Future Perspective: Modelling the Brain-Liver-Immune Axis

At IfADo we analyse the effect of important work-related factors on cognition, human behavior and health (Figure 1). To understand these processes we need to analyse different organs. Chemicals are metabolised in the liver, stress can influence the central nervous system and the immune system is essential for maintaining health – and importantly, at IfADo we have the combined expertise to analyse all these different organ systems. However, work-related factors do not just influence one or the other organ in isolation. There is clear evidence that these systems are tightly interconnected. Well-known examples are liver damage due to exposure to hepatotoxic chemicals, which leads to increased systemic levels of ammonia and compromised cognitive functions; acute or chronic stress which influences not only the central nervous system, but also compromises immune functions and results in higher susceptibility to infections; immune cells within the liver can protect from infections but can also cause adverse effects depending on cell type and microenvironment. However, most of these complex interactions remain unclear. Therefore, one of our goals in future is to analyse the interdependencies between the brain, the liver and the immune system, or in short the brain-liver-immune axis. A deeper understanding of the brain-liver-immune axis is of high relevance, since work-place related factors, stress and chemicals, as well as modifying factors, such as aging, influence its balance with major consequences for working humans.

General working strategy
The general working strategy is based on iterative cycles of data generation, mathematical modelling, hypothesis generation and experimental validation. In our previous work, we have repeatedly observed that a first cycle of data generation and modelling does not necessarily result in simulations that correctly reflect the real human situation, thereby leading to discrepancies to experimentally obtained data. However, this represents an important challenge, since it means that either the individual components of the model or the functions by which they have been linked are incomplete or incorrect or both of these. In some cases this led to the discovery of novel mechanisms that after integration into the model resulted in a quantitative agreement of simulated and experimentally obtained data. As soon as a model correctly predicts responses of a human system, especially for interventions that have not been used for training of the model, it may be utilised to perform virtual experiments with a ‘human in silico’, which may even lead to on first sight counterintuitive but correct predictions and the discovery of new mechanisms. This would improve our understanding of the brain-liver-immune axis, constitute an important step towards an integrative, holistic quantitative view on their interplay, and deepen our understanding of its responses to challenges relevant for workplaces (Figure 2).

Work package concept

The long-term goal of our work is modelling of the brain-liver-immune axis. Research work will be organised into closely interacting work packages (WPs). The results obtained in these WPs will be evaluated every 2 years by the scientific advisory board that will be supported by additional external experts. In these evaluations, decisions will be made whether the work in individual WPs shall be continued, modified or terminated and whether additional WPs should be included. The first two WPs (WP1, WP2) will focus on the influence of dopamine and choline metabolism on cognitive functions (Figure 3). These WPs, which address the interaction of metabolic and cognitive functions, will be supported by a next generation of the Virtual Liver (WP3), a spatio-temporal model recently established by a BMBF program, to simulate the impact of the normal and chemically stressed liver on systemic cytokine and metabolite levels and their influence on cognition, thereby closely interacting with WP1 and WP2. The Virtual Liver (WP3) will also integrate the most relevant immune cells of the liver (tissue resident and circulating macrophages, NK cells, neutrophils, ILCs) to simulate the influence of the normal and pathologic liver environment on the activation and function of these immune cells. In line with this, WP4 will focus on signal processing in immune cells and modelling of the critical cell fate decision whether immune cells stay tolerant or become active. Results of WP4 will form the basis of WP5 which focusses on the influence of workplace related factors, such as stress or ageing on cell fate decisions of immune cells and their resulting enhanced or compromised capacity to prevent infections. A mutual challenge of all WPs1-5 is to link results obtained in vitro (or in mice) to the human in vivo situation. Therefore, WP6 will support WPs1-5 by models extrapolating in vitro as well as mouse data to humans, which will be achieved by integrated physiologically-based-pharmacokinetic (PBPK)/spatio-temporal models, where specific model domains (e.g. metabolism, physiological parameters, plasma protein binding) are systematically adjusted to the human situation. Finally, WP7 will open the possibility to understand, whether mechanisms addressed in the previous WPs have an impact on human behaviour, cognition, motivation, emotion, as well as health/disease under workplace-relevant conditions. For this purpose the ‘Dortmund Vital Study’ has been started in which up to 800 individuals will be followed in five year intervals over a period of 20 years. This will give us the possibility to address specific hypotheses resulting from the previous WPs, e.g. do specific forms of stress compromise immune functions and human health? Are specific variants in choline metabolism (EDI3) key factors of cognition? Does the Virtual Liver correctly predict when chemicals, besides compromising functional liver micro-architecture, also influence immune and brain functions? WP1-7 presented in the next paragraph represent a closely interacting, fine-tuned work programme to obtain answers to key questions of workplace-associated research.

