Metabolic Flux Analysis of the Synechocystis sp. PCC 6803 ΔnrtABCD Mutant Reveals a Mechanism for Metabolic Adaptation to Nitrogen-Limited Conditions

Metabolic flux redirection during nitrogen-limited growth was investigated in the Synechocystis sp. PCC 6803 glucose-tolerant (GT) strain under photoautotrophic conditions by isotopically non-stationary metabolic flux analysis (INST-MFA). A ΔnrtABCD mutant of Synechocystis sp. PCC 6803 was constructed to reproduce phenotypes arising during nitrogen starvation. The ΔnrtABCD mutant and the wild-type GT strain were cultured under photoautotrophic conditions by a photobioreactor. Intracellular metabolites were labeled over a time course using NaH¹³CO₃ as a carbon source. Based on these data, the metabolic flux distributions in the wild-type and ΔnrtABCD cells were estimated by INST-MFA. The wild-type GT and ΔnrtABCD strains displayed similar distribution patterns, although the absolute levels of metabolic flux were lower in ΔnrtABCD. Furthermore, the relative flux levels for glycogen metabolism, anaplerotic reactions and the oxidative pentose phosphate pathway were increased in ΔnrtABCD. This was probably due to the increased expression of enzyme genes that respond to nitrogen depletion. Additionally, we found that the ratio of ATP/NADPH demand increased slightly in the ΔnrtABCD mutant. These results indicated that futile ATP consumption increases under nitrogen-limited conditions because the Calvin–Benson cycle and the oxidative pentose phosphate pathway form a metabolic futile cycle that consumes ATP without CO₂ fixation and NADPH regeneration.

Introduction

Cyanobacteria regulate their metabolic state to adapt to various environmental conditions. It has been reported that the metabolism of the model cyanobacteria, Synechocystis sp. PCC 6803 glucose-tolerant (GT) strain, was drastically changed through the redirection of metabolic flow in central metabolism, when shifted between photoautotrophic and photoheterotrophic conditions (Yoshikawa et al. 2013, Nakajima et al. 2014). Metabolic adaptation to nitrogen-depleted conditions has also been reported for the wild-type GT strain when transferred to a medium containing a limited source of nitrogen (Krasikov et al. 2012). Analyses showed that the amount of 2-oxoglutarate (2OG), a precursor for the nitrogen assimilation reaction, increased in response to nitrogen depletion (Muro-Pastor et al. 2001). The 2OG accumulation triggered a signaling cascade involving NtcA, SigE and Rre37 transcription factors, which resulted in the down-regulation of Calvin–Benson cycle-related genes, as well as an up-regulation of genes associated with glycogen metabolism (glgX, glgP, glgA and glgC) and the oxidative pentose phosphate pathway (oxPPP; zwf and gnd) (Aichi et al. 2001, Osanai et al. 2005, Osanai et al. 2006, Azuma et al. 2011, Joseph et al. 2014). Quantitative real-time PCR analyses showed that tricarboxylic acid (TCA) cycle- and anaplerosis-related transcripts (acnB, icd, ppc, pyk1, me and pta) increased after 4 h of nitrogen depletion in the wild-type GT strain (Iijima et al. 2014). Additional phenotypes such as glycogen accumulation, bleaching from blue-green to yellow upon degradation of phycobilisomes (the light-harvesting apparatus) and slower cell growth were observed after 12 h of nitrogen starvation (Krasikov et al. 2012, Osanai et al. 2014).

These results suggested that the metabolic flux distribution in the central metabolism of Synechocystis sp. PCC 6803 is likely to be drastically reprogrammed or redirectred by the global regulation of gene expression in response to nitrogen starvation. However, since the metabolic flux distribution is hardly estimated from gene expression or metabolite pool size data, the effect of changes in gene expression on the flow levels of the metabolic network remains unclear. A relationship between gene expression regulation and the mechanisms of metabolic adaptation under nitrogen-limited conditions should be investigated by examining the metabolic flux distribution in the central metabolism of cyanobacteria.

¹³C-Metabolic flux analysis (¹³C-MFA) is a method that estimates the metabolic flux distribution in microbial cells under heterotrophic conditions (Antoniewicz 2013). By culturing cells with ¹³C-labeled glucose as a carbon source, the metabolic flux distribution has been estimated by measuring the ¹³C-labeling patterns of several intermediates. ¹³C-MFA of the lipid-producing alga, Chlorella protothecoides, under photoheterotrophic conditions revealed that the oxPPP flux for glucose uptake increased from 3% to 20% during nitrogen-limited growth to meet the NADPH demands required for increased lipid production (Gopalakrishnan et al. 2015).

