Introduction

Cyanobacteria are photosynthetic microbes that serve as an attractive platform for the sustainable production of chemicals and fuels, mainly due to their capability of converting atmospheric carbon dioxide into organic compounds by using solar energy, their relative rapid growth rate, and the readily available genetic toolbox for various species1. In the past years, the unicellular cyanobacterium Synechocystis sp. PCC 6803 (hereafter Synechocystis) has been successfully used as a cell factory for the production of commodity chemicals, including, n-butanol2, bisabolene3 and phenylpropanoids4. Besides autotrophy, cyanobacteria are able to grow under mixotrophic conditions. Mixotrophic metabolism allows the cells to catabolize organic carbon parallel to CO2 being assimilated photosynthetically. Addition of organic substrates in light growth conditions, though increasing the risk of contamination and cost, is considered as a promising cultivation strategy for the commercialization of cyanobacteria5,6, owing to enhanced biomass7 and chemical yield8,9.

The biosynthesis of various cell metabolites involves biochemical reactions, which require cofactors, such as adenosine triphosphate (ATP) and reduced nicotinamide adenine dinucleotides [NAD(P)H]. ATP is the major energy carrier in the cell, driving numerous metabolic pathways. The pyridine nucleotides NADH and NADPH serve as reducing equivalents in redox reactions. While NADH participates mainly in catabolic reactions, NADPH plays a role in anabolic reactions. In cyanobacteria, the photosynthetic electron transport chain is the major source for ATP and NADPH generation, when grown under light conditions10. The different glycolytic routes and the TCA cycle generate reducing agent in the form of NADH. In addition, transhydrogenases catalyze the interconversion between NADH and NADPH11.

Optimal performance of photosynthesis requires that the light-energy conversion by the photosystems and the downstream metabolic pathways are fine-tuned, to support the ATP/NADPH output ratio and the cellular energy economy12. The photosynthetic linear electron flow (LEF) in Synechocystis generates an ATP/NADPH ratio of approximately 1.28, which is less than required for CO2 fixation by the Calvin–Benson–Bassham (CBB) cycle and downstream biochemical pathways12. To deal with this imbalance, alternative pathways contributing to ATP production exist13,14. In addition, re-oxidation of NAD(P)H to NAD(P)+ and phosphorylation-dephosphorylation of reducing equivalents are important for balancing the ATP/NADPH budget and maintaining redox homeostasis15. Despite the importance of meeting the energy and redox requirement to achieve an efficient cellular performance16,17, only few metabolic engineering efforts of Synechocystis have targeted this designing strategy18, while most studies are concerned with controlling carbon partitioning1.

Constraint-based modelling (CBM) has been employed for the characterization of physiological capabilities of Synechocystis, for which several detailed genome-scale metabolic reconstructions are available19,20,21,22,23,24,25. CBM is based on imposing of a set of constraints that govern the operation of a metabolic network at steady state. These include, for example, the stoichiometry of the biochemical reactions, mass balance and thermodynamic laws. Flux balance analysis (FBA) is a common CBM-based approach to calculate, under given constraints, the intracellular flux distributions within the stoichiometric network, by optimizing an objective function26. As such, FBA can be used to predict the maximum (or minimum) growth rate or the production of desired metabolites, as well as to identify genetic interventions that force carbon flux toward chosen compound. Case studies demonstrating model-driven metabolic engineering of Synechocystis include the overproduction production of isoprene27, n-butanol28 and ethanol29.

In this work, we focus on alkenes, commercially valuable platform chemicals, which are traditionally used as detergents, lubricants and rubbers. In addition, they are compatible hydrocarbon fuels, due to their high energy content30. The development of gas-to-liquid technologies, enabling the oligomerization of small gaseous substrates into liquid chemicals, together with the environmental concerns associated with the use of fossil fuels and carbon dioxide emissions31, are driving the necessity for an environmentally-friendly, yet economically-feasible, manufacturing process of bio-alkenes32,33.

We addressed the question of how the model cyanobacterium Synechocystis balances its anabolic-catabolic processes, with respect to central cofactor metabolites [ATP and NAD(P)H], during growth under autotrophic and mixotrophic conditions. To this end, we employed genome-scale metabolic modelling, and analyzed the metabolism of Synechocystis strains overproducing alkenes as a case study.

