AB

STRA

CT B

OO

K

Page 1

WELCOME

On behalf of the SYSMED 2012 Programme Committee, it is my privilege and pleasure to welcome you to SYSMED 2012 here at Carton House Hotel from the 10th-13th September.

Systems Medicine has emerged as a new approach to personalised health care, utilising biological/medical data integrated with mathematical and computational modelling, to understand the underlying mechanisms of disease and to develop new strategies for individualised diagnosis, treatment and prevention. The key challenge in Systems Medicine is to bridge the gap between experimental data and medical knowledge and to evaluate both in terms of clinical utility and patients' benefit. This is an ambitious field that will challenge how we study complex diseases and is anticipated to have a far-reaching, positive impact on modern therapeutics and medical practice, resulting in a more comprehensive and systematic patient care.

We are pleased to present a dynamic conference programme including thematic sessions, posters, roundtable discussions and workshops. The sessions are led by a distinguished group of researchers and experts who are interested in applying systems level approaches to address medical problems. Of particular interest is a pre-conference workshop showcasing the ASSET Project, funded under the FP-7 Health Programme, which uses systems biology approaches to improve therapies for childhood cancers.

I would like to take this opportunity to thank our generous sponsors and supporters, listed overleaf, without whom this conference could not take place.

Thank you also to the programme committee, reviewers, speakers, session chairs, panel participants, Carton House Hotel, industry supporters, delegates and Systems Biology Ireland staff.

I look forward to meeting you over the coming days and on behalf of my colleagues on the local organising and programme committees, I extend a warm welcome to you all.

Professor Walter Kolch SYSMED 2012 Conference Chair

Page 2

CONTENTS Programme/Organising Committee and Sponsor Details

3

Pre-conference Workshop Programme (EU FP7 ASSET Project)

4

SYSMED 2012 Conference Programme

5

Abstracts for Invited Speakers and Oral Presentations

10

Session I: Systems Medicine - Hype or Hope?

10

Session II: Frontiers in Medicine

13

Session III: Potential for Therapeutics

15

Session IV: Drug Discovery

19

Session VI: Omics to Systems Medicine

23

Session VII: Computational Modelling

28

Abstract for Closing Keynote

33

Abstracts for Poster Presentations (Session V)

34

* MEDIA QUERIES: Media queries relating to all abstracts/presentations should be addressed to sbiadmin@ucd.ie. The authors consent must be obtained for any media coverage.

Page 3

PROGRAMME COMMITTEE

Walter Kolch, Systems Biology Ireland and Conway Institute, University College Dublin, Ireland

Rudi Balling, Luxembourg Centre for Systems Biomedicine (LCSB), University of Luxembourg

Claus Bendtsen, Computational Biology, AstraZeneca, UK

Hans Westerhoff, Manchester Centre for Integrative Systems Biology (MCISB), University of Manchester, UK/VU University Amsterdam, Netherlands/IT Future of Medicine

Olaf Wolkenhauer, Systems Biology and Bioinformatics, University of Rostock, Germany

Ruedi Aebersold, ETH Zürich Institute of Biotechnology/Institute for Systems Biology, Seattle/SystemsX.ch

Pierre de Meyts, De Meyts R&D Consulting SPRLU, Belgium

Boris Kholodenko, Systems Biology Ireland, University College Dublin, Ireland

ORGANISING COMMITTEE

Eadaoin Mc Kiernan, Systems Biology Ireland, University College Dublin, Ireland

Ian Barwick, Conway Institute, University College Dublin, Ireland

Philip Smyth, Systems Biology Ireland, University College Dublin, Ireland

Yvonne Smith, Systems Biology Ireland, University College Dublin, Ireland

Aimee Carmody, Systems Biology Ireland, University College Dublin, Ireland

Brendan McCann, Systems Biology Ireland, University College Dublin, Ireland

Susan Yeates, Systems Biology Ireland, University College Dublin, Ireland

SPONSORS

Page 4

Sunday 9th September Pre-conference Workshop - EUFP7 ASSET** Programme Carton Suite III, Conference Centre 11:30 - 13:15 Registration (including light lunch) 12:30 - 13:15 Lunch in “The Linden Tree Restaurant” 13:15 - 13:30 Workshop Opening and Welcome

Walter Kolch, Systems Biology Ireland and Conway Institute, University College Dublin, Ireland

13.30 - 15.30 Title: Cell Cycle Perturbations in Embryonal Tumours (Part I) 13:30 - 14:00 Neuroblastoma and Cell Cycle Overview (TBC)

Frank Westermann, Department of Tumor Genetics, Cancer Research Center- DKFZ, Heidelberg, Germany

14:00 - 14:15 N-myc Dependent Tumour Models Johannes H. Schulte, Pediatric Oncology Research Lab, University Hospital of Essen, Germany

14:15 - 14:45 Copy number alterations in neuroblastoma preferentially target MYCN down-stream genes: implications for modeling and study of MYCN regulated signaling pathways Frank Speleman, Center Medical Genetics, Ghent University, Belgium

14:45 - 15:00 Application of Omics Technologies to Elucidate the MYCN Transcriptional Network in Neuroblastoma

David Duffy, Systems Biology Ireland, University College Dublin, Ireland 15:00 - 15:30 Medulloblastoma Overview

Alexandre Arcaro, Paediatric Oncology, Department of Clinical Research, University of Bern, Germany

15:30 - 16:00 Coffee break 16.00 - 18.00 Title: Cell Cycle Perturbations in Embryonal Tumours (Part II) 16:00 - 16:30 Ewing Sarcoma Overview

Heinrich Kovar, Children’s Cancer Research Institute, Vienna, Austria 16:30 - 16:45 Gene Networks in EWS Loredana Martignetti, Institut Curie, Paris, France 16:45 - 17:00 Data Integration

Ramneek Gupta, University of Copenhagen, Denmark 17:00 - 17:15 Modelling approaches Florian Lamprecht, Cancer Research Center - DKFZ, Heidelberg, Germany 17:15 - 18:00 Annual Report Preparatory Meeting (ASSET Partners only) 19.00 BBQ reception in the Coach House 21.00 Traditional Irish Music in the Coach House

** ASSET: Analysing and Striking the Sensitivities of Embryonal Tumours (http://www.ucd.ie/sbi/asset)

Page 5

Monday 10th September SYSMED 2012 Conference (inclusive of Sponsor Exhibition in Carton Suite I) Carton Suite, Conference Centre 8.30 - 9.30 Registration 9.30 - 10.00 Conference Opening and Welcome Desmond Fitzgerald, University College Dublin, Ireland

Christina Kyriakopoulou, Health Directorate, DG Research & Innovation, European Commission, Belgium

10.00 - 13.10 Session I: Systems Medicine - Hype or Hope? Chair: Desmond Fitzgerald, University College Dublin 10.00 - 10.40 Systems Medicine: How to Capitalise on Molecular Medicine

Hans Westerhoff, Manchester Centre for Integrative Systems Biology (MCISB), University of Manchester, UK/VU University Amsterdam, Netherlands/IT Future of Medicine

10.40 - 11.10 Coffee break 11.10 - 11.50 Systems Pathology: How much more can tissue tell us?

David Harrison, St Andrew’s University, UK 11.50 - 12.30 The way to Systems Medicine Approaches in neuroblastoma

Angelika Eggert, Department of Pediatric Hematology/Oncology, Pulmonology and Cardiology, University Hospital of Essen, Germany

12.30 - 13.10 Systems Biology and Translational Medicine- Maximising the Synergy Seamas Donnelly, Education and Research Centre, St. Vincent's University Hospital and School of Medicine, University College Dublin, Ireland

13:10 - 14:00 Lunch in The Linden Tree Restaurant

14.00 - 16.00 Session II: Frontiers in Medicine Chair: Pierre De Meyts, De Meyts R&D Consulting SPRLU, Belgium 14:00 - 14:40 Interactome Networks and Human Disease

Marc Vidal, Centre for Cancer Systems Biology, Department of Genetics, Dana-Farber Cancer Institute/Harvard Medical School, Boston, USA

14.40 - 15.20 Cancer Response to Targeted Agents: Dynamics and Variability from Single Cell Data Vito Quaranta, Integrative Cancer Biology Center, Vanderbilt University Medical School, Nashville, USA

15.20 - 15.40 A Brief View of Systems Medicine and Genetics Jie (Bangzhe) Zeng, Institute of Systems Biological Engineering, China

15.40 - 16.00 Pathway-GPS and SIGORA: Identifying relevant pathways based on the over-representation of their gene-pair signatures David Lynn, Teagasc, Ireland

16.00 - 16.30 Coffee break 16.30 - 20.00 Panel Discussion, Exhibition and Networking Session 16:30 - 16:50 Introduction to panel discussion

Adriano Henney, German Virtual Liver Network, University of Heidelberg, Germany (Chair)

Page 6

16.50 - 18.00 Panel Discussion - "Systems Biology and the Future of Medicine?" William Finlay, Global Biologic Therapeutics, Pfizer Rudi Balling, Luxembourg Centre for Systems Biomedicine, University of Luxembourg Barry Heavey, Life Sciences, IDA Ireland Axel Polack, TVM Capital Life Science Walter Kolch, Director, Systems Biology Ireland and Conway Institute, University College Dublin, Ireland Lukas Huber, Cell Biology Division, University of Innsbruck and Oncotyrol, Austria Seamas Donnelly, Education and Research Centre, St. Vincent's University Hospital and School of Medicine, University College Dublin, Ireland Hans Westerhoff, Manchester Centre for Integrative Systems Biology (MCISB), University of Manchester, UK/VU University Amsterdam, Netherlands/IT Future of Medicine

18.00 - 20.00 Research and Technology Discovery Path in The Morrison Room - Exhibition and Networking event including wine and canapé reception

Exhibitors: Systems Biology Ireland (UCD & NUIG), Centre for Systems Medicine (RCSI), REMEDI (stem cell research, NUIG), Digital Enterprise Research Institute (NUIG), Complex and Adaptive Systems Laboratory (UCD), Conway Institute (UCD), Marine Biodiscovery Research Programme (NUIG), Shannon Applied Biotechnology Centre (LIT and ITT) and Alimentary Glycoscience Research Cluster (NUIG)

Tuesday 11th September SYSMED 2012 Conference (inclusive of Poster Session in Carton Suite I) Carton Suite, Conference Centre 9.00 - 12.50 Session III: Potential for Therapeutics Chair: Mark Ferguson, Science Foundation Ireland 9.00 - 9.40 Systems Approaches to Parkinson’s Disease

Rudi Balling, Luxembourg Centre for Systems Biomedicine (LCSB), University of Luxembourg

9.40 - 10.20 Systems Medicine Approaches to improve the Therapeutic Potential of Mesenchymal Stromal Cells in Critical Limb Ischemia

Timothy O’Brien, Regenerative Medicine Institute, National University of Ireland Galway, Ireland

10.20 - 11.00 Biomarker Discovery and Validation for Prostate Cancer patient stratification into Appropriate treatment options: Addressing the Clinical Dilemma

William Watson, School of Medicine and Medical Science, University College Dublin, Ireland

11.00 - 11.30 Coffee break 11.30 - 12.10 Neuropilin as a therapeutic target in Cancer

Ian Zachary, UCL Centre for Cardiovascular Biology and Medicine and Consultant to Ark Therapeutics Ltd, UK

12.10 - 12.30 Unravelling the path to Mesenchymal Stem Cell migration: A co-ordinated effort of beta gamma subunit, PI3K, Rac and Rho signalling Caroline Ryan, Regenerative Medicine Institute, National University of Ireland Galway, Ireland

Page 7

12.30 - 12.50 Differential localization of A-Raf kinase regulates MST2-mediated apoptosis and differentiation

Jens Rauch, Systems Biology Ireland, University College Dublin, Ireland 12.50 - 13.50 Lunch in The Linden Tree Restaurant 13:50 - 17.00 Session IV: Drug Discovery

Chair: Jochen Prehn, Department of Physiology and Medical Physics and Centre for Systems Medicine, Royal College of Surgeons in Ireland

13.50 - 14.30 The Cellular uptake of Pharmaceutical Drugs: A problem not of Biophysics but of Systems Biology

Douglas Kell, Biotechnology and Biological Sciences Research Council, UK and School of Chemistry and the Manchester Institute of Biotechnology, The University of Manchester, UK

14.30 - 15.10 New Drug Screening Strategies for Multiple Myeloma Treatment Lukas Huber, Cell Biology Division, University of Innsbruck and Oncotyrol, Austria

15.10 - 15.30 Uncoupling oncogenic receptor kinases from their intracellular nanocomputers Stephan Feller, University of Oxford, UK

15.30 - 16.00 Coffee break 16.00 - 16.40 Cell Signaling by receptor tyrosine kinases; From basic principles to cancer therapy Joseph Schlessinger, Yale School of Medicine, Connecticut, USA 16.40 - 17.00 Discovering new lead compound candidates and combination therapy model using

traditional Oriental Medicine Jongwook Jeon, Korea Advanced Institute of Science & Technology

17.00 - 18.00 Session V: Poster session in Carton Suite I 19.30 - 20.00 Pre-Conference Dinner Drinks Reception in Carton House Lobby and The Kitchen Bar 20.00 Conference Dinner in Carton Suite 21.30 Traditional Irish Dancing Showcase in Carton Suite 22.30 Traditional Irish Music in The Kitchen Bar

Wednesday 12th September SYSMED 2012 Conference (inclusive of Social Outing) Carton Suite, Conference Centre 9.00 - 12.50 Session VI: Omics to Systems Medicine

Chair: Walter Kolch, Systems Biology Ireland and Conway Institute, University College Dublin, Ireland

9.00 - 9.40 From proteomics and mathematical models to new therapies for colorectal cancer Walter Kolch, Systems Biology Ireland and Conway Institute, University College Dublin, Ireland

9.40 - 10.20 Apoptosis modelling predicts therapy responses in Cancer patients: Towards Personalised Oncology Approaches

Jochen Prehn, Department of Physiology and Medical Physics and Centre for Systems Medicine, Royal College of Surgeons in Ireland

Page 8

10.20 - 10.40 Systems Medicine of Neurogenic Hypertension James Schwaber, Department of Pathology, Anatomy & Cell Biology, Thomas Jefferson University, USA

10.40 - 11.10 Coffee break 11.10 - 11.50 Multi-Levelness as the Final Frontier of Systems Medicine Olaf Wolkenhauer, Systems Biology and Bioinformatics, University of Rostock, Germany 11.50 - 12.30 SWATH-MS provides a deep, reproducible and permanent quantitative proteomic

reference for biomarker studies Yansheng Liu, Institute of Molecular Systems Biology, ETH Zürich, Switzerland 12.30 - 12.50 Identification of molecular mechanisms underlying MCF-7 cell differentiation using

proteomics and systems biology approaches Natalia Volinsky, Systems Biology Ireland, University College Dublin, Ireland

12.50 - 13.50 Lunch in The Linden Tree Restaurant 15:00 - 21.00 Social Outing: Dublin City Walking Tour and Visit to Guinness Storehouse 15.00 Buses depart for Dublin City Centre 16.00 Dublin City Walking Tour (to include Trinity College Dublin, Book of Kells, Dublin Castle,

Christchurch) 18.30 Bus transport to Guinness Storehouse 19.00 Guinness Storehouse Tour and Buffet Reception 20.30 Buses depart for Carton House

Thursday 13th September SYSMED 2012 Conference (inclusive of Closing Keynote) Carton Suite, Conference Centre 9.00 - 12.00 Session VII: Computational Modelling

Chair: Boris Kholodenko, Systems Biology Ireland, University College Dublin, Ireland 9.00 - 9.40 Modelling of Cancer Kinome Networks

Rune Linding, Department of Systems Biology, Technical University of Denmark 9.40 - 10.00 Agent-based modelling approach of immune defense against opportunistic human

pathogenic fungi Christian Tokarski, Friedrich Schiller University Jena, Germany

10.00 - 10.20 Gene regulatory network controllability to identify cellular reprogramming driver genes Antonio del sol, Luxembourg Centre for Systems Biomedicine, University of Luxembourg

10.20 - 10.40 Coffee break 10.40 - 11.20 Signalling Ballet in Four Dimensions Boris Kholodenko, Systems Biology Ireland, University College Dublin, Ireland 11.20 - 12.00 The future of Medicine (And Health) Hans Lehrach, Max Planck Institute for Molecular Genetics, Berlin, Germany 12.00 - 13.00 Closing Keynote

Systems Pharmacology and Systems Medicine

Page 9

Garret FitzGerald, University of Pennsylvania and Institute for Translational Medicine & Therapeutics, Philadelphia, USA

13.00 - 13.15 Presentation of Prizes and Awards Walter Kolch, Systems Biology Ireland and Conway Institute, University College Dublin, Ireland

13.15 - 14.00 Lunch in The Linden Tree Restaurant

Page 10

ABSTRACTS FOR INVITED SPEAKERS AND ORAL PRESENTATIONS

CATEGORISED BY SESSION TOPIC

Monday 10th September

10.00 - 13.10

Session I: Systems Medicine - Hype or Hope?

Des Fitzgerald, University College Dublin

Carton Suite

Systems Medicine: How to Capitalise on Molecular Medicine Hans Westerhoff Manchester Centre for Integrative Systems Biology (MCISB), University of Manchester, UK/VU University Amsterdam, Netherlands/IT Future of Medicine For 1 k€ one can have one’s genome sequenced. As compared to one’s lifetime medical expenses this is affordable, as is the 500 G€ for all human genomes in Europe alone. Transcriptomics, proteomics and metabolomics are equally feasible, for blood, urine, fecal, tumor and certain tissue samples. The finding of diverse RNAs in serum further shows that one may readily obtain images of the state of the human body, healthy, sickly or diseased, and of the underlying causality of that state. The diversity of the images will confirm that individual humans are very different both in terms of their state on the axis between health and disease, and in terms of the causalities. This will suggest that the best therapy for disease as well as the best advice for a lifestyle that enables a desired performance (such as running marathons comfortably or memorizing Hamlet’s text), differs between individuals. It will be difficult to continue not to make use of these new technologies when faced with individuals requesting to be cured of a debilitating disease and with sufficient motivation to get their genome sequenced. Medicine is bound to yield to this upcoming request for personalization. Already, tumors are being typed on the basis of EGF receptor mRNA and patients are treated differently depending on the result; the same was already based on the histology of lymph nodes. In this sense, personalized medicine, is already practiced. A problem is that the success of this type of personalization is limited. Although the procedure often has success in terms of a retardation or alleviation of the disease, it may to provide few definitive cures. Systems biology has revealed why this is: health and disease emerge from the networking of the molecules of the body rather than from just an individual ‘elixir’ molecule. Medicine has a long tradition of dealing with the complex system that the human body is. This does not make existing medicine ‘systems medicine’ however. Much like systems biology, systems medicine acknowledges that much of health and disease does not exist in individual genetic and environmental factors and not even in the sum thereof, but mostly emerges in the networking of multiple of those factors. Consequently, it expects a single-factor-based therapy to have limited success, and diseases to be multifactorial. In this presentation we shall argue that systems medicine is neither hype nor hope; it is the logical next step in the endeavor to make best use of the molecular life sciences in medicine. Systems medicine is happening whether one likes it or not. The issue worth debating is how to prevent the development of a multitude of systems medicines, disconnected from each other in the arguably best traditions of the life sciences, with every subdiscipline speaking its own language. Will it be possible to mimic the physicists in their quest for the humble Higg’s boson in coordinating much of the activities towards systems medicine? The reward would be many times more significant: (i)

Page 11

understanding for the first time in human history something as complex as the human body, (ii) being able to calculate the best possible cures for disease and management of health with (iii) possibly- we do not wish to (over)promise- a significant improvement of health, reduction in health-care cost and stimulation of a new healthcare ICT economy. The 1.3 G€ ITFOM flagship program on the funding agenda of the EU takes the initiative to integrate the plethora of biomedical information, by using dynamic replica of the human body that can then be individualized to enable individualized medicine. I will discuss examples of our own, tiny contributions to this worldwide initiative: (i) a model and some of its experimental validation that addresses the toxicity of drugs, and (ii) new ways in which the success of stem-cell engineering may be assessed in the systems-biology sense. Systems medicine is neither hype nor hope: it is a challenge, but a rewarding one.

Systems Pathology: how much more can tissue tell us? David Harrison University of St Andrews, Scotland Classical pathology has in some ways tried to practice systems medicine, in as much the diagnosis has been a model, albeit empirical, seeking to derive dynamical meaning from static, secondary information presented in the artefact that is a tissue section. Despite this, or perhaps because of it, pathology to some extent has been a bystander for much of the advent of –omics technologies. More recently it has become apparent that the integration of pathology with other streams of data can significantly add value and also to challenge and help form strategies which seek to draw together disparate strands to lead to a diagnosis, inform prognosis and increasingly allow prediction of likely response to therapy. There are some key principles learned from pathology that may help facilitate discussion on wider aspects of the application of systems biology approaches in medicine. Firstly, datasets derived from patients are often small or incomplete, and may fail completely to capture some areas. This prompts the need for standardisation, quality assurance and consistency but at the heart of it there will be some situations where it is simply not possible to derive all the information desired. The addition of high throughput technologies such as transcriptomics certainly produces data but the sifting of this data, to decide what is needed and what is not, provides additional challenges. The use of human tissue comes with its own challenges and it is often the case that clinical material will not meet the highest standards, irrespective of how hard the investigator tries. This prompts the community to develop strategies that are pragmatic and flexible, to maximise the information derived from clinical studies. Secondly, pathology has emphasised the heterogeneity of disease processes. So for example, some models have extrapolated from a cell-based biochemical signalling pathway to a cancer tissue without perhaps accounting in full for the heterotypic nature of cancer (composed of cancerous and non-cancerous elements) and its heterogeneity (variation in genetic aberration in different parts of the tumour). Pathology has itself not adequately resolved how heterogeneity affects outcome (prognosis or prediction) despite years of empiricism, in part due to problems of lack of standardisation and consistency. This is an exciting challenge for systems biology on a multiscale level. Thirdly, pathology for much of its existence as a discipline has relied upon a two dimensional artefact, stained with antiquated histochemical techniques and viewed by transmission light microscopy to deduce behaviour, both past and future in a dynamical, three dimensional way. The addition of monoclonal antibodies and latterly RNA and DNA have allowed in depth analyses of some components, but bizarrely pathology for all of this still it is the best working model available of many diseases. It is

Page 12

apparent that the artefact of a tissue section is in fact a distillation of complex phenotype; a surrogate of DNA, RNA, proteome, methylome, metabolome all rolled into one without any clear guidebook of how to balance different features. The application of quantitative, image analysis techniques, for example morphometry and quantitative immunofluorescence, encourages the pathologist to be able to provide more raw data to the modeller aware of the spatial and presumably temporal variation that occurs. Fourthly, pathology is essentially visual, confirming that meaning inferred may be more rich and complex than simply multiple, measured datasets. Thus, the future of systems pathology may be to visualise processes, building in methods that allow filling-in where data is incomplete and prompting focus of research efforts where gaps appear in the picture.

The way to Systems Medicine Approaches in neuroblastoma Angelika Eggert Department of Pediatric Hematology/Oncology, Pulmonology and Cardiology, University Hospital of Essen, Germany Cancer chemotherapy is in evolution from non-specific cytotoxic drugs to more specific targeted agents and immunotherapy approaches. Targeted agents are directed at unique molecular features of cancer cells, and aim towards greater effectiveness with less general toxicity. The development of high-throughput technologies able to comprehensively assess DNA, RNA and proteins in patient tumors has fueled efforts to tailor medical care for patients with many types of cancers, including neuroblastoma. This lecture will provide an overview of the state-of-the-art in personalized medicine approaches for neuroblastoma patients. Molecular profiling of neuroblastomas has indeed provided us with a wealth of data for optimized risk stratification and potential targets for novel therapies. However, the explosion of knowledge presents both an opportunity for understanding cancer treatments and a challenge for organizing the information. The masses of data generated by high-throughput technologies are challenging to manage and convert to the knowledge required to improve patient outcomes. A cross-disciplinary systems biology effort will be necessary to convert the information into useful biomarkers that can classify patient tumors by prognosis and identify the drivers of tumor behavior that are optimal targets for therapy. The major tasks in the next few years will be to harmonize the existing biological datasets and functional experimental data, validate results in sophisticated cell culture and mouse models, and, finally, define algorithms for molecular target prioritization to support therapy choices guided by biology. Validated molecular tests assessing neuroblastoma tumor tissue will then successfully drive therapeutic decision-making.

Systems Biology and Translational Medicine- Maximising the Synergy Seamas Donnelly Education and Research Centre, St. Vincent's University Hospital and School of Medicine, University College Dublin, Ireland Systems Medicine is the application of systems biology to clinical research ultimately to the benefit of patients and their families. The genomic/array era has enhanced our understanding of specific diseases but has disappointed in the delivery of a relatively low number of novel therapies. The scientifically “low lying fruit” of the pathogeneisis monogenetic diseases has been defined but the majority of chronic diseases are complex and multifactorial. While avoiding overstating the case, Systems Medicine offers an opportunity to address the diagnosis, treatment and management of selected chronic diseases.