In this study, the metabolic flux redirection during nitrogen-limited growth was investigated in the Synechocystis sp. PCC 6803 GT strain under photoautotrophic conditions by isotopically non-stationary metabolic flux analysis (INST-MFA) (Jazmin and Young 2013). INST-MFA is another method for analyzing metabolic flux. It estimates metabolic flux distributions in photoautotrophic organisms, using time course measurements of the ¹³C-labeling patterns of several intermediates after treating cells with ¹³CO₂ (Jazmin et al. 2014). To obtain precise INST-MFA measurements, the metabolic system of interest must be kept at steady state during the ¹³C-labeling procedure (Young et al. 2011). This cannot be achieved in cells that are transferred to a nitrogen-depleted medium because transient protein degradation causes the recycling of nitrogen and carbon to generate an endogenous source of nitrogen under these conditions (Krasikov et al. 2012, Hasunuma et al. 2013). A stable cultivation system using a medium with a low concentration of some nitrogen source is not experimentally feasible because the nitrogen would be consumed too quickly and depleted before the end of the experiment. Thus, in this study, we used a deletion mutant of Synechocystis sp. PCC 6803 that lacks the nitrate transporter operon (nrtABCD). It has been reported that cyanobacteria such as Synechocystis sp. PCC 6803 actively incorporate nitrate and nitrite by employing the nitrate transporter (Ohashi et al. 2011). It was also demonstrated that mutant strains of Synechococcus elongatus PCC 7942 lacking nitrate transporter could grow on BG-11 medium in a manner dependent on nitrate concentrations. This indicated that nitrate was incorporated into the cyanobacteria by diffusion across the cell membrane (Omata et al. 1989).

In this study, a ΔnrtABCD mutant of Synechocystis sp. PCC 6803 was constructed to recapitulate nitrogen depletion phenotypes. Wild-type GT and ΔnrtABCD strains were cultured under photoautotrophic conditions, using a photobioreactor to perform ¹³C-labeling experiments for INST-MFA. A comparison of the metabolic flux distributions in the ΔnrtABCD mutant with that in the wild-type GT strain revealed that the metabolism of cyanobacteria employs a metabolic futile cycle under nitrogen-limiting conditions.

Results

Construction and phenotype analysis of the ΔnrtABCD mutant

The Synechocystis sp. PCC 6803 GT mutant strain, ΔnrtABCD, was constructed by replacing the nrtABCD operon with the kanamycin resistance gene using a disruption cassette (Supplementary Fig. S1). The flask-scale cultivation of ΔnrtABCD showed a positive relationship between the growth rate of ΔnrtABCD and nitrate concentrations in the modified BG-11 medium, whereas the growth rate of the wild-type GT strain was constant (Fig. 1). This phenotype was similar to that of the Synechococcus sp. PCC 7942 mutant strain, ΔnrtA (Omata et al. 1989), indicating that the ΔnrtABCD strain probably lacked nitrate transporters and instead incorporated nitrate by diffusion.

Fig. 1

Growth curves for the wild-type GT strain and the ΔnrtABCD mutant in flask culture. The deletion mutant of the nitrate transporter operon nrt (ΔnrtABCD) was cultured for 144 h in 20 ml of modified BG-11 medium supplemented with 10, 25 or 40 mM NaNO₃ under photoautotrophic conditions. The OD₇₃₀ was measured at the indicated time intervals. Experiments were performed in triplicate. Data represent the mean ± SD.

Large-scale cultures, using a photobioreactor and the modified BG-11 medium supplemented with 25 mM sodium nitrate, were established to investigate nitrogen starvation phenotypes (first cultivation in Table 1). The wild-type GT and ΔnrtABCD strains were pre-cultured in 500 ml flasks with 100 ml of modified BG-11 medium supplemented with 5 mM NH₄Cl, then transferred to the 1 liter photobioreactor with 550 ml of modified BG-11 medium supplemented with 25 mM NaNO₃ (see the Materials and Methods for the detailed procedure). While the specific growth rate of ΔnrtABCD (0.015 h^–1) was one-sixth that of the wild-type GT strain (0.089 h^–1), both strains showed exponential growth until reaching OD₇₃₀ = 1.0 (Fig. 2a). ΔnrtABCD cells were pale yellow in color, which is similar to bleached cells obtained from nitrogen-limited culture conditions (Fig. 2b). It has been reported that the expression of the antenna protein complex (phycobilisome) was decreased under nitrogen-limited conditions (Richaud et al. 2001). A comparison of ultraviolet–visible (UV-VIS) spectra showed that a peak corresponding to the phycocyanin subunit of phycobilisome (630 nm) was not observed in the UV-VIS spectrum of the ΔnrtABCD mutant at 6 h after the start of cultivation. Additionally, the absorbance was reduced across the whole spectrum including the spectral range associated with Chl a (68 nm) (Fig. 2c). These results indicated that the ΔnrtABCD mutant has a reduced amount of photosystems including antenna proteins.

