various pathways can lead to formate biosynthesis in R. denitrificans. It is less likely that the labeled histidine using D-[1-13 C]glucose was produced exclusively from the 13 C-labeled formate through the oxidative PP pathway.
Alternatively, the non-oxidative PP pathways can also generate D-ribose-5-phosphate through steps of isomerization, epimerization and transketolation ( Figure S3). If the metabolic flux goes through the non-oxidative PP pathway, more than one 13 C-labeled sources are available using D-[6-13 C]glucose (i.e. labeled GAP from the ED pathway and labeled fructose-6-phosphate (F6P)), while only F6P is labeled using D-[1-13 C]glucose (GAP is not labeled via the ED pathway using D-[1-13 C]glucose) ( Figure 4). Since the EMP pathway in R. denitrificans is likely to be inactive, based on the experimental data and genomic information presented above, the labeled F6P is only led into the non-oxidative PP pathway. It is the most straightforward explanation for the detected 13 C-labeling abundance in histidine (M+0 value is 0.28) with D-[1-13 C]glucose (Table S1e) and for smaller M+0 value in histidine using D-[6-13 C]glucose (0.10) (Table S1i) than using D-[1-13 C]glucose (0.28). Together, our studies imply that R. denitrificans uses the nonoxidative PP pathway for histidine, ATP, coenzymes, and nucleic acids biosynthesis, as well as supplying GAP for the ED pathway (Figure 2a and 4).

Carbon fixation in R. denitrificans
Previous reports suggested that R. denitrificans [19] and other AAPs [11,20] can fix CO 2 . In our studies, no growth of R. denitrificans was observed in the minimal medium with CO 2 (or bicarbonate) as the sole carbon source (Figure 1c), consistent with the finding that the key genes required for the Calvin cycle are missing in the genome of R. denitrificans OCh114 [8]. Moreover, no CO 2 -enhanced growth was obtained in the minimal medium containing pyruvate or glucose (Figure 1d and 1e), and 13 Cisotopomer distribution of protein-based amino acids suggests that the presence of bicarbonate has a rather small effect on pyruvate and glucose metabolism in R. denitrificans (Table S1c and S1f). These experimental data indicate a carbon-fixation pathway other than the Calvin cycle is utilized by R. denitirificans OCh114. Swingley et al. [8] proposed that R. denitrificans could fix CO 2 by pyruvate or/and PEP to form oxaloacetate (OAA). Using uniformly 13 C-labeled D-glucose (D-[U-13 C 6 ] glucose) (Table S1j and S1k), our studies show less fully labeled amino acids derived from OAA (aspartate (M+4, 0.83), methionine (M+5, 0.78) and threonine (M+4, 0.85)) than the amino acid derived from pyruvate (alanine (M+3, 0.93)) ( Table S1k). Also, a smaller M+0 value can be seen in threonine (0.86) and aspartate (0.87) than alanine (0.94) using 13 C-NaHCO 3 and unlabeled pyruvate (Table S1d). Similar results are also obtained using 13 C-NaHCO 3 and unlabeled glucose (Table S1g). Together, our studies imply that R. denitrificans fixes approximately 10-15% of protein carbon from CO 2 via the anaplerotic pathways, despite lacking the Calvin cycle. These experimental evidences support the CO 2 fixation pathways by R. denitrificans proposed by Swingley et al. [8].

