Genome Scale-Differential Flux Analysis reveals deregulation of lung cell metabolism on SARS-CoV-2 infection

Abstract

The COVID-19 pandemic is posing an unprecedented threat to the whole world. In this regard, it is absolutely imperative to understand the mechanism of metabolic reprogramming of host human cells by SARS-CoV-2. A better understanding of the metabolic alterations would aid in design of better therapeutics to deal with COVID-19 pandemic. We developed an integrated genome-scale metabolic model of normal human bronchial epithelial cells (NHBE) infected with SARS-CoV-2 using gene-expression and macromolecular make-up of the virus. The reconstructed model predicts growth rates of the virus in high agreement with the experimental measured values. Furthermore, we report a method for conducting genome-scale differential flux analysis (GS-DFA) in context-specific metabolic models. We apply the method to the context-specific model and identify severely affected metabolic modules predominantly comprising of lipid metabolism. We conduct an integrated analysis of the flux-altered reactions, host-virus protein-protein interaction network and phospho-proteomics data to understand the mechanism of flux alteration in host cells. We show that several enzymes driving the altered reactions inferred by our method to be directly interacting with viral proteins and also undergoing differential phosphorylation under diseased state. In case of SARS-CoV-2 infection, lipid metabolism particularly fatty acid oxidation, cholesterol biosynthesis and beta-oxidation cycle along with arachidonic acid metabolism are predicted to be most affected which confirms with clinical metabolomics studies. GS-DFA can be applied to existing repertoire of high-throughput proteomic or transcriptomic data in diseased condition to understand metabolic deregulation at the level of flux.

Fig 1. Reconstruction of biomass equation (VBOF) of SARS-CoV-2 virus from its stoichiometric make-up.

(A) The stoichiometry of various proteins, nucleic acid and their locations in the SARS-CoV-2 virus. The molar composition of constituent molecules in lipid fraction (B), nucleic acid (C) and proteins (D). PC-Phosphatidylcholine; PE-Phosphatidylethanolamine; PS-Phosphatidylserine; PI-Phosphatidylinositol; SM-Sphingomyelin; CS-Cholesterol. (E) Sensitivity analysis of growth rate of SARS-CoV-2 with respect to coefficient (fractional composition) of biomass precursor. The coefficients are varied by ±10% and y-axis shows growth rates simulated by Flux Balance Analysis (FBA).

Fig 2. Generation of context-specific models of SARS-CoV-2 infected cells and normal cells using gene-expression data and tINIT algorithm.

(A) The distribution of TPM normalized gene expression in Mock NHBE cells, SARS-CoV-2 infected NHBE cells, Normal Lung Tissue (From Biopsy) and Infected Tissue (From Biopsy). (B) The Hamming distance between all the generated context specific models based on the differences in reactions in the model. (C) The pathways over-represented while comparison between models. The reactions over-represented in each comparison are present exclusively in one of the models. (D) The agreement of model simulated specific growth rate of the virus with the experimental reported specific growth rate. n.s. stands for non-significant differences as estimated by t-test. The model simulated growth rate was calculated under ±5% uncertainty in biomass composition and uptake rates. The experimental growth rates were derived from Bojkova et al. (Nature), 2020.

Fig 3

(A) and (B) show the metabolic condition for low and high arginine uptake rates respectively. A high arginine uptake increases the flux through NOS (Nitric Oxide Synthase) pathway which scavenges oxygen away from aerobic metabolism and impairs viral growth. (C) and (D) show the metabolic states for high and low lysine uptake. A low lysine uptake reduces the flux through lysine degradation pathway (Saccharopine pathway) and reduces the supply for Acetyl-CoA required for viral growth. (E) The robustness of specific growth rate of SARS-CoV-2 with respect to changes in specific uptake rates of arginine and lysine. Refer to S3 Fig for robustness analysis with respect to specific uptake rates of all nutrients. Fig 3A–3D were created using BioRender.com.

Fig 4. Boolean relation between activity of enzymes and the gene encoding the proteins in the complex makes the prediction of flux from expression difficult.

Gene A and Gene B encodes subunits of the enzyme which catalyzes ‘Reaction 1’. Gene C and Gene D encode two alternative isozymes which catalyzes ‘Reaction 2’. The width of the arrow for ‘Reaction 1’ and ‘Reactions 2’ represents the level of flux. The black arrow is in the direction of cause to effect. A fold change in the expression of A is sufficient to change the flux through reaction 1 which further changes through reaction 2. All the enzymes are assumed to have their concentration above the threshold. (Created with BioRender.com).

Fig 5. The pipeline for GS-DFA beginning from RNA-seq of infected and normal cell line, followed by construction of context specific models, flux sampling, reaction filtering and over-representation analysis of pathways.

The ‘star’ symbol indicates a multiple hypothesis correction step through Benjamini-Hochberg False Discovery Rate. A two-sample Kolmogorov-Smirnov test is used to differentiate between the probability distribution of flux between diseased and non-diseased state.

Fig 6. GS-DFA reveals altered pathways in infected cells and post-translational modifications in infected cells are enriched in altered reactions.

(A) Enriched pathways (subsystems) in the SARS-CoV-2 infected NHBE cells as revealed by GS-DFA analysis. The presented reactions have flux change (as described in materials and methods) greater than 0.82 and have adjusted p-value less than 0.05 (B) Some of the pathways altered in the SARS-CoV-2 infected cells that could have therapeutic relevance (Created with BioRender.com).

Fig 7

(A) A pathway wise description with the flux changes of the constituent altered reactions in the pathways along with the number of such reactions affected by differential phosphorylation and protein-protein interactions with viral proteins. (B) The prospective mechanism of flux regulation by post-translational modification or allosteric effects on enzymes mediated by viral proteins. (C) The enrichment of differentially phosphorylated enzymes and enzymes interacting with viral proteins in the set of altered reactions identified by GS-DFA.