Identifying metabolic shifts in Crohn's disease using' omics-driven contextualized computational metabolic network models

Abstract

Crohn's disease (CD) is a chronic inflammatory disease of the gastrointestinal tract. A clear gap in our existing CD diagnostics and current disease management approaches is the lack of highly specific biomarkers that can be used to streamline or personalize disease management. Comprehensive profiling of metabolites holds promise; however, these high-dimensional profiles need to be reduced to have relevance in the context of CD. Machine learning approaches are optimally suited to bridge this gap in knowledge by contextualizing the metabolic alterations in CD using genome-scale metabolic network reconstructions. Our work presents a framework for studying altered metabolic reactions between patients with CD and controls using publicly available transcriptomic data and existing gene-driven metabolic network reconstructions. Additionally, we apply the same methods to patient-derived ileal enteroids to explore the utility of using this experimental in vitro platform for studying CD. Furthermore, we have piloted an untargeted metabolomics approach as a proof-of-concept validation strategy in human ileal mucosal tissue. These findings suggest that in silico metabolic modeling can potentially identify pathways of clinical relevance in CD, paving the way for the future discovery of novel diagnostic biomarkers and therapeutic targets.

Figure 1

Schematic Overview of the performed metabolic modeling experiments: (A) On publicly-available RISK transcriptomic dataset and (B) Enteroid cultures that were generated for CD pediatric patients and controls.

Figure 2

Metabolic pathways with an altered flow in patients with Crohn's disease (CD) versus control patients. The Flux Balance Analysis/Random Forest framework outlined in Fig. 1 were applied to the RISK dataset. After obtaining the list of top metabolic reactions that were altered between CD and control groups, these reactions were grouped into "families" (A–D) based on the biological processes they were involved in: (A) ATP transport processes were found to have significant rates of flux. (B) Mevalonate pathway. Several enzymes within the mevalonate pathway were altered in patients with CD, including the exchange of R- mevalonate, hydroxymethylglutaryl coenzyme A reversible mitochondrial transport, transport of R- mevalonate, and R- mevalonate NADP + oxidoreductase (CoA acylating), (C) Fatty acid oxidation. Long-chain-acyl Coenzyme A dehydrogenase, linoleic acid transport, and alpha-linolenoyl-CoA exchange (D) Uridine transport and exchange. For all graphs, the x-axis describes the reaction that is altered between controls (orange) and diseased states (blue). The y-axis shows the flux values generated by RIPTiDe by analyzing the flow of metabolites through an ileal-specific metabolic network reconstruction. The scale of flux values (y-axis) varies with the reactions as the efficiency of different metabolic pathways in generating biomass varies in a given biological system. A Mann–Whitney U test was done to compare reactions that varied between patients with Crohn's disease and controls. The stars (*) are used to flag levels of significance. *p < 0.05, **p < 0.01, ***p < 0.001. AMP: adenosine monophosphate; ATP: adenosine triphosphate; ER: endoplasmic reticulum; NADP + : nicotinamide adenine dinucleotide phosphate CoA: Coenzyme A.

Figure 3

Metabolic pathways with an altered flow in enteroids generated from patients with Crohn's disease (CD) versus control patients: Our Flux Balance Analysis/Random Forest framework was applied to RNA sequencing data from enteroid models generated from n = 16 patients with Crohn's disease and n = 12 healthy controls. After obtaining the list of top metabolic reactions that were altered between the CD and control groups, these reactions were grouped into "families" (A–D) based on the biological processes they were involved in. (A) ROS detoxification: superoxide dismutase and catalase, (B) Glycerophospholipid metabolism: choline phosphatase and the transport of phosphatidylserine. (C) Fatty acid oxidation: RE3121R is involved in pentaenoyl coenzyme A metabolism and palmitoyl coenzyme A hydrolase (D) Sphinganine transport. For all graphs, the x-axis describes the reaction that is altered between controls (orange) and diseased states (blue). The y-axis shows the flux values generated by RIPTiDe by analyzing the flow of metabolites through an ileal-specific metabolic network reconstruction. The scale of flux values (y-axis) varies with the reactions as the efficiency of different metabolic pathways in generating biomass varies in a given biological system. A Mann–Whitney U test was done to compare reactions that were varied between patients with Crohn's disease and controls. The stars (*) are used to flag levels of significance. *p < 0.05, **p < 0.01, ***p < 0.001, ns: not significant. ROS: reactive oxygen species.

Figure 4

Overlap between gene count data and pathways with an altered flow in the RISK dataset and enteroids: (A) Overlap between the reaction "families" that were found to be altered in the RISK dataset and enteroids. After conducting in silico RIPTiDe studies and performing random forest analysis, reactions involved in fatty acid oxidation were found to be altered in both the RISK dataset and the enteroids as detailed in Table 1. (B). Gene overlap between the RISK cohort and the ileal organoids- "enteroids." Gene count analysis showed that 99% of the genes present in enteroids were also present in the RISK dataset. Only 96 genes were found to be unique to the enteroids.

Figure 5

Ileal tissue from inflamed pediatric CD patients vs. non-inflamed have distinct metabolic signatures. (A) A dendrogram depicting relative abundances of the 50 most significant compounds (top to bottom) compared between inflamed ileal tissue (red, left) and non-inflamed ileal tissue (green, right) patients (left to right). Data was obtained with reverse phase liquid chromatography mass spectrometry (LC–MS) in positive ion mode. Detailed LC–MS compound information is detailed in Supplemental Table 2. Supplemental Table 2 clarifies the y-axis labels, which are the feature numbers, the compounds, their elemental formulas, their molecular weight, as their chromatographic retention time. (B) Pathway analysis of pilot data using the mummichog analytic approach showed the enrichment of multiple metabolites including vitamin B3 metabolism, pyrimidine metabolism, nitrogen metabolism and other lipid and amino acid pathways in CD compared to control tissue groups, in alignment with our in silico findings.

Figure 6

Overview of Methods Framework: (A) Risk Stratification and Identification of Immunogenetic and Microbial Markers of Rapid Disease Progression in Children with Crohn’s Disease (RISK) is a prospective inception cohort study, which enrolled 1,276 pediatric patients with IBD at diagnosis at 28 sites in North America between 2008 and 2012. It includes n = 163 patients with ileal Crohn's disease and n = 42 healthy controls. These transcriptomic abundances were overlaid onto Recon3D. Recon 3D is a comprehensive human metabolic network model consisting of three-dimensional (3D) metabolite and protein structure data. (B) Flux Balance Analysis workflow. Reaction Inclusion by Parsimony and Transcript Distribution (RIPTiDe) Flux Balance Analysis algorithm was used to generate flux values for the flow of metabolites through our ileal-specific metabolic network reconstruction. RIPTiDe generates stoichiometric equations for reactions and utilizes the transcriptomic expression of critical enzymes and metabolites to quantify flux within a specific disease state. A growth constraint is built into this algorithm to identify the reactions most efficient in generating biomass. (C) Random Forest Classifier Overview. Discriminating reactions between CD patient subtypes were determined using Random Forests (RF). The RF classifier used multiple classification trees to sample a given dataset and asked a series of T/F questions related to flux values to "learn" which metabolic reaction fluxes distinguished Crohn's disease versus normal tissues.