The history of Biology abounds in discoveries where the integral of precise theory and precise experimentation, the trade mark of the ISGSB, has made the difference. These discoveries include no less than the structure of DNA and the mechanisms underlying biological free-energy transduction and the multiplicity of oncogenes. This opening lecture of the 2024 International Study Group on Systems Biology (ISGSB) will sketch the essence and history of ISGSB/BTK. The essence will be identified as ISGSB’s intensive and informal discussions of controversial issues in biology. The latter will be described not in terms of numbers, but in terms of leaps in (my) understanding, produced by ISGSB’s core methodologies.
NAD is a vital coenzyme participating in a multitude of metabolic reactions. Moreover, it serves as a signaling molecule to mediate fundamental cellular processes including DNA repair, cell cycle progression, transcriptional, epigenetic and metabolic regulation. In these processes, NAD is cleaved to liberate nicotinamide, and the ADP-ribosyl moiety is used to perform protein or nucleic acid modifications or to generate messenger molecules. To maintain cellular NAD levels, the released nicotinamide is recycled into NAD synthesis through the salvage pathway.
Here, we aimed to understand the potential interaction between different subcellular NAD pools with a main interest in mitochondria. These organelles represent a major pool with the highest concentration of the dinucleotide. Following targeted overexpression of an NAD consumer in a variety of subcellular compartments we measured NAD turnover using stable isotope-labeled precursors. Remarkably, turnover was hardly affected by the induced increase of NAD-consuming activity, irrespective of its subcellular expression. Accordingly, no upregulation of NAD synthesis was observed and therefore, NAD levels were chronically decreased to limit NAD consumption to the original value. We hypothesize that these observations might provide a mechanistic background for age-dependent cellular NAD decline.
The mitochondrial NAD pool is known to have a certain degree of autonomy, and this has been linked to the mitochondrial localization of NMNAT3, an enzyme catalyzing the reversible, final step of NAD formation from NMN and ATP. However, NMNAT3 is dispensable in mice, and the recent identification of SLC25A51, or MCART1, as a mitochondrial NAD+ transporter, seems to finally have settled the question regarding the establishment and maintenance of the mitochondrial NAD pool in mammals. Based on a large set of experiments including genetically modified cell lines, we propose a key role of NMNAT3 in the mechanism how mitochondria maintain a balanced NAD pool. We posit that the reversible cleavage of imported NAD into NMN and ATP establishes an equilibrium between NAD and NMN (and ATP) that can be shifted to either side upon demand. For example, high cytosolic NAD increases mitochondrial uptake and subsequent cleavage of NAD, whereas high NAD consumption activity would favor NAD synthesis. This mechanism would establish a buffer to compensate fluctuations in mitochondrial NAD concentrations. Moreover, in concert with reversible NAD exchange through SLC25A51, this buffer would be functional for NAD pools in other subcellular locations. Indeed, we demonstrate that the mitochondrial NAD pool is “tapped” when the NAD consumer is overexpressed outside mitochondria. We conclude that subcellular NAD pools are interconnected with a major role of mitochondria in maintaining the cellular homeostasis of this coenzyme.
Lactococcus cremoris is a lactic acid bacterium that is used in dairy applications such as a strarter for cheese production, where lactate acidifies the milk as the key product produced from sugars. The central metabolism of L. cremoris has been studied intensively as a model system, in particular because of its interesting metabolic switching behaviour: Under sugar excess - or more precisely, high glycolytic flux conditions - lactate is produced by homolactic fermentation. However, when glucose is limited (e.g. in a glucose limited chemostat), or in the presence of a “slow” sugar, mixed acid fermentation occurs with formate, acetate and ethanol as products.
Since enzyme concentrations hardly vary in the chemostat when the switch occurs, metabolic regulation is believed to underlie the shift. However, a satisfying and unifying explanation of the switch, and how glycolysis is regulated in this bacterium remains elusive, despite a number of kinetic models in the literature. We revisited these models and designed a kinetic model of glycolysis with a focus on growth-associated ATP supply and demand, phosphate homeostasis and enzyme kinetics that captures the most important regulatory mechanisms known from literature. Our model reproduces the metabolic shift and is able to reproduce and explain many different experimental results regarding to the control and regulation of the glycolytic flux and its branches.
The acetylation of lysine residues in histones are an important regulatory mechanism. But still relatively little is known about the dynamics of histone acetylation and their dependency on metabolic processes. We therefore developed a stable isotope labelling approach based on ^13^C-glucose that allows the quantitative analysis of histone acetylation dynamics in human cell lines. The substrate for histone acetylation is acetyl-CoA which first needs to be synthesized from ^13^C-glucose causing a delay in the label incorporation of histone-acetyl-lysins. To correctly determine histone-acetylation dynamics we therefore simultaneously extracted proteins and metabolites and measured the time-dependent incorporation of ^13^C into both acetyl-CoA and acetylated histone peptides. We than used an ODE based modelling approach to calculate the correct histone acetylation and deacetylation rates. We show that without correction alterations in metabolic fluxes would erroneously be interpreted as changes in histone acetylation dynamics while our approach allows to discriminate between both processes.
Cell growth relies on anabolism synthesizing precursors for macromolecule biosynthesis from nutrients and from energy equivalents (synthesized from nutrients by catabolism). Since anabolism involves catabolic pathways – i.e. anabolic precursors (e.g. pyruvate) are intermediates of catabolism – catabolism and anabolism are severely intertwined and not readily disentangled. The fact that energy equivalents are then also synthesized by anabolism (e.g. pyruvate synthesis from glucose yields 2 ATP and 2 NADH) obscures the exact contribution of energy supply by catabolism for cell growth. For instance, how much of the ATP needed to make 1 gram biomass is actually supplied by catabolism? When is catabolism providing NAD(P)H in addition to ATP for anabolism? Since genome-scale stoichiometric models (GSSMs) study catabolism and anabolism as entangled processes, we need to develop a method first for separating catabolism and anabolism.
To address this, we developed a computational method, which is general, unbiased and enables us to understand the role of catabolism across diverse microbes. For this contribution, we analysed GSMMs of seven microbial species across 50 growth conditions, utilizing various organic and inorganic carbon, electron, and nitrogen sources. We found that the amount of ATP produced in anabolism varies from none at all to all, and that redox cofactors, such as NAD(P)H, are exchanged between catabolism and anabolism in certain cases.
Given these results, we reasoned that catabolism is “driven” (determined) by anabolism. Accordingly, we expect that the stoichiometry of the net anabolic reaction can be determined prior to that of the catabolic reactions and its energy need (ATP and/or NAD(P)H) subsequently dictates the net stoichiometry of the catabolic reaction. Therefore, we aimed to identify the factors that determine the amount of ATP produced during anabolism. Typically, donating electrons to a terminal electron acceptor yields ATP. Accordingly, the presence of an electron acceptor in the anabolic overall reaction correlates positively with the fraction of ATP produced in anabolism. The need for reduced redox cofactors in anabolism depends on the electron balance within this process. Anabolism converts a carbon substrate into biomass and potentially also into carbon-containing byproducts, with the electron balance of these components dictating the use of electron acceptors or donors. If an electron donor supplied by catabolism and an external electron acceptor cannot exchange electrons, both are consumed in anabolism. The energy carriers that are consumed in anabolism, must be supplied by catabolism using the same carbon and/or energy source.
To conclude, the fraction of energy supplied by catabolism for cell growth varies depending on the nature of the involved energy sources and is determined by the pathway stoichiometry of anabolism.
Abstract Established genome-scale modelling methods primarily predict reaction fluxes while established high-throughput experimental technologies primarily measure molecular species concentrations. This paradoxical situation has arisen because of the problem to implement the nonlinear constraints that represent reaction kinetic rate equations without recourse to convenient yet inaccurate approximations. We present a mathematically elegant and computationally tractable solution to this problem. First we introduce a mathematical reformulation of established knowledge on metabolic reactions and reaction kinetics in matrix vector notation. Then we introduce variational kinetics, a novel approach to satisfy steady-state reaction kinetics at genome scale using novel mathematical and numerical optimisation techniques. We illustrate how this approach may be used to simultaneously optimise over the set of steady-state reaction fluxes, thermodynamically feasible kinetic parameters and kinetically feasible elementary and phenomenological rate laws, with solution times competitive with linear optimisation.
The dry mass in the cytosol of a bacterial cell is composed of molecules of diverse sizes, spanning from tiny metabolites to large ribosomes. The dry mass density determines the diffusion of macro- and small-metabolites and also the catalytic efficiency of enzymes, and thereby determines the reaction fluxes. As a bacterial cell is optimized for fast growth, it must strike a balance between competing factors such as the density and distribution of the constituent molecules of the dry mass when allocating resources for their synthesis. Here we simulate a model cell to investigate how bacterial cells optimize their cytosolic density, accounting for the effects of molecular crowding on metabolic reactions. Our simplified model classifies molecular interactions into two groups: ribosomal reactions involving larger ribosomes and tRNAs, and metabolic reactions involving smaller globular enzymes and metabolites. We find that while higher density enhances encounter rates for metabolic reactions involving small molecules, lower density is preferred for ribosomal reactions involving larger molecules to facilitate better diffusion. Notably, our model predicts that large deviations from the optimal density lead to small reductions in growth rate. Moreover, our model's predictions of optimal density across different growth rates are consistent with the trends observed in experiments, such as the cytosolic density in E. coli cells cultured in both minimal and rich media. In sum, the cytosolic density of bacterial cells is governed by an optimality principle that aims at maximizing cellular efficiency.
