|
RA1. Mass spectrometric resolution of protein phosphorylation in hormonal signalling
The reversible phosphorylation of specific proteins participates in the regulation of virtually all aspects of cell physiology and development. The importance of this process is illustrated by the many hundreds of protein kinases and phosphatases detected in eukaryotic genomes. Reversible protein phosphorylation is also the major mechanism for external control. Signal transduction by cell surface receptors operate via a variety of protein kinase cascades that in turn mediate numerous protein-protein interactions which eventually lead to effector protein phosphorylation control. A significant amount of information has recently been accumulated on multiple protein phosphorylation reactions downstream from various receptors and on interaction of multiple branching pathways in signalling. In order to understand signalling and cellular control the focus is now moving towards an understanding of the spatial and temporal relations between pathways. Insulin signal transduction is a major research area because of its medical and socio-economic importance in connection with diabetes, which is known to affect over 140 million people worldwide.
The Department of Medical Cell Biology, University of Linköping has developed a mass spectroscopic methodology for specific detection, identification and characterization of phosphopeptides en masse and in very complex mixtures of biological peptides. This methodology for the first time allows a characterisation of naturally phosphorylated proteins without using radioactive labelling, antibodies or any other exogenous tracer.
The purpose of the project to be conducted in connection with the BioSim Network is (i) to characterise adipocyte phosphorylation and changes in the phosphorylation process in response to insulin treatment, (ii) to identify proteins that undergo reversal phosphorylation and their phosphorylation sites, (iii) to determine the stoichiometry of hormone-induced protein phosphorylation and the spatial-temporal variations in this process, (iv) to examine cells from patients exhibiting insulin resistance to compare their phosphorylation processes with those of normal adipocytes, and (v) to develop mathematical models that can integrate and evaluate the large amount of data generated by the experimental activities.
Human adipocytes will be obtained during elective surgery. Isolated cells incubated with and without maximally effective concentration of insulin (10 nM) for 20 min. Cells will then be subfractionated into purified plasma membrane, nuclear/mitochondrial, internal membrane, and cytosolic fractions by differential and gradient centrifugation. These fractions will be prepared in the presence of appropriate phosphatase inhibitors. They will be proteolytically digested with trypsin and subjected to analyses for phosphopeptide contents and patterns using two complementary MS methods. First, the phosphopeptides will be identified in the complex peptide mixtures by detection of characteristic metastable ions appearing upon the loss of phosphoric acid during MALDI-TOF mass spectrometry. Second, the peptide mixtures will be separated by liquid chromatography (LC) with online detection using QSTAR hybrid mass spectrometer. The diagnostic fragments produced during LC-MS will allow detection of phosphorylated peptides. These analytical screenings will identify molecular masses of the peptides, which become phosphorylated or de-phosphorylated in each subcellular fraction in response to the hormone treatment.
To obtain the peptide sequence information we will first enrich for phosphopeptides by immobilised metal [Fe(III) and/or Ga(III)] affinity chromatography (IMAC). This enrichment will simplify MS sequencing of the peptides undergoing hormone-dependent phosphorylation as identified during the first stage of the analysis. These phosphopeptides will be sequenced using MALDI-TOF PSD and complementary ESI MS/MS analyses using QSTAR spectrometer. The complete or partial sequence information will allow identification of corresponding peptides and proteins in the human genomic database, as well as the sites of phosphorylation in these proteins.
The phosphopeptides identified in the first two stages will be mapped in the LC-MS separations of total tryptic peptide mixtures from each subcellular fraction. This approach can detect with high sensitivity both the phosphorylated and corresponding non-phosphorylated forms of the peptides simultaneously and it measures stoichiometry directly without the need for an exogenous tracer. Measuring of the phosphorylation state for every identified protein in each subfraction of adipocytes at distinct time points after insulin application will reveal the spatio-temporal mechanism of hormonal signalling by protein phosphorylation. Regardless of interaction of multiple branching pathways, our approach will disclose the central protein phosphorylation events in normal adipocytes subjected to hormone treatment.
Cells from patients with insulin resistance, i.e. with a reduced ability of target cells to respond properly to increasing concentrations of insulin, as found, e.g., in different forms of type-2 diabetes, in polycystic ovarian syndrome, and after trauma, will be mapped and compared to normal cells. Elucidation of the differences in phosphorylation of distinct proteins in pathological and normal cells should open new directions in application of drugs targeted to cellular processes downstream from the insulin receptor.
Model development will occur along several lines: (i) development of algorithms and software for automatic phosphopeptide identification and calculation of phosphorylation stoichiometries from MS raw data, (ii) development of tools to visualise the multidimensional arrays of phosphorylation states of hundreds of proteins that change in time and space as well as within the individual protein, and (iii) development of nonlinear dynamic analyses to help us understand the mechanisms behind the changes in phosphorylation patterns.
The dynamic spatio-temporal mapping and visualisation of the information flow in response to insulin can provide a hitherto unknown level of understanding of how this hormone works, in health and disease. To our knowledge, this kind of investigation has never been accomplished before for any mammalian hormone or extracellular controller. The investigation will establish a technique that is applicable to studies of protein phosphorylation in any type of living cell and in response to any hormone. This could make a tremendous advancement in many aspects of live sciences and drug development. The approach to identify molecular sites of dysfunction in cells from insulin resistant patients is also unique and may provide novel insights into the disease and indicate targets for drug development. It will furthermore allow in silico testing of the effects of developing drugs aiming at different cellular targets.
RA2. Metabolic fates of pharmaceuticals in living cells (PharmBiosim)
Pharmaceuticals are often polyfunctional molecules and their metabolic fates are difficult to predict. In the quest to replace animal tests by other approaches, microbial models of drug metabolism have been applied for more than 30 years. However, due to the complexity of the molecules, the obtained results have not yet allowed us to extract general principles of eukaryotic degradation of drugs. Hence, there is a significant demand for a modelling system that can describe the most basic interactions of the enzymes involved in drug biotransformation. The PharmBiosim approach has been developed at the Technical University of Dresden. The approach aims at identifying general principles of drug metabolism - also for complex pharmaceuticals - by attributing enzyme actions and xenobiotic stress phenomena to functional groups and, thereby, reducing the complexity of the drug molecule to those functionalities provoking chemical modifications.
