PoLiMeR Final Symposium

PoLiMeR Final Symposium


Polymers in the Liver – Metabolism and Regulation

28-31 March 2022

Düsseldorf, Germany


Metabolic diseases are a burden on the European population and health care system. It is increasingly recognised that individual differences with respect to history, lifestyle, and genetic make-up affect disease progression and treatment response. A Systems Medicine approach, based on computational models fed with individual patient data, has the potential to provide the basis for a personalised diagnosis and treatment strategy. The PoLiMeR consortium (Polymers in the Liver: Metabolism and Regulation) has identified the inherited, liver-related diseases of glycogen and lipid metabolism as the ideal starting point for innovative research training in personalised ‘Systems Medicine’. These diseases are life-threatening for children. Since each specific disease is rare, research efforts are diluted. Our system-based perspective opens possibilities for the application of novel drugs and diagnostic tools to a range of different diseases. This symposium presents the results of the PoLiMeR consortium and recent research of related research activities.

The symposium has been organised as a live event, featuring:

  • keynote lectures
  • invited presentations
  • presentations from the PoLiMeR PhD students
  • selected presentations
  • poster sessions



The program started on Monday afternoon 28th of March with a welcome by the PoLiMeR project coordinator Prof. Barbara Bakker followed by the opening lecture. On Tuesday, Wednesday, and Thursday, we had plenary sessions starting with a presentation by an invited speaker followed by presentations of the PoLiMeR PhD students and selected presentations. During the lunch breaks on Tuesday and Wednesday there were poster sessions. The symposium was closed on Thursday afternoon by a closing lecture and lunch. There were sessions on:

Session I – Enzyme catalysis at the polymer surface
Experimental and computational tools that can resolve spatiotemporal aspects of enzyme catalysis at the complex polymer surface.

Session II – Pathway dynamics
Experimental tools and computational models to analyse the impact of substrate competition on pathway dynamics.

Session III – Combinatorial explosion of molecular species
Analytical methods and genome-scale modelling approaches to analyse and interpret complex lipid profiles.

Session IV – Spatial and hierarchical regulation
Experimental and computational methodologies to follow polymers in the cell and to integrate this spatial regulation with classical levels of regulation.

Session V – Natural Language Processing for the Life Science
Improving the findability of data  by using information retrieval methods and how we can enable users to store their results in a way that the results are most easily found, even if the query is imprecise.

Besides the scientific program, we will also have a social program which consists of a get-together on Monday evening with drinks and snacks and a symposium dinner on Wednesday evening. More information will follow soon.

The program can be found here.



The following speakers were invited:

Sabine Fuchs
Paediatrician metabolic diseases at University Medical Center Utrecht.

Thomas Hankemeier
Full professor of Analytical BioSciences at the Leiden Academic Centre for Drug Research, Leiden University.

Sophia Ananiadou
Professor in the School of Computer Science at the University of Manchester and is the director of The National Centre for Text Mining (NaCTeM) the only centre of its type in the world.

Ronan Fleming
Senior Lecturer at the National University of Ireland, Galway, and assistant professor in the group of Systems Medicine and Pharmacy at the University of Leiden.

Grant Mitchell
Clinician and a biochemical and molecular geneticist at the CHU Sainte-Justine Mother and Child Hospital Center.



Session I – Enzyme catalysis at the polymer surface

Impact of substrate structural properties in enzymatic reactions, the case of lignocellulose

Adélaïde Raguin
Computational Biophysics, Heinrich Heine Universität, Düsseldorf

Enzymatic processes are widely spread in living and biotechnological systems. For decades, mathematical and computational models have contributed to the interpretation of enzyme assays, leading to various models, in particular based on deterministic methods. With the advent of High Performance Computing, stochastic simulations have also gained much recognition as they for instance allow to model the dynamics of a system at a chosen mesoscopic scale, while taking into consideration its physical properties (structure and interactions). This approach is interesting to specifically study complex systems in which distinct enzymes act simultaneously on a highly structured and large substrate (e.g. glycogen or starch). As an illustration, here we model the enzymatic degradation of a plant cell wall microfibril composed of cellulose and surrounded by hemicellulose and lignin, with various relative abundances and arrangements. Our model highlights the synergistic action of enzymes, and confirms the linear decrease of sugar conversion when either lignin content or crystallinity of the substrate increases. Importantly, we show that considering the structural properties of the substrate in addition to its composition is essential to interpret experimental saccharification data.

