Fructose-1,6-bisphosphatase is involved in heterotrophic growth and glycogen metabolism in cyanobacteria

Abstract

Cyanobacteria switch between photoautotrophic and heterotrophic carbon metabolism during diurnal cycles. A classical control point is characterized by two glycolytic phosphofructokinases (PFKs) and a bifunctional fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase (F/SBPase; slr2094), which catalyses two reactions in the Calvin–Benson–Bassham (CBB) cycle. In addition, Synechocystis possesses a fructose-1,6-bisphosphatase (FBPase; slr0952) with yet unknown physiological function and biochemical properties. Our aim was to elucidate the physiological role of FBPase, in combination with the above-mentioned enzymes. We discovered that FBPase is specific for fructose 1,6-bisphosphate (FBP), showing no SBPase activity, and unlike F/SBPase does not exhibit any biochemical regulatory properties. In contrast to F/SBPase, FBPase is not involved in the CBB cycle, but instead affects growth and glycogen metabolism under heterotrophic conditions. We hypothesize that FBPase may influence glycogen turnover by controlling cellular levels of its substrate, FBP, since FBP is involved in the formation of glucose 1,6-bisphosphate, which is a regulatory metabolite for the control point between glycogen and central carbon metabolism at the level of phosphoglucomutases. Our data indicate that PFK and FBPase act as an antagonistic enzyme couple in darkness. Furthermore, we found redox-insensitive FBPases from plant chloroplasts to be closely related to Synechocystis FBPase, indicating that they might serve a similar function.

Synechocystis fructose-1,6-bisphosphatase specifically hydrolyses fructose 1,6-bisphosphate and is involved in heterotrophic growth and glycogen metabolism in darkness, whereas the bifunctional fructose-1,6-biphosphatase/sedoheptulose-1,7-bisphoshatase catalyses two Calvin–Benson–Bassham cycle reactions in the light.

Introduction

Cyanobacteria are the only prokaryotes performing oxygenic photosynthesis and are widely recognized as ancestors of this process in the plant kingdom. During the day, cyanobacteria photosynthesize, fix CO₂ to generate carbohydrates via the Calvin–Benson–Bassham (CBB) cycle, also known as the reductive pentose phosphate cycle, and store excess carbon intracellularly as glycogen. In darkness, the direction of carbohydrate metabolism is reversed. In the heterotrophic mode, glycogen or extracellular glucose is degraded through various catabolic pathways, primarily the oxidative pentose phosphate (OPP) pathway and glycolysis via the catabolic Embden–Meyerhoff–Parnass (EMP) pathway (Yang et al., 2002; Wan et al., 2017). Besides having an either anabolic or catabolic carbohydrate metabolism, cyanobacteria also enter conditions under which mixed forms are required. Especially in transition stages between light and darkness or in the presence of external carbohydrates in the light, glycolytic pathways can be shortened to glycolytic shunts [phosphoglucose isomerase (PGI) and OPP shunt] that feed carbohydrates into the CBB cycle for replenishment to enhance CO₂ fixation and growth (Makowka et al., 2020; Schulze et al., 2022). Similarly, in plants carbohydrate reserves from the vacuole and the cytosol are converted into ribulose 5-phosphate via the so-called cytosolic glucose 6-phosphate (G6P) shunt, which comprises the first two enzymes of the catabolic OPP pathway. Ribulose 5-phosphate is then transported into the chloroplasts and further into the CBB cycle. Under stress conditions replenishment of the CBB cycle also occurs via a chloroplastic G6P (OPP) shunt (Sharkey and Weise, 2016; Xu et al., 2022; Sharkey, 2024).

The anabolic CBB cycle and the catabolic OPP and EMP pathways share several enzymes and reactions, enabling the opportunistic reversal of carbon flow based on environmental conditions such as light intensities, CO₂ levels, and metabolic demands (Yang et al., 2002; Schulze et al., 2022). While some non-rate-limiting enzymes function bidirectionally under physiological conditions, key enzymes catalysing irreversible reactions are often tightly regulated to ensure precise control and fine-tuning of metabolic fluxes. The classical key control point for regulating the anabolic and catabolic directions of carbon flow, i.e. glycolysis and gluconeogenesis, is catalysed by the antagonistic unidirectional enzyme couple phosphofructokinase (PFK) and fructose-1,6-bisphosphatase (FBPase) in bacteria and eukaryotes (Sabnis et al., 1995; Lenzen, 2014). PFK, recognized as a key glycolytic enzyme, catalyses the phosphorylation of fructose 6-phosphate (F6P) to fructose 1,6-bisphosphate (FBP) in the catabolic direction, and FBPase catalyses the reversed anabolic reaction, the dephosphorylation of FBP to F6P. Another important regulatory control point unique to photoautotrophs is the dephosphorylation of sedoheptulose 1,7-bisphosphate (SBP) to sedoheptulose 7-phosphate (S7P), catalysed by sedoheptulose-1,7-biphosphatase (SBPase) (Fig. 1).

Fig. 1.

Reactions of the central carbohydrate metabolism in Synechocystis. Reactions of the CBB cycle are colored in purple, those of the OPP pathway in blue, and those of upper EMP pathway (glycolysis) in red. Many reactions of the CBB cycle overlap with the OPP or EMP pathway. AGP, ADP-glucose pyrophosphorylase; CBB, Calvin–Benson–Bassham; DHAP, dihydroxyacetone phosphate; E4P, erythrose 4-phosphate; EMP, Embden–Meyerhoff–Parnass; ENO, enolase; F6P, fructose 6-phosphate; FBA, fructose-bisphosphate-aldolase; FBP, fructose 1,6-bisphosphate; FBPase, fructose-1,6-bisphosphatase; F/SBPase, fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase; GAP, glyceraldehyde 3-phosphate; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GBP, glucose 1,6-bisphosphate; GND, 6-phosphogluconate dehydrogenase; GP, glycogen phosphorylase; GS, glycogen synthase; HK, hexokinase; OPP, oxidative pentose phosphate; PFK, phosphofructokinase; PGAM, phosphoglycerate mutase; PGI, glucose-6-phosphate isomerase; PGK, phosphoglycerate kinase; PGL, 6-phosphogluconolactonase; PGM, phosphoglucomutase; PRK, phosphoribulokinase; PYK, pyruvate kinase; R-5P, ribose 5-phosphate; RPE, ribulose-5-phosphate epimerase; RPI, ribose-5-phosphate isomerase; TalB, transaldolase; TKT, transketolase; TPI, triosephosphate isomerase; X-5P, xylulose 5-phosphate; ZWF, glucose-6-phosphate dehydrogenase.

The latter reaction of SBPase is unique to the CBB cycle and has a strong control on the flux through the cycle (De Porcellinis et al., 2018). This notion is supported by findings that overexpression of SBPase, e.g. using a bifunctional cyanobacterial enzyme, increased the CO₂ fixation and growth of plants (Miyagawa et al., 2001). No enzyme is known in cyanobacteria or plants that catalyses the corresponding reverse reaction (the phosphorylation of S7P), whereas some clostridia possess a PP_i-PFK that is able to convert S7P to SBP (Koendjbiharie et al., 2020). However, transaldolase (TalB; slr1793), which is mainly attributed to the OPP pathway, is theoretically able to bypass the SBPase reaction in the catabolic direction, by converting glyceraldehyde 3-phosphate (GAP) and S7P to F6P and erythrose 4-phosphate (E4P) (Fig. 1). However, this reaction operates reversibly and is close to equilibrium. It is not well established if transaldolase is restricted to the OPP pathway or might in addition be involved in carbon fixation via the CBB cycle (Fridlyand and Scheibe, 1999). So-called transaldolase variants of the CBB cycle that lack SBPase and use transaldolase instead were reported in autotrophic bacteria such as Thermosulfobium acidiphilum (Antonovsky et al., 2016; Frolov et al., 2019). A similar pathway is likely used by Escherichia coli strains that have been genetically engineered to grow autotrophically (Antonovsky et al., 2016; Frolov et al., 2019).

Unlike plants, the cyanobacterial cell has no defined compartments. Hence all anabolic and catabolic routes of carbon metabolism occur in the cytoplasm, which increases the need of regulatory mechanisms to avoid futile cycles (Lucius and Hagemann, 2024). Although two PFK isoenzymes, an FBPase (slr0952), and a bifunctional fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase (F/SBPase; slr2094) have been identified in the genome of Synechocystis sp. PCC 6803 (hereafter Synechocystis), this glycolytic control point was regarded as being absent in the organism (Knowles and Plaxton, 2003). However, our recent characterization of the Synechocystis PFK-A family isoenzymes, PFK-A1 (sll1196) and PFK-A2 (sll0745), revealed that both specifically utilize ADP as co-substrate instead of ATP and exhibit distinct allosteric regulation (Shen et al., 2024). PFK-A1 is inhibited by 3-phosphoglycerate (3PG), a product of the CBB cycle, while PFK-A2 is inhibited by ATP generated during the photosynthetic light reaction. Physiological experiments with Synechocystis showed diminished levels of polyhydroxybutyrate production in the single deletion mutants Δpfk-A1 and Δpfk-A2, and the double deletion mutant Δpfk-A1Δpfk-A2 (called Δpfk from here on) under nitrogen starvation (Koch et al., 2019). In other experiments, the Δpfk mutant showed growth similar to that of wild type (WT) under photoautotrophic and photomixotrophic conditions including different CO₂ levels, but reduced growth under heterotrophic conditions (Chen et al., 2016; Makowka et al., 2020; Lucius et al., 2021). Growth data for the single Δpfk-A1 and Δpfk-A2 mutants are still lacking.