Work packages

WP1 – Choline metabolism and cognition. Recently, IfADo has identified a so far uncharacterised key enzyme in choline metabolism, EDI3 (GDE5; GPCPD 1) (Stewart et al., 2012). This enzyme cleaves glycerophosphocholine to glycerol-3-phosphate and choline. Because of IfADo’s involvement in the Human Protein Atlas (Uhlen et al., 2015) and studies on genome-wide genetic variants (Kiemeney et al., 2010; Rothmann et al., 2010), we could clarify that EDI3 is highly expressed in neuronal tissues and contains at least two relatively frequent variants. Interestingly, the group of Dr. Anders in San Diego, CA could demonstrate that one genetic variant of EDI3 is associated with the size of the human visual cortex (Bakken et al., 2012). Further work at IfADo demonstrated that knockdown of EDI3 in neuronal cells leads to reduced levels of acetylcholine. This leads to the question whether EDI3 represents a key factor of acetylcholine availability, which may influence neurotransduction in the CNS. Therefore, the Toxicology and Psychology & Neurosciences Departments will study possible associations of EDI3-polymorphisms and brain activity by combining physiological and cognitive interventions, including plasticity induction, and cognitive parameters including memory formation, working memory and attentional processes, with fMRI and EEG. The project requires modelling and integration of complex time-resolved fMRI and EEG data in relation to genetic variants under conditions of workplace-relevant challenges.

WP2 – Influence of dopamine on cognitive functions. In the last years, the Dept. Psychology & Neurosciences has made considerable advances in understanding the role of the neuromodulator dopamine, on neuroplasticity of the human brain, a physiological key factor for learning and memory formation. We showed that depending on receptor subtype, dosage, and the focality of plasticity induction, dopamine has a complex non-linear impact on plasticity (Kuo et al., 2008, Monte-Silva et al., 2009, Fresnoza et al., 2014a,b). Beyond this evidence, knowledge on the mechanisms by which dopamine affects plasticity in the human brain is still limited. Furthermore, the cognitive consequences, which are relevant for human work performance, which requires life-long learning, are not well understood. Together with the Toxicology Department, the role of genetic polymorphisms involved in dopamine metabolism will be studied in humans and animal/cellular models. With these models, we will avoid one of the major disadvantages of the studies performed to date, i.e. that the results depend on artificially enhanced or reduced components of the dopaminergic system. With regard to physiological mechanisms in the human model, a combination of non-invasive brain stimulation with functional imaging, including fMRI, magnetic resonance spectroscopy, and EEG, will be crucial to explore respective regional and network mechanisms, as well as the dependency of dopaminergic effects from the modulation of GABAergic and glutamatergic mechanisms at the whole brain level, which is critical for the understanding of dopaminergic effects on brain functions, given its relevant cortico-subcortical interconnections. For example, it has been shown that electrophysiological brain activity during complex cognitive tasks depends on dopamine availability (Seer et al. Neurosci Biobehav Rev, 2016), particularly during error processing as reflected by the error negativity (Ne; Falkenstein et al. 1991). In addition, disturbed dopamine metabolism due to the wide spread Toxoplasma gondii infection in humans affect memory functions and the underlying brain activity (Gajewski et al. 2016). Using already established neuronal in vitro systems (Krug et al., 2013; Frey et al., 2014), the corresponding metabolic activities will be studied including calcium imaging, a main substrate of neuroplasticity, in cellular models. Thus, combining animal with human data will deliver a complete picture ofthe effects of dopamine on brain physiology. . Additionally, as already described for WP1, where key enzymes of choline metabolism will be studied, we will also use the physiological knowledge about dopamine effects on cerebral activity and plasticity to model its impact on cognitive processes, such as learning and memory formation, working memory, and attention. The validity of these models will then be tested by combining cognitive task performance with functional imaging, such as fMRI, and EEG. Subsequently, the impact of work-relevant factors, such as distress, age, motivation, which involve dopaminergic modulation, on cognitive performance, and respective physiological foundations, including functional connectivity, and transmitter availability, as outlined above will be modelled and empirically validated.