A simple methodology to investigate glycogen structure and the activity of glycogen-active enzymes in Glycogen Storage Diseases (GSDs)

Gaia Fancellu1,2, M. J. Marín1, S. Dedola2, R. Field2,3
1School of Chemistry, University of East Anglia (UEA), Norwich Research Park, Norwich, UK; 2Iceni Glycoscience, Norwich Research Park, Norwich, UK; 3Manchester Institute of Biotechnology, University of Manchester, Manchester, UK

Glycogen is a complex polysaccharide composed of linear glucose chains linked via α(1,4)-glycosidic bonds containing α(1,6)-linked branching points. The alteration of glycogen generated by metabolic disorders results in the accumulation of aberrant insoluble structures, as observed in Glycogen Storage Diseases (GSDs). The aim of this project is to develop easy-to-use methodologies to analyse and better define the structure of glycogen and the activity of glycogen-active enzymes from healthy and GSDs sources. The chain length distribution, the average degree of polymerisation, and branching were investigated following a top- down approach on commercially available glycogens to establish the methodologies. A first methodology consisted in the enzymatic digestion of glycogen performed using debranching enzymes, such as pullulanase and Ps. isoamylase, to estimate the branches chain lengths and their relative abundance. A second methodology consisted in a stepwise enzymatic digestion. Initially, the external layers were digested with either glycogen phosphorylase or β-amylase, cleaving the unbranched chain up to the first branching point, followed by treatment of the substrate with the same debranching enzymes employed in the first methodology. A decrease in the average degree of polymerization and a greater relative abundance of short chains was observed in the data obtained in the second treatment compared to the first one. The results from both methodologies were then integrated with theoretical models of short, branched oligosaccharides (DP15) to speculate on the arrangement of the branching points prior to their digestion and on the branching and debranching activity of the glycogen-active enzymes.

Chemoenzymatic synthesis of glucans to assess enzymes of glycogen metabolism

Evaldas Simanavicius1, Robert A. Field1
1Manchester Institute of Biotechnology, The University of Manchester, 131 Princess Street, Manchester, M1 7DN, UK

Chemoenzymatic Synthesis of Glucans to Assess Enzymes of Glycogen Metabolism Glycogen is a highly branched polysaccharide that makes the analysis of its structure technically challenging. Glycogen storage diseases (GSDs) are caused by different enzyme deficiencies affecting either glycogen synthesis or its breakdown. There are still many unknowns in glycogen function which is determined by its structure. This study is developing a bottom-up approach to produce artificial glycogens that emulate the natural biopolymers and defective forms found in GSDs. Firstly, the reliable enzymatic synthesis of linear amylose with the average 18 and 30 degree of polymerization (DP) was successfully achieved with recombinant (expressed in Escherichia coli) starch phosphorylase (PHS2). The DP was controlled by manipulating substrate concentration and incubation time. Secondly, the branch points will be initiated by using recombinant bacterial glycogen branching enzyme (GBE) which has both α-1,4-amylase and α-1,6-transferase activities. However, GBEs from various sources have different preferences for the lengths of transferred chains [2]. The products obtained after this enzymatic reaction will be the subject for Pseudomonas sp. isoamylase which exclusively cleaves α-1,6-glucosidic linkages of chains of DP≥3 and leaves linear glycan chains. The length of donor and acceptor chains, the distance between two branch points, the relative occurrence of intra- versus inter-chain transfer, will be assessed through analytical techniques: thin-layer chromatography, bicinchoninic acid method, size-exclusion chromatography, MALDI-ToF mass spectrometry, HPAEC with pulsed amperometric detection. In addition, the experimental results from bottom-up approach will be provided for the computational glycogen structure modelling.