Page 13

A key stakeholder in this Systems Medicine consortium is the clinical scientist. They will identify clinical unmet needs and contribute to formulating the right questions to be addressed by Systems Medicine. This in turn will drive prioritisation of selected hypotheses which are then taken through in silica, cellular and in vivo studies. In this talk a clinical case study will be presented of a disease which has no specific therapy and which is invariably fatal. Idiopathic Pulmonary Fibrosis (IPF) represents a disorder which is associated with aberant remodelling and repair within the lung, progressive pulmonary fibrosis and a median survival of 2-3 years. Maxmising synergy bewteen Systems Biology and Translational Medicine represents a universal challenge. Embedding clinical scientists in a Systems Medicine faremework represents a crucial objective in delivering on the promise of Systems Medicine.

Monday 10th September

14.00 - 16.00

Session II: Frontiers in Medicine

Pierre De Meyts, De Meyts R&D Consulting SPRLU, Belgium

Carton Suite

Interactome Networks and Human Disease Marc Vidal Centre for Cancer Systems Biology, Department of Genetics, Dana-Farber Cancer Institute/Harvard Medical School, Boston, USA For over half a century it has been conjectured that macromolecules form complex networks of functionally interacting components, and that the molecular mechanisms underlying most biological processes correspond to particular steady states adopted by such cellular networks. However, until a decade ago, systems-level theoretical conjectures remained largely unappreciated, mainly because of lack of supporting experimental data. To generate the information necessary to eventually address how complex cellular networks relate to biology, we initiated, at the scale of the whole proteome, an integrated approach for modeling protein-protein interaction or “interactome” networks. Our main questions are: How are interactome networks organized at the scale of the whole cell? How can we uncover local and global features underlying this organization, and how are interactome networks modified in human disease, such as cancer?

Cancer Response to Targeted Agents: Dynamics and Variability from Single Cell Data

Vito Quaranta12, Darren R. Tyson1,2, Shawn P. Garbett1,2, Peter L. Frick1,2 1Department of Cancer Biology, Vanderbilt University Medical Center, Nashville, TN, USA, 2Center for Cancer Systems Biology @ Vanderbilt, Vanderbilt University Medical Center, Nashville, TN, USA. Typical cell proliferation assays estimate cell counts at fixed time-points, not dynamically. Moreover, in the presence of drugs affecting proliferation, they provide little information on underlying individual cellular behaviors (e.g., apoptosis, decreased cell division rate, etc.). We present Fractional Proliferation, an integrated method, based on extended time-resolved automated microscopy that quantifies cell proliferation dynamics in response to drugs by integrating population- and single-cell data. Direct cell

Page 14

count data, collected every 6 minutes, are initially fit with a novel Quiescence-Growth mathematical model, based on three parameters: division, death and quiescence rates. This model is then substantiated by extracting these rates from experimental observations of hundreds of single cells, fitted with an Exponentially Modified Gaussian model. In the final output graphs, Fractional Proliferation describes the underlying behavioral dynamics that result in proliferative changed by perturbations. Using this method, we discovered that the response of lung cancer cell lines to erlotinib, an epidermal growth factor receptor tyrosine kinase inhibitor, is a nonlinear process dominated by an increased rate of cell entry into quiescence. Even in highly sensitive, homogeneous “oncogene-addicted” cell populations, we observed variability of cell-to-cell response. Up to 72 hours of treatment, quiescence prevailed, with only a modest increase in death rate. After 72 hours, the population of treated cells reaches a new steady state in the presence of erlotinib characterized by an overall rate of proliferation, which is a composite of death, quiescence and division rates. Similar results were obtained with oncogene-addicted melanoma and breast cancer cell lines treated with the respective targeted oncogene inhibitors. In contrast to our results, drug targeting of addicting oncogenes has thus far been thought to result in massive cell death. Instead, our findings indicate that it may cause response behaviors other than death, underscoring the realistic in vitro representation of cell proliferative response to perturbations provided by Fractional Proliferation, and providing means to optimize and improve discovery and deployment of targeted therapy in cancer.

A Brief View of Systems Medicine and Genetics Jie (Bangzhe) Zeng, Xiao-Xue Zeng Institute of Systems Biological Engineering, China To investigate natural and artificial biosystems, systems biology which included systems genetics and systems medical study of diseases, systems biological engineering worked as synthetic biology and systems biotechnology, were established on the theoretical fundamentals of evolution, systems and structure theories. The science and engineering of biosystems are disciplines by integrative methodology of computational, experimental and engineering biology. Systems dynamics of cytogenesis and bio-molecular networks is explored as the mechanism for the evolution of genomic structures and cell lineages mapping during pattern formation of organisms. For drug discovery, micro-fluidic chips are useful for the functional analysis, identification of expression genes among difference cell-types and drug treatments, and also used for the designing of artificial cells such as neurons and neuronal communication networks etc. Refs: 1. Zeng (B.) J., On the holographic model of human body, 1st National Conference of Comparative Studies Traditional Chinese Medicine and West Medicine, Medicine and Philosophy, April, 1992 (on the concept of systems medicine). 2. Zeng (B.) J., On the concept of system biological engineering, Communication on Transgenic Animals, No. 6, June, 1994. 3. Zeng (B.) J., Transgenic animal expression system – transgenic egg plan (goldegg plan), Communication on Transgenic Animal, Vol.1, No.11, 1994 (on the concept of system genetics). 4. Zeng (B.) J., From positive to synthetic medical science, Communication on Transgenic Animals, No.11, 1995 (on systems medicine). 5. Zeng(B.)J., The structure theory of self-organization systems, Communication on Transgenic Animals, No.8-10, 1996. Etc.

Page 15

Pathway-GPS and SIGORA: Identifying relevant pathways based on the over-representation of their gene-pair signatures. David Lynn1, Amir Foroushani1, Fiona Brinkman2

1Teagasc, Ireland 2Simon Fraser University, Canada Predominant pathway analysis approaches treat pathways as collections of individual genes and consider all pathway members as equally informative. As a result, at times spurious and misleading pathways are inappropriately identified as statistically significant, solely due to components that they share with the more relevant pathways. We introduce the concept of Pathway Gene-Pair Signatures (Pathway-GPS) as pairs of genes that, as a combination, are specific to a single pathway. We devised and implemented a novel approach to pathway analysis, Signature Over-representation Analysis (SIGORA), which focuses on the statistically significant enrichment of Pathway-GPS in a user-specified gene list of interest. In a comparative evaluation of several published datasets, SIGORA outperformed traditional methods by delivering biologically more plausible and relevant results. Availability: An efficient implementation of SIGORA, as an R package with precompiled GPS data for human and mouse pathway annotations can be downloaded from CRAN. (http://cran.rproject.org/web/packages/sigora/)

Tuesday 11th September

09.00 - 12.50

Session III: Potential for Therapeutics

Mark Ferguson, Science Foundation Ireland

Carton Suite

Systems Approaches to Parkinson’s Disease Rudi Balling Luxembourg Centre for Systems Biomedicine (LCSB), University of Luxembourg Parkinson´s disease is one of the major neurodegenerative diseases and primarily due to the loss of dopaminergic neurons in the substantia nigra. Pathogenesis of PD is not well understood, however it is thought to be multi-factorial and age-related with many genetic and environmental factors involved. In order to capture the rapidly increasing information and inter-relationships between different factors contributing to PD we are in the process of establishing a “PD-disease map”. The map captures and visualizes all major molecular pathways involved in PD pathogenesis and can serve as a resource for further computational analyses and as a platform for community level collaborations. We have also begun to study the role of neuroinflammation in the development of neurodegeneration. For this purpose we analyzed the metabolomic response of microglia cells to inflammatory stress and identified a new metabolite and pathway involved in metabolic mediated immunity.

Systems Medicine Approaches to improve the Therapeutic Potential of Mesenchymal Stromal Cells in Critical Limb Ischemia Timothy O’Brien Regenerative Medicine Institute, National University of Ireland Galway, Ireland

Page 16

Critical limb ischemia is a common cause of amputation in patients with diabetes mellitus. While revascularization is the treatment of choice for these patients, it has been estimated that approximately 30% of patients are not candidates – so called no option patients. Stem cell transplantation may be an option for these patients by inducing therapeutic angiogenesis. We have shown that bone marrow derived mesenchymal stromal cells (MSCs) have angiogenic properties and demonstrated pre-clinical efficacy and safety of transplantation of MSCs to ischemic limbs. We have developed GMP methodology to manufacture MSCs and have prepared the regulatory submission for a phase 1 dose escalation study of MSC transplantation in no-option critical limb ischemia. This approach however has some limitations such as issues of cell engraftment, cell survival, cell persistence, cell migration and mode of cell delivery. REMEDI is working with SBI to enhance our understanding of MSC biology in order to refine therapy and develop next generation cell therapeutics. While intravenous delivery of MSCs initially targets the lungs followed by injured areas, migration of cells to the site of injury is relatively inefficient. The mechanisms and signalling events responsible for MSC migration have not been fully elucidated. We have shown that MCP-1 plays a pivotal role in MSC migration in the context of cancer and myocardial infarction and will present data on the signalling pathways responsible for this effect. We also define a role for ROCK/Rho signaling in MSCs that is essential for actin-mediated tail retraction and efficient migration. Together these results provide novel insights into the important mechanisms orchestrating MSC migration, which will be vital in understanding and maximizing the therapeutic potential of MSCs. In summary, success of cell transplantation strategies for critical limb ischemia may be increased by understanding the fundamental biology of this cell type using systems medicine approaches.

Biomarker Discovery and Validation for Prostate Cancer patient stratification into appropriate treatment options: Addressing the Clinical Dilemma William Watson, Fan Yue, Murphy Brendan, Boyce Susie, Pennington Stephen, Fitzpatrick John and the Prostate Cancer Research Consortium School of Medicine and Medical Science, Conway Institute, University College Dublin, Ireland Prostate cancer remains the most common form of male cancer in the US and Europe. Despite PSA screening decreasing cancer mortality it has been associate with the over detection and over treatment of the disease, impacting on patients quality of life. Determining the most appropriate treatment strategy represents a significant dilemma for the patient and clinician. The Prostate Cancer Research Consortium focus of research has been driven by the clinical need to identify biomarkers that will inform appropriate treatment options. The consortiums approach is to interrogate biomarkers in serum and tissue at the genomic, transcriptomic and proteomic level using novel statistical and bioinformatics approaches to identify panels of biomarkers which will address clinically relevant questions. Preclinical samples were collected from men with different grades and stages of prostate cancer as part of the Prostate Cancer Research Consortium Bioresource. Their analysis by 2-D DIGE and label-free LC-MS/MS and the use of novel statistical software has identified panels of proteins with acceptable predictability for further validation. This panel has also been informed by our epigenetic, transcriptomics and other proteomics projects. Using multiple reaction monitoring assays these panels are being validated in independent samples and then key panel developed as multiplex antibody assays for larger scale validation with samples from the Prostate Cancer Research Consortium Bioresource and our international collaborators which show AUC values in line for clinical utility. With careful validation these panels will be combined with current clinical tools using novel statistical approaches leading to

Page 17

biomarker panels which could improve diagnosis and thus patient outcome through the selection of appropriate treatment options.

Neuropilin as a therapeutic target in Cancer Ian Zachary1, Paul Frankel1,2, Haiyan Jia1,2, Angela Barrett1,, Ashley Jarvis5, Lili Cheng1,2, David Selwood3 and Snezana Djordevic4 1Centre for Cardiovascular Biology & Medicine, BHF Laboratories, Division of Medicine, University College London; 2Ark Therapeutics Ltd, Rayne Building, 3Biological and Medicinal Chemistry Group, Wolfson Institute for Biomedical Research, Division of Medicine, UCL, 4Structural and Molecular Biology, UCL, 5Domainex Ltd (NCE discovery) Neuropilin-1 (NRP1) is a non-tyrosine kinase receptor for vascular endothelial growth factor and class 3 Semaphorins, essential for developmental angiogenesis. NRP1 is highly expressed in many human neoplasms, and is implicated in the pathogenesis of cancer and other human diseases. A growing body of evidence indicates that NRP1 is important in endothelial and tumour cell migration mediated via tyrosine phosphorylation of the adapter protein p130Cas in response to the chemoattractant stimuli of VEGF, HGF and PDGF. These findings suggest that NRP1 is a potential target for the development of novel anti-cancer therapeutics. We have developed small molecule inhibitors of the VEGF-A interaction with NRP1 and structurally characterized NRP1 complexes with these inhibitors by NMR spectroscopy and X-ray crystallography. Mutagenesis studies localized VEGF-A binding in a specific pocket of the NRP1 b1 domain and demonstrated the relevance of the small molecule binding site for endothelial cell functions. Studies in tumour cells and in an in vivo model of tumour growth indicate that small molecule inhibitors of VEGF binding to NRP1 have therapeutic potential as anti-cancer agents. NRP1 is attractive as a target for anti-cancer therapy because its inhibition may circumvent pathways that underlie development of resistance to anti-VEGF drugs. Our findings indicate that this inhibition of NRP1-dependent signalling is attainable with small molecule ligands of the NRP1 VEGF binding site.

Unravelling the path to Mesenchymal Stem Cell migration: A co-ordinated effort of beta gamma subunit, PI3K, Rac and Rho signalling Caroline Ryan, Ainé Prendergast, James Brown, Timothy O’Brien, Frank Barry Regenerative Medicine Institute, National University of Ireland Galway, Ireland Mesenchymal stem cells (MSCs) have generated much interest as a potential source of cells for cell-based therapeutic strategies, particularly as MSCs have the ability to home to damaged tissue and contribute to tissue repair and regeneration. MSCs are thought to migrate to sites of injury in response to chemokine gradients. The mechanisms and signaling events responsible for MSC migration have not been fully elucidated. Understanding these mechanisms is of great clinical relevance. We found that MSCs rapidly migrated in response to a gradient of MCP-1. This directional migration was supported by a polarised phenotype and altered actin and α-tubulin dynamics. MCP-1 induced a rapid redistribution of the chemokine receptor CCR2 and an accompanied clustering of the adapter molecule FROUNT. Using specific chemical inhibitors we found that MSC migration and associated CCR2 and FROUNT clustering was dependent not only on CCR2 activation, but also on the βγ subunits of GPCRs. Analysis of downstream signaling cascades revealed that MCP-1 induced MSC migration was dependent on PI3K signaling and the RhoGTPases, Rac1 and RhoA. We found that MCP-1 induced a rapid phoshorylation of AktSer473 that was abrogated following inhibition of βγ subunits, signifying that MCP-1 specifically activated PI3Kγ. MCP-1 also induced the phosphorylation of the Rac effector PAK and ERK in a βγ subunit-PI3Kγ dependent manner. We also define a role for ROCK/Rho signaling in MSCs that is essential for actin-mediated tail retraction and efficient migration. Together these results provide novel insights

Page 18

into the important mechanisms orchestrating MSC migration, vital in understanding and maximizing the therapeutic potential of MSCs.

Differential localization of A-Raf kinase regulates MST2-mediated apoptosis and differentiation. Jens Rauch1, Drieke Vandamme1, Natalia Volinsky1, Brigitte Mack2, Oliver Gires2, Walter Kolch1 1Systems Biology Ireland, University College Dublin, Ireland 2Clinical Cooperation Group Molecular Oncology, Department of Head and Neck Research, Ludwig-Maximilians-Universitaet A-Raf belongs to the family of oncogenic Raf kinases that are involved in mitogenic signaling by activating the MEK–ERK pathway. Low kinase activity of A-Raf toward MEK suggested that A-Raf might have alternative functions. We recently identified A-Raf as a potent inhibitor of the MST2 tumor suppressor pathway in carcinoma cells. Independent of kinase activity, A-Raf binds to the proapoptotic mammalian sterile 20-like kinase (MST2) thereby efficiently inhibiting apoptosis. MST2 and A-Raf were concomitantly overexpressed and colocalized at mitochondria in cancer cell lines and also in primary human tumors. Here, we show that A-Raf re-localizes to the plasma membrane upon epithelial differentiation in vivo. While in proliferating normal cells and tumour cells A-Raf localizes to the mitochondria, differentiated non-carcinogenic cells of head and neck epithelia express A-Raf at the plasma membrane. We can demonstrate, that constitutively plasma-membrane-localized A-Raf loses its ability to efficiently sequester and inactivate MST2 thus allowing etoposide-mediated apoptosis. These results were corroborated using a transient model system based on the ARIAD heterodimerization system. In addition to differentiation of head and neck epithelia, A-Raf re-localization to the plasma membrane was found during mammary differentiation. Using the MCF7 cell differentiation system, we could demonstrate that overexpression of A-Raf in MCF7 cells induces differentiation. Using quantitative proteomics we are trying to identify novel interaction partners of A-Raf mediating the observed effects. Our findings offer a new paradigm to understand how differential localization of Raf complexes affects diverse signaling functions in normal cells and carcinoma. Acknowledgements: This work was supported by Cancer Research UK and Science Foundation Ireland under grant no. 06/CE/B1129. References: Rauch, J., Moran-Jones, K., Albrecht, V., Schwarzl, T., Hunter, K., Gires, O., and Kolch, W. (2011). c-Myc regulates RNA splicing of the A-Raf kinase and its activation of the ERK pathway. Cancer research 71, 4664-4674. Rauch, J., O'Neill, E., Mack, B., Matthias, C., Munz, M., Kolch, W., and Gires, O. (2010). Heterogeneous Nuclear Ribonucleoprotein H Blocks MST2-Mediated Apoptosis in Cancer Cells by Regulating a-raf Transcription. Cancer Res 70, 1679-1688. Matallanas, D., Romano, D., Yee, K., Meissl, K., Kucerova, L., Piazzolla, D., Baccarini, M., Vass, J.K., Kolch, W., and O'Neill, E. (2007). RASSF1A elicits apoptosis through an MST2 pathway directing proapoptotic transcription by the p73 tumor suppressor protein. Mol Cell 27, 962-975. O'Neill, E., Rushworth, L., Baccarini, M., and Kolch, W. (2004). Role of the kinase MST2 in suppression of apoptosis by the proto-oncogene product Raf-1. Science 306, 2267-2270.

Page 19

Tuesday 11th September

13.50 - 17.00

Session IV: Drug Discovery

Jochen Prehn, Department of Physiology and Medical Physics and Centre for Systems Medicine, Royal College of Surgeons in Ireland

Carton Suite

The Cellular uptake of Pharmaceutical Drugs: A problem not of Biophysics but of Systems Biology Douglas Kell Biotechnology and Biological Sciences Research Council, UK and School of Chemistry and the Manchester Institute of Biotechnology, The University of Manchester A fundamental question remains as to whether xenobiotic drugs cross cellular membranes mainly (or exclusively) by ‘passive’ (transporter-independent) diffusion across cellular membranes, or whether they normally (or exclusively) ‘hitchhike’ rides using the carriers normally involved in the metabolism of natural metabolites. The former would involve a biophysical mechanism, based mainly on lipophilicity, while the latter requires a mechanistic understanding of which carriers are involved, and is thus a problem of network or systems biology. In other words [1], is carrier-mediated transport of pharmaceutical drugs the exception or the rule? A huge amount of literature [1-5, and references therein], that I shall summarise, indicates that there is no serious evidence for transbilayer-mediated transfer of pharmaceutical drugs across biological membranes, while there is abundant and increasing evidence for the carrier-mediated route. A recent approach in yeast illustrates this experimentally [6], while the digital availability of principled metabolic network models [7,8] allows one to determine [9], consistent with this, that successful pharmaceutical drugs are much more like metabolites than are the ‘Lipinski-compliant’ molecules typically available in drug discovery libraries. This suggests that cellular drug uptake is more or less exclusively transporter-mediated, and that knowledge of both the metabolome and of the transporters used by individual xenobiotics will be of much value in designing better drugs. [1] Dobson, P. D. & Kell, D. B. (2008). The carrier-mediated cellular uptake of pharmaceutical drugs: an exception or the rule? Nat Rev Drug Discov 7, 205-220. [2] Kell, D. B. & Dobson, P. D. (2009). The cellular uptake of pharmaceutical drugs is mainly carrier-mediated and is thus an issue not so much of biophysics but of systems biology. In Proc Int Beilstein Symposium on Systems Chemistry (ed. M. G. Hicks and C. Kettner), pp. 149-168 - http://www.beilstein-institut.de/Bozen2008/Proceedings/Kell/Kell.pdf. Logos Verlag, Berlin. [3] Dobson, P., Lanthaler, K., Oliver, S. G. & Kell, D. B. (2009). Implications of the dominant role of cellular transporters in drug uptake. Curr Top Med Chem 9, 163-184. [4] Kell, D. B., Dobson, P. D. & Oliver, S. G. (2011). Pharmaceutical drug transport: the issues and the implications that it is essentially carrier-mediated only. Drug Disc Today 16, 704-714. [5] Kell, D. B., Dobson, P. D., Bilsland, E. & Oliver, S. G. (2012). The promiscuous binding of pharmaceutical drugs and their transporter-mediated uptake into cells: what we (need to) know and how we can do so. Drug Disc Today, submitted. [6] Lanthaler, K., Bilsland, E., Dobson, P., Moss, H. J., Pir, P., Kell, D. B. & Oliver, S. G. (2011). Genome-wide assessment of the carriers involved in the cellular uptake of drugs: a model system in yeast. BMC Biology 9, 70. [7] Herrgård, M. J., Swainston, N., Dobson, P., Dunn, W. B., Arga, K. Y., Arvas, M., Blüthgen, N., Borger, S., Costenoble, R., Heinemann, M., Hucka, M., Le Novère, N., Li, P., Liebermeister, W., Mo, M. L.,

Page 20

Oliveira, A. P., Petranovic, D., Pettifer, S., Simeonidis, E., Smallbone, K., Spasić, I., Weichart, D., Brent, R., Broomhead, D. S., Westerhoff, H. V., Kırdar, B., Penttilä, M., Klipp, E., Palsson, B. Ø., Sauer, U., Oliver, S. G., Mendes, P., Nielsen, J. & Kell, D. B. (2008). A consensus yeast metabolic network obtained from a community approach to systems biology. Nature Biotechnol. 26, 1155-1160. [8] Heavner, B. D., Smallbone, K., Barker, B., Mendes, P. & Walker, L. P. (2012). Yeast 5 - an expanded reconstruction of the Saccharomyces cerevisiae metabolic network. BMC Syst Biol 6, 55. [9] Dobson, P. D., Patel, Y. & Kell, D. B. (2009). "Metabolite-likeness" as a criterion in the design and selection of pharmaceutical drug libraries. Drug Disc Today 14, 31-40.

New Drug Screening Strategies for Multiple Myeloma Treatment Lukas Huber Cell Biology Division, University of Innsbruck and Oncotyrol, and Austrian drug screening institute (ADSI)

Austria The ADSI (Austrian Drug Screening Institute) and ONCOTYROL have been established as open-innovation platforms. ADSI and ONCOTYROL bring together the expertise of translational researchers, clinicians and pharmaceutical industry in focused projects to apply innovative approaches for the product-oriented development of novel personalized cancer therapies. Multiple Myeloma (MM) is an incurable disease with rapidly growing prevalence. Despite the introduction of novel agents, treatment often fails. Consequently, it is our goal to seek out novel strategies for the development of novel diagnostic and therapeutic options. MM development involves not only genetic changes within the tumour cells but also the emergence of supportive conditions by the bone marrow microenvironment (BMM). To target the essential components of this support system we establish preclinical in vitro and in vivo models of MM that include functionally relevant elements of the BMM. We analyse clinical data to correlate the presence of particular MM-BMM interactions with the development of MM, with its intrinsic therapy resistance as well as with disease relapse due to the development of acquired drug resistance. This correlative data are validated by an autologous MM-BMM co-culture assay, reverse translated into in vitro screening and in vivo validation models, which are subsequently used to develop lead compounds that target myeloma cells within their microenvironment. The clinical expertise of several Haematology/Oncology Divisions, the research experience of academic laboratories and the pharmaceutical knowhow of small and medium sized enterprises as well as biotech industry are united to drive this important development and to ensure translation towards clinical trials. We expect this work to impact on the establishment of better diagnostics and new drug screening approaches for MM ultimately leading to reduced patient mortality by introducing novel personalised therapies based on individual ex vivo phenotyping. In addition, results will open new markets for industry partners, given that envisaged drug screening methods are applicable to other areas of research, drug discovery, and development of products and services personalized medicine.