In contrast to plants, which use separate enzymes for dephosphorylation of either FBP or SBP during the CBB cycle in chloroplasts, Synechocystis employs a bifunctional F/SBPase. During the day, this enzyme is photoactivated by reduced thioredoxin, which breaks disulfide bonds between conserved cysteines in the enzyme. Furthermore, the enzyme is inhibited by AMP (K_i(FBP) 34 µM), ensuring activity in the light at high energy charge of the cell and minimal activity during darkness (Tamoi et al., 1998; Feng et al., 2014; Mallén-Ponce et al., 2021; Sporre et al., 2023). A similar photoactivation mechanism by reduction of cysteine-disulfide bonds is described for plant chloroplast cpFBPase and cpSBPase, but those are insensitive to AMP (Nishizawa and Buchanan, 1981; Ladror et al., 1990; Buchanan, 2016). Deletion of slr2094 encoding the bifunctional F/SBPase in Synechocystis abolished photoautotrophic growth, which could only be restored if both plant enzymes, the cpFBPase and cpSBPase were expressed in the mutant simultaneously (Yan and Xu, 2008; García-Cañas et al., 2022).

Notably, Synechocystis encodes an additional, distinct fructose-1,6-bisphosphatase (fbpase; slr0952). However, unlike F/SBPase, the function of this FBPase protein in Synechocystis remains enigmatic, as the enzyme has not been characterized, and no growth conditions have yet been identified under which FBPase plays a role. Deletion mutants of slr0952 (Δfbpase) consistently exhibited growth comparable to that of the WT in the light (García-Cañas et al., 2022). Furthermore, the expression level of slr0952 (fbpase) is significantly lower than that of slr2094 (f/sbpase) (Jackson et al., 2023). The function of FBPase slr0952 has therefore remained unknown to date.

Plants possess a cytosolic cyFBPase in addition to the chloroplastic cpFBPase. This cytosolic enzyme plays a critical role by catalysing the first irreversible step in sucrose synthesis, which is tightly regulated. Plant cyFBPases lack the conserved cysteine residues and thus do not undergo photoactivation. However, they are inhibited by the carbon regulatory compound fructose 2,6-bisphosphate and by AMP, similar to gluconeogenetic FBPases from mammals and yeast (Stitt et al., 1985; Ladror et al., 1990; Khayat et al., 1993; Zhou and Cheng, 2004). Interestingly, in some plants a second chloroplastic FBPase, which is named cpFBPase II, has been identified, which also lacks the conserved cysteine residues required for photoactivation. Its physiological role remains unclear (Serrato et al., 2009; Li et al., 2020).

The main aim of this study is to clarify the enzymatic regulation and physiological function of FBPase in the cyanobacterium Synechocystis. We furthermore seek to understand its role in a wider biochemical and physiological context including the antagonistic enzyme pairs PFK-A1, PFK-A2 and FBPase, F/SBPase and TalB in regulating the switch between catabolic and anabolic carbon metabolism in cyanobacteria.

Materials and methods

Gene cloning and protein overexpression

The FBPase (slr0952) and F/SBPase (slr2094) encoding genes were amplified from Synechocystis sp. PCC 6803 genomic DNA, using the primer sets slr0952_forward_NdeI_5′ and slr0952_reverse_BamHI_5′ and slr2094_forward_XhoI_5′ and slr2094_reverse_BamHI_5′, respectively (for sequences, see Supplementary Table S1). Subsequently, slr0952 and slr2094 were cloned into the expression vector pET15b with N-terminal 6×histidine-tag (Novogene, Beijing, China). Successful cloning was confirmed by DNA sequencing (LGC genomics, Berlin, Germany). The expression plasmid (pET16b_bmmga3_16125) for fructose-1,6-bisphosphate aldolase from Bacillus methanolicuss (BmFBA) used for synthesis of sedoheptulose 1,7-bisphosphate (SBP) was kindly provided by V. Wendisch (University Bielefeld, Germany) (Stolzenberger et al., 2013b). For expression, the respective plasmids were transformed into the E. coli strain Rosetta (DE3, Agilent Technologies) and overexpression was performed in 1 liter of terrific broth (TB) medium (yeast extract 22 g l⁻¹, tryptone 12 g l⁻¹, and glycerol 4 ml l⁻¹) containing 100 µg ml⁻¹ ampicillin and 30 µg ml⁻¹ chloramphenicol. Cells were cultivated at 37 °C with shaking at 180 rpm (Unitron, INFORS HT, Bottmingen, Switzerland) and protein expression was induced at an OD₆₀₀ of 0.6–0.8 by addition of 1 mM isopropyl-β-D-thiogalactopyranoside. After induction, the cultures were further incubated at 18 °C and 180 rpm for 18–22 h. Cells were collected by centrifugation (15 min, 8630×g, 4 °C) and stored at −70 °C until use.

Protein purification

Both the recombinant Synechocystis FBPase and F/SBPase, as well as BmFBA were purified by immobilized metal ion affinity chromatography and size exclusion chromatography (SEC). For purification, frozen cells from FBPase, F/SBPase, and BmFBA expression were resuspended in 50 mM HEPES–NaOH (pH 7.8, 30 °C), 300 mM NaCl [1 g cells (wet weight)/3 ml buffer]. The cells were disrupted using sonication (3×5 min, amplitude 50, cycle 0.5) (UP200S, Hielscher Ultrasonics, Brandenburg, Germany). Cell debris was removed by centrifugation (45 min, 21 130×g, 4 °C) and the histidine-tagged proteins were purified from the supernatant using nickel (tris(carboxymethyl)ethylenediamin) columns (Ni-TED) (Macherey-Nagel, Dueren, Germany) according to the manufacturers’ instructions. Elution fractions containing the recombinant protein were collected and concentrated using centrifugal concentrators (Vivaspin^®20, Satorius Stedium Biotech, cut off size 30 kDa). Afterwards, the concentrated protein samples of FBPase, F/SBPase, or BmFBA (6, 3.5, and 6 mg, respectively) were applied onto a SEC column (HiLoad 16/600 Superdex 200 prep grade, GE Healthcare) pre-equilibrated with 50 mM HEPES–NaOH (pH 7.8, 30 °C), 300 mM NaCl. Protein fractions for FBPase, F/SBPase, or BmFBA were collected after SEC (5, 3, and 4.5 mg, respectively) and analysed by activity measurements and SDS-PAGE. Proteins were stored at −70 °C in the presence of 25% (v/v) glycerol. The protein concentration was determined using the Bradford assay (Zor and Selinger, 1996) with BSA (Merck, Darmstadt, Germany) as standard.

Determination of the native molecular mass

After purification by metal ion affinity chromatography, 1 mg of FBPase was loaded onto a SEC column (Superose 6 Increase 10/300 GL column, GE Healthcare). As divalent metal requirement is reported for other FBPases (Donahue et al., 2000; Stolzenberger et al., 2013a), we observed an effect of MgCl₂ on the oligomeric structure of FBPase. After 6 h of incubation with 10 mM MgCl₂ in 50 mM HEPES–NaOH (pH 7.8, 30 °C), 300 mM NaCl, the enzyme separation revealed two peaks, representing either a dodecameric or a tetrameric structure. After 48 h incubation, only the tetrameric form was observed. Using the same SEC column, the calibration curve was generated with five proteins [aprotinin (6.5 kDa), ovalbumin (43 kDa), aldolase (158 kDa), ferritin (440 kDa), and thyroglobulin (669 kDa)] and blue dextran from the LMW and HMW gel filtration calibration kits (GE Healthcare), using the same buffer as for the purification of recombinant FBPase. The native molecular mass of FBPase was calculated using the generated calibration curve.

In vitro fructose-1,6-bisphosphatase activity measurements

The FBPase activity of FBPase and F/SBPase was determined using a continuous, coupled enzymatic assay to monitor the conversion of FBP to F6P. This reaction was facilitated by two auxiliary enzymes: PGI (from Saccharomyces cerevisiae; Merck, Darmstadt, Germany), which converts the formed F6P to G6P, and glucose-6-phosphate dehydrogenase (rabbit muscle; from Saccharomyces cerevisiae; Merck, Darmstadt, Germany), which oxidizes G6P to 6-phosphogluconate while reducing NADP⁺ to NADPH. The increase in NADPH was detected by measuring the absorbance at 340 nm in 96-well plates (BRANDplates^®, BRAND, Wertheim, Germany). A NADPH calibration curve (0–0.7 mM NADPH) was used for quantification, and measurements were conducted using a Tecan Infinite M200 plate reader (Tecan Group AG, Männedorf, Switzerland) at 30 °C.

The assay mixture (200 µl total volume) contained 1 μg FBPase or 1.6 μg F/SBPase, 50 mM HEPES–NaOH (pH 7.8 at 30 °C), 1 U PGI, 1 U glucose-6-phosphate dehydrogenase, 10 mM MgCl₂, 5 mM NADP⁺, and 3 mM FBP. For F/SBPase 10 mM dithiothreitol (DTT) was included in the assay. The reaction was started by adding the enzyme.

