Cyanobacterial carboxysome mutant analysis reveals the influence of enzyme compartmentalization on cellular metabolism and metabolic network rigidity

Abstract

Cyanobacterial carboxysomes encapsulate carbonic anhydrase and ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO). Genetic deletion of the major structural proteins encoded within the ccm operon in Synechococcus sp. PCC 7002 (ΔccmKLMN) disrupts carboxysome formation and significantly affects cellular physiology. Here we employed both metabolite pool size analysis and isotopically nonstationary metabolic flux analysis (INST-MFA) to examine metabolic regulation in cells lacking carboxysomes. Under high CO₂ environments (1%), the ΔccmKLMN mutant could recover growth and had a similar central flux distribution as the control strain, with the exceptions of moderately decreased photosynthesis and elevated biomass protein content and photorespiration activity. Metabolite analyses indicated that the ΔccmKLMN strain had significantly larger pool sizes of pyruvate (> 18 folds), UDPG (uridine diphosphate glucose), and aspartate as well as higher levels of secreted organic acids (e.g., malate and succinate). These results suggest that the ΔccmKLMN mutant is able to largely maintain a fluxome similar to the control strain by changing in intracellular metabolite concentrations and metabolite overflows under optimal growth conditions. When CO₂ was insufficient (0.2%), provision of acetate moderately promoted mutant growth. Interestingly, the removal of microcompartments may loosen the flux network and promote RuBisCO side-reactions, facilitating redirection of central metabolites to competing pathways (i.e., pyruvate to heterologous lactate production). This study provides important insights into metabolic regulation via enzyme compartmentation and cyanobacterial compensatory responses.

1. Introduction

Cyanobacteria are promising production systems to meet renewable resource demands because they are photosynthetic, genetically tractable, and capable of rapid growth. Photosynthetic organisms assimilate carbon using ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO). However, in addition to its slow carboxylation kinetics, RuBisCO also oxygenates ribulose-1,5-bisphosphate (RuBP) producing 2-phosphoglycolate (2PG), a compound that is toxic to cells at high concentrations (Kern et al., 2011). Therefore, cyanobacteria have evolved CO₂ concentrating mechanisms (CCM) around RuBisCO, encapsulating the enzyme within clathrate structures comprised of multiple proteins that form microcompartments referred to as carboxysomes. Carboxysomes result from protein-protein interactions between structural proteins (ccmKLMNO) and RuBisCO (rbcL, rbcS), which promotes the formation of an icosahedral shell that also encases carbonic anhydrase (CA, cccA) (Price et al., 1992; Cameron et al., 2013). The hierarchy of protein-protein interactions required for carboxysome assembly has previously been elucidated (Cameron et al., 2013). These polyhedral structures concentrate and increase CO₂ availability to RuBisCO (Cai et al., 2009; Dou et al., 2008), without saturating CA (Mangan and Brenner, 2014).

Bacteria, as simple systems, are presumed to have little or no spatial subcellular organization, and thus the cytosol is assumed to be homogenous. However, evidence of metabolons (a multi-enzyme complex formed by non-covalent bonds between enzymes in a metabolic pathway) in bacteria exists (Ishikawa et al., 2004; Broddrick et al., 2016), indicating that enzymes may be more self-organized than previously suspected (Sweetlove and Fernie, 2018). Carboxysome compartmentalization and photosystem co-localization on cyanobacterial membranes are an obvious exception, and the advanced spatial organization confers enhanced growth and productivity. A small fraction of un-encapsulated RuBisCO is also localized outside of carboxysomes, often in association with the cell periphery near the thylakoid membranes in procarboxysome assembly centers (Agarwal et al., 2009; Cameron et al., 2013). It is possible that a cascade of enzymes proximal to carboxysomes might form metabolons that channel CO₂ fixation reactions. The direct passing of metabolites among cascade enzymes may uniquely contribute to metabolic regulation (Abernathy et al., 2017b). In eukaryotes, the formation of metabolons are associated with cellular organelles and membrane compartments, such as the mitochondria or peroxisomes (Graham et al., 2007). In prokaryotes, evidence suggests that enzyme proximity channeling may be widely present in central pathways (Abernathy et al., 2019). Cyanobacterial RuBisCO is the starting point for reduced carbon metabolism in the cell, and the encapsulating carboxysome may offer an anchoring point for cascade enzymes. Removal of carboxysomes can change the cellular locations of both RuBisCO and downstream enzymes (Fig. 1A). Currently, it is unclear how the redistribution of enzymes in space affects the metabolic network.

Fig. 1. Strain physiology under high and low CO₂ and light conditions.

