Investigations of Ligand Interactions with Nuclear Receptors and Implications for Human Health

Miriam Naomi Jacobs BSc (Hons), MSc, MSc, RPHNutr.

Submitted for the degree of Doctorate of Philosophy

School of Biomedical and Molecular Sciences University o f Surrey

August 2003

©Miriam Naomi Jacobs ProQuest Number: 27598754

INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. uest

ProQuest 27598754

Published by ProQuest LLO (2019). Copyright of the Dissertation is held by the Author.

ProQuest LLO. 789 East Eisenhower Parkway P.Q. Box 1346 Ann Arbor, Ml 4 8 1 0 6 - 1346 Acknowledgements

I have many people to thank for giving me the opportunity to spend the last three years conducting research for this PhD thesis.

Prof Dave Lewis, my academic supervisor, believed in my abilities at the outset, initially backed me, and helped pursue and obtain funding for my PhD so that I could change from part-time to full-time research. I must also thank the BBSRC and GSK at the outset for the opportunity they have provided me by their support to my project and provision of a wide variety of training opportunities. I thank Maurice Dickins, my industrial supervisor for facilitating most of my time at

GSK, for really helpful comments and discussion. My grateful thanks to Chris Luscombe, who introduced and guided me with multivariate analytical tools, to Gail Nolan and Amanda

Woodrooffe who supervised the wet work study aspects, to Sandeep Modi and Sami Raza for guidance with computational chemistry study aspects and access to receptor crystal structures, and to Steve Hood who replaced Maurice Dickins as my industrial supervisor during this last year of writing up my PhD, thanks for also giving me helpful feedback when I needed it.

My heartfelt thanks also go to the scientists who have collaborated with me on the different projects that originally grew independently from my PhD thesis, but helped the directions taken, particularly Joe Ferrario of the US EPA and Adrian Covaci, at the University of Antwerp,

Belgium, for their analytical support on the food safety aspects, and Guosheng Chen now at

Health Canada for providing data for a QSAR study. My thanks to Maged Younes for being my

‘unofficial’ mentor throughout my PhD, always ready with constructive advice and encouragement.

Thanks also to Becky Lloyd, Charlotte Ashby, Jen Riley, Sandy Baldwin and Jo Bramhall for the social support (!) whilst at GSK, PK for social support at Surrey and Nicola Roberts who has been both a comrade in arms, and a great help with managing my bibliography. Thanks also to Hugh

Hegarty in reprographics at GSK (William Lea) for his help with printing, and willingness to stay late to get it all done.

My family and friends I thank for their love and support. Special thanks go to my mother, Liselott

Jacobs, who has had to endure the frustrations of not just one, but two daughters intent with their PhDs at the same time, to Wendy and Mick McLeod Gilford, and to Prof Dennis Parke who gave me so much support in the last years of his life. Finally I must thank my children, Joe and

Lani, who started me on this road of exploration in the first place, and have travelled along with me, being my inspiration. I thank them for their tolerance and lack of tolerance (sorry Lani!) for their zest for life, and wish a healthy environment for all our children.

My thanks to everyone who has helped me along this road. With the completion of this thesis, I can finally move on. Abbreviations

XICES 7PCBS Sum of ICES 7 PCB congeners XPCDD/FTEQ Sum of WHO TEQ for individual PCDD and PCDF congeners 'Lnon-ortho PCB TEQ Sum of WHO TEQ for individual non-ortho PCB congeners 'Lortho PCB TEQ Sum of WHO TEQ for individual ortho PCB congeners 'Lnon-ortho PCB TEQ Sum of WHO TEQ for individual non-ortho PCB congeners XWHO TEQ Sum of XPCDD/F TEQ ILnon-ortho PCB TEQ, "Lortho PCB TEQ using WHO assigned TEFs aa Amino Acid AA Arachidonic acid ABC ATP binding cassette ACT pActin ADME Absorption, distribution, metabolism and excretion AhR Aryl hydrocarbon receptor ALB Albumin ALNA a-linolenic acid ANOVA Analysis of variance AP Alkaline phosphatase AR Androgen Receptor Arg Arginine ARNT Aryl Hydrocarbon Receptor Nuclear Transporter protein ARPl Apoliopoprotein regulating protein 1 B-Gal P-Galactosidase bp Base pairs BNP P-naphthaflavone BRG-1 Brahma/SWI2-related Gene 1 Protein BSA Bovine serum albumin BTE Basic transcription element cAMP Cyclic adenosine 3’,5’-monophosphate CAR Constitutively activated / androstane receptor CAT Chloramphenicol acetyltransferase CBP Cyclic adenosine 3 ’ ,5 ’ -monophosphate response element binding protein - binding protein cDNA Complementary deoxyribonucleic acid crrco 6-(4-chlorophenyl)imidazo[2,l-^][l,3]thiazole-5-carbaldhyde O- (3,4-dichlorobenyl)oxime cLogP calculated LogP COT Committee on Toxicity of Chemicals in Food, Consumer Products and the Environment, UK COUP-TFl Chicken ovalbumin upstream promoter transcription factor 1 COX-1 Cyclooxygenase-1 COX-2 Cyclooxygenase-2 CRM Certified reference material CV Cross-validation CYP cytochrome P450 CYP3A Cytochrome P450 3A subfamily DBD DNA-binding domain DBP D-element binding protein DDD 1,1 -dichloro- 2,2-^/5(4-chlorophenyl)-ethane DDE 1,1 -dichloro- 2,2-^/5'(4-dichlorodiphenyl)ethylene DDT 1,1,1 ,trichloro- 2,2-&/5(4-chlorophenyl)ethane DEFRA Department of the Environment, Food and Rural Affairs Dex Dexamethasone DHA Docosahexaenoic acid DMEM Dulbecco’s Modified Eagle Medium DMSO Dimethylsulfoxide EDC Endocrine Disrupting Compound EDTA Ethylenediaminetetraacetic acid ER Estrogen receptor ER6 Everted repeat separated by 6 nucleotides EPA Eicosapentaenoic acid ERE Estrogen Response Element FAO Food and Agriculture Organisation FBS Foetal bovine serum FSA Food Standards Agency FXR Famesoid X Receptor GC Gas chromatograph GC/HRMS Gas chromatography -high resolution mass spectrometry GC/LRMS Gas chromatography -low resolution mass spectrometry GC/MS Gas chromatography -mass spectrometry GH Growth hormone GIT Gastrointestinal Tract Gin Glutamine Glu Glutamate GR Glucocorticoid receptor GRE Glucocorticoid response element GRIPl Glucocorticoid receptor interacting polypeptide 1 GSK GlaxoSmithKline H Human HBS HEPES buffered saline HCB Hexachlorobenzene HCH Hexachlorohexane HEPES 40(2-Hydroxyethyl)-1 -piperazineethanesulfonic acid hERa human oestrogen receptora hERp human oestrogen receptor^ hEST human oestrogen sulphotransferase His Histidine HNFl Hepatocyte nuclear factor 1 HNF3 Hepatocyte nuclear factor 3 HNF4 Hepatocyte nuclear factor 4 HNF5 Hepatocyte nuclear factor 5 HRE Hormone response element hsp90 Heat shock protein 90 ICES International Committee for the Exploration of the Sea IL6 Interleukin-6 iNOS human inducible Nitric Oxide Synthase gene lUPAC International Union of Pure and Applied Chemistry kb Kilobase pairs LA Linoleic acid LED Ligand binding domain LCPUFA Long chain polyunsaturated fatty acids LH Leutinising Hormone LOD Limit of detection LXR Liver X receptor m Mouse MAFF Ministry of Agriculture Fisheries and Food (UK), now the FSA and DEFRA (UK) MDRl Multi-drug resistance-1 MEM Minimal essential medium MFO Microsomal mixed function oxidase system MIF 3D Molecular interaction fields mLogP Measured Log P MLR Multiple linear regression MO Mono-ortho mRNA Messenger ribonucleic acid MRP Multi-drug resistance associated protein MT Metric Ton MW Molecular weight NA Not analysed NcoR Nuclear receptor corepressor ND Not detected NEAA Non essential amino acids NFl Nuclear factor 1 NF kB Nuclear factor kappa B NIH National Institute of Health NO Non-ortho NR Nuclear receptor NSAIDs Nonsteroidal anti-inflammatory drugs Oat rat organic anion transporter 0ATP2 Hepatic organic anion transporting polypeptide 2 OATP-C Hepatic organic anion transporting polypeptide-C OC Organochlorine P450 Cytochrome P450 PB Phénobarbital PBDE Polybrominated Diphenyl Ether (flame retardant) PBS Phosphate buffered saline PC Principal Components PCA Principal components analysis PCB Polychlorinated Biphenyl PCDDs Polychlorinated Dibenzo-p-Dioxins PCDFs Polychlorinated Dibenzofurans PCN Pregnenolone 16a-carbonitrile PCR Polymerase chain reaction PEPCK Phosphoenolpyruvate carboxykinase PGC-la Peroxisme proliferator activated receptory co-activator-la P-gP P-Glycoprotein PLS Partial Least Squares projection to latent structures PLS-DA Partial Least Squares -Discriminant analysis POPs Persistent Organic Pollutants PPAR Peroxisome Proliferator Activator Receptor PPARa human Peroxisome Proliferator Activated Receptor a PPARy human Peroxisome Proliferator Activated Receptor y PR Progesterone Receptor PRESS Predictive Residual Sum of Squares PUPA Polyunsaturated fatty acids PXR Pregnane-X-Receptor QA Quality Assurance QSAR Quantitative Structure-Activity Relationship r Rat RA Retinoic acid RAR Retinoic acid receptor Rif Rifampicin RIP140 Receptor interacting protein 140 RMSEE Root Mean Square Error of Estimation RMSEP Root Mean Square Error of Prediction RSD Residual Standard Deviation RNA Ribonucleic acid RT Reverse transcriptase RXR Q-cw-Retinoic Acid Receptor SDEP Standard Deviation Error of Prediction Ser Serine SF-1 Steroidogenic factor 1 SHR Steroid Hormone Receptors SRC-1 Steroid hormone coactivator-1 SREBP-lc Sterol regulatory element binding protein -Ic TCDD 2,3,7,8-tetrachlorodibenzo-p-dioxin TCPOBOP 1,4-bis[2-{3,5-dichloropyridyloxy)]benzene TDI Tolerable Daily Intake TE Trypsin EDTA TEF Toxic Equivalent Factor TEQ Sum of toxic equivalents (TCDD equivalent) TR Thyroid receptor Tyr Tyrosine TZDs Thiazolidinediones US EPA United States Environmental Protection Agency UV Ultra violet VDR Vitamin D3 Receptor vHNFl Variant hepatocyte nuclear factor 1 VIP Variable Influence on Projection WHO World Health Organisation XREM Xenobiotic responsive enhancer module

Throughout this thesis the lUPAC nomenclature is used to identify the PCBs and PBDES discussed. The full names of the individual congeners can be obtained from the lUPAC web site; www.chem.qmul.ac.uk/IUPAC Table of Contents Page

Acknowledgments 2 List of Abbreviations 4 Table of Contents 8 Summary 16 Publications in peer reviewed journals IS Presentations at scientifîc meetings 19

Chapter 1. Nuclear receptors and dietary ligands -An introduction and selected review 20

1. Abstract 21 1.1. Introduction 22 1.1.2. Endocrine disrupting chemicals 23 1.1.3. Hormone dynamics 23 1.1.4. Hormone mimicking chemicals and steroid hormone receptors 24 1.1.5. The biological differences in hormones and hormone metabolism 24

1.2. Molecular mechanics of endocrine action 25 1.2.1. Receptor-based mechanisms 25 1.2.2. Receptors interacting with enzymes 26

1.3. Most research is based on the oestrogen receptors (ER) 27 1.3.1. Tissue differences in ER distribution in adults 29 1.3.2. Serum oestradiol levels in prepubertal children 29 1.3.3. ERs and the brain; relationships with neurodegenerative conditions 30

1.4. Receptors and ‘cross-talk’ 30 1.4.1. Differences in activation; How much inhibition, how much activation? 31 1.4.2. ERp and men 33 1.4.3. The many biological roles of ERp 34 1.4.4. ERp and ovarian cancer 35 1.4.5. Dietary exposure; Synthetic oestrogens versus phytoestrogens 35

1.5. Key nuclear receptors: a multifaceted communication system 38 1.5.1. Aryl hydrocarbon Receptor (AhR) 40 1.5.2. Pregnane X Receptor (PXR) 42 1.5.3. The Constitutive Androstane Receptor (CAR) 43 1.5.4. Human Peroxisome Proliferator Activated Receptor (hPPAR) 44

1.6. Co-modulators that enhance or suppress ER transcriptional activity 45 1.6.1. Cytochrome P450 enzymes; 45 1.6.1.1. CYP 17A1/19A1 (Aromatase) 46 1.6.1.2. Aromatase, breast cancer and hormonal disruption in women 47

1.7. Molecular modelling 47 1.8. Conclusions 50 Table of Contents

Page Chapter 2. 51 An Environmental and Dietary Case study: The Aquaculture food chain 51 2.1 Introduction 52

2.2. Dioxins and dioxin like PCBs in Salmon 56 2.2.1 Abstract 56 2.2.2. Materials and methods 56 2.2.2.1. Sites and sampling 56 22.2.2. Sample preparation 58 2.2. 2.3. Analytical procedures 58 2.2.2.4. Sample preparation for analysis 58 2.2.2.5. Instrumental analysis 59 2.2.2.Ô. Quality Control and calibration 60 2.2.3. Results 61 2.2.4. Discussion 63

2.3. PCBs, PBDEs and pesticides in the salmon aquaculture food chain 70 2.3.1. Abstract 70 2.3.2. Materials and methods 70 2.3.2.1. Sites and sampling 70 2.3.2.2. Sample preparation 71 2.3.2.3. Fish tissue 73 2.3.2.4. Fish feed 73 2.3.2.5. Fat content determination 73 2.3.2.Ô. Organochlorine analysis 74 2.3.2.7. PBDE analysis 74 2.3.2.8. Quality Control/Quality Assurance 74 2.3.3. Results 2.3.3.1. Salmon samples 75 2.3.3.2. Aquaculture feeds 76 2.3.3.3. Fish oils 78 2.3.3.4. Comparison between the three types of samples 78 2.3.4. Discussion 79

2.4. Investigation of selected PCBs, PBDEs and organochlorine pesticides 89 in omega-3 fatty acid rich dietary fish oil and vegetable oil supplements 2.4.1. Abstract 89 2.4.2. Materials and methods 90 2.4.2.1. Sites and sampling 90 2.4.2.2. Sample preparation 90 2.4.2.3. Sample analysis 91 2.4.2.4. PCB and OCP determination 91 2.4.2.5. PBDE determination 91 2.4.2.Ô. Quality Control 92 2.4.2.7. Oil density 92

2.4.3. Results and Discussion 92 Table of Contents

Page Chapter 2 continued

Section 2.5. Principal Components Time Trend Analysis Of Organochlorine Contaminants Detected In Omega- 3 Rich Fish And Vegetable Oil Dietary Supplements from UK outlets 103 2.5.1. Abstract 103 2.5.2. Introduction 103 2.5.3. Materials and methods 2.5.3.1 Sample sources 104 2.5.3.2 Multivariate analytical procedures 104 2.5.4. Results 105 2.5.5. Discussion 109 2.5.6. Risk assessment 110

Chapter 3. Homology Modelling Of Nuclear Receptors 112 3.1. Abstract 113 3.2. Introduction: Mechanisms of action 113 3.2.1. The human oestrogen receptora (hERa) and oestrogen receptor^ (hERP) 117 3.2.2. The Pregnane-X-Receptor (PXR) 118 3.2.3. The Constitutive Androstane Receptor (CAR) 120 3.2.4. The Aryl Hydrocarbon Receptor (AhR) 121 3.2.5. The human Peroxisome Proliferator Activated Receptors (PPAR) 123

3.3. Molecular modelling methods 126 3.3.1 Introduction 126 3.3.2. Materials and Methods 3.3.2.1. Receptor alignments 127 3.3.2.2. Building in silico models 127 3.3.2.3. Range of ligand probes 128

3.4. Results and Discussion 131 3.4.1. Promiscuity of ligands 132 3.4.2. Evidence of receptor ‘cross-talk’ 132 3.4.3. The Pregnane-X-Receptor (PXR) 133 3.4.4. The Constitutive Androstane Receptor (CAR) 136 3.4.5. The Ah Receptor (AhR) 136 3.4.6. The Human Peroxisome Proliferator Activated Receptor (PPAR) 138

3.5. Conclusion 139

10 Table of Contents

Page

Chapter 4. 141 In Vitro Biological Parameter Estimation In Nuclear Receptors: Experimental Case Study: Activation Of PXR By Dietary Ligands, In Vitro Transient Transfection Assays 142 4.1. Introduction: PXR activation 142 4.2. Cell Based Transfection Assay Materials and Methods 143 4.2.1. Selection of chemicals 143 4.2.2. Cell based transfection assay 144 4.2.2.1 Preparation of nuclear receptor constructs 145 4.2.2.2. PXR Species differences: Rat versus Human 146 4.2.3. Cell culture and maintenance 4.2.3.1. Chemicals and supplies 147 4.2.3.2. Cotransfection assays 147 4.2.3.2.1. Cell maintenance reagents 147 4.2.3.2.2. Cell media 147 4.2.3.2.3. Nuclear receptor constructs for transfection 147 4.2.4. Methods for media preparation 148 4.2.5. Cell culture and maintenance 149 4.2.6. Methods: Daily experimental procedure 152 4.2.6.1. Day One- Cell plating 152 4.2.Ô.2. Day Two - Transfection 153 4.2.Ô.3. Day Three- Compound treatment of transfected cells 155 4.2.Ô.4. Day Four- Alkaline phosphatase (SPAP) assay 155 4.2.7. Data Analysis 158

4.3. Compounds tested 159 4.3.1. Flavonoids 162 4.3.2. Herbs 162 4.3.3. Lignans and polyphenols 163 4.3.4 Persistent Organic Pollutants (POPs) 164 4.3.5. Essential Nutrients: Vitamins 165

4.4. Results 166 4.4.1. The test compounds 166 4.5. Discussion 172 4.5.1 Lignan potency 172 4.5.2 Potency of OH compounds: tamoxifen, vitamin D3 172 4.5.3. Persistent organic pollutants 172 4.5.4. Differential effects of increasing dosage between compounds 173 4.5.5. The role of transporters and other cofactors 174 4.5.5.1. Transporters: P-Glycoprotein 174 Multi-drug resistance associated protein 175 Organic anion transporting protein 175 4.5.5.2. Receptor complements in cell lines 175 4.5.5.3. Example of other factors, Albumin 177 4.6. Conclusions 177

11 Table of Contents

Page Chapter 5. Quantitative Structure Activity Relationships (QSARs) of Nuclear Receptors 179 5. Abstract 180 5.1. Introduction 181

5.2. Methods 183 5.2.1. Collection of data for QSAR studies 183 5.2.2. Generation of Smiles 183 5.2.3. Descriptors 184 5.2.3.1. Abraham descriptors 185 5.2.3.2. GSK in-house descriptors 185 5.2.3.3. VolSurf descriptors 186 5.2.4. Statistical tools 188 5.2.4.1. Principal Component Analysis (PCA) 188 5.2.4.2. Partial Least Squares regression (PLS) 189 5.2.4.3. Model Validity 190 5.2.4.3.1. External validation using a designed validation set 191 5.2.4.3.2. Splitting of parent data set into training and validation sets 191 5.2.4.3.3. The randomization test 192 5.2.4.3.4. Cross validation 192 5. 2.4.4. Predictive models 194 5.2.5. Molecular modelling 195

5.3. Investigation of structure-activity relationships using PLS regression and VolSurf descriptors for a large diverse set of natural, synthetic and environmental compounds in ERa 197 5.3.1. Abstract 197 5.3.2. Introduction 197 5.3.3. Methods 198 5.3.3.1. Source of biological data 198 5.3.3.2. Generation of VolSurf descriptors. 198 5.3.3.3. Statistical tools: Partial Least Squares Regression (PLS) 199 5.3.3.4. Molecular modelling 199 5.3.4. Results and Discussion 199 5.3.4.1. The key VolSurf variables 202 5.3.4.2. Molecular modelling: Comparison with selected compounds docked in the ERa crystal structure 204 5.3.5. Discussion 206

12 Table of Contents

Chapter 5 continued Page 5.4. Building a QSAR for PXR 208

5.4.1. Abstract 208 5.4.2. Introduction 208 5.4.3. QSAR methods 209 5.4.3.1. Generation and collation of biological data Using an in vitro transient transfection assay 209 5.4.3.2. Generation of descriptors 210 5.4.3.3. Statistical tools 210 5.4.3.4. Molecular modelling: in silico investigations of ligand binding in PXR 211 5.4.4. Results 212 5.4.5. Discussion 222 5.4.6. Conclusions 225

5.5. Multivariate (QSAR) Modelling Of Polybrominated Diphenyl Ethers Activity in Different Species Cell Lines and Their Capacity to Induce CYPIA by the Ah Receptor Mediated Pathway 227

5.5.1. Introduction 227 5.5.2. Methods 228 5.5.2.1. Generation of PBDE biological data using in vitro assays 229 5.5.2.2. Statistical analysis for the evaluation and interpretation of the data 229 5.5.2.5. Utilization of a homology model of AhR Based on the ERa crystal structure 229 5.5.3. Results and Discussion 229 5.5.4. Conclusions 234

5.6. Overall Discussion and Conclusions 234

Chapter 6. Nuclear Receptor Ligand-Gene Interactions And Implications for Human Health 237

6.1. Nuclear receptors and human health; 6.1.1. Summary 238 6.1.2. The development of steroid receptors; an evolutionary perspective 240 6.1.3. Receptor-ligand interactions 242 6.1.4. Role of Protein and Enzyme Coregulators 242 6.1.4.1. Receptor-receptor and receptor-enzyme interactions 243 6.1.4.2. Heterodimerization partner: RXRa 243 6.1.4.3. Heterodimerization partner: ARNT 244 6.1.4.4. Chaperone Proteins 244 6.1.4.5. Selected examples of other activators and cellular responses 245 6.1.4.6. CYP17A1/19A1 (Aromatase) 245

6.1.5. Considerations and conclusions 246

13 Table of Contents

Chapter 6 continued Page

6.2. Future work 246 6.2.1. Future research with respect to contaminants in the aquatic food chain 247 6.2.1.1. Future research: A European Union policy perspective 247 6.2.1.2. Potential Impact 248 6.2.2. In silico molecular modelling and QSARs: Further research 249 6.2.3. Applications: Tier testing 250

6.3. Conclusion 251

7. Bibliography 253

Appendices 295 Appendix 1 Table i) Batch sample details for omega 3 rich dietary supplement samples, Chapter 2 296 Table ii) Sample details and dosage instructions for omega 3 rich dietary supplement samples. Chapter 2 297

Appendix 2 302 Table i) Raw data generated by the PXR transient transfection Assay, HuH7 cells. Chapter 4 302 Table ii) Raw data generated by the PXR transient transfection Assay, CV-1 cells. Chapter 4 306

Appendix 3 Table i) The VolSurf Descriptors used for the QSAR studies. Chapter 5 311

14 Table of Contents

Appendices continued Page

Appendix 4 Selection of posters presented during the course of this thesis 312

Chapter 2 Poster 1: Investigation of PCDDs, PCDFs and selected coplanar PCBs in Scottish farmed Atlantic Salmon, Dioxin 2000. 313

Poster 2: Investigation of selected PCBs and organochlorine pesticides in farmed Atlantic salmon (Salmo Salar), Salmon aquaculture feed and fish oil components of the feed. Dioxin 2001. 314

Poster 3: Principal Components Time Trend Analysis Of Organochlorine Contaminants Detected In Omega- 3 Rich Fish And Vegetable Oil Dietary Supplements from UK outlets. BTS Annual Congress 2003. 315

Chapter 3 Poster 4: Homology modelling of the human oestrogen receptorp (hER) from the human oestrogen receptora (hERa) crystal structure, EUROTOX 2000. 316

Poster 5: Homology modelling of the aryl hydrocarbon receptor (AhR) ligand binding domain from the human oestrogen receptor alpha crystal structure, EUROTOX 2000 317

Poster 6: Homology modelling of the human peroxisome proliferator activated receptora (PPARa) ligand binding domain from the human PPARy crystal structure, EUROTOX 2000. 318