Uncoupling oncogenic receptor kinases from their intracellular nanocomputers Stephan Feller University of Oxford, UK Oncogenic receptor tyrosine kinase like c-Met, EGFR and ErbB2 transmit many of their signals through protein complexes that contain large multi-site docking (LMD) proteins. The LMD proteins function as essential assembly platforms for sophisticated computational units (a.k.a. as ‘signalosomes’ or ‘cellular

Page 21

nanocomputers’) in the cytoplasm of metazoan cells. In normal cells, such molecular computers integrate multiple kinase and phosphatase signals and subsequently steer several well-orchestrated functional outputs. In cancer cells, however, the same computational units drive tumour malignancy. Well-studied examples of LMD proteins with critical roles in major cancers (breast, colorectal etc.) are the Gab family proteins Gab1 and Gab2. Gab2 is overexpressed in a subset of breast cancers and promotes oncogene actions in vivo. Gab1 is crucial for in vivo signaling of the c-Met receptor, which is deregulated in a wide range of cancers. Gab family proteins display well folded N-termini in the form of pleckstrin homology (PH) domains, but these are followed by tail regions of several hundred amino acids which are supposedly entirely unstructured, at least according to standard structure prediction programs. The molecular modes of how Gab proteins facilitate the computation of cross-talking pathways have remained a mystery. How is it possible that largely disordered proteins organize effective and highly sophisticated signal computation units? The answer seems to be, that they are not as ‘chaotic’ as previously thought. We have recently made two observations of fundamental importance that would seem to challenge the current structural view of entirely ‘intrinsically disordered’ protein tails in Gab proteins and similar LMD proteins. Firstly, we have shown by biophysical methods that segments of the Gab tail regions can indeed form secondary structure elements with helical properties that are distinct from those of alpha-helices (Harkiolaki et al. 2009, Structure). Secondly, we have evidence that the long tail of the Gab1 protein can interact with the N-terminal PH domain of Gab1 at several sites (N-terminal folding nucleation [NFN] hypothesis; Simister et al. 2011, PLoS Biol). This should generate loop regions where functionally defined subcomplexes can assemble and hence facilitate effective cross-talk between multiple pathways, i.e. biological computation (Lewitzky et al. 2012, FEBS Lett). Interestingly, the helical tail segments we identified serve as critically important docking sites for adaptor proteins that are necessary for coupling the intracellular nanocomputers to oncogenic receptors. We believe that they may be an ‘Achilles heel’ in the system that could be targeted in the context of human cancers. Uncoupling an oncogenic receptor from its intracellular signalosome should affect several pathways at once, thereby in essence creating a multi-pathway inhibitor drug, which simultaneously affects cell survival, proliferation, motility and invasion. Moreover, since receptors like c-Met exert the majority of their actions during embryogenesis, a clinically relevant ‘therapeutic window’ for such drugs may exist. In order to explore this possibility we have, based on our crystal structures and in collaboration with the group of Andrew Hamilton (Oxford University), begun to synthesise and functionally analyse new helicomimetic scaffold molecules (unpublished data). These molecules, aimed at uncoupling oncogenic receptors from their cellular nanocomputers, explore an entirely new section of the chemical space of drug-like molecules.

Cell Signaling by receptor tyrosine kinases; From basic principles to cancer therapy Joseph Schlessinger Yale School of Medicine, Connecticut, USA Receptor tyrosine kinases (RTKs) comprise a family of cell surface receptors that control many critical cellular processes. Various human diseases are caused by dysfunction in RTKs or in their intracellular

Page 22

signaling pathways. Stem Cell Factor (SCF) initiates its multiple cellular responses by binding to the extracellular region of the RTK KIT resulting in receptor dimerization and tyrosine kinase activation. We have determined the crystal structure of the entire extracellular region of KIT before and after SCF stimulation. The structures show that KIT dimerization is driven by SCF binding whose sole role is to bring two KIT molecules together. We have used X-ray crystallography, electron microscopy (EM) and biochemical experiments to determine the three dimensional structure of intact, SCF-stimulated KIT dimers. Several forms of dimeric KIT molecules with different asymmetric arrangements of the two tyrosine kinase domains were identified. These asymmetric contacts may represent specific interactions occurring between two KIT tyrosine kinase domains poised towards trans autophosphorylation. We propose that cooperative interactions mediated by multiple weak homotypic contacts between the extracellular, transmembrane and cytoplasmic regions of KIT are responsible for tyrosine kinase activation and cell signaling in normal and transformed cells. Sutent is a drug that blocks the tyrosine kinase activities of several RTKs including KIT and VEGFR2. Sutent has been approved by the FDA for the treatment of advanced kidney cancers, gastrointestinal tumors and for endocrine pancreatic cancers . A scaffold-based drug discovery approach was developed enabling the development of new families of inhibitors for protein kinases and other enzymes that play a role in cell signaling. V600E B-RAF mutant is an oncogenic mutation identified in approximately 50% of melanomas and in 4% of all solid tumors. By using a scaffold-based drug discovery approach, a potent inhibitor of V600E B-RAF mutant has been discovered designated PLX 4032 (Vemurafenib). Phase-I and-II clinical trials in multiple medical centers revealed 80% responses such as tumor shrinkage, delay in tumor progression and survival of melanoma patients harboring the B-RAF mutation. Primary endpoints of phase-III clinical trials in multiple medical centers were achieved to support Vemurafenib registration. On August 17, 2011 FDA approved Vemurafenib with a brand name Zelboraf for treatment of melanomas harboring B-RAF activating mutation.

Discovering new lead compound candidates and combination therapy model using traditional Oriental Medicine Jongwook Jeon, Lee Jungsul Korea Advanced Institute of Science & Technology System Medicine and traditional Oriental Medicine (OM) have considerable common properties. Interpreting life as system composed of non-linear, multiple pathways, and several intercalated causalities is supposed as the principle of them. Here we add one interesting phenomenon observed universally in the traditional OM. Specificity to a certain symptom of one herb, for example Jinseng, is lower than that of two herbs, Jinseng and Danggui. This means cocktail of the two herbs is necessary for the improved therapeutic efficacies, which implies the synergistic effects of the each herb. We focused on the big data contained in the traditional oriental medicine documents or folk remedy text in Korea, such as Dongui-bogam(UNESCO heritage of record) or Inje-ji. They accumulate clinically observed effective treatments with little mechanistic details how the treatments work, but it is likely that the treatments contain chemical compounds therapeutically effective for the symptoms the treatments aim to resolve. We performed statistical analysis of traditional OM text 1) to give clues how the synergism work with two or more herbs in the regular treatments (cocktail treatments) and 2) to provide promising lead compound candidates from the cocktails. Over five thousand cocktail treatments included in prominent text source were processed, and all symptoms and medicinal herbs were identified. After taking advantage of active components included in herbs we connected the chemicals with symptoms to which the chemicals are likely to have therapeutic effects by integrating TCM@Taiwan, Chembl, DrugBank, and ClinicalTrials databases. As a representative example we present chemicals specifically used for

Page 23

diabetes mellitus (DM) in the text. Of the specific compounds baicalein, berberine, and chrysin share target genes with approved drugs for DM. Especially berberine was turned out to be successful in lowering blood glucose level in type 2 DM patients with dyslipidemia at clinical trial phase 3 (ClinicalTrials.gov Identifier NCT00462046). It suggests that several other specific compounds for DM such as 25(R,S)-ruscogenin, columbianadin, jatrorrhizine, and schisandrol as a few examples, are promising lead compound candidates for DM. We provided here a method by which compounds specific for particular diseases can be identified. Moreover it can provide promising combination of compounds with therapeutic effects only when used in the combination. Eventually the method we present can be used to identify promising lead compounds or combination of the compounds for disease of interest from the accumulated clinical data of traditional OM.

Wednesday 12th September

09.00 - 12.50

Session VI: Omics to Systems Medicine

Walter Kolch, Systems Biology Ireland and Conway Institute, University College Dublin

Carton Suite

From proteomics and mathematical models to new therapies for colorectal cancer Walter Kolch, David Romano, David Matallanas-Gomez, Lan Nguyen, Boris Kholodenko Systems Biology Ireland and Conway Institute, University College Dublin, Ireland Ras is frequently mutated in human cancers. Ras signalling has been intensely studied, revealing a complex network of downstream effectors that can mediate multiple outputs including cell proliferation, malignant transformation, but also apoptosis. An unsolved question is how the different Ras effector pathways are coordinated, and how distinct cell fate decisions arise from such coordination. In order to address these questions we are using a combination of quantitative interaction proteomics and mathematical modelling to reconstruct and analyse the Ras dependent signalling network. K-Ras, the most frequently mutated Ras isoform in human tumours, can activate cell transformation via the Raf-MEK-ERK and PI3K – AKT pathways and apoptosis via the RASSF1A – MST2 pathway. We have mapped pro-apoptotic pathways using quantitative interaction proteomics. Surprisingly, in colon cancer the transforming activity of the mutant K-Ras allele requires the inhibition of the MST2 pathway by the wildtype K-Ras allele or the loss of MST2 expression. Downregulation of MST2 signalling seems to be more generally important for cellular transformation as suggested by mathematical modelling and confirmed through experimentation. Using dynamic models to analyse how the signal distribution between transforming and apoptotic pathways is coordinated showed that competing protein interactions play a major role, and in conjunction with posttranslational modifications that change affinities work like Boolean gates that can compute cell fate decisions. The results also make predictions about therapies tailored to tumour stages and individual mutation patterns.

Apoptosis modelling predicts therapy responses in Cancer patients: Towards Personalised Oncology Approaches Jochen Prehn1, Andreas Lindner1, Gerhardt Boukes1, Mary Cannon1, Suzanne Hector1, Caoimhin G. Concannon1, Markus Rehm1 Frank Murray2, Deborah McNamara2, Elaine W. Kay1,2, Heinrich J. Huber1

Page 24

1Centre for Systems Medicine, Royal College of Surgeons in Ireland, 2Beaumont Hospital, Beaumont, Ireland.

Apoptosis deficiency, or the ability of cancer cells to escape programmed cell death, contributes to carcinogenesis, and plays a crucial role in the resistance of tumour cells to chemo- and radiotherapy. Most common chemotherapeutic drugs such as the genotoxic agents 5-FU or Oxaliplatin activate the so-called mitochondrial apoptosis pathway. In this pathway, mitochondrial outer membrane permeabilisation (MOMP) and the activation of caspases represent two key control steps required for the execution of apoptosis. MOMP is regulated by the interplay of several anti- and pro-apoptotic members of the BCL2 protein family and hence is best studies at a systemd level. MOMP is induced by homo-oligomerisation of the pro-apoptotic BCL2 family proteins BAK and BAX that form pores in the outer mitochondrial membrane. To gain insights into the apoptotic machinery upstream of MOMP, we constructed an ODE-based computational model that studies BCL2 protein interactions during genotoxic stress. The model demonstrated that the 'indirect activation model' (which disregards an explicit Bax or Bak activation step) requires the assumption that Bax or Bak oligomerisation is either reversible, or that oligomers need to get rapidly degraded to prevent MOMP in the absence of stress. In contrast, the direct activation model (which proposes direct Bax or Bak activation by pro-apoptotic BH3-Only proteins) demonstrated stability, and was subsequently applied to predict the amount of apoptosis in populations of different colon cancer cell lines subjected to genotoxic stress. Using experimentally determined protein levels of BCL2 family members as input, model predictions coincided with apoptosis susceptibility. We next explored the ability of the model to predict therapy responses in stage 3 colorectal cancer patients who received adjuvant 5-FU/Oxaliplatin chemotherapy. Bcl-2 family proteins were quantified in tumour and matched normal tissue (n=20 patients). Modelling demonstrated an increased sensitivity of tumor vs. matched normal tissue to undergo MOMP in response to genotoxic therapy. Furthermore, tumour tissue of patients with a good treatment response (4 years disease free survival) showed a significantly enhanced sensitivity to undergo MOMP in response to genotoxic therapy than patients with poor outcome (recurrence or death from disease within four years) (p<0.02). Even in cells that successfully undergo MOMP in response to genotoxic stress, cells may still evade apoptosis when caspase activation is inhibited. Indeed, the application of an ODE-based model of the caspase activation pathway (‘APOPTO-CELL’) demonstrated that the likelihood of colorectal tumours to undergo apoptosis decreases with advancing disease stage. Systems-level analysis correctly predicted positive or negative outcome in 85% (P = 0.004) of colorectal cancer patients receiving 5-fluorouracil based chemotherapy and significantly outperformed common uni- and multi-variate statistical approaches. Furthermore, modelling of individual patient responses to novel apoptosis-inducing therapeutics such as Smac mimetics or Bcl-2 antagonists revealed markedly different inter-individual responses. In conclusion, we have developed computational models that provided new mechanistic insights into the process of apoptosis. Such models may be developed into new clinical tools that predict treatment responses in cancer patients. Furthermore, such models can be used as stratification tools for the design of future clinical trials that test novel therapeutics such as apoptosis sensitisers. Supported by Science Foundation Ireland, the Health Research Board, and FP7 APO-SYS.

Page 25

Systems Medicine of Neurogenic Hypertension James Schwaber Department of Pathology, Anatomy & Cell Biology, Thomas Jefferson University, USA

High-throughput data are extremely variable. The variability is such that it does not appear to represent deviation around a mean, with samples drawn from a single distribution - it appears more consistent with multiple distinct subtypes or states within the population. “Precision medicine” requires we treat this variability and complexity “not as a glitch but as a feature”. In this regard we view the genome, and thus cell phenotype, as highly adaptive, dynamic in response to changing environment. Our approach is to use the transcriptome, defined as RNA and RNA regulatory proteins, as a surrogate for phenotype, and to both analyze and drive the variability in neuronal phenotype. We take the transcriptome as a summary record of the unique experience of the organism, and indeed of each cell, over its history in the environment – and as not an end state but as highly malleable, including responses leading to disease. Thus, our goal is to develop a molecular physiology that incorporates variability. We study the neural regulation of visceral homeostasis and take regional, pooled cell, and single cell samples in the central neural control system involved. We assay these neural samples at rest and in response to elevated blood pressure. Analysis of the resulting high-dimensional dataset reveals a remarkable complexity and heterogeneity of the samples and of their response to changes in blood pressure. These results demonstrate differential expression modules for each anatomically distinct neuronal group and neuronal type. We have been able to analyze the variability in a way that connects it to higher-level neurophysiological and neuroanatomical features of the system. The results suggest a novel concept of neuronal regulation of cardiovascular homeostasis arising from variability in the molecular physiology. Our view is this approach may now enable a mechanistic molecular physiology of neural function and disease, involving constrained gene regulatory network models and multiscale models. Our multiscale models include the several cell types comprising brain tissue as they interact in innate neuroimmune inflammatory responses during the development of hypertension. We are developing the approach with the aim of discovering the adaptive responses involved in the development of neurogenic hypertension, thought to underlie the vast majority of all hypertension.

Multi-Levelness as the Final Frontier of Systems Medicine Olaf Wolkenhauer Systems Biology and Bioinformatics, University of Rostock, Germany Deciphering and learning to represent multiple levels of structural and functional organization is arguably the fundamental problem of systems medicine. High impact advances will require an understanding of how malfunctioning at the physiological level is related to the functioning of cells at the molecular level. In recent years much has been learned about molecular components and subcellular processes but an important question that remains to be answered is how to integrate data and models across a wide range of spatial and temporal scales. Thus, it is likely that a better understanding of the emergence, progression and treatment of a disease will need to involve novel mathematical approaches. Multi-levelness is a hallmark of biological complexity and, in my view, the final frontier of and the greatest hurdle for systems medicine. To begin to understand the coupling of structure and function across system levels, I will argue that rather than, or complementary to ontological approaches of mechanistic modeling, representing biochemical and biophysical properties, what is also needed is a

Page 26

mathematics that helps in developing epistemological arguments about tissue organization. My argument is however not one for a theory but, rather, for a change of perspective that is less gene, pathway and cell-centric and instead puts the search for organizing principles back into the spotlight that guides our experiments. While much progress has been made in modeling sub-cellular processes, we largely lack understanding of how the functioning of cells is related to their tissue environment. Taking cancer research as an example, an important open question that influences the design of experiments and treatments, is whether carcinogenesis, tumor progression or metastasis are best understood as a malfunctioning of cells or a problem of tissue organization. I argue, the characteristic whole-part relationship of tissues spoils our pragmatic-reductionist approaches by which we consider cells in isolation. In tissues, the properties of every constituent part are largely determined by their functioning in the whole. In other words, a tissue is self-organizing, that is, each cell is at once acting (cause) and reacting (effect), a means and an end. An antireductionist consequence of this is that one cannot separate an understanding of the cell from understanding the tissue that provides the environment and context for cellular function. One important objective for research in systems medicine is then to construct models that can represent both, progressive and regressive determination in tissue organization. The main challenge is that the bottom-up development of a tissue structure, together with the emergence of physiological function is coupled to the topdown coordination of cellular function in response to needs in tissue maintenance and repair. On this backdrop, I present a new ‘tissue functor model’ (TFM) that for the first time allows the representation of the reciprocal determination between cells and tissue. Due to its mathematical roots, the ‘model’ I have developed, is more accurately described as a conceptual framework’. In fact, the ‘model’ is a generalization and abstraction of mechanistic modeling, thereby specifying a category of systems, to which a mechanistic model of interest can belong. Importantly, this conceptual framework enables us therefore to formulate and prove organizing principles: Akin to numerical simulations of mechanistic models to validate hypotheses about cell functions, for the TFM, non-numerical approaches, like theorem proving, are used to formulate and demonstrate a logical argument about the functional organization of tissues. The TFM demonstrates how cells and tissue are characterized by the existence of domains of functioning within which the levels do not interact, even though they are interdependent within the same overall system. The TFM allows for a natural representation of this autonomy of levels and can therefore be used to explain how a distinct physiological behavior at the tissue level can emerge “bottom-up”, from interactions at the cellular level, while simultaneously the tissue level coordinates cell functioning “top-down”, in response to needs for maintenance, repair or regeneration of the tissue.

SWATH-MS provides a deep, reproducible and permanent quantitative proteomic reference for biomarker studies Yansheng Liu1, Ruth Hüttenhain1, Silvia Surinova1, Ludovic CJ Gillet1, Ruedi Aebersold1,2 1Department of Biology, Institute of Molecular Systems Biology, ETH Zurich, Switzerland 2 Faculty of Science, University of Zurich, Switzerland In systems biological and biomedical research, an essential task is to detect and quantify proteins or sets of proteins with high precision across multiple samples. Previously, Selected reaction monitoring (SRM) has emerged as a promising technique for such precise quantification of targeted proteins. Currently, 50–100 proteins can be measured in parallel in the same LC-SRM run. We recently developed a novel SWATH-MS technology as a data independent acquisition (DIA) mass spectrometric method in which the data are acquired on a fast, high resolution Q-TOF instrument by repeatedly cycling through

Page 27

sequential isolation windows (swaths). The SWATH-MS acquisition generates, in a single measurement, a complete recording of the fragment ion spectra of all analytes of 400-1200 m/z precursor range detectable in a biological sample. These complex fragment ion maps are then mined using a targeted data extraction strategy conceptually derived from selected SRM. The pilot datasets suggested that SWATH-MS permits the quantification of as many as compounds that can be identified by regular shotgun proteomics with the accuracy and reproducibility of SRM across many samples. As one application example, we investigated the quantitative and qualitative capabilities of SWATH-MS in human plasma, an important generic resource for biomarker discovery. Specifically, we analyzed the subproteome of the N-linked glycopeptides. We benchmarked the LOQ, reproducibility and quantification accuracy of SWATH-MS as compared to the state-of-the-art SRM. Compared to SRM, SWATH acquisition coupled with a targeted analysis strategy allows the quantification of plasma glycoproteins with only a slightly higher LOQ (3 times less sensitive), an equal variability, similar accuracy and dynamic range. Expressing the peptide limits at a protein level and assuming a 100% glyco-capture efficiency, we estimate the LOQ of SWATH-MS to reach down to 5 ng/mL for plasma glycoproteins. Importantly, as a DIA method, all the biomarker features recorded in SWATH-MS can be reproducibly profiled for one sample and easily matched between samples. This is a crucial requirement of clinical proteomics. Our results suggest that SWATH-MS analysis can provide a reproducible, deep and quantitative proteomic reference for clinical samples. Future studies will focus on further applications of this approach across multiple human samples to address biologically and clinically important questions. References: [1] Gillet, L. C., et al., Targeted data extraction of the MS/MS spectra generated by data independent acquisition: a new concept for consistent and accurate proteome analysis. Mol Cell Proteomics. 2012, 11(6): O111.016717. [2] Reiter, L., et al., mProphet: automated data processing and statistical validation for large-scale SRM experiments. Nat Methods, 2011 (8): 430-435.

Identification of molecular mechanisms underlying MCF-7 cell differentiation using proteomics and systems biology approaches Natalia Volinsky1, Alex VonKriegsheim1, Christian Preisinger2, Tapesh Santra1, Nina Saban3, Marc Birtwistle1, Albert Heck2, Walter Kolch1, Boris Kholodenko1

1 Systems Biology Ireland, University College Dublin, Ireland, 2 Utrecht University, Netherlands 3 Rudjer Boskovic Institute, Croatia In MCF-7, a breast cancer cell line, ErbB receptor ligands induce different cell phenotypes. Epidermal Growth Factor (EGF) induces cell proliferation, whereas Heregulin (HRG) induces an irreversible differentiation process accompanied by lipid droplets formation within the cells. So far, no signalling or transcriptional events regulating MCF-7 differentiation have been identified. Using a set of specific chemical inhibitors we identified PI3 kinase activation as a key event in MCF-7 differentiation. Unlike some other well studied differentiation models, no Erk 1/2 activation is required for successful MCF-7 differentiation. Supporting this finding, Insulin, a strong activator of PI3 kinase but only a weak activator of MAPK pathway, was demonstrated to induce lipid accumulation in MCF-7 cells. Additionally, knocking down PTEN, a negative regulator of the PI3 kinase signalling pathway was shown to be sufficient to induce MCF-7 differentiation. Based on this initial data, we decided to identify further signalling events involved in the MCF-7 differentiation process. We employed a quantitative phosphoproteomics approach, stable isotope

Page 28

labeling with amino acids in cell culture (SILAC) followed by TiO2 enrichment, to identify and quantify differentiation specific changes in the proteome and phosphoproteome. The phosphoproteomics data acquired was further analyzed by bioinformatical tools and lead us to validate the involvement of two disparate signalling pathways. Pathway mapping lead us to investigate mTOR signalling and we could confirm that mTOR signalling is required for cell differentiation. Further, analysis of consensus sequence of deferentially phosphorylated peptides, predicted the involvement of several other kinases, including CK2 (former Casein Kinase II), in cell differentiation. This prediction was validated by using both, a CK2 selective chemical inhibitor and by a CK2-targeted siRNA knock-down. In both cases, CK2 inhibition resulted in increased lipid accumulation in differentiating MCF-7 cells. Moreover, CK2 inhibition promoted differentiation of cells grown under basal conditions. These data demonstrates that CK2 acts as a negative regulator of lipid accumulation during MCF-7 cell differentiation. Here we demonstrate that bioinformatical analysis of phosphoproteomics data is a useful tool for finding non-intuitive candidate kinases involved in biological processes.

Thursday 13th September

09.00 - 12.00

Session VII: Computational Modelling

Boris Kholodenko, Systems Biology Ireland, University College Dublin, Ireland Carton Suite

Modelling of Cancer Kinome Networks

Rune Linding Department of Systems Biology, Technical University of Denmark Biological systems are composed of highly dynamic and interconnected molecular networks that drive biological decision processes. The goal of network biology is to describe, quantify and predict the information flow and functional behaviour of living systems in a formal language and with an accuracy that parallels our characterisation of other physical systems such as Jumbo-jets. Decades of targeted molecular and biological studies have led to numerous pathway models of developmental and disease related processes. However, so far no global models have been derived from pathways, capable of predicting cellular trajectories in time, space or disease. The development of high-throughput methodologies has further enhanced our ability to obtain quantitative genomic, proteomic and phenotypic readouts for many genes/proteins simultaneously. Here, I will discuss how it is now possible to derive network models through computational integration of systematic, large-scale, high-dimensional quantitative data sets. I will review our latest advances in methods for exploring phosphorylation networks. In particular I will discuss how the combination of quantitative mass-spectrometry, systems-genetics and computational algorithms (NetworKIN [1] and NetPhorest [4]) made it possible for us to derive systems-level models of JNK and EphR signalling networks [2,3]. I shall discuss work we have done in comparative phospho-proteomics and network evolution[5-7]. Finally, I will discuss our most recent work in analysing genomic sequencing data from NGS studies and how we have developed new powerful algorithms to predict the impact of disease mutations on cellular signaling networks [8,9]. References: http://www.lindinglab.org Linding et al., Cell 2007.

Page 29

Bakal et al., Science 2008. Jørgensen et al., Science 2009. Miller et al., Science Signaling 2008. Tan et al., Science Signaling 2009. Tan et al., Science 2009. Tan et al., Science 2011. Creixell et al., Nature Biotechnology Sep 2012.

