Final Report Summary - UNICELLSYS (Eukaryotic unicellular organism biology – systems biology of the control of cell growth and proliferation)
The overall objective of UNICELLSYS was a quantitative understanding of fundamental characteristics of eukaryotic unicellular organism biology: how cell growth and proliferation are controlled and coordinated by extracellular and intrinsic stimuli. UNICELLSYS integrated quantitative experimentation with simulations of dynamic mathematical models in a systems biology approach.
Unicellular biology can be described at different levels of organisation: cell population; single cell; “whole-cell” molecular networks; large interconnected systems of biomolecules operating at different time scales and/or different types of molecular reactions; defined functional modules. UNICELLSYS addressed and integrated these levels. A major deliverable of the Project are computational simulations based on predictive mathematical models that enable the investigator to observe the response to different stimuli of a cell population as well as at levels of increasing detail. These simulations allow the investigator to make predictions of the effects of physiological, genetic and pharmacological perturbations.
UNICELLSYS employed baker’s yeast to study the control of proliferation (increase in cell number) and cell growth (increase in volume and mass) in response to external and intrinsic stimuli: nutrient availability; stress; hormone. Many details of the metabolic pathways (central carbon, energy and nitrogen metabolism), signalling pathways (PKA: protein kinase A; TOR: target of rapamycin; Snf1: AMP-activated kinase; HOG, PHD, PKC, STE: Mitogen-Activated Protein Kinase (MAPK) pathways; Snf3/Rgt2: sugar sensing) as well as the cell cycle machinery are highly conserved across eukaryotes. It is generally accepted that not only the molecules and their connections but also many system-level principles are conserved from yeast to human. Since these molecular systems also play central roles in human biology at the level of module, system and subsystems, network, cell, and organ, they are highly important for medical conditions ranging from the metabolic syndrome to cancer. For these reasons, a better understanding of quantitative system properties and behaviour of fundamental cellular control mechanisms gained in UNICELLSYS will have significant biomedical importance.
UNICELLSYS operated at the forefront of the field of systems biology and its results, new technologies for experimentation and modelling and the insight generated will have long lasting effects, in particular with respect to:
1. New tools for dynamic modelling in computational systems biology.
2. New and optimised tools of generic suitability for quantitative measurements as required for dynamic modelling, including at the level of single living cells.
3. New knowledge on the quantitative and system properties of the central regulatory mechanisms of eukaryotic cells.
4. A large number of different networks, ranging from genome-scale to detailed kinetic models of smaller sub-networks, for further use in research.
5. Mechanisms for interdisciplinary collaboration between experimentalists and theoreticians in systems biology.
UNICELLSYS integrated data-driven and model-driven approaches. Project activities can be summarised as: (i) computational reconstruction and modelling the different levels of biological organisation of unicellular eukaryotic biology, (ii) generation of quantitative data for modelling, (iii) dissemination of project results and new knowledge.
UNICELLSYS brought together a consortium of leading experimental and theoretical biologists. It created new opportunities for preventing, diagnosing and treating perhaps complex diseases and designing bio-engineering approaches. The project has trained a large number of young scientists in interdisciplinary systems biology and has advanced the careers of several of its researchers. The field will benefit from results, approaches, tools and trained work force for the years to come.
Project Context and Objectives:
UNICELLSYS has achieved many, though not all of its ambitious. Main focus in the second half of the project period has been the ultimate project challenge, the connection and integration of dynamic models of different cellular pathways (metabolism, signalling, cell cycle) and for modelling different levels of organisation (from cell population to single cell and defined molecular systems).
UNICELLSYS has produced and further improved a consensus genome-wide reconstruction of the metabolic network of yeast. The project has also generated reconstructions of the yeast signalling networks (based on protein-protein interaction data as well as literature data, respectively) as well as of the cell cycle. These reconstructions were guided by existing data as well as by topological data generated within UNICELLSYS. The reconstructions serve as important guidance for dynamic models and are continuously updated with new data/information and guide further activities within and outside UNICELLSYS. They will have lasting effects for research after UNICELLSYS.
Data for dynamic models have being produced at three levels:
(1) Global proteomics, phosphoproteomics transcriptomics and metabolomics data have been generated in different partner laboratories under highly controlled conditions. This work required also the development of new methodology. Several different growth conditions were employed and time course data suitable for modelling have been generated. Focus has mainly been on TOR and osmostress signalling.
(2) Targeted specific datasets on particular cellular modules (signalling pathways) using specific shifts of growth conditions have been generated for all modules included in the scope of UNICELLSYS. The number of experimental data available through the data portal has further increased during the reporting period and pushed ahead the work on dynamic models of specific cellular modules.
(3) Generating data at single cell level involved substantial method development and optimisation. Numerous reporters for single cell analyses have been generated and employed during the reporting period. UNICELLSYS measured population data on cell-to-cell variation followed cellular processes in real time. This latter approach employed measurements on protein movements and diffusion (FRAP, FLIP) as well as microfluidics for highly controlled conditions. Main focus has been on the osmostress HOG, the Snf1 and the PKA pathways, but also on the pheromone and cell wall integrity MAPK pathways.
A large set of initial mathematical models has been developed over the entire project period. These models are based on experimental data retrieved in the respective WPs or on published experimental data. Highlights included a dynamic model of central metabolism, a model connecting HOG signalling with glycerol and central metabolism, models connecting Snf1 signalling with cell cycle control, and models connecting cell cycle regulation with other cellular processes. These models form the basis for further refinement and for integration towards large-scale models. Advanced model decomposition methods relying on monotone systems represent a clear progress beyond the state-of-the-art. Moreover, aspects of the overall concept for modular decomposition and linking of different network representations were developed.
Significant efforts went into developing the concepts and approaches for generating large scale dynamic models that connect different signalling systems. Methodology has been developed for multiscale modelling to integrate cell population and cell-to-cell variability data with molecular network dynamics.
UNICELLSYS has generated a wealth of new biological insight concerning the mechanisms of signal transduction, metabolism and cell growth and proliferation, documented by a large number of published and upcoming papers. Highlights include mechanistic links between signaling pathways but also between signaling, metabolism and cell cycle.
The overall outcome of UNICELLSYS is a better quantitative and integrative understanding of central and conserved control mechanisms of cell growth and proliferation in eukaryotic cells. In this way UNICELLSYS contributes to the field of Systems Biology, which aims at understanding the properties and phenotypes that emerge through the interaction of bio-molecules, cells and organs in the body. It is expected that understanding those properties will have major impact on diagnosis and treatment of diseases. The ability to use computational models of cells, tissues and eventual the entire human body is expected to enable predictive and personalised treatment of diseases. UNICELLSYS has already inspired initiatives that aim at generating computational models of human as well as building a European Systems Biology infrastructure.
Specifically, main expected results include:
1. Systematically reconstructed cellular networks on the basis of genomics, omics and literature data. This includes metabolic, signaling and other regulatory networks.
2. High-resolution coupled data sets available through a data repository for dynamic modelling of cellular systems and functional modules, including new and improved tools for quantitative data generation.
3. Data sets available through a data repository generated from single cells employing suitable reporters and probes. Those data sets have already been used to capture population distribution profiles and cell to cell variability and to measure in living cells signalling in real time. This also includes new and improved tools for quantitative data generation and data handling/interpretation.
4. High resolution dynamic models of functional modules that have been optimized following iterative system perturbations, possess predictive value and are suitable for integration into large-scale models of cellular systems. This also includes new tools and approaches for dynamic modelling.
5. Dynamic models of the larger cellular systems connecting metabolism, signaling, cell cycle and cell growth optimized by systematic iterative experimentation and system perturbations. This also includes new tools and approaches for dynamic modelling, especially concerning connectivity of modules operating at different time scales, definitions of granularity and system identification.
6. Computational models that allow simulations and predict the effects of perturbations through the different levels of biological organisation from population to functional module. This also includes new tools and software for modelling and simulation through the different scales of biological organization.
7. Detailed knowledge about higher-order properties of the different levels of functional organization under study with effects for solving biomedical research problems of wider impact.
8. A strongly advanced understanding of the dynamic properties of metabolism, signaling, growth and cell cycle control and in particular how those different process affect each other and integrate to generate cell behavior and phenotypes.
9. Tools for experimental planning, data collection and retrieval based on other EC-funded projects and existing infrastructure. This includes infrastructure for tool and software development and dissemination.