The TCA cycle and anaplerotic pathways
Our studies imply that R. denitrificans has an active TCA cycle and metabolic fluxes in the anaplerotic pathways. PEP can be converted into OAA, the precursor of aspartate, asparagine, methionine, threonine, lysine and isoleucine. OAA can be also generated from pyruvate and through the TCA cycle. Using D-[1-13 C] glucose, the M+0 value is similar in aspartate (0.58), threonine (0.47), methionine (0.49) and alanine (0.56) but smaller than serine (0.74) (Table S1e), suggesting that 13 C-isotopomer abundance in the amino acids derived from OAA is largely from pyruvate (via the anaplerotic pathway) and/or the TCA cycle flux. Four anaplerotic enzymes are known to assimilate CO 2 and replenish the intermediates of the TCA cycle, and genes encoding these anaplerotic enzymes can be all found in the genome of R. denitrificans OCh114; malic enzyme (tme, RD1_0421, EC 1.  1.31, PEP=OAA). To further investigate carbon fixation by R. denitrificans, we used gene expression studies to confirm if the pathways are active and to identify the transcriptional level of the genes encoding enzymes for the pathways under various growth conditions, as well as 13 C-isotopomer abundance to probe the anaplerotic flux. In the conditions we tested, the transcript level is higher for pckA (RD1_1376) and tme (RD1_0421) than for pyc (RD1_3376) and ppc (RD1_4248) (Figure 3). The activity of these four anaplerotic enzymes can be detected in cell-free crude extracts, consistent with the gene expression profiles. The activity is higher in PEP carboxykinase (6 umole/min?mg protein) and pyruvate carboxylase (2.5 umole/min?mg protein) than in PEP carboxylase (0.7 umole/ min?mg protein) in the cell extracts from the rich dark cultures, while the activity of pyruvate carboxylase and PEP carboxylase are comparable (1-2 umole/min?mg protein) in the minimal medium supplied with pyruvate.
To investigate metabolic fluxes through the anaplerotic pathways, [3-13 C]pyruvate was used and the 13 C-isotopomer abundance in aspartate (from OAA), serine (from PEP, 3-PGA) and alanine (from pyruvate) was analyzed (Table S1h). Both QRT-PCR and 13 C-labeling pattern suggest that the flux of pyruvate to PEP is weak (different isotopomer distribution between alanine and serine), consistent with low gene expression of ppdK (RD1_1948, pyruvate phosphate dikinase, EC 2.7.9.1, pyruvateRPEP) and that the pathway from OAA to PEP is likely to be active and pyruvate to OAA is not as strong. This is in agreement with the gene expression profile of pckA (RD1_1376) and pyc (RD1_3376). Hence, most of the 13 C-labeling in OAA likely comes from the TCA cycle flux (from malate), in which the pyruvate to malate pathway (via malic enzyme) should be rather active, compatible with higher expression of tme (RD1_0421) than pyc (RD1_3376). The proposed anaplerotic flux is summarized in Figure 5.

The isoleucine and leucine biosynthesis pathways
In many organisms, isoleucine is converted from threonine by threonine deaminase (EC 4.3.1.19) in the threonine-dependent pathway [21], and leucine is made from valine. The 13 C-isotopomer abundance in [1-13 C]pyruvate cultures showed higher M+0 value in leucine (0.91) and isoleucine (0.86) compared to valine (0.25) and threonine (0.70) ( Table S1a). Most of the leucine should be unlabeled because the first (labeled) carbon in 13 C-(2R,3S)-3isopropylmalate is released as CO 2 catalyzed by 3-isopropylmalate dehydrogenase (LeuB, EC 1.1.1.85) ( Figure 6). However, highly unlabeled isoleucine is not expected if 2-oxobutanoate, a precursor of isoleucine biosynthesis, is derived from threonine in the threonine-dependent pathway ( Figure 6). The results suggest that isoleucine is not exclusively derived from threonine. If it is, the M+0 values in threonine and isoleucine should be similar.
In the studies of the threonine-dependent pathway, QRT-PCR showed that the gene encoding theronine deaminase (ilvA, RD1_0416) was active in various growth conditions (Figure 3). The activity of citramalate synthase (CimA, EC 2.3.1.182) (360.5 nmole/min?mg protein) and theronine deaminase (6065 nmole/min?mg protein) can be detected in the cell-free extracts of the minimal medium containing pyruvate. Further, smaller M+0 values in isoleucine (0.83-0.85) than in leucine (0.90-0.91) using [1-13 C]pyruvate (Table S1a and S1b) imply that some isoleucine molecules are derived from the threonine rather than pyruvate and acetyl-CoA. If R. denitrificans utilizes the citramalatedependent pathway exclusively to generate isoleucine, the 13 Clabeling pattern in leucine and isoleucine will be identical (and the difference is expected to be within the error of GC-MS measurements: ,0.01), since the same precursors (pyruvate and acetyl-CoA) are employed in the biosynthesis of both amino acids. Therefore, both threonine-and citramalate-dependent pathways should be contributed to synthesize isoleucine, considering that the difference in 13 C-isotopomer abundance between leucine and isoleucine (0.06-0.07) is larger than the instrumental error (,0.01). Further, the calculated values of all 13 C-isotopomer abundance (from M+0 to M+5) of isoleucine indicate that 80% of the in vivo metabolic flux for isoleucine biosynthesis comes from the citramaltedependent pathway, 20% from the threonine-dependent pathway, when [3-13 C]pyruvate is the carbon source (Table S1h and Figure  S3). Together, our studies suggest that both the citramalate-and threonine-dependent pathways are active for isoleucine biosynthesis in R. denitrificans, similar to the report in Geobacter sulfurreducens [23]. As the putative cimA gene is not annotated in R. denitrificans, identification of the proposed citramalate synthase is in progress.