WP3 – The Virtual Liver: communication with CNS and immune cells. The Virtual Liver is based on work of the Virtual Liver Network, a project that was funded by the BMBF with 42 M Euros. This model is based on the reconstruction of functional units of the liver, where the position of each cell and vessel in a three-dimensional space is known (Hoehme et al., 2010; Drasdo et al., 2014). Into each individual cell of this reconstructed tissue functional principles, such as signalling networks or metabolic pathways can be programmed, which create cell fate decisions or metabolic consequences at the single cell level, whose consequences at organ level can be simulated. Previously, we used this type of spatio-temporal model to address questions at different levels within the liver: at the subcellular level – how small GTPases control size and number of endosomes (Zeigerer et al., 2012); at the cellular level – which coordination principles allow re-establishment of a functional liver microarchitecture after toxic tissue damage (Hoehme et al., 2010); at the organ level – what are mechanisms by which the entire system adapts to toxic stress and when the adaptive capacity is overwhelmed, how does this lead to a breakdown of organ functions (Ghallab et al., 2016; Vartak et al., 2016). In the present project, the Virtual Liver model will be extended to integrate the most important immune cells into the spatio-temporal model, specifically tissue (Kupffer cells) and circulating macrophages, neutrophils, NK cells and ILC, in order to simulate their interactions with hepatocytes and stellate cells of the liver under normal conditions and different types of cell stress, including exposure to chemicals and under conditions of steatosis. One of the key questions in liver toxicology is, when does perfect (scar-free) liver regeneration switches to a situation where normal parenchymal liver tissue is replaced by scare/fibrotic tissue. It has become clear that this shift to fibrotic regeneration is a consequence of a shift in the balance between pro- and anti-fibrotic mechanisms, for example when the capacity of some immune cells, e.g. NK cells, to remove profibrotic activated stellate cells is reduced. Understanding this complex situation will be facilitated by quantitative spatio-temporal modelling. These simulations and experiments will be performed in close cooperation with the Immunology Department. Together with the Psychology & Neurosciences and Ergonomics Departments liver CNS interactions will be integrated into the model. In a first step, the project will focus on the release of metabolites and cytokines from the liver. In our previous work, we have already established models to simulate the consequences of hepatotoxic chemicals on endogenous metabolism of the liver, and how these changes influence metabolite concentrations in the systemic circulation (Ghallab et al., 2016). From our ongoing work in the BMBF funded project DEEP we also know that steatosis, which is prevalent in more than 20 % of the working European population, causes major changes of circulating metabolites and cytokines. Only little is known how these changes influence cognitive and immune functions. In this WP we will use our already established in vitro systems of the normal and steatotic liver, neuronal in vitro systems, and in vitro systems of immune cells, followed by mouse experiments to study whether metabolites and cytokines controlled by the liver influence immune and cognitive functions and the responsible underlying mechanisms. Finally, we will model the impact of altered hepatic metabolism (e.g. of human steatosis and associated systemic changes of metabolites and cytokines) on cognitive processes, such as attention and working memory, fatigue, learning and memory formation, and test the validity of these models by combination of cognitive task performance with functional imaging, such as EEG and fMRI, similarly as described in WP2. The deliverable of this WP will be a model that reliably simulates the influence of metabolites and cytokines released from the liver on functions of the immune and central nervous system.