(A) Representative fluorescence images of the Synechococcus PCC 7002 control and ΔccmKLMN. Chlorophyll fluorescence is red, and GFP-labeled RuBisCO is cyan. Scale bars are 1 μm. (B) Growth under 0.2% CO₂ and 25 μmol m⁻²•s⁻¹. (C) Growth under 1% CO₂ and 125 μmol m⁻²•s⁻¹. Glycerol at a concentration of 0.2% (w/w) or sodium bicarbonate (NaHCO₃) at a concentration of 50 mM were added to cultures to investigate photo-mixotrophic growth rates.

Isotopic experiments are regularly used to investigate pathway activity (Antoniewicz et al., 2007; Boyle et al., 2017) and metabolite channeling (Shearer et al., 2005; Williams et al., 2011; Zhang et al., 2017). Isotopically nonstationary metabolic flux analysis (INST-MFA) uses time-dependent mass isotopomer distributions (MID) of free metabolites to estimate fluxes through photosynthetic metabolism (Young et al., 2011; Abernathy et al., 2017a; Hendry et al., 2017; Xiong et al., 2015a; Ma et al., 2014). The same labeling experiments can also identify heterogeneous intracellular environments (i.e. metabolically inactive pools of central metabolites) (Abernathy et al., 2017a). In this study, INST-MFA on a carboxysome (ΔccmKLMN) mutant offers a strategy to understand Synechococcus 7002 metabolism in the absence of an innate microcompartment. Specifically, a ΔccmKLMN mutant was generated (Gordan et al., 2016), eliminating the protein structure around RuBisCO (Fig. 1). Due to the lack of a carbon concentrating mechanism, the mutant was grown under 1% v/v CO₂ environments (hereby referred to as the high CO₂ condition). The mutant phenotypes were measured and compared to the control strain, including spatial locations of RuBisCO, flux organizations, metabolite concentrations, biomass compositions, labeling dynamics of cascade metabolites, and photosynthetic capabilities. The results obtained here provide holistic insights into the effect of microcompartments on cyanobacterial metabolism.

2. Materials and methods

2.1. Strains and cultivation conditions

The Synechococcus PCC 7002 control strain and ΔccmKLMN mutant were cultured and maintained in A+ medium without citrate (18 g/L NaCl, 0.6 g/L KCl, 1 g/L NaNO₃, 5 g/L MgSO₄•7H₂O, 0.05 g/L KH₂PO₄, 0.37 g/L CaCl₂, 0.03 Na₂EDTA•2H₂O, 1 g/L Tris HCl, 0.555 mM H₃BO₃, 2.3 μM ZnCl₂, 0.21 μM MoO₃, 3 μM NH₄Fe(SO₄)₂•12 H₂O, 22 μM MnCl₂, 0.012 μM CuSO₄, 0.05 CoCl₂, 0.0004 g/L vitamin B12) (Stevens et al., 1973). 100 μg/mL kanamycin or 25 μg/mL gentamicin were added to the ΔccmKLMN and control strain cultures, respectively. Concentrated glycerol, acetate, pyruvate, and sodium bicarbonate stocks were filtered and added to denoted cultures for studying carbon substrate effects. The same temperature (37 °C), shaking speed (200 rpm), and culture volume (20 mL of medium in 125 mL shake flasks) were used in all experiments. Cultivation occurred at either 0.2% v/v CO₂ or 1% v/v CO₂ with continuous light. In order to balance the minimal CO₂ fixation that occurs in the low CO₂ condition (0.2%), the light intensity was adjusted to 25 μmol of photons m⁻²•s⁻¹. To assess the metabolism in the high CO₂ condition (1%), light intensity was scaled linearly to 125 μmol of photons m⁻²•s⁻¹. During cell growth, optical density was measured at 730 nm (OD₇₃₀).

2.2. Synechococcus PCC 7002 ΔccmKLMN mutant generation and imaging

The Synechococcus PCC 7002 mutant was generated previously (Gordan et al., 2016) and kindly provided by the Pfleger group (University of Wisconsin-Madison). Briefly, the control strain was created by inserting RbcL-GFP in the SYNPCC7002_A2842 neutral site along with a gentamicin resistance cassette. The ΔccmKLMN mutant was generated by replacing ccmKLMN with a kanamycin resistance cassette along with RbcL-GFP (Gordan et al., 2016). The GFP marker enabled us to confirm the presence or absence of carboxysomes in the control and ΔccmKLMN, respectively (Fig. 1). Fluorescence microscopy images were acquired on a custom Nikon TiE inverted wide-field ORCA Flash 4.0 V2+ Digital sCMOS camera (Hamamatsu) using a Nikon CFI60 Plan Apochromat Lamda 100× oil immersion objective (1.45 N.A.) using an excitation wavelength of 488 nm (GFP) or 640 nm (Cy5/chlorophyll) with an LED source (Spectra X Light Engine, Lumencor, Beverton, OR) and emission through standard GFP and Cy5 filter sets. Prior to imaging, cells were grown in ambient or high CO₂ conditions in an AL-41L4 Environmental Chamber (Percival Scientific, Perry, IA) maintained at 37 °C with continuous illumination (~150 μmol photons m⁻² s⁻¹) provided by cool white fluorescent lamps.