Mechanisms of intracellular xenobiotic metabolism
In drug metabolism the living cell acts as a complex catalytic system capable of selectively converting pharmaceuticals. At the present, the mechanisms of intracellular xenobiotic metabolism are mostly unknown. In many cases, there is more than one enzyme actively converting the substrate, and in the majority of all biotransformations, these competing enzymes exhibit divergent stereoselectivities. As this factor is critical for the metabolic fate of a drug, there have been several incentives to elucidate selectivities towards the investigated xenobiotic. This is usually done by heuristically inhibiting enzyme activity, often without exact notion of the targeted enzymes. Such an approach suffers from high cytotoxicities of the additives and side effects on the cell secondary metabolism. Methods of metabolic engineering, on the other hand, attempt to overexpress inborn or foreign genes in a cell. The intracellular enzyme network is severely affected by genetic modifications and limited information can, therefore, be derived about natural enzyme interactions. Recent approaches to a more detailed understanding of the intracellular processes have, however, revealed effects of cell physiology and cell stress responses on the interfering action of enzymes converting ethyl 4-chloro-acetoacetate.
The PharmBiosim project is a combined experimental and theoretical work on the factors governing the biotransformation of drugs in S. cerevisiae and mammalian cells. As initial model compounds serve ethyl (S)-3-hydroxybutyrate (2) which is the common reaction product of the biotransformations of ethyl acetoacetate (1), ethyl 2-chloro-acetoacetate (3), ethyl 4-chloro-acetoacetate (4), and ethyl 4,4,4-trifluoro-acetoacetate (5). Biotransformations of (3-5) will subsequently complement the model. With its (chiral) functionalities (ketone, ester, halogen, doubly activated methylene group, carboxylic acid and β-halohydrin) these substrates and their products represent a variety of essential drug functionalities and activated positions which finally are strategic positions for enzymatic attack.
Applied techniques
Thanks to recent progress in elucidating cell stress phenomena in whole-cell biotransformations, almost every biotransformation reaction observed for the above substrates can be attributed to distinct enzymatic activities, and the involved enzymes are mostly known. The innovative approach of the PharmBiosim project is to assign enzymatic activities to cell physiological and culture conditions. Under precisely defined conditions, exactly two dehydrogenases with opposite stereoselectivities have been found to partake in the reduction of the carbonyl group in the above mentioned substrates. Further, for these substrates there is no example known so far in which enzymes with comparable substrate specificity and stereoselectivity participate in the biotransformation of the above β-keto esters. Each reaction proceeds in a highly stereoselective manner and the stereoselectivity of a reaction will, therefore, provide the necessary information about the involved dehydrogenases.
After enzymes and/or metabolic pathways have been identified and quantified in their contribution to the biotransformation of a drug, the impact of this xenobiotic on cellular metabolism will be investigated with the aid of glycolytic oscillations, i.e. periodic changes in cellular NADH concentration. This methodology profits from the circumstance that dehydrogenases as well as dehalogenating and carbon backbone modifying enzymes use NAD(P)H/H+. The upshot on the oscillation is a direct measure for the extent of perturbation on the metabolic network upon the uptake of a pharmaceutical. Glycolytic oscillations that are systematically perturbed by altered environmental conditions, e.g., the exposure to the xenobiotic, constitute an easily accessible measure of the intracellular behaviour since the frequency and amplitude of oscillating metabolite concentrations and fluxes depend on both the perturbation and most intracellular processes due to the coupled energy (ATP) and redox (NADH) balances.
The direct access to the Complex Pathway Simulator developed at the European Media Lab will ensure a head start for the PharmBiosim project. The project will also make use of advanced concepts from applied mathematics including algorithms for continuation and bifurcation analysis. These combined expertises are essential for the analysis of complex mathematical models. More specifically, we will describe the distribution of each relevant metabolite by a single time-dependent variable, typically the concentration.
Today the most common approach to numerically determine the unknown parameter values involves massively repeated simulations to follow the model's behaviour for many combinations of parameter values. The parameters are updated depending on the model's performance for previous choices. This "optimisation" strategy is automated by Copasi and allows a much faster convergence to a reasonable set of values than it would be possible by random or dense sampling of the high-dimensional parameter space. The PharmBiosim project will combine the above simulation-based strategy with the application of novel algorithms based on continuation and bifurcation analysis.
The programme for this project includes: (i) Initial investigations of the metabolic fates of small mono- and trisubstituted organic compounds. This will allow us to directly attribute cellular responses to the presence of functional groups in an organic substrate. (ii) Attempts to formulate general principles for xenobiotic metabolism. (iii) Establishment of a model for the bioinformation of ethyl acetoacetate. (iv) Studies of the biotransformations of ethyl 2-chloro-acetoacetate, ethyl 4-chloro-acetoacetate, and 4,4,4-trifluoro-acetoacetate.
RA3. Microcompartments associated with microtubular networks
Glucose is the principal energy source in brain and red blood cells and it is metabolized via glycolysis. In the brain the microtubular system is of special importance and it comprises almost all neuronal volume and surface area. The Institute of Enzymology in Hungary has demonstrated and quantitatively characterized the heteroassociations of some selected glycolytic enzymes with each other and with microtubular proteins and revealed the formation of distinct superstructure resulting in alterations in the enzyme properties and fluxes. The formation of superstructures are highly specific for the enzyme binding. We have begun to explore the in vivo localization of these enzymes in eukaryotic cells and have found that the pyruvate kinase that impedes the microtubule assembly shows perinuclear localisation in fibroblast cells. Recently, we have characterized the glucose conversion by multiple pathways and the consequence of microcompartmentation of enzymes by tubulin in brain extract. We plan to extend the micropathway analysis to the whole glycolytic and pentosephosphate pathways, accounting for the form-specific (isoform, oligomeric form and conformation) enzyme associations and the effects of other endogenous and exogenous factors (e.g., drugs) present in the brain tissue.
Development of drugs targeting microtubular network and calmodulin
Rapidly growing tumor cells are characterized by uncontrolled division of the cells and by a high aerobic glycolysis. Both processes are the targets of anti-tumor agents which, on one hand, decrease glucose uptake, and on the other hand destroy the dynamics of the cytoskeletal network essential for mitosis. Our data reveals that the glycolytic enzymes associated with the tubulin/microtubule can cause distinct alterations in the cytoarchitecture, which might be significantly different in normal and tumor cells, and could open new possibilities and strategies in the design of anti-tumor drugs. We have developed a complex screen system to test standard and new molecules for anti-mitotic as well as anti-calmodulin (anti-CaM) activities at molecular level. The rationale for including anti-CaM tests in the screen is that severe toxic side effects of antimitotic drugs may be related to the action of these drugs on cellular targets other than the tubulin/MT system. The promiscuity of CaM-ligand interactions makes CaM a potential additional target for chemotherapeutical agents.
Calmodulin (CaM) is a ubiquitous, Ca2+-dependent regulatory protein with multifarious functions, which has been shown to be involved in the regulation of numerous Ca2+-mediated events. We have characterized the binding of various drugs to CaM in solution as well as in crystal form, and determined the conformational changes of CaM caused by the drug binding and identified the drug-binding domain(s). We identified a new arylalkylamine derivative, N-(3,3-diphenylpropyl)-N0-[1-R-(3,4-bis-butoxyphenyl)ethyl]-propylene-diamine (AAA) with potent anti-CaM activity characterised in solution and crystal. These data indicate that each of the two AAA molecules anchors in the hydrophobic pockets of CaM, (which is the binding domain for target proteins), and the two AAA molecules interact with each other as well and mimic the structural arrangement and occupancy at the enzyme binding domain caused by target peptides.