To determine the metal ion dependency, various divalent metal ions were tested at concentrations of 1 and 5 mM using the standard assay. For the best performing metal ions, the optimal concentration was determined, for MgCl₂ within a range of 1–20 mM and for MnCl₂ in a narrower range of 0.01–5 mM. The enzyme characterization was performed in the presence of 10 mM MgCl₂. To calculate V_max and K_m values, the Michaelis–Menten equation was fit using the NonlinearModelFit function in Wolfram Mathematica v14.

Effector studies were performed using the standard assay in the presence of 1 mM fructose 1-phosphate, phosphoenolpyruvate, citrate, isocitrate, malate, 2-phosphoglycolate, ATP, ADP, AMP, and DTT with a half-saturation concentration of FBP (0.2 mM). A more detailed characterization was performed for ATP, AMP, and DTT, using 1–10 mM of the effectors. Additionally, for ATP, the experiment was repeated with 30 mM MgCl₂.

All assays were performed in triplicate. Negative controls were conducted by omitting either enzyme or FBP. One unit (1 U) of enzyme activity is defined as 1 µmol of NADP⁺ being reduced to NADPH via the auxiliary enzyme glucose-6-phosphate dehydrogenase per minute.

In vitro sedoheptulose-1,7-biphosphatase activity measurements

Since SBP was not commercially available, BmFBA was employed for synthesis. BmFBA catalyses the reversible conversion of FBP to GAP and dihydroxyacetone phosphate (DHAP), as well as the reversible condensation of E4P and DHAP to SBP (Stolzenberger et al., 2013b). The activity of BmFBA was confirmed in cleavage direction using FBP as the substrate. The reaction mixture (200 µl total volume) consisted of 2 μg BmFBA, 50 mM HEPES–NaOH (pH 7.8 at 30 °C), 0.7 mM NADH, 5 mM FBP and 1 U glycerol-3-phosphate dehydrogenase (from rabbit muscle; Merck). In the cleavage reaction, DHAP, one of the products, was converted to glycerol 3-phosphate resulting in the oxidation of NADH to NAD⁺. The oxidation of NADH was monitored as a decrease in absorbance at 340 nm in 96-well plates (BRANDplates). A NADPH calibration curve (0–0.7 mM NADPH) was used for quantification, and measurements were conducted using a Tecan Infinite M200 plate reader at 30 °C. A specific activity of 4 Umg⁻¹ was determined for BmFBA using this assay.

To analyse the SBPase activity of FBPase, three different assays were performed. Each reaction contained 3 mM E4P, 3 mM DHAP, and 50 mM HEPES–NaOH (pH 7.8 at 30 °C). The specific compositions of the reactions were as follows: reaction A included 10 μg BmFBA; reaction B 10 μg BmFBA, 10 mM MgCl₂, and 5 μg FBPase; and reaction C 10 μg BmFBA, 10 mM MgCl₂, 10 mM DTT, and 5 μg F/SBPase. The reactions were incubated at 30 °C for 2 h. After incubation each reaction was diluted 30-fold to achieve a final concentration of 0.1 mM of substrates/products. From the diluted mixtures, 0, 10, 20, 30, 40, and 50 μl of the reaction were mixed with 30 μl of malachite green phosphate reagent (Phosphate calorimetric assay kit; Merck, Darmstadt, Germany) and H₂O was added to a final volume of 200 μl. Absorbance at 650 nm in 96-well plates (BRANDplates) was measured using a Tecan Infinite M200 plate reader.

To exclude potential inhibitory effects of DHAP and E4P on Synechocystis FBPase activity, appropriate control assays were conducted. For DHAP, a continuous standard FBPase assay was performed with increasing concentrations of DHAP (0–5 mM) using a half-saturating concentration of FBP (0.2 mM). In the case of E4P, due to interference with the continuous assay, a discontinuous malachite green assay was employed (see above). The reaction mixture contained 3 mM FBP, 10 mM MgCl₂, 2 μg FBPase, and 50 mM HEPES–NaOH (pH 7.8 at 30 °C), with E4P added at concentrations ranging from 0 to 5 mM. The specific activity (U mg⁻¹) was determined using a PO₄³⁻ calibration curve ranging from 0 to 5 nmol per well.

³¹P-Nuclear magnetic resonance analysis of fructose-1,6-bisphosphatase and fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase

For qualitative analysis, the enzyme reactions of FBPase and F/SBPase were analysed via ³¹P-nuclear magnetic resonance (NMR) spectroscopy. Control samples, including 3 mM FBP, 3 mM phosphate (PO₄³⁻), or 3 mM F6P, were prepared in 50 mM HEPES–NaOH (pH 7.8, 30 °C), 300 mM NaCl, and 10 mM MgCl₂. The enzyme assays were performed in the same buffer with 3 mM FBP, and 5 µg FBPase or F/SBPase. For F/SBPase, the reaction mixture also included 10 mM DTT. All reaction mixtures were incubated at 30 °C for 2 h. The ³¹P-NMR spectrum was obtained using a Bruker Avance Neo 400 device, and data visualization was performed using TopSpin 3.7.0 software.

Bioinformatic analysis

Crystal structures were retrieved from the Protein Data Bank (PDB) and structural models from the AlphaFold Server version 3.0 (Abramson et al., 2024). Molecular docking to infer the SBP binding site in SBPase was performed using SwissDock (Grosdidier et al., 2011). Structural analyses, comparative assessments, and visualizations were conducted using the UCSF ChimeraX software suite, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (Pettersen et al., 2021). Amino acid sequences for sequence alignment were retrieved from the UniProt database, and alignments were generated using the EMBL-EBI Job Dispatcher framework for sequence analysis tools (Madeira et al., 2024). The BioEdit software suite was used for manual refinement and editing of sequence alignments. Protein alignments of selected FBPase and SBPase proteins from heterotrophic bacteria, cyanobacteria, green algae, and streptophytes were used for phylogenetic analyses using the software package MEGA. The protein sequence of GlpX from Corynebacterium glutamicum served as outgroup. GenBank accession numbers of the used sequences are given in Supplementary Table S2.

Generation and verification of deletion and complementation mutants of Synechocystis

To generate deletion mutants, Synechocystis was transformed with constructs in which a resistance cassette was fused to around 200 bp directly up- and downstream of the gene to be deleted. The fragments for these constructs were amplified by PCR and subsequently cloned into pBluescript vector by Gibson assembly (Gibson et al., 2009); alternatively, a pUC vector already containing the up- and downstream regions was ordered from Genewiz (Genewiz Germany GmbH, Leipzig, Germany), and opened with EcoRV between the up- and downstream regions, and the resistance cassette amplified by PCR was inserted by Gibson assembly.

Complementation mutants were constructed to complement Δfbpase. This was achieved by introducing into Δfbpase a plasmid containing the fbpase gene under the control of a rhamnose induced promoter, as well as spectinomycin resistance (Δfbpase::pSSR_fbpase). As a negative control, we additionally generated a mutant that was complemented with the same vector containing the mVenus gene instead of fbpase (Δfbpase::pSSRV).

For testing the segregation of the mutants, PCRs were performed (Supplementary Protocol S1; Supplementary Figs S1–S6).

All mutants that were used are listed in Supplementary Table S3, along with the primers that were used for their generation and for testing their segregation. Some of them had already been generated previously. All primer sequences are listed in Supplementary Table S1.

Growth conditions for Synechocystis

Strains were maintained on BG11 plates at 28 °C under constant illumination (50–100 µmol m⁻² s⁻¹), mutants with antibiotics.

For growth experiments, Synechocystis sp. PCC 6803 WT and mutants were grown in 200 ml BG11 medium (pH 8) at 28 °C in glass tubes and equally aerated with filter-sterilized air, as described previously (Makowka et al., 2020). Cultures were inoculated at an OD₇₅₀ of around 0.1 to 0.15 and illuminated from two sides with constant light of 50 µmol m⁻² s⁻¹). To monitor growth, OD₇₅₀ was measured every day, using a 96-well plate and a plate reader (Infinite M Nano+, Tecan Group AG). In each experiment, three biological replicates were used and independent experiments for every analysed condition were conducted at least three times. Pre-cultures for growth experiments were grown autotrophically, first in baffled shake flasks for 1 d (in 50 ml BG11), then in glass tubes for further 6 d (diluted with 50 ml BG11). For mutants, antibiotics were used in the pre-cultures, but not in the growth experiments, unless noted otherwise. Cultures were illuminated from two sides with constant light of 50 µmol m⁻² s⁻¹); only heterotrophic cultures were kept in darkness, except for daily illumination of around 10 min. For photomixotrophic, photoheterotrophic, and heterotrophic cultures, 10 mM glucose was added. For photoheterotrophic cultures, 40 µM 3-(3,4-dichlorophenyl)-1,1-dimethylurea was added additionally.

The following antibiotic concentrations were used: kanamycin 50 µg ml⁻¹, spectinomycin 20 µg ml⁻¹, gentamycin 10 µg ml⁻¹ in agar plates, 2.5–10 µg ml⁻¹ in liquid culture.

Determination of fructose-1,6-bisphosphatase activity in crude cell extracts of Synechocystis

The method for measuring FBPase activity in crude cell extracts of Synechocystis was based on (Yan and Xu, 2008), but some adaptations were made. Like in the in vitro measurements, the substrate of FBPase (FBP) was supplied and PGI and glucose-6-phosphate dehydrogenase (ZWF) were used as auxiliary enzymes to convert its product, fructose 6-phosphate, to 6-phosphogluconate. The NADPH generated in the last step was quantified by measuring the absorbance at 340 nm.

