The Chemical and Computational Biology of Inflammation

The chemical and computational biology of inflammation

A thesis submitted to the University of Manchester for the degree of Ph.D. in the Faculty of Engineering and Physical Sciences

2011

Ben Small

Doctoral Training Centre Integrative Systems Biology Molecules to Life within the School of Chemical Engineering and Analytical Science

Table of Contents

Title Page………………………………………………………………………………1 Table of Contents ...... 2 List of Figures ...... 3 Abbreviations and Symbols ...... 4 Abstract ...... 9 Lay Abstract – ‘Sifting drug combinations to treat disease’ ...... 10 Declaration ...... 11 Copyright Statement ...... 11 Acknowledgements ...... 12 Preface and Quotes ...... 14

Chapter 1 Introduction…………………………………………………………………. 15 1.1 Inflammation: a central problem for medicine...... 15 1.2 Interleukin-1; a pro-inflammatory cytokine ...... 17 1.2.1 The release of IL-1 ...... 18 1.3 Toll-like receptor 4 and the synthesis of IL-1β ...... 20 1.3.1 Consideration of the TLR4 signaling network ...... 21 1.5 Inflammatory transformation of tissues: an issue of robustness ...... 24 1.6 Systems biology a definition and its impact in innate immunity ...... 25 1.7 The Problem ...... 27 1.8 The Question ...... 28 1.9 An Introduction to evolutionary computation ...... 29 1.10 A biochemists guide to In silico modeling of biochemical systems ...... 32 1.11 Combinatorial solutions to complex problems? ...... 36 1.12 Aims and objectives ...... 36

Chapter 2…………………………………………………………………………………..38 Publication number: 01 Efficient discovery of anti-inflammatory small molecule combinations using evolutionary computing ...... 38

Chapter 3 Analysis of the LPS directed signaling network ………………………….39 3.1 Introduction………………………………………………………………. 39 3.2 Network map construction………………………………………………...40

Chapter 4…………………………………………………………………………………..41 Publication number: 02 A network map of lipopolysaccharide-dependent IL-1β expression in macrophages and monocytes ...... 41

Chapter 5 Discussion ...... 42 5.1. Introduction ...... 42 5.2 Efficient discovery of anti-inflammatory drug combinations using evolutionary computing ...... 42 5.3 A network map of lipopolysaccharide dependent IL-1β expression in macrophages and monocytes...... 46 5.4 Conclusion ...... 46 5.5 Back to the future ...... 47 5.6 Future work ...... 47 5.6.1 Mixed messages: drug efficacy…………………………………………………48

Bibliography ...... 50 Word count: 18351 2 List of Figures

Figure 1.1 A wordle generated from the extracted text of reviews on the science of 28 systems biology presented above. Figure 1.2 The basic principles behind the workings of evolutionary algorithms. 32 Figure 1.3 An idealized uni-molecular reaction detailing the conversion of species 34 A into B with corresponding mass fluxes ( Γ). Figure 1.4 Simulation of the uni-molecular reaction A = B (outlined in Figure 1.3) 37 performed in COPASI.

3 Abbreviations and Symbols ∆∆∆G Gibbs free energy ΓΓΓ Mass flux ∞∞∞ Infinity µµµ micro (10 -6 of quantity) µµµg microgramme µµµM micromolar µµµL microlitre ο C degrees celsius [nX] Heavy atom; where n is mass number and X is the element labelled

A xxxnm Absorbance measured at a given 'xxx' wavelength in nm A20 Tumor necrosis factor alpha-induced protein 3 Ab Antibody AIM2 Absence In Melanoma receptors Akt v-akt murine thymoma viral oncogene homolog AlogP Atomic contribution calculation of Partition coefficient AMS Academy of Medical Sciences ANOVA Analysis Of VAriance ASC Apoptosis-associated Speck-like protein containing a CARD ASK1 Apoptosis Signal regulating Kinase 1 ATP Adenine TriPhosphate BBSRC Biotechnology and Biological Sciences Research Council BMS345541 (4(2 ′-aminoethyl)amino-1,8-dimethylimidazo(1,2-a)quinoxaline) bZIP basic leucine ZIPper domain 2+ [Ca ]i Intracellular calcium concentration cAb Capture Antibody CAPS Cyropyrin Associated Periodic Syndromes CARD CAspase Recruitment Domain CAT ChlorAmphenicol Transferase CDX Cluster of Differentiation (where X is the subtype of this molecule) cDNA Complementary DeoxyriboNucleic Acid C/EBP-βββ CAAT Enhancer Binding Protein-beta ChEMBL a database of bioactive drug-like small molecules hosted by EMBL ChIP Chromatin ImmunoPrecipitation CLR C-type Lectin Receptors CIAS1 Cold Induced Autoinflammatory Syndrome 1 gene (Human) CIITA Class II, major histocompatibility complex, TransActivator gene (Human) CKII Casein Kinase II CO 2 Carbon dioxide COPASI Complex Pathway SImulator COPD Chronic Obstructive Pulmonary Disease CREBP cAMP Response Element Binding Protein CSBP Cytokine Supressive anti-inflammatory Binding Protein CSP CAP Site Proximal dAb Detection Antibody DAMP Danger Associated Molecular Patterns DECTIN-1 C-type lectin domain family 7, member A DMEM Dulbecco’s Modified Eagle Medium DMSO DiMethyl SulphOxide DNA DeoxyriboNucleic Acid

4 DTC Doctoral Training Centre - Integrative Systems Biology Molecules to Life DUSP1 / MKP-1 DUal Specificity Phosphotase 1 / MAPK Phosphotase 1 EA (Multi-Objective) Evolutionary Algorithm EGCG Epigallocatechin Gallate ELISA Enzyme Linked Immunosorbant Assay EMBL European Molecular Biology Laboratory EPSRC Engineering and Physical Sciences Research Council ERK Extracellular signal-Regulated Kinases FBS / FCS Foetal Bovine Serum / Foetal Calf Serum g times gravity (x g: units of relative centrifugal force) GA Genetic Algorithm GSK3 Glycogen Synthase Kinase 3 GST Glutathione-S-Transferase GTPase Guanine Tri Phosphotase h hours HBA Hydrogen Bond Acceptor HBD Hydrogen Bond Donor HETE HydroxyEicosaTetraEnoic acid HMGB1 High Mobility Group Box 1 HMG-CoA 3-hydroxy-3-methyl-glutaryl-CoA reductase reductase HSE Heat Shock Element HSF1 Heat Shock Factor 1 IBEA Indicator Based Evolutionary Algorithm IC 50 The concentration of a reagent at which the 50 % of the effect / response is inhibited. IFN-βββ Interferon-β IKKi BMS345541 (see abbreviation) IKKX Inhibitor of Kappa B Kinase (where the suffix X represents a subtype of this kinase) IL Interleukin IL-1 Interleukin-1 IL-6 Interleukin-6 IL-10 Interleukin-10 IL-1ααα Interleukin-1 Alpha (also ILF1) IL-1βββ Interleukin-1 Beta (also ILF2) IL1B Interleukin-1 Beta gene (Human) il1b Interleukin-1 Beta gene (Mouse) IL-1RA Interleukin-1 Receptor Antagonist IL-1RAcP Interleukin-1 Receptor Accessory Protein IL-1RI Interleukin-1 Receptor I IL-1RII Interleukin-1 Receptor II IL-F11 Interleukin-33 IRAKX Interleukin-1 Receptor Associated Kinase (where the suffix X represents a subtype of this kinase) IRF3 Interferon Regulated Factor 3 IRF4 Interferon Regulated Factor 4 IRF8 Interferon Regulated Factor 8 IκκκBααα Inhibitor of Kappa B alpha J Concentration flux JNK c-jun N-terminal Kinase k Rate constant kDa kiloDalton 5 K + Potassium ion LAF Leuckocyte Activating Factor LBP Lipopolysaccharide Binding Protein LDH Lactate DeHydrogenase LILRE LPS and IL-1 Responsive Element log 10 Logarithm to the base 10 LP Leukocytic Pyrogen LPS LipoPolySaccharide LRR Leucine Rich Repeats LY294002 2-morpholin-4-yl-8-phenylchromen-4-one MAL MyD88 Adaptor Like MAP2K Mitogen Activated Protein Kinase Kinase MAP3K Mitogen Activated Protein Kinase Kinase Kinase MAPK14 p38 α MAPK MAPK Mitogen Activated Protein Kinase MAPKK Mitogen Activated Protein Kinase Kinase MAPKKK Mitogen Activated Protein Kinase Kinase Kinase MCPIP1 Monocyte Chemotactic Protein–Induced Protein 1 MD-2 Lymphocyte antigen 96 MEKK3 Mitogen-activated protein kinase kinase kinase 3 min minute/s mM Millimolar mL Millilitre MKK Mitogen Activated Protein Kinase Kinase MKKK Mitogen Activated Protein Kinase Kinase Kinase MOOP Multi-Objective Optimization Problem mRNA messenger RiboNucleic Acid MSU MonoSodium Urate MyD88 Myeloid Differentiation primary response gene 88 n number of moles of substance (unless defined otherwise) NAC N-AcetylCysteine NACHT Domain present in NAIP, CIITA, HETE and TP1 NADPH oxidase Nicotinamide Adenine Dinucleotide Phosphate oxidase NALP3 NACHT-, LRR- and Pyrin-domain-containing protein 3 NAIP NLR family Apoptosis Inhibitory Proteins NCD Non communicable disease NEMO NF-κ B essential modulator NF-κκκB Nuclear Factor- kappa-light-chain-enhancer of activated B cells NIK NF-κB inducing kinase NLR Nod Like Receptor NLRP Nod Like Receptor Protein nm nanometres NOD Nucleotide-binding Oligomerisation Domain NSGA2 Nondominated Sorting Genetic Algorithm 2 O2 Oxygen ODE Ordinary Differential Equation

P2X 7 The ionotropic purinergic subtype 7 receptor PAMP Pathogen Associated Molecular Pattern PBS Phosphate Buffered Saline PDTC Pyrrolidine dithiocarbamate PGE 2 Prostaglandin E2 PIP PhosphatidylInositol Phosphate PI3K Phosphatidyl Inositol-3-Kinase

6 PI3K δδδ Phosphatidyl Inositol-3-Kinase delta PKC εεε Protein Kinase C epsilon PMA Phorbol 12-Myristate 13-Acetate PSA Polar Surface Area PTP4A3 Protein Tyrosine Phosphatase type IVA, member 3 PYD Pyrin Domain Pro-IL-1ααα Pro-Interleukin-1 Alpha Pro-IL-1βββ Pro-Interleukin-1 Beta Pro-IL-18 Pro-Interleukin-18 PRR Pattern Recognition Receptors PU.1 PUrine box binding protein 1 QC Quality Control RAE Royal Academy of Engineering RAGE Receptor for Advanced Glycation End products RIG-1 Retinoic acid-inducible gene 1 RING Really Interesting New Gene RLR Retinoic acid-inducible gene 1 like receptors ROS Reactive Oxygen Species RO5 Rule Of Five rpm revolutions per minute RT Room Temperature RT-PCR Real Time Polymerase Chain Reaction SARM Sterile α - and ARmadillo-Motif-containing protein SB203580 4-[4-(4-fluorophenyl)-2-(4-methylsulfinylphenyl)-1H-imidazol-5- yl]pyridine SD Standard Deviation SEM Standard Error of the Mean SIGRR Single IgG IL-1 Related Receptor SIH salicylaldehyde isonicotinoyl hydrazone Simvastatin (1 S,3 R,7 S,8 S,8a R)-8-{2-[(2 R,4 R)-4-hydroxy-6-oxotetrahydro-2H- pyran-2-yl]ethyl}-3,7-dimethyl-1,2,3,7,8,8a-hexahydronaphthalen- 1-yl 2,2-dimethylbutanoate SKI-II 4-[4-(4-chloro-phenyl)-thiazol-2-ylamino]-phenol SmadX conjunction of SMA protein and mothers against decapentaplegic (X represents the subtype of protein) SPEA2 Strength Pareto Evolutionary Algorithm 2 SRP System Response Profile ST2L Interleukin 1 receptor-like 1 sTLR4 soluble TLR4 Strep-HRP Streptavidin-Horse Radish Peroxidase t Time TABX TAK1 binding protein (where X is the subtype of protein) TAK1 Transforming growth factor-β Associated Kinase 1 TBD To Be Determined TBK1 Transforming Growth Factor (TGF) β -activated kinase 1 TBP TATA Box Protein TF Transcription Factor TFIID Transcription Factor IID TIFAX TRAF-Interacting protein with a Forkhead-Associated domain (where X represents the subtype of this protein) TIR Toll/Interleukin-1 Receptor (usually in the context of a protein domain) TLRX Toll Like Receptor (where X denotes a number indicating the

7 receptor sub-type) TMB 3, 3', 5, 5' tetramethyl benzidine TNF-ααα Tumour Necrosis Factor alpha Tollip Toll interacting protein TP1 Transistion Protein 1 TPEN N,N,N',N'-Tetrakis-(2-pyridylmethyl)ethylenediamine TRAF6 Tumor necrosis factor Receptor-Associated Factor 6 Triad3A E3 ubiquitin ligase RING finger protein TRIKA1 TRAF-6 regulated IKK activator 1 TICAM1/2 TIR Containing Adaptor Molecule 1/2 TIRP TIR domain containing Protein TIRAP TIR domain containing Adaptor Protein TRAM TRIF related adaptor molecule TRIF TIR domain containing adaptor protein inducing Interferon-β tSRI Temporal System Response Indicator U0126 1,4-diamino-2,3-dicyano-1,4-bis(2-aminophenylthio)butadiene Ubc1A / UBC13 E2 ubiquitin conjugating complex UIS Upstream Induction Sequence VK-28 (5-[4-(2-hydroxyethyl) piperazine-1-ylmethyl]-quinoline-8-ol) V Volume v / v volume for volume (expressed as a percentage) wHTH winged Helix Turn Helix domain wt Wild Type w / v weight for volume (expressed as a percentage) YFP Yellow Fluorescent Protein ZCCHC11 zinc finger, CCHC domain containing 11

The University of Manchester

Benjamin Gavin Small

Doctor of Philosophy (PhD)

The chemical and computational biology of inflammation

Monday, January 23, 2012

Abstract

Non-communicable diseases (NCD) such as cancer, heart disease and cerebrovascular injury are dependent on or aggravated by inflammation. Their prevention and treatment is arguably one of the greatest challenges to medicine in the 21 st century. The pleiotropic, proinflammatory cytokine; interleukin-1 beta (IL-1β) is a primary, causative messenger of inflammation. Lipopolysaccharide (LPS) induction of IL-1β expression via toll-like receptor 4 (TLR4) in myeloid cells is a robust experimental model of inflammation and is driven in large part via p38-MAPK and NF-κB signaling networks. The control of signaling networks involved in IL-1β expression is distributed and highly complex, so to perturb intracellular networks effectively it is often necessary to modulate several steps simultaneously. However, the number of possible permutations for intervention leads to a combinatorial explosion in the experiments that would have to be performed in a complete analysis. We used a multi-objective evolutionary algorithm (EA) to optimise reagent combinations from a dynamic chemical library of 33 compounds with established or predicted targets in the regulatory network controlling IL-1β expression. The EA converged on excellent solutions within 11 generations during which we studied just 550 combinations out of the potential search space of ~ 9 billion. The top five reagents with the greatest contribution to combinatorial effects throughout the EA were then optimised pairwise with respect to their concentrations, using an adaptive, dose matrix search protocol. A p38 α MAPK inhibitor (30 ± 10% inhibition alone) with either an inhibitor of I κB kinase (12 ± 9 % inhibition alone) or a chelator of poorly liganded iron (19 ± 8 % inhibition alone) yielded synergistic inhibition (59 ± 5 % and 59 ± 4 % respectively, n=7, p ≤0.04 for both combinations, tested by one way ANOVA with Tukey’s multiple test correction) of macrophage IL-1β expression. Utilising the above data, in conjunction with the literature, an LPS-directed transcriptional map of IL-1β expression was constructed. Transcription factors (TF) targeted by the signaling networks coalesce at precise nucleotide binding elements within the IL-1β regulatory DNA. Constitutive binding of PU.1 and C/EBP-β TF’s are obligate for IL-1β expression. The findings in this thesis suggest that PU.1 and C/EBP-β TF’s form scaffolds facilitating dynamic control exerted by other TF’s, as exemplified by c-Jun. Similarly, evidence is emerging that epigenetic factors, such as the hetero-euchromatin balance, are also important in the relative transcriptional efficacy in different cell types. Evolutionary searches provide a powerful and general approach to the discovery of novel combinations of pharmacological agents with potentially greater therapeutic indices than those of single drugs. Similarly, construction of signaling network maps aid the elucidation of pharmacological mechanism and are mandatory precursors to the development of dynamic models. The symbiosis of both approaches has provided further insight into the mechanisms responsible for IL-1β expression, and reported here provide a platform for further developments in understanding NCD’s dependent on or aggravated by inflammation 9

The University of Manchester

Benjamin Gavin Small

Doctor of Philosophy (PhD)

The chemical and computational biology of inflammation

Monday, January 23, 2012

Lay Abstract – ‘Sifting drug combinations to treat disease’ A computer program which can select effective drug combinations from a billion others is set to improve our understanding of cancer, heart disease, stroke and many other diseases.

Inflammation-best known in arthritis, is common to many chronic diseases. Inflammation causes the release of a whole array of molecules that help our bodies fight infection but can also be very damaging in long term diseases. A key molecule called IL-1 is produced by blood cells called macrophages and these cells can be studied in the laboratory. Only 30 drugs are required to generate over a billion possible combinations, so it would take over 100 years to test all possible combinations even with the assistance of the lab robot! We bypassed this problem with a computer program: an evolutionary algorithm (EA) that compared the results of a few combinations that we tested in the lab, and selected those that were most effective in blocking IL-1 production. We repeated this cycle until the best combination was found, but in 10 weeks rather than 100 years! (see Figure) Although the promise of future remedies is a distant one, using this approach, of computer program with experiments will vastly speed up the search for remedies and provide a better understanding of chronic diseases such as cancer, cerebrovascular injury and heart disease.

Figure. The search for new drug combinations inhibiting the production of an inflammatory protein. In this graphic the sieve represents how the computer program chooses good (inhibitory) combinations through repeated rounds of combination testing and selection .

10 Declaration

No portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning:

i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes.

ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made.

iii. The ownership of certain Copyright, patents, designs, trade marks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions.

iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://www.campus.manchester.ac.uk/medialibrary/policies/intellectual- property.pdf ), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.manchester.ac.uk/library/aboutus/regulations ) and in The University’s policy on presentation of Theses

11 Acknowledgements

I would like to thank my project supervisors; Dr. David Brough and Dame Prof. Nancy Rothwell FRS of the Faculty of Life Sciences and Prof. Pedro Mendes (School of Computer Science) and Prof. Douglas Kell (School of Chemistry) of the Faculty of Engineering and Physical Sciences for giving me the opportunity to pursue this line of questioning in my thesis and providing invaluable direction, encouragement, patience, support and advice in what I hope has provided a relevant answer.

Equally, I should like to thank Dr. Barry McColl (now at the Roslin Institute), Dr. Nadia Luheshi and Dr. Gloria López-Castejón (Faculty of Life Sciences) and Dr. Jürgen Pahle (School of Computer Science) for their generosity in terms of time, coaching and advice in all matters related to biological research and computational biology research respectively.

Many thanks also go to Mr. Richard Allmendinger and Dr. Joshua Knowles of the School of Computer Science who have been unstinting in giving their time and expertise in formulating and implementing a evolutionary algorithm that has enabled us to locate a evolutionarily derived solution to combinatorially inhibiting IL-1βexpression.

Prof. Gerold Baier, Deputy Director of Doctoral Training Centre - Integrative Systems Biology Molecules to Life (DTC) is particularly worthy of mention, in providing generously of his time in explaining the logic behind the mathematics taught and applied within Systems Biology and in instilling the confidence in myself to tackle these mathematical problems. I would also like to thank the staff (particularly Prof. Hans Westerhoff, Dr. Mara Nardelli, Dr. Hanan Messiha and Mrs. Lynne Davies) and students of the DTC and the Manchester Centre for Intergrative Systems Biology at the University of Manchester. They have provided help, encouragement and interesting and thought- provoking insights into the field of systems biology.

One of my first scientific mentors' deserves a mention. Dr. W.R.Robertson (formerly of this University) taught me how to write scientifically and reigned in my naturally verbose style. Any decay in the intervening years has been all of my own `letting go' in his exemplary standard.

12 I should also like to thank Prof. Andy Brass (University of Manchester) and Dr. Clare Bryant (University of Cambridge) for conducting a thought-provoking examination of this thesis dissertation that has allowed me to make minor corrections to its final format.

Finally, my immediate and extended family have been continuously supportive, particularly my long suffering wife - even during the stressful period prior to beginning my studies where my indecision about whether to leave my employment filled a two year period. Without them, none of this would have been possible.

13 Preface and Quotes

I was minded of the words of the famous pharmacologist Sir Jack Gaddum who in an article outlined the then achievements of pharmacology and its likely future directions. (Gaddum, 1954). Within this forward looking précis he eludes to the interdisciplinarity required for advancement in the field of pharmacology. I think this has particular resonance within systems biology (and has recently been echoed (Dollery et al., 2007); (Sharp et al., 2011)) which is undoubtedly the most recent effort to provide an integrated approach to a mechanistic biology.

“Believe nothing, no matter where you read it, or who said it, no matter if I have said it, unless it agrees with your own reason and your own common sense.”

Hindu Prince Gautama Siddharta, the founder of Buddhism, 563-483 B.C.

The scientist has a lot of experience with ignorance and doubt and uncertainty, and this experience is of very great importance, I think. When a scientist doesn’t know the answer to a problem, he is ignorant. When he has a hunch as to what the result is, he is uncertain. And when he is pretty damn sure of what the result is going to be, he is still in some doubt. We have found it of paramount importance that in order to progress, we must recognize our ignorance and leave room for doubt. Scientific knowledge is a body of statements of varying degrees of certainty — some most unsure, some nearly sure, but none absolutely certain. Now, we scientists are used to this, and we take it for granted that it is perfectly consistent to be unsure, that it is possible to live and not know. But I don’t know whether everyone realizes this is true. Our freedom to doubt was born out of a struggle against authority in the early days of science. It was a very deep and strong struggle: permit us to question — to doubt — to not be sure. I think that it is important that we do not forget this struggle and thus perhaps lose what we have gained.

Richard Feynman "The Value of Science," address to the National Academy of Sciences (Autumn 1955)

“Three months in the lab, can save you a whole afternoon in front of the computer”

Douglas B Kell (http://blogs.bbsrc.ac.uk/)

And finally, I am an aspirant of the sentiment expounded in this verse.

“…What in me is dark Illumin, what is low raise and support; That to the highth of this great Argument I may assert Eternal Providence, And justifie the wayes of God to men”.

John Milton “Paradise Lost”, Book 1, 22 – 26 (1674).

14 Introduction

1.1 Inflammation: a central problem for medicine

Understanding the inflammatory process is perhaps, the critical challenge for medicine within the 21st century. Inflammation is an adaptive, physiological and hence protective response of cells and tissues to noxious chemical or physical stimuli; danger or pathogen associated molecular patterns (DAMP's or PAMP's; e.g. viral particles, virulence factors, bacterial lipopolysaccharides or fungal membrane peptides or proteins) and tissue injury (Weiss, 2008, Medzhitov, 2008, Medzhitov, 2010). Inflammation can often be described as deleterious when reflecting on the time taken for the condition to develop, where often identification of pathology precedes that of a corresponding physiological inflammatory episode that successfully resolves, if indeed these exist at all. A notable exception to this includes the wound healing and repair response.

Inflammatory responses occur over two ill-defined timescales; acute and chronic (Mosser and Edwards, 2008). Acute inflammation resulting from an initial infection (e,g, PAMPs) or injury (e.g. release of sterile particulates and intracellular cytokines from necrotic cells) is characterised by the activation of tissue resident macrophages via pattern recognition receptors (PRR), specifically; toll-like receptors (TLR), nucleotide-binding oligomerisation domain (NOD)-like receptors, nucleic acid sensor or RIG-1 (retinoic acid-inducible gene 1)-like receptors (RLR), absence in melanoma receptors (AIM2) and C-type lectin receptors (CLR) (Barton, 2008, Chen and Nunez, 2010, Iwasaki and Medzhitov, 2010). This leads to the subsequent generation of a variety of signaling molecules (e.g. cytokines, pro-inflammatory mediators, chemokines and eicosanoids inter alia ). These mediators affect the integrity of post capillary venules also causing their vasoconstriction, whilst vasodilitatory effects are seen in pre capillary beds. Additionally, these signaling molecules recruit leukocytes (e.g. neutrophils and monocytes) to the endothelial surface of the capillaries proximal to the site of injury or infection, allowing a 'staging post' for the migration of these cells to this site. The combination of increased hydrostatic pressure within the capillary bed and its now porous nature causes an exudate of leukocytes and proteins of the complement system to the inflamed area resulting in redness and oedema of the tissue or organ. Continued swelling owing to this leukocyte invasion of the tissue results in local increases in temperature, leading to pain and potential loss of function of the tissue or organ if left untreated. However, in the majority of acute inflammatory conditions, resolution is realised via the recognition and phagocytosis of apoptotic neutrophils by monocytes and macrophages, with subsequent phagocyte mediated

15 secretion of anti-inflammatory mediators (i.e. interleukin-10 (IL-10). prostaglandin E2

(PGE 2)) that aid repair (Murphy, 2007, Soehnlein and Lindbom, 2010). The activation of macrophages by versican (a proteoglycan found in the extracellular matrix and up- regulated in tumours) via TLR2 receptors results in their expression of cytokines (e.g. interleukin-6 (IL-6) and tumour necrosis factor-alpha (TNF-α)) that contribute to tumour metastasis (Mantovani, 2009, Kim et al., 2009). However, pro-inflammatory cytokines and particularly interleukin-1 (IL-1) (Dinarello, 2011) are implicated in a variety of chronic and therefore non-resolving inflammatory conditions such as atherosclerosis, obesity, chronic obstructive pulmonary disease, asthma, inflammatory bowel disease, neurodegeneration and rheumatoid arthritis (Nathan and Ding, 2010). These non-resolving inflammatory conditions appear to be driven by an intersection between tissue injury and subsequent exposure to microbial signals (i.e. infection) or extensive tissue damage that the host interprets as being part microbial (i.e. sterile inflammation) (Nathan and Ding, 2010, Chen and Nunez, 2010). Despite the maturity of a field whose recognition transcends and is blurring its boundaries, a summary of current thinking with respect to the genesis and pathophysiology of inflammation is in order.

The maxim 'you are what you eat' although clichéd is no truer of macrophages whose phenotype changes in response to not only extracellular mediator signals but also the material they phagocytose. Mosser and Edwards (Mosser and Edwards, 2008) suggest that the binary classification of macrophage populations as classically or alternatively activated is too simplistic (Gordon, 2003) and suggest viewing macrophage activation in terms of a functional spectrum encompassing wound healing, host-defence and immune response. The macrophages used in the studies presented herein are heterogenous populations, likely consisting of both classic and alternatively activated macrophages. The ‘functional spectrum’ concept becomes evident when one considers the macrophage and dendritic cell precursor population (MDP) that developmentally transit via monocytes to form macrophages (Auffray et al., 2009). The potential combinatorial complexity of differentially expressed markers used in the classification of human monocytes (e.g. (cluster of differentiation (CD)14, CD16, CD64 and CD32 positive cells are responsible for lipopolysaccharide (LPS)-dependent production of pro-inflammatory cytokines; (Auffray et al., 2009)) is large. Similarly, the spectrum of functions that macrophages embrace (i.e. host defence, wound healing and immune regulation) can be superimposed onto a colour wheel, thus capturing the subtler aspects of macrophage activation and their unique physiologies. Secreted mediators inform macrophage and target tissue physiological function, controlling the macrophage mediated repair of tissue damage.

16 Conversely, chronic inflammation is depicted by a prolonged, poorly regulated damage to tissue with periodic attempts at tissue repair. Causative principles of chronic systemic inflammatory processes (or as Medzhitov labels this - 'para-inflammation') that switch tissue function from the physiological to pathological is the aberrant synchronisation and placement of mediator secretion leading to tissue injury. The underlying development of chronic inflammation has recently been reviewed (Renz et al., 2011) and increasingly epigenetic factors (i.e. deoxyribonucleic acid (DNA) modifications, e.g. methylation inter alia ) are thought to play a role in linking an individuals genetic susceptibility and their environment. Thus understanding the biology of these inflammatory mediators and in particular the cytokines involved is a key focus for beginning a search for the elusive 'smoking gun' of chronic systemic inflammation.

1.2 Interleukin-1; a pro-inflammatory cytokine

Cytokines are responsible for orchestrating host-defence responses to injury and infection with their expression up-regulated in disease (Dinarello, 1996, Steinman, 2008). The interleukin-1 family is composed of 11 members (i.e. ILF1 – ILF11) (Dinarello, 2011) and interleukin-1 alpha (IL-1α) and interleukin-1 beta (IL-1β) are regarded as `proto-typical' pro-inflammatory cytokines. The measurement and description of IL-1β undertaken in this thesis refers to both pro and mature forms unless explicitly stated otherwise. IL-1β concentrations are elevated in human plasma (albeit to lower levels than IL-6 or TNF-α) in chronic systemic inflammatory states (Dinarello, 1996) (Rothwell and Luheshi, 2000). The discovery of IL-1 is the story of the hunt for the causative principle of endotoxin induced fever which was initially labeled ’endogenous pyrogen’ and discovered by Elisha Atkins in 1955 (Dinarello, 2010). Bacterial endotoxin and particularly LPS derived from gram negative bacteria are potent inducers of IL-1 cytokines from the monocytes / macrophages which are primarily responsible for driving auto-inflammatory syndromes (Dinarello, 2011) such as Cyropyrin Associated Periodic Syndromes (CAPS) (McDermott and Tschopp, 2007). The realisation that macrophages were primary cellular factories for IL-1β was first uncovered in 1972 by Gery and Waksman (Gery and Waksman, 1972). They demonstrated that splenically derived mouse cells (that consisted largely of macrophages) produced supernatants in response to LPS that stimulated the incorporation of [ 3H]-Thymidine (a surrogate measure of mitosis) into thymocytes. Similarly, removal of macrophages from human leukocyte preparations destroyed their ability to produce these active supernatants (Gery and Waksman, 1972). However, it was not until thirteen years later that that the complementary DNA (cDNA) for human IL-1α and β isoforms 17 was cloned into Xenopus oocytes and the supernatants shown to stimulate IL-2 release from a murine lymphoma cell line stimulated with phytohaemagglutin, an assay reported to be 1,000 times more sensitive for IL-1 activity than the thymocyte mitogenesis assay used by Gery and Waksman (March et al., 1985).

Physiological roles for IL-1 have been difficult to establish owing to its low expression, although in addition to its pyrogenic properties, neurologically it's thought to impair long- term potentiation (a proposed cellular correlate of learning and memory) and influence sleep patterns (Steinman, 2008, Rothwell and Luheshi, 2000). Both molecules are synthesised as precursor forms in the cytosol. Pro-IL-1α remains intracellularly associated with microtubules and chromatin and is converted to its mature 17 KDa form by the calpain family of calcium-dependent proteases. Often pro-IL-1β is released during necrotic cell death and converted into its mature form by extracellular proteases secreted from neutrophils (e.g. particularly proteinase 3, but also elastases and chymases) (Dinarello, 2011). IL-1α has been postulated as an autocrine growth factor being involved in the differentiation of cells. A number of lines of evidence point to this, its association with chromatin and the existence of a nuclear localisation sequence in its N-terminal domain that engages in transcription (Luheshi et al., 2009). In contrast, IL-1β is as a paracrine signaling molecule which is released from macrophages and extensively negatively regulated via competitive interactions with Interleukin receptor antagonist (IL- 1RA) to its cognate receptor; interleukin receptor I (IL1-RI).

Transduction of the IL-1β and IL-1α response via IL-1RI is mediated by co-association of the receptor with the interleukin-1receptor accessory protein (IL-1RAcP). Similarly, decoy receptors termed IL-1RII and SIGIRR (single IgG IL-1 related receptor; which also acts on TLR4 dependent signals) (Dinarello, 2011) both sequester IL-1β from the circulation and in the former instance also IL-1RAcP from the membrane. Differential expression of these receptors regulates the inflammatory response.

1.2.1 The release of IL-1

The intracellular cascade of events leading to release of IL-1β involves the assembly of a multiprotein complex: the inflammasome. Signals that mediate the processing and release of IL-1β have been reviewed for bacterial infection (Mariathasan and Monack, 2007) and for certain inherited autoimmune disorders, such as the cryopyrin associated periodic

18 syndromes (CAPS) (Ogura et al., 2006). These syndromes are of interest as they are distinct from other autoimmune disorders (e.g. rheumatoid arthritis) in that affected individuals don't display high titres of auto-antibodies or antigen specific T-cells (McDermott and Tschopp, 2007) and have been termed auto-inflammatory (Dinarello, 2011). These hereditary syndromes are associated with mutations in the cold induced autoinflammatory syndrome 1 ( CIAS1 ) gene which encodes a NACHT-, LRR- and Pyrin- domain-containing (NALP3) protein 3. The NACHT (domain present in NAIP, CIITA, HETE and TP1 (see abbreviations) ) domains within NALP3 mediate NALP3 oligomerization which exposes previously hidden pyrin domains (PYD) that in turn, via homotypic protein - protein interactions, complex with an adaptor molecule: apoptosis- associated speck-like protein containing a CARD (ASC). This is the final step in recruiting pro-caspase-1 via its caspase recruitment domain (CARD) prior to its cleavage and activation. Structurally, this complex unit forms the 'inflammasome' and once activated, caspase 1 is capable of cleaving pro-IL-1β and pro-IL18 resulting in the release of mature IL-1. Thus, mutations in CIAS1 result in the spontaneous oligomerisation of NALP3 and its constitutive activation of the chain of events that release IL-1β.

Understanding events downstream of the inflammasome is aided by these CAPS which provide interesting parallels with sterile inflammation, albeit in the absence of an endogenous stimulus. During sterile inflammation which can arise from tissue injury, trauma and ischaemia, necrotic cell death results in the release of DAMPs that include heat shock proteins, hyaluronan, adenine triphosphate (ATP), monosodium urate (MSU) and the high mobility group box 1 (HMGB1) ligand (McDermott and Tschopp, 2007), (Rock and Kono, 2008). However, confirming a 'pure' sterile inflammatory episode is hard to achieve, because minor infections normally contaminate the inflammatory background. Interestingly, in a murine model of ischaemic stroke HMGB1 is thought to signal via a receptor for advanced glycation end products (RAGE) enhancing ischaemic brain damage via inflammation. Addition of an anti-HMGB1 antibody or HMGB1 box A (an antagonist of RAGE) decreased ischaemic brain damage (Muhammad et al., 2008). The complex assembly and activation of the inflammasome reveals a tight regulation of its function; that is the unit acts as a coincidence detector (i.e. two inputs or more are required in order to achieve an output). The first signal is responsible for expression of the cytokine, detected via activation of the TLR / NLR network and the secondary signal which often involves a + depolarising ion flux (e.g. K efflux via P2X 7) necessary for release (Mariathasan and Monack, 2007). However, stimuli that trigger the synthesis of IL-1 cytokines via their

19 interactions with toll-like receptors are of particular interest in understanding inflammation.

1.3 Toll-like receptor 4 and the synthesis of IL-1βββ

TLR’s akin to other PRR’s have evolved to recognise the penetrance of microbial pathogens (via PAMP’s) across host immune barriers (e.g. the skin and mucous membranes). Classes of TLR respond to a variety of PAMP and DAMP signals. TLR’s 1, 2 , 4, 5, 6 and 11 detect lipids, proteins and lipoproteins of bacterial, fungal and viral origin and are localised at the plasma membrane in contrast to TLR3, 7, 8 and 9 that are nucleic acid sensing and are selectively distributed across intracellular compartments (e.g. endoplasmic reticulum, endosomes and lysosomes) to monitor non-self signals and prevent auto-inflammatory events from occurring (Kawai and Akira, 2010). LPS-induced TLR4 stimulation of macrophages is a widely used experimental model (Lu et al., 2008) that mimics key aspects of inflammation including IL-1β expression (Hsu and Wen, 2002, Matsuzawa et al., 2005). Nevetheless, the expression of IL-1β mRNA has been demonstrated in primary murine (C3H/OuJ) macrophages using Pam3Cys, a TLR2 agonist. However, it appears that TLR4 agonism triggers a wider range of responses in C3H/OuJ macrophages via STAT1 phosphorylation and subsequent transcription of MCP-5, iNOS and IP-10 (Toshchakov et al., 2002). The recognition of LPS by TLR4 was demonstrated by Poltorak (Poltorak et al., 1998) who showed that an LPS unresponsive strain of mouse; C3H/HeJ was homozygous for the co-dominant Lps d allele (heterozygotes showing an impaired LPS response). Additionally, the Lps allele mapped to the Tlr4 locus , strongly suggesting that the encoded proteins were one and the same (Poltorak et al., 1998).

TLR4-dependent signaling is described by a convoluted and sequential assembly of proteins beginning with the titrating out of circulating LPS by the soluble protein; LPS- binding protein (LBP). This interaction facilitates the binding of the LPS-LBP complex to either the glycosylphosphatidylinositol-anchored protein CD14 or its soluble counterpart, which in turn facilitates recruitment of either lymphocyte antigen 96 (MD2) alone or a TLR4-MD2 dimer (Bryant et al., 2010). The evolutionary maintenance of these proteins suggests that the LPS signal and unsurprisingly by proxy, that bacterial infection is sufficiently noxious to warrant their co-ordinated expression. Dimerisation of TLR4 facilitates the binding of adaptor proteins via the toll/interleukin-1 receptor (TIR) domain to the complex. There are five TIR domain containing proteins or adaptors (highlighted in bold in the following text) that are relevant to TLR4 mediated signaling, these include: myeloid differentiation primary response gene 88 ( MyD88); TIR domain containing 20 adaptor (TIRAP) also often called MyD88 adaptor like (MAL) which is essential for the MyD88 dependent production of pro-inflammatory cytokines (Yamamoto et al., 2002). IL-1 is capable of autocrine signaling, acting via this MyD88 dependent route providing positive feedback that probably enhances the inflammatory process. TIR domain containing adaptor protein inducing interferon-β (TRIF) also called (TICAM -1) which couples TLR3 / TLR4 heterodimers to MyD88-independent signaling pathways mediating the inhibitory kappa kinase-epsilon (IKK ε) and transforming growth factor (TGF) β- activated kinase 1 (TBK1) activation of the interferon regulated factor 3 (IRF-3) transcription factor that leads to IFN-β induction (Yamamoto et al., 2003). Similarly, this MyD88 independent pathway appears to be important for IL-1β signaling in neurones via the protein kinase; v-akt murine thymoma viral oncogene homolog (Akt) which has anti- apoptotic effects (Kenny and O'Neill, 2008) (Davis et al., 2006). Other adaptor proteins also play an important role in signaling; TRIF related adaptor molecule (TRAM) or TIRP (TIR domain containing adapter protein) when over-expressed activates the transcription factor; nuclear factor- kappa-light-chain-enhancer of activated B cells (NF-κB), but unlike TRIF, it does not produce interferon-β (IFN-β), TRAM is also critical to the TLR4 mediated transduction of MyD88 independent signaling leading to IL-1RI activated kinase (IRAK1) (Kenny and O'Neill, 2008). Finally, to date, the sterile - and armadillo-motif- containing protein (SARM ) has a contentious role in TLR signaling with some positing that SARM negatively regulates TRIF and others who contend that it plays no role in TLR signaling (Kenny and O'Neill, 2008).

1.3.1 Consideration of the TLR4 signaling network

Oda and Kitano (Oda and Kitano, 2006) have manually mapped the trans-cellular TLR network from 411 literature sources involved in innate and longer term adaptive immunity (Oda and Kitano, 2006). They suggest that understanding the logic of TLR directed immune circuitry and its phenotypic consequences are arguably a crux issue for the comprehensive understanding of immunological regulation. A prominent sub-system identified within the network is the MyD88-IRAK4-IRAK1-TRAF6 system that forms the 'tie' of the 'bow tie' motif that in turn signals to nuclear-factor kappa-B (NF-κB) and mitogen activated protein kinase (MAPK) cascades activating transcription of cytokine targets such as IL-1. Conversely, guanine nucleotide triphosphatase (GTPase) and phosphatidylinositol phosphate (PIP) signaling modules are considered separate from the MyD88 signaling protein core as they are capable of integrating inputs as described above

21 and sending signals to NF-κB and MAPK systems downstream of this protein. This is an appropriate juncture to move away from the network level to examine the modular complexities of TLR mediated NF-κB and MAPK signaling cascades.

Numerous studies point to the dependence of cytokine expression on NF-κB and p38 mitogen activated protein kinase (MAPK) cascades in a variety of tissues including BV2 microglia (Jung et al., 2009), rat derived microglia (Xu et al., 2009), murine bone marrow derived neutrophils (Asehnoune et al., 2004), bovine articular chondrocytes (Mendes et al., 2003), J774A.1 macrophages (Hsu and Wen, 2002), RAW264.7 macrophages, splenocytes and dendritic cells (Matsuzawa et al., 2005).

The NF-κB cascade shows well established oscillatory behaviour that in turn switches NF- κB mediated gene transcription on and off with inhibitors of kappa B (IκB; isoforms α, β, and ε) tethering the common p50/p65 NF-κB heterodimer in the cytoplasm thus inhibiting its activity. Stimulation of the NF-κB cascade occurs via receptor dependent (e.g. TLR4) phosphorylation of the IκB moiety by IκB kinases (IKK), which are associated with the scaffold protein; NF-κB essential modulator (NEMO) (O'Neill, 2006). After phosphorylation IκB is targeted for ubiquitination and degradation. This allows the p50/p65 heterodimer to translocate to the nucleus and associate with the promoter regions of target genes such as IκB, which in turn promotes the dissociation of NF-κB from DNA binding elements and its relocation to the cytoplasm. Hence, repeated oscillatory cycles can be established that in part are regulated via the differential assembly of structural components (e.g. NF-κB dimer composition, promoter sequences) and provide specific and selective spatio-temporal patterns of gene expression within the cell. TRAM dominant negative mutants transfected into 293 cells and coupled to NF-κB luciferase reporter constructs whilst over-expressing either IL-1 receptors, TRAF6, IKK β, TLR2 or TLR4 showed significant inhibition of NF-κB activation by IL-1RI (and it's accessory protein) but only marginal inhibition was observed for TLR2 over-expressing cells (Bin et al., 2003). However, this experiment must be viewed with caution given that there was no comparison made with wild-type TIRP and 293 cells do not endogenously express this protein. Bin et al, (Bin et al., 2003) measured luciferase catalysed chemiluminescence 16 h after mutant TRAM transfection of 293 cells and therefore TLR4 mediated MyD88 dependent effects appear to be independent of TRAM at this time. This result indicates that TLR4 can activate NF-κB in the absence of TRAM. This maybe a compensatory effect of MyD88 as Kenny and O'Neil (Kenny and O'Neill, 2008) point out that TRIF /

22 TRAM and MyD88 / MAL are thought to be involved in late versus early TLR4 signaling respectively.

The p38 MAPK's are serine / threonine kinases and at least in their interactions with apoptosis signal regulating kinase (ASK1) are considered evolutionarily older than the NF- κB cascade (Matsuzawa et al., 2005). The considered opinion on p38 MAPK's are that they are detectors of cellular stress (Raman et al., 2007). The p38 family consists of α, β, γ, and δ subtypes and are parts of a larger family of MAPKs including extracellular signal- regulated kinases (ERKs) and c-Jun N-terminal Kinases (JNKs). Gamma and delta p38 isoforms catalyse an initial phosphorylation of protein kinase C epsilon (PKC ε) resulting in this becoming a substrate for glycogen synthase kinase 3 (GSK3) after which further phosphorylation by GSK3 renders PKC ε able to interact with 14-3-3 proteins for further downstream signaling events. These are TLR4 and MyD88 dependent events as demonstrated by MyD88 gene knockouts where PKC and 14-3-3 phosphorylation were markedly reduced. Also, PKC phosphorylation of TRAM is also required for TLR4 transduction to the transcription factor; IRF3 via TRIF (Kenny and O'Neill, 2008).

