To capture the phenotypical outcomes of Synechocystis sp. PCC 6803, we measured the cellular growth rate under four different light intensitites (40, 80, 480 and 960 μmol m^–2 s^–1), with the goal of capturing light-limited, light saturated and light-inhibited growth (Figure 1). Under light-limited conditions, growth of cyanobacteria is at least partially constrained by energy supply, while under light-inhibiting conditions, the production of radical oxygen species can cause photoinhibition or photobleaching in photoautotrophic organisms (Tripathy and Oelmüller, 2012; Hsieh et al., 2014).

Global non-targeted proteomics revealed distinct clustering (Supplementary Figure 1) under different light conditions, demonstrating significant global changes in the proteome. Three-dimensional principal component analysis was done on the proteomic data (Supplementary Figure 2), demonstrating distinct and reproducible clusters for each light condition.

Several CBB cycle enzyme abundances are well correlated with light level (Figure 2). RPIa, ENO, PGK, RPE, TKT2, FBA-II, GPM, PGM, GAPDH1, and both subunits of RBC are largely positively associated with increased light levels and growth rate. In contrast, G6PDH, FBA-I, and PKT are negatively associated with light level and growth rate. Meanwhile, GAPDH2, G6PI, TKT, TPI, FBP/SBPase, TA, and both isoforms of phosphoribulokinase (PRK) do not change significantly over the studied light intensities. Additionally, the majority of detected abundances for photosystem subunits were negatively associated with light level, while TCA cycle enzyme abundances appeared to be largely unaffected (Supplementary Figure 3). These results demonstrate an individualized enzyme abundance response to changes in light level. Rather than simply increasing the overall enzyme abundance uniformly, this result is more consistent with a redistribution of enzyme abundances.

To further quantify the relationship between metabolic fluxes and individual enzyme abundances, the Pearson correlation coefficient and associated p-value was calculated for each enzyme and its corresponding reaction (Table 1). Based on these results, the enzymes ENO, GAPDH1, RBC_L, PGM1, FBA-I, FBA-II, RPE, and RBC_S have the highest correlations with their corresponding flux. In cases where a relationship between enzyme abundance and corresponding flux could be observed (p < 0.1), a “pseudo-flux control coefficient” (PFCC), PEiJ, was also calculated.

In metabolic control analysis, a flux control coefficient, CEiJ, is the scaled infinitesimal change of flux, J, in response to some infinitesimal change in enzyme level, E_i, around a reference state (Ap Rees and Hill, 1994). Typically, a flux control coefficient is most accurate when a minimal number of enzymes are perturbed and other enzyme abundances are held constant to minimize deviations from linearity, a quality that is not met in these experimental conditions. Furthermore, due to a scarcity of data points, and because these experimental perturbations are not infinitesimal, these coefficient estimates are likely to be inaccurate (Pettersson, 1996; Wu et al., 2004; Link and Weuster-Botz, 2007), and do not abide by the summation or connectivity theorems typically applied in metabolic control analysis. As a result, we relied on an inexact approximation for the FCC, the PFCC, where these requirements are relaxed. PFCCs can be a useful way to determine if there is a strong relationship between flux and enzyme abundance around a reference state. In this case, the reference state was chosen to be the 80 μmol m^–2 s^–1 steady state. The non-linear fitting method (Small, 1993) was used to estimate the coefficients for each enzyme. In prior literature, enzymes with high flux control coefficients have been suggested as potential targets for overexpression (Fell, 1998), as they potentially offer high positive control over a pathway. Of the selected enzymes, PGM1, ENO, RPE, and FBA-II enzymes have the highest PFCCs, with FBA-II and both RBC subunits in particular corroborating experimental results which demonstrate enhanced growth rates in Synechocystis (Liang and Lindblad, 2016). Interestingly, these PFCCs were calculated to be above unity. In such cases, the flux increases proportionally more than the enzymatic abundance around the reference state, indicating the presence of other confounding variables that positively affect flux. For instance, substrate concentrations may be altered after perturbation of light level, altering flux. Additionally, CBB cycle enzymes are known to adopt more active conformations as a result of redox-sensitive post-translational modifications (Michelet et al., 2013), meaning that the same amount of enzyme concentration can support a larger flux under favorable reducing conditions. The flux control coefficients were not calculated for enzymes where the correlation between enzyme abundance and flux were ambiguous or negligible (Table 1, p > 0.1), as it was not possible to perform any accurate non-linear fitting between J and E.