2.3. Dynamic labeling experiments

INST-MFA was conducted at 1% v/v CO₂ with 125 μmol of photons m⁻²•s⁻¹ to investigate phenotypic differences between the control and the mutant strains. For each experiment, cultures were inoculated from a seed culture to an OD₇₃₀ of 0.05 in A+ medium and grown to exponential phase. Prior to the labeling experiment, a 2 mL aliquot of saturated NaH¹³CO₃ (> 98% purity, Sigma Aldrich, St Louis) was injected into the medium for a final concentration of 4 g/L of NaH¹³CO₃. Following the ¹³C-pulse, the cultures were quenched at different time intervals (30, 60, 120, 180, 300, 150, 600, 900 s) with ice cold medium and a liquid nitrogen bath (Abernathy et al., 2017a). The time-courses samples were centrifuged for 5 min (8000×g, ~0 °C). Biomass pellets were then frozen at −80 °C until metabolite extraction and analysis. Biological duplicates were used to generate standard deviations for experimentally measured MID values. Relative metabolite pool sizes were analyzed by an MS isotopomer ratio approach using fully labeled cell extracts as internal standards, specifically, fully labeled E. coli K-12 (cultured with uniformly labeled ¹³C glucose and ¹³C sodium bicarbonate). The estimation of metabolite pool sizes was performed by mixing a known amount of ¹³C labeled E. coli biomass with unlabeled cyanobacteria (three biological replicates). Then, the mixtures were extracted for LC-MS analysis. The isotopic ratio of each metabolite (labeled vs. unlabeled) was normalized by the amount of E. coli and cyanobacterial biomass, respectively (Abernathy et al., 2017a).

2.4. Metabolite extraction and analysis

The pelleted samples were extracted with a chloroform–methanol method (Ma et al., 2014). Supernatant aliquots were taken for analysis by lyophilizing 5 mL of medium. All samples were re-suspended in 1 mL of ddH₂O for LC-MS/MS. Ion-pairing LC–MS/MS was performed at the Proteomics and Mass Spectrometry Facility, Donald Danforth Plant Science Center, St. Louis. Succinic acid in the supernatant was precisely quantified using an UV-method enzyme kit from R-Biopharm (Darmstadt, Germany) with A+ medium as controls (Table S1).

2.5. Isotopically nonstationary ¹³C-MFA

A cyanobacterial metabolic network was constructed with carbon atom transitions, which included the Calvin cycle, the oxidative pentose phosphate pathway, the TCA pathway, anaplerotic pathways, and a grouped photorespiration pathway (Table S2). A lumped biomass equation based on biomass composition analysis and biochemical equations was used to fit the model to biomass production. The INCA platform used a custom MATLAB ODE solver for a least squares regression to fit the experimentally determined free metabolite labeling data for Synechococcus PCC 7002 control and the ΔccmKLMN mutant (Young, 2014). INCA provided a goodness of fit via a Chi square statistical test as well as 95% confidence intervals of all estimated parameters by evaluating the sensitivity of the sum of squared residuals to parameter variations. Dilution parameters from INCA were used to account for labeling dilutions due to metabolically inactive pools (Young et al., 2011).

2.6. Biomass composition analysis

The cells were harvested during exponential growth by centrifugation, washed with a 0.9% NaCl solution and then freeze-dried. Protein and amino acid compositional analyses were performed by the Molecular Structure Facility, University of California (Davis, CA). Total carbohydrate was measured using the Anthrone Method, as described in our previous paper (Abernathy et al., 2017a). Fatty acid content and profile analyses were performed by Microbial ID (Newark, USA). Chlorophyll, pigments, RNA, DNA, and inorganics were estimated from previously published literature (Hendry et al., 2017). The biomass measurements were used to fine-tune the biomass equation in the INCA model. p values were calculated using a two-tailed student’s t-test.

2.7. GC–MS analysis of proteinogenic amino acids from mixotrophic cultures

To assess the capability of Synechococcus 7002 to uptake glycerol, acetate, or pyruvate, unlabeled seed cultures were grown at 1% CO₂ and used to inoculate A+ medium (5% v/v ratio). The A+ medium was supplemented with either 25 mM glycerol-2-¹³C (second position labeled), 25 mM fully labeled acetate-¹³C₂, or 25 mM pyruvate-3-¹³C (third position labeled). The ¹³C-cultures were transferred to either 1% CO₂ or 0.2% CO₂ conditions and grown until mid-log phase. The labeling in proteinogenic amino acids was then analyzed using a TBDMS method (You et al., 2012).