Metabolic networks need to meet more requirements than single enzymes, in order to be functional. Aware of the heterogeneity of (tumor) cell populations, we went after this principle and engaged in network based drug design. We thought that by identifying the metabolic potential of individual cells, we could identify which targets could be used to most effectively incapacitate most individuals of a tumor cell population.
We projected mRNA sequence counts obtained for >3000 cells out of a growing tumor-cell population onto the genome-wide metabolic map after converting the numbers to Vmax’s. We used Flux Balance Analysis to predict the pathways the individual cells could be using and thereby their vulnerabilities to potential metabolic drugs. That is at least what we thought we would do.
Much to our surprise however, none of the cells was predicted to be able to grow.
We then considered whether this could be due to the cells being social metabolically, i.e. massively exchanging metabolites, with some cells taking care of the upper part of glycolysis, others the TCA cycle, yet others the lower part of glycolysis. This was a nice and social idea, but apparently not realistic: subdividing the cells into subpopulations and offering metabolites synthesized by one subpopulation as substrates to the others, did not lead to growth of either subpopulation.
In this presentation we shall discuss what explanation of the growth in the absence of mRNA for the metabolic pathways, we did come up with.
And, we discuss how the actual resulting model did identify cholesterol and asparagine synthesis pathways as relevant, though complex, drug targets.
In silico tools such as genome-scale metabolic models have shown to be powerful for metabolic engineering of microorganisms. Saccharomyces pastorianus is a complex aneuploid hybrid between the mesophilic Saccharomyces cerevisiae and the cold-tolerant Saccharomyces eubayanus. This species is of biotechnological importance because it is the primary yeast used in lager beer fermentation and is also a key model for studying the evolution of hybrid genomes, including expression pattern of ortholog genes, composition of protein complexes, and phenotypic plasticity. Here, we created the iSP_1513 GSMM for S. pastorianus CBS1513 to allow top-down computational approaches to predict the evolution of metabolic pathways and to aid strain optimization in production processes. The iSP_1513 comprises 4,062 reactions, 1,808 alleles, and 2,747 metabolites, and takes into account the functional redundancy in the gene-protein-reaction rule caused by the presence of orthologous genes. Moreover, a universal algorithm to constrain GSMM reactions using transcriptome data was developed as a python library and enabled the integration of temperature as parameter. Essentiality data sets, growth data on various carbohydrates and volatile metabolites secretion were used to validate the model and showed the potential of media engineering to improve specific flavor compounds. The iSP_1513 also highlighted the different contributions of the parental sub-genomes to the oxidative and non-oxidative parts of the pentose phosphate pathway. Overall, the iSP_1513 GSMM represent an important step toward understanding the metabolic capabilities, evolutionary trajectories, and adaptation potential of S. pastorianus in different industrial settings.
Peroxiredoxins (Prxs) play central roles in the detoxification of reactive oxygen species. These proteins exist in multiple oligomeric forms, depending on their state of oxidation/reduction. The most common of these states are dimers and decamers, with decamers predominating under reduced conditions. Prxs have been modelled across multiple organisms using a variety of kinetic methods. However, their dimer-to-decamer transition has been underappreciated in these studies despite the 100-fold difference in activity between dimers and decamers. This is due to the lack of available kinetics and theoretical framework for modelling this process. Using published isothermal titration calorimetry data, we were able to obtain association and dissociation rate constants for the dimer-decamer transition of human PRDX1. We developed an approach that greatly reduces the number of reactions and species needed to model the peroxiredoxin decamer oxidation cycle. Using these data, we simulated horse radish peroxidase competition and NADPH-oxidation linked assays and found that the dimer-decamer transition had an inhibition-like effect on peroxidase activity. Further, we incorporated this dimer-decamer topology and kinetics into a published and validated in vivo model of PRDX2 in the erythrocyte and found that it almost perfectly reconciled experimental and simulated responses of PRDX2 oxidation state to hydrogen peroxide insult. This allowed us to mechanistically resolve a discrepancy between experimental data and kinetic simulations by showing that reduced Prx sites can be sequestered in a less active dimeric form, thus obviating the need to postulate an "inhibited" form of Prx as had been done in earlier models. Additionally, we have demonstrated that Prx decamer dissociation occurs within a time-frame relevant to peroxidase assays and other oxidation experiments and needs to be considered when working with Prx in a laboratory. Using computational modelling, we were able to to combine and organise different types of experimental data into a single framework to better understand the dynamics of these important antioxidant proteins.
One of the goals of bottom-up systems biology is to generate high-quality predictive models that enhance our understanding of cellular behaviour. For mathematical models of metabolism to accurately simulate experimental data, the conditions under which enzyme parameter values are obtained should closely resemble the actual in vivo environment. Traditionally, this alignment is often lacking, as many enzyme kinetic studies are conducted under optimal conditions for the enzyme, which may significantly differ from the enzyme’s native conditions. A frequently overlooked aspect of the intracellular environment is macromolecular crowding—the influence, through the excluded volume effect, of large quantities of different macromolecules occupying the cell.
To better understand how the complex heterogeneous environment of the cell influences enzyme kinetics, we exposed kinetic assays of various enzymes in the glycolytic pathway of Saccharomyces cerevisiae to inert synthetic polymers of different shapes and sizes at two concentrations, thus mimicking in vivo crowded conditions. Kinetic data were acquired from spectrophotometric assays with microtitre plates or from Nuclear Magnetic Resonance (NMR) spectroscopy time courses. Enzyme kinetic parameters were estimated by fitting initial rate kinetics and NMR time-course data to kinetic models based on rate equations for each enzyme.
The presence of synthetic polymers (Dextran70, Ficoll70, and PEG35) influenced the Vmax and KM-values for different enzymes to varying extents. In some cases, significant changes in kinetic parameters were observed in crowded solutions relative to baseline uncrowded solutions; for instance, high concentrations of crowding agents decreased the Vmax values of numerous enzymes. The changes in kinetic parameters depended on the size and shape of the crowding agent used. Current work focuses on determining the effect of these parameter changes on the kinetic behaviour of the entire pathway network, allowing us to assess the broader impact of macromolecular crowding on the network and its emergent properties under crowded versus dilute conditions.
Cellular constraints and limited resources govern the metabolic strategies of cells to adapt to environmental conditions. Under excess glucose conditions, many yeasts switch from high-yield respiratory metabolism to low-yield fermentation, a phenomenon called the Crabtree effect in yeast, or the Warburg effect in mammalian cells. Which constraints cause this effect is still under debate.
Here we study the Crabtree-negative, fully respiratory yeast Pichia kluyveri and compare it to the Crabtree-positive yeast Saccharomyces cerevisiae from a resource allocation perspective. By integrating quantitative physiology and proteomics into genome-scale proteome-constrained models, we find that the Crabtree effect is determined by the composition of the electron transport chain and is rather sensitive to (often poorly characterized) catalytic constants of mitochondrial enzymes and complexes. This suggests that the “proteome efficiency” - a concept in need of a proper definition that will be addressed - of respiration versus fermentation varies between species. This variation likely reflects evolutionary and ecological history and remains to be explained.
This study advances our understanding of the role of proteome constraints and proteome efficiency in governing cellular metabolism of yeasts, and that of eukaryotic cells at large.
In many cells, thiol-based redox systems are primarily responsible for hydrogen peroxide reduction but also trigger signalling cascades leading to the induction of oxidant-specific transcriptional programs by redox-sensitive transcription factors. Cells lacking these transcription factors are extremely sensitive to oxidative stress, but their constitutive activation is also detrimental. However, the quantitative relationship between these oxidative inputs and transcriptional outputs has remained obscure because we lacked tools to quantify the dynamics of redox signals. Using the fission yeast Pap1 transcription factor pathway as a model, we show how hydrogen peroxide and tert-butyl hydroperoxide triggered quantifiably distinct Pap1 activation profiles and transcriptional responses. Based on these findings, we propose that different oxidants and oxidant concentrations modulate the Pap1 dynamic profile, leading to specific transcriptional responses. We further show how the effect of combination and pre-exposure stresses on Pap1 activation dynamics can be quantified using this approach. Our work advances the concept that redox signalling dynamics are a key aspect of the oxidative stress response.
Christoff Odendaal1$, Ligia Akemi Kiyuna1$, Madhulika Singh2$, Albert Gerding1,3, Miriam Langelaar-Makkinje1, Marianne van der Zwaag4, Asmara Drachman2, Vladimíra Cetkovská2, Gaby Liem Foeng Kioen2, Anne-Claire M.F. Martines1, Nicolette C. A. Huijkman1, Hein Schepers4, Bart van de Sluis1, Dirk-Jan Reijngoud1, Ody C.M. Sibon4, Amy C. Harms2, Thomas Hankemeier2, Barbara M. Bakker1
1Laboratory of Pediatrics, 3Departments of Laboratory Medicine and 4Biomedical Sciences of Cells and Systems, University of Groningen, UMCG, The Netherlands
2Division of Analytical Biosciences, Leiden Academic Centre for Drug Research, Leiden University, The Netherlands.