Previous isotopic labeling studies (Huege et al., 2011; Young et al., 2011) found notable ¹³C labeling kinetics wherein downstream metabolites had higher ¹³C labeling than their precursors at the steady state condition. A proposed explanation for this phenomenon is that CBB cycle metabolites may be subjected to metabolite channeling, a phenomenon where enzymes are organized into complexes that pass metabolites to each other, thereby potentially increasing the local concentration resulting in higher enzymatic reaction rates (Süss et al., 1993; Anderson et al., 2005; Abernathy et al., 2017a). The most obvious example of metabolite channeling in cyanobacteria occurs in the carboxysome, where RuBisCO and carbonic anhydrase are localized (Yu et al., 1994; Rae et al., 2013; Long et al., 2018; Abernathy et al., 2019) to facilitate carbon fixation by raising the local concentration of CO₂ around RuBisCO.

Our results reinforce these observations, as we consistently found lower ¹³C labeling enrichments in the first of sequential pairs of metabolites such as FBP and F6P, as well as RU5P and RUBP (Figure 3A). The presence of metabolite channeling would result in a more quickly labeled ‘active’ pool as well as a slowly labeled ‘inactive’ pool that dilutes the overall labeling enrichment measurement. With the incorporation of dilution parameters, it is also possible to estimate the fraction of active and inactive intermediates using INST-MFA (Figure 3B) (Ma et al., 2014). In general, the inactive pool fractions were minimal during the highest growth condition (480 μmol m^–2 s^–1), and significantly higher when light was either limited or inhibiting growth.

For the INST-MFA model, we designated the metabolic network to encompass reactions in the CBB cycle, TCA cycle, photorespiration, and the phosphoketolase shunt (Supplementary File 2). Although biomass composition may be altered slightly across light levels (Zavřel et al., 2019), these differences are not statistically significant at the measured growth rates, and we failed to identify statistically significant differences in our own carbohydrate and protein assays. Therefore, an additional simplifying assumption was made that biomass composition is not significantly altered for WT cells grown under varying light levels. The net carbon fixation was computed from the overall growth rate, allowing for the individual fluxes to be represented on an absolute basis in units of μmol C gDW^–1 h^–1 (assuming 0.0412 mol C gDW^–1 based on the WT biomass equation). In all tested conditions, as expected, the majority of carbon (∼95%) was fixed from RuBisCO activity as opposed to other CO₂ incorporating reactions such as PEP carboxylase. In general, it is difficult to statistically differentiate between small CBB cycle fluxes due to relatively large error bounds (Supplementary File 2). This indicates that relative carbon flux can be constantly regulated to partition itself into different biomass components regardless of the light level.

A relatively small amount of oxidative pentose phosphate pathway flux is observed under light-limited conditions, which results in differences in isotopic labeling of R5P and RU5P. However, this activity is nearly zero at light saturating or light inhibiting conditions, corroborating prior observations of efficient photoautotrophic metabolism (Ueda et al., 2018).