UNICELLSYS has played an important role in the development and penetrance in the Life Sciences of the emerging field of Systems Biology. This was achieved by the intense collaboration and staff exchange within the project, involvement in the organization of conferences such as the yearly International Conference of Systems Biology, the involvement in organizing international courses in Systems Biology, communication with different bio-industries, involvement in various international initiatives and interaction with the wider publi
Achievements of Work Package 1 (PI: Jens Nielsen, CHALMERS)
This work package makes use of available information in the public domain and data generated in WP2 to
• Generate genome-wide reconstructions of metabolism and the regulatory network of the cell.
• Define the scope of the dynamic modelling effort and the experimental strategies for collecting suitable datasets in WP3 through WP5.
• Implement an information gathering system for the purpose of updating networks/models and the experimental and modelling strategies.
The work package operates through the entire life time of the project to ensure that new information becoming available from research outside UNICELLSYS can be implemented into network models and dynamic models on a regular basis. For this purpose the project will establish an information gathering system based on text mining and it will collaborate with existing databases such as the Saccharomyces Genome Database SGD and BIOGRID.
Task 1.1: Reconstruction of large scale metabolic network.
The ground work for the majority of the tasks was laid with the construction of a highly reliable metabolic network for yeast, which has been a collaboration among many partners and other labs outside the consortium. This metabolic network was further updated by including many reactions in the lipid pathways and thereby, improving the reliability of the biomass equation. This updated model, iIN800, is available in SBML format and is demonstrated to have excellent predictive power.
The Manchester, Cambridge and Aberystwyth teams made a further update on the consensus model, merging it with iIN800 and expanding the connectivity (ie several new transporters). The paper describing this has been accepted for publication and the data (sbml and a searchable database) can be found online.
Task 1.2: Systematic reconstruction of regulatory networks.
A regulatory network that involved the signalling pathways mediated by four key kinases, Snf1,Pka1, Tor1 and Hog1 was also constructed by using information from public databases such as BioGrid and SGD. This reconstructed model is available in CellDesigner format and can be easily converted to other formats such as SBML, etc. This will be used as the scaffold to expand the signaling network to include other protein kinases and their targets. This was done in a jamboree in September 2009, where partner CHALMERS hosted members from other interested labs and worked towards this goal. Detailed strategies for modelling the dynamic phenomena, etc have also been discussed with the experimentalists. Deliverable 1.2.
Task 1.3: Reconstruction of a global regulatory network model.
An annotated reconstruction of the protein-protein interactions (PPI) around four key nutrient-sensing and metabolic regulatory signal transduction pathways (STP) operating in Saccharomyces cerevisiae was developed. The reconstructed STP network includes a full protein-protein interaction network including the key nodes Snf1, Tor1, Hog1 and Pka1. A number of proteins were identified having interactions with more than one of the protein kinases. The reconstructed interaction model serves as a platform for integrated systems biology studies of nutrient sensing and regulation in S. cerevisiae and this approach could serve as first step towards generation of an extensive annotated PPI interaction network of signal transduction and metabolic regulation in this yeast.
Task 1.4: Implementation of information gathering systems.
After internal discussions within the consortium it has been decided not to establish a specific database for UNICELLSYS as this will simply be redundant with already existing databases. The objective has therefore been changed to use a Wiki collecting key data from different experiments. The first set of experiment was already conducted within the YSBN consortium (large overlap with the UNICELLSYS consortium), and these data have been made available to all partners. This dataset involves the following data from growth in batch and chemostat cultures of two different strains, CEN.PK and YSBN (a prototrophic strain derived from the BY strain series):
• Gene expression analysis using Affymetrix Yeast 2.0 arrays.
• Genome wide expression analysis using tiling arrays.
• Proteomics using LC-MS.
• Metabolomics using a range of different analytical platforms and different extraction methods.
• Enzyme activity measurements for key glycolytic enzymes.
• RT-PCR for quantitative analysis of gene expression of a few selected genes.
• TRAC analysis for quantitative mRNA levels of key genes.
• Detailed physiological data in the form of measurements of carbon dioxide evolution rate and oxygen uptake rate and measurements of all key medium components over time.
Task 1.6: Definition of experimental and computational strategies for dynamic modelling.
After discussion with other partners involved in WP6 and WP7, a detailed strategy for dynamic modelling was planned. Two kinds of dynamic experiments are planned using the haploid yeast, YSBN6: (a) Aerobic to anaerobic shift in carbon and nitrogen limited chemostats and (b) carbon to nitrogen shift under aerobic and anaerobic conditions. The details of the protocols are given in Deliverable 1.6.
Achievements of Work Package 2 (PI: Steve Oliver, UCAM)
Task 2.1: Systematic quantitative phenotypic profiling
A detailed phenotypic comparison between yeast growing under a number of different defined nutrient limitations was made. A library of high-resolution growth responses was generated by microcultivation of the prototrophic YSBN strain. A wide range of growth conditions was analysed, including 15 different carbon sources (e.g. glucose, galactose, maltose, trehalose, mannitol, sorbitol, and ethanol), 18 nitrogen sources (e.g. glutamine, leucine, ammonium sulphate, allantoin, urea, proline) and 27 stress-inducing agents (e.g. paraquat, diamide, CdCl, NaCl, LiCl, DTT, hydroxyurea, rapamycin, caffeine). For each of the conditions, 10-15 different concentrations were analysed, thus providing a quantitative phenotypic library of 600 growth curves. This library of growth dynamics formed the basis for quantitative analyses and verifications of metabolic models with growth as the read-out (see Task 2.3).
A newly formed temperature-sensitive (TS) collection for essential genes was analysed in collaboration with Prof. C. Boone (Toronto). The collection consists of almost 800 TS strains representing 500 different genes. Many of these essential genes are part of the various signalling paths under analysis/modelling in the UNICELLSYS project. The details of temperature dependence provided an important resource for setting specific inhibition levels of these essential genes. The collection was also used in double-deletion screens for genetic interaction for genes of relevance to UNICELLSYS.
Task 2.2: Assessment of genetic interactions
A collaborative effort with the Papp, Lercher, and Boone laboratories generated quantitative measurements of the genetic interactions between ~185,000 metabolic gene pairs in Saccharomyces cerevisiae and superposed the data on a Stoichiometric Model of yeast metabolism. A machine-learning method was employed in an attempt to reconcile empirical interaction data with model predictions. This automated method suggested several modifications that, together, considerably improved both the recall and the maximum precision of negative interaction predictions. This automated procedure indicated that one of two alternate routes to NAD biosynthesis contained in the Stoichiometric Model is unlikely to occur in yeast. Several additional erroneous genetic interaction predictions between NAD pathway genes has led to a complete revision NAD biosynthesis in the Model. The validity of these revisions was confirmed by empirical experiments that involved feeding pathway intermediates to single-gene deletion mutants. This work has been published.
Task 2.3: Network reconstruction and gap-filling
Improvements were made to the consensus genome-scale Stoichiometric Model of the yeast metabolic network. The model has been expanded with improved representation of metabolite transport and lipid metabolism, involving 1102 unique metabolic reactions, 924 unique metabolites located in 15 cellular compartments, and with improved representation of lipid metabolism and other pathways. An iterative semi-automated procedure for model validation and refinement was developed, using computational tools within the framework of flux balance analysis (FBA) for model validation and optimisation. This can assist the revision process of hypothesis generation, evaluation and testing via model simulation, literature/bioinformatics evidence mining and biological experiments. Fully automated experiments were conducted by the Robot Scientist Adam to investigate the metabolism of yeast in utilising amino acids (AA) as the sole carbon/nitrogen (C/N) source, and to survey the phenotypes for isoenzyme deletants using growth curve analysis. The experimental data, together with other gene deletion study data from literature, were used to validate and refine the metabolic model. Furthermore, targeted experiments with single-gene deletants have been conducted to validate the model revision hypotheses generated from the constraint-based optimisation analyses. The improved Model was used to predict haploinsufficient and haploproficient phenotypes, and simulations of dynamic behaviour carried out.
Task 2.4: Gain-of-function/Overexpression effects
The objective of task 2.4 was to determine the robustness of networks to the overexpression of individual signal transduction components, in the presence and absence of various environmental perturbations. Initial experiments characterised robustness against overexpression in the HOG MAP kinase pathway, and this was followed by the determination of robustness patterns in the following UNICELLSYS pathways: Protein kinase A (PKA), Target of Rapamycin (TOR), Protein Kinase C (PKC), High Osmolarity Glycerol (HOG), Mating & Filamentous growth (STE/PHD), Snf3/Rgt2 and Snf1 pathways. Crosstalk between the pathways was analysed by quantifying the suppression of overexpression toxicity effects byknock-out mutations of key components in other pathways. Such suppression proved to be rare, but was found to occur between certain pathways and cell cycle regulators. Thus only a minor part of the toxicity is caused by crosstalk. A third phase of the study encompassed the transcriptional network, while a fourth phase included all kinases and phosphatases in yeast.