Modelling glycogen: structure, metabolism, and enzyme mechanistic

Yvan Rousset1, Adélaïde Raguin2, Oliver Ebenhöh1
1Quantitative and theoretical biology, Heinrich Heine Universität, Düsseldorf; Computational cell biology, Heinrich Heine Universität, Düsseldorf

Glycogen storage diseases are rare metabolic disorders caused by enzyme defects that disturb both synthesis and degradation of glycogen as well as export of glucose to the blood. A major challenge in the investigation of glycogen metabolism is that the structural properties of the molecule directly affect the dynamics of the biochemical pathways involved in its synthesis and degradation. Indeed, different structures are achieved during glycogen synthesis depending on enzyme activities. Likewise, during glycogen breakdown the availability of glucose chains and their branching pattern affect the release of glucose. We developed a model that reproduces the synthesis and the degradation of glycogen using stochastic simulations of the involved enzymatic reactions, coupled with a 3D structural model for glycogen. We present the model and demonstrate its capability to reproduce experimentally observed properties of glycogen in vivo, despite the small number of enzymes involved in our model. Then we will show that different structures can be obtained depending on enzyme concentrations and properties. Our numerical simulations show that the chain length distribution (CLD) is highly sensitive to the way we model the branching process. We propose a method to infer mechanistic properties of the enzymes involved in glycogen synthesis and breakdown using the CLD as the molecular fingerprint resulting from the enzyme activities. Results suggest a considerably broader substrate specificity of the branching enzyme compared to what is usually assumed. We later extend the framework to other metabolic reactions, which allows us to discuss different scenarios in the context of glycogen diseases.

Session II – Pathway dynamics

Inborn Errors of Coenzyme A (CoA) Metabolism

Grant Mitchell
Medical Genetics Service, CHU Sainte-Justine, Montreal, Canada

CoA thioesters are obligatory intermediates of fatty acid oxidation in mitochondria and peroxisomes, of amino acid degradation and of the Krebs cycle. Acetyl-CoA (AcCoA) acts in biosynthesis and acylation in other cell compartments. Dozens of diseases arise from hereditary deficiencies of single steps of intermediary metabolism involving acyl-CoA derivatives (CoA Metabolic Diseases, CAMDs). Comparing the clinical signs of CAMDs and known properties of acyl-CoAs, we proposed hypotheses for mechanisms underlying CAMDs. Many CAMDs produce both chronic problems in one or more organs and acute episodes of organ dysfunction (“crises”), often with hypoglycemia, hyperammonemia and ketoacidosis. Because acyl-CoAs are intracellular and tissue samples are required for acyl-CoA measurement, we study mouse models of CAMDs. For example, in humans, HMG-CoA lyase deficiency (HLD) causes episodic hypoglycemia and hyperammonemia. In mice with liver-specific HLD, hypoglycemia and hyperammonemia occur, but only if high liver HMG-CoA plus low liver Ac-CoA levels are present. HMG-CoA levels corresponded to those known in vitro to inhibit AcCoA-mediated stimulation of pyruvate carboxylase, in gluconeogenesis. Similarly, AcCoA is necessary for synthesis of N-acetylglutamate, activator of the rate-limiting step of ureagenesis, carbamylphosphate synthetase I. A functional, AcCoA-independent NAG analogue reduces hyperammonemia in liver HLD mice. Similar findings in propionic acidemia (PA) mice suggest AcCoA insufficiency as a possible common denominator for acute crises. To test an extrahepatic organ, we studied mice with cardiomyocyte-specific HLD, and found acute transient contractile insufficiency after injection of the HMG-CoA precursor, ketoisocaproate. To date, mouse CAMDs resemble their corresponding human disease and support acyl-CoA-determined pathophysiology.

Personalised computational modelling of liver mitochondrial β- oxidation suggests a reduced risk of free CoA depletion as a rescue mechanism in MCADD

Johannes C.W. Odendaal1,*, Emmalie A. Jager2,*, Anne-Claire M.F. Martines1, Marcel A. Vieira-Lara1, Nicolette C.A. Huijkman1, Ligia A. Kiyuna1, Albert Gerding1,3, Justina C. Wolters1, Rebecca Heiner-Fokkema2, Terry G.J. Derks2, Karen van Eunen1, Barbara M. Bakker1
1Laboratory of Paediatrics, University of Groningen, University Medical Centre Groningen, The Netherlands; 2Section of Metabolic Diseases, Beatrix Children’s Hospital, University Medical Centre Groningen, The Netherlands; 3Department of Laboratory Medicine, University Medical Centre Groningen, The Netherlands; 4Department of Analytical Biochemistry, Groningen Research Institute of Pharmacy, University of Groningen, The Netherlands; *These authors contributed equally.