Calculating fluxes for the TCA cycle precisely is difficult due to its comparatively low values under photoautotrophic conditions (Ma et al., 2014), resulting in no statistically significant differences in TCA cycle flux between all light intensities (Supplementary File 2). On the other hand, there is a generally slight increase in the ¹³C labeling of TCA cycle intermediates malate and succinate as the light level increases, which is consistent with an increase in TCA cycle flux. Similarly, a slight increase in labeling of the photorespiration intermediates 2PG and GA is observed at saturating and light inhibiting conditions, but photorespiration fluxes are again poorly resolved due to its relatively low values as compared to the CBB cycle. Such a result indirectly supports the notion that photorespiration may serve as a way to protect against oxidative stress (Sunil et al., 2019). We note that predicted photorespiration activity has non-zero lower bounds even under low light conditions (Figure 4), corroborating prior literature (Burnap et al., 2015; Klemke et al., 2015). Furthermore, as shown in flux maps (Abernathy et al., 2017b) and in previous genome scale modeling (Nogales et al., 2012), photorespiration can be beneficial for high light photosynthesis, and was also found to play an important role during maximal Synechococcus elongatus UTEX 2973 growth (Abernathy et al., 2017b), despite carbon loss.

Although it has been shown that the total protein concentration of cyanobacteria is relatively invariant (Zheng and O’Shea, 2017), individual abundances of light harvesting complexes (LHC) and metabolic enzymes as revealed by proteomic studies are at least partially dependent on light level. Specifically, the relative abundances of LHC proteins detected in this study generally decreased with increasing light level, in agreement with previous proteomic studies (Jahn et al., 2018; Zavřel et al., 2019). In contrast, many Calvin cycle enzymes generally increased with increasing light level. For example, the amount of RPIa, ENO, PGK, both subunits of RBC, RPE, TKT2, FBA-II, GPM, PGM, and GAPDH1 are largely positively associated with increased light levels and growth rate, which is the expected behavior in a light dependent carbon fixation pathway like the Calvin cycle (Figure 2). On the other hand, enzymes such as G6PDH, FBA-I, and PKT are also notable for being negatively associated with light level and growth rate, but these results are not surprising. G6PDH, as the first step of the oxidative pentose phosphate pathway (OPPP), is known to be more active in the dark (Ueda et al., 2018). Synechocystis contains two FBA isozymes (Class I and Class II) (Nakahara, 2003), of which FBA-II is thought to be the more active and abundant enzyme. Additionally, phosphoketolase (PKT) showed an inversely correlated expression profile compared to the majority of CBB cycle enzymes. Previously, PKT was found to be non-essential, but slightly beneficial to growth rate during photoautotrophic conditions (Xiong et al., 2015). PKT catalyzes an alternate entry point into the tricarboxylic acid (TCA) cycle. Interestingly, phosphoglycerate mutase (PGM) and enolase (ENO) do not share this behavior, despite also catalyzing an entry point into the TCA cycle. It was earlier speculated that PKT may be able to improve the carbon efficiency since the phosphoketolase pathway does not lose carbon but utilizes different amounts of cofactors (Xiong et al., 2015). The balance between the phosphoketolase-dependent pathway and the phosphoketolase-independent pathway may perhaps reflect a different balance of energetic cofactors under different light levels. These results also indicate that some steady-state enzyme abundances are more light-dependent than others despite being part of the same pathway, suggesting possible differences in regulatory behavior. These differences in sensitivity between individual CBB cycle enzymes and light level were not fully apparent when grouping enzymes, as was done in previous proteomic studies (Jahn et al., 2018; Zavřel et al., 2019).

Although integrating proteomic and fluxomic datasets has been previously discussed in the literature (Zhang et al., 2010; Winter and Krömer, 2013), actual studies incorporating these techniques together remain uncommon. Prior studies (Saha et al., 2016; Nakajima et al., 2017) examined transcriptomic results together with fluxes to describe regulation in Synechocystis 6803, but the use of proteomics over transcriptomics represents an iterative improvement. The use of proteomics and fluxomics together presents an opportunity to examine correlations between fluxes and enzyme abundances. Based on statistical correlation analysis (Table 1), many enzyme abundances appear to be uncorrelated with their actual corresponding flux. There may be several reasons for the apparent insensitivity of the abundance of enzymes like GAPDH2, G6PI, TKT, TPI, FBP/SBPase, TA, and both isoforms of PRK to light intensity. First, the isomerases (G6PI and TPI) have generally been considered poor candidates of exerting control over the pathway based on kinetic parameters (Zhu et al., 2007), having relatively rapid kinetics. The relatively insensitive behavior of GAPDH2 is also not unexpected when considering that the more photosynthetically active GAPDH1 isoform is positively associated with growth. Another possible contributing reason may be due to a slow protein turnover rate, which implicates post-translational regulation as a way to control flux (Karlsen et al., 2021).