CO 2 fixation in R. denitrificans and other AAPs
In this paper, we reported that R. denitrificans uses the anaplerotic pathways to fix 10-15% of protein carbon from CO 2 . The amount of CO 2 fixation is lower in R. denitrificans than in autotrophic bacteria, supporting previous observations that R. denitrificans cannot grow autotrophically. Other than R. denitrificans, CO 2 fixation in several AAPs producing BChl a aerobically was reported previously by 14 CO 2 incorporation studies: Erythrobacter sibiricum (a very low level of CO 2 (0.4%) assimilated) [7], Erythrobacter longus [7], Acidiphilium rubrum [20] and Erythrobacter sp. strains [15]. Some difference in CO 2 fixation can be detected between R. denitrificans and A. rubrum. In A. rubrum, CO 2 fixation is completely inhibited with the glucose-growth culture, and is enhanced 3-to 5-fold in the light (,3.0 nmoles/mg dry cell weight) [20]. In contrast, our studies suggest that CO 2 incorporation can be detected in R. denitrificans with either pyruvate-or glucose-growth cultures in either light or dark growth, and is not enhanced by light exposure.
Among the wide-spread marine Roseobacter clade, many of them cannot generate photosynthetic pigments. To the best of our knowledge, no CO 2 assimilation data have been reported for these unpigmented strains of the Roseobacter clade. It will be very useful to compare the difference in the CO 2 assimilation between R. denitrificans and un-pigmented Roseobacter clade bacteria.

Carbohydrate metabolism in R. denitrificans
The ED pathway has generally been viewed as an alternative to the EMP pathway. In examining the carbohydrate catabolism of both Rhodobacter capsulatus [25] and Rhodobacter sphaeroides [26] under anaerobically phototrophic and aerobically heterotrophic growth conditions, Conrad and Schlegel reported that fructose degradation was shifted to the ED pathway from the EMP pathway during the transition from anaerobically phototrophic to aerobically heterotrophic growth conditions, and that glucose catabolism was going through the ED pathway under both phototrophic and heterotrophic growth. In agreement with their experimental evidences, all of the genes in the EMP and ED pathways can be found in the genome of several R. sphaeroides strains, while the 6-phosphogluconate dehydrogenase gene (pgd) in the oxidative PP pathway cannot be annotated. Our studies in this report indicate that the ED pathway plays a significant role in glucose metabolism of R. denitrificans, and that the non-oxidative PP pathway contributes to histidine, ATP, coenzyme and nucleic acid biosynthesis, and that no metabolic flux going through the EMP pathway and the oxidative PP pathway. Our experimental evidences are consistent with the genomic information of R. denitrificans, in which the ED pathway and the non-oxidative PP pathway are complete, but the pfk gene, essential for the EMP pathway, and the pgd gene, required for the oxidative PP pathway, are not annotated.
Searching the genomes of the Roseobacter clade [5,[8][9][10], all of them have genes in the ED pathway annotated. The ED pathway is identified mainly in prokaryotes, and almost all of them are aerobes and facultative bacteria, consistent with the physiological features of AAPs. Moreover, note that the pgd gene is also missing in Silicibacter pomeroyi DSS-3, Silicibacter sp. TM1040 and Dinoroseobacter shibae DFL12, and that the pfk gene is not found in Jannaschia sp. CCS1, Silicibacter pomeroyi DSS-3, and Silicibacter sp. TM1040. Alternatively, Dinoroseobacter shibae DFL12 has all of the genes in the EMP pathway, and Jannaschia sp. CCS1 has complete oxidative PP pathway. It will be interesting to learn carbohydrate metabolisms in these two bacteria.