Cells from 50 ml liquid culture were harvested by centrifugation (1200×g, 10 min, 4 °C), the pellet was washed in potassium phosphate buffer once (50 mM, pH 8.0; 6000×g, 10 min, 4 °C) and the pellet was resuspended in 2 ml extraction buffer (50 mM potassium phosphate pH 8.0, 2.5 mM DTT, 1 mM glutathione, 10% v/v sucrose). The suspension was distributed into two reaction tubes containing around 400 µl glass beads (0.17–0.18 mm) and vortexed (3 min, 4 °C). Two centrifugation steps followed (maximum speed, 1 min, 4 °C; then 16 000×g, 20 min, 4 °C). The supernatant (crude extract) was diluted with extraction buffer to the equal phycobilisome concentration for all samples, determined by measuring the absorbance at 650 nm. Crude extract of 20 µl was mixed with 160 µl enzyme mix (100 mM Tris–HCl pH 8.0, 10 mM MgCl₂, 0.5 mM EDTA, 1.5 mM NADP⁺, 0.5 U ml⁻¹ ZWF, 1.5 U ml⁻¹ PGI, 10% v/v glycerol). The baseline absorption at 340 nm was measured over 10 min in a 96-well plate with a plate reader (Infinite M Nano+). Substrate mix of 20 µl was added (100 mM Tris–HCl pH 8.0, 10 mM MgCl₂, 0.5 mM EDTA, 20 mM FBP), and after shaking, the NADPH production was quantified by measuring the absorbance at 340 nm over 30 min. FBPase enzyme activity was quantified with a calibration curve of NADPH (0–0.16 µM).

For testing reducing conditions, the measurements were conducted as described; for oxidizing conditions, extraction buffer without DTT was used. We verified that the presence or absence of DTT did not influence the auxiliary enzymes and that extraction without DTT did indeed represent reducing conditions (Supplementary Protocol S2; Supplementary Fig. S7). Furthermore, we observed that addition of DTT did not increase the activity after extracting under oxidizing conditions, which indicates that the reduction for activating F/SBPase requires a cellular mediator (Supplementary Protocol S2; Supplementary Fig. S8).

To test for significant differences, a one-way analysis of variance (ANOVA) was performed for the two influence factors strain and extraction method, for the whole dataset and also stratified on the other influence factor. If the factor strain was significant, post-hoc comparisons were performed according to the Tukey methods, which means that these are adjusted for multiple testing. The significance level for all tests was 0.05 and all tests were two-sided. All analyses were performed with the statistical software R, version 4.4.0 (R Core Team, 2024).

Determination of the cellular glycogen content in Synechocystis

The glycogen content in Synechocystis cells was measured as described by Makowka et al. (2020) with some modifications. Glycogen was extracted and digested into glucose, which was subsequently converted to 6-phosphogluconate by hexokinase and ZWF. The NADPH generated in the last step was quantified by measuring the absorbance at 340 nm.

Cells from 10 ml liquid culture were harvested by centrifugation (8000–15 000×g, 10 min, 4 °C), the pellet was resuspended in 300 µl 30% (w/v) KOH, and incubated for 2 h at 95 °C for cell lysis. To precipitate the glycogen 900 µl 100% ethanol was added and the sample was incubated at −20 °C overnight. After subsequent centrifugation (10 000×g, 10 min, 4 °C) the pellet was washed first with 1 ml 100% ethanol (10 000×g, 5 min, 4 °C), then with 1 ml 70% ethanol (10 000×g, 5 min, 4 °C), then dried at 50 °C for 20–30 min. To digest the glycogen 500 µl digestion buffer (100 mM sodium acetate pH 4.5, 1.5 mg ml⁻¹ amyloglucosidase, 10% v/v glycerol) was added to the pellet, and the sample was vortexed and incubated for 90 min at 60 °C. After centrifugation (10 000×g, 10 min, room temperature), 50 µl of the supernatant was transferred into a new reaction cup and 940 µl NADP buffer (100 mM Tris–HCl pH 7.4, 3.189 mM MgCl₂, 2.127 mM ATP, 2.127 mM Na₂NADP) was added. If necessary, the supernatant was diluted and 50 µl of the dilution was used instead. Glucose standards (0–10 mM) were prepared, and 50 µl was mixed with 940 µl NADP buffer. Samples of 198 µl and standard were transferred onto a 96-well plate (four replicates) and the baseline absorption at 340 nm was measured with a plate reader (Infinite M Nano+). Two microliters of enzyme solution (100 mM Tris–HCl pH 7.4, 17 U ml⁻¹ hexokinase, 8.5 U ml⁻¹ ZWF, 10% v/v glycerol) was added to each standard and sample and mixed by shaking in the plate reader. After incubation at 37 °C for 80 min the endpoint absorption at 340 nm was measured. To quantify the cellular glycogen content, the increase in absorption was calculated and converted into glucose concentration with the calibration curve generated from the values of the glucose standard. The glucose concentration (mM) was converted into the glycogen mass (µg) contained in the sample by multiplying with the lysate volume (0.5 ml), the molar mass of glucose (180.16 g mol⁻¹), and the molar mass ratio of glycogen to glucose (162.16/180.16). Finally, the glycogen mass was converted into the specific cellular glycogen content (µg OD₇₅₀⁻¹ ml⁻¹) by dividing by the OD₇₅₀ and the volume (10 ml) of the sample.

Measurement of the glucose concentration in Synechocystis cultures

Glucose concentration in Synechocystis cultures was measured by the same principle as the cellular glycogen content, namely by using the absorbance at 340 nm to quantify the NADPH production upon conversion of glucose to 6-phosphogluconate by hexokinase and ZWF.

Ten milliliters of liquid culture was centrifuged (8000–15 000×g, 10 min, 4 °C); 50 µl of the supernatant was transferred to a new reaction cup and mixed with 940 µl NADP buffer (100 mM Tris–HCl pH 7.4, 3.189 mM MgCl₂, 2.127 mM ATP, 2.127 mM Na₂NADP). Glucose standards (0–10 mM) were prepared and 50 µl was mixed with 940 µl NADP buffer. As for the glycogen measurement, 198 µl of the samples and standard were transferred onto a 96-well plate (four replicates) and the baseline absorption at 340 nm was measured with a plate reader (Infinite M Nano+). Enzyme solution of 2 µl (100 mM Tris–HCl pH 7.4, 17 U ml⁻¹ hexokinase, 8.5 U ml⁻¹ ZWF, 10% v/v glycerol) was added to each standard and sample and mixed by shaking in the plate reader. After incubation at 37 °C for 80 min the endpoint absorption at 340 nm was measured. The increase in absorption was calculated and the glucose concentration of the samples was determined with the calibration curve generated from the values of the standard.

Results

Overexpression and purification of the recombinant fructose-1,6-bisphosphatase

For the biochemical characterization of the FBPase from Synechocystis, the encoding gene (slr0952, fbpase) was expressed in E. coli and subsequently purified (Supplementary Fig. S9). The molecular mass of FBPase under denaturing conditions in SDS-PAGE was approximately 40 kDa, which aligns well with the calculated molecular mass of 39 kDa. The native molecular mass after preincubation with 10 mM MgCl₂, determined by SEC, showed a single peak at 163 kDa, suggesting a homotetrameric structure for FBPase (Supplementary Fig. S10). This observation is consistent with previous studies reporting that class I FBPases typically exhibit a homotetrameric structure (Hines et al., 2006).

Biochemical characterization of fructose-1,6-bisphosphatase

Like most characterized FBPases (Stieglitz et al., 2003; Brown et al., 2009), Synechocystis FBPase is a metal-dependent enzyme that requires divalent metal ions for activity. In the absence of divalent metal ions, the enzyme showed no activity with FBP as substrate (Supplementary Fig. S11A). Significant FBPase activity was observed only in the presence of Mg²⁺. Mn²⁺ and Co²⁺ could partially activate the enzyme, though to a lesser extent. This metal response differs from previously characterized bifunctional F/SBPases, such as the one from Synechocystis, which can utilize both Mg²⁺ and lower concentrations of Mn²⁺ to shift from an inactive homodimeric form to an active tetrameric form (Hines et al., 2006; Feng et al., 2014). In contrast, Synechocystis FBPase specifically requires Mg²⁺. The inclusion of 10 mM Mg²⁺ in the reaction buffer resulted in the highest FBPase activity, reaching approximately 12 U mg⁻¹ protein (Supplementary Fig. S11B). To determine kinetic parameters, FBPase was incubated with varying FBP concentrations. The enzyme follows a classical Michaelis–Menten kinetics with a V_max of 12.7±0.19 U mg⁻¹ protein and a K_m of 0.18±0.01 mM (Fig. 2). In comparison with F/SBPase, FBPase has a 2-fold lower affinity towards FBP (K_m) and catalyses its dephosphorylation with a slightly lower turnover number (K_cat), resulting in lower catalytic efficiency (K_cat/K_m) values (Table 1).

Fig. 2.

Enzymatic characterization of FBPase. The enzymatic activity of FBPase was assessed with varying concentrations of FBP (0–3 mM) in the presence of 10 mM MgCl₂. Data are presented as means ±SD from three independent measurements (technical replicates) of two biological replicates (n=6). The shaded region shows the 95% confidence interval. To determine V_max (12.7 U mg⁻¹) and K_m (0.18 mM) values, the Michaelis–Menten equation was fit using the NonlinearModelFit function in Wolfram Mathematica v14.2.