As d⁢Jd⁢Ei can vary widely around different reference states, the analysis of the relationship between enzyme abundance and flux can be complicated. One interpretation of uncorrelated proteomic and fluxomic data is that these reactions are regulated at a post-translational level (Bro et al., 2003). These results appear to contradict previous estimates of CBB cycle enzymes with higher FCCs as derived from kinetic models (Janasch et al., 2019; Wu et al., 2020) especially for PRK and FBP/SBPase, strongly implying that the control of these fluxes, in wild-type, are not based on enzyme abundances. One explanation may be that the overexpression of these enzymes is conditionally beneficial to growth depending on the light condition due to different FCCs at different reference states. Indeed, our prior study indicated significant growth advantages when FBP/SBPase was overexpressed in Synechocystis 6803 at low light conditions, but a much smaller advantage when the light level was increased (Hing et al., 2019). Some alternative forms of regulation that could explain the changes in fluxes are discussed in the section below.

There is compelling evidence of localization of CBB cycle enzymes near the thylakoid membrane (Dani and Sainis, 2005; Agarwal et al., 2009). Specifically, RPI, PRK, RuBisCO, PGK, and GAPDH, which comprise five sequential enzymes of the CBB cycle were found to be preferentially, but not completely, localized near the thylakoid membranes in Synechocystis sp. PCC 6803 (Süss et al., 1993; Dani and Sainis, 2005; Agarwal et al., 2009). Similarly, aldolase, SBPase, and FBPase have been noted to be preferentially localized around the thylakoid membrane in pea leaves (Alscher-Herman, 1982; Hermoso et al., 1989; Andres et al., 1990; Anderson et al., 2005) even though these enzymes have functional activity in soluble in vitro assays devoid of any membranes. These enzymes can benefit from the production of energetic cofactors from the light reactions of photosynthesis occurring around the thylakoid membrane, and are able to quickly interact with the thioredoxin-ferredoxin system to transition between their redox regulated post-translational states (Michelet et al., 2013). In the aforementioned immunolabeling studies, it was demonstrated that at least a fraction of the detected CBB enzymes were cytosol localized, although an explanation for the partial split between cytosol and thylakoid membrane localized fractions was not explicitly presented therein. Yet, as shown previously in chloroplasts, exposure to sufficient quantities of light and/or reducing factors can induce a release of enzymes from the thylakoid membranes (Ben-Bassat and Anderson, 1981; Tanaka et al., 1990), suggesting a strong link between redox environment and enzyme localization.

Consequently, it is reasonable to conclude that the co-localization of CBB cycle enzymes onto the periphery of the thylakoid membranes can result in the observed ¹³C enrichment pattern, by forming a metabolite channel. The formation of a metabolite channel can regulate flux for several reasons. First, the local concentration of metabolites can be higher than the overall measured intracellular concentration. Second, the kinetic parameters of thylakoid-bound enzymes can be significantly different from their soluble isoforms. Our data indicates that while metabolite channeling is generally occurring under photoautotrophic conditions, the fractional size of the inactive pools is not constant across all conditions. Most notably, the predicted fractional inactive pool size of RU5P is significantly higher at the light-inhibiting condition and to a lesser extent under the low light conditions as opposed to the optimal light condition because of a stark difference in RU5P enrichment (0.84 and 0.35 at 480 and 960 μmol m^–2 s^–1, respectively) between those conditions despite a similarly labeled RUBP (0.90 and 0.93 at 480 and 960 μmol m^–2 s^–1, respectively) (Figure 3). To a lesser extent, this trend is also present for FBP and F6P enrichment. This result implies a movement of PRK and FBP/SBPase to the cytosol to turn over the inactive cytosolic pool at highly reducing conditions. Thus, we infer from this data that the inactive pool sizes are related to the redox environment of the cell. During light inhibiting conditions, the production of radical oxygen species can alter redox balance to favor oxidation (Hsieh et al., 2014). Similarly, the lower amount of reducing cofactors produced during low light conditions also results in a non-optimal redox balance. It is only around the optimal light condition where redox conditions are ideal that PRK and FBP/SBPase most effectively turns over the pool of their respective substrates.