WP4 – Signal processing in immune cells. The activity of immune cells is regulated by external signals that are typically sensed through cell surface receptors. These receptors can bind soluble factors (e.g. cytokines or chemokines) or interact with specific ligands on the surface of other cells and thereby transduce different signals into the cell. Mathematical modelling can be a powerful tool to understand this complex regulatory system (Watzl et al., 2012). The microenvironment can additionally influence the activity of immune cells. For example, NK cells represent a major fraction of the immune cells in human liver. Interestingly, they differ in phenotype and function from typical NK cells found in the blood (Fasbender et al., 2016). Our data show that some ligands for activating NK cell receptors are up-regulated on hepatocytes upon exposure to different chemicals. Therefore, we hypothesise that the microenvironment of the liver influences the phenotype and the activity of NK cells, which is additionally modified by the presence of chemicals. To investigate this, we will perform a comprehensive analysis of surface markers to identify and characterise circulating and liver resident NK cells and subsequently study how the liver microenvironment modifies signal processing of NK cells. We will additionally study how this changes in the presence of different chemicals. In the past we have successfully used mathematical modelling to understand the signal integration within NK cells (Mesecke et al., 2011). We will extend these models to the specific liver environment in order to get a better understanding of the regulation of NK cells in the liver and to study how the activity of these cells may influence liver function.

WP5 – Impact of stress and ageing on the immune system and cognition. Stress is an integral part of modern life. Stressful situations comprise a wide range of internal or environmental conditions or events, e.g. caregiving for a relative with chronic disease, interpersonal conflicts, juggling many roles and responsibilities, job strain, unemployment, financial worries, over-exercising, and many others. While physical stressors elicit a stress response rather directly, psychological stressors first require a cognitive appraisal by the individual, which then elicits a response. Thus, depending on the subjective perception and interpretation of a stressor due to previous experiences and coping strategies, responses to stress can be different. This may explain why studies find different effects of psychological stress on immune functions and why there is not just one immunological parameter, which could serve as a biomarker for psychological stress and exhaustion. By combining the expertise of the Department of Psychology & Neurosciences, the Department of Ergonomics and the Department of Immunology, we can comprehensively measure changes in the immune system, determine the individual stress response and also measure cognitive performance parameters. The goal of such studies would be to carefully determine which parameters of the individual stress response elicit changes in immune parameters and possible alterations in CNS parameters (EEG, fMRI) and cognitive performance. In proof-of-concept studies we could already demonstrate that such a comprehensive interdisciplinary approach can be very successful. For example, we found a novel correlation between the ‘immunological age’ and the cognitive performance of healthy individuals in two independent studies, which was independent of the chronological age. In other words, older individuals with a ‘younger immune system’ showed a higher cognitive performance compared to younger individuals with an ‘older immune system’.

The comprehensive analysis of immune parameters, the individual stress response and the determination of cognitive performance (including EEG and fMRI parameters, and thus physiological correlates of cognition) produces large data sets. Standard statistical methods are often insufficient to analyse such large quantities of data with complex correlations. Therefore, the expertise in analysing such complex systems provided by the novel modelling unit described in this application will be instrumental for the future success of these approaches.

