A refined genome-scale reconstruction of Chlamydomonas metabolism provides a platform for systems-level analyses

Abstract

Microalgae have reemerged as organisms of prime biotechnological interest due to their ability to synthesize a suite of valuable chemicals. To harness the capabilities of these organisms, we need a comprehensive systems-level understanding of their metabolism, which can be fundamentally achieved through large-scale mechanistic models of metabolism. In this study, we present a revised and significantly improved genome-scale metabolic model for the widely-studied microalga, Chlamydomonas reinhardtii. The model, iCre1355, represents a major advance over previous models, both in content and predictive power. iCre1355 encompasses a broad range of metabolic functions encoded across the nuclear, chloroplast and mitochondrial genomes accounting for 1355 genes (1460 transcripts), 2394 and 1133 metabolites. We found improved performance over the previous metabolic model based on comparisons of predictive accuracy across 306 phenotypes (from 81 mutants), lipid yield analysis and growth rates derived from chemostat-grown cells (under three conditions). Measurement of macronutrient uptake revealed carbon and phosphate to be good predictors of growth rate, while nitrogen consumption appeared to be in excess. We analyzed high-resolution time series transcriptomics data using iCre1355 to uncover dynamic pathway-level changes that occur in response to nitrogen starvation and changes in light intensity. This approach enabled accurate prediction of growth rates, the cessation of growth and accumulation of triacylglycerols during nitrogen starvation, and the temporal response of different growth-associated pathways to increased light intensity. Thus, iCre1355 represents an experimentally validated genome-scale reconstruction of C. reinhardtii metabolism that should serve as a useful resource for studying the metabolic processes of this and related microalgae.

Keywords: Chlamydomonas reinhardtii; constraint-based analysis; flux balance analysis; lipid accumulation; metabolic modeling; photosynthesis; systems biology.

Figure 1. Distribution of genes and reactions in iCre1355

(A) Subsystem distribution of 312 newly identified metabolic genes incorporated in iCre1355. (B) Distribution of the subcellular localization of reactions associated with the 312 newly identified metabolic genes. (C) Subsystem distribution of all reactions in iCre1355. (D) Subcellular localization of all reaction in iCre1355. (E) Subsystem distribution of blocked reactions in iCre1355. Blocked reactions were identified using flux variability analysis with all exchange reactions made reversible. (F) Venn diagram depicting the overlap in predicted essential genes in iCre1355 under photoautotrophic, mixotrophic and heterotrophic conditions.

Figure 2. Pathway-level expression of genes in iCre1355 across conditions

Heatmap depicting the proportion of genes within each pathway in iCre1355 that are highly expressed across 6 groups of experiments. Pathways were grouped into very highly, highly, moderately and lowly expressed categories based on hierarchical clustering.

Figure 3. Assessing the performance of iCre1355

(A) ROC curves comparing the predictions of iCre1355 to iRC1080 for gene deletion phenotype data. * Only 48 out of 81 metabolic mutants could be assessed with iRC1080. iCre1355 was also evaluated with this set of 48 mutants. ** iCre1355 predictions using biomass reactions from iRC1080. (B) Assessment of predicted maximal TAG yield on CO₂ and photon uptake by iCre1355 and iRC1080. * Yields range from 1.33 to 1.43 depending on the number of c subunits (14 to 12 respectively) assumed for ATP synthase. A value of 12 used for this simulation. (C) Comparison of iCre1355 predicted growth rate to observed growth rates from chemostat grown cells. Measured uptake rates for carbon (CO₂ or Acetate), nitrogen (NH₄) and phosphorus (PO₄) were used as constraints. (D) Predicted growth rates when only CO₂ or PO₄ used as initial constraints.

Figure 4. Predicting growth rate and TAG flux during nitrogen starvation

Predicted growth rate (A) and TAG flux (B) during the time course of nitrogen starvation. Predictions were made, with transcriptional data used to set flux capacity bounds, either by maximizing for biomass (E-flux) or maximizing the agreement between gene expression and predicted flux (E-flux+iMAT). The 2 minute time point was an outlier in the TAG flux analysis and thus was omitted to permit better visualization of the other time points.

Figure 5. Impact of nitrogen starvation on pathway activity in iCre1355

Heat map depicting predicted changes in pathway activity across time during nitrogen starvation. Darker shades indicate greater predicted pathway activity based on an increased number of “high” activity state reactions.

Figure 6. Pathway-level impact of change in light intensity on flux capacity in iCre1355

Heatmap depicting predicted changes in pathway activity during the shift from low to a higher light intensity. Pathway-level changes were predicted based on predicted changes in flux capacities of constituent reactions relative to time 0 m.