Recently, a synthetic metabolite channel was engineered in Bacillus subtilus to enhance N-acetylglucosamine synthesis (Lv et al., 2020), which provides concrete empirical evidence that metabolite channeling can significantly alter reaction efficiency. The creation of metabolite channels is a promising metabolic engineering strategy because it is orthogonal to many common metabolic engineering strategies such as enzyme overexpressions or knockdowns, allowing for a more complex or fine-tuned approach. While the exact mechanism remains unclear, it is apparent that the fraction of the inactive pool was minimized under high growth rates, and therefore may provide a future target for metabolic engineering. For example, addition of a thylakoid-binding domain to other enzymes outside the CBB cycle to make use of channeled metabolites can potentially increase the rate of specific reactions outside of the high growth rate condition. Both the proteomic and labeling data in combination with prior literature suggest a complex regulatory system where light availability simultaneously influences enzyme activity and enzyme localization, as has already been known to occur with carboxysomes (Sun et al., 2016), resulting in control of CBB fluxes. Efforts to capture these important regulatory mechanisms with mechanistic kinetic models should begin to consider these previously neglected elements, as thylakoid-bound enzymes may have different kinetic parameters than their unbound forms.

Phosphoribulokinase is known to form a complex together with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and the small protein CP12, forming a multi-enzyme complex that appears to be conserved in both cyanobacteria and plants (Trost et al., 2006; Marri et al., 2010; López-Calcagno et al., 2014). This complex is also known to form under an oxidative redox environment (Trost et al., 2006), which inhibits both PRK and GAPDH activity, and dissociates under highly reducing conditions. In Synechocystis, two GAPDH genes exist, gap1 and gap2, of which gap2 appears to be responsible for anabolic CBB cycle activity, and gap1 for catabolic glycolytic activity (Koksharova et al., 1998). Of these genes, gap2 appears to share sequence similarity with chloroplastic gapAB genes which constitutes the PRK/CP12/GAPDH complex in higher plants (Koksharova et al., 1998). Curiously, our proteomic study suggested that both PRK and GAPDH2 (gap2) expression was relatively insensitive to light level, a result corroborated in a recent proteomic study (Zavřel et al., 2019). This contrasts with many other CBB cycle enzymes, which are positively correlated with light-intensity. Therefore, these results suggest that the formation of the PRK/CP12/GAPDH is a significant way that cells regulate PRK and GAPDH activity as opposed to simply varying enzymatic abundance. One advantage of forming such a complex is that the formation or dissociation of this complex is much more rapid than de novo synthesis or degradation of proteins (Howard et al., 2008).

During light-inhibiting conditions, the production of radical oxygen species becomes significant and may promote PRK/CP12/GAPDH formation similarly to a low light or dark condition as reflected by the lower growth rate. Therefore, under light-inhibiting conditions where oxidative stress could be high, the formation of PRK/CP12/GAPDH complex can be an additional way to explain the reduced flux through PRK and GAPDH. The PRK/CP12/GAPDH complex has much lower activity as compared to the dimeric PRK or GAPDH alone (Howard et al., 2008). Based on these findings, both the formation of metabolic channels and the formation of the PRK/CP12/GAPDH complex are coordinated by the redox environment (Figure 5).