Possible physiological roles of the ED pathway in R. denitrificans
The physiological significance for R. denitrificans and possibly some other resourceful heterotrophic marine Roseobacter clade [12] to depend on the ED pathway, which produces 1 ATP, 1 NADH and 1 NADPH, compared to 2 ATP and 2 NADH from the more common EMP pathway, is an interesting question. The most straightforward answer could be that R. denitrificans demands more NADPH than ATP in the cells when the oxidative PP pathway is not active. The oxidative PP pathway, not the non-oxidative PP pathway, can generate two molecules of NADPH from one molecule of glucose, and can produce approximately 60% of the NADPH required inside the cells.
NADPH is not only the reducing power for synthesizing energyrich molecules and biofuels, but also prevents oxidative stress by reducing glutathione via glutathione reductase, which converts reactive H 2 O 2 into H 2 O by glutathione peroxidase [27]. The reactive oxygen species inside the cell can be fatal for R. denitrificans since it is an aerobic anoxygenic bacterium that contains highly absorbing pigments that can sensitize reactive oxygen species formation. When the oxidative PP pathway is not active, the ED pathway may be the best option to generate NADPH required by the cells. It was not established whether an anoxygenic photosynthetic bacterium, like R. denitrificans, can produce glutathione, which can be synthesized by oxygenic phototrophs, such as cyanobacteria and higher plants. Note that genes for glutathione biosynthesis (glutamate-cysteine ligase, gshA, RD1_4077, glutathione synthetase, gshB, RD1_1192), glutathione reductase (gor, RD1_1919), glutathione peroxidase (gpo, RD1_0599), and some glutathione transferases can be found in R. denitrificans genome, suggesting possible roles of glutathione in reducing oxidative stress in R. denitrificans.
The other possibility for R. denitrificans adapting the ED pathway could be that many marine Roseobacter-lineage bacteria live in coastal seawater, an environment enriched with a wider variety of organic acids, and the ED pathway is a better option for digesting the hexuronic acids (e.g. glucuronic acid). This hypothesis may be supported by the broad substrate specificity of KDPG aldolase identified biochemically [28] and structurally [29]. As gene sequencing is currently in progress for 40 different Roseobacter strains [2], it will be interesting to learn if the patterns of carbohydrate metabolism revealed in this work may be a general theme in the Roseobacter clade, especially the significance of the ED/non-oxidative PP pathway in these physiologically diverse, widespread and high abundance marine microorganisms.

Gene expression under different growth conditions
Throughout the QRT-PCR data analysis, we observed that the transcript level for all of the genes tested is higher in cells grown in the minimal medium containing either pyruvate or glucose than in the rich medium ( Figure 3). The discrepancies may be attributed to the following possible scenarios: (1) metabolic regulation: it has been thought in some bacteria, E. coli included, that the anaplerotic and metabolic pathways are not as much in demand in the rich medium as in the minimal medium containing defined carbon source, or some pathways may be shut down due to metabolic regulation [30,31], so genes in those metabolic pathways are not highly expressed in rich growth medium; and/or (2) the abundance of the 16S rRNA gene in different cell growth conditions: as reported in previous studies [32][33][34], the abundance of the 16S rRNA gene is dependent on the growth rate of the cells. If it is the case for R. denitrificans, lower copy numbers of the 16S rRNA gene will be expected in the minimal medium with glucose (grow slowest), and highest number in the rich medium grown under dark conditions (grow fastest). Together, both possible scenarios can explain lower transcript level for all of the target genes we tested in the rich medium. Further investigations are required for addressing this important issue.