Some people with medium-chain acyl-CoA dehydrogenase deficiency (MCADD) develop life-threatening hypoketotic hypoglycaemias, while others remain asymptomatic. Diagnostic tools for effective risk stratification are as of yet unavailable. We built and validated a computational model of the human liver mitochondrial fatty acid oxidation (mFAO). This model suggests that MCADD causes simultaneous free CoA (CoASH) depletion and reduced mFAO flux. We hypothesise that the interplay between these two effects determines the efficacy of different rescue mechanisms and might help to explain why certain MCADD patients develop symptoms while others manage to avoid them. In our model, the enzymes with the largest absolute flux control were identified and then investigated as potential rescue mechanisms. Changes in enzyme expression did not always simultaneously increase CoASH and flux. For instance, increased VLCAD expression above certain levels actually exacerbated CoASH sequestration while still increasing mFAO pathway flux. Among the changes that did increase both pathway flux and free CoASH concentration were SCAD and MTP, which were also increased in the fibroblast proteomics from an asymptomatic MCADD individual. Personalised models based on these proteomics data confirmed that the asymptomatic patient might have been able to compensate for pathway flux and CoASH sequestration under metabolic stress and that both MTP and SCAD contributed to this alleviated phenotype. Such personalised models of the mFAO might serve as a future prognostic tool for MCADD which accounts for complex adaptations that might not be apparent from data alone.

In-vitro reconstitution of NAD core metabolism

Eugenio Ferrario1
1University of Bergen

Nicotinamide Adenine Dinucleotide (NAD) is a relatively simple biomolecule ubiquitously present in all domains of life. It is mainly known as an electron carrier involved in many metabolic pathways such as glycolysis, β-oxidation, Krebs cycle which fuels oxidative phosphorylation furnishing reducing equivalents into the mitochondrial electron transport chain. In recent years, in addition to its redox function, NAD has emerged also as a key regulator of cellular and organismal homeostasis, involved in signaling processes in which it doesn’t act anymore as a cofactor but as a substrate that is cleaved into different moieties. This signaling function is mediated by different classes of NAD-signaling enzymes such as poly-ADP-polymerases (PARPs) and sirtuins (SIRTs), each located in specific cell compartments regulating fundamental processes like cell cycle, DNA repair, calcium signaling, and epigenetics. NAD bioenergetic pathways, connected with NAD’s role as a signaling molecule and the complexity of NAD biosynthesis and compartmentalization highlight how tightly regulated NAD levels are in cell compartments, and how any change in NAD homeostasis is a readout of non-physiological conditions. In order to better investigate these dynamics, we remodeled in an in-vitro system NAD biosynthetic, and degrading pathways using purified recombinant proteins, thereby reconstituting NAD metabolism in a simplified and controlled cyclic reaction. Using LC-MS we developed a sensitive method to determine the concentrations of each intermediate as well as the turnover rates of each enzyme. This system of in vitro NAD metabolism provides a sensitive tool able to quantify the impact of added enzymes on NAD homeostasis.

Respectful modelling of mFAO: metabolite solubility, transparent parameter documentation, and ensemble modelling

Johannes C.W. Odendaal1, Anne-Claire M.F. Martines1, Fentaw Abegaz1,2, Karen van Eunen1, Barbara M. Bakker1
1Laboratory of Paediatrics, University of Groningen, University Medical Centre Groningen, The Netherlands; 2Statistics and Probability Unit, University of Groningen, Groningen, The Netherlands

We built and validated a computational model of the human liver mitochondrial fatty acid oxidation (mFAO). We built on the methodology of computationally recreating this pathway in two ways: by adding a solubility attribute to the metabolites and by implementing the respectful modelling framework. The solubility attribute was based on the observation that acyl-CoAs diffuse laterally within the membrane, anchored by their hydrophobic tails, with increased likelihood of diffusion into the bulk medium as the chains are shortened. Some authors have therefore postulated that mFAO intermediates are transferred between active sites by “surface crawling” along the inner-mitochondrial membrane. The solubility attribute simulates this chain-length-dependent effect and led to more accurate model predictions of various experimental measurements. The respectful modelling framework with consists of an input and an output aspect. On the input side, we produced a nested set of documentation for the model which would make the parameter choices more transparent and amenable to scrutiny and improvement. This includes detailed kinetic data extracted from literature, including considered alternatives and the argumentation underlying our modelling decisions. For the output, we objectivised our parameter choices by generating an ensemble model, in which parameters do not have single values, but rather vary across a range of retrieved literature values, with probabilities expressed as weighted distributions. Together, these three methodological elements represent a clear qualitative step forward in modelling mFAO and also makes the model a more interoperable tool for use by other investigators.