It was not possible to resolve any statistically significant differences in the relative carbon flux distribution throughout the CBB cycle under different light conditions because of the large upper and lower bounds, implying a strict level of regulation inherent in Synechocystis central metabolism. In comparison with earlier INST-MFA flux maps performed under several different light conditions (Young et al., 2011; Jazmin et al., 2017; Hing et al., 2019), these error bounds were much larger, likely due to the addition of more inactive pools in the stoichiometric model. This contrasts with the estimated active and inactive fractional pool sizes of the metabolites themselves, which are frequently significantly different (Figure 3). In other words, the CBB cycle can regulate carbon flow to maintain biomass growth, under photoautotrophic conditions, at least in part via modulating the fractional size of the active intermediate metabolite concentrations. Small changes in inactive pool sizes can result in large differences in absolute flux. Recent metabolomic studies confirm significant changes in intermediates across the diurnal cycle to maximize growth during the day (Werner et al., 2019; Jaiswal and Wangikar, 2020). The ability to modulate the fractional size of active intermediates during different light conditions provide an additional means to control the rate of CBB cycle reactions.

Synechocystis sp. PCC 6803 cells were cultured in batch in BG-11 PC media (van Alphen et al., 2018) with no organic carbon source in a photobioreactor (PBR101, Phenometrics, MI, United States) at 30°C, with a magnetic stir bar (300 rpm) and 600 mL total volume with constant sparging of air (1 L/min). Continuous light at four different intensities (40, 80, 480 and 960 μmol m^–2 s^–1) was provided by a cold LED light (PBR101, Phenometrics, MI). The light spectra is provided as a Supplementary Figure 4. Optical density at 730 nm was measured to provide cell density estimates at each condition, and the growth rate was estimated using the Levenberg Marquardt algorithm during early exponential growth phase, ranging from OD₇₃₀ 0.1 to 0.7. ¹³C was introduced to the bioreactor after OD₇₃₀ reached at least 0.6.

Isotopically non-stationary metabolic flux analysis (INST-MFA) allows for estimation of flux distribution throughout photoautotrophic central carbon metabolism by tracking the transient incorporation of ¹³C into cellular metabolism (Ma et al., 2014, 2017; Abernathy et al., 2017b; Hendry et al., 2017; Gopalakrishnan et al., 2018). NaH¹³CO₃ in water was injected into the bioreactor to a total concentration of 1 g NaH¹³CO₃/L to quickly introduce ¹³C into cellular metabolism during exponential phase. Culture samples (10 mL) were removed and quenched at separate time intervals (0, 0.5, 1, 1.5, 2, 3, 5, 10, and 20 min) into 40 mL pre-chilled 60% methanol in a dry ice-ethanol bath and rapidly pelleted at 8000 g, −20°C and then extracted. The MATLAB-based software INCA v1.8 (Young, 2014) was used to estimate pathway fluxes at each light condition (Supplementary File 2). Flux estimates were only accepted if the calculated χ² goodness of fit is within the expected 95% confidence interval. Upper and lower bounds on fluxes were estimated using the parameter continuation method.

Pellets were extracted using 500 μL chloroform/methanol (3:7 v/v) under −20 °C for 2 h followed by two rounds of 500 μL 0.2 M NH₄OH (Prasannan et al., 2018). The upper methanol-water layer was collected. Samples were dried under N₂ and then reconstituted in 60 μL ultrapure water. An QTRAP 5500 (AB Sciex) linked to HPLC (Shimadzu) was used to identify central carbon metabolites using an established ion-pairing method (Shastri and Morgan, 2007; Ma et al., 2017) using a Polaris C18 column (150 × 2 mm) (Agilent, CA, United States).