$ These authors contributed equally
Coenzyme A (CoA) is a vital cofactor that is involved in 8-10% of all metabolic reactions in human cells. In so-called ‘CoA Sequestration, Toxicity, and Redistribution’ (CASTOR) diseases, specific enzyme deficiencies lead to the accumulation of a corresponding CoA ester that is not efficiently metabolised. Common symptoms include acidosis, hypoglycaemia and hyperammonaemia. It has been proposed that a depletion of free, non-esterified CoA (CoASH) is underlying these symptoms, but there is limited direct evidence for this hypothesis. Here, we focus on medium-chain acyl-CoA dehydrogenase deficiency (MCADD), the most prevalent fatty-acid oxidation (mFAO) disorder, in which patients accumulate medium-chain acylcarnitine esters. The aim of this study is to investigate if the loss of MCAD leads to the accumulation of medium-chain acyl-CoA esters, sequestration of CoASH, and remodelling of CoA metabolism.
In agreement with the CASTOR hypothesis, kinetic computational simulations of the mFAO pathway predicted elevated medium-chain C8-acyl-CoA levels and reduced CoASH and short-chain acyl-CoA esters in MCAD knockout (-KO) versus wild-type (WT) hepatocytes. Remarkably, the model predictions were replicated experimentally in MCAD-KO HepG2 cells. Moreover, long-chain acyl-CoA esters, upstream of the deficient enzyme, were also reduced in both the MCAD-KO computational model as well as in MCAD-KO HepG2 cells. According to the model, this may point to a limitation imposed by reduced CoASH, as the generation of new long-chain acyl-CoA esters and their entry into the mFAO pathway also require CoASH. Incorporation of 13C315N1- labelled pantothenate (vitamine B5, the precursor of CoA) showed no difference in the CoA biosynthesis rate between MCAD-KO and WT HepG2 cells. In MCAD-KO mice exposed to severe energetic stress (14h overnight fasting at room temperature followed by 4h fasting at 4°C), however, the total CoA concentration (free plus esterified) was increased. This was accompanied by the upregulation of genes involved in CoA biosynthesis (pantothenate kinases) and CoA release (carnitine acyltransferases and acyl-CoA thioesterases (ACOTs)). Computational simulations showed that the combined effect of elevated CoA and ACOT activity is an effective way to increase free CoASH, while relieving the excessive accumulation of C8-acyl-CoA.
To our knowledge, these results represent the first experimental evidence of the CASTOR mechanism in MCADD. Furthermore, using in vivo and computational models of MCADD, this study provides insights into a potential compensatory remodelling of CoA metabolism, activated under catabolic stress.
Head and neck squamous cell carcinoma (HNSCC), the most common form of head and neck cancer, is diagnosed in almost 900 000 patients annually with a mortality rate of approximately 40% within 5 years of diagnosis. Early diagnosis and effective treatment strategies with limited toxicity are urgently awaited. Alterations in cellular metabolism is one of the hallmarks of cancer and could reveal potential diagnostic options and/or therapeutic targets. We have found that HNSCC cells exhibit a remarkable aerobic fermentation, which we termed the ‘super-Warburg effect’. This means that their lactate production was >2.0 times higher than their glucose consumption, while the full fermentation of one molecule of glucose can maximally yield two lactate molecules.
To study the possible origin of this surplus lactate in a systemic context, we used proteomics from HNSCC cell lines to make context-specific HNSCC models from an existing genome-scale reconstruction of human metabolism (Human1). Flux balance analysis and flux variability analysis revealed anaplerotic amino acids – in addition to glucose – as likely sources of extra carbon. These amino acids enter the Krebs cycle and exit again, for instance as malate, to be decarboxylated to pyruvate, and subsequently to lactate. Our model analysis showed glutamine to be by far the most abundant source of extra carbon, and also showed it to be an obligatory substrate for the cells to grow the at the measured rate with the substrates available in the medium, while maintaining super Warburg lactate production. Malic enzyme 1 (ME1) and serine dehydratase (SDH) were indicated as important nodes in this conversion.
Surprisingly, cultured cells did not take up glutamine from the medium as measured by nuclear magnetic resonance spectroscopy. This could be due to the uptake fluxes simply being undetectably low. One factor causing this might be the availability of alternative amino acid sources. A prominent candidate is the medium component, foetal calf serum (FCS), which contains proteins. It has been shown previously that some cancers can take up proteins by macropinocytosis and convert it into lactate.
We adjusted our model to test the hypothesis that albumin – comprising about two-thirds of FCS protein – could reduce the required glutamine uptake to explain the lack of a detectable glutamine uptake flux in the data. However, the amount of albumin approximated to be present in a growth medium containing 5% of FCS is very low and millimolar changes in glutamine were still necessary to account for the growth rate and lactate production in our simulation. We will further investigate the dependency of the cells on proteins in the medium, and analyse possible flux distributions at low uptake of glutamine. We will also manipulate the genes for ME1 and SDH test their importance in the production of lactate from glutamine.
Adrenal steroids, which include corticosteroids (glucocorticoids like cortisol and mineralocorticoids like aldosterone), and adrenal androgens such as androstenedione (A4), play an important role in regulating electrolyte and water levels in the kidneys. Originally thought of having marginal biological significance, the 11-oxygenated androgens are increasingly recognized as potent steroids with significant roles in human health and disease, particularly in disorders associated with androgen excess or androgen dependence such as polycystic ovary syndrome, congenital adrenal hyperplasia and castration-resistant prostate cancer.
We constructed a detailed mathematical model for the interconversion of the oxygenated androgens (11-hydroxyandrostenedione, 11OHA4; 11-ketoandrostenedione, 11KA4; 11-ketotestoterone, 11KT; 11-hydroxytestosterone, 11OHT) based on in vitro kinetics of the individual enzymes (11-hydroxysteroid dehydrogenase type 2, HSD11B2; aldo-keto reductase type 1C3, AKR1C3; 11-hydroxysteroid dehydrogenase type 1, HSD11B1; and 17-hydroxysteroid dehydrogenase type 2, HSD17B2) and validated the model with experimental data for reconstituted systems with varying enzyme levels at the cellular level, and with inhibitor titrations of HSD11B1 in adipose tissue using the Astra-Zenica drug AZD4017.
We subsequently used the model to analyze clinical data of Chronic Kidney Disease (CKD), and PolyCystic Ovary Syndrome (PCOS) patients, in terms of relative expression levels of the four enzymes, based on plasma concentrations of the oxygenated steroids. For the CKD patients the different disease states could be described by varying HSD11B2 while keeping the other enzymes at the healthy control group values. Interestingly the estimated HSD11B2 levels correlated proportionally to the independent clinically measured eGFR values (estimated Glomerular Filtration Rate), normally used for evaluation of kidney function. The PCOS clinical data could be well described by adapting the AKR1C3 expression (together with a smaller adjustment of HSD17B2, and the implicit total androgen levels). Model simulations showed that an inhibition of AKR1C3 could bring the oxygenated androgens back to wild type levels.
Patients with severe malaria can experience parasitemia levels exceeding 10%, while also suffering from symptoms such as fever, anemia and renal impairment. The symptoms can be attributed to the disease as well as the immune response to disease. Indicative of a poor prognosis, some patients also present with hyperlactatemia and hypoglycemia, attributed to changes to the host metabolism during infection as well as an increase in glycolytic flux through the parasite.
To investigate the effects of disease and parasite metabolism on disease manifestation and metabolic abnormalities a multi-scale model was developed. The model linked two existing models: 1) a within-host disease model that incorporates the immune response and, 2) a whole-body glucose metabolism model that includes parasite metabolism. The models were linked in two steps. The top-down linking of the models used comparable variables and mapped the red blood cell (RBC) populations from the disease model to the metabolic model. The bottom-up linking mapped the parasite ATP production rate to biomass formation in the disease model. A Monte-Carlo simulation was performed and model predictions were compared to clinical data from literature.
Sensitivity analysis on the linked model suggested that processes that affect the disease and immune response have the largest effects on parasitemia and hematocrit, while parasite glycolysis and the total number of RBCs have the largest effects on blood lactate and glucose levels. Additionally, the model also suggested that the innate immune response, and more specifically innate immune cell longevity, has a greater effect on disease outcome than the adaptive immune response.
Seven treatments were added to the model targeting specific processes. Two treatments affect the disease directly by blocking the invasion of healthy RBCs and by reducing the number of merozoites released per bursting iRBC. Two treatments enhance the innate immune response, one by increasing its efficiency and one by increasing its production. The last three treatments all targeted parasite glycolysis by inhibiting the glucose transporter, hexokinase or phosphofructokinase.
Comparison of the treatment effects indicated that targeting the proliferation of merozoites within the infected RBC is most efficient, while glycolytic inhibitors, although less effective against disease itself, led to the best treatment of hyperlactatemia and hypoglycemia.
Infections caused by common human pathogenic fungi result in over 1.5 million deaths worldwide every year. The main source of this problem is the increasing emergence of species resistant to widely used antifungal agents. They emerge due to selection events induced by the over-use of these agents in agriculture and medicine. This issue could be tackled by targeting mechanisms important for fungal pathogenicity and survival. The regulation of carbon metabolic pathways is crucial to adapting cellular processes, like energy production and cellular component synthesis, to changing environmental conditions for fungal cells.