Poster 7: Homology modelling of the pregnane-X receptor (PXR) from the human oestrogen receptor alpha crystal structure, EUROTOX 2000. 319

Poster 8: Homology Modelling Of The Nuclear Orphan Constitutive Androstane Receptor (CAR) Ligand Binding Domain From The Human Oestrogen Receptora (HERa) Crystal Structure. BTS Annual Congress 2002. 320

Chapters 4 and 5 Poster 9: Investigation of structure-activity relationships using PLS regression and VolSurf descriptors for a large diverse set of natural, synthetic and environmental compounds in Estrogen receptor a. EURO QSAR 2002. 321

Poster 10: Building a QSAR Model For Pregnane X Receptor Ligands. ISSX 2003. 322

Poster 11: The comparative use of quantitative structure activity relationships (QSARs) and molecular modelling for understanding receptor mediated mechanisms of toxicity, receptor cross-talk and implications for endocrine disruption. EUROTOX 2003. 323

15 Summary

This course of research examines ligand interactions with nuclear receptors and cross- talk between receptors and implications for human health by using a number of different scientific research tools. These include in silico molecular modelling using nuclear receptor crystal structures;in vitro assays and multivariate analytical techniques to derive robust models of receptor activation. The selection of ligands were based both on literature reviews, which constitutes Chapter 1, (Jacobs and Lewis, 2002) and Chapter 2, an environmental pollutant case study of contaminants in the aquaculture food chain and dietary oil studies (Jacobs et al., 2002a; 2002b). The data from these studies informed likely relative dietary exposure of individual environmental contaminant congeners of European consumers of oily fish and oils (Jacobs et al., 2002a; 2002b; Jacobs et al., 2002 c; Jacobs, 2002; Jacobs, 2003). While they were conducted in addition to the molecular work, as they contained information pertinent to my PhD project, and the studies were conducted at similar times, they have been included here as a legitimate and informative avenue of study. However they were not funded by my PhD funding through the BBSRC and GSK, University of Surrey facilities were not used, and they were conducted independently, in my own time, outside the time allocated for my PhD work. For support and collaboration with these studies I am extremely grateful to Joe Ferrario of the US Environmental Protection Agency, and Adrian Covaci, of the University of Antwerp, Belgium. Chapter 3 presents the receptor models built and the relevant crystal structures used for the ERp, AhR, PXR, CAR, and PPAR models (Jacobs et al., 2003a). These models or the crystal structure if available later, are shown with ligands of interest docked in the ligand binding domains of the relevant receptors. Differences in the ligand binding domains and modes of binding were observed. These studies provided an additional basis for selection of experimental compounds to test in an in vitro nuclear receptor assay described in Chapter 4, and supported the receptor based Quantitative Structure Activity Relationships (QSARs) described in Chapter 5. Chapter 4 describes how the biological data was generated experimentally utilising the in house GSK PXR transient transfection assay at the GSK site in Ware, UK. Additional biologically relevant data for PXR and ERa and AhR was obtained from GSK in house data sets, the literature and unpublished data was also kindly provided by Guosheng Chen from the University of Guelph, Canada (Jacobs et al. 2003b). These

16 data sets were utilised to generate the QSARs presented in Chapter 5. The QSARs presented were supported by further exploration of the modes of ligand binding in the crystal structures and/or homology models of the SHRs under study. Chapter 6 discusses the applications of the studies documented in the preceding chapters and future avenues of related research with particular reference to the implications of the studies presented here for human health. Endocrine systems need to be allowed to maintain their equilibrium. Perturbing the endocrine system can result in ill health, reduced quality of life, and loss of self; Tfyou skew the endocrine system, you lose the pathways to self Hilary Mantel in Giving up the Ghost: A Memoir. Publ Fourth Estate 2003. This thesis is a starting point for the development of scientific tools that add to and are alternatives to animal in vivo models of endocrine disruption, and tools that can suggest alternatives in the synthesis of new chemical entities, for better protecting human health and the wider global environment upon which we all depend.

17 Papers published and submitted from this thesis

MN Jacobs, M Dickins, DFV Lewis. Homology Modelling Of The Nuclear Receptors: Human Oestrogen ReceptorP (hERP), The Human Pregnane-X-Receptor (PXR), The Ah Receptor (AhR) And The Constitutive Androstane Receptor (CAR) Ligand Binding Domains From The Human Oestrogen Receptora (hERa) Crystal Structure, And The Human Peroxisome Proliferator Activated Receptora (PPARa) Ligand Binding Domain From The Human PPARy Crystal Structure.Journal of Steroid Biochemistry and Molecular Biology 2003, 84/2. 117-132.

DFV Lewis, MN Jacobs, M Dickins and BG Lake. Molecular Modelling Of The Peroxisome Proliferator-Activated Receptor a (PPARa) From Human, Rat And Mouse, Based On Homology With The Human PPARy Crystal Structure. Toxicology in Vitro 2002,16, 275-280.

DFV Lewis, MN Jacobs, M Dickins and BG Lake. Quantitative Structure-Activity Relationships For Inducers Of Cytochromes P450 And Nuclear Receptor Ligands Involved In P450 Regulation within the CYPl, CYP2, CYP3 and CYP4 families. Toxicology. 2002,176,51-57.

MN Jacobs, A Covaci, P Schepens. Investigation Of Selected Persistent Organic Pollutants In Salmon (Salmo Salar), Salmon Feed And Fish Oil Components Of Salmon Feeds. Environmental Science and Technology. 2002,36, 2797-2805.

MN Jacobs and DFV Lewis. Steroid Hormones Receptors And Dietary Ligands - A Selected Review. Invited paper for The Proceedings of the Nutrition Society, 2002,61 (1), 105-122.

MN Jacobs, J Ferrario, C Byrne. Investigation of PCDDs, PCDFs and selected coplanar PCBs in Scottish farmed Atlantic Salmon Chemosphere 2002,47 (2), 183-191.

Book Chapters

MN Jacobs ‘Steroid hormone receptors and dietary ligands’ in ‘Nutrients and Cell Signaling’, editors: J Zemplini and K Dakshinamurti, ‘Oxidative Stress and Disease’ series, publ. Marcel Dekker Inc, New York 2004/5.

MN Jacobs ‘Unsafe sex, how endocrine disrupters work’ in MN Jacobs and B Dinham ed. Silent Invaders: Pesticides, Livelihoods and Women’s Health. Publ. Zed books 2002.

Abstracts in peer reviewed journals/conference proceedings

MN Jacobs, C Luscombe, M Dickins, S Hood, DFV Lewis. The Comparative Use Of Quantitative Structure-Activity Relationships (QSARs) and Molecular Modelling For Understanding Receptor Mediated Mechanisms Of Toxicity, Receptor Cross-Talk And Implications For Endocrine Disruption. Toxicology Letters. 2003,144 Suppl. 1 s52-s53

MN Jacobs, G Chen, C Luscombe, NJ Bunce. Multivariate (QSAR) Modelling of Polybrominated Diphenyl Ethers activity in different species cell lines and their capacity to induce CYPIA by the Ah Receptor Mediated Pathway. Organohalogen Compounds. 2003,61, 96-99.

MN Jacobs, M Dickins, C Luscombe, G Nolan, S Hood. Building a QSAR Model For Pregnane X Receptor Ligands. Drug Metabolism Reviews, 2003, 35 SI, 116.

18 MN Jacobs. Principal Components Time Trend Analysis Of Organochlorine Contaminants Detected In Omega- 3 Rich Fish And Vegetable Oil Dietary Supplements from UK outlets. Toxicology Letters, in press. 2003

MN Jacobs and C Luscombe. Investigation of structure-activity relationships using PLS regression and VolSurf descriptors for a large diverse set of natural, synthetic and environmental compounds in Estrogen receptor a. EURO QSAR 2002, Bournemouth, UK In proceedings from EUROQSAR 2002, in press. 2003

MN Jacobs, A Covaci, A, Ghoerghe and P Schepens. Investigation of selected PCBs, PBDEs and organochlorine pesticides in omega-3 fatty acid rich dietary fish oil and vegetable oil supplements. Organohalogen Compounds, 2002,57, 225-229.

MN Jacobs, M Dickins, DFV Lewis. Homology Modelling Of The Nuclear Orphan Constitutive Androstane Receptor (CAR) Ligand Binding Domain From The Human Oestrogen Receptora (HERa) Crystal Structure. Poster presentation to the Annual British Toxicology Society April 2002. Published in BTS Programme and Abstracts p 99. and Toxicology Letters, 2002.

MN Jacobs and DFV Lewis. Homology modelling of the human oestrogen receptor^ (hER) from the human oestrogen receptora (hERa) crystal structure Toxicology Letters, 2000, 116 (Supplement 1) 58.

MN Jacobs and DFV Lewis. Homology modelling of the human peroxisome proliferator activated receptora (PPARa) ligand binding domain from the human PPARy crystal structure. Toxicology Letters, 2000,116 (Supplement 1) 59.

MN Jacobs and DFV Lewis. Homology modelling of the pregnane-X receptor (PXR) from the human oestrogen receptor alpha crystal structure. Toxicology Letters, 2000,16 (Supplement 1) 58.

MN Jacobs and DFV Lewis. Homology modelling of the aryl hydrocarbon receptor (AhR) ligand binding domain from the human oestrogen receptor alpha crystal structure. Toxicology Letters, 2000,116 (Supplement 1) 59.

MN Jacobs, A Covaci, P Schepens Investigation of polybrominated diphenyl ethers in Scottish and European farmed Atlantic Salmon (Salmo salar), Salmon aquaculture feed and fish oils. Organohalogen Compounds, 2001,51, 239-242.

MN Jacobs, A Covaci, P Schepens, E Millstone Investigation of selected PCBs and organochlorine pesticides in farmed Atlantic Salmon (Salmo salar), Salmon aquaculture feed and fish oil components of the feed. Organohalogen Compounds, 2001,51, 314-318.

MN Jacobs, J Ferrario and C Byrne. Investigation of PCDDs, PCDFs and selected coplanar PCBs in Scottish Farmed Atlantic Salmon Organohalogen Compounds, 2000, 47,338-341.

19 C hapter 1

Nuclear Receptors and Dietary Ligands

- An Introduction and a Selected Review

20 C hapter 1

Nuclear Receptors and Dietary Ligands

- An Introduction and a Selected Review

1. Abstract

Members of the nuclear steroid hormone receptor superfamily mediate essential physiological functions. Nuclear receptors (NRs) and steroid hormone receptors (SHRs) act directly upon DNA, regulate the synthesis of their target genes, and are usually activated by ligand binding. Both endogenous and exogenous compounds and their metabolites may act as activators of NRs and disruptors of endocrine, cellular and lipid homeostasis. The endogenous ligands are generally steroids such as l?P-oestradiol, androgens, progesterone and pregnenolone. The exogenous compounds are usually delivered through the diet and include non-steroidal ligands. Examples of such ligands include isoflavonoids or phytoestrogens and food contaminants such as exogenous oestrogens from hormone treated cattle, pesticides, polychlorinated biphenyls (PCBs) and plasticisers. Certain drugs are also ligands; so nuclear receptors are also important drug targets for intervention in disease processes. This chapter summarises recent reports on ligand activated NRs that describe the selective regulation of a tightly controlled, cross-talking network involving exchange of ligands, and the control of major classes of cytochrome P450 (CYP) isoforms which metabolise many bioactive exogenous compounds. Many P450s have broad substrate activity and appear to be integrated into a co-ordinated metabolic pathway, such that whilst some receptors are ligand specific, other sensors may have a broader specificity and low ligand affinity to monitor aggregate levels of inducers. They can then trigger production of metabolising enzymes to defend against possible toxic nutrients and xenobiotic compounds. The influence of dietary intakes of nutrients and non nutrients upon the human oestrogen receptors (hERa and (3), the Aryl hydrocarbon Receptor (AhR) the Pregnane-X-Receptor (PXR), the Constitutive Androstane Receptor (CAR), and the Peroxisome Proliferator-Activated Receptors (PPARa and y), can be examined by utilising computer generated molecular models of the ligand-receptor interaction, based on information generated from crystallographic data and sequence homology. In

21 relation to experimental and observed data, molecular modelling can provide a scientifically sound perspective on the potential risk and benefits to human health from dietary exposure to hormone mimicking chemicals, providing a useful tool in a complex situation of considerable public concern.

1.1. Introduction 1.1.2. Endocrine disrupting chemicals The endocrine system is complex, with many organs producing different hormones, contributing to a multifaceted feedback regulatory system that is deficient in the developing fetus and infant. Endocrine disrupting compounds (EDCs) are substances that can cause adverse effects by interfering in some way with the body’s hormones or chemical messengers that constitute the endocrine system. Under the control of the Central Nervous System (CNS), hormones are secreted by the endocrine glands and exert control on other cells of the body. Endocrine-disrupting chemicals are differentiated from classical toxicants such as carcinogens, neurotoxicants and heavy metals, because they can interfere with normal blood hormone levels or the subsequent action of those hormones, but do not have a classical toxic effect. The effects can influence and disrupt the hormonal regulation and hormonal imprinting (in the fetus) of normal cell differentiation growth, development, metabolism and reproduction throughout life. Endocrine disruption can occur at levels far lower than those of traditional concern to toxicologists. Sometimes high doses shut off effects that occur at low levels, and sometimes low and intermediate doses produce greater effects than those observed at high doses. Human populations are exposed to complex mixtures of environmental and endogenous agents, which may act together or modulate one another to produce biological effects. While exogenous endocrine modulators can involve any hormonal system and be affected by normal physiological states (such as menstruation and menopause in women, the andropause in men, and puberty, the somatopause and adrenopause in both sexes), diet, stress and other lifestyle factors, the principal focus and starting point of this Chapter is on sex-hormone, and especially oestrogen-mediated effects upon the steroid hormone and nuclear receptors in humans where the ligands are delivered through the diet. The intention is that this Chapter will provide a basis for understanding some of the critical issues involved in hormone receptor mediated toxicity, and the broad context involved.

22 1.1.3. Hormone dynamics Accurate assessment of a chemical’s potential to alter the endocrine system depends upon consideration of the entire hormone dynamic pathway together with changes in hormone activity. Figure 1.1 presents a simplistic model of hormone dynamics, following hormones from production to excretion. After a hormone is produced, circulating and intracellular binding proteins regulate the hormones’ bioavailability. Then the hormone triggers action by binding to a specific cellular receptor. The hormone is then either excreted in the urine after conjugation reactions in the liver, or biotransformed into another hormone, which will begin the cycle of bioavailability, action, and excretion/biotransformation. Hormones, and hormone mimicking chemicals, can potentially affect each point in the hormone dynamic cycle, and each point in the steroid hormone pathway, and the enzyme families associated with the pathway. Assessing the mode of action of an EDC is further complicated by many feed-back mechanisms within and between the different hormone systems, as well as interconnections with the nervous and immune systems. Current scientific knowledge of these systems is fragmentary but the emerging picture suggests an interlinked fabric of hormone dynamics where the delicately balanced compensatory systems are easily perturbed. Hormone dynamics have evolved over a long period of time to deal with hormone and dietary phytochemical exposure, but they are not a rapid response system able to deal effectively with the synthetic chemicals of the twentieth and twenty first centuries. At the molecular level, receptors mediate alterations of hormone availability, action, excretion and biotransformation, in concert with the cytochrome P450 enzyme system. Many receptors have been identified and are awaiting the recognition of specific ligands and functions, and it is likely that more receptors will be discovered in the future. Each type of receptor has the potential to regulate a distinct endocrine signalling pathway, of which we only have a rudimentary knowledge.

23 Figure 1.1. A simplified model of hormone dynamics (after Crain et al., 2000)

— Hormone Production -Endogenous hormones, Endogenous -Exogenous hormones. -Environmental chemicals, -Hormonal compounds of Hormone natural origin (e.g. Biotransformation Hormone Phytoestrogens ) Availability

Enzyme action Receptor Hormone (i.e. P450) Excretion Hormone Action Receptor < y

1.1.4. Hormone mimicking chemicals and steroid hormone receptors

In recent years there have been many reports on the increasing incidence of breast cancer in women (Davis et al., 1993; B radio w et al., 1995; Bradlow et al., 1997; Davidson and Yager, 1997; Davis et al., 1997; DeBruin and Josephy, 2002; Green et al., 2003) and decreased sperm counts and increasing incidence of testicular cancer in men (Sharpe and Skakkebaek, 1993; Sharpe, 2001; Auger et al., 1995; Adami et al., 1994), together with adverse wildlife effects that include birth defects, reproductive failures, and sexual abnormalities (Colburn et al., 1993; Guillette and McLachlan, 1996; Colburn et al., 1999; Hayes et al., 2002). These have stimulated research into both the chemical and molecular actions, and the clinical and epidemiological effects, of a large variety of natural and synthetic oestrogens present in the environment (Falck et al., 1992; Hunter and Kelsey, 1993; Wolff et al., 1993; Kreiger et al., 1994;Stevens et al., 1994; Ahlborg et al., 1995; Wolff and Toniolo, 1995; Gray Jr et al., 1997; Hunter et al., 1997; Safe, 1997a; Bernstein and Press, 1998; Petreas et al., 1998; Krstevska-Konstantinova et al., 2001). The growing body of evidence on the hormone-like effects of many synthetic chemicals and by-products in fish, wildlife and humans, has led to the instigation of endocrine disruption screening programmes relating to persistent organic pollutants (POPs), and the international POPs treaty signed in Stockholm, Sweden in May 2001.

1.1.5. The biological differences in hormones and hormone metabolism

In utero, our natural sex hormones are largely responsible for our development into females, or totally responsible for male development, as the default pathway for fetal development is phenotypically female. In the male fetus, androgens stabilise the Wolfian

24 duct development in the fetus, and actively remove the Müllerian duct (the precursor to the female uterine system) (Sharpe 2001), The most important factors guaranteeing the homeostasis of female and male sexual functions include differentiation and reproduction. Main target tissues include bone and skin, the cardiovascular system, the brain and central nervous and immune system. Many endocrine hormones are lipophilic (fat loving), moving easily through cell membranes to activate or suppress the nuclear receptors (i.e. receptors in the nucleus of the cell) that directly act upon DNA. In concert with each other they contribute to the control of broad aspects of growth, development and adult organ physiology. The NR and SHR family is extensive but currently the emphasis is on a sub-classification, the oestrogen receptors (ER) (Colburn et al., 1993; Soto et al., 1995; Guillette and McLachlan, 1996; Colburn et al., 1999) with most effects attributable to pubertal or adult levels of steroids. Other members of the SHR family and other nuclear receptors of unknown functions (not discussed here) also play key roles in endocrine disruption (Parker, 1993). For example, EDCs are known to affect the thyroid, adrenal glands and pancreas, a clear link between DDT and pancreatic cancer was recently reported (Porta et al., 1999), and the persistent DDT metabolite p, p '-DDE has been shown to have anti- androgenic potential (Kelce et al., 1995). EDCs such as phthalic acid and nonylphenol are now known to activate PXR, both in vitro and in vivo, as observed in the ERs, while bisphenol A has been both reported to have no effect on PXR-mediated transcription (Masuyama et al., 2000), and later reported to be a ligand for PXR (Takeshita et al., 2001) while also significantly enhancing ER-mediated transcription (Masuyama et al.,

2000).

1.2. Molecular mechanics of endocrine action

1.2.1. Receptor-based mechanisms There are many different steroid hormones and therefore many different SHRs and NRs. These receptors are proteins which are synthesized by the cell in the same way as other proteins are synthesized, under genetic control. Thus all cells have the ability to synthesize all receptors, but variations in the extent of gene expression means that an individual cell or tissue probably does not synthesize receptors for more than three or four hormones (Gard, 2001). This is considered to be part of the selectivity of the hormone mechanism to produce effects on selected target tissues.

25 The nuclear receptor family have structural features in common. These include a central highly conserved DNA binding domain (DBD) that targets the receptor to specific DNA sequences, termed hormone response elements (HREs). A carboxyterminal portion of this receptor includes the ligand binding domain (LED) that interacts directly with the hormone. Embedded within the LED is a hormone dependent transcriptional activation domain. The LED acts as a molecular switch that recruits coactivator proteins and activates the transcription of target genes when flipped into the active conformation by hormone binding. The currently accepted theory of steroid hormone binding suggests that in the absence of the hormone, each receptor is associated with certain ‘chaperone’ proteins (these are other proteins which protect and aid the receptor) (Weigel, 1996). Binding of the steroid hormone with the receptor protein causes a conformational change. This molecular switch results in the removal of the heat shock complex and allows the receptors to dimerise. Then DNA binding to a hormone response element (HRE) occurs, to produce a complex that can trigger or suppress the transcription of a selected set of genes, see figure 1.2 for a simple illustration of the mechanism of action of nuclear receptors (Weigel, 1996; Alberts et al., 1997). Figure 1.2. Diagram showing the simplified mechanism of action of nuclear receptors LIGAND - Hormone NUCLEUS - EDO pesticide - Phytoestrogen - Plasticisers - Drugs

Dimerization KEY R = receptor: ERa. .ACTIVATION ERp TARGET PxR HRE EFFECT AR GENE CAR AMR hsp = heat shock protein and co-factors ‘chaperone complex’ leading to EFFECTS: - increase in DNA proliferation mutation and HRE = hormone response - alteration of protein expression J CARCINOGENESIS elem ent

1.2. 2. Receptors interacting with enzymes The hormone receptors and chaperone cofactors also mediate hormonal homeostasis by the coordinated release and degradation of bioactive hormones. Steroid hormones, their metabolites, ingested plant and animal steroids and bioactive xenobiotic

26 compounds are primarily metabolised by cytochrome P450 (P450s) enzymes reduction and oxidation in the liver. Many P450s have broad substrate activity and appear to be integrated into a coordinated metabolic pathway, such that whilst some receptors are ligand specific, other sensors may have a broader specificity and low ligand affinity. In this way they can monitor aggregate levels of inducers to trigger production of metabolizing enzymes, and thereby mount a defense against toxic compounds in the diet. This hypothesis is strongly supported by the expression of a receptor sensing system in the digestive tissue (Jones et al., 2000). P450 induction by EDCs and other xenobiotics may therefore lead to alterations of endogenous regulatory pathways, with associated physiological consequences harmful to health.

1.3. Most research is based on the oestrogen receptors (ER) Due to the early identification of ERa and links between DDT and breast cancer, the ERs have received most EDC research attention. So far they are known to exist as two subtypes, each one encoded by a separate gene. These are ERa (Brzozowski et al.,

1997), see figure 1.3, and the recently discovered ER(3 (Kuiper et al., 1997; Kuiper et

al., 1998) and its isoforms, of which a spliced isoform, ER|3/2 appears to be equally

expressed in tissue density studies (Petersen et al., 1998). Figure 1.3. 17(3 oestradiol docked in the LBD of the ERa crystal structure (using SYBYL biopolymer software from Tripos associates, St Louis Missouri) on a Silicon Graphics Indigo” IMPACT 10000 Unix workstation. (Jacobs, 1998) *

* The pink ribbon denotes the protein backbone of the receptor, key amino acid residues are shown in blue and red, 17 P oestradiol is shown in orange. Hydrogen bonding is indicated by the dotted pale blue lines.

27 Figure 1.4. Schematic diagram showing differences in tissue distribution of the oestrogen receptors (Gustafsson, 1999; Warner et al., 1999)

Brain & CNS ERa e r (3 Pituitary ERa Hypothalamus ERp Thymus ERp

Breast Cardiovascular ERa ERp system Liver ERa ERp ERa ERp Lungs i:Rp

Adrenals and kidneys ERa ERP Ovaries ERa i:Rp Gastrointestinal tract ERp Uterus ERa ERp Urogenital tract ERa ERp Bone Bladder ERp [:Ra i:Rp Testes Skin ERp ERa ERp

Prostate ERP

The classical ERa subtype and ERP receptors and isoforms apparently evolutionary diverged over 450 million years ago (Kelley and Thackray, 1999), suggesting that although they have evolved in parallel, this ancient duplication was to facilitate unique roles in vertebrate physiology and reproduction. The ERs differ in tissue distribution and relative ligand binding affinities (Kuiper et al., 1997; Petersen et al., 1998), which may help explain the selective action of oestrogens in different tissues (Gustafsson, 1999). Indeed there is evidence that ERp plays an important role in the organization and adhesion of epithelial cells in breast tissue and is therefore instrumental in the differentiation of tissue morphology, particularly if stimulated at an early age (Forster et al., 2002). This has important gender implications due to the differences in tissue physiology between women and men and important disease implications, as different spliced variants are observed in malignant as opposed to normal tissues that lead to poor patient prognosis related to oestrogen refractoriness (Fujimoto et al., 2000a; Fujimoto et al., 2000b; Omoto et al., 2002).