Immediately prior to labeling experiments, a 20 mL aliquot of culture was withdrawn and pelleted for use in non-targeted proteomics. Cell pellets were washed twice in 20 mM PBS and then stored at −80°C until sample preparation for proteomic analysis. Cell pellets were re-suspended in 400 μL of 100 mM ammonium bicarbonate containing 1 mM phenylmethylsulfonyl fluoride and homogenized at high pressure (20,000 psi) at 6500 rpm for 90 s (Bertin Technologies SAS). Homogenate was transferred to new tubes and centrifuged at 13,500 rpm for 15 min at 4°C. The supernatant was transferred to a new tube and treated as a soluble fraction, and the pellets were treated as insoluble fraction. Proteins in the soluble fractions were precipitated overnight at −20°C using 4 volume of cold (−20°C) acetone and precipitated proteins were pelleted by centrifugation at 13,500 rpm for 15 min at 4°C. The soluble and insoluble fractions were dissolved in 8 M urea at room temperature for 1 h with continuous vortexing. The protein concentration was determined by Bicinchoninic Acid BCA assay and 50 μg of the protein (equivalent volume) was used for proteomics sample preparation. Samples were incubated in 10 mM dithiothreitol at 37°C for an hour for reduction followed by incubation in 20 mM iodoacetamide for an hour at room temperature in the dark for alkylation. Samples were digested using Trypsin/LysC protease mix at 1:25 (enzyme to substrate) ratio at 37°C. After 3 h of digestion, additional Trypsin/LysC protease mix was added at 1:50 (enzyme to substrate) ratio and digestion was allowed to proceed overnight at 37°C. Peptides were desalted using C18 micro spin desalting columns (The Nest Group, Inc.). Eluted peptides were dried and re-suspended in 3% acetonitrile/0.1% formic acid to a final concentration of 1 μg/μL and 1 μL (1 μg) was loaded used for LC-MS/MS analysis.

Peptides were analyzed in an Dionex UltiMate 3000 RSLC nano System coupled on-line to Q Exactive Orbitrap High Field Hybrid Quadrupole Mass Spectrometer (Thermo Fisher Scientific, Waltham, MA, United States) as described previously (Aryal et al., 2018; Mohallem and Aryal, 2020). Briefly, reverse phase peptide separation was accomplished using a trap column (300 μm ID × 5 mm) packed with 5 mm 100 Å PepMap C18 medium coupled to a 50-cm long × 75 μm inner diameter analytical column packed with 2 μm 100 Å PepMap C18 silica (Thermo Fisher Scientific). The column temperature was maintained at 50°C. Sample was loaded to the trap column at a flow rate of 5 μL/min and eluted from the analytical column at a flow rate of 300 nL/min using a 120-min LC gradient. The column was washed and equilibrated by using three 30 min LC gradient before injecting next sample. Precursor ion (MS1) scans were collected at a resolution of 120,000 and MS/MS scans at a resolution of 15,000 at 200 m/z in data dependent acquisition mode.

LC-MS/MS data were analyzed using MaxQuant (version 1.6.3.4) against the Synechocystis sp. PCC 6803 database downloaded from the NCBI¹. We edited the following parameters for our search: precursor mass tolerance of 10 ppm; enzyme specificity of trypsin/Lys-C enzyme allowing up to 2 missed cleavages; oxidation of methionine (M) as a variable modification and carbamidomethylation (C) as a fixed modification. False discovery rate (FDR) of peptide spectral match (PSM) and protein identification was set to 0.01. Proteins with LFQ >0 and MS/MS (spectral counts) ≥2 were considered as true identification and used for further downstream analysis. In total, 1,202 proteins were detected in at least one of the four tested conditions.

To compare data between different light conditions, the label free quantification (LFQ) intensities were used. The intensities of the soluble and insoluble fractions were combined for each replicate. InfernoRDN² (Polpitiya et al., 2008) was used to perform a Kruskal-Walis test to determine p-values and associated q-values to correct for background false discovery rate for the 294 detected proteins that were detected at least twice per triplicate at all conditions (Supplementary File 8).