Task 2.5: Protein-protein interaction data
This task involves studies at both the bioinformatics and experimental levels.
2.5.1 Bioinformatics. Several high-throughput datasets for protein-protein interactions are available from many databases e.g. BioGRID, IntAct. However, this information is frequently not well structured and semantically defined, which makes them difficult to be used further in, for example, model construction and integrated data analysis. This task aimed to establish a comprehensive protein-protein interaction network that is well annotated and semantically represents protein complexes and functional reactions. The project represented a collaboration between 5 partner and 5 international laboratories (Kitano, Ideker, Vidal, Tyers, Boone) and concentrated on apoptosis. as an example process due to its relatively small scale and simplicity. Reactome schema were used to store interaction information and a data model designed to represent the network of protein-protein interactions, which were then curated by experts.
2.5.2 Experimental. A directed yeast-two-hybrid analysis was employed to produce specific sets of novel protein-protein interactions relevant to cell cycle regulation with the purpose of network reconstruction. This permitted an exploration whether protein-protein interactions predicted from computational models could be validated experimentally. Further support for the protein-protein interactions observed by the yeast-2-hybrid approach was obtained by independent validation using GST pull-down analyses.
Achievements of Work Package 3 (PI: Uwe Sauer, ETH)
The purpose of WP3 “High-throughput quantitative time course data” was to develop the necessary analytical methods and to generate proteomics (absolute protein levels, protein modification, protein interaction), transcriptomics (mRNA levels, translated mRNAs, DNA-protein interaction) and metabolomics data (metabolite levels, fluxes) of relevance for model structure and parameter estimation.
On the technological side, several cutting edge methods where developed by the expert analytical labs.
A mass spectrometric method was developed for quantification of absolute levels of metabolic and signaling proteins down to 100 copies per cell based on selective reaction monitoring (SRM). This method enabled, for the first time, comprehensive quantification of yeast metabolic and signaling proteomes with nearly 100% coverage. This SRM method was then extended to phosphoproteome analysis where quantification was achieved using synthetic AQUA peptides. Both methods were used extensively for time-course analyses in UniCellSys.
Protein complexes (Serrano):
A highly accurate method to analyze specific complexes was developed.
By comparison of about 10 work flows, two complementary and broad-scope mass spectrometric methods were developed for maximum complementarity in determining metabolite level changes in yeast. Since the quantitative LC-MS/MS method encompassed most of the metabolites relevant for computational modelling, it was used extensively for time-course data generation within UniCellSys.
On the biological side, the consortium decided to generate large and dynamic data sets for 3 so called standard experiments that would severe as a data basis for model generation. All 3 experiments were performed with the YSBN 6 strain and extensive data analysis with multiple omics methods was performed on samples send to the different partner labs. High level papers on these experiments are either published or are in preparation:
1. Chemostat shifts (led by the Nielsen Lab, Chalmers Gothenburg)
Gradual transition from glucose starvation to ammonia starvation with wild-type, SNF1 and TOR1 mutants (Zhang J et al. 2011. Mol Syst Biol. 7:545). We show that Snf1 regulates a much broader range of biological processes than TORC1 under both glucose- and ammonium-limited conditions. Surprisingly TORC1 also regulated fatty acid metabolism, likely through modulating the peroxisome and β-oxidation.
2. TOR signaling (led by the Sauer lab, ETH Zurich)
Two nitrogen nutrient shifts (up and down) plus a rapamycin pulse in batch culture. In particular computational analyses of metabolite and transcriptome changes identified several candidate metabolites as causes for TOR dependent and independent changes of transcription factor activity. Phosphoproteomics mapped the TOR dependent and independent phosphorylation network, including several novel metabolic targets of the TOR kinase that are currently being verified.
3. Osmotic stress in chemostats (led by Teusink lab, VU Amsterdam)
We identified 3 immediate and transient responses that are thus far undescribed: i) accumulation of 4 different amino acids (alanine, aspartate, glutamine, glutamate), ii) accumulation of the protective storage sugar trehalose, and iii) instantaneous switch from complete respiration to respirofermentation. Accumulation of the 4 amino acids was completely dependent on osmotic stress signaling. The accumulation of trehalose was largely dependent on this signaling pathway and is important for cellular survival upon osmotic stress.
Achievements of Work Package 4 (PI: Francesc Posas, UPF)
Building computational reconstructions and predictive dynamic models of biological processes involve different precise quantitative measurements. Work package 4 has addressed the objective of generating high-quality quantitative time-course data sets in a dedicated manner to capture different levels of regulatory events. The purpose was to generate data that either cannot be generated in high-throughput or where high-throughput analyses do not provide data of sufficient precision. During the study, the objective was reached and form part of the foreground of the project.
Quantitative data for predictive dynamic models have included kinetic parameters of the proteins involved, as well as changes over time of both absolute levels and activities of system components, protein modifications or subcellular localisation. Several time course data generated in the context of the previous EC-funded projects QUASI, CELLCOMPUT and AMPKIN have been transferred to UNICELLSYS. This includes mainly time course data on activation and deactivation of the MAPK pathways (pheromone response and HOG) as well as the Snf1 pathway.
More particularly, during the project, the following specific functional modules were analyzed:
1. Central carbon and energy metabolism. Mechanisms that control dynamic adaptive responses of metabolic flux, including compartmentalization, and how metabolic regulation is integrated into the control of growth and proliferation. Described in Deliverables 4.1 and 4.2.
2. MAPK signalling. Integration into the control of metabolism, signalling, growth and division. Described in Deliverables 4.3 and 4.4.
3. Glucose sensing and Snf1 kinase. Integration with growth and proliferation control. Described in Deliverables 4.5 and 4.6.
4. PKA signalling. Integration with metabolism, signalling as well as cell growth and proliferation. Described in Deliverables 4.7 and 4.8.
5. The cell-cycle machinery. Impact of signalling pathways on cell cycle progression. Described in Deliverable 4.11.
6. Generation of cell mass, defined as macromolecular biogenesis, including transcription, mRNA turnover, translation, protein turnover and ribosome biogenesis. Described in Deliverable 4.12.
WP4 has achieved all the Scientific & Technological target objectives established at begin of the activities. As main result, high resolution data sets of functional modules have been generated, initially from standard experiments and subsequently from systematic system perturbations. Finally, a set of new and improved tools for quantitative data generation has been generated (Deliverable 4.14).
Achievements of Work Package 5 (PI: Mattias Goksör, UGOT)
The purpose of WP5 “Single cell quantitative time course data” was to attain high-quality quantitative time-course data sets on both population level (FACS) and single cell level (microfluidics approach). By using different probes and reporter proteins, population profiles of regulatory events such as signaling, protein movements and interactions as well as gene transcription was captured and population characteristics, cell-to-cell variability versus noise and threshold properties were assessed.
The generated yeast strains expressing different fluorescent reporter proteins (Task 5.1.) have been used to meet deliverable 5.1 to 5.8. Generated FACS data from cells exposed to various levels of osmotic stress revealed bimodality existence of the HOG pathway and threshold dependency of pathway specific proteins (Task 5.2). Comparison between population and relocation microscope data, respectively, revealed heterogeneity within each FACS population and the underlying mechanisms was investigated (D5.1 and D5.5).
The FRAP experiments delineating rates of diffusion (Task 5.3) was focused on the Hog1 pathway and Mig1 repressor since these proteins display a relocation upon activation (D5.2 and D5.6). Hog1 displays a delayed nuclear localization in cells treated with an increased osmotic stress, while the localization and migration behavior of the Mig1 repressor has been shown to depend on the initial growth conditions. Also, none expected behavior of Mig1 has been identified which in collaboration with modelers will constitute a spin-off project and continuation of these data sets. The FRAP (and FLIP) experiments suggest that both for Hog1 as well as Mig1, two mechanisms of nuclear entry exists; one fast acting and one slower. The Hog1 nuclear delay could possibly be assigned the slower diffusion rate in high osmotic conditions and/or be explained by an anchor mechanism of Hog1 to the cell membrane in the corresponding condition. Novel methods for image analysis and subsequent quantification of reporter proteins have been established.