Table 1.

Kinetic parameters of the Synechocystis F/SBPase and FBPase

Parameter	F/SBPase (Feng et al., 2014)	FBPase (this work)
K cat (FBP) (s−1)	10.5	8.4
K cat (SBP) (s−1)	4.2	NA
K cat/Km (FBP) (s−1 mM−1)	131.25	46.7
V max (FBP) (U mg−1)	16.1a	12.7
K m (FBP) (mM)	0.08	0.18
K m (SBP) (mM)	0.24	NA
Activators	DTT	None
Inhibitors	AMPb	None

^a Calculated using the given K_cat value in Feng et al. (2014).

^b K _i(FBP)=34 µM.

DTT, dithiothreitol; FBP, fructose 1,6-bisphosphate; FBPase, fructose-1,6-bisphosphatase; F/SBPase, fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase; NA, not active; SBP, sedoheptulose 1,7-bisphosphate.

Previous studies on class I FBPases have demonstrated that several metabolic intermediates such as fructose 1-phosphate, malate, citrate, isocitrate, and AMP can inhibit the enzyme (Babul and Guixé, 1983; Van Praag, 1997; Donahue et al., 2000; Wolf et al., 2018). The intriguing regulation of the Synechocystis F/SBPase, which is activated by DTT and inhibited by AMP, has revealed its relevance for the CBB cycle under light conditions (Feng et al., 2014), and serves as a foundation for investigating the biochemical regulation of Synechocystis FBPase. In this study, we tested several effectors under half-saturating conditions for FBP. None of the tested effectors caused a significant effect, except for a slight reduction in activity by ATP, DTT, and AMP (93, 91, and 94% residual activity, respectively, in the presence of 1 mM of these effectors) (Fig. 3). To further explore these inhibitory effects, we varied the concentration of ATP, DTT, and AMP under half-saturating FBP conditions. ATP was the only effector that caused significant inhibition, with a K_i of 5.3 mM (Supplementary Fig. S12). However, further studies with increased Mg²⁺ concentrations (30 mM) revealed that the inhibitory effect of ATP was due to competition for Mg²⁺ required for enzyme activity. Hence, ATP inhibition of the Synechocystis FBPase is probably not relevant under in vivo conditions. In summary, none of the tested effectors had a regulatory effect on FBPase. Notably, AMP and DTT, which are known to affect the activity of Synechocystis F/SBPase, did not impact the activity of FBPase, even at concentrations up to 10 mM.

Fig. 3.

Influence of different metabolites and effectors on FBPase activity. The effects of various metabolites and effectors (1 mM) on FBPase activity was assessed under subsaturating (0.2 mM FBP) conditions. Relative activity (%) is expressed as a percentage of the control activity measured in the absence of effector (100%). None of the metabolites has a relevant effect on FBPase activity. The specific activity of the control under subsaturating conditions was 6 U mg⁻¹ protein. Data represent the means ±SD from three technical replicates (n=3). 2PG, 2-phosphoglycolate; DTT, dithiothreitol; F1P, fructose 1-phosphate; FBP, fructose 1,6-bisphosphate; FBPase, fructose-1,6-bisphosphatase; PEP, phosphoenolpyruvate.

Monofunctional fructose-1,6-bisphosphatase lacks sedoheptulose-1,7-biphosphatase activity

To test if the Synechocystis FBPase is mono- or bifunctional, we compared its possible SBPase activity with the bifunctional F/SBPase as positive control. First, the FBPase activity of recombinant F/SBPase was confirmed in the presence of 10 mM Mg²⁺ and 10 mM DTT. Consistent with previous reports (Feng et al., 2014), its activity was almost completely inhibited by 50 µM AMP (Supplementary Fig. S13). Additionally, ³¹P-NMR analysis clearly verified that both, FBPase and F/SBPase are able to irreversibly cleave FBP to F6P and PO₄³⁻ (Supplementary Fig. S14).

To test the SBPase activity of the enzymes, the substrate SBP was produced by recombinant fructose-1,6-bisphosphate aldolase from Bacillus methanolicus (BmFBA), as SBP was not commercially available. BmFBA catalyses the reversible condensation of E4P and DHAP to SBP (Stolzenberger et al., 2013b). Only in the presence of the Synechocystis F/SBPase did we observe release of PO₄³⁻ using a malachite green assay (Fig. 4A), whereas the FBPase did not dephosphorylate SBP. Additionally, we verified that DHAP consumption, indicative of SBPase activity, only occurred in the presence of both BmFBA and F/SBPase, but neither with F/SBPase alone nor with BmFBA and FBPase from Synechocystis (Fig. 4B). To assess potential inhibition of FBPase by DHAP and E4P, enzyme activity was measured in the presence of increasing concentrations of each compound. DHAP had no inhibitory effect on FBPase activity at concentrations up to 5 mM. In contrast, E4P (used at 3 mM as the starting concentration for SBP synthesis) caused a moderate inhibitory effect, with FBPase retaining 65% of its activity at this concentration (Supplementary Fig. S15). Although this slight inhibition was observed at the initial E4P concentration, the gradual decrease in E4P levels during SBP formation suggests that its impact on FBPase activity under assay conditions is minimal. These results support the conclusion that the FBPase is monofunctional and lacks SBPase activity.

Fig. 4.

FBPase (slr0952) possesses no SBPase activity. (A) SBP was synthesized through the reversible condensation of 3 mM E4P and DHAP catalysed by BmFBA. The cleavage of SBP to S7P and PO₄³⁻ by FBPase was assessed spectrophotometrically using a malachite green assay. The bifunctional F/SBPase (slr2094) served as positive control. (B) As an alternative to PO₄³⁻ release, the consumption of DHAP was measured after 2 h. DHAP levels were quantified via NADH oxidation upon the addition of 5 U glycerol-3-phosphate dehydrogenase. Significant DHAP depletion, corresponding to PO₄³⁻ and S7P formation, was observed only in the presence of both BmFBA and F/SBPase, but not in the presence of FBPase. Data represent the means ±SD from three technical replicates (n=3). BmFBA, fructose-1,6-bisphosphate aldolase from Bacillus methanolicuss; DHAP, dihydroxyacetone phosphate; E4P, E4P, erythrose 4-phosphate; FBPase, fructose-1,6-bisphosphatase; F/SBPase, fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase; S7P, sedoheptulose 7-phosphate; SBP, sedoheptulose 1,7-bisphosphate.

To investigate substrate specificity and possible regulatory effects, we tested glucose 1,6-bisphosphate (GBP). However, it was not utilized as a substrate by either enzyme nor did it act as an effector on FBPase or F/SBPase activity (Supplementary Fig. S16).

Structural comparison of Synechocystis fructose-1,6-bisphosphatase and fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase

FBPases are members of the large superfamily of lithium-sensitive phosphatases. They are metal-dependent enzymes that are classified into five distinct groups based on their amino acid sequences (Brown et al., 2009). Sequence alignments (Supplementary Fig. S17) and comparison of the AlphaFold structural model of the Synechocystis FBPase with the crystal structure of the well-characterized class I FBPase from E. coli (P0A993) revealed that the Synechocystis FBPase adopts a typical class I FBPase fold. This enzyme exhibits a high degree of sequence similarity to class I FBPases, retaining conserved residues at both the metal-binding and substrate-binding sites (Fig. 5A–C). In contrast, the conserved AMP-binding site present in E. coli class I FBPase [TGELT motif, highlighted in the structural comparison (Fig. 5) and in the sequence alignments (Supplementary Fig. S17)] is absent in the Synechocystis FBPase. Consistent with this observation, no inhibitory effect of AMP was detected at concentrations up to 10 mM (Supplementary Fig. S12). Also, the class I cpFBPases from Arabidopsis and Fragaria×ananassa exhibit a similar fold to the Synechocystis FBPase, except for an additional 20–30 amino acid sequence in the regulatory domain of the plastidial enzymes. This region, referred to as the ‘loop 170,’ contains three cysteine residues, two of which form a disulfide bridge that is reduced by thioredoxin f during activation (Chiadmi et al., 1999; Fig. 5D–F; Supplementary Fig. S8). Consistent with these structural differences, no regulatory effect of DTT was observed on the Synechocystis FBPase (Supplementary Fig. S12).

Fig. 5.

Structural comparison of the Synechocystis FBPase with class I FBPases from E. coli and Arabidopsis. (A–C) A ribbon representation is shown for the monomer of Synechocystis class I FBPase (slr0952, yellow, Alpha fold model) (A) in comparison with the crystal structure of the class I FBPase from E. coli (2q8 m, gold; Hines et al., 2007b) (B), as well as their superimposition (C). (D–F) The AlphaFold models from Synechocystis class I FBPase (yellow) (D) and Arabidopsis class I chloroplastic FBPase (cpFBPase, P25851, grey) (E) are compared and their superimposition is shown (F). Substrate, metal, and AMP binding sites are highlighted in red, blue, and green, respectively. Regulatory cysteines are marked in yellow. The figure was created using AlphaFold server 3 and Chimera X. FBPase, fructose-1,6-bisphosphatase.

