Ensemble Modeling of Cyanobacteria Metabolism Using Multi-Omics Data

Cyanobacteria is a promising host for a sustainable bio-production from carbon dioxide. The dynamic behavior of the central metabolism in cyanobacteria has been uncovered by the transcriptome, metabolome, and fluxose analyses of a glucose-tolerant strain of the Synechocystis sp. PCC 6803 under auto-, mixo- and heterotrophic conditions. However, its control mechanism remains unclear due to lack of the enzyme expression data and a kinetic metabolic model required for the metabolic control analysis. In this study, protein expression profile data were obtained from Synechocystis sp. PCC 6803 by the targeted proteome analysis. In addition to an integrated analysis with other omic data, kinetic models of the central carbon metabolism was constructed by the ensemble modeling technique with using the proteome, metabolome and fluxome dataset.

Method

Synechocystis sp. PCC 6803 cells were cultivated in 500 mL Erlenmeyer flasks with 100 mL modified BG11 medium (with 5 mM glucose and 20 µM atrazine) under continuous light (40 µmol of photons/m²/sec). The peptide samples were analyzed using the nanoLC-MRM system (LC: LC-20ADnano (Shimadzu, Japan), MS: LCMS-8040 (Shimadzu, Japan)).  A metabolic model including 62 reactions with the Michaelis-Menten and mass action type kinetics was used for the ensemble modeling. Flux control coefficients were determined by the metabolic control analysis (Fell et al. 1992). All in silico analyses were performed by in-house scripts written with Python2.7+Numpy1.9.

Results and discussions

Expression profiles of 93 central metabolism and photosystem related proteins were obtained from Synechocystis sp. PCC 6803 under the auto-, mixo- and heterotrophic conditions by the targeted proteome analysis using nanoLC-MS. Levels of rbcS protein in the mixotrophic and heterotrophic conditions were lower than that of the autotrophic condition. Comparisons with other omics data showed that the expression patterns of rbcS protein was correlated with that of metabolic flux levels in RuBisCo reaction. On other hands, there were poor similarities with rbcS gene expressions and ribulose-1,5-biophosphate accumulations.

An ensemble including 100,000 models of Synechocystis sp. PCC 6803 metabolism with randomly sampled parameters were produced using the proteome, metabolome and fluxsome data at the autotrophic condition as a reference. Photon flux was determined by minimizing  ATP futile cycling. Following a modulation of the enzyme expression levels to that of heterotorophic condition according to the proteome data, the metabolic flux and metabolite concentration levels were predicted and compared with that of measured data of the heterotorophic condition. A subset of 100 metabolic models with better predictability were selected from the 100,000 models. The metabolic control analysis (MCA) using the 100 models showed that the expression levels of RuBisCo (rbcS) protein had low flux control coefficient (less than 0.01 in all models) for the RuBisCo flux level under the autotrophic condition studied here. The MCA also suggest a dispersed control of RuBisCo flux by several enzyme activities. These results suggested that the ensemble modeling using multi-omics data could be a strategy to investigate a control of the cyanobacteria metabolism. This work was supported by NSF/JST.