We have demonstrated in vivo and in vitro that KAR-2, a semisynthetic bisindol derivative is a potent anti-microtubular, anti-mitotic as well as anti-cancer agent, which displays low neurotoxic side effect as compared to the know chemotherapeutic drug, vinblastine. Both drugs bind to CaM, however, KAR-2 does not show anti-CaM activity. Combination of the solution and crystal structures of CaM-drug complexes will provide evidence for the functional difference. Preliminary data indicate that it is uniquely localized in the N-terminal domain of CaM in which interdomain site might involved as well. This structural arrangement could explain why KAR-2 does not display anti-CaM activity. Concerning the molecular base of the anti-mitotic potency of KAR-2, we have shown that it induces the formation of aberrant mitotic spindles, with no apparent effect on interphase microtubules, whereas vinblastine partially destroyed interphase microtubules coexisting with normal and aberrant mitotic spindles.
This programme involves the following projects: (i) Extension of the micropathway analyses to the full glycolytic and pentosephophate pathways and investigation of the consequences of microcompartmentation of enzymes by tubulin in brain cells. (ii) Development of a mathematical model that includes the mutant enzyme functionalities and can be used to examine whether the mutation alone can account for the altered glycolytic flux observed in the presence of tubulin. (iii) Structural analyses of the TPPP/p25 protein using both isolated protein from bovine brain and recombinant human protein. (iv) Investigations of the physiological and pathological properties of the new protein family TPPP/.
RA4. Regulation of pancreatic α- and β-cells
The pancreatic cells play a central role in blood glucose homeostasis. Glucose increases insulin secretion from β-cells and somatostatin release from δ-cells, but suppresses glucagon release from α-cells. In all three cell types, exocytosis is stimulated by Ca2+ influx and the local elevation of [Ca2+]i that results from changes in the electrical activity. Thus electrical activity plays an essential role in the regulation of islet hormone release. Indeed, it is because of their different complements of ion channels that glucose has opposite effects on β- and α-cell secretion.
Glucose influences islet cell electrical activity via metabolically induced changes in the activity of ATP-sensitive K+ (KATP) channels. These channels open when metabolism is low and close when metabolism is stimulated by elevated plasma glucose. Precisely how metabolism is coupled to KATP channel closure remains unclear, but it is believed to involve metabolically induced changes in intracellular [ATP] and [ADP]. This is because ATP is found to block KATP channel activity, whereas MgADP (and MgATP) stimulates channel opening. The KATP channel is also the molecular target for common anti-diabetic drugs, such as the sulphonylureas which have been used to treat type-2 diabetes for more than half a century. Sulphonylureas close these channels independently of a low glucose metabolism, thereby eliciting β-cell electrical activity and insulin secretion.
Electrical activity and insulin release
In pancreatic β-cells, opening of KATP channels at low plasma glucose levels keeps the membrane hyperpolarized so that voltage-gated Ca2+ channels remain closed. Consequently, [Ca2+]i is low and insulin secretion is prevented. When metabolism increases, KATP channels close, depolarising the β-cell membrane, opening voltage-gated Ca2+ channels and initiating β-cell electrical activity. Electrical activity in β-cells exhibits a characteristic bursting pattern, which consists of slow oscillations in membrane potential between a depolarised plateau, on which Ca2+ dependent action potentials are superimposed, and a hyperpolarized electrically silent interval (Fig.4.2.4). As glucose is increased, the duration of the plateau phase increases and that of the silent interval decreases so that electrical activity becomes continuous at approximately 20mM glucose. This produces a graded increase in action potential firing and Ca2+ influx that contributes to the glucose-dependence of insulin secretion. A complex interaction between KATP channels, voltage-gated Ca2+ channels, Ca2+ pumps and metabolism is responsible for this bursting pattern of β-cell electrical activity and its modulation by glucose. The importance of KATP channels in regulating electrical activity (and thus insulin release) at supra-threshold glucose concentrations is illustrated by the fact that sulphonylureas, stimulate electrical activity even at 10 mM glucose. Furthermore, β-cells from mice, in which either the Kir6.2 or SUR1 gene has been deleted, fire action potentials continuously in the absence of glucose. Similar results are found for β-cells from patients with loss-of-function mutations in KATP channel genes.
Figure RA4. A. Consensus model for glucose-stimulated electrical activity and insulin secretion from the pancreatic β-cell. The β-cell membrane potential is controlled by the activity of KATP channels, via metabolically induced changes in the cytoplasmic ATP and ADP concentrations. In turn, the membrane potential controls the opening of voltage-gated Ca2+-channels, which regulate Ca2+-influx into the cell and thereby insulin release. B. Glucose-induced electrical activity. Glucose was elevated from 5 to 10 mM as indicated by the step in the top trace.
Electrical activity and glucagon release
A different complement of ion channels ensures that glucose inhibits glucagon secretion from α-cells. As in β-cells, secretion is triggered by an increase in [Ca2+]i. When α-cell metabolism is low, KATP channels are open but because α-cells possess voltage-gated Na+ channels and T-type Ca2+ channels, which are activated at more negative membrane potentials, α-cells fire action potentials in the absence of glucose. Consequently, [Ca2+]i rises and secretion is stimulated. When metabolism increases, KATP channels close, producing membrane depolarisation. This leads to inactivation of voltage-gated Na+ channels and T-type Ca2+ channels and inhibition of electrical activity. Consequently, Ca2+ influx and glucagon secretion cease.
The questions that we want to address in this project are: (i) Complete characterisation of ion channels expressed in pancreatic islet cells (molecular and biophysical characterisation) in situ, (ii) Regulation of the islet cell ion channels by cytoplasmic metabolites and intracellular Ca2+ concentrations (iii) Relationship between electrical activity and changes in the cytoplasmic free Ca2+-concentration, (iv) Control of exocytosis by rapid and extensive increases in the near-membrane Ca2+-concentration, (v) Replenishment of the release-competent pool of granules for release, (vi) Modelling the kinetics of peptide release via the narrow pore (fusion pore) connecting the granule lumen to the extracellular space, (vii) Elucidation of the role of paracrine (hormonal) and electrical (gap junctions) signalling on electrical activity and secretion (viii) Establishment of the mechanisms by which small changes in the metabolic regulation and/or activation/inactivation properties of islet cell ion channels (such as those occurring in diabetes) influence islet cell electrical activity and secretion, (ix) Modelling of the effects of drugs through drug mediated modifications of the electrical activity.
