Current status and applications of genome-scale metabolic models

Abstract

Genome-scale metabolic models (GEMs) computationally describe gene-protein-reaction associations for entire metabolic genes in an organism, and can be simulated to predict metabolic fluxes for various systems-level metabolic studies. Since the first GEM for Haemophilus influenzae was reported in 1999, advances have been made to develop and simulate GEMs for an increasing number of organisms across bacteria, archaea, and eukarya. Here, we review current reconstructed GEMs and discuss their applications, including strain development for chemicals and materials production, drug targeting in pathogens, prediction of enzyme functions, pan-reactome analysis, modeling interactions among multiple cells or organisms, and understanding human diseases.

Fig. 1

A phylogenetic tree of all of the GEMs reconstructed to date at the family level. GEMs for 434, 40, and 117 taxonomic families of bacteria (light blue), archaea (light purple), and eukarya (pink), respectively, are marked in the phylogenetic tree. Organism names are labeled with circles of different colors outside the circular phylogenetic tree, depending on the development methods used: manual, Path2Models [13], AGORA [14], and CarveMe [15]. For manually reconstructed GEMs, the relevant PubMed identifier (PMID) or digital object identifier (DOI) for the latest GEM version is additionally indicated. The phylogenetic tree was prepared as follows. First, organism names were collected from BioModels for the GEMs from Path2Models, from Virtual Metabolic Human (VMH) for AGORA models, and from a GitHub repository (https://github.com/cdanielmachado/embl_gems/blob/master/model_list.tsv) for CarveMe. Next, National Center for Biotechnology Information (NCBI) taxids at the species level and taxonomic lineages for all of the organisms subjected to the GEM reconstruction were obtained from a dataset available (as of May 14, 2019) at the NCBI Taxonomy FTP (ftp://ftp.ncbi.nih.gov/pub/taxonomy/taxdump.tar.gz). Finally, a Newick file for all of the organisms with taxids was subsequently generated using an in-house Python script at the family level, and this file was used to create a phylogenetic tree using iTOL (https://itol.embl.de/) [16]. A phylogenetic tree of GEMs at the species level is available as Additional file 1. A full list of organisms subjected to the GEM reconstruction and preparation of phylogenetic trees is available as Additional file 2

Fig. 2

Applications of GEMs for the production of chemicals and materials, drug targeting in pathogens, the prediction of enzyme functions, and pan-reactome analysis. a. Flux-response analysis using the Escherichia coli GEM iJO1366 [20] has been used to identify gene manipulation targets for the enhanced production of a monomer of an aromatic polymer d-phenyllactic acid in E. coli [122]. The final strain has two additional genes (tyrB and aspC) knocked out from an E. coli base strain XB201T that expresses AroG^fbr, PheA^fbr, and FldH. The final strain produced 1.62 g/L of d-phenyllactic acid, much more than the 0.55 g/L produced by the base strain. b Reconstruction of the Yarrowia lipolytica GEM iYLI647 and its application for the prediction of reaction engineering targets using four different in silico strain-design strategies [123]. c Identification of stage-specific antimalarial drug targets for Plasmodium falciparum using stage-specific GEMs that represent five different life cycle stages [124]. d Reconstruction of a GEM for Acinetobacter baumannii using several databases and its application for the prediction of condition-specific drug targets to combat antibiotic-resistant A. baumannii [125]. e Discovery of new isozyme functions for genes that have been shown to be nonessential in experiments but that were predicted to be essential in a gene essentiality simulation of the E. coli iJO1366 GEM (i.e., false-negative prediction for the aspC gene) [126]. It was found that tyrosine aminotransferase, which is encoded by tyrB (red line), can compensate for the loss of aspartate aminotransferase, encoded by aspC, which catalyzes the conversion of l-aspartate (l-Asp) and α-ketoglutarate (Akg) to oxaloacetate (Oaa) and l-glutamate (l-Glu). gDCW/L grams dry cell weight per liter. f The PROmiscuity PrEdictoR (PROPER) method identifies promiscuous enzymes at a genome-scale in a target organism [127]. Promiscuous functions for all of the genes in the target organism (E. coli) were predicted using the PROPER method and an E. coli GEM from the Model SEED, which identified 98 alternative routes for the biosynthesis of various metabolites. For example, the product of the thiG gene in E. coli was newly found to biosynthesize pyridoxal 5′-phosphate, which is also known to be biosynthesized by the product of the pdxB gene. g Analysis of the pan-reactome and accessory reactome of 410 Salmonella strains spanning 64 serovars using their respective GEMs [128]. Simulation of the GEMs under various nutrient conditions revealed the different catabolic capabilities of the different strains as well as their preferred growth environments. h Analysis of the pan-reactome and accessory reactome of 24 Penicillium species by using their respective GEMs [129]. Hierarchical clustering of the 24 GEMs revealed additional insights into the biosynthetic pathways of secondary metabolites, which successfully differentiated the metabolic clades

Fig. 3

Applications of GEMs for interspecies metabolic interactions and understanding human diseases. a GEM-based simulation of the effects of costless metabolites (i.e., metabolites that have no effects on the producing organism’s growth rate) secreted by at least one of two paired microorganisms on their growth under anaerobic and aerobic conditions [144]. The number of growth-supporting environments was increased as a result of cross-feeding. b Prediction of the metabolites (e.g., short-chain fatty acids [SCFAs]) required or produced by four representative gut microbiota species, Escherichia sp., Akkermansia muciniphila, Subdoligranulum variabile, and Intestinibacter bartlettii, which are known to be affected by the type 2 diabetes (T2D) drug metformin [146]. c Prediction of the metabolites produced by gut microbiota species from malnourished children using community GEMs that describe the metabolism of multiple gut microbiota species [147]. The prediction results were consistent with the children’s plasma metabolite profiles. d Prediction of the suppressed photosynthesis of a potato plant (Solanum tuberosum) upon infection by the plant pathogen Phytophthora infestans, which triggers the plant’s defense responses against pathogen attack through oxygenation of ribulose1,5-bisphosphate (RuBP) and subsequently increases in the intracellular levels of reactive oxygen species (ROS) [148]. Formation of glyceraldehyde-3-phosphate (GAP) and starch were also decreased as a result of the infection. e Identification of metabolic differences between liver cancer stem cells (LCSCs) and non-LCSCs, and of the transcription factors responsible for the metabolic changes, by using GEMs integrated with transcriptome data [149]. f Characterization of the reprogrammed metabolism of the endothelium cells of sepsis patients using a human endothelium GEM, iEC2812 [150]. Context-specific GEMs were created using transcriptome and metabolome data obtained from human umbilical vein endothelial cells (HUVECs) treated with lipopolysaccharide (LPS) and/or interferon-γ (IFN-γ). Simulation of the context-specific GEMs indicated that increased glycan and fatty acid metabolism led to increased glycocalyx shedding and endothelial permeability where there was endothelial inflammation. HPAEC human pulmonary artery endothelial cell, HMVEC human microvascular endothelial cell