Returning to the network, Oda and Kitano mention that as MyD88 is the central element of the 'bow-tie' structure, diverse inputs that converge on this protein can only change its level of activation and in principle cannot produce differential responses (Oda and Kitano, 2006). However, this observation and the fact that the network furnishes varying outcomes dependent on its combinatorial input can be reconciled by their invocation of the non- obvious notion of mathematical hyperspace. This concept is most easily conceived in terms of Cartesian x, y and z co-ordinates, where, for example each axis corresponds to a signaling protein and the relative distance along these axes representing the level of activation / inhibition of the protein. Discrete levels of network response can be mapped onto this three dimensional space and this idea can be extended to encompass N- dimensional (i.e. many proteins or signaling modules) hyperspace, elegantly describing different response outcomes to diverse inputs. In essence, signal transduction networks can be thought of classifiers (as Oda and Kitano describe) that integrate diverse inputs and categorise them in hyperspace which in turn produce a set of responses (Oda and Kitano, 2006). Clearly, the interactions between MyD88 dependent and collateral and MyD88 independent pathways remains to be defined, but provides an interesting focus for the study of biological network behaviour.

23 Therefore the selective spatio-temporal activation or silencing of network components may result in aspects of tissue stress with consequent disruptions in homeostasis that result in a bifurcation in tissue behaviour. That is the inflamed tissue either restores its previous homeostatic 'set' points or shifts these variables to a new steady-state. To effectively understand these disease state(s) within a complex biological network -we need to understand the key concept of robustness (Aderem, 2005).

1.5 Inflammatory transformation of tissues: an issue of robustness

Wells et al (Wells et al., 2005) describe inflammation as failure of resolution rather than an induction of the inflammatory process and point to the `inappropriately prolonged life of macrophages' to sites where they have been actively recruited as likely to lead to para-inflammation. They point out that sets of genes that negatively regulate macrophage activation maybe considered 'inflammation suppressor genes' and that at many steps within the pathway of macrophage activation are subject to negative feedback regulation. The majority of genes encoding these negative regulators are produced late on in the transcriptional response of macrophages to LPS and include regulators of receptor complexes, signaling and transcription. One such candidate is a gene encoding the p47phox subunit of the NADP(H) oxidase. Mice homozygous p47phox (-/-) for its absence displayed exacerbated lesions and slowed recovery to experimental arthritis. Conversely, this gene maybe regarded as pro-inflammatory as it is a prima facie candidate for the oxidative damage caused by the production of reactive oxygen species (ROS) in macrophages during chronic inflammation (Halliwell, 1999). These seemingly contradictory observations between in vivo and in vitro models corroborate the chasm that exists between our reductionist understanding and that of the holism reflected in the study of a network or complete system.

In the above example we can see that despite a plethora of genes that encode negative regulators of the LPS-TLR4 pathway in macrophages, seemingly mutation of any of these candidates directs the macrophage towards a para-inflammatory phenotype. The reason for this probably lies in the polymorphic nature of these genes and conference of a evolutionary advantage to individuals that are maximally heterozygous at these `inflammatory suppressor' loci (Wells et al., 2005). This ensures that a host response to a wide variety of different pathogens or chemical insults are robust. Nevertheless, it's likely that only a subset of these inflammation suppressor genes are invoked within a particular 24 macrophage, selectively switching off a particular pathway within the network. Hence, homozygous mutations at particular loci within alleles may result in the adoption of a para- inflammatory phenotype (Wells et al., 2005). This is a good example of a trade off between being robust to an enormous number of potential insults but extremely fragile to mutations in those genes mediating the response. The themes of phenotypic stability and maintenance of states versus functions are defining characteristics of robust networks as espoused by Kitano (Kitano, 2007) which he defines thus:

`Robustness is a property that allows a system to maintain its functions against internal and external perturbations.' After (Kitano, 2004).

Kitano makes the distinction between homeostasis and robustness (i.e. maintaining states vs. functions). However, this distinction is somewhat laboured as states ultimately form a continuum that defines and describes function. However, Kitano eludes to the fact that robustness is a more general concept insofar that a system is 'robust' provided it maintains functionality even if it transitions through a new steady state or instability. Robustness can be thought of as a general concept and stability and homeostasis as specific instances of this. However, robustness although important, is only one such concern of systems biology and it is worth digressing to review the field, its decade long existence and how it can impact our understanding of the innate immune system.

1.6 Systems biology a definition and its impact in innate immunity

Much has been said of systems biology, but arriving at a definition of this science is still proving problematic some 43 years since this term was first coined by Mesarovic (Mesarovic et al., 2004). Mesarovic’s intent in 1968 was to explain biological phenomena using decision making / control concepts from mathematical general systems theory. Likely, through the inability to make high-throughput parallel measurements at this time, this effort appears to have stalled. Mesarovic also acknowledges the parallels of the systems biology challenge and complexity science (Kauffman, 2008, Kauffman, 1995), whereas other system biologists have emphasized evolution of the ‘two roots’ of traditional molecular biology and non-equilibrium thermodynamics that have led to systems approaches in biology (Westerhoff and Palsson, 2004). In an attempt to arrive at a definition, a limited and far from exhaustive survey of reviews that discuss systems biology was performed. Excerpts from these reviews are provided below and then

25 compiled for the respective frequency of words that occur in the text using wordle (Figure 1.1):

‘Systems biology is a comprehensive quantitative analysis of the manner in which all the components of a biological system interact functionally over time.’ (Aderem, 2005)

‘…systems biology involves an iterative interplay between more or less high-throughput and high-content ‘wet’ experiments, technology development, theory and computational modelling, and that it is the involvement of computational modelling, in particular, in the process that sets systems biology apart from the more traditional and more reductionist molecular biology.’ (Kell, 2006a)

‘Systems Biology is the science that aims to understand how biological function that is absent from macromolecules in isolation, arises when they are components in the system.’ (Westerhoff et al., 2008)

‘The central task of systems biology is (a) to comprehensively gather information from each of these distinct levels [(i.e. molecules, cells, tissues, organs, organisms)] for individual biological systems and (b) to integrate these data to generate predictive mathematical models of the system.’ (Ideker et al., 2001)

‘[Systems Biology]…is an emerging field of biological research that aims at a system level understanding of genetic or metabolic pathways by investigating interrelationships (organisation or structure) and interactions (dynamics or behaviour) of genes, proteins and metabolites.’ (Wolkenhauer, 2001)

Figure 1.1. A wordle generated from the extracted text of reviews on the science of systems biology presented above. Although a trivial analysis it provides recurrent themes in definitions of systems biology.

In conclusion the definition used in this thesis is ‘Systems biology is the spatio-temporal analysis of networks of biological components to elucidate their emergent function’

The innate immune system is particularly suited to a systems biology analysis. One does not have to disrupt physical syncytial connections at least at certain stages of development of the macrophage – aiding cellular isolation. Aderem and Zak have reviewed the systems biology of innate immunity (Zak and Aderem, 2009) emphasizing three founding principles of systems biology: emergence ; the likely recognition of multiple PRR’s by a pathogen, robustness ; the redundancy of PRR’s in detecting PAMP’s and modularity ; activation of specific PRR networks that themselves have specific feedback architectures.

1.7 The Problem

Clearly, the transformation of the systems phenotype to an inflammatory condition, or moreover a para-inflammatory state, is a problem of unraveling an extremely complex and dynamic network of primarily proteins and nucleic acids that drive IL-1β expression and cannot necessarily be rectified by time alone. This presents an inverse problem (Tarantola, 2006); where one uses observations (e.g. expression of a protein) to infer parameters of the system (e.g. identities and concentrations of signaling molecules). Given the degeneracy of the situation (i.e. a specific level of protein expression may correspond to many different sets of parameters) some posit that these problems cannot be solved (Brenner, 2010). However, this assertion is itself flawed; the outcome of testing any hypothesis does not ‘solve’ the problem. It is possible only to disprove a falsifiable, null hypothesis with a specified statistical confidence ( α), at least according to the prevailing scientific worldview espoused by Popper (Popper, 2002). Indeed, other systems biologists have remarked on the relevance of data-driven approaches as being complementary to hypothesis driven research and these are allied arguments for pursuing inverse problems (Kell and Oliver, 2004).

Other means need to be sought to readdress the balance and one such way is via the use of chemical genetics: ‘[which is] the application of small molecule probes as tools for understanding and manipulating biological pathways’ (Bucci et al., 2010). Using 27 combinations of ligands or reagents that are selective for different signaling proteins in order to modulate their activities and the outcome of the measured variable is a primary concern of pharmacology. A review on the field of combination chemical genetics (Lehar et al., 2008) outlines the advantages this approach as follows:

1. The targeting of a single domain of a multi-domain protein 2. Allows timely control of biological processes 3. Targets orthologous and paralogous proteins 4. Does not directly alter the concentrations of targeted species obviating indirect effects

Pharmacology, although arguably a relatively modern science has a murky legacy in therapeutics and thankfully so, as new paradigms for drug development (Fitzgerald et al., 2006) are harking back to original preparations for the treatment of disease which were in many cases unwittingly combinatorial (Rang, 1995, Ruiz-Mesia L Fau - Ruiz-Mesia et al., 2005).

1.8 The Question

Nevertheless, the inverse problem still remains: how do we combine reagents in a coherent manner to target combinations of unknown signaling protein nodes that drive IL-1β expression? Given the phenomena of combinatorial explosion that describes functions that grow rapidly in combination it is worth reviewing the science of combinatorics. Consider the following formula that states a k combination of a set S is a subset of k distinct elements of S. If the set has n elements then the number of combinations is equal to the binomial coefficient.  n  n!   = (1)  k  k (! n − k)! This enumeration holds true whenever k ≤ n. For instance from a set ( S) of thirty three elements ( n), the subset ( k) of all five membered combinations from this group number 237,336. However, this formula reduces to the following if considering all possible k. n  n  ∑  = 2n (2) k =0  k  Therefore, for S = n = 33 (i.e. evaluating the entire set) the calculation simply becomes 2 33 = 8.59 x 10 9 where a reagent can either be present or absent resulting in some 8.5 billion possible combinations. Despite the progress made in high-throughput screening technologies (Bajorath, 2002, Feng et al., 2009), the complete analysis of even modestly 28 sized chemical libraries is prohibitive due to this combinatorial explosion that occurs in pharmacological space (Paolini et al., 2006). However, a solution for reducing this space and consequently the size of the inverse problem is at hand in the form of a widely used optimization technique in computer science – genetic or evolutionary algorithms (Mendes and Kell, 1998, Kell, 2006b).

1.9 An Introduction to evolutionary computation

The field of evolutionary computation studies and attempts to replicate and exploit biologically inspired methods (e.g. growth and development) of evolution in silico , one method encompassed within this broad church of techniques is that of the genetic algorithm (GA). The original development of the GA by John Holland in the 1960's has been reviewed by Melanie Mitchell (Mitchell, 1996);

[The purpose of the GA is to]`...formally study the phenomenon of adaptation as it occurs in nature and develop ways in which the mechanisms of natural adaptation might be imported into computer systems.'

The development of evolutionary computation and it's sub-fields (e.g. genetic algorithms, evolutionary programming and strategies) has blurred Holland's initial definition of what a GA was, with researchers often citing use of a GA which is less recognizable from Holland's original idea. To this end I will refer to the methods outlined in this dissertation as evolutionary algorithms (EA), with our main aim being to use these methods as an optimization strategy in the search for a unique and efficacious set of reagent combinations that inhibit IL-1β expression. The essence of the EA is a computational distillation of the process of natural selection; whereby heritable traits that positively influence the survival and fecundity of an organism are preferentially selected in that population. Whitley (Whitley, 1994) summarizes the EA by stating that:

`These algorithms encode a potential solution to a specific problem on a simple chromosome-like data structure, and apply recombination operators to these structures in such a way to preserve critical information'

The `chromosomes' mentioned in Whitley's quote refer to in silico lines or `strings' of one's and zero's, which are individually referred to as `bits'. Alone or in combination (depending on the problem one is trying to encode) these bits represent `genes' and the binary variation 29 of each bit (1 or 0) within a gene an `allele'. A basic overview of the methods used by the EA for the generation of new solutions (i.e. combinations) is given in figure 1.2 and the relative balance between the contributing experimental and computational components given. EA methods assume that high quality (i.e. high fitness) parent candidate solutions are available to undergo recombination and generate new offspring (Mitchell, 1996).

Figure 1.2. The basic principles behind the workings of evolutionary algorithms.

A tick against a bit string (e.g. combination) represents its comparative fitness to other chromosomes in the population and subsequent selection for reproduction. The string amongst others is then subjected to recombination (cross over) with other strings and mutation that determine the new generation. This new generation of strings is then evaluated experimentally and the cycle repeated iteratively until a suitable solution is located.

Other important concepts in the use of EA's are those of `search space' and `fitness landscapes' (Mitchell, 1996). Simply put the `search space' is a set of different candidate solutions or chromosomes. It may at first appear strange to those unfamiliar with EA's, why if one is encoding a solution then why go and look for it? The underlying issue is now familiar to the reader: that there are many numerous solutions to the potential problem (i.e. combinatorial explosion), more than it is often possible to evaluate sequentially. Conversely, the `fitness landscape' is a representation of all possible chromosomes and their fitnesses (Mitchell, 1996) and can be thought of as a l+1-dimensional plot where l is the length of the bit string encoded and the fitness is plotted on the l+1 st axis. So, for a 30 reagent combination we have a 31-dimensional problem! This `landscape' can be thought to have peaks (corresponding to local maxima) and troughs (minima), with changes made

30 in the bit strings via the action of recombination and mutation operators that allow exploration of new areas of this landscape.

Although the discussion thus far has focused on the search for a unique and efficacious reagent combination, within this searches biological context it's also critically important we do not obviate the cells viability and to this end we are actually presented with conflicting objectives. This is a problem that can be solved via evolutionary multi-objective optimization (EMO) (Knowles, 2009, Knowles et al., 2008, Handl et al., 2007), any optimization is concerned with finding optimal values within functions and more often than not owing to the symmetrical nature of maximizing or minimizing a function; minimization is chosen (Mendes and Kell, 1998). Given that it's necessary to minimize the values of both objectives without a worsening in either the set of optimal solutions lie on what is termed the Pareto front . Solutions within the search space that are non-dominated are so called because they have a minimal value compared to any other solution in at least one objective.

Knowles et al (Knowles et al., 2008) emphasize the applications of EMO which are perfectly suited to what in essence is a drug discovery initiative such as this. Mendes and Kell (Mendes and Kell, 1998) have compared and applied a variety of numerical optimization methods to a hypothetical metabolic network. In their comparison they optimized the network to maximize a flux parameter (J5) whilst either constraining or varying another specified flux parameter (J8). Without constraining J8 the GA and evolutionary programming methods were found to locate a value for J5, 30-fold higher than the wild-type flux, this was the maximal value attained by any method of optimization. Similarly, whilst constraining J8 within ~ 8 % of its wild-type value revealed the same result with evolutionary programming slightly superior to the optimal solution located by the GA (Mendes and Kell, 1998).

A review covering the application of search algorithms to the discovery of novel drug combinations (Feala et al., 2010) classifies approaches to the search into six categories. These are; non-systematic, exhaustive enumeration, statistical association, explicit model based methods, model-free biological search algorithm (the approach adopted here) and model plus biological search algorithm (which is a combination of the 3 rd , 5 th and 6th approaches respectively). This latter method provides the strengths of both in silico modeling and experimental analysis providing a dual assessment of prediction and confirmation respectively that can be coupled in iterative cycles to successively focus on

31 unique, selective and potent combinations. Prediction using in silico models has been a neglected component in this discussion and warrants further scrutiny.

1.10 A biochemists guide to In silico modeling of biochemical systems

The preceding discussions of the TLR4 signaling network have focused on the protein players or nodes, their activation / deactivation and consequent effects on signaling. These transformations are conducted via reactions or edges within these networks (Barabasi and Oltvai, 2004). For the biochemist it is critical to consider the physical basis of these changes which are often abstracted and expressed in in silico representations as rates of change using ordinary differential equations (ODEs) (Mendes et al., 2009, Aldridge et al., 2006, Sreenath et al., 2008, Pahle, 2009). The modeling of dynamical systems using ODEs in biology and especially physiology (Hodgkin and Huxley, 1952) and ecology (Haefner, 2005b) has a long and distinguished history. An ODE describes the change (formally the derivative of a function and this is often concentration of a protein) with respect to an independent variable (nearly always time for the purposes of these discussions). However, these abstractions naturally arise from a consideration of the law of mass conservation, put simply for any reactant:

Rate of change = rate of appearance – rate of disappearence

in Γ + in ΓA r ΓB A B - Γr

Figure 1.3. An idealized uni-molecular reaction detailing the conversion of species A into B with corresponding mass fluxes ( Γ). (Beard and Qian, 2008).

32 For the example above (Figure 1.3) where we assume constant volume and that spatial gradients of species are negligible, with concentrations of species being continuous it becomes easy to derive ODE’s describing the system (Beard and Qian, 2008). We know that mass (n) is calculated from the volume ( V) multiplied by concentration ([X]; where X is any species), thus:

Rate of change = V.d [A] / dt (3)

Rates of appearance and disappearance are calculated from mass fluxes ( Γ; in units mass / unit time), therefore, the concentrations of species A ([A]) and B ([B]) can be described by the following ODEs

d[A] − + V = (Γin Γ+ ) Γ− (4) dt A r r

d[B] + − V = (Γin Γ+ ) Γ− (5) dt B r r

Summing both (4) and (5) leads to:

 d A][ d[B] V  +  = Γin Γ+ in (6)  dt dt  A B

A closed system (one where matter is not transferred) do not have transport fluxes and therefore (6) is rendered equal to zero (the masses of both reactants remain constant). This is an important result as we shall see later. In biological and thermodynamic senses this can be interpreted as the system being dead or at equilibrium respectively. It is often more convenient to express fluxes in terms of concentration ( J) and this is simply established via the following relation (J = Γ / V) .

Thermodynamically (and independently of the kinetics of the reaction scheme), it can be shown that the change in Gibbs free energy (∆G – a measure of the chemical work done on a system in this context) is proportional to the ratio of forward and reverse concentration fluxes.

∆G −= RT ln(J + / J − ) (7)

Often, simulations of biochemical systems are built on the simplistic kinetic scheme of mass action kinetics which assumes that molecules of the same species do not affect their ability to react. For our simple reaction (Figure 1.3), the law of mass action implies:

+ J = k+[A] (8) - J = k- [B] (9)

Where k+ and k- are forward and reverse rate constants of the reaction. Using equations (8) and (9) it is possible to re-write equations (4) and (5) in terms their concentration fluxes.

d[A] in = (J + k− [B]− k+ [A]) (10) dt A

d[B] in = (J + k+ [A]− k− [B]) (11) dt B

Open systems with constant transport fluxes may approach stable steady states that are not in −= in = = equilibrium states (e.g. (10) and (11)), when J A J B J constant, then:

d[A] −= (k− + k+ )[ A] + J + k− X (12) dt 0

Where X0 is a constant ([A] + [B] = X0). Previously, as we had done for (6), constraining (12) to zero yields the steady state concentrations of the reactants.

(J + k X ) = → ∞ = − 0 [A]ss (tA ) (13) ()k− + k+

Numerical integration of systems of ODEs (see Figure 1.4 for this example) allows us to arrive at approximate solutions as their solution analytically (in contrast to the example above) is often not an option. In the instance of this uni-molecular reaction, it’s possible to intuitively arrive at the solution, but in systems that have increased dimensionality with respect to the number of ODE’s they have it most often is not. Integration is usually carried out using specialised software for biochemical modeling and simulation (i.e. the computational implementation of a formal model) such as COPASI (Hoops et al., 2006).

34 0.65

0.60

0.55 [A] 0.50 [B] 0.45 [Species] (mM) [Species] 0.40

0.35 0 1 2 3 4 Time (secs)

Figure 1.4. Simulation of the uni-molecular reaction A = B (outlined in Figure 1.3) -1 -1 performed in COPASI. Parameters were set as follows: k+ = 1 s and k- = 1 s , V = 1 mL, J -1 = 0.2 mM s . Initial conditions were [A] 0 = [B] 0 = 0.5 mM (After (Beard and Qian, 2008)).

ODE modeling in the context of biochemical systems has to span large changes in space and time and hence this provokes problems if there are reactions present within the sets of ODE’s that occur on markedly different time scales (e.g. hours vs. seconds) or in different cellular compartments. Problems of this nature are termed 'stiff', fortunately, complex pathway simulation COPASI solves the temporal problem by using LSDODA (Mendes et al., 2009). Methods and solutions for obviating these issues have been reviewed (Dada and Mendes, 2011, Martin et al., 2009).

The purpose or objectives of modeling per se as summarised by Mendes (Mendes et al., 2009) could be to challenge assumptions, think about a problem in depth, aid experimental design and to identify unknown parameters that are important in determining the system. Haefner (Haefner, 2005a) similarly makes the point that models have one of three purposes either to understand, predict or control a system. Similarly, questions asking what is the model to be compared to and how will the model output be used? are also pertinent to this project.

35 1.11 Combinatorial solutions to complex problems?

Given the complexity biological signaling during inflammatory episodes and divergent signals generated, studying the downstream signaling events of activating just a single receptor subtype (i.e. TLR4), highlights the need for a systems approach to address how the network responds to inputs and modulates outputs. Using a ‘reverse’ combination chemical genetic approach (Lehar et al., 2008) our aim was to optimize combinations of reagents that minimize LPS-induced macrophage IL-1β expression, and simultaneously minimize the number of component reagents in the combination and their propensity to induce macrophage cell death. Subsequently we sought to optimise reagents to inhibit IL-1β expression at concentrations lower than the component reagents used in isolation. This was achieved by application of an EA-directed, semi-automated robotic assay of IL-1β expression to a dynamic chemical library of a total of 33 reagents (see Methods, and also (Knight et al., 2008, O'Hagan et al., 2005, Knowles, 2009)). The specific algorithm used here was the indicator based evolutionary algorithm (IBEA) (Zitzler and Künzli, 2004), as in preliminary simulations (Allmendinger and Knowles, 2009) this proved superior to a variety of other multi-objective optimization algorithms. This was followed by a dose matrix search of resultant top ranked reagents from the EA-directed search. We demonstrate that the EA converges efficiently on good solutions, and that dual p38 MAPK inhibition and either I κB kinase inhibition or iron chelation yields synergistic and biologically-relevant inhibition of macrophage IL-1β expression. Similarly, the construction of an IL-1β transcriptional map from the primary literature in human and murine monocytes and macrophages, specifically focused on proteins and regulatory DNA elements where evidence existed for their dynamic transcriptional effects provided insights into the modulation of the experimentally observed inhibition of IL-1β expression.

1.12 Aims and objectives

1. To locate combinations of signaling protein inhibitors, pro / anti-oxidants and / or iron chelators that inhibit IL-1β expression using an EA directed search coupled to a semi-automated robotic assay of IL-1β expression in J774.A1 macrophages.

36 2. To optimise pairs of five top-ranked reagents (from the post-hoc analysis of the EA-directed search) in combination with respect to their concentration using an adaptive dose-matrix search protocol in J774.A1 macrophages.

3. To construct a human / mouse TLR4 - IL-1β signaling map from the primary literature.

37 Publication number : 01

Publication title : Efficient discovery of anti-inflammatory small molecule combinations using evolutionary computing.

Page number : 38

Published: doi:10.1038/nchembio.689

Journal Nature Chemical Biology

Efficient discovery of anti-inflammatory small molecule combinations using evolutionary computing

Ben G Small 1,8,$ , Barry W M cColl 6,$ , Richard Allmendinger 4, Jürgen Pahle 4, Gloria López- Castejón 3, Nancy J Rothwell 3, Joshua Knowles 4, Pedro Mendes 2,4,5 , David Brough 3, Douglas B Kell 2,7*.

1Doctoral Training Centre, Integrative Systems Biology Molecules to Life and 2Manchester Centre for Integrative Systems Biology, Manchester Interdisciplinary Biocentre, University of Manchester, 131 Princess Street, Manchester, M1 7DN, United Kingdom. 3NeuroSystems, Faculty of Life Sciences, AV Hill Building, University of Manchester, Oxford Road, Manchester, M13 9PT, United Kingdom. 4School of Computer Science, University of Manchester, Oxford Road, Manchester, M13 9PL, United Kingdom. 5Virginia Bioinformatics Institute, Virginia Polytechnic Institute and State University, Washington Street, MC0477, Blacksburg, Virginia, United States, 24061-0477. 6The Roslin Institute, University of Edinburgh, Easter Bush, Midlothian, EH25 9RG, United Kingdom. 7School of Chemistry, University of Manchester, 131 Princess St, Manchester M1 7DN, United Kingdom. 8School of Chemical Engineering and Analytical Science, University of Manchester, 131 Princess St, Manchester M1 7DN, United Kingdom.

*Corresponding Author: [email protected] $These authors contributed equally to this work Character count: 31898

Running Title: Chemical genetics of IL-1β expression

Abstract The control of biochemical fluxes is distributed and to perturb complex intracellular networks effectively it is often necessary to modulate several steps simultaneously.

However, the number of possible permutations leads to a combinatorial explosion in the number of experiments that would have to be performed in a complete analysis. We used a multi-objective evolutionary algorithm (EA) to optimize reagent combinations from a dynamic chemical library of 33 compounds with established or predicted targets in the regulatory network controlling IL-1β expression. The EA converged on excellent solutions within 11 generations during which we studied just 550 combinations out of the potential search space of ~ 9 billion. The top five reagents with the greatest contribution to combinatorial effects throughout the EA were then optimized pairwise. A p38 MAPK inhibitor with either an inhibitor of I κB kinase or a chelator of poorly liganded iron yielded synergistic inhibition of macrophage IL-1β expression. Evolutionary searches provide a powerful and general approach to the discovery of novel combinations of pharmacological agents with potentially greater therapeutic indices than those of single drugs.

Acute or chronic (non-resolving) inflammation is a well established mediator of major diseases including vascular disease (e.g. atherosclerosis, stroke), chronic obstructive pulmonary disease (COPD), cancer, diabetes, obesity, rheumatoid arthritis, psoriasis and inflammatory bowel disease 1,2 . Each of these conditions exhibits an elevated expression of the potent and pleiotropic, pro-inflammatory cytokine interleukin-1β (IL-1β) 2,3 .

Pharmacological therapies targeted against IL-1β are largely focused on biologics that are only likely to act extracellularly 4 (e.g. Anakinra; an IL-1 receptor antagonist), which have shortcomings such as poor CNS penetration. Small molecule modulation of IL-1β may offer benefits in certain conditions such as in cerebrovascular injury where IL-1β mediates significant cerebral damage during acute ischaemia and excitotoxic insult 5.

The lipopolysaccharide- (LPS-) induced Toll-like receptor 4 (TLR4) stimulation of macrophages is a widely used experimental model that mimics key aspects of inflammation including IL-1β expression 6. Signal transduction induced by TLR4 stimulation proceeds through the activation of a complex array of multi-protein signalling networks (e.g. MyD88, TRAF6, p38 MAPK, NF-κB), ultimately resulting in the expression of the IL-1B gene. Possibly based on the ‘one gene, one drug, one disease paradigm’ 7 a plethora of reagents has been developed to modulate individual proteins within these signalling networks. However, it is well known that multiple steps must be modulated simultaneously to have significant effects on biochemical fluxes 8. Thus, a single-target approach is unlikely to be optimal in inhibiting the expression of a proinflammatory cytokine such as IL-1β, where there is both inherent degeneracy and considerable complexity within the signal transduction network 9. More generally, there is an increasing recognition of the need to target multiple steps within signalling networks for their effective pharmacological modulation 7.

Combinatorial chemical genetics 10 uses combinations of small molecules that allow dissection of cellular phenomena via their selective modulation of individual biological targets. Despite progress made in high-throughput screening technologies 11 , the analysis iii of even modestly sized chemical libraries is prohibitive due to the combinatorial explosion that occurs in pharmacological space 12 (2 33 or ~ 9 billion combinations for all possible combinations of the chemical library explored here). Thus we sought heuristic solutions

(i.e. reagent combinations) that are good but not provably globally optimal.

The terms ‘evolutionary computing’ and ‘evolutionary algorithms’ describe a set of approaches based loosely on Darwinian evolution by the natural selection of individuals and populations. In this case the population consists of individuals that each encode a candidate solution to the problem at hand. The ‘fitness’ of each solution is reflected in the objective function(s) designed by the experimenter, but normally includes the concept that fitter individuals provide more accurate solutions. There may be multiple fitness functions.

For instance a simpler solution may be deemed to be a fitter solution, and algorithms with multiple objectives (multiple fitnesses), as in this work, are known as multi-objective evolutionary algorithms. Multi-objective evolutionary algorithms (EA) allow for the specification of multiple and distinct optimization objectives and their simultaneous handling 13-15 . Based on the fitness(es), a selection step determines which individuals will be allowed to remain under consideration for the next generation. Some of these individuals are retained by simply being copied unchanged into the subsequent generation(s), but new diversity, based on the parents selected, is then produced from them by processes analogous to mutation and recombination. The fitnesses of these new individuals are then evaluated as above, and the algorithm continues cycling through the steps of selection, breeding and fitness evaluation until an acceptable solution is found.

Decades of research within the field of evolutionary computing (e.g. 16,17 ) have revealed that optimization of multivariate problems can be highly effective using small numbers of experimental tests. Although the present study used an EA and an adaptive dose matrix search strategy, we recognize that other kinds of combinatorial optimization approaches might also prove effective.

Using a ‘reverse’ combination chemical genetic approach 10 our aim was to optimize combinations of reagents that minimize LPS-induced macrophage IL-1β expression, and simultaneously minimize the number of component reagents in the combination and their propensity to induce macrophage cell death. Subsequently we sought to optimize reagents to inhibit IL-1β expression at concentrations lower than the component reagents used in isolation. This was achieved by application of an EA-directed, semi-automated robotic assay of IL-1β expression to a dynamic chemical library of a total of 33 reagents (see

Methods, and also 18-20 ). The specific algorithm used here was the Indicator Based

Evolutionary Algorithm (IBEA) 21 , as in preliminary simulations 22 this proved superior to a variety of other multi-objective optimization algorithms. This was followed by a dose matrix search of top-ranked reagents resulting from the EA-directed search. We demonstrate that the EA converges efficiently on good solutions, and that dual p38 MAPK inhibition and either I κB kinase inhibition or iron chelation yields synergistic and biologically-relevant inhibition of macrophage IL-1β expression.

Results

Rapid convergence of Multi-objective IBEA to near-optimal solutions

Concentration-effect curves for a selection of reagents with known or predicted targets in the IL-1β expression network were determined in order to identify the most appropriate concentration for use in the EA (Supplementary Fig. 2). On the basis of these data we selected a sub-optimal dose (3 µM) which would provide scope for observing combinatorial synergy.

The Indicator Based Evolutionary Algorithm (IBEA) 21 directed a semi-automated robotic assay utilizing chemical combinations to inhibit IL-1β expression was initialized (Fig. 1,

Loop 1). As described in Methods (Implementation of the Indicator Based Evolutionary

Algorithm (IBEA)), the IBEA generates subsets of combinations from the library and then assesses the performance of these subsets with respect to inhibition of IL-1β expression, decreases in LDH release (a marker for cell death) and the number of member reagents within the combination. In the present case, we confined the number of experiments in each of the first and subsequent generations to 50 and those for generation 1 were selected randomly by the EA from the first-generation library of chemicals. Superior combinations are retained and recombined with other library components in successive subsets and assessed iteratively until satisfactory combinations are found. Assay of successive generations of chemical combinations from a dynamic chemical library (a total of 33 reagents, each at a concentration of 3 µM, see supplementary methods) revealed their convergence towards a set of highly effective cocktails (Fig 2; see also Supplementary Fig.

3 and Supplementary Table 3 and 4 (top) for data on the individual generations). In addition, Supplementary Spreadsheet 1 gives all of the data on both reagent combinations and experimental measurements in tabular form.

Convergence of solutions derived from the IBEA’s search of the chemical library was assessed by measuring the population average rank of the IBEA hypervolume (see supplementary methods) and the three objective functions (IL-1β expression, number of component reagents within combinations and LDH release (Fig. 3). The observed plateau in IL-1β expression between generations 9 and 10 and very marginal decreases in LDH decrease (Fig 3 top and bottom left respectively) were triggers to halt the IBEA-directed search. By this stage, although not provably globally optimal, almost all IL-1β expression had been ablated by some combinations, with negligible toxicity.

A particular strength of this algorithm is the ability to add and remove reagents to/from the library during the evolution of the combinations 23 (see supplementary methods – reagent removal / addition and data analysis), and generations 10 and 11 explored these a little further. We also noted that many of the more successful reagent combinations contained the p38 MAPK inhibitor SB203580.

Post-hoc analysis of the IBEA search of combinatorial chemical space

The search of combinatorial chemical space yielded 51 and 188 reagent combinations that exhibited an inhibition of IL-1β expression that was greater than or equal to either 95 % or

70 % of the control response, respectively (Supplementary Table 3 and 4, (Top)) across all generations. We chose to explore the inhibitory activities of other reagents independently of SB203580 because these effects might have been ‘masked’ in the EA by the dominance of SB203580 (70% (35/50)) of sampled combinations at generation 10 contained

SB203580). Thus, a ‘post hoc’ analysis was conducted of all data to assess the fitness contributions of single reagents (their overall score against our defined objectives) alone and in two- and three- component combinations (in the presence and absence of

SB203580) for all reagents (Fig. 4 and see Methods). These component reagents included the p38 MAPK inhibitor (SB203580), a sphingosine kinase inhibitor, (SKI-II), a statin

(simvastatin), an iron chelator (SIH) and an inhibitor (BMS-345541) of the inhibitory κB kinase (IKKi). Inhibitors of p38 MAPK and IKK are well established as inhibitors of IL-

1β expression, and (as well as its effects on HMG-CoA reductase) simvastatin is a known anti-inflammatory and an abundance of evidence implicates poorly liganded iron in inflammatory processes 24,25 . However, the appearance of SKI-II may have been unexpected, albeit that there is evidence for the involvement of at least one sphingosine kinase in inflammation 26 . Despite the appearance of both wortmannin and the mitochondrial uncouplers in the post-hoc analysis, these compounds were not pursued further owing to their lack of specificity and potential toxicity, respectively. The ability to

vii observe substantial inhibition of IL-1β expression with just pairs of inhibitors

(Supplementary Table 3 and 4 (bottom)) led us to study the concentration-dependent pairwise optimization of the top-ranked reagents.

Concentration-dependent search reveals combinatorial synergism

The search over defined concentration ranges of all pairs of five top-ranked reagents was assessed using an adaptive dose matrix search protocol (Figure 1, Loop 2 and

Supplementary Fig. 4). Briefly, this protocol adaptively changes the concentrations of chosen reagents (See methods and compare SB203580 and SIH (Supplementary Fig. 4) versus the same combination presented here (Fig. 5)). Reagents were assessed alone and in pairs (Fig 5). Modes of pharmacological effect driven by reagent combinations have their own nomenclature 27-29 . Hence, additivity is the linear superposition of two different reagent effects and synergy a non-linear (excess) inhibition from a reagent combination beyond that expected for simple additivity 27 . Evidence of synergy (Fig 5(IKKi; c) or

5(SIH; f)) was determined by subtraction of predicted additive effects (Fig 5(IKKi; b) or

5(SIH; e)) of each combination (based on single reagent efficacy) from the actual experimental data (Fig 5(IKKi; a) or 5(SIH; d)). Synergy was observed for most combinations; however, we noted in some instances that this was in fact attributable to a loss of a potentiating effect on IL-1β expression that occurred with either of the reagents when used alone. In these cases, the net (i.e. resulting) magnitude of IL-1β inhibition was minimal and therefore we focus here on examples of combinatorial synergy generating biologically-relevant levels of IL-1β inhibition. SB203580 (0.1 µM) and IKKi (1 µM) alone inhibited IL-1β expression by 28 ± 7 % and 12 ± 9 % respectively (mean ± SEM, n=7) and SB203580 (0.1 µM) and IKKi (1 µM) in combination achieved significantly greater inhibition than did either drug alone (59 ± 5 %, n=7, p< 0.02 and p< 0.0005, one way ANOVA with Tukey’s multiple test correction versus SB203580 or IKKi alone respectively). The inhibitory effect of SB203580 (0.1 µM) and IKKi (1 µM) in viii combination was 19% greater than that predicted for a purely additive effect thus demonstrating a marked synergistic interaction at these concentrations (Fig. 5(c)).

Similarly, SB203580 (0.1 µM) and SIH (3 µM) alone inhibited IL-1β expression by 31 ±

10 % and 19 ± 8 % respectively (n=7 plates) and combinatorially inhibited IL-1β expression by a significantly greater magnitude (59 ± 4 %, n=7 plates, p<0.04 and p<

0.004, one way ANOVA with Tukey’s multiple test correction versus SB203580 and SIH alone, respectively). This combinatorial inhibitory effect was 9% greater than that predicted for pure additivity therefore indicating a synergistic interaction (Fig. 5(f)). We also observed marked synergistic effects for combinations of IKKi and SKI-II, IKKi and

SIH and SB203580 and SIH (Supplementary Fig. 4), To assess whether a triple combination of SB203580, IKKi and SIH could provide an inhibition of IL-1β expression beyond the synergy already observed for both paired combinations (i.e SB203580 with either IKKi or SIH) we superimposed increasing concentrations of SIH (0, 0.1, 0.3, 1 and 3

µM) onto SB203580 and IKKi dose matrices (Supplementary Fig. 5). We did not observe any further synergy with the triple-combination.

Discussion

There is a growing recognition that drugs, to be effective whether singly or in combination, must affect multiple steps simultaneously 7,10,30,31 . This, however, leads immediately to a combinatorial explosion of experimental possibilities that limits the number of drugs that can reasonably be tested exhaustively. We have applied a multi-objective evolutionary algorithm (EA) to the optimization of reagent combinations, using a panel of candidate reagents selected from our own studies and the literature 32 as targeting the proinflammatory IL-1β expression network. The objective assessment of reagent combinations arising from the IBEA-directed search utilizing the ‘post hoc’ analysis of reagent fitness contributions is useful as it removes a layer of decision making 20 , reducing bias and potentially enhancing ‘hidden’ phenomena (e.g. off target effects of reagents) that may have beneficial effects on the system output (i.e. IL-1β expression). In this regard, it is worth mentioning the increasingly effective use of adaptive dosing regimes in clinical trials of pharmaceutical drugs both singly and in combinations 33 .

Combination therapy is now returning to the fore with a greater understanding of pharmacological mechanism being uncovered by advances in parallel measurements of biological endpoints 11,34,35 The use of the Gur-game stochastic search algorithm has been reported 36,37 in elucidating the anti-viral and NF-κB-activating efficacy of drug and cytokine combinations, respectively. Briefly, this algorithm functions by generating a random number (e.g. one representing a specified anti-viral activity) and switching the concentration of component drugs if their efficacy is below this value. In contrast, stochastic and deterministic elements of our search were based on experimental data output and recombination of reagents in new cocktails that had not yet been evaluated. In addition, the multi-objective nature of the EA-driven search presented here allows assessment of a number of biological endpoints. Following this approach with an adaptive dose-matrix driven search enabled a search of pharmacological space using just top-ranked x

‘hits’. Applications of search algorithms and the use of machine learning in the optimization of combinatorial therapies have recently been reviewed 38 , and within their categorization our method would fall into “E – Model-free Biological Search”.

Recently, two papers have cited the pairing of reagent combinations as indicative / predictive of higher order effects 39,40 . The latter 40 used a series of chemotherapeutic reagents to monitor for additivity, antagonism or synergy of combinations on YFP-tagged protein dynamics in the H1299 cell line. The authors propose that a linear superposition of weighted sums from the effect single drugs have on protein dynamics can predict higher order (i.e. combinatorial) effects for these reagents, although they were unable to demonstrate this for Wortmannin (a PI3K inhibitor).

39 2+ The former looked for an enhancement in the [Ca ]i signal of platelets using a pairwise agonist scanning approach of six reagents at three different concentrations. Using these data the authors trained a neural network model to predict higher order effects and were successful in doing so. However, the rapid and non-transcriptional signal transduction required for Ca 2+ mobilization may not be entirely reflective of multi-protein signalling networks, where there is extensive cross-talk and feedback loops that modulate responses on timescales ranging in minutes to hours and beyond.

In the present case, we found a substantial inhibition of IL-1β production by the p38

MAPK inhibitor SB203580, that was enhanced synergistically by either the IKK inhibitor

BMS 345541 or the iron chelator SIH 41 . Perhaps surprisingly, the triple combination of

SB203580, IKKi and SIH did not reveal additional effects beyond those observed for pairwise combinations of SB203580 and either of the other two reagents. The effects of iron chelation are of especial interest here, as there is abundant but widespread evidence for the role of unliganded iron in a variety of inflammatory disorders 24,25,42 .

xi p38 inhibitors have proved disappointing in clinical trials, but this is not so much because they are not active in vivo but because (at the doses used) they lose specificity for the α isoform of p38 and are toxic 43 . The particular benefit of the synergy when combining a p38 inhibitor such as SB203580 with BMS345541 or SIH observed here is that we can use them at lower concentrations in combinations than those at which they would be used alone (Fig 5). We note too that such inhibitors seem not to have been designed to exploit the specificity of pharmaceutical drug transporters 44 . Finally, it was noted too the efficacy of some combinations that did not involve the p38 inhibitor.

In conclusion, the application of an EA in conjunction with semi-automated assay of a dynamic chemical library enables a rapid scanning of reagent combinations without the need of initial hypotheses 45 about likely higher-order effects. Our results show that synergistic combinations can be revealed quickly, and that these combinations survived further experimental scrutiny, leading to pairwise combinations that seem promising to use in practice. Synergism of the SB203580 and either IKKi or SIH combination presented here, in contrast to the comparatively marginal effect of the individual reagents at the same concentrations, shows that pharmacological modification of biological targets and processes may be effected at concentrations that are more likely to avoid toxicities. This has particular relevance to the treatment of chronic inflammatory conditions such as irritable bowel syndrome 46 and chronic obstructive pulmonary disease 47 where treatment may be maintained for extensive periods. Our approach is essentially generic, and the time required per generation is determined by the time needed for setting up and running the assays (typically 3 days, one each for cell preparation, combination preparation and ELISA analyses), since the time needed for the algorithm to analyse the results and then to choose the cocktails for the next generation was negligible in comparison. Overall, our new method substantially decreases the time taken for the triage of pharmacologically useful chemical diversity within chemical libraries 48 . Additionally, we demonstrate how

xii combinations of known drugs or reagents could allow them to be repurposed 49 and could provide an elegant adjunct to existing therapeutic strategies in chronic inflammatory conditions.

xiii

Methods All procedures, protocols and methods were carried out under aseptic conditions where deemed necessary.