We intend to model the individual islet cells in several stages. In the case of the β-cell, for example, we will first construct a mathematical model of electrical activity based on experimentally measured ion channel parameters. Subsequently, we will incorporate a model of β-cell metabolism, so that we can address how electrical activity is modulated in response to glucose and other nutrients and, conversely, how electrical activity influence metabolism. This will involve modelling the dynamics of intracellular ATP syntheses and hydrolysis, as well as changes in nucleotide concentrations in time and space. New experiments may be required to define some of the key parameters. A more detailed knowledge of the metabolic regulation of the KATP channel, particularly the interaction of nucleotides with nucleotide-binding domains of SUR1, will also be required. In parallel, we will model intracellular Ca2+ and its relationship to electrical activity and exocytosis. This will require a detailed knowledge of release kinetics, transmembrane Ca2+ flux, Ca2+-buffering by cytosolic proteins and intracellular organelles, and spatial gradients.
The next stage will be to generate similar models for the other islet cell types. Finally, the separate models of the individual islet cells will be synthesized to produce a unified model of the whole islet. This will necessitate inclusion of electrical coupling between β-cells (non-β-cells appears not to be electrically coupled) and paracrine effects. All models will be designed so that they can easily incorporate new biological information as it becomes available. The model will be validated through their ability to reproduce experiments with compounds that are known to affect specific ion channels.
RA5. Neuronal and systemic models of mental diseases and sleep regulation.
Mental disorders are among the most common diseases. Especially the depressive disorders have a very high prevalence of about 15%. Besides substantial individual suffering this also entails enormous socioeconomic costs. Depressive disorders can be treated to some extent by anti-depressant drugs, an often used type being the selective serotonine re-uptake inhibitors (SSRIs). These substances enhance the neuronal actions of serotonine. Serotonine is a neurotransmitter which contributes to information processing in many areas of the brain – including those that are involved in mood control and sleep-wake cycles. Remarkably, the anti-depressive drug effects occur with long-time delays (two-weeks or more) which indicates that the desired effects are caused by secondary changes.
The cause of depressive disorders is only known in general terms. Pathophysiological factors which are discussed include disturbances of different neurotransmitter systems, dysfunctions in neuroendocrine systems for hormone release, and disorders of autonomous systems, e.g. in the control of sleep-wake cycles.
All these factors are intimately linked to each other and it is probably a combination of different factors that leads to a depression. In addition to the neuronal, humoral and genetic factors, also psychosocial stressors, at least initially, play an essential role. However, depressive disorders are recurrent and progressive. Each disease episode increases the probability of successive episodes and continuously shortens the time-intervals. In such severe states of chronic depression, the efficiency of conventional drug therapy is considerably reduced.
Anti-depressive effects can also be observed after sleep deprivation. The effects occur much faster than on any standard drug treatment - already the following day. However, the improvement does not last very long. It would be a significant progress for the treatment of depressive disorders to develop a drug therapy which combines the advantages of fast onset (sleep deprivation) and long duration (anti-depressant drugs). To identify the relevant neurobiological principles it will be necessary to examine the interactions between the different systems at different functional levels and time-scales. Considering the complexity of these systems, with the meshing of a manifold of nonlinear feedback loops, it is evident that a reliable understanding can only be obtained with the help of computer simulations.
To define the functional principles which are relevant for anti-depressive drug treatment we will develop computer models at different functional levels and in connection with a number of different phenomena:
Anti-depressant drug treatment
Whereas depressive disorders can be treated effectively to some extent by anti-depressant drugs that interact with relevant neurotransmitter systems (e.g., selective serotonine re-uptake inhibitors "SSRIs"). Problems associated with drug treatment are long delays in the onset of anti-depressive action, poorer outcome in severe depressive states and, in particular, chronic depression as well as definition of drug selection criteria. Optimization of the effective, safe and selective anti-depressant drug treatment is a major issue which can benefit from the use of computational approaches including data banking, biosimulation and system analysis. Our approach will address two specific issues - regulation and plasticity - with respect to different biological levels as well as with respect to different levels of analysis (i.e. detailed vs. simplified simulations).
Network regulation
One of the challenges arising from the different pathophysiological factors is - besides characterisation of the major pathobiological players - to understand the regulatory issues and network dynamics that arise from the interplay of these factors. For example, alteration of a specific neurotransmitter system such as serotonin has impact on cellular and subcellular systems (from neural transmission via second messenger systems to gene expression) as well as system level effects on HPA-axis regulation and sleep architecture. Moreover, interactions and links exist to systems involved in vegetative regulations such as temperature, body weight, heart rate variability, platelet function etc. These latter interactions are of particular interest with regard to side effects of drugs and thus with regard to safety issues of drug development as well as with respect to drug prescription to patients with co-morbid disorders such heart diseases. Additional, more mathematical oriented aspects of network regulation are considered in RA11.
Adaptation and plasticity
The second specific task - closely associated to regulatory mechanisms - is understanding of the longterm adaptive and maladaptive effects relevant for the pathology of mood disorders as well as with regard to the actions of drugs. Examples are up- and down-regulation of receptors (e.g. serotonergic, noradrenergic) as resulting from disease mechanisms and drug intervention, sensitisation and kindling-like effects (e.g., changes of NMDA-dependent neurotransmission, kindling-increased protein kinase C (PKC) - Ca2+ feedforward activity), sensitisation of neuroendocrine systems (e.g., HPA-axis modification by altered GABA/glutamate equilibrium at the level of the hypothalamus).
RA6. Synchronisation of nephron pressure and flow regulation
The kidneys play an important role in regulating the blood pressure and maintaining a proper environment for the cells of the body. The process of glomerular ultrafiltration is highly sensitive to variations in the blood pressure, and a proper regulation of the excretion of water and salts involve mechanisms that can compensate for variations in the arterial blood pressure. Two different physiological mechanisms that operate at the level of the individual nephron are known to be important in this regulation:
The tubuloglomerular feedback (TGF) mechanism regulates the resistance of the afferent arteriole in response to changes in the tubular NaCl concentration at the macula densa. Experiments with rats have shown that this mechanism tends to be unstable and produce self-sustained oscillations in the proximal tubular pressure with a period of 30-40 s. With different amplitudes and phases, similar oscillations are observed in the distal tubular pressure and in the chloride concentration at the terminal part of the loop of Henle.
The myogenic mechanism is an intrinsic response of the preglomerular vessels in which an increase of the transmural pressure results in a vascular constriction. This mechanism also produces oscillations, although with a period of 4-6 s. Detailed analyses of experimental pressure data have revealed that the two oscillatory modes can synchronise with one another, and that variations in the blood pressure can lead the individual nephron to jump from one state of synchronisation to another.