Results and discussion

To enable a system analysis of alkene-overproducing Synechocystis strains, and assess the resulting energy and redox (im)balance, we chose four alkenes whose production using cyanobacteria has been previously demonstrated: isoprene34, isobutene35, ethylene36 and 1-undecene37. The four alkenes stem from different metabolic routes and possess different cofactors requirements (Fig. 1, Table 1).

Fig. 1: Non-native alkene biosynthesis pathways reconstructed in the iJN678_AK Synechocystis sp. PCC 6803 genome-scale model for cofactor balance analysis.
figure 1

TCA tricarboxylic acid, BCAA branched-chain amino acids. Irreversible reactions are indicated by one-headed arrows; reversible reactions are indicated by two-headed arrows. Metabolic reactions and metabolites (except heterologous ones) are indicated by their BiGG identifier73.

Full size image
Table 1 Evaluated turnover rates of energy and redox carriers.
Full size table

Evaluation of maximal production capability

First, we examined how increasing growth rate affected alkene production, using phenotypic phase plane analysis38. Under both autotrophic and mixotrophic growth conditions, we noticed a trade-off between growth (biomass production) and product synthesis. Predicted maximal alkene production rates resulted in zero biomass accumulation, and vice-versa (Fig. 2a–d). Therefore, we aimed to simulate a more real scenario, in which a cell produces high amounts of alkenes, whilst sustaining minimal biomass synthesis (Fig. 3a). We performed the computational analysis in a two-step optimization strategy (see Methods section).

Fig. 2: Alkene-biomass phenotypic phase planes.
figure 2

a Isoprene. b Isobutene. c Ethylene. d 1-undecene. Phenotypic phase planes were obtained by varying these fluxes in the range from zero to maximal rate, under constant photon flux. Auto autotrophic, Mixo mixotrophic.

Full size image
Fig. 3: Factors influencing the stoichiometric evaluation in this study.
figure 3

a As a trade-off occurs between biomass and alkene production, a two-step optimization strategy was performed, as described in the Methods section. b Cellular resources are allocated to biomass, cell maintenance and alkene production.

Full size image

Subsequently, we evaluated the theoretical maximum production capability, by analyzing the maximum productivity, maximum mass yield and maximum c-mol yield (Table 2); all are parameters of interest from both biotechnological and biochemical aspects. By comparing carbon-source-dependent bioproduction, our computational analysis illustrated that the maximum theoretical mass yield is approximately doubled under mixotrophic conditions, as compared to autotrophic condition, for the production of the examined alkenes. Except for 1-undecene, autotrophic conditions resulted in a comparable productivity for all objectives. Under both growth conditions, the yields were similar for all alkenes, except for ethylene, for which these values were lower. Under autotrophic growth, the release rate of CO2 is 3.81 mmol/gDW/h for biomass production, whereas ethylene production resulted in a release rate of 5.27 mmol/gDW/h (Supplementary Table 6). Thus, the low yields found for ethylene can be owed to the high carbon loss.

Table 2 Theoretical maximum productivity, mass yield and c-mol yield.
Full size table

Evaluation of ATP and NAD(P)H demands

We then analyzed Synechocystis metabolism from the metabolite-centric point of view, also termed as flux-sum analysis39. This approach is useful for understanding metabolite essentiality40, and for studying the intracellular network robustness to perturbations in metabolite turnover rates39. Even though the net accumulation of a metabolite might be zero, the overall turnover rate is an indicator of the importance of the specific metabolite.

Under both autotrophic and mixotrophic growth, a higher usage of NADPH (turnover rate of 3.87–5.49 mmol/gDW/h), and ATP (turnover rate of 7.24–8.61 mmol/gDW/h) was observed, as compare to the usage of NADH (turnover rate of 0.01–0.49 mmol/gDW/h) (Table 1). These results are in line with previous experimental studies, showing that NADPH is being preferentially used over NADH in Synechocystis, when grown under autotrophic conditions10. In addition, all strains exhibited a comparable ATP requirement, although with some variations. For example, less ATP was required for 1-undecene production under autotrophic conditions, and for isoprene under mixotrophic conditions. Under mixotrophic conditions, lower turnover rates were observed for both NADPH and NADH. An explanation for these observations is that the introduced biosynthesis pathways are mainly anabolic, and therefore use more NADPH than NADH.

