Characterizing metabolism from bulk and single-cell RNA-seq data using METAFlux

Abstract

Cells often alter metabolic strategies under nutrient-deprived conditions to support their survival and growth. Characterizing metabolic reprogramming in the TME (Tumor Microenvironment) is of emerging importance in ongoing cancer research and therapy development. Recent developments in mass spectrometry (MS)-based technologies allow simultaneous characterization of metabolic features of tumor, stroma, and immune cells in the TME. However, they only measure a subset of metabolites and cannot provide in situ measurements. Computational methods such as flux balance analysis (FBA) have been developed to estimate metabolic flux from bulk RNA-seq data and have recently been extended to single-cell RNA-seq (scRNA-seq) data. However, it is unclear how reliable the results are, particularly in the context of tissue TME characterization. To investigate this question and fill the analytical gaps, we developed a computational program METAFlux (METAbolic Flux balance analysis), which extends the FBA framework to infer metabolic fluxes from either bulk or single-cell transcriptomic TME data. We benchmarked the prediction accuracy of METAFlux using the exometabolomics data generated on the NCI-60 cell lines and observed significant improvement over existing approaches. We tested METAFlux in bulk RNA-seq data obtained from various tumor types including those in the TCGA. We validated previous knowledge, e.g., lung squamous cell carcinoma (LUSC) has higher glucose uptake than lung adenocarcinoma (LUAD). We also found a novel subset of LUAD samples with unique metabolic profiles and distinct survival outcome. We further examined METAFlux on scRNA-seq data obtained from coculturing tumor cells with CAR-NK cells and observed high consistency between the predicted and the experimental (i.e., Seahorse extracellular) flux measurements. Throughout our investigation, we discovered various modes of metabolic cooperation and competition between various cell-types in TMEs, which could lead to further target discovery and development.