Nephrons are typically arranged in pairs or triplets with their afferent arterioles branching off from the same interlobular artery. This structure allows the flow regulations of one nephron to interfere with the regulations in adjacent nephrons in at least two different ways (a) via a vascularly propagated interaction where signals from one nephron spread to neighbouring nephrons through gap junction mediated cell-to-cell interactions of the smooth muscle cells of the arteriolar wall, and (b) via a direct hemodynamic coupling that arises from a displacement of part of the blood flow from a nephron that contracts its afferent arteriole to its neighbours.
Our experimental investigations have shown that these interactions can cause neighbouring nephrons to operate in synchrony. Both in-phase and anti-phase synchronization have been reported for the regular pressure oscillations in normotensive rats, and signs of chaotic synchronization have been demonstrated for hypertensive rats. Moreover, detailed analyses of the tubular pressure data for neighbouring nephrons have shown that they can operate in a state where only the slow (TGF-mediated) oscillations are synchronized, or in a state where both fast and slow oscillations are synchronized. We have recently developed a detailed model of two coupled nephrons. This model can account for all the observed synchronization phenomena. We have also performed detailed data analyses of the pressure recordings from neighbouring nephrons to elucidate the influence of arterial pressure variations on the bi-modal dynamics associated with the pressure of slow and fast regulatory processes.
The questions we want to address are: (i) What is the length scale of the synchronisation phenomena, i.e., to what extent can the synchronisation spread to nephrons situated along adjacent interlobular arteries or to juxtamedullary nephrons? (ii) What role do synchronisation phenomena play in connection with development of hypertension? It is well-known that the vascularly mediated interactions are stronger for hypertensive rats than for normotensive rats, and (iii) How variable are the synchronization patterns, and to what degree are they affected by anti-hypertensive drugs or gap junction inhibitory peptides?
It is well-known from cross-transplantation studies that the kidneys play a central role in the pathogenesis of hypertension. The mechanisms underlying this effect are still largely unknown. By analogy with similar phenomena in other types of tissue, synchronisation between the nephrons may play a hitherto unexamined role in the kidneys response to external perturbations such as, for instance, variations in the arterial blood pressure. One would expect large groups of synchronised nephrons to give a stronger and faster response than the same nephrons in the absence of synchronisation.
The project involves: (i) Development of new experimental techniques to investigate the synchronisation patterns of surface nephrons, and techniques to study the regulatory dynamics of juxtamedullary nephrons. (ii) Development of new data analysis techniques to reveal the details of various synchronisation states. (iii) Formulation of a large-scale simulation model of 20-30 interacting nephrons, a model that can account both for the complex dynamics of the individual nephron and for the complex structure of the arteriolar tree. (iv) Detailed modelling of signal propagation along the arteriolar vessels. This part of the project will be performed in collaboration with a project on coupled smooth muscle cells. (v) A series of experiments to investigate the influence of various antihypertensive drugs or gap junction inhibitory peptides on both the function of the individual nephron and on the coupling between the nephrons.
We have previously demonstrated that introduction of 2 kidney 1 clip (2K-1C) hypertension in rats is accompanied by a transition from regular to chaotic oscillations in the unclipped kidney. We plan to follow the global pattern of synchronisation in 2K-1C hypertensive rats during 4-5 weeks after clipping when the arterial pressure rises. We will record from the surface of both the unclipped kidney (which is exposed to high blood pressure) and of the clipped kidney (which operates under normo- to hypotensive conditions).
Since the arteriolar network can be mapped out, and the lengths and diameters of the various vessels determined, it is possible to obtain an independent estimate of the typical strengths of both the hemodynamic and the vascularly propagated coupling. Using this information we will build a model of an interlobular artery together with the 20-30 nephrons associated with the artery. Nephrons positioned near the top and bottom of this tree will operate under different conditions (e.g., difference in perfusion pressure and in the strength of the TGF mechanism), and we intend to account for this variation in the model. In this way the model will allow a detailed examination of the possible modes of synchronisation between nephrons.
The coupled nephron model will provide an important contribution to (i) our understanding of kidney regulation, (ii) our understanding of how an ensemble of biological oscillators, each displaying complicated nonlinear dynamic phenomena, can interact to produce different forms of coherent behaviour on a higher structural level, and (iii) of the possible effects of anti-hypertensive drugs or gap junction inhibitory peptides on kidney function.
RA7. Models of full-scale cardiac arrhythmias
Research performed at The Department of Physiology, University of Oxford and at the School of Medical Sciences, University of Leeds creates the platform on which we can now construct models of full-scale cardiac arrhythmia, including re-entrant arrhythmias using detailed biophysical cell models. To date most work in this area has been done with greatly simplified cell models. The computing resources now available to us (including the Oxford Supercomputer Centre) make it possible to attempt such reconstructions with more biophysically detailed models.
The following arrhythmia mechanisms are ready for incorporation into tissue and whole organ models:
Sodium Channel Mutations leading to Early After-depolarizations (EADs). A number of mutations are known that trigger repolarization failure, some of which have been, at least partially, modelled. Two of these are mutations in the sodium channel. The simplest from a biophysical point of view is a missense mutation on the outer surface of domain 4 of the channel protein that alters the charge on that domain and so influences the voltage sensitivity of the inactivation gate formed by the intracellular loop between domains 3 and 4. This shifts the inactivation curve in a depolarizing direction by around 10-20 mV, so increasing the overlap between the opening and closing functions and increasing the sodium "window" current. This in turn triggers EADs, which have been modelled by the Oxford group in interaction with plasma potassium level and in interaction with K channel blocking drugs. This is the easiest way to trigger these EADs and could readily be incorporated into organ level models. The real situation at the molecular level is more complex, and involves a multi-state sodium channel, but this would be computationally much heavier. We will start with the simplest way of inducing the phenomenon. We have, however, coded up the multi-state sodium channel in CellML format, so this could eventually be plugged in.
Drug-induced EADs Class III drugs, acting to inhibit K channels, are well-known to induce torsade de pointes arrhythmias. These are re-entrant arrhythmias where the vector shifts from cycle to cycle so producing a "rotating dance" (torsade) of the main vector - this generates the characteristic slow oscillations in the amplitude of the large triangular ECG waves. All the cell models show EADs or repolarization failure when iKr (hERG + mink) is inhibited by around 80-90% and it is presumed that this is the cellular basis for triggering torsades de pointes arrhythmia. Although this presumption is widespread there is no systematic proof that this is the case. Incorporating iKr block into cell models in whole heart simulations is one of the ways in which this hypothesis could be tested. An important question will be the role of pre-existing inhomogeneity in predisposing towards this type of arrhythmia. Proof of this hypothesis is also of great importance to understanding the way in which large numbers of drugs produce QT side-effects since iKr is the target of many drugs with this kind of side-effect.