Construction and composition of the chemical library

The choice of reagents with which to populate the chemical library searched here was guided in part by Oda and Kitano’s TLR signalling network 32 and via identification of suitable ligands from single reagent studies in peritoneal macrophages. The following pharmacological classes of reagents were used: Iron chelators, TPEN; a zinc chelator, anti / pro-oxidants, NADPH oxidase inhibitors, PI3Kinase inhibitors, MAPK pathway inhibitors, NF-κB pathway inhibitors, Genistein; a tyrosine kinase inhibitor, mitochondrial uncouplers (removed after generation 3), statins and small-GTPase inhibitors. In the evolutionary optimization process each of these reagents corresponds to a single binary variable indicating whether the reagent is included in a combination or not; the combination itself represents a candidate solution to the problem. Detailed information regarding the construction, storage, maintenance and removal and replacement of reagents within the chemical library can be found in the supplementary methods (Reagent removal / addition and data analysis).

xiv

Implementation of the Indicator Based Evolutionary Algorithm (IBEA)

In order to select a suitable multi-objective evolutionary algorithm (EA) for addressing the search of the fitness landscape of the chemical library, a comparison of four EAs; IBEA,

SPEA2 and NSGA2 with binary tournament and NSGA2 with probabilistic selection 50 was undertaken 23 . All EAs were assessed on a family of test problems used to simulate the reagent combination problem. The test problems model a scenario where pharmacological interactions among reagents can be described by single, binary, and ternary effects only. A reagent combination can effect minimal IL-1β expression by killing cells; measured as a large release of LDH. To de-risk the detection of lethal versus effective and benign combinations respectively, positive, negative, and no correlations were considered between these two objectives. The different levels of correlation were realized by assigning certain probabilities to effect values; first for IL-1β expression and dependent on this value a subsequent probability determining an effect value for the LDH release. Depending on the correlation level, the effect values were drawn uniformly from the interval [-1,0) and/or (0,1].

All EAs tested were capable of locating combinations of compounds of similar quality in the presence of 80 % and 10 % variability in IL-1β expression and LDH release measurements respectively. However, IBEA had the best performance at finding effective compound combinations that contained only a few compounds (although its search was unrestricted and could have used any number of compounds). This was the only EA tested that was not based on Pareto ranking; rather IBEA searches for those solutions that maximize their hypervolume within objective space. Initialization of the 1 st generation of reagent compounds in IBEA was conducted by fixing the probability of compound selection to 3 / 33 to ensure a random selection of compounds from across the library, with on average three compounds in a cocktail. See the Supplementary methods (IBEA directed xv evolutionary search of combinations inhibiting IL-1β expression) and results

(Supplementary table 2) for other details of the algorithm and its parameters.

Production and assay of reagent combinations

A Sciclone ALH3000 laboratory robot (Caliper Life Sciences) under the indirect control of an Indicator Based Evolutionary Algorithm (IBEA) enabled the semi-automated assay of chemical combinations (Supplementary methods – Production of combinatorial chemical cocktails) in LPS stimulated J774.A1 macrophages. The iterative searching and analysis of incremented generations of combinations was conducted via measurements of an IL-1β expression ELISA (R&D Systems; DY401) and LDH release (Promega) (see

Supplementary methods – Measurement of LDH and IL-1β expression).

Treatment of Peritoneal and J774.A1 macrophages with single reagents and combinations

Peritoneal and J774.A1 macrophages were prepared and cultured (see supplementary methods) for either single reagent or combinatorial and dose matrix studies respectively.

Peritoneal macrophages were exposed to either single reagents (0.01 µM - 100 µM) or

DMSO (0.5 % v/v) for 0.5 h prior to stimulation with LPS (1 µg / mL). Similarly, J774.A1 macrophages were treated with chemical combinations (3 µM or varying concentration) or

DMSO (0.5 % v/v or 0.1% v/v) during the EA-directed and adaptive dose-matrix search respectively for 10 min prior to stimulation with LPS (1 µg / mL). After 4 h (Peritoneal) or

2 h (J774.A1) aliquots of well supernatants were taken for the measurement of LDH during single-reagent and EA driven combinatorial assessment respectively (see supplementary methods) prior to disposal of remaining supernatant, lysis of cells and freezing before

xvi measurement of IL-1β expression (see Supplementary methods – Measurement of LDH and IL-1β expression).

Post-hoc analysis of IBEA search and calculation of reagent fitness

Calculation of the fitness contribution of a single reagent within a combination was assigned as follows (1):

− − = − Fi X 1 X n (1)

− Where; Fi is the fitness contribution of any given single reagent ( i), and where X 1 and

− β X n are the mean IL-1 expression values of all combinations where this single reagent ( i) was present or absent respectively. Thus, a larger fitness contribution Fi indicates that a

β reagent is more efficient in decreasing IL-1 expression.,

Concentration-dependent optimization of paired reagent combinations using an adaptive dose matrix search protocol

Upon completion and post-hoc prioritization of reagent combinations from the IBEA search a concentration-dependent optimization step was implemented. Briefly, to assess the potentially synergistic effects of paired combinations on IL-1β expression we serially and logarithmically decreased the test concentrations of reagents from those used during the EA-directed search. Similarly, after this initial optimization step, we extended the scanned concentration ranges of promising combinations by adding in test concentrations of reagents at approximate 0.5 log 10 spacings within the dose-matrix. This allowed effect xvii

(i.e. IL-1β expression) comparisons at multiple doses of paired reagents. Pseudocolor mappings were performed by linear interpolation between samples; those mappings that move away from the blue end of the spectrum within combination response shape plots are indicative of synergistic inhibition of IL-1β expression between two reagents. (Figure 5,

Supplementary Fig 4).

Acknowledgements

We would like to thank the BBSRC for funding this project (BB/F018398/1) and the

BBSRC/EPSRC and Wellcome Trust for the funding of studentships and fellowships. We thank the Manchester Centre for Integrative Systems Biology for providing access to the

Sciclone instrument. BGS would like to thank Dr. Susan Wasley and Dr. Darren Stubbs of

Caliper Life Sciences for their help and assistance in commissioning and their image of the

Sciclone instrument (Figure 1) respectively. The images of the computer and 96-well plate

(Figure 1) were used under the creative commons CC0 public domain dedication and obtained from Clker.com.

Author Contributions

BGS designed, performed and analysed the data from the single reagent, combinatorial, evolution directed and adaptive dose matrix studies respectively and wrote the first draft of the manuscript.

BM designed, performed and analysed the data from single reagent and evolution directed combinatorial studies.

JP suggested and was involved in the design of the concentration-dependent adaptive dose matrix search protocol of paired combinations and wrote the code in R to produce the combination response shape plots. xviii

GLC conducted adaptive dose matrix studies.

DBK, DB and NJR conceived the idea of applying an evolutionary algorithm to select combinations of reagents for inhibition of IL-1β expression, wrote the grant application, co-investigated / co-supervised BGS and advised on experimental design.

JK supervised RA, generally advised on evolutionary algorithms and multiobjective optimization, and with RA evaluated and selected the evolutionary algorithm (IBEA) applied here. RA wrote the IBEA code in Java, helped implement its use and generated post-hoc reagent fitness measurements.

PM supervised BGS and advised on experimental design and data analysis.

All authors contributed to the writing of the manuscript and approved its final form.

Competing Financial Interests Statement

DBK transferred his responsibility as Principal Investigator of BBSRC Grant

BB/F018398/1 to NJR and PM prior to taking up his appointment as Chief Executive of the BBSRC on 1 st October 2008. NJR is a non-executive director of AstraZeneca. BGS as a former employee of AstraZeneca and Eli Lilly and Company is a shareholder of stock from both companies.

References

1. Nathan, C. & Ding, A. Nonresolving inflammation. Cell 140 , 871-82 (2010). 2. Dinarello, C.A. Interleukin-1 in the pathogenesis and treatment of inflammatory diseases. Blood 117 , 3720-3732 (2011). 3. Dinarello, C.A. Biologic basis for interleukin-1 in disease. Blood 87 , 2095-147 (1996). 4. Luheshi, N.M., Rothwell, N.J. & Brough, D. Dual functionality of interleukin-1 family cytokines: implications for anti-interleukin-1 therapy. Brit J Pharmacol 157 , 1318-29 (2009). 5. Relton, J.K. & Rothwell, N.J. Interleukin-1 receptor antagonist inhibits ischaemic and excitotoxic neuronal damage in the rat. Brain Res Bull 29 , 243-6 (1992). 6. Lu, Y.-C., Yeh, W.-C. & Ohashi, P.S. LPS/TLR4 signal transduction pathway. Cytokine 42 , 145-51 (2008). 7. Hopkins, A.L. Network pharmacology: the next paradigm in drug discovery. Nat Chem Biol 4, 682-690 (2008).

xix

8. Fell, D.A. & Thomas, S. Physiological control of metabolic flux: the requirement for multisite modulation. Biochem J 311 ( Pt 1), 35-9 (1995). 9. Jeon, Y.J. et al. Dexamethasone inhibits IL-1beta gene expression in LPS- stimulated RAW 264.7 cells by blocking NF-kappaB/Rel and AP-1 activation. Immunopharmacology 48 , 173-183 (2000). 10. Léhar, J., Stockwell, B.R., Giaever, G. & Nislow, C. Combination chemical genetics. Nat Chem Biol 4, 674-81 (2008). 11. Feng, Y., Mitchison, T.J., Bender, A., Young, D.W. & Tallarico, J.A. Multi- parameter phenotypic profiling: using cellular effects to characterize small- molecule compounds. Nat Rev Drug Discov 8, 567-78 (2009). 12. Paolini, G.V., Shapland, R.H., van Hoorn, W.P., Mason, J.S. & Hopkins, A.L. Global mapping of pharmacological space. Nat Biotechnol 24 , 805-15 (2006). 13. Coello, C.A.C., Lamont, G.B. & Veldhuizen, D.A.V. Evolutionary algorithms for solving multi-objective problems , (Springer, 2007). 14. Handl, J., Kell, D.B. & Knowles, J. Multiobjective optimization in bioinformatics and computational biology. IEEE/ACM Trans Comput Biol Bioinform 4, 279-92 (2007). 15. Knowles, J., Corne, D. & Deb, K. Multiobjective problem solving from nature: from concepts to applications , (Springer, Heidelberg, 2008). 16. Bäck, T., Fogel, D.B. & Michalewicz, Z. Handbook of evolutionary computation , (Institute of Physics Pub., 1997). 17. Goldberg, D.E. The design of innovation: lessons from and for competent genetic algorithms , (Kluwer Academic Publishers, 2002). 18. Knight, C.G. et al. Array-based evolution of DNA aptamers allows modelling of an explicit sequence-fitness landscape. Nucleic Acids Res 37 (2008). 19. O'Hagan, S., Dunn, W.B., Brown, M., Knowles, J.D. & Kell, D.B. Closed-loop, multiobjective optimization of analytical instrumentation: gas chromatography/time-of-flight mass spectrometry of the metabolomes of human serum and of yeast fermentations. Anal Chem 77 , 290-303 (2005). 20. Knowles, J. Closed-Loop Evolutionary Multiobjective Optimization. IEEE Comput Intell M 4, 77-91 (2009). 21. Zitzler, E. & Künzli, S. Indicator based selection in multi-objective search. in Parallel problem solving from nature - PPSN VIII : 8th international conference, Birmingham, UK, September 18-22, 2004 : proceedings (ed. Yao, X.) 832 - 842 (Springer, Berlin; [London], 2004). 22. Allmendinger, R. & Knowles, J. Analysis of Several Evolutionary Algorithms on the Noisy Three-Objective Chemical Mixture Optimization Problem. Technical report MLO-12009, 1 - 9 (2009). http://www.cs.man.ac.uk/~allmendr/publications.html 23. Allmendinger, R. & Knowles, J. Evolutionary Optimization on Problems Subject to Changes of Variables. in Parallel Problem Solving from Nature – PPSN XI , Vol. 6239 (eds. Schaefer, R., Cotta, C., Kolodziej, J. & Rudolph, G.) 151-160 (Springer Berlin / Heidelberg, 2010). 24. Kell, D. Iron Behaving Badly: Inappropriate Iron Chelation as a Major Contributor to the Aetiology of Vascular and Other Progressive Inflammatory and Degenerative Diseases. BMC Med Genomics 2, 2 (2009). 25. Kell, D.B. Towards a unifying, systems biology understanding of large-scale cellular death and destruction caused by poorly liganded iron: Parkinson's, Huntington's, Alzheimer's, prions, bactericides, chemical toxicology and others as examples. Arch Toxicol 84 , 825 - 889 (2010). 26. Puneet, P. et al. SphK1 regulates proinflammatory responses associated with endotoxin and polymicrobial sepsis. Science 328 , 1290-4 (2010).

27. Greco, W.R., Bravo, G. & Parsons, J.C. The search for synergy: a critical review from a response surface perspective. Pharmacological reviews 47 , 331-85 (1995). 28. Léhar, J. et al. Synergistic drug combinations tend to improve therapeutically relevant selectivity. Nat Biotechnol 27 , 659-66 (2009). 29. Zimmermann, G.R., Léhar, J. & Keith, C.T. Multi-target therapeutics: when the whole is greater than the sum of the parts. Drug Discov Today 12 , 34-42 (2007). 30. Ejim, L. et al. Combinations of antibiotics and nonantibiotic drugs enhance antimicrobial efficacy. Nat Chem Biol 7, 348-50 (2011). 31. Knight, Z.A., Lin, H. & Shokat, K.M. Targeting the cancer kinome through polypharmacology. Nat Rev Cancer 10 , 130-137 (2010). 32. Oda, K. & Kitano, H. A comprehensive map of the toll-like receptor signaling network. Mol Syst Biol 2, 2006 0015 (2006). 33. Bornkamp, B. et al. Innovative approaches for designing and analyzing adaptive dose-ranging trials. J Biopharm Stat 17 , 965-95 (2007). 34. Jia, J. et al. Mechanisms of drug combinations: interaction and network perspectives. Nat Rev Drug Discov 8, 111-128 (2009). 35. Fitzgerald, J.B., Schoeberl, B., Nielsen, U.B. & Sorger, P.K. Systems biology and combination therapy in the quest for clinical efficacy. Nat Chem Biol 2, 458-466 (2006). 36. Sun, C.-P. et al. Integrative systems control approach for reactivating Kaposi's sarcoma-associated herpesvirus (KSHV) with combinatory drugs. Integrative Biology 1, 123-130 (2009). 37. Wong, P.K. et al. Closed-loop control of cellular functions using combinatory drugs guided by a stochastic search algorithm. Proc Natl Acad Sci USA 105 , 5105- 10 (2008). 38. Feala, J.D. et al. Systems approaches and algorithms for discovery of combinatorial therapies. Wiley interdisciplinary reviews. Systems biology and medicine 2, 181-93 (2010). 39. Chatterjee, M.S., Purvis, J.E., Brass, L.F. & Diamond, S.L. Pairwise agonist scanning predicts cellular signaling responses to combinatorial stimuli. Nat Biotech 28 , 727-32 (2010). 40. Geva-Zatorsky, N. et al. Protein dynamics in drug combinations: a linear superposition of individual-drug responses. Cell 140 , 643-51 (2010). 41. Horackova, M., Ponka, P. & Byczko, Z. The antioxidant effects of a novel iron chelator salicylaldehyde isonicotinoyl hydrazone in the prevention of H 2O2 injury in adult cardiomyocytes. Cardiovas Res 47, 529-36 (2000). 42. Sindrilaru, A. et al. An unrestrained proinflammatory M1 macrophage population induced by iron impairs wound healing in humans and mice. J Clin Invest 121 , 985- 997 (2011). 43. Hammaker, D. & Firestein, G.S. "Go upstream, young man": lessons learned from the p38 saga. Ann Rheum Dis 69 , 77-82 (2010). 44. Dobson, P.D. & Kell, D.B. Carrier-mediated cellular uptake of pharmaceutical drugs: an exception or the rule? Nat Rev Drug Discov 7, 205-20 (2008). 45. Kell, D.B. & Oliver, S.G. Here is the evidence, now what is the hypothesis? The complementary roles of inductive and hypothesis-driven science in the post- genomic era. Bioessays 26 , 99-105 (2004). 46. Scully, P. et al. Plasma cytokine profiles in females with irritable bowel syndrome and extra-intestinal co-morbidity. Am J Gastroenterol 105 , 2235-43 (2010). 47. Löfdahl, C.-G. COPD and co-morbidities, with special emphasis on cardiovascular conditions. Clin Respir J 2 Suppl 1 , 59-63 (2008). 48. Akella, L.B. & DeCaprio, D. Cheminformatics approaches to analyze diversity in compound screening libraries. Curr Op Chem Biol 14 , 325-30 (2010).

xxi

49. Tobinick, E.L. The value of drug repositioning in the current pharmaceutical market. Drug News Perspect 22 , 119-25 (2009). 50. Zitzler, E. et al. Evolutionary Multi-objective Ranking with Uncertainty and Noise. in Evolutionary Multi-Criterion Optimization , Vol. 1993 329-343-343 (Springer Berlin / Heidelberg, 2001).

xxii

Figure legends

Figure 1. Combinatorial evolutionary inhibition of IL-1β expression (Loop 1, clockwise, black arrows). Known drugs were tested alone before being used at a single concentration

(3µM) in a chemical library. Initialization of the Indicator Based Evolutionary Algorithm

(IBEA) creates a random selection of combinations that are incubated with stimulated cells before measurement of cell death (LDH release) and IL-1β expression. Evaluation of these data against the number of compounds in the combination (All data n = 3) is performed by

IBEA prior to a new generation of combinations being computed and tested. After 11 generations, concentration-dependent optimization (Loop 2) of five top-ranked reagents was undertaken. Synergy was detected in novel dual-combinations.

Figure 2. Analysis of successive generations (Generations 1; initialization, 5 and 10) of reagent combinations reveals their convergence to a subset of highly effective combinations reflecting the inhibition of IL-1β expression with concomitant decreases in

LDH release and the number of member reagents. All data presented are the means of 3 determinations. Data points appearing as zero on the number of (#) reagents axis were reflective of positive control responses (LPS (1 µg / mL) and DMSO (0.5 % v/v).

Figure 3. Population average rank for inhibition of IL-1β expression (top left), number of component reagents in combinations (top right), LDH release (bottom left) and overall

IBEA hypervolume (bottom right) respectively. Error bars are the standard errors. The

IBEA hypervolume is a composite (see Methods; Implementation of the Indicator Based

Evolutionary Algorithm (IBEA)) of the performance of the different generations with regard to the three objectives, a smaller number being better.

xxiii

Figure 4. Analysis of all EA generations (1 – 11) in the presence (top) and absence

(bottom) of SB203580 yielded a rank order for the fitness contribution (see Methods; Post- hoc analysis of IBEA search and calculation of reagent fitness) of each reagent within the library. Only five top ranked reagents are displayed here either alone (i.e. single) or in double or triple combinations in the presence and absence of SB203580.

Figure 5. Concentration-dependent adaptive dose matrix optimization of paired reagent combinations was achieved by adaptively changing the concentrations of reagents after assessing the inhibition of IL-1β expression. A p38 MAPK inhibitor (SB203580) and an

Iκ Kinase inhibitor (IKKi) (a) or an iron chelator (SIH) (d) were assessed alone and as paired combinations. Potential synergy of the SB203580 & IKKi (c) and SB203580 &

SIH (f) combinations were revealed by subtraction of simple additive effects ((b) and (e); calculated from single reagent data in the absence of the other reagent respectively) of the respective combinations from the experimental data (a) & (d). Synergistic inhibition of IL-

1β expression was revealed with the combinations of SB203580 & IKK i (c) and SB203580

& SIH (f) as an ‘extra’ inhibition additional to the additive inhibitions of the individual reagents taken together.

xxiv

FIGURES

Page ii Figure 1

Page iii Figure 2

Page iv Figure 3

Page v Figure 4

Page vi Figure 5 (a-c)

Page vii Figure 5 (d-f)

xxv

xxvi

xxvii

xxviii

xxix

a) b) c)

xxx

d) e) f)

xxxi

SUPPLEMENTARY INFORMATION

Efficient discovery of anti-inflammatory small molecule combinations using evolutionary computing

*Corresponding Author: [email protected] $These authors contributed equally to this work Character count: 32864

SUPPLEMENTARY METHODS 4 SINGLE REAGENT STUDIES ...... 4

Preparation of primary peritoneal macrophages ...... 4 Preparation of single reagents and lipopolysaccharide (LPS) stimulation of macrophages for assessment of inhibiton of IL-1βββ expression ...... 5

IBEA DIRECTED EVOLUTIONARY SEARCH OF COMBNATIONS INHIBITING IL-1β EXPRESSION ...... 6

Multi-objective optimization of a chemical library and the implementation of an evolutionary algorithm: IBEA...... 6 Construction, storage and maintenance of the chemical library ...... 8 Production of combinatorial chemical cocktails ...... 9 Preparation of J774.A1 macrophages ...... 10 Treatment of J774.A1 macrophages with chemical combinations ...... 10

MEASUREMENT OF LDH & IL-1β EXPRESSION ...... 11

Lactate dehydrogenase (LDH) cell viability measurements ...... 11 IL-1βββ ELISA ...... 11

REAGENT REMOVAL / ADDITION AND DATA ANALYSIS ...... 14

Removal and replacement of reagents within the chemical library during the IBEA search ...... 14 Data analysis ...... 14

SUPPLEMENTARY RESULTS ...... 16 THE EFFECT OF SINGLE REAGENTS ON IL-1β EXPRESSION IN PERITONEAL MACROPHAGES ...... 16

REAGENT & ASSAY QC / VALIDATION ...... 17

Reagent physicochemical properties ...... 17 Assessment of assay discrimination ...... 18

TABLES ...... 19 Supplementary Table 1. Identities and physicochemical properties of reagents populating the chemical libraries used in the assessment of the effects of single reagents and evolutionary (IBEA) directed combinations on LPS stimulated IL-1β expression in peritoneal macrophages and J774 macrophages respectively...... 21

Supplementary Table 2. Summary of screening parameters for the assay of the evolution directed (IBEA) semi-automated robotic production of chemical combinations inhibiting LPS stimulated J774.A1 macrophage IL-1β expression...... 25

FIGURES ...... 29 Supplementary Figure 1. Reagent dilutions and the percentage of DMSO in the semi- automated production of chemical combinations ...... 29

Supplementary Figure 2. Pharmacological activities of iron chelators, anti / pro- oxidants, MAPK inhibitors, NADPH inhibitors, mitochondrial uncouplers, PI3Kinase inhibitors, NF-κB inhibitors, a statin, calpain inhibitor, COX2 inhibitor, cathepsin B inhibitor and matrix metalloproteinase inhibitor on LPS (1µg / mL) stimulated IL-1β expression in mouse primary peritoneal macrophages...... 32

Supplementary Figure 4. Adaptive dose-matrix search of paired reagent combinations for potential synergy in their inhibition of IL-1β expression...... 41

Supplementary Figure 5. Dose-matrix search of the triple reagent combination (SB203580, IKKi and SIH) for potential synergy in its inhibition of IL-1β expression...... 44

Supplementary Figure 7. Pseudo-color heat mapping of positive control (LPS (1µg / mL) & DMSO (0.5 % v/v)) responses (% IL-1β expression) to plates of J774.A1 macrophages during the assay of chemical combinations from all generations (Gen 1 – Gen 11)...... 46

Supplementary Figure 8. Quantile-Quantile plots of the positive control (LPS & DMSO) responses plotted against theoretical quantiles drawn from a normal distribution confirmed the data were normally distributed...... 47

REFERENCES ...... 48

Supplementary dataset 1 …………………………………………………………. 49

SUPPLEMENTARY METHODS All procedures, protocols and methods were carried out under aseptic conditions where deemed necessary.

SINGLE REAGENT STUDIES

Preparation of primary peritoneal macrophages

Batches (4 per batch) of C57BL/6J male mice (6 - 8 weeks old) were sacrificed by rising carbon dioxide asphyxiation in accordance with schedule 1 of the Animals (Scientific

Procedures) Act 1986 as issued by the Home Office, U.K. Slow intra-peritoneal (i.p) injection of pre-warmed (37°C, 8 mL) supplemented Dulbeccos Modified Eagles Medium

(DMEM: penicillin and streptomycin (5 µg / mL); and foetal calf serum; 2.5%) was performed into the abdominal area of cadavers. Peritoneal lavage was carried out and an incision made into the abdominal surface to create a pocket to drain the lavage fluid.

Lavage fluid was aspirated using a hypodermic syringe and titrated into a conical tube (15 mL).

A cell count of peritoneal macrophages was conducted using an Improved Neubauer haemocytometer. Macrophages were discriminated (large, circular, phase bright) from other cell types by their morphology. Lavage fluid was centrifuged (250g, 5 min) and the resultant pellet re-suspended in fresh, pre-warmed (37°C) DMEM to a cell concentration of

5.0 X 10 5 cells / mL. Aliquots (200 µL) of cell suspension were subsequently pipetted into

96-well plates (Costar, Corning; 3596) to ensure 100,000 cells / well. Plates of cells were maintained at 37°C in a humidified 95 % O 2 / 5 % CO 2 atmosphere for up to 96 h prior to use.

Preparation of single reagents and lipopolysaccharide (LPS) stimulation of macrophages for assessment of inhibiton of IL-1βββ expression

Stock reagents (Supplementary table 1) were prepared gravimetrically wherever possible, reconstituted to 20 mM in dimethylsulphoxide (DMSO) and serially diluted in log 10 spacings in DMSO down to concentrations of 2 µM. The five stock concentrations of reagent were further diluted (1 : 20) into DMEM at 10 X their final concentration (0.1 µM

- 1 mM). Peritoneal macrophages were exposed to reagents for 30 min prior to LPS stimulation. LPS (Sigma, serotype O26: B6) was diluted from a 1 mg / mL stock (in

Phosphate Buffered Saline; PBS). LPS stimulated reactions were terminated after 4 h by addition of homogenising buffer (pH 7.6, 0.02 % NaN) supplemented with Triton X-100 (1

% v/v) and protease inhibitor cocktail (1 % v/v; Calbiochem). Plates were left on ice with regular agitation (to ensure maximal lysis of cells) for 20 min prior to freezing.

IBEA DIRECTED EVOLUTIONARY SEARCH OF COMBNATIONS INHIBITING IL-

1β EXPRESSION

Multi-objective optimization of a chemical library and the implementation of an evolutionary algorithm: IBEA.

A multi-objective optimization problem (MOOP) of generic form can be defined as 1:

Minimize

ρ ρ ρ ρ ρ = Κ f x)( ( f1 x),( f 2 x),( , f m (x)). subject to x in X (1)

ρ where x = (x1,...,x n) is a solution vector (here representing a particular combination of reagents) consisting of n decision variables (i.e. different reagents) for which values must be found and X the feasible search space. The function f:X → Y represents a mapping from

X into the objective space Y in Rm. In general, a MOOP admits no single optimal solution, but rather a set of optimal trade-off solutions exist (also called Pareto-optimal solutions), none of which can be improved in all objectives simultaneously. The aim is to find this solution set, or an approximation of it, called the approximation set.

Our solution vectors represent whether or not to include a particular reagent in a combination or not (the concentration of reagents is fixed), and there are no constraints ρ restricting the search space, hence the space X is binary or x ∈ }1,0{ n where n is the number of reagents. A 1 in the vector at position j means that the jth reagent is included in the cocktail, a 0 means do not include it. There are three objectives fi, i = 1, 2, 3, to be minimized: IL-1β expression, LDH release and the number of component reagents in a given combination. The aim of the search process is to find reagent combinations that display the best trade-off between these objectives.

The field of evolutionary computation has proposed several techniques to solve MOOPs efficiently: after preliminary experimentation, we settled on using the indicator-based EA

(IBEA) 2. The aim of IBEA is find a set of Pareto-optimal solutions that that maximizes some unary performance measure (indicator), which here is the hypervolume of the objective space dominated by the approximation set.

Our experimental run of IBEA used the parameters given in Supplementary Table 2

To monitor the progress of the evolutionary search, we use a population average rank measure (Figure 1(C)). This is a nonparametric statistic related to the familiar rank sum tests, Mann-Whitney U and Kruskal-Wallis. All reagent combinations evaluated up to the current generation G are labeled by the generation in which they were tested. These samples are then ranked by fitness, and, for each generation g=1 to G, the average of the ranks associated with generation g is computed. The standard error of these ranks is also shown in the plots. This ranking statistic is computed separately for each of the three objectives, and also for the IBEA indicator (i.e. the three objectives combined).

Construction, storage and maintenance of the chemical library

Reagents were purchased from either Sigma-Aldrich, Tocris, Chembridge Cheng-Du

Biopurify Phytochemicals or Merck Chemicals and displayed purities ≥ 95 % as assessed by HPLC or TLC (except SIH; ≥ 90 % assessed by NMR). All compounds were prepared wherever possible gravimetrically and dissoluted into DMSO (Supplementary table 1) to a stock concentration of 15 mM. Subsequently, reagents were pipetted into a DMSO resistant, 2.0 mL, 96 deepwell plate (Starlab, E2896-0200) and sealed with a chemically resistant silicone mat (Starlab, E2896 3200) before being stored in an airtight plastic box partially filled with silica gel (Fisher Scientfic, S/0761/53) and kept in the dark at room temperature. Stock reagents were transferred periodically to new deep well plates to ensure their pharmacological integrity. A list of reagents and associated information (i.e. physicochemical properties) is given in Supplementary table 1.

Production of combinatorial chemical cocktails

The generation of poly-combinatorial mixtures of compounds was conducted using a

Sciclone ALH 3000 Liquid Handling Workstation under the control of Maestro and

HitPicking (v9.0) software (Both Caliper Life Sciences) running on a Windows XP platform on a personal computer. Compound combinations were specified in a 'Hitlist.csv' file; each line specifying the source and destination plate locations on the ALH3000 deck, source and destination well identities (e.g. D1, H12) and the volume (in µL) of compound required for transfer from the source to the destination plate. Populations of combinations and vehicle wells were constrained to 50 and 5 wells per plate respectively (with an additional 5 empty wells per plate) and randomly assigned to positions D1 – H12 across the plate to control for position-dependent plate effects (Supplementary figure 7) in accordance with the following equation (2).

= − −1 Pi (k n) (2)

th Where Pi is the probability of the i well position being assigned to one of the following treatments: a combination well, vehicle well or empty well. The total number of k wells available (i.e. in this instance; 60) and n the number of wells currently assigned to any treatment. Low volume ( ≥ 5 µL) dispense operations were executed after higher volumes of solvent ( ≥ 90 µL) had already been pipetted to ensure complete dispensing of these low volumes. Combinations were tested at a final concentration of 3 µM (The detail behind the serial dilutions of combinations and the % solvent at each step of the transfer are detailed in Supplementary figure 1).

Preparation of J774.A1 macrophages

J774.A1 macrophages were a kind gift from Prof. Anne-Marie Suprenant to DB’s lab, used between passages 12 – 19 and maintained in Dulbeccos modified eagles medium (DMEM) supplemented with high serum (10%, FBS) and penicillin / streptomycin (5 µg / mL).

Cells were split once 80 % confluency had been attained. Briefly, cells were removed from the base of a T75 flask (Costar) by aspirating existing media and replacing with a pre- warmed (37°C) aliquot (5 mL) of supplemented DMEM prior to using a cell scraper to remove the cell monolayer. After counting the resulting cell suspension was centrifuged

(250 g, 5 min), the supernatant discarded and the resultant pellet re-suspended in fresh, pre-warmed (37°C) high serum DMEM to a cell concentration of 6.5 X 10 5 cells / mL.

Aliquots (200 µL) of cell suspension were subsequently pipetted into 96-well plates

(Costar, Corning; 3596) and were maintained at 37°C in a humidified 95 % O 2 / 5 % CO 2 atmosphere 24 h prior to use.

Treatment of J774.A1 macrophages with chemical combinations

Previously prepared J774.A1 macrophages were treated with either chemical combinations

(all component compounds at 3 µM) or IKK inhibitor (100 µM) or DMSO (0.5 % v/v) alone for 10 minutes prior to stimulation with LPS (1 µg / mL) or PBS (0.1% v/v).

Subsequently, cells were maintained at 37°C in a humidified 95 % O 2 / 5 % CO 2 atmosphere for 2h. After this time, triplicate wells on each plate were exposed to lysis buffer (20 µL, 1X final concentration, Promega) prior to being returned to the incubator for a remaining ten minutes. Immediately after this time, aliquots (50 µL) of well supernatants were taken and pipetted into a new 96-well plate for the measurement of LDH release (see below). Termination of IL-1β expression and lysis of cells was conducted identically to that during the single reagent studies.

MEASUREMENT OF LDH & IL-1β EXPRESSION

Lactate dehydrogenase (LDH) cell viability measurements

This protocol is adapted from the recommended procedure provided with the CytoTox™

96 Non-Radioactive Cytotoxicity Assay kit (Promega ; G1780). Well supernatants (50 µL aliquots) were pipetted into a fresh 96-well plate prior to the addition of aliquots (50 µL) of substrate mix and covering and wrapping of the plate with laboratory film (Parafilm) and foil respectively. The plate was placed on a mini-orbital shaker (50 rpm) for 30 minutes and subsequent addition of stop solution (acetic acid; 50 µL) added to each well prior to reading on a plate reader (Biotek Synergy HT; single reagent studies or FLUOstar Omega for combinatorial studies respectively) at an absorbance of 490 nm (A 490nm ). During single reagent studies, treated wells (LPS and compounds) were compared to control wells

(DMSO alone) to assess if there had been a reduction in cell viability. Assessment of treated wells during the combinatorial study was conducted by expressing the A490nm of each well to that observed from the mean A490nm (of triplicate wells) that had been exposed to lysis buffer and then normalising this value relative to that seen for positive control

(quintuplicate) wells.

IL-1βββ ELISA

This protocol is adapted from the recommended procedure provided with the mouse IL-

1β/IL-1F2 DuoSet ELISA development kit (R&D Systems; DY401). Coating of 96-well plates (Maxisorp) with aliquots (50 µL) of rat anti-mouse IL-1β capture antibody (cAb; 4.0

µg / mL) was conducted and the plates left on a mini orbital shaker (50 rpm) overnight to ensure an even coating of cAb on all wells prior to the analysis of macrophage lysates.

Plates coated with cAb were aspirated of their cAb solution, washed three times with wash buffer (PBS with 0.05 % v/v Tween 20) after which any remaining well contents were blotted onto an absorbent paper towel until dry. This procedure was followed for all wash steps forthwith. Following this, blocking buffer (PBS and 1 % w/v BSA, Sigma) was aliquoted (50 µL) onto all wells, after which plates were covered with laboratory film and placed on a mini orbital shaker (50 rpm) until dilutions for the IL-1β standard curve had been made.

All IL-1β standards were made by reconstitution into lysis buffer (homogenising buffer supplemented with TX-100; 1% v/v). A recombinant mouse IL-1β standard (360 ng / mL,

3 µL) was diluted (537 µL) to produce a final concentration of 2 ng / mL of a top standard

(S1), this was further serially diluted by halving its concentration, this procedure was repeated to produce 9 standards in total, descending to a concentration of 7.8 pg / mL (S9).

The blocked plate was subsequently aspirated of the blocking buffer prior to the aliquoting

(50 µL) of samples and standards onto the plate Plates were covered with laboratory film and placed on a mini orbital shaker (50 rpm) for 2 hours at room temperature (RT) prior to detection.

Detection of bound IL-1β was carried out by aspirating plate wells of their samples and standards. A biotinylated goat anti-mouse IL-1β detection antibody (dAb; 108 µg / mL) was diluted (10 µL) in blocking buffer (7.5 mL) to furnish a final concentration of (144 pg

/ mL) and aliquoted (50 µL) onto all wells prior to covering with laboratory film and placing this plate on a mini orbital shaker (50 rpm) for 1 h. Subsequently, plates were aspirated of their dAb solution, Streptavidin-Horse Radish Peroxidase (Strep-HRP) solution (50 µL) was diluted into blocking buffer (10 mL) and aliquoted (50 µL) onto wells before the plate was covered with laboratory film and foil and placing on a mini- orbital shaker (50 rpm) for 20 mins. 3, 3', 5, 5' tetramethyl benzidine (TMB) substrate 12 reagents A and B were mixed in equal volumes (2.5 mL : 2.5 mL) in a foil covered tube

(15 mL), plates were aspirated of their Strep-HRP solution and aliquots (50 µL) of the mixed TMB substrate reagents added to all wells of the plate, covered with laboratory film and foil and left on a mini-orbital shaker for 20 min. Finally, plates were incubated with aliquots (50 µL) of dilute sulphuric acid (1M) and the resultant chromophore measured at an absorbance of 450 nm (A 450nm ) and background subtracted for non-specific fluorescence by measuring at 570 nm (A 570nm ) on a plate reader (Biotek Synergy HT).

Data analysis was conducted by normalizing absorbance readings to LPS (1 µg / mL) and

DMSO (0.5 % v/v) positive controls in order to ascertain % inhibitions / potentiations of combinations. In instances where the majority of wells had not exceeded the IL-1β concentration of the top standard (2 ng / mL) measured, a standard curve was constructed and IL-1β concentrations interpolated for sample wells. However, in all cases % changes versus positive control responses are quoted.

REAGENT REMOVAL / ADDITION AND DATA ANALYSIS

Removal and replacement of reagents within the chemical library during the IBEA search

During the course of the evolutionary search it was necessary to remove and introduce reagents respectively into the chemical library. A simulation study revealed that it is possible to substitute small numbers of reagents (< 5) without undue effects on the course of the evolutionary search 3. To this end, solutions where these single reagents were present after generation 3 were discarded from the retained elite solutions and a new generation (4) built. Similarly, Epigallocatechin gallate (ECGC) and Simvastatin were added to the chemical library from generation 4. Melatonin, was not introduced until generation 10, in order to compensate for this reagents late arrival into the library it was assigned a higher probability of selection (1 / 5) across all bits present.

Data analysis

Quantile-Quantile plots of the positive control (LPS & DMSO) responses plotted against theoretical quantiles drawn from a normal distribution confirmed the data were normally distributed (Supplementary Figure 8) 4.

Means and standard errors of the mean (SEM) of data are given as a function of the number of preparations of cells (where primary cells were used). That is 2 - 4 animals were used per preparation or plate of cells (n=1). Measurements were averaged across a minimum of triplicate wells per batch. Significance was assessed using a one-way

ANOVA and appropriate post-hoc multiple comparison tests reporting significance at the

P<0.05(*), P<0.01 (**) and P<0.0001 (***) levels.

Means and standard error of the mean of data are given as a function of the number of plates of cells per generation tested and were used for making inferences about the relative

14 efficacy of combinations during the evolution directed and dose-matrix adaptive search strategies. A statistical measure of assay quality; Z’-factor 5 was calculated retrospectively for each generation of combinations tested in accordance with the following equation (3).

) ) (3 σ p + σ ) Z'= 1− ) ) n (3) µ − µ p n

) ) Where Z’ is an estimate of the size of the statistical effect, σ p and σ n are the sample ) ) µ µ standard deviations of the positive and negative controls respectively and p and n .the sample means for positive and negative controls.

SUPPLEMENTARY RESULTS

THE EFFECT OF SINGLE REAGENTS ON IL-1β EXPRESSION IN PERITONEAL

MACROPHAGES

We characterized the potency and efficacy of single reagents selected to affect nodes (i.e. signalling proteins and processes respectively in this context)within the TLR4 signalling network and thought important for the expression of IL-1β6 in LPS-stimulated murine peritoneal macrophages. In total, 24 different reagents were assessed over a 0.01 – 100

µM concentration range from the following pharmacological classes: NF-κB, MAPK and

PI3kinase inhibitors, pro / anti-oxidants and iron chelators (which indirectly effect antioxidant activity 7,8 ). In summary, VK-28, α-tocopherol succinate, PDTC, SB203580,

LY294002, U0126 and IKK inhibitor III were effective inhibitors of IL-1β expression at concentrations between 1 and 100 µM (statistically significant at P<0.05 (*) and P<0.01

(**) levels by one-way ANOVA and Dunnett’s test; see supplementary methods and

Supplementary Fig. 2). Our subsequent experiments were designed to discover combinations of these reagents that have inhibitory effects at lower concentrations than when used in isolation. A fixed concentration of 3 µM was adopted, that was broadly sub- optimal in inhibiting IL-1β expression by any of these reagents used alone.

REAGENT & ASSAY QC / VALIDATION

Reagent physicochemical properties

In order to check whether there were common physicochemical parameters that may have influenced the biological activity of combinations within the search we attempted to mine these parameters for all reagents within the chemical library (Supplementary table 1). This was made possible by the recent provision of the publically accessible ChEMBL database of bioactive drug like small molecules ( https://www.ebi.ac.uk/chembl/ ) AlogP is a calculated measure of the partition coefficient or lipophillic character of a drug molecule and is a ratio of a molecule’s dispersion between organic solvent and aqueous phases.

There was no obvious trend in AlogP for the prioritized reagents (i.e. SB203580; 3.88,

SIH; 1.29 and Simvastatin; 4.63.) we derived from the IBEA search. However, all shared similar polar surface area (PSA) estimates, higher numbers of hydrogen bond accepting

(HBA) moieties versus their hydrogen bond donating (HBD) counterparts and no rule of 5

(RO5) violations. Unfortunately, owing to SKI-II’s absence from the ChEMBL database we were unable to derive parameters for this molecule. Clearly, assessment of physicochemical parameters can form another criterion for proposing reagents to populate chemical libraries.

