Analysis of human metabolism by reducing the complexity of the genome-scale models using redHUMAN

Genome-scale metabolic models in translational medicine: the current status and potential of machine learning in improving the effectiveness of the models

The genome-scale metabolic model (GEM) has emerged as one of the leading modeling approaches for systems-level metabolic studies and has been widely explored for a broad range of organisms and applications. Owing to the development of genome sequencing technologies and available biochemical data, it is possible to reconstruct GEMs for model and non-model microorganisms as well as for multicellular organisms such as humans and animal models. GEMs will evolve in parallel with the availability of biological data, new mathematical modeling techniques and the development of automated GEM reconstruction tools. The use of high-quality, context-specific GEMs, a subset of the original GEM in which inactive reactions are removed while maintaining metabolic functions in the extracted model, for model organisms along with machine learning (ML) techniques could increase their applications and effectiveness in translational research in the near future. Here, we briefly review the current state of GEMs, discuss the potential contributions of ML approaches for more efficient and frequent application of these models in translational research, and explore the extension of GEMs to integrative cellular models.

PGE2 inhibits TIL expansion by disrupting IL-2 signalling and mitochondrial function

Expansion of antigen-experienced CD8+ T cells is critical for the success of tumour-infiltrating lymphocyte (TIL)-adoptive cell therapy (ACT) in patients with cancer1. Interleukin-2 (IL-2) acts as a key regulator of CD8+ cytotoxic T lymphocyte functions by promoting expansion and cytotoxic capability2,3. Therefore, it is essential to comprehend mechanistic barriers to IL-2 sensing in the tumour microenvironment to implement strategies to reinvigorate IL-2 responsiveness and T cell antitumour responses. Here we report that prostaglandin E2 (PGE2), a known negative regulator of immune response in the tumour microenvironment4,5, is present at high concentrations in tumour tissue from patients and leads to impaired IL-2 sensing in human CD8+ TILs via the PGE2 receptors EP2 and EP4. Mechanistically, PGE2 inhibits IL-2 sensing in TILs by downregulating the IL-2Rγc chain, resulting in defective assembly of IL-2Rβ-IL2Rγc membrane dimers. This results in impaired IL-2-mTOR adaptation and PGC1α transcriptional repression, causing oxidative stress and ferroptotic cell death in tumour-reactive TILs. Inhibition of PGE2 signalling to EP2 and EP4 during TIL expansion for ACT resulted in increased IL-2 sensing, leading to enhanced proliferation of tumour-reactive TILs and enhanced tumour control once the cells were transferred in vivo. Our study reveals fundamental features that underlie impairment of human TILs mediated by PGE2 in the tumour microenvironment. These findings have therapeutic implications for cancer immunotherapy and cell therapy, and enable the development of targeted strategies to enhance IL-2 sensing and amplify the IL-2 response in TILs, thereby promoting the expansion of effector T cells with enhanced therapeutic potential.

Overexpression of CD2/CD27 could inhibit the activation of nitrogen metabolism pathways and suppress M2 polarization of macrophages, thereby preventing brain metastasis of breast cancer

OBJECTIVE

Our study aimed to reveal the possible molecular mechanisms of CD2 and CD27 in influencing the tumor microenvironment of breast cancer (BC) brain metastasis based on the TCGA (The Cancer Genome Atlas) and SRA (Sequence Read Archive) databases.

METHODS

We calculated the proportions of tumor-infiltrating immune cells and the immune and stromal cell scores in 1222 BC samples from the TCGA-BRCA dataset, followed by identification of candidate DEGs. We further screened for BC brain metastasis-related DEGs in the BC brain metastasis dataset SUB12911144 from the SRA database. Finally, we established a mouse breast cancer brain metastasis model for in vivo validation.