Sympathetic overdrive, including exercise. This is thought to act by increasing iCaL. We have done an integrative study of this mechanism in the context of exercise arrhythmias showing how the combination of up-regulation of iCaL and increased extracellular potassium protects against such arrhythmias. The converse of this situation is that, in low potassium, such action would be dangerous. Post exercise, there is a fall in [K]o. Low potassium has also been examined in the context of sodium channel mutation effects.
Delayed after-depolarizations. Sodium-calcium overload produced by sodium pump inhibition, ischaemia and other conditions generates intracellular calcium oscillations which are well-reproduced by the models. These oscillations then stimulate oscillatory inward currents carried by sodium-calcium exchange which, if large enough, can initiate ectopic beats. Computations done on a 2D ventricular network show that this mechanism can produce propagated ectopic beats but it has not yet been incorporated into whole organ modelling. The interest here would be to investigate the conditions under which the collision of an ectopic beat with the normal wave of excitation could trigger a re-entrant arrhythmia.
Slowed conduction. Slowed conduction is well-known to produce re-entrant arrhythmia and this is the mechanism by which class I drugs (sodium channel blockers) produce their arrhythmic effects. The process concerned almost certainly involves re-entry arising from longer pathway times due to the slowed conduction. Genetic manipulation of the sodium channel has also been shown to trigger this kind of arrhythmia. It will be possible to reproduce this in whole ventricle simulations. One of the interesting questions is what effect inhomogeneity of gene expression levels may have on this kind of arrhythmia.
The role of anatomical detail, including pathological changes, and of electrophysiological inhomogeneity needs to be assessed in all these forms of arrhythmic mechanism. There is a lot of modularity here in the sense that, although there are many different mechanisms of arrhythmia, the basic mechanisms (EAD, DAD, re-entry) keep recurring. There will therefore be a lot of transfer from one project to the others. The time is now ripe to do this. There is enough understanding of the cellular processes and there is now enough computing power. The latter development is very recent, which is why very little 3 dimensional reconstruction has been done so far.
RA8. Spatio-temporal organization of intra- and inter-cellular Ca2+ dynamics
Calcium is a widespread second messenger, mediating important physiological responses in all types of organisms, from bacteria to specialised neurons. It has been known for about 15 years that the calcium increases induced by an external stimulation are highly organized, both in time and space. Indeed, the rise in cytosolic calcium concentration occurs in the form of repetitive calcium spikes. These calcium oscillations are observed in most cell types and are considered as a prototype of oscillating system in cellular biology. Moreover, each calcium spike is also organised at the spatial level; the rise in calcium concentration is first restricted to a portion of the cell, and later invades the whole cell as a wave.
It is well-established that calcium oscillations result from a periodic exchange of calcium between the cytosol and the internal calcium stores (endoplasmic reticulum). In response to the external stimulation, inositol 1,4,5-trisphosphate (InsP3) is synthesized in the cytoplasm. InsP3 receptors are calcium channels located in the membrane of the endoplasmic reticulum. Periodic release of calcium from the reticulum can be ascribed to the autocatalytic regulation by which calcium can activate its own release through the InsP3 receptors.
Stochastic simulations of the activity of the InsP3 receptor
In response to a stimulation of appropriate amplitude, the InsP3-regulated calcium channels open in a coordinated manner through the whole cell. However, when InsP3 concentration is low, transient and non-coordinated increases in cytosolic calcium concentration can be observed. The latter increases are very localized, both in time (0.5 ms) and space (1µm); they probably correspond to the opening of a few calcium channels.
In order to study the behaviour of a small number of InsP3 receptors, it is necessary to resort to new types of numerical simulations based on a quantitative description of the stochastic transitions between the various molecular states of a calcium channel. Recently, we have used the latter type of stochastic simulations to model the behaviour of an isolated channel in its physiological cytoplasmic environment. We have been able to show that, although the mean opening time of a InsP3 receptor/ calcium channel is equal to 3 ms , the activity of such a channel in physiological conditions is much prolonged due to the repetitive opening of this channel before desensitization. Because InsP3 receptors possess regulatory calcium binding sites, an open channel can send out a signal to adjacent calcium channels. Such a concerted opening may occur only if the distance between neighboring channels is small enough.
We plan to carry on our quantitative analysis of the intracellular calcium dynamics based on the detailed properties of individual channels. Stochastic simulations will be used to study the behaviour of several interacting sites; such a system could thus represent the whole cell. Our final aim is to understand how the stochastic behaviour observed at the level of a restricted number of channels can give rise to regular and periodic calcium spikes when several sites are supposed to interact.
Models for intercellular Ca2+ waves in liver cell plates
Some years after the discovery of intracellular calcium waves, calcium signals have been seen to propagate intercellularly in tracheal epithelial cells or glial cells. A model based on the passive diffusion of InsP3 through gap junctions can account for the existence and the main characteristics of these intercellular calcium waves.
Intercellular calcium waves have also been observed in the liver. These waves propagate along a unidirectional set of connected hepatocytes known as 'liver cell plate'. The latter plays an important role in physiological phenomena, such as bile secretion and canicular contraction. Three main characteristics of intercellular calcium wave propagation have thus been identified. First, each cell of the multiplet requires the presence of the stimulus in order to relay the intercellular wave. Second, the liver cell plate is characterized by a gradient of sensitivity to the calcium-mobilizing hormones. Finally, gap junctional connectivity is required for intercellular propagation of the calcium signal. However, from an experimental point of view, the nature of the messenger diffusing through gap junctions (calcium or InsP3) is still unknown.
We have developed a theoretical model based on the coupling between several oscillators (i.e. the individual cells of the multiplet), whose dynamics is described by a model previously proposed to account for intracellular calcium oscillations. Numerical integration of the model shows that it is possible to coordinate calcium spiking among connected hepatocytes when it is assumed that InsP3 can somewhat diffuse through gap junctions; calcium spiking however occurs with a slight phase-shift among connected cells, giving rise to the appearance of a phenomenon of wave propagation. The direction of the wave is imposed by the direction of the sensitivity gradient. Models of intra- and intercellular calcium waves will also be developed for pancreatic β-cells (see RA4) and for smooth muscle cells in the arteriolar wall (see RA6).
Frequency coding of pulsatile Ca2+ signals
In most cell types, the level of external stimulation imposes both the frequency of calcium oscillations and the intensity of the physiological responses that are controlled by the level of cytosolic calcium. Various cellular responses such as secretion, contraction or genetic expression are thus controlled by variations in the level of intracellular calcium. The molecular mechanism by which the frequency of calcium oscillations can be decoded by the cell remains poorly understood. We plan to analyze two examples of frequency coding of calcium oscillations based on distinct mechanisms: (i) Decoding by the calmodulin-dependent protein kinase II (CaMKII), (ii) Role in phosphorylase-mediated degradation of glycogen.