Assessment of assay discrimination

The ability of the IBEA 2 directed semi-automated robotic assay of chemical combinations generated to inhibit IL-1β expression was scrutinized retrospectively via calculation of the

Z’-factor (see data analysis) statistic (Supplementary table 2 and Supplementary figure 6) to determine its ability to locate ‘true’ efficacious combinations and discriminate against trivial solutions. Although in two instances (at generations 2 and 5) this statistic fell beneath the acceptable range 9 across all other generations the assay was judged to have been of an acceptable or excellent standard and further to this the mean and median values for this statistic were 0.33 and 0.35 respectively, indicating overall that the assay was of an acceptable quality. In summary, this statistic in tandem with the canonical abilities of the

IBEA to locate combinations in the presence of 80 % noise in IL-1β expression (see main paper; methods) point to a robust assay for determining novel combinatorial solutions

TABLES

Name Supplier Cat# Purity Lot# Mwt Pharmacological chEBI Parent AlogP PSA HBA HBD RO5 (≥≥≥ %) Class* ID Mwt alpha-tocopherol Sigma T3251 96.0 078K1882 430.7 tioxidant An 103345 430.7 10.44 29.46 2 1 1 EGCG ChengDu Biopurify E07018 99.3 E090715 458.4 Antioxidant 167987 458.4 3.10 197.36 11 8 2 Phytochemicals alpha-tocopherol Sigma T3126 98.0 14H0063 530.8 Prooxidant 229103 0.8 53 10.30 72.83 5 1 2 succinate NAC Sigma A9165 99.0 80K13495 163.2 Antioxidant 881 120 163.2 -0.58 105.20 4 3 0 Rotenone Calbiochem 557368 98.0 D00055180 394.4 ioxidant Ant 102812 394.4 3.93 63.22 6 0 0 TEMPOL Tocris 3082 NA 1A/85506 172.2 Antioxidant 2703 29 173.3 0.32 43.70 3 2 0 Calpain III Calbiochem 208722 95.0 B29649 382.5 Enzyme inhibitor 131775 382.5 3.53 84.50 4 2 0 CA-074 Sigma C5732 99.0 NA 383.4 Enzyme inhibitor 60554 5 383.4 0.26 128.34 6 3 0 NS-398 Calbiochem 349254 98.1 D00017723 314.4 Enzyme inhibitor 439249 328.4 2.68 100.81 5 0 0 GM-6001 Calbiochem 364205 98.0 B69073 388.5 Enzymeinhibitor 123466 388.5 1.62 123.32 4 5 0 Melatonin Sigma M5250 99.0 098K1117 232.3 Hormone 03012 1 232.3 1.56 54.12 2 2 0 Desferrioxamine Sigma D9533 100.0 128K1205 656.8 Iron chelator 330444 560.7 -0.89 205.83 9 6 2 mesylate SIH Chembridge 5109995 90.0 610065 225.3 Iron chelator 434094 241.3 1.29 74.58 4 2 0 VK-28 Sigma V4264 97.0 034K46251 287.4 Iron chelator NA SR11302 Tocris 2476 99.3 1A/89800 376.5 MAPK pathway NA inhibitor PD98059 Calbiochem 513000 98.0 D00048465 267.3 MAPK pathway 150222 267.3 2.37 61.55 4 1 0 inhibitor U0126 Calbiochem 662005 98.0 D00058335 403.5 MAPK pathway 257660 380.5 2.18 202.26 8 4 0 monoethanolate inhibitor U0126 Tocris 1144 99.4 4A/93229 426.5 MAPK pathway 257660 monoethanolate inhibitor U0126 Tocris 1144 99.4 4A/97041 426.5 MAPK pathway 257660 monoethanolate inhibitor IRAK 1/4 inhibitor Calbiochem 407601 99.9 D00050155 395.4 MAPK pathway 449907 395.4 2.60 105.21 6 1 0 inhibitor IRAK 1/4 inhibitor Calbiochem 407601 99.9 D00057214 395.4 MAPK pathway 449907 inhibitor SP600125 Calbiochem 420119 98.0 D00002116 220.2 MAPK pathway 431500 220.2 2.99 45.75 2 1 0 inhibitor SB203580 Sigma S8307 98.0 029K4619 377.4 MAPK pathway 100250 377.4 3.88 77.84 3 1 0 inhibitor

Name Supplier Cat# Purity Lot# Mwt Pharmacological chEBI Parent AlogP PSA HBA HBD RO5 (≥≥≥ %) Class* ID Mwt SB203580 Tocris 1202 98.7 4A/93107 377.4 MAPK pathway 100250 inhibitor 2,4-dinitrophenol Sigma D19,850- 97.0 554589-308 184.1 mitochondrial 110392 184.1 1.38 111.87 5 1 0 1 uncoupler CCCP Sigma C2759 97.0 088K1511 204.6 mitochondrial 474035 204.6 2.47 71.97 4 1 0 uncoupler FCCP Tocirs 0453 99.0 2A/93205 254.2 mitochondrial 615435 254.2 3.92 81.20 5 1 0 uncoupler Apocynin Calbiochem 178385 99.9 D00055492 166.2 NADPH oxidase 364242 166.2 1.31 46.53 3 1 0 inhibitor diphenylene Calbiochem 300260 100.0 D00057661 314.6 NADPH oxidase 491098 279.1 iodonium (DPI) inhibitor IKK inhibitor III Calbiochem 401480 96.0 D00034862 255.3 NF-kB pathway 513905 255.3 1.00 68.23 4 2 0 (BMS-345541) inhibitor IKK inhibitor III Sigma B9935 98.0 077K462Z2 291.8 NF-kB pathway 513905 (BMS-345541) inhibitor PDTC Sigma P8765 99.0 1299391 / 164.3 NF-kB pathway 521017 147.3 2.48 74.13 2 1 0 106K1856 inhibitor CAPE Calbiochem 211200 97 NA 284.3 tyrosine kinase 271563 284.3 3.57 66.80 4 2 0 inhibitor Wortmanin Sigma W1628 99.0 029K4036 428.4 PI3-Kinase inhibitor 111985 428.4 1.71 109.11 7 0 0 LY294002.HCl Sigma L9908 98.0 017K4616 343.8 PI3-Kinase inhibitor 257152 307.3 3.34 38.76 4 0 0 LY294002.HCl Tocris 1130 99.4 3*A/93270 343.8 PI3-Kinase inhibitor 257152 NSC23766 Tocris 2161 99.0 2A/93470 558.0 Small GTPase 525412 421.6 3.22 91.99 7 3 0 inhibitor Y27632 Tocris 1254 98.0 12A/92215 329.3 Small GTPase 150239 247.3 1.11 68.01 3 2 0 dihydrochloride inhibitor di-methyl Calbiochem 310500 98.0 D00033039 327.6 sphingosine kinase 277963 327.6 5.79 43.70 3 2 1 sphingosine inhibitor SKI-II Calbiochem 567731 98.0 Unknown 339.2 sphingosine kinase NA inhibitor SKI-II Sigma S5696 99.0 077K46181 302.8 sphingosine kinase NA inhibitor Mevastatin Sigma M2537 95.0 0001423869 390.5 Statin 184814 390.5 3.97 72.83 5 1 0 Simvastatin Calbiochem 1965 98.2 2B/96951 418.6 Statin 238562 418.6 4.63 72.83 5 1 0 Genistein Sigma G6649 98.0 018K1203 270.2 tyrosine kinase 102658 270.2 2.14 86.99 5 3 0 inhibitor DMSO Sigma D8418 100.0 108K01864 78.1 Vehicle 110009 78.1 -0.32 36.28 1 0 0 TPEN Sigma P4413 99.0 078K1429 424.6 Zinc chelator NA Supple 21

Supplementary Table 1. Identities and physicochemical properties of reagents populating the chemical libraries used in the assessment of the effects of single reagents and evolutionary (IBEA) directed combinations on LPS stimulated IL-1β expression in peritoneal macrophages and J774 macrophages respectively. Reagents highlighted in bold were utilized in both single reagent and combination studies. Those reagents highlighted in red and italicised (i.e. CA-074, NS-398, GM-6001, PD98059 and CAPE) were not pursued during the course of the evolution driven search for combinations inhibiting IL-1β expression. Discrepancies between ‘Molecular Weight’ (supplied by the manufacturer) and ‘Parent Molecular Weight’ (provided courtesy of the ChEMBL database) fields are attributable to the salt form of the compounds (i.e. NS-398, Desferrioxamine Mesylate, SIH, U0126, diphenylene iodonium, LY294002 hydrochloride, IKK inhibitor III, NSC23766, Y27632 dihydrochloride). * Pharmacological class is an arbitrary assignation defining biological targets / processes where we assume these molecules will exert their effects in the TLR4 stimulated macrophage IL-1β signalling network. Cat#; catalogue number, Mwt; Molecular weight, chEBI ID; Chemical Entity of Biological Interest (database) Identity, AlogP; calculated partition coefficient, PSA; polar surface area, HBA; number of hydrogen bond accepting moieties, HBD; number of hydrogen bond donating moieties, RO5; number of rule of five violations, NA; Not Identified.

Category Parameter Description Assay Type of assay In vitro , evolution directed (IBEA) semi-automated robotic assay utilizing chemical combinations to inhibit LPS stimulated J774.A1 macrophage IL-1β expression.

Target IL-1β, UniProt ( IL1B_MOUSE; P10749 )

Primary measurement IL-1β expression measured via ELISA

Key reagents All antibodies and proteins were obtained from R&D systems.

Capture Antibody (Part 840134, 1 vial) - 720 µg / mL of rat anti-mouse IL-1β when reconstituted with 1.0 mL of PBS.

Detection Antibody (Part 840135, 1 vial) - 108 µg / mL of biotinylated goat anti-mouse IL- 1β when reconstituted with 1.0 mL of blocking buffer (see supplementary methods for recipe).

Standard (Part 840136, 1 vial) - 360 ng / mL of recombinant mouse IL-1β when reconstituted with 0.5 mL of blocking buffer.

Streptavidin-HRP (Part 890803, 1 vial) - 1.0 mL of streptavidin conjugated to horseradish peroxidase.

Assay protocol This protocol is adapted from the recommended procedure provided with the mouse IL-1β/IL-1F2 DuoSet ELISA development kit (R&D Systems; DY401)

Additional comments see supplementary methods Library Library size 33 reagents in total, never exceeding 30 reagents during the course of the iterative, IBEA directed search.

Library composition Small molecule reagents: drug like inhibitor molecules of biological targets (e.g. NF-κB and MAPK inhibitors) and processes (e.g. redox reactions – antioxidants and iron chelators).

Source Sigma-Aldrich, Tocris, Chembridge Cheng-Du Biopurify Phytochemicals or Merck Chemicals

Additional comments see supplementary table 1

Screen Format 96-well plates (Nunc; Maxisorp)

Concentration(s) tested All reagents evaluated either alone or in combinations at 3 µM.

Plate controls Randomized quintuplicate positive control wells (LPS, 1 µg / mL & DMSO, 0.5 % v/v) and negative control (C11) well (IKK inhibitor III, 100 µM)

Reagent/ compound dispensing system Sciclone ALH3000 (Caliper Life Sciences)

Detection instrument and software BioTek Synergy HT plate reader used for the measurement of the IL-1β ELISA absorbance (A450nm – A570nm ) readings under the control of Gen 5 software.

FLUOstar Omega (BMG LabTech) plate reader was used for the measurement of LDH reaction absorbance (A490nm ).

Assay validation/QC The median Z’ factor for the assay was 0.35

indicating the assay was acceptable.

Plate edge effects were assessed by the heat mapping of randomized positive control wells (n = 3 minimum / data point) (see supplementary figure 7).

Correction factors NA

Normalization J774.A1 cell IL-1β expression data was expressed as a percentage of the measured IL-1β concentration (pg / mL) interpolated from the standard curve for a given well / plate to the mean IL-1β concentration (pg / mL) derived from five positive control wells for that plate. Populations of generations were assayed in plates in triplicate.

Additional comments See supplementary information; supplementary figure 6.

Post-HTS analysis Hit criteria Objective evaluation of all combinations across all generations (i.e. generation 1 – 11) was performed by measuring (post-hoc) the fitness contribution of reagents from combinations and deriving a ranked list.

Hit rate Reagents were selected from top five ranked reagents when assessed either singularly or as double or triple combinations in either the presence or absence of the dominant solution (i.e. SB203580).

Additional assay(s) Concentration-dependent adaptive dose matrix search of paired combinations from five selected reagents.

Confirmation of hit purity and structure Manufacturer specification and certificates of analysis were used as confirmatory evidence of hit purity and structure.

Additional comments See supplementary information, particularly supplementary table 1 (confirming reagent identities & physicochemical properties) and supplementary figure 7 detailing the variation in positive control responses across the plate. IBEA parameter Population size 50 reagent combinations were generated by IBEA at settings each generation.

Parental selection Solutions (parents) for reproduction were selected using binary tournament selection with replacement.

Crossover probability Offspring were generated by applying uniform crossover with a probability of 0.7 to the selected parents. In situations where crossover is not applied, offspring are simply copied versions of the parents.

Per-bit mutation probability Each decision variable (reagent) of an offspring (reagent combination) was mutated or flipped independently with a probability of 1/33 (i.e. on average one solution variable was flipped).

Additional comments Recall that the initial population was generated at random such that each reagent combination contains on average 3 reagents; i.e. any reagent of the library was included with a probability of 1/33 into a reagent combination. A more detailed explanation of IBEA including a pseudocode of the algorithm can be found in Reference 8. We use the adaptive version of IBEA from that reference. The value of the scaling parameter, κ, was 0.05, and the hypervolume reference point used was (2,2,2), as in the original source reference.

Supplementary Table 2. Summary of the screening and EA parameter settings for the assay of the evolution directed (IBEA) semi-automated robotic production of chemical combinations inhibiting LPS stimulated J774.A1 macrophage IL-1β expression.

Generation Number of Mean number of Mean fractional combinations per component LDH release of generation showing reagents per combinations per ≥≥≥ 95% inhibition of generation generation IL-1βββ expression (rounded to integer) 2 1 3 0.96 4 4 5 1.01 6 8 5 1.08 7 3 4 0.99 8 1 4 1.00 9 27 3 1.05 10 3 5 1.00 11 4 4 0.97

Number of reagents per Mean inhibition of IL-1βββ Number of combinations combination expression for combinations where inhibition was already ≥≥≥ 95 % of the positive control response 7 98.51 4 6 99.10 2 5 99.09 8 4 99.20 16 3 98.93 13 2 100.00 5

Supplementary Table 3. Characteristics of reagent combinations found from the IBEA directed search that inhibited IL-1β expression in the LPS stimulated J774 macrophages by ≥ 95% of positive control responses. Overall, successive generations of reagent combination yielded decreases in the number of component reagents and fractional LDH release (Top). Despite decreases in component reagent number the search still located five very efficacious dual combinations (Bottom).

Generation Number of Mean number of Mean fractional combinations per component LDH release of generation showing reagents per combinations per ≥≥≥ 70% inhibition of generation generation IL-1βββ expression (rounded to integer) 1 8 4 1.06 2 5 5 1.03 3 32 4 1.02 4 12 5 1.02 5 18 4 1.01 6 29 4 1.04 7 27 4 1.06 8 24 4 1.04 9 28 4 1.05 10 32 3 1.01 11 18 3 0.98

Number of reagents per Mean inhibition of IL-1βββ Number of combinations combination expression for combinations where inhibition was already ≥≥≥ 70 % of the positive control response 8 90.52 1 7 91.51 10 6 89.92 18 5 91.93 35 4 92.29 57 3 90.20 58 2 87.88 34

Supplementary Table 4. Characteristics of reagent combinations found from the IBEA directed search that inhibited IL-1β expression in the LPS stimulated J774 macrophages by ≥ 70% of positive control responses. Overall, successive generations of reagent combination yielded decreases in the number of component reagents and fractional LDH release (Top). Similarly, despite

decreases in component reagent number the search still located thirty four dual combinations (Bottom).

FIGURES

Robot Manual

Mix Final

Stocks @ 15 mM

[Compound] 15 mM 600 µM 30 µM 3 µM DMSO (v/v) 100 % 100 % 5 % 0.5 %

Supplementary Figure 1. Reagent dilutions and the percentage of DMSO in the semi-automated production of chemical combinations

ααα-Tocopherol Succinate SIH VK-28 1.5 1.5 1.5

1.0 1.0 1.0 *

** ** ** 0.5 * 0.5 0.5 ** (fold induction vs. ctrl) vs. induction (fold (fold induction vs. ctrl) vs. induction (fold (fold induction vs. ctrl) vs. induction (fold β β β β β β β β β β β β 0.0 0.0 0.0 IL-1 IL-1 IL-1 1 0 1 0 O .1 10 0 S 10 S O 1 1 0 1 PS .01 0.1 10 BS P S 01 0. 10 PBS LPS 0.01 PB L 0 P L 0. 100 DMSO µ DM + DMS SIH ( µµM) VK-28 ( µµµM) ααα-Toc Succ ( µµµM) PS + LPS LPS + L

TEMPOL NAC PDTC 1.5 1.50 1.5

1.25

1.0 1.00 1.0

0.75

0.5 ** 0.50 ** 0.5 ** ** **

(fold induction vs. ctrl) vs. induction (fold ctrl) vs. induction (fold 0.25 ctrl) vs. induction (fold β β β β β β β β β β β β 0.0 0.00 0.0 IL-1 IL-1 IL-1 1 1 1 1 0 1 1 10 S 10 10 BS SO 0.1 100 B PS 0.1 10 BS SO 0.1 100 P LPS 0.0 P L MSO 0.0 P LPS 0.0 D + DM TEMPOL ( µµµM) NAC + DM PDTC ( µµµM) PS PS L LPS + L

CAPE SB203580 LY294002 1.50 1.5 1.5

1.25

1.00 1.0 1.0

0.75

* 0.5 0.50 0.5 ** ** ** (fold induction vs. ctrl) vs. induction (fold (fold induction vs. ctrl) vs. (fold induction

0.25 ctrl) vs. induction (fold β β β β β β β β β β β β 0.0

0.00 IL-1

IL-1 0.0 IL-1 1 1 1 1 0 S 10 00 1 S 1 0 0 PS SO .0 0.1 1 BS PS 0.1 100 0.1 1 0 PB L 0 P L 0.0 PB LPS 0.01 1 CAPE µµµ + DM µµµ PS + EtOH SB203580 ( M) PS LY294002 ( M) L LPS + H2O L

Wortmannin CA-074 Genestein 1.5 1.5 1.5

1.0 1.0 1.0

0.5 0.5 0.5 (fold induction vs. ctrl) vs. induction (fold ctrl) vs. induction (fold ctrl) vs. induction (fold β β β β β β β β β β β β 0.0 0.0 0.0 IL-1 IL-1 IL-1 S O 1 1 0 0 h O 1 h O .1 1 0 .0 0.1 1 0.1 10 00 0 1 PBS LP 0 10 ve LPS 0.01 1 ve LPS 0.01 100 DMS DMS DMS + µ + + Genestein S Wortmannin ( µµM) S CA-074 S LP LP LP

GM-6001 NS-398 Calpain inhibitor 1.5 1.5 1.5

1.0 1.0 1.0

0.5 0.5 0.5 (fold induction vs. ctrl) vs. induction (fold ctrl) vs. induction (fold ctrl) vs. induction (fold β β β β β β β β β β β β 0.0 0.0 0.0 IL-1 IL-1 IL-1 1 0 1 0 h 1 0 S 10 0 O 10 0 e O .1 10 veh P 0.1 1 veh 0.1 1 v S 0 10 L MSO 0.01 LPS MS 0.01 LPS 0.01 D + + D NS-398 S + DM Calpain inhibitor GM-6001 P LPS LPS L

Mevastatin Apocyanin PD98059 1.5 1.5 1.5

1.0 1.0 1.0

0.5 0.5 0.5 (fold induction vs.ctrl) induction (fold vs.ctrl) induction (fold (fold induction vs. ctrl) vs. induction (fold β β β β β β β β β β β β ** 0.0 0.0

0.0 IL-1 IL-1 IL-1 1 1 0 1 0 S O 10 00 S .1 1 S 1 00 veh .01 0.1 1 P SO 0 100 P 0.1 1 LP 0 PBS L M 0.01 PBS L 0.01 DMS Apocyanin + D µµµ Mevastatin S PD98059 ( M) P LPS + L LPS + DMSO

U0126 JNK Inhibitor IRAK 1/4 Inhibitor 1.5 1.5 1.5

1.0 1.0 1.0

0.5 * 0.5 0.5 ** (fold induction vs.ctrl) induction (fold vs.ctrl) induction (fold (fold induction vs.ctrl) induction (fold ** β β β β β β β β β β β β ** ** 0.0 0.0 0.0 IL-1 IL-1 IL-1 1 1 1 1 0 0 1 1 0 O 10 00 S O .1 1 0 10 0 PS S 0. 1 P S .0 0 1 .0 0.1 1 PBS L 0.01 PBS L M 0 PBS LPS 0 DM + U0126 ( µµµM) JNK Inhibitor ( µµµM) IRAK 1/4 Inhibitor PS L LPS + D LPS + DMSO

2,4-Dinitrophenol IKK Inhibitor III 1.5 1.5

1.0 1.0

0.5 0.5 (fold induction vs.ctrl) induction (fold vs.ctrl) induction (fold β β β β β β β β ** ** ** 0.0 0.0 IL-1 IL-1 1 1 1 0 1 0 10 S .1 1 PS .0 0. 10 P 0 100 PBS L MSO 0 PBS L 0.01 D + 2,4-Dinitrophenol µµµ S S + DMSO IKK Inhibitor III ( M) P LP L

Supplementary Figure 2. Pharmacological activities of selected reagents on LPS (1µg / mL) stimulated IL-1β expression in mouse primary peritoneal macrophages. Comparison of different treatments to positive control (LPS and DMSO) was assessed by means of a one-way ANOVA and a post-hoc Dunnett's multiple comparison test reporting significance at the P<0.05 (*) and P<0.01 (**) levels. Data are expressed as mean ± SEM (n=3 preparations).

Generation 1 250

200

150

expression 100 β β β β

IL-1 50 (% of & DMSO) LPS

0 i 1 2 3 4 5 6 7 8 9 0 0 0 0 0 O O K 1 11 12 13 14 15 16 17 18 19 2 21 22 23 24 25 26 27 28 29 3 31 32 33 34 35 36 37 38 39 4 41 42 43 44 45 46 47 48 49 5 S K M I D + + S + DMSS B LPS P LP

1 (vs. LPS & DMSO) fractional LDH LDH release fractional 0 1 2 3 4 5 6 7 8 9 O Ki 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 S K MSO M I D D + + + S S BS P LP P L Treatment / Combination

Generation 2 300

200 expression β β β β 100 IL-1 (% of & DMSO) LPS

0 i 1 2 3 4 5 6 7 8 9 1 1 1 1 O O K 10 1 12 13 14 15 16 17 18 19 20 2 22 23 24 25 26 27 28 29 30 3 32 33 34 35 36 37 38 39 40 4 42 43 44 45 46 47 48 49 50 S K M I D + + S + DMSS B LPS P LP

Generation 3 200

150

100 expression β β β β 50 IL-1 (% of DMSO) & LPS

0 O 1 2 3 4 5 6 7 8 9 5 2 1 8 7 S 10 11 12 13 14 1 16 17 18 19 20 21 2 23 24 25 26 27 28 29 30 3 32 33 34 35 36 37 3 39 40 41 42 43 44 45 46 4 48 49 50 IKKi DMSO + + S B LPS P LPS + DM

Generation 4 200

150

100 expression β β β β 50 IL-1 (% of & DMSO) LPS

0 i 1 2 3 4 5 6 7 8 9 0 0 0 0 0 O O K 1 11 12 13 14 15 16 17 18 19 2 21 22 23 24 25 26 27 28 29 3 31 32 33 34 35 36 37 38 39 4 41 42 43 44 45 46 47 48 49 5 S K M I D + + S + DMSS B LPS P LP

1 (vs. LPS & DMSO) fractional LDH LDH release fractional 0 O 1 2 3 4 5 6 7 8 9 S SO 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 M M D + IKKi + + D PS LPS PBSL Treatment / Combination

Generation 5 200

150

100 expression β β β β 50 IL-1 (% of & DMSO) LPS

0 i 1 2 3 4 5 6 7 8 9 0 0 0 0 0 O O K 1 11 12 13 14 15 16 17 18 19 2 21 22 23 24 25 26 27 28 29 3 31 32 33 34 35 36 37 38 39 4 41 42 43 44 45 46 47 48 49 5 S K M I D + + S + DMSS B LPS P LP

Generation 6 200

150

100 expression β β β β 50 IL-1 (% of & DMSO) LPS

0 i 1 2 3 4 5 6 7 8 9 0 0 0 0 0 O O K 1 11 12 13 14 15 16 17 18 19 2 21 22 23 24 25 26 27 28 29 3 31 32 33 34 35 36 37 38 39 4 41 42 43 44 45 46 47 48 49 5 S K M I D + + S + DMSS B LPS P LP

Generation 7 200

150

100 expression β β β β 50 IL-1 (% of & DMSO) LPS

0 i 1 2 3 4 5 6 7 8 9 0 0 0 0 0 O O K 1 11 12 13 14 15 16 17 18 19 2 21 22 23 24 25 26 27 28 29 3 31 32 33 34 35 36 37 38 39 4 41 42 43 44 45 46 47 48 49 5 S K M I D + + S + DMSS B LPS P LP

Generation 8 200

150

100 expression β β β β 50 IL-1 (% of & DMSO) LPS

0 i 1 2 3 4 5 6 7 8 9 0 0 0 0 0 O O K 1 11 12 13 14 15 16 17 18 19 2 21 22 23 24 25 26 27 28 29 3 31 32 33 34 35 36 37 38 39 4 41 42 43 44 45 46 47 48 49 5 S K M I D + + S + DMSS B LPS P LP

Generation 9 200

150

100 expression β β β β 50 IL-1 (% of & DMSO) LPS

0 i 1 2 3 4 5 6 7 8 9 0 0 0 0 0 O O K 1 11 12 13 14 15 16 17 18 19 2 21 22 23 24 25 26 27 28 29 3 31 32 33 34 35 36 37 38 39 4 41 42 43 44 45 46 47 48 49 5 S K M I D + + S + DMSS B LPS P LP

Generation 10 200

150

100 expression β β β β 50 IL-1 (% of & DMSO) LPS

0 i 1 2 3 4 5 6 7 8 9 0 0 0 0 0 O O K 1 11 12 13 14 15 16 17 18 19 2 21 22 23 24 25 26 27 28 29 3 31 32 33 34 35 36 37 38 39 4 41 42 43 44 45 46 47 48 49 5 S K M I D + + S + DMSS B LPS P LP

Generation 11 200

150

100 expression β β β β 50 IL-1 (% of& DMSO) LPS

0 e 4 6 i 8 II 0 n n n 6 n i O O n in C K 10 11 12 14 16 17 18 32 2 21 22 PI G IH 29 ne 34 ne i i 4 S o e A K 6 D S i 580 pol 302 at 12 002 at 059 / MS n K-28 N I KI- GC IKKi si stei m ynin 4 8 1 + IKKi e ist V S TPENE a-TOS am o 03 PDTC i c st U0 st K D t n Y27 a-TOC tempol x g te 29 C23766 + DM o po va Y va Ge Gen SR 11a e S PD9 LPS R SP600125SB203580 ortmanni rrio SB2 L m N SB203580IRA Melatonin e m si PBSLPS + W lsphin desf di-methy

Supplementary Figure 3. Efficacy of chemical combinations with respect to LPS stimulated (1µg / mL) IL-1β expression (upper plot per generation) and LDH release (lower plot per generation) in J774.A1 macrophages for all tested generations (Gen 1 – Gen 11). Vehicle (PBS; 0.1 % v/v) and positive control (LPS; 1µg / mL & DMSO; 0.5 % v/v) responses were tested in triplicate or quintuplicate per plate respectively. Negative control (LPS; 1µg / mL & IKKi; 100 µM, except Gen 11 where IKKi; 3 µM data was substituted) responses were tested in single wells per plate. Fifty combinations were tested per plate and all data is expressed as n = 3 plates per generation ± SEM.

Supplementary Figure 4. Adaptive dose-matrix search of paired reagent combinations for potential synergy in their inhibition of IL-1β expression. Combination response shape plots shown are for a fully enumerated search of all possible pair wise combinations of the five

reagents prioritized from the post-hoc analysis of the IBEA search (i.e. SB203580 & SIH, SIH & SKI-II, SB203580 & SKI-II, SB203580 & Simvastatin, SIH & Simvastatin, Simvastatin & SKI-II, IKKi &. Simvastatin, IKKi &. SKI-II, IKKi &. SIH). Three plots are shown (from left to right) representing: the experimental data; simple additive effects of the combination calculated from single reagent data in the absence of the other reagent; and detection of synergy in the paired combination by subtraction of additive effects from the experimental data. Pseudocolor mappings were performed by linear interpolation between samples. All data are expressed as n = 3 - 4 plates.

Supplementary Figure 5. Dose-matrix search of the triple reagent combination (SB203580, IKKi and SIH) for potential synergy in its inhibition of IL-1β expression. Combination response shape plots show the effect of increasing concentration (by row) of SIH (0, 0.1, 0.3, 1 and 3 µM) Three plots are shown (from left to right) representing: the experimental data; simple additive effects of the combination calculated from single reagent data in the absence of the other reagent; and detection of synergy in the paired combination by subtraction of additive effects from the experimental data. Pseudocolor mappings were performed by linear interpolation between samples. All data are expressed as n = 5 plates.

1.0 Excellent

0.5 Acceptable 0.0 2 3 4 5 6 7 8 9 10 11 Z'-Factor Generation -0.5 Unacceptable

-1.0

Supplementary Figure 6. Calculation of the Z’-Factor statistic retrospectively for the IBEA directed search revealed mean and median values of 0.33 and 0.35 respectively indicating overall that the assay was of an acceptable quality. The Z’-Factor was determined by four parameters (see data analysis); positive and negative control, mean and standard deviation respectively. At each generation there were 15 positive control wells and 3 negative control wells.

0.000

25.00 H 50.00

75.00

G 100.0

125.0

150.0 F Plate row

1 2 3 4 5 6 7 8 9 10 11 12 Plate column

Supplementary Figure 7. Pseudocolor heat mapping of positive control (LPS (1µg / mL) & DMSO (0.5 % v/v)) responses (% IL-1β expression) to plates of J774.A1 macrophages during the assay of chemical combinations from all generations (Gen 1 – Gen 11). Quintuplicate positive control responses were randomly assigned across wells (D1 – H12) for each generation tested. Data are expressed as a % of the mean positive control response of the plate they were assayed on (n=5 wells) with purple colors representing high positive control responses to mauve representing lower responses. Orange space represented wells that were not sampled.

REFERENCES

1. Deb, K. Multi-objective optimization using evolutionary algorithms , (John Wiley & Sons, New York, 2001). 2. Zitzler, E. & Kunzli, S. Indicator based selection in multi-objective search. in Parallel problem solving from nature - PPSN VIII : 8th international conference, Birmingham, UK, September 18-22, 2004 : proceedings (ed. Yao, X.) 832 - 842 (Springer, Berlin ; [London], 2004). 3. Allmendinger, R. & Knowles, J. Evolutionary Optimization on Problems Subject to Changes of Variables. in Parallel Problem Solving from Nature – PPSN XI , Vol. 6239 (eds. Schaefer, R., Cotta, C., Kolodziej, J. & Rudolph, G.) 151-160 (Springer Berlin / Heidelberg, 2010). 4. Crawley, M.J. Statistics : an introduction using R , xiii, 327 p. (John Wiley & Sons, Chichester, 2005). 5. Zhang, J.H., Chung, T.D. & Oldenburg, K.R. A Simple Statistical Parameter for Use in Evaluation and Validation of High Throughput Screening Assays. J Biomol Screen 4, 67-73 (1999). 6. Oda, K. & Kitano, H. A comprehensive map of the toll-like receptor signaling network. Mol Syst Biol 2, 2006 0015 (2006). 7. Kell, D. Iron Behaving Badly: Inappropriate Iron Chelation as a Major Contributor to the Aetiology of Vascular and Other Progressive Inflammatory and Degenerative Diseases. BMC Med Genomics 2, 2 (2009). 8. Kell, D.B. Towards a unifying, systems biology understanding of large-scale cellular death and destruction caused by poorly liganded iron: Parkinson's, Huntington's, Alzheimer's, prions, bactericides, chemical toxicology and others as examples. Archives of toxicology (2010). 9. Sui, Y. & Wu, Z. Alternative statistical parameter for high-throughput screening assay quality assessment. Journal of biomolecular screening : the official journal of the Society for Biomolecular Screening 12 , 229-34 (2007).