Session III – Combinatorial explosion of molecular species

Combinatorial implosion via constraint-based modelling

Ronan Fleming
School of Medicine, National University of Ireland, Galway; Leiden Academic Centre for Drug Research, University of Leiden, NL

Constraint-based modelling can mechanistically simulate biochemical systems at genome-scale, permitting hypotheses generation, experimental design and integration of experimental data, with numerous applications, including modelling of metabolism. However, to date, integration of tracer-based metabolomic data with constraint-based modelling has been intractable due to combinatorial explosion in the potential number of isotopically labelled metabolites and difficulty with implementing nonlinear thermodynamic constraints. We present a novel approach to integration of tracer-based metabolomic data with genome-scale, constraint-based models of metabolism that avoids combinatorial explosion and implements thermodynamic constraints via an efficient convex optimization formulation. Subsequently, the relation between this work and combinatorial explosion in lipid metabolism is introduced.

Stoichiometric modularisation of combinatorial lipid modelling

Hadjar Rahou1, Ronan Fleming1
1National University of Ireland, Galway

Lipid metabolism involves many promiscuous enzymes which leads to a combinatorial explosion of the number of lipid species. Chemical reactions can be considered in terms of conserved and reacting moieties. A conserved moiety is a group of atoms that are invariant in all reactions of a metabolic network. A reacting moiety is the group of atoms related to the bond changes in a reaction. We hypothesise that the computation of the conserved and reacting moieties for all available metabolic reactions catalysed by the same enzyme would allow us to define promiscuity constraints on the set of reactions catalysed by a promiscuous enzyme and use this to develop a new method to avoid combinatorial explosion. Acyl-CoA Dehydrogenase Medium Chain (ACADM) is a gene that encodes the medium-chain acyl-CoA dehydrogenase (MCAD) enzyme, which is mutated in patients with MCAD deficiency. We constructed the expanded ACADM network from Recon3 by extracting the reactions associated with the ACADM gene and adding 30 new metabolites and 73 new reactions, when lumped reactions were required to be split. By using, atom mappings for each reaction in the ACADM network, we identified the conserved moieties by analysis of the corresponding atom transition network, and the reacting moieties by the study of the bond transition network. The analysis of enzymatic reactions according to their conserved and reacting moieties opens up the prospect of representing a metabolic network susceptible to a combinatorial explosion as a constrained combination of conserved and reacting moieties constrained by enzyme-specific promiscuity patterns.

Quantitative and miniaturized assays for structure-resolved lipidomics in metabolic diseases

Madhulika Singh1, K.A. Krishnamurthy2, M.H. Oosterveer2, A.C. Harms1, T. Hankemeier1
1Division of System Biomedicine and Pharmacology, LACDR, Leiden University, The Netherlands; 2Laboratory of Pediatrics, University of Groningen, University Medical Center Groningen, The Netherlands

Inborn errors of metabolism are group of disorders which occur due to deficient activity of enzymes or transporter proteins in a metabolic pathway and are inherited in an autosomal recessive manner. Glycogen storage disease (GSD) and mitochondrial fatty acid oxidation disorders are examples of these rare disease conditions which lead to series of complications that can be fatal. Many of these disorders are diagnosed based on plasma/urine metabolic tests performed in laboratories. Identification of suitable novel biomarkers can help in predicting the future risk of the disease and can be translated and integrated in various ways to develop a ‘systems medicine’ approach. Lipids are important biomarkers which are associated with inborn errors and can be very helpful in understanding the mechanism behind them. The lipidomics profiling of these inborn errors needs an efficient bioanalytical method that includes a maximum number of lipid classes requiring lower sample volumes with appropriate sensitivity. We are working on various LC-MS technologies using hydrophilic interaction chromatography coupled with mass spectrometry to have comprehensive coverage of lipids as well as important intermediate compounds involved in the metabolic pathways such as acyl CoA and acyl carnitines. These methods have been applied to mouse models of GSD disorders to observe the difference in lipid concentrations between control and disorder samples. Approximately, 472 and 483 lipids shows fold change of 1.5 or more in GSD 1a and 1b samples respectively. Further, the role of all these metabolites will be considered and linked for further exploration of the possible pathways.