RESULTS

We further screened two immune-regulatory DEGs (CD2 and CD27). GSEA analysis showed that the downregulation of CD2 and CD27 expression was closely related to the activation of nitrogen metabolism pathways. CIBERSORT algorithm analysis showed a correlation between the expression of 16 types of tumor-infiltrating immune cells and CD2 and 19 types of tumor-infiltrating immune cells and CD27. In addition, CD2 and CD27 expression were negatively associated with the proportion of M2 macrophages. In vivo experimental results demonstrated that overexpression of CD2/CD27 could suppress the M2 polarization of macrophages and inhibit breast cancer brain metastasis.

CONCLUSION

In the tumor microenvironment, overexpression of CD2/CD27 inhibited the activation of nitrogen metabolism pathways and suppressed M2 polarization of macrophages, thereby preventing brain metastasis of breast cancer.

Dynamic partitioning of branched-chain amino acids-derived nitrogen supports renal cancer progression

Metabolic reprogramming is critical for tumor initiation and progression. However, the exact impact of specific metabolic changes on cancer progression is poorly understood. Here, we integrate multimodal analyses of primary and metastatic clonally-related clear cell renal cancer cells (ccRCC) grown in physiological media to identify key stage-specific metabolic vulnerabilities. We show that a VHL loss-dependent reprogramming of branched-chain amino acid catabolism sustains the de novo biosynthesis of aspartate and arginine enabling tumor cells with the flexibility of partitioning the nitrogen of the amino acids depending on their needs. Importantly, we identify the epigenetic reactivation of argininosuccinate synthase (ASS1), a urea cycle enzyme suppressed in primary ccRCC, as a crucial event for metastatic renal cancer cells to acquire the capability to generate arginine, invade in vitro and metastasize in vivo. Overall, our study uncovers a mechanism of metabolic flexibility occurring during ccRCC progression, paving the way for the development of novel stage-specific therapies.

The hidden simplicity of metabolic networks is revealed by multireaction dependencies

Understanding the complexity of metabolic networks has implications for manipulation of their functions. The complexity of metabolic networks can be characterized by identifying multireaction dependencies that are challenging to determine due to the sheer number of combinations to consider. Here, we propose the concept of concordant complexes that captures multireaction dependencies and can be efficiently determined from the algebraic structure and operational constraints of metabolic networks. The concordant complexes imply the existence of concordance modules based on which the apparent complexity of 12 metabolic networks of organisms from all kingdoms of life can be reduced by at least 78%. A comparative analysis against an ensemble of randomized metabolic networks shows that the metabolic network of Escherichia coli contains fewer concordance modules and is, therefore, more tightly coordinated than expected by chance. Together, our findings demonstrate that metabolic networks are considerably simpler than what can be perceived from their structure alone.

Machine learning-guided evaluation of extraction and simulation methods for cancer patient-specific metabolic models

Genome-scale metabolic model (GEM) has been established as an important tool to study cellular metabolism at a systems level by predicting intracellular fluxes. With the advent of generic human GEMs, they have been increasingly applied to a range of diseases, often for the objective of predicting effective metabolic drug targets. Cancer is a representative disease where the use of GEMs has proved to be effective, partly due to the massive availability of patient-specific RNA-seq data. When using a human GEM, so-called context-specific GEM needs to be developed first by using cell-specific RNA-seq data. Biological validity of a context-specific GEM highly depends on both model extraction method (MEM) and model simulation method (MSM). However, while MEMs have been thoroughly examined, MSMs have not been systematically examined, especially, when studying cancer metabolism. In this study, the effects of pairwise combinations of three MEMs and five MSMs were evaluated by examining biological features of the resulting cancer patient-specific GEMs. For this, a total of 1,562 patient-specific GEMs were reconstructed, and subjected to machine learning-guided and biological evaluations to draw robust conclusions. Noteworthy observations were made from the evaluation, including the high performance of two MEMs, namely rank-based ‘task-driven Integrative Network Inference for Tissues’ (tINIT) or ‘Gene Inactivity Moderated by Metabolism and Expression’ (GIMME), paired with least absolute deviation (LAD) as a MSM, and relatively poorer performance of flux balance analysis (FBA) and parsimonious FBA (pFBA). Insights from this study can be considered as a reference when studying cancer metabolism using patient-specific GEMs.