28 1.3.1. Tissue differences in ER distribution in adults Whilst women have ERa and ER(3 in breast, uterine and ovarian tissue, men have

ERp in the prostate and ERa in the testes. Both women and men have ERa in the

adrenals and kidneys, but ERp in the brain, thymus, lung, vascular system, bladder and bone and skin (Kuiper et al., 1997; Paech et al., 1997; Tetsuka et al., 1997; Thornton et al., 2003). ERa dominates specifically in the reproductive tissues, while ERp plays an important role in the physiology of several tissues, as indicated in figure 1.4. Female and male sex hormones can be understood to act as functional antagonists, such that an excess of oestrogenic hormones may depress male development or male functions. There are also situations in which male and female hormones may act synergistically (and, for example, show effects upon bone density or the promotion of liver tumours). This complexity is the consequence of the multiple targets of these hormones within mammals, including target tissue other than the sex organs, as the tissue distribution of the ERs in adults indicate.

1.3.2. Serum oestradiol levels in prepubertal children Oestrogens are considered reversible cellular signals for adults, but when given to newborn mice, at least one gene under oestrogen control is persistently expressed, even in the absence of oestrogen (Nelson et al., 1991). While very little is known about the role of hormones in gene imprinting, potent oestrogens have been observed in vivo to contribute to DNA méthylation or déméthylation (McLachlan et al., 2001). Low doses of oestrogen have been observed to have a significant biological effect in girls with Turner syndrome. A dose of lOOng/kg/day ethinylestradiol administered orally for 5 weeks resulted in a significantly increased growth velocity (P<0.001), with no effects upon vaginal maturation or breast tissue. Higher doses did not increase the growth rate, indicating that oestrogen has a biphasic dose response relationship for epiphyseal growth. Similar effects have been reported in boys, but there are clear gender differences in serum oestradiol levels prepubertally, with girls having approximately eightfold higher oestradiol levels than boys (Klein et al., 1994). This may explain why girls mature faster than boys, but also, in a population exposed to excessive but low levels of oestrogen, one may expect to see a slightly younger age for pubertal onset (Andersson and Skakkebaek, 1999; Krstevska-Konstantinova et al., 2001) and an increase in oestrogen related diseases and cancers (Newbold et al., 2001).

29 1.3.3. ERs and the brain: relationships with neurodegenerative conditions The brain and the role of ERp in the brain is attracting increasing interest. The expression of ERs in the adult cortex and hippocampus, areas associated with learning and memory, have been observed experimentally to be responsive to neuronal injury, suggesting a link may exist between early onset of senile dementia or Alzheimer’s disease in post menopausal women with significantly reduced oestrogenic activity, compared with normal post menopausal levels (Shughrue and Merchenthaler, 2000; Wise et al., 2001). During fetal and postnatal development, peak neuronal cell proliferation occurs within specific brain regions, including the hippocampus and cortex. Unlike other organ systems, the brain and CNS development has a limited capacity to compensate for cell loss, and environmentally induced cell death can lead to a permanent reduction in cell number. Combined with higher relative cerebral blood flow and the immaturity of the blood brain barrier, this can give rise to increased exposure of the brain to potential neurotoxins and EDCs (Faustman et al., 2000; ten Tusscher and Koppe, 2003). The involvement of ERa and ER(3 in the brain learning and memory centres suggest a logical mechanism of action of EDCs in the brain that could trigger learning difficulties, as reported in male ERP deficient mice. Whilst this is speculation, in vivo work on ER deficient and aromatase deficient mice (Simpson et al., 2000) should shed more light on the mechanism of the disruption of ERs and aromatase upon brain function.

1.4. Receptors and ‘cross-talk’ The receptors often have ligands in common (albeit with different binding affinities), and there is also a great deal of ‘cross-talk’ and ‘ligand promiscuity’. That is, the same chemical may be able to bind with different receptors, and if they do, the binding strength may differ greatly between different receptors. This can occur for endogenous hormones, exogenous hormones and environmental chemicals. For example, forms of the oestrogenic (oestradiol) and androgenic (androstanes) hormones are both ligands for ERp. Oestradiol is less potent in ERP than in ERa, whilst the natural ligands for ERP may actually be androgens: 5a androstane diol and 3p androstane diol (Gustaffson, 2000). The organochlorine pesticides trans-nonachlor and chlordane are known to activate both known ERs, and PXR, but with different affinities (Kuiper et al., 1997; Barkhem et al., 1998; Waxman, 1999; Jones et al., 2000). As these receptors are present in different ratios in different cell types and tissues, the response on a cellular, tissue and systemic level may be quantitatively very different, and may vary over time.

30 Receptor modulation has been seen with lactation, when a form of ERp has been observed to increase in the rat mammary glands (Gaido et al., 1999; Gustafsson, 2000) and in breast tissue hyperplasias, where a frequent mutation in the ERa gene shows increased sensitivity to oestrogen compared with wild type ERa, by affecting the border of the hinge and hormone binding domains (HBD) (Fuqua et al,, 2000). EDCs may act on some but not all of the receptors and their isoforms in the tissues of these organs, or act to different affinities, as methoxychlor and its analogue DDT do in ERa and ERp. More specifically, ERp is dependent on pure agonists for the activation of transcription from its target promotors, while ERa can be activated by agonists, partial agonists (such as tamoxifen, which is used in the treatment of ER positive breast cancer) and ligand independent mechanisms. Appropriate test systems must be strategically designed to cover such multiple actions and interactions. There is also evidence that isoforms of different receptors modulate each other at a functional level, attempting to retain a balance. The modulation is aided at low (Hall and McDonnell, 1999) and high hormone levels (Petersen et al., 1998) by different ERp isoforms, which may enable a tissue to govern its own responsiveness to oestradiol, oestradiol metabolites and related hormones such as progestins, and oestrogen mimics or EDCs. This speculation is supported by what appears to be an emerging pattern in nuclear receptor signalling, as similar balancing acts have been observed also in the a and p forms of the human glucocorticoid receptor (GR), the progesterone receptor (PR), and more recently between the pregnane X receptor (PXR) and the constitutive androstane receptor (CAR) (Honkakoski et al., 2003).

1.4.1. Differences in activation: How much inhibition, how much activation? The proliferative role of ERa in breast cancer is well established, but there is increasing evidence that ERp is part of the disease picture, through antiproliferative activity, which has therapeutic implications. The distinct differences in activational mechanisms between the different ERs and their isoforms will have a direct effect upon possible disease scenarios, such as ER positive and ER negative breast cancers, testicular and prostate cancers, and perhaps also brain injury. EDCs may act on some but not all of the receptors and their isoforms in the tissues of these organs, or act to different affinities, as many phytoestrogens (e.g. coumestrol and genistein), synthetic chemicals such as bisphenol A (Matthews et al..

31 2001) and organochlorine pesticides (e.g. methoxychlor and its analogue DDT) do in

ER a and ERp. These differences between the two isoforms are of great importance with respect to the endocrine disrupting ability of a ligand. The ability of ERP to function both as an inhibitor or activator depending on the agonist concentration suggests that totally different patterns of gene expression may be observed at different hormone levels, and may be a mechanism by which cellular sensitivity to hormones is controlled. In tissues where ERa and ERp co-localise, fluctuations in the bioavailability of receptor activating ligands may have a greater impact, so characterizing the interactions of EDCs with both ERa and ERp (see figure 1.4.). There can also be additional non- hormonal pathways influenced by specific compounds. A substance may act in a synergistic way on one target and in an antagonistic way on another one (Paech et al., 1997). Such opposing effects may also occur at different dose levels of the same substance, or as combinations of, for example, phytoestrogens and PCBs (Jacobs and Lewis, 1999), and other EDCs (Payne et al., 2001) but experimentally there have been difficulties replicating these findings, and consequently one study has been retracted (Arnold et al., 1996; Arnold et al., 1997; McLachlan, 1997). The in vitro evidence of additive potency of combined EDC pesticides in ERa is far stronger (Ramamoorthy et al., 1997; Payne et al, 2001; Silva et al., 2002), and has been noted for other receptors also (Carpenter et al., 2002). Phytoestrogens appear to have a greater affinity for ERP than ERa (Kuiper et al., 1997), but they have an ERa selective efficacy (Pike et al., 1999; Gustafsson, 2000), while EDCs such as the hydroxylated metabolites of methoxychlor appear to be an ERa specific agonist, and ERP antagonist (Gaido et al., 1999), but with about the same affinity for both isoforms (Kuiper et al., 1997).

32 Table 1.1. Relative Binding Affinities of suspected endocrine disrupters for ERa and ERp, adapted from solid phase competition experiments (Kuiper et al., 1997).

Compound RBA RBA ERa ER(3

17(3 oestradiol 100 100 Isoflavones: Coumestrol 20 140 Genistein 4 87 Daidzein 0.1 0.5 Pesticides: o,p’-DDT 0.01 0.02 Chlordecone 0.06 0.1 Endosulphan <0.01 <0.01 Methoxychlor* <0.01 <0.01

RBA of each competitor was calculated as a ratio of concentrations of oestradiol and competitor required to reduce the specific radioligand binding by 50% (=ratio of IC 50 values). RBA value for oestradiol was arbitrarily set at 100. *The metabolite of methoxychlor, 2,2-bis-(p-hydroxyphenyl)-l,l,l-trichloroethane (HPTE) is approximately 100-fold more active at ERa than methoxychlor (Gaido et al., 1999).

1.4.2. ERp and men Perinatal exposure of the male fetus to potent oestrogens is known to increase the incidence of cryptochordism and hypospadias at birth, and small testes, reduced sperm counts, epididymal cysts, prostatic abnormalilites and testicular cancer in adult animals. ERP appears to be preferentially expressed at significant levels in spermatocytes at various developmental stages in the testes, specifically the gonocytes, spermatogonia and spermatocytes. This suggests that EDCs with an affinity for ERP may find their way into precursors of sperm and cause disturbances in their function, disrupting male reproductive functions (Giwercman et al., 1993; Hess et al., 1997; Sharpe, 1997; Sharpe et al., 1998; Shughrue and Merchenthaler, 2000; Sharpe, 2001; Richtoff et al., 2003). Timing is all-important. During fetal testicular development, if one phase of development is out of phase with the following step ( such as Müllerian duct regression for instance), developmental problems ensue in the adult that may not become apparent until later developmental stages, such as puberty, have been completed. This is reflected particularly in the reports of falling sperm counts (Auger et al., 1995) the incidence of testicular cancer (Adami et al., 1994), and decreased in vitro fertilization rates in couples with paternal pesticide exposure (Tielemans et al., 1999). Coupled with the likelihood

33 that the endogenous ligand for ERP is probably not oestradiol, but is an androgenic metabolite, investigation of the androgenic interactions with ERp is critical for the understanding the increasing adverse reproductive effects seen in men in relation to dietary exposure to hormonally active compounds. ERp is also reported to play a central role in oestrogen signalling in normal and malignant prostate epithelial cells (Lau et al., 1999), while the roles played by oestrogens in the transformation from a healthy prostate cell to a cancerous one remain controversial, whereas the role of androgens in this transformation is clearer. However ERp, AR, androstane and CYP7B1 are known to regulate an endocrine pathway in the prostate (Weihua et al., 2002). Peroxisome proliferator-activated receptor a (PPARa) also has a role, it is functionally present in human prostate, but is down regulated by androgens and over expressed in advanced prostate cancer (Collett et al., 2000).

1.4.3. The many biological roles of ERp Data addressing the various biological roles of ERp are being generated from in vivo studies by developing mice with a deleted ERP gene (Gustafsson, 1999; Forster et al., 2002). These ERP knockout (BERKO) mice display significantly reduced fertility in the female, and show the essentiality of ERp for normal functioning ovaries. At the beginning of the estrus cycle in particular, ERP expression is high, but after the

Luteinizing Hormone (LH) surge, ERp is rapidly downregulated. The ratio between ERa and ERp in the ovary is about 1: 9. Another phenotypic characteristic of BERKO animals of both sexes is that the bladder epithelium, the epithelium of the dorsal prostate, the coagulation glands and the urethra show signs of hyperproliferation, suggesting that the growth control of these tissues is impaired if ERp levels are compromised, and that ERP may have a protective role against hyperproliferation and carcinogenesis. For both male and female BERKO mice, reproductive behaviour appears normal, although the males appear to be more aggressive under certain conditions (compared to wild type mice). Kuiper, Gustafsson and coworkers are continuing to investigate the phenotypic characteristics of BERKO animals with particular reference to the cardiovascular system, bone, the immune system, sexually differentiated liver metabolism and reproductive and non-reproductive behaviour. Research into defects in ERp expression and activity should also yield useful data for disease syndromes involving excess androgens in women, particularly as seen in women with polycystic ovarian syndrome (PCN), with

34 symptoms such as secondary male characteristics, menstrual disruption and difficulty conceiving (Franks et al., 1999; Abbott et al., 2002).

1.4.4. ERp and ovarian cancer Steroid hormone expression in ovarian surface epithelial cells (the tissue of origin for over 90% of ovarian cancers) has been observed to be disrupted in ovarian cancer cells taken from postmenopausal women (Lau et al., 1999). In the cancerous cells, the normal healthy coexpression of ERa and ERp mRNA (along with other receptors) were disrupted with an ensuing loss of ERa, PR and AR mRNA, suggesting that these receptors may be responsible for the neoplastic transformation of this cell type, but not

ERp, whose mRNA levels appear to be unaffected by this malignant state. The researchers also suggest that ERP action may depend upon functional ERa levels. Taken together, these findings implicate the regulation of normal ovarian cells by oestrogens (Britt and Finday, 2002), androgens and progestins. The emergence of sex hormone resistance, via down-regulation or mutational inactivation of receptors (including the receptors ER, AR, CAR and PXR described below) may be a key feature of ovarian epithelial transformation, which may lead to diseases such as endometriosis, some cases of sterility and uterine/ovarian cancers.

1.4.5. Dietary exposure: Synthetic oestrogens versus phytoestrogens Phytoestrogens are found in the diet, and are widely available from plant foods, including nitrogen fixing plants and legumes such as clover, lentils and soya, grains such as rye, and lignans such as those found in linseed, and hops (Aldercreutz, 1998; Milligan et al., 1998; Milligan et al., 1999; Cassidy and Faughnan, 2000). They do not rapidly accumulate and are water soluble. They are readily excreted in urine (6-8 hours), and they are probably the source of greatest dietary exposure to environmentally derived oestrogen mimics. On one hand, they can be regarded as a defense mechanism for the plant to reproductively impair herbivore and omnivore endocrine systems and thus be an effective strategy to reduce local herbivore populations, as first noted in the 1950’s when Australian clover was found to be the feminising agent responsible for impairing sexual performance in rams with a consequent dramatic effect in lambing (Bradbury and White, 1954). However, on the other hand scientific evidence indicates that counter defenses have evolved such that adult diets rich in phytoestrogens are associated with a reduced incidence of cardiovascular disease, breast cancer, prostate cancer and osteoporosis.

35 Asian women and vegetarians have a lower than average breast cancer risk, together with a relatively higher excretion of urinary phytoestrogens (Cassidy, 1996; Aldercreutz, 1998; Bingham et al., 1998; Cassidy and Milligan, 1998; Milligan et al., 1998; Milligan et al., 1999; Cassidy and Faughnan, 2000). Some of the possible SHR mechanisms of action of EDCs and phytoestrogens such as genistein are structure specific (figure 1.5), but others combine with structurally diverse ligands, for example hydroxylated PCB 153 (figure 1.6), DDT (figure 1.7), and polybrominated diphenyl ether flame retardant (PBDE) 47 (2,2’4,4’PBDE) (figure 1.7). When the structures of many phytoestrogen compounds are overlaid with the oestrogen structure, they are virtually superimposable, the distance between the hydroxyl groups at each end of the molecules are almost identical. These distances determine hydrogen bond interaction with the receptor. The more structurally diverse synthetic compounds have far greater flexibility, so they can move to, bind with, and activate a wider range of receptors. The less flexible phytoestrogens such as genistein, however, have sub-optimal interactions in the LBD of the ERs (figure 1.6), especially in ERp, giving a partial agonist activity compared to antagonists such as raloxifene (Pike et al., 1999). Synthetic hormone mimics, such as PCBs, dioxins, polybrominated flame retardants (PBDEs) and organochlorine pesticides, accumulate within the food chain and, due to their fat solubility and long half-lives, persist and bioaccumulate in adipose tissue and bone for many years, unlike phytoestrogens. DDT, for example, has a half-life of 10 years. While dietary intakes of synthetic xenoestrogens are in far smaller doses compared to phytoestrogen intakes, they remain available as ligands to the SHR system for a far longer period of time, and the metabolised forms, particularly when hydroxylated have greater affinity for the ERs (Jacobs, 1998; Meerts et al., 2001). The ubiquitous EDCs represent potential risk factors for a number of human cancers and other detrimental health effects. They are present at detectable levels in food and water even those such as DDT, which have been banned in many countries. PCBs, DDT metabolites, dioxins and PBDEs are the predominant POPs contaminants of concern in fatty animal based foods such as oily fish, fish oils, meat and dairy products (Liem and Theelen, 1997; Liem et al., 2000; Jacobs et al 2002a; 2002b; 2002c; Solomon and Weiss, 2002). Phytoestrogens may influence carcinogenesis via their hormonal, antihormonal and antioxidant actions in a beneficial way, when administered throughout life. Substantial evidence indicates that diets high in plant-based foods may explain

36 epidemiological variance of many hormone dependent diseases that are major causes of mortality and morbidity in Western populations. Phytoestrogens such genistein (from soya) have been shown to stimulate cell proliferation and bind to ERs at relatively low levels (i.e. a low affinity) both in vitro (Cassidy, 1996; Bingham et al., 1998;Cassidy and Milligan, 1998; Jacobs et al., 1999) and in vivo (Milligan et al., 1998), and have been reported to have a protective effect if given before puberty (Lamartinière et al., 1995) and an inhibitive effect upon the growth of cancerous prostate cells (Hillman et al., 2001), but a detrimental effect, with increased risk of uterine cancer, if given in infancy (Newbold et al., 2001). EDC pesticides have a more flexible molecular structure compared with phytoestrogens and thus greater affinity with the cellular receptors. It is frequently argued in the literature that the EDC pesticides bind very weakly to the oestrogen receptor compared with natural oestrogens and some phytoestrogens (Safe, 1995; Safe, 1997; Safe et al., 1997). QSAR studies have revealed more fully the importance of understanding differences in ligand binding within the ER (Jacobs, 1998; Jacobs and Lewis, 1999), and these differences need to be related to differences in serum levels of hormones at all stages of human development. EDC pesticides binding affinity in the classical oestrogen receptor (ERa) differs markedly from those of ERp and other receptors (e.g. PXR), and there are additional modes of action taking place. For example, organochlorine pesticides such as o, p ’ DDT and alachlor have been reported to partially mimic oestradiol and function to suppress apoptosis (programmed cell death) in ER responsive cells. Apoptosis appears to play a critical role in the generation and progression of cancer, and is probably regulated by steroid hormones in hormone responsive tissues (Burow et al., 1999). p ,p ’~ DDT has been found to be capable of activating cellular signalling events in ER- negative breast cancer cells (Shen and Novak , 1997a; Shen and Novak, 1997b), so it is highly likely that some organochlorine pesticides may function through other signalling pathways. Chronic exposure to large quantities of phytoestrogens in foods might have a direct binding effect upon the ER and other hormone receptors. Indeed, coumestrol, a very potent phytoestrogen, has uterotropic activity in the immature rat that is typical of the activity of the endogenous oestradiol (Ashby et al., 1999b). Newbold and co-workers (2001) report that in vivo genistein exposure in newborn mice, a time at which the developing organism would normally be using natural oestrogen signals to guide development, increase the risks of uterine cancer in adult life. The amount of genistein

37 used by Newbold et al. was slightly higher than the amount consumed by infants, but was within one order of magnitude of the level of human exposure (approximately 27 mg of genistein per day for infants feeding on formula vs. 50 mg/day in the experiment). Outbred female CD-I mice were treated on days 1-5 with equivalent estrogenic doses of diethylstilbestrol (0.001 mg/kg/day) or genistein (50 mg/kg/day). At 18 months, the incidence of uterine adenocarcinoma was 35% for genistein and 31% for DES. These data suggest that genistein is carcinogenic if exposure occurs during critical periods of differentiation. Other impacts observed from genistein exposure included reductions in fertility during adulthood following exposure as newborn. Inappropriate oestrogens can alter development by changing the intensity of the oestrogen signal whether they are from natural sources or synthetic sources. In the adult, such an intake may also have indirect modulating effects upon associated and related factors (e.g. chaperone heat shock proteins), such that even whilst acting as an oestrogen at certain doses, phytoestrogens such as genistein may be also be acting upon multiple sites having indirect anti-oestrogenic, and anti-cancer effects. A significant but neglected dietary intake avenue of oestrogens occurs in hormone residues from meat treated with sex steroids for growth promotion (Andersson and Skakkebaek, 1999). While the Joint Food and Agriculture Organisation/World Health Organisation (FAO/WHO) expert committee on food additives (JECFA) and the US Food and Drug Administration (FDA) consider that the residues found in meat from treated animals are safe for consumers, they have not considered the sensitivity of healthy prepubertal children to low levels of oestradiol (Andersson and Skakkebaek 1999) or oestrogen mimicking compounds (Howdeshell et al., 1999), and pre and post natal infants to gene imprinting (McLachlan et al., 2001) from this dietary source of oestrogens. Possible adverse effects on human health by consumption of meat from hormone-treated animals cannot be excluded.

1.5. Key nuclear receptors: a multifaceted communication system Members of the same nuclear receptor family share a common heterodimerization partner, retinoid-X-receptor (RXR), and there is cross talk with other nuclear receptors and with a broad range of intracellular signalling pathways. There may be competition for RXR for the dimerization stage of receptor activation of DNA. There may even be a cascade effect, where metabolites produced through the activities of one

38 receptor are specific signalling molecules (and ligands) to modulate the next receptor, providing a link in the nuclear receptor intercommunication web of the body (figure 1.8). Figure 1.5. a) Daidzein and b) Genistein docked in the LBD of the hERa crystal structure (Jacobs, 1998). The photos are oriented to show the respective ligands hydrogen bonding (dashed orange lines) with key amino acids (in blue and red). The ribbon denotes the protein backbone of the receptor.

Figure 1.6. a) Genistein (in green) and PCB 153 (in orange) docked in the LBD of the hERa crystal structure and h) hydroxy PCB 153 docked in the LBD of the hERa crystal structure. (Jacobs, 1998)

Figure 1.7. a) DDT (space-filled) docked in the LBD of the hERa crystal structure and h) PBDE 47 docked in the LBD of the hERa crystal structure. (Jacobs, 1998)

39 Figure 1.8. Schematic diagram showing gender differences in tissue distribution of steroid hormone receptors (collated from the literature).

Brain & CNS E R a E R p Breast E R a E R P Cardiovascular AhRPXRPPARy system Liver ERa ERP PPARa P P A R y ERa ERP AhR PXR CAR I.XR FXR . Lungs ERP A h R P P A R a P Y

Adrenals R E R a E R p a r c a k Ovaries Kidneys E R a E R p E R a E R p Gastrointestinal tract ERP PXR PPARy Uterus E R a ER P hi Urogenital tract A liR P X R E R a E R p Bladder E R p Bone E R a ER P . Testes ERa ERP AR I AR

Prostate

ERP PPARa AR

Activation of an NR may either increase or decrease the synthesis of that or another receptor. A given receptor population is not constant and may be influenced by the concentrations of circulating ligands and also the state of the receptor population. To demonstrate how important it is to consider the spectrum of receptors when assessing the dietary input of endocrine disrupters, a brief description of selected receptors and their known interactions with ERs follows. The interactions are far more complex than the brief descriptions may imply.