Agent-based modelling approach of immune defense against opportunistic human pathogenic fungi Christian Tokarski1, Sebastian Germerodt1, Sabine Hummert1,2, Franziska Mech2, Marc Thilo Figge2, Anja Schroeter1, Stefan Schuster2

1Friedrich Schiller University Jena, 2Leibniz Institute for Natural Product Research and Infection Biology, Hans Knoell Institute, Jena Opportunistic human pathogenic fungi like the ubiquitous fungus Aspergillus fumigatus are a major threat to immunocompromised patients. A weakened immune defence, due to diseases like infection with human immunodeficiency virus (HIV) or in the course of medical treatment, renders the body vulnerable to invasive mycoses that often lead to the death of the patient. While the number of immunocompromised patients is rising, the process and dynamics of defence against invaded and ready to germinate fungal conidia are still insufficiently understood. Here, we focus on neutrophil granulocytes which, besides macrophages, form an important line of defence in that they can engulf and degrade conidia. Live imaging shows the interaction of those phagocytes and conidia as a dynamic process of touching, dragging and phagocytosis. To elucidate the dynamics of the interaction of Aspergillus fumigatus conidia and neutrophils, live imaging data have been recorded and analyzed. The goal is to make a first step in analyzing these data by computer simulations. To unravel strategies of phagocytes on the hunt for conidia an agent-based modelling approach is used to test the impact of different "hunting" strategies of neutrophils on their phagocytosis efficiency. It is known that neutrophils have the ability to recruit other neutrophils. A central question tackled in this work is whether chemical communication and chemotaxis of neutrophils improve the clearing efficiency. Therefore, different modes of movement of phagocytes are tested regarding their clearing efficiency. Simple random walk as well as a short-term persistence in keeping direction, both without sensing of any chemotactic molecules, were tested. Then, perception of secreted metabolites either of germinating conidia or signal molecules from complement system was considered as well as positive feedback activation via chemical communication between neutrophils. In a large-scale modeling approach representing a part of lung tissue we additionally tested the efficiency of chemotactic communication between neutrophils clearing clustered distributions of conidia simulating an infection scenario. The "short-term persistence" hunting strategy turned out to be more efficient than the simple random walk. Following a gradient of chemokines released by conidia is even better, depending on the diffusion parameters of the chemokines. We show that the efficiency in clearing conidia using communication between neutrophils depends on the spatial distribution of conidia, so that two cases can be distinguished. If conidia are distributed randomly, communication does not pay off because it leads to an aggregation of neutrophils, so that many conidia are unaffected from neutrophilic attacks. If conidia are clustered, communication of neutrophils results in a higher clearing efficiency, since attracting other neutrophils for a faster cooperative clearing pays off. Furthermore, distinct strategies using communication are shown resulting in either a quick or a thorough success of the neutrophils.

Page 30

Gene regulatory network controllability to identify cellular reprogramming driver genes Antonio del sol1, Isaac Crespo1, Hiroaki Kitano2

1 Luxembourg Centre for Systems Biomedicine, University of Luxembourg 2 Systems Biology Institute, Tokyo, Japan

Controllability of complex networks relates to the ability to drive these networks from any initial state to any desired final state. In particular, understanding the molecular mechanisms of normal cell fate decisions shall provide us with valuable insights on how to control transitions between different cellular phenotypic states, and therefore develop strategies for cellular reprogramming. In this respect, an important goal is to understand the dynamics of molecular circuits that play a key role in the regulatory process; and elucidate how these circuits control cell transitions. Furthermore, the stability of molecular circuits in gene regulatory networks has been used to model transitions from healthy to disease states. Liu, Slotine and Barabasi have recently introduced a method to identify subsets of nodes in complex networks to achieve full controllability of these networks. Application of this method to gene regulatory networks would imply that in order to exert full control over these networks we roughly must control 80% of their genes. Therefore, although this approach guarantees full controllability of complex networks, it does not constitute a feasible strategy for cellular reprogramming. Furthermore, it has been experimentally shown that few driver genes are able to lead cellular systems from one stable phenotype to another. Thus, since cellular reprogramming concerns driving the system between pairs of stable cellular states, represented as stable steady states in the gene expression landscape (termed as attractors), it constitutes a particular case of the definition of full controllability, and therefore it is needed to introduce a different procedure for the identification of genes driving such transitions. Here, we introduce a network topology-based method that aims at identifying sets of genes, which drive transitions between pairs of attractors. Our method relies on the fact that a necessary condition for multistability, i.e. the existence of multiple attractors is the existence of positive circuits in gene regulatory networks (the sign of a circuit is defined by the product of the signs of its edges). Therefore, some positive circuits play a crucial role in maintaining the stability of these attractors. Following this rationale, our method identifies a minimal set of nodes belonging to differentially expressed positive circuits (circuits are considered differentially expressed if all their genes are differentially expressed between two attractors) whose perturbations can trigger the transition between these attractors. In order to validate this strategy, we applied our method to 1000 randomly generated networks with the same topological properties of a well characterized gene regulatory network (E. Coli). Results showed that perturbations of these minimal sets of driver genes were able to trigger transitions between all pairs of attractors. Further, we applied our method to five different gene regulatory networks describing relevant biological examples of cell fate and reprogramming (T-helper differentiation, HL60-Neutrophil differentiation, EMT, Fibroblast-hepatocyte differentiation, Fibroblast-cardiomyocyte differentiation). These examples nicely illustrate the experimental validation of our predicted driver genes, and therefore the utility of our method. Furthermore, we predict cellular transitions, which can be validated in the future. It is worth mentioning that the predictive power of our method allows us to identify new driver genes in networks reconstructed based on knowledge about different driver genes. For example, in the T-Helper differentiation example, IFN-gamma and IL-4 were identified experimentally as driver genes for the transition from Th0 to Th1 and Th2 respectively, and the gene regulatory network was reconstructed according to this knowledge. After applying our method we found T-bet and GATA3 as key transcription factors in reprograming of T-helper cells into Th1 and Th2 lineages as well as redirection from Th2 to Th1, and there is experimental support for these driver genes.

Page 31

In summary, our method can be used to control transitions between attractor states corresponding to stable cellular phenotypes by perturbing few key genes. Moreover, it can be useful for disease modelling and therapeutic intervention in regenerative medicine.

Signalling Ballet in Four Dimensions Boris Kholodenko Systems Biology Ireland, University College Dublin, Ireland The advancements in “omics” (proteomics, genomics, and metabolomics) technologies have yielded large inventories of genes, transcripts, proteins, and metabolites. The challenge is to find out how these entities work together to regulate cellular responses to external and internal cues. Computational models provide insight into the intricate relationships between stimuli and responses, revealing mechanisms that enable networks to amplify signals and reduce noise and generate discontinuous bistable dynamics or oscillations. In this talk, we review experimental and theoretical progress towards better understanding how the cellular functions are encoded by the spatiotemporal dynamics of downstream signalling networks. We focus on how cellular networks integrate the temporal and spatial information to determine specific biological outcomes, and how the design features of the networks specify biological decisions. In addition to mechanistic computational modeling, a top-down approach to inferring the structure of cellular signaling and gene networks will be presented. We demonstrate how dynamic connections leading to a particular network node can be retrieved from experimentally measured network responses to perturbations influencing other nodes. References. 1. Kholodenko, B. N., Hancock, J. F. & Kolch, W. (2010) Signalling ballet in space and time. Nat Rev Mol Cell Biol. 11, 414-426. 2. Nakakuki, T., Birtwistle, M. R., Saeki, Y., Yumoto, N., Ide, K., Nagashima, T., Brusch, L., Ogunnaike, B. A., Okada-Hatakeyama, M. & Kholodenko, B. N. (2010) Ligand-specific c-Fos expression emerges from the spatiotemporal control of ErbB network dynamics, Cell. 141, 884-896. 3. Nguyen, L. K., Munoz-Garcia, J., Maccario, H., Ciechanover, A., Kolch, W. & Kholodenko, B. N. (2011) Switches, Excitable Responses and Oscillations in the Ring1B/Bmi1 Ubiquitination System, PLoS Comput Biol. 7, e1002317. 4. Borisov, N., Aksamitiene, E., Kiyatkin, A., Legewie, S., Berkhout, J., Maiwald, T., Kaimachnikov, N. P., Timmer, J., Hoek, J. B. & Kholodenko, B. N. (2009) Systems-level interactions between insulin-EGF networks amplify mitogenic signaling, Mol. Syst. Biol. 5, 256. 5. Tsyganov, M. A., Kolch, W. & Kholodenko, B. N. (2012) The topology design principles that determine the spatiotemporal dynamics of G-protein cascades, Molecular BioSystems. 8, 730-43. 6. Munoz-Garcia, J., Neufeld, Z. & Kholodenko, B. N. (2009) Positional information generated by spatially distributed signaling cascades, PLoS Comput. Biol. 5, e1000330. 7. Kholodenko, B., Yaffe, M. B. & Kolch, W. (2012) Computational approaches for analyzing information flow in biological networks, Sci Signal. 5, re1.

The future of Medicine (And Health) Hans Lehrach Max Planck Institute for Molecular Genetics, Berlin, Germany The solution of many medically important problems depends primarily on being able to predict the behaviour of complex networks (e.g. the biological networks acting within a tumor, but also in the other tissues of the patient) under complex disturbances (e.g. a particular therapy). Targeted drugs used in

Page 32

oncology therefore typically help only one quarter of the patients they are used on, with the others suffering from severe side effects, without any medical benefits. Decades of molecular cancer research, but also the recent genome revolution, have however still not been able to provide this urgently needed power to predict the response of individual patients. We are currently sequencing the genome and transcriptome of the tumor and the genome of the patient for individual cancer patients, as the basis of a ‘virtual patient’ models, which can then be used to predict effect and side effects of specific therapies on the individual patient (www.treat1000.org). In addition, we have proposed IT Future of Medicine (ITFoM), www.ITFoM.eu), with the goal to develop integrated molecular/physiological/anatomical models of every individual in the health care system, on the basis of –omics (genomics, epigenomics, transcriptomics, proteomics, metabolomics, metagenomics etc), imaging and sensor data, as the basis of a new, data rich, computation based individualised medicine of the future.

Page 33

ABSTRACT FOR CLOSING KEYNOTE

Systems Pharmacology and Systems Medicine Garret FitzGerald University of Pennsylvania and Institute for Translational Medicine & Therapeutics, Philadelphia, USA Drug discovery and development has traditionally focused on individual signaling pathways using clinical trials to establish large average signals to detect efficacy or hazard, but ignoring the potential for substantial interindividual variability in drug response. Examples of selective efficacy of anticancer drugs based on individual gene variants have afforded proof of principle for a more personalized approach to therapy. However while individual gene polymorphisms have been associated with variance in response to drugs such as warfarin and clopidogrel, they have had little impact on clinical practice. Systems biology offers the prospect of developing a more thorough understanding of the possibility of benefit or risk from prospective pharmacological interventions and also a more personalized approach of drugs commonly used for heterogeneous conditions such as pain and inflammation than ever can be guided by assessment of single gene variants. An example of the former is the question of how to, if at all, pharmacologically perturb the molecular clock. Presently, we understand that this is a highly conserved, highly regulated system. A central clock in the suprachiasmatic nucleus governs peripheral tissue clocks that have the ability also to function autonomously or to regulate as well as be regulated by central clock function. The clock network appears key to linking networks across tissues and to play an integrative role particularly with respect to metabolism and the inflammatory response. Although chemical screens have yielded small molecules that will perturb clockworks in various ways, studies relevant to therapeutic exploitation of the system have yet to be performed. An example of the latter is an attempt to personalize progressively the use of nonsteroidal antinflammatory drugs (NSAIDs) which are amongst the commonest drugs consumed worldwide. They target the prostanoid biosynthetic pathway to relieve pain and inflammation and to cause – in maybe 1-2% of patients exposed to chronic therapy – serious cardiovascular adverse events. However, there is no science based approach to predicting therapeutic efficacy or cardiovascular risk. Comparison of inter- and intra- individual variability of biochemical response to the COX-2 selective NSAIDs, celecoxib and rofecoxib in healthy volunteers suggests a maximal contribution to interindividual variation from fixed sources ( such as the genome) of ~30%. Thus, even in such controlled circumstances, most variability results from unrecognized environmental factors impacting on individual genomes. PENTACON (the Personalized NSAID Therapeutics Consortium) was formed to develop a science based predictive paradigm for analgesic efficacy and cardiovascular risk. NSAIDs preferential for inhibition of COX-1 and COX-2 will be used as pharmacological probes to seek discriminant signatures using genomics, epigenomics, lipidomics, imaging, microbiomics and metabolomics (GEPLIMM) approaches across 5 model systems – yeast, mammalian cells, zebrafish, mice and humans – progressively to develop an understanding of contexts within which NSAIDs perturb the prostanoid pathway. Novel informatics and statistical tools will be developed to display, integrate and be used to evaluate such heterogeneous data drawn from different species. Hence, we hope to construct biological networks, progressively informed by human data, from which we will derive novel hypotheses relating to prediction of efficacy and cardiovascular risk. These will be then tested prospectively at scale in clinical trials.

Page 34

ABSTRACTS FOR POSTER PRESENTATIONS

Tuesday 11th September

17.00 - 18.00

Session V: Poster Presentations

Carton Suite I

P01: Differential localization of A-Raf kinase regulates MST2-mediated apoptosis and differentiation Jens Rauch1, Drieke Vandamme1, Natalia Volinsky1, Brigitte Mack2, Oliver Gires2, Walter Kolch1 1Systems Biology Ireland, University College Dublin, Ireland 2Clinical Cooperation Group Molecular Oncology, Department of Head and Neck Research, Ludwig-Maximilians-Universitaet, Munich, Germany A-Raf belongs to the family of oncogenic Raf kinases that are involved in mitogenic signaling by activating the MEK–ERK pathway. Low kinase activity of A-Raf toward MEK suggested that A-Raf might have alternative functions. We recently identified A-Raf as a potent inhibitor of the MST2 tumor suppressor pathway in carcinoma cells. Independent of kinase activity, A-Raf binds to the proapoptotic mammalian sterile 20-like kinase (MST2) thereby efficiently inhibiting apoptosis. MST2 and A-Raf were concomitantly overexpressed and colocalized at mitochondria in cancer cell lines and also in primary human tumors. Here, we show that A-Raf re-localizes to the plasma membrane upon epithelial differentiation in vivo. While in proliferating normal cells and tumour cells A-Raf localizes to the mitochondria, differentiated non-carcinogenic cells of head and neck epithelia express A-Raf at the plasma membrane. We can demonstrate that constitutively plasma-membrane-localized A-Raf loses its ability to efficiently sequester and inactivate MST2 thus allowing etoposide-mediated apoptosis. These results were corroborated using a transient model system based on the ARIAD heterodimerization system. In addition to differentiation of head and neck epithelia, A-Raf re-localization to the plasma membrane was found during mammary differentiation. Using the MCF7 cell differentiation system, we could demonstrate that overexpression of A-Raf in MCF7 cells induces differentiation. Using quantitative proteomics we are trying to identify novel interaction partners of A-Raf mediating the observed effects. Our findings offer a new paradigm to understand how differential localization of Raf complexes affects diverse signaling functions in normal cells and carcinoma. References: Rauch, J., Moran-Jones, K., Albrecht, V., Schwarzl, T., Hunter, K., Gires, O., and Kolch, W. (2011). c-Myc regulates RNA splicing of the A-Raf kinase and its activation of the ERK pathway. Cancer research 71, 4664-4674. Rauch, J., O'Neill, E., Mack, B., Matthias, C., Munz, M., Kolch, W., and Gires, O. (2010). Heterogeneous Nuclear Ribonucleoprotein H Blocks MST2-Mediated Apoptosis in Cancer Cells by Regulating a-raf Transcription. Cancer Res 70, 1679-1688. Matallanas, D., Romano, D., Yee, K., Meissl, K., Kucerova, L., Piazzolla, D., Baccarini, M., Vass, J.K., Kolch, W., and O'Neill, E. (2007). RASSF1A elicits apoptosis through an MST2 pathway directing proapoptotic transcription by the p73 tumor suppressor protein. Mol Cell 27, 962-975. O'Neill, E., Rushworth, L., Baccarini, M., and Kolch, W. (2004). Role of the kinase MST2 in suppression of apoptosis by the proto-oncogene product Raf-1. Science 306, 2267-2270.

Page 35

P02: Pathway-GPS and SIGORA: Identifying relevant pathways based on the over-representation of their gene-pair signatures David Lynn1, Amir Foroushani1, Fiona Brinkman2

1Teagasc, Ireland 2Simon Fraser University, Canada Predominant pathway analysis approaches treat pathways as collections of individual genes and consider all pathway members as equally informative. As a result, at times spurious and misleading pathways are inappropriately identified as statistically significant, solely due to components that they share with the more relevant pathways. We introduce the concept of Pathway Gene-Pair Signatures (Pathway-GPS) as pairs of genes that, as a combination, are specific to a single pathway. We devised and implemented a novel approach to pathway analysis, Signature Over-representation Analysis (SIGORA), which focuses on the statistically significant enrichment of Pathway-GPS in a user-specified gene list of interest. In a comparative evaluation of several published datasets, SIGORA outperformed traditional methods by delivering biologically more plausible and relevant results. Availability: An efficient implementation of SIGORA, as an R package with precompiled GPS data for human and mouse pathway annotations can be downloaded from CRAN. (http://cran.rproject.org/web/packages/sigora/).

P03: Identification of molecular mechanisms underlying MCF-7 cell differentiation using proteomics and systems biology approaches Natalia Volinsky1, Alex VonKriegsheim1, Christian Preisinger2, Tapesh Santra1, Nina Saban3, Marc Birtwistle1, Albert Heck2, Walter Kolch1, Boris Kholodenko1

1Systems Biology Ireland, University College Dublin, Ireland 2Utrecht University, Netherlands 3Rudjer Boskovic Institute, Croatia In MCF-7, a breast cancer cell line, ErbB receptor ligands induce different cell phenotypes. Epidermal Growth Factor (EGF) induces cell proliferation, whereas Heregulin (HRG) induces an irreversible differentiation process accompanied by lipid droplets formation within the cells. So far, no signalling or transcriptional events regulating MCF-7 differentiation have been identified. Using a set of specific chemical inhibitors we identified PI3 kinase activation as a key event in MCF-7 differentiation. Unlike some other well studied differentiation models, no Erk 1/2 activation is required for successful MCF-7 differentiation. Supporting this finding, Insulin, a strong activator of PI3 kinase but only a weak activator of MAPK pathway, was demonstrated to induce lipid accumulation in MCF-7 cells. Additionally, knocking down PTEN, a negative regulator of the PI3 kinase signalling pathway was shown to be sufficient to induce MCF-7 differentiation. Based on this initial data, we decided to identify further signalling events involved in the MCF-7 differentiation process. We employed a quantitative phosphoproteomics approach, stable isotope labeling with amino acids in cell culture (SILAC) followed by TiO2 enrichment, to identify and quantify differentiation specific changes in the proteome and phosphoproteome. The phosphoproteomics data acquired was further analyzed by bioinformatical tools and lead us to validate the involvement of two disparate signalling pathways. Pathway mapping lead us to investigate mTOR signalling and we could confirm that mTOR signalling is required for cell differentiation. Further, analysis of consensus sequence of deferentially phosphorylated peptides, predicted the involvement of several other kinases, including CK2 (former Casein Kinase II), in cell differentiation. This prediction was validated by using both, a CK2 selective chemical inhibitor and by a CK2-targeted siRNA

Page 36

knock-down. In both cases, CK2 inhibition resulted in increased lipid accumulation in differentiating MCF-7 cells. Moreover, CK2 inhibition promoted differentiation of cells grown under basal conditions. These data demonstrates that CK2 acts as a negative regulator of lipid accumulation during MCF-7 cell differentiation. Here we demonstrate that bioinformatical analysis of phosphoproteomics data is a useful tool for finding non-intuitive candidate kinases involved in biological processes.

P04: Guanine Nucleotide Inhibitor binding modifies the temporal dynamics of GTPases Elena Nikonova1, Dirk Fey1, Mikhail Tsyganov2, Boris Kholodenko1 1Systems Biology Ireland, University College Dublin, Ireland 2Institute of Theoretical and Experimental Biophysics, Pushchino, Russia Monomeric G proteins control intracellular signaling by cycling between inactive GDP-bound and active GTP-bound states. The reaction is catalyzed by guanine nucleotide exchange factors (GEFs) and by GTPase activating proteins (GAPs) respectively. Guanine nucleotide dissociation inhibitors (GDIs) also regulate GTPase cycling by binding to the active and inactive sites and sequestering the formed complex from the membrane to the cytosol, thus inhibiting GTPase transformation. This paper provides a first GTPase model with GDI binding and explores its dynamics by regulating GDI binding parameters. In particular, our results show that by varying GTPase-GDI affinities the model can exhibit transitions between dynamical behaviors.

P05: Regulated Trafficking of APP by SORLA in Alzheimer’s Disease Angelyn Lao1, Vanessa Schmidt2, Yvonne Schmitz1, Thomas E. Willnow2, Olaf Wolkenhauer1 1University of Rostock, Germany 2Max-Delbrueck-Center for Molecular Medicine, Berlin-Buch, Germany Proteolytic breakdown of the amyloid precursor protein (APP) by secretases is a complex cellular process that results in formation of neurotoxic Aβ peptides, causative of neurodegeneration in Alzheimer’s disease (AD). Processing involves monomeric and dimeric forms of APP that traffic through distinct cellular compartments where the various secretases reside. Amyloidogenic processing is also influenced by modifiers such as sorting receptor-related protein (SORLA), an inhibitor of APP breakdown and major AD risk factor. This study aims to (i) model the neuronal factors central to the proteolytic processing of amyloid precursor protein (APP), (ii) trace the trafficking of APP in various compartments, and (iii) evaluate the influence of the SORLA on those factors. Using experimental data and literature-based, information we developed a multi-compartment model to simulate the complexity of APP processing in neurons, and to accurately describe the effects of SORLA on these processes. Our model enables regulation of trafficking of APP by SORLA through intracellular compartments. We have successfully confirmed our hypothesis that blockade of APP dimerization is an important aspect of SORLA action on AD. Using this model, we are able to uncover that SORLA not only affects amyloidogenic processing through interaction with APP but also specifically targets β-secretase - the enzyme responsible for initial amyloidogenic cleavage. Our model represents a major conceptual advancement by identifying APP dimers and β-secretase as the two distinct targets of the inhibitory action of SORLA in AD. Reference: A. Lao, V. Schmidt, Y. Schmitz, T.E. Willnow, and O. Wolkenhauer. Multi-compartmental modeling of SORLA’s influence on amyloidogenic processing in Alzheimer’s disease. BMC Syst. Biol., 6(1):74, June 2012.

Page 37

P06: Protein-Protein interaction in living cell using fluorescence cross-correlation spectroscopy Sivaramakrishnan Ramadurai1, Chiara Saladino2, Jens Rauch1, Walter Kolch1, Heinz peter Nasheuer1, 2 1National University of Ireland Galway, Ireland 2Systems Biology Ireland, University College Dublin, Ireland Quantifying interactions and dynamics of biomolecules are crucial for understanding important biomolecular interactions in living systems. System biology attempts to model the dynamics and structure of complete biological systems. The insight gained from a system biology approach can then be used to design in vivo and in vitro experiments, and in turn further develop models in an ever more refined description of physical and biological interactions. Recent developments in fluorescent-based imaging assays will allow us to study the molecular interaction in a biological pathway at single molecule resolution. Fluorescence microscopy and spectroscopy are often used in systems biology to study temporal and the spatial distributions of biological processes throughout an intact living cell. Fluorescence correlation spectroscopy [FCS] is a fluorescent-based single molecule technique which measures fluorescent fluctuations in a tiny observation volume, defined by confocal optics. Correlating fluctuations signals provides quantitative data on both mobility and number of molecules in the observation volume. Fluorescence cross correlation spectroscopy [FCCS] is an extension of the FCS priniciple and measures interactions or dynamics of co-localization of two molecules, by saptio-temporal coincidence of their fluorescence signals. When distinctly labelled molecules interact to each other, they synchronously diffuse through the observation volume and induce a cross-correlation signal, the amplitude of which gives information on the level of interaction or dynamic co-localization. FCCS, quantitative imaging technique, will be used to study the interaction of proteins in the Mitogen-activated protein kinase pathway, a key pathway involved in cancer development. This approach can potentially accelerate the achievement of a systems level understanding of biological complexity. The integration of FCCS and modelling that can enable new advances in oncology and other fields in the biomedical sciences.