To further investigate the basis of substrate specificity between class I FBPases and SBPases, multiple sequence alignments and structural comparisons including substrate docking analyses were performed (Supplementary Figs S17, S18). These analyses allowed us to identify residues potentially critical for substrate recognition. Specifically, in class I FBPase, including the Synechocystis enzyme, we identified two residues, asparagine (N223) and tyrosine (Y271), which are highly conserved in FBPases and whose side chains bind the phosphate moiety at C6 of FBP and sterically hinder binding of the larger SBP molecule. In contrast, these residues are absent in the Physcomitrium patens (PDB: 5IZ3) and other SBPases, resulting in a more spacious substrate-binding pocket that can accommodate the seven-carbon SBP, as previously reported (Gütle et al., 2016). Our structural comparisons complement and extend these findings by identifying specific sequence differences likely responsible for substrate discrimination (for further details see Supplementary Protocol S3).

As previously reported, the class II F/SBPase from Synechocystis adopts a similar fold that is distinct from that of class I FBPases and instead resembles the structure of the class II FBPase from E. coli (Supplementary Fig. S19A–C) (Feng et al., 2014). Unlike class I FBPases, which possess a different overall architecture, class II enzymes are characterized by the presence of a specialized regulatory domain. This domain includes AMP-binding sites and two conserved cysteine residues that participate in thioredoxin-mediated redox regulation (Feng et al., 2014). Notably, these regulatory cysteines are absent in class II FBPases from non-photosynthetic bacteria such as E. coli (Brown et al., 2009), highlighting functional divergence within the class II enzyme family.

Phylogenetic relation of Synechocystis fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase and fructose-1,6-bisphosphatase to plant enzymes

We performed phylogenetic analyses to investigate the relation of Synechocystis class II F/SBPase and class I FBPase to enzymes from other organisms, in particular to plant cytoplasmic FBPase (cyFBPase), the two chloroplastic FBPases (cpFBPase, cpFBPase II), and the chloroplastic SBPase (cpSBPase) (Hines et al., 2007a, b; Gütle et al., 2016). Sequences from green algae and proteobacteria were also included. The Synechocystis class I FBPase clusters together with FBPases from other cyanobacteria and in close proximity to proteobacterial FBPases as well as all plant enzymes (cyFBPase, cpFBPase, cpFBPase II, and cpSBPase), which each form a distinct cluster (Fig. 6). In contrast, Synechocystis class II F/SBPase is more distantly related to the plant enzymes. It clusters with bifunctional enzymes from other cyanobacteria near the outgroup and SBPases from green algae. Overall, the tree topology indicates that (cyano)bacterial monofunctional FBPases, rather than bifunctional F/SBPases, are the origin of all plant FBPases and SBPases.

Fig. 6.

Phylogenetic tree of class I FBPases and SBPases and class II F/SBPases from photoautotrophic and heterotrophic organisms. Enzymes from cyanobacteria are colored in blue (Synechocystis in bold), from proteobacteria in red, from chlorophyta (green algae) in light green, and from streptophyta (plants) in dark green. As outgroup, the multifunctional sugar phosphatase (GlpX) from Corynebacterium was used. The tree was generated in MEGA with a maximum-likelihood analysis of aligned and trimmed sequences. Accession numbers of all sequences used in the phylogenetic cluster analysis can be found in the Supplementary Table S2. FBPase, fructose-1,6-bisphosphatase; F/SBPase, fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase; SBPase, sedoheptulose-1,7-biphosphatase.

In vivo fructose-1,6-bisphosphatase activity in deletion mutants

To analyse the in vivo function of FBPase in a cellular context, a collection of deletion mutants was created. In contrast to other groups (Yan and Xu, 2008; García-Cañas et al., 2022), our attempts to segregate the slr2094 deletion (Δf/sbpase) were not successful, similar to earlier studies that reported the deletion of F/SBPase to be lethal (Tamoi et al., 1999). However, for slr0952 (Δfbpase) complete deletion was successful (Supplementary Figs S1–S3). In the following, a merodiploid f/sbpase/Δf/sbpase and a fully segregated Δfbpase mutant were characterized.

First, we analysed the in vivo contribution of the two enzymes to total FBPase activity in crude cell extracts by performing an enzyme activity assay with FBP as substrate. As shown above, the bifunctional F/SBPase in contrast to FBPase has to be activated by reducing conditions. Therefore, cells were disrupted under either reducing or oxidizing conditions, enabling us to distinguish the activities of the two enzymes in the different strains.

For all strains, FBPase activity under reducing conditions was much higher than under oxidizing conditions (P<0.05 for each strain) (Fig. 7). This can be explained by the specific activation of F/SBPase by reduction, which is expressed at a much higher level than FBPase (Saha et al., 2016). Under reducing conditions FBPase activity was similar in WT, f/sbpase/Δf/sbpase, and Δfbpase. The fact that we could not detect a difference in total cellular FBPase activity between the WT and the Δfbpase mutant indicates that the activity of FBPase is negligible compared with that of F/SBPase under those conditions. The WT-like FBPase activity of the merodiploid f/sbpase/Δf/sbpase mutant can likely be attributed to an up-regulation of the expression of the remaining copies of f/sbpase. An up-regulation of fbpase expression cannot explain this observation, because also under oxidizing conditions the f/sbpase/Δf/sbpase mutant showed the same activity as the WT, and FBPase is not activated by reduction. In previous studies FBPase activity was virtually lacking in a fully segregated f/sbpase deletion mutant, which led to the conclusion that F/SBPase might be the only active FBPase in vivo (Yan and Xu, 2008). Interestingly, we observed that deletion of fbpase resulted in a significant reduction of FBPase activity by around two-thirds under oxidizing conditions (P<0.001, Fig. 7). This suggests that, unlike what was previously assumed, the monofunctional FBPase is indeed active under oxidizing conditions in vivo, albeit at low rates. The remaining activity in Δfbpase under oxidizing conditions can likely be attributed to residual activity of F/SBPase due to incomplete inactivation.

Fig. 7.

Cellular FBPase activity in deletion mutants. FBPase activity was measured in crude cell extracts of WT, the merodiploid f/sbpase/Δf/sbpase mutant and Δfbpase that were obtained under either oxidizing (Ox.) or reducing (Red.) conditions and normalized to the same phycobilisome concentration. Mean, maximum, and minimum values of two (WT) or three (f/sbpase/Δf/sbpase and Δfbpase) biological replicates are depicted. FBPase, fructose-1,6-bisphosphatase; WT, wild type.

Growth of f/sbpase, fbpase, and talB deletion mutants

To further investigate the physiological role of the two F(/S)BPases, we investigated the growth of the fbpase and merodiploid f/sbpase deletion mutants under different conditions. Additionally, we included deletion mutants of talB (encoding the OPP pathway-specific transaldolase) to better understand its relation to FBPase, F/SBPase, and fructose-1,6-bisphosphate aldolases (FBA). In the OPP pathway, TalB basically reverts the CBB cycle reactions of SBPase and FBA. However, the substrates and products are not identical between the two directions, as TalB converts GAP and S7P to F6P and E4P, while the combined activity of SBPase and FBA converts E4P and DHAP to S7P (see Fig. 1). Since the reaction catalysed by TalB is close to thermodynamic equilibrium, it is theoretically conceivable that it operates in CBB cycle direction. In fact, there are organisms with a CBB cycle that lack SBPase and in which SBPase is functionally replaced by a transaldolase (Frolov et al., 2019). We were therefore interested to investigate if TalB is confined to the OPP pathway in Synechocystis or might also be able to participate in the CBB cycle in vivo.

Growth of the merodiploid f/sbpase/Δf/sbpase mutant was not affected under any of the light conditions we tested, which is well in line with the interpretation that this mutant up-regulated F/SBPase activity of the bifunctional enzyme to WT levels (compare Figs 7, 8A–D). Similarly, the Δfbpase mutant grew like the WT under photoautotrophic and photomixotrophic conditions in constant light (Fig. 8A, C). These results are consistent with previous studies, which showed that the absence of FBPase affected neither photoautotrophic, nor photoheterotrophic, nor photomixotrophic growth, whereas F/SBPase was essential under photoautotrophic conditions (García-Cañas et al., 2022). Furthermore, in diurnal day/night cycles photoautotrophic and photomixotrophic growth was not affected in Δfbpase (Fig. 8B, D). However, under heterotrophic conditions (darkness with glucose supplied), we observed differences. The mutant f/sbpase/Δf/sbpase reached higher optical densities than the WT, whereas surprisingly growth of Δfbpase was diminished (Fig. 8E). These results indicate that FBPase, in contrast to the role of F/SBPase in the CBB cycle during the day, apparently plays a role in the dark rather than in the light.

Fig. 8.

Growth of Δfbpase and merodiploid f/sbpase/Δf/sbpase deletion mutants compared with wild type (WT). Photoautotrophic and photomixotrophic growth in continuous light (A, C) and day/night cycles (B, D), and light activated heterotrophic growth (E) were measured. Mean, maximum, and minimum of three biological replicates are depicted. Each experiment was conducted at least three times and the results of one representative repetition are shown.

This fits well with our biochemical data, which show that FBPase retains its activity under oxidizing conditions, whereas F/SBPase is activated under reducing conditions and inhibited by AMP (low energy charge; Figs 3, 7; Supplementary Fig. S13). The likely role of FBPase in the dark phases is further supported by the fact that its expression is up-regulated during the night in contrast to F/SBPase (Saha et al., 2016).