By the use of microfluidic devices and optical tweezers, single cell data of reporter proteins’ transduction in real time has been acquired (Task 5.4). Focus has been on the dynamic localization of Hog1, Mig1 and Msn2 (D5.3) both during static conditions and upon fast environmental changes (D5.7). Threshold levels of the added stress revealed dynamic signaling profiled on single cell level, hence the cell-to-cell variations could be captured and used for systems modeling.
The final task (Task 5.5) were met in the deliverables D5.4 and 5.8 where two methods were employed; BiFC and FLIM-FRET. The first method was used to detect direct interactions between the cyclin-dependent kinase inhibitor Sic1 and the Clb cyclins, and between the last and specific transcription factors involved in the regulation of the Clb synthesis. The combined method of FLIM-FRET monitors dynamic changes in protein-protein interactions by the use of mCFP and mYFP reporters. Upon interaction, the subcellular localization of Sic1 and Clb was identified and the affinity constant of this protein interaction could be calculated.
Achievements of Work Package 6 (PI: Edda Klipp, HUB)
Work package 6 was devoted to the development of computational models for different cellular pathways. These cellular pathways comprise metabolic pathways, signaling pathways, cell cycle controlling pathways as well as pathways for maintenance of cellular infrastructure.
To this end, the research in Work package 6 was guided by the definition of common principle for detailed models. This included model format, model annotation, assumptions, parameter estimation, and simulation information. Major principles entail that the model format must allow for reproducibility and exchange. We recommended using SBML and upload to databases (JWS online). All other information must explain the model sufficiently such that elements can be identified and simulations repeated.
Further, we defined a set of standards for quality control, entailing in-sample-fits, out-sample-fits, predictions for different input scenarios, predictions for perturbation of key molecules as well as predictions for genetic perturbations. These measures allowed us in the future to compare models and use them for integration, guiding the following modeling processes.
Work package 6 has created a series of models with explanatory power. For yeast metabolism, both small-scale models (mainly for glycolysis) and a large-scale model comprising 285 and 294 metabolites were created.
A series of signaling pathways including the MAPK pathways, the TOR pathway, the cAMP/PKA pathway, glucose sensing and Snf1 energy sensing, have been computationally described at different levels of granularity. Cell cycle progression has also been analyzed in different models, again also at different levels of granularity. Here, focus was on mitotic exit and on the regulation through signaling.
Based on the modeling approaches for distinct type of cellular processes, integrated models have been developed at interfaces of metabolism, signaling and/or cell cycle. These models have been fitted to previously measured data. They have shown to exhibit predictive value for hitherto not considered data and to guide further experiments.
Infrastructure models combining cell cycle progression with information on cellular growth, metabolic capacity, and volume as well as cell age have allowed to simulate the behavior of cell populations and to analyze the distribution of a set of properties under various conditions. These predictions could again be tested experimentally.
In conclusion, a set of tested and improved mathematical models of various cellular regulation pathways is available and can be used in further research.
Achievements of Work Package 7 (PI: Jörg Stelling, ETH)
The main aims of WP7 consisted of the development of methods and their application for the construction of large-scale predictive dynamic models, including approaches aiming at the refinement of models by increasing detail and at the connection of detailed models into larger systems. With respect to these objectives, main results were achieved in the following areas:
Standards for model development: Detailed guidelines for the development of standardized, and thus combinable, dynamic models in the form of ordinary differential equations (ODEs) were developed. They covered naming conventions, criteria for model correctness and completeness, and mandatory model annotation. In particular, the concept of a ‘global’ module for state variables relevant to all composite models allowed for the formal interfacing of partial (pathway or module) models developed by different partners of the consortium.
Network decomposition, modularization, and interfaces of mathematical frameworks: We developed general methods for dynamic network decomposition based on monotone systems theory, including a multi-stage algorithm for computing a decomposition of a model into modules. Technical modules are computed to be monotone with regard to tasks that require simplifying the description of a module's dynamics for model integration, including the minimization of interfaces between modules. Proof-of-concept studies were performed for smaller-scale examples in metabolism – signalling interactions (e.g. for glucose signalling) as well as for a large-scale signalling network map (see WP1). The latter was transformed into a representation for further analysis, for example, of signalling cross-talk. Importantly, with the monotone systems framework, model integration does not need to follow a strictly incremental approach because it focuses on qualitative features of dynamic interactions. Pair-wise integration of signalling models was achieved by this approach, but overall limitations in the availability of validated models for modules prevented further scaling in terms of large-scale model integration.
Model development, integration, and experimental validation: To enable model-experiment iterations under conditions of high uncertainty (e.g. on inputs for signalling pathways), we developed novel computational methods for topology identification as well as for optimal experimental design (partly in collaboration with nationally funded projects). Major advances were made in the coupling of metabolic dynamics and signalling, for example by unravelling novel signalling mechanisms in nutrient sensing (e.g. via pH as intracellular messenger) and gene expression control (e.g. identification of the transcription factor Msn2 as a potential rate-of-change sensor in nutrient stress signalling). Other examples of combined theory-experimental investigations concerned model-based approaches for the identification of TOR input signals mediating gene expression control using a novel computational approach of ‘prototypic’ dynamic models and Bayesian inference; it lead to the identification of a small set of candidate metabolites as inputs to the TOR pathway as well novel interactions in transcriptional control. However, due to unanticipated limitations in validated dynamic models and methodological challenges, strategic revisions implied a focussing on example subprojects for coupled processes across levels of cellular organization instead of full model integration (without feasible model validation).
Achievements of Work Package 8 (PI: Pedro Mendes, UNIMAN)
Work package 8 was devoted to the development of computational models that cross several scales of organization of growth and proliferation. These models aimed at predicting the effects of perturbations through the different levels of organization, as well as cell-to-cell variation. The Work package also aimed at constructing software to simulate such multiscale models.
The research in Work package 8 was based on models constructed in Work packages 6 and 7 and depended on the standards and principles approved thereof (such as the use of SBML). The main focus was on the operation of the entire cell cycle, at the level of biochemical networks, and how this drives proliferation at the level of the cell culture. Two models were simulated which focused on different aspects of growth: the first one included the detailed network of cyclin control and described cell size through a variable that reflects its mass; the second had a lower resolution cyclin network but described cell size through a variable that reflects the cell volume. Multiscale simulations of these models have shown that the distribution of cell ages is fairly constant through different phases of culture growth, and that even in the presence of large cell-to-cell variation of cell volumes the average cell volume of a culture is fairly stable. Equally the average cell mass of a culture is also stable.
Work package 8 also contained a component of software development to create a package for multiscale simulations spanning different levels of organization. This effort resulted in a software pacakge called ManyCell, and which has been released publicly as open source. ManyCell uses ordinary differential equations (ODEs) to model the internal biochemistry of cells (cell level); each cell is then modelled as an agent, where its state changes depend on discrete events that are triggered by the ODEs (culture or tissue level). The system also allows other entities, such as the extracellular medium or other extrinsic factors, to be modelled as other agents. The software allows researchers to observe results at the level of the culture and to zoom in to any cell to investigate its behaviour at the network level. ManyCell has an innovative architecture, where the management of agents is carried out by a transactional (relational) database management system, and where the multiscale aspect is implemented by a tabulation mechanism. The latter was shown to achieve a level of up to 100x speedup in the simulations mentioned above.
In conclusion, a software system capable of simulating ensembles of cells that can actively divide generating new cells has been made available as open source. Multiscale models and simulations of cell growth and division in a proliferating cell culture were also made available and can be used in further research.
Achievements of Work Package 9 (PI: Hans Westerhoff, VUA)
Task 9.1: Detailed analysis of realistic computational models to reveal novel and emerging topological of biological systems properties
Stealthy engineering is an example of a new concept that had emerged in the previous reporting period from the systems biology of S. cerevisiae developed. In this reporting period a manuscript on this topic was refined and resubmitted. Deliverable D9.11: one revised manuscript (abstract thereof).
Task 9.2: Detailed analysis of computational models to reveal novel and emerging kinetic/dynamic properties
Biology honours the tradition of making discoveries empirically: data are collected experimentally and used to generate hypotheses that are then tested again experimentally. Occasionally theory and hypothesis come first and biology engages in hypothesis driven experimental research. Systems Biology has furthered the combination of the two approach into a spiral turning between experiment and modelling.
Nonlinear, complex systems offer potential to a third type of discovery, i.e. discovery in silico. This is relevant for the case where experiments have defined all the important individual and interactive properties of all components, and where the network behaviour can be computed. If the network is sufficiently nonlinear, then the behaviour of the network cannot be extrapolated from the known individual behaviour of its components. In that case, sheer computation can lead to discoveries of properties of the real network. If these discoveries are general enough they do not require validation. Whether the discovered principle apply to the actual system under study depends on the correctness of the model and this applicability is served with experimental validation.