Requirement of yeast extract or exogenous vitamin B 12
In this report, we demonstrate that vitamin B 12 can successfully replace the undefined carbon source, yeast extract, for the growth of R. denitrificans in our developed minimal medium containing defined organic carbon source. While it is not known why the undefined carbon source yeast extract is required for R. denitrificans and several other photosynthetic bacteria, it is not clear why exogenous vitamin B 12 is required for R. denitrificans, because all of the genes required for vitamin B 12 biosynthesis are annotated in R. denitrificans (based on the information in the KEGG (Kyoto Encyclopedia of Genes and Genomes) database). It is also unclear how cobalamin can be utilized by R. denitrificans, since no known cobalamin transporter genes (btuB, btuC, btuD and btuF) can be recognized in the genome. Nonetheless, the effect of exogenous vitamin B 12 is clear and its inclusion permitted most of the results reported in this paper. More experimental evidences are required for understanding the roles of yeast extract and exogenous vitamin B 12 reported herein for enhancing cell growth.

A common feature in Roseobacter clade?
Our studies suggest approximately 10-15% CO 2 fixation in protein carbon through the anaplerotic enzymes by R. denitrificans, supporting the proposal by Swingley et al. [8]. We also confirmed that unlike carbon fixation by autotrophs, R. denitrificans cannot live with CO 2 as the sole carbon source, consistent with lack of genes for key enzymes of the Calvin cycle. Missing genes in the Calvin cycle are also common features in the other four Roseobacter clade bacteria with complete sequences [2,5,9], suggesting that carbon fixation mechanisms other than the Calvin cycle are needed for the AAPs.
Other than the ED pathway, our studies demonstrate that R. denitrificans uses the non-oxidative PP pathway as the other alternative pathway for sugar degradation to make the ribose-5phosphate for histidine, cofactors, ATP and nucleic acid biosynthesis, as well as to produce GAP for carbon metabolism. In contrast to the oxidative PP pathway, no CO 2 is released in the non-oxidative PP pathway, which is an important feature for carbon fixation by autotrophs using the Calvin cycle. In these autotrophs, the non-oxidative PP pathway is part of the Calvin cycle. Although R. denitrificans cannot use the Calvin cycle for fixing CO 2 , it does have a complete non-oxidative PP pathway and relies on the pathway for converting hexose phosphate into pentose phosphate.

Materials
The DNA oligomers in this report are from Integrated DNA Technology (IDT) and were used without further purification. The 13 C-labeled glucose, pyruvate and sodium bicarbonate were purchased from Cambridge Isotope Laboratories (CIL), Inc (MA, USA).
The growth of bacterial strains R. denitrificans OCh114 (a gift from the laboratory of Dr. Beatty at University of British Columbia, Vancouver, Canada) was grown aerobically (20-30%-filled Erlenmeyer flasks) on either a nutrition-rich medium (Difco TM Marine Broth 2216; Becton, Dickson and Company) (pH 7.5) or a reported minimal medium [13] with modifications (illustrated in next paragraph). No undefined carbon sources (yeast extract) was included in our minimal growth medium. The seawater source in our minimal medium was prepared by 3.2% Instant OceanH artificial sea salts (United Pet Group, Cincinnati, OH). The minimal growth medium was adjusted to pH 7.5. All of the cultures were grown at 28uC and the growth rates were estimated by OD 600 . We used 2% cultures (50-fold dilution) in the late exponential growth phase grown in the rich medium to inoculate the minimal medium containing defined carbon source(s). To minimize the carryover of the rich medium, inoculated cells were centrifuged and supernatant liquid was removed. Cell pellets were washed with the minimal medium (no carbon source included) twice, resuspended in the minimal medium, and then transferred to the minimal medium containing defined carbon source(s).
In the minimal medium supplied with defined 13 C-isotopically labeled carbon source, 0.1% pyruvate (labeled on either the first or third carbon position), 0.1% glucose (on the first, sixth or uniform labeled), or 0.2% NaH 13 CO 3 were added. Due to the requirement of a growth medium without including undefined carbon sources for GC/MS studies, ferric citrate [13] was replaced by ferric chloride hexahydrate, and yeast extract was substituted by 20 mg/L vitamin B 12 . The cultures were grown to an exponential phase, representing the pseudo steady-state, and then harvested for analyzing the isotopomer distribution in key metabolites.