1.5.1. Aryl hydrocarbon Receptor (AhR) The AhR, a member of the PAS (Per-Arnt-Sim) family of nuclear regulatory basic helix-loop-helix proteins, has been detected in nearly all vertebrate groups examined (Hahn et al., 1998; Hahn, 2002a) and appears to have a fundamental role in cellular physiology, neurodevelopment and circadian rhythmicity (Poellinger, 2000). Predominantly found in hepatocytes, but also in breast cancer cells (Nguyen et al., 1999), and the lung (Song et al., 2002). The AhR regulates the expression of a number of genes, including cytochrome P450 1 Al, 1A2, IBl and glutathione S transferase M (GSTM) in a

40 ligand dependent manner. AhR is also upregulated during cell division, and in patients with cardiomyopathy (Mehrabi et al., 2002). Exposure to 2,3,7,8-tetrachlorodibenzo-p- dioxin (TCDD), the most potent AhR ligand known, results in a wide variety of species- and tissue-specific toxic and biological responses. Animals treated with 2,3,7,8-TCDD have developed abnormalities in several organs including the thyroid, thymus, lung and liver, immune and endocrine function. Wasting, lethality and induction of gene expression by TCDD have also been shown to be AhR dependent (Abbott et al., 1999; Diliberto et al., 2000). Within the cytosol of the cell, the AhR is associated with a heterodimeric transporter protein partner, termed the Aryl Hydrocarbon Receptor Nuclear Transporter protein (ARNT). The unliganded AhR may also act through other mechanisms by being phosphorylated to key regulatory proteins such as hsp90, which appears to be required for maintaining the receptor in a non- activated ligand binding conformation (Pongratz et al., 1992, cited in Poellinger 2000), as with the ERs, and other proteins such as p37, AIP, XAP2, src, rel, and Rb (Bimbaum, 2000; Bimbaum and Tuomisto, 2000). AhR knock out mice display reduced fertility, reduced viability, and liver and immune deficits (Diliberto et al. 2000) in some independently generated line mice, but not others (for a review see Poellinger 2000). The best characterised high affinity ligands for the AhR include a variety of ubiquitous man-made toxic and hydrophobic chemicals, including halogenated aromatic hydrocarbons (HAHs), such as the polychlorinated dibenzo-p-dioxins, dibenzofurans, and biphenyls, and polycyclic aromatic hydrocarbons (PAHs) such as benzo(a)pyrene, 3- methylcholanthrene, benzoflavones and other chemicals. Certain dietary indole derivatives, such as indolo[3,2-6]carbazole, appear to bind to the AhR with the same affinity to that of 2,3,7,8-TCDD (Gillner et al., 1993), but a physiological receptor ligand has not been identified yet. Weaker ligands include diaminotoluene, omeprazole, brevetoxin, indole carbinols (found in cruciferous vegetables), and endogenous weak ligands such as bilirubin (a hydrophobic heme degradation product metabolised in the liver), biliverdin (Sinai and Bend, 1997; Phelan et al., 1998), and water soluble metabolites of tryptophan, tryptamine and indole acetic acid (Heath-Pagliuso et al., 1998), equol, a daidzein metabolite and compounds produced endogenously for prostaglandin pathways, including arachidonic acid, prostaglandins and lipoxin A4 (Schaldach et al., 1999; Denison et al., 2002). More recently, a further endogenous ligand, with an indole thiazole structure, ITE (2-(l’H- indole-3’carbonyl)-thiazole-4-

41 carboxylic acid methyl ester) has been proposed, after isolation from porcine lung, it appears to have relatively high affinity with the AhR, with similar activity to BNF (Song et al., 2002). It is possible that this compound is not endogenous; it may have formed in a reaction with the extraction and purification acid treatment combination. Clearly there is a wide gamut of structural diversity of AhR ligands, many of which are ligands for other NRs. Many of these ligands activate other receptors, dioxins for example have anti- oestrogenic activity in the ER, while equol is strongly oestrogenic. The endogenous ligands hint further at the developmental role of AhR, as persistent CYPlAl and CYP1A2 gene expression has been observed in congenitally jaundiced Gunn rats (Lorenzen and Kennedy, 1993, cited in Phelan et al., 1998). Both enzymes play a role in the oxidative metabolism of bilirubin, decreasing its toxicity and enhancing its elimination, an important detoxication role in the newborn infant where jaundice commonly occurs due to less developed liver function and excretion. Increased AhR induction is observed with increased oxidative stress, and it is probable a route for antioxidant activity is through the AhR to down regulate oxidative signaling.

1.5.2. Pregnane X Receptor (PXR) An important requirement for homeostasis is the detoxication and removal of endogenous hormones and xenobiotic compounds with biological activity. PXR (Lehmann et al., 1998a; Kliewer et al., 1998; Blumberg et al., 1998) is involved in activating the expression of several P450 detoxifying enzymes, including CYP3A4 in the adult and CYP3A7 in the fetus in response to xenobiotics and steroids (Pascussi et al., 1999). CYP3A4 is the major human hepatic P450, and has been suggested as being involved in the metabolism of over 60% of drugs in clinical use (Maurel, 1996). Wide species differences exist in PXR activation due to differences in the LBD (Jones et al., 2000). The major site of PXR expression is in the liver cells and the gastrointestinal tissues, but they are also present in both normal and neoplastic breast tissue. Indeed a statistically inverse relationship between the level of PXR mRNA expression and ER status has been observed (P=0.04) (Dotzlaw et al., 1999). PXR can be activated by a variety of chemically distinct ligands, in a species dependent manner, including endogenous hormones such as oestrogen, pregnenolone (figure 1.7), and progesterone, their synthetic derivatives such as pregnenolone 16a carbonitrile (PCN), bile acids.

42 rifampicin, dexamethasone, corticosterone, spironolactone, phénobarbital, antiretroviral drugs, simvastatin (reviewed in Plant and Gibson, 2003), vitamin D 3 (l,25-(OH)2-D3), and plant-sourced ligands such as hyperforin (Moore et al., 2000a), coumestrol and genistein (Chapter 4). Organochlorine pesticides and the synthetic food contaminants, phthalic acid and nonylphenol (Masuyama et al., 2000) with known ER activity (Harris et al., 1997) also activate PXR. Many of these compounds activate other receptors from liver receptors such as LXR and FXR, to the ERs, AhR and the vitamin D receptor. The percentage inhibition in competition binding assays for the pesticide trans-nonachlor is close to one hundred percent, with seven times greater activation than the activation by PCN in human PXR transfection assays (Jones et al., 2000). PXR is also essential in mediating transcriptional activation of CYP3A by environmental contaminants (Schuetz et al., 1998). It appears that there is a specific regulatory pathway where the accumulation of steroidal PXR ligands, including xenobiotics, results in increased CYP3A transcription and steroid catabolism, possibly providing the route for excess steroids to be eliminated from the body. So not only is PXR a xenobiotic sensor, it is also a key player in the regulation of steroid homeostasis, steroid metabolism, by involvement in the expression of steroid hydroxylases and detoxication (Kliewer et al., 1999). By implication, EDCs affecting PXR may also have indirect effects upon the regulation of steroid homeostasis. Some NRs have quite different, unanticipated mechanisms of action which EDCs may also influence directly or indirectly, as with the constitutive androstane receptor.

1.5.3. The Constitutive Androstane Receptor (CAR) Present in the liver and intestine, CAR belongs to the subfamily of NRs that includes the Vitamin D receptor (VDR), and PXR. CAR similarly heterodimerises with RXR (Baes et al., 1994; Choi et al., 1997; Honkakoski et al., 2003) and interacts with, and is inhibited by, two endogenous steroids, androstanol and androstenol, via a mechanism that involves a widely expressed nuclear receptor coactivator, SRC-1. Both PXR and CAR are involved in the expression of steroid hydroxylases, and consequently regulate key steps in steroid metabolism (Kliewer et al., 1999). However, unlike most NRs, including PXR and ER, CAR is constutively active in nature and is sequestered in the cytoplasm of non-exposed cells. Ligand activation of CAR inhibits receptor- dependent gene transcription. It does this through the ligand-independent recruitment of

43 transcriptional co-activators. CAR functions in a complex manner, a manner opposite to that of the conventional NR pathways. The physiological relevance of CAR is unknown as yet, although evidence is accumulating that CAR is a drug-activated receptor integral to the elimination of drugs mediated by P450s. There are significant sex differences in plasma androstane levels and CAR has been recently implicated as a transcriptional regulator of the gene governing the steroid hydroxylase CYP2B (Forman et al., 1998; Honkakoshi et al., 1998). The PB responsive DNA elements (PBREM) reporter gene has been shown to be activated by CAR (Honkakoshi et al., 1998). CAR is highly responsive to phénobarbital (PB) type compounds, including certain PCBs, pesticides, drugs, solvents and other xenobiotics, such that endogenous CYP2B is responsive to PB and other inducers only in the presence of suppressed CAR. At the systemic level, hormonal imbalances that affect androstanol and androstenol levels may impart a perturbation at the molecular level, preventing or inducing the inhibition of CAR and consequently the steroid hydroxylase CYP2B.

1.5.4. Human peroxisome proliferator activated receptor (hPPAR) The Peroxisome Proliferator-Activated Receptors (PPARs) are a family of orphan receptors with fundamental roles in regulating energy balance (Johnson et al., 1997; Blumberg and Evans, 1998; Uauy et al., 2000; Willson et al., 2000; Bar-Tana, 2001). A number of prevalent metabolic disorders such as obesity, atherosclerosis and type 2 diabetes are associated with a shift in this balance. The PPARs are activated by chemicals which elicit increases in the number and size of peroxisomes when administered to rodents (Kliewer et al., 1999). Three closely related PPARs, a, (3/8 and y, are found in the liver, kidney, heart, haematopoietic and adipose tissue, but with different expression patterns. PPAR a is found in liver, kidney, heart and muscle, PPAR (3/8 is expressed in nearly all tissues and

PPAR y is expressed in fat cells, the large intestine, and monocyte lineage cells (Memon et al., 2000). They each play key roles in lipid metabolism and homeostasis; they are responsible for CYP4A induction, peroxisomal enzyme induction and hepatic peroxisome proliferation. PPARa has a central role in hepatogenesis, and PPARy a central regulatory role in adipogenesis (Willson et al., 2000). PPARa regulates key steps in lipid and fibrate metabolism and this receptor is the molecular target for naturally occurring plant fatty acids (pristinic acid and phytanic acid) present at physiological concentrations (Zomer et

44 al., 2000), long chain polyunsaturated fatty acids (LCPUFA), eicosanoids (Uauy et al., 1999), and peroxisome proliferators, which include drugs such as the fibrates (used widely to lower high triglyceride levels, a risk factor in coronary heart disease) and synthetic chemicals such as the phthalate ester plasticisers, and pesticides (Kliewer et al. 1999). PPARy ligands include fatty acids, prostaglandins and the antidiabetic thiazolidinedione (TZD) drugs (Memon et al., 2000; Willson et al., 2000). Pristinic acid and phytanic acid are branch-chained fatty acids obtained through the diet from the chlorophyll in plants. Present at micromolar concentrations in healthy individuals, they can accumulate in a variety of inherited disorders. Potent binding of pristinic acid and phytanic acid in the LBD of PPARa (Zomer et al., 2000) indicates a primary mechanism for metabolising these dietary fatty acids. The PPARs have a far larger ligand binding pocket than the receptors so far discussed, and there are differences in the shape of each PPAR ligand binding pocket, giving broad ligand specificity on a structural basis (see Chapter 3). Another factor to be considered is modulation through cross-talk between PPAR and other nuclear receptors/signalling molecules. For example, thyroid hormone suppresses hepatic PP responses and exhibits inhibitory cross-talk with PPARa, due in part to competition between the thyroid receptor and PPAR for their common heterodimerization partner RXR (Miyamoto et al., 1997) (See figure 1.1.). Rosiglitazone occupies a fraction of the available LBD space in PPARa, and even less than that in PPARy, particularly the rosiglitazone TZD head-group, and thus comparatively reduced selectivity is observed. This has also been observed for different ligands in the PPAR family, and is a clear descriptor for PPAR selectivity.

1.6. Co-modulators that enhance or suppress transcriptional activity

NRs interact with a group of novel nuclear proteins, including steroid hormone receptor coactivator-1 (SRC-1), Steroidogenic factor 1 (SF-1), and receptor interacting protein 140 (RIP 140) (Luo et al., 1999). RIP 140 has been shown to interact with the ER LBD in the presence of oestrogen, amplifying and potentiating ER dependent transcriptional activity (Sheeler et al., 2000). SF-1 is a key regulator of the tissue specific expression of the cytochrome P450 steroid hydroxylases, and is essential for the embryonic survival of the primary steroidogenic organs and the regulation of reproductive function at all three levels of the hypothalamic-pituitary-gonadal axis from

45 the earliest stages of gonadogenesis (Luo et al., 1999). In vitro studies suggest that exposure to o,p -DDT (and other oestrogenic compounds) at sufficient concentrations or in the presence of an ER coactivator could have deleterious effect upon normal cell function due to the untimely activation of oestrogen regulated genes (Sheeler et al. 2000). Similarly, in a yeast two hybrid protein interaction assay, PXR has been observed to interact with SRCl and RIP 140 (Masuyama et al., 2000). Cellular efflux transporter proteins present in the intestine and liver, such as Organic Anion Transporting Polypeptides (e.g. OATP-C, OATP-1), P-glycoprotein (P-gp) and multidrug resistant proteins (e.g. MDRl, MDR2 ) can have major effects upon PXR ligand availability by interacting with PXR pathways. It has also been suggested that the differences in potency of phytoestrogen ligands in binding studies for ERs, especially ERp, compared with the lower potency detected in whole cell assays may be a consequence of interactions with binding proteins. Perturbation of the endocrine system hormones can also affect certain transport proteins, many of which are gender specific. However as part of the complicated nature of hormonal interactions it is relevant to note that sex steroids are known to regulate certain excretory transport proteins, such as the mRNA levels of Oats in in vivo rat models, in a gender specific manner. Androgens have been reported to increase rat organic anion transporter 1 (Oatl) mRNA in kidney, while the female GH (growth hormone) secretion pattern increases a different Oat, Oat2, with both androgens and GH regulating Oat3 mRNA levels in the liver (Buist et al., 2003). The implications of the interactions with transporting proteins, such as OATP, P- gp, MRDl and MRD2 are significant with respect to ligand availability and this is discussed further with particular reference to PXR in Chapter 4.

1.6.1. Cytochrome P450 enzymes: CYP17A1/19A1 (Aromatase) Aromatase is a key P450 in the production of oestrogen, catalyzing the conversion of androgens, androstenedione and testosterone via three hydroxylation steps to estrone and estradiol (Martucci and Fishman, 1993; Brodie et al., 1999). Aromatase is expressed in many tissues, including the ovaries, the testis, the placenta, the brain, and adipose tissue of the breasts, abdomen, thighs, buttocks and bone osteoblasts. Phytoestrogens are potent inhibitors of aromatase (reviewed in Whitten and Patisaul,

2001).

46 1.6.2. Aromatase, breast cancer and hormonal disruption in women Adipose tissue is the major site of oestrogen biosynthesis in postmenopausal women, with the local production of oestrogen in breast adipose tissue implicated in the development of breast cancer. Inhibition of this pathway is one method that is exploited pharmacologically in ER positive breast cancer treatments to inhibit oestrogen production. Another approach is to inhibit oestrogen action by antioestrogens, which interact with the oestrogen receptor in the tumours. Adipose tissue is also the most significant storage site in the human body for environmental pollutants such as PCBs and DDT, compared to blood and the liver. lida et al. (1999) reports measured values for PCB 118, for instance, as follows; blood: 59.2 pg/g tissue, liver: 1692 pg/g tissue, and fat: 13,896 pg/g tissue. Recently, rosiglitazone and troglitazone, compounds known to activate another receptor, PPARy (a key factor in adipocyte differentiation), used in the control of insulin -resistant diabetes, have also been found to inhibit aromatase expression in human breast adipose stromal cells (Simpson et al., 2000; Rubin et al., 2000). They have also been observed to stimulate breast cancer cell lines (Mueller et al., 1998). Developmental inhibition of the aromatase pathway can give rise to polycystic ovarian syndromes (PCN) (Franks et al., 1999). EDCs that may affect this pathway include those that have been observed to bind with the ER. Figure 1.7 indicates current knowledge of the gender differences in the distribution of steroid hormone receptors. It is likely that there are also polymorphisms to consider, as inter-individual variations in gene sequences (Masahiko and Honkakoski, 2000), suggest that individuals may vary in terms of amount and function of all the receptors discussed herein. The hierarchy of ligand activation differs between the receptors as well as for receptors isolated from different species, and in many instances molecules that were previously regarded as metabolic intermediates are in fact “intracrine” signalling molecules within tightly coupled metabolic pathways for altering gene expression.

1.7. Molecular modelling Using computer generated molecular models of the ligand receptor interaction, one can examine the likely mechanisms of action of NRs. By modelling such interactions and evaluating the activity of potentially hormone mimicking materials by way of Quantitative Structure Activity Relationships (QSARs), it is possible to examine and

47 estimate the potency differences within these compounds in human hormone receptor ligand-interactions. This thesis aims to improve the development of this area, examining the differences in affinity of NRs for a given ligand. In relation to experimental data, it can provide a useful tool in a complex situation of considerable public concern. Whilst used extensively by the pharmaceutical industry, QSARs are under-utilised by nutritionists, food scientists and environmental toxicologists. By using crystal structure co-ordinates it is possible to examine the mode of binding of selected chemicals and calculate the values for the ligand binding affinity, when compared to in vitro data. Thus, a measure can de derived of the potency and action of a chemical in a given receptor, compared with another chemical, at the molecular level. This information can be related to physiological and epidemiological health effects observed, providing an essential part of the whole picture of endocrine function and dysfunction. Data from animal and in vitro studies provide convincing evidence for the potential of phytoestrogens in influencing hormone dependent states, although many in vivo studies can be confounded by the oestrogenic activity of components of the rodent diets (Boettger-Tong et al., 1998; Ashby, 2001), housing and species (Anon, 1999). While the clinical application of high phytoestrogen human diets is in its infancy, data from preliminary studies suggest beneficial effects of importance to women and men’s health. Chapter 3 reports the generation by homology modelling of the three-dimensional (3-D) structures of the ligand-binding domain (LBD) of several interrelated human nuclear and steroid hormone receptors. These are the oestrogen receptor(3 (hER(3), the pregnane-x- receptor (PXR) (figure 1.7), the Ah receptor (AhR) and the constitutive androstane receptor (CAR). They were produced by homology modelling from the human oestrogen receptora (hERa) crystallographic coordinates ( figures 1.3 and 1.5) (Brzozowski et al.,

1997) as a template, together with the amino acid sequences for hERp (Mosselman et al., 1996), PXR (Lehmann et al., 1998), AhR (Burbach et al., 1992) and CAR (Forman et al., 1998), respectively. The selective endogenous ligand was docked interactively within the putative ligand binding site using the position of oestradiol in hERa as a guide, and the total energy was calculated. In each receptor model a number of different ligands known to fit closely within the ligand binding site were interactively docked and binding interactions noted. Specific binding interactions included combinations of hydrogen bonding and hydrophobic contacts with key amino acid sidechains, which varied depending on the nature of the ligand and receptor concerned. With the AhR an

48 important facet of the ligand binding process involves a n-n stacking interaction between benzene rings on selected PCB molecules and aromatic amino acid residues in the AhR ligand binding site. Differences in the binding affinities for the same ligands in ERa compared to ERP are under investigation, further to the report published by Pike et al. (1999). We also produced PPARa by homology modelling using the human PPARy (hPPARy) LBD crystallographic coordinates reported by Nolte and co-workers (Nolte et al., 1998) as a template together with the amino acid sequence for hPPARa (Sher et al., 1993; Lewis and Lake, 1998). Models of selected ligands were docked interactively within the putative ligand binding site using the position of rosiglitazone as a guide, and the total energy was calculated. This work comprises Chapter 3, it was conducted prior to the publication of the PPARa crystal structure, and has been published (Jacobs and Lewis, 2000; Jacobs et al., 2003a). The models provide a useful tool in unravelling the complexity in the physiologic response to xenobiotics, by examining the ligand binding interactions, differences and cross talk between the steroid hormone receptors and activation or inactivation by their ligands. Nuclear receptors are important drug targets for intervention in disease processes. Exogenous compounds that target these receptors can therefore disrupt both normal and abnormal functioning of these key metabolic pathways. While environmental hormone mimics contribute to detrimental health effects by activating certain receptors and disturbing normal function, there are therapeutic uses from both dietary and pharmacological treatment for abnormal functioning of the hormone pathways and hormone dependent diseases. The promiscuous ability of many cellular receptors to bind to ligands of different chemical structures is a major mechanism by which dietary ligands, including environmental chemicals, can enhance or inhibit receptor activation, as well as being the basis for the pharmaceutical development of receptor regulating drugs. However, endocrine effects are mediated through multiple sites of action, and EDCs are able to alter endogenous hormone pathways, as well as acting directly upon receptors. Fetal exposure to EDCs at critical time points will have harmful health effects that do not become evident until puberty and adulthood, and again there will be gender differences and not just age-related susceptibilities. Vulnerability to the adverse effects of EDC exposure and protective effects from phytoprotectants such as phytoestrogens escalates during dynamic periods of growth and development. Children may metabolize

49 compounds faster, but they detoxify more slowly and have greater body burdens due to higher dietary intakes in relation to body size, compared to adults. The effect of ligand binding to a steroid hormone receptor needs to be considered as a synthesis of the entire endocrine system, over time, according to gender and reproductive status and with due consideration for diurnal rhythms and environmental factors.

1.8. Conclusion The nuclear receptors modelled display a spectrum of ligand specificities, ranging from the highly specific, as seen in CAR, which binds 5a-androstan-3a-ol

(androstanol) but not 5 a-androstan-3p-ol, and in PPAR selectivity, to the highly non specific, such as hPXR, which is very flexible, and can bind with a large number of wide ranging molecules, from rifampicin to steroidal structures. They also display a spectrum of binding modes within the LED, from hydrogen bonding with different key amino acids (as observed in all the receptors) to n-n stacking, as seen in the AhR and also binding outside the LED, as seen with antagonists such as tamoxifen in ERa. The receptor models can be used to explore modes of binding of different ligands in the LED, and to generate Quantitative Structure Activity Relationships in conjunction with experimental ligand binding data from the literature. The identification and in silico assessment of the different ligands both in isolation and within the receptor models will add to a better knowledge of their specificity and may help explain the selective action of oestrogens, phytoestrogens and EDCs in different tissues, and expand knowledge on the role of nutrition and pharmacology for therapeutic intervention in various endocrine, developmental and energy homeostasis functions that involve the ERs, AhR, CAR, PXR and the PPARs. Pervasive EDCs that target these receptors, disrupting normal functioning of key metabolic pathways, contribute to detrimental side effects in just the same way as observed pharmacologically, with the inadvertent activation of orphan receptors by certain drugs. Without considering the whole steroid hormone receptor super family, current EDC research can only be considered piecemeal, insufficient and ineffective for risk assessment and not representative of the hormonal effects of EDCs. The full range of effects of EDCs, whether of synthetic or plant origin, from exposure routes to health outcomes, upon women, men, and subsequent generations can only be unravelled with an integrated multidisciplinary approach.