Table 1

Culture profiles of wild-type and ΔnrtABCD strains of Synechocystis sp. PCC 6803 using the photobioreactor

Strains	Specific growth rate (h–1)		Specific glycogen production rate (mg glycogen g–1 DCW h–1)
	Firsta	Secondb	Firsta	Secondb
Wild type	0.089	0.082	–10.1	–10.2
ΔnrtABCD	0.015	0.017	1.6	2.7

^a Data obtained from the first trial. Specific rates were determined from 15–21 h data for the wild type and 24–96 h data for ΔnrtABCD.

^b Data obtained from the second trial. Specific rates were determined from 15–18 h data for the wild type and 24–72 h data for ΔnrtABCD.

Fig. 2

Nitrogen starvation phenotypes of the ΔnrtABCD mutant. Data were obtained from large-scale cultures using a photobioreactor with 550 ml of modified BG-11 medium supplemented with 25 mM sodium nitrate. (a) Cell growth curve. (b) Comparison between the wild-type GT strain and the ΔnrtABCD mutant. (c) UV-VIS absorbance spectra. Maximal absorption wavelengths for phycocyanin and Chl a are 630 and 680 nm, respectively. (d) Glycogen concentrations.

The culture profile data showed that the glycogen content of the wild-type GT strain decreased at the beginning of culture, and then it increased at 24 h of cultivation (Fig. 2d). In contrast, the glycogen level was elevated in the ΔnrtABCD mutant in a time-dependent manner. The specific glycogen production rate during the exponential growth phase was –10.1 mg glycogen g^–1 dry cell weight (DCW) h^–1 for the wild-type strain (calculated from the 15–21 h data) and 1.6 mg glycogen g^–1 DCW h^–1 for the ΔnrtABCD mutant (calculated from the 24–96 h data). A comparison of 2OG levels between the wild-type and ΔnrtABCD strains collected at 18 and 72 h of culture, respectively, revealed that 2OG levels were 1.21 times higher in the ΔnrtABCD mutant than in the wild-type strain (Supplementary Fig. S2). These results indicated that the ΔnrtABCD mutant cultured in the modified BG-11 medium supplemented with 25 mM sodium nitrate displayed phenotypes similar to those of the wild-type GT strain cultured under nitrogen-limited conditions.

¹³CO₂-labeling experiment

The large-scale cultures using the photobioreactor were performed again for a ¹³CO₂-labeling experiment (second cultivation in Table 1). The wild-type GT and ΔnrtABCD strains were cultured in the modified BG-11 medium supplemented with 25 mM sodium nitrate until mid-log phase. After culturing for 18 and 72 h, the aeration by CO₂ gas was stopped and the ¹³CO₂-labeling began by adding NaH¹³CO₃ to the medium (see the Materials and Methods for the detailed procedure). A constant pH of 7.5 and a bicarbonate ion concentration >1.0% were maintained throughout the labeling experiment by adding buffer and excess amounts of NaH¹³CO₃ (data not shown).

Cells were iteratively collected from the identical photobioreactor for 600 s after the initiation of ¹³CO₂-labeling. For the ΔnrtABCD strain, the mass distribution vector (MDV) at 600 s could not be obtained due to technical limitations. Free metabolites were extracted from the cells and MDVs were determined by gas chromatography–mass spectrometry (GC-MS) and liquid chromatography–tandem mass spectrometry (LC-MS/MS). In this study, time course MDV data were successfully obtained for 15 fragments of 13 free metabolites (Supplementary Tables S1, S2).