Thermodynamic Genome-Scale Metabolic Modeling of Metallodrug Resistance in Colorectal Cancer

Simple Summary

Cancer, but also its treatment, can lead to a reprogramming of cellular metabolism. These changes are observable in metabolite abundances, which can be unbiasedly measured via mass spectrometry metabolomics. However, even when the metabolome changes strongly, a (mechanistic) interpretation is difficult as metabolite levels do not necessarily directly correspond to pathway activities. Here we measure the changes of the cellular metabolome in colorectal cancer cell lines sensitive and resistant to the ruthenium-based drug BOLD-100/KP1339 and the platinum-based drug oxaliplatin. We map these changes onto a cancer-specific genome-scale metabolic model, which allows us not only to compute intracellular flux distributions, but also to disentangle drug-specific effects from growth differences from differences in metabolic adaptations due to resistance. Specifically, we find that resistance to BOLD-100/KP1339 induces more extensive reprogramming than oxaliplatin, especially with respect to fatty acid and amino acid metabolism.

Abstract

Background: Mass spectrometry-based metabolomics approaches provide an immense opportunity to enhance our understanding of the mechanisms that underpin the cellular reprogramming of cancers. Accurate comparative metabolic profiling of heterogeneous conditions, however, is still a challenge. Methods: Measuring both intracellular and extracellular metabolite concentrations, we constrain four instances of a thermodynamic genome-scale metabolic model of the HCT116 colorectal carcinoma cell line to compare the metabolic flux profiles of cells that are either sensitive or resistant to ruthenium- or platinum-based treatments with BOLD-100/KP1339 and oxaliplatin, respectively. Results: Normalizing according to growth rate and normalizing resistant cells according to their respective sensitive controls, we are able to dissect metabolic responses specific to the drug and to the resistance states. We find the normalization steps to be crucial in the interpretation of the metabolomics data and show that the metabolic reprogramming in resistant cells is limited to a select number of pathways. Conclusions: Here, we elucidate the key importance of normalization steps in the interpretation of metabolomics data, allowing us to uncover drug-specific metabolic reprogramming during acquired metal-drug resistance.

Modular dynamic biomolecular modelling with bond graphs: the unification of stoichiometry, thermodynamics, kinetics and data

Renewed interest in dynamic simulation models of biomolecular systems has arisen from advances in genome-wide measurement and applications of such models in biotechnology and synthetic biology. In particular, genome-scale models of cellular metabolism beyond the steady state are required in order to represent transient and dynamic regulatory properties of the system. Development of such whole-cell models requires new modelling approaches. Here, we propose the energy-based bond graph methodology, which integrates stoichiometric models with thermodynamic principles and kinetic modelling. We demonstrate how the bond graph approach intrinsically enforces thermodynamic constraints, provides a modular approach to modelling, and gives a basis for estimation of model parameters leading to dynamic models of biomolecular systems. The approach is illustrated using a well-established stoichiometric model of Escherichia coli and published experimental data.

TCA Cycle and Its Relationship with Clavulanic Acid Production: A Further Interpretation by Using a Reduced Genome-Scale Metabolic Model of Streptomyces clavuligerus

Streptomyces clavuligerus (S. clavuligerus) has been widely studied for its ability to produce clavulanic acid (CA), a potent inhibitor of β-lactamase enzymes. In this study, S. clavuligerus cultivated in 2D rocking bioreactor in fed-batch operation produced CA at comparable rates to those observed in stirred tank bioreactors. A reduced model of S. clavuligerus metabolism was constructed by using a bottom-up approach and validated using experimental data. The reduced model was implemented for in silico studies of the metabolic scenarios arisen during the cultivations. Constraint-based analysis confirmed the interrelations between succinate, oxaloacetate, malate, pyruvate, and acetate accumulations at high CA synthesis rates in submerged cultures of S. clavuligerus. Further analysis using shadow prices provided a first view of the metabolites positive and negatively associated with the scenarios of low and high CA production.