RA9. Modelling of molecular regulatory mechanisms of circadian rhythms
Circadian rhythms occur with a period close to 24 h in all eukaryotic and some prokaryotic organisms. These rhythms have a profound influence on human biological processes. In recent years The Department of Physical Chemistry in Brussels has proposed some of the first experimentally based theoretical models for circadian rhythms. We intend to pursue this line of research by investigating new aspects of these rhythms. We intend to focus on the detailed molecular mechanism of circadian clocks in Drosophila and mammals. We also plan to investigate the origin and consequences of circadian disorders in human physiology.
Suppression of circadian rhythms by light pulses
A remarkable property of circadian rhythms is that a single pulse of light can suppress them in a transient or even permanent manner. Permanent suppression of the rhythm has been observed in insects, plants, and mammals including humans. Once the rhythm is suppressed, it is often possible to restore it by a second pulse of the same duration and amplitude. The suppression phenomenon is of particular interest as it provides unique informations on the dynamics responsible for circadian oscillations. Permanent suppression can be interpreted in terms of a coexistence between a stable limit cycle and a stable steady state; the effect of a critical light pulse suppressing oscillatory behaviour would be to bring the system back to its singularity, as proposed by Winfree, although the situation discussed by him primarily pertains to the case where a stable limit cycle surrounds an unstable steady state (only transient suppression can then occur in response to a critical light pulse).
The model proposed by Leloup and Goldbeter describes in detail the molecular mechanism responsible for circadian rhythms in Drosophila. This model incorporates the effect of light which is to induce degradation of the TIM protein. By means of this model one can therefore determine the effect of a light pulse. Our study showed that critical light pulses can suppress circadian rhythms in a permanent manner and that the rhythm can be restored by a second perturbation identical to the first pulse. The model proves useful in allowing quantification of the duration and amplitude of the effect of a light pulse necessary to suppress rhythmic behaviour. Numerical simulations of the model also indicate the phases of the limit cycle at which light pulses can suppress circadian rhythmicity. If the explanation turns out to be correct, these results provide a clear example where a computational model is required to account for experimental observations, since without a model there is no way to predict the coexistence between a stable steady state and a stable rhythm.
Molecular mechanism of circadian rhythms in mammals
Remarkable progress has recently been made in unravelling the molecular mechanism of the circadian clock in mammals, where the circadian pacemaker is located in neurons of the suprachiasmatic nuclei in the hypothalamus. As in Drosophila, the molecular mechanism of circadian rhythms relies on indirect negative autoregulation of gene expression. The genes per, tim, bmal1 (homolog of cyc) and clock have been identified in mammals. However, marked differences exist with respect to Drosophila as to the effect of light. Three per genes (per1, per2, per3) have thus been characterized in mammals; light acts by inducing their expression rather than by triggering the degradation of the TIM protein as in Drosophila. A complex between the PER and CRY proteins binds to, and thereby inactivates, a complex between the BMAL1 and CLOCK proteins, which promote the expression of the per and cry genes.
Circadian rhythms and chronopharmacology
An important aspect of circadian rhythms pertains to the implications of these rhythms for pharmacology. Given that most physiological functions vary in a circadian manner, it is not surprising that the toxicity of many drugs as well as their efficacy vary in the course of the day with a circadian period. This aspect has long remained practically unnoticed in pharmacology but is increasingly gaining interest as shown by the slow but sure development of chronopharmacology whose goal is to determine the optimal timing for administration of medications, as a function of the physiological rhythms of the patient. The most convincing advances in this field are probably those made in cancer therapy, where multicentric trials of phase III are under way for the treatment of colon cancer. We think that modelling studies based on the pharmacokinetics of the drug and of the circadian rhythms involved in drug action and degradation should contribute to optimize the patterns of drug delivery.
RAS-regulated signalling cascades
We will also study signalling cascades in mammalian cells in order to understand oncogenic transformations due to RAS mutations. Using nylon filters, customized glass chips, and Affymetrix arrays we will analyse the differential gene expression in normal, transformed and revertant cells. The role of DNA methylation and histone modifications will be explored by studying cells treated with inhibitors of methyltransferases and histone acetylases. Promoters of coregulated genes are predicted in upstream regions and analysed. For this purpose we look for overrepresented motifs, clusters of motifs and regions conserved in mammalian genomes. Furthermore, we study time-series data of target genes and (activated) signalling proteins in order to refine mathematical models of MAPK and PI3K-cascades. In particular, we analyse ultrasensivity and bistability predicted by models of the MAPK-cascade.
Huntington's Disease The neurodegenerative disease Chorea Huntington will be studied in appropriate model systems. The proteome of a transgenic mouse model has been compared with the proteome of a control strain. It turned out that only few proteins are clearly differentially expressed. The quantitative analysis of large 2D-gels will be refined in order to detect also less pronounced changes of protein concentrations. In a cell model of chorea Huntington we will monitor the formation of Huntington aggregates. At subsequent days we will extract RNA and hybridise Affymetrix-chips in order to get hints for dysregulatory processes. In parallel, we design a customised array to dissect the possible signalling cascades and feedback loops affected by Huntington aggregation. A preliminary model of aggregate formation has been developed that includes feedback loops due to proteasome inhibition, caspase activation, and transcription factor depletion.
RA10. Deep brain stimulation and medication
In several neurological diseases like Parkinson's disease (PD) or essential tremor brain function is severely impaired by synchronization processes. Parkinsonian resting tremor appears to be caused by a population of neurons located in the thalamus and the basal ganglia. These neurons fire in a synchronized and intrinsically periodical manner at a frequency similar to that of the tremor, regardless of any feedback signals. In contrast, under physiological conditions these neurons fire incoherently. In patients with PD this cluster acts like a pacemaker and activates premotor areas and the motor cortex, where the latter synchronize their oscillatory activity.
In patients with advanced PD or essential tremor who no longer respond to sufficiently drug therapy, depth electrodes are chronically implanted in target areas like the thalamic ventralis intermedius nucleus or the subthalamic nucleus. Electrical deep brain stimulation (DBS) is performed by administering a permanent high-frequency (> 100 Hz) periodic pulse train via the depth electrodes. High-frequency DBS has been developed empirically, and its mechanism is not yet fully understood. Permanent high-frequency stimulation basically mimics the effect of tissue lesioning by suppressing neuronal firing, which, in turn, suppresses the peripheral tremor. High-frequency DBS is reversible and has a much lower rate of side effects than lesioning with thermocoagulation. Nevertheless, e.g., due to current spread, high-frequency DBS may lead to severe side effects like dysarthria, dysesthesia, cerebellar ataxia.