2.8. Measuring photosynthetic parameters and oxygen evolution

Cells were harvested during log growth and adjusted based on cell numbers. The chlorophyll fluorescence parameter, maximum quantum yield (F_V/F_M) of PSII, was measured using a fluorescence fluorometer, FL-200 (Photon Systems Instruments, Brno, Czech Republic). Cells were dark adapted for 90 s before F₀ and F_M were measured using red LED measuring light. Using a custom-built Clark-type electrode, light induced oxygen evolution was measured for 2 min at 37 °C following the addition of 10 mM NaHCO₃ as described by a previous protocol (Ungerer et al., 2018). For respiratory O₂ uptake, O₂ consumption was measured for 2 min under dark conditions. Chlorophyll a content was measured by removing one milliliter of cell culture, extracting with methanol, and measuring the resulting absorbance at 665 and 652 nm as determined by a previous method (Porra et al., 1989).

3. Results

3.1. Growth comparison between the control strain and ΔccmKLMN mutant

The Synechococcus control strain and the ΔccmKLMN mutant were both grown at high CO₂ (1% v/v) and under continuous light (125 μmol photons m⁻²•s⁻¹) (Fig. 1 B&C). The control and ΔccmKLMN mutant growth rates in the exponential phase were 0.11 ± 0.005 h⁻¹ and 0.085 ± 0.005 h⁻¹, respectively. In the high CO₂ concentration, the addition of NaHCO₃ reduced mutant growth while there was no growth effect on the control strain. Under low CO₂ conditions (0.2% v/v CO₂), inclusion of 50 mM NaHCO₃ reduced growth rate for the control strain. When the CO₂ concentration was 0.2% or below, the ΔccmKLMN mutant exhibited little growth and the addition of 50 mM NaHCO₃ was harmful to mutant survival. Under high CO₂ conditions, neither the control strain nor the ΔccmKLMN strain utilized glycerol to promote cell growth. The mutant showed slight growth with glycerol under low CO₂ circumstances. These observations confirmed the deleterious impact on cyanobacterial growth if the cyanobacterial ability to concentrate CO₂ in microcompartments is compromised.

3.2. Biomass composition analysis

The biomass compositions (Fig. 2) were measured to reveal the differences in carbon partitioning between the control strain and the ΔccmKLMN mutant. The fatty acid profile was compositionally similar between the two strains, with the exception of statistically significant (p value < 0.05) increases of 18:2 cis-9,12 and 18:1 cis-11 fatty acids and a decrease of 18:1 cis-9 fatty acid in the ΔccmKLMN mutant. Protein-based amino acid profiles were also similar, with the exceptions of slightly elevated levels of leucine and lysine and slightly reduced levels of serine and threonine in the mutant. The biomass of the ΔccmKLMN strain exhibited a higher protein fraction (p value < 0.01), 51.2 ± 1.7 wt% protein of dry cell weight (DCW), compared to the control strain (43.7 ± 0.4%). The total glycogen/carbohydrate and lipid content were comparable with the ΔccmKLMN mutant comprised of 10.4 ± 3.1 wt% carbohydrate and 7.3 ± 3.0 wt% lipids and the control strain containing 12.3 ± 3.4 wt % carbohydrate and 6.8 ± 4.9 wt% lipids. The remaining components were expected to be chlorophyll and other pigments, soluble metabolites, and ashes.

Fig. 2. Biomass compositional analysis of Synechococcus PCC 7002 control and ΔccmKLMN mutant.

(A) Fatty acid profile found in each strain as a relative abundance of the total fatty acid content. (B) Amino acid profile found in each strain as the relative mole percentage. asx: aspartate and asparagine; glx: glutamate and glutamine. (C) Macromolecule biomass composition as a percentage of dry cell weight. * indicates a significant difference (p value < 0.05). Both control and mutant strains were cultivated in high CO₂/light conditions.

3.3. Flux topology in PCC 7002 control and ΔccmKLMN mutant strains

Using the biomass constraints as a sink reaction in the metabolic model, MFA was performed with the control and ΔccmKLMN mutant strains under the high CO₂ and high light conditions. Fig. 3 represents a simplified flux network and is scaled with net fluxes normalized to a CO₂ uptake of 100 units using a network derived from annotated genes found in the KEGG database. The full model can be found in Table S2. The INST-MFA model produced statistically acceptable sum-of-squared-residual fits for both the ΔccmKLMN mutant and the control strain (Tables S3 and S4, respectively).

Fig. 3. Synechococcus PCC 7002 control and ΔccmKLMN flux maps as determined by INCA (under high CO₂/light conditions.

Net fluxes are normalized to a CO₂ uptake of 100 mmol and reported as mmol/gDCW/h. ΔccmKLMN flux values are reported in blue, and PCC 7002 control flux values in red. Green arrows represent biomass synthesis reactions. Percentages of total ¹³C enrichment at steady state of key metabolites are represented by bar graphs. Total ¹³C enrichment was calculated at isotopic steady state. Standard deviations are based off biological replicates (n = 2). The complete metabolic network and full flux results can be found in Tables S2–S4.