One of the most prevalent human pathogenic fungi is the airborne saprophytic fungus Aspergillus fumigatus. This pathogen can cause infections in the lower respiratory tract, lungs, sinuses and skin. Aspergillus fumigatus uses a complex signalling network system to make changes in carbon metabolic processes when exposed to osmotic or cell wall stress. These pathways are key to infecting the host and surviving in the human organism. Many existing antifungal drugs, like azoles and echinocandins, target biomolecules in these pathways. However, the problem of increasing A. fumigatus resistance to the available drugs is causing a surge of deaths among immunodeficient patients with invasive aspergillosis.
Knowledge about the networks that regulate core cellular processes in A. fumigatus can be used to look for new fungicide targets and help reduce the problem of antifungal drug resistance. Here, we designed two Boolean models that show how carbon metabolism signalling pathways respond to osmotic and cell wall stress in A. fumigatus. Then, we used these models to identify new antifungal drug targets in these networks. Our results suggest that osmotic and cell wall stress both induce the synthesis of carbon-based cell wall components in order to defend against stressors. Moreover, we show how these processes can be disrupted by targeting the RlmA transcription factor under cell wall stress and the SskB kinase under osmotic stress. Our models show that targetting RlmA of the cell wall integrity pathway is a way to inhibit 1,3-beta-D glucan synthase and increase echinocandin drug effectiveness, while targetting SskB of the HOG pathway is a possible way to inhibit fungal cell wall component synthesis.
Parasite growth and metabolism are crucial areas of study in parasitology, particularly concerning the development of effective treatments. We focused on the model protozoan parasite Plasmodium falciparum, the causative agent of malaria, known for its high metabolic activity and reliance on glycolysis for ATP production. This study investigates the impact of Spinosad, a natural insecticide derived from Saccharopolyspora spinosa, on parasite growth, glycolytic flux, and ATP levels. Spinosad has been primarily used against insect pests, but its effects on parasites, especially protozoans, are less understood. This research aims to elucidate the biochemical and physiological responses of P. falciparum to Spinosad treatment.
Compared to untreated controls parasites treated with Spinosad showed a marked decrease in specific growth rate, and in glucose conversion to lactate. The decrease in glycolytic flux suggests that Spinosad might target glycolytic enzymes or regulatory mechanisms within the parasite. To further understand the mechanism underlying these observations, we conducted kinetic assays for the glycolytic enzymes, in Spinosad-treated parasites. Preliminary data suggest a direct inhibitory effect of Spinosad on phosphofructokinase and to a lesser extent glucose-isomerase activity, while none of the other glycolytic enzymes were affected. PFK in yeast and in red blood cells were not inhibited by Spinosad indicated some specificity of the inhibitor for Plasmodium glycolysis. A dose dependent inhibition of PFK by Spinosad indicated a high flux control by the enzyme and confirmed the mechanism of Spinosad inhibition of glycolysis.
In conclusion, Spinosad disrupted glycolytic metabolism, leading to decreased ATP production and impaired parasite growth. These findings highlight the potential of Spinosad as a novel antimalarial agent and underscore the importance of targeting parasite metabolism in the development of new therapeutic strategies. Further studies are warranted to fully elucidate the molecular targets of Spinosad in P. falciparum and to evaluate its efficacy in vivo. This research contributes to the broader understanding of how metabolic interventions can be leveraged to combat parasitic diseases.
In 2022, the ELIXIR Systems Biology community was established within the broader ELIXIR infrastructure to represent and advocate for systems biology within European scientific infrastructures. While ELIXIR focuses on improving the infrastructure for "data for life," the Systems Biology Community is dedicated to "models for life." Over the past couple of years, community-led activities have begun to yield tangible results.
One such results is the publication of a revised whitepaper in June 2024, which outlines the major short- and long-term infrastructural challenges in systems biology. These challenges include identifying and overcoming barriers to the broader adoption of systems biology, making data more model-friendly, ensuring the interoperability of systems biology resources, and enhancing education and training.
The first Systems Biology implementation study has initiated efforts to address some of these challenges. This includes preparing a whitepaper on FAIR PBPK modeling, the integration of two existing systems biology ontologies (EDAM and SBO), annotating (a set of ) systems biology tools in the bio.tools repository, and planning a comprehensive analysis of the interoperability of these tools to be conducted later in 2024.
A second implementation study is currently being prepared, focusing on how AI can enhance various aspects of systems biology. Despite these positive steps over the past two years, many challenges remain. The success of these endeavors will be greatly enhanced by a larger and more engaged community. We invite you to join us in our efforts: https://elixir-europe.org/communities/systems-biology.
Constraint-based modelling and genome scale models (GSM's) are ubiquitous in Systems Biology with applications ranging from agriculture to human health (1). Fundamental to this methodology is the ability to create and exchange models, a process facilitated through the use of the Systems Biology Markup Language (SBML) (2) and its Flux Balance Constraints Package (FBC) extension.
Released in September 2015, FBC version 2 has become the de facto standard for encoding GSM's and is widely used in model repositories (BioModels, BiGG), software (e.g. COBRAPy, CBMPy) and curation pipelines (MEMOTE, FROG). It extends SBML by adding the components necessary for building typical GSM, including a linear objective function, reaction flux bounds and gene-protein-reaction associations (3). However, more recent model types, such as community and macromolecular expression (ME) models, could not be fully encoded in FBC version 2.
To address these shortcomings a working group, including members of both the SBML and constraint-based modelling community, have been working towards a new version of FBC. Recently finalised, the FBC version 3 specification (4) builds on FBC version 2 by allowing the definition of:
- objective functions with mixed quadratic and linear terms, that allows the definition of QP based models,
- user-defined (UD) constraints that are not defined as part of the stoichiometric matrix,
- UD constraints can contain quadratic terms that allow the definition of quadratic constraint (QC) models,
- UD constraint can also contain "artificial" variables that are defined as non-constant parameters,
- species that have chemical formulas with generic terms (e.g. R, X) and non-integer charges,
- KeyValuePairs, a simple, flexible annotation type that supplements the existing SBML annotations.
The FAIR (findable, accessible, interoperable, reusable) data principles form the basis of data management practices that are focussed on the reuse of research data (5). In this context models are also considered research data that should be reusable by yourself and others thus enabling good and reproducible research practices. While much of the focus on FAIR data is on the findability and accessibility, SBML provides a structured data format with built-in support for metadata that is primarily focussed on interoperability. The FBC 3 KeyValuePair annotations introduce support for less metadata that allows for a wider range of information to be stored as model annotation, and documentation, potentially leading to enhanced model reusability.
References
- https://doi.org/10.1038/s41579-019-0264-8
- https://doi.org/10.15252/msb.20199110
- https://doi.org/10.1515/jib-2017-0082
- https://github.com/sbmlteam/sbml-specifications/blob/develop/sbml-level-3/version-1/fbc/spec/sbml-fbc-version-3-release-1.pdf
- https://doi.org/10.1038%2Fsdata.2016.18
LabNexus – An open-source enzyme kinetics data automation web application based on FAIR principles and STRENDA guidelines.
Authors: C.E. de Beer and J.M. Rohwer
In response to the reproducibility crisis in enzymology and the principles of FAIR (Findable, Accessible, Interoperable, and Reusable) data, we present LabNexus, a novel and user-friendly platform for standardizing procedures related to the collection, processing, and storage of both data and metadata with respect to enzyme kinetics experiments.
LabNexus features a fully integrated web interface that automates the corroboration of raw experimental data with corresponding metadata with minimal user input. This includes integrated search tools for cross-referencing external databases such as PubChem, ChEBI, and UniProt, and efficient data processing capabilities. The platform utilizes the Enzyme Markup Language (EnzymeML), a derivative of SBML, to package data and metadata in adherence to STRENDA guidelines. Users can submit various types of data, including EnzymeML documents and raw spectrophotometric data, both of which can be read into the platform with minimal user intervention.
LabNexus is designed to be accessible to all enzyme kinetics researchers, regardless of their programming literacy, ensuring a broader adherence to FAIR principles in published literature. Users can view their submitted data for validation purposes, ensuring accuracy and consistency in their experimental results. The platform's modular design facilitates the expansion of supported spectrophotometric models and output formats. Users can output their virtual documents (referred to as workspaces) in various formats, including EnzymeML, YAML, or Markdown, enhancing the flexibility and utility of the data management ecosystem.
Optionally, LabNexus offers automated synchronization of spectrophotometric instrument output files to the server via a companion application on the instrument host machine, adding redundancy to experiments.
The platform supports operation in both closed and open networks, incorporating network security protocols to safeguard data. In an open network configuration, automated synchronization allows users to access, view, edit, and annotate their data from anywhere in the world via an internet connection.
Systems biology is largely becoming transformed by the fourth industrial revolution; the integration of artificial intelligence and machine-learning within scientific research practice. Parameter estimation within enzyme kinetics entails determining the values of the parameters in a kinetic model that best fit the experimental data. However, experimentally, this process can be laborious, prone to error, and expensive. Therefore, a machine-learning system approach might provide some benefits.