For this reason stimulation techniques have been developed which aim at desynchronizing the pacemaker's pathologically synchronized firing in a demand-controlled way instead of simply suppressing the neuronal firing. These methods share one particular feature: Each stimulus consists of two qualitatively different stimuli. The first stimulus is stronger and resets (restarts) the cluster, whereas the second, weaker stimulus is a single pulse which is administered after a constant time delay and desynchronizes by hitting the cluster in a vulnerable state. The goal of the reset is to control the dynamics of the cluster by restarting the cluster in a stereotypical way. The reset may be achieved by means of a strong single pulse, a high-frequency pulse train or a low-frequency pulse train.
A desynchronising single pulse is only effective provided it hits the population very precisely at a vulnerable phase. Hence, the composite stimulation techniques only work with an effective reset which may require high stimulation intensities, and minor variations of critical stimulation parameters (especially the delay between resetting stimulus and desynchronizing single pulse) may abolish the desynchronizing effect. Thus, these methods are not robust against larger variations of the model parameters (frequencies of the neurons and strength of synaptic interactions). Such variations may be closely related to variations of blood concentrations of drugs like DOPA. Furthermore, these techniques require a time-consuming calibration procedure lasting for more than 30 min (based on series of test stimuli), which complicates the testing in patients during operation (i.e., electrode implantation).
For this reason, further demand-controlled stimulation techniques have been developed, which are based on a different stimulation principle. Instead of directly desynchronizing a synchronized population of neurons with well-calibrated stimuli, these techniques shift the synchronized population to a state which can be reached in a robust and reliable way: For this, a coordinated reset of neural subpopulations via several or multiple electrodes is performed. As soon as it reaches the cluster state, the population spontaneously relaxes from the cluster state to a nearly uniformly desynchronized state. This spontaneous transition is due to the self-organizing principles governing the pathological synchronization process.
The goal of the BioSim project is to study the interplay between medication (e.g., DOPA) and demand-controlled DBS in order to improve the clinical outcome. Specifically, by means of a modelling and computer simulation approach, we shall investigate the impact of variations of the drug concentration on both the stimulation outcome and on how the stimulator may compensate for this by appropriate learning algorithms. The ultimate goal of our approach is an intelligent device for demand-controlled deep brain stimulation that reacts on changes of system parameters (related to changes of the stimulation outcome) by appropriately adapting stimulation parameters.
The project will be divided into three main parts: (i) Derive a physiologically realistic model of a neural network in the typical target areas of DBS (e.g., subthalamic nucleus and thalamus). (ii) Investigate how electrical stimulation via macroelectrodes can be modelled appropriately on such a microscopic level of description. (iii) Study how variations of the blood concentration of drugs (such as DOPA) show up in terms of varying model parameters (e.g., synaptic strength). (iV) Incorporate learning algorithms into the demand-controlled stimulation techniques in order to compensate for physiologically realistic variations of model parameters. In particular, test the performance of different demand-controlled stimulation techniques (see above) under the influence of such variations and supported by appropriate learning algorithms.
RA11. Biological networks, data analysis, and pharmacokinetic models
A central theme for this project is the integration of genomes and high-throughput data with mathematical modelling of cellular processes. We study statistical dependences and periodicities in complete genomes, quantify reproducibility of DNA-chips and large 2D-gels, and develop tools for the analysis of promoters of co-regulated genes. These bioinformatic techniques are applied to specific cellular systems as a prerequisite of mathematical modelling. The project includes the following themes:
Regulatory networks: statistical links between models and data
This project aims to establish, in a common conceptual framework, statistical links between theoretical formulations of a regulatory network model and observed multivariate time series. The approach is based on previous experience with biological systems. Following the development of the appropriate adapted stochastic models for use with the estimation machinery, we will use computer algorithms to identify the parameter values and network structures which provide the best approximation to the data corresponding to the model chosen. Depending on the complexity of the model, it may be possible to estimate the parameters directly by maximising the likelihood using a nonlinear optimisation procedure. In more complex systems, data imputation algorithms such as MCMC provide a natural and unifying approach. A Bayesian model formulation will be used to incorporate priors and we will investigate the use of bioinformatic databases to set some of the structural priors. This framework will incorporate the uncertainty in the estimation of the parameters and the imputed processes. We will also develop a more consistent approach to uncertainties in the measurement process.
Modelling transport and protein sorting across cell membranes and compartments
The project studies the effects of spatial heterogeneity and spatial segregation (of parts) of biological regulatory networks in the cell. The eukaryotic cell is subdivided into many membrane-enclosed compartments, which form functional subunits with their own characteristic sets of proteins and other molecules. It is of central importance to understand the additional degrees of freedom or robustness such a spatial organisation gives to the cell, especially from an evolutionary point of view. The project will provide and develop the analytical and mainly numerical tools needed for such an analysis. It focuses on two examples, the gated transport of molecules between the nucleus and the cytosol, and targeting of thylakoid lumen proteins. The long-term aim is a more direct comparison of spatially explicit simulations of cellular substructures with combined data from protein-protein interaction studies and experimental visualisatio
Virtual populations
The current process of drug development involves activities that are organised sequentially for the purpose of drug selection. A parallel approach to testing all possible compounds is difficult to apply currently considering the cost, time and huge number of compounds that are produced by combinatorial chemistry techniques. However, surrogate in silico techniques for screening tools can be applied simultaneously to all candidate drugs. The parallel nature of this approach means that the attrition rate of potentially useful compounds would depend on the most non-specific test amongst all of the tests. Thus, if the least specific in silico test is 40% reliable then an attrition of 60% is expected, which still is better than the previous sequential model despite the high specificity of real tests as opposed to in silico tests. Clinical studies are among the most expensive steps during drug development. Thus, the objective of this project is to build a "Virtual Human Population" pertinent to testing drugs for their pharmacokinetic properties and to explore the possibility of a parallel approach to the current drug research & development process. The components of the project will utilise current scientific knowledge to build in silico models as surrogates for clinical studies. The models cover a wide range of processes from predictive models of drug absorption to models of disease groups with specific attributes related to the absorption, metabolism and excretion of drugs.
Nonlinear signal analysis
To characterise the behaviour of complex feedback systems and regulatory networks it is essential to develop data-driven mathematical models that can enable a systematic test of biological hypotheses. We intend to contribute through work in (i) pre-processing and nonlinear analysis of multivariant data by means of Independent Component Analysis (ICA), Recurrent Plot Analysis (RPA), and multiscale analysis (e.g., wavelets). The main aims will be to identify and quantify nonlinear correlations in experimental data and to compare the data with simulation results, and to (ii) analyse complex synchronisation processes. This work will be particularly important to the research activities on synchronisation of nephron pressure and flow regulation (RA6) and on models of full-scale cardiac arrhythmia (RA7).
|