50 C hapter 2

Food Animal - Environmental Case Study

51 C hapter 2

Food Animal - Environmental Case Study

2.1 Introduction

A major route of exposure for compounds that are ligands for the NRs, as reviewed in chapter 1, is dietary. This chapter uses a case study format to examine the dietary routes of exposure to environmental contaminants that are also ligands for SHRs and are key suspected endocrine disrupters. The salmon aqauculture food chain is the subject of this case study. The POPs levels are determined in mainly farmed salmon, feed and fish oils. In recent years, the salmon industry has grown significantly in Europe, and salmon is becoming increasingly available to the consumer, so is a common dietary source for many European consumers. Until very recently, and at the time of this investigation was conducted and published, there were very little data available on contamination pathways in aquaculture systems, such as that for farmed salmon farmed fish, although there is extensive literature documenting the bioaccumulation of persistent organic pollutants in the marine environment. Data are available on the concentrations of polychlorinated biphenyls (PCBs), and other organochlorine contaminants such as polychlorinated dibenzo-p-dioxins (PCDDs) and furans (PCDFs), in fish from wild stocks, where exposure is associated with chronic contamination due to leaching of these chemicals from treated or contaminated soils into surface waters and the global distribution and deposition by atmospheric transport (North Sea Task Force, 1993). With farmed fish, potential hazards associated with the ingredients and additives used in aquaculture feed were not considered significant until very recently (World Health Organisation, 1999; European Commission, 2000; Scientific Committee on Food, 2000a; Scientific Committee on Food, 2000b). Fish oil is a by-product of the fish meal manufacturing industry and comes from many different parts of the world. As chlorinated compounds accumulate in the lipid compartment of the animal, oil extracted from fish caught in polluted waters may be contaminated with chlorinated hydrocarbons (Jacobs et al., 1997; Ministry of Agriculture Fisheries and Food, 1997; Jacobs et al., 1998). Existing data on the levels of PCBs, PCDDs, PCDFs and organochlorine pesticides consumed in the UK have been mainly derived from total diet surveys and surveillance data for specific food types (Ministry of Agriculture Fisheries and Food, 1996; Food Standards Agency, 2000). These data indicate the presence of elevated organochlorine contamination of farmed fish. Salmon and trout samples tested in 1997 were

52 found to contain residues of PCBs at concentrations between 23 and 620 ng/g lipid weight (Veterinary Medicines Directorate, 1998; Ministry of Agriculture Fisheries and Food / Department of Health, 1998). DDT and HCB were the predominant organochlorine pesticides detected in these samples. The widely used flame retardants, polybrominated diphenyl ethers (PBDEs), are also persistent and ubiquitous organic pollutants (POPs) that biomagnify and may have endocrine disrupting effects (World Health Organisation, 1994; De Boer et al., 1998; Bergman, 2000; Damerud et al., 2001; Branchi et al., 2003). Concern with the risks to human health, particularly infants, is increasing due to observations of increasing PBDE concentrations in human breast milk (Branchi et al., 2003; Meironyté et al., 1999), although available data, including that reported here, suggest that current levels of PBDEs in the UK salmon industry are an order of magnitude lower than those of PCBs. While PBDEs are in extensive production and use, existing data on the levels of PBDEs consumed in the UK are very limited, although surveys to determine the distribution and fate of PBDEs in marine life in waters around the UK have been recently reported (Allchin et al., 2000; Boon et al., 2002). The determination of these contaminants in fish and identification of the original sources of the contamination is, therefore, important for dietary exposure assessment and the protection of public health, particularly in view of the increasing availability to the consumer of farmed salmon. The farmed salmon industry is rapidly expanding; it has trebled in production from 1990 to 1998 (FAO Fishery Information, 1998). The three major producers are Norway, the United Kingdom and Chile, although other countries, notably the US, are also increasing the growth rate of their salmon farming industries. While the use of fish meal in aquaculture feeds for farmed fish and shrimp continues to increase (FAO Globefish, 2000), it should be noted that twice as much fish meal is used in poultry and pig feeds (FAO Globefish, 2000; FAO Fishery Information, 1998). However, the feed costs to produce farmed salmon appear to be significantly greater on a live weight basis (Naylor et al., 2000). The proportion of fish meal supplies used for farming fish rose from 10% in 1988 to 33% in 1997 (FAO Globefish, 2000; Forster, 1999). The use of fish meal and fish oil of European origin are considered to be of critical concern when used in diets for farmed fish and other food producing animals (Scientific Committee on Food, 2000a; Scientific Committee on Food, 2000b). The aquaculture sectors most dependent on the use of fish oil (commonly herring oil) as a source of dietary lipids are the salmon and marine finfish industries. Fish oil is now being strongly promoted by the aquaculture industry as an aquaculture additive in a ‘High Energy diet’ that encourages fast

53 growth of fish whilst minimising protein utilisation as an energy source, and improves the nutritional value to humans through raised omega-3 fatty acid levels. Its use has increased significantly since 1995, from less than 30% to 33% and even 36.3% of the total diet for Salmon, in recent years (Royce, 1996; Bell, 1998; BOCM, 1998; Trouw Aquaculture., no date). The change in use of fish oil is indicated in Table 2.1. Table 2.1. Manufacturers prediction of UK consumption of fish oil (Barlow 1995).

Thousand tonnes

1995 2010

Hardening 1,050 500

Farming 350 750

Supplements/Industry 100 200

Refined oil 0 50

TOTAL 1,500 1,500

Section 2.2 presents the results of analysis of farmed Scottish salmon samples for PCDDs, PCDFs and selected coplanar PCBs. These values are compared with those obtained from salmon reported in other studies. The potential contribution of fish oils to the bioaccumulation of PCDDs and PCDFs and PCBs in farmed fish is discussed, together with suggestions on how to reduce these contaminant levels whilst also achieving nutritional and environmental benefits in an economically viable way for both industry and the consumer (Jacobs et al., 2000; Jacobs et al., 2002a). To our knowledge, section 2.3 presents the first independently published assessment of levels of organohalogenated compounds in the farmed salmon food chain (Jacobs et al., 2001a; 2001b; Jacobs et al., 2002b). Farmed European Salmon, feed and fish oil samples were analysed for a wide range of PCBs, organochlorine pesticides and PBDEs. This chapter builds on the initial study, presented in section 2.2, by including data on a further five salmon samples, eight aquaculture feed samples and five fish oils. These values are compared with those for fish and feed data reported in the literature, and the potential contribution to the bioaccumulation of PCBs, organochlorine pesticides and PBDEs in farmed salmon is

54 discussed. Since publication, the findings of these studies have been confirmed through testing by the UK Food Standards Agency and extensive testing conducted by the aquaculture industry (Nutreco, 2003). Section 2.4 is a case study examining the organhalogen contaminants in dietary omega- three rich supplements, which as with animal feeds is considered a major growth area by the fish oil industry (see Table 2.1). Multivariate analytical techniques are utilised to examine time trends in the contaminant loads between samples tested in 2002 compared to eight years previously (Jacobs et al., 1997). Section 2.5 discusses the changing contaminant patterns observed and the issues, implications and further data needs and tools required to provide realistic dietary exposure estimations on POPs contaminants, and realistic cost-benefit scenarios.

55 2.2 Investigation of polychlorinated dibenzo-p-dioxins, dibenzofurans and selected coplanar biphenyls in Scottish farmed Atlantic Salmon (Salmo salar).

2.2.1. Abstract This study determines the levels of PCDDs, PCDFs, and selected coplanar PCBs in Scottish farmed Atlantic salmon {Salmo salar). This study initiates an investigation of the changing dietary pathways of fish oil and fish meal, and bioaccumulation of organochlorines such as dioxins in fish tissue destined for human consumption, which is developed further in Chapter 2.2. where the same salmon samples and aquaculture feeds and oils are further analysed for a wider range of organhalogen pollutants. Twelve farmed and wild Scottish Atlantic Salmon fish were obtained from retail suppliers, producers and Stirling University, in Scotland during January 1999 for determination of 17 PCDDs and PCDFs, and 7 non-ortho and mono-ortho PCBs. The study confirms previous reports of relatively high concentrations of PCDDs, PCDFs and, especially, PCBs in farmed Scottish Salmon. The results indicate that high consumption of salmon, particularly by children under 5 years, could lead to intakes above the tolerable daily intake (TDI) and temporary tolerable weekly intake (tTWI) for these chemicals, especially PCBs, in combination with mean or high level intakes from the UK diet. These results suggest further investigation of farmed salmon and salmon feed, including feed fortified with fish oil and feed fortified with selected vegetable oils, are warranted.

2.2.2. Materials and methods 2.2.2.1. Sites and sampling Nine British salmon samples (of which seven samples were from individual fish, and two samples were composites of two fish each) and one Norwegian salmon {Salmo salar) sample that enter the European fish market were analysed. Salmon fish samples of variable age, both farm raised and wild, were obtained from eight different sites in January 1999, and were essentially samples of opportunity. The timing of the sample collection was in accord with the seasonal variation in the oiliness of fish fed diets containing greater than 28% lipid, as the exudation of free oil affects the ease of handling and preparation of the fish samples. The sampling times to avoid are September to mid November, and the month of April (Royce, 1996; Bell et al. 1998). The fresh samples were obtained from commercial farms, supplying the local and national markets and national supermarket chains, and the Glasgow fish market. They were wrapped in polyethylene bags and packed in ice (for less than 3 hours) prior to

56 being placed in a deep freeze. Frozen samples, originally sourced from a supermarket, Scottish salmon farms and Norway, were kindly provided by Dr S. George and Dr. G. Bell, of Stirling University, Scotland. Upon collection, each sample was given a unique laboratory reference number, and the sample details logged into a database. Laboratory reference numbers and GC-MS data file codes are included in the results tables. When all samples were collected (3 days), and frozen, they were packed in ice and transported directly by M. Jacobs to the laboratory where they were stored at -20°C. Random sampling was not possible, but unbiased representation of the situation of interest was achieved, by obtaining fish from the Northwest Scottish Highlands, the Western Scottish Highlands, the lowlands surrounding Stirling, plus a wild fish from the Scottish border with England, two samples for which no reliable information was available (but one may have been wild), and a Norwegian sample for which no information was available. Whole body weights of the fish ranged from 400g to 5.2kg, the fish ranged in age from 1 year to 3+ years. Table 1 gives sample details for the salmon and it is important to note that the ‘wild’ fish may not have been genuinely wild, but were probably farm escapees (S. George, personal communication). Figure 2.1 indicates the spread of the sample sites in Scotland. Sample identity numbers beginning 213 refer to the sample codes allocated by the US ERA analytical laboratory. The sample identity numbers in brackets beginning M refer to the relative duplicate sample analyzed in collaboration with Adrian Covaci at the University of Antwerp, reported in section 2.3.

L.Tain: 213-02; 213-04; 213-05 (M il, M24, M25)

Fort William

L. Leven: 213-07 ( Ml 2) R.Devon:213-09 '

Stirling: 213-10; 213-11; 213-12, (M14, M28, M31)

Allen: 213-06

57 2 2 2 2 . Sample preparation The samples were thawed, filleted, skinned and the epaxial muscle homogenised before being subdivided into smaller replicate portions of approximately 100 grams. The portions were weighed, stored in tightly sealed polythene bags and frozen at -20°C. Two samples consisted of homogenised samples of two fish each (from the same source and of the same age and size) to ensure similar sample quantities of tissue. One set of the frozen samples remained at the laboratory for subsequent analysis. The other set of samples were packed in dry ice and taken directly by courier to the US EPA/OPP laboratory the following day, where they were inspected, logged-in and stored at -60°C prior to analysis.

2.2.2. j. Analytical Procedures Many of the sample preparation procedures, analytical techniques, and quality control strategies described in this paper parallel those defined in U.S. EPA’s Method 1613 (Ferrario et al., 1996; Ferrario et al., 1997).

2.2.2.4. Sample preparation for analysis One hundred grams of homogenised muscle tissue were extracted using a mixture of hexane/methylene chloride/ 2-isopropanol to obtain the lipid and an aliquot representing a 7-8 g subsample (when possible) taken for analysis. An aliquot was also removed to determine the percent lipid. The crude lipid extract was fortified with 13C labelled analogs of the various 2,3,7,8-CL substituted congeners and selected coplanar PCBs. The crude extract was cleaned up using acid/base-modified silica gel, alumina, and graphitized carbon column chromatography. The eluent was then concentrated, fortified with 13C labelled internal standards, and analysed by high-resolution gas chromatography/high-resolution mass spectrometry (HRGC/HRMS). Though the methodology used in this study followed U.S. EPA Method 1613 closely, there were several significant changes, which should be noted: (1) Standard solutions were prepared at lower concentrations than suggested in 1613. Method 1613's lowest calibration standard contains 500 fg/pL TCDD and 100 pg/pL of 13C-labeled surrogates. This study's lowest calibration standard contained 50 fg/pL TCDD and 5 pg/pL of 13 labelled surrogates. The samples in this study fortified with 13C-labeled surrogates to deliver 5 pg/pL, where Method 1613 delivers 100 pg/pL from the same 20-pL final volume. This lower 13C surrogate fortification level at 10 ppt allowed for a more realistic

58 approximation of the actual recovery of native analytes at low part per trillion (ppt) levels and better approximates the behaviour of trace levels of natives during extraction, concentration, and chromatography. (2) A DB-5MS column was used in place of the DB-5 specified by Method 1613. The DB- 5MS has a superior separation of the 2,3,7,8-TCDD from the other tetra isomers and better resolves the 2,3,7,8 -Cl-substituted dioxins and furans. It should be noted that the elution order of the 1,2,3,7,8,9-HxCDF and the 2,3,4,6,7,8-HxCDF is reversed on the DB-5MS relative to the DB-5. (3) Method 1613 uses an AX21 and Celite 545 mixture for the graphitized carbon column cleanup; this study used a mixture consisting of 9.5 g of BioSil A silica gel and 0.5 g of Amoco PX-21 carbon. The eluting solvents were also different: the column was conditioned with 10.0 mL of 50% dichloromethane/benzene solution, 10 mL of toluene, and 5 mL of hexane. The sample extract was then added plus two 0.5ml rinses of hexane. The column was then eluted with 4.5 ml of 25/75 methylene chloride (MeCl)/hexane (which is discarded) followed by 5.5 ml of MeCl and 12 ml of 75/25 MeCl/benzene solution which are collected to obtain the mono- and non-ortho PCBs. The column was inverted, and the dioxins/furans eluted with 12 mL of toluene. Two microliters of tetradecane was added, and the samples concentrated to <10 pL. The samples were then stored in a refrigerator until HRGC/HRMS analysis.

2.22.5. Instrumental Analysis All analyses were performed on a Kratos Concept high-resolution mass spectrometer using isotope dilution. The HRMS was operated in the electron impact ionization mode using selected ion monitoring. Chromatographic separations were achieved using a Hewlett Packard 5890 Series II high-resolution gas chromatograph, utilizing a 60-m x 0.32-mm (0.25- pm film thickness) DB-5MS capillary column. The GC conditions were optimised to completely separate the various 2,3,7,8-Cl-substituted dioxins/furans: initial oven temperature, 130°C; injector temperature, 270°C; interface temperature, 300°C; temperature programming, time 1: 1.0 min, rate 1: 5 °C/min, time 2: 15.0 min, rate 2: 6 °C/min, temperature 3: 295 °C; injector: splitless, 1.0 min, split flow: 30-40 mL/min; purge flow, 1-2 mL/min; and temperature equilibrium time: 2 min. The mass spectrometer was tuned and calibrated prior to all analyses. It was tuned to a minimum resolution of 10 000 ppm (10% valley) using m/z = 330.9792 (or any suitable reference peak) at full accelerating voltage of 8000 V. Pertinent MS parameters were as follows: cycle time for each congener group, -1.0

59 s; ESA sweep (analytes), 10 ppm; native ion dwell, -100 ms; 13C-labeled ion dwell, -35ms; lock mass sweep, 200 ppm; lock mass dwell, 50 ms; ionization voltage, -35 eV; source temperature, 250 °C; accelerating voltage, 8000 V; and trap current 500 pA.

2.22.6. Quality Control and Calibration Between four and six calibration standards with native analyte concentrations bracketing the expected analyte concentrations were analysed prior to analysing samples. Dioxin/furan calibration solutions contained all 13C 2,3,7,8-Cl substituted recovery surrogates (5 pg/pL) and two internal standards (10 pg/pL) each. Native analyte concentrations varied depending on the calibration standard number. The native analyte concentrations in Calibration Standard 1 were 100 fg/pL for the tetras (TCDD/TCDF); 500 fg/pL for the pentas (PeCDDs/PeCDFs), hexas (HxCDDs/HxCDFs), and heptas (HpCDDs/HpCDFs); and 1000 fg/pL for the octas (OCDD/OCDF). Native analyte quantities in other Calibration Standards were multiples of the calibration standard 1 as follows: 0.5x, 2x, 4x, 8x, and 16x. The analyses of these calibration standards permitted the response factors to be determined as a function of concentration using linear regression. The response factor (RF) for each native analyte at each concentration was calculated relative to its 13C-labeled analog. The relative standard deviation (RSD) for the average response factor for each of the native analytes had to be <20%. Similarly, the RF for each 13C recovery surrogate relative to the appropriate internal standard was also calculated. The RSD for the average RF for each labelled surrogate had to be <35%. The calibration curves were considered linear under these conditions, and the analytical system was considered calibrated when these conditions had been satisfied. If these conditions could be not satisfied, corrective actions were taken. The average RFs were used for subsequent quantitations. Prior to sample analysis, the linearity of the calibration curve was verified by analysing Calibration Solution 2 (200 fg of TCDD) and calculating the RF as described previously. The percent difference between the new RF and the average had to be <20% for the native analytes and <35% for the 13C recovery surrogates. The mass chromatogram was also examined to ensure that all the 2,3,7,8-Cl-substituted congeners were clearly separated. If the S/N values were =10, the ion abundance ratios were 15% of the theoretical, and the RF and isomer separations were within specified limits, then sample analyses proceeded. Corrective actions were initiated if specified control limits were exceeded. The concentrations of the various PCB congeners in pg/ul in Cal.Std. #3 were as follows: 77, 2.2; 118, 50.0; 105, 25.0; 126, 0.5; 156, 5.0; 157, 1.0 and 169, 0.5. C13 labelled recovery standards were present in all calibration standards at 5.0 pg/ul, and the two internal

60 standards, PCB 111 and 128, were present at 10 pg/ul. All the native compounds in the remaining calibration standards were multiples of Cal. Std. #3 as follows: 0.25x, 0.5x, 2.5x, 5.Ox, and lOx. The GC and Mass Spec acquisition parameters as well as the QA/QC conditions required prior to analyses for the PCBs were as described above. On the days that samples were analysed, 10 pL of the internal standard solution (20 pg/pL) was added to each sample, and the final volume was adjusted to 20 pL. Once all QA/QC parameters had been verified to be within specified limits, sample analyses proceeded. The mass spectrometer was operated in a mass drift correction mode using PFK to provide lock masses. The selected ion current profile (SICP) areas for the characteristic ions for each native and labelled analyte were measured. Native analyte concentrations were determined by isotope dilution. Peak areas from the characteristic ions for each native analyte and its 13C-labeled analog were used in conjunction with RFs from the internal calibration data to determine concentrations directly. Labelled surrogate concentrations (expressed as % recovery) were similarly calculated using an internal standard method. Samples were organised and analysed in sets: method blank, matrix blank, laboratory control spike (LCS), and the nine samples. Peak identification criteria were as follows: S/N >3.5; the isotope ratio of the two characteristic ions for each congener class within 15% of the theoretical value; the peak maxima for the molecular cluster ions coincide within 2 s; and native analytes elute within 3 s of their corresponding 13C-labeled analogs. Method blanks were examined for the presence of interfering background, which, if present, was subtracted from the sample amount prior to reporting. For furans, an ion for the appropriate chlorinated diphenyl ether was monitored and the ion chromatogram examined to ensure the absence of chlorinated diphenyl ether contamination. The amount of any native analyte detected was listed on the quantitation report, along with the recovery of its labelled analog. Recoveries of I3C-labeled analogs for the samples were between 25 and 150%. Sample sets were reviewed by the QA/QC officer to ensure compliance with QA/QC guidelines/criteria. Analyte concentrations were calculated on a pg/ul basis and, after background subtraction, were converted to ppt in both whole weight and lipid adjusted. Results were transcribed to an Excel spreadsheet for additional processing to derive TEQ values for each sample.

2.2.3. Results The concentrations and Toxic Equivalent contributions (TEQ) from the analyses of the fish tissue for the 2,3,7,8-Cl-substituted dioxins and furans and selected coplanar PCBs are listed in Table 2.2 on a lipid adjusted basis together with summaries of the toxic equivalents

61 (TEQ) results for each sample together with weight, age and lipid content data. All values were adjusted to the lipid content of the sample by dividing the whole weight concentration by the percent lipid in each sample. The percent lipid was determined according to Method 1613. The rounded result for each congener was multiplied by the appropriate I-TEF (International toxic equivalent factor) (van den Berg et al., 1998) and summed (TEQ) for all seventeen congeners. In cases where congeners were reported as non detections, lower bound TEQs were calculated by treating the result as if zero and upper bound TEQs by treating the result as if present at half of the limit of detection (LOD). Quality control (QC) samples were included in batches of samples analysed. Data were rounded to three decimal places for all analytes and are reported on a lipid adjusted basis. The whole weight detection limits may be calculated by dividing the detection limit by the appropriate weight given in Table 2.2. TEQs were also calculated for non-ortho and ortho-PCBs using internationally agreed TEE values by following the same procedures as above. All PCBs analysed for were detected. The summed concentration of selected PCBs is the sum of concentrations of the congeners measured and is an underestimate of the total PCB concentration as other isomers were present which were not measured. The concentration of lipid found in the samples ranged from 3.43% to 14.7%. The immature fish are fed lower levels of oils, and tend to store lower levels of lipid in the flesh compared to mature fish and this is reflected in the lipid content of the samples, with one unusual exception (sample 213-10). Whether this fish had a similar dietary lipid intake to the other fish sampled is not known, but it has been observed that there can be marked variation in flesh lipid content within fish fed the same dietary oil level (the percentage of lipid in flesh of fish fed 28% oil varied from 3% to 17%,) such that certain individual fish utilise high- energy diets but deposit little lipid in their flesh while others tend toward greater adiposity (Bell, 1998). Table 2.3 compares age specific mean lipid adjusted PCDD/F and selected coplanar PCBs TEFs with the means of all samples in this study and the MAFF 1999 report (Parsley et al. 1999). The data was disaggregated by both age and whether the fish was farmed or wild. Significant differences were observed between the 2 year old + fish and 1 year olds, shown in Table 2.2, with no statistically significant difference between the 3 and 2 year old fish (not shown). There was also no statistically significant difference between the farmed and wild fish (or farm escapees).

62 All fish contained detectable levels of PCDD/Fs and mono-ortho and non-ortho PCBs. Levels varied between samples, generally increasing with age, and showing a skewed distribution for the PCDD/Fs resulting from a minority of relatively high values. The main dioxins and furans detected were 2,3,7,8-TCDD, 1,2,3,7,8-PeCDD, 2,3,7,8-TCDF, 2,3,4,7,8- PeCDF and to a lesser extent 1,2,3,7,8-PeCDF. Apart from one fish (sourced via Stirling University, 213-10) which had been frozen a year previously, the highest levels of PCDDs, PCDFs and PCBs were observed in the oldest fish, 3 years +, as expected. The 2,3,7,8- TCDD concentrations ranged from 0.51 pg/g lipid adjusted in the post smolt fish, around a year old, (Sample 213-12) to 1.09 pg/g lipid adjusted in the most mature fish, between 4 to 5 years old, (Sample 213-05), and 1.39 pg/g lipid adjusted in Sample 213-10. This is equivalent to 0.03 pg/g wet weight in the post smolt to 0.15 pg/g in the most mature fish, and 0.05 pg/g in Sample 213-10. This fish had the lowest lipid content of all the samples including the post smolts (See Table 2.2). 1,2,3,7,8-PeCDD concentrations ranged from 0.8 pg/g lipid adjusted to 4.16 pg/g lipid adjusted in the Sample 213-10 fish. All the selected coplanar PCBs were detected, and they were quantitatively the dominant contaminants. The consistent congener patterns were markedly higher than one would expect from the range of sources of the samples, unless they were fed on similarly sourced feeds. The highest PCB concentration detected in all samples was PCB 118 with a highest lipid adjusted value of 31900 pg/g (1093 pg/g wet weight: Sample 213-10) and lowest lipid adjusted value of 14800 pg/g (1380 pg/g wet weight: Sample 213-07). The highest total lipid adjusted PCB concentration was 45100 pg/g (Sample 213-10), but the lipid adjusted TEQ contribution for this fish was 24.0 pg/g, lower than the lipid adjusted TEQ contribution for one of the largest fish (Sample 213-05) with a value of 24.8 pg/g. On a wet weight basis the largest fish PCB TEQ remained high at 3.40 pg/g, whilst the wet weight PCB TEQ value for Sample 213-10 fell to 0.82 pg/g due to the lower lipid content in the fish tissue.

2.2.4. Discussion The results from this study are comparable and of a similar order of magnitude to the values reported in the recent MAFF study (Parsley et al., 1999), see Table 2.3 and also data reported more recently by the Irish Food Safety Authority of Ireland (Food Safety Authority of Ireland, 2003). The results presented here indicate a contamination problem, especially in the case of the selected PCBs. Comparisons with background data from the MAFF study (using the pre 1998 TEF values, (Kutz et al., 1990) suggest that for wild species such as cod.