Deletion of talB did not affect photoautotrophic or photomixotrophic growth in continuous light, or phototrophic growth in day/night cycles, neither in the single mutant nor in the double mutants in combination with f/sbpase/Δf/sbpase or Δfbpase (Supplementary Fig. S20A–C). However, growth of mutants with deleted talB was diminished under photomixotrophic conditions in day/night cycles (Supplementary Fig. S20D) and completely abolished under heterotrophic conditions (Supplementary Fig. S20E). Additional deletion of f/sbpase/Δf/sbpase or Δfbpase did not suppress growth further. On the contrary, the ΔtalBΔfbpase seemed to grow slightly better than the ΔtalB single mutant in photomixotrophic day/night cycles. The observation that deletion of talB only impacts growth in conditions with dark phases indicates that the enzyme is confined to its role in the catabolic OPP pathway in darkness and is rather not involved in the CBB cycle in light in Synechocystis.

Heterotrophic glucose consumption and cellular glycogen content of Δfbpase

For a more detailed investigation of the metabolism of Δfbpase in darkness, we measured glucose consumption from the medium and cellular glycogen content. Glucose consumption was similar in WT and Δfbpase (Fig. 9A), as was the glucose consumption rate per biomass (estimated by OD₇₅₀) and day (Supplementary Fig. S21). In contrast, the cellular glycogen content was higher in the Δfbpase mutant than in the WT over the whole time of heterotrophic cultivation (Fig. 9B). As a control, complementation mutants were constructed and characterized, confirming that the observed phenotypes are attributable to the deletion of FBPase (Supplementary Fig. S22A–C). These data show that a larger fraction of glucose is stored as glycogen rather than directed to growth in Δfbpase, indicative of a modified balance between glycogen and central carbon metabolism.

Fig. 9.

Heterotrophic glucose consumption and glycogen content of Δfbpase. (A) Glucose concentration in the medium. The accompanying growth curve can be found in the Supplementary Fig. S21A. (B) Specific cellular glycogen content (µg glycogen OD₇₅₀⁻¹ ml⁻¹) in Δfbpase and WT. The strains were pre-cultivated photoautotrophically and shifted to heterotrophic conditions (darkness, addition of 10 mM glucose) at time point 0. Data represent the means ±SD of three biological replicates. The experiment was conducted four times and the results of one representative repetition are shown.

Growth of pfk deletion mutants

To further clarify the relevance of the two isoenzymes of PFK and their relation to FBPase, we conducted growth experiments with WT, the double (Δpfk) and single (Δpfk-A1, Δpfk-A2) deletion mutants. Under photoautotrophic and photomixotrophic conditions, Δpfk grew like the WT but was impaired in growth under heterotrophic conditions (Fig. 10A, B, D), consistent with previous studies (Makowka et al., 2020). As expected, also both single deletion mutants grew like the WT under photoautotrophic and photomixotrophic conditions (Fig. 10A, B). Under photoheterotrophic conditions the double and single pfk deletion mutants grew again similar to the WT, but Δpfk-A2 reached reproducibly lower final optical densities (Fig. 10C). Under heterotrophic conditions, Δpfk-A2 growth was more strongly affected than that of Δpfk-A1 (Fig. 10D). Interestingly, the growth of the double deletion mutant lay between that of the two single deletion mutants.

Fig. 10.

Growth of the pfk single and double deletion mutants compared with WT under different conditions. Photoautotrophic (A), photomixotrophic (B), and photoheterotrophic (C) growth in continuous light, and light activated heterotrophic growth (D) were measured. The double deletion mutant Δpfk-A1Δpfk-A2 is abbreviated as Δpfk. Data represent the means ±SD of three biological replicates. Each experiment was conducted at least three times and the results of one representative repetition are shown.

Apparently both PFK isoenzymes are active and required for optimal heterotrophic growth in the dark with external glucose. They do not seem to be able to (fully) compensate the absence of each other, even though PFK-A2 seems to be more important.

Discussion

In summary, our results provide the following comprehensive picture of the switch of the cyanobacterial carbohydrate metabolism between light and darkness, which is shown in Fig. 11.

Fig. 11.

Proposed model of the cyanobacterial carbohydrate metabolism in light and darkness. (A) In the light, photosynthesis and anabolic reactions of the CBB cycle, including gluconeogenesis, are active. The bifunctional F/SBPase is activated by reducing conditions and catalyses two key reactions of the CBB cycle. The PFKs are inhibited by ATP and 3PG. (B) In darkness, F/SBPase is inhibited by AMP and oxidizing conditions, while inhibition of PFKs is lifted. The carbon flow is reversed in the catabolic direction via the OPP pathway (dependent on TalB) and glycolysis. FBPase is active in parallel. We hypothesize that FBPase influences the balance between glycogen and central carbon metabolism by controlling the level of its substrate FBP, which is involved in a regulatory network around phosphoglucomutase (PGM1) (Neumann et al., 2022). 3PG, 3-phosphoglycerate; CBB, Calvin–Benson–Bassham; E4P, erythrose 4-phosphate; F6P, fructose 6-phosphate; FBA, fructose-1,6-bisphosphate aldolase; FBP, fructose 1,6-bisphosphate; FBPase, fructose-1,6-bisphosphatase; F/SBPase, fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase; GAP, glyceraldehyde 3-phosphate; GBP, glucose 1,6-bisphosphate; OPP, oxidative pentose phosphate; PFK, phosphofructokinase; PGM, phosphoglucomutase; S7P, sedoheptulose 7-phosphate; SBP, sedoheptulose 1,7-bisphosphate; TalB, transaldolase.

Under photoautotrophic conditions, photosynthesis provides ATP and NADPH to fuel the CBB cycle. The bifunctional F/SBPase is activated by reduced thioredoxin. In light, AMP levels are probably below F/SBPase inhibiting concentrations. We showed that FBPase does not possess SBPase activity and its deletion does not affect photoautotrophic growth (Figs 4, 8A). Therefore, F/SBPase is most likely the sole enzyme responsible for dephosphorylating not only SBP, but also FBP to drive the CBB cycle. This interpretation is furthermore supported by a previous study in which only the simultaneous expression of plant cpSBPase and cpFBPase restored photoautotrophic growth in the fully segregated Δf/sbpase mutant (García-Cañas et al., 2022). The cell's own FBPase was apparently unable to replace the missing FBPase activity of the deleted f/sbpase in the light. Based on our measurements of FBPase activity in crude cell extracts, a likely explanation is that its activity in reducing conditions is negligible compared with that of F/SBPase (Fig. 7). The low activity of FBPase may, among other reasons, result from its approximately 30-fold lower protein abundance compared with F/SBPase (Jackson et al., 2023).

The growth behavior of various talB deletion mutants verifies that this enzyme only plays an essential role under heterotrophic conditions and is confined to the catabolic OPP pathway (Supplementary Fig. S20). While some chemoautotrophic bacteria that lack F/SBPase can use transaldolase instead (Frolov et al., 2019), Synechocystis is apparently unable to replace F/SBPase with transaldolase. Attempts to delete f/sbpase either resulted in mutants that were no longer able to grow photoautotrophically (Yan and Xu, 2008; García-Cañas et al., 2022) or the segregation of the mutants did not succeed (Tamoi et al., 1999). TalB is therefore unable to functionally replace F/SBPase in the CBB cycle.

Similarly, both PFK isoenzymes are inactive in the light, as evidenced by the growth behavior of the deletion mutants (Fig. 10A–C). The relatively high level of ATP produced in the photosynthetic light reaction inhibits PFK-A2, while 3PG, the first stable product of photosynthetic CO₂ fixation by Rubisco, inhibits PFK-A1 (Shen et al., 2024). Additionally, the concentration of their co-substrate ADP is relatively low. By utilizing ADP instead of ATP, the latter being the common co-substrate of PFKs from heterotrophic organisms, activation of PFKs is prevented under photoautotrophic conditions in cyanobacteria (Shen et al., 2024).

In the dark, oxidizing conditions prevail, so that F/SBPase is inactivated (Fig. 7). Due to the absence of the photosynthetic light reaction, the energy charge is presumably low in darkness, removing the inhibition of PFK-A2 by ATP and providing ADP as co-substrate for both PFKs. The inactivity of the CBB cycle results in lower 3PG concentrations, which removes inhibition of PFK-A1 (Shen et al., 2024). The activity of PFKs provides AMP, which further supports inhibition of F/SBPase. Thereby, the carbon metabolism is switched from an anabolic to a catabolic direction via glycolysis and the OPP pathway. Flux analyses show that under heterotrophic conditions carbohydrates are primarily metabolized via the OPP pathway, while glycolysis is used to a lesser extent (Yang et al., 2002; Wan et al., 2017). This explains the strong effect of talB deletion on heterotrophic growth. Accordingly, the PFKs are less important, although both PFK-A1 and PFK-A2 are required for optimal heterotrophic growth (Fig. 10D). Possessing two isoenzymes might have the advantage of providing greater flexibility and enabling a finer regulation of metabolism (Jablonsky et al., 2014).