Session IV – Spatial and hierarchical regulation

Tissue-specific characterization of deoxy-glucose and palmitoyl- carnitine uptake in MCAD KO mouse under fasting and cold challenge

Ligia Akemi Kiyuna1, Miriam Langelaar-Makkinje1, Anne-Claire M.F. Martines1, Albert Gerding1, Justina C. Wolters1,2, Maaike H. Oosterveer1, Dirk-Jan Reijngoud1, Barbara M. Bakker1
1Laboratory of Pediatrics, University of Groningen, University Medical Center Groningen, The Netherlands; 2Department of Analytical Biochemistry, Groningen Research Institute of Pharmacy, University of Groningen, The Netherlands

Background: Medium-chain acyl-CoA dehydrogenase deficiency (MCADD) is the most prevalent fatty-acid oxidation disorder. Among its symptoms, life-threatening hypoketotic hypoglycemia remains poorly understood. To gain insights into the role of peripheral tissues in this severe phenotype, we used 14C-labelled substrates to assess tissue-specific energetic demand in mice under metabolic stress. Methods: Wild-type (WT) and MCAD KO mice (C57BL/6) were exposed to 14h-overnight fasting followed by 4h-fasting at either 4°C or 25°C. In separate experiments, animals were injected with 14C-deoxy-glucose (DOG) or 14C-palmitoyl- carnitine (PC). Blood glucose (BG), blood ketone bodies (KB) and glycogen stores were quantified. Finally, ACADs expression was evaluated in the liver and skeletal muscles. Results: At 4°C, MCAD KO mice presented lower BG and KB than WT counterparts. No differences in ACADs expression and 14C-DOG uptake were observed between both groups in any condition. Regarding cold response, 14C-DOG uptake by the brown adipose tissue was increased at 4°C in both groups. In all tissues 14C-PC detection was lower at 4°C than at 25°C, except for the liver of MCAD KO mice, suggesting the accumulation of 14C-PC and/or intermediates by this tissue. At 4°C, glycogen was reduced in quadriceps of MCAD KO mice compared to the WT group, which correlated with changes in BG and body temperature. Conclusion: A combination of fasting and cold triggered a mild hypoketotic hypoglycemic phenotype in MCAD KO mice, but this could not be attributed to an increased peripheral glucose demand. This suggests that hypoglycemia is largely due to impaired glucose production by the liver.

The metabolic master regulator mTORC1 in glycogen storage disease type 1

Maria Rodríguez Peiris1,2, Jose Miguel Ramos Pittol1, Kishore Alagere Krishnamurthy3, Maaike H. Oosterveer3, Alexander Martin Heberle1,3, Kathrin Thedieck1,2,3
1Institute of Biochemistry and Center for Molecular Biosciences Innsbruck, University of Innsbruck, Innsbruck, Austria; 2Department for Neuroscience, School of Medicine and Health Sciences, Carl von Ossietzky University Oldenburg, Oldenburg, Germany; 3Laboratory of Pediatrics, Section Systems Medicine of Metabolism and Signaling, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands

Glycogen storage disease type 1 (GSD1) is an autosomal recessive metabolic disease with an incidence of 1/100.000 births. GSD1 is characterized by mutations in glucose-6-phosphatase (G6Pase), which catalyses the last step of glycogenolysis and gluconeogenesis. Patients present hypoglycaemia and impaired glucose homeostasis, and up to 8% of patients older than 25 years develop hepatocellular carcinoma (HCC). While dietary treatments prevent hypoglycaemia and improve survival into adulthood, HCC incidence persists. Thus, new therapeutic strategies are needed. Mechanistic/mammalian target of rapamycin complex 1 (mTORC1) is a protein serine-threonine kinase complex that promotes cell growth and inhibits autophagy. mTORC1 activity is often linked to tumour formation. Recent evidence indicates mTORC1 hyperactivity and autophagy inhibition in GSD1, which could act as a driver of HCC development. We assessed the link between G6Pase and mTORC1 in the frame of the PoLiMeR consortium.