P07: Crosstalk between EGFR and Integrin signalling in organised 3D structures of MCF10A mammary cells through JNK Benedikt Minke1, Drieke Vandamme1, Sara McNally2, Boris Kholodenko1, Finian Martin2 1Systems Biology Ireland, University College Dublin, Ireland, 2Conway Institute, University College Dublin, Ireland MCF-10A cells are an immortalized but non transformed human breast epithelial cell line. Grown in a laminin-rich extracellular matrix (ECM), in the presence of EGF, glucocorticoids and insulin, MCF-10A cells differentiate and form 3-dimensional acinar structures. This complex 3-dimensional morphogenesis has been shown to involve a range of signalling pathways and occurs through different stages. Each acinus arises from cell polarisation upon contact with the ECM, followed by formation of cell-cell junctions, exit of the cell cycle and clearing of the luminal cells. Two of the major pathways involved are the signalling downstream of EGFR (through Erk activity) and integrin signalling. Previous work also suggests that glucocorticoid signalling regulates acinar integrity through JNK. We have shown that treatment with the JNK inhibitor SP600125 results in reduced SRC phosphorylation, as well as increased ERK phosphorylation. This suggests a link between these members of the network and possibly links integrin and EGFR signalling pathways through JNK. Our work is currently focusing on testing that hypothesis through identifying players involved in this crosstalk with proteomics based experiments.

Page 38

P08: Charge profile analysis reveals that activation of pro-apoptotic regulators Bax and Bak relies on electrostatically driven allostery Heinrich Huber1, Crina Ionescu2, Radka Svobodova-Varekova2, Jochen Prehn1, Jaroslav Koca2 1Royal College of Surgeons in Ireland, 2Masaryk University BRNO, Czech Republic Mitochondrial outer membrane permeabilization is a hallmark of programmed cell death and executed by Bcl-2 family proteins Bak and Bax. However, the structural aspects of the activation mechanisms are not fully understood. We provide here a method for estimating partial atomic charges, to the study of structural alterations and their functional consequences during Bax activation. Using this method, we investigated the changes in the Bax charge profile upon activation by analyzing previously reported structural data for inactive Bax, and for Bax in complex with a functional peptide of its natural activator protein, Bim. We found that charge reorganizations upon activator binding induced an Arg94-mediated abrogation of the Ser184-Asp98 interaction. This results in a decreased affinity of the C-domain for its hydrophobic groove, which is responsible for Bax activation. We further identified a network of charge reorganizations that confirms previous speculations of allosteric sensing, whereby the activation information is conveyed from the activation site, through the hydrophobic core of Bax, to the C-domain binding groove. This charge transfer network was found to be mediated by a central hub of three residues on helix 5 of the hydrophobic core of Bax. Sequence and structural alignment revealed that this hub was conserved in Bak amino acid sequence, and in the 3D structure of folded Bak.. Our method provides the an easy analysis of intra-molecular charge redestribution, therereby putting forward the notion of intra-molecule signalling pathways that allow a swift reaction to diverse molecular stimuli.

P09: Modelling MYCN-amplified neuroblastoma chemoresistance Axel Kuehn1, Dirk Fey1, Boris Kholodenko1 1Systems Biology Ireland, University College Dublin, Ireland Neuroblastoma is the most common extra cranial solid tumour in childhood and the most common cancer in infancy. It is an extremely heterogeneous disease stratified in low-, intermediate- and high-risk tumours. Whereas low-risk tumours often undergo spontaneous regression, high-risk tumours can be very aggressive despite multi-modal treatments. The MYCN gene encodes for MYCN transcription factor and its amplification is commonly associated with high-risk tumours although it has been shown that MYCN gene amplification also correlates with apoptosis sensitization. HMGA1 is one of the MYCN target genes and is involved in regulating apoptosis in response to DNA damage in a pathway involving ataxia-telangiectasia-mutated (ATM), HIPK2 and p53. On one hand, HMGA1 increases the DNA damage response by inducing ATM gene expression in a positive feedback loop. On the other hand HMGA1 prevents the activation of p53 by binding HIPK2 and trans-locating it to the cytosol. Here we propose a dynamic model in which MYCN protein regulates this DNA damage system with implications for the chemoresistance in MYCN-amplified neuroblastoma. In addition to the intricate regulation of the DNA damage response outlined above, the analysis of this model points towards an important role MYCN-independent HMGA1 expression and subsequent HIPK2 nuclear/cytoplasmic localization.

P10: Mathematical Investigation of c-FOS Synergistic Expression Induced by Prolactin and EGF Andrea Degasperi1, Edita Aksamitiene2, Aimee Carmody1, Alex Cheong1, Till Frank1, Anatoly Kiyatkin2, Natalia Volinsky1, Walter Kolch1, Boris Kholodenko1 1Systems Biology Ireland, University College Dublin, Ireland 2Thomas Jefferson University, Philadelphia, USA

Page 39

Transcription factor c-FOS plays a central role in cell response to external cues. Its gene activation follows shortly growth factor stimulation and its regulation can lead to either cell proliferation or differentiation. We study the integration of prolactin and growth factor stimulations which produce a synergistic (more than additive) expression of c-FOS in T47D breast cancer cells. Western blot and qRT-PCR time series data are used to train and validate a mathematical model of the signalling and the promoter activity. We show how this mathematical model helps elucidating the dynamics of complex non-linear interactions, suggesting candidate mechanisms that explain the observed synergistic c-FOS expression.

P11: Systems Medicine Approaches to Chronic Fatigue Syndrome/ME Graham Smith, Thomas Kirkwood, Julia Newton Newcastle University, UK CFS (Chronic Fatigue Syndrome), also known as ME (Myalgic Encephalopathy/Encephalomyelitis), is a debilitating condition with a prevalence of about 0.2% in the UK, characterised by severe long-lasting fatigue and usually other symptoms such as unrefreshing sleep, muscle pain, cognitive impairments and post-exertion malaise. Despite extensive efforts, a cause has not yet been discovered. The absence of objective diagnostic criteria and heterogeneity of patients' symptoms leads to difficulty of definition and risk of misdiagnosis. Immune and neuroendocrine system abnormalities are frequently seen in CFS/ME and metabolic/bioenergetic defects may also be present. Predisposing factors such as stress and a precipitating factor such as a viral infection, probably interacting with the genotype of a susceptible host, are commonly thought to be implicated. This complexity of CFS/ME underlines the potential that a systems approach could have for unravelling the apparently disparate features of the condition, and for resolving the difficulties of understanding the possible causal relationships between them. Following such an approach, we present analysis of detailed cardiovascular measurements made on a cohort of 195 CFS patients and 68 normal controls when resting and under conditions of mild stress by orthostasis (standing) and the Valsalva manoeuvre (blowing against a closed airway). We describe the extent to which multivariate statistical techniques can be used to discriminate CFS cases from normal controls and also from patients with other fatigue-associated conditions. One frequently described abnormality in CFS patients is abnormality of the autonomic and cardiovascular system. We also show preliminary results for the modelling of cardiovascular stresses using a mechanistic model of the circulatory system [1], which may be used to relate the cardiovascular abnormalities seen in the majority of CFS patients to alterations in the CV system such as changes in autonomic feedbacks, the permeability of compartments and the elasticity of peripheral vessels. We conclude that a systems medicine approach, drawing on recent successes in systems biology, is likely to prove essential for understanding the underlying pathogenesis of CFS. Such an approach will allow us to explore the likely effects of potential interventions targeting particular dysfunctional organ systems and to assess the value of multiple combinations of therapies. Heldt, T., Shim, E. B., Kamm, R. D. and Mark, R. G. (2002) Computational modeling of cardiovascular response to orthostatic stress. J Appl Physiol. 92, 1239-1254

Page 40

P12: A dynamic model of MAPK crosstalk and signalling switches Dirk Fey, David R Croucher, Walter Kolch, Boris Kholodenko Systems Biology Ireland, University College Dublin, Ireland Mitogen-activated protein kinase (MAPK) cascades integrate and process intra- and extracellular signals, thereby regulating pivotal cell fate decisions, such as proliferation, differentiation and apoptosis. However, depending on cell type, signal strength and dynamics, similar MAPK kinetic profiles can be associated with opposing cellular decisions. This implies that signalling by each individual MAPK cascade has to be considered in the context of the entire MAPK network. Here, we develop a dynamic model of feedback and crosstalk for the three major MAPK cascades; extracellular signal-regulated kinase (ERK), p38 mitogen-activated protein kinase (p38), c-Jun N-terminal kinase (JNK), and also include input from protein kinase B (AKT). The model incorporates mechanistic details of positive feedback from JNK to its own MAP3Ks and negative crosstalk from and to other pathways. In the proposed model, JNK can switch from a transient to sustained activity due to multiple positive feedback loops. Once activated, positive feedback locks JNK into a highly active state that promotes cell death. The switch is differentially regulated by the ERK, p38 and AKT pathways. AKT activation decreases the value of the JNK-on state by inhibiting the JNK positive feedback, thus abrogating the apoptotic switch and allowing only proliferative signalling. In contrast, ERK activation shifts the threshold of the apoptotic switch to higher inputs by enhancing the dual specificity phosphatase (DUSP) mediated dephosphorylation of JNK. In non-transformed cells, activation of p38 restores the threshold by inhibiting ERK activity via the PP1 or PP2A phosphatases. Transformed cells lack the p38-ERK crosstalk, which increases the apoptotic threshold to unphysiological levels, thus rendering transformed cells biologically immortal. Further, our model predicts a critical role for DUSP1 and DUSP2 expression patterns, as both expression of DUSP1 and deletion of DUSP2 are necessary for preventing the JNK apoptotic switch (the nominal model is robust to dysregulation of either DUSP in isolation). The result is particularly interesting in the context of a) cancer, as many cancers show increased expression of DUSP1 and reduced expression of DUSP2, and b) tumour related conditions such as hypoxia, where low oxygen levels upregulate DUSP1 and downregulate DUSP2. In conclusion, our model facilitates understanding of how cancerous deregulations disturb MAPK signal processing and provides explanations for the complex and tumour specific behaviour of MAPK systems and certain drug resistances. Reference: Fey D,Croucher DR, Kolch W and Kholodenko BN. Crosstalk and signalling switches in mitogen-activated protein kinase cascades. Frontiers in Systems Physiology. Paper pending published: 15 Jul 2012.http://www.frontiersin.org/Systems_Physiology/abstract/31406.

P13: Multi-scale, disease response modeling of influenza infections Jason Shoemaker1, Satoshi Fukuyama1, Yoshihiro Kawaoka2, Hirako Kitano3

1Japanese Science and Technology Agency, 2Influenza Research Institute, University of Wisconsin, Madison, 3The Systems Biology Institute For virus infections, several drugs have been developed to successfully target aspects of the virus itself, but the role of the host, another key factor behind a virus’ potency has - for the most part - been ignored. There are two options to manipulating the host with the intention of minimizing the impact of an infection. Some suggest targeting host’s proteins essential to the virus replication cycle while others support the development of immunomodulatory therapies which seek to mitigate inflammation resulting from an aggressive immune response. We focus on the second approach, developing immunomodulatory therapies, which is ultimately an optimization problem which seeks to maximize virus clearance while minimizing inflammation. A major challenge for this goal is the development of

Page 41

accurate mathematical representations of the immune response, capable of integrating critical intracellular signalling events with the changes and activity of key lymphoid cells. Here, we developed a collection of cell signature enrichment and gene expression clustering tools which allow us to associate changes in gene expression with the influx of immune related cell types and/or intracellular signalling. Applying these tools to dynamic gene expression data developed from the lungs of mice infected with several influenza viruses, we found distinct signatures of several lymphocytes (macrophages, B cells, T cells and natural killer cells). The cell signatures were verified using flow cytometry. Fitting the expression data to a collection of first order and Hill equations, we show that that the immune response was a highly conserved and predictable mechanism that was surprisingly independent of the survivability of the infection. Interestingly, we found a lack of any feedback mechanisms to protect the host from severe inflammation during infection. We also provide evidence that to prevent accidental activation, the extracellular, virus-detection receptor pathways employ ultrasensitive-like mechanisms; i.e., below a critical threshold concentration of virus, these pathways are inactive but activate quickly once the threshold is exceeded. While this protects the cell from accidental stimulation, it ultimately promotes virus growth by delaying the immune response. These finding has several implications in the design next generation therapeutics and may explain why some people experience asymptomatic infections while others suffer more severe symptoms. In the future, we will apply this technique to additional pathogens and evaluate how the host response differs between the infections and how to optimally intervene in the host response so as to minimize collateral inflammation without disturbing an efficacious immune response.

P14: A dynamic Renin-Angiotensin-Aldosterone System model integrated into a physiological whole-body pharmacokinetic model framework to simulate the effects of ACE inhibitors like enalapril Karina Claassen1, Stefan Willmann2, Thomas Eissing2, Tobias Preusser1, Michael Block2

1Jacobs University Bremen, School of Engineering and Science, Germany, 2Bayer Technology Services GmbH, Technology Development, Enabling Technologies, Germany Objectives: The renin-angiotensin-aldosterone system (RAAS) plays an important role in physiology and the pathogenesis of cardiovascular disorders. It operates at the junction of blood pressure, blood sodium and extracellular volume homeostasis. The RAAS is a widely used drug target for a number of antihypertensive drugs. Introducing the RAAS into a physiologically-based pharmacokinetic (PBPK) model allows for the inclusion of a priori information. PBPK models allow a mechanistic description of the distribution of substances within a WB context, and further, the consideration of their pharmacodynamic interaction at the relevant sites within the body [1]. PBPK models can enhance our understanding of the underlying biology and allow for a quick adaptation to different problems, making these models generally applicable. Thus, a whole body physiologically-based pharmacokinetic (WB-PBPK) model integrating this endocrinological cascade and the hormonal effects of inhibition is highly desirable. This work is aimed at developing a WB-PBPK and hormonal pharmacodynamic (PD) model including a dynamic RAAS. Methods: The PBPK model was based on data for physiological factors determining the RAAS and the circulation. It was established by using the software PK-Sim® and MoBi® [2]. First, a PBPK model was developed and physiological data of the RAAS [3-6] have been included. This results in a coupled WB-PBPK model containing the temporal evolution of the main RAAS actors renin, angiotensinogen, angiotensin 1 and 2,

Page 42

angiotensin converting enzyme (ACE), angiotensin 2 receptor type 1 and aldosterone. The real life case study of the ACE inhibitor enalapril is presented to exemplify the influences of xenobiotics on hormonal levels involved in the regulation of blood pressure. The parent drug enalapril is converted to the active metabolite enalaprilat in the liver. A coupled dynamic parent-metabolite PBPK model was developed. Second, a mechanistic representation of the RAAS was built based on available data and then integrated into a WB-PBPK/PD model. Results: All physiological RAAS hormone profiles are consistently described at steady state and during pharmacological intervention with an ACE inhibitor in healthy male volunteers. The model simultaneously represents the exemplary drug enalapril and its conversion to the active form enalaprilat, even as it represents the effect of the drug on the hormone levels, offering a detailed mechanistic insight into the hormone cascade and its inhibition. Conclusions: A WB-PBPK/PD model of the RAAS is developed. It is able to represent the pharmacokinetics of a drug interacting with the RAAS and the hormone system well. This indicates a sufficient description of the underlying physiology. The model constitutes a first major step towards establishing a PBPK-PD-model for a wide-range of drugs acting on blood pressure. It will be expanded by developing a blood pressure model and coupling of both models, which will lead to a mechanistic representation of the PBPK-PD relationship in cardiovascular diseases. The WB-PBPK/PD model of the RAAS helps to improve the understanding of interactions between drugs and the target system, thus facilitating dose and dosing regimen decisions in the area of cardiovascular diseases. References: [1] Kuepfer L, Lippert J, Eissing T: Multiscale mechanistic modeling in pharmaceutical research and development. Adv Exp Med Biol. 2012;736:543-61. [2] Eissing T, Kuepfer L, Becker C, et al.: A computational systems biology software platform for multiscale modeling and simulation: integrating whole-body physiology, disease biology, and molecular reaction networks. Front Physiol. 2011;2:4. [3] Juillerat L, Nussberger J, Menard J, et al.: Determinants of angiotensin II generation during converting enzyme inhibition. Hypertension. 1990 Nov;16(5):564-72. [4] Nussberger J, Wuerzner G, Jensen C, Brunner HR: Angiotensin II suppression in humans by the orally active renin inhibitor Aliskiren (SPP100): comparison with enalapril. Hypertension. 2002 Jan;39(1):E1-8. [5] Keshava Prasad TS, Goel R, Kandasamy K, et al.: Human Protein Reference Database--2009 update. Nucleic Acids Res. 2009;37(Database issue):6. [6] Wishart DS, Knox C, Guo AC, et al.: HMDB: a knowledgebase for the human metabolome. Nucleic Acids Res. 2009;37(Database issue):25.

P15: Unravelling signaling crosstalk at the promoter of the c-fos proto-oncogene Aimee Carmody, Alex Cheong, Till Frank, Walter Kolch, Boris Kholodenko Systems Biology Ireland, University College Dublin, Ireland The elucidation of regulatory mechanisms controlling the expression of proto-oncogenes is crucial in order to understand the transformation of normal cells into cancer cells. One such proto-oncogene, c-fos, is a well characterised immediate early gene that is induced by many different growth factors and other extracellular stimuli. However, relatively little is known about the crosstalk between transcription factors that regulate the c-fos promoter and produce transcriptional responses to various stimuli. We

Page 43

focus here on unravelling the crosstalk between the Elk-1, CREB and STAT-3 transcription factors in the regulation of the c-fos promoter in MCF-7 breast cancer cells. A fragment of the c-fos gene promoter was cloned into a Gaussia princeps luciferase reporter system, and the binding sites for Elk-1, CREB and STAT-3 were mutated to generate a series of constructs where these binding sites are disabled individually or in different combinations. We also used different stimuli (EGF, heregulin and prolactin) alone and in combination to activate distinct pathways that induce transcription of the c-Fos promoter. The multivariate temporal luciferase data was used to construct a thermostatistical model of the transcriptional crosstalk. Our data indicate that specific stimuli activate a complex network of pathways leading to promoter activity. Analysis of the network using thermostatistical modelling indicates a level of co-operation of transcription factors in generating specific transcriptional responses. We conclude that expression of c-fos can be differentially regulated by transcription factor crosstalk at the level of the gene promoter.

P16: T-MedFusion: a biomedical data fusion platform Imad Abugessaisa, Jesper Tegnér Karolinska Institutet, Stockholm, Sweden Data access and integration are two major bottlenecks in system biology, bioinformatics and biomedical research (Garny et al, 2010). Data fusion system can provide efficient solution and enable effective research that aim at bridging bench to the bedside.We present T-MedFusion, a data fusion system based on a JDL data fusion model (Steinberg and white, 1999). Our aims are to provide the desirable capabilities and functionalities of biomedical data fusion system. T-MedFusion is capable to fuse different sources of biomedical data (intermediate phenotype, clinical phenotype, life style, biobank, genomics, proteomics etc...) for different therapeutic areas.We implement the system in chronic inflammatory and cardiovascular diseases and we present the case of Rheumatoid arthritis (RA).Basic functionalities provided by data fusion system are data acquisition and management of patient cohort, workflow management ; identify patients according to specific biological, clinical, or statistical features. Data fusion aim at combining and integrate data from multiple sensors and databases to allow end-users to perform operations and generate inferences that is not possible from a single database source. The Joint directorate of laboratories (JDL) developed the first data fusion model that categorizes data fusion complements, functions, and interfaces. The current version of JDL consists of five process levels. We applied each of the levels to desired system behavior to all the process of data source fusion. The central object in our case study is the patient (i.e. his clinical phenotype, biological sample, genomics, and proteomics). During this research we dealt with technical challenges related to semantics and definition across the sources, lacking of integrity constraints (IC), accuracy and consistency among the sources, and information flows and the lack of a meta-model for the different sources. To solve the above challenges we revised and applied the well-known JDL data fusion model. Requirement-engineering methods were used to determine users’ requirements. To overcome semantic and definitional problems, we developed a meta-model, which consists of a common vocabulary. The model defines all classes, attributes, and the relationships that occur between different sources. T-MedFusion based on JDL data fusion model, provides data fusion functionalities for multiple data sources at the Center for Molecular Medicine (CMM), Karolinska Institute (KI), Sweden (research center) in one hand and the clinical data at the Karolinska University Hospital (KUS) in the other hand. The data sources at are, Biobank, cell registry, genomics, serology and life-style for cohort of patients and healthy

Page 44

control. At the clinic data about disease duration, treatment, disease activity, laboratory results and specification of the disease are stored in the electronic medical record or clinical registry. A data fusion system that integrate all sources of data are the key to make research even more translational, also it will power current research for finding predictive markers such as immunological phenotypes and genetic markers (single nucleotide polymorphism (SNPs)). Our first implementation of T-MedFusion with our collaborators at the Rheumatology research laboratory, previous RA studies have been performed on donated patient material from Karolinska hospital. Patients donate several types of samples, primarily blood and synovial fluid, from which e.g. serum, DNA and viable cells are retrieved and deposited in the Biobank. The inclusion criteria for sample for a specific study is typically based on serology and the availability of cells, while clinical information with regard to the samples have only been collected retrospectively when needed. To enable cellular and immunological research for each set of experiments researcher (end-user) have to select samples based on different clinical and genetic parameters. They should be able to select research material based on the availability of the cell samples from the biobank, suitable genotype (SNPs) or serology status and on disease parameters from the medical record. The following examples show simplified workflows and demonstrate the need of data fusion capabilities: 1- Find how many patients are available with a certain HLA-type & antibody pattern & disease severity from which cryopreserved cells, serum, and DNA sample are donated to the biobank. This kind of workflow will allow the researcher to select sub-cohort from the entire research cohort to run specific study and answer research question. 2- Another workflow will help to answer a research questions like, are there clinical differences (such as duration) between Immunglobulin G (IgG) single positive and Immunoglobulin A(IgG) /(IgG) double positive anti-citrullinated protein antibodies (ACPA) patients? Currently T-MedFusion implemented to manage and fuse clinical and research databases for the following research areas I-Psoriasis and Rheumatoid Arthritis. II-Congenital heart block (CHB) III-VEBIOS Venous thrombus Embolism- atherosclerosis We considered user interaction and usability of the system and provide an easy to use and intuitive interface for the biomedical researcher & clinician. The system provides gateway to different data sources and support “complex” search & workflow capabilities. T-MedFusion enables cohort discovery through data visualization techniques. We considered the regulation w.r.t. patients (Personal health information) hence we applied security techniques de-identification method. In conclusion, there are huge research efforts within the system biology community in interpreting and correlating molecular data with selected parts of a phenotype profile of different diseases. It is difficult to conceive how to tackle this grand challenge without resorting to computational infrastructure to manage and fuse different sources of data, which in turn by their nature should be open to different types of omics data and methods for integrating data at different spatial and temporal scales up to the level of clinical phenotypes. In our view, the implementation of JDL data fusion model will have an impact as a vehicle to bridge between the wealth of molecular information and numerous clinical phenotypes available through EMR and patient’s life style data collected via questionnaire.

Page 45

References: C. L. B. A.N. Steinberg., F.E. White., "Revisions to the JDL data fusion model," Proceedings of SPIE, Sensor Fusion: Architectures, Algorithms, and Applications vol. 3719, pp. 430-441., 1999. A. Garny, J. Cooper, and P. J. Hunter, "Toward a VPH/Physiome ToolKit," Wiley interdisciplinary reviews. Systems biology and medicine, vol. 2, pp. 134-47, Mar-Apr 2010 P17: Iterative mathematical and experimental analysis captures the dynamic regulation of HIF transcriptional activity by hydroxylases

Alex Cheong1, Lan Nguyen1, Miguel Cavadas1, Carsten Scholtz1, Susan Fitzpatrick1, Ulrike Brunin2, Murtaza Tamuwala2, Mario Manresa2, Boris Kholodenko1, Cormac Taylor1

1Systems Biology Ireland, University College Dublin, Ireland 2Conway Institute, University College Dublin, Ireland Activation of the hypoxia-inducible factor (HIF) pathway is a critical step in the transcriptional response to hypoxia. While many of the key proteins involved have been characterised, the dynamics of their interactions in generating this response remain unclear. We have generated a comprehensive mathematical model of the HIF pathway based on core validated components and dynamic experimental data, and confirm the described connections within the predicted network topology. Our model confirms that the steps leading to optimal HIF transcriptional activity require sequential inhibition of both prolyl- and asparaginyl-hydroxylases. We also predict and show that there is residual activity of asparaginyl-hydroxylase FIH at low oxygen tension. Furthermore silencing FIH results in increased HIF transcriptional activity but decreased stability. Using a core module of the HIF network and mathematical proof supported by experimental data, we propose that asparaginyl hydroxylation confers resistance to proteosomal and lysosomal degradation. Thus, through in vitro eAxperimental data and in silico predictions, we provide a comprehensive model of the dynamic regulation of HIF transcriptional activity by hydroxylases and use its predictive and adaptive properties to explain counter-intuitive biological observations.