We here address these challenges by implementing the use of Neural Differential Equations (NDEs) for modeling experimental enzymatic time series data, using the phosphoglycerate mutase (PGM) and enolase enzymes in the glycolytic pathway as an example model system. NDEs combine deep learning with differential equations by modeling the change of a neural network’s unknown position repeatedly over time. This allows the handling of convoluted, irregular and time-dependent time series data. This research utilized 52 experimental time-series datasets from our own laboratory. Validation of the NDE models is performed through classic machine learning procedures in order to test integrity and reliability. This includes a 20/80 test/train dataset split, to evaluate model performance, cross validation using k-fold cross validation methods as well as comparisons to established baseline models.
NDEs present numerous advantages in comparison to traditional Bayesian parameter estimation techniques. These include greater efficiency, enhanced scalability for handling large and complex datasets and the ability to accurately represent and capture underlying processes of dynamical systems in a continuous manner.
This research utilizes the application of Catalax, a JAX-based framework developed by Jan Range at the University of Stuttgart, Germany, which supports simulation and parameter inference through Bayesian parameter estimation. Catalax leverages Markov Chain Monte Carlo (MCMC) sampling to infer posterior distributions of model parameters with the inclusion of the training framework for the NDEs.
Many experimental procedures in biochemical laboratory practice are repetitive, time-consuming, and prone to human error due to manual processing. One procedure is the spectrophotometric assay for analysing enzymatic activity using microtiter plates. Currently, transferring experimental data from the instrument to the analysis platform requires manual copying and pasting. Recently, a specialized markup language EnzymeML and the associated PyEnzyme software were created to automate parts of this workflow, specifically fitting models to experimental data and transferring kinetic parameters to databases. However, the programming of the experimental protocol and the transfer of metadata have not yet been automated. Moreover, in enzymology research, scientists are frequently facing reproducibility issues, mainly because crucial metadata are not reported. These can include the precise reaction conditions, comprehensive experimental results, and procedures used for data analysis. Therefore, the aim of this study was to implement and test tools and interfaces to automate aspects of the experimental workflow and associated data management. These interfaces will assist in enabling data to be Findable, Accessible, Interoperable and Reusable (FAIR). The tools are tested in the laboratory by performing enzymatic assays using an OT2 liquid handler to characterize the bi-functional enzyme complex consisting of phosphopantothenoylcysteine synthetase and phophopantothenoylcysteiene decarboxylase (CoaBC), the second and third enzymes in the coenzyme A biosynthesis pathway. In Staphylococcus aureus, CoaBC has not yet been kinetically characterised; this process is extremely challenging due to the complexity of the enzyme system, as it involves multiple substrates, products and sequential reactions. To mitigate this, we aim to link the activity of CoaBC to the activity of the fourth enzyme CoaD, which is responsible for catalysing the conversion of 4’-phosphopantetheine to dephospho-Coenzyme A with the release of pyrophosphate. The kinetics of the first step catalysed by CoaBC and the kinetics of CoaD were investigated by measuring the production of pyrophosphate in an enzyme-linked assay. During the second step of the reaction catalysed by CoaBC, CO2 is released, which cannot be measured spectrophotometrically. Therefore, we aim to perform kinetic assays either with CoaBC on its own, or by combining both enzymes CoaBC and CoaD. In the case where we combine both enzymes, we expect the rate to be twice as fast, since two pyrophosphate molecules are released compared to the case with only CoaBC where only one pyrophosphate is released. Lastly, we construct a kinetic model using PySCeS, where datasets with CoaBC and CoaD and datasets without CoaD are combined and the kinetic parameters for previously obtained for the first step of CoaBC and CoaD are included, to fit the kinetic parameter values for the second step catalysed by CoaBC.
Microbiota plasticity, the ability of microbial communities to adapt to changing environments, is crucial for understanding gut health. We develop a spatiotemporal model to simulate small intestinal microbiota and their interactions with host cells. We begin by constructing a community model of various microbial species using metabolic reconstructions and Flux Balance Analysis (FBA). By optimizing community growth, we investigate species interactions, applying L2-regularization and alternative objective formulations. The next phase incorporates enterocytes into the microbial community model. Utilizing metabolic reconstructions of S. thermophilus, F. prausnitzii, B. caccae, and E. rectale, we simulate interactions under conditions resembling an average Western diet. Our findings highlight significant interactions, including cross-feeding and competition among species. Finally, we expand the model into a spatiotemporal framework, simulating microbial dynamics along the small intestine. These simulations reveal how species abundance varies with distance and time, influenced by community composition and medium conditions.
Despite challenges in parameterization and validation, our model offers insights into the plasticity of small intestinal microbiota and their interactions with enterocytes, enhancing our understanding of gut microbiome dynamics.
Large collaborative projects need to share data during and after, within and beyond the consortium. FAIRDOM-SEEK (https://fairdomseek.org/) is an open-source software for storing, cataloguing, sharing and reusing research outcomes designed to support the principles of FAIR (Findable, Accessible, Interoperable, and Reusable) research data management. Originally developed for the needs of systems biology of microorganisms, SEEK is used in numerous projects of systems biology, systems medicine, and related domains. All data types can be handled and the use of files or references to files is possible. Users can change the visibility of files and references, making it a platform for projects and data publication. Its properties make it an interoperability resource for combining different tools for scientific work and subsequent publication of the outcomes.
The systems medicine approach to quantification and characterization of large complex systems involves integration of multipledata types (e.g. genomics, proteomics, metabolomics, phenomics, images, patient related data, etc.), stored in several specialized systems used within one project.
LiSyM-Cancer for example, uses REDCap (https://www.project-redcap.org/) as a clinical data system that manages information about patients and samples; openBIS (https://openbis.ch/) as primary system for experimental raw data and its metadata; Nextcloud (https://nextcloud.com/) for short-term raw data exchange; and OMERO for microscopic images. The harmonisation and integration of (meta)data between these platforms is mandatory to make the data comparable and publishable in open data repositories.
Here, we describe our experience in combining multiple open-source data repository systems for the benefit of large collaborative projects.
In physical chemistry, Gibbs free energies and chemical potentials (not energies) are the descriptors for the energetics of compounds and processes. They integrate energy, entropy and volume-work. Changes in Gibbs energy and between chemical potentials equal the useful work at constant temperature and pressure. However, this systematics requires one to be explicit about the protonation, Mg-complexation and hydration states of the molecules, and to correct of activity coefficients differing from 1. Moreover the standard chemical potentials are defined for biologically irrelevant situations such as pH=pMg=0, concentrations of 1 Molar, hydrogen gas at 1 atmosphere and crystalline phosphorous. In addition, one must monitor the number of protons and Mg ions liberated or consumed by the reaction. Rather than a single ATP synthesis reaction, this methodology requires describing some 20 reaction variants, explicating the various protonation, Mg-complexation and hydration states of the three molecules involved. Indeed, multiple literature studies have presented highly complex ways of calculating the standard Gibbs energy of this highly important reaction, which then typically differed by 17 kJ/mol from its actual work potential.
We here present new ‘metabolic’ energies and potentials. These should replace the Gibbs free energy and the chemical potential for life processes and metabolites under in-vivo like conditions. We also present a new ‘Thermotable’, which contains the standard metabolic potentials for up to a thousand metabolites of interest to systems biology. The Thermotable enables the direct computation of standard (i.e., concentration-independent) reaction energies by simple subtraction. These are immediately relevant for the in vivo reference conditions of pH=7, pMg=3, ionic strength 0.15 M, and T=310 K, and concentrations of 1 mM. There is no need for correcting concentrations to activities. The new standard reaction energies are very good approximations towards the actual reaction energies, for when concentrations are unknown (yet on the order of 1 mM rather than 1 M).
The metabolic potentials given in the Thermotable are much better descriptors of the energetic potential of biochemicals than the chemical potentials were, because they are taken relative to a growth medium of biological significance, i.e., 1 mM of (total) bicarbonate, ammonium, phosphate, and sulfate, as well as liquid water, H+ at pH7, and Mg2+ at pMg3. This will be illustrated by plotting the metabolic energy landscape for travelling down major metabolic pathways. Whilst the maps of chemical potentials were rugged and irregular, the magnitudes of the potentials are realistic (in terms of numbers of ATP energies) and the new maps are smoothly down-hill except for steps up of about 50 kJ/mol where ATP energy is invested (and down where ATP is made). With the new metabolic energies, the calculation of free energy differences and equilibrium constants becomes facile, omitting the many usual points of confusion in calculating from standard chemical potentials.
The new thermodynamics will be useful for turning metabolic network maps into metabolic energy landscapes and for warning against perpetua mobilia proposed by Flux Balance Analysis.
Low hydrogen peroxide concentrations are essential for eukaryotic cell physiology, but high concentrations trigger an antioxidant response. The redox-dependent activation of specific transcription factors is a critical feature of this response. Deleting these transcription factors increases the sensitivity of cells to hydrogen peroxide stress, but their constitutive activation is also harmful. Curiously, many redox transcription factors require multiple oxidation events for full activation. E. coli OxyR needs four events, mammalian Nrf2-Keap1 requires three, Yap1 in baker’s yeast needs three to four, and Pap1 in S. pombe requires at least two oxidation events. We investigated the purpose of these multiple oxidation events using computational modelling of a basic system and the fission yeast Pap1 system. Our results demonstrated that multiple oxidation steps increased the system’s ability to attenuate signal activation at low hydrogen peroxide concentrations and amplify it at higher peroxide concentrations. This high-pass filtering property, in part, explains how eukaryotic cells can tolerate low hydrogen peroxide levels without triggering an adaptive response.