RNA extraction and quantitative real-time PCR (QRT-PCR)
The methods used to extract RNA and perform QRT-PCR were described previously [35]. QRT-PCR was performed to profile the gene expression under different growth conditions of R. denitrificans OCh114. cDNA was synthesized from 1 mg RNA and 100 mM random 9mer DNA using Superscript III reverse transcriptase (Invitrogen). The QRT-PCR reactions were performed via ABI 7500 Real-Time PCR System (Applied Biosystems). The primers (shown in Table S2) for QRT-PCR reactions were designed using Primer Express 2.0 software (PE Applied Biosystems) and analyzed by OligoAnalyzer 3.0 software (IDT). The Power SYBR Green Master Mix (PE Applied Biosystem) was used for amplifying DNA [35]. The cycle threshold (Ct) was calculated as the cycle number at which a fluorescence threshold (DRn) crossed the baseline. Data were normalized by analyzing DCt = Ct of the target gene -Ct of the internal control gene (16S rRNA), and each relative gene expression value is calculated with 2 2(DCt) . Three biological replicates, with six technical replicates for each biological sample, and total eighteen technical replicates were performed for validation, and the mean value was reported (Figure 3). The amplified DNA fragments were verified by 1% agarose gel electrophoresis, and a single fragment was obtained for all amplicons (data not shown).

Measurements of pyruvate/glucose uptake
The amount of D-glucose in solution was determined by hexokinase/glucose-6-phosphate dehydrogenase coupling assay with a D-glucose assay kit (Roche, Mannheim, Germany), and the reduction of NAD + was followed spectrophotometrically by the increase at 340 nm in absorbance. The amount of pyruvate in the culture was estimated by lactate dehydrogenase (EC 1.1.2.3) (Sigma-Aldrich) with 25 mM NADH in 0.1 M Tris-HCl buffer at pH 8.0 [36]. The reaction was followed by the oxidation of NADH (the decrease at 340 nm in absorbance).

Gas chromatography (GC)/mass spectrometry (MS)
The methods for GC/MS are reported previously [44,[47][48][49]. In brief, biomass was collected from 50 mL culture by centrifuge and then was hydrolyzed in 6 M HCl at 100uC. The resulting amino acids were dried and derivatized in tetrahydrofuran and N-(tert-butyl dimethylsilyl)-N-methyl-trifluoroacetamide (Sigma-Aldrich) at 70uC for 1 h. A gas chromatograph (GC) (Hewlett-Packard, model 7890A, Agilent Technologies Inc., Ballwin, MO) equipped with a DB5-MS column (J&W scientific, Folsom, CA) and a mass spectrometer (5975C, Agilent Technologies Inc.) was used for analyzing metabolite labeling profiles. We report four types of MS fragment data: (1) Figure S1 Spectra, culture, and growth curves of R. denitrificans under different growth conditions. The culture of R. denitrificans grown in the rich medium under the day-light cycles (a). The spectrum (b) and growth curve (c) of the cells grown in the rich medium under dark versus under light. The spectrum of the cells grown under dark and day-light cycle is shown in the inset. The formation of the light harvesting antenna complex II antenna complex (807-nm peak) is repressed in the light, while the level of the reaction center-light harvesting complex I (RC-LHI) core complex (872-nm peak) is similar. Data fit of pyruvate uptake with or without NaHCO3 in the defined medium, and the uptake rate is 2.5610226561024 mmole per hour (d).