48 Generation # Combination # Combination reagent composition Objective 1_%IL-1b Objective 2_cell death Objective 3_# component reagents Reagent_1 Reagent_2 Reagent_3 Reagent_4 Reagent_5 Reagent_6 Reagent_7 Reagent_8 Reagent_9 P1 P2 P3 P1 P2 P3 1 1 Diphenylene iodoniumFCCP (DPI) 22.60 31.89 38.34 1.02 1.04 1.03 2 1 2 TPEN N-acetyl cysteine tempol SR 11302 CCCP 39.13 57.11 61.97 1.02 0.96 0.99 5 1 3 N-acetyl cysteine 110.53 125.92 118.35 1.00 0.96 1.09 1 1 4 rotenone SR 11302 PDTC 2-4-dinitrophenol mevastatin NSC23766 (Rac1 inhibitor) 43.64 59.31 58.66 1.00 1.00 1.02 6 1 5 alpha-tocopherol U0126 PD98059 61.30 71.90 85.43 0.99 0.98 0.98 3 1 6 VK-28 NSC23766 (Rac1 inhibitor) 110.53 104.65 120.11 0.99 1.00 0.99 2 1 7 DMSO 116.89 101.55 115.70 0.97 0.97 0.98 2 1 8 alpha-tocopherol N-acetyl cysteine U0126 mevastatin Y27632 dihydrochloride (ROCK inhibitor) 27.01 41.16 88.83 1.06 0.97 1.03 5 1 9 SIH wortmannin IRAK 1 FCCP NSC23766 (Rac1 inhibitor) 9.52 22.03 13.80 1.16 1.06 1.05 5 1 10 tempol apocynin LY294002 hydrochloride 193.03 224.10 227.19 1.00 1.10 0.99 3 1 11 SIH tempol SR 11302 NSC23766 (Rac1 inhibitor) 130.53 129.92 121.00 0.87 0.97 1.08 4 1 12 desferrioxamine 116.18 131.53 132.59 1.00 1.00 0.98 1 1 13 alpha-tocopherol LY294002 hydrochloride mevastatin 172.93 180.36 155.31 1.00 0.99 0.96 3 1 14 alpha-tocopherol wortmannin succinate CCCP 13.23 22.73 19.27 1.06 1.03 1.06 3 1 15 di-methyl sphingosine 81.43 83.93 87.99 0.98 0.98 0.98 1 1 16 U0126 PDTC NSC23766 (Rac1 inhibitor) 35.92 46.93 44.79 1.00 1.00 0.96 3 1 17 rotenone 58.00 61.52 86.28 0.97 0.99 1.01 1 1 18 tempol PDTC SKI-II 25.75 40.45 31.93 1.01 0.96 0.97 3 1 19 SKI-II 28.91 41.89 39.95 0.99 0.97 1.04 1 1 20 desferrioxamine SB203580 30.18 29.77 27.95 1.02 1.04 0.99 2 1 21 desferrioxamine wortmannin 28.91 32.60 27.95 1.07 1.05 1.09 2 1 22 TPEN alpha-tocopherol SKI-II succinate 50.78 45.48 45.59 1.03 0.97 0.98 3 1 23 wortmannin 23.23 29.77 27.16 1.06 1.09 1.04 1 1 24 LY294002 hydrochloride PD98059 genistein FCCP 20.72 33.31 29.54 1.02 1.02 1.02 4 1 25 SIH apocynin SB203580 15.72 25.54 22.42 0.99 1.00 1.00 3 1 26 N-acetyl cysteine LY294002 hydrochlorideIRAK 1 SKI-II di-methyl sphingosine2-4-dinitrophenol 11.99 21.33 21.63 1.10 0.96 1.05 6 1 27 alpha-tocopherol U0126 succinate IRAK 1 CCCP 11.99 19.94 13.80 1.06 0.99 1.08 4 1 28 apocynin IRAK 1 Y27632 dihydrochloride (ROCK inhibitor) 32.09 42.60 43.97 1.05 0.98 1.03 3 1 29 LY294002 hydrochlorideIRAK 1 di-methyl sphingosine 32.72 48.38 52.92 1.05 1.12 1.03 3 1 30 alpha-tocopherol tempol succinate Diphenylene SRiodonium 11302 (DPI) 61.96 86.20 98.28 1.00 1.02 1.03 4 1 31 apocynin SKI-II 32.09 35.44 35.12 1.00 1.04 1.14 2 1 32 DMSO 126.20 134.75 127.23 1.03 1.05 1.02 2 1 33 alpha-tocopherol tempol U0126 PD98059 82.79 68.18 61.14 1.04 1.02 0.99 4 1 34 DMSO 109.11 116.40 138.89 1.00 0.98 1.00 2 1 35 alpha-tocopherol LY294002 hydrochlorideIRAK 1 65.29 63.73 58.66 1.02 0.98 1.02 3 1 36 PD98059 di-methyl sphingosine genistein 76.02 83.17 82.04 1.01 1.01 1.01 3 1 37 U0126 SR 11302 CCCP 21.34 19.94 36.73 1.02 1.00 1.04 3 1 38 Diphenylene iodoniumSP600125 (DPI) SKI-II genistein 43.64 63.00 40.75 1.09 1.05 1.06 4 1 39 IKK inhibitor (BMS-345541) 54.71 51.27 62.79 1.03 0.98 1.04 1 1 40 alpha-tocopherol Diphenylene succinate wortmanniniodonium (DPI) di-methyl sphingosine 16.96 24.84 28.75 1.18 1.20 1.21 4 1 41 PDTC di-methyl sphingosine genistein FCCP 18.21 35.44 28.75 1.03 1.03 1.09 4 1 42 TPEN genistein 134.16 133.95 167.37 1.00 1.02 1.05 2 1 43 U0126 PDTC SKI-II 2-4-dinitrophenol 18.21 22.03 24.00 1.04 1.04 1.01 4 1 44 SIH alpha-tocopherol rotenone DiphenyleneIRAK iodonium 1 (DPI) SR 11302 PDTC 27.01 32.60 25.58 1.10 1.06 1.07 7 1 45 desferrioxamine 78.03 132.34 156.24 1.09 1.00 1.00 1 1 46 U0126 di-methyl sphingosine genistein 52.09 53.46 66.11 1.08 1.00 1.00 3 1 47 VK-28 wortmannin 16.34 34.74 25.58 1.19 1.15 1.10 2 1 48 tempol CCCP 27.01 44.76 43.17 1.00 0.98 1.02 2 1 49 LY294002 hydrochloride NSC23766 (Rac1 inhibitor) 114.06 105.43 252.46 1.07 1.05 1.03 2 1 50 genistein 110.53 142.05 101.74 1.01 0.99 1.06 1 2 1 LY294002 hydrochloride 289.30 202.58 156.58 0.99 1.01 0.98 1 2 2 U0126 PDTC NSC23766 (Rac1 inhibitor) 91.26 69.62 59.74 0.97 1.08 0.99 3 2 3 N-acetyl cysteineIRAK 1 SKI-II 2-4-dinitrophenol 35.74 35.50 22.13 0.98 1.04 0.97 4 2 4 desferrioxamine 142.39 82.81 92.67 1.00 1.00 0.98 1 2 5 alpha-tocopherol alpha-tocopherol U0126 succinate IRAK 1 CCCP mevastatin 23.26 24.66 7.51 0.96 1.08 1.03 6 2 6 diphenylene iodonium wortmannin (DPI) mevastatin 28.44 42.79 17.50 1.04 1.19 0.98 3 2 7 tempol U0126 PDTC SKI-II 20.15 37.32 8.28 1.03 0.99 0.98 4 2 8 wortmannin 2-4-dinitrophenol 23.26 41.87 22.90 1.00 1.01 1.00 2 2 9 alpha-tocopherol U0126 SKI-II di-methyl sphingosine genistein 2-4-dinitrophenol 65.35 25.56 0.00 1.02 1.01 0.98 6 2 10 diphenylene iodonium di-methyl (DPI) sphingosine 69.62 78.08 31.44 1.03 1.01 1.05 2 2 11 alpha-tocopherol rotenone apocynin 102.21 85.65 74.11 1.05 1.05 1.05 3 2 12 alpha-tocopherol wortmannin succinate 69.62 65.88 73.31 1.08 1.14 1.11 2 2 13 SIH LY294002 hydrochloride wortmannin IRAK 1 di-methyl sphingosine genistein FCCP NSC23766 (Rac1 inhibitor)36.78 25.56 0.00 1.14 1.14 1.06 8 2 14 apocynin IRAK 1 Y27632 dihydrochloride (ROCK inhibitor) 69.62 51.05 42.38 0.97 1.00 0.96 3 2 15 alpha-tocopherol U0126 PD98059 SR 11302 85.81 71.50 65.32 0.99 1.02 0.98 4 2 16 wortmannin genistein NSC23766 (Rac1 inhibitor) 25.33 40.96 22.13 1.01 1.15 0.97 3 2 17 alpha-tocopherol SB203580 succinate PDTC SKI-II 13.97 18.38 12.11 1.02 1.00 0.95 4 2 18 di-methyl sphingosine 73.92 70.55 56.58 0.98 1.01 0.99 1 2 19 SR 11302 IKK inhibitor (BMS-345541) di-methyl sphingosine 20.15 37.32 22.90 0.99 1.02 1.00 3 2 20 DMSO 121.02 117.45 83.76 0.98 0.99 0.99 0 2 21 desferrioxamine 125.49 184.83 117.24 1.03 1.03 1.04 1 2 22 tempol LY294002 hydrochloride PDTC SKI-II di-methyl sphingosine FCCP 90.16 31.87 12.11 1.04 1.02 0.99 6 2 23 U0126 PDTC FCCP NSC23766 (Rac1 inhibitor) 55.75 28.26 12.88 1.01 1.04 1.02 4 2 24 desferrioxamine rotenone wortmannin 57.88 45.53 26.78 1.05 1.12 1.05 3 2 25 diphenylene iodonium U0126 (DPI) genistein 2-4-dinitrophenol 92.34 60.29 32.22 1.00 1.04 1.28 4 2 26 di-methyl sphingosine 70.70 72.43 45.53 1.00 1.02 1.02 1 2 27 diphenylene iodoniumIRAK 1 (DPI) SKI-II 0.00 24.66 0.00 1.06 1.08 1.01 3 2 28 tempol U0126 CCCP mevastatin 15.00 64.02 34.56 0.98 1.10 1.01 4 2 29 TPEN rotenone SR 11302 CCCP 115.46 109.65 95.93 1.15 1.14 1.06 4 2 30 alpha-tocopherol U0126 PD98059 SP600125 di-methyl sphingosine mevastatin 66.41 59.37 64.52 1.06 1.01 1.02 6 2 31 IRAK 1 91.26 69.62 60.54 1.03 1.01 1.07 1 2 32 tempol wortmannin IRAK 1 CCCP 42.03 92.32 15.96 1.15 1.24 1.30 4 2 33 alpha-tocopherol N-acetyl succinate cysteine PDTC genistein FCCP 57.88 35.50 17.50 1.04 1.00 1.00 5 2 34 IRAK 1 SR 11302 NSC23766 (Rac1 inhibitor) 67.48 33.68 23.68 0.99 1.06 1.00 3 2 35 rotenone PDTC 2-4-dinitrophenol 82.56 62.16 55.78 1.02 1.06 1.06 3 2 36 U0126 PD98059 63.20 80.92 62.93 0.98 1.10 0.95 2 2 37 desferrioxamine alpha-tocopherol diphenylene succinate iodonium U0126 (DPI) 45.18 54.74 24.45 1.07 1.04 0.96 4 2 38 VK-28 U0126 NSC23766 (Rac1 inhibitor) 50.46 43.70 40.82 1.01 1.00 1.00 3 2 39 alpha-tocopherol tempol succinate apocynin LY294002 hydrochloride 159.49 144.13 138.89 0.99 1.02 1.01 4 2 40 desferrioxamine wortmannin SB203580 genistein FCCP 0.00 16.59 0.00 1.12 1.03 1.18 5 2 41 desferrioxamine NSC23766 (Rac1 Y27632 inhibitor) dihydrochloride (ROCK inhibitor) 166.38 94.23 75.71 1.02 1.01 0.97 3 2 42 N-acetyl cysteine PDTC 142.39 88.50 91.87 1.01 1.10 1.03 2 2 43 SIH tempol 137.86 100.00 77.32 0.99 1.04 0.93 2 2 44 alpha-tocopherol tempol succinate diphenylene iodonium SR 11302 (DPI) genistein 107.70 90.42 58.16 1.04 1.04 1.02 5 2 45 VK-28 SIH alpha-tocopherol U0126 SR 11302 di-methyl sphingosine mevastatin NSC23766 (Rac1 Y27632 inhibitor) dihydrochloride 0.00 (ROCK inhibitor) 47.37 9.81 0.99 1.12 1.05 9 2 46 SB203580 PDTC di-methyl sphingosine FCCP 12.95 64.02 0.00 1.01 1.16 0.91 4 2 47 alpha-tocopherol alpha-tocopherol U0126 succinate PDTC SKI-II CCCP 21.19 28.26 6.74 1.07 0.98 0.92 6 2 48 N-acetyl cysteine LY294002 hydrochloride SR 11302 FCCP 2-4-dinitrophenol 30.52 40.96 55.78 1.00 0.93 1.07 5 2 49 SIH desferrioxamine tempol SR 11302 NSC23766 (Rac1 inhibitor) 68.55 127.25 84.57 0.99 0.94 0.94 5 2 50 tempol CCCP 51.51 63.08 43.17 0.97 1.02 1.00 2 3 1 TPEN U0126 48.64 34.47 19.66 1.06 1.05 1.06 2 3 2 tempol CCCP 27.13 13.59 19.10 1.00 1.05 1.02 2 3 3 apocynin PDTC SKI-II 13.90 8.61 11.92 1.11 1.03 1.03 3 3 4 DMSO 91.37 82.76 52.14 1.02 0.98 0.97 0 3 5 wortmannin 2-4-dinitrophenol 42.34 42.85 49.81 1.04 0.99 1.02 2 3 6 tempol SP600125 SKI-II CCCP 9.55 7.38 10.83 1.03 1.01 1.05 4 3 7 alpha-tocopherol U0126 PD98059 69.08 80.04 48.06 0.98 1.00 0.96 3 3 8 VK-28 PDTC FCCP 7.93 7.38 6.46 0.98 1.01 0.98 3 3 9 VK-28 wortmannin PDTC IKK inhibitor (BMS-345541) SKI-II 5.77 0.00 19.66 1.10 1.08 1.28 5 3 10 tempol apocynin U0126 PDTC FCCP 11.18 11.10 11.92 1.05 1.05 1.09 5 3 11 SIH apocynin IRAK 1 31.60 35.12 41.70 1.02 1.00 1.03 3 3 12 SIH tempol apocynin SB203580 PDTC 15.54 13.59 14.13 1.00 1.01 1.00 5 3 13 SB203580 PDTC 17.18 12.35 11.38 0.98 1.00 0.99 2 3 14 desferrioxamine N-acetyl cysteine SP600125 SR 11302 SKI-II CCCP 2-4-dinitrophenol NSC23766 (Rac1 inhibitor)9.55 8.61 10.28 1.00 1.00 1.01 8 3 15 apocynin U0126 PDTC 2-4-dinitrophenol Y27632 dihydrochloride (ROCK inhibitor) 70.26 78.67 50.97 0.99 0.99 0.97 5 3 16 desferrioxamine tempol SB203580 13.36 12.35 13.58 1.00 0.95 1.02 3 3 17 DMSO 76.23 78.67 108.86 0.97 1.00 1.01 0 3 18 rotenone 34.41 40.91 28.05 1.00 1.03 0.97 1 3 19 wortmannin 2-4-dinitrophenol 22.69 23.64 20.21 1.03 1.02 0.98 2 3 20 SIH SB203580 SKI-II 0.00 0.00 0.00 0.97 0.98 0.93 3 3 21 SB203580 PDTC di-methyl sphingosine genistein FCCP 0.00 0.00 6.46 0.98 1.00 1.04 5 3 22 2-4-dinitrophenol 121.42 185.91 191.90 0.99 1.01 1.06 1 3 23 VK-28 wortmannin NSC23766 (Rac1 inhibitor) 33.29 38.98 41.70 1.02 1.00 1.06 3 3 24 tempol U0126 PDTC 87.71 88.93 91.88 1.03 1.00 1.03 3 3 25 alpha-tocopherol SB203580 succinate U0126 PD98059 SKI-II genistein 5.77 0.00 5.92 1.02 0.99 1.00 6 3 26 di-methyl sphingosine 79.23 65.22 86.95 1.00 0.98 1.02 1 3 27 tempol wortmannin SB203580 genistein 14.99 16.72 13.03 0.98 1.00 0.97 4 3 28 SIH alpha-tocopherol SB203580 succinate SR 11302 PDTC NSC23766 (Rac1 inhibitor) 11.72 9.86 11.38 0.97 0.95 0.99 6 3 29 SIH SB203580 12.81 13.59 14.13 0.98 0.96 1.01 2 3 30 SIH alpha-tocopherol wortmannin succinate 38.37 40.26 35.41 1.00 1.05 0.97 3 3 31 SIH tempol U0126 PDTC SKI-II di-methyl sphingosine 8.47 8.61 7.55 0.97 0.99 0.96 6 3 32 TPEN alpha-tocopherol wortmannin succinate SKI-II 9.01 6.76 8.09 1.39 1.56 1.40 4 3 33 alpha-tocopherol SB203580 PDTC 2-4-dinitrophenol 13.90 13.59 16.33 1.00 0.97 1.02 4 3 34 VK-28 desferrioxamine tempol LY294002 hydrochloride SKI-II Y27632 dihydrochloride (ROCK inhibitor) 18.28 16.09 17.99 1.02 1.02 1.02 6 3 35 SKI-II 12.27 11.10 13.58 1.00 0.98 1.01 1 3 36 alpha-tocopherol apocynin succinate wortmannin PD98059 CCCP mevastatin 0.00 9.86 11.92 1.03 1.05 1.01 6 3 37 wortmannin genistein 40.63 42.85 46.32 1.00 0.97 1.02 2 3 38 desferrioxamine N-acetyl cysteine tempol NSC23766 (Rac1 inhibitor) 62.59 75.29 62.74 0.95 0.99 0.95 4 3 39 SIH tempol SR 11302 88.92 84.81 78.37 0.95 0.97 0.97 3 3 40 IRAK 1 SKI-II NSC23766 (Rac1 inhibitor) 10.63 12.35 11.38 1.04 1.02 1.09 3 3 41 tempol U0126 PDTC SKI-II NSC23766 (Rac1 inhibitor) 6.85 8.00 8.64 1.04 1.05 1.07 5 3 42 SKI-II Y27632 dihydrochloride (ROCK inhibitor) 14.45 13.59 7.55 1.01 1.03 0.97 2 3 43 VK-28 SIH alpha-tocopherol SKI-II succinate 13.36 14.22 15.78 0.97 1.00 1.05 4 3 44 CCCP 18.83 24.28 25.24 0.98 1.03 1.02 1 3 45 alpha-tocopherol PD98059 SKI-II 11.72 11.10 13.58 1.00 0.98 1.00 3 3 46 tempol FCCP 9.55 10.48 11.92 1.00 0.99 1.00 2 3 47 SR 11302 CCCP 24.35 20.49 21.89 1.00 1.02 0.94 2 3 48 U0126 PDTC 82.85 91.68 59.78 0.99 1.01 0.98 2 3 49 U0126 SR 11302 CCCP 21.04 23.64 19.10 1.02 1.02 0.98 3 3 50 rotenone SP600125 66.12 70.57 61.56 1.05 1.02 1.06 2 4 1 TPEN alpha-tocopherol U0126 succinate 45.45 102.61 65.41 0.87 1.03 1.03 3 4 2 PD98059 107.75 148.58 83.08 0.97 1.02 1.01 1 4 3 apocynin SKI-II EGCG 44.99 50.34 42.37 1.01 1.12 0.98 3 4 4 tempol SKI-II 18.64 35.98 45.59 0.92 1.02 1.00 2 4 5 SIH desferrioxamine diphenylene iodonium LY294002 (DPI) hydrochloride PD98059 IRAK 1 SKI-II 11.14 13.23 9.95 1.05 1.11 1.02 7 4 6 VK-28 TPEN alpha-tocopherol EGCG succinate 73.19 126.38 112.61 0.95 1.09 1.01 4 4 7 VK-28 desferrioxamine SB203580 5.25 8.97 7.21 0.99 1.12 0.97 3 4 8 U0126 SKI-II 61.11 37.25 36.97 1.03 0.98 0.95 2 4 9 SIH TPEN apocynin IRAK 1 SKI-II 19.05 49.79 25.07 0.95 1.09 0.99 5 4 10 apocynin PD98059 SR 11302 genistein 47.76 78.04 59.10 0.91 0.99 0.98 4 4 11 SIH desferrioxamine tempol SR 11302 NSC23766 (Rac1 Y27632 inhibitor) dihydrochloride (ROCK inhibitor) 56.18 107.14 57.38 1.04 1.01 0.92 6 4 12 diphenylene iodonium di-methyl (DPI) sphingosine 22.13 55.74 37.29 0.91 1.10 1.09 2 4 13 tempol rotenone SR 11302 EGCG 50.13 42.78 36.89 1.04 1.05 1.06 4 4 14 SR 11302 IKK inhibitor (BMS-345541) di-methyl sphingosine 71.03 65.83 40.93 1.02 1.08 0.96 3 4 15 DMSO 44.72 121.99 93.24 0.96 1.06 1.00 0 4 16 apocynin PDTC SKI-II 13.16 48.37 27.42 0.92 1.10 0.97 3 4 17 wortmannin SB203580 SKI-II genistein 3.32 4.32 3.66 1.04 0.99 1.01 4 4 18 PDTC SKI-II 9.75 54.14 23.89 0.93 1.13 1.00 2 4 19 desferrioxamine N-acetyl cysteine tempol di-methyl sphingosine NSC23766 (Rac1 simvastatin inhibitor) 35.18 67.47 50.42 0.92 1.00 0.97 6 4 20 SIH desferrioxamine PDTC 80.31 78.18 83.73 1.04 0.99 0.92 3 4 21 alpha-tocopherol apocynin succinate SKI-II simvastatin 12.85 41.24 21.50 0.90 1.05 1.01 4 4 22 SIH TPEN N-acetyl cysteine SB203580 PD98059 SR 11302 7.82 7.42 6.93 1.03 1.01 1.01 6 4 23 alpha-tocopherol PD98059 70.33 108.51 73.77 0.99 1.04 0.95 2 4 24 SIH alpha-tocopherol SB203580 PDTC SKI-II 2.02 3.97 3.93 0.88 0.99 0.99 5 4 25 SIH tempol SB203580 SR 11302 SKI-II NSC23766 (Rac1 inhibitor) 4.47 5.46 3.45 1.00 1.04 0.95 6 4 26 apocynin PDTC SKI-II 10.83 40.94 25.01 0.91 1.04 0.97 3 4 27 desferrioxamine rotenone SKI-II genistein 19.38 20.67 18.08 1.08 1.02 1.03 4 4 28 tempol SB203580 PDTC simvastatin 3.81 10.10 6.38 0.90 1.07 0.95 4 4 29 SIH tempol 57.00 111.81 94.79 0.89 0.99 0.99 2 4 30 IRAK 1 NSC23766 (Rac1 inhibitor) 33.16 25.38 34.82 1.04 0.98 0.97 2 4 31 SIH apocynin SB203580 10.53 10.63 9.71 1.06 0.99 0.99 3 4 32 apocynin SB203580 SP600125 PDTC 8.87 11.61 5.55 1.01 1.06 1.03 4 4 33 apocynin PD98059 SKI-II 31.04 36.66 20.90 1.00 1.03 0.99 3 4 34 SIH rotenone 21.57 38.34 43.40 0.92 1.04 1.06 2 4 35 SIH SB203580 SKI-II 3.76 6.11 3.66 1.02 1.06 1.01 3 4 36 N-acetyl cysteine apocynin 53.15 151.72 76.99 0.92 1.03 0.97 2 4 37 SIH SB203580 SKI-II 2.56 5.11 3.45 0.93 1.05 1.01 3 4 38 SIH 113.26 109.03 82.95 1.01 0.97 0.99 1 4 39 TPEN tempol U0126 Y27632 dihydrochloride (ROCK inhibitor) 58.66 177.34 104.52 0.93 1.07 0.94 4 4 40 wortmannin U0126 IRAK 1 45.45 36.66 42.28 1.03 0.96 0.98 3 4 41 SIH SB203580 SKI-II 2.94 3.76 3.18 1.07 0.99 1.02 3 4 42 IKK inhibitor (BMS-345541) 49.75 97.44 47.65 1.00 1.13 1.05 1 4 43 VK-28 alpha-tocopherol SKI-II succinate EGCG 38.58 19.17 31.19 1.00 0.97 1.03 4 4 44 SB203580 PD98059 SKI-II di-methyl sphingosine NSC23766 (Rac1 simvastatin inhibitor) EGCG 2.51 3.62 2.81 1.03 1.09 1.07 7 4 45 diphenylene iodonium (DPI) 20.26 58.66 32.03 1.07 1.15 1.10 1 4 46 SKI-II 37.74 49.57 28.01 0.98 1.08 0.95 1 4 47 desferrioxamine alpha-tocopherol U0126 PD98059 IKK inhibitor (BMS-345541) simvastatin 36.66 30.41 46.58 1.11 1.00 1.03 6 4 48 SIH SB203580 SP600125 IRAK 1 3.98 6.11 4.52 1.08 1.13 0.94 4 4 49 tempol wortmannin SB203580 9.70 5.82 7.83 1.19 0.99 0.98 3 4 50 rotenone 30.34 32.36 28.30 1.10 1.04 1.07 1 5 1 apocynin SB203580 SR 11302 16.47 15.44 13.58 1.01 1.05 1.01 3 5 2 SIH SKI-II di-methyl sphingosine 35.99 35.58 13.67 0.99 1.04 0.98 3 5 3 SIH tempol SKI-II 48.78 35.28 22.38 0.98 1.03 1.02 3 5 4 tempol wortmannin PDTC simvastatin 111.97 28.97 19.28 0.99 1.06 1.01 4 5 5 tempol SP600125 PDTC SKI-II di-methyl sphingosine 38.77 39.81 35.51 0.97 1.05 1.05 5 5 6 PDTC IKK inhibitor (BMS-345541) SKI-II 26.77 28.53 16.17 1.02 1.12 1.03 3 5 7 alpha-tocopherol diphenylene succinate iodonium SB203580 (DPI) PDTC SKI-II Y27632 dihydrochloride (ROCK inhibitor) 5.64 5.61 4.46 0.99 1.02 1.10 6 5 8 TPEN tempol apocynin SB203580 IRAK 1 SR 11302 genistein 16.82 15.04 9.72 1.00 1.02 0.99 7 5 9 SB203580 SKI-II 7.08 6.51 4.63 1.03 1.00 0.98 2 5 10 alpha-tocopherol tempol succinate SB203580 PDTC genistein simvastatin 9.66 8.44 7.37 1.01 0.99 1.04 6 5 11 SKI-II mevastatin 26.89 34.24 17.95 1.01 1.04 1.02 2 5 12 SIH tempol SR 11302 SKI-II Y27632 dihydrochloride (ROCK inhibitor) 36.78 38.44 12.49 0.98 1.03 0.95 5 5 13 apocynin PD98059 SR 11302 genistein 126.67 116.96 100.00 0.95 1.03 1.03 4 5 14 SIH PD98059 SKI-II genistein 30.81 35.58 26.78 1.00 1.04 1.04 4 5 15 VK-28 desferrioxamine SB203580 SR 11302 Y27632 dihydrochloride (ROCK inhibitor) 12.39 12.24 9.11 0.99 1.05 1.01 5 5 16 SKI-II 29.41 33.35 15.89 1.00 1.03 0.98 1 5 17 SKI-II di-methyl sphingosine 22.94 31.59 21.89 0.98 1.06 1.05 2 5 18 LY294002 hydrochloride di-methyl sphingosine 149.98 122.44 74.77 1.00 1.02 0.97 2 5 19 TPEN SKI-II 34.03 42.73 44.41 1.05 1.01 1.08 2 5 20 U0126 SR 11302 EGCG 151.80 129.18 139.91 1.00 1.03 1.01 3 5 21 SIH apocynin SB203580 11.59 14.37 10.70 0.99 1.02 1.00 3 5 22 tempol 105.71 139.67 103.44 0.98 1.03 0.96 1 5 23 tempol SR 11302 SKI-II di-methyl sphingosine EGCG 23.67 27.96 25.87 1.02 1.04 1.07 5 5 24 wortmannin PD98059 di-methyl sphingosine 87.12 90.33 69.38 0.99 1.06 1.00 3 5 25 DMSO 104.04 146.92 94.82 0.96 0.91 0.93 0 5 26 tempol 101.45 122.88 122.55 0.97 1.00 0.98 1 5 27 SKI-II 30.17 24.83 16.07 0.96 0.98 0.94 1 5 28 SB203580 SKI-II 6.41 6.12 4.63 1.00 0.99 0.96 2 5 29 alpha-tocopherol mevastatin succinate 100.36 77.72 118.70 0.99 1.01 0.99 2 5 30 SIH SB203580 PD98059 SKI-II 5.97 6.76 4.63 0.99 1.05 0.95 4 5 31 SIH SB203580 SKI-II Y27632 dihydrochloride (ROCK inhibitor) 6.30 6.76 5.31 0.97 0.99 1.01 4 5 32 DMSO 83.39 132.18 113.64 0.97 1.03 0.99 0 5 33 SIH SKI-II 25.27 26.96 32.50 0.99 0.97 1.09 2 5 34 tempol SB203580 PDTC simvastatin 9.10 10.39 8.50 1.01 1.09 1.03 4 5 35 desferrioxamine alpha-tocopherol 86.95 92.65 107.48 0.95 1.01 1.04 2 5 36 wortmannin SB203580 simvastatin 10.22 12.50 8.93 0.98 0.98 0.96 3 5 37 tempol SKI-II 22.57 34.24 26.68 0.92 1.00 1.02 2 5 38 SIH SB203580 PDTC 9.43 8.70 7.89 0.94 1.01 0.98 3 5 39 SIH SB203580 SP600125 SKI-II 9.66 8.57 6.93 1.00 0.98 1.01 4 5 40 rotenone wortmannin 55.95 33.79 36.93 1.04 1.00 1.11 2 5 41 rotenone PD98059 SKI-II 17.53 18.56 16.45 1.06 1.05 1.13 3 5 42 TPEN wortmannin SB203580 10.00 11.58 13.22 0.97 1.00 1.05 3 5 43 alpha-tocopherol tempol succinate U0126 PDTC 46.68 67.72 80.82 0.99 1.06 0.98 4 5 44 SB203580 SKI-II 5.53 5.99 5.31 1.02 1.09 1.00 2 5 45 apocynin SR 11302 genistein 47.10 65.12 118.13 0.99 0.97 1.08 3 5 46 VK-28 55.08 75.72 96.63 0.98 1.02 1.01 1 5 47 SIH SKI-II 19.67 33.06 21.99 0.94 1.02 1.02 2 5 48 PDTC NSC23766 (Rac1 inhibitor) 46.40 44.44 76.52 0.96 0.97 0.99 2 5 49 SIH SB203580 8.87 7.92 11.95 1.01 0.96 1.10 2 5 50 wortmannin SKI-II genistein Y27632 dihydrochloride (ROCK inhibitor) 26.64 16.39 16.35 0.97 0.96 1.04 4 6 1 wortmannin SB203580 SKI-II 0.00 4.87 2.66 1.06 1.04 1.03 3 6 2 N-acetyl cysteine NSC23766 (Rac1 inhibitor) 119.95 133.27 112.91 1.02 1.06 0.96 2 6 3 IKK inhibitor (BMS-345541) SKI-II 17.10 27.56 11.38 1.08 1.05 0.96 2 6 4 desferrioxamine tempol wortmannin SB203580 PD98059 Y27632 dihydrochloride simvastatin (ROCK inhibitor) 0.00 2.97 2.66 1.27 1.23 1.18 7 6 5 SIH alpha-tocopherol SB203580 PDTC SKI-II 3.10 4.39 0.00 0.98 1.07 0.89 5 6 6 SIH SB203580 SP600125 SKI-II 6.78 6.46 5.02 0.93 1.00 0.86 4 6 7 desferrioxamine tempol LY294002 hydrochloride Y27632 dihydrochloride (ROCK inhibitor) 132.94 159.43 95.96 0.96 1.06 0.93 4 6 8 wortmannin SB203580 PD98059 simvastatin 0.00 0.00 2.39 1.26 1.29 1.26 4 6 9 SB203580 simvastatin 5.30 8.06 6.56 1.00 1.04 0.93 2 6 10 N-acetyl cysteine NSC23766 (Rac1 inhibitor) 74.83 97.98 55.48 1.04 1.02 0.93 2 6 11 SIH alpha-tocopherol SB203580 SP600125 simvastatin 7.52 14.39 7.82 1.04 1.05 0.99 5 6 12 VK-28 131.21 137.57 122.23 1.01 1.01 1.00 1 6 13 wortmannin SKI-II simvastatin 16.64 18.20 7.54 1.02 1.03 0.94 3 6 14 SB203580 7.52 7.89 8.81 0.99 1.00 0.98 1 6 15 desferrioxamine SKI-II 25.01 27.21 14.26 1.03 1.08 0.92 2 6 16 SIH desferrioxamine tempol SB203580 7.08 7.09 5.30 0.99 0.98 0.91 4 6 17 SB203580 di-methyl sphingosine Y27632 dihydrochloride (ROCK inhibitor) 3.98 5.18 5.16 0.97 1.04 0.93 3 6 18 VK-28 tempol 103.08 83.38 95.16 0.93 0.97 1.04 2 6 19 SIH alpha-tocopherol SB203580 PDTC SKI-II 2.96 3.60 2.25 0.99 1.04 1.00 5 6 20 SIH LY294002 hydrochloride SB203580 PDTC 9.62 9.83 7.12 1.03 1.03 0.96 4 6 21 tempol wortmannin 98.69 95.31 118.73 1.01 0.98 1.06 2 6 22 alpha-tocopherol SB203580 succinate genistein 10.22 19.54 10.23 1.01 1.02 0.96 3 6 23 SIH tempol SR 11302 SKI-II 24.37 32.63 12.82 1.00 1.01 0.91 4 6 24 VK-28 SP600125 96.31 102.71 95.76 1.00 0.98 0.98 2 6 25 SIH SB203580 PDTC 7.97 8.22 4.88 1.05 1.02 0.93 3 6 26 rotenone IRAK 1 SKI-II 9.32 12.10 5.58 1.09 1.05 0.97 3 6 27 SIH tempol SR 11302 PDTC SKI-II 23.56 29.99 12.67 0.99 1.04 0.92 5 6 28 SIH SB203580 6.48 8.22 8.39 0.94 0.95 1.09 2 6 29 SB203580 SKI-II simvastatin 0.00 4.55 0.00 1.01 1.03 1.03 3 6 30 SIH tempol wortmannin SKI-II 7.23 9.83 7.12 1.52 1.52 1.26 4 6 31 SKI-II di-methyl sphingosine 9.92 32.98 26.01 1.03 1.05 1.01 2 6 32 tempol SKI-II NSC23766 (Rac1 inhibitor) 15.09 17.04 17.33 1.04 1.03 0.96 3 6 33 apocynin SKI-II 23.72 22.41 17.62 1.00 1.01 0.99 2 6 34 VK-28 SIH tempol diphenylene iodonium SB203580 (DPI) SKI-II 0.00 3.13 2.25 1.07 1.11 0.97 6 6 35 SIH alpha-tocopherol tempol succinate 102.63 83.81 59.98 0.92 0.96 0.94 3 6 36 wortmannin SKI-II 22.77 37.26 20.43 0.94 1.05 0.91 2 6 37 wortmannin SB203580 mevastatin simvastatin 0.00 2.97 4.60 1.18 1.15 1.20 4 6 38 desferrioxamine tempol SB203580 6.93 9.50 5.44 1.00 0.99 0.94 3 6 39 SIH SB203580 SR 11302 SKI-II 2.52 14.23 6.56 1.01 1.05 1.04 4 6 40 tempol SP600125 SR 11302 PDTC NSC23766 (Rac1 inhibitor) 83.29 75.22 43.72 1.01 1.02 0.99 5 6 41 rotenone NSC23766 (Rac1 inhibitor) 29.40 37.80 36.77 1.07 1.05 1.14 2 6 42 SIH desferrioxamine tempol SB203580 SR 11302 PDTC 7.23 6.77 8.39 0.96 0.99 0.99 6 6 43 SIH tempol SB203580 SR 11302 genistein 6.04 8.54 8.39 0.99 0.98 0.94 5 6 44 tempol SB203580 PDTC Y27632 dihydrochloride simvastatin (ROCK inhibitor) 4.72 7.74 7.68 0.99 0.99 0.96 5 6 45 SIH N-acetyl cysteine SB203580 U0126 SKI-II 0.00 3.45 0.00 1.12 1.11 0.97 5 6 46 SKI-II Y27632 dihydrochloride EGCG (ROCK inhibitor) 14.17 26.00 15.42 0.99 0.97 0.93 3 6 47 SB203580 U0126 PDTC simvastatin 2.38 0.00 0.00 1.02 0.96 1.05 4 6 48 LY294002 hydrochloride wortmannin SKI-II 35.21 33.69 18.51 1.03 1.01 0.98 3 6 49 IRAK 1 108.88 131.03 49.03 1.01 1.01 0.95 1 6 50 SKI-II EGCG 33.87 34.93 40.63 1.03 1.02 1.09 2 7 1 tempol SP600125 SKI-II 16.06 15.77 20.31 1.15 1.13 1.01 3 7 2 wortmannin SB203580 simvastatin 4.04 0.00 9.03 1.32 1.41 1.69 3 7 3 SIH SB203580 SP600125 SKI-II NSC23766 (Rac1 inhibitor) 4.91 3.57 8.39 0.99 1.02 1.07 5 7 4 SKI-II 25.64 13.44 15.13 1.03 0.97 1.06 1 7 5 apocynin SB203580 10.75 7.83 8.39 1.01 0.98 0.96 2 7 6 SB203580 SKI-II simvastatin 4.91 3.57 4.58 1.05 1.02 0.94 3 7 7 alpha-tocopherol wortmannin succinate SKI-II di-methyl sphingosine 4.04 4.22 6.48 1.09 1.32 0.99 4 7 8 Y27632 dihydrochloride (ROCK inhibitor) 90.13 58.00 110.88 0.98 0.93 1.04 1 7 9 genistein 76.11 54.82 59.01 1.06 0.89 0.92 1 7 10 SB203580 SKI-II 0.00 0.00 5.21 1.05 1.06 1.00 2 7 11 SB203580 SKI-II 4.04 0.00 8.07 0.99 0.96 1.08 2 7 12 rotenone SB203580 Y27632 dihydrochloride (ROCK inhibitor) 9.58 8.82 14.48 1.05 1.05 1.10 3 7 13 tempol 90.13 73.77 121.18 0.99 0.96 1.07 1 7 14 SIH SB203580 SP600125 SKI-II 7.24 6.85 7.43 1.00 1.00 0.93 4 7 15 alpha-tocopherol wortmannin SB203580 simvastatin 5.49 4.22 4.89 1.59 1.29 0.97 4 7 16 SB203580 Y27632 dihydrochloride (ROCK inhibitor) 9.58 8.82 11.91 0.99 1.00 1.04 2 7 17 SIH apocynin SP600125 SKI-II Y27632 dihydrochloride EGCG (ROCK inhibitor) 82.07 94.74 72.37 0.96 1.00 0.93 6 7 18 SIH SB203580 SP600125 SKI-II NSC23766 (Rac1 inhibitor) 6.65 6.19 11.91 1.01 0.92 1.06 5 7 19 SKI-II NSC23766 (Rac1 inhibitor) 89.79 52.00 49.35 1.00 0.93 0.92 2 7 20 diphenylene iodonium SB203580 (DPI) IKK inhibitor (BMS-345541) SKI-II simvastatin 0.00 0.00 3.94 1.16 0.97 0.97 5 7 21 VK-28 SIH desferrioxamine alpha-tocopherol diphenylene iodonium SB203580 (DPI) SKI-II 0.00 0.00 6.48 0.99 0.98 1.15 7 7 22 apocynin SB203580 SKI-II 3.46 0.00 7.43 0.99 1.02 1.01 3 7 23 apocynin mevastatin simvastatin 63.39 43.97 88.22 0.97 0.93 1.01 3 7 24 SIH wortmannin IKK inhibitor (BMS-345541) SKI-II 0.00 0.00 8.39 1.09 1.52 1.53 4 7 25 SKI-II 16.36 22.79 29.50 0.99 0.99 1.02 1 7 26 IRAK 1 SKI-II di-methyl sphingosine 0.00 4.22 4.89 0.98 0.97 0.95 3 7 27 SIH rotenone SB203580 6.95 5.21 12.23 1.01 0.95 1.09 3 7 28 LY294002 hydrochloride wortmannin SB203580 SKI-II 6.36 0.00 4.89 1.48 1.05 1.03 4 7 29 SKI-II di-methyl sphingosine 25.94 8.49 10.95 1.06 0.90 0.92 2 7 30 tempol rotenone 42.48 28.19 56.58 1.03 0.97 1.16 2 7 31 SKI-II 103.79 125.69 105.62 1.03 1.05 0.96 1 7 32 SIH SB203580 PDTC SKI-II 7.82 7.50 11.59 1.04 1.07 1.08 4 7 33 EGCG 86.42 75.58 134.79 1.02 1.00 1.11 1 7 34 SB203580 di-methyl sphingosine 8.41 8.82 9.35 1.00 0.99 0.93 2 7 35 apocynin 90.47 92.51 86.76 1.03 1.06 0.95 1 7 36 SIH SB203580 SKI-II mevastatin simvastatin 0.00 0.00 4.58 0.97 0.99 0.93 5 7 37 alpha-tocopherol SB203580 SKI-II 10.46 14.11 20.97 0.98 0.90 1.05 3 7 38 tempol mevastatin 121.69 68.35 65.31 0.99 0.90 0.96 2 7 39 SIH SB203580 PDTC IKK inhibitor (BMS-345541) 6.07 0.00 5.53 1.02 0.94 0.96 4 7 40 SIH desferrioxamine alpha-tocopherol tempol SB203580 Y27632 dihydrochloride (ROCK inhibitor) 7.53 5.21 14.81 0.99 0.91 1.09 6 7 41 SIH LY294002 hydrochloride SB203580 IRAK 1 SKI-II 14.29 16.10 17.39 1.06 1.05 0.92 5 7 42 apocynin SKI-II 95.21 98.49 129.70 1.01 1.04 1.07 2 7 43 desferrioxamine 97.61 99.62 137.56 1.04 1.02 1.07 1 7 44 desferrioxamine SKI-II genistein Y27632 dihydrochloride EGCG (ROCK inhibitor) 86.76 112.54 76.29 1.02 1.04 0.96 5 7 45 SIH alpha-tocopherol SB203580 PDTC SKI-II 5.78 0.00 5.53 0.99 1.01 0.91 5 7 46 DMSO 86.76 93.26 156.78 1.04 1.07 1.06 0 7 47 LY294002 hydrochloride SR 11302 SKI-II Y27632 dihydrochloride (ROCK inhibitor) 21.73 25.49 18.69 1.08 1.02 0.94 4 7 48 SB203580 PDTC 10.46 7.50 10.31 1.06 1.01 0.96 2 7 49 TPEN wortmannin SB203580 SKI-II simvastatin 5.78 0.00 5.85 1.54 1.42 1.23 5 7 50 SIH desferrioxamine tempol SB203580 IKK inhibitor (BMS-345541) 8.11 0.00 9.03 1.02 0.95 1.14 5 8 1 SB203580 SP600125 SKI-II mevastatin 17.42 12.03 13.51 1.23 1.04 1.00 4 8 2 N-acetyl cysteine diphenylene iodonium Y27632 dihydrochloride(DPI) (ROCK inhibitor) 47.47 51.12 52.55 1.09 1.11 1.01 3 8 3 SIH LY294002 hydrochloride SB203580 PDTC SKI-II genistein 17.42 17.11 14.47 1.13 1.08 1.02 6 8 4 tempol SB203580 SKI-II 14.48 11.53 13.51 1.06 1.04 1.04 3 8 5 tempol PD98059 simvastatin 96.73 88.58 80.62 1.09 1.10 1.05 3 8 6 SIH SB203580 SP600125 IKK inhibitor (BMS-345541) SKI-II di-methyl sphingosine mevastatin 10.97 9.51 10.15 0.99 1.01 1.04 7 8 7 SB203580 PDTC genistein 22.15 20.18 21.23 1.08 1.07 1.07 3 8 8 SR 11302 genistein 60.43 83.51 102.17 0.99 0.99 1.05 2 8 9 VK-28 SIH alpha-tocopherol wortmannin succinate SB203580 14.48 10.52 11.11 1.17 1.17 1.12 5 8 10 SIH SB203580 mevastatin simvastatin 16.25 17.11 15.91 0.97 1.04 0.99 4 8 11 SIH desferrioxamine SB203580 IKK inhibitor (BMS-345541) 12.14 0.00 10.15 1.06 0.97 0.96 4 8 12 IRAK 1 SKI-II 19.19 13.05 15.91 1.37 1.03 0.98 2 8 13 SIH SB203580 SKI-II 12.72 12.03 12.55 1.01 1.07 0.97 3 8 14 SP600125 di-methyl sphingosine 127.24 118.64 109.33 1.13 1.03 1.01 2 8 15 SB203580 SP600125 di-methyl sphingosine 22.15 18.65 19.29 1.05 1.02 1.04 3 8 16 apocynin SB203580 PDTC NSC23766 (Rac1 inhibitor) 22.15 19.67 16.87 1.10 1.07 1.04 4 8 17 SIH PD98059 75.51 75.68 78.50 1.03 1.01 1.03 2 8 18 SB203580 genistein 19.78 21.21 20.74 1.03 1.04 1.01 2 8 19 diphenylene iodonium wortmannin (DPI) genistein 13.90 11.53 15.43 1.12 1.11 1.12 3 8 20 diphenylene iodonium SB203580 (DPI) 15.07 14.57 12.55 1.12 1.06 1.01 2 8 21 N-acetyl cysteine tempol SB203580 simvastatin 19.19 20.69 16.87 0.99 1.00 1.01 4 8 22 SB203580 SKI-II genistein NSC23766 (Rac1 inhibitor) 10.97 0.00 11.59 1.02 0.97 0.99 4 8 23 SB203580 IRAK 1 simvastatin 0.00 0.00 0.00 1.00 0.98 1.06 3 8 24 SIH tempol 147.76 105.76 116.01 1.19 1.03 0.97 2 8 25 IKK inhibitor (BMS-345541) 52.38 45.77 41.94 1.10 1.03 1.02 1 8 26 SB203580 genistein 23.34 21.21 21.23 1.05 1.02 1.00 2 8 27 VK-28 SIH TPEN SB203580 SP600125 SKI-II mevastatin 16.25 17.62 14.47 1.01 0.95 0.98 7 8 28 apocynin LY294002 hydrochloride SB203580 28.11 24.81 23.66 1.07 1.00 0.99 3 8 29 VK-28 SB203580 SP600125 SR 11302 SKI-II 14.48 12.03 14.47 0.96 0.98 0.98 5 8 30 SB203580 di-methyl sphingosine 24.53 14.57 15.43 0.99 0.95 0.92 2 8 31 SIH SB203580 SKI-II 0.00 0.00 10.15 0.98 0.99 0.98 3 8 32 apocynin SB203580 32.30 25.33 25.13 1.13 1.03 0.98 2 8 33 SIH apocynin SB203580 20.97 25.33 19.77 0.98 1.05 0.99 3 8 34 SB203580 23.34 21.72 23.66 1.07 1.02 1.02 1 8 35 tempol SB203580 IRAK 1 13.90 11.02 15.91 1.05 1.00 1.02 3 8 36 alpha-tocopherol SB203580 simvastatin 20.38 17.11 14.95 1.11 1.04 0.99 3 8 37 SIH SB203580 IRAK 1 SKI-II 12.14 9.00 11.59 1.28 1.06 1.03 4 8 38 SB203580 SKI-II 13.31 11.02 11.59 0.96 1.01 0.98 2 8 39 genistein simvastatin 89.60 55.42 57.15 1.01 0.93 0.91 2 8 40 VK-28 N-acetyl cysteine rotenone SKI-II Y27632 dihydrochloride EGCG (ROCK inhibitor) 16.25 15.59 21.23 0.98 1.02 0.99 6 8 41 SB203580 SKI-II 14.48 11.53 11.11 1.11 1.04 1.02 2 8 42 rotenone apocynin SB203580 Y27632 dihydrochloride (ROCK inhibitor) 13.31 15.08 14.47 1.12 1.14 1.06 4 8 43 SB203580 SKI-II 13.31 14.06 13.03 1.16 1.05 1.05 2 8 44 TPEN genistein 103.27 109.83 135.82 1.12 1.03 1.04 2 8 45 apocynin SB203580 PD98059 21.56 18.14 16.39 1.13 1.10 0.99 3 8 46 N-acetyl cysteine SB203580 NSC23766 (Rac1 inhibitor) 26.32 18.14 20.26 1.14 0.98 0.99 3 8 47 SB203580 SKI-II simvastatin 12.14 10.52 11.11 1.09 1.06 1.00 3 8 48 SIH desferrioxamine tempol SB203580 SKI-II 12.72 9.51 11.11 1.14 0.98 0.96 5 8 49 SIH SB203580 PDTC IKK inhibitor (BMS-345541) 11.55 9.51 0.00 1.14 1.06 1.04 4 8 50 tempol SKI-II 31.10 28.95 30.03 1.12 1.02 1.04 2 9 1 diphenylene iodonium SB203580 (DPI) IKK inhibitor (BMS-345541) SKI-II simvastatin 0.00 0.00 0.00 1.24 1.25 1.24 5 9 2 VK-28 alpha-tocopherol tempol PD98059 IKK inhibitor (BMS-345541) mevastatin simvastatin 19.22 19.77 8.23 1.02 1.13 1.06 7 9 3 IKK inhibitor (BMS-345541) NSC23766 (Rac1 inhibitor) 27.95 16.48 16.85 0.98 0.96 0.95 2 9 4 tempol 141.29 179.90 94.85 1.03 1.03 0.98 1 9 5 SB203580 PDTC di-methyl sphingosine 0.00 0.00 0.00 1.03 0.99 1.05 3 9 6 SB203580 SP600125 di-methyl sphingosine Y27632 dihydrochloride (ROCK inhibitor) 0.00 0.00 0.00 1.01 1.02 0.97 4 9 7 alpha-tocopherol tempol SB203580 SP600125 NSC23766 (Rac1 simvastatin inhibitor) 0.00 0.00 0.00 0.98 0.98 1.02 6 9 8 PDTC 78.24 87.64 88.52 0.94 0.95 0.99 1 9 9 VK-28 SIH tempol 97.08 141.14 115.94 1.11 1.10 1.13 3 9 10 DMSO 97.08 138.29 73.50 0.98 1.05 0.98 0 9 11 SKI-II NSC23766 (Rac1 inhibitor) 6.33 11.45 14.36 1.01 1.08 1.07 2 9 12 SIH desferrioxamine SB203580 0.00 0.00 0.00 1.03 1.03 1.06 3 9 13 IRAK 1 Y27632 dihydrochloride (ROCK inhibitor) 14.20 15.58 8.47 0.98 0.96 0.92 2 9 14 SKI-II NSC23766 (Rac1 inhibitor) 10.13 0.00 11.40 1.02 1.00 1.04 2 9 15 IKK inhibitor (BMS-345541) SKI-II NSC23766 (Rac1 inhibitor) 0.00 0.00 0.00 1.10 1.05 0.98 3 9 16 DMSO 108.61 121.23 112.90 0.97 0.96 1.00 0 9 17 SB203580 genistein 0.00 0.00 0.00 0.97 0.94 0.95 2 9 18 SIH tempol SB203580 IKK inhibitor (BMS-345541) 0.00 0.00 0.00 1.04 1.11 0.94 4 9 19 tempol SB203580 PD98059 SR 11302 0.00 0.00 0.00 1.09 1.07 1.10 4 9 20 SB203580 0.00 0.00 0.00 1.00 0.99 0.95 1 9 21 SB203580 Y27632 dihydrochloride (ROCK inhibitor) 0.00 0.00 0.00 1.01 1.04 1.00 2 9 22 apocynin SB203580 SP600125 SKI-II mevastatin 0.00 0.00 0.00 1.00 1.05 1.01 5 9 23 SB203580 0.00 0.00 0.00 0.98 1.08 0.99 1 9 24 TPEN tempol U0126 34.40 40.12 27.02 0.98 0.98 0.91 3 9 25 SP600125 SKI-II genistein 57.99 59.30 35.40 1.03 1.02 0.88 3 9 26 SIH alpha-tocopherol SB203580 succinate EGCG 0.00 0.00 0.00 1.02 0.99 1.00 4 9 27 genistein 85.13 88.71 62.02 1.00 0.97 0.90 1 9 28 U0126 genistein 44.85 34.15 33.02 0.99 0.94 0.94 2 9 29 N-acetyl cysteine wortmannin SB203580 SKI-II di-methyl sphingosine 0.00 0.00 0.00 1.43 1.42 1.23 5 9 30 SIH LY294002 hydrochloride PD98059 IRAK 1 simvastatin 17.46 36.02 34.08 1.12 1.10 1.14 5 9 31 SB203580 0.00 0.00 0.00 0.94 1.00 0.92 1 9 32 desferrioxamine alpha-tocopherol wortmannin succinate SKI-II Y27632 dihydrochloride (ROCK inhibitor) 0.00 0.00 0.00 1.31 1.30 1.37 5 9 33 SB203580 di-methyl sphingosine Y27632 dihydrochloride (ROCK inhibitor) 0.00 0.00 0.00 0.99 1.02 0.98 3 9 34 apocynin 86.53 123.56 131.95 0.99 1.08 0.97 1 9 35 SB203580 SR 11302 0.00 0.00 0.00 0.98 0.98 0.90 2 9 36 tempol SB203580 SKI-II simvastatin 0.00 0.00 0.00 1.05 1.05 0.90 4 9 37 wortmannin SKI-II Y27632 dihydrochloride (ROCK inhibitor) 0.00 0.00 0.00 1.41 1.18 1.38 3 9 38 SIH SB203580 0.00 0.00 0.00 1.00 1.00 0.92 2 9 39 TPEN alpha-tocopherol SB203580 succinate Y27632 dihydrochloride (ROCK inhibitor) 0.00 0.00 0.00 0.96 0.94 0.95 4 9 40 SKI-II NSC23766 (Rac1 EGCG inhibitor) 16.37 21.27 9.93 0.98 1.05 0.94 3 9 41 IKK inhibitor (BMS-345541) EGCG 39.81 70.12 70.81 1.12 1.13 1.12 2 9 42 diphenylene iodonium SB203580 (DPI) IKK inhibitor (BMS-345541) SKI-II simvastatin 0.00 0.00 0.00 1.11 1.19 0.95 5 9 43 SIH SB203580 U0126 SP600125 0.00 0.00 0.00 1.06 1.06 1.05 4 9 44 apocynin SB203580 IRAK 1 0.00 0.00 0.00 1.12 1.12 1.05 3 9 45 SB203580 di-methyl sphingosine 0.00 0.00 0.00 1.02 0.99 0.92 2 9 46 genistein Y27632 dihydrochloride (ROCK inhibitor) 87.65 118.90 76.51 1.00 1.03 0.92 2 9 47 SIH SB203580 IRAK 1 di-methyl sphingosine genistein 0.00 0.00 0.00 1.08 1.05 0.92 5 9 48 alpha-tocopherol rotenone succinate SB203580 IRAK 1 SKI-II di-methyl sphingosine simvastatin 0.00 0.00 0.00 1.13 1.11 1.12 7 9 49 N-acetyl cysteine tempol PDTC genistein simvastatin 72.05 67.39 59.14 0.98 0.97 0.98 5 9 50 rotenone NSC23766 (Rac1 inhibitor) 25.46 46.84 19.11 1.03 1.09 1.02 2 10 1 SIH SB203580 9.23 11.38 11.83 0.98 1.02 1.03 2 10 2 SB203580 SKI-II di-methyl sphingosine 2.97 5.47 4.32 1.00 1.06 1.02 3 10 3 SB203580 U0126 3.42 4.47 5.01 1.00 0.99 1.04 2 10 4 SB203580 PDTC di-methyl sphingosine NSC23766 (Rac1 inhibitor) 7.84 8.99 7.82 1.03 1.06 1.00 4 10 5 SB203580 11.57 12.58 9.67 1.01 1.05 1.01 1 10 6 TPEN diphenylene iodonium Y27632 dihydrochloride(DPI) (ROCK inhibitor) 23.43 39.88 35.80 1.04 1.09 1.06 3 10 7 SIH SB203580 Y27632 dihydrochloride (ROCK inhibitor) 9.69 10.86 9.67 1.00 1.05 0.99 3 10 8 SIH SB203580 NSC23766 (Rac1 inhibitor) 9.00 8.49 6.98 0.99 1.03 0.96 3 10 9 SB203580 SR 11302 9.92 10.18 7.26 0.95 0.99 0.93 2 10 10 TPEN alpha-tocopherol SB203580 succinate PD98059 Y27632 dihydrochloride (ROCK inhibitor) 6.01 6.97 7.97 0.96 0.98 1.02 5 10 11 SB203580 9.23 14.48 9.67 0.99 1.03 0.96 1 10 12 SB203580 U0126 IRAK 1 genistein Y27632 dihydrochloride EGCG (ROCK Melatonin inhibitor) 0.00 0.00 3.34 1.02 1.02 1.05 7 10 13 PD98059 Y27632 dihydrochloride EGCG (ROCK Melatonin inhibitor) 109.94 142.32 135.89 0.98 1.07 1.02 4 10 14 TPEN alpha-tocopherol SB203580 succinate SP600125 Y27632 dihydrochloride (ROCK inhibitor) 10.98 8.15 8.96 0.95 1.00 0.99 5 10 15 alpha-tocopherol apocynin IKK inhibitor (BMS-345541) 22.79 41.04 37.41 0.97 1.06 1.01 3 10 16 rotenone SB203580 di-methyl sphingosine 3.76 5.80 3.90 1.07 1.11 1.04 3 10 17 genistein 82.38 143.45 124.22 0.96 1.04 1.01 1 10 18 SR 11302 Melatonin 88.68 70.37 55.36 1.00 1.00 0.93 2 10 19 alpha-tocopherol SB203580 succinate 8.65 12.58 9.25 0.98 1.03 0.97 2 10 20 SB203580 di-methyl sphingosine Y27632 dihydrochloride (ROCK inhibitor) 8.65 8.32 7.26 0.97 0.97 0.93 3 10 21 LY294002 hydrochloride SB203580 di-methyl sphingosine 12.87 15.52 10.82 0.96 0.98 0.93 3 10 22 TPEN alpha-tocopherol Y27632 succinate dihydrochloride (ROCK inhibitor) 55.66 53.75 74.05 0.97 0.99 1.00 3 10 23 TPEN SR 11302 genistein 125.38 169.99 75.97 0.99 1.06 0.95 3 10 24 TPEN diphenylene iodonium SB203580 (DPI) genistein 3.98 5.63 6.70 1.03 1.10 1.05 4 10 25 SIH TPEN SB203580 10.39 12.92 12.41 0.99 1.05 1.02 3 10 26 TPEN alpha-tocopherol SB203580 succinate Y27632 dihydrochloride (ROCK inhibitor) 7.50 10.86 7.12 1.01 1.03 0.98 4 10 27 alpha-tocopherol U0126 succinate 47.71 86.05 76.93 0.97 1.02 1.01 2 10 28 SB203580 8.88 8.83 6.70 1.02 1.01 0.95 1 10 29 SB203580 genistein 10.16 15.69 11.68 1.02 1.03 1.01 2 10 30 SIH SB203580 IKK inhibitor (BMS-345541) 3.20 4.30 3.21 0.96 0.99 0.92 3 10 31 TPEN SB203580 di-methyl sphingosine Y27632 dihydrochloride EGCG (ROCK inhibitor) 7.27 8.83 6.41 0.96 0.96 0.94 5 10 32 di-methyl sphingosine 136.40 170.29 86.99 1.01 1.04 0.97 1 10 33 EGCG 105.90 105.38 123.99 1.00 1.02 1.01 1 10 34 TPEN SB203580 SR 11302 PDTC Y27632 dihydrochloride (ROCK inhibitor) 8.65 13.44 11.83 0.97 1.07 1.01 5 10 35 SB203580 di-methyl sphingosine genistein 10.51 13.96 9.67 1.00 1.04 0.98 3 10 36 SB203580 SP600125 genistein Y27632 dihydrochloride simvastatin (ROCK inhibitor) 8.65 13.44 12.55 0.95 1.06 1.00 5 10 37 genistein 118.25 156.73 81.61 1.01 1.07 0.97 1 10 38 SB203580 U0126 di-methyl sphingosine simvastatin 2.53 4.14 3.76 1.00 1.04 1.03 4 10 39 VK-28 SB203580 Melatonin 8.30 11.03 6.27 0.98 1.02 0.93 3 10 40 SIH apocynin wortmannin SKI-II 4.43 7.31 5.01 1.09 1.14 1.08 4 10 41 SIH SB203580 7.15 10.69 6.70 0.95 0.96 0.92 2 10 42 SIH SB203580 9.23 8.83 10.53 1.02 1.02 1.07 2 10 43 SB203580 SR 11302 10.27 15.00 8.39 0.98 1.06 0.97 2 10 44 DMSO 109.94 154.67 117.11 1.00 1.07 1.03 0 10 45 alpha-tocopherol wortmannin succinate SB203580 Y27632 dihydrochloride (ROCK inhibitor) 4.77 5.13 3.48 1.01 1.06 1.00 4 10 46 wortmannin SB203580 di-methyl sphingosine NSC23766 (Rac1 inhibitor) 4.09 5.63 3.48 1.09 1.14 1.07 4 10 47 desferrioxamine alpha-tocopherol SB203580 succinate Y27632 dihydrochloride (ROCK inhibitor) 7.96 9.16 6.84 0.98 1.03 0.95 4 10 48 SIH genistein 73.88 131.82 105.36 1.03 1.07 1.04 2 10 49 SB203580 PDTC di-methyl sphingosine 7.61 12.41 6.55 1.01 1.03 0.98 3 10 50 TPEN alpha-tocopherol SB203580 succinate Y27632 dihydrochloride (ROCK inhibitor) 5.67 9.16 5.29 1.08 1.03 0.95 4 11 1 rotenone 10.89 36.18 36.16 0.94 0.99 1.04 1 11 2 genistein 116.62 123.79 81.31 0.99 0.99 1.00 1 11 3 VK-28 92.99 58.81 88.46 1.02 0.94 0.99 1 11 4 SB203580 di-methyl sphingosine Y27632 dihydrochloride (ROCK inhibitor) 0.00 8.69 7.01 0.90 1.05 0.99 3 11 5 N-acetyl cysteine 37.72 68.07 103.46 0.92 0.96 1.03 1 11 6 VK-28 SIH SB203580 IKK inhibitor (BMS-345541) 0.00 0.00 4.43 0.94 0.95 1.01 4 11 7 IKK inhibitor (BMS-345541) 33.87 17.73 19.32 1.00 0.97 0.94 1 11 8 IRAK 1 SKI-II 0.00 0.00 5.90 0.93 0.95 1.00 2 11 9 SKI-II 10.89 21.21 11.08 0.99 1.01 0.93 1 11 10 LY294002 hydrochloride SB203580 5.98 10.40 7.75 0.93 1.02 0.95 2 11 11 SIH TPEN SB203580 SP600125 IKK inhibitor (BMS-345541) 0.00 5.28 4.43 1.01 1.02 0.96 5 11 12 VK-28 TPEN SB203580 SR 11302 genistein 5.98 14.70 14.81 0.89 0.93 1.04 5 11 13 SP600125 95.80 113.29 87.62 0.98 0.98 0.99 1 11 14 alpha-tocopherol PDTC succinate mevastatin 64.34 50.58 59.51 1.00 0.94 0.99 3 11 15 SB203580 5.98 10.40 10.71 0.91 0.95 1.01 1 11 16 N-acetyl cysteine SB203580 U0126 0.00 0.00 0.00 0.95 1.00 1.02 3 11 17 SIH SB203580 IKK inhibitor (BMS-345541) NSC23766 (Rac1 inhibitor) 0.00 0.00 0.00 0.97 0.95 0.97 4 11 18 SIH SB203580 genistein 0.00 6.13 9.97 0.93 0.94 0.98 3 11 19 Y27632 dihydrochloride (ROCK inhibitor) 57.25 68.07 78.80 0.95 0.99 0.95 1 11 20 N-acetyl cysteine SB203580 5.98 13.41 13.32 0.90 0.93 1.02 2 11 21 VK-28 IRAK 1 Y27632 dihydrochloride (ROCK inhibitor) 33.45 35.73 22.35 0.96 0.99 0.99 3 11 22 SB203580 PDTC mevastatin 7.61 8.26 8.49 1.00 0.93 0.99 3 11 23 diphenylene iodonium (DPI) 12.53 43.79 29.98 0.96 1.05 0.97 1 11 24 TPEN 61.23 86.48 117.07 0.95 0.97 1.00 1 11 25 EGCG 32.17 59.26 98.71 0.89 0.94 1.01 1 11 26 alpha-tocopherol 94.86 58.81 77.55 0.96 0.94 0.92 1 11 27 SIH 33.02 50.58 77.13 0.92 0.95 0.99 1 11 28 alpha-tocopherol succinate 60.79 76.04 64.38 0.96 0.99 1.01 1 11 29 SB203580 PD98059 SR 11302 di-methyl sphingosine 0.00 0.00 0.00 0.92 0.99 0.95 4 11 30 wortmannin 20.82 27.77 21.59 1.06 1.05 1.08 1 11 31 tempol 57.69 95.15 112.65 0.91 0.94 1.06 1 11 32 IKK inhibitor (BMS-345541) 48.49 50.58 37.32 0.96 0.98 0.99 1 11 33 desferrioxamine 111.36 77.92 98.71 1.03 0.96 0.98 1 11 34 wortmannin SB203580 di-methyl sphingosine 0.00 4.85 5.90 0.95 1.07 1.01 3 11 35 di-methyl sphingosine 35.15 42.44 77.13 0.91 0.97 1.04 1 11 36 SB203580 8.43 9.97 11.45 0.99 1.00 0.96 1 11 37 PDTC 30.48 62.03 99.57 0.89 0.96 0.98 1 11 38 genistein 92.99 61.57 103.03 0.96 0.97 0.95 1 11 39 tempol 25.00 56.97 103.90 0.89 0.96 0.96 1 11 40 SR 11302 67.46 71.80 76.71 0.98 0.99 0.95 1 11 41 apocynin 35.57 81.72 70.93 0.91 1.02 0.95 1 11 42 mevastatin 28.36 74.15 76.30 0.92 0.99 1.05 1 11 43 U0126 15.83 45.14 41.22 0.93 1.03 0.99 1 11 44 LY294002 hydrochloride 34.30 100.98 129.60 0.94 1.01 1.07 1 11 45 simvastatin 85.61 92.72 69.70 1.03 1.06 0.99 1 11 46 NSC23766 (Rac1 inhibitor) 29.63 61.57 70.52 0.96 0.99 0.98 1 11 47 PD98059 66.57 51.94 51.90 1.03 0.96 0.97 1 11 48 SB203580 0.00 9.12 14.07 0.93 0.98 1.00 1 11 49 IRAK 1 18.74 28.21 28.83 1.02 1.02 1.03 1 11 50 Melatonin 33.87 101.47 82.98 0.95 1.02 0.96 1 1 DMSO 116.89 101.55 115.70 0.97 0.97 0.98 1 DMSO 126.20 134.75 127.23 1.03 1.05 1.02 1 DMSO 109.11 116.40 138.89 1.00 0.98 1.00 1 IKK inhibitor (BMS-345541) 0.00 21.00 17.00 1.08 1.06 1.11 2 DMSO 64.27 81.86 53.41 0.94 1.01 1.02 2 DMSO 129.97 86.61 92.67 1.02 1.00 1.01 2 DMSO 116.57 131.21 125.52 0.98 1.06 0.96 2 DMSO 87.99 106.75 112.29 1.01 0.96 0.97 2 DMSO 105.51 97.12 119.73 1.05 0.96 1.03 2 IKK inhibitor (BMS-345541) 57.97 28.74 23.74 1.19NA NA 3 DMSO 128.48 150.32 162.43 1.06 1.05 1.07 3 DMSO 120.13 97.90 122.99 1.03 0.97 1.04 3 DMSO 78.63 84.12 56.83 1.00 1.01 0.95 3 DMSO 80.44 71.25 96.87 0.96 0.96 0.97 3 DMSO 93.21 97.90 68.11 0.96 1.01 0.96 3 IKK inhibitor (BMS-345541) 0.00 0.00 0.00 1.65 1.45 1.91 4 DMSO 137.18 130.31 113.27 1.06 1.03 1.00 4 DMSO 132.36 118.28 115.09 1.05 0.99 0.98 4 DMSO 104.05 94.44 81.93 0.97 1.00 1.03 4 DMSO 71.27 89.11 82.83 0.98 0.98 0.98 4 DMSO 72.11 74.91 111.63 0.94 1.01 1.01 4 IKK inhibitor (BMS-345541) 1.00 2.00 2.00 1.24 1.33 1.22 5 DMSO 159.47 121.78 92.20 1.02 1.01 1.02 5 DMSO 124.41 171.31 95.81 0.99 1.03 0.93 5 DMSO 117.22 74.10 121.97 1.01 0.97 1.03 5 DMSO 64.48 75.17 84.32 0.97 1.06 0.97 5 DMSO 53.20 73.20 108.73 1.01 0.93 1.05 5 IKK inhibitor (BMS-345541) 5.00 5.00 4.00 1.45 1.32 1.29 6 DMSO 92.68 111.47 107.84 0.99 1.01 1.01 6 DMSO 105.51 110.53 107.64 1.00 0.99 1.01 6 DMSO 104.62 99.55 75.87 1.02 1.01 0.98 6 DMSO 120.88 111.94 106.38 1.01 0.99 0.99 6 DMSO 77.62 68.88 103.69 0.98 1.00 1.01 6 IKK inhibitor (BMS-345541) 0.00 0.00 0.00 1.51 1.55 1.56 7 DMSO 86.08 84.73 88.22 1.01 0.99 1.07 7 DMSO 107.95 121.02 92.97 1.03 1.04 0.92 7 DMSO 89.45 106.42 55.54 1.00 1.00 0.92 7 DMSO 92.15 89.91 116.97 0.97 1.00 1.03 7 DMSO 124.53 98.87 150.71 0.98 0.96 1.07 7 IKK inhibitor (BMS-345541) 0.00 0.00 0.00 1.43 1.07 1.09 8 DMSO 111.18 145.72 109.88 0.94 1.03 1.00 8 DMSO 99.34 91.98 106.02 1.02 0.98 1.02 8 DMSO 79.33 86.33 94.01 0.99 0.99 1.00 8 DMSO 89.60 93.13 94.55 1.01 0.99 0.97 8 DMSO 122.52 84.62 95.09 1.05 1.01 1.01 8 IKK inhibitor (BMS-345541) 12.00 11.00 11.00 1.12 1.05 1.00 9 DMSO 107.41 102.60 94.22 1.00 1.06 1.03 9 DMSO 80.70 87.28 90.09 0.99 1.07 1.01 9 DMSO 102.34 107.45 118.34 0.97 0.95 0.97 9 DMSO 100.29 86.92 83.23 0.98 0.96 1.00 9 DMSO 110.42 116.20 115.94 1.06 0.96 0.99 9 IKK inhibitor (BMS-345541) 0.00 0.00 0.00 1.20 1.19 1.38 10 DMSO 145.60 98.06 87.99 1.00 1.00 0.98 10 DMSO 102.73 92.35 132.02 1.09 0.97 1.00 10 DMSO 87.05 101.94 60.34 0.96 0.97 0.92 10 DMSO 91.45 89.76 120.52 0.97 1.05 1.07 10 DMSO 79.04 119.60 105.57 0.99 1.02 1.03 10 IKK inhibitor (BMS-345541) 0.00 0.00 0.00 1.48 1.33 1.21 11 DMSO 86.06 82.66 104.33 1.02 0.99 0.99 11 DMSO 114.22 137.53 121.07 1.02 1.01 0.99 11 DMSO 76.48 82.19 87.20 0.96 0.98 0.97 11 DMSO 106.61 127.34 94.85 1.00 1.04 1.03 11 DMSO 119.04 74.62 91.86 1.00 0.99 1.02 11 IKK inhibitor (BMS-345541) 18.74 75.09 54.69 0.94 1.06 0.96 Construction of an LPS directed signaling network map