63 the concentrations on a lipid adjusted basis are similar. However, due to the higher fat content of salmon, the concentrations reported here are comparatively higher on a whole weight basis. Species with a higher fat content in the body tissues, such as herring, are reported to have consistently higher concentrations both on a lipid adjusted (mean 83 pg/g WHO-TEQ lipid) and whole weight basis, but this is not seen in other fatty fish such as mackerel (mean 17 pg/g WHO-TEQ lipid). Flat fish with a low fat content, such as plaice, approach the concentrations seen in herring on a lipid adjusted basis (mean 67 pg/g WHO-TEQ lipid). Interestingly, the TEQ levels reported for another UK farmed species, trout, are in very good agreement with the salmon concentrations on a lipid adjusted basis, again suggesting that the probable source of contamination lies in the aquaculture feed, although other environmental factors could not be ruled out at this stage. Recent, as yet unpublished, data from fish uptake and diet analytical studies, strongly indicate that diet is the primary source on PCB and pesticide contamination in young farmed salmon (Fernandez et al., unpublished data, in review 2003). Further research conducted by the regulatory bodies is required, and action to address the contaminant problem is now being taken by the aquaculture and aquaculture feed industry.

64 o\ oo cn —< COCOCOCO-^CNOsVOCOCOCOCOCOCO"^ CO

CN C4 r-4 CN CN CN (N OO VO CNCN CNCN CN oo CNCN I O Os sqcNcococococN^co COCOCNCOCOCO CNSO Tj- cq CN CN 0(3 Tf f c< rn O O O O O VO ' d d d d d d 4—4 VÔVÔ d r— A CO od VO Os so 00 r— 00 00 VO J CO so CO CN CO c/5 4'Cj- so —4

k: m oo OS Os so VO Os r— r-' r- CN VO VO OO 00 r- c— r~ r— 2 - ^ 'cf Os ,oq CO CO ’-H 0C5 CO CO CO Tt VO sq Os so r-; CO cq CO Tt CN '=q cq OS 4—4 Tq c/5 CN cn o od CO O O —^ od v5 CO d d T—I O O O cc oo CN d CO CO so Os O CN -rt r— 00 CO oo CO 5 S so CN oo ^ Tf so Os 4^ CN CO

r— CN r— CN t— c— C7\ t" CO Os r- r- r- Os so oo 9 = oo ^ CN .—4 \o VO so VO o o 00 c q ^-4 s o OO C/5 CN CN CD o r- O O O O O oo so d d d d d d d A A A c n A A r - • 9 VO ^ _( O VO m g CN CN s 5

A in CO CO 00 c— oo oo CN iO oo oo oo oo 00 CN CO so A k P o , C3s so CN CO 4—1 c— Os q t- CN CN q 4—4 •3; CNCN Os 9 d d so O'^OOOOOvo-^ v5 d d d d d d d SO SO 0\ A A ■9 d A O V) o o VO Os VO ■3- CN CN so o OS ■3- 3 f VO t-~ 1 OO so so so so so CNSO so CN so so so so so so 0^ CO T-4 so so . CN so o Os tq tq CN 4—4 + Os

Î , „ 00 CN -H so so so so so so so so so so so so so Os ^ 00 so so 4—* 4-H —4 —4 SO os so o OO —4 VO Os Os VO c CN c— o oo r— d r-4 d d d d d A d d d d d d d A •4^ 0C5 r- A VÔ OS od ^ O Os CN VO o CN C— Os VO 0\ CN VO < so CN I A ■ IiII Ü II 3 "Tf Os so oo CO 00 oo CO C<0 oo so os 00 oo 00 cn VO 0C3 ^0 O 4—I —4 T—( C-4 so CN r - VO 4^ (Os CO SO ë 5 1 1 ?O O so 4-^ CN d d d d d d d d d CD d so d A AA ^ r t CO O VO 9 9 CO CN C3s o so VO ^3> CN H VOCN OS CN so JS t" H-j -^s s 3 < % VO CO CN oo t— t— oo r- C" r- Os r~ c- f" oo V 'c" >s r-% r-; oc) CO so CO 4—4 CN 4—4 —H VO cq so 9 q CN r- so oo 0C3 C^ '= + -rf Os Os C< d —^ d d d d d Tj- d d d d d d d so A od os d 0\ od CO ^ Os T f CN H oo VO CNCO r- p oo CN s so -o c As k ll (w cn 00 so oo t— r- so r- r— Os CN r- r- so r- r- c- CO VO ^ 0^ -H 0C3 oo CN so CN 4—1 cN —4 4—4 VO V) OO 3 oo r- CN CN CN VO n "3 -t- CN OS Os c— o o o o o t- d d d d d d d so so d d (Os A os A 1 1 ^ CO ^ OS VO os t—OO o oo CO (Os CO oo VO s VO CN '2 13 c VOVO o o O o o 0 VO -4 -H O o o O o o si O VO —4 o o o ii o o O o o 1 '2 oooooooooooo d d d d d d d d d 3 n

I p -nS A1/3 A 0Û bû A A Q 0 ^ 0 I 0 0 8 AAA 11 IT3 II gg 12 CO i X A oo l§i oq crs •o s A oo 00 OS A A oq ill00 oc^ os N In oo > 9 oo c-T r- 0C5 VO 00 oo r— so VO 00 so so I—■ Os r4 R r-{ t-[ ocT 11 O —4 CN VO VO so 0) (D K < < K < Q ri ri < so VO r< A ? '5- y ■a CO CO CO CO CO CO CO k « e 3 ‘a CN CN CN CN CO CN CN CO CN 3 O i m 338 3 3 8 Ô3 CO C/1 -T 4-T O ' CN ^ ^-4 CN 4—4 A A O II g I I A VO oo m m oo oo CO VO CO CO CO 4^ o \ Os VO VO O n q CN o o o q VO o q CN o o O CN Os VO 0 CN O) ( d d d d d OS s o od AA 00 %% %%% %% +12 % ? +1 +] •fi +1 ■H VO oo 43- +1 T)- (N +1 i CN VO +1 +1 +1 +1 +1

I T)- t "3- Tj- -T)- "Tf Tj- (N VO •4^ 44^ 44(j- a VO 00 r — VO 8 VO 8 q 9 2 CN cq CN T f d d d d d d d d d d d d so 4^ A 44^ od A 00 00 VO 00 Ji H 44^ 00 CO 44^ g

CN VO 4^ 44(j- VO VO VO Os VO ■4^- r - VO VO VO VO T f o \ 2: *o q CO q 44^ VO 0 0 CO CO q co co 2 O O o cq VO -q O CN VO A d d d d d d VO d d 0 d d d d d VÔ vd CO 44^ cd +1 +\ +1 ■fi •H +\ -t-i ■H +\ •n +\ +1 +1 —4 r f VO VO +1 +l ^ +"+l'5

- ë ' 3 oo VO VO 0 0 o s CN CO r - CO 0 0 0 0 4Tf 43- 00 00 CN 44^1- VO '3 CN 00 t— CO 00 q r- 0 CO 00 q VO CN q r - Os d o o d d d A vd d d d d d (d d vd A vd VO Os d CN 00 CN VO 00 4^ CN S Jl CO 4^}- VO CN OS VO A VO CN ^ k CN

CN oo CO ^ CN ^ covooo 4'^r~'\oo\cN a\co^cor-oo 4'^ c N v o c N c q c N V 0 CNCNCN 4< ^ C N r '^ V 0 V0 — ^ O C N —4 CN 0C5 V0 0 0 O c 0 0 0 V0 O +!%%%%%% % +1 2 + i? ‘?2 2 2 ? 2 § O sv o ^ O o "^ II +I - g ■n 3 •n ë ai 004>^4'^OVO'4:4"44^VOOOCOCN-^'^ CO Os 00 VO VO CO r - CO r-- CN VO 1| VOOO'^OVOOOOCOVOr.'CNCN CN 0 CN 00 00 9 9 9 q cddAcoddddcdATfdd d d d d d A A 0 A A CN VO CO VO CO lit 44^ 0 g VO s Os

VO 00 r - VO Os (Os 4^ VO 00 CO 44^ Os VO crs VO CN 0 (Os q CN (Os CO '( t 0 0 CN q q CN CN CO 0 0 0 CN CO 0 q q q 00 A d d d d d A d d d d d d d A 0 6 A A (Os A +\ +\ ■H +1 +1 +1 -H ■H +1 +1 +1 +1 +\ +1 +1 •fi +1 VO VO VO CN 44^ ■n +1 T|- I i +i +Î +Î ii +i +Î It 0 a ” >> (Os •SÈ CO VO VO CO T t VO r - CO CN 0 (N V—4 T—I l/q VO VO 44:f 9 2 = s i VO r- VO O CN 00 VO q CN 0 0 CN v—4 004'qvovq4^_tq44^vqvo VO o< d A d d ( d d Os vd d d c 5 O O O o oôvôAAAAd’^ § f . s 9 oo 44^ 4,^f CN CN CO CN CO —4 CO —4 CT\ VO I S ■ VO 4-^ 4-H CN 1

O iO Q A g Q I® - § ■ a A U Q Q Q Q 8 y 0 O4 A E 3 o O X U a d o Q 8(U A t 1oill q o q Os A U A m w S ITS 00 VO VO r- VO A A o q A r-. c CN VO VO I, °o 9 9 VO

1 U 'g oo m "Tj- (N VO q q q q q q q q es A d A d d d AAavddd

^ VO CN CN CN

CN r~- —^ t— ^ VO oo q q —H cN r - di cô d> cô d) di cô A A III V V "0 ^ ^ ^i 2

u CÜ co q q q q C3V q q q CN —I vo vo d O o ovAAdddcjA^ijcj I V V S' li % Z I CN -H c—• r—r- t—r~ r- r- t— ■'(t -H cN o o r" vo vo r— Ov vo oo vo "Tf o 0 ^ 0 0 0 0 0 . q q vo o iis q q q q q q q t, d d d d d d d d d d d d d A A A A A A A en —4 oo en ov vo • S | i i f vo O —4 CN ü a A A A A A 88 ffi ffi 0) si m o q q II i?oo 00 o v A q q Ov A o q iiH q o 6 v d llq A A A o q v q A VO OO VD VO r - o \ h Ï il f " o —H CN v o vo vo îc /l c o q A q A v q v q A t-H #—4 *—H V—4 0 2 - 9 < Q 2 2 A A e n c q c q co c o Q A c q q q c q q c q e n e n O ffl A A m m pq CQ c q CN A A A A U c q CN cq CN CN c q ( A A A U u u u u u u u II A 1 II CN I A A O A A A A —s CN — A A O A A A A A A A vo 2.2.4. Discussion (continued) More detailed comparison with other existing data sets is confounded by the fact that many previous studies quote only means and ranges, or they use a different TEF system or the WHO-TEF system prior to 1998. Thus, direct comparison with Norwegian data for Atlantic salmon was not possible as the TEQ values reported in the literature use the Nordic TEFs, and not the revised WHO- TEFs (Vuorinen et al., 1997). As the values for the individual congeners are not given, a corrected TEQ value cannot be recalculated, as undertaken in this report with the UK MAFF study. Table 2.4 provides comparative data for PCDD/Fs and coplanar PCBs in Salmon and Herring fish from studies where individual congener levels were reported. Both this study and the MAFF study report surprisingly high detections of PCB congeners. From the literature one might speculate that the source of contamination was from the herring and fish oil based aquaculture feed consumed by the salmon (Bergqvist et al., 1990). A fish oil study found relatively high levels of PCBs 52, 101, 118, 138, 149, 153, 170, 180 in both the herring oil sample from Germany and salmon oil sample, e.g. PCB 118 detection of 60 pg/g Iwt in Salmon oil and 158 pg/g Iwt Herring oil (Jacobs et al., 1998). (The coplanar PCBs, TCDDs and TCDFs were not analysed in this fish oil study.) Preliminary data on the analyses of salmon aquaculture feeds and fish oil components of the feeds for organochlorine pesticides and selected PCBs indicate that this is a probable route of contamination (Jacobs et al., 2002b). It is unlikely that the ash fraction could account for the high PCB detections, although it might also contribute to the contaminant profile in aquaculture feed as contaminated ball clay was found to contribute the PCDD and PCDF concentrations detected in samples of US farmed catfish and chickens (Ferrario et al., 1999; Ferrario and Byrne, 2000). Both ash and ball clay are used as fillers and mineral sources in animal feeds, and can be expected to have a specific PCDD/F congener profile that can be used to identify the source of contamination when compared to biological matrices. However, the PCDD/F congener profile from the fish in this study did not match that found in the ball clay. Previous reports have detected significant levels of PCDDs, PCDFs and PCBs in fatty fish such as herring and salmon, particularly from the Baltic Sea, see Table 4 (Rappe et al., 1989; Bergqvist et al., 1990; Svensson et al., 1991; De Boer et al., 1992; Vuorinen et al., 1997; Liem and Theelen, 1997; Atuma et al., 1998). The potential contribution to the human diet of PCBs, PCDDs and PCDFs from Scottish salmon will vary according to the age of the fish, whether the individual fish had a predisposition to adiposity, or not, and the frequency of consumption, and portion size.

68 cooking practises (it is more usual to retain the oils when cooking salmon, for their nutritional fatty acid benefits), and age of the consumer. The UK Ministry of Agriculture Fisheries and Food (Ministry of Agriculture Fisheries and Food, 1999), using pre 1998 WHO TEF values, give an estimate of upper bound dietary intake for adults of PCDD/Fs and PCBs from salmon of 2.3 pg TEQ/kg bw/day. Upper bound adult dietary exposure for high level consumers is 5.6 pg TEQ/kg bw/day, the average consumer intake is estimated at 2.6 pg TEQ/kg bw/day. However, young children are particularly susceptible to higher intakes, as they have a more limited diet than older children and adults, together with a low body weight. This leads to higher consumptiombodyweight ratios than that of the of the average consumer. Currently the World Health Organisation recommends a Tolerable Daily Intake (TDI) of 1-4 pgTEQ/kg bw/day (van Leeuwen and Younes, 1998) and the European Commission Scientific Committee on Food recommend a temporary tolerable weekly intake (tTWI) of 7 pg WHO-TEQ/kg bw (Scientific Committee on Food, 2000a). The equivalent of this recommendation was adopted by the UK Committee on Toxicity of Chemicals in Food, Consumer Products and the Environment (COT) in 2001, reducing the TDI to 2pg/kg from 10 pg/kg bw/day for 2,3,7,8-TCDD, based on lifetime averages. The COT has acknowledged that ‘for people in the UK and many other countries, the daily intake of dioxins could be close to or even above the guideline value’ (UK Committee on Toxicity of Chemicals in Food, 2000). One gram of sample 213-05, the largest fish, would give a contribution of 4.98 pg of the congeners analysed, towards the total TEQ/day. The possible contribution to dietary intakes of organochlorines from farmed fish could be significant for high fish consumers, and may be congener specific, similar to the species being consumed, as observed in Finnish studies (Kiviranta et al., 2000; Kiviranta et al., 2002), and also observed in the livers of Canadian salmon, compared to their feed (Dconoumu et al., 2003, paper in review).

69 2.3. Investigation of selected persistent organic pollutants in farmed Atlantic Salmon {Salmo salar), Salmon aquaculture feed and fish oil components of the feed

2.3.1. Abstract

This study builds on the dioxin and dioxin-like PCB study of farmed salmon (Chapter 2.2.), taking a closer look at a more extensive range of contaminants detected in aquaculture feed and components of the feed, particularly fish oil, as introduced in section 2.1. This study reports on the determination of a wide range of polychlorinated biphenyls (PCBs), organochlorine pesticides and polybrominated diphenyl ethers (PBDEs) in farmed and wild European Atlantic salmon fish, aquaculture feeds and fish oils used to supplement the feeds. It confirms previous reports of relatively high concentrations of PCBs, and indicates moderate concentrations of organochlorine pesticides and PBDEs in farmed Scottish and European Salmon. Concentrations of the selected persistent organic pollutants varied among the samples: PCBs (salmon, 145-460 ng/g lipid; salmon feeds, 76-1153 ng/g lipid; fish oils, 9-253 ng/g lipid), X DDTs (salmon, 5-250 ng/g lipid; salmon feeds, 34-52 ng/g lipid; fish oils, 11-218 ng/g lipid) and PBDEs (salmon, 1-85 ng/g lipid: salmon feeds, 8-24 ng/g lipid; fish oils, ND-13 ng/g lipid). Comparison of the samples for all groups of contaminants, except for HCHs, showed an increase in concentration in the order fish oil < feed < salmon. Homologue profiles were similar, with an increase in contribution of hepta- and octa-PCBs in the fish, and profiles of DDTs were similar in all three types of samples. With a constant contribution to the total PCB content, the ICES 7 PCBs appear to be reliable predictors of the PCB contamination profile through all the samples. For PBDEs, BDE 47 dominated the profiles, with no significant difference in the PBDE profiles for the three matrices. Samples with higher PCB contents generally showed higher levels of the pesticide residues, but this was not the case with the PBDEs, indicating the existence of different pollution sources.

2.3.2. Materials and methods 2.3.2.1. Sites and sampling Seven British salmon samples (of which six samples were from individual fish, and one sample was a composite of two fish from the same source and of the same age and size to ensure similar sample quantities of tissue), and one Norwegian salmon {Salmo salar) sample that enter the European fish market were analysed. The samples were of variable age, both farm raised and wild, and were obtained from five different Scottish

70 sites, and were duplicates of the samples documented in Chapter 2.2 and shown in figure 2.1. Whole body weights of the fish ranged from 400g to 5.2kg. In addition five salmon samples, two originating from Ireland (one wild, one farmed) and three farmed samples for which no further information was available, purchased from the Belgian market, were analysed. The fresh and frozen samples were wrapped in polyethylene bags and frozen immediately at Eight salmon feeds were analysed (from four different Scottish sources), five fish oils and one vegetable oil were analysed, (all but one of which were obtained from the same source, on the same date, but originally from widely varying sources). The fish oil and feed were not samples fed directly to the salmon collected as these were not available, but were collected from Scottish sites in January 1999. Table 2.5 gives complete sample details.

23.2.2. Sample preparation The samples were thawed, filleted, skinned and the epaxial muscle homogenized before being subdivided into smaller replicate portions of approximately 100 grams. The portions were weighed, stored in tightly sealed polythene bags and frozen at -20°C. The samples were sent packed in dry ice to the Toxicological Center, Antwerp, where they were logged-in and stored at -20°C prior to analysis. All samples were analysed for HCH isomers, HCB, DDT and metabolites (six op- and /7/7-isomers) (Table 2.6) and for nine PBDE congeners (lUPAC no. 28, 47, 66, 71, 75, 99, 100, 153 and 154). The samples obtained from Scotland were analysed for 59 PCB congeners (see Table 2.6), including mono-ortho PCBs with an assigned toxic equivalency value (TEF) (van den Berg et al., 1998). All dioxin-like PCB levels compared favourably with those found in duplicate salmon samples analyzed earlier by the US Environmental Protection Agency laboratory at the John C. Stennis Space Center, MS, USA , section 2.2 (Jacobs et al., 2000; Jacobs et al, 2002a). The samples obtained in Belgium were analysed for 23 PCB congeners. The procedures used to analyze the salmon samples were similar to those used to analyze organochlorine compounds in chicken and pork fat as described elsewhere (Schepens et al., 2001; Covaci et al., 2002).

71 Table 2.5. Sample details for the salmon, feed and fish oil.

Sample Sample Sample type % Lipid Information Weight codes and collection lipid Weight (g) (g)* M il (f) 3+ yrs fresh 13.5 8.19 1.10 M12 (f) 2+ yrs fresh 14 8.13 1.14 M13 Norway frozen 13.7 8.39 1.15 M14 Salmon: (w) 2+yrs frozen 4.9 9.73 0.47 M24 Scotland (f) 3+ yrs fresh 19.9 9.69 1.93 M25 Jan 1999 (f) 3+ yrs fresh 19.1 9.04 1.73 M28 (w) 12.3 8.19 1.17 M31 (f) N=2, Smolt, 9.3 8.13 0.85 fresh M18 (f) sm Bio. 3.9 4.60 0.18 M19 Salmon: (0 Fresh Bio. 16.5 5.15 0.84 M20 Belgian Market (f) Fresh 11.0 5.48 0.60 M21 March 2001 (w) Ireland, sm. 11.6 4.51 0.52 M22 (f) Ireland, sm. 14.8 7.25 1.07 MCI Fry l-5g (A) 18.1 3.00 0.54 M02 1000-2200g (B) 30.1 3.00 0.90 M03 350-1000g (B) 31.0 3.03 0.94 M04 S. feed 1000-2200g (C) 35.9 3.07 1.10 M05 Jan 1999 Fry(C) 19.8 2.26 0.45 M15 500- 1300g (D) 28.1 2.00 0.56 M16 1300-2200g (D) 34.3 2.11 0.72 M17 500-1300g (D) 32.7 1.97 0.64 M06 Fish oil (C) 100 0.5 M07 Fish oil (C) 100 0.5 M08 Fish oil (C) 100 0.5 Fish oils M09 Fish oil (C) 100 0.5 Jan 1999 MIG Vegetable Oil (C) 100 0.5 MD Fish oil (E) March 100 0.5 EPA 01

*= Sample weight taken for analyses; ND= Non Detects (treated as V2 LOD for Xcalculations); (w)=wild; (f)=farmed; Sources A, B, C, D, E denoted in brackets; S. Feed=Scotland aquaculture salmon feed; Sm=smoked; Bio=bioculture

The methods used to analyze PBDEs are described in Covaci et al. 2002 (Covaci et al., 2002a). PCB 46, PCB 143, e-HCH and ^^C-BDE 47, 99 and 153 were used as internal standards, while 1,2,3,4-tetrachloronaphthalene (TCN) and bromobiphenyl (PBB 80) were used as recovery standards. All individual standards of PCBs and pesticides were obtained from Dr. Ehrenstorfer Laboratories (Augsburg, Germany). PBDE native and labelled standards were available from Wellington Laboratories (Guelph, Canada). Acetone, hexane, dichloromethane and iso-octane were of pesticide grade (Merck,

72 Germany). Anhydrous sodium sulphate, basic aluminium oxide and silica gel (Merck) were used after heating overnight at 120°C. An accelerated Soxhlet extractor B-811 (Buchi, Switzerland) was used for the extraction of target compounds from fish tissues and feed.

2.3.2.3. Fish tissue The fish fillet was defrosted at room temperature and approximately lOg precisely weighed were ground with 20g anhydrous sodium sulphate, and placed into an extraction thimble. Internal standards (20ng of PCB 46, PCB 143, lOng of e-HCH and 1.5ng of a mixture of ^^C-BDE 47, 99 and 153) were added and the mixture was extracted for 2 h with 75 ml hexane: methylene chloride: acetone=3:l:l(v/v) into a hot Soxhlet manifold. After concentration and fat content determination, the extract was subjected to clean-up on 2 successive solid phase extraction (SPE) cartridges containing 5g acid silica and acid silica (5g): neutral silica (Ig): deactivated basic alumina (2g) (from top to bottom), respectively. Fifty (50) ml of hexane were used for complete elution of PCBs, pesticides and PBDEs in the same fraction. The final eluate was concentrated under nitrogen and recovery standards were added prior to GC analysis.

2.3.2.4. Fish oil 0.5g fish oil were precisely weighed, and solubilised in 3 ml hexane. Internal standards were added and the mixture equilibrated in an ultrasonic bath for 5 minutes, after which it was ready for application to the SPE cartridges.

2.3.2.5. Fish feed Approximately 3g of fish feed were weighted accurately and placed into an extraction thimble. Internal standards were added and the feed was extracted for 2h with 60ml hexane: methylene chloride: acetone=3:l:l (v/v) into a hot Soxhlet manifold. After concentration and fat content determination, the extract was ready to be applied to the SPE cartridges

2.3.2.6. Fat content determination The Soxhlet extract was concentrated and then transferred to a preweighed tube. The extract was completely dried under a gentle nitrogen stream and then kept at 60°C until constant mass. The measured weight represented the fat content of the sample.