Averaged ¹³C-enrichment was calculated for each intermediate from the MDV data (Fig. 3). The averaged ¹³C-enrichment is the ratio of ¹³C atoms to the total amount of carbon atoms in the intermediate. The time course data showed that the averaged ¹³C-enrichment of all intermediates was close to zero at the beginning of the labeling experiment. The averaged ¹³C-enrichment of PGA (a mixture of 2-phosphoglycerate and 3-phosphoglycerate) in the wild-type GT strain quickly increased to 53.4% at 180 s. This gradually increased further during the course of the labeling experiment. Although the averaged ¹³C-enrichment at 600 s differed among the intermediates, all intermediates (except for TCA cycle metabolites) showed similar labeling kinetics. These results suggested a rapid turnover of the metabolic intermediates of the Calvin–Benson cycle in cyanobacteria. In the case of the ΔnrtABCD mutant, the labeling kinetics were slower than those of the wild-type GT strain. Furthermore, the ¹³CO₂-labeling kinetics of TCA cycle metabolites such as glutamate (Glu) and succinate (Suc) were significantly slower than that of other metabolites.

Fig. 3

Kinetics of the ¹³C-labeling of intracellular metabolites. Cells of the wild-type GT (a) and ΔnrtABCD (b) strains were iteratively collected at 1, 30, 60, 90, 120, 180, 300 and 600 s after the addition of NaHCO₃ to the culture using the photobioreactor. Free metabolites were extracted and the mass distribution vectors (MDVs) were determined by GC-MS and LC-MS/MS. Averaged ¹³C-enrichment (the ratio of ¹³C atoms to the total amount of carbons in the metabolite) was determined from the MDV data. For the ΔnrtABCD mutant, the MDV data at 600 s could not be obtained due to technical limitations. Glu57, [M-57]⁺ fragment of glutamate including carbons 1-2-3-4-5; Glu85, [M-85]⁺ fragment of glutamate including carbons 2-3-4–5; PGA117, [M-117]⁺ fragment of 3-phosphoglycerate including carbons 2-3.

Determination of metabolic flux distribution

A metabolic flux distribution was estimated by non-linear fitting among predicted and measured time course MDV data (Jazmin and Young 2013). For this purpose, a metabolic model of Synechocystis sp. PCC 6803 is needed to calculate a predicted MDV from a metabolic flux distribution and intracellular metabolite levels. Here, a metabolic model was created considering the previous research including the Calvin–Benson cycle, glycolysis, the oxPPP, glycogen biosynthesis and degradation, the anaplerotic reactions catalyzed by phosphoenolpyruvate (PEP) carboxylase and malic enxyme, the glyoxylate shunt and two TCA bypass reactions (Young et al. 2011, Xiong et al. 2014). The metabolic flux for the biomass synthesis was determined from a specific cell growth rate and the biomass composition described in a previous report (Supplementary Table S3) (Young et al. 2011).

An optimal metabolic flux distribution and metabolite pool size were estimated using non-linear fitting of predicted and measured time courses of MDV (Supplementary Table S4). The residual sum of squares (RSS) of the best-fit results were 545.8 and 530.1 for the wild-type GT and ΔnrtABCD strains, respectively, and they passed the χ² test with α = 0.05 (Fig. 4; Supplementary Fig. S3) (Antoniewicz et al. 2006). The 95% confidence intervals were determined for all metabolic flux levels and pool sizes by the Monte-Carlo method with 1,000 iterations (Supplementary Table S4). The estimated metabolic flux distribution of the wild-type GT strain in this study was similar to that reported in a previous study (Young et al. 2011).

Fig. 4

Metabolic flux distributions of the wild-type GT (a) and ΔnrtABCD (b) strains determined by isotopically non-stationary metabolic flux analysis (INST-MFA). The metabolic flux levels for the best-fit results are shown for each reaction in red. Arrows represent metabolic levels and the directions of metabolic flow. The pool sizes of each metabolite are indicated by the circle sizes (more details are shown in Supplementary Table S4).

Effect of nitrogen limitation on the metabolic flux distribution

The central carbon metabolism of Synechocystis sp. PCC 6803 consists of the Calvin–Benson cycle, its bypass via the oxPPP, and two branching pathways that involve glycogen metabolism starting with glucose 6-phosphate (G6P) and the TCA cycle starting with PGA. The results of INST-MFA showed that the metabolic flux levels of the carbon fixation reaction catalyzed by RuBisCO were 4.63 and 1.19 mmol g^–1 DCW h^–1 for the wild-type GT and ΔnrtABCD strains, respectively (Fig. 4). While the specific growth rate for the ΔnrtABCD mutant (0.015 h^–1) was 17% that of the wild-type GT strain (0.089 h^–1), the metabolic flux levels of RuBisCO remained at 26% that of the wild type. This indicated that carbon utilization efficiency decreased under nitrogen-limited conditions.