Since the overall consumption and production rates are equal, under the steady-state assumption, the flux-sum serves as a proxy of the metabolite pool size in the system, by which the cellular ATP/NADPH ratio could be examined. The analysis revealed a narrow range of intracellular ATP pool (Table 1). In contrast, a larger variation in NADPH levels was determined. We computed that biomass accumulation required an ATP/NADPH ratio of 2.11 and 19.49, under autotrophic and mixotrophic conditions, respectively. In comparison, all the examined alkenes required ATP/NADPH ratios below the values for biomass (Table 1). Therefore, in order to attain a growth-coupled production, genetic interventions are needed to manipulate the cofactor pools41. A lower ATP availability could be achieved by introducing an “ATP-wasting” mechanism (e.g., ATPM)42. A higher cofactor availability of NAD(P)H could be achieved by the introduction of non-native NADH-dependent oxidoreductases, or protein engineering of enzymes for swapped cofactor-dependency43.

It should be noted that, the ATP/NADPH ratios presented here slightly differ from earlier published results44,45. These differences could be justified by that: (i) Here, we considered flux in the entire metabolic network, whereas previous studies mainly considered photosynthetic reactions (e.g., ATPSu and NDH-1 and FNOR). (ii) Different metabolic network sized were examined, covering different number of reactions acting on ATP and NAD(P)H: 760 reactions44, 780 reactions45, and 883 reactions in this study. (iii) We simulated a synergistic interaction between biomass formation (10%) and alkene biosynthesis, whereas earlier studies modelled 0%44 or 75%45 of growth, which affected the evaluation of ATP/NAD(P)H demand.

Evaluation of intracellular flux distributions

In order to identify the metabolic reactions contributing to the computed turnover rates, we employed a reaction-centric approach by using pFBA. By comparing biomass and alkene-overproducing strains, potential reactions whose metabolic fluxes constrain alkenes synthesis in Synechocystis could be recognized. We would like to note that FBA and 13C-metabolic flux analysis (MFA) are two different approaches to determine the intracellular fluxes46. Further, our model does not include information regarding cellular regulation or enzyme kinetics, which influence the conclusion on the flow of carbon in the cell.

Cofactor energy metabolism

Under both trophic growth modes, the majority of ATP generation was performed by the photosynthetic ATP synthetase (ATPSu) (Table 3). Under autotrophic conditions, a residual respiratory activity (ATPS4rpp_1) was determined for all objectives, but for the 1-undecene production. Our data are in line with earlier studies, showing that the respiratory pathway in Synechocystis is active also during light, albeit at low capacity47. Isoprene and ethylene production required a higher contribution of respiratory oxidative phosphorylation, whereas for isobutene, pyruvate kinase (PYK), from lower glycolysis, contributed to ATP generation (Table 3).

Table 3 Predicted flux distributions of cofactors-producing reactions.
Full size table

Reactions involved in ATP consumption were more diverse than for ATP generation and differed between growth conditions. Under autotrophic growth, ATP consumption was dictated by reactions within CBB cycle and glycolysis/gluconeogenesis (phosphoglycerate kinase, PGK and phosphoribulokinase, PRUK) for all strains (Table 4). Earlier in vivo studies also determined PGK and PRUK as major ATP-consumers in Synechocystis48,49. Our data also revealed specific flux distributions depending on the objective function. For ethylene production, the ATP-consuming reactions were within nitrogen metabolism pathways (e.g., glutamine synthetase, GLNS), whereas for isoprene production, the terpenoid biosynthesis pathways contributed the most to ATP consumption. Bicarbonate transport (BCT1_syn) exhibited a large ATP consumption for production of isoprene and 1-undecene. Acetyl coenzyme A (acetyl-CoA) carboxylase (ACCOAC) was mainly involved in 1-undecene accumulation, probably due to high amount of carbon assimilated during the elongation process of fatty acyl chain for 1-undecene synthesis.

Table 4 Predicted flux distributions of cofactors-consuming reactions.
Full size table

Reactions involved in ATP consumption were more diverse than for ATP generation and differed between growth conditions. Under autotrophic growth, ATP consumption was dictated by reactions within CBB cycle and glycolysis/gluconeogenesis (phosphoglycerate kinase, PGK and phosphoribulokinase, PRUK) for all strains (Table 4). Earlier in vivo studies also determined PGK and PRUK as major ATP-consumers in Synechocystis48,49. Our data also revealed specific flux distributions depending on the objective function. For ethylene production, the ATP-consuming reactions were within nitrogen metabolism pathways (e.g., glutamine synthetase, GLNS), whereas for isoprene production, the terpenoid biosynthesis pathways contributed the most to ATP consumption. Bicarbonate transport (BCT1_syn) exhibited a large ATP consumption for production of isoprene and 1-undecene. Acetyl coenzyme A (acetyl-CoA) carboxylase (ACCOAC) was mainly involved in 1-undecene accumulation, probably due to high amount of carbon assimilated during the elongation process of fatty-acyl chain for 1-undecene synthesis.