3.1 Introduction

In the previous chapter (Publication number 01) there was no assumption made concerning the hierarchical or topological structure of protein nodes within the signaling network of LPS-activated macrophages. Despite the limited chemical diversity of the library explored here, an initial unbiased random selection of reagents was evolved iteratively (in a biased fashion) to reflect a satisfactory outcome against the objectives to optimise against. Subsequent optimization of component reagents in paired combinations with respect to concentration yielded combinations that are theoretically, primarily inhibiting their intended pharmacological targets (i.e. SB203580 = p38 MAPK and BMS345541 = IKK).

Barabasi and Oltvai (Barabasi and Oltvai 2004) provide a thorough overview of network architecture analysis and how the application of mathematical graph theory underpins this study. Briefly, biological networks can be abstracted to nodes (e.g. transcription factors, metabolites) and edges (the links between the nodes, often transformative and encoded by enzymes). These networks of nodes (hubs) and edges (links) can have directionality (e.g. where a signal is transduced via a number of cytoplasmic intermediates from the membrane to the nucleus). The number of edges that feed in or out of a node define the degree of the node, often represented by the nomenclature k in the literature, correspondingly represents the average degree of all nodes within the network. The architecture of biological networks is characterized by their 'scale free' nature, conversely, random networks have nodes with an that equates to a Poisson distribution - all nodes have approximately the same number of links. Scale free networks exhibit a power law distribution, where, the probability of any node displaying exactly k links ( P(k)) is given by the relationship (14).

P(k) ≈ k-γ (14)

Where γ is the degree exponent and often takes a value between 2 and 3. Nodes within scale free networks display a non-uniform distribution of links, with few nodes being highly connected and the majority displaying low numbers of links. Another property of scale free networks is that they are ultra-small (i.e. they exhibit small world properties), with any two nodes being connected to each other by only a few edges. The significance of this in the context of a signal transduction

39 network could be interpreted in the sense that any marked perturbation to a node is quickly propagated to the remaining network. The evolution or emergence of biological scale free networks is thought attributable to two phenomena (Barabasi and Oltvai 2004):

1. Disassortativity - where highly connected nodes avoid connecting to one another. 2. Preferential attachment - where nodes will attach to nodes that have a greater number of links than themselves.

Modularity of networks can be defined in terms of clustering coefficients, which is an index of the networks potential modularity. Recurrent modules often referred to as motifs are key signatures of biological networks. However, the genesis of such network maps is dependent on data from the literature and associated datasets.

3.2 Network map construction

Mining published data concerning the LPS directed signaling and molecular biology of IL- 1β expression from the last ~ 30 years enables a model or map to be created. Delineating publications where dynamic roles for transcription factors in IL-1β expression have been shown limits the mapped space of the network to those nodes and edges that have a demonstrable role in IL-1β expression. This is almost certainly an underestimate of the interactions. Network maps, provide a reference and address topological and hierarchical issues of node positioning. Furthermore qualitative network analysis is a necessary precursor to the building of quantitative models of biological phenomena capable of simulation as their study can lead to the detection of emergent behaviour.

40 Publication number : 02

Publication title : A network map of lipopolysaccharide-dependent IL-1β expression in macrophages and monocytes.

Page number : 41

Published: In preparation

Journal TBD

41 A network map of lipopolysaccharide-dependent IL-1βββ expression in macrophages and monocytes

Ben G Small 1,7 , Holly Summersgill 1,7 , David Brough 3, Nancy J Rothwell 3, Douglas B Kell 2,6 , Pedro Mendes 2,4,5 , Jürgen Pahle 4

1Doctoral Training Centre, Integrative Systems Biology Molecules to Life and 2Manchester Centre for Integrative Systems Biology, Manchester Interdisciplinary Biocentre, University of Manchester, 131 Princess Street, Manchester, M1 7DN, United Kingdom. 3NeuroSystems, Faculty of Life Sciences, AV Hill Building, University of Manchester, Oxford Road, Manchester, M13 9PT, United Kingdom. 4School of Computer Science, University of Manchester, Oxford Road, Manchester, M13 9PL, United Kingdom. 5Virginia Bioinformatics Institute, Virginia Polytechnic Institute and State University, Washington Street, MC0477, Blacksburg, Virginia, United States, 24061-0477. 6School of Chemistry, University of Manchester, 131 Princess St, Manchester M1 7DN, United Kingdom. 7School of Chemical Engineering and Analytical Science, University of Manchester, 131 Princess St, Manchester M1 7DN, United Kingdom.

Character count: 53035

Running title: IL-1β expression map

1 Abstract

Interleukin-1-beta (IL-1β) is a potent, pleiotropic, pro-inflammatory cytokine and its transcriptional regulation and expression are key steps in the mounting of robust inflammatory responses. Lipopolysaccharide (LPS)-induced IL-1β expression requires toll-like receptor 4 (TLR4) in myeloid cells and is driven in large part via the MyD88-IRAK-1-IRAK4-TRAF6 complex and consequently p38-MAPK, JNK and NF-κB signaling networks. LPS triggered responses involve signaling networks that are utilised by many other inflammatory stimuli and therefore presents a very useful experimental model. Here we present a signaling network map of LPS- induced IL-1β transcription. Transcription factors (TF) targeted by these signaling networks coalesce at precise nucleotide binding elements within the IL-1β regulatory DNA. Constitutive binding of PU.1 and C/EBP-β TF’s are obligate for IL-1β expression. We suggest that PU.1 and C/EBP-β TF’s form scaffolds for the tailored modulation of dynamic control exerted by other TF’s, exemplified by c-Jun. Similarly, IL-1β expression induced via NF-κB, CREBP and Fox01 provides a further layer of modulation. High-throughput parallel measurements of interactions show that combinations of TF regulate transcription.. Epigenetic factors, such as the hetero-euchromatin balance, are also important in the relative transcriptional efficacy in different cell types. The development of network maps such as these are a mandatory first step in the building of dynamic models capable of being simulated in silico . A deeper understanding of the transcriptional regulation of IL-1β will inform a broader understanding of cytokine modulation at the cellular level and the development of chronic inflammatory diseases.

Number pages: 38

2 Introduction Pro-inflammatory cytokines and particularly interleukin-1 (IL-1) 1 are implicated in acute and chronic inflammatory conditions associated with common diseases such as atherosclerosis, asthma, cancer, stroke 2, chronic obstructive pulmonary disease, obesity, inflammatory bowel disease, and rheumatoid arthritis 3. IL-1 family consists of 11 members (ILF1 – ILF33) 1. The prototypical and pleiotropic IL-1 receptor I (IL-RI) agonists IL-1 alpha (IL-1α) and IL-1 beta (IL-1β) are synthesised primarily by cells of the innate immune system such as macrophages and monocytes. They are first translated as 31 kDa precursors and later converted to their 17 kDa mature forms by proteolysis promoted by either calpain and caspase-1, respectively 4. Mature IL-1β is released in response to stimuli that activate a potassium ion efflux (e.g. via P2X 7) of myeloid cells and acts as a paracrine mediator of inflammation. IL-1α is also thought to play a role as an autocrine growth factor 5, 6 . The three step process of expression, activation and release of IL- 1β compared to the two step process of other pro-inflammatory cytokines (i.e. expression, release) is indicative of the important role of this protein in innate immune signaling and inflammation. Nevertheless, this still belies the extensive negative feedback that takes place at TLR4 and downstream complexes to further attenuate IL-1 mediated signaling (for review see 7).

Lipopolysaccharide (LPS) is a primary constituent of the Gram-negative bacterial outer membrane and mediates mammalian endotoxic responses; a major complicating factor in sepsis 8. The recognition of LPS by toll-like receptor 4 9 (TLR4) was demonstrated by Poltorak . The high affinity interaction (EC 50 = 4.41 ± 0.16 pg / mL, 10 ) between TLR4 and LPS, convoluted and sequential assembly of proteins (i.e. LBP, CD14, MD-2) and their evolutionary conservation suggests that the LPS signal and, unsurprisingly by proxy, that bacterial infection is sufficiently noxious to warrant their co-ordinated expression. Interest in TLR4 mediated responses extends beyond the paradigm of infection to include sterile injury 11 , where it is becoming arguably recognized that intracellular molecules (e.g. HSP60, HMGB1) also transduce their effects via TLR4 12 and particularly macrophage

3 expression of IL-1β 13, 14 . The requirement for an integrated and coherent map of TLR4 mediated IL-1β expression in myeloid cells is of vital importance to an understanding of both fundamental mechanisms of cell signaling and the pathogenesis of inflammation.

Here, we present an integrated and mechanistic signaling network map detailing the detection of LPS, transduction of the signal from the cytoplasm to the nucleus and consequent expression of IL-1β in macrophages and monocytes of human and murine origin. This includes an overview of the transcription factors (TF) that bind the regulatory DNA and modulate pro-IL-1β gene expression. Further, to describe regulation of IL-1β gene expression by indirect kinase mediated signaling (p38 MAPK 15 and JNK 16 ) and direct transcriptional control via nuclear factor kappa- light-chain-enhancer of activated B cells (NF-κB) mediated signaling 17 . This review focuses on the molecular interaction map presented as a systems biology graphical notation (SBGN) 18 process diagram (Fig 1) and supplementary systems biology mark up language (SBML) 19 file.

Owing to the degenerate nomenclature of transcription factors over preceding decades, it is important to specify these entities in a clear way. Table 1 lists the relevant TFs uniquely identified through the UNIPROT database 20 naming conventions and identifiers, as well as their related synonyms (Table 1). This should facilitate unique identification of these molecules.

4 Table 1. Transcription factors (TFs) involved in IL-1β transcription. Transcription Human Synonyms Human / murine factor gene name UniProt Accession ID PU.1 SPI1 Spi1, NFIL-1βA (NF-βA) P17947 / P17433

NF-κB p65 NFKB3 p65, RelA, NFKB3p50 Q04206 / Q04207 (p105/NFKB1 precursor)

NF-κB p50 NFKB1 KBF1, NF-kappaB, P19838 / P25799 NFkappaB, NFKB-p50, p50, p105

C/EBP-β CEBPB NF-IL6, LAP, TCF5, P17676 / P28033 PP9092 IRF8 IRF8 NF-β1, ICSBP1 Q02556 / P23611

IRF4 IRF4 NF-β2, NF-EM5, LSIRF, Q15306 / Q64287 MUM1

Methods and search criteria

A comprehensive map of TLR signaling has been constructed, but provided no detail regarding transcriptional events 21 . We have interrogated published data using literature already known to us 21, 89, 90, 103 . Additionally, we show the initial genesis of targeted PubMed 135 search queries (Table 2) that were used to update known post- translational modifications (PTMs) of signaling proteins from 21 and the actions of TF’s at IL-1β regulatory DNA where an effect has been observed on a reporter (e.g. luciferase). We present the IL-1β regulatory DNA sequence (Figure 1) in accordance with the structure of the BDC454 clone (a 14 kb sequence containing the entire transcriptional unit of the IL-1β gene) 22 . It has been necessary to assume that

5 biochemical interactions validated in cell types (e.g. HEK293) other than myeloid cells and often in the absence of LPS stimulation also occur in macrophages and monocytes under these conditions.

Table 2. Publications and search terms used in PubMed for extracting literature concerning LPS-directed IL-1β transcription. Author Reference search terms References Reference # returned #

Toll-like receptor signal transduction a

Oda and 21 7, 9, 14, Kitano 19, 27, 30, 2006 31, 41, 49, 55, 56, 60, 73, (LPS activation of MKK) OR 51 29, 82, 41 ((TLR4 and TRAF6) AND and macrophage) OR (TLR4 activation of MKK) OR (traf6 activation of MKK) OR (TAK1 activation of MKK)

LPS OR 2042 9, 10, 55, lipopolysaccharide[Title/Abstr 56, 27, 31, act] AND Toll- 63, 25, 12, like[Title/Abstract] AND 67, 126, receptor*[Title/Abstract] AND TLR4 [Title/Abstract]

LPS AND TAK1 OR TAB1 92 49 AND phosphorylation

LPS OR lipopolysaccharide 12 41 OR TLR4 AND MKK6 AND p38

IL-1β transcription

Zhang et 89 14 104, 100, al., 2008 107, 110, 115, 103, 90, 116,

6 102, 109, 121, 101, 111, 88 Unlu et 103 10 22, 97, 98, al., 2007 106, 90, 104, 100, 101, 102, 110

Liang et 90 10 100, 107, al., 2006 108, 106, 116, 102, 104, 88, 111, 109

LPS OR lipopolysaccharide 471 117, 118, OR TLR4 OR toll-like receptor 113, 120, 4 AND (p38 MAPK OR NF- 114, 121, κBb) AND interleukin-1 beta 104, 111, OR IL-1 beta AND expression 109, 106, AND transcription AND 98, 72 macrophages

((Interleukin-1 beta OR IL-1 118 117, 118, beta) AND and Promoter) 113, 120, AND and macrophages 114, 121, 104, 109, 106, 99 a Additional search terms relating to Toll-like receptor signal transduction included "Mal TLR4", "LPS TLR4 MD-2", "IRAK TLR4", "TLR4 Pellino", "Pellino IRAK", "TLR4 TAK1", "TLR4 Tollip", "TRAF6 UeV1A Ubc13" b Replacement of ‘p38 MAPK’ with ‘NF-κB’ yielded an identical set of search results.

7 Structural overview of the signaling network map effecting IL-1βββ expression

The sequence of events that convert the LPS sequestered signal via lipopolysaccharide binding protein (LBP) 23 and glycosylphosphatidylinositol (GPI) linked glycoprotein: CD14 24 into an IL-1β output are presented in Figure 1. Control mechanisms facilitating the numerous feedbacks are encapsulated in Table 2. Leucine rich repeats (LRR) in TLR4 aid the detection of LPS and modulate TLR4 interaction with lymphocye antigen 96 (MD2) respectively 12 ,25 . This leads to the recruitment of signaling adaptor proteins containing Toll-like/IL-1 receptor (TIR) domains.

TIR domain signaling adaptors specify initial output signal

Signaling catalysed by the TIR containing adaptor protein, myeloid differentiation primary response gene (88) (MyD88) 26 in conjunction with the TIR domain containing adaptor protein (TIRAP, also called Mal) 27-29 is of primary concern here. These early MyD88 dependent signaling events activate IL-1β expression via p38 MAPK, JNK and NF-κB dependent transcription 30 . Endosomal internalisation of later TLR4 signaling occurs via the complexation of TLR4 with TIR-domain- containing adapter-inducing interferon-β (TRIF) and TRIF-related adapter molecule (TRAM) 31.initiating interferon expression. We make no attempt to review TRIF / TRAM dependent responses here 28 .

IRAK-1/4 transduces MyD88-dependent signals

Interleukin-1 receptor associated kinase (IRAK) proteins are recruited via their N- terminal death domains to co-localise with MyD88 32 , 33 . The activation of IRAK occurs via rapid phosphorylation, (potentially by IRAK4 in vivo ) and subsequent auto-phosphorylation 34 . Hyperphosphorylation of key serine, proline and threonine residues in the proST domain 32 of IRAK1 achieve full kinase activation. This

8 facilitates IRAK1 uncoupling from MyD88 and subsequent interaction with tumor 35 necrosis factor receptor-associated factor 6 (TRAF6) .

Ubiquitination; an essential post-translational modification for transduction and feedback

Polyubiquitination of the Lys-63 residue of IRAK1 is mediated by the interaction of Pellino E3 ubiquitin ligases 36 and TRAF6 37 enabling IRAK1 binding to NF-κB essential modulator (NEMO) and consequent NF-κB activation 37, 38 . Ubiquitination is a vital post-translational modification necessary to ensure the efficient transduction of the LPS signal via IRAK proteins.

TRAF6 and TAK1 specify the signaling sub-system

TRAF6 functions as a central component of the LPS signaling complex which is bound to IRAK1 via interactions of both proteins with TRAF-interacting protein with a forkhead-associated (FHA) domain (TIFA) 39 . TIFA is required for inhibitory kappa kinase (IKK) activation and is thought to play a role in oligomerisation and polyubiquitination of TRAF6 40 . The E3 ubiquitin ligase activity of TRAF6 is conferred by its really interesting new gene (RING) domain and is necessary for its interaction with the dimeric E2 enzyme Ubc13/Uev1A (also called TRAF-6 regulated IKK activator 1 (TRIKA1) 41 ).

Transforming growth factor-β associated kinase-1 (TAK1) activation occurs at the plasma membrane via interactions with TAK1 binding protein 1 (TAB1) and TAB2. Phosphorylation of TAK1 occurs following TAB1 and TAB2 activation 42 by Lys- 63 poly-ubiquitination of TAB2 43 . A bifurcating output may occur here where a putative Lys-63 poly-ubiquitinated complex of .IRAK4.IRAK1.TRAF6.UeV1A.UBC13 that condenses with a TAK1 -TAB1/2 complex in the absence of Mitogen-activated protein kinase kinase kinase 3

9 (MEKK3) 44 (Figure 1) may activate IKK 45 ,46 via NF-κB inducing kinase (NIK) 47 , 48 and p38 α MAPK directly via its auto-phosphorylation 49 . This may occur ahead of MKK6, MKK7, and direct canonical IKK 50, 51 activation which has a requirement for MEKK3.

The plethora of negative feedback loops (Table 2) in the upper echelon of the TLR4- IL-1β network are indicative of the structured and coordinated regulation of this cytokine. Ubiquitination and phosphorylation are two vital PTMs in mediating these effects. Although TLR inputs other than TLR4 are not presented explicitly, it is reasonable to assume our map conforms to the ‘bow tie’ structure elucidated by Oda and Kitano 21 . Where signals are transmitted via the active TAK1.TAB1/2.IRAK4.IRAK1.TRAF6.UeV1A.UBC13 (here on in the TRAF6 active) complex and ‘kinase pivot’ (e.g. all tiers of MAPK and NIK and IKK) (Figure 1) differentially to receiver TFs. A member of the Nod-like receptor family, NLRP12 inhibits both the incorporation of IRAK1 and catalyses NIK degradation 52 . The NLR family is involved with inflammasome formation, leading to IL-1β processing 53 . Although, not demonstrated as yet, NLRP12 may catalyse the maturation of IL-1β, whilst inhibiting its transcription. This opposing feedback control exerted by NLRP12 at transcriptional and translational levels may function to resolve inflammation.

10 Table 3. Feedback control of lipopolysaccharide stimulated signal transduction mediating IL-1β expression Complex / Protein Type of feedback Source Target Notes References affected (see Fig 1) Bound TLR4 Negative sTLR4 CD14, MD-2 and Competes for 54, 55 LPS targets, inhibiting TLR4 signaling Bound TLR4 Negative MD-2 TLR4 MD-2 dimerises, 56 , 57 limiting interaction with TLR4 TLR4 Negative Triad3A TLR4 Ubiquitinates TLR4 58 triggering proteasomal degradation Bound Receptor Negative SIGRR TLR4-MyD88- Sequesters the 59 TIR complex TIRAP complex preventing further signaling Bound Receptor Negative ST2L TLR4-MyD88- Competes with 60 TIR complex TIRAP TLR4 for MyD88 and TIRAP

11 Table 3. Feedback control of lipopolysaccharide stimulated signal transduction mediating IL-1β expression Complex / Protein Type of feedback Source Target Notes References affected (see Fig 1) IRAK1.IRAK4.TIFA. Negative 186 aa of N-term IRAK1 / IRAK4 N-terminal 61 TRAF6 residues of IRAK sequence facilitates the degradation of the IRAK protein IRAK1 Negative Tollip IRAK1 / IRAK4 Thr 66 mediates 62, 63 , 64 interaction with Tollip, thus limiting interaction with MyD88 and IRAK4 Bound Receptor TIR Positive IRAK Pellino Phosphorylation of 38 complex Pellino by IRAK amplifying signal Bound Receptor TIR Negative viral-Pellino IRAK Lacks E3 activity 65 complex and inhibits signaling

12 Table 3. Feedback control of lipopolysaccharide stimulated signal transduction mediating IL-1β expression Complex / Protein Type of feedback Source Target Notes References affected (see Fig 1) Bound Receptor Positive Smad6 / Smad7 Pellino Differential binding 66 TIR complex may enhance IRAK mediated signals Bound Receptor Negative IRAK1 / IRAK4 TIRAP Degradation of 67 TIR complex TIRAP by IRAKs Bound Receptor Negative IRAK-M TRAF6 IRAK-M prevents 68 TIR complex formation of an active TRAF6 complex Bound Receptor Negative TIFAB IRAK1 Impedes IRAK-1 – 69 TIR complex TRAF6 interaction Bound Receptor Negative ZCCHC11 TRAF6 Zinc finger proteins 70 TIR complex MCPIP1 that de-ubiquitinate 71 TRAF6

13 Table 3. Feedback control of lipopolysaccharide stimulated signal transduction mediating IL-1β expression Complex / Protein Type of feedback Source Target Notes References affected (see Fig 1) TAK1.TAB1. Negative p38 α MAPK TAB1 TAB1 72 TAB2.IRAK4. phosphorylation IRAK1.TIFA. impairs TAK1 TRAF6.UeV1A. UBC13 TAK1.TAB1. Negative A20 TRAF6 A20 (Tnfaip3) de- 73 TAB2.IRAK4. ubiquitinates IRAK1.TIFA. TRAF6 inactivating TRAF6.UeV1A. this complex UBC13 p38 MAPK Negative DUSP1 p38 MAPK DUSP1 (MKP-1) 74 JNK1 JNK1 attenuates the 75 phosphorylation of both p38 and JNK1 p38 MAPK Negative PTP4A3 p38 MAPK Decreases p-p38 76

14 The activation of indirect and direct transcriptional control of IL-1βββ expression

The activation of indirect MAPK (c-jun N-terminal kinases (JNK) and p38) and direct NF-κB mediated signaling provides layers controlling IL-1β transcription. . The activation of both JNK and p38 MAPK sub-types is initiated via cellular stresses such as arsenite, ansiomycin, IL-1 and TNF 16 ,15 and potently by LPS stimulation 77 ,78 . Both JNK 79 and p38 MAPK 14 are regulated via the mitogen activated protein kinase kinase kinase (MAP3K or MKKK), apoptosis signal- regulating kinase 1 (ASK1) 80 and MAP2K (or MKK) isoforms that include MKK7 81, 82 , and MKK4 83 ,84 , MKK3/6 83, 85 respectively. Negative feedback mediated by dual specificity MAPK phosphatase 1 (DUSP1 or MKP-1) 86 , 75 and the protein tyrosine phosphatase – PTP4A3 76 is summarised in table 2.

A direct association of p38 MAPK with casein kinase II (CKII) under stress conditions in HeLa cells, albeit in the absence of LPS has been observed 87 . Lodie et al 88 established the LPS induced phosphorylation of PU.1 at a ser 148 residue that sits within a CKII motif. CKII inhibitors (e.g. apigenin, disogenin) have also been used attenuate LPS-induced IL-1β transcription 89-91

NF-κB is the proto-typical inducible TF that initiates either cannonical (via IKK phosphorylation of I κBα) or non-cannonical signaling via NIK) freeing NF-κB dimers to engage in gene transcription 17 . Activation of this signaling complex involves stresses similar to that observed for MAPK signaling networks (e.g cytokines, protein kinase C activators, oxidants and viruses) 92

The structure of IL-1βββ regulatory DNA

The isolation of human and mouse complementary DNAs (cDNA) that initiated the study of IL-1 gene regulation has been well documented 93 ,94 ,95 . Identification of TF binding elements (i.e.TATA (-31) and CCAAT (-126 and -75) motifs) 22 , TATAAA (-32 to -27) and CCAAG (-74 to -71) 96 , the stimulation of human monocytes by

15 LPS 97 , phorbol 12-myristate 13-acetate 98 and definition of the CAP site proximal (CSP) promoter element (-131 to +1) and phorbol ester inducible enhancer (-2982 to -2795) 99 are presented in the references herein. Enhancer (-3,757 to -2599) and promoter (-1807 to +1) regions are defined arbitrarily here for the sake of clarity (Figure 1)

The LPS sensitivity of the IL-1βββ enhancer

Regulatory regions that effect the LPS sensitivity of IL1B transcription are recognised as either overlapping or upstream of the phorbol ester sensitive element 99 . Shirakawa et al 100 defined three regulatory regions in the IL1B enhancer from the human BDC454 clone (a 14 kb sequence containing the entire transcriptional unit of the IL-1β gene) situated between -3757 to -2923 nucleotides that were LPS sensitive. These nucleotide locations were later updated by the same group 101 (legend of Table 3). Regions were assigned as A (-3757 to -3132), B (-3132 to - 2987) and C (-2987 to -2923) (Table 3 and Figure 1) and thought responsible for 20, 60 and 20 % of the LPS driven transcriptional response of IL1B, respectively. Although not necessary for an LPS-inducible signal, deletions of a downstream sequence (-2729 to -131) revealed impaired expression of the CAT reporter. Deletion of a region between -2729 to -2599 enhanced CAT expression, suggesting negative feedback on expression was mediated by elements within this region. Shirakawa et al summarise their findings by sub-dividing the -3757 to -2599 region into sub-regions denoted A - J (Table 3 and Figure 1) with varying degrees of importance in supporting LPS-inducible expression. Regions C and E are specified as being 'required', regions B, H and I as being strongly supportive and region A as weakly supportive of a role in LPS-inducible expression. This study 101 denoted the distal nucleotide of region B (-3134) to the proximal nucleotide of region I (-2729) as the upstream induction sequence (UIS).

An update of the function of the UIS of the IL1B enhancer, specifically between regions F - G (named the LPS and IL-1 responsive element; (LILRE) 101 ,102 ) has been

16 outlined 103 . A trimer of a non-tyrosine phosphorylated Stat 1, PU.1 and IRF8 is bound constitutively between a narrow element encompassing -2862 and -2831 nucleotides in the LILRE (encompassed by elements -2866 through to -2833; Figure 1). The authors were able to attribute ~ 90 % of the reporter expression to IRF8 phosphorylation, when comparing wild-type and a non-phosphorylatable mutant IRF8 Tyr -211-Phe expression vectors, co-transfected with TRAF6, PU.1 and an IL1B XT (-3757 to +12) – luciferase coupled reporter in LPS (10 µg / mL, 4 h) stimulated RAW264.7 cells. Although knowledge of dimeric interactions between IRF and PU.1 proteins at the IL-1β enhancer had been noted much earlier 104.

17 Table 4. Enhancer regulatory domains of the human IL-1β gene. An earlier report 100 had missed the presence of two cytosine (C) nucleotides between positions -2748 to -2747 that were later accounted for 101 . The corrected regions are shown. Nucleotide positions within regions correspond to their relative position to the transcriptional start site (+1) here and in the main body of the text. Region Binding Site Identity References -2982 to -2795 Putative AP-1 (-2892 to -2883) & IFN-β 99 (-2853 to -2845) -3757 to 3132 Region A 101 -3134 to -2989 Region B, Putative AP-1 -2989 to -2925 Region C -2925 to -2896 Region D, Putative AP-1 -2896 to -2866 Region E, NF-IL6 -2866 to -2846 Region F, NF-β1 -2846 to -2833 Region G, NF-β2 -2833 to -2782 Region H -2782 to -2729 Region I, CREB / AP-1 & NF-κB site -2729 to -2599 Region J, Putative AP-1

18 The role of PU.1 and C/EBP-βββ TF’s in the IL-1βββ promoter

Human IL1B and murine Il1b gene promoter architectures reside in open or accessible configurations and are not complexed with either acetylated or unmodified histone H3 90 . The constitutive association of both PU.1 and C/EBP β in both mouse (RAW264.7) and human (MM6) macrophages was demonstrated via chromatin immunoprecipitation (ChIP) 90 . Within the map (Figure 1) these associations are demonstrated by interactions with elements from -199 through to +1. Similarly, the TATA box binding protein (TBP) is constitutively bound to the IL1B promoter in MM6 monocytes 89 , further exemplifying the transcriptional poise of this promoter.