In this regard, the oxPPP bypasses the Calvin–Benson cycle, in which a hexose phosphate is converted to a pentose phosphate while discarding CO₂ and regenerating two molecules of NADPH (You et al. 2015). The metabolic flux analysis revealed that there was metabolic flow in the oxPPP bypass measuring 1.14 and 0.35 mmol g^–1 DCW h^–1 for the wild-type GT and ΔnrtABCD strains, respectively. The carbon loss rate by the oxPPP bypass during carbon fixation by RuBisCO was 24.6% (=1.14/4.63) in the wild-type GT strain. The carbon loss rate slightly increased to 29.4% (=0.35/1.19) in the ΔnrtABCD mutant, suggesting that NADPH regeneration by the oxPPP was activated by up-regulation of the zwf and gnd genes as reported in a previous study (Osanai et al. 2006).

Similar results were also observed for glycogen metabolism. The metabolic flux levels for glycogen biosynthesis and degradation in the wild-type GT strain were 0.37 and 0.43 mmol g^–1 DCW h^–1, respectively (Fig. 4). In contrast, the metabolic flux levels for glycogen biosynthesis and degradation in the ΔnrtABCD mutant were 0.39 and 0.3 mmol g^–1 DCW h^–1, respectively (Grundel et al. 2012). The rate of glycogen metabolism relative to the carbon fixation flux levels increased in the ΔnrtABCD strain compared with that in the wild-type GT strain, probably because of the activation of the glgX, glgP, glgA and glgC genes (Osanai et al. 2006).

As mentioned above, PGA is the second branch point in the metabolic network. The split ratio of the Calvin–Benson cycle to the TCA cycle was 7.9 : 1.0 (=8.12 : 1.03) in the wild-type GT strain (Fig. 4). This ratio increased to 11.4 : 1.0 (=2.17 : 0.19) in the ΔnrtABCD mutant, suggesting that the relative metabolic flow towards the TCA cycle was reduced under nitrogen-limited conditions. Although the expression levels of TCA cycle-related genes increased following nitrogen depletion in the wild-type GT strain compared with those in the ΔnrtABCD strain (Iijima et al. 2014), the low metabolic flux levels of the TCA cycle in the ΔnrtABCD strain were coincident with a slower turnover of ¹³C-labeled Glu and Suc (Fig. 3). On the other hand, up-regulation of the ppc, pyk1 and me genes during nitrogen starvation (Iijima et al. 2014) was consistent with the metabolic flux data. This is because the metabolic flux levels of anaplerotic reactions, such as PEP carboxylase (encoded by ppc), malic enzyme (me) and pyruvate kinases (pyk1 and pyk2), increased in the ΔnrtABCD mutant compared with the levels in the wild-type GT strain.

Cofactor balances

Fig. 5 represents comparisons between cofactor regeneration and consumption in the central metabolic network including NADPH and ATP. These were obtained by stacking the metabolic flux levels of regenerating and consuming reactions. Here, NADPH and NADH are not discriminated because strict cofactor usages remain unclear for several enzymes. Here, cofactor balances required for glutamine synthetase and biomass synthesis were not considered. In the metabolic network of the wild-type GT strain, NADPH was mainly consumed in the glyceraldehyde-3-phosphate dehydrogenase (encoded by gap2) reaction in the Calvin–Benson cycle (Fig. 5a). In contrast, some NADPH was metabolically regenerated by several reactions, including those mediated by two dehydrogenases (encoded by zwf and gnd genes) in the oxPPP and the reaction catalyzed by malic enzyme. The gap between the NADPH production and consumption rates or a net NADPH consumption rate in the central carbon metabolism of the wild-type GT strain was 4.30 mmol g^–1 DCW h^–1. This was probably supplied from the light reaction mediated by the photosystems.

Fig. 5

Cofactor regeneration and consumption in the central metabolic network. The specific rates for regeneration and consumption of NADPH (A) and ATP (B) were obtained from the best-fit metabolic flux distributions (Fig. 4). NADPH and NADH are not distinguished since cofactor usage of several enzymes was unclear. Gaps between the net production and consumption rates are shown.

The ATP regeneration and consumption in the wild-type GT strain indicated that a large amount of ATP was used by phosphoglycerate kinase and phosphoribulokinase (encoded by pgk and prk, respectively) for carbon fixation in the Calvin–Benson cycle (Fig. 5b), and the net ATP consumption rate in the metabolic network was 12.7 mmol g^–1 DCW h^–1. The estimated ratio of the ATP/NADPH demand is 2.94 (=12.7 ATP/4.3 NADPH). Since the ATP/NADPH ratio generated by the linear electron flow from eight absorbed quanta is 1.29 (=2.57 ATP/2 NADPH), these results suggests that the required amounts of ATP and NADPH were likely to be supplied by employing other mechanism such as the cyclic electron flow (Kramer and Evans 2011).

The cofactor regeneration and consumption in the ΔnrtABCD mutant revealed that the net ATP and NADPH consumption levels decreased to 4.19 and 1.28 mmol g^–1 DCW h^–1, respectively. The ratio of the ATP/NADPH demand (3.29) was also slightly increased in the ΔnrtABCD mutant compared with that in the wild-type GT strain.

Discussion

In this study, INST-MFA was employed to investigate a metabolic redirection in the nitrogen-limited growth of Synechocystis sp. PCC 6803 under photoautotrophic conditions. Since nitrogen starvation caused by a medium transfer is too transient to be analyzed by INST-MFA, a method for the large-scale cultivation of the ΔnrtABCD strain using the photobioreactor was established (Fig. 2). The ΔnrtABCD strain showed a series of phenotypes related to nitrogen limitation including slower growth rates (Fig. 2a), glycogen accumulation (Fig. 2d), and a pale yellow color due to the low phycobilisome level (Fig. 2b, c). The cultivation of the strains using the photobioreactor allowed us to maintain an exponential growth phase required for the INST-MFA experiment under nitrogen-limited conditions (Fig. 2a).

The time course MDV data for intracellular metabolites were obtained from the ¹³C-labeling experiment using NaH¹³CO₃ as a carbon source (Fig. 3). Based on the data, the metabolic flux distributions of the wild-type GT strain and the ΔnrtABCD mutant were estimated by INST-MFA (Fig. 4). The metabolic flux distribution in the wild-type GT strain showed that the metabolic flux levels for reactions related to the Calvin–Benson cycle were the largest in the metabolic network (Fig. 4a). The metabolic activity of the TCA cycle was relatively low under the photoautotrophic conditions because the metabolic flux was also at a low level. In the case of the ΔnrtABCD strain, the metabolic flux levels of the Calvin–Benson cycle were reduced to approximately one-third of that for the wild-type GT strain (Fig. 4b). These results were consistent with previous observations that the expression levels of genes associated with the Calvin–Benson cycle were significantly reduced during nitrogen starvation (Osanai et al. 2006, Krasikov et al. 2012).

The results of INST-MFA showed that the metabolic flux distribution patterns were similar in both strains, although the absolute metabolic flux levels were lower in the ΔnrtABCD mutant. These results indicated that the central metabolism slowed down during nitrogen limitation in the ΔnrtABCD strain without drastic redirection and reprogramming of the metabolic distribution. However, a metabolic flux towards the TCA cycle was reduced in nitrogen-limited conditions, suggesting a slow metabolic flow in the TCA cycle observed as the accumulation of TCA cycle metabolites in the previous study (Osanai et al. 2014). In contrast, the relative flux levels in the oxPPP, glycogen metabolism and anaplerotic reactions were elevated. This was probably due to the increased expression of genes required to respond to nitrogen starvation (Osanai et al. 2006).

One function of the oxPPP is to supply NADPH for cellular biosynthesis under photoheterotrophic conditions. The ¹³C-MFA of C. protothecoides demonstrated that the metabolic flux levels of the oxPPP drastically increased during photoheterotrophic conditions (Gopalakrishnan et al. 2015). The regenerated NADPH contributed to an increase in lipid biosynthesis. However, the metabolic function of the oxPPP bypass under photoautotrophic conditions is unlikely to produce an excess supply of NADPH because the Calvin–Benson cycle and the oxPPP bypass form a metabolic futile cycle consuming ATP without CO₂ fixation and a net NADPH regeneration. The metabolic flux analysis revealed that the relative levels of the oxPPP bypass flux were slightly increased in the ΔnrtABCD strain. The metabolic flux analysis also indicated that futile ATP consumption increased under nitrogen-limited conditions because glycogen synthesis and degradation as well as anaplerotic reactions constitute other metabolic futile cycles in cyanobacterial metabolism (Fig. 4), and the ratio of the demand for ATP/NADPH was slightly increased in the ΔnrtABCD mutant compared with that in the wild-type GT strain (Fig. 5).

The futile ATP consumption seems to be a reasonable strategy to adapt to nitrogen-limited environments by cyanobacteria. When an alternative nitrogen source suddenly becomes available, the re-establishment of nitrogen assimilation requires an additional supply of NADPH and ATP from the light reactions to activate the Calvin–Benson cycle and the nitrogen assimilation reactions. If there was no futile consumption in the metabolic network, the ATP supply would not be increased without activation of the light reaction. On the other hand, excess ATP consumed by the futile cycle can be rapidly diverted to facilitate nitrogen assimilation. Thus, futile ATP consumption could be a metabolic buffer for cyanobacterial metabolism, allowing them to respond to sudden changes in nutritional and environmental conditions. The results of this study indicate that this metabolic buffer tended to increase in ΔnrtABCD, and contribute to adapting to the nitrogen-limited conditions. Furthermore, it was also demonstrated that metabolic flux analysis is a promising tool for quantitatively investigating the mechanisms that regulate metabolism in cyanobacteria, complementing transcriptomic and metabolomic approaches.

Materials and Methods

Strains and culture conditions

The GT strain of Synechocystis sp. PCC 6803 was used in this study. A DNA construct consisting of the 5′-open reading frame (ORF) of the nrtABCD operon (500 bp), the kanamycin resistance gene from pHSG298 (Km^R) and the 3′-ORF of the nrtABCD operon (500 bp) was prepared by overlapping PCR. All primers for this cloning strategy are shown in Supplementary Fig. S1. This construct was used to establish the ΔnrtABCD strain according to a previously described transformation method (Joseph et al. 2014).

Cyanobacteria were cultured in modified BG-11 medium [2.7 µM EDTA disodium salt, 46 µM H₃BO₃, 20 mM HEPES, 1.6 µM Na₂MoO₄·2H₂O, 220 µM K₂HPO₄, 300 µM MgSO₄·7H₂O, 260 µM CaCl₂, 9.1 µM MnCl₂·4H₂O, 0.77 µM ZnSO₄·7H₂O, 0.32 µM CuSO₄·5H₂O, 0.17 µM Co(NO₃)₂·6H₂O, 16 µM FeCl₂·4H₂O], which was adjusted to a pH of 7.5 using 1 M KOH. Kanamycin (20 µg ml^–1) was added to cultures of the ΔnrtABCD strain. The wild-type GT and ΔnrtABCD strains were pre-cultured for 2 d in 500 ml flasks with 100 ml of modified BG-11 medium supplemented with 5 mM NH₄Cl. Following centrifugation, pelleted cells were washed with the modified BG-11 medium, and then inoculated with 20 ml of modified BG-11 medium containing nitrogen sources in 100 ml Erlenmeyer flasks. Cells were cultivated at 34°C with rotary agitation of 150 r.p.m. (BR-43FL, TAITEC) under continuous light-emiting diode (LED) illumination [about 40 µmol (photons m^–2) s^–1; LC-LED 450 W, TAITEC]. For cultivation in the photobioreactor, the pre-cultured cells were inoculated in a 1 liter reactor (BMZ-01NP3S, ABLE) with 550 ml of modified BG-11 medium supplemented with 25 mM NaNO₃ (pH 7.5) at 34°C with agitation at 200 r.p.m. Cells were exposed to continuous light illumination from two LEDs emitting at 660 and 475 nm at a ratio of 5 : 1. Total light intensity on the surface of the bioreactor was about 120 µmol photons m^–2 s^–1). Aeration was achieved using an air contactor producing 2.5% CO₂ at 1 v.v.m. Cell concentrations in culture were measured by optical density at 730 nm (OD₇₃₀) using a spectrophotometer (UVmini-1240, Shimadzu Co.). OD₇₃₀ was converted into dry cell weight using the following formula: dry cell weight (g DCW l^–1) = OD₇₃₀ × 0.16 (g DCW l^–1/OD₇₃₀).

Glycogen assay

Culture medium was centrifuged at 10,000 r.p.m. for 10 min at 4°C, and then washed with 1 ml of water. The cell pellets were treated with 200 µl of 30% (w/v) KOH for 90 min at 95°C. The mixture was frozen and the residue was washed with 600 µl of ethanol and dried with a vacuum. The residue was then resuspended in water and the glycogen concentration was determined using a glycogen assay kit (EnzyChrom, BioAssay Systems)

¹³C-labeling experiments

The gas supply for the batch cultivation in the photobioreactor was stopped during the mid-log phase growth (OD₇₃₀ ∼0.6), and 20 ml of an aqueous solution of 0.83 M Na₂¹³CO₃ (Cambridge Isotope Laboratories) was added to the medium. A consistent amount of cells [cell density (OD₇₃₀) × culture volume (ml) = 25] was harvested using a 50 ml syringe at 1, 30, 60, 90, 120, 180, 300 and 600 s after the start of labeling (SS-50ESZ, TERMO). The medium was immediately removed by vacuum filtration using a PTFE-type membrane filter (H100A090C, ADVANTEC), and the cells were soaked in 1.6 ml of pre-cooled methanol on the filter. Cells were then stored at –80°C until analysis. This sample collection procedure was performed within 20 s.

For metabolite extraction, 540 µl of water, 1.6 ml of chloroform and 100 µl of 0.1 mM aqueous ribitol were added to the cells. The suspension was vortexed for 1 min. The cell debris and the membrane filters were removed by centrifugation at 4,500 r.p.m. for 40 min at 4°C. The upper layer (1.6 ml) was collected and dried using a SpeedVac^™ system (SAVANT SPD1010, Thermo Scientific). The dried cell extracts were resuspended in 200 µl of water, and these samples were subsequently analyzed by GC-MS and LC-MS/MS. The GC- and LC-MS/MS analyses were performed according to previously described methods (Nishino et al. 2015, Okahashi et al. 2015). The results from the selected reaction monitoring method for quantifying ¹³C-enrichment for each metabolite are shown in Supplementary Table S1. The effect of naturally occurring isotopes was removed from the raw MS data to obtain corrected ¹³C-labeling patterns for carbons in the metabolic intermediates (van Winden et al. 2002).

Isotopically non-stationary metabolic flux analysis (INST-MFA)

INST-MFA was performed using a metabolic model including 62 reactions in the Calvin–Benson cycle, glycolysis, oxPPP, glycogen biosynthesis and degradation, anaplerotic reactions, glyoxylate shunt, TCA cycles without 2-OG dehydrogenase-mediated reactions, and two TCA bypass reactions (Supplementary Table S3) (Young et al. 2011, Xiong et al. 2014).

The metabolic flux for biomass synthesis was determined from a specific cell growth rate and from the biomass composition described in a previous report (Young et al. 2011). The metabolic flux distribution was estimated using OpenMebius software (Kajihata et al. 2014). The optimal metabolic flux distribution and metabolite pool sizes were investigated by fitting predicted time course curves of MDV with that of measured MDV data. For this purpose, the elementary metabolite unit framework (Antoniewicz et al. 2007) was used to calculate the ¹³C-enrichment for each metabolite from the metabolic fluxes and the labeling pattern of the substrate. Metabolic flux distributions were iteratively tuned by minimizing the following RSS using a non-linear optimization method:

$$RSS={\displaystyle \sum }_i{\displaystyle \sum }_t{\left(\frac{{\text{M}}_{i,\hspace{1em}t,\text{exp}}-{\text{M}}_{i,\hspace{1em}t,\text{sim}}}{{\text{σ}}_{i,\hspace{1em}t}}\right)}^2$$

where M_i,t,exp and M_i,t,sim represent the experimentally measured and simulated ¹³C-enrichment values of the ith metabolites at time t, respectively, and σ represents the SD (0.03). The 95% confidence intervals were determined using the Monte Carlo method with 1,000 iterations (Quek et al. 2009).

Supplementary data

Supplementary data are available at PCP online.

Funding

This work was supported by the Japan Science and Technology Agency [Strategic International Collaborative Research Program]; Strategic International Research Cooperative Program (SICORP) for JP–US Metabolomics; and the Japan Society for the Promotion of Science [KAKENHI grant No.16H06559].

Abbreviations

Abbreviations
 
• DCW

  dry cell weight

 
• Glu

  glutamate

 
• G6P

  glucose 6-phosphate

 
• GT

  glucose-tolerant

 
• INST-MFA

  isotopically non-stationary metabolic flux analysis

 
• LED

  light-emitting diode

 
• MDV

  mass distribution vector

 
• MFA

  metabolic flux analysis

 
• 2OG

  2-oxoglutarate

 
• ORF

  open reading frame

 
• oxPPP

  oxidative pentose phosphate pathway

 
• PEP

  phosphoenolpyruvate

 
• PGA

  mixture of 2-phosphoglycerate and 3-phosphoglycerate

 
• RSS

  residual sum of squares

 
• Suc

  succinate

 
• TCA

  tricarboxylic acid

Acknowledgments

The authors thank Professor Masayoshi Matsuoka (Sojo University) for kindly providing the glucose-tolerant strain of Synechocystis sp. PCC 6803.

Disclosures

The authors have no conflicts of interest to declare.