Under mixotrophic conditions, the ATP consumption was mainly governed by cyanophycin metabolism (cyanophycin synthetase, CYPHYS) during biomass, isobutene and 1-undecene synthesis (Table 4). For isoprene accumulation, ATP maintenance (ATPM) accounted for the majority of ATP consumption (Table 4). As ATPM contributes to cellular functions other than growth (e.g., turgor pressure) (Fig. 3b), it can be inferred that a hyper-producing isoprene strain would more likely become ATP-limited when grown mixotrophically, as compared to the other alkene-producing strains.

We additionally investigated how the ATP precursors (ADP, AMP and Pi) are recycled within the metabolic network (Fig. 6, Supplementary Tables 9–12 and Supplementary Figs. 11–19). The stoichiometric analysis revealed that, under both autotrophic and mixotrophic growth, the de novo adenosine biosynthesis, downstream to phosphoribosylpyrophosphate synthetase (PRPPS), contributed equally to all objectives but ethylene production (Fig. 6, Supplementary Table 9 and Supplementary Figs. 11–19). The salvage adenosine pathway through adenylate kinase (ADK1) contributes the most to ADP-AMP cycling during ethylene (Supplementary Tables 7 and 10 and Supplementary Figs. 16–17) and biomass overproduction (Fig. 6 and Supplementary Table 7). For ethylene synthesis, argininosuccinate synthase (ARGSS) in arginine metabolism showed enhanced flux to produce AMP (Fig. 5a, Supplementary Fig. 10 and Supplementary Table 9). Of note, phosphoketolase (PKETF) was shown to be a common consumer of Pi for the production of all objective functions (Fig. 4, and Supplementary Table 12).

Fig. 4: Metabolic flux map of central carbon metabolism for Synechocystis sp. PCC 6803.
figure 4

Synechocystis was simulated to grow under autotrophic conditions, with biomass set as objective. Reaction rates (mmol/gDW/h) were predicted using pFBA69. Note that, the colors associated with the fluxes are relative to the other reactions rates presented in the map. Irreversible reactions are indicated by one-headed arrows; reversible reactions are indicated by two-headed arrows. For reaction directionality, refer to the data availability section. For metabolic flux maps of central carbon metabolism for biomass simulated to grow mixotrophic conditions, and other objectives simulated to grow under auto- and mixotrophic conditions, refer to Supplementary Figs. 1–9. The map was generated with Escher web-tool72. Metabolic reactions and metabolites are indicated by their BiGG identifier73. TCA tricarboxylic acid, GABA branched-chain amino acids γ-aminobutyrate, CBB Calvin–Benson–Bassham, ED Entner–Doudoroff, EMP Embden-Meyerhof-Parnas, PPP pentose phosphate pathway.

Full size image

The enhanced ADK1 flux during biomass (Fig. 6, Supplementary Fig. 11 and Supplementary Tables 7 and 10) and ethylene production (Supplementary Figs. 16–17 and Supplementary Tables 7 and 10) presumably sustains the increase in respiration rate, by providing more ADP to the oxidative phosphorylation pathway. A prior study revealed that low respiration rate is a possible limiting factor of ethylene production through ethylene-forming enzyme (EFE) in Saccharomyces cerevisiae50, and could explain the low ethylene production found experimentally in Synechocystis36.

Based on our studies and others51, we suggest that a combinatorial metabolic engineering, with the aim to form an efficient ATP cycling for fueling the CBB cycle, could be based on the ATP-consuming enzymes (e.g., PRUK and PGK), and on the ATP-producing enzymes (i.e., PYK). In addition a high ATP turnover would be beneficial for the heterologous production of alkenes, and probably other products, in Synechocystis, as previously confirmed for E. coli42.

We propose that, under mixotrophic conditions, the regeneration of cellular nitrogenous polymers (e.g., cyanophycin) is essential for optimized nitrogen and energy metabolism in the cell. The data are in agreement with previous studies, arguing that the transition in growth conditions affects nitrogen metabolism through cyanophycin52.