This type of systems biology requires the determination of components properties of the system under standardized conditions that are relevant for the in vivo situation. Because UNICELLSYS has achieved much (though not yet all) of this, and because this application of systems biology is hardly appreciated by the field of biology (even though it is completely accepted in physics (e.g. statistical thermodynamics) and chemistry (molecular modelling)), we have developed 2 illustrations of such in silico discovery. These were the discovery of the turb effect and the discovery of the relative importance of phosphatases and kinases in signal transduction. We developed both into workflows that used the live model repository JWS online. Because discovery is exciting to students, the workflows can be (and have meanwhile been) used in training courses on unicellular systems biology for biology and medical students.
Deliverable D9.7: discovery workflows using JWS (report)
Task 9.3: Detailed analysis of computational models to reveal novel and emerging control and regulation properties
UNICELLSYS has produced a framework for the systems biology of living cells, well grounded in experimental data sets obtained in yeast. Through dynamic modelling and hierarchical regulation analysis, insights were obtained in the multidimensional adaptation of unicellular organisms to altered environmental conditions, for instance carbon and nitrogen starvation. Unequal distributed patterns of gene expression and metabolic regulation have been observed. Control engineering is a discipline where control structures are designed that lead to optimal regulation behavior of networks. Control engineering distinguishes between proportional, integral and differential control loops. We have tried to identify which regulatory phenomena in the biology of yeast might correspond to the former two types of loops. Metabolic regulation was identified as a, nonlinear, generalization of proportional control and regulation through gene expression as a ditto generalization of integral control. We found that the perfect adaptation that integral control loops proposed in control engineering are able to provide, may be rare in biology as the require zero order decay kinetics of proteins. Deliverable D9.1: Report, to become manuscript submitted for publication.
The standardised conditions for enzyme activity assays that were developed by UNICELLSYS have been published (Van Eunen 2010). Now the dependency of the enzymes of both pathways on ATP, ADP, and AMP under thjese in vivo conditions have been analysed. For this the activity assays of some of the enzymes needed to be redesigned. The nucleotides needed to be excluded from the assay mixture where they initially served as co-factors for so-called coupling enzymes, but were affecting the overall reaction rate. (The assays are based on the monitoring (at 340 nm) of NADH / NAD+ / NADP+ / NADPH oxidation / reduction by the enzymatic reaction taking place. Doing this we observed that adenine nucleotides also affected many reactions for which they are neither substrates nor products, nor known allosteric regulators. The effecst occurred at relatively high, millimolar concentrations and involved Michaelis constants for the enzymes’ substrates and products. This made us postulate that around their physiological concentrations, adenine nucleotides and hence the free energy state of the cell, regulate many intracellular reactions pleiotropically. Deliverable D9.5: Submitted manuscript (the abstract thereof).
Task 9.5: Detailed analysis of the extent of coupling of metabolism, cell cycling, gene expression and signal transduction, and vice versa.
On the basis of the previous finding we inserted the newly found enzyme kinetic effects of the adenine nucleotides under in vivo conditions into one of the most established computational models of the yeast glycolytic pathway. We observed that these effect significantly affected the flux and the ATP/ADP ration(both by a factor of more than 2). Deliverable D9.9: Model with the effects inserted.
Task 9.6: Detailed analysis of how reduction in substrate concentration, environmental stress and hormones affect gene expression and cell cycle progression
Several pieces of relevant research have been outlined in previous work packages: the osmotic stress response at limited sugar concentrations, osmo stress response delay at high osmotic stress as well as Snf1 pathway behaviour at different sugar concentrations at single cell level. Those studies have provided a wealth of new mechanistic insight.
Task 9.7: Detailed analyses of pathway crosstalk
Various pieces of work have been reported with other work packages. Recently partner UGOT published a study that analysed in detail cross talk between the three yeast MAPK pathways. It became apparent that in this case, a response instigated by one pathway resulted in the activation of a second pathway and so on, until all three pathways were activated consecutively. This work demonstrated that pathway cross talk should consider not just the pathways themselves but also their targets and responses.
Task 9.8: Detailed analysis of the mechanisms of integration of different regulatory cues to control of growth, division and differentiation
A number of partners of the UNICELLSYS consortium were involved in the first consensus genome metabolic map of any organism, i.e. that of S. cerevisiae. During the UNICELLSYS project this result inspired the same and other groups to produce the analogous genome wide metabolic map for the human. The genome wide map of yeast could already be used to circumvent mutations in the metabolic map that reduced growth rate or growth yield, by growing on different substrates. In this reporting period we generalized, elaborated and publicized the corresponding approach towards personalized, diet-based medicine for the human. We herein highlight that human individuals differ in multiple mutations in their map, that these may be dealt with by medicines or by dietary changes, both in personalized strategies, enabled by the availability of the personal metabolic maps. Deliverable: powerpoint presentation D9.6.
Task 9.9: Detailed analysis of how cell cycle stage generate cell-to-cell variation in cell populations
Inspired by experimental findings in UNICELLSYS, a law was proposed and then proven mathematically that relates regulation to control. Deliverable D9.3: report.
Task 9.10: Estimation of the dependence of cell functions such as growth rate on the properties of heterogeneity and noise
Robustness coefficients were calculated for a cell cycle model and for concentrations of metabolites. Also in this case networking enhances robustness. Deliverable D9.4: report.
Task 9.12. Free research making use of models and experimentation to address
As most other catabolic pathways, the glycolytic pathway in yeast depends on Gibbs free energy input to get started. This takes the form of the ATP driven phosphorylations of glucose and fructose 1 phosphate. One of the consequences of this type of pathway structure is the liability for metabolic explosions, known as the turbo-explosion effect and none of our examples of possible discoveries through in silico systems biology (see above). A second consequence is the expectation that if for some reason the network comes into a situation where there is virtually no ATP, catabolism cannot get started and the cell dies in a necrotic process. We have examined this issue in the in silico models of yeast glycolysis such as the ones available through JWS online, and confirmed earlier insights that under normal conditions a zero ATP state does not lead to such necrosis, as the zero ATP state is not an attractor; the network tends to recover. We have identified reasons for this recovery and thereby found conditions where a zero ATP state is an attractor. Because of the conditionality of this type of predicted cell death, we refer to this as metabolic apoptosis. Deliverable D9.12: report
Task 9.13. Analysis of data of and models for mammalian cells and their parasites to examine whether some of the observed properties also apply there.
Achievements of Work Package 10 (PI: Chris Workman, DTU)
A UNICELLSYS specific implementation of the SysMO-DB (UNICELLSYS-DB) was made by adaptation of SysMO-DB open-source software. The URL http://unicellsys-db.cbs.dtu.dk was established and is linked from the main UNICELLSYS webpage. The UNICELLSYS-DB is based on SysMO-DB version 0.10.1. In total, 4 investigations, 5 studies and 2 assays have been registered for UNICELLSYS-DB. These were associated with 32, 27 and 5 data files respectively.
Time course experiments of growth in liquid media 132 viable protein kinase and phosphatase deletion backgrounds (plus 7 wild type strains) were conducted by Partner 3 for various stress conditions including methyl methanesulfonate (MMS), hydrogen peroxide (H2O2) and Rapamycin on the BioLector mini-bioreactor. These studies provided on-line growth estimates every 3 to 5 minutes in 48 or 96 well microtiter plates. The growth phenotypes for kinases such as HOG1 and SNF1 were of interest to this project and could be integrated with other similar efforts by Partner 1. These data were up-loaded to the UNICELLSYS-DB.
Overview of data sets that are available in UNICELLSYS-DB (as of March 2012) grouped roughly by signalling pathway.
Achievements of Work Package 11 (PI: Mats Jirstrand, FCC)
Generating, Improving, and Disseminating computational tools for systems biology
The main science and technology results for WP 11 are reported below according to their relation to the tasks of the work package.
Task 11.1 Design of infrastructure architecture in terms of modules, exchange formats, and tools.
An inventory among the project partners of currently used software and computational platforms was performed and examples of workflows involving two or more tools were listed. The result was given in terms of a report, which constitutes deliverable D11.1.