73 23.2.7. Organochlorine analysis One \il was injected in pulsed splitless mode on a Hewlett Packard 6890 GC connected via direct interface with a HP 5973 mass spectrometer. A 50m x 0.22mm x 0.25pm, HT-8 capillary column (SGE, Zulte, Belgium) was used with helium as carrier gas at a constant flow of 0.7 ml/min. Injector and interface temperatures were set at 270°C and 300°C, respectively. The oven temperature program began at 90°C, was kept for 1 min, then with 15°C/min to 170°C, held 3 min, then with 4°C/min to 270°C, held 1 min and further with 10°C/min to 290°C, held 15 min. The mass spectrometer was operated in electron impact ionisation mode. Three most abundant ions were monitored for each level of chlorination for PCBs or for each pesticide. Method limits of detection (LOD) for individual PCB congeners ranged between 0.1-0.5 ng/g lipid. For HCHs and DDTs, the detection limit was 0.2 ng/g lipid for each isomer. Recoveries of target compounds ranged between 72 and 80%.

2.3.2.8. PBDE analysis The analysis, previously described by Covaci and coworkers (Covaci et al., 2002a), was performed on a Hewlett Packard 6890 GC equipped with a 10m x 0.10mm x 0.10pm AT-5 (5% phenyl polydimethyl siloxane) capillary column (Alltech, Belgium) and connected via direct interface with a HP 5973 mass spectrometer. Helium was used as carrier gas at constant flow (0.4 ml/min). The low-resolution quadrupole mass spectrometer (El) was operated at 70 eV in the selected ion monitoring (SIM) mode. The ion source and quadrupole temperatures were 230 and 250°C, respectively. The electron multiplier voltage was set at 2300 V. Dwell times were set at 10 ms. Two most abundant ions (M^ and [M+2]"^ for the tri-, tetra-, and penta- congeners and [M-2Br]^ and [M- 2Br+2]^ for the hexa-congeners) were monitored for each level of bromination for native and labeled PBDEs. Recoveries of internal standards, ^^C-labeled BDEs, ranged between 81 and 103% (RSD < 21%). Method limits of detection (LOD) for individual PBDE congeners ranged between 0.1 and 0.4 ng/g lipid.

2.3.2.9. Quality Control and Quality Assurance Retention times, ion chromatograms and intensity ratios of the monitored ions were used as identification criteria. A deviation of the ion intensity ratios within 20% of the mean values of the calibration standards was considered acceptable. The method performance was assessed through rigorous internal quality control, which included daily check of calibration curves and regular analysis of procedural blanks and certified material CRM 350 (PCBs in mackerel oil). The method was tested by participation in an

74 interlaboratory test organized by the Institute for Reference Measurements and Materials (IRMM, Geel, Belgium). Seven PCB congeners (no. 28, 52, 101, 118, 138, 153 and 180) were determined in non-spiked, medium- and high-level spiked pork fat. The results of the individual PCB congeners deviated less than 10% from the target values. For PBDEs, the analysis of two samples of biota (eel and porpoise liver) used for the first interlaboratory test on PBDE (de Boer, 2000), showed a variation of less than 10 % from mean values obtained from eighteen participating laboratories.

2.3.3. Results

All samples examined contained detectable residues of organochlorine contaminants. Summary results are presented in Table 2.6, figure 2.2, with total sum (E) values given using a value equal to Vi LOD where there were no detections, as recommended by the EC Scientific Committee on Animal Nutrition in 2000 (Scientific Committee on Food, 2000b). Detailed results of PCB congeners are presented in Table 2.7 and figure 2.3. All concentrations were adjusted to the lipid content of the sample. o,p' isomers of DDTs were not found in any sample. PBDEs were detected in all fish samples, all feed samples but only one of the fish oil samples (see Table 2.8). The total concentration of selected PBDEs is the sum of concentrations of the congeners measured and is an underestimate of the total BDE concentration as other isomers present were not measured.

2.3.3.1. Salmon samples PCB concentrations in the farmed samples were between 145.1 and 459.9 ng/g lipid weight. Profiles and concentrations of PCBs were similar and of the same order of magnitude for all the farmed samples, but lower in a Norwegian sample (M l3) and a magnitude lower in a wild Irish sample (M21). In the sample M13, the PCB profile is dominated by the hexa- and hepta-PCB congeners, while in the other samples, the lower chlorinated congeners (tri-, tetra- and some penta-PCB congeners) had a higher contribution to the EPCBs. The mean of the 7 ICES marker PCBs was 129.5 ± 38.1 ng/g lipid and they constituted 41.1 ± 1.8 % of the total EPCB congeners for the Scottish salmon samples. The mean for HCB for all the salmon samples (n=13) was 11.6 ± 11.9 ng/g lipid, but was much lower in sample M13 (under the detection limit of 0.2 ng/g lipid). The mean concentration for HCHs (Ea, p, y-HCH) was 13.8 ± 7.6 ng/g lipid for the samples obtained in Scotland and 8.5 ±9.1 ng/g lipid for all the salmon samples (n=13). y-HCH was the predominant HCH isomer with a mean ratio y-HCH/EHCHs of 0.69 ± 0.01. The mean concentration of DDTs (E/7/7-DDE, /7/7-DDD and pp-DDT) was 69.6 ± 58.4 ng/g

75 lipid for all the samples (n=13) and pp-DDE was the predominant contributor to the EDDTs. The ratio of /7/?-DDT/EDDTs ranged from 0.1 to 0.31 (mean 0.22). Concentration of DDTs in M13 was much lower (5.1 ng/g lipid). The PCB, organochlorine pesticide and PBDE contaminant loading of the bioculture farmed salmon (n=2) appears to be of the same magnitude as that of the farmed salmon, however the sample size is too small to draw any firm conclusions. Bioculture salmon diets are similar to those of farmed salmon with regard to fishmeal content, but the maximum fish oil content is 28% of the total diet (Soil Association., 2000). All fish contained detectable levels of PBDEs, the sum of the BDEs (EBDEs) ranged from 1.1 to 85.2 ng/g lipid for the thirteen salmon samples. The highest levels were found in a wild (or possible farm escapee) salmon sample. As expected, BDE 47 dominated the profiles, with a mean ratio BDE 47/sum BDEs was 0.53 ± 0.04. The levels of PBDEs found the salmon analysed in this study are comparable with those found in other cultured oily fish, particularly gray mullets and yellow tails, as reported recently in a Japanese study (Akutsu et al., 2001), and also wild fish (de Boer et al., 1998).

2.3.3.2. Aquaculture feed Concentrations of PCBs in the eight feed samples ranged between 75.6 and 1153.2 ng/g lipid (Table 2.6, 2.7 and figure 2.3.). Six feed samples presented the same profile of PCBs, with the penta- and hexa-PCB congeners representing up approximately 80% of the total PCB content. The 7 ICES marker PCBs constituted approximately 40% of the total EPCB congeners. As expected, the concentration of PCBs was high (147 ±41 ng/g lipid), probably due to the high fish oil content of some samples (up to 35%). Two samples from the same manufacturer (M2 and M3) presented a totally different profile of PCB congeners when compared with the other six feed samples. Tri and tetra congeners (approximately 55 and 28%, respectively) dominated, and the profile was close to that of Aroclor 1242. This different profile correlated with the high PCB concentrations found for these two samples (the highest concentrations detected in all the samples) and could point to the existence of local PCB sources or illegal dumping of used Aroclor mixtures in the animal food chain (Schepens et al., 2001). The mean for HCB was 9.2 ± 2.9 ng/g lipid, while for HCHs, the mean was 23.7 ± 14.9 ng/g lipid, with y-HCH being the predominant HCH isomer (mean ratio y-HCH/EHCHs of 0.63 ± 0.16). The mean for the

EDDTs was 42 ± 8 ng/g lipid, with pp-DDE the predominant contributor to the EDDTs

(Table 2.6). The mean ratio pp-DDT/EDDTs was 0.13 ± 0.04. The levels of PBDEs ranged from 8.1 to 23.9 ng/g lipid for the eight feed samples (Table 2.8, figure 2.2). The

76 fry feed samples, with a far lower fish oil content, had lower residue levels than the feeds

designed for smolts and adult salmon with a fish oil content over 20%. Figure 2.2: Median concentrations of HCB, HCHs, DDTs, PBDEs and marker PCBs

in salmon, feed and fish oil samples

H salmon (n=13) afeed (n=8) □ fish oil (n=5)

% 100

op 80

2 60

H 40

HCB HCHs DDTs PBDEs Marker PCBs

Figure 2.3: Profiles of individual PCB congeners in salmon, feed and fish oil samples

□ salmon a feed □ fish oil

CDIO

Hci

PCB congener

77 2.3.3.3. Fish oils The fish oil samples showed varying PCB contents between 8.8 and 252.5 ng/g lipid (ND = Vi LOD). Hexa and penta-PCB congeners dominated the PCB profile for all samples. The same profile was found in fish oil used for diet supplement (sample MDEPA), which contained higher PCB amounts (450 ng/g lipid). However, the tri- and tetra-PCB congeners were higher contributors to the total sum in the fish oil M06 than in the fish oils M08 and MDEPA. Furthermore, the hepta- and octa- congeners were more important for the last two oils. The difference in profiles can be due to different species used for collection of the oil and to different processing procedures. During the oil processing (usually a steam distillation), some of the low chlorinated PCB congeners can be lost through their increased volatility. Mean HCB concentrations were 2.5 ±1.5 ng/g lipid, while mean HCHs levels were 10 ± 7 ng/g lipid. y-HCH was the predominant HCH isomer with a mean ratio y-

HCH/EHCHs of 0.50 ±0.21. Mean ZDDT concentrations were 66 ± 87 ng/g lipid and pp-

DDE was the predominant contributor and the mean ratio pp-DDT/ZDDTs was 0.15 ± 0.07. The levels and ratios of the HCB, HCHs and DDT are reflected in the aquaculture feed, and the DDT levels were less wide ranging in the feeds. PBDE concentrations ranged from ND to 12.7 ng/g lipid.

2.3.3.4. Comparison between the three types of samples The three types of samples were compared using median values for all groups of compounds (Figure 2.2). Except for HCHs, all other groups of contaminants show an increase in concentration in the order fish oil < feed < salmon. Samples with concentrations close to the median value can be considered as representative for each type of sample: M06 for fish oil, M04 for feed and M24 for salmon. For individual congeners (Figure 2.3), a higher contribution of PCB 18, 28, 31 (tri- PCBs), 66 (tetra-PCB), 101, 105 (penta-PCBs), 138 (hexa-PCBs) and all hepta and octa- PCB congeners was found for the salmon, together with a lower contribution of PCB 47/48, 49, 52 (tetra-PCBs), 87, 110 (penta-PCBs), 132, 146 and 149 (hexa-PCBs). The profiles of homologues were similar, with an increase of contribution of hepta- and octa- PCBs in the fish. Profiles of DDTs were similar in all three types of samples. For HCHs, the a-HCH isomer had a higher contribution in the fish oil, while the y-HCH isomer had a lower contribution. With a constant contribution to the total PCB content, the ICES 7 PCBs appear to be reliable predictors of the PCB contamination profile through all the samples. For PBDEs, BDE 47 dominated the profiles, as expected, ranging between 50 to 100% of the XBDEs included in the analyses. There was no significant difference in the

78 PBDE profiles for the three matrices. Samples with higher PCB contents generally showed higher levels of the pesticide residues, but this was not the case with the PBDEs, indicating the existence of different pollution sources.

2.3.4. Discussion The PCB results from this study are in good agreement and of a similar order of magnitude to the values for salmon reported in recent studies (Parsley et al., 1999; Ministry of Agriculture Fisheries and Food, 1999) (see Table 2.9). The comparability of the PCB data suggests that both the data sets can be considered fairly representative of PCB concentrations in British farmed salmon. Previous reports have detected significant levels of PCBs and organochlorine pesticides in fatty fish such as herring and salmon (Table 2.3), (Parsley et al., 1999; Ministry of Agriculture Fisheries and Food, 1999; Vuorinen et al., 1997; Atuma et al., 1998; Becher et al., 1998). The 7 marker PCB concentrations in the farmed salmon samples from this study and a Dutch study (RIVO Fisheries Institute of Holland, 2000) were very similar. The Dutch study found PCB concentrations of 110 and 170 ng/g lipid in farmed salmon from Norway, while farmed salmon from Scotland contained 135 and 210 ng/g lipid, ranging from 61 to 183 ng/g lipid. The Dutch study also found farmed salmon from Norway or Scotland had relatively low TEQ values (< 4 pg TEQ/g whole weight). Assuming a mean value of 20% of the lipid content (RTVO Fisheries Institute of Holland, 2000), the TEQ value becomes approximately 20 pg TEQ/g lipid, similar with or somehow lower than values found in our study (see Table 2.3).

79 o^>nvO'—ivoo\>n o >—I ov cn ov ^o4o4o4voo4incn ^ O l/l /-N o iCS^CSCSCNfS 04 cn 04 04 ^ 04 ^ 1-^ y 0 0 0 0 0 C 5 0 0 C3 cd c> CD c 5 CDCDOOOCDCDCD CD o ' CD CD ^ O I II o\ V) Tf vn Tf Tf O' in in p p p p p p CO p T t O OV 1 ^ ON \d 00 vi \d 1—I O 00 i n 00 iri 00 c d I—i 1—i CD vd . o g | CQ » n T f (N Tf Tf Tf VI VO v£) i n T t CO CO CO r o CO i n i n CO I i n TJ "I ë t" 0 P VO CO CO Ov oo O ' VO Ov 04 oo t J- 00' 00 P 0C5 0C3 cd cd i i i i i i n 0'^ o4 OV vd 23 — I - ^

1 VO ov CO ^ CO p p 04 p p p p p p p - ^ p p p oo in VO O . 0 0 00 ^ 0^ 2h OV 2h O'TfTf-Ti-invdcsvd c d c d vd Z ag . èi

O \0 O r-i p p p Tf p p Tt i GO 04 00 04 B o O' 04 VO 00 OV CN in cd ^ cd in CD I i n c n c n f o O4 04 04 04 O 4 C O C O C O

•43 0 Ov VO ov rv Ov 00 O ' o o VO P VO C wS in 0 00 OV 0 in VO O' CO '- i 00 S H R V i 2 o< CQ X R Z 5 i i i i i I CS T f (N r—I T—I

CO O' p O' VO VO Q l~' 1 Tf CD vd 04 vd vd vd i 2 23 % z % i i i i i CO z Ov 23

QQQ CD 00 z Z Z 04 i i i i i z iii2 g g g g II is

oo o 25 00 p p p p 04 OV CO «> p vd Ti" o lg iig g g i n ■^' 04 v d i n vd vd ^ T f 04 si

% VO VO VO O' 00 VO Û p p p p p p p p p p 04 CN P p ov cd P i CO vd vd c d c s O4T|-Tfvd(D0'Ov0' c d ^ CD CD Z 04 00 p g T f §gg 5 04 CD 8 . .. 00 «3 I CD 00 a r t p p VO VO p VO O'O' g o 4 ^ 04 V/“|(SO\—(cnoo^oino —|'-|'0/^_I_1(SJQ 00 VO VO VO CO a 04 CD O ' t 53

VO VO 1—1 OS 1—1 CD 00 cs Th p p p p VO VO l l w g vd O n vd oS O' 8 8 04 VO ^ t R s 00 3 00 VO g + I CO CO S0 s CO s OS 04 i~' 5 =

S 3 f3 00 CD 00 VO VO VO 00 VO VO Ov 04 O O ' p p p p Q p oo Tj- w y I o cd 04 cd cd T|- 04 CD 04 T-i rH cd c d o c d c d Z c d c d c d i i 3: is sq VO T f CO p CD P O' p p p p p I— I CO OOPcD^O'Ari^ p Q ,0 i n I W 00 oS oS vd (D vd 04 rd 9 RCO CO CO R 04 "Tf" 04 I ^ c^Sod^cd-^S Si ii 1c § p- CO CS 04 VO p P 04 O' VO p O' T^ 00 p O s VO T f 00 OV 0 04 04 O T|- N g vd S CO S m 04 Ov 00 CD CO OV I 2h t~' 8 23 3 9 K 00 8 l> O' VO CO 82 T 5 8 5 § VOfSlO'^'^'OTf 04 CD ov (S T f VO CD VO 00 CO S;2 CO H 2 ON i n o CT\ t-h ’ ov CD oS CO Æ R t~' a t 3 a R VO VO 9 5 i i u CL CL

Cm 0£ CO Th 0 c n r o '-H 04 i n p ^ oo lOv 00 00 Ov p Q 04 " o b o vd vd T|- T f o \ r d 00 vd oS 04 ? i n Tj- cn cn 04 cn 04 CS CS 23 i i t r 3 II 1 1 T 3 ^ p p VO 04 04 T f VO s CO VO p p 04 m Tf 00 in CO CD p 00 Ov CO 04 3 3 W • £ cd c d CD CD CD ov o~ 0~ Tf 04 vi cd CS cd cd CD CD O z % 9 CO CN IS 3 VO r i 3 (D C4 " a 04 CO Tj- Ti VO 00 00 Ov CD ^ 04 CN CO T|- VO VO VO O' VO t ' ' 00 e s 04 (S CO CD CD CD CD CD CD CD Z i 3 II s s § § § § § s H * 0

uouijBS P 99J no qsH CL CL iizizzziiiiiziiiêiiiiiiiiiêiiiizz 3 iiiiiiiiggigiiiisisissisisiiiiisg

I Q Q Q Q Q Q Q P en O'. «N Q Q o -H en o i vq vo vq Z Z Z Z Z Z Z < o

QQÛ'OS'Qc'vO'Oco d d d -■ Z t5 '-:

I OQQoio\QQf'»-:o:pTÇ

S es en es 00 © 3 : O 2

00 es 00 r t vn I d d d d vd en vd Tf P d d ? ao r f q s: p d d d d d r f d d d 'H en o

p p 'H p p vo P g S s d d d ■d d vo es ov 2 icv O 2 d en d % 0 0 vo es o o VO vo vo p p Q 00 00 \o VI 8PL d d d Ov Tf CD t> d d d o< d 2 R o< i f i Q p p g g g g g g g gg gg gg sS S g Z S-' d d d •§ p cN vq en 00 un o g- d d vd c4 '«t vd en d vd

Tf Ov C: 1-H 1-d es en es es l d d d d d d d (D s- 2 cK es* vS o vd es ' s : d a es en Uo

04 ü H p q p q q Q p q p s- en 2 vo en Z es (S S 2 i m a

es < < < < < zzzzzzzzzzzzzzzzzzzzzzz g g d z z z z z gs§ssgi2iiiiiiiiisgsgii g g

<—iin'5es'5Qt''enQesovooovr~oot''envO'^'~lQoO'^ en <;■<<;< <1 en 00 ^ vd ^ Z d 2: d d d cs‘ d en d es' o\ en 2 Z d d d z z z z z

p p es < < < < < ZZdZdZZZZZZZZZZZZZZZZZZ g g00 es d Z Z Z Z Z ooesovvO'^n d

< ■ < < !<

2 3g5g2 2gg2ggSg:gS2S3ggg g g Tf d

2;2sSss3gSgg§gSgg5§3ggg g g g g g ssssgssagggggggggsisggg g gS én ^

2“aS§2§SgSgg§g2§gS2“ggg g g en o r-in ^ . Qr'OvQQQQQQr'QQooQoQQQ 2Zos-ZZZZZZoZZ-ZenZZZ g g pppQqOQpQqQQeNQpQQQ oo>oZ-'ZZoZoZZcsZenZZZ g g 55PwS3S3gSggggSggSSSggg g g g g g fl 8| es Q rf d SI il OV es

en 00 Ov es g# es o es es p Q es un es ^ o\ ^ es o es' d d d es O Z d d

"t ov en p o p 2 l es vo S" d d 3 d ; ; ^ "o q Q ^ d d es d Z CS en vo vo Tf l/V p p es vS d 00 d es d d o

q p Ov p Q «n CS en 00

00 es q 'H p q rf es Ov O en es r- o es o 2 CN >n d O

P P q p p p p p Q es vo es d d d ci es d es 00 d o Z d d M ^ 8 p q ^ p lA Z S- 2 .H t' o ov Q es es 2 es d o Z "S2 g §11. H 5 i Table 2.8: Selected PBDEs in European Atlantic salmon, aquaculture feed and fish oils: lipid adjusted (ng/g, ppb)

Sample BDE BDE BDE BDE BDE BDE BDE BDE BDE sum BDE codes 28 71 47 75 66 100 99 154 153 BDEs 47 /sum M 11 1.3 7.1 29.3 0.3 2.0 6.5 4.9 1.3 1.0 53.7 0.55 M12 0.9 4.4 14.7 ND ND 2.9 2.8 0.7 0.8 27.2 0.55 M13 ND ND 0.4 ND ND ND 0.7 ND ND 1.1 0.37 M 14 0.8 1.7 12.9 ND 0.9 2.4 5.9 ND ND 24.6 0.52 M24 1.5 7.1 28.9 0.2 2.2 6.4 5.0 1.6 ND 52.9 0.55 M25 1.5 7.1 28.0 0.4 1.8 6.1 4.4 1.0 ND 50.3 0.56 M28 2.0 7.9 43.0 0.3 3.2 9.9 14.0 3.6 1.3 852 0.50 M31 1.3 6.7 25.0 0.5 1.6 5.6 4.0 0.9 ND 45.6 0.55

M18 ND ND 3.1 ND ND ND ND ND ND 3.1 - M 19 0.5 2.8 10.2 ND 0.8 2.8 1.6 0.2 ND 18.9 0.54 M20 0.5 2.4 9.4 ND 0.4 1.5 1.5 ND ND 15.7 0.60 M21 0.2 0.7 5.0 ND 0.1 1.1 1.3 ND ND 8.4 0.59 M22 1.0 8.1 25.8 0.7 1.7 7.9 5.6 1.3 ND 52.1 0.49 MOl 0.3 1.4 7.0 ND ND ND ND ND ND 8.7 0.81 M02 0.5 3.1 8.2 0.2 0.5 1.5 1.7 ND ND 15.6 0.53 M03 0.5 3.1 8.2 ND 0.7 1.5 1.4 ND ND 15.4 0.54 M04 0.7 4.3 12.3 0.3 0.7 1.8 2.7 ND ND 228 0.54 M05 ND 1.4 5.1 ND ND 0.9 0.7 ND ND 8.1 0.63 M15 0.6 4.6 13.2 ND ND 2.3 23 ND ND 220 0.57 M 16 0.6 4.4 13.0 0.3 0.9 2.4 2.2 ND ND 23.9 0.54 M 17 0.5 3.0 9.5 ND ND 1.8 1.4 ND ND 16.2 0.59 M06 0.3 2.7 7.1 ND ND 1.4 1.2 ND ND 127 0.56 M07, 08, 09, ND ND ND ND ND ND ND ND ND ND 10

N D = not detected

83 Table 2.9. Comparison of mean congener specific PCB (ng/g, ppb, lipid weight) with means of Salmon samples from this study (n=8) and Parsley et al. 1999 (n=12)

Salmon Herring Mean ± SD Mean ± SD (ng/g lipid weight) (ng/g lipid weight) PCB This study Parsley et al Parsley et al congener (n=8) 19 99 19 99 (n=12) (n=10) 18 2.8 ± 1.8 1.3 ±0.6 1.2 ±0.5

28 7.4 ± 8.7 4.3 ± 1.3 8.0 ±3.6

31 4.0 ±1.6 2.7 ± 0.8 3.5 ±1.5

47 3.0* ± 1.0 2.1 ±0.6 4.4 ± 2.2

49 4.4 ±1.9 4.0 ±1.2 4.4 ±2.1

52 10.2 ± 3.7 10.2 ±3.4 17.8 ±8.7

99 10.2 ±4.2 9.6 ±2.9 29.8 ± 14.2

101 19.0 ±9.0 22.8 ± 6.8 55.7 ±25.9

105 5.5 ±1.8 6.3 ± 1.8 20.3 ± 10.5

114 0.8 ± 0.4 0.4 ±0.1 0.9 ± 0.5

118 18.3 ±6.5 19.0 ±5.8 60.3 ± 28.7

123 1.4 ±0.5 0.9 ± 0.3 2.4 ±1.2

128 2.7 ± 0.6 4.6 ± 1.4 14.7±7.2

138 25.8 ± 5.4 34.3 ± 10.0 110.5 ±46.8

153 37.0 ± 8.7 37.2 ±11.1 129.7 ± 70.0

156 1.9 ±0.4 1.8 ±0.6 5.3 ±3.1

167 1.0 ±0.3 1.1 ±0.3 2.5 ± 1.3

180 10.9 ± 2.7 11.6 ±4.5 40.7 ± 20.4

Lipid (%) 13.3 ±4.9 13.6 ±3.3 9.6 ± 2.9 *PCB 47/48

Wild salmon from polluted waters contain higher concentrations of PCBs. Up to 3500 ng/g lipid for the sum of 6 PCB congeners are reported for Baltic salmon (Asplund et al., 1999). In Lake Michigan Coho and Chinook salmon {Oncorhynchus kisutch and Oncorhynchus tshawytscha), mean concentrations of PCBs (78 congeners) were 39,870 and 53,638 ng/g lipid (Jackson et al., 2001), much higher than in the farmed salmon analysed in this study.