Metabolic rewiring is observed in almost all cancer types and is considered one of the hallmarks of cancer. The Warburg effect, also named aerobic glycolysis, is characterised by an increased conversion of glucose to lactate and was first observed by Otto Warburg in the 1920s. Studies have shown a correlation between aerobic glycolysis and metastatic properties of cancer. The cancer-specific metabolic rewiring raises the question of whether the flux control distribution over the glycolytic pathway has changed compared to normal cells, and whether such a redistribution could be exploited for drug target identification. Work by Shestov et al. (2016) highlighted glyceraldehyde 3-phosphate dehydrogenase (GAPDH) as a good glycolytic target to perturb aerobic glycolysis in cancer. In a combined experimental and modelling approach we investigated aerobic glycolysis in the MDA-mb-231 cell line - a highly invasive and hormone-independent breast cancer cell line, to study the flux control distribution, particularly the control by GAPDH.
For the construction of a detailed mathematical model for the glycolytic pathway, we kinetically characterised all 12 glycolytic enzymes. Kinetic parameters were obtained by performing a global fit on the enzyme’s initial rate kinetics using the corresponding rate equations. A set of ordinary differential equations was defined with parameterised rate equations based on the kinetic data. The model was successfully tested in its capacity to predict intermediate dynamics upon a pulse of 14C labelled glucose to cell free extracts and in its prediction of the effect of inhibition of GAPDH by iodoacetic acid.
Subsequently, the glucose transporter was characterised, and we are busy integrating the transporter kinetics with the glycolytic enzymes to simulate glucose metabolism in intact cells.
Inhibitor titrations of intact cells revealed that the flux control coefficient of GAPDH is very low, Kouril et al. (2023), making the enzyme a poor therapeutic target in this highly invasive breast cancer cell line. We are currently investigating whether the high flux control observed by Shestov is cell line dependent, by analysing more sensitive cell lines. Once the intact cell model has been validated, we will apply MCA for drug target identification.
Shestov, A.A., Liu, X., Ser, Z., Cluntun, A.A., Hung, Y.P., Huang, L., Kim, D., Le, A., Yellen, G., Albeck, J.G. and Locasale, J.W., 2014. Quantitative determinants of aerobic glycolysis identify flux through the enzyme GAPDH as a limiting step. elife, 3, p.e03342.
Kouril, T., October, C., Hollocks, S., Odendaal, C., van Niekerk, D.D. and Snoep, J.L., 2023. Inhibitor titrations reveal low control of glyceraldehyde 3-phosphate dehydrogenase and high control of hexokinase on glycolytic flux in an aggressive triple-negative breast cancer cell line. Biosystems, 231, p.104969.
Type 2 diabetes (T2D) is a metabolic disease that negatively impacts the health of many individuals worldwide. It accounts for roughly 95% of diabetes cases and was found to be responsible for 6.7 million diabetes related deaths in 2021, with approximately 536.6 million people suffering from this disease globally. Despite the advancements that have been made in our understanding of T2D, the molecular mechanisms underlying this disease remain poorly understood. Therefore, it is necessary to investigate the cellular mechanisms underlying conditions such as insulin resistance and T2D to understand the extent of the metabolic dysfunction. There are three main tissues implicated in T2D; skeletal muscle, adipocytes and the liver. Among these, skeletal muscle is responsible for 75% of insulin-dependent glucose uptake and is therefore important for maintaining glucose homeostasis.
Core mathematical models were constructed for the insulin signalling pathway, glucose transport and glucose metabolism to simulate insulin dependent glucose metabolism in control C2C12 skeletal muscle cells, and in insulin insensitive cells. The model parameters were fitted to experimental data for dose and time dependent data for insulin addition and removal. Analysis of the individual models and their interactions will reveal the effected points leading to insulin insensitivity. Although the mouse cell model might not accurately reflect the T2D disease state in patients, the concept to analyse the point(s) of interference is still applicable.
Diabetes is a growing global epidemic currently affecting 537 million adults worldwide, with 90% of these cases attributed to type 2 diabetes mellitus. Individuals with diabetes suffer from glucose intolerance, hyperinsulinaemia and hyperglycaemia. Additionally, target tissues and organs of insulin have a diminished response to insulin, known as insulin resistance, which results in the dysregulation of insulin signalling. Prolonged insulin resistance, in turn, leads to pancreatic β cell dysfunction leading to decreased insulin production and ultimately the cessation of insulin secretion. Several target tissues and organs are implicated in type 2 diabetes mellitus namely the liver, brain, skeletal muscle and adipose tissue. Skeletal muscle cells are primarily responsible for insulin-dependent glucose uptake and homeostasis. Ergo, in vitro models of skeletal muscle cells often are utilised in studies examining insulin signalling and glucose uptake. In the highly complex insulin signalling pathway, protein kinase B more commonly known as Akt, plays an essential role where its dysregulation has been implicated in several deceases. Specifically, the overactivation of Akt can result in tumour growth, whereas metabolic conditions such as insulin resistance can occur when Akt phosphorylation is impaired. Thus, inhibitors of this enzyme are of interest as a tool for research into Akt phosphorylation and insulin signalling.
Recently, our group has developed a kinetic mechanistic model to describe the dynamics of insulin signalling, glucose uptake and its metabolism in C2C12 mouse skeletal muscle. The activities of these modules were analysed dose- and time-dependently upon induction with insulin. This model serves as the reference state to further study mechanisms leading to insulin resistance as a function of known agents causing type 2 diabetes mellitus.
Accordingly, the effect of Akt inhibition on insulin signalling, glucose uptake and glycolytic flux was investigated in C2C12 mouse skeletal muscle myotubes. The current phase II trial Akt allosteric inhibitor, MK-2206, was utilised. The inhibitory effect of MK-2206 on Akt phosphorylation was analysed using our group’s minimal mathematical model. As before, 100 nM insulin stimulation of untreated cells increased Akt Ser473 and Thr308 phosphorylation dose- and time-dependently. Treatment with MK-2206 decreased the dose-dependent increase of Akt phosphorylation of both Ser473 and Thr308 regulatory sites after stimulation with 100 nM insulin. In addition, MK-2206 treatment delayed the time dynamics of insulin-induced Akt phosphorylation. Model simulations were able to describe the delay and decrease of Akt phosphorylation that was observed experimentally. In control cells, stimulation with 100 nM insulin increased glucose uptake rate and lactate production rate by 2-fold and 1.9-fold, respectively. MK-2206 treatment did not affect the basal rate of glucose uptake or lactate production, but a strong decrease of insulin stimulation of glucose metabolism was observed in MK-2206 treated cells.
Metabolic alterations are associated with many especially age related diseases and they are caused by changes in the expression of the corresponding enzymes. Although analysis of gene expression data has become a standard approach, predicting large scale metabolic alterations is still challenging. Here, we present a mathematical framework GEMCAT, Gene Expression-based Metabolite Centrality Analysis Tool, that is based on the centrality of nodes in a directed graph. Through integration of differential expression data from either transcriptomics or proteomics we can predict metabolic alterations for a large set of metabolites but avoid artificial biomass or energy constraints required for other genome-scale modelling approaches. We demonstrate the predictive efficacy of GEMCAT using two distinct datasets: one involving RNA sequencing data and metabolomics from a human cell line featuring a deleted mitochondrial NAD-transporter, and another comprising proteomics and metabolomics data from patients afflicted with inflammatory bowel disease. We furthermore extended our approach to trace the experimentally confirmed metabolic alterations back to the expression changes enabling network based multi-omics integration.
Three components of sustainability (environmental, economic and societal) of various bioprocesses become increasingly important. Neglecting the sustainability issues may cause adverse environmental and societal problems proportional to the production volumes. Governmental bodies are expected to pay more attention to environmental and societal parameters of industry in near future.
Sustainable metabolic engineering (SME) is defined by Stalidzans and Dace (2021) as optimization of metabolism where economic, environmental and societal sustainability parameters of all incoming and outgoing fluxes and produced biomass of the applied organisms are considered.
Constraint-based stoichiometric modelling framework can be adapted for the calculation of the integrated sustainability score (ISS) that consists of weighed economic (Econ), environmental (Env) and social (Soc) components. ISS= WeconEcon + WenvEnv + Wsoc*Soc.
The calculations are carried out introducing economic and environmental sustainability indicator vectors for all exchange metabolites that have to be multiplied by the metabolic flux vector of all exchange fluxes accounting for the contribution of each exchange metabolite. The economic indicator vector (Econ) contains prices of substrates, products and by-products. The environmental vector (Env) contains characteristics of environmental impact parameters of substrates, products, by-products and biomass. The value of the social component (Soc) is calculated for the whole design of the organism (does not depend directly on the values of exchange fluxes). Social component takes into account characteristics like health and safety risks, newly created working places (depend on the estimated financial turnover), social acceptance of genetic engineering methods and others.
Growth-coupled production approach where the production of the target metabolite is coupled with the production of biomass has been selected as an option to implement SME. In case of SME the biomass production is coupled with the ISS that can be introduced as an objective function. An important side-effect of growth-coupling is the reduction of exchange flux variability that reduces also the variability of ISS.