Task 11.2 Development and implementation of tools for defining models and model structures
The biochemical reaction network modelling tool PathwayLab has been extended with drag-and-drop libraries for component based biochemical reaction network modelling using the Systems Biology Graphical Notation (SBGN) standard, which constitutes deliverable D.11.2. We have also developed a format for textual specification of continuous time dynamic systems with discrete time measurements that follows standard mathematical notion. The format permits representation of both ordinary differential equations (ODEs) and stochastic differential equations (SDEs), which permit probabilistic descriptions of system and measurement noise including coloured noise models and covariance matrices. In addition, so called non-linear mixed effects (NLME) models can be represented. This is key to describing cell-cell variability and uncertainty in quantitative terms for dynamic system models. Cell-cell variability has recently gained an increasing interest and the tools and methods for parameter estimation in such models will be of large applicability to the systems biology community. This model/system description has been prototyped in Mathematica and constitutes deliverable D.11.3. An integrated framework, called the DynamicModel framework, has been created as a set of Mathematica packages which facilitate the specification and manipulation of the above described classes of models.
Task 11.3 Development and implementation of tools for pre-processing and management of measurement data for the purpose of parameter estimation
A data structure for input/output data has been developed. In this way input/output data that belong together are kept in one and the same place. The data structure can be used to store both time series data generated by a real experiment and to store simulation results. Furthermore, tools for basic pre-processing operations on measurement data have been developed such as outlier detection, filtering, and scaling. The input output/data structure and tools for pre-processing of measurement data constitute the deliverable D11.4.
Task 11.4 Development and implementation of algorithms and tools for parameter estimation
A top-level function, DynamicModelFit has been included in the DynamicModel framework. This function performs parameter estimation of parameters in models specified using the framework. The parameter estimation capabilities range from global optimization of trivial objective functions, such as the sum of squared errors between the simulated model and the observations to advanced gradient based optimization schemes of regularized objective functions based on norms of prediction errors. Advanced algorithms for parameter estimation based on nonlinear filtering and prediction error minimization (PEM) have been implemented as part of the DynamicModel framework. Here numerical schemes such as extended Kalman filters, unscented Kalman filters, and particle filters have been used. Furthermore, algorithms for computing the gradients of prediction errors based on these filters have been derived. This permits the application of highly efficient gradient based optimization schemes for maximum likelihood or maximum a posteriori based parameter estimation. The PEM algorithms above have been extended to also cover NLME models. We have derived numerical schemes for computations of likelihood function gradients for the NLME methodology. This framework, including the DynamicModelFit function, constitutes deliverable D.11.5.
Task 11.5 Development and implementation of tools for model validation
The DynamicModel framework contains functions for visualization of residuals between a model simulation and the corresponding observations, which is one of the most common methods of validating a model. A statistical test for whiteness of model residuals is also included in the framewok. This functionality constitutes deliverable D.11.6.
Task 11.6 Development of algorithms and tools for computational analysis
A package named SensitivityAnalysis has been developed in Mathematica, providing functions for deriving sensitivity equations of differential and algebraic equations with respect to a set of parameters. This functionality can be used with the model specifications obtained through the DynamicModel framework to derive and numerically solve the sensitivity equations for both the differential equations specifying the system model, and the algebraic measurement equations. Identifiability of a model structure (i.e. if one can guarantee that there are not two sets of parameter values that correspond to identical measured time-series) is an important property needed for any parameter estimation procedure to return a sensible result. We have implemented tools for both structural identifiability analysis and practical identifiability analysis. The structural identifiability analysis is based on probabilistic rank computations of large matrices using modular integer arithmetics, which is capable of handling very large dynamic system models. The practical identifiability analysis is based on so called profile likelihoods, which are computed at a specific point in parameter space around which the local properties of the likelihood can be visualized.
Task 11.7 Development and implementation of tools for visualization of simulation results
Click-select support in pathway diagrams for visualization of time-course, phase plane, and metabolic control analysis data has been developed. This supports a rapid “what-if” workflow to understand how biochemical entities co-evolve over time and metabolic processing capabilities of a specific metabolic pathway model. This functionality has been implemented in PathwayLab.
Task 11.8 Development of tools supporting the system identification workflow
The modelling and simulation software PathwayLab provides means to encapsulate complex reaction mechanisms into modules. This is performed by the possibility to represent an underlying reaction network by, e.g. a named box. The functionality also provides means to implement cross-talk between modules, as a species can occur in several modules, while still referring to the same mathematical variable. This functionality constitutes deliverable D.11.7. We have also initiated the development of a stand-alone tool for modelling and simulation of biochemical reaction networks – Simfony – focusing on high performance in combination with streamlined and highly intuitive workflows for different tasks making it an attractive system for biologists with little or no prior experience with modelling and simulation tools. In applied projects with other UNICELLSYS partners the developed and implemented tool chain, supporting the system identification workflow, has been applied and its usefulness has been validated. Most recently a model for Mig1 nuclear shuttling has been developed and used for quantifying cell-cell variability using simultaneously collected multiple single cell data.
The emerging field of Systems Biology is anticipated to have a major impact on the development of biosciences in the beginning of the 21st century. It is generally expected that the use of mathematical models, i.e. computational reconstructions of biological systems and processes, will result in a new level of understanding: the elucidation of the basic and presumably conserved “design” and “engineering” principles of bio-molecular systems. Thus Systems Biology will move biology from a phenomenological to a predictive science. The ability to predict the response of biological extrinsic or intrinsic perturbations should allow us to accurately predict the outcome of therapeutic interventions with individual patients or to regulate and optimise industrial bioprocesses more precisely than has been possible before. Therefore, the results of Systems Biology are expected to have major impact on both medicine and industrial production in the future. Developing the research field and ensuring exploitation of its results therefore is of major economic interest for the European Union.
There are two principal strands in Systems Biology:
- The top-down or data-driven approach, which focuses on building and analysing network models derived from genome-wide high-throughput data (e.g. expression data).
- The bottom-up or model-driven approach focuses on building and analysing detailed dynamic models of cellular pathways identified by classical molecular studies. These computational reconstructions capture biological processes over time and hence attempt to reproduce, and eventually predict, the quantitative dynamic behaviour of cellular modules.
A major challenge in Systems Biology where UNICELLSYS has made major contributions, concerns the construction of comprehensive molecular dynamic models that, eventually, will encompass all molecules and reactions in a cell. For this to become a reality, both top-down and bottom-up approaches have to be pursued in a coordinated manner. UNICELLSYS has moved the field towards this goal. In addition to generating new knowledge on the function of biological systems, UNICELLSYS has provided solutions to mathematical and computational problems arising from building large-scale dynamic models and to advance our capability to generate the necessary quantitative, time-resolved data.
Building precise, comprehensive and predictive dynamic models of specific biological processes and performing insightful analyses requires quantitative data (physicochemical properties of biomolecules and their activities, absolute numbers of specific biomolecules in the cell, reaction rates, and changes over time). The more accurate a model, the better it is able to predict the outcome of drug treatments or genetic alterations. Hence, predictive models are needed, for instance, to anticipate the effects of personalised patient treatment or to guide metabolic engineering. To produce such predictive, data-rich models requires close collaboration between theoreticians that build models and experimentalists that generate data. This tight integration of diverse, and often dispersed, capabilities has been a very strong feature of UNICELLSYS.
UNICELLSYS has major long-lasting impact in the following areas:
- Integrating dispersed capabilities and establishment of critical mass to enable a Systems Biology approach and for solving complex research problems, such as the connection of metabolism, signalling, growth and cell cycle
- Networking the necessary expertise in Systems Biology in Europe and beyond
- Generating research results with impacts on human health
- Generating research results with impacts on industrial biotechnology
- Developing the field of Systems Biology in Europe, for instance by participating in new initiatives, such as the ESFRI project ISBE
Integrating dispersed capabilities and assembling the critical mass to enable systems approaches
UNICELLSYS will integrated the following dispersed capabilities:
- Large-scale data collection for systems biology. UNICELLSYS involved European laboratories at the cutting edge of transcriptome, proteome, and metabolome/fluxome analyses. These laboratories worked together to generate coupled datasets (i.e. from single experiments). Of particular value are the datasets generated in UNICELLSYS that monitor processes over time. In addition, UNICELLSYS generated datasets on phenotypic profiling and genetic interactions. At present, datasets of this type can only be produced with yeast as a eukaryotic model organism but are of major importance for understanding the genetic interactions involved in complex human diseases.
- Specific data collection for modelling. UNICELLSYS involves leading molecular biological laboratories that generated high-quality quantitative time-course data on key components in different pathways. The UNICELLSYS Consortium significantly advanced abilities in single-cell analyses, allowing the assessment of cell-to-cell variations and the quantitative analysis of individual components in the cell (followed up in a dedicated Marie Curie ITN). This opens the path to the precise assessment of important system properties, such as noise in cell regulation and response thresholds.