Cofactor redox metabolism

In agreement with earlier reports10,53, the flux analysis revealed that, under both autotrophic and mixotrophic growth, the photosynthetic ferredoxin-NADP+ reductase (FNOR) contributed largely to the supply of NADPH (Table 3). An exception was noted for biomass production, under mixotrophic conditions, for which NADPH was produced by the oxidative pentose phosphate pathway (Table 3). In relation to the calculated ATP/NADPH ratios (Table 1), we presume that the high ratio determined for biomass production under mixotrophic conditions stems from the lower flux through the FNOR, as compared to that as seen for the other objectives.

In the ethylene-overproducing strain, the flux through isocitrate dehydrogenase (ICDHyr) was significantly higher than that in the other modeled strains (Table 3 and Supplementary Figs. 6–7). We suggest that ICDHyr is a good candidate for enhanced production of TCA-derived molecules, such as ethylene, supported by the results from studies on engineered ethylene-producing strain of Synechocystis54.

The flux distribution of reactions catalyzing NADPH consumption revealed that for all objectives, NADPH was mainly consumed by glyceraldehyde-3-phosphate dehydrogenase (GAPDi_nadp), a key enzyme within the CBB cycle and gluconeogenesis (Table 4). As could be expected, the flux though ketol-acid reductoisomerase (KARA1) within the branched-chain amino acids (BCAA) biosynthesis was enhanced for isobutene production, and reaction fluxes though the terpenoid and fatty acids pathways (i.e., fatty-acyl reductases) were increased for isoprene and 1-undecene.

Though carrying low flux under both growth conditions, production of NADH were distributed among BCAA biosynthesis pathway (3-isopropylmalate dehydrogenase, IPMD), TCA cycle (malate dehydrogenase, MDH) and L-arginine and L-proline metabolism (phosphoglycerate dehydrogenase, PGCD) (Table 3). Notably, an enzyme from the L-arginine and L-proline metabolism (Δ1-pyrroline-5-carboxylate dehydrogenase, P5CD) was found to be predominant during ethylene overproduction (Fig. 5b).

Fig. 5: Metabolic flux map of nitrogen metabolism for Synechocystis sp. PCC 6803.
figure 5

Synechocystis was simulated to grow under autotrophic conditions, with ethylene set as objective. Reaction rates (mmol/gDW/h) were predicted using pFBA69. a L-glutamate and L-arginine regeneration through the urea cycle. b L-glutamate and L-proline regeneration. Note that, the colors associated with the fluxes are relative to the other reactions rates presented in the map. Irreversible reactions are indicated by one-headed arrows; reversible reactions are indicated by two-headed arrows. For reaction directionality see data availability section of the GitHub repository. For metabolic flux maps of nitrogen metabolism for Synechocystis sp. PCC 6803 with ethylene set as objective, simulated to grow under mixotrophic conditions, refer to Supplementary Fig. 10. The map was generated with Escher web-tool72. Metabolic reactions and metabolites are indicated by their BiGG identifier73.

Full size image

We observed that under autotrophy, for all objectives but 1-undecene, the photosynthesis (NAD(P)H dehydrogenase, NDH1_2u) is the main NADH consumer (Table 4). For 1-undecene, L-lysine metabolism (L-homoserine dehydrogenase, HSDxi) contributed the most. HSDxi and L-alanine metabolism (L-alanine dehydrogenase, ALAD_L) were major NADH consumers for biomass, isoprene, isobutene and 1-undecene production. For ethylene production, nitrogen metabolism was a key NADH consumer (e.g., L-glutamate synthase, GLUSx).

The data imply that the regeneration of the NADH pool in the system is facilitated mainly by amino acids metabolism. The reduction of nitrate to ammonia requires a considerable amount of NADPH. We suggest that designing an alternative NADH-dependent pathway for ammonium assimilation (for example by introducing ALAD_L) could reduce the competition for NADPH. Hence, increasing the potential for NADPH-dependent product formation (Fig. 1), while simultaneously enabling nitrogen cycling in the cell.

We additionally investigated the redox balancing mechanism by the (de)phosphorylating enzymes, and identified that NAD kinase (NADK) had a negligible contribution to the overall flux of NAD+ consumption (Supplementary Table 7). As NADP+ serves as the final electron acceptor of photosystem I10, this finding was rather surprising. The role of NADK only begins to be elucidated55,56, and the functionality of NADP+ phosphatase remains to be discovered.

Considering the large variation in NADPH turnover rates (Table 1), we postulate that, under light-limited conditions, the ATP/NADPH ratios are mainly governed by the reduction of NADP+, rather than by the regeneration of ATP. Our data are consistent with a previous study, demonstrating that, under steady-state photosynthesis, CO2 assimilation is not limited by the availability of ATP, but by the rate of NADPH formation57.