OptGene tool is adapted for the implementation of ISS and applied for deletion-based design development. Sustainable designs of E.coli and some other organisms are proposed and compared indicating advantages and disadvantages of designs for production of particular metabolites pointing at some patterns in sustainability analysis. We have observed that the environmental and societal sustainability components can be improved without harming the economic parameters of target metabolite production.
Currently SME can be used as a tool for preliminary selection of appropriate organisms and designs to study them in more detail. Further development of SME approach depends on the availability of suitable data for environmental assessment of metabolite production and consumption impact on the environment.
An integrated systems toxicological approach was taken to identify physiologically relevant biomarkers of perfluorooctane sulfonate (PFOS) toxicity in fish. PFOS is a ubiquitous pollutant in global aquatic ecosystems with increasing concern for its toxicity to aquatic wildlife. An in silico stoichiometric metabolism model of zebrafish was used to integrate available metabolomics and transcriptomics datasets from in vivo toxicological studies with 5 days post fertilized embryo-larval zebrafish. The experimentally derived omics datasets were used as constraints to parameterize the in silico zebrafish model. In silico simulations using flux balance analysis (FBA) showed prominent effects of PFOS exposure on the carnitine shuttle and fatty acid oxidation. Further analysis of impacted metabolites indicated carnitine to be the most highly represented cofactor metabolite. Taken together, our results showed dyslipidemia effects under PFOS exposure and uniquely identified carnitine as a candidate metabolite biomarker. Subsequently, verification of this prediction was sought through an in vivo environmental monitoring study which showed carnitine to be a modal biomarker of PFOS exposure in wild-caught fish and marine mammals sampled from the northern Gulf of Mexico. Therefore, we highlight the efficacy of FBA to integrate multi-omics datasets to study the properties of large-scale metabolic networks and identify biomarkers of exposure and likely adverse effects.
Chinese Hamster Ovary (CHO) cells are cell factories for a variety of pharmaceutical outputs, namely monoclonal antibody production and biotherapeutics. Having been utilised for the synthesis of biological compounds for over a century, CHO cells facilitate a wide range of post-translational modifications, including glycosylation, making them a popular choice for human drug development. Furthermore, CHO cells can tolerate shifts in their culture conditions, such as pH, oxygen levels, temperature and cell density. These characteristics can be predicted and explained using constraint-based modelling, which through the integration of ‘omics data, invites the simulation of changing environmental conditions favouring improved productivity.
Here, the recently published iCHO2441 genome-scale model (GEM) (Strain et al, 2023) has been modified to closely match a specific cell line used for production by FUJIFILM Diosynth Biotechnologies (FDB), as part of our industry partnership. A wealth of time-course ‘omics data (transcriptomics, proteomics and metabolomics) has been generated to enable constraint-based modelling and provide opportunities for creative model inputs and validations. Initial work has seen the flux sampling of models constrained using transcriptomics data to represent specific time points over a two-week culture period. Flux sampling for bioengineering is a novel approach, and here we have used it in conjunction with both unsupervised and supervised methods to identify metabolic signatures of low- and high-producing cell lines. Results from this analysis suggested a range of metabolic subsystems associated with high productivity, including bile, eicosanoid and steroid metabolism, fatty acid metabolism and vitamin metabolism, amongst others. Furthermore, we have used these models as a framework for media simulations, to validate predictions against experimental metabolite uptake rates, production rates and growth measurements. From here we have been able to begin predicting media supplements which could favour high productivity.
Results generated here could have exciting implications for bioengineering, expanding our knowledge of CHO cell metabolism and pathways underlying recombinant protein production. In addition, we are developing a workflow involving more unconventional ‘omics constraints and the interpretation of complex flux sampling results. This project could be translated to other bioengineering systems and our simulations aim to maximise the efficiency, cost-effectiveness and predictive potential of sample-specific constraint-based models.
Strain, B. et al. (2023) ‘How reliable are Chinese hamster ovary (CHO) cell genome-scale metabolic models?’, Biotechnology and Bioengineering [Preprint]. Available at: https://doi.org/10.1002/bit.28366.
Genetically encoded redox sensors, such as the Hyper and roGFP sensor families, are powerful tools for enumerating the real-time dynamics of hydrogen peroxide in cells. In typical experiments, a dynamic profile of sensor oxidation and reduction is obtained following an external hydrogen peroxide perturbation. Using these profiles to characterise the quantitative relationship between the hydrogen peroxide concentration and sensor outputs is challenging as non-linearity in sensor responses to hydrogen peroxide may not be evident. Further, it is unclear how different sensors could be compared. We tested whether these profiles could be characterised by the area under the curve (AUC), signal amplitude, signal time and signal duration parameters. In baker’s yeast, the Hyper7 AUC and amplitude showed a strong linear correlation (r>0.9) to a wide range of hydrogen peroxide concentrations (1-1000 μM). These responses were higher than roGFP2-Tsa2ΔCR parameters at hydrogen peroxide concentrations greater than 100 μM. By contrast, the roGFP2-Tsa2ΔCR AUC and amplitude plots presented distinct linear correlation equations for lower (<100 μM) and higher hydrogen peroxide (>100 μM) concentrations establishing that this sensor’s output is range specific. The signal time and duration for Hyper7 were lower than roGFP2-Tsa2ΔCR at higher hydrogen peroxide concentrations (>100 μM), showing that its activation/deactivation cycle was faster. By contrast, in the fission yeast, the AUC and amplitude for Hyper7 and roGFP2-Tpx1.C169S both showed distinct linear correlations for lower (<50 μM) and higher (>50 μM) concentrations, and the signal time and duration were constant in this background. These results show that any purported correlation between hydrogen peroxide input and sensor output depends on the sensor, cell type and the hydrogen peroxide concentration range chosen. In summary, this method facilitates the characterisation signalling data generated by redox sensors.
Climate change is galvanising plant science and agricultural research to develop stress resilient crops, highlighting the importance of a holistic understanding of the complex signalling responses and growth tradeoffs in plants under various stresses (biotic and abiotic). Knowledge on these molecular interactions and processes is scattered across various sources, in diverse formats and levels of detail, and thus not easily accessible for downstream analysis and modelling. Knowledge graphs provide a powerful and flexible platform for integrating information from diverse datasets with complex relationships. Stress Knowledge Map (SKM) is a knowledge graph resource that integrates previously dispersed information on molecular interactions in A. thaliana and S. tuberosum into a single entrypoint for plant stress response investigations. The knowledge graph implementation of SKM, together with our library of analysis tools (SKM-tools) allows for the systematic generation of context specific formulations, providing a multi-purpose platform for complex analyses to gain new insights behind experimental observations, as well as the development of systems biology models. Given the frequently missing kinetics for such models, a Boolean graph formulation circumvents the need for exhaustive mechanistic details. Together with our Python package for Boolean and semi-quantitative modelling (BoolDog), we are exploring signalling dynamics underlying plant responses to stress, with the long term aim of developing a digital twin. We will describe the development and features of SKM, showcase examples of system-wide data contextualisation and hypothesis generation, and our forays into systematically generating a basis for a molecular digital twin in plants.
Metabolic reconstruction is a challenging and time-consuming process, often resulting in slow progress in developing new reactions. An open-source and accessible interactive interface that allows users to test their curated reactions quickly and accurately can significantly improve efficiency.
We present Reconstructor, a manual and AI-enabled reconstruction interface that allows researchers to analyze curated reactions and generate predicted reactions. Reconstructor accurately provides mass and charge balance information related to curated reactions, offers detailed information on metabolites, and generates atom mapping of the reaction. Reconstructor features a GPT-enabled prediction function where users can input a gene and receive predicted reactions. This predictive capability has been demonstrated to be somewhat accurate in certain types of metabolism when compared to known reactions in a reliable database.
Reconstructor allows various input types for metabolites from reference databases such as VMH, ChEBI, SwissLipids, and MetaNetX formats. It also includes a drawing function using ChemDoodle, with standardisation performed using RDKit. Reactions are created by verifying metabolites, generating RXN files, and performing atom mapping via Reaction Decoder. Mass and charge balance are checked and detailed molecular information, including formula and atom counts, is provided.
Users can save their curated reactions, name them (and optionally add them to a group) for easy access, and edit them through the interface.
Reconstructor supports integration with metabolic databases like VMH. As a test of this integration, we used a copy of the VMH database and have successfully integrated the interface with it, ensuring smooth operation. The integration process includes validation to avoid duplicates, preparation of reaction data, submission to the database, and confirmation to the user. This integration ensures up-to-date datasets and facilitates efficient data management across platforms.
Availability and Implementation: The interface is available at http://reconstructor.humanmetabolism.org/ and the source code can be found at https://github.com/opencobra/reconstructor.