84 PCB profiles are more species dependent for fish and birds than for mammals, due to selective metabolism depending upon the degree of chlorination. Fish species can accumulate lower chlorinated PCB congeners (such as tetra- and penta-PCBs), some of which have the highest TEF values among dioxin-like PCBs (van den Berg et al., 1998). Indeed, levels of PCBs are higher in fish compared with other farm raised animals (such as chicken, beef or pork) intended for human consumption. The concentrations of organochlorine pesticides in salmon compare favourably with those reported for farmed salmon obtained during 1997 in the UK. XDDT concentrations ranged from 10 to 40 ng/g whole weight (n=14), y-HCH concentrations ranged from ND in 11 samples to 2 ng/g whole weight in 3 samples, and HCB ranged from 2-5 ng/g whole weight (n=14) (Ministry of Agriculture Fisheries and Food / Health and Safety Executive, 1998). The UK does not have statutory Maximum Residue Levels (MRLs) for these pesticides in fish, although they exist for other food products (Ministry of Agriculture Fisheries and Food / Health and Safety Executive, 1998). The average concentration across all salmon samples of the XBDE congeners was 33.7 ng/g lipid. The results from this study are an order of magnitude lower than the respective PBDE values for salmon reported in recent studies from polluted open waters. Previous reports have detected significant levels of PBDEs in fatty fish such as sprats, herring and salmon from the Atlantic and more polluted waters such as the Baltic (Bergman, 2000; Damerud et al., 2001), showing age related accumulation and biomagnification. The concentrations of PBDEs in the Atlantic salmon were markedly lower than Lake Michigan Coho and Chinook salmon (Manchester-Neesvig et al., 2001), (mean 2440 ng/g lipid, range 773 - 8120 ng/g lipid). Salmon from Lake Michigan tributaries in the US appear to have the highest PBDE concentrations so far reported worldwide, several orders of magnitude greater than the concentrations reported here, but with a similar congener distribution (Manchester-Neesvig et al., 2001). PCB contamination of the farm-raised salmon comes via aquaculture feed, while for the wild salmon, sources are more diffuse, but clearly biomagnify through trophic levels (Jackson et al., 2001). One of the routes of food chain contamination may be partly as a consequence of ingestion of plastic resin pellets (Mato et al., 2001) by the marine wild fish used in the manufacture of fish meal. There is very little literature available for PCB and organochlorine pesticide comparison in fish feed. Data from the 1970’s exists for Aroclor 1254 in nine feed samples (range 0.05 - 0.23 pg/g whole weight) and p,p ' DDE

(range 0.05 - 0.34 pg/g whole weight) (Mac et al., 1979). Fish meal, a principal component of fish feed, shows a much higher degree of contamination with organochlorines if sourced from North Atlantic stocks than those from South Pacific

85 (Scientific Committee on Food, 2000b). Thus, fish meal originating from Denmark and the Faroe Islands contained 120 and 250 ng/g lipid for the 7 marker PCBs, respectively, while fish meal from Peru had a concentration of 10 ng/g lipid (Schepens et al., 2001). The PCB concentrations in fish oils from this study are consistent with the study reported by the UK Ministry of Agriculture, Fisheries and Food (now the Food Standards Agency) (Ministry of Agriculture Fisheries and Food, 1997), with surprisingly high levels of PCB congeners. A previous study (Jacobs et al., 1998) has found relatively high levels of PCBs 101, 118, 138, 149, 153, 170, 180 in both a herring oil sample from Germany and a salmon oil sample. For example, the concentration of PCB 118 was 158 ng/g lipid in the herring oil and 60 ng/g lipid in the salmon oil. The average concentration of the EBDE congeners in the analysed feed samples was 16.6 ng/g lipid. At the present time, comparative data are not available in the public domain, although since this work was published the aquaculture industry state that they monitor routinely, but results of PBDEs detected in samples are not reported (Nutreco 2003). The target PBDEs were not detected in all but one of the oil samples, a fish oil from the Northern Hemisphere, with a sum PBDE concentration of 12.7 ng/g lipid. The data available in the literature concerning organohalogen contamination of fish oils vary widely, with higher contamination levels observed from more polluted Northern hemisphere waters (Scientific Committee on Food, 2000b; Jacobs et al., 1998; Falandysz et al., 1994; Nutreco, 2003). Wild salmon from Lake Michigan have been shown to have a significant relationship with the congener concentrations in lower trophic levels of invertebrates. PCB congeners biomagnify 20-30 fold, with a more efficient transfer of penta- and hexa-PCBs through the pelagic food web (Jackson et al., 2001). The risks of exposure to organochlorine contaminants from eating relatively contaminated fish have been known for many years, but how to quantify the risks to human health is frequently disputed. While PCBs and p,p'-DDE provide the bulk of organochlorine residues in human tissues (Longnecker et al., 1997), there is a lot of variability in regulations for PCB and pesticide concentrations in food worldwide. Many EU member states do not have national maximum limits for concentrations of PCBs in food (European Commission, 2000). However in Holland, the current norm for fish designated for human consumption is 620 ng/g lipid for the sum of the 7 marker PCBs (RIVO Fisheries Institute of Holland, 2000), a value lower than the concentrations observed in Baltic or Lake Michigan salmon. Due to comparatively lower toxicity, the risk assessments of non-dioxin-like PCBs have been given a lower priority than dioxin like PCBs, but with the proviso that the PCB risk assessment (IPCS, 1993) should be updated on a congener specific toxicological evaluation to provide an adequate risk

86 assessment in the near future (Scientific Committee on Food, 2000b). In terms of risk assessment, the potential contribution to the human diet of PCBs, organochlorine pesticides and PBDEs from farmed European Salmon will vary according to the diet and age of the fish, the frequency of consumption, portion size, cooking practices and the age of the consumer. The possible contribution to dietary intakes of organohalogen compounds from farmed salmon could clearly be significant for high consumers, particularly if pregnant or breast feeding. As the PCB data are corroborated (Parsley et al. 1999), table 2.9., national extrapolations could be suggested on the basis of this study, but not for the other target contaminants, due to the relatively small sample size, and lack of available comparative data. This study recommends that further investigation of farmed salmon and salmon feed, including feed fortified with fish oil and feed fortified with selected vegetable oils, is warranted, together with consistent regular vigilance by health authorities for PCBs, organochlorine pesticides and PBDEs, to monitor and reduce current human dietary exposure. While PCBs and many organochlorine pesticides have been banned in most of the world, they are still being delivered in the diet and European farmed salmon can be a significant source. In addition, in some cases there are suggestions of recent usage, particularly of DDT. Whilst diets based on marine fish oils are currently favoured by the aquaculture industry, it is likely that these oils are contributing greatly to the contamination of farmed salmon by POPs. The degree of variations in levels of contaminants for the feed and fish oil data indicate the significance of changes in feedstock and variations in oil sources and processing procedures that can result in batch-to-batch differences in contaminant levels even in the final product from a single manufacturer. The generation of additional data is needed to further assess the importance of batch variation in animal feeds, this being conducted now by the aquaculture industry (Nutreco, pers comm.) in accordance with EU requests. The aquaculture industry are also conducting analysis to detect PBDEs, but the data are not easily accessible, and are not provided in the Nutreco Social and Economic report published in May 2003 (Nutreco 2003). There is experimental evidence that suggests aquaculture diets utilizing vegetable oils (with both n-3 and n-6 fatty acids), and having fatty acid compositions which resemble those of invertebrates that comprise the natural diet of salmon, could be more beneficial in accommodating successful sea water adaptation than diets based on marine fish oils (Bell, 1998). Vegetable oil based diets could also facilitate the requirement for high- energy aquaculture feeds on an economically competitive basis, whilst reducing the problems of organochlorine contamination. This would reduce the reliance on fish meal and oils from non-sustainable natural resources. Contamination problems do exist with

87 vegetable oils but to a far lesser degree than fish oils and this is examined further in chapter 2.4 (Jacobs et al., 1998; Jacobs et al., 2002c; Jacobs, 2003). 2.4 Investigation of selected PCBs, PBDEs and organochlorine pesticides in omega -3 fatty acid rich dietary fish oil and vegetable oil supplements

2.4.1. Introduction

Fish oil is a by-product of the fish meal manufacturing industry and comes from many different parts of the world. Fish oils are sold for dietary purposes due to the presence of long chain omega-3 (n-3) fatty acids which have certain health benefits and are essential to the human diet. Specific vegetable oils and foods are sources of the short chain form that is the precursor to the long chain form utilised in human metabolism. Essential fatty acids are needed for growth and repair of nervous tissue and for the maintenance of its structure. The consumption of long chain omega-3 essential oils from fatty fish or concentrated sources such as fish oil supplements are thought to have associated benefits in reducing mortality from heart disease and improving symptoms of a number of diseases including multiple sclerosis, rheumatoid arthritis and osteoporosis. However, oily fish and extracted fish oils are a major source of lipophilic contaminants entering the human food chain (MAFF., 1997; Jacobs et al., 1997). Chlorinated compounds accumulate in the lipid compartment of the animal, and oil extracted from fish caught in polluted waters may be contaminated with chlorinated hydrocarbons. UK government officials have met with the fish oil industry to discuss how dioxins and polychlorinated biphenyls (PCBs) can be reduced in dietary supplements and medicinal products containing fish oil (MAFF 1997). The market for fish oil as a supplement to large sections of the population has been noted in the UK (Hamilton and Rice, 1995; Gregory et al., 1990); see table 2.10 for manufacturers predictions. The Diet and Nutritional Survey of British Adults 1990 found that 17% of the women respondents took dietary supplements, commonly fish oil (Gregory et al., 1990). Fish oils have also been traditionally given to children to protect against Vitamin D deficiency and rickets. As quantities consumed can vary enormously, it is important that close monitoring is undertaken to ensure contaminant levels are greatly reduced.

89 There is an extensive and growing body of data on the presence of persistent organic pollutants (POPs), including polybrominated diphenyl ethers (PBDEs), in the aquatic environment, biota (de Boer and Allchin, 2001; Jackson et al., 2001; Manchester-Neesvig et al., 2001; Jacobs et al., 2002) and fish oils designated for aquaculture feed (Jacobs et al., 2002). However, there are few data available in the public domain, for PBDEs in fish oils, and other essential fatty acid containing dietary oils. This study reports the levels of selected PCBs, organochlorine (OC) pesticides and PBDEs, in 21 omega-3 rich dietary fish and vegetable oil supplements obtained through retail and company outlets in London, UK. The sources and brands of the dietary fish oil supplements were the same as those previously reported (Jacobs et al., 1998).

2.4.2. Materials and Methods 2.4.2.1. Sites and sampling 21 dietary supplements rich in omega-3 fatty aeids (7 cod liver oil supplements, 7 whole body fish oil based supplements, 3 vegetable and fish oil combination supplements and 4 vegetable oils) were obtained from retail and practitioner suppliers of the UK market in December 2001 and January 2002 (Sample descriptions are given in Table 2.10). Samples were chosen to cover the products available to the UK consumer, including a neighbourhood pharmacy, a national high street chain pharmacy, whole food retailers and dietary supplement companies selling direct to nutritional therapists. While the samples cover the leading brands available on the London/UK market, this was not a comprehensive survey of all brands available.

2.4.2.2. Sample preparation The samples were logged and prepared in duplicate pre-washed glass vials. They were kept at room temperature in the dark. All samples were analysed for 28 PCB congeners n°: 18, 28, 31, 52, 74, 95, 99, 101, 105, 110, 118, 128, 132, 138, 149, 153, 156, 163, 167, 170, 174, 177, 180, 183, 187, 194, 196, 199. All samples were

90 also analysed for HCHs (3 isomers), HCB, DDT and metabolites (5 op- and pp- isomers) and 7 PBDEs congeners (n°: 28,47,99,100,153,154,183).

2.4.2.3. Sample analysis 0.5-0.7 g of the sample oils were solubilised in 3 ml hexane, internal standards

(10 ng of PCB 46 and 143, 5 ng of e-HCH and 1 ng of BB 80, BB 103 and BB 155) were added and the mixture equilibrated in an ultrasonic bath for 5 min. The extract was applied to a hexane pre-washed SPE cartridge filled with 5 g acidified silica and eluted with 15 ml of hexane and 10 ml of dichloromethane. The final eluate was concentrated with a rotary evaporator and under nitrogen to approximately 100 pi (ready for GC analysis).

2.4.2.4. PCB and OCP determination One pi of extract was injected in pulsed splitless mode into a GC-pECD equipped with a 50m x 0.22mm x 0.25pm, HT-8 capillary column. For confirmation, one pi was injeeted in pulsed splitless mode into a GC/MS equipped with a 25m x 0.22mm x 0.25pm, HT-8 capillary column. The MS was operated in electron impact ionisation and SIM mode. Two most abundant ions were monitored for each level of chlorination for PCBs or for each pesticide. Method limits of determination ranged for individual compounds ranged between 0.1- 0.2 ng/g oil. Recoveries of internal standards ranged between 75 and 85%.

2.4.2.5. PBDE determination One pi of extract was injected in cold splitless mode into a GC/NCI-MS equipped with a 10m x 0.10 x 0.10pm, HT-8 capillary column. The MS was operated in the negative chemical ionisation in SIM mode (ions: 79 and 81). Method limit of determination ranged between 0.05 and 0.1 ng/g oil for individual PBDE congeners.

91 2A.2.6. Quality Control All samples were analysed with GC-pECD and GC-MS systems. The lowest result for each compound obtained in both systems was considered for further calculations. The procedure was validated through regular analysis of procedural blanks, certified material CRM 350 (PCBs and organochlorine pesticides in mackerel oil) and through successful participation to Quasimeme interlaboratory studies (PCB determination in sediment and fish).

2.4.2.7. Oil density Densities of oils were determined gravimetrically and were found to lie between 0.87 and 0.90 g/ml. Thus, direct comparisons given in pg/1 with those given in pg/kg or ng/g (ppb) should allow for an approximate 10% underestimate of concentrations expressed on a volumetric basis.

2.4.3. Results and Discussion

All fish oil samples contained detectable residues of organochlorine contaminants, with cod liver oil (Samples 1-7) having similar levels as previously reported (MAFF, 1997; Jacobs et al., 1997; 1998) and the greatest levels of POPs contaminants compared to the other supplement groups in this study. Fish and vegetable oil mixtures (Samples 14-17) had lower levels than the whole fish body oils, while no PCBs were detected in the vegetable oils (Samples 18-21). On the whole, the fish oils showed less contamination compared with past reports (MAFF, 1997; Jacobs et al., 1998; Falandysz et al., 1994; Falandysz, 1994), suggesting that improved sourcing, developments in processing such as deodorisation and nutritional profile development in creating specific ‘optimum’ essential fatty acid balances (Samples 8, 14-17, 21), are improving the quality of such supplements for the consumer. With most of the samples, the predominant pesticide isomers were p,p'-DDE and y-HCH. Summary results are presented in Table 2.10, with total sum (X) values given using the LOD equal to zero where there were no detections. Table 2.11 gives full congener specific data for each sample. In comparison with the samples purchased

92 eight years ago (Jacobs et al., 1997, 1998), there is an evident reduction of contaminant loadings. Most fish oil samples showed evidence of steam distillation, and this probably contributed to the reduction in contaminant levels. HCHs and HCB levels were very low, only the a- and y-HCH isomers were detected. DDTs were present in all but one sample (Sample 21, a vegetable oil mix from Canada), while concentrations of PCBs were under the detection limit for some samples. The oils with higher concentrations of PCBs had low concentrations of volatile compounds, an indication of thermal processing. The PCB profile was dominated by hexa and penta congeners followed by hepta, with some variation in percentages of different PCB congeners depending upon the type of fish oil and sample formulation. In the fish oils, higher levels of DDTs were correlated with higher PCB concentrations (r^= 0.78), Figure 2.4. PCB concentrations in the fish oil samples were between ND and 90.9 ng/g lipid. These concentrations are generally of the same order of magnitude, but are not as high, as previously reported based on ICES 7 PCBs (Jacobs et al. 1998). This is probably a consequence of the reduction of OC contaminants during refining (Hilbert et al., 1998), as requested by UK government departments (MAFF, 1997). Comparisons with unrefined fish oils would indicate whether a real reduction in the steady state of PCBs has occurred (Falandysz et al., 1994; Falandysz, 1994). The ICES 7 PCBs consistently constituted over 50% of the sum of PCB congeners in all cod liver oil and fish oil samples where PCBs were detected. The PBDE levels in cod liver oils ranged from 14.6 to 34.2 ng/g and were greater than those reported in samples collected three years previously, as shown in sections 2.3 and 2.4 (Jacobs et al., 2002). The fish and fish/vegetable formulations showed PBDEs levels which were an order of magnitude lower than the cod liver oils and all but one fish/vegetable oil sample (Sample 17) had detectable levels of PBDEs. No PBDEs were detected in the pure vegetable oils. The most abundant PBDE (BDE 47) was detected at concentrations comparable to PCB 118. PBDEs showed the same trend as PCBs (r^= 0.94), Figure 2.5. These results are in good agreement with the levels of PCB and OC pesticides previously reported for fish oil supplements (Jacobs et al., 1998; Fromberg, 2001), and for congener specific PCBs reported by the UK

93 Food Standards Agency and the Irish Food Safety Authority for fish liver oils (Fernandes et al., 2002; Food Safety Authority of Ireland, 2002), but with higher levels of PBDEs than those recently reported (Jacobs et. al., 2002). A more detailed comparison with other studies is hampered by the lack of congener specific data (e.g. Fromberg 2001), or because different contaminants, such as dioxins and dioxin-like PCBs, were analysed and reported in other relevant and recent studies (Fernandes et al., 2002; Food Safety Authority of Ireland, 2002). More detailed time trend comparisons are made in the next chapter, 2.5. Although fish are a significant source of POPs in the diet, fish and fish oil consumption are associated with numerous health benefits due to the long chain omega-3 essential fatty acids. There are alternative sources of omega-3 fatty acids (n-3 PUFA) with evident lower contamination levels, a-linolenic acid (ALNA), abundant in rapeseed and flax (linseed) oil, is the short chain natural precursor found in the vegetable oil based nutritional supplements used in this study. These sources offer an abundant and accessible source of dietary n-3 PUFA that can be further elongated and unsaturated in vivo, to contribute to the health benefits associated with a reduction in the n-6:n-3 ratio, which is currently considered to be too high for most of the UK population (de Longeril et al., 1999; Hu et al., 1999). There is growing but still limited evidence of nutritional benefits of dietary ALNA, raising questions about the extent and efficiency of its long chain conversion, especially its conversion to DHA. A number of human clinical studies have demonstrated good conversion of a-linolenic acid to the metabolically active eicosapentaenoic acid (EPA) in infants and adults, but decosahexaenoic acid (DHA) synthesis is reported to be much lower (Emken et al., 1994; Sauerwald et al., 1996; Wilkinson et al., 1999; Uauy et al., 2000; Pawlosky et al., 2001; Emken et al., 2001; Millward and Griffin, 2002). ALNA long chain conversion rate is known to vary with the dietary ALNA intake and dietary LA inhibits the conversion. ALNA can be converted to EPA and increase phospholipid EPA levels. Indeed, stable isotope studies show that ALNA can serve as a significant source of EPA, with a very approximate equivalence of 10:1 with dietary EPA, in subjects where the conversion is optimised by a low n6:n3 ratio (Millward and Griffin

94 2002). This observation is also reflected in the dosages recommended by the manufacturers and producers of the omega 3 rich supplements, as can be seen in Appendix 1, Tables i and ii. Millward and Griffin, 2002, observed extensive and rapid conversion of ALNA to EPA and in contrast to the compositional data which showed no detectable increase in phospholipid DHA, the isotopic results showed that some DHA formation did occur in all subjects, albeit at a very low rate. In the study reported by Millward and Griffin, ALNA was given orally to a small cohort of normal, healthy male subjects. The subjects received either a diet enriched with flaxseed-oil (high ALNA, n=6) or a diet enriched with sunflower oil (high n-6 diet, n=5) and the conversion to EPA and DHA was measured over a two week period from their enrichments in phosphatidylcholine fractions isolated from plasma and erythrocytes. There was substantial conversion of ALNA into EPA with very much less further conversion to DP A or DHA, although significant enrichment was observed at 24 hrs in plasma DHA on both diets and after 7 days in red blood cell DHA in all subjects. The 14 day mean total distribution between ALNA, EPA, DP A and DHA, was 28%, 52%, 16% and 4%, respectively, with no differences between the two diets. Millward and Griffin, 2002, report an interesting observation, with respect to conversion efficiency, in one subject studied after a fish oil diet conversion was very much reduced. Such stable isotope studies are limited due to the high cost involved, but are generating very useful data for public health purposes. In the last ten years there has been clear shift in accepted wisdom, from the traditional position that humans do not synthesise long chain fatty acids efficiently, to the current situation, where robust scientific evidence suggests that efficient conversion to EPA can and does occur. At the bottom of the food chain, DHA is originally synthesised by phytoplankton and krill {Euphausiapacified) (Saito et al., 2002). Species further along the food chain preferentially accumulate DHA, but appear to be unable to synthesise it effectively. DHA is of nutritional importance in fetal development, so nutritional sources of DHA are often promoted for pregnant and breast feeding women. However, dietary DHA is not necessarily essential throughout life, it appears that DHA turns

95 over very slowly, much more slowly than EPA, which is consistent with its low rate of conversion to eicosanoids. DHA synthesis may be regulated through additional pathways or independently from EPA. Metabolic demands for DHA appear to be very low, especially once tissue pools are sufficient. Indeed, existing metabolic levels of DHA probably represent what is appropriate for normal healthy metabolism, with levels rising where metabolic demands for DHA may be increased such as infection and pregnancy. A well designed ethical study of pregnant women, particularly at term, when the cord blood could be sampled, would be needed to confirm such a hypothesis. DHA has been shown to be toxic in vivo in advanced age. Supplementation with DHA resulted in oxidative DNA damage in the bone marrow of aged rats (Gress et al., 2002). Taken together these studies suggest that if DHA is not found to be synthesized to any significant degree in the studies conducted so far, perhaps this is because it is not required in significant amounts for the general population’s healthy metabolism, and is only produced when required. In terms of risk assessment, the potential contribution to the human diet of POPs from omega-3 fatty acids food sources increases if sourced from fish, particularly cod liver. The vegetable oils studied here presented little or no POP contamination, but approximately ten times the quantity needs to be consumed to provide comparative levels of EPA, and this increase in quantity needs to be accounted for when calculating comparative dietary intakes of POPs contaminants from these dietary sources. Furthermore, supplementation with fish oils may impair a person’s subsequent metabolic ability to convert EPA from vegetable sources of ALNA ( as observed for one of the subjects observed in the Millward and Griffin study). Although this is probably a transient effect, for a complete risk benefit analysis of the health benefits of fish oils versus the pollutants contained within the fish oils, this would need to be taken into account when reviewing the literature. The possible contribution to dietary intakes of POPs from cod liver oils currently available in UK and international markets could clearly be significant for high consumers, as has been shown previously (Jacobs et al., 1997; 1998; MAFF,

96 1997). Although preliminary, this investigation strongly supports the need for monitoring POPs, including the newer POPs, in fish oils designated both as nutritional supplements for consumers and as animal feed ingredients.

97 l a a s /ia a < f ^ ooooooooooooo o o d liî ON ■ÿ lï 5 s 3 5 | 00 On 1-H o s 2253 1 sia a z 0\ v-4 vM Tf m m m ^ 2S3i P o\ r- 00 r- M t" es aaa-o r- lev en O ‘^a a - d d g B'S .S s ^ 'H f—I rH t-H vo es d d d Z .-s Q o 00 Tf N û Q Q Q Ov en vo en o a vd u SHDHZ 00 Æ un z o\ T-i vi