Energy and redox balance are interlinked with the production of central metabolites

Our findings show that balancing the intracellular ratio of ATP and NADPH is directly interconnected with and influence the regeneration of central carbon metabolites (i.e., pyruvate and acetyl-CoA). We identified that, under autotrophic conditions, carbon is channeled back to the Embden-Meyerhof-Parnas (EMP) pathway to synthetize glycogen through glucose-1-phosphate adenylyltransferase (GLGC) and glycogen synthase (GLCS1) reactions (Fig. 4 and Supplementary Figs. 2, 4, 6). In Synechocystis, glycogen metabolism was shown to act as an ATP-buffering system, and the inactivation of GLGC led to an increased NADPH pool58,59. It has furthermore been reported that glycolytic routes form anaplerotic shunts that replenish metabolites for carbon fixation by the CBB cycle60. We suggest that gluconeogenesis and glycogen turnover should be considered within future bioengineering research. This is in contrast to a number metabolic engineering studies, in which a glycogen-deficient Synechocystis strain is used as chassis for production61,62.

Pyruvate kinase (PYK) and Malic enzyme (ME2) are the major contributors to pyruvate synthesis, although at distinct ratios, depending on the objective function (Supplementary Table 13). The data are in accordance with previous reports, suggesting that due to allosteric-inhibition of PYK by its product, ATP63, ME2 functions as an additional route for synthesizing pyruvate in Synechocystis7. Fluxomic studies pinpointed that PYK reaction represents a bottleneck for isobutyraldehyde production in Synechococcus elongatus PCC 7942, which was overcome by heterologous overproduction of PYK64. Under mixotrophic conditions, the Entner–Doudoroff pathway also provided pyruvate for biomass synthesis (Supplementary Table 13). Hence, based on pure stoichiometry, and in agreement with previous research, we suggest that PYK and ME2 are good candidates for metabolic engineering of Synechocystis for hydrocarbon production. The CO2 released by ME2 can be further assimilated by phosphoenolpyruvate carboxylase (PPC) (Fig. 4), thus resulting in a net zero carbon loss and a high product yield.

Interestingly, and less intuitively, IPMD within the isobutene biosynthesis pathway (Fig. 1), showed a high contribution for redox recycling in biomass, isoprene and 1-undecene biosynthesis as well (Table 3). In cancer research, it was shown that the TCA cycle in tumor cells is heavily fed by acetyl-CoA derived from BCAA65. Taken together, we presume that, the catabolism of BCAA provides an alternative route for the synthesis of acetyl-CoA - a key precursor molecule for many cellular biosynthesis pathways, including terpenoids and lipids.

Our results further emphasize the essentiality of metabolic shortcuts in central carbon metabolism for attaining effective metabolic states66. We show that acetate metabolism through phosphoketolase serves as an efficient carbon- and energy-recycling pathway for cell maintenance, as suggested in a biochemical study focusing on the role of phosphoketolase in cyanobacteria67.

Serine and glycine metabolism represents an uncharacterized pathway for cellular energy and redox homeostasis in Synechocystis

We show that FBA is a powerful tool to investigate cellular metabolism globally and highlight flux patterns, by which new routes could be found and serve as basis for future studies. We learned that reactions participating in the L-serine synthesis, one-carbon metabolism and the glycine cleavage system (SOG pathway) function as an auxiliary pathway to regenerate ATP, NADPH and NADH in Synechocystis. Fluxes through SOG were identified when maximizing biomass and alkenes production (Fig. 6, Supplementary Tables 9–12 and Supplementary Figs. 11–19). The discovery that SOG pathway contributes to the biosynthetic requirements of ATP and NADPH of high-proliferating cells68, strengthens our suggestion that SOG deserves a thorough investigation for metabolic engineering of Synechocystis.

Fig. 6: Metabolic flux map of ATP and NAD de novo and salvage metabolism for Synechocystis sp. PCC 6803.
figure 6

Synechocystis was simulated to grow under autotrophic conditions, with biomass set as objective. Reaction rate (mmol/gDW/h) were predicted using pFBA69. Note that, the colors associated with the fluxes are relative to the other reactions rates presented in the map. Irreversible reactions are indicated by one-headed arrows; reversible reactions are indicated by two-headed arrows. For reaction directionality, refer to the data availability section. For metabolic flux maps of central carbon metabolism for biomass simulated to grow mixotrophic conditions, and other objectives simulated to grow under auto- and mixotrophic conditions, refer to Supplementary Figs. 11–19. The map was generated with Escher web-tool72. Metabolic reactions and metabolites are indicated by their BiGG identifier73. NAD nicotinamide adenine dinucleotide, SOG serine, one-carbon cycle, glycine synthesis.