P18: Future Challenges for Systems Medicine Marc Goodfellow, Yusur Al-Nuaimi, Ben Small University of Manchester, UK The use of Systems Biology promises to significantly benefit our understanding of health and disease. It is therefore vital to advance Systems Medicine research and overcome any barriers to its success. The aim of the “Future Challenges for Systems Medicine” initiative (www.fcsysmed.com) is to identify these barriers and bring together early career scientists and clinicians across disciplines. In June 2012 we launched this initiative by hosting a Young Life Scientists’ Symposium, sponsored by the University of Manchester and a triumvirate of U.K based societies. Here we report on the outcomes of this event, data from participants regarding Systems Medicine as a discipline and plans for our wider initiative. We successfully attracted 65 registrants from different academic, industrial and clinical backgrounds. Our international and local keynote speakers covered diverse aspects of Systems Medicine, highlighting issues in i) translating basic understanding through animal and computer models of disease, ii) conceptualisation of disease in terms of mathematical models at different levels of abstraction and iii) working in interdisciplinary teams of experimental scientists, theoretical scientists and clinicians. This complemented several diverse talks and posters from early career researchers and clinicians. Audience participation for talks was higher than many other conferences we had attended and so the main

Page 46

objective of interactivity was certainly met. Discussion reached its peak during the “Question time” debate, which could have given rise to many more hours of discussion. Further driven by excellent feedback on the event, we hope to continue to engage a large pool of interested parties through future meetings and community resources, including expanding our internet presence (www.fcsysmed.com).

P19: Controlling bipartite networks with power-law degree distributions using dominating sets Jose Nacher1, Tatsuya Akutsu2

1Department of Information Science, Toho University, Japan, 2Bioinformatics Center, Kyoto University, Japan The existence of non-random network structures in disparate fields from living cells to technological and social systems has motivated researchers to investigate complex networks. But once we have characterized most of the available real-world networks, the natural step is to learn how to control and modify the intrinsic dynamics that govern them at will. There are many examples of controlling networks. Internet routers transmit an amount of bytes per second, transcriptional factors regulate expression level of genes as well as drugs bind and repress target proteins in specific human disorders. The question that naturally arises is whether the non-randomness observed in natural and technological systems may help to control these networks. Recent works have addressed the problem of controllability in complex networks [1, 2]. In particular, by using a maximum matching algorithm, they have shown that homogeneous (i.e., random networks) can be controlled easily because the number of nodes that need to be controlled externally is small [1]. Hence, it seems that scale-free structures observed in nature do not help to control complex networks. However, there are several ways to define controllability in complex networks. Nepusz and Vicsek have analyzed this problem from the angle of edge dynamics and their conclusion is that networks with scale-free degree distributions have better controllability properties [3,4]. In parallel, we have recently introduced the minimum dominating set (MDS) approach to control complex networks, which conceptually has similarities with the controlling link dynamics [5]. First, our theoretical findings suggest that scale-free networks with small scaling exponent values (gamma< 2), where high-degree nodes are present, require relatively few nodes to be controlled [5,6]. Conversely, networks with large degree exponent values (gamma > 2) or faster exponential decay, where hubs are weakly connected or almost absent, require more nodes to be fully controlled. Therefore, inside of scale-free networks class, the scaling exponent also determines a critical boundary with relevant controllability effects. In short, MDS targets high-degree nodes, whereas a maximum matching approach would avoid the high-degree nodes. However, many important real-world systems can be better represented by bipartite graphs instead of simple graphs. In biochemical and medical complex systems, metabolic pathways consist of chemical compounds and chemical reactions catalyzed by enzymes and drugs have the purpose of regulating target proteins in human disorders. Few drugs are known to have many targets. In social systems, research collaborations are formed by joint teams of scientists, where some of them control a high number of research projects. Facebook users tend to participate and lead conversations in a large number of topics. By following our concept of MDS applied to controllability of complex networks, we have theoretically analyzed the covering set (CS) size of bipartite networks [7]. Our results show how the size of the covering set changes with the topology of bipartite networks, where both top and bottom nodes follow a power-law with specific scaling exponents. We then performed computer

Page 47

simulations with large networks to compare with our theoretical analysis as well as analyzed some of the above mentioned real-world networks, with emphasis on the drug - target gene network. In this case, for example, we see that the covering set is around 20% of all the approved drugs with an average degree of 2.44. In contrast, the average degree of all the approved drugs connected to disease genes is 1.75. Though here we report on our recent findings on controlling bipartite networks [7], we approach to the problem by using the MDS. In simple and bipartite graphs, our analysis demonstrates that the MDS tends to target highly connected nodes, whereas the previous study [1] suggested that driver nodes tend to avoid high-degree nodes in simple graphs. Our approach, however, can be connected to Liu et al by assuming that every edge in a network is bi-directional and every node in MDS can control all of its outgoing links separately [5]. Then, the network is structurally controllable by selecting the nodes in MDS as the driver nodes. Though Muller and Schuppert [2] recently suggested that iPS cells can be controlled by a few driver nodes, they did not show general results or consider structural properties of networks. It is also worth mentioning that our approach has some conceptual resemblance with controlling edge dynamics [3]. However, we tackle the problem from MDS point of view point rather than using switchboard-dynamics and its associated mapping to the line graph [8]. References: [1] Liu Y-Y, Slotine J-J and Barabsi A-L (2011), Nature 473, 167-173. [2] Muller F-J and Schuppert A (2011) Nature 478, E4. [3] Nepusz T and Vicsek T (2012) Nature Physics 8, 568-573. [4] Slotine J-J and Liu Y-Y (2012) Nature Physics 8, 512. [5] Nacher JC and Akutsu T (2012) New Journal of Physics 14, 073003 (24 pages) [6] Nacher JC and Akutsu T (2012) To appear in Proc. International Conf. on Math. Modeling in Physical Sciences, Budapest, Hungary, Sept. 3-7, 2012. [7] Nacher JC, Akutsu T (2012) In preparation. [8] Nacher JC, Yamada T, Goto S, Kanehisa M and Akutsu T (2005) Physica A 349, 349-363.

P20: Opposing positive and negative feedback loops regulate JNK signalling dynamics upon activation by multiple stimuli David Croucher, Dirk Fey, Boris Kholodenko, Walter Kolch Systems Biology Ireland, University College Dublin, Ireland Within the tumour micro-environment cancer cells receive signals from a multitude of different sources. This includes exposure to extrinsic molecules that promote tumour growth and survival, such as growth factors and certain cytokines, but also signals related to the harsh tumour micro-environment which can cause oxidative, metabolic, osmotic and hypoxic stress. How these often opposing stimuli are processed by the cell into a coherent cell fate decision is largely unknown, although the c-Jun N-terminal Kinase (JNK) is likely to play a central role in this decision making process due to its adaptive role in promoting either apoptosis or cell survival and proliferation. The JNK signalling pathway can respond to a wide variety of these stimuli and accordingly displays a broad array of dynamic behaviours. Utilising the SH-SY5Y Neuroblastoma cell line that expresses all three isoforms of JNK, we have observed that growth factors induce transient JNK activity that responds in a linear manner to increased growth factor concentration. Conversely, cell stress induces sustained JNK activity that responds in an ultrasensitive, bistable manner to increasing stimuli.

Page 48

Analysis of the crosstalk and feedback structures underlying this fundamentally different network behaviour has revealed that the linear growth factor response is regulated by a negative feedback loop acting via Akt mediated MKK4 inhibition, whilst the ultrasensitive response to cell stress is produced by activation of a positive feedback loop. Herein we characterise the network architecture required to facilitate the stress induced JNK positive feedback loop and present a dynamic model of JNK activation which examines the interplay between these two opposing feedback loops upon exposure to multiple stimuli. These findings have many implications for both the understanding of how cells process information from multiple stimuli and how tumour cells respond to environmental and therapeutically induced stress.

P21: The Effect of Mesenchymal Stem Cells on Prostate Cancer Bone Metastasis: A Systems Biology Approach Sarah Ridge2, Sonia Prado-Lopez1, Georgina Shaw1, Rhodri Ceredig1, Frank Sullivan2, Frank Giles2, Sharon Glynn2

1Regenerative Medicine Institute, 2Prostate Cancer Institute, National University of Ireland Galway, Ireland The most frequent site of metastasis in prostate cancer (PC) is to the bone. Most deaths in PC occur following metastasis to the bone. However, the interactions that promote the metastasis remain largely unknown. Mesenchymal stem cells (MSCs) are multipotent cells that reside in the bone marrow. These stromal cells have the capacity to home to inflammatory sites and have previously been implicated in the advancement of tumour growth at the primary site. Here, we propose a role for MSCs in the promotion of prostate cancer metastasis to the bone. Using a systems biology approach, interactions between the metastatic PC cell line, PC3, and human MSCs from 3 male donors will be analysed. Genomic, proteomic and metabolomic profiles will be obtained following co-culture of PC3 with the donor MSCs. Data obtained from the proteins, metabolites, DNA, mRNA, miRNA, lncRNA will be combined and assessed using biostatistics and bioinformatics. This holistic approach could lead to the discovery of novel molecular pathways that would be predictive of increased potential for the patient to develop bone metastasis.

P22: Time resolved proteomic profiling of epithelial to mesenchymal transition (EMT) Alex von Kriegsheim1, Jennifer Farrell1, Phillip Cotter1, Jeremy Simpson2

1Systems Biology Ireland and 2Conway Institute, University College Dublin, Ireland EMT is a fundamental differentiation/dedifferentiation process by which epithelial cells are reprogrammed into motile mesenchymal cells. This process is crucial during embryonic development, such as allowing the formation of the neural crest. Additionally, evidence suggests that EMT is taking place as part of a pathological programme. Data supporting a malignant role is especially strong in cancer development. EMT is proposed to be the initiating step of metastasis formation, by enabling epithelial cells to break free from the cancer mass and to invade into the surrounding tissue. Given the importance of this transition it is unsurprising that EMT has been the focus of intensive research over the past years. Much of our current understanding of EMT has been shaped by monitoring signalling and mRNA expression changes in model cellular systems. Madin-Darby Canine Kidney Epithelial Cells (MDCK) represents one of these model systems. In vitro, MDCK cells form tightly packed, polarised cell clusters, which undergo EMT upon stimulation with

Page 49

Hepatocyte Growth Factor (HGF). The transition is fast and is completed within 24 hours. Additionally, the cells are easily transfected making this model attractive to high throughput analysis. Recent evidence has shown a poor correlation between mRNA and protein expression, thus we hypothesised that monitoring mRNA expression may miss essential components of the transition. We therefore decided to use global, quantitative mass spectrometry to reveal expression changes at the proteome level and to cross-correlate it to mRNA levels. We monitored protein changes at 5 time points over 24 hours and were able to quantify over 4000 proteins. Quantification of mRNA and protein expression revealed that a significant proportion of proteins showed little correlation. Proteins involved in cell death and the ubiquitination machinery were enriched in this fraction. In depth analysis of some candidates confirmed the protein expression changes and revealed that the pro-apoptotic MST2/Hippo pathway is inhibited during EMT. This inhibition occurs at different levels and is essential for EMT. Taken together, our data suggest that an integrative approach, combining quantitation of mRNA and protein expression is more informative and further, that for EMT to occur, suppression of pro-death pathways is required.

P23: Dynamics of DNA damage induced pathways to cancer Jean-Marc Schwartz, Kun Tian, Marija Krstic-Demonacos University of Manchester, UK Although more than 62,000 papers about p53 have been published since its discovery, we are still far from understanding the details of its role in the cellular response to DNA damage. A vast amount of biological knowledge on p53 is available in the form of qualitative information that must be converted into mathematical models. We constructed a large-scale logical model of the p53 interactome from database and literature information; this model contains 205 nodes representing genes or proteins, DNA damage input and apoptosis output, and 677 logical interactions. Predictions were generated and tested by performing in silico knock-outs, and a subset of the major changes was validated using literature searches and in vitro based experimental analyses. Finally, the comparison of model simulations with genome wide microarray data demonstrated a significant rate of successful predictions. In summary, we show that a systematic compilation of biological knowledge into dynamic logical models can provide predictive value and better understanding of complex p53 pathways induced by DNA damage and leading to cancer, with potential applications for medicine.

P24: Bistability in the Rac1, PAK and RhoA network governs cancer cell motility Kate Byrne, Lan Nguyen, Naser Monsefi, Alex Von Kriegsheim, Boris Kholodenko Systems Biology Ireland, University College Dublin, Ireland Cell migration plays a vital role in the invasion and metastasis stages of cancer, the latter being the most frequent cause of death in cancer patients. This cell migration is governed by the dynamic interaction between RhoA and Rac1, members of the RhoGTPase family. The current project aims to elucidate the biochemical mechanism underlying this interplay and subsequently, to manipulate cancer cell’s motility. We developed a mathematical model that describes the interaction of Rac1 and RhoA via the intermediary PAK using ordinary differential equations with Michaelis-Menten and mass action kinetics. The model was validated using experimental data generated in the MDA-MB-231 breast cancer cell line. Computational analysis of our mathematical model predicted that Rac1 and RhoA activity would react to a PAK inhibitor in a bistable manner, i.e., where there are two possible states of the system for a

Page 50

given input. We confirmed this experimentally by assaying changes of RhoA, Rac and PAK activity in response to PAK inhibition. Interestingly, cell migration was found to behave identically. Analysis of the model showed that bistability was found for a wide range of parameters surrounding the chosen set. From this we hypothesised that the stability of the system could be manipulated by forcing the system out of the bistable region by changing system parameters. RhoA and Rac are interlocked in a mutual negative feedback-loop via PAK. PAK inhibition locks the cells into a high RhoA state and arrests cell migration completely. Due to the bistable nature of the network, this arrest is maintained at low inhibitor concentrations. Our research indicates that PAK is a desirable drug target as migration is completely dependent on PAK activity. Further, once PAK is inhibited, it remains locked in this state despite a decrease in inhibitor concentration, allowing for reduced inhibitor dosages after an initial treatment.

P25: Under-appreciated Network Architectures that Generate Population Heterogeneity Maciej Dobrzynski1, Dirk Fey1, Lan Nguyen1, Marc Birtwistle2, Alex Von Kriegsheim1, Alex Cheong1, Walter Kolch1, Boris Kholodenko1

1Systems Biology Ireland, University College Dublin, Ireland 2Mount Sinai School of Medicine, New York, USA Phenotypic heterogeneity within a population of cells has been demonstrated to confer fitness advantage in systems ranging from bacteria treated with antibiotics to cancer cells subjected to chemotherapeutic drugs. Bimodal protein distributions are a prime indicator of such heterogeneity. We discuss two network architectures that can lead to bimodality. We measure heterogeneous response of HIF network and propose a probabilistic model as the first mechanism that gives rise to cell-to-cell variability in a system without feedbacks. The second mechanism, heterogeneous oscillations, is addressed analytically and numerically in a model of feedback-connected GTPases. Background: Protein levels across a cellular population, be it a bacterial colony or tumor cells, at any given point in time exhibit a distribution rather than a precise concentration value. The source of this variability lies in thermal noise -- an inevitable factor affecting all biochemical reactions. Of particular interest are bimodal protein distributions that indicate a temporal or steady-state phenotypic division of an isogenic cellular population. Bimodality reflects the existence of two subpopulations, each capable of performing a different task [1] or having an altered survival rate to stress [2] and drug treatment [3]. Bimodal distributions of protein activities in signaling systems are often interpreted as indicators of underlying stochastic switch-like responses [4] or bistable dynamics that stems from feedback regulation [5]. We investigate theoretically and experimentally the emergence of bimodality by analyzing two less appreciated scenarios: noisy networks with nonlinear dose response and heterogeneous oscillations. We address the first mechanism by performing flow cytometry measurements of hypoxia response network in stable HCT 116 cells transfected with oxygen-dependent domain tagged with GFP (ODD-GFP). Cells exhibit steep, sigmoidal response upon stimulation with prolyl-4-hydroxylase inhibitor (DMOG), which upregulates ODD-GFP -- an active domain of hypoxia-inducible factor (HIF). Consistently with our theoretical prediction, broad, bimodal-like distributions are observed. This heterogeneous response stems largely from noise in gene expression of HIF network components, which in turn, introduces differences in steepness and threshold levels of the input/output relationship between cells.

Page 51

The second of the mechanisms capable of generating bimodal protein distributions is an oscillating signaling system with varying amplitude, phase and frequency due to biochemical noise. We support our analysis by analytical derivations for basic oscillators and numerical simulations of a GTPase signaling cascade, which displays sustained oscillations in protein activities. Importantly, we show that the time to reach the bimodal distribution depends on the magnitude of cell-to-cell variability. We quantify this time using the Kullback-Leibler divergence. Conclusions: Given wide implementation of oscillatory networks and networks with sigmoidal dose response in eukaryotic systems, both mechanism may play an important role in generation of phenotypic variability. Our mathematical analysis demonstrates conditions under which broad or bimodal protein distributions may arise. Our study is a convincing argument that single-cell measurements are required to reveal the underlying dynamics of networks that contribute to randomization of cell phenotypes. Bulk measurements such as Western blots where proteins are detected in cell lysates, only estimate the average (per-cell) concentration of the entire population. Protein distributions can only be assessed as population snapshots in fluorescence-activated assays using flow cytometry or cell imaging. References: 1. Ackermann, M., Stecher, B., Freed, N.E., Songhet, P., Hardt, W.D., Doebeli, M.: Self-destructive cooperation mediated by phenotypic noise. Nature 454(7207), 987– 990 (Aug 2008) 2. Blake, W.J., Balazsi, G., Kohanski, M.A., Isaacs, F.J., Murphy, K.F., Kuang, Y., Cantor, C.R., Walt, D.R., Collins, J.J.: Phenotypic consequences of promoter- mediated transcriptional noise. Molecular Cell 24(6), 853–865 (Dec 2006) 3. Spencer, S.L., Gaudet, S., Albeck, J.G., Burke, J.M., Sorger, P.K.: Non-genetic origins of cell-to-cell variability in TRAIL-induced apoptosis. Nature 459(7245), 428–432 (May 2009) 4. Acar, M., Mettetal, J.T., van Oudenaarden, A.: Stochastic switching as a survival strategy in fluctuating environments. Nat Genet 40(4), 471–475 (Apr 2008) 5. Samoilov, M., Plyasunov, S., Arkin, A.P.: Stochastic amplification and signaling in enzymatic futile cycles through noise-induced bistability with oscillations. PNAS 102(7), 2310–2315 (Feb 2005)

P26: A systems approach to studying the signaling network regulating bone- marrow derived mesenchymal stem cell chemtoaxis Drieke Vandamme1, Laetitia Kurzawa1, Alex Von Kriegsheim1, Marc Birtwistle2, Boris Kholodenko1, Walter Kolch1

1Systems Biology Ireland, University College Dublin, Ireland 2Mount Sinai School of Medicine, New York, USA Bone marrow derived mesenchymal stem cells (MSCs) have generated much interest as a potential source of cells for cell-based therapeutic strategies. However, to date, the knowledge of the molecular background of MSC chemotaxis is rather poor. PDGF induces a concentration-dependent migration response in MSCs. Our goal is to reconstruct the spatiotemporal signalling network upon activation of the PDGF receptor in MSCs at a systems level. Over the past 10 years, a high number of genetically encoded FRET-based biosensors for real time monitoring of various biochemical activities in live cells has been published. Most of these probes share a general structure where a sensing unit, which is conformationally responsive to the biochemical activity of interest, is positioned in between a “blue-shifted” and a “red-shifted” fluorescent protein (FP), capable of FRET. Thus, changes in the biochemical activity of interest change the distance between the FPs, which leads to detectable changes in FRET. We have generated immortalized hMSC pools which

Page 52

stably express FRET/FLIM biosensors that measure different migration related signalling activities. These will be used in different imaging based chemotaxis assays.

P27: PHD1 and FIH regulate IL-1β-induced NF-κB activity linking key hypoxic and inflammatory signaling pathways Carsten Scholz1, Murtaza Tambuwala2, Emily Hams2, Alex Cheong1, Ulrike Bruning3, Padraic Fallon2, Eoin Cummins3, Cormac Taylor1, 3

1Systems Biology Ireland, University College Dublin, Ireland 2Institute of Molecular Medicine, Trinity College Dublin, Ireland, 3School of Medicine and Medical Science, Conway Institute, University College Dublin, Ireland Low oxygen concentration (hypoxia) is a feature of chronically inflamed tissues. Hydroxylases are oxygen-sensing enzymes, which control the transcriptional response to hypoxia and influence the course of inflammation through the regulation of HIF- and NF-κB-dependent pathways. Four different oxygen-sensing hydroxylases are known: Prolyl Hydroxylase 1, 2 and 3 (PHD1, 2, 3) and Factor Inhibiting HIF (FIH). Pharmacologic hydroxylase inhibition reduces inflammation in animal models of colitis, reperfusion injury and sepsis, however, the underlying mechanisms remain unclear. IL-1β is a major pro-inflammatory cytokine that activates NF-κB-dependent transcriptional pathways and is associated with numerous inflammatory pathologies. We investigated a role for hydroxylases in regulating IL-1β-dependent inflammatory signalling. We show that hydroxylase inhibition reduces IL-1β-induced NF-κB activity in a manner, which is dependent upon the combinatorial inhibition of PHD1 and FIH and results in repression of NF-κB-dependent genes. Hydroxylase inhibition reduced IL-1β-induced signalling upstream of JNK, p38 and IKK. Thus, combinatorial inhibition of PHD1 and FIH represents a new approach to the inhibition of inflammatory signalling pathways activated by IL-1β, which may be of benefit in inflammatory disorders.

P28: Maximum Likelihood Estimation for fitting Parameters of a Stochastic Model Aonghus Collins, Till Frank, Boris Kholodenko Systems Biology Ireland, University College Dublin, Ireland Likelihood based methods are an alternative to Linear Regression methods such as least squares that can be used to estimate the parameters of a model from experimental data. Due to their probabilistic nature, they are in some ways better suited to fitting stochastic models, as they can attempt to fit both drift and volatility directly from the stochastic data, rather than separately fitting the averaged data to calculate drift and deviation to calculate volatility. There are a number of advantages to the MLE method, not least of which is that for large datasets it can provide an extremely accurate fit. In addition, it also performs a simultaneous fit as mentioned above, and is also relatively transferable, as all that is required to change the model is to "slot in" drift and volatility functions. The purpose of the research we were doing was to establish whether the accuracy and speed of computation was comparable to least squares methods. We took as a starting point the work of Finkenstaadt et al (2008), who were using these methods to infer parameters for transcriptional dynamics, and attempted to apply these techniques to thermostatistic models of transcriptional regulation. One of the potential problems with fitting thermostatistic models is that they typically involve several unknown parameters, and we wanted to determine if MLEs would provide a good way of estimating all of these.

Page 53

In addition, we have also looked at applying the method to a simpler Chemical langlevin model, using both dependant and independent drift and volatility. At present, the models we are using remain simple enough that it is possible to solve analytically rather than numerically, which provides a stronger verification of the value of this method, in addition to allowing it to be used in a more computationally effective manner. While our results to date on Thermostatistic transcriptional regulation have not been promising, with a number of the parameters being incorrectly estimated, we are currently exploring a number of ideas to improve this. However, the results for the simpler models look extremely good, and have the advantage of being able to accurately calculate drift and volatility from a single dataset, as expected, provided there are enough points to work with.

