Simultaneous Integration of Gene Expression and Nutrient Availability for Studying the Metabolism of Hepatocellular Carcinoma Cell Lines

Abstract

How cancer cells utilize nutrients to support their growth and proliferation in complex nutritional systems is still an open question. However, it is certainly determined by both genetics and an environmental-specific context. The interactions between them lead to profound metabolic specialization, such as consuming glucose and glutamine and producing lactate at prodigious rates. To investigate whether and how glucose and glutamine availability impact metabolic specialization, we integrated computational modeling on the genome-scale metabolic reconstruction with an experimental study on cell lines. We used the most comprehensive human metabolic network model to date, Recon3D, to build cell line-specific models. RNA-Seq data was used to specify the activity of genes in each cell line and the uptake rates were quantitatively constrained according to nutrient availability. To integrated both constraints we applied a novel method, named Gene Expression and Nutrients Simultaneous Integration (GENSI), that translates the relative importance of gene expression and nutrient availability data into the metabolic fluxes based on an observed experimental feature(s). We applied GENSI to study hepatocellular carcinoma addiction to glucose/glutamine. We were able to identify that proliferation, and lactate production is associated with the presence of glucose but does not necessarily increase with its concentration when the latter exceeds the physiological concentration. There was no such association with glutamine. We show that the integration of gene expression and nutrient availability data into genome-wide models improves the prediction of metabolic phenotypes.

INTEGRATE: Model-based multi-omics data integration to characterize multi-level metabolic regulation

“…These constraints also reflect the slight differences among the growth medium of the five cells. Similarly to what was done in [49], if the concentration of metabolite X in the medium of cell A is, for instance, 20% higher than in the medium of Cell B, the maximum uptake flux allowed for metabolite X in cell A will be 20% greater than that of cell B.…”

“…As commonly done (e.g., in [39,49,65]), we combined the RNA-seq datasets with the GPR rules through the employment of the Reaction Activity Score (RAS). The RAS is based on the assumption that enzyme isoforms contribute additively to the reaction activity, whereas the least expressed enzyme subunit is limiting.…”

INTEGRATE: Model-based multi-omics data integration to characterize multi-level metabolic regulation

“…These constraints also reflect the slight differences among the growth medium of the five cells. Similarly to what was done in [43], if the concentration of metabolite X in the medium of cell A is, for instance, 20% higher than in the medium of Cell B, the maximum uptake flux allowed for metabolite X in cell A will be 20% greater than that of cell B.…”

Constraint-Based Reconstruction and Analyses of Metabolic Models: Open-Source Python Tools and Applications to Cancer

“…The resulting reaction expression levels are used subsequently to extract a context-specific metabolic model of active reactions from the whole-organism GEM to reflect a phenotypic state specific to cell type and condition, such as disease state or nutrient level. The simplest transcriptome constraints can be applied by setting associated expression levels as the reaction upper bound, as demonstrated in E-flux and other studies ( 47 , 105 , 106 ). Instead of constraining all genes, PRIME is method that adjusts reaction upper bounds of phenotype-associated genes that are correlated with phenotypic data such as growth rate ( 90 ).…”