Full size image

To conclude, we demonstrate the significance of high flexibility in activity of energy- and carbon-converting enzymes in order to achieve the required ATP/NADPH output ratio. It was apparent that the overall activity of the Synechocystis metabolism is governed by several high-flux core reactions, which probably allows the cell to adapt rapidly to perturbations in environmental conditions and genetic modifications. We also show that assessment of ATP and NAD(P)H balance can neither be done in isolation from each other and their precursors, nor from carbon and nitrogen metabolism. Even though the predictions are yet to be validated in vivo, our results provide hints for strain design strategies enabling efficient bioproduction of valuable compounds derived from core metabolic pathways in Synechocystis.

Methods

Computational tools

Parsimonious enzyme usage FBA (pFBA)69 was employed on COBRApy v0.20.070 using the commercial solver Gurobi v10.0.1 (Gurobi Optimization, Inc., Houston, TX, United States) in Python v3.8.5, using the packages Pandas v1.3.0, NumPy v1.19.4, Matplotlib v3.3.3 and Pickle v4.0. Model consistency was tested with MEMOTE v0.12.071. Visualization of the metabolic reaction network and the obtained fluxes was done using the Escher web-tool72.

Metabolic network reconstruction

The iJN678 genome-scale metabolic model of Synechocystis sp. PCC 680319, downloaded from the BiGG database73, was used for the computational analysis. The model reconstruction includes 622 genes, 863 reactions and 795 metabolites. iJN678 also contains three different biomass compositions, allowing for simulating autotrophic, mixotrophic, and heterotrophic growth conditions, in addition to constraining the appropriate carbon and photon uptake rates.

The metabolic network was updated in accordance with the literature, yielding iJN678_AK, (Supplementary Table 1), as follows: addition of the tricarboxylic acid (TCA) cycle shunt reactions74,75,76, addition of the phosphoketolase reactions67,77, addition of the Entner–Doudoroff (ED) pathway78, addition of the light-independent L-serine biosynthesis pathway79, addition of prephenate transaminase, prephenate dehydratase and arogenate dehydrogenase80, transhydrogenase (NADTRHD) was changed to a reversible reaction11, addition of NADH dehydrogenase 2, alternative respiratory terminal oxidase, and flavodiiron proteins 2 and 4, and constraining cytochrome b6f, cytochrome c oxidase and NAD(P)H dehydrogenase to zero47,81. In addition, non-growth associated ATP maintenance (ATPM)82,83 reaction was added to the model. The ATPM coefficient is represented by a pseudo-reaction that degrades ATP into ADP and orthophosphate (Pi) (Eq. 1).

$${\rm{ATP}}+{{\rm{H}}}_{2}{\rm{O}}\to {\rm{ADP}}+{\rm{Pi}}+{\rm{H}}$$
(1)

For the evaluation of heterologous alkene production in Synechocystis, four biosynthesis pathways34,35,36,84 were separately implemented into the iJN678_AK model, yielding iJN678_AK_isoprene, iJN678_AK_isobutene, iJN678_AK_ethylene and iJN678_AK_1-undecene (Supplementary Tables 2–5). In each reconstructed model, a cytoplasmatic export reaction for each alkene was introduced and set as the objective function. The resulting metabolic network reconstructions are encoded in SBML format (Systems Biology Markup Language)85, and are available in the GitHub repository https://github.com/amitkugler/CBA.

Constraint-based modeling

To evaluate the steady-state flux distributions, we employed pFBA69, in which the objective function is optimized using FBA, followed by the minimization of total absolute flux through all gene-associated reactions (Eqs. 2–5).

$${maximize}\,Z={c}^{T}\,\cdot\, v$$
(2)

s.t

$$\mathop{\sum }\limits_{j=1}^{n}{S}_{{ij}}\,\cdot\, {v}_{j}=0$$
(3)
$$\,{{v}_{j}}^{{LB}}\le {v}_{j}\le {{v}_{j}}^{{UB}}$$
(4)

s.t

$${minimize}\mathop{\sum }\limits_{j=1}^{m}\left|{v}_{{irrev},\,j}\right|$$
(5)