P29: Omics of Mycs: Application of Omics Technologies to Elucidate the MYCN Transcriptional Network at the Systems Level in Neuroblastoma David Duffy1, Thomas Schwarzl1, Aleksandar Krstic1, Jai Prakash Mehta1, Norma Coffey1, Elisa D’Arcangelo2, Walter Kolch2

Systems Biology Ireland, University College Dublin, Ireland We are utilising deep sequencing to comprehensively map the entire MYCN transcriptional network in Neuroblastoma (NB). NB is an embryonal tumour which predominantly arises during infancy and displays a heterogeneity of outcomes from spontaneous differentiation to fatal morbidity. NB arises from the improper terminal differentiation of neural crest cells during embryonic development. These neural progenitors are highly proliferative migratory cells, which enables them to populate the developing embryonal body with peripheral nerves. Therefore, their transformation to cancerous phenotypes is particularly undesirable, as they are already geared to allow rapid tumour growth and metastasis. Genomic amplification of the MYCN oncogene has been shown to correlate with poor patient prognosis. The MYCN transcription factor is a key gene in driving high risk NB development. Overexpression of MYCN alone in neuronal progenitor cells (neural crest cells) subsequently inserted into nude mice was sufficient to induce NB tumour formation. However, despite being widely studied, the complete transcriptional network of MYCN in NB has not been well defined. This is mainly due to the fact that MYCN is a broad but weak transcription factor with the response of target genes largely being close to the detection cut-off levels of our technologies. However, recent advances in deep sequencing are allowing us to overcome this limitation. Utilising deep sequencing we have sequenced the transcriptomic response of inducible stable MYCN overexpression in the non-amplified MYCN NB cell line SH-SY5Y. We are using a range of omics techniques to examine many levels of expression: gene level (mRNA-seq), alternatively spliced transcript level (mRNA-seq), miRNA expression (miRNA-seq) as well as detecting the presence and expression of fusion genes (paired-end mRNA-seq) and the identification of MYCN DNA binding sites (ChIP-seq). While the importance of MYCN in NB has long been known it is not an easily drugable protein. Therefore, using our datasets to identify its transcriptional network and understanding its basic dynamics may provide important insights into potential drug targets to combat this cancer in high risk patients. To date, our time course analysis of MYCN overexpression has detected differential expression of a range of known and novel MYCN target genes (444 differentially expressed genes by 48h MYCN overexpression) including oncogenes and tumour suppressors. We have also detected differential expression of many gene clusters involved in a variety of processes including neuronal development,

Page 54

RNA processing and splicing, cell cycle progression, transcription factors and other developmentally important genes. Our results have highlighted the importance of the role MYCN plays in regulating alternatively spliced transcripts. Low average fold changes at the gene level often mask more dramatic results when individual isoforms are examined. Different isoforms of the same gene can have opposing effects; for example MYCN may down-regulate a repressive isoform while up-regulating the equivalent activating isoform. Analysis at the gene level only would miss such changes even though the biological function has been reversed. We must be conscious of this fact in all transcriptional profiling experiments, and alternative splicing needs to be given broader consideration, otherwise important information will be lost and apparent contradictions will persist. In addition to our MYCN overexpression analysis, we have sequenced a number of MYCN amplified NB cell lines to greater understand their transcriptional similarities and differences and how these compare with the effects of MYCN overexpression. To further expand our basic understanding of NB, we have sequenced the transcriptome of differentiating NB cells; 431 differentially expressed genes were detected by 24h after the initiation of differentiation. Spontaneous differentiation occurs in many tumours, effectively curing the patient. When the genes altered in response to differentiation are compared with those altered in response to MYCN overexpression, genes present in both datasets largely show opposing transcriptional responses, further indicating that MYCN directs cells away from a differentiating fate. Our goal is to reconstruct the MYCN transcriptional network in an embryonal tumour. To achieve this we are utilising the power of omics technologies to obtain the deepest insight yet into the cellular role of MYCN in NB with potential implications for systems medicine.

P30: Identification of Immediate Early Response Genes for Chondrogenesis in Human Mesenchymal Stem Cells using RNA-seq Thomas Schwarzl1, Elizabeth Clayton1, 2, Andrea Keogh1, 2, Frank Barry1, 2, Walter Kolch1, Des Higgins1

1Systems Biology Ireland, University College Dublin, Ireland 2Regenerative Medicine Institute, National University of Ireland Galway, Ireland Chondrogenesis is a highly complex process that takes place over a 2-3 week period and can be triggered through a combination of pelleting and TGFb treatment. By 24 hours, however, commitment and differentiation are evident by expression of known markers of chondrogenesis (e.g. SOX9, fibromodulin, and aggrecan). We are interested in identifying immediate early response genes for chondrogenesis, which give an understanding of the underlying mechanisms and have potential to be used as early markers. This may lead to the development of pharmaceutical compounds that would control mesenchymal stem cell fate in vitro and in vivo. RNA was extracted at 15 minutes, 30 minutes, 1 hour, 2 hours, 4 hours, 8 hours, 16 hours, and 21 days after initiating differentiation. As control, untreated samples (0 hour) and untreated samples in the absence of pelleting (monolayer) were used. The mRNA level of 3 biological replicates per time-point were measured using RNA-sequencing. A general linear model approach was used for the identification of differentially expressed genes in the time-course data. We will show the changes of the transcriptome over time of differentiation as well as the effect of pelleting.

Page 55

P31: Stem Cell Resources on the Web – a semantic web approach toward integrative bioinformatics for stem cells Arindam Halder1, Stefan Decker1, Frank Barry1, Helena F Deus3

1Digital Enterprise Research institute, National University of Ireland Galway, Ireland 2Regenerative Medicine Institute, National University of Ireland Galway, Ireland Stem cells have a promising role to play in multiple areas of medicine. As a result, research in the stem cell area has very actively engaged in the era of high-throughput technologies, high quality imaging and microscopy. The data made available from such studies is huge but there is very little integration. The importance of integrating these dataset is like putting together the pieces of a jigsaw puzzle. The pieces (datasets) are known but complete the puzzle is a complex challenge. It is becoming increasingly clear that the key to efficiently handling the massive amounts of data from such domain of research lies not only in human interpretation but also in machine readability. Supporting this feature can greatly accelerate knowledge discovery and sharing [1]. Semantic web and linked data technologies (SWLDT) consist of a set of best practices for ensuring that data representation on information systems is made machine-readable [2]. In this work, we briefly describe our proposed architecture to integrate various stem cell resources available on the web using SWLDT as well as enabling the seamless integration of newer datasets in a ‘plug-and-play’ manner. We discuss the importance of interlinking various publicly available datasets and the role it plays in knowledge discovery in a faster and more efficient manner than with current methods. Herein we highlight the key research issues of implementing and automating the process of interlinking heterogeneous datasets. We will update on the current status of the work and the future roadmap for implementing the architecture. Stem cell experiments are information-intensive experiments, which can include results from genomics, transcriptomics and proteomics. Interpreting these experiments is, in turn, dependent on the availability and the efficient use of information publicly available in multiple bioinformatics databases. The resources nowadays available to a scientist are immense - there are hundreds of tools for data analysis and hundreds of data repositories available for the scientific community [3]. Mastering each and every tool and data resource requires significant time and resources. One of the other major problem scientists often face is the lack of time to report data generated in standard formats. Due to these challenges, scientists often devise workarounds for reporting their data, often resulting in data annotations using non-standards descriptors. Although supporting tools for some standards reporting exist [4], scientists prefer to pursue the necessary data reporting and analysis using tools with which they are familiarized with but which may not be suited for standardized data reporting and analysis. The core deliverable of the work presented here will be the unification of available stem cell resources at a single unified location to ensure that new data collection methods do not become an obstacle to scientific progress. Our approach attempts to do this by reducing the learning curve in mastering new tools and methods. In this work, we will describe our method, using the federated query approach, as it provides significant advantages, in terms of data synchronization and annotation [5]. The following set of steps will be to match the ontologies being generated from the current data sources to the already existing bio-medical ontologies [6]. In addition to enabling the seamless integration of data available in different formats from multiple sources, this strategy will also give researchers the freedom to creatively combine and query data from ad-hoc data sources. References: 1. T. Berners-Lee and J. Hendler, Publishing on the semantic web,Nature, vol. 410, pp. 1023-1024,

2001.

Page 56

2. S.R. Bratt, Toward a Web of Data and Programs, in IEEE Symposium on Global Data Interoperability - Challenges and Technologies, 2005.

3. Sujansky W., Heterogeneous database integration in biomedicine J Biomed Inform. 2001 Aug;34(4):285-98

4. P.L. Whetzel, H. Parkinson, H.C. Causton, L. Fan, J. Fostel, G. Fragoso, L. Game, M. Heiskanen, N. Morrison, P. Rocca-Serra, S.-A. Sansone, C. Taylor, J. White and C.J. Stoeckert, Jr., The MGED Ontology: a resource for semantics-based description of microarray experiments, Bioinformatics, vol. 22, pp. 866-873, 2006.

5. Peter Hasse,Tobias Mathab, Michael Ziller, An evaluation of approaches to federated query processing over linked data Proceedings of the 6th International Conference on Semantic Systems Article No. 5, ACM,2010

6. Michael Ashburner et. al, Gene Ontology: tool for the unification of biology Nature Genetics 25, 25 - 29 (2000)

P32: The role of Src in MCF-7 cell differentiation. Cormac McCarthy, Natalia Volinsky, Boris Kholodenko, Walter Kolch Systems Biology Ireland, University College Dublin, Ireland MCF-7 is a mammary epithelial luminal carcinoma cell line and is used as a model system for cellular differentiation. Upon stimulation with the ErbB receptor ligand, Heregulin, MCF-7 cells differentiate and accumulate lipid droplets. Unlike some previously described differentiation systems, we have shown that differentiating MCF-7 cells retain their ability to proliferate. We have found that chemical inhibition of Src, a tyrosine kinase acting downstream to Receptor Tyrosine Kinases (RTK), prevents differentiation of MCF-7 cells. We hypothesised that by knocking-down CSK, a negative regulator of Src, the catalytic activity of endogenous Src would increase and result in increased cell differentiation. Our results show an increase in differentiation upon CSK-targeted siRNA knock-down in basal conditions and upon cell stimulation. This result suggests that upregulation of Src catalytic activity promotes differentiation of MCF-7 cells. MCF-7 cell differentiation is mediated by sustained activity of PI3 kinase and can also be induced by other strong activators of PI3 Kinase, such as Insulin or phorbol ester, 12-O-tetradecanoylphorbol-13-acetate (TPA). We have found that Src catalytic activity is also required for differentiation induced by Insulin and TPA. We hypothesise that Src acts on two levels within signalling pathways: regulation of RTK kinase activity by their direct phosphorylation and specific regulation of PI3 kinase activity. Our preliminary data demonstrates that Src inhibition during Heregulin or Insulin stimulation blocked activation of two major signalling pathways: MAP Kinase and the PI3 Kinase pathways. On the contrary, Src inhibition of TPA stimulated cells, where MAP Kinase and PI3 Kinase pathways are activated independently of RTKs, prevented PI3 Kinase activation but had no effect of MAP Kinase activity. Based on this data, we suggest that Src functions both, on RTK level and within the PI3 Kinase pathway, and we are currently investigating the role of Src using specific chemical inhibitors and proteomics tools.

P33: Experimental and theoretical modelling of the mitogen activated protein kinase pathway Louise Maddison1, Nils Blüthgen2, Kathleen Carroll1, Hans Westerhoff1

1Manchester University, UK 2Charité - Universitätsmedizin Berlin, Germany

Page 57

The MAPK pathway plays a crucial role in regulating the cellular response to external stimuli. The binding of growth factors and other mitogenic signals to cell surface receptors initiates a phosphorylation-dependent relay of protein activation, resulting in altered transcription, ultimately regulating cell proliferation and differentiation. Signalling through this pathway is regulated by the coordinated function of specific protein kinases and protein phosphatases. As perturbation of this signalling system is often associated with diseases such as cancer, modelling is a useful means to improve our understanding of the outcomes that may result following changes in component levels or activity. Expansion of current models of the MAPK pathway is necessary to facilitate accurate modelling by providing quantitative information on both total protein levels and stoichiometry of protein components. Mass spectrometry and the QConCAT approach is being used to provide absolute quantitative data for ~30 protein kinases, phosphatases, scaffolds and substrates involved in MAPK signalling (MEF2C, c-Myc, Elk-1, Stat3, EST1/2, DUSP1/2/4/5/6/7/9/10/14/16/18, KSR1/2, Mp1, p14, Paxillin, MORG1, β-arrestin1/2, PEA-15, Sef1, IQGAP1, MVP, RKIP). QConCAT constructs of selected tryptic peptides for each of the proteins of interest were designed and expressed to generate [13C6]Arg/Lys labelled peptides as internal standards for absolute protein quantification. Multiple Reaction Monitoring (MRM) assays were developed to quantify the proteins from mammalian colon cancer cell lines HT29 and HCT116 and cell line HEK293 as a control. Absolute quantification data for the scaffold proteins was obtained in these cell lines and scaffold behaviour was predicted by mathematical modelling.

P34: Personalised medicine in pancreatic cancer: use of systems level approaches to identify therapeutic targets Bryan Miller, Lynn McGarry, Emma Shanks, Jennifer Morton, Owen Sansom The Beatson Institute for Cancer Research, Glasgow, UK Therapeutic approaches to pancreatic cancer are complicated by the heterogeneous nature of the disease. While activating mutation of Kras is an early events in pancreatic cancer progression, and is associated with over 90% of pancreatic tumours, numerous further mutations are associated with later-term disease. Addressing these diverse genetic backgrounds of patients with personalised medicine approaches represents a promising strategy for the development of novel therapies. We have combined the use systems level approaches and transgenic mouse models to identify new therapeutic targets that are specific for certain subtypes of pancreatic cancer. We have established mouse models of pancreatic cancer that couple activating mutation of Kras with mutation of p53 (p53R712H) or deletion of Pten (Ptenflox). Both models display greatly accelerated tumourigenesis when compared to Kras mutation alone. Microarray analysis shows that there are specific patterns of altered gene expression in each individual model, with distinct subsets of genes upregulated in each. In order to address which of these upregulated genes are required for tumour cell survival, we established cell models from p53R712H and Ptenflox tumours and have used them to undertake a siRNA-based cell viability screen. We have identified a subset of genes required for the maintenance of viable p53R712H cells that have no effect on Ptenflox cells. Overlaying the cell viability and microarray data sets has allowed us to identify specific subsets of genes that display aberrant expression and are required for tumour cell viability. Furthermore, the clear differences obtained with the two models, suggest that there are distinct therapeutic targets in each.

Page 58

Lysyl oxidase (Lox) was identified as a target specific to p53R712H tumours. Further investigation in cell models suggested that abrogation of Lox expression sensitises p53R712H cells to the chemotherapeutic agent gemcitabine. In mouse models, we have undertaken a co-treatment regime with a blocking antibody to Lox and gemcitabine and have shows that it significantly impairs tumourigenesis. This is associated with stromal alterations and increased blood flow and immune cell infiltration in the tumour. The application of systems-level approaches with transgenic mouse technology has led to the identification of potential therapeutic targets specific for distinct pancreatic cancer models. This raises the prospect on progress towards developing new personalised medicine approaches for pancreatic cancer therapies in the future.

P35: Protein expression changes in cancer motility, invasion and metastasis Mohammad Naser Monsefi, Alexander Von Kriegsheim Systems Biology Ireland, University College Dublin, Ireland Cancer deaths are mainly due to tumour invasion that escapes therapy. While progress has been made to reduce tumour mass by surgical, radiotherapeutic or chemotherapeutic interventions, we still lack therapies that can prevent cancer cell invasion. This gap reflects our lack of molecular mechanistic knowledge of how cancer cells invade. Cell motility has been extensively studied with cell line models plated on plastic or other two-dimensional environments. Although convenient, results obtained by this simplification sometimes poorly correlate with the in-vivo data. Thus we propose focuses on cancer cell motility in three-dimensional environments which provide cells with a surrounding similar to the situation in-vivo. A2780 well characterised ovarian cancer cells and its stably expressing Rab25 derivative is used for modelling invasion. Rab25 is a member of the RAS family of small GTPase that has been linked to tumour aggressiveness and metastasis. Interestingly when these cells are plated on plastic; Rab25 over-expression does not alter cell velocity or morphology; while the cells cultured in-vivo, in-vitro in a three-dimensional matrix or on cell derived matrix (CDM); a pseudo-3D environment produced by fibroblasts; Rab25 not only alter the migratory speed but also changes the cell morphology and promotes invasiveness. This has been shown using biological analysis including migration assays and live cell imaging alongside new findings while inhibiting different kinases activity. In order to measure protein expression changes between invasive and non-invasive cells quantitative mass spectrometry was used and samples were prepared base on standard triple SILAC protocol followed by strong anion exchange (SAX) fractionation. Then they were analysed by LC-MS/MS on an Orbitrap mass spectrometer and protein groups were identified using MaxQuant software. Preliminary protein expression results reveal indirect information about regulated pathway mostly related to migration and focal adhesion.

P36: Sequencing approach to identify miRNA targets for MYCN over-expression Jai Prakash Mehta, David J. Duffy, Thomas Schwarzl, Walter Kolch Systems Biology Ireland, University College Dublin, Ireland Neuroblastoma is a childhood cancer which arises from the developing sympathetic nervous system. Previous studies have shown MYCN to be associated with the most aggressive form of the cancer. Here we aim to use next generation sequencing to identify miRNAs that are regulated by the expression of MYCN.

Page 59

Materials and Methods: miRNA sequencing was performed in the neuroblastoma cell line SH-SY5Y-MYCN, which contains a stable inducible MYCN overexpression construct. Biological replicates were generated in duplicate at 0, 1, 4 and 24 hours after induction of MYCN overexpression. The sequences obtained were de-multiplexed, adapters removed, aligned to mirBASE and reference genome, quantified, normalised and differentially expressed miRNA was identified using DGEseq. Additionally ANOVA followed by clustering approach was also used to identify miRNAs of interest. Results: The total number of reads obtained was 22891194 with an average of 2861399 reads per sample. On average the alignment was 86% with the mirBASE. A total of 1073 miRNA were expressed in either of the samples. Differential expression analysis identified 6, 21 and 30 Differentially expressed (p < 0.05) miRNAs comparing Control with Time 1 hr., 4 hr. and 24 hr. respectively. ANOVA analysis among the various time points identified 47 miRNA to be differentially expressed. Clustering miRNA identified the biggest changes at time 24 hour. A novel prediction algorithm identified a number of previously unknown miRNAs among the samples.

P37: Detection of MYCN targets in neuroblastoma by ChIP seq Aleksandar Krstic1, David Duffy1, Thomas Schwarzl1, Elisa D’Arcangelo2, Walter Kolch1

1Systems Biology Ireland, University College Dublin, Ireland 2Conway Institute, University College Dublin, Ireland Neuroblastoma (NB) is a specific type of embryonal tumours that arises from precursor cells of the sympathetic nervous system. This malignancy exerts heterogeneous clinical behaviour, from treatable or spontaneously regressing to untreatable or relapsing forms. Neuroblastoma is extensively studied, but still represents a major challenge for paediatric oncologists, the pharmaceutical industry and researchers. It is diagnosed in approximately eleven out of million children in Europe, predominantly affecting infants. Although improved clinical protocols are constantly introduced, five-year survival for patients is only up to 67%. The MYC gene family members, which encode leucine zipper transcription factors (TFs), are crucial for understanding the biology of NB, since they affect almost every aspect of their behaviour. Amplification of the MYCN gene has been associated with poor survival in NB. It has also been shown that overexpression of MYCN, in the absence of amplification, can result in spontaneous NB regression. Our experiments are aimed at the identification of genomic regions bound by MYCN, with particular emphasis on its target genes. This question is being addressed by a ChIP seq (Chromatin Immuno Precipitation Sequencing) experiment in the human neuroblastoma SH-SY5Y cell line with inducible MYCN expression. Sequencing data obtained on Illumina GIIx platform will provide in-depth information on MYCN DNA binding activity. Planned analyses will include distribution of binding sites relative to position of transcription start points, MYCN binding motif enrichment and discovery of novel target motifs, as well as, identification of targets bound by MYCN with high affinity. Gene Ontology and pathway analyses will further enable investigation of targets which are involved in key aspects of NB behaviour, such as apoptosis, chemoresistance, metastatic potential and differentiation. The ChIP seq data will feed into the generation of mathematical models of the MYCN TF network in NB which will aid in the identification of vulnerable nodes as potential therapeutic targets.

Page 60

P38: The role of Paxillin on migration and phenotype in human breast cancer cells Andrew Murphy, Walter Kolch Systems Biology Ireland, University College Dublin, Ireland Changes of phenotype, actin cytoskeleton, cell adhesion and motility accompany metastasis. These changes are accomplished via signalling pathways which in turn influence cell behaviour and phenotype. Crucial to these signalling pathways are scaffolds proteins which can regulate spatial and temporal orchestration of signalling. Since our focus is the migration and invasion of breast cancer cells, here we discuss the central role of the actin-regulating adaptor protein, Paxillin (PXN). We examined the role of PXN in migration, cellular phenotype, mRNA expression and transcription factor activity in the mesenchymal, triple negative, breast cancer cell line, MDA-MB-231 cells. We found that in agreement with previous observations siRNA mediated knock down of PXN inhibited the migratory capacity of MDA-MB-231 cells as measured by scratch heal wound assays and transwell migration assays. In addition to regulation of the migration of MDA-MB-231 cells knock down of PXN also influenced the expression of a number of markers of mesenchymal phenotype including Vimentin and fibronectin. In addition to this PXN knock down also regulated the expression of mRNA of a number of molecules crucial to metastatic/migratory behaviour. In order to assess how these alterations in expression were occurring we employed a unique reporter assay system to assess transcription factor activity. The reporter assay system in called the Molecular Nose ™ (Jiwaji et al., 2011). The Molecular nose comprises plasmids with individual transcription factor binding sites upstream of unique reporter sequences. As transcription factor DNA binding is increased or decreased (by post-translation modification) and increased or decreased amount of transcription factor is bound to the plasmid producing increasing or decreasing amounts of the unique reporter sequence. The quantity of Unique Reporter sequence produced is compared between samples using high-throughput quantitative real time PCR. We examined the effect of siRNA mediated knock down of PXN on the activity of over 50 transcription factors. We found that PXN knock down influenced the activity of number of transcription factors in a manner consistent with the observed changes in mRNA and protein expression. Using these results we have attempted to create a clearer understanding of the role PXN in breast cancer phenotype and metastasis.

P39: The Kinase Suppressor of Ras dynamic interactome Brendan McCann, Jens Rauch, Walter Kolch

Systems Biology Ireland, University College Dublin, Ireland

The Ras – Raf – MEK – ERK/MAPK pathway is an important conserved kinase signalling pathway implicated in the cellular response to extracellular stimuli. Activation of the pathway can result in a variety of specific outcomes for the cell, affecting fundamental processes such as the cell cycle, proliferation, differentiation, transformation, apoptosis, metabolism, the cytoskeleton, adhesion, and motility. Disruption of this pathway has been implicated in a variety of cancers, and has long been of therapeutic interest. Spatio-temporal organisation of the ERK/MAPK pathway contributes to this diversity in outcome. ERK signalling is facilitated at different locations by several scaffold proteins, including Kinase Suppressor of Ras (KSR). KSR is a scaffold protein that interacts with the kinase isoforms of the ERK signalling cascade. KSR1 regulates the intensity and duration of ERK signalling and is located in the cytosol with MEK constitutively bound. Upon activation of Ras, KSR1 is recruited to the plasma membrane, and Raf and ERK are recruited into the complex, resulting in ERK activation. Importantly, KSR1 is known to regulate only a subset of Raf function, as it binds only 5% of cellular Raf. In addition, KSR1-/- knockout mice were shown to have no major detrimental effects and little impairment in bulk ERK activation. However, KSR1

Page 61

knock-out mice, when treated with oncogenic retrovirus, surprisingly do not get cancer. These results suggest that KSR1 mediated Raf signalling plays a crucial role in cancer development. Therefore, the aim of our project is to better understand the role of KSR1 in MAPK signalling. As interactions with KSR1 seem to be crucial for this process, we want to identify dynamic interactors of KSR1 upon growth factor stimulation. Serum-starved HEK293 cells stably expressing GFP tagged KSR1 or KSR2 were stimulated with EGF at different time points, followed by lysis and immunoprecipitation of tagged KSR complexes. Components of the KSR complexes were then identified and quantified by mass spectrometry revealing the change in complex composition. This quantitative proteomics approach will allow for the identification of novel interactions and profiling of interaction dynamics, which can then be further characterised to establish their significance and potential role in oncogenesis.