Multi-omics analysis of tissue-specific metabolic crosstalk after CMA supplementation on rat NAFLD models

Hong Yang1, Cemil Bayram2, Xiangyu Li1, Hasan Türkez2, Adil Mardinoglu1,3
1Science for Life Laboratory, KTH – Royal Institute of Technology, Stockholm, Sweden; 2Department of Medical Biology, Faculty of Medicine, Atatürk University, Erzurum, Turkey; 3Centre for Host-Microbiome Interactions, Faculty of Dentistry, Oral & Craniofacial Sciences, King’s College London, London, United Kingdom

Non-alcoholic fatty liver disease (NAFLD) is a leading cause of chronic liver disease worldwide, and it is considered as a hepatic manifestation of metabolic syndrome. One of the prevailing hallmarks of NAFLD is mitochondrial dysfunction and metabolic abnormalities due to altered NAD+ and glutathione (GSH) metabolism. In placebo-controlled clinical trials, we have reported supplementation of NAD+ and GSH precursors, named combined metabolic activators (CMAs, consisting of L-serine, N-acetyl-L-cysteine, L- carnitine tartrate, and nicotinamide riboside), can be used as a promising treatment strategy in patients with metabolic abnormalities, including NAFLD, COVID-19, Parkinson’s disease and Alzheimer’s disease. Here, we first tested the effect of individual and combined metabolic activators on two rat NAFLD models and observed that CMA significantly attenuates lipidosis, hydropic degeneration, necrosis and hyperemia in liver tissue of animals. Next, we performed a comprehensive investigation on the underlying molecular mechanism of CMA supplementation through multi-tissue (liver, muscle, and adipose tissues) transcriptomics and plasma metabolomics analysis. By generating tissue-specific biological networks, we observed both common and specific biological responses to CMA supplementation across a range of metabolically active tissues.

Hepatic ChREBP knockdown aggravates hepatic glycogen accumulation and hypoglycemia while enhancing glycogen cycling in a mouse model for acute Glycogen Storage Disease type Ib

Kishore A. Krishnamurthy1, M.G.S. Rutten1, J.A. Hoogerland1, T.H. Van Dijk2, T. Bos1, M. Koehorst2, M.P De Vries1,3, N.J. Kloosterhuis1, R. Havinga1, J.C. Wolters1,3, B.M. Bakker1, M.H. Oosterveer1
1Laboratory of Pediatrics, University of Groningen, University Medical Center Groningen, The Netherlands; 2Department of Laboratory Medicine, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands; 3Department of Analytical Biochemistry, Groningen Research Institute of Pharmacy, University of Groningen, The Netherlands

Glycogen Storage Disease type 1 (GSD I) is an inborn error of metabolism caused by mutations in Glucose-6- Phosphatase (G6pc; GSD Ia) or the Glucose-6-Phosphate Transporter (Slc37a4; GSD Ib). GSD I is characterized by intracellular accumulation of Glucose-6-Phosphate (G6P) and glycogen, which contribute to hepatomegaly. Accumulated G6P in turn activates Carbohydrate Response Element Binding Protein (ChREBP), a transcription factor that induces adaptive metabolic responses in GSD I. Predicted ChREBP target genes include Liver Glycogen Synthase (Gys2) and Liver Glycogen Phosphorylase (Pygl); however, the regulatory role of ChREBP in hepatic glycogen metabolism remains largely unresolved. Therefore, we assessed the consequence of hepatic ChREBP knockdown (ChREBP-KD) on hepatic glucose and glycogen metabolism in acute GSD Ib induced by S4048, a pharmacological SLC37A4 inhibitor. Hepatic ChREBP activity was markedly increased by S4048, and normalized by ChREBP-KD. Pygl and Gys2 mRNA levels and PYGL protein levels were significantly reduced by ChREBP-KD in S4048-treated mice, while GYS2 protein levels tended to decrease. Stable isotope studies showed that GYS2 flux was enhanced in these mice. ChREBP-KD furthermore increased Glucokinase (GCK) mRNA and protein expression independently of S4048 treatment, and GCK flux in vehicle-treated mice. Expression of Glucokinase regulatory protein (GCKR) remained unchanged upon ChREBP-KD. These changes were paralleled by more pronounced hepatic G6P and glycogen accumulation and hypoglycemia upon ChREBP-KD in S4048-treated mice. In conclusion, hepatic ChREBP-KD enhances glycogen cycling and aggravates hepatomegaly and hypoglycemia in a mouse model for acute GSD Ib.