The critical involvement of PU.1 alone 105, 106 and in tandem with C/EBP-β 107, 108 , in driving IL-1β expression has been reported widely. Structurally, there are three PU.1 sites in the CSP promoter (-120 to -93; duplicate overlapping sites, 106 -49 to - 38 105, 106, 108, 109 ) and two C/EBP-β sites (-90 to -82 107 and -41 to -33 108 ) (Figure 1). The promoter proximal PU.1 binding element and its position in relation to the TATA box motif is 100 % conserved between human and mouse sequences, suggesting a primary role of PU.1 in directing the initiation of IL-1β transcription, likely via transcription factor IID (TFIID) 105 . Co-transfection studies of wt PU.1 with constructs encoding a mutated C/EBP-β element revealed decreased binding to a glutathione-s-transferase-basic leucine zipper domain (GST-bZIP) column. Similarly, co-transfection of a mutated PU.1 domain and wt C/EBP-β showed a 45 % reduction in wt il1b (-59 to +12) coupled luciferase reporter expression 110 . Yang et al., suggest the synergistic enhancement of il1b expression via the PU.1 winged helix turn helix (wHTH) DNA binding domain tethering the bZIP domain of C/EBP- β. They further speculate interactions with PU.1 bound to its distal CSP site and the same protein bound to the UIS in the enhancer and modulation of this proximal PU.1/C/EBP-β overlapping element by p38 MAPK. The transcriptional targets of p38 α include the C/EBP-β TF via the sensitivity of IL-1β (-50 to 1) promoter and

19 CAT constructs display concentration-dependent sensitivity to the bicyclic imidazole; SB203580 111 . Heat shock factor (HSF) has also been shown to compete with PU.1s interaction with the bZIP domain of C/EBP-β, effectively decreasing the efficacy of these TFs 112 in THP-1 cells. However, heat shocked (39.5 °C) RAW264.7 cells show an increased IL-1β mRNA expression compared to a decreased THP-1 release of this cytokine 113 . These observations can be reconciled by the presence of a single adenine nucleotide present in the mouse (but not human) heat shock response element (HSE) that likely impairs binding of HSF to the promoter 113 . Conversely, high mobility group box protein 1 (HMGB1) has been shown to enhance IL-1β promoter (HT fragment) luciferase reporter expression in RAW 264.7 macrophages 114 .

The lack of a nucleosome associated promoter and focus of the p38 MAPK and HMGB1 and HSF TFs on the constitutively bound PU.1 and C/EBP β promoter complexes points to the rapid and acute regulation of IL-β expression. However, modulation can also be achieved via inducible signals.

The role of inducible TF signaling in IL-1βββ transcription

NF-κκκB

A decameric nucleotide NF-κB binding motif (-297 to -288) is located in the IL1B promoter and concentration-dependently enhances CAT reporter expression 115 . Similarly, NF-κB p50/p65 heterodimers have the greatest efficacy in driving CAT reporter expression from an IL-1β x2-κB plasmid 115 . Electrophoretic mobility shift assay (EMSA) analysis confirmed the binding of LPS (10 µg / mL, 1 h) stimulated THP-1 nuclear extracts to a labeled 25 bp oligonucleotide probe (-305 to -280) that was effectively displaced by competition for by unlabeled Ig NF-κB oligonucleotide 116 .

20 The role of a novel IκB protein IκBβ has recently been established suggesting this protein maybe selective in modulating IL-1β transcription alone 117 (Figure 1). ChIP analysis of LPS (100 ng / mL) stimulated RAW 264.7 (2h) and BMDM (1h) macrophages revealed association of NF-κB p65 and c-Rel with the Il1b promoter (- 300). In contrast IκBβ (-/-) mutants have decreased c-Rel recruitment to their κB sites. LPS stimulated (100 ng / mL, 8 h) IL-1β-luciferase reporter constructs containing variable numbers of -κB sites (-518; x2, -399; x1 and -133; 0) were transfected into RAW264.7 cells. The -133 construct revealed a marked reduction in its reporter expression in comparison to both -518 and -399 constructs.

Phosphorylation of Ser 536 of the NF-κB p65 subunit is not necessary for this subunit’s recruitment to the IL1B promoter as measured by ChIP in PMA (10 ng / mL) differentiated U937 macrophages stimulated with LPS (1 µg / mL; for 2 or 6h) 118 . Leptomycin (10 nM: an inhibitor of I κBα nuclear export) markedly affected LPS (10 ng / mL; 2 h) stimulated IL-1β transcription and release in U937 macrophages. Similarly, leptomycin ablated the NF-κB DNA binding activity of U937 macrophage nuclear extract, but did not affect their ability to bind AP-1 and CREB DNA sequences 118 . NF-κB p65 subunit appearance in the nucleus and I κBα disappearance from the cytoplasm has also been observed in RAW264.7 macrophages stimulated with LPS (1 µg/mL) at 15 and 30 min time points 119 .

21 FoxO1

A role for the FoxO1 TF (which is also implicated in insulin signaling) in LPS directed IL-1β expression has recently been discovered 120 (Figure 1 elements -1807 through to -1296). RAW264.7 cells transfected with an IL-1β luciferase reporter and vectors encoding FoxO1 and either p50, p65 or both revealed the synergistic interaction of FoxO1 and NF-κB p50 in response to LPS stimulation (100 ng / mL, 24 h) 120 . cAMP response element binding protein (CREBP)

Enhancer located cAMP response elements were identified (Figure 1, Table 3) in the BDC454 clone 101 . The over-expression of constructs co-expressing C/EBP-β and CREBP in LPS (10 ng / mL, 24 h) stimulated THP-1 cells inhibited this LPS stimulated I/ fos CAT reporter expression 101 . However, the treatment of reporter transfected THP-1 cells with LPS and either di-buturyl cAMP (100 µM) or the catalytic subunit of protein kinase A enhanced CAT reporter expression in a concentration-dependent manner 101 . c-Jun

Attenuated real-time (RT) PCR measurements of c-Jun were made in LPS stimulated (10 ng / mL; 45 min) JNK1 knockout (vs. wt) bone marrow derived macrophages (BMDM) 74 . The small molecule JNK Inhibitor II, SP600125 suppressed expression of the luciferase reporter in RAW264.7 cells from a wild-type PU.1 lipocalin-prostaglandin D synthase (L-PGDS) promoter but not that of a PU.1 S148A mutant 119 . Treatment with ChIP also confirmed a direct interaction between PU.1 and c-Jun in RAW264.7 macrophages and BMDM's from C57BL/6 mice. Nuclear extracts derived from either cell type treated in the presence or absence of LPS (1 µg / mL, 4h) were immuno-precipitated with either anti-PU.1 or anti-c-Jun antibodies and the corresponding DNA extracted and amplified by PCR (using

22 primers flanking the PU.1 site). In both LPS stimulated macrophage types, PU.1 and c-Jun were detected by immunoblots . A ChIP assay in RAW264.7 macrophages 119 established that c-Jun binding to the PU.1 site in the L-PGDS promoter was reduced by addition of SP600125 (10 µM) or SB-220025 (10 µM) reflecting predominant, if likely non-selective JNK and p38 MAPK inhibition. However, neither JNK or p38 MAPK inhibition affected PU.1 binding to the L- PGDS promoter in LPS stimulated RAW macrophages.

The binding (compared to wt) of [35 S] labeled mutant c-Jun homodimers to GST- C/EBP-β columns revealed a relatively decreased binding compared to that observed in the GST-PU.1 column. However, co-transfection of these homdimers with PU.1, C/EBP-β and an IL-1β131 luciferase reporter construct showed a marked activation of reporter expression vs. wt. In summary, c-Jun co-activation of PU.1 and C/EBP- β complexes via the DNA binding domains of c-Jun necessitate their interaction and preclude c-Jun engaging in direct transcriptional activation 121 .

Induced binding of the NF-κB, c-Jun and CREBP TFs presented here point to an additional layer of regulation that may reflect the phased fine-tuning of IL-1β expression in response to environmental stimuli. For example, LPS stimulated CREBP-C/EBP-β complexes may provide a tonic inhibition of IL-1β expression in the absence of protein kinase A mediated phosphorylation of CREBP. Enhancement of IL-1β expression may commence once this CREBP phosphorylation has been completed 101 . The co-dependency of FoxO1 and NF-κB p50 in regulating IL-1β expression also suggests that combinatorial TF binding and action being a general strategy by which TF achieve their aims (see next section - Parallel measurements of LPS stimulated transcription in myeloid cells)

23 Parallel measurements of LPS stimulated transcription in myeloid cells

Network maps compiled from the literature often present the observer with a biased set of interactions with respect to environment and experimental perturbation where often only limited sets of TF’s are monitored. Increasingly the use of high- throughput, parallel measurements allow measurement of genome-wide changes in TF expression, progressively alleviating this bias (Table 4).

The degenerate nature of TF binding motifs and the often overlapping nature of these sites in conjunction with the native chromatin structure of the gene make it difficult to assign a predominant transcriptional regulation by any one factor. Indeed, the combinatorics of TF binding and dynamics of these interactions make understanding IL-1β expression very complex (e.g. sub-proteins of the NF-κB family). 122 explored combinatorial interactions of TFs using mammalian-2-hybrid (M2H) and qRT-PCR assays in tandem. The authors provide a classification of TFs into ‘facitlitators’ or ‘specifiers’.The former have a broad tissue expression that in combination with the narrower tissue expression of ‘specifier’ TFs dictates development and or embryogenesis dependent on the cellular context. Progress already made by applying these parallel measurement technologies will hopefully continue to provide even more detail concerning IL-1β transcription.

24 Table 5. LPS directed changes in innate immune cells of myeloid type measured via parallel measurement methods Cell type LPS Time Method Results Conclusion Ref (ng / (hours) mL) BMDM (p300) 10 2 ChIP-seq Identification of 2432 (p300 acetyl-transferase) inducible PU.1 association is largely 123 RAW264.7(PU.1) peaks (vs.control). constitutive. PU.1 association not different between LPS treated p300 plays a role in inductive (44,243) and untreated (44,4453). gene transcription. p300 acetyltransferase binding is LPS induced at promoter (-184) and enhancer (-39461) regions, but not other enhancers (-9901 and -2466) relative to the il1b gene. BMDM 10 2 cDNA Identification of 2892 transcripts. Transcripts clustered based on 124 microarray Notable transcripts down-regulated 2h after LPS promoter structure were qPCR stimulation: Tlr4, Stat4, Sirt1 significantly co-expressed. Notable transcripts up-regulated 2h after LPS stimulation: Nfkb2, Tlr2, MAP3K8, Bcl3, Cxcl2. Notable TF’s up-regulated 2h after LPS stimulation: Atf4, Cited2 (p300 interacting protein), Nfkbiz, Bcor Abbreviations. BMDM: Bone marrow derived macrophage

25 Table 5. LPS directed changes in innate immune cells of myeloid type measured via high-throughput methods Cell LPS Time Method Results Conclusion Ref type (ng / (hours) mL) BMDC 100 6 Microarray, Identification of 2322 significant regulatory Delineated transcriptional programs. 125 shRNA perturbation interactions noted. shRNA perturbation of LPS PAM (TLR2) inflammatory response stimulated cells revealed that Nfkbiz, Nfkb1, and a poly(I:C) (TLR3/MDA5) Socs6, Runx1, Bhlhb2, Hat1, Hhex, Trim12, antiviral response. Both are induced Cebpb,Cited2, Bcl3 were all up-regulated for il1b by LPS (TLR4). expression. BMDM 100 0.75 SILAC and phosphopeptide in silico promoter analysis of 50 TF families with Merging of TF phosphorylation data 126 enrichment based MS and in phosphorylated members identified enrichment of alongside experimental and in silico silico promoter analysis. NF-κB, CREB and IRFF in LPS induced genes. transcriptome analysis identified non siRNA knockdown siRNA knockdown of Cebpz, Hsf1, Atf7and Cic redundant roles for TFs in Il1b suggested Il1a and Il1b as target genes. expression. Abbreviations. BMDM: Bone marrow derived macrophage. BMDC: Bone marrow derived dendritic cell.

26 Discussion

Here we present a map of the LPS stimulated signaling network in macrophages and monocytes of human and murine origin (Figure 1). We have limited the scope of our map at the transcriptional level to TF’s with demonstrated functional roles in driving IL1 β transcription. Nevertheless, we are aware that these efforts are subject to the ‘cartographers curse’ where a map is made redundant by an inevitable and perpetual increase in further information.

The prevailing view of IL1B expression is that the enhancer mediates inducible signaling in conjunction with the proximal CAP site proximal (CSP) promoter region. This is thought to occur via topographical interaction of TF bound UIS enhancer elements that are able to fold onto the accessible promoter structure 110 . In contrast, the CSP promoter appears to be required for basal expression 89 . Although this picture is largely accurate, it belies the acute regulation of the CSP promoter by p38 MAPK and c-Jun, HSF and HMGB1 as discussed previously. A20 (also called TNFAIP3 ) displays ubiquitinating and de-ubiquitinating activities that activate proteins and target the same for proteasomal degradation respectively. These processes operate above and below the ‘kinase pivot’. A20 deactivates the TRAF6 active complex 73 and modulates the ubiquitination status of the IKK complex 127 . On-going investigations should attempt to resolve how macrophage signaling networks interpret the similarities and differences in PTM’s.

It is becoming increasingly clear that epigenetic parameters also impact the expression of IL-1β. The equilibrium between hetero and euchromatin respectively can mediate IL-1β transcription and LPS tolerance in sepsis 128 . Demethylation of undifferentiated and differentiated HL-60 monocytic cells caused increases in IL-1β expression that was augmented by LPS (250 ng / mL; 4 h) stimulation 129

Inevitably, attempting to represent a four dimensional dynamic system (x,y,z and time) in two dimensions (x,y) is a limiting factor in the utility of these maps.

27 Dynamic reconstructions of the network map, in the form of event based models 130 ,131 could be used to investigate the temporal element of this control. Systems biology approaches are likely to significantly enhance our ability to generate testable hypotheses 132 that will allow verification of basic disease mechanisms that contribute to chronic inflammatory conditions and particularly non-communicable disease 133, 134 .

Acknowledgements

We would like to thank the BBSRC for funding this project (BB/F018398/1) and the

BBSRC/EPSRC and Wellcome Trust for the funding of studentships and fellowships.

Author Contributions

BGS searched the literature concerning the indirect and direct transcriptional control of IL-1β expression, analysed the data from these articles, created the map and wrote the first draft of the manuscript.

HS searched the literature concerning the LPS directed signalling downstream of

TLR4, analysed the data from these articles and created the map.

DB, NJR and DBK wrote the grant application, co-investigated / co-supervised BGS and HS and advised on the analysis and creation of the map.

PM supervised BGS and HS and advised on experimental design and data analysis.

JP supervised BGS and HS, directed the investigation, oversaw the writing of the manuscript and advised on the analysis and creation of the map.

28 All authors contributed to the writing of the manuscript and approved its final form.

Competing Financial Interests Statement

DBK transferred his responsibility as Principal Investigator of BBSRC Grant

BB/F018398/1 to NJR and PM prior to taking up his appointment as Chief

Executive of the BBSRC on 1 st October 2008. NJR is a non-executive director of

AstraZeneca. BGS as a former employee of AstraZeneca and Eli Lilly and Company is a shareholder of stock from both companies.

29 References

1. Dinarello, C.A. Interleukin-1 in the pathogenesis and treatment of inflammatory diseases. Blood 117, 3720-3732 (2011). 2. Brough, D., Tyrrell, P.J. & Allan, S.M. Regulation of interleukin-1 in acute brain injury. TiPS In press (2011). 3. Nathan, C. & Ding, A. Nonresolving inflammation. Cell 140, 871-82 (2010). 4. Luheshi, N.M., Rothwell, N.J. & Brough, D. The dynamics and mechanisms of interleukin-1α and β nuclear import. Traffic 10, 16-25 (2009). 5. Luheshi, N.M., Rothwell, N.J. & Brough, D. Dual functionality of interleukin-1 family cytokines: implications for anti-interleukin-1 therapy. Brit J Pharmacol 157, 1318-29 (2009). 6. Werman, A. et al. The precursor form of IL-1α is an intracrine proinflammatory activator of transcription. Proc Natl Acad Sci USA 101, 2434-9 (2004). 7. Liew, F.Y., Xu, D., Brint, E.K. & O'Neill, L.A.J. Negative regulation of Toll-like receptor-mediated immune responses. Nat Rev Immunol 5, 446-458 (2005). 8. Cohen, J. The immunopathogenesis of sepsis. Nature 420, 885-91 (2002). 9. Poltorak, A. et al. Defective LPS Signaling in C3H/HeJ and C57BL/10ScCr Mice: Mutations in Tlr4 Gene. Science 282, 2085-2088 (1998). 10. Du, X., Poltorak, A., Silva, M. & Beutler, B. Analysis of Tlr4-Mediated LPS Signal Transduction in Macrophages by Mutational Modification of the Receptor. Blood Cells, Molecules, and Diseases 25, 328-338 (1999). 11. Chen, G.Y. & Nunez, G. Sterile inflammation: sensing and reacting to damage. Nat Rev Immunol 10, 826-37 (2010). 12. Bryant, C.E., Spring, D.R., Gangloff, M. & Gay, N.J. The molecular basis of the host response to lipopolysaccharide. Nat Rev Microbiol 8, 8-14 (2010). 13. Hsu, H.Y. & Wen, M.H. Lipopolysaccharide-mediated Reactive Oxygen Species and Signal Transduction in the Regulation of Interleukin-1 Gene Expression. J Biol Chem 277, 22131 - 22139 (2002). 14. Matsuzawa, A. et al. ROS-dependent activation of the TRAF6-ASK1-p38 pathway is selectively required for TLR4-mediated innate immunity. Nat Immunol 6, 587-92 (2005). 15. Cuadrado, A. & Nebreda, A.R. Mechanisms and functions of p38 MAPK signalling. Biochem J 429, 403-17 (2010). 16. Raman, M., Chen, W. & Cobb, M.H. Differential regulation and properties of MAPKs. Oncogene 26, 3100-12 (2007). 17. Hayden, M.S. & Ghosh, S. NF-κB in immunobiology. Cell Res 21, 223-244 (2011). 18. Novere, N.L. et al. The Systems Biology Graphical Notation. Nat Biotech 27, 735-741 (2009). 19. Hucka, M. The systems biology markup language (SBML): a medium for representation and exchange of biochemical network models. Bioinformatics 19, 524-531 (2003).

30 20. Apweiler, R. et al. UniProt: the Universal Protein knowledgebase. Nucleic Acids Res 32, D115-9 (2004). 21. Oda, K. & Kitano, H. A comprehensive map of the toll-like receptor signaling network. Mol Syst Biol 2, 2006 0015 (2006). 22. Clark, B.D., Collins, K.L., Gandy, M.S., Webb, A.C. & Auron, P.E. Genomic sequence for human prointerleukin 1 beta: possible evolution from a reverse transcribed prointerleukin 1 alpha gene. Nucleic Acids Res 14, 7897-7914 (1986). 23. Wright, S.D., Tobias, P.S., Ulevitch, R.J. & Ramos, R.A. Lipopolysaccharide (LPS) binding protein opsonizes LPS-bearing particles for recognition by a novel receptor on macrophages. J Exp Med 170, 1231-41 (1989). 24. Gegner, J.A., Ulevitch, R.J. & Tobias, P.S. Lipopolysaccharide (LPS) signal transduction and clearance. Dual roles for LPS binding protein and membrane CD14. J Biol Chem 270, 5320-5 (1995). 25. Park, B.S. et al. The structural basis of lipopolysaccharide recognition by the TLR4-MD-2 complex. Nature 458, 1191-5 (2009). 26. Medzhitov, R. et al. MyD88 is an adaptor protein in the hToll/IL-1 receptor family signaling pathways. Molecular Cell 2, 253-8 (1998). 27. Yamamoto, M. et al. Essential role for TIRAP in activation of the signalling cascade shared by TLR2 and TLR4. Nature 420, 324-329 (2002). 28. Kenny, E.F. & O'Neill, L.A.J. Signalling adaptors used by Toll-like receptors: An update. Cytokine 43, 342-349 (2008). 29. Verstak, B. et al. MyD88 adapter-like (Mal)/TIRAP interaction with TRAF6 is critical for TLR2- and TLR4-mediated NF-κB proinflammatory responses. J Biol Chem 284, 24192-203 (2009). 30. Horng, T., Barton, G.M., Flavell, R.A. & Medzhitov, R. The adaptor molecule TIRAP provides signalling specificity for Toll-like receptors. Nature 420, 329-333 (2002). 31. Yamamoto, M. et al. TRAM is specifically involved in the Toll-like receptor 4-mediated MyD88-independent signaling pathway. Nat Immunol 4, 1144- 1150 (2003). 32. Flannery, S. & Bowie, A.G. The interleukin-1 receptor-associated kinases: critical regulators of innate immune signalling. Biochemical Pharmacol 80, 1981-91 (2010). 33. Swantek, J.L., Tsen, M.F., Cobb, M.H. & Thomas, J.A. IL-1 receptor- associated kinase modulates host responsiveness to endotoxin. J Immunol 164, 4301-6 (2000). 34. Kollewe, C. et al. Sequential autophosphorylation steps in the interleukin-1 receptor-associated kinase-1 regulate its availability as an adapter in interleukin-1 signaling. J Biol Chem 279, 5227-36 (2004). 35. Gohda, J., Matsumura, T. & Inoue, J.-I. Cutting edge: TNFR-associated factor (TRAF) 6 is essential for MyD88-dependent pathway but not toll/IL-1 receptor domain-containing adaptor-inducing IFN-β (TRIF)-dependent pathway in TLR signaling. J Immunol 173, 2913-7 (2004).

31 36. Smith, H. et al. Identification of the phosphorylation sites on the E3 ubiquitin ligase Pellino that are critical for activation by IRAK1 and IRAK4. Proc Natl Acad Sci U S A 106, 4584-90 (2009). 37. Conze, D.B., Wu, C.-J., Thomas, J.a., Landstrom, A. & Ashwell, J.D. Lys63- linked polyubiquitination of IRAK-1 is required for interleukin-1 receptor- and toll-like receptor-mediated NF-κB activation. Mol Cell Biol 28, 3538-47 (2008). 38. Ordureau, A. et al. The IRAK-catalysed activation of the E3 ligase function of Pellino isoforms induces the Lys63-linked polyubiquitination of IRAK1. Biochem J 409, 43-52 (2008). 39. Takatsuna, H. et al. Identification of TIFA as an adapter protein that links tumor necrosis factor receptor-associated factor 6 (TRAF6) to interleukin-1 (IL-1) receptor-associated kinase-1 (IRAK-1) in IL-1 receptor signaling. J Biol Chem 278, 12144-50 (2003). 40. Ea, C.-K., Sun, L., Inoue, J.-I. & Chen, Z.J. TIFA activates I κB kinase (IKK) by promoting oligomerization and ubiquitination of TRAF6. Proc Natl Acad Sci USA 101, 15318-23 (2004). 41. Wang, C. et al. TAK1 is a ubiquitin-dependent kinase of MKK and IKK. Nature 412, 346-51 (2001). 42. Landstrom, M. The TAK1-TRAF6 signalling pathway. Int J Biochem Cell Biol 42, 585-9 (2010). 43. Xia, Z.-P. et al. Direct activation of protein kinases by unanchored polyubiquitin chains. Nature 461, 114-9 (2009). 44. Huang, Q. et al. Differential regulation of interleukin 1 receptor and Toll-like receptor signaling by MEKK3. Nat Immunol 5, 98-103 (2004). 45. Lamothe, B. et al. Site-specific Lys-63-linked tumor necrosis factor receptor- associated factor 6 auto-ubiquitination is a critical determinant of I κB kinase activation. J Biol Chem 282, 4102-12 (2007). 46. Lamothe, B. et al. The RING domain and first zinc finger of TRAF6 coordinate signaling by interleukin-1, lipopolysaccharide, and RANKL. J Biol Chem 283, 24871-80 (2008). 47. Fischer, C. et al. Differential effects of lipopolysaccharide and tumor necrosis factor on monocytic I κB kinase signalsome activation and I κB proteolysis. J Biol Chem 274, 24625-32 (1999). 48. Ninomiya-Tsuji, J. et al. The kinase TAK1 can activate the NIK-IκB as well as the MAP kinase cascade in the IL-1 signalling pathway. Nature 398, 252- 6 (1999). 49. Ge, B. et al. MAPKK-independent activation of p38 α mediated by TAB1- dependent autophosphorylation of p38 α. Science 295, 1291-4 (2002). 50. Deacon, K. & Blank, J.L. MEK kinase 3 directly activates MKK6 and MKK7, specific activators of the p38 and c-Jun NH2-terminal kinases. J Biol Chem 274, 16604-10 (1999). 51. Zhang, D. et al. Identification of MEKK2/3 serine phosphorylation site targeted by the Toll-like receptor and stress pathways. EMBO J 25, 97-107 (2006).

32 52. Ye, Z. et al. ATP binding by monarch-1/NLRP12 is critical for its inhibitory function. Mol Cell Biol 28, 1841-50 (2008). 53. Schroder, K. & Tschopp, J. The inflammasomes. Cell 140, 821-32 (2010). 54. Hyakushima, N. et al. Interaction of soluble form of recombinant extracellular TLR4 domain with MD-2 enables lipopolysaccharide binding and attenuates TLR4-mediated signaling. J Immunol 173, 6949-54 (2004). 55. Iwami, K.I. et al. Cutting edge: naturally occurring soluble form of mouse Toll-like receptor 4 inhibits lipopolysaccharide signaling. J Immunol 165, 6682-6 (2000). 56. Re, F. & Strominger, J.L. Monomeric recombinant MD-2 binds toll-like receptor 4 tightly and confers lipopolysaccharide responsiveness. J Biol Chem 277, 23427-32 (2002). 57. Mullen, G.E.D. et al. The role of disulfide bonds in the assembly and function of MD-2. Proc Natl Acad Sci USA 100, 3919-24 (2003). 58. Chuang, T.-H. & Ulevitch, R.J. Triad3A, an E3 ubiquitin-protein ligase regulating Toll-like receptors. Nat Immunol 5, 495-502 (2004). 59. Qin, J., Qian, Y., Yao, J., Grace, C. & Li, X. SIGIRR inhibits interleukin-1 receptor- and toll-like receptor 4-mediated signaling through different mechanisms. J Biol Chem 280, 25233-41 (2005). 60. Brint, E.K. et al. ST2 is an inhibitor of interleukin 1 receptor and Toll-like receptor 4 signaling and maintains endotoxin tolerance. Nat Immunol 5, 373- 9 (2004). 61. Hu, J., Jacinto, R., McCall, C. & Li, L. Regulation of IL-1 receptor- associated kinases by lipopolysaccharide. J Immunol 168, 3910-4 (2002). 62. Burns, K. et al. Tollip, a new component of the IL-1RI pathway, links IRAK to the IL-1 receptor. Nat Cell Biol 2, 346-51 (2000). 63. Didierlaurent, A. et al. Tollip regulates proinflammatory responses to interleukin-1 and lipopolysaccharide. Mol Cell Biol 26, 735-735 (2006). 64. Neumann, D. et al. Threonine 66 in the death domain of IRAK-1 is critical for interaction with signaling molecules but is not a target site for autophosphorylation. J Leukocyte Biol 84, 807-13 (2008). 65. Griffin, B.D. et al. A poxviral homolog of the Pellino protein inhibits Toll and Toll-like receptor signalling. Eur J Immunol 41, 798-812 (2011). 66. Lee, Y.S. et al. Smad7 and Smad6 bind to discrete regions of Pellino-1 via their MH2 domains to mediate TGF-beta1-induced negative regulation of IL- 1R/TLR signaling. Biochemical and biophysical research communications 393, 836-43 (2010). 67. Dunne, A. et al. IRAK1 and IRAK4 promote phosphorylation, ubiquitination, and degradation of MyD88 adaptor-like (Mal). J Biol Chem 285, 18276-82 (2010). 68. Kobayashi, K. et al. IRAK-M is a negative regulator of Toll-like receptor signaling. Cell 110, 191-202 (2002). 69. Matsumura, T. et al. TIFAB inhibits TIFA, TRAF-interacting protein with a forkhead-associated domain. Biochem Biophys Res Commun 317, 230-4 (2004).

33 70. Minoda, Y. et al. A novel Zinc finger protein, ZCCHC11, interacts with TIFA and modulates TLR signaling. Biochem Biophys Res Commun 344, 1023-30 (2006). 71. Liang, J. et al. MCP-induced protein 1 deubiquitinates TRAF proteins and negatively regulates JNK and NF-κB signaling. J Exp Med 207, 2959-73 (2010). 72. Cheung, P.C.F., Campbell, D.G., Nebreda, A.R. & Cohen, P. Feedback control of the protein kinase TAK1 by SAPK2a/p38alpha. EMBO J 22, 5793-805 (2003). 73. Boone, D.L. et al. The ubiquitin-modifying enzyme A20 is required for termination of Toll-like receptor responses. Nat Immunol 5, 1052-60 (2004). 74. Sánchez-Tilló, E. et al. JNK1 Is required for the induction of Mkp1 expression in macrophages during proliferation and lipopolysaccharide- dependent activation. J Biol Chem 282, 12566-73 (2007). 75. Zhao, Q. et al. The role of mitogen-activated protein kinase phosphatase-1 in the response of alveolar macrophages to lipopolysaccharide: attenuation of proinflammatory cytokine biosynthesis via feedback control of p38. J Biol Chem 280, 8101-8 (2005). 76. Tang, X., Woodward, T. & Amar, S. A PTP4A3 peptide PIMAP39 modulates TNF-alpha levels and endotoxic shock. J Innate Immun 2, 43-55 (2009). 77. Hambleton, J., Weinstein, S.L., Lem, L. & DeFranco, A.L. Activation of c- Jun N-terminal kinase in bacterial lipopolysaccharide-stimulated macrophages. Proc Natl Acad Sci USA 93, 2774-8 (1996). 78. Lee, J.C. et al. A protein kinase involved in the regulation of inflammatory cytokine biosynthesis. Nature 372, 739-46 (1994). 79. Nishitoh, H. et al. ASK1 is essential for JNK/SAPK activation by TRAF2. Molecular Cell 2, 389-95 (1998). 80. Noguchi, T. ROS-dependent Activation of ASK1 in Inflammatory Signaling. J Oral Biosci 50, 107-114 (2008). 81. Tournier, C. et al. MKK7 is an essential component of the JNK signal transduction pathway activated by proinflammatory cytokines. Genes Dev 15, 1419-26 (2001). 82. Bhattacharyya, A. et al. Mitogen-activated protein kinases and nuclear factor-κB regulate Helicobacter pylori -mediated interleukin-8 release from macrophages. Biochem J 368, 121-9 (2002). 83. Brancho, D. et al. Mechanism of p38 MAP kinase activation in vivo. Genes Dev 17, 1969-78 (2003). 84. Winston, B.W., Chan, E.D., Johnson, G.L. & Riches, D.W. Activation of p38 mapk , MKK3, and MKK4 by TNF-α in mouse bone marrow-derived macrophages. J Immunol 159, 4491-7 (1997). 85. Remy, G. et al. Differential activation of p38MAPK isoforms by MKK6 and MKK3. Cell Signal 22, 660-7 (2010). 86. Dickinson, R.J. & Keyse, S.M. Diverse physiological functions for dual- specificity MAP kinase phosphatases. J Cell Sci 119, 4607-15 (2006).

34 87. Sayed, M., Kim, S.O., Salh, B.S., Issinger, O.G. & Pelech, S.L. Stress- induced activation of protein kinase CK2 by direct interaction with p38 mitogen-activated protein kinase. J Biol Chem 275, 16569-73 (2000). 88. Lodie, T.A. et al. Stimulation of macrophages by lipopolysaccharide alters the phosphorylation state, conformation, and function of PU.1 via activation of casein kinase II. J Immunol 158, 1848-56 (1997). 89. Zhang, Y., Saccani, S., Shin, H. & Nikolajczyk, B.S. Dynamic protein associations define two phases of IL-1β transcriptional activation. J Immunol 181, 503-12 (2008). 90. Liang, M.D., Zhang, Y., McDevit, D., Marecki, S. & Nikolajczyk, B.S. The interleukin-1β gene is transcribed from a poised promoter architecture in monocytes. J Biol Chem 281, 9227-37 (2006). 91. Jung, D.H. et al. Diosgenin inhibits macrophage-derived inflammatory mediators through downregulation of CK2, JNK, NF-κB and AP-1 activation. Int Immunopharmacol 10, 1047-54 (2010). 92. Barnes, P.J. & Karin, M. Nuclear Factor-κB -- A Pivotal Transcription Factor in Chronic Inflammatory Diseases. N Engl J Med 336, 1066-1071 (1997). 93. Auron, P.E. et al. Nucleotide sequence of human monocyte interleukin 1 precursor cDNA. Proc Natl Acad Sci U S A 81, 7907-11 (1984). 94. Lomedico, P.T. et al. Cloning and expression of murine interleukin-1 cDNA in Escherichia coli. Nature 312, 458-62 (1984). 95. March, C.J. et al. Cloning, sequence and expression of two distinct human interleukin-1 complementary DNAs. Nature 315, 641-7 (1985). 96. Telford, J.L. et al. The murine interleukin 1 β gene: Structure and evolution. Nucleic Acids Res 14, 9955-9955 (1986). 97. Fenton, M.J. et al. Transcriptional regulation of the human prointerleukin 1 β gene. J Immunol 138, 3972-9 (1987). 98. Fenton, M.J., Vermeulen, M.W., Clark, B.D., Webb, A.C. & Auron, P.E. Human pro-IL-1β gene expression in monocytic cells is regulated by two distinct pathways. J Immunol 140, 2267-73 (1988). 99. Bensi, G. et al. An inducible enhancer controls the expression of the human interleukin 1 β gene. Cell Growth Differ 1, 491-7 (1990). 100. Shirakawa, F. et al. The human prointerleukin 1 β gene requires DNA sequences both proximal and distal to the transcription start site for tissue- specific induction. Mol Cell Biol 13, 1332-44 (1993). 101. Tsukada, J., Saito, K., Waterman, W.R., Webb, A.C. & Auron, P.E. Transcription factors NF-IL6 and CREB recognize a common essential site in the human prointerleukin 1 β gene. Mol Cell Biol 14, 7285-97 (1994). 102. Tsukada, J., Waterman, W.R., Koyama, Y., Webb, a.C. & Auron, P.E. A novel STAT-like factor mediates lipopolysaccharide, interleukin 1 (IL-1), and IL-6 signaling and recognizes a gamma interferon activation site-like element in the IL1B gene. Molecular and cellular biology 16, 2183-94 (1996).

35 103. Unlu, S. et al. Phosphorylation of IRF8 in a pre-associated complex with Spi- 1/PU.1 and non-phosphorylated Stat1 is critical for LPS induction of the IL1B gene. Mol Immunol 44, 3364-79 (2007). 104. Marecki, S., Riendeau, C.J., Liang, M.D. & Fenton, M.J. PU.1 and multiple IFN regulatory factor proteins synergize to mediate transcriptional activation of the human IL-1β gene. J Immunol 166, 6829-38 (2001). 105. Hunninghake, G.W. et al. The functional importance of a cap site-proximal region of the human prointerleukin 1 β gene is defined by viral protein trans- activation. Mol Cell Biol 12, 3439-48 (1992). 106. Kominato, Y., Galson, D., Waterman, W.R., Webb, A.C. & Auron, P.E. Monocyte expression of the human prointerleukin 1 β gene ( IL1B ) is dependent on promoter sequences which bind the hematopoietic transcription factor Spi-1/PU.1. Mol Cell Biol 15, 58-68 (1995). 107. Zhang, Y. & Rom, W.N. Regulation of the interleukin-1 beta (IL-1β) gene by mycobacterial components and lipopolysaccharide is mediated by two nuclear factor-IL6 motifs. Mol Cell Biol 13, 3831-7 (1993). 108. Buras, J.A., Monks, B.G. & Fenton, M.J. The NF-βA-binding element, not an overlapping NF-IL-6-binding element, is required for maximal IL-1β gene expression. J Immunol 152, 4444-54 (1994). 109. Buras, J.A., Reenstra, W.R. & Fenton, M.J. NF βA, a factor required for maximal interleukin-1β gene expression is identical to the ets family member PU.1. Evidence for structural alteration following LPS activation. Mol Immunol 32, 541-54 (1995). 110. Yang, Z., Wara-Aswapati, N., Chen, C., Tsukada, J. & Auron, P.E. NF-IL6 (C/EBP β) vigorously activates il1b gene expression via a Spi-1 (PU.1) protein-protein tether. J Biol Chem 275, 21272-7 (2000). 111. Baldassare, J.J., Bi, Y. & Bellone, C.J. The role of p38 mitogen-activated protein kinase in IL-1β transcription. J Immunol 162, 5367-73 (1999). 112. Xie, Y., Chen, C., Stevenson, M.A., Auron, P.E. & Calderwood, S.K. Heat shock factor 1 represses transcription of the IL-1β gene through physical interaction with the nuclear factor of interleukin 6. J Biol Chem 277, 11802- 10 (2002). 113. Cooper, Z.A. et al. Febrile-range temperature modifies cytokine gene expression in LPS-stimulated macrophages by differentially modifying NF- κB recruitment to cytokine gene promoters. Am J Physiol Cell Physiol 298, C171-81 (2010). 114. Mouri, F. et al. Intracellular HMGB1 transactivates the human IL1B gene promoter through association with an Ets transcription factor PU.1. Eur J Haematol 80, 10-9 (2008). 115. Hiscott, J. et al. Characterization of a functional NF-κB site in the human interleukin 1 beta promoter: evidence for a positive autoregulatory loop. Mol Cell Biol 13, 6231-40 (1993). 116. Monks, B.G., Martell, B.A., Buras, J.A. & Fenton, M.J. An upstream protein interacts with a distinct protein that binds to the cap site of the human interleukin 1 beta gene. Mol Immunol 31, 139-51 (1994).

36 117. Scheibel, M. et al. I κBβ is an essential co-activator for LPS-induced IL-1β transcription in vivo. J Exp Med 207, 2621-30 (2010). 118. Ghosh, C.C. et al. Gene-specific repression of proinflammatory cytokines in stimulated human macrophages by nuclear I κBα. J Immunol 185, 3685-93 (2010). 119. Joo, M. et al. Lipopolysaccharide-dependent interaction between PU.1 and c- Jun determines production of lipocalin-type prostaglandin D synthase and prostaglandin D2 in macrophages. Am J Physiol Lung Cell Mol Physiol 296, L771-9 (2009). 120. Su, D. et al. FoxO1 links insulin resistance to proinflammatory cytokine IL- 1β production in macrophages. Diabetes 58, 2624-33 (2009). 121. Grondin, B. et al. c-Jun homodimers can function as a context-specific coactivator. Mol Cell Biol 27, 2919-33 (2007). 122. Ravasi, T. et al. An atlas of combinatorial transcriptional regulation in mouse and man. Cell 140, 744-52 (2010). 123. Ghisletti, S. et al. Identification and characterization of enhancers controlling the inflammatory gene expression program in macrophages. Immunity 32, 317-28 (2010). 124. Nilsson, R. et al. Transcriptional network dynamics in macrophage activation. Genomics 88, 133-42 (2006). 125. Amit, I. et al. Unbiased reconstruction of a mammalian transcriptional network mediating pathogen responses. Science 326, 257-63 (2009). 126. Weintz, G. et al. The phosphoproteome of toll-like receptor-activated macrophages. Mol Syst Biol 6 (2010). 127. Perkins, N.D. Integrating cell-signalling pathways with NF-κB and I κK function. Nat Rev Mol Cell Biol 8, 49-62 (2007). 128. Yoza, B.K. & McCall, C.E. Facultative heterochromatin formation at the IL- 1 beta promoter in LPS tolerance and sepsis. Cytokine 53, 145-52 (2011). 129. Wessels, I., Fleischer, D., Rink, L. & Uciechowski, P. Changes in chromatin structure and methylation of the human interleukin-1β gene during monopoiesis. Immunology 130, 410-7 (2010). 130. Fisher, J. & Henzinger, T.A. Executable Cell Biology. Nat Biotechnol 25, 1239 - 1249 (2008). 131. Mendes, P. et al. Computational modeling of biochemical networks using COPASI. Methods Mol Biol 500, 17-59 (2009). 132. Kell, D.B. & Oliver, S.G. Here is the evidence, now what is the hypothesis? The complementary roles of inductive and hypothesis-driven science in the post-genomic era. Bioessays 26, 99-105 (2004). 133. Tunstall-Pedoe, H. Preventing chronic diseases : a vital investment. Int J Epidemiol 35 (2006). 134. Reardon, S. A world of chronic disease. Science 333, 558-9 (2011). 135. PubMed. US National Library of Medicine, NIH. http://www.ncbi.nlm.nih.gov/pubmed/

Figure legends

Figure 1. A network map of lipopolysaccharide (LPS) directed signaling in macrophages and monocytes of human and murine origin. LPS directed signals are received via the TLR4-CD14-MD-2 complex (top, left-hand side of diagram) before ejection of CD14 and dimerisation of the TLR4-MD-2 complex. Further signal transduction proceeds through a series of TIRAP and IRAK proteins inter alia (top, middle, through to mid section). Subsequently, JNK, MAPK or NIK, IKK proteins are then capable of phosphorylating transcription factors (TFs) or I κB proteins respectively enabling direct interaction of TFs or I κB with IL-1β regulatory DNA. The thicker beige box bounding the diagram represents the cell membrane and contains the cytoplasmic signaling reactions. Similarly, the thinner beige box represents the nuclear membrane and details the interactions of TFs and I κB proteins with the IL-1β regulatory DNA therein.

Discussion

5.1 Introduction

The major findings of this thesis were that machine learning and design of experimental methods coupled to wet experimental techniques in an iterative loop revealed that a p38 α MAPK inhibitor in conjunction with either an iron chelator or IKK inhibitor could synergistically inhibit LPS-dependent IL-1β expression in macrophages. Levels of inhibition were achieved at concentrations below those that any compound at an identical concentration alone could achieve. Additionally, construction of a TLR4 signalling map of IL-1β expression has revealed the modular complexity of its transcriptional activation. This general discussion aims to summarise the research context in a way that has not been explored previously in either of the chapters presented above.

5.2 Efficient discovery of anti-inflammatory drug combinations using evolutionary computing

The discovery that a p38 α MAPK inhibitor, the bicyclic pyridinyl-imidazole - SB203580 (Cuenda et al., 1995), was the common component between both paired combinations causing synergistic inhibition of IL-1β expression is confirmatory of the importance of p38 α MAPK in driving expression of this cytokine rather than insightful. When p38 α kinase (MAPK14) was found to be sensitive to this class of compound (Lee et al., 1994) it was labelled as a cytokine suppressive anti-inflammatory drug binding protein (CSBP). Similarly, the synergistic inhibition of IL-1β expression via SB203580 and the IKK inhibitor, BMS-345541 (Burke et al., 2003)combination was consistent with reports that these two signalling networks are necessary for IL-1β expression in macrophages (Baldassare et al., 1999, Basak et al., 2005, Hsu and Wen, 2002, Scheibel et al., 2010). However, discovery of the iron chelator; SIH (Horackova et al., 2000) as a highly ranked reagent in the post-hoc analysis of the EA directed search of reagent combinations was novel. The disproportionation (4.1) of superoxide by superoxide dimutase to hydrogen peroxide and the subsequent production of hydroxyl radicals ( ñOH) by the Fenton reaction (4.2) (Wardman and Candeias, 1996) may tie the labile ferrous (Fe 2+ ) iron pool to effects on the NF-κB signalling network (Droge, 2002, Kell, 2009).

Disproportionation: + - H + O 2 → HO 2·

- + HO 2· + O 2 + H → H 2O2 + O 2 4.1

42 Fenton reaction: 2+ 3+ - Fe + H 2O2 → Fe + OH + OH · 4.2

Similarly, the regeneration of ferrous iron ( Fe 2+ ) from its corresponding ferric form by antioxidants (e.g. ascorbate) thus creates a pool of pro-oxidant capable of catalysing the further production of the hydroxyl radical that has a deleterious effect on cell viability.