Synaptic transmission is the main process providing cross-connecting activity among neurons in the central nervous system (CNS). It is commonly accepted that a synaptic contact consists of the complex of pre- and post-synaptic membranes contacting with astrocytes. The transmission is accomplished through several steps. Initially, neuromediators stored in pre-synaptic vesicles release into a synaptic cleft. Then they diffuse to post-synaptic membrane where neuromediators bind to specific receptors cause their activation. Finally, in the most cases except acetylcholine the mediators will be removed from the cleft either by convectional diffusion or by re-uptake. There is much experimental evidence indicating a structural ordering of both vesicles with pre-synaptic contacts and receptor localization. In particular, the number and eventual position of glycine receptors (GlyRs) on a post-synaptic membrane are defined by the structural data of the GlyR-gephyrin complex. In this case the membrane cluster of GlyR can have central, divided and rear localizations. In the present study, the 3D mathematical model of a neuronal bouton with a cluster localization of glycine receptors (GlyRs) on the post-synaptic membrane was developed. GlyR provides a transmembrane current of Cl- mediating a hyperpolarization of neuronal membranes. The forming of inhibitory postsynaptic potential (IPSP) and an electro-diffusion of chloride ions were evaluated by applying the boundary problems for a Poisson’s equation and a non-steady-state diffusion equation, respectively. The local changes ion concentration near the post-synaptic membrane, mediated by GlyRs activation, can raise up to 80–110% from the initial level. It is remarkable that the central spatial localization of GlyRs in the cluster had a considerable difference both in the chloride ion concentration changes (6%) and IPSP (17%) compared to the divided or rear localization. Thus, a spatial polymorphism of the post-synaptic density of GlyRs is important to form a physiological response to a neuromediator release.
The aim of the study was to investigate the effect of dietary restriction on the development of insulin resistance, an established precursor to Type 2 Diabetes. This was investigated in conjunction with the effects of both age and diet in a mouse model. The mice were separated into 16 cohorts according to the implementation of dietary restriction or unrestricted access to food (DR or AL, respectively), being fed either a high-fat (HF) or a low fat (LF) diet, as well as the age at which the Oral Glucose Tolerance Test (OGTT) was performed (4, 9, 15, or 21 months of age). The OGTT was performed using a glucose bolus consisting of both unlabelled (natural) glucose as well as deuterium labelled glucose ([6,6-2H2]-glucose). Blood plasma glucose and insulin concentrations were measured for a 2-hour interval following administration of the glucose bolus. A mathematical model consisting of 2 compartments, the Gastrointestinal (GI) and Plasma compartments was used to mathematically model the appearance and disappearance of glucose in the blood plasma. This was achieved using mass action kinetics to describe the transport of glucose from the GI compartment into the blood plasma compartment, as well as its subsequent removal. The contribution of the liver to plasma glucose concentrations was also included as the endogenous glucose production (EGP). Plasma glucose concentrations were then described as functions of time and fit to the OGTT time course data. The parameters obtained from these fitted functions were then used in conjunction with the measured plasma insulin concentrations to calculate Peripheral and Liver specific insulin sensitivity indices (ISP and ISL, respectively). The ISP is calculated using the quotient of the rate of elimination of labelled glucose from the plasma compartment as well as the average insulin concentration throughout the OGTT. In this way, lowered insulin concentrations in conjunction with elevated rates of glucose elimination indicate elevated sensitivity to insulin. The calculated ISP showed DR strongly increased sensitivity, with advanced age and a HF diet only dampening this effect but not preventing it. The ISL is calculated using the average measured plasma insulin concentrations and the average calculated EGP during the time course. Consequently, reduced EGP in conjunction with reduced plasma insulin concentrations indicate an elevated response of the liver to insulin and subsequently a high ISL. The calculated ISL showed DR strongly elevated sensitivity independent of both advanced age and diet. This study highlights the substantial improvement in insulin sensitivity caused by DR, its subsequent reduction in insulin resistance and lowered risk of developing Type 2 Diabetes, as well as the differential effect DR has on both peripheral and hepatic insulin sensitivity.
One of the hallmarks of cancer is a deregulated energy metabolism. A well-known characteristic is the Warburg effect describing an increased glucose uptake and lactate release of the cells, identified first 100 years ago by Otto Warburg. In recent years, more global changes in the metabolism of cancer cells have been described, in particular in biosynthetic pathways, e.g. the amino acid and nucleotide metabolism. Also, the impact of oncogenes, e.g. MYC family members, on metabolic processes has been studied in great detail.
A powerful tool to study metabolic changes in cancer cells are computational models. These often are either relatively small-scale ordinary differential equation models (ODEs) or genome-scale metabolic stoichiometric models (GEMs). While ODE models give detailed insights on the pathways kinetic behavior, they include a limited number of metabolites and often neglect the multitude of connections to neighboring pathways. GEMs on the other hand usually focus on the stoichiometry of the reactions and the steady state flux distribution, but aim to include all known metabolic processes.
In our work, we developed a detailed ODE model of the energy metabolism of a childhood cancer and analyzed the influence of the high-risk oncogene MYCN. Moreover, we aim to embed detailed kinetic models of the energy metabolism into large-scale GEMs. Therefore, we applied an algorithmic reduction of a GEM to derive stoichiometrically correct rates for the withdrawal and import of biosynthetic precursors of an ODE model during growth.
Objective: The objective of this study was to investigate how different weight-loss interventions result in metabolic changes, at different time scales.
Methods: Mathematical models (differential equations) of energy metabolism were used to study weight loss trajectories and changes in postprandial dynamics in response to diet restriction, Roux-en-Y gastric bypass (RYGB) surgery and semaglutide (Ozempic, Wegovy) interventions. Personalized models and Virtual Patients were created and analyzed.
Results and Conclusions: Model for long-term obesity-driven development and progression of diabetes based on the 'twin cycle hypothesis' (liver cycle, pancreas cycle) expanded with inflammation contributing to glucolipotoxicity. The model identifies a window of opportunity for remission through weight loss.
RYGB surgery increases glucagon-like peptide 1 (GLP-1) and improves glucose levels, but also increases (postprandial) insulin. Strikingly, postprandial hypoglycemia is a common problem after RYGB. Model trajectory simulations suggest that interplay between changed anatomy, GLP-1 kinetics and changes in insulin sensitivity may explain the emergence of post-bariatric hypoglycemia months or years after surgery.
Both RYGB surgery and GLP-1 receptor agonism interventions weaken the appetite feedback control circuit that regulates body weight. Treatment with semaglutide not only lowers body weight, but also glucose levels, an effect that warrants further investigation for non-diabetic individuals.
The persistence of endemic malaria infections and the increasing occurrence of antimalarial resistance necessitates the search for treatment strategies to combat and ultimately eradicate the disease. When considering treatment regimens, it also becomes important to consider the specific stages of the parasite life cycle that are affected by treatments. Of interest are the distinct within-host asexual and sexual stages which are associated with disease symptoms and transmission respectively. Ideally one should aim to target both stages to treat an ill patient and to prevent the further spread of the disease.
In this study, we investigated the use of monotherapies and dual therapy antimalarial treatment regimens by using published disease, pharmacokinetic and pharmacodynamic data and models. First, we constructed linked PKPD-disease models that contain a newly developed gametocyte description to account for observed time delays between parasite forms in clinical trial data. Candidate models were validated using clinical data to assess their ability to predict treatment outcomes in vivo, which was used to identify the most appropriate model for further analysis.
Our research also delved into the potential of a novel treatment approach. We evaluated disease outcomes in the model using a CDC-recommended administration of artesunate, used to treat severe P. falciparum malaria. We also explored the use of artemisinin-free combination therapy. This approach involved the repurposing of tafenoquine (a gametocytocidal) to target sexual forms, in combination with lumefantrine (a blood schizonticidal) to target asexual forms. Our model construction, validation, and analysis results will be presented.
Energy metabolism is essential for all living cells, particularly during periods of rapid growth or stress.
Cancer cells, activated immune cells, and yeasts predominantly rely on aerobic glucose fermentation to
generate ATP. This phenomenon is termed the ”Warburg effect” in cancer cells and the ”Crabtree effect”
in yeast cells [1]. Recently, several mathematical models have been proposed to theoretically explain the
Warburg effect.
Beyond glucose, glutamine is an important substrate for eukaryotic cells, playing significant role not
only in biosynthesis but also in energy metabolism. In this study, we develop a minimal constraint-based
stoichiometric model to explain the Warburg effect, incorporating the experimentally observed utilization
of glutamine (the WarburQ effect) [2]. Our model considers both glucose and glutamine respiration, as
well as the fermentation of these metabolites. By accounting for enzyme masses when calculating the
ATP production rates, our resource allocation model reflects the costs associated with different pathways.
Our results indicate that glucose fermentation is a superior energy-generating pathway in human
cancer cells. However, the characteristics of yeast homologues diminish this advantage or lead to the
situation when glucose respiration is more effective. The latter observation is consistent with the behavior of the fungal pathogen Candida albicans, which is known to be a Crabtree-negative yeast. Our results also demonstrate that glutamine serves as a valuable energy source under glucose-limited conditions, in addition to its role as a carbon and nitrogen source in eukaryotic cells. Moreover, the results effectively explain the observed simultaneous uptake of glucose and glutamine.
References
[1] Noureddine Hammad et al. “The Crabtree and Warburg effects: Do metabolite-induced regulations participate in their induction?” In: Biochimica et Biophysica Acta (BBA)-Bioenergetics 1857.8
(2016), pp. 1139–1146.
[2] Jing Fan et al. “Glutamine-driven oxidative phosphorylation is a major ATP source in transformed
mammalian cells in both normoxia and hypoxia”. In: Molecular Systems Biology 9.1 (2013), pp. 1–11.