Based on the INST-MFA analysis, both the control and mutant strains showed similar flux topology under high light and high CO₂ conditions. The Calvin cycle in PCC 7002 made use of ~110 mmol CO₂. Due to the malic shunt (Malate→Pyruvate + CO₂) and pyruvate dehydrogenase (Pyruvate→AcCoA + CO₂), ~15% of the fixed carbon was lost. The resulting biomass synthesis was 2.5 gDCW for ΔccmKLMN mutant and 2.8 gDCW for the control strain. Nearly 11% of the carbon flux from CO₂ was directed to citrate synthesis for initiating the TCA pathway, while major carbon flux was cycled through the CBB cycle to regenerate RuBP. The relative fluxes through most central metabolic pathways (including the glycolysis, the TCA and the Calvin cycle pathways) were statistically indistinguishable between the ΔccmKLMN strain and the mutant (despite small differences in fluxes directed towards biomass synthesis and extracellular organic acids). This result indicated that, under sufficient light/CO₂, ΔccmKLMN largely maintains a similar flux network to the control strain.

Compared to the other model cyanobacterium strains, Synechocystis 6803 (Young et al., 2011) and Synechococcus 7942 (Jazmin et al., 2017), PCC 7002 showed subtle metabolic network differences. In the TCA cycle, PCC 7002 lacks fumarate hydratase which interconverts malate and fumarate such that fumarate is solely generated as a byproduct of purine biosynthesis (Fig. 3). PCC 7002 also lacks the typical γ-aminobutyric acid shunt to close the TCA cycle. However, the TCA cycle can be functionally connected via the succinic semialdehyde bypass route (Zhang et al., 2016). Under our growth condition, the flux from AKG to succinate was close to zero and thus the TCA pathway was mainly used for biosynthesis. Furthermore, phosphoketolase, phosphate acetyltransferase, and acetate kinase are not present in the PCC 7002 genome annotation. Thus PCC 7002 may not be able to generate acetyl-CoA from the pentose phosphate pathway. In contrast, Synechocystis 6803 has a phosphoketolase shunt that splits xylulose-5-phosphate to acetyl phosphate and G3P (note: acetyl phosphate can be converted into acetyl-CoA or acetate to promote bio-production) (Knoop et al., 2013; Xiong et al., 2015b; Chwa et al., 2016; Moriyama et al., 2014). As INST-MFA is limited in its ability to distinguish between these pathways, we were unable to determine any non-annotated homologues or new enzymes in PCC 7002 central pathways. These enzymes may further fine-tune the flux network.

3.4. Photorespiration and photosynthesis

Carboxysome knock-out may impact photorespiration and photo-system activities. Photorespiration occurs when RuBisCO oxygenates RuBP to 2PG and 3-phosphoglycerate (3PGA). 2PG can be metabolized to glyoxylate and then to glycine, serine, and glycerate. In the control strain, photorespiration can be repressed by the carboxysomes which provide an efficient CO₂ concentrating mechanism. The flux map showed (Fig. 3) that the control strain produced 4.5 mol of 2PG per 100 C-moles of biomass under the high light/high CO₂ condition. Best estimates of photorespiratory flux for the ΔccmKLMN mutant indicated a net flux of 7.6 mol of 2PG per 100 C-moles of carbon fixation. This increase in photorespiration is consistent with increased free RuBisCO without a carboxysome’s physical protection against O₂ (Cameron et al., 2013). To investigate the functions of the photosystems, chlorophyll fluorescence (F_V/F_M) under high CO₂ conditions was measured for both strains (Fig. 4a). The F_V/F_M ratio determines the quantum efficiency of photosystem II (PSII). Here, the variable fluorescence (F_V) was increased in the control strain compared to the ΔccmKLMN mutant while the maximum fluorescence (F_M) remained similar. Thus, there was a 15% higher F_V/F_M ratio in the control strain (i.e., higher photosynthetic efficiency in the control strain). Furthermore, photosynthesis activity is associated with O₂ evolution. The ΔccmKLMN mutant exhibited a lower net O₂ evolution than the control strain (397 ± 68 μmol O₂/mg chl a/hr vs 496 ± 74 μmol O₂/mg chl a/hr). But respiratory O₂ uptake by the control and the mutant strains was identical under dark condition (Fig. 4b), which accounted for 30 ± 7% of the O₂ evolution in the ΔccmKLMN strain and 20 ± 3% in the control strain. In cyanobacteria, respiratory O₂ uptake is due to photorespiration, oxidative phosphorylation (Wan et al., 2017), and the Mehler reactions (formation of H₂O₂ by photosynthesis) (Allahverdiyeva et al., 2011). Therefore, photo-respiration/photosynthesis analyses suggested that removal of carboxysomes moderately increased cyanobacterial photorespiration and decreased the photosystem efficiency under high CO₂ growth conditions.