Reaction lumping in metabolic networks for application with thermodynamic metabolic flux analysis

Thermodynamic metabolic flux analysis (TMFA) can narrow down the space of steady-state flux distributions, but requires knowledge of the standard Gibbs free energy for the modelled reactions. The latter are often not available due to unknown Gibbs free energy change of formation [Formula: see text], of metabolites. To optimize the usage of data on thermodynamics in constraining a model, reaction lumping has been proposed to eliminate metabolites with unknown [Formula: see text]. However, the lumping procedure has not been formalized nor implemented for systematic identification of lumped reactions. Here, we propose, implement, and test a combined procedure for reaction lumping, applicable to genome-scale metabolic models. It is based on identification of groups of metabolites with unknown [Formula: see text] whose elimination can be conducted independently of the others via: (1) group implementation, aiming to eliminate an entire such group, and, if this is infeasible, (2) a sequential implementation to ensure that a maximal number of metabolites with unknown [Formula: see text] are eliminated. Our comparative analysis with genome-scale metabolic models of Escherichia coli, Bacillus subtilis, and Homo sapiens shows that the combined procedure provides an efficient means for systematic identification of lumped reactions. We also demonstrate that TMFA applied to models with reactions lumped according to the proposed procedure lead to more precise predictions in comparison to the original models. The provided implementation thus ensures the reproducibility of the findings and their application with standard TMFA.

Quantifying the roles of space and stochasticity in computer simulations for cell biology and cellular biochemistry

Most of the fascinating phenomena studied in cell biology emerge from interactions among highly organized multimolecular structures embedded into complex and frequently dynamic cellular morphologies. For the exploration of such systems, computer simulation has proved to be an invaluable tool, and many researchers in this field have developed sophisticated computational models for application to specific cell biological questions. However, it is often difficult to reconcile conflicting computational results that use different approaches to describe the same phenomenon. To address this issue systematically, we have defined a series of computational test cases ranging from very simple to moderately complex, varying key features of dimensionality, reaction type, reaction speed, crowding, and cell size. We then quantified how explicit spatial and/or stochastic implementations alter outcomes, even when all methods use the same reaction network, rates, and concentrations. For simple cases, we generally find minor differences in solutions of the same problem. However, we observe increasing discordance as the effects of localization, dimensionality reduction, and irreversible enzymatic reactions are combined. We discuss the strengths and limitations of commonly used computational approaches for exploring cell biological questions and provide a framework for decision making by researchers developing new models. As computational power and speed continue to increase at a remarkable rate, the dream of a fully comprehensive computational model of a living cell may be drawing closer to reality, but our analysis demonstrates that it will be crucial to evaluate the accuracy of such models critically and systematically.

A Graph-Based Framework for Multiscale Modeling of Physiological Transport

Trillions of chemical reactions occur in the human body every second, where the generated products are not only consumed locally but also transported to various locations in a systematic manner to sustain homeostasis. Current solutions to model these biological phenomena are restricted in computability and scalability due to the use of continuum approaches in which it is practically impossible to encapsulate the complexity of the physiological processes occurring at diverse scales. Here, we present a discrete modeling framework defined on an interacting graph that offers the flexibility to model multiscale systems by translating the physical space into a metamodel. We discretize the graph-based metamodel into functional units composed of well-mixed volumes with vascular and cellular subdomains; the operators defined over these volumes define the transport dynamics. We predict glucose drift governed by advective-dispersive transport in the vascular subdomains of an islet vasculature and cross-validate the flow and concentration fields with finite-element-based COMSOL simulations. Vascular and cellular subdomains are coupled to model the nutrient exchange occurring in response to the gradient arising out of reaction and perfusion dynamics. The application of our framework for modeling biologically relevant test systems shows how our approach can assimilate both multi-omics data from in vitro-in vivo studies and vascular topology from imaging studies for examining the structure-function relationship of complex vasculatures. The framework can advance simulation of whole-body networks at user-defined levels and is expected to find major use in personalized medicine and drug discovery.