Optimization and characterization of precision-cut liver slices as an experimental model to study hepatic glucose production ex vivo

Ligia Akemi Kiyuna1, Kishore Alagere Krishnamurthy1, Miriam Langelaar-Makkinje1, Albert Gerding1, Peter Olinga2, Karen van Eunen1, Maaike H. Oosterveer1, Barbara M. Bakker1
1Laboratory of Pediatrics, University of Groningen, University Medical Center Groningen, The Netherlands; 2Faculty of Science and Engineering, University of Groningen, The Netherlands

Background: Hypoglycemia is a severe symptom of various inborn errors of metabolism. To unravel the mechanisms underlying perturbed hepatic glucose homeostasis, an in vitro/ ex vivo model is needed. In view of the limitations of in vitro hepatocyte cultures, the goal of this study is to optimize and characterize precision-cut liver slices (PCLS) as a model to study glucose production. Methods: PCLS were prepared from liver tissue collected from fed and 12h-fasted male C57BL/6 mice. They were kept in culture in glucose-free medium supplemented or not with different gluconeogenic precursors: glycerol 10 mM, lactate 20 mM/ pyruvate 2 mM or dihydroxyacetone (DHA) 10 mM. Net cumulative glucose production was assessed at different time intervals and glycogen content was quantified. Finally, responsiveness of glucose production to hormonal and pharmacological stimulation and the effect of 24h-pre-incubation in complete medium were evaluated. Results: PCLS derived from fed and fasted mice produced glucose. Highest net glucose production was observed by PCLS derived from fed mice supplemented with DHA. Net glucose production was lower after 24h-pre-incubation as compared to no pre-incubation. After 5h in culture, PLCS glycogen content was equally reduced in all conditions tested. Both forskolin and dibutyryl-cAMP increased net cumulative glucose production. However, glucagon and insulin did not alter PCLS glucose production. Conclusion: PCLS can be used as an ex vivo platform to study glucose production. Analysis of net glucose production is possible up to 24h in culture and is maximum with DHA supplementation. PCLS glucose production can be stimulated by glucagon mimetics.

Session V – Natural language processing for the life science

Enhancing Deep-Learning-based named entity recognition for biomedical applications via modified training

Ghadeer Mobasher1,2 Wolfgang Müller1 Michael Gertz2 Olga Krebs1
1Heidelberg Institute for Theoretical Studies (HITS), Heidelberg, Germany; 2Institute of Computer Science, Heidelberg University, Heidelberg, Germany

In biomedical text mining, named entity recognition (NER) is an important task used to extract information from biomedical articles. Improving the NER’s performance will directly have a positive impact on extracting relations between those entities. In recent years, deep learning has become the main research direction of NER due to the development of effective models. However, training the model naively on targeted datasets without considering the class distribution would be problematic, especially with dealing with imbalanced datasets. As a result, the model will be having a high prediction performance to majority classes. However, rare classes are at the centre of our application interest. Language transformer models like e.g. BERT are frequently used because they enable the specialisation of models by domain-specific training based on pre- training, yielding models like e.g. BioBERT. However, It is worth investigating the performance of continual models that combine training with specialized and with general corpora against models that were trained from scratch in biomedical literature only. Therefore, in our proposed approach, we combine a specialized training data collection with a modified loss function of the model during training. We introduce coefficients that penalize the majority classes and give more weight to the rare classes. Our experimental results support these approaches, i.e. modifying the loss function and using domain-specific training from scratch. We measure an increased performance with respect to state-of-the-art results.



The PoLiMeR Final Symposium took place at Haus der Universität in Düsseldorf, Germany.

Haus der Universität
Schadowplatz 14
40212 Düsseldorf, Germany

Go to venue website