Fig. 4. Comparison of photosynthetic efficiency in PCC 7002 control and ΔccmKLMN mutants.

(A): PSII activities (F_V/F_M). (B): O₂ evolution and uptake. F₀: initial fluorescence; Fv: variable fluorescence; F_M: maximum fluorescence. Technical (n = 2) and biological replicates (n = 3) were used for standard deviations. *p < 0.001. Percentages represent the respiratory O₂ uptake (as measured in the dark) as a fraction of the total net O₂ evolution.

3.5. Change in pool sizes in response to carboxysome knockout

Metabolite concentrations impact reaction free energy and flux organization (Park et al., 2016). This study further investigated ΔccmKLMN compensatory responses at metabolite levels (Heise et al., 2015). Pool sizes were measured using fully ¹³C labeled E. coli cell extracts as an internal standard (Fig. 5a). The ΔccmKLMN mutant had significantly increased levels of intracellular pyruvate compared to the control (18-folds higher). The mutant also had significantly higher levels of intracellular UDPG (2.1-fold), aspartate (2.5-fold), and glycerate (1.6-fold) and reduced levels of ADPG (0.5-fold). UDPG was also more rapidly labeled in the ΔccmKLMN mutant (Fig. S1), which may be beneficial for sucrose synthesis (Qiao et al., 2018). The metabolite pool size of RuBP was lower and the levels of 2PG and 3PGA were slightly decreased compared to the control. The ΔccmKLMN mutant also had an increased pool size and labeling rate of glycerate (a downstream photorespiration intermediate), which provides further evidence for the elevation of photorespiration flux. In the ΔccmKLMN mutant, the flux from PEP to OAA (aspartate precursor) was expected to decrease in comparison to the control strain (Hackenberg et al., 2012). However, free aspartate accumulated in the mutant, possibly due to a slower anabolism for utilizing free amino acids. Increased expression of nblA1, involved in degradation of the phycobilisome light-harvesting apparatus, has been previously observed in the cyanobacteria lacking carboxysomes (Hackenberg et al., 2012). This suggests that protein turnover could also influence metabolite pools in the ΔccmKLMN. Overflow metabolites in the supernatant were also detected by LC-MS. The ΔccmKLMN mutant excreted a small amount of organic acids (e.g., succinate and malate) (Fig. 5b). We utilized a quantitative enzyme-based spectroscopic assay and found that the ΔccmKLMN mutant produced extracellular succinate up to 2.8 ± 0.3 mg/L in the late growth phase (Table S1).

Fig. 5. Comparison of metabolite pool size between Synechococcus PCC 7002 control and ΔccmKLMN.

(A) Relative pool size comparisons between the two strains. A y-axis of 1 represents equal metabolite concentrations while a y-axis > 1 indicates the factor of pool size in ΔccmKLMN greater than PCC 7002 control. The relative pool size was normalized using a¹³C-labeled internal standard and cell density. (B) Succinate and malate in the culture supernatant were detected by LC-MS, and the peak intensity (counts per second, cps) was proportional to their concentrations. Exact succinate concentration was further quantified using an enzyme kit (Table S1).

Intriguingly, ¹³C-pulse experiments indicated that central metabolites in the mutant strain had high ¹³C-enrichments at the end of tracing experiments (Fig. 3 and Fig. S1). For example, ¹³C-enrichments in ASP, UDPG, and glycerate from the mutant reached > 80% at isotopic steady state while enrichment in same metabolites from the control strain was between 40 and 60% (see discussion section for further explanations).

4. Discussion

4.1. Compensatory response in cyanobacteria conserves flux networks

Carboxysomes are key microcompartments in cyanobacteria for carbon concentration and assimilation. The loss of carboxysomes significantly impairs cell growth under atmospheric and low CO₂ conditions. Under high CO₂ concentration (1% v/v), the ΔccmKLMN mutant maintains flux ratios in central pathways similar to the control strain. In previous reports, quantitative RT-PCR and microarrays were used to monitor the abundance of gene transcripts in both ΔccmM and wild type cyanobacterial cells grown in high CO₂ conditions (Emlyn-Jones et al., 2006; Hackenberg et al., 2012). Despite large transcriptional changes seen in genes involved in photosynthesis, high-light stress, and N assimilation in the ΔccmKLMN mutant, little-to-no transcriptional change was detected in its carbon uptake (i.e., genes encoding carbon transporter systems and transcriptional regulators), the Calvin Cycle, and other central metabolic pathways. These reports indicate that these central metabolic pathways are likely post-transcriptionally regulated (Hackenberg et al., 2012). The loss of the CCM impairs the carbon supply and energy-balance, leading to significant accumulations of intracellular metabolites (i.e., UDPG, glycerate, and pyruvate) as well as secretion of succinate and malate. The CCM is also closely tied to allosteric metabolites (Burnap et al., 2015), and a loss of carboxysomes may cue metabolite flux shifts towards free amino acids (i.e. aspartate). Macromolecule partitioning in biomass was also changed and the ΔccmKLMN mutant contains a higher protein content. Potentially, the ΔccmKLMN mutant may accumulate more stress proteins than the wild type strain (Hackenberg et al., 2012).