- Expertise on different biological systems. UNICELLSYS brought together world experts in experimental and theoretical research of key and conserved biological processes, including cell cycle control, signalling, gene expression, RNA and protein turnover and metabolism. This made possible important biological discoveries that will be followed up in medically important or biotechnologically relevant systems.
- Diverse modelling and computational expertise. UNICELLSYS brought together many of Europe’s leading centres in computational Systems Biology, thus integrating expertise in network biology, mathematics, theoretical physics, systems theory, analysis and engineering, dynamic modelling, and computer sciences.
- Software development for systems biology. Systems biology requires software tools that visualise the results of simulations, enable advanced analyses and optimisation of mathematical models, and allow even experimentalists with limited insight into the mathematical and computational basis to perform simulations. UNICELLSYS members have previously provided major tools such as PathwayLab, CellDesigner, and Copasi and developed for instance ManyCell in UNICELLSYS. Moreover, implementation of UNICELLSYS software on the web was done via the JWS Online Cellular Systems Modelling resource.
Impact on human health:
The last thirty years have seen a constant decline in the productivity of pharmaceutical research and development. This has occurred despite accelerating investment in biomedical research on the part of both industry and governments. While there are many reasons for this trend, one is undoubtedly our limited understanding of how different pathways interact within living cells. The ability of systems biology to solve this problem is widely recognised, including by the US Food and Drug Administration. It can make a major contribution in all these areas:
- Drug target identification. It is especially anticipated that simulation of models will allow identification of targets via analyses of the structure of biomolecular networks as well as on the basis of quantitative effects on cellular processes. It is also anticipated that simulations using computational models will facilitate moving from the traditional “one disease-one target-one drug” towards treatment concepts where a combination of compounds affects more than target in a system to cause the desired effect. It is further expected that modelling and simulation will enable design of efficient drugs that cause fewer side effects.
- Recognition and avoidance of adverse properties. System approaches are expected to allow the design of drugs that show proper adsorption and pharmacokinetics characteristics while not adversely interacting with other components of the human ‘ecosystem’.
- Individualised therapy. Systems modelling and simulation will facilitate developing treatments that are specific for a given patient, based on their phenotypic profile and accurate prediction the appropriate drug composition.
UNICELLSYS will have a substantial impact in this area by developing mechanisms that will allow the quantitative study and prediction of system features such as response times and amplitudes, thresholds, and noise. These properties are underexploited in drug target identification.
In addition, UNICELLSYS will address the causes of cell-to-call variation, enabling assessment of their consequences for drug treatments in the future. Moreover, the variation between the cells in a population is also a simple model for the variation underlying drug effects on human individuals.
New knowledge on basic biological processes relevant to health and diseases
The history of biology shows that novel fields require simple model organisms. The UNICELLSYS strategy to employ the budding yeast (Saccharomyces cerevisiae) to accomplish “the proof of principle” with specific measurement and modelling procedures was very successful. The budding yeast has been a major driver of eukaryotic genetics and genomics for at least 40 years and Systems Biology is no exception.
UNICELLSYS has contributed to a quantitative understanding of the mechanisms with which cells integrate different internal signals (e.g. generated by metabolism) and external signals (as generated by stress or hormone) to take the fundamental decisions of growth, division or development. Those, by definition, have significant impact on very many human diseases as well as disease symptoms. Hereby the results of UNICELLSYS have impact the development of personalised nutrition, as well as impact the field of nutragenomics, where the focus is to obtain quantitative understanding of the interactions between nutrition and human health.
Standards for both data and models
UNICELLSYS set out detailed standards for experimental protocols, data and metadata, and model quality. Consortium partners played leading roles in the establishment of proteomic, metabolomic, and model software standards. The critical mass and scientific standing of UNICELLSYS will ensure that Europe has a major voice in the future standards development in the field of Systems Biology.
Combine, integrate and extend existing data sources and screen the different heterogeneous data resources
UNICELLSYS made use for computational reconstructions and mathematical modelling of existing data from a number of different sources, including the scientific literature:
Existing data resources pertinent to this effort include protein-protein interactions (IntAct http://www.ebi.ac.uk/intact MINT http://mint.bio.uniroma2.it/mint/ DIP http://dip.doe-mbi.ucla.edu/ BIND http://bond.unleashedinformatics.com etc.), protein-DNA interactions (GABI-DB, currently under development for the YSBN project), genetic and chemical-genetic interactions (the BioGRID (http://www.thebiogrid.org/) signalling pathways (KEGG http://www.genome.jp/kegg/ MIPS pathways http://mips.gsf.de/proj/yeast/pathways/ Science's Signal Transduction Knowledge Environment (STKE) http://stke.sciencemag.org/) and protein kinase-kinase substrate interactions (Mike Snyder lab). In addition, the Project collaborated with the Saccharomyces Genome Database (Stanford), which traditionally stores relevant yeast genome information.
Impact on industrial biotechnology
In order for the new knowledge generated by UINCELLSYS to be further exploited in research and for European industry, appropriate dissemination of results has to be ensured. UNICELLSYS enables computer-driven design of cell factories for production of fuels, chemicals, enzymes, food ingredients and pharmaceuticals. The Consortium played and plays an important role in the development of industrial biotechnology, which is forecasted to lead to the establishment of more than 200,000 new jobs in Europe. While yeast itself is a very important cell factory and is likely to be used for production of a wide range of products in the future, the concepts and methods developed within UNICELLSYS will also have substantial impact on the exploitation of other micro-organisms and cells as cell factories.
UNICELLSYS required a European (rather than a national or local) approach
Europe is presently competing reasonably well with the US and Japan in Systems Biology. This is due to both national investments but also EC-funding in FP6 and 7. Europe has traditionally been in a leading position with respect to model-driven system approaches, which, in fact, have their foundations at least fifty years ago. Model-driven system approaches so far suffer from the lack of suitable data and hence both data- and model-driven Systems Biology will eventually work more closely together to advance large-scale dynamic modelling. This is the approach where UNICELLSYS advanced the field and is expected to give the European research base a competitive advantage. Moreover, due to EC framework programmes the European research base has developed a sense to collaborate towards common goals, which is needed to advance Systems Biology, adding to a potential competitive advantage.
Account was taken of other national or international research activities
UNICELLSYS will make excellent use of previous investments in genomics, functional genomics, bioinformatics and systems biology thereby illustrating the European Commission’s strategy of advancing research fields across Framework Programmes. The EC supported in the 1990s budding yeast genome sequencing and functional genomics, forming the basis for the fields of genomics and functional genomics. FP6 invested in developing systems biology via a number of projects; UNICELLSYS uses expertise, capacities and infrastructure generated in those projects:
- QUASI (STREP 2004-2007) formalised integration of dynamic modelling and quantitative measurements, integration of different biological pathways into single models and addressed crosstalk between different MAPK pathways and cell cycle control (partners 1, 4, 5, 11, 13).
- AMPKIN (STREP, 2006-2009) addresses comparative dynamic modelling of a pathway in yeast and mammalian cells as well as integrating metabolism and signalling (partners 1, 15).
- RIBOSYS (STREP, 2006-2009) provides the basis for the systems biology approach to RNA metabolism and ribosome biogenesis in the Project (partner 14).
- DIAMONDS (STREP, 2005-2008) studies cell cycle networks in different organisms, including budding yeast, providing expertise to the Project (partner 3).
- 3D REPERTOIRE (IP, 2005-2009) studies yeast protein complexes in a structural genomics approach, forming a basis for protein interaction studies in the Project (partner 6).
- ENFIN and EMBRACE (NoEs, 2005-2010) provides database and data access infrastructure for systems biology that will be employed by the project (partners 3, 8, 11).
- BIOSIM (NoE, 2004-2009) is building infrastructure and expertise on quantitative models of normal and dysfunctional biological systems (partners 2, 3, 7).
- YSBN (CA, 2006-2008) provides infrastructure, competence and an organisational basis in yeast systems biology for the Project (partners 1, 2, 3, 4, 7, 8, 10, 11, 12, 13, 15, 16).
There are links to additional international activities such as the EUREKA project EUSysBio (partner 7), the project NucSys (also partner 7) supported by the EC as well as the international SysMO programme (Systems Biology of Micro-organisms; several partners are involved in different projects) instigated by the German research ministry.
UNICELLSYS mobilised national programmes for its research goals, including
- The two BBSRC-funded Systems Biology Centres in Manchester and Edinburgh, which both have strong yeast research components.
