Systems-based approaches to study immunometabolism

Abstract

Technical advances at the interface of biology and computation, such as single-cell RNA-sequencing (scRNA-seq), reveal new layers of complexity in cellular systems. An emerging area of investigation using the systems biology approach is the study of the metabolism of immune cells. The diverse spectra of immune cell phenotypes, sparsity of immune cell numbers in vivo, limitations in the number of metabolites identified, dynamic nature of cellular metabolism and metabolic fluxes, tissue specificity, and high dependence on the local milieu make investigations in immunometabolism challenging, especially at the single-cell level. In this review, we define the systemic nature of immunometabolism, summarize cell- and system-based approaches, and introduce mathematical modeling approaches for systems interrogation of metabolic changes in immune cells. We close the review by discussing the applications and shortcomings of metabolic modeling techniques. With systems-oriented studies of metabolism expected to become a mainstay of immunological research, an understanding of current approaches toward systems immunometabolism will help investigators make the best use of current resources and push the boundaries of the discipline.

Keywords: GSMM; Immunometabolism; Metabolic modeling; Metabolic techniques; Systems biology.

Fig. 1

Network-based methods complement and add to enrichment-based workflows for interrogating immunometabolism. Top panel: High-throughput data, such as transcriptomics, metabolomics, or systemic CRISPR screens, are used to generate data-driven hypotheses in the form of differentially expressed targets [119, 120]. Pathways and gene sets are knowledge-based representations of shared biological activity derived from established gene ontologies and databases (e.g., GO [121, 122] and KEGG [61]). Subsequent experiments validate and refine the data-driven hypotheses, add mechanistic insight, and may lead to another cycle of high-throughput data collection and analysis. Bottom panel: Network-based approaches augment enrichment-based approaches. Biological networks, such as genome-wide metabolic networks, are generated from the annotated genome of a species of interest together with functional genomic data and computational gap-filling where appropriate [–106]. Network algorithms, such as flux balance analysis for genome-scale metabolic models, integrate global information agnostic of pathway divisions and can therefore predict targets that will not be prioritized based on the workflow described above. Similar to pathway- and gene-set-based approaches, networks organize extant knowledge and allow the contextualization of gathered high-throughput data within extant knowledge bases. However, network approaches may capture systemic effects that are missed by pathway-focused approaches.

Fig. 2

Schematic overview of a constraint-based modeling approach to study metabolism. a Annotated genes from a species of interest are combined with metabolic knowledge bases to generate a draft for the metabolic reactions available to a cell. b This draft is refined based on existing knowledge bases, and computational gap filling [–106] is applied to ensure desired properties, such as the ability to generate ATP from a given substrate. c The product of this phase is a stoichiometric matrix (S) wherein entries are the stoichiometric coefficient of a particular metabolite (row) in a particular reaction (column). Reactions that have only negative or positive entries are exchange reactions that allow metabolite intake into and secretion out of the system (e.g., R5 in the illustrated matrix). d, e Imposing the assumptions of mass balance and of biochemical steady-state (i.e., constant metabolite concentrations) leads to a feasible space of metabolic flux distributions v (mathematically, this is the kernel of the stoichiometric matrix, namely, the solution space of S · ν = 0). The addition of (f) thermodynamic and (g) capacity constraints further restrict the feasible flux distribution space into a convex cone and bounded convex cone, respectively [71, 123, 124]. h The optimization of objective functions is used to detect mechanistically relevant flux distributions (i.e., the assignment of predicted flux for each reaction). The optimization of the pertinent objective (e.g., synthesis of biomass molecules) allows finding the vertices of the convex cone (namely, specific flux distributions) of interest, although the system is often underconstrained, and as a result, the solution to the optimization problem is nonunique. i The final outcome is predicted flux distributions, namely, the assignment of flux values to each reaction, that achieve the optimum of the stated objective, subject to the stated constraints. Often, the optimization function and/or constraints are informed by empirical high-throughput data, such as gene expression of different phenotypes/cells as denoted in the figure.

Fig. 3

Preclinical and clinical applications of genome-scale metabolic models (GSMMs): Cartoon depiction of the current use and potential applications of GSMMs. The left panel represents the preclinical applications of GSMMs in mice and cell lines. The right panel summarizes the clinical applications in humans.