Free RuBisCO was reported to produce a small level of pyruvate instead of 3PGA via beta elimination of a phosphate ion and oxidative stress conditions (e.g., H₂O₂) increased this RuBisCO based pyruvate formation (Andrews and Kane, 1991). It has also been proposed that the loss of the carboxysome limits cyanobacterium ability to dissipate excess light energy. The resulting stresses (e.g., production of reactive oxygen species) on ΔccmKLMN mutant may lead to a significant increase of intracellular pyruvate accumulation through this RuBisCO side-reaction. Additionally, high-expression of nblA1, involved in degradation of the phycobilisome light-harvesting apparatus, was also observed in the cyanobacterial strain lacking carboxysomes (Hackenberg et al., 2012). This suggests that protein turnover in the ΔccmKLMN can be another source of pyruvate as well as free aspartate. Furthermore, the ΔccmKLMN mutant had lower [ATP]/[AMP] than the control strain (Fig. S2). An elevated ratio of [NADPH]/[NADP] was also observed in the ΔccmKLMN mutant (Fig. S2), which may indicate saturated terminal electron acceptors for photosynthesis and thus lead to lower net O₂ evolution (Wang et al., 2016). In summary, ΔccmKLMN mutant demonstrated a suboptimal balance between the rates of photosynthesis and carbon fixation after spatial re-organizations of RubisCO. However, the compensatory responses at different cellular processes as well as metabolite overflow were able to restore a conserved flux network under high CO₂ conditions (Cano et al., 2018).

4.2. Using carboxysome-deficient mutants to redirect carbon fluxes

Cyanobacterial species are promising photobiorefineries (Yu et al., 2015). Often, cyanobacterial engineering efforts target pathways that act as carbon sinks (Xiong et al., 2015a). For example, blocking the synthesis of glycogen is considered a useful strategy for rerouting carbon flux towards bioproduction (Ungerer et al., 2018; Hendry et al., 2017). However, such a strategy does not necessarily increase product titers (Li et al., 2014). In this study, the ΔccmKLMN mutant was able to increase UDPG pool sizes (by two-fold) as well as UDPG synthesis rates (Fig. 5 and Fig. S1). As UDPG is the precursor for sucrose and other valuable polysaccharides like heparosan (Sarnaik et al., 2019), the ΔccmKLMN mutant can potentially be utilized for production of value-added products. Additionally, the ΔccmKLMN mutant under high CO₂ growth conditions accumulates pyruvate and organic acids, while largely maintaining a similar growth rate as the control strain. Common metabolic engineering strategies for increasing pyruvate concentration typically include N-deprivation and knocking out glycogen synthesis, but both methods cause significant growth inhibition (Benson et al., 2016; Jackson et al., 2015). Thus, knocking out carboxysomes may represent an alternative strategy to increase intracellular pyruvate (Figs. 3 and 5). For example, the removal of carboxysomes from a lactate producing PCC 7002 strain was recently shown to increase productivity by 24% and to increase titer by 125% (Clark et al., 2018). As carboxysome-free cyanobacterial mutants offer a unique method of for bio-containment of genetically engineered cell factories, these findings offer additional benefits for using a ΔccmKLMN mutant as a phototrophic platform.

4.3. Impact of carboxysome loss on photomixotrophic metabolism

To probe photomixotrophic metabolism in the WT and carboxysome mutant, labeled glycerol, acetate, and pyruvate were used for isotopic experiments (Fig. 6). Both strains incorporated acetate and resulted in labeled Leu (derived from acetyl-CoA) and Glu (derived from ketoglutarate), which indicates that organic carbons are incorporated to metabolites of the TCA cycle. The incorporation of the ¹³C-carbon from acetate labeling into leucine but not isoleucine suggests PCC 7002 uses a threonine dependent pathway rather than the citramalate pathway for isoleucine synthesis (Wu et al., 2010). Under the high CO₂ conditions, a mixotrophic environment had minimal impact on ΔccmKLMN mutant growth. Under the low CO₂ condition, acetate supplementation rescued growth of the ΔccmKLMN mutant and resulted in the incorporation of acetate carbons into the biomass (revealed by isotopic labeling of amino acids). The increase of acetate usage benefitted photomixotrophic metabolism when cyanobacterial carboxysomes were disrupted. Glycerol was also found to be incorporated into biomass in the ΔccmKLMN mutant under the low CO₂ condition. Literature has indicated that glycerol increased NAD(P)H in PCC 7002 cells growing at atmospheric conditions (Ludwig and Bryant, 2012). Here, neither strain utilized significant amounts of labeled glycerol under high CO₂ conditions. Finally, exogenous pyruvate can be co-utilized by both strains for synthesis of some amino acids (such as leucine, isoleucine and glutamate). However, pyruvate impaired ΔccmKLMN mutant survival under low light and low CO₂ conditions.