The decision to not include variable concentration of each component as a search objective in the decision space of the EA was appropriate (see Chapter 2). Had just three concentrations been included (the minimum number required to define an IC 50 measurement) of each reagent then the search space would have increased to 2 99 (6.3 x 10 29 ) possible combinations. Retaining the same population size would have slowed the search, required more iterations to reach convergence in any given objective and still, as a stochastic search strategy, not delineated a specific set of reagent concentrations that we could confidently suggest as optimal in inhibiting IL-1β expression. Rather, the adaptive dose matrix search protocol, enumerated the activity of sets of paired reagents over a range of concentrations lower than that used in the stochastic algorithm (and hence theoretically increased the pharmacological specificity of the component reagents). This ‘bracketing’ of concentration ensured, overall location of an optimal combination with respect to both mode of action and concentration. Others (Wong et al., 2008, Sun et al., 2009) have employed the ‘Gur-game’ stochastic search algorithm and varied concentration but with a limited set of component reagents (i.e. five). However, positing that reagent concentration drives target interaction specificity, then activation or inhibition of ‘off-target’ processes by component reagents may have shielded the true extent of the combinatorial effect.

The complete scan of all possible combinations in the search space of our assay (2 33 or 8.6 billion combinations), even with ultra-high-throughput screening (uHTS) techniques that can screen in excess of 100,000 compounds a day (Wunder et al., 2008), would likely take over 235 years. The use of machine learning in drug discovery has recently been reviewed (Murphy, 2011) and more pertinently to combinatorial strategies (Feala et al., 2010). This latter group has demonstrated previously the use of 'top down' and 'bottom up' search algorithms for determining optimal drug combinations in a D.melanogaster model of decline in cardiovascular function and exercise capacity with age (Calzolari et al., 2008). The authors located 3 out of 5 best drug combinations performing approximately a third of the tests required in a fully factorial search (i.e. 24 - 27 vs. 81).

43 The application of EA directed approaches to finding optimal drug combinations are superior with respect to the number of objectives that can be specified relative to the stack sequential and top down methods used by Calzolari et al (Calzolari et al., 2008) on their data tree. Similarly, progress in the application of the tandem approaches of particle swarm optimisation to fit phosphoproteomic data to an ODE model of IGF-1 signaling in

MDA-MB231 cells (Iadevaia et al., 2010), and IC 50 data from 60 tumourigenic cell lines to 45,344 compounds derived from the National Cancer Institute, were fitted to minimal hitting set models via use of a simulated annealing algorithm (Vazquez, 2009).

Hsueh et al., (Hsueh et al., 2009)examined whether increasing the number of inputs or stimuli (LPS, IFN-β, TGF-β, IL-6, isoproterenol and a non-hydrolysable cAMP analogue) in RAW264.7 cells would elicit unique and iteratively complex responses that scaled with input number. The authors demonstrated that most higher-order (>3) combinations produced less than additive responses in their individual cytokine outputs relative to the synergy observed for pair-wise combinations. This is concordant with the results of the study presented here, in which the triple combination of p38 α MAPK inhibitor, iron chelator and IKK inhibitor did not exhibit a synergistic inhibition of IL-1β expression. The authors interpret these observations as evidence for the encoding of multiple-signals along ‘sparse and discrete’ routes and the ‘limitation of interaction complexity’. However, these conclusions are not supported by the data, where we might expect maintenance of a response type (i.e. synergy) if the effects were mediated via few types of signaling complex. An alternative interpretation of the data might posit that additional stimuli activate either inhibitory networks or processes that compete for a common signaling molecule. In the sparse instances where synergistic responses were observed in higher order combinations, the authors reflect this maybe representative of important phenotypes. Similarly, a report using a multiplexed fluorescent bead immunoassay method has shown the synergistic interactions of TLR8 and TLR3 or TLR4 activating NF-κB, IRF, MAPK and PI-3K signaling networks in human macrophages (Mäkelä et al., 2009).

These observations in conjunction with the biological literature allow a more general question to be asked. Cellular responses are multi-variate outputs that drive phenotype. Phenotype is likely constrained by the epigenetic landscape within the cell as described by Waddington in 1957 where cellular decisions or phenotypes are arrived at by occupation of distinct and specific permitted developmental trajectories (as reviewed by Goldberg et al , (Goldberg et al., 2007)). More recently this theory has been described in the quantitative terms of networks of high-dimensional attractors that drive gene and protein expression 44 (Huang, 2006). The choreographed expression of signaling proteins may result in them coalescing into distinct oligomeric protein- (Gibson, 2009) or cellular complexes (e.g. the inflammasome), ensuring that a signal is communicated efficiently. In metabolic networks, the concept of ‘metabolic channeling’(Mendes et al., 1992), where a substrate is passed directly between active sites of enzymes, undergoing stepwise transformations is analogous to the sequence of reversible covalent modifications (e.g. phosphorylation, ubiquitination) that are transduced in signaling. Dynamic construction and disassembly of discrete signaling complexes would reduce uncertainty, still allow for modularity (i.e. allow common components to be used in a variety of complexes), enhance the robustness of signal transduction (Aderem, 2005), explain the degeneracy of outputs and hence why only small sets of phenotype are realised macroscopically for low ( ≤2) numbers of stimuli. The question then becomes ‘can discrete models of low input number represent phenotype adequately?’ A preliminary examination of this area yielded data suggesting this to be the case (Saez-Rodriguez et al., 2009).

This study described in chapter 2 would have been enhanced by the introduction of mass- spectrometric techniques for the monitoring of multiple protein endpoints (e.g TNF-α, IL-6 inter alia ) via measurement of the mass/charge ratio of these. The evolution of the selected reaction monitoring technique now makes this a possibility (Lange et al., 2008) with the potential to measure up to 200 different proteins. Measurement of multiple protein outputs would have enabled determination of the selectivity and specificity of combination mechanism of action. Although the EA can embrace this number of objectives (i.e. variables (proteins) that change after stimulus), the time and complexity required to run the algorithm is scaled approximately to an exponential relationship (formulated in big ‘O’ notation: O(dn (3) ) where d is the number of objectives and n is the population size). Thus, the time taken to run this problem could obviate its use dependent on machine constraints (e.g. cache size, number of processors). A solution might be found partly via the application of high performance computing . Also, using an expansive range of experimental methods available to the pharmacologist, elucidating the mechanism of action of these combinatorial inhibitors will present a valuable addition to the literature.

45 5.3 A network map of lipopolysaccharide dependent IL-1βββ expression in macrophages and monocytes

The motivation for construction of a TLR4 signalling map impacting IL-1β transcription was to better understand the underlying reactions and feedbacks and provide an essential framework that will allow translation into dynamic models. Previous and more extensive mapping of TLR networks were notably absent of any mapping of transcriptional effects. In the construction of such a map or network there are likely sources of error, namely; factual inaccuracies within papers and errors of interpretation (Oda and Kitano, 2006). We attempted to limit factual inaccuracies by only utilising publications where there were demonstrable dynamic roles for transcription factors (TFs). The likelihood of having obviated other TFs is high, as an in silico scan of the promoter region of the IL-1β gene has revealed sites for additional TFs (Hashimoto et al., 2009) that were not included here owing to an absence of dynamic data.

Advantages of the précis of literature data in a network map form are numerous. Data visualization (Tufte, 2001), allows an immediate appreciation of the relationship (or its absence) between one chemical species and another, that can’t be conveyed in the written word. Similarly, they can condense (and abbreviate) a literature from the past 30 years with little loss of information (e.g. expression of complex relationships, spatial locations / transitions) except in one important respect - time. A limiting factor in presenting a network map is that it necessarily attempts to compress four dimensional (space (x, y, z), time) data via a two dimensional representation (space (x, y)). Although different, biochemical network maps are analogous to circuit diagrams and process flow diagrams in electrical and chemical engineering respectively. However, in systems biology elucidating a network map is a necessary and very important precursor to the further development of a dynamic model, whether as a discrete or continuous representation (Janes and Yaffe, 2006, Fisher and Henzinger, 2008, Kholodenko, 2006).

5.4 Conclusion

This thesis has attempted to define signaling components responsible for TLR4 mediated IL-1β expression in macrophages and reconcile the complementary approaches of an iterative machine learning - ‘wet’ experimentation with the construction of a human / murine structural model from literature data. Acquisition of the EA-directed and adaptive dose matrix search data confirmed literature reports on the importance of p38 α MAPK in IL-1β expression and informed the construction of a structural model. 46

5.5 Back to the future

Reflection on the philosophies and methodologies employed within systems biology (Kell, 2006b) leads to the conclusion that their ongoing evolution conforms to a ‘back to the future’ trajectory. AV Hill (Hill, 1910). RA Fisher and Hodgkin and Huxley’s (Hodgkin and Huxley, 1952) work are all prescient examples of the great impact mathematical and statistical thinking has had in the biological sciences. A common theme emerging here is that new experimental techniques and the application of mathematical tools have always been developed to make innovative discoveries in biology (e.g. biochemistry, genetics, physiology, pharmacology, ecology). The arrival of ‘omics technologies in conjunction with improved data processing capabilities provided by computational platforms and programming languages is allowing the statistical modeling of datasets and the subsequent development and parameterization of in silico models. This is just the latest iteration of a longer chain of technology development, clearly, with statistics as a hub science. Quoting from (Wilkinson, 2009):

‘Statistics is the science concerned with linking models to data [and vice versa , BGS], and as such it is absolutely pivotal to the success of the systems biology vision.’

5.6 Future work

Addressing important problems and posing important questions (Rothwell, 2002, Alon, 2009) that are often derived from the humanities (we are often most interested in the propagation and welfare of our own species), is a fundamental first step within the arena of scientific discovery. The tools and technologies described here within an ever burgeoning literature (Ananiadou et al., 2006) have the capacity to answer these questions with greater efficiency, with greater breadth (i.e. across species) and detail (increased resolution of temporal and spatial measurements) than has been previously examined. Cooperative alliances, consolidation of data, its interpretation via these alliances and translation of these findings for publication and hence a reduction in the literature will be of benefit to all. The Royal Academy of Engineering and Academy of Medical Sciences published a joint report outlining the importance of integrating the principles and methods of engineering into biology (Dollery et al., 2007). Recently, a white paper from a cross-faculty panel at the Massachusetts Institute of Technology has renewed calls for greater efforts to support interdisciplinarity (Sharp et al., 2011).

47 5.6.1 Mixed messages; drug efficacy

To conclude, the most pressing global healthcare problem is that of non-communicable disease (NCD) (e.g. atherosclerosis, diabetes, cancer) which affects a large proportion of populations in both the developing and developed world (Tunstall-Pedoe, 2006),(Reardon, 2011). A manuscript authored by a large and internationally diverse panel of scientists has proposed a systems medicine framework to combat the prevalence of NCDs (Bousquet et al., 2011). The discovery and development of medicines is still seen as being critical for treating NCDs. However, given the ethical concerns, expense and stringent regulations governing the use of animal models to discover remedies, the triage of potential molecules is conducted in in vitro cellular models. Despite this, an understanding of drug efficacy at the cellular level still eludes us. Signal transduction and drug efficacy are the result of the complex dynamics of the cell. A better understanding of cells as complex systems would result in more effective treatments and decreased toxicity, both of which would save the lives of many individuals worldwide, especially in developing nations.

Macrophage cells would form the biological model of choice in which to study the efficacy concept. Their isolation can be achieved without disrupting any syncytial connections and small molecule control of their internal environment can be performed. The evolved function of a macrophage is broadly two-fold; to phagocytose pathogens and cellular debris (innate immune function) and also to provide a critical link to the adaptive arm of the immune system, via secretion of signaling molecules to activate lymphocyte populations. This latter signaling function provides a convenient biological model for understanding the efficacy concept.

Two successive tracks of experimentation could be advanced. Utilising a MOEA we could determine combinations of stimuli and associated application regimes (concentration threshold, frequency and duration of application) and how these scale with the output (i.e. temporal profile of concentration, multiplicity). This can be achieved by defining temporal system response indicators (tSRIs) (van der Greef & McBurney 2005) that are parallel, multi-variate measures of a response estimated as a ratio at a particular time. Collectively, these tSRI’s could then be integrated into a system response profile (SRP) that could be shaped pharmacologically using the same MOEA method outlined in this dissertation (Publication number 01). However, concentrations of secreted signaling molecules (e.g. cytokines) that are expressed via transcriptional mechanisms will vary over minutes to hours. Additionally, monitoring intracellular Ca 2+ mobilization allows a comparison

48 between relatively fast and slow, intracellular versus extracellular (secreted) and non- transcriptional versus transcriptional signaling events respectively. Further, in defining what is a ‘desirable’ SRP, in vivo experiments of acutely inflamed mice (i.e. via intra-peritoneal administration of LPS and / or combinations of stimuli) would allow sampling of endogenous macrophages and their inflammatory outputs with varying time to gauge their tSRI’s and an organismal SRP of inflammation.

The recent announcement by Harvard Medical School of the launch of an initiative in systems pharmacology and corresponding white paper (Sorger et al. 2011) on this subject has reinvigorated the search for basic and generalised principles governing signal transduction. This proposal both strategically and tactically addresses the long-standing problem of efficacy in classic analytical pharmacology using novel systems methods. Similarly, non-communicable disease (Reardon 2011), the current and continuing dearth of prospective new drug candidates (the bulk of which fail in Phase II and III clinical trial stages via their lack of efficacy (Thomson-Reuters 2011) and the long-standing resolution of the efficacy concept in pharmacology are just three problems whose resolution could be realised via undertaking the work set out here.

Bibliography

ADEREM, A. (2005) Systems biology: its practice and challenges. Cell, 121, 511-3.

ALDRIDGE, B. B., BURKE, J. M., LAUFFENBURGER, D. A. & SORGER, P. K. (2006) Physicochemical modelling of cell signalling pathways. Nat Cell Biol, 8, 1195-203.

ALLMENDINGER, R. & KNOWLES, J. (2009) Analysis of Several Evolutionary Algorithms on the Noisy Three-Objective Chemical Mixture Optimization Problem. Technical report MLO-12009, 1 - 9.

ALON, U. (2009) How To Choose a Good Scientific Problem. Molecular Cell, 35, 726- 728.

ANANIADOU, S., KELL, D. B. & TSUJII, J. (2006) Text mining and its potential applications in systems biology. Trends Biotechnol, 24, 571-9.

ASEHNOUNE, K., STRASSHEIM, D., MITRA, S., KIM, J. Y. & ABRAHAM, E. (2004) Involvement of reactive oxygen species in Toll-like receptor 4-dependent activation of NF-kappa B. J Immunol, 172, 2522-9.

AUFFRAY, C., SIEWEKE, M. H. & GEISSMANN, F. (2009) Blood monocytes: development, heterogeneity, and relationship with dendritic cells. Annu Rev Immunol, 27, 669-92.

BAJORATH, J. (2002) Integration of virtual and high-throughput screening. Nat Rev Drug Discov, 1, 882-894.

BALDASSARE, J. J., BI, Y. & BELLONE, C. J. (1999) The role of p38 mitogen-activated protein kinase in IL-1β transcription. J Immunol, 162, 5367-73.

BARABASI, A. L. & OLTVAI, Z. N. (2004) Network biology: understanding the cell's functional organization. Nat Rev Genet, 5, 101-13.

BARTON, G. M. (2008) A calculated response: control of inflammation by the innate immune system. J Clin Invest, 118, 413-20.

BASAK, C., PATHAK, S. K., BHATTACHARYYA, A., MANDAL, D., PATHAK, S. & KUNDU, M. (2005) NF-kappaB- and C/EBPbeta-driven interleukin-1beta gene expression and PAK1-mediated caspase-1 activation play essential roles in interleukin-1beta release from Helicobacter pylori lipopolysaccharide-stimulated macrophages. J Biol Chem, 280, 4279-88.

BEARD, D. A. & QIAN, H. (2008) Chemical biophysics: quantitative analysis of cellular systems , Cambridge University Press.

BIN, L.-H., XU, L.-G. & SHU, H.-B. (2003) TIRP, a Novel Toll/Interleukin-1 receptor (TIR) Domain-containing Adapter Protein Involved in TIR Signaling. J Biol Chem, 278, 24526-24532.

BOUSQUET, J., ANTO, J. M., STERK, P. J., ADCOCK, I. M., CHUNG, K. F., ROCA, J., AGUSTI, A., BRIGHTLING, C., CAMBON-THOMSEN, A., CESARIO, A., ABDELHAK, S., ANTONARAKIS, S. E., AVIGNON, A., BALLABIO, A., 50 BARALDI, E., BARANOV, A., BIEBER, T., BOCKAERT, J., BRAHMACHARI, S., BRAMBILLA, C., BRINGER, J., DAUZAT, M., ERNBERG, I., FABBRI, L., FROGUEL, P., GALAS, D., GOJOBORI, T., HUNTER, P., JORGENSEN, C., KAUFFMANN, F., KOURILSKY, P., KOWALSKI, M. L., LANCET, D., PEN, C. L., MALLET, J., MAYOSI, B., MERCIER, J., METSPALU, A., NADEAU, J. H., NINOT, G., NOBLE, D., OZTÜRK, M., PALKONEN, S., PRÉFAUT, C., RABE, K., RENARD, E., ROBERTS, R. G., SAMOLINSKI, B., SCHÜNEMANN, H. J., SIMON, H.-U., SOARES, M. B., SUPERTI-FURGA, G., TEGNER, J., VERJOVSKI-ALMEIDA, S., WELLSTEAD, P., WOLKENHAUER, O., WOUTERS, E., BALLING, R., BROOKES, A. J., CHARRON, D., PISON, C., CHEN, Z., HOOD, L. & AUFFRAY, C. (2011) Systems medicine and integrated care to combat chronic noncommunicable diseases. Genome medicine, 3, 43-43.

BRENNER, S. (2010) Sequences and consequences. Philos Trans R Soc Lond B Biol Sci, 365, 207-12.

BRYANT, C. E., SPRING, D. R., GANGLOFF, M. & GAY, N. J. (2010) The molecular basis of the host response to lipopolysaccharide. Nat Rev Microbiol, 8, 8-14.

BUCCI, M., GOODMAN, C. & SHEPPARD, T. L. (2010) A decade of chemical biology. Nat Chem Biol, 6, 847-854.

BURKE, J. R., PATTOLI, M. A., GREGOR, K. R., BRASSIL, P. J., MACMASTER, J. F., MCINTYRE, K. W., YANG, X., IOTZOVA, V. S., CLARKE, W., STRNAD, J., QIU, Y. & ZUSI, F. C. (2003) BMS-345541 is a highly selective inhibitor of I kappa B kinase that binds at an allosteric site of the enzyme and blocks NF-kappa B-dependent transcription in mice. J Biol Chem, 278, 1450-6.

CALZOLARI, D., BRUSCHI, S., COQUIN, L., SCHOFIELD, J., FEALA, J. D., REED, J. C., MCCULLOCH, A. D. & PATERNOSTRO, G. (2008) Search algorithms as a framework for the optimization of drug combinations. PLoS Comput Biol, 4, e1000249.

CHEN, G. Y. & NUNEZ, G. (2010) Sterile inflammation: sensing and reacting to damage. Nat Rev Immunol, 10, 826-37.

CUENDA, A., ROUSE, J., DOZA, Y. N., MEIER, R., COHEN, P., GALLAGHER, T. F., YOUNG, P. R. & LEE, J. C. (1995) SB 203580 is a specific inhibitor of a MAP kinase homologue which is stimulated by cellular stresses and interleukin-1. FEBS letters, 364, 229-33.

DADA, J. O. & MENDES, P. (2011) Multi-scale modelling and simulation in systems biology. Integr Biol (Camb), 3, 86 - 96.

DAVIS, C. N., MANN, E., BEHRENS, M. M., GAIDAROVA, S., REBEK, M., REBEK, J., JR. & BARTFAI, T. (2006) MyD88-dependent and -independent signaling by IL-1 in neurons probed by bifunctional Toll/IL-1 receptor domain/BB-loop mimetics. Proc Natl Acad Sci U S A, 103, 2953-8.

DINARELLO, C. A. (1996) Biologic basis for interleukin-1 in disease. Blood, 87, 2095- 147.

DINARELLO, C. A. (2010) IL-1: Discoveries, controversies and future directions. Eur J Immunol, 40, 599-606.

51 DINARELLO, C. A. (2011) Interleukin-1 in the pathogenesis and treatment of inflammatory diseases. Blood, 117, 3720-3732.

DOLLERY, C., KITNEY, R., CHALLIS, R., DELPY, D. & EDWARDS, D. (2007) Systems biology: A vision for engineering and medicine. Academy of Medical Sciences and The Royal Academy of Engineering.

DROGE, W. (2002) Free radicals in the physiological control of cell function. Physiol Rev, 82, 47-95.

FEALA, J. D., CORTES, J., DUXBURY, P. M., PIERMAROCCHI, C., MCCULLOCH, A. D. & PATERNOSTRO, G. (2010) Systems approaches and algorithms for discovery of combinatorial therapies. Systems biology and medicine, 2, 181-93.

FENG, Y., MITCHISON, T. J., BENDER, A., YOUNG, D. W. & TALLARICO, J. A. (2009) Multi-parameter phenotypic profiling: using cellular effects to characterize small-molecule compounds. Nat Rev Drug Discov, 8, 567-78.

FISHER, J. & HENZINGER, T. A. (2008) Executable Cell Biology. Nat Biotechnol, 25, 1239 - 1249.

FITZGERALD, J. B., SCHOEBERL, B., NIELSEN, U. B. & SORGER, P. K. (2006) Systems biology and combination therapy in the quest for clinical efficacy. Nat Chem Biol, 2, 458-466.

GADDUM, J. H. (1954) The Science of Pharmacology. Nature, 173, 14-15.

GERY, I. & WAKSMAN, B. H. (1972) Potentiation of the T-lymphocyte response to mitogens. II. The cellular source of potentiating mediator(s). J Exp Med, 136, 143- 55.

GIBSON, T. J. (2009) Cell regulation: determined to signal discrete cooperation. TiBS, 34, 471-482.

GOLDBERG, A. D., ALLIS, C. D. & BERNSTEIN, E. (2007) Epigenetics: a landscape takes shape. Cell, 128, 635-8.

GORDON, S. (2003) Alternative activation of macrophages. Nat Rev Immunol, 3, 23-35.

HAEFNER, J. W. (2005a) Chapter 1: Models of Systems. Modeling Biological Systems: Principles and Applications. Springer.

HAEFNER, J. W. (2005b) Chapter 2: The Modeling Process. Modeling Biological Systems: Principles and Applications. 2nd ed., Springer.

HALLIWELL, B., GUTTERIDGE, J.M.C. (1999) Free Radicals in Biology and Medicine , Oxford University Press.

HANDL, J., KELL, D. B. & KNOWLES, J. (2007) Multiobjective optimization in bioinformatics and computational biology. IEEE/ACM Trans Comput Biol Bioinform, 4, 279-92.

HASHIMOTO, K., OREFFO, R. O. C., GIBSON, M. B., GOLDRING, M. B. & ROACH, H. I. (2009) DNA demethylation at specific CpG sites in the IL1B promoter in response to inflammatory cytokines in human articular chondrocytes. Arthritis Rheum, 60, 3303-13. 52 HILL, A. V. (1910) The possible effects of the aggregation of the molecules of haemoglobin on its dissociation curves. J Physiol, 40, iv-vii.

HODGKIN, A. L. & HUXLEY, A. F. (1952) A quantitative description of membrane current and its application to conduction and excitation in nerve. J Physiol, 117, 500-544.

HOOPS, S., SAHLE, S., GAUGES, R., LEE, C., PAHLE, J., SIMUS, N., SINGHAL, M., XU, L., MENDES, P. & KUMMER, U. (2006) COPASI--a COmplex PAthway SImulator. Bioinformatics, 22, 3067-3074.

HORACKOVA, M., PONKA, P. & BYCZKO, Z. (2000) The antioxidant effects of a novel iron chelator salicylaldehyde isonicotinoyl hydrazone in the prevention of H2O2 injury in adult cardiomyocytes. Cardiovas Res, 47, 529-36.

HSU, H. Y. & WEN, M. H. (2002) Lipopolysaccharide-mediated Reactive Oxygen Species and Signal Transduction in the Regulation of Interleukin-1 Gene Expression. J Biol Chem, 277, 22131 - 22139.

HSUEH, R. C., NATARAJAN, M., FRASER, I., POND, B., LIU, J., MUMBY, S., HAN, H., JIANG, L. I., SIMON, M. I., TAUSSIG, R. & STERNWEIS, P. C. (2009) Deciphering Signaling Outcomes from a System of Complex Networks. Science Signaling, 2, ra22 (1 - 13).

HUANG, S. (2006) Multistability and multicellularity: Cell fates as high-dimensional attractors of gene regulatory networks. IN KRIETE, A. & EILS, R. (Eds.) Computational Systems Biology. Elsevier.

IADEVAIA, S., LU, Y., MORALES, F. C., MILLS, G. B. & RAM, P. T. (2010) Identification of Optimal Drug Combinations Targeting Cellular Networks: Integrating Phospho-Proteomics and Computational Network Analysis. Cancer Res, 70, 6704-6714.

IDEKER, T., GALITSKI, T. & HOOD, L. (2001) A new approach to decoding life: systems biology. Annu Rev Genomics Hum Genet, 2, 343-72.

IWASAKI, A. & MEDZHITOV, R. (2010) Regulation of Adaptive Immunity by the Innate Immune System. Science, 327, 291-295.

JANES, K. A. & YAFFE, M. B. (2006) Data-driven modelling of signal-transduction networks. Nat Rev Mol Cell Biol, 7, 820-8.

JUNG, H. W., YOON, C. H., PARK, K. M., HAN, H. S. & PARK, Y. K. (2009) Hexane fraction of Zingiberis Rhizoma Crudus extract inhibits the production of nitric oxide and proinflammatory cytokines in LPS-stimulated BV2 microglial cells via the NF-kappaB pathway. Food Chem Toxicol .

KAUFFMAN, S. A. (1995) At home in the universe: the search for laws of self- organization and complexity , Viking.

KAUFFMAN, S. A. (2008) Reinventing the sacred : a new view of science, reason and religion, New York, Basic Books.

KAWAI, T. & AKIRA, S. (2010) The role of pattern-recognition receptors in innate immunity: update on Toll-like receptors. Nat Immunol, 11, 373-84.

53 KELL, D. (2009) Iron Behaving Badly: Inappropriate Iron Chelation as a Major Contributor to the Aetiology of Vascular and Other Progressive Inflammatory and Degenerative Diseases. BMC Med Genomics, 2, 2.

KELL, D. B. (2006a) Systems biology, metabolic modelling and metabolomics in drug discovery and development. Drug Discov Today, 11, 1085-92.

KELL, D. B. (2006b) Theodor Bucher Lecture. Metabolomics, modelling and machine learning in systems biology - towards an understanding of the languages of cells. Delivered on 3 July 2005 at the 30th FEBS Congress and the 9th IUBMB conference in Budapest. Febs J, 273, 873-94.

KELL, D. B. & OLIVER, S. G. (2004) Here is the evidence, now what is the hypothesis? The complementary roles of inductive and hypothesis-driven science in the post- genomic era. Bioessays, 26, 99-105.

KENNY, E. F. & O'NEILL, L. A. J. (2008) Signalling adaptors used by Toll-like receptors: An update. Cytokine, 43, 342-349.

KHOLODENKO, B. N. (2006) Cell-signalling dynamics in time and space. Nat Rev Mol Cell Biol, 7, 165-176.

KIM, S., TAKAHASHI, H., LIN, W. W., DESCARGUES, P., GRIVENNIKOV, S., KIM, Y., LUO, J. L. & KARIN, M. (2009) Carcinoma-produced factors activate myeloid cells through TLR2 to stimulate metastasis. Nature, 457, 102-6.

KITANO, H. (2004) Biological robustness. Nat Rev Genet, 5, 826-37.

KITANO, H. (2007) Towards a theory of biological robustness. Mol Syst Biol, 3, 137.

KNIGHT, C. G., PLATT, M., ROWE, W., WEDGE, D. C., KHAN, F., DAY, P. J., MCSHEA, A., KNOWLES, J. & KELL, D. B. (2008) Array-based evolution of DNA aptamers allows modelling of an explicit sequence-fitness landscape. Nucleic Acids Res, 37.

KNOWLES, J. (2009) Closed-Loop Evolutionary Multiobjective Optimization. IEEE Comput Intell M 4, 77-91.

KNOWLES, J., CORNE, D. & DEB, K. (2008) Multiobjective problem solving from nature: from concepts to applications , Springer, Heidelberg.

LANGE, V., PICOTTI, P., DOMON, B. & AEBERSOLD, R. (2008) Selected reaction monitoring for quantitative proteomics: a tutorial. Mol Syst Biol, 4.

LEE, J. C., LAYDON, J. T., MCDONNELL, P. C., GALLAGHER, T. F., KUMAR, S., GREEN, D., MCNULTY, D., BLUMENTHAL, M. J., HEYS, J. R. & LANDVATTER, S. W. (1994) A protein kinase involved in the regulation of inflammatory cytokine biosynthesis. Nature, 372, 739-46.

LEHAR, J., STOCKWELL, B. R., GIAEVER, G. & NISLOW, C. (2008) Combination chemical genetics. Nat Chem Biol, 4, 674-81.

LU, Y.-C., YEH, W.-C. & OHASHI, P. S. (2008) LPS/TLR4 signal transduction pathway. Cytokine, 42, 145-51.

54 LUHESHI, N. M., ROTHWELL, N. J. & BROUGH, D. (2009) Dual functionality of interleukin-1 family cytokines: implications for anti-interleukin-1 therapy. Brit J Pharmacol, 157, 1318-29.

MÄKELÄ, S. M., STRENGELL, M., PIETILÄ, T. E., OSTERLUND, P. & JULKUNEN, I. (2009) Multiple signaling pathways contribute to synergistic TLR ligand- dependent cytokine gene expression in human monocyte-derived macrophages and dendritic cells. J Leukocyte Biol, 85, 664-672.

MANTOVANI, A. (2009) Cancer: Inflaming metastasis. Nature, 457, 36-37.

MARCH, C. J., MOSLEY, B., LARSEN, A., CERRETTI, D. P., BRAEDT, G., PRICE, V., GILLIS, S., HENNEY, C. S., KRONHEIM, S. R. & GRABSTEIN, K. (1985) Cloning, sequence and expression of two distinct human interleukin-1 complementary DNAs. Nature, 315, 641-7.

MARIATHASAN, S. & MONACK, D. M. (2007) Inflammasome adaptors and sensors: intracellular regulators of infection and inflammation. Nat Rev Immunol, 7, 31-40.

MARTIN, M.-S., IAIN, D. C. F. & FREDERICK, K. (2009) Multiscale modeling for biologists. Systems Biology and Medicine, 1, 4-14.

MATSUZAWA, A., SAEGUSA, K., NOGUCHI, T., SADAMITSU, C., NISHITOH, H., NAGAI, S., KOYASU, S., MATSUMOTO, K., TAKEDA, K. & ICHIJO, H. (2005) ROS-dependent activation of the TRAF6-ASK1-p38 pathway is selectively required for TLR4-mediated innate immunity. Nat Immunol, 6, 587-92.

MCDERMOTT, M. F. & TSCHOPP, J. (2007) From inflammasomes to fevers, crystals and hypertension: how basic research explains inflammatory diseases. Trends Mol Med, 13, 381-8.

MEDZHITOV, R. (2008) Origin and physiological roles of inflammation. Nature, 454, 428-35.

MEDZHITOV, R. (2010) Inflammation 2010: new adventures of an old flame. Cell, 140, 771-6.

MENDES, A. F., CARAMONA, M. M., CARVALHO, A. P. & LOPES, M. C. (2003) Differential roles of hydrogen peroxide and superoxide in mediating IL-1-induced NF-kappa B activation and iNOS expression in bovine articular chondrocytes. J Cell Biochem, 88, 783-93.

MENDES, P., HOOPS, S., SAHLE, S., GAUGES, R., DADA, J. & KUMMER, U. (2009) Computational modeling of biochemical networks using COPASI. Methods Mol Biol, 500, 17-59.

MENDES, P. & KELL, D. (1998) Non-linear optimization of biochemical pathways: applications to metabolic engineering and parameter estimation. Bioinformatics, 14, 869-83.

MENDES, P., KELL, D. B. & WESTERHOFF, H. V. (1992) Channelling can decrease pool size. Eur J Biochem / FEBS, 204, 257-66.

MESAROVIC, M. D., SREENATH, S. N. & KEENE, J. D. (2004) Search for organising principles: understanding in systems biology. Syst Biol, 1, 19-27.

55 MITCHELL, M. (1996) An Introduction to Genetic Algorithms, Cambridge, Mass. ; London, MIT Press.

MOSSER, D. M. & EDWARDS, J. P. (2008) Exploring the full spectrum of macrophage activation. Nat Rev Immunol, 8, 958-69.

MUHAMMAD, S., BARAKAT, W., STOYANOV, S., MURIKINATI, S., YANG, H., TRACEY, K. J., BENDSZUS, M., ROSSETTI, G., NAWROTH, P. P., BIERHAUS, A. & SCHWANINGER, M. (2008) The HMGB1 receptor RAGE mediates ischemic brain damage. J Neurosci, 28, 12023-31.

MURPHY, K. M., TRAVERS, P. AND WALPORT, M. (2007) Janeway’s Immunobiology , Garland Science.

MURPHY, R. F. (2011) An active role for machine learning in drug development. Nat Chem Biol, 7, 327-330.

NATHAN, C. & DING, A. (2010) Nonresolving inflammation. Cell, 140, 871-82.

O'HAGAN, S., DUNN, W. B., BROWN, M., KNOWLES, J. D. & KELL, D. B. (2005) Closed-loop, multiobjective optimization of analytical instrumentation: gas chromatography/time-of-flight mass spectrometry of the metabolomes of human serum and of yeast fermentations. Anal Chem, 77, 290-303.

O'NEILL, L. A. J. (2006) Targeting signal transduction as a strategy to treat inflammatory diseases. Nat Rev Drug Discov, 5, 549-563.

ODA, K. & KITANO, H. (2006) A comprehensive map of the toll-like receptor signaling network. Mol Syst Biol, 2, 2006 0015.

OGURA, Y., SUTTERWALA, F. S. & FLAVELL, R. A. (2006) The inflammasome: first line of the immune response to cell stress. Cell, 126, 659-62.

PAHLE, J. (2009) Biochemical simulations: stochastic, approximate stochastic and hybrid approaches. Brief Bioinform, 10, 53-64.

PAOLINI, G. V., SHAPLAND, R. H., VAN HOORN, W. P., MASON, J. S. & HOPKINS, A. L. (2006) Global mapping of pharmacological space. Nat Biotechnol, 24, 805- 15.

POLTORAK, A., HE, X., SMIRNOVA, I., LIU, M. Y., VAN HUFFEL, C., DU, X., BIRDWELL, D., ALEJOS, E., SILVA, M., GALANOS, C., FREUDENBERG, M., RICCIARDI-CASTAGNOLI, P., LAYTON, B. & BEUTLER, B. (1998) Defective LPS Signaling in C3H/HeJ and C57BL/10ScCr Mice: Mutations in Tlr4 Gene. Science, 282, 2085-2088.

POPPER, K. R. (2002) The logic of scientific discovery , Routledge.

RAMAN, M., CHEN, W. & COBB, M. H. (2007) Differential regulation and properties of MAPKs. Oncogene, 26, 3100-12.

RANG, H. P., DALE, M.M. AND RITTER, J.M. (1995) Pharmacology , Churchill Livingstone.

REARDON, S. (2011) A world of chronic disease. Science, 333, 558-9.

56 RENZ, H., VON MUTIUS, E., BRANDTZAEG, P., COOKSON, W. O., AUTENRIETH, I. B. & HALLER, D. (2011) Gene-environment interactions in chronic inflammatory disease. Nat Immunol, 12, 273-277.

ROCK, K. L. & KONO, H. (2008) The inflammatory response to cell death. Annu Rev Pathol, 3, 99-126.

ROTHWELL, N. (2002) Who wants to be a scientist?: choosing science as a career , Cambridge University Press.

ROTHWELL, N. J. & LUHESHI, G. N. (2000) Interleukin 1 in the brain: biology, pathology and therapeutic target. Trends Neurosci, 23, 618-25.

RUIZ-MESIA L FAU - RUIZ-MESIA, W., RUIZ-MESIA W FAU - REINA, M., REINA M FAU - MARTINEZ-DIAZ, R., MARTINEZ-DIAZ R FAU - DE INES, C., DE INES C FAU - GUADANO, A., GUADANO A FAU - GONZALEZ-COLOMA, A. & GONZALEZ-COLOMA, A. (2005) Bioactive cinchona alkaloids from Remijia peruviana. J Agric Food Chem, 53, 1921 - 1926.

SAEZ-RODRIGUEZ, J., ALEXOPOULOS, L. G., EPPERLEIN, J., SAMAGA, R., LAUFFENBURGER, D. A., KLAMT, S. & SORGER, P. K. (2009) Discrete logic modelling as a means to link protein signalling networks with functional analysis of mammalian signal transduction. Mol Syst Biol, 5.

SCHEIBEL, M., KLEIN, B., MERKLE, H., SCHULZ, M., FRITSCH, R., GRETEN, F. R., ARKAN, M. C., SCHNEIDER, G. & SCHMID, R. M. (2010) I κBβ is an essential co-activator for LPS-induced IL-1β transcription in vivo. J Exp Med, 207, 2621-30.

SHARP, P. A., COONEY, C. L., KASTNER, M. A., LEES, J., SASISEKHARAN, R., YAFFE, M. B., BHATIA, S. N., JACKS, T. E., LAUFFENBURGER, D. A., LANGER, R., HAMMOND, P. T. & SUR, M. (2011) The Third Revolution : The Convergence of the Life Sciences , Physical Sciences , and Engineering. MIT-AAAS Forum on Convergence. AAAS auditorium, Washington, DC MIT press.

SOEHNLEIN, O. & LINDBOM, L. (2010) Phagocyte partnership during the onset and resolution of inflammation. Nat Rev Immunol, 10, 427-439.

SORGER, P.K., ALLERHEILIGEN, S.R.B., ABERNETHY, D. R., ALTMAN, R. B., BROUWER, K.L.R., CALIFANO, A., DAVID, Z., ARGENIO, D., IYENGAR, R., JUSKO, W.J., LALONDE, R., LAUFFENBURGER, D.A., SHOICHET, B., STEVENS, J.L., SUBRAMANIAM, S., GRAAF, P.V.D. & WARD, R. (2011). Quantitative and Systems Pharmacology in the Post-genomic Era : New Approaches to Discovering Drugs and Understanding Therapeutic Mechanisms An NIH White Paper by the QSP Workshop Group – October , 2011. 1 - 47.

57 SREENATH, S. N., CHO, K. H. & WELLSTEAD, P. (2008) Modelling the dynamics of signalling pathways. Essays Biochem, 45, 1-28.

STEINMAN, L. (2008) Nuanced roles of cytokines in three major human brain disorders. J Clin Invest, 118, 3557-63.

SUN, C.-P., USUI, T., YU, F., AL-SHYOUKH, I., SHAMMA, J., SUN, R. & HO, C.-M. (2009) Integrative systems control approach for reactivating Kaposi's sarcoma- associated herpesvirus (KSHV) with combinatory drugs. Integr Biol, 1, 123-130.

TARANTOLA, A. (2006) Popper, Bayes and the inverse problem. Nat Physics, 2, 492- 494.

THOMSON-REUTERS. (2011). 2011 CMR international pharmaceutical R&D factbook executive summary. 1 - 8.

TOSHCHAKOV, V., JONES, B.W., PERERA, P-Y., THOMAS, K., JOSHUA CODY, M., ZHANG, S., WILLIAMS, B.R.G., MAJOR, J., HAMILTON, T.A., FENTON, M.J. & VOGEL, S.N. (2002). TLR4, but not TLR2, mediates IFN-β-induced STAT1 α/β-dependent gene expression in macrophages. Nat Immunol, 3, 392 - 398.

TUFTE, E. R. (2001) The visual display of quantitative information , Graphics Press.

TUNSTALL-PEDOE, H. (2006) Preventing chronic diseases: a vital investment. Int J Epidemiol, 35.

VAN-DER GREEF, J. & Mc BURNEY, R. (2005). Innovation: Rescuing drug discovery: in vivo systems pathology and systems pharmacology. Nat Rev Drug Discov, 4, 961 - 967.

VAZQUEZ, A. (2009) Optimal drug combinations and minimal hitting sets. BMC Syst Biol, 3, 81.

WARDMAN, P. & CANDEIAS, L. P. (1996) Fenton chemistry: an introduction. Radiat Res, 145, 523-31.

WEISS, U. (2008) Inflammation. Nature, 454, 427.

WELLS, C. A., RAVASI, T. & HUME, D. A. (2005) Inflammation suppressor genes: please switch out all the lights. J Leukocyte Biol, 78, 9-13.

WESTERHOFF, H. V., KOLODKIN, A., CONRADIE, R., WILKINSON, S. J., BRUGGEMAN, F. J., KRAB, K., VAN SCHUPPEN, J. H., HARDIN, H., BAKKER, B. M., MONE, M. J., RYBAKOVA, K. N., EIJKEN, M., VAN LEEUWEN, H. J. & SNOEP, J. L. (2008) Systems biology towards life in silico: mathematics of the control of living cells. J Math Biol .

WESTERHOFF, H. V. & PALSSON, B. O. (2004) The evolution of molecular biology into systems biology. Nat Biotech, 22, 1249-52.

WHITLEY, D. (1994) A Genetic Algorithm Tutorial. Stat Comput, 4, 65-85.

WILKINSON, D. J. (2009) Stochastic modelling for quantitative description of heterogeneous biological systems. Nat Rev Genet, 10, 122-133.

58 WOLKENHAUER, O. (2001) Systems biology: the reincarnation of systems theory applied in biology? Brief Bioinform, 2, 258-270.

WONG, P. K., YU, F., SHAHANGIAN, A., CHENG, G., SUN, R. & HO, C. M. (2008) Closed-loop control of cellular functions using combinatory drugs guided by a stochastic search algorithm. Proc Natl Acad Sci USA, 105, 5105-10.

WUNDER, F., KALTHOF, B., MULLER, T. & HUSER, J. (2008) Functional cell-based assays in microliter volumes for ultra-high throughput screening. Comb. Chem. High Throughput Screening, 11, 495-504.

XU, Y., XUE, Y., WANG, Y., FENG, D., LIN, S. & XU, L. (2009) Multiple-modulation effects of Oridonin on the production of proinflammatory cytokines and neurotrophic factors in LPS-activated microglia. Int Immunopharmacol, 9, 360-5.

YAMAMOTO, M., SATO, S., HEMMI, H., SANJO, H., UEMATSU, S., KAISHO, T., HOSHINO, K., TAKEUCHI, O., KOBAYASHI, M., FUJITA, T., TAKEDA, K. & AKIRA, S. (2002) Essential role for TIRAP in activation of the signalling cascade shared by TLR2 and TLR4. Nature, 420, 324-329.

YAMAMOTO, M., SATO, S., HEMMI, H., UEMATSU, S., HOSHINO, K., KAISHO, T., TAKEUCHI, O., TAKEDA, K. & AKIRA, S. (2003) TRAM is specifically involved in the Toll-like receptor 4-mediated MyD88-independent signaling pathway. Nat Immunol, 4, 1144-1150.

ZAK, D. E. & ADEREM, A. (2009) Systems biology of innate immunity. Immunol Rev, 227, 264-82.

ZITZLER, E. & KÜNZLI, S. (2004) Indicator based selection in multi-objective search. IN YAO, X. (Ed.) Parallel problem solving from nature - PPSN VIII : 8th international conference, Birmingham, UK, September 18-22, 2004 : proceedings. Berlin; [London], Springer.