P40: Meta-analysis of pediatric tumour transcriptional profiles based on independent components Loredana Martignetti1, Zinovyev Andrei1, Karine Laud-Duval2, Olivier Delattre2, Emmanuel Barillot1

1Institut Curie / INSERM U900 / Mines ParisTech, France 2Institut Curie / INSERM U830, Paris, France Motivations: Most pediatric tumours develop from embryonic tissue and are characterized by a genetic profile with one or few key genetic events being clearly responsible for tumorigenesis. They represent an excellent system for understanding the complex picture of deregulations happening in cancer cells. The availability of high-throughput transcriptome data for multiple pediatric tumours allows us to identify activation/inhibition patterns common to various pediatric diseases as well as those specific for a single tumour type. However, exploitation of large amount of data generated by high-throughput expression profiles is a difficult task. Gene expression is controlled by many cellular variables and the derived expression profiles are complex and noisy. We expect each profile to be influenced by several factors, like different transcription regulators, cellular processes and biological responses. Independent component analysis (ICA) is a statistical and computational technique for revealing hidden factors that underlie sets of measurements or signals [1]. It relies on the idea of a combinatorial control, describing the expression profiles of genes as linear functions of common hidden variables. According to the ICA model, the hidden variables exert linear influences on the gene profile with minimal statistical dependences between them. Once independent components have been identified, this analysis also reveals those individual genes that have the strongest contribution to each component. Robust components define to groups of commonly influenced genes and we expect to find among them some which are relevant in tumour biology. Methods: We apply ICA to decompose gene expression profiles from different pediatric tumours, including Ewing sarcoma, rhabdoid tumors, neuroblastoma, medulloblastoma and desmoplastic round cell tumors. Data from each tumor dataset were analyzed separately and independent components were extracted using FastICA algorithm [2]. Robust components for each dataset were defined by a vector of gene projection values. Then, reproducible components in different datasets were identified through a correlation graph. The most reproducible components in this graph correspond to strongly connected subgraphs. Results:

Page 62

To assess the biological relevance of gene groups identified by ICA, we investigated multiple data sources, including manually curated protein-protein interactions from HPRD database and disease-gene associations. The analysis of gene groups shows some components reproducible in all independent datasets. These have been associated to tumour related processes such as cell cycle, cell growth, chromatin remodelling and invasion. A component common to all brain tumours is characterized by multiple ELAV-like family genes, known to increase MYCN mRNA stability in neuroblastoma. These findings confirm the potential of ICA to extract information from large and complex expression data. The analysis helps in distinguishing between hidden regulatory factors shared by different pediatric tumours as well as disease-specific ones. [1] Comon P, Signal Processing, 1994 [2] Hyvarinen A, IEEE Trans Neural Netw., 1999

P41: Understanding Cell Signaling through Linked Data Laleh Kazemzadeh1, Frank Barry2, Stefan Decker1, Helena F. Deus1

1Digital Enterprise Research Institute (DERI), National University of Ireland Galway, Ireland 2Regenerative Medicine Institute (REMEDI), National University of Ireland Galway, Ireland Cellular processes are the result of signalling orchestras, transforming environmental effectors through specific pathways in order to regulate specific cellular functions. These pathways are often overlapping, indicating that they interact with each other, forming a biological network1. As such, perturbation of one may affect regulation of others. Investigation into the dynamics of signalling networks relies on the background information of the network structure and how proteins or genes may interact together. Such background information can be obtained from advanced techniques in genomics, proteomics and metabolomics. Even though these techniques facilitate the accumulation of experimental data, analyzing these datasets require application of a combination of different disciplines such as mathematics, informatics, chemistry and statistics in order to process the data, highlight the possible outcomes and prove the hypotheses. However, these datasets are not always consistent because they were generated using different experimental protocols or techniques for very different purposes. Hence, a core challenge towards enabling efficient analysis is to make these datasets interoperable by transforming the data into a unified format. Linked Data can be used to address the interoperable data challenge. Linked Data is a set of best practices for linking related data sets, distributed across the web and ultimately, enable the creation of a complete and global dataset enabling the discovery of hidden/missing data. A key concept in Linked Data is the Uniform Resource Identifier (URI), which represents the identifier of any element in a dataset and is accessible worldwide via HTTP2. The fact that several large biological data sets have been published in Linked Data format is a clear indication that Linked Data in life science has achieved wide acceptance3. Among the several types of experimental biological data worth considering for understanding biological networks, Protein–Protein interaction (PPI) data is typically the most accurate and precise type of information to use when devising a network structure. However, networks built from PPI are often not complete due to lack of accurate experimental data; some connections might be missing in the network structure. It is often possible to formulate hypotheses for missing interactions in the network and thus suggest and test whether new links should be added to the network. Computational modeling offers great promise in ranking and narrowing down the hypothesis before they need to be experimentally validated. To broaden the scope of data available for assembling the networks, we propose an approach whereby data from the literature, gene expression and structural prediction are also processed and integrated in the model using different and separate layers, one per data type. Applying Linked Data

Page 63

principles greatly facilitates integration, transformation and analysis of the data in a uniform format while keeping data in different layers according to their sources. The non-overlapping portions of the layers can be used to suggest new interactions, which might not have been detected by a particular type of experiment but highlighted in several others. The aim of our work is to develop a model capable of predicting missing information in PPI network of a given pathway from different available data of multiple types. To this end, it is necessary to start by assembling a Boolean model of the selected pathway using literature research findings. The core model will then be expanded by addition of new links to the network structure based on suggested connections from different layers of data. We will also devise a formal model by specifying the rules of PPI in order to filter validate whether the asserted interactions are viable. By applying this method, we will attempt to devise and complete the model for mitogen-activated protein kinase (MAPK) PPI network. MAPK network is a very well studied pathway and misregulated in various types of cancer. MAPK pathways are highly conserved pathways involve in regulation of various processes such as apoptosis, survival, growth, differentiation and migration in cell and disruption of each may lead to formation of tumor cell4. Their role in directing growth factor and mitogens is of particular interest in cancer studies. As such, our main focus will be on the role of critical elements involved in regulating these pathways. Our ultimate aim is to build a network structure based on available PPI knowledge and compare the network with the network built from experimental data in order to predict the missing information in the network topology. References: 1. Eungdamrong, N.J. & Iyengar, R. Modeling Cell Signaling Networks. Biology of the Cell 94, 355-362 (2004). 2. Bizer, C., Heath, T. & Berners-Lee, T. Linked Data - The Story So Far. International Journal on Semantic Web and Information Systems 5, 1-22 (2009). 3. Cyganiak, R. & Jentzsch, A. Linking Open Data cloud diagram. at <http://lod-cloud.net> 4. Dhillon, A., Hagan, S., Rath, O. & Kolch W. MAP Kinase signaling pathways in cancer. Oncogene 26, 3279-3290 (2007).

P42: Footprinting Microbial Metabolites in Nature and Medicine Sónia Carneiro, Daniela Correia, Eugénio Ferreira, Isabel Rocha

University of Minho, Portugal

The study of metabolic alterations in response to genetic and environmental perturbations has been a central topic in microbial metabolomics (Fiehn, 2002; Kol et al., 2010; Villas-Boas et al., 2008). Some of these alterations can be readily detected by changes in their surroundings, normally associated with metabolites that are released by cells as by-products of the metabolism or as extracellular signalling molecules to mediate cross-talk within microbial communities. The analysis of these metabolites, also known as metabolic footprinting, has been widely applied with different purposes: discriminating between metabolic phenotypes in order to classify and identify mutant strains (Villas-Boas et al., 2008); monitoring bioprocesses with the aim to detect specific metabolites that indicate alterations in the culture performance (Carneiro et al., 2011; Sue et al., 2011); and identifying quorum-sensing metabolites that indicate potential targets to annihilate pathogens (Birkenstock et al., 2012). These metabolic readouts have been also useful to give insights into intracellular metabolic activities and provide a straightforward way to analyse simultaneously multiple metabolic activities, since no extraction procedures are required to analyse the endometabolome (i.e., intracellular metabolites).

Page 64

Thus, through metabolic footprint analysis we can assess central metabolic activities that characterize the reproduction and survival of organisms. We have developed a methodology to evaluate the metabolic state of microbial cultures by analysing the footprints of two microbial systems: the bacterium Escherichia coli and the human pathogen Helicobacter pylori that increases the risk of gastric cancer. Strategies for sampling and sample preparation were developed, as well as the analytical procedures based on gas chromatography with mass spectrometry (GC-MS). A wide variety of metabolites was detected, including fatty, amino and organic acids, which allowed us to address changes in most central metabolic pathways, such as the tricarboxylic acid cycle (TCA cycle), the biosynthesis of amino and fatty acids, as well as other energy generating metabolic reactions. The analysis of extracellular metabolites of E. coli cultures at different growth conditions were first performed to discriminate the physiological state of cultures and to evaluate the metabolic alterations produced at different growth conditions. According to our results in these experiments, metabolic footprints are good indicators of alterations in the intracellular metabolism. Next, the metabolic footprints of H. pylori cultures were investigated to get insights on the catabolism of this human pathogen. Overall, fifteen amino acids were detected in extracellular medium; six of them were confirmed as essentials for H. pylori growth, four amino acids were identified as non-essentials and can be used as carbon source, whilst five amino acids were identified as non-essentials and non-carbon source. In addition, some organic acids were also identified as carbon sources for H. pylori. This metabolic footprint analysis of H. pylori cultures allowed us to uncover key metabolic activities, mainly related with amino acids catabolism and to get insight on the metabolic behaviour of this organism. The characterization of catabolic pathways, as well as of possible metabolic constraints, is of major importance to understand the dynamic basis of the interactions host–microbe in the human gut, and in particular to discover potential ‘diagnostic’ biomarkers. It is well-known that pathogen's metabolism can influence the host health and may affect drug metabolism, toxicity and the efficacy of therapies (Holmes et al., 2011). However, little is known about their metabolic structure and behaviour. Our methodology allows uncovering part of the metabolic structure of H. pylori metabolism and undisclosed catabolic activities. Acknowledgments: This work was partially supported by the MIT-Portugal Program in Bioengineering (MIT-Pt/BS-BB/0082/2008), the research project HeliSysBio-Molecular Systems Biology Helicobacter pylori (FCT PTDC/EBB-EBI/104235/2008) and a Post-doc grant from Portuguese FCT (Fundação para a Ciência e Tecnologia) (ref. SFRH/BPD/73951/2010). References: 1. Kol S, Merlo ME, Scheltema RA, de VM, Vonk RJ, Kikkert NA, Dijkhuizen L, Breitling R, Takano E.

2010. Metabolomic characterization of the salt stress response in Streptomyces coelicolor. Applied and Environmental Microbiology 76: 2574-81.

2. Villas-Boas SG, Moon CD, Noel S, Hussein H, Kelly WJ, Cao M, Lane GA, Cookson AL, Attwood GT. 2008. Phenotypic characterization of transposon-inserted mutants of Clostridium proteoclasticum B316(T) using extracellular metabolomics. Journal of Biotechnology 134: 55-63.

3. Sue T, Obolonkin V, Griffiths H, Villas-Boas SG. 2011. An exometabolomics approach to monitoring microbial contamination in microalgal fermentation processes by using metabolic footprint analysis. Appl Environ Microbiol 77: 7605-10.

4. Carneiro S, Villas-Boas SG, Ferreira EC, Rocha I. 2011. Metabolic footprint analysis of recombinant Escherichia coli strains during fed-batch fermentations. Molecular Biosystems 7: 899-910.

Page 65

5. Birkenstock T, Liebeke M, Winstel V, Krismer B, Gekeler C, Niemiec MJ, Bisswanger H, Lalk M, Peschel A. 2012. Exometabolome analysis identifies pyruvate dehydrogenase as a target for the antibiotic triphenylbismuthdichloride in multiresistant bacterial pathogens. J Biol Chem 287: 2887-95.

6. Holmes E, Li JV, Athanasiou T, Ashrafian H, Nicholson JK. 2011. Understanding the role of gut microbiome-host metabolic signal disruption in health and disease. Trends Microbiol 19: 349-59.

P43: Large-scale modelling of cancer related pathways Maria Luisa Guerriero, Lan Nguyen, Tapesh Santra

Systems Biology Ireland, University College Dublin, Ireland

Deregulations of signalling pathways are known to be involved in the emergence and progression of cancers. Reductionist approaches over the last few decades have provided us with good understanding of the main mechanistic steps in many pathways, which are traditionally viewed as cascades of events that translate a ligand-mediated extracellular signal into phenotypic outcomes. In reality, however, a ligand can trigger multiple pathways and multiple ligands can trigger overlapping pathways via feedback and crosstalk interactions which operate by means of transcriptional and non-transcriptional mechanisms. It is through these complex, integrated networks of signalling pathways that cells process and make fate-related decisions, rendering it difficult to dissect the contribution to phenotypic outcomes of different extracellular signals. To gain further understanding of how different signalling pathways exchange information to translate multiple extracellular signals into cellular phenotypes, we are developing a large-scale computational model which integrates many well-known signalling pathways, such as the MAPK/ERK pathway, PI3K/AKT pathway, JAK/STAT pathway, p53/Mdm2 pathway, and the Caspase signalling pathway. These pathways are triggered by a variety of stimuli such as the Epidermal Growth Factor (EGF), Cytokines, Insulin and Tumor Necrosis Factor (TNF), and lead to a wide range of phenotypic outcomes such as proliferation, differentiation, cell division, adhesion and apoptosis. Complex diseases such as cancers often result from multiple mutations, many of which occur in the pathways considered in our computational model. Our large-scale model, therefore, has the potential to reveal how carcinogenic mutations modulate the signalling mechanisms and cause inappropriate cellular phenotypes which may lead to tumour development. This information can then be used to pinpoint parts of the signalling mechanisms that can be targeted for more efficient cancer therapy. The model would also be a valuable tool to explore novel therapies where combinations of different drugs are used to target multiple molecules.

P44: Analysis of the mTOR/DEPTOR signalling system reveals a novel mechanism of oscillations based on protein-protein interactions Thawfeek Mohamed, Lan Nguyen, Boris Kholodenko

Systems Biology Ireland, University College Dublin, Ireland

Background: The mammalian target of rapamycin (mTOR) signalling pathway plays central roles in the control of cell survival, proliferation and tumourogenesis. Recently, DEPTOR has been identified as a new mTOR-interacting protein whose degradation is mediated by both mTORC complexes (mTORC1/2). Their interplay endows the mTOR pathway with intricate interlinked feedback mechanisms capable of

Page 66

displaying rich dynamical behaviours. Oscillations have been thought to arise mainly through negative feedback regulations with sufficient nonlinearity. Methodology: We constructed a mathematical model which describes the dynamic interactions within the mTOR/DEPTOR network. Using this model, we investigated conditions under which oscillations can arise. To identify the nodes and connections required for oscillations, a computational framework based on random parameters sampling and stability analysis was developed. The network was thus strategically reduced to explain the emergence of oscillations. Results: We showed that the mTOR/DEPTOR interaction network can display oscillatory behaviour. The mechanism of oscillations was unravelled which relies only on simple protein-protein interactions without the need of having explicit negative feedback loops. General protein-protein interaction motifs and parametric regimes for oscillations were described. We also identified a minimal mTOR/DEPTOR interaction network which is capable of oscillations. Conclusions: A possibility of oscillations exists in the mTOR/DEPTOR signalling network. Importantly, such oscillations can arise simply from protein-protein interactions without the requirement of explicit negative feedback regulation. The present work indicates that the conditions for oscillations may be less stringent after all. Implications: The protein-protein binding pattern that was found to be oscillating is actually quite common in cell signalling systems. The current findings have important significance for other signalling networks in general and suggest that oscillations may be much more prevalent than previously thought.

P45: “Pathway hopping”, a novel method to map signalling pathways Bernadetta Turriziani, Amaya Garcia, Susan Kennedy, Alex Von Kriersheim

Systems Biology Ireland, University College Dublin, Ireland The ErbB family of tyrosine kinase receptors play an important role throughout the early stages of development, and beyond, governing physiological processes as disparate as cell proliferation and differentiation. The enzymatic cascades activated by ErbB receptors involve numerous proteins and kinases which not only mediate the effects of the receptor, but also enable communication between ErbB and other receptor pathways. This cross-talk partially explains the diverse cell fates which have been linked to ErbB signalling. De-regulation of these receptors is a main cause of malignant transformation and neoplastic carcinoma. Prominently, overexpression of ErbB2 s is one of the leading factors in breast cancer and other cancer types, such as lung, ovarian,stomach, and bladder carcinomas. Numerous studies indicate that amplification or overexpression of ErbB2 disrupts normal cell-control mechanisms and gives rise to aggressive tumour cells which are more prone to metastasise. Using a label-free quantitative interaction- proteomic approach, we have mapped the dynamic interactome of the ErbB signalling pathway downstream of its physiological activator, Heregulin. Our approach allows us to follow the signal and to reverse engineer the signalling cascade in an unbiased manner. This will deepen our understanding of how the signal downstream of Heregulin is relayed from the plasma membrane to the nucleus. Additionally, the reconstructed network will highlight pivotal interactions and reactions, some of which may be suitable as new therapeutic targets in cancer types where the ErbB pathways are deregulated.

Page 67

P46: The tumor suppressor DiRas3 regulates dimerization and kinase activity of C-RAF Angela Baljuls1, Matthias Beck2, Ayla Oenel3, Armin Robubi3, Ruth Kroschewski2, Mirko Hekman3, Thomas Rudel3, Ulf R. Rapp4, Walter Kolch1

1Systems Biology Ireland, University College Dublin, Ireland 2Institute of Biochemistry, Swiss Federal Institute of Technology (ETH), Zurich 3Theodor Boveri Institute of Bioscience, University of Wuerzburg, Germany 4Max Planck Institute for Heart and Lung Research, Bad Nauheim, Germany Upregulation of the classical MAPK signal transduction pathway occurs in approximately 30% of all human cancers (1). In light of the experimental evidence, this pathway represents an attractive target for the pharmacological intervention in proliferative diseases. Therefore, understanding how the naturally-occurring inhibitors intercept the signal transduction through the MAPK pathway is crucial for the development and improvement of anti-cancer drugs. The Ras-related tumor suppressor gene DiRas3 is lost or down-regulated in more than 60% of ovarian and breast cancers. The anti-tumorigenic effect of DiRas3 is achieved through several mechanisms including inhibition of cell proliferation, motility and invasion, as well as induction of apoptosis and autophagy (2). Despite considerable progress the molecular mechanisms of the DiRas3 tumor-suppressive activity are not sufficiently elucidated. Using transient expression of DiRas3 in COS7 cells we could show that DiRas3 interacts with H-Ras and that activation of H-Ras enforces its association with DiRas3, indicating that the tumor-suppressive activity of DiRas3 is achieved, at least in part, at the level of Ras signaling. Although associated with DiRas3, H-Ras is able to bind to its effector C-RAF and the multimeric complex consisting of DiRas3, C-RAF and H-Ras is more stable than the two-protein complexes H-Ras/C-RAF or H-Ras/DiRas3, respectively. Binding of DiRas3 to H-Ras/C-RAF complex disrupts the H-Ras-induced heterodimerization of C-RAF with B-RAF and suppresses the kinase activity of C-RAF. Acknowledgements: This work was funded by Deutsche Forschungsgemeinschaft (DFG) Grant SFB 487, Project C3 and personal research fellowship. The work at ETH-Zurich was supported by an ETHZ research council fellowship, the Novartis Foundation for medical-biological research, and the Roche Research Foundation (2000-2006). References: 1. Wellbrock, C., Karasarides, M., and Marais, R. (2004) Nat Rev Mol Cell Biol 5, 875-885 2. Yu, Y., Luo, R., Lu, Z., Wei Feng, W., Badgwell, D., Issa, J. P., Rosen, D. G., Liu, J., and Bast, R. C., Jr. (2006) Methods Enzymol 407, 455-468.

P47: hCNK1: tuning the MST2/LATS1 pathway Alexandro Membrino1, Roby Urcia2,Walter Kolch1,Andrew Pitt2, David Gomez Matallanas1 1Systems Biology Ireland, University College Dublin, Ireland 2Sir Henry Wellcome Functional Genomics Facility, University of Glasgow, UK hCNK1 is a scaffold protein made of different binding modules, none of them with enzymatic activities. It is involved in the regulation of different biological functions such as cell growth and apoptosis. hCNK1 is able to regulate these processes, in part, by interacting with Raf-1 and RASSF1A. Since Raf-1 and RASSF1A are upstream components of the MST2/LATS1 pathway we investigated hCNK1 role in its regulation. MST2/LATS1 pathway can lead to cell growth or apoptosis depending on its tuning. Here we present experimental evidence showing that hCNK1 acts as the hub of Raf-1-MST2-RASSF1A interaction.

Page 68

Depending on its concentration, hCNK1 can keep MST2 in an inactive state bound to Raf-1, or facilitates a pro-apoptotic interaction of MST2 with RASSF1A, followed by MST2-LATS1 interaction. Furthermore, our data indicates that hCNK1 has an effect in the regulation of apoptosis mediated by the MST2/LATS1 pathway. Thus hCNK1 is required for Raf-1 inhibitory binding to MST2 but also for MST2/RASSF1A pro-apoptotic interaction. In summary our results reveal a switch of MST2 between MST2/RASSF1A and MST2/LATS1 complex regulated by hCNK1 which coordinates assembly of appropriate pathway components to regulate cell fate decisions.

P48: Role of hydroxylation in the regulation of Glycolysis Javier Rodriguez-Martinez, Alexander Von Kriegsheim Systems Biology Ireland, University College Dublin, Ireland Posttranslational modifications (PTMs) can cause profound changes in protein function. One of the most common PTMs in humans is the hydroxylation of a proline residue. Hydroxyproline (HyP) is implicated in the regulation of different cellular processes. It has been shown to be necessary during the synthesis of collagen, where it stabilises the final protein assembly and folding. More recently, hydroxylation of proline has been demonstrated to be regulating the hypoxic response by de-stabilising HIF1-alpha, its transcriptional master regulator of hypoxia. An essential part of the response is the switch of several metabolic pathways to anaerobic glycolysis. The main aim of this project is try to see if the deregulation of glycolysis under hypoxic conditions could be associated with the presence of hydroxylations within the canonical pathway. Working with mass spectrometry techniques we identified several proline hydroxylations in the glycolytic pathway and we quantified oxygen dependent changes of hydroxylation of glycolytic enzymes. So if we can corroborate the implication of this hydroxylation over the deregulation of glycolysis in hypoxia, in the future these residues could be a good target to control this metabolic alteration.

P49: Friend or foe? Progesterone-Induced Blocking Factor differentially regulates physiological trophoblast and pathological tumour invasion Melinda Halasz1, Gergely Berta2, Livia Czimbalek3, Beata Polgar1, Julia Szekeres-Bartho1 1Dept. of Medical Microbiology and Immunology, Medical School, University of Pecs, Pecs, Hungary, 2Dept. of Medical Biology, Medical School, University of Pecs, Pecs, Hungary, 3Dept. of Biophysics, Medical School, University of Pecs, Pecs, Hungary Introduction: Progesterone-Induced Blocking Factor (PIBF) mediates the immunological effects of progesterone during pregnancy and supports immune tolerance towards the fetus by blocking maternal NK activity and altering the cytokine balance. Recently it was shown that PIBF is not only produced by maternal lymphocytes, but it is also present in pregnancy-associated tissues including trophoblast and in a set of malignant tumours. Trophoblast and tumour cells have much in common, among others, both are highly invasive. However, while trophoblast invasion is strictly regulated, tumour invasion is uncontrolled. Our aim was to investigate the role of PIBF in regulation of invasion. Materials and Methods: Cell invasion assay was used to determine the invasive behaviour of PIBF-silenced tumour (HT-1080, HCT116, A549, PC-3) and trophoblast (HTR-8/SVneo) cells. Cell-conditioned media from the invasion assay were subjected to gelatine zymography to measure matrix metalloproteinase (MMP) activity. To further confirm the role of PIBF in invasion, zebrafish (Danio rerio) embryos were microinjected with fluorescently-labelled PIBF-silenced trophoblast or tumour cells, and

Page 69

dissemination of invasive cells was monitored by confocal microscopy. Since STAT3 has a central role in invasion, the effect of PIBF treatment on STAT3-phosphorylation was tested by Western blotting. IL-6 and leptin are capable to activate STAT3, thus secreted IL-6 was measured in the supernatants of PIBF silenced cells by cytometric bead array and leptin-receptor expression was monitored by Western blotting in normal and PIBF knock down cells. Protein arrays were performed on PIBF-silenced and PIBF-treated lysates of trophoblast and tumour cells. To verify that PIBF is capable to bind to the promoter region of certain genes, chromatin immunoprecipitation (ChIP) was performed with anti-PIBF antibody. Results: In trophoblast cells, PIBF knock down resulted in increased MMP-2 and MMP-9 activity together with increased invasivity; while PIBF silenced tumour cells showed reduced levels of MMP-2 and MMP-9 thus decreased invasive behaviour. In case of microinjection with the PIBF-silenced trophoblast cells, significantly more cells were disseminated within the zebrafish embryo compared to the scrambled-injected embryos. PIBF-silenced tumour cells showed decreased invasive potential in zebrafish embryos compared to the scrambled-injected control. In tumour cells PIBF treatment resulted in late Ser- and Tyr-phosphorylation of STAT3, suggesting an indirect role of PIBF in STAT3 induction. PIBF inhibited Tyr-phosphorylation of STAT3 in trophoblast cells. IL-6 production and leptin-receptor expression increased in PIBF knock down trophoblast cells, while IL-6 production and leptin-receptor expression were downregulated in PIBF-silenced tumour cells, suggesting that PIBF differentially regulates IL-6 and leptin-receptor in trophoblast and tumour cells. Based on the protein array, PIBF induced the transcription of HB-EGF, PlGF, FGF1 and inhibited TIMP-1 in tumour cells; while PIBF inhibited the transcription of HB-EGF, PlGF, FGF1 and induced TIMP-1 in trophoblast cells. ChIP revealed that PIBF has the capacity to bind to the promoter of IL-6, EGF and FGF1 both in trophoblast and fibrosarcoma cells, however, the promoter-binding complex differed in composition. In tumor cells the protein/DNA complex included the full-length PIBF, in addition to the 50-kDa and 67-kDa PIBF isoforms found in the trophoblast. Conclusions: Our data allow the assumption that PIBF controls physiological trophoblast invasion by gene suppression; while in tumour cells PIBF promotes pathological invasion by gene activation (e.g. EGF, IL-6). The transcribed and secreted proteins bind then to their own receptors and induce phosphorylation of STAT3 which in turn further activates the transcription of invasion promoting molecules (e.g. MMPs). We hypothesize that the different composition of the DNA-binding PIBF complex might underlie the differential regulation of trophoblast and tumour invasion by PIBF.