Fig. 6. Steady-state proteinogenic amino acid labeling from labeled glycerol, acetate and pyruvate to monitor photomixotrophic growth of PCC 7002 control (ctrl) and ΔccmKLMN.

GC-MS data for proteinogenic amino acids under photomixotrophic conditions with 25 mM labeled carbon. Standard deviations are estimated based on two biological replicates (n = 2). Percent ¹³C increase is graphed (i.e., evidence of organic carbon utilization). (A) Labeling of amino acids and (B) Ratio of biomass accumulation under mixotrophic and low light/CO₂ conditions. (C) Labeling of amino acids and (D) Ratio of biomass accumulation under mixotrophic and high light/CO₂ conditions. Note: “ratio of biomass accumulation per day” is defined as: $\frac{{\text{Maximal OD}}_{\text{730}}\text{ increase over time (with organic carbon)}}{{\text{Maximal OD}}_{\text{730}}\text{ increase over time (photoautotrophic)}}$.

4.4. Insights into microcomponent based flux regulation

Presence of bacterial metabolons or metabolite channeling may cause a heterogeneous mixture of intracellular metabolites. In INST-MFA, dilution parameters are typically introduced to aid in model fitting due to observations that metabolites never reach fully isotopic enrichment. These fitted dilution parameters for multiple metabolites represent unlabeled biomass reflux or heterogeneous metabolite pools that arise from enzyme localization and/or metabolite channeling. Removal of carboxysomes resulted in higher steady state ¹³C enrichment in central metabolites (particularly sugar phosphate metabolites) and smaller dilution parameters from INST-MFA. This indicates that metabolite pools become more homogeneous after disruption of metabolons. Recent literature has focused on metabolite channeling via enzyme clusters as a way for cells to regulate fluxes and cell homeostasis (Zhang et al., 2017; Sweetlove and Fernie, 2018; Abernathy et al., 2017b). In cyanobacteria, the Calvin cycle enzymes may channel intermediates through enzyme co-localization, as evidenced through ultra-centrifugation and immunoelectron microscopy studies (Agarwal et al., 2009). Disruption of native channels and redistribution of enzymes may decrease inactive metabolite pools, leading to more ¹³C enrichment in some metabolites at the end of pulse-tracing experiments. The channeling concept offers a new method for loosening metabolic rigidity and for flux redirection towards products. Deciphering channeling associated flux regulation can improve microbial feedstock utilization and biomanufacturing prospects (Abernathy et al., 2017b).

5. Conclusions

Carboxysome removal in PCC 7002 alters the spatial organization of CO₂ assimilation. Enzyme organization and protein-protein interactions affect flux networks, but the quantitative details remain poorly understood. The information provided here on cell metabolons/channeling can improve INST-MFA modeling results and minimize the isotopic dilution effect. The outcomes from this study also offers insights into micro-compartment based metabolic regulations. Future studies are needed to elucidate the rules dictating the formation of metabolons or metabolite channeling in central pathways, and as such will enable bio-designers to add an extra level of metabolic control to current methods that rely mostly on gene copy number, transcriptional and translational regulations.

Abbreviations:

2PG

2-phosphoglycolate

2PGA

2-phosphoglycerate

3PGA

3-phosphoglycerate

6PG

6-phosphogluconate

ACA(or AcCoA)

acetyl-CoA

AKG

α-ketoglutarate

ADPG

adenosine diphosphate glucose

CARB

carbohydrate

CIT

citrate

DHAP

dihydroxyacetone phosphate

E4P

erythrose-4-phosphate

F6P

fructose-6-phosphate

FBP

fructose bisphosphate

FUM

fumarate

G1P

glucose-1-phosphate

G6P

glucose-6-phosphate

G3P

glyceraldehyde-3-phosphate

GA

glycerate

gDCW

gram dry cell weight

GOX

glyoxylate

MAL

malate

OAA

oxaloacetate

PEP

phosphoenolpyruvate

PYR

pyruvate

R5P

ribose-5-phosphate

RuBisCO

ribulose-1,5-bisphosphate carboxylase/oxygenase

Ru5P

ribulose-5-phosphate

RuBP

ribulose-1,5-bisphosphate

S7P

sedoheptulose-7-phosphate

SBP

sedoheptulose-1,7-bisphosphate

SUC

succinate

UDPG

uridine diphosphate glucose

X5P

xyulose-5-phosphate