Even though catabolic pathways are operating in darkness, the FBPase is apparently nevertheless active as well, as its deletion leads to reduced growth and increased glycogen content with unchanged glucose consumption from the medium (Figs 8E, 9; Supplementary Figs S21, S22). This is further supported by the fact that unlike F/SBPase, FBPase is not inactivated by oxidation (Fig. 7) and its expression is up-regulated in darkness, compared with light (Saha et al., 2016). The question arises as to which function the FBPase could fulfill. As an antagonistic enzyme couple, FBPase and the ADP-dependent PFK-As catalyse exergonic irreversible biochemical reactions between two substrates (F6P and FBP) in opposite directions and their simultaneous operation results in an energy dissipating futile cycle (Fig. 1). Such substrate cycles between PFKs and FBPases are well known from eukaryotes for thermogenesis (Staples et al., 2004; Sharma et al., 2024). Another assumed function of substrate cycles is that they permit a better fine-tuning of metabolic fluxes by enhancing the sensitivity of enzymes to changes in effector concentrations (Newsholme et al., 1984). In particular, switching between zero to moderate flow rates can be achieved more easily in the presence of regulated forward and back reactions. The Synechocystis FBPase could therefore have the task of fine-tuning the flow through the EMP pathway under heterotrophic conditions by catalysing its back reaction.

Another conceivable scenario is that FBPase participates in a cyclic OPP pathway in darkness by dephosphorylating FBP produced from GAP by fructose-1,6-bisphosphate aldolases (FBA) to F6P, which can subsequently be converted to G6P to start a new round of the OPP pathway (see Fig. 11).

However, heterotrophic flux analyses clearly show that the predominant flux through a cyclic OPP pathway directly leads from F6P produced by TalB to G6P, bypassing FBA and FBPase (as shown in Fig. 11) (Wan et al., 2017). A small flux via FBA and FBPase has been observed, but accounting for only 1% of the glucose uptake rate it seems of minor importance, compared with the flux via transaldolase which was determined to be 163% of the glucose uptake rate (Wan et al., 2017). In another study, no flux at all was detected from GAP to FBP through FBA, which instead operated in the opposite glycolytic direction under heterotrophic conditions (Yang et al., 2002). This is well in line with our observation that the EMP pathway, requiring FBA to convert FBP to GAP, is relevant under heterotrophic conditions. Taken together, the disruption of a potential minimal cyclic OPP flux through FBA and FBPase is unlikely to be the reason for the observed differences between WT and Δfpase mutant.

Instead, our data indicate that FBPase affects the balance between glycogen storage and growth (Figs 8E, 9). We hypothesize that this may be achieved by adjusting the concentration of its substrate, FBP, which participates in a regulatory network around phosphoglucomutase (PGM1) (Fig. 11) (Neumann et al., 2022).

PGM1 operates the control point for distributing carbon to either glycogen storage or growth by catalysing the interconversion of glucose 1-phosphate, the substrate for glycogen synthesis, and G6P, the substrate for glycolysis and the OPP pathway. It is activated by GBP, which is produced by a second regulatory phosphoglucomutase 2 (PGM2) from either glucose 1-phosphate or G6P, using FBP as a phosphodonor (Neumann et al., 2022). In addition, FBP was also shown to directly influence PGM1 activity, whereby its exact impact awaits further clarification (Neumann et al., 2022). While FBPase is not able to utilize GBP as a substrate directly (Supplementary Fig. S16), it could regulate PGM1 by adjusting the levels of FBP in collaboration with the PFKs, thus controlling the flow of glycogen into glycolysis and the OPP pathway in darkness. This hypothesis is supported by our observation that Δfbpase is not disturbed in its ability to consume glucose under heterotrophic conditions but stores a larger fraction of glucose as glycogen instead of utilizing it for growth (Fig. 9; Supplementary Fig. S22).

Phylogenetic analyses revealed that Synechocystis class I FBPase is more closely related to all plant FBPases and SBPase than Synechocystis class II F/SBPase (Fig. 6). The redox regulation in Synechocystis F/SBPase and plant cpFBPase might thus be the result of convergent evolution and an adaptation to a photoautotrophic lifestyle. The cpFBPase II occurring in some but not all plant chloroplasts is redox-insensitive like the Synechocystis FBPase (Serrato et al., 2009; Li et al., 2020). Therefore, it would be interesting to test if it might fulfil a similar function in dark metabolism of plants.

In conclusion, F/SBPase and FBPase perform distinct functions in the cyanobacterium Synechocystis. F/SBPase is highly abundant and is regulated by various biochemical effectors (AMP and redox charge), which ensure its activation in light and inactivation in darkness (Table 1; Supplementary Fig. S13). FBPase, on the other hand, occurs in lower concentrations and is not regulated by effectors (Fig. 3), but at the transcriptional level, with its concentration increasing in darkness. Accordingly, F/SBPase catalyses reactions in the CBB cycle in light, while we found that FBPase influences the balance between glycogen storage and growth in darkness. This could be achieved by the formation of a regulatory futile substrate cycle with the PFKs to fine-tune the flow through the EMP pathway and/or by controlling PGM1 activity via the FBP level to adjust the distribution of carbon flow between glycogen synthesis or growth via the EMP and OPP pathways. The classic control point in the EMP pathway, which is known to be mediated by the antagonistic enzyme pair PFK and FBPase in heterotrophic bacteria and eukaryotes, is therefore also present in Synechocystis. However, adaptation to the dual heterotrophic and photoautotrophic metabolism requires a more complex regulatory mechanism involving two ADP-dependent PFK-A isoenzymes, a FBPase, and a bifunctional F/SBPase. The complex and delicately tuned network of differently expressed and regulated isoenzymes at the control points of anabolic and catabolic carbon metabolism enables cyanobacteria to switch the direction of their carbon flow and to coordinate glycogen storage and growth.

Abbreviations

3PG

3-phosphoglycerate

CBB

Calvin–Benson–Bassham

DHAP

dihydroxyacetone phosphate

DTT

dithiothreitol

E4P

erythrose 4-phosphate

EMP

Embden–Meyerhoff–Parnass

F6P

fructose 6-phosphate

FBA

fructose-1,6-bisphosphate aldolase

FBP

fructose 1,6-bisphosphate

FBPase

fructose-1,6-bisphosphatase

F/SBPase

fructose-1,6-biphosphatase/sedoheptulose-1,7-biphosphatase

G6P

glucose 6-phosphate

GAP

glyceraldehyde 3-phosphate

GBP

glucose 1,6-bisphosphate

OPP

oxidative pentose phosphate

PFK

phosphofructokinase

PGI

phosphoglucose isomerase

PGM

phosphoglucomutase

S7P

sedoheptulose 7-phosphate

SBP

sedoheptulose 1,7-bisphosphate

SBPase

sedoheptulose-1,7-biphosphatase

SEC

size exclusion chromatography

TalB

transaldolase

WT

wild type

ZWF

glucose-6-phosphate dehydrogenase

Supplementary data

The following supplementary data are available at JXB online.

Table S1. Primer list.

Table S2. Accession numbers of protein sequences used for phylogenetic analyses.

Table S3. Mutant list.

Fig. S1. Segregation for f/sbpase in the single and double f(/s)bpase deletion mutants.

Fig. S2. Segregation for f/sbpase in different mutants.

Fig. S3. Segregation for fbpase in the single and double deletion mutants.

Fig. S4. Segregation of the fbpase complementation mutants.

Fig. S5. Segregation of the talB deletion mutants.

Fig. S6. Segregation of the pfk single and double deletion mutants.

Fig. S7. Test of redox conditions in the FBPase activity assay with cell extracts.

Fig. S8. FBpase activity in cell extracts under different redox conditions.

Fig. S9. Purification of the recombinant Synechocystis FBPase (slr0952).

Fig. S10. Native molecular mass of FBPase.

Fig. S11. Metal dependency and optimal concentration for FBPase activity.

Fig. S12. Activity of FBPase under varying concentrations of ATP, AMP and DTT.

Fig. S13. Enzymatic activity and regulatory properties of F/SBPase (slr2094).

Fig. S14. NMR spectra of FBPase activity by FBPase and F/SBPase.

Fig. S15. Effect of DHAP and E4P on FBPase activity.

Fig. S16. Glucose-1,6-bisphosphate interaction with FBPase and F/SBPase.

Fig. S17. Multiple sequence alignment of FBPases and SBPases.

Fig. S18. Structural comparison of substrate-binding sites in class I FBPase and SBPase enzymes.

Fig. S19. Structural comparison of Synechocystis F/SBPase, FBPase, and E. coli class II FBPase.

Fig. S20. Growth of the talB single and double deletion mutants.

Fig. S21. Heterotrophic growth and glucose consumption rate in Δfbpase.

Fig. S22. Characterization of the fbpase complementation mutants.

Protocol S1. Mutant segregation.

Protocol S2. Verification of the redox-regulation of F/SBPase observed in cell extracts.

Protocol S3. Discussion on structural basis of substrate specificity in class I FBPase and SBPase

Author contributions

FC, RO, LS, MG, and MB performed the experiments; FC, RO, LS, CB, JLS, KF, MH, BS, and KG analysed data; FC and KG wrote the original draft; RO, BS, and MH, wrote, reviewed, and edited; KF, MH, BS, and KG conceptualized the study; BS and KG supervised; KF, MH, BS, and KG acquired the funding.

Funding

We acknowledge funding by the German Research Foundation, Bonn (Deutsche Forschungsgemeinschaft, DFG) for the Research Unit SCyCode (FOR2816) (grant SI 642/14–1 and SI 642/14-2 to BS, HA2002/23-2 to MH, GU 1522/5-1 to KG, and FO195/16-2 to KF), DFG grant GRK2749/1 as well as financial assistance from the Department of Science and Innovation (DSI)/National Research Foundation (NRF) in South Africa (grant NRF-SARCHI-82813 to JLS).

Data availability

The primary data supporting this study were not made publicly available at the time of publication. The data that support the findings of this study are available from the corresponding author upon request.