- The German HepatoSys project where partner 11 is involved in.
- The Dutch Vertical Genomics project with partner 7.
- The Swiss Systems X programme which supports partner 4.
UNICELLSYS partners are involved in a number of initiatives and activities that are partly and mainly inspired by the project. This includes for instance projects in SysMO (financed by national agencies jointly), various ERA-Net projects as well as the ESFRI project ISBE.
The research base and training and development of researchers
UNICELLSYS financed more than 200 person/years. Hence, the project used significant human resources, generating an opportunity to develop this resource. By training through research UNICELLSYS reached more than 100 young scientists and thereby has significant impact on the development of the European Systems Biology research base and the future progress of the European industry in both the biomedical and white biotechnology arenas. Several research career have been sparked by the project and young researchers employed by the project now have independent positions. One example if Matteo Barberis, who worked for three different partners during the project and is now employed as assistant professor with a forth partner.
Use and dissemination of foreground
UNICELLYS primarily addressed basic biological questions and generated a vast amount of novel results and new biological information. Because UNICELLSYS pursued a highly interdisciplinary approach, the results have effect on various disciplines and cross-over technologies, which are considered the fundaments of systems biology, including mathematical modelling, engineering and experimental technologies such as proteomics, imaging, metabolomics, and genomics to name the most relevant ones.
Hence, UNICELLSYS generated project results of potential economic value, including:
- software and computational tools for the entire systems biology community
- experimental design principles and specific experimental tools
- novel methods and technologies for single-cell analysis of proteins
- new or optimised instrumentation for quantitative data generation
- a mechanistic understanding of complex diseases and consequences for therapeutic action
- new knowledge potentially facilitating drug target identification and drug discovery
- uncover biological principles useful for improving health and human well-being
- knowledge that can be used in the design of novel foods containing bioactive ingredients
This said, none of the project results is of immediate commercial relevance.
UNICELLSYS has disseminated its project results in the first place via common scientific channels, i.e. a huge number of publications, many of these in high impact and absolute top journals, as well as via presentations at conferences and lectures. In addition, UNICELLSYS disseminated, or attempted to disseminate, results in the following ways.
Systems Biology Industrial Advisory Platform
This task has been abandoned and stopped, since industry has not indicated a thorough and long-term interest in establishing such a platform to communicate with participating academic institutions.
UNICELLSYS and projects it has integrated have now led to a substantial amount of training and illustration material of the subject. Yet, the problem remains that systems biology tends to be BIG SCIENCE. Molecules are networked to almost all other molecules in an organism and it is thereby hard for a research unit of limited size, even if it is an institute, to begin to work o systems biology in order to address a limited issue, such as the finding of a drug target vis-a-vis a disease. We have formulated a workflow for introducing the research unit to the topic.
Deliverable D9.2: Recipes for making systems biology work - advice for industry or large research units new to systems biology (report)
Participation in the organisation of international conferences and workshops.
The resources set aside for ICSB2008 were not used. Instead those resources have been used to support the STSM exchange programme in Task 12.3 to increase the interaction and training efforts. At ICSB2010 in Edinburgh, UNICELLSYS arranged and sponsored the parallel Session 2 "Functional Genomics and Biological Networks", chaired by Jean Beggs and Mike Tyers. Moreover, a number of UNICELLSYS group leaders, as well as students have been participating as invited speakers, lecturers and participants, respectively at international conferences, with a regular presence at the most important meetings, such as the ICSB2010 in Edinburgh, the ICSB2011 in Mannheim, the ICSB2012 in Toronto and (after the project) ICSB2013 in Copenhagen. UNICELLSYS funds have been used to support student and post-doctoral attendance at these international meetings to increase networking and presence. Numerous short talks and posters were presented by UNICELLSYS participants at these events.
Systems Biology as practised in UNICELLSYS engages in understanding the functional properties that emerge in the existing networks of living cells. The understanding is tested by predicting and then measuring the effects of perturbations of the existing networks. This bottom-up systems biology is one of the parent disciplines of Synthetic Biology, which engages in much stronger modulations of existing networks so as to create new networks with new properties that do not exist yet in living Nature. As in 2011, in the first week of October 2012 the (second) European iGEM meeting was hosted and co-organized by the Amsterdam UNICELLSYS team, and an opening lecture was given by the UNICELLSYS-Amsterdam PI. This lecture again highlighted the importance of unicellular systems biology, but now emphasized the importance of the availability of well understood microorganisms such as S. cerevisiae as ’chassis’ for synthetic biology. In particular the importance of stealthy synthetic biology on a robust chassis environment for which a computational systems biology model is available, was emphasized. Deliverable D9.10: lecture (ppt file), first week of October 2012, Amsterdam.
We have examined whether and how UNICELLSYS and JWS online can motivate biology and medical students, both in a local setting in Amsterdam and in an international setting in Ljubljana (a systems medicine course, organized together with CaSyM, ISBE and ELIXIR).
This has led to an evaluation of the UNICELLSYS database by the students as one deliverable and to an illustration of JWS online and the UNICELLSYS database as discovery tool.
Training and Education of Students/Pos-tdocs through STSMs.
The STSM programme has been continuously used by co-workers and students/postdoctoral fellows of the project to foster exchange and collaborations between theoretical and experimental groups within the consortium. During the entire project period, a total of 7 STSMs have been awarded and completed by students and/postdoctoral fellows. In addition, a total of 26 travel fellowships and other fellowships were granted to postdoctoral fellows and PhD students to attend international courses, conferences as well as meetings or workshops.
International Lecture Courses and Conferences.
After FEBS-SysBio2009, the FEBS-SysBio2011 course (http://cdl.univie.ac.at/sysbio2011/) was again organized and held in Innsbruck. Some 160 participants were attending 120 young scientists (PhDs & Postdocs) and 40 senior scientists and invited lecturers). The lecture course has been organized by several UNICELLSYS partner groups (Westerhoff, Sauer, Klipp, Kuchler). In addition, consortium partners also organized in Sweden ICYSB2009, 2011 and 2013 (International Course on Yeast Systems Biology), another FEBS Advanced Practical /Lecture Course, attracting 24 participants from within as well as outside the consortium.
Then, FEBS received an application for the course in 2013, but declined funding due to strongly reduced budgets in the Charity. Because SystemsX in Switzerland and the German BMBF, had offered support, the main organizers for 2013 (Sauer, Kummer, Klipp & Kuchler) decided to move the course into 2014 after the end of UNICELLSYS. This course, could therefore not be held in conjunction with the final UNICELLSYS meeting in Innsbruck.
However, the SysBio2014 lecture course will be held again in March 2014 in Innsbruck. Due to the financial cuts at FEBS, it was also not possible to obtain a sufficient budget to organize the grand final meeting of UNICELLSYS as a FEBS Special Meeting.
The final UNICELLSYS meeting was organized in March 2013 in Innsbruck. Most group leaders and partners were attending to present their past and future research activities in systems biology. As mentioned above, the FEBS Special Meeting was not possible as the financial crisis interfered with our efforts to obtain financial support, and the available UNICELLSYS budget was not sufficient to organize an international conference with several hundred people attending. Therefore, the available budget in UNICELLSYS for meetings and conferences was used to support the ICSB, other local courses and participation of students at international conferences during the last period.
Spreading Excellence through Public Perception and Outreach to Media.
This has remained as the most difficult task and has not achieved in full or on a regular basis. There are a number of reasons, including lag phase from submission to (online) publication (sometimes more than year), PU/RE dissemination issues and communication. While the number of publications from UNICELLSYS kept increasing constantly from all groups, we have not reached a level where we could generate a gapless pipeline filling a regular newsletter or consortium journal with big news of general interest.
Moreover, media journalists are very reluctant or hesitant to publish in daily or weekly papers news that are purely scientific since they are not, at least in their view, of major broad or public interest, on top of the difficulties explaining top science to the public.
Nonetheless, several group leaders have been active in providing easy understandable information through their personal web pages linked to the UNICELLSYS webpage as well as through local activities in schools and in public education institutions through lectures. Journalists interested to present some of the scientific data from UNICELLSYS, they were invited to contact the scientists directly via the homepage.
Because of the limited interest expressed by journalists in yeast systems biology, we also refrained from producing a costly brochure about the scientific challenges and outcome UNICELLSYS has been facing. A detailed list of all scientific publications that emerged from UNICELLSYS from each partner is of course available and included in the final report.
List of Websites:
University of Gothenburg
Dept of chemistry and molecular biology
PO Box 462