WP6 – In vitro techniques and human extrapolation. A mutual challenge when modelling the brain-liver-immune axis is extrapolation to the human in vivo situation. Our central interest is how workplace relevant factors influence humans in vivo. Naturally, the possibilities of experimentation with humans are limited, which hampers the possibilities to understand the mechanisms responsible for a particular phenotype. To nevertheless be able to understand the human situation, a combination of mouse in vivo experiments with mouse and human in vitro systems will be used. Recently, scientists of IfADo obtained the Ebert price 2016 (the oldest price of the American Pharmacists Association) for models extrapolating from mouse in vivo as well as human in vitro data to the human in vivo situation. Modelling is required, since in vitro systems do not perfectly recapitulate all in vivo mechanisms (e.g. metabolism; signal transduction), but are biased by quantitative and for some mechanisms even qualitative differences. However, based on global (e.g. genome-wide) analyses these differences can be adjusted by integrated (e.g. PBPK) models which nevertheless allows correct predictions of the human in vivo situation. Specifically, the model differentiates parameter domains, e.g. metabolism, physiological key parameters (e.g. volume and perfusion of organs) and plasma protein binding, which are systematically humanised. This modelling technique with adjusted parameter domains allowed precise predictions of the human in vivo pharmacokinetics (Thiel et al., 2015). In the present WP this modelling strategy will be used to address the following topics: (1) Currently more than 100,000 chemicals are produced with more than 1 ton per year in Europe. However, toxicological characterisation is insufficient for a relatively large fraction of these chemicals. Based on in vitro systems and integrated spatio-temporal/PBPK modelling we will establish alternative techniques that are more precise, faster and cost efficient than conventional animal experiments. (2) We further aim to study the correlations between invasive and non-invasive imaging modalities to better understand the extent to which information from invasive techniques, such as bright field or confocal microscopy, may be inferred from non-invasive techniques, such as fMRI. This will be achieved by a comparison of the same objects using both imaging techniques (e.g. fMRI and confocal microscopy), whereby functional interventions (e.g. inhibitors of mitochondrial metabolism or metabolic processes) will be performed in the in vitro systems. For this purpose, organoids of liver and neuronal tissue are available. Moreover, comparisons between invasive and non-invasive imaging will be performed in mouse models to avoid false interpretations that may be due to possible deviations of the in vitro and in vivo situation. The WP will be performed by the Toxicology Department in close co-operation with the Department of Immunology, to integrate immune mechanisms into the simulations and with the Departments of Psychology & Neurosciences and Ergonomics to apply EEG as well as fMRI technologies, and non-invasive brain stimulation techniques to assess the influence of chemical agents (e.g. solvents, which are of relevance at workplaces) on cognitive functions and activity of specific brain regions.

WP7 – Dortmund Vital Study. Perceptual, cognitive and motor abilities usually decline with increasing age. Nevertheless, the demographic change in Germany and most Western countries will lead to a situation in which employees will be retiring at a later age. Thus, preservation of abilities with increasing age is a central topic of occupational research in the near future. A central prerequisite for better understanding of healthy ageing is to increase our knowledge about the interplay of ageing on basic cognitive functions and the role of factors modulating age-related changes. The identification of endogenous (e.g., immunological status, infections, diseases) and environmental (e.g., education, work conditions, stress, lifestyle, exposure to chemicals) modulating factors on cognitive ageing is therefore of central interest. In this WP we will investigate age-relevant processes and functions at different levels – from the biochemical and cellular to the behaviour level. By collaboration of all the departments of the IfADo a number of previously unknown modulating factors on healthy ageing could already be revealed, for example functional TNF-α polymorphism and its modulation of attentional selection in elderly (Gajewski et al., 2013), and ultra-slow NAT2*6A haplotype and its effect on higher cognitive functions (Selinski et al., 2015). In the course of the next years, data from a cohort of 800 younger, middle-aged, and older subjects who has undergone an extensive test battery of EEG-based tasks and fMRI analyses will be collected. In this comprehensive, multi-year project the influences of a wide range of modulating factors (e.g., work conditions, stress, lifestyle, physical fitness, infections, personality traits) on cognitive functions and their neurophysiological correlates will be tested. Functional imaging would complete the portfolio of methods and provide important data for the overall project. With follow-up measurements (every 5 years) in combination with elaborate experimental paradigms reflecting basic cognitive functions (such as attention, executive control, memory updating), modern EEG methodology, and the analysis of relevant biochemical data, the ‘Dortmund Vital Study’ will allow us to develop and evaluate highly specific hypotheses on the mechanisms of healthy aging. Work-related human functions and the role of work conditions (like stress and job satisfaction) will be of particular interest. Based on an extensive, model-based analysis of the data obtained, specific investigations of the role of the brain-liver-immune axis upon work-related capabilities will follow, among others in virtual workplace simulations that allow controlled experiments in settings close to real-life scenarios.