Where \(Z\) is the objective function, \(v\) is the flux vector of the metabolic reactions, \({c}^{T}\) is the transposed vector of the objective coefficient. \({S}_{{ij}}\) refers to the stoichiometric coefficient of metabolite i participating in reaction j, and the \({v}_{j}\) refers to the vector of reaction flux (mmol/gDW/h; gDW, gram dry weight) of reaction j at steady state. The flux \({v}_{j}\) is the j-th component of an n-dimensional flux vector v, where n is the total number of fluxes, m is the number of gene-associated irreversible reactions (\({v}_{{irrev}}\)). LB and UB correspond to the lower bound and upper bound, respectively, of the j-th reaction in the flux vector v. Equation 3 can also be used to distinguish between reversible and irreversible reactions, where \({{v}_{j}}^{{LB}}=0\) for the latter.

To explore the flux solution space, flux variability analysis (FVA)86 was performed. The fraction of the optimum was set to 95%, which accounts for 5% variation around the best-known objective value. The FVA revealed infinite bounds for transhydrogenase (NADTRHD), irreversible leucine transaminase (LEUTAi) and glycine cleavage system (GLYCL and GLYCL_2). It was reported earlier that, the transhydrogenase reaction was poorly resolved by 13C measurements87. Therefore, we chose to model redox balance without an active NADTRHD, as previously suggested88. At the same time, we could learn how the entire metabolic network balances between NADPH/NADH. In addition, the upper and lower bounds of LEUTAi and GLYCL and GLYCL_2 were constrained to zero, and the glycine cleavage system was modeled by introducing three separate reactions (Supplementary Table 1).

In order to avoid thermodynamic infeasible cycles, the “add_loopless” function in COBRApy was applied, that makes thermodynamically infeasible loops impossible89. In addition, we coupled our analysis to FVA.

According to previous studies7,90, autotrophic metabolism was simulated by constraining the carbon dioxide, bicarbonate and glucose uptake rates to 0, 3.7 and 0 mmol/gDW/h, respectively. Mixotrophic metabolism was simulated by constraining the inorganic carbon (carbon dioxide and bicarbonate) uptake rates to 0 and glucose uptake rate to 0.38, respectively. For both trophic conditions, the photon uptake rate to was set to 45 mmol/gDW/h, to simulate a light-limited condition. These settings resulted in maximum specific growth rates of 0.082 h−1 and 0.056 h−1, under autotrophic and mixotrophic conditions, respectively, which are in agreement with experimental growth values7,22. In addition, the biomass objective function91 was set to BIOMASS_Ec_SynAuto, BIOMASS_Ec_SynMixo, depending on the simulated growth conditions, to account for changes in macromolecular composition.

Two-step optimization strategy

Following earlier studies92,93, showing that a feasible overproduction of hydrocarbons in Synechocystis entails a carbon partitioning to biomass of 10%, we simulated a scenario where the cellular growth is limited, while achieving a high alkene production rate. We used a two-step optimization strategy, where, first, the biomass objective function was maximized. Then, the minimum flux through the biomass objective function was set to 10% of the obtained value, followed by the maximization of alkene production.

Evaluation of production strains

For providing a global perspective on the relation between growth rate and alkene production a phenotypic phase plane analysis38 was carried out.

The flux-sum analysis was used to quantify the cofactors turnover rates among the metabolic network39. The flux-sum values for a metabolite represent the sum of flux in all the reactions that consume or produce it, multiplied by the corresponding stoichiometric coefficients (Eq. 6).

$$\,{\varPhi }_{i}=\sum _{j\varepsilon {P}_{i}}{S}_{{ij}}\bullet {v}_{j}=\sum _{j\varepsilon {C}_{i}}{S}_{{ij}}{\bullet v}_{j}=0.5\sum _{j}{\rm{|}}{S}_{{ij}}\bullet {v}_{j}{\rm{|}}$$
(6)

Where \({P}_{i}\,\) denotes the set of reactions producing metabolite i, and \({C}_{i}\) denotes the set of reactions consuming metabolite i. Under the steady-state conditions, the consumption and production rates for any metabolite are equal. Thus, the turnover rate of metabolite i is half the absolute sum of consumption and the generation rates.

The theoretical maximum productivity is expressed h−1 or mmol/gDW/h for biomass and alkenes, respectively. The theoretical maximum yield is expressed as gram product produced per gram of substrate consumed. The carbon-conversion efficiency was estimated by calculating carbon-mole (c-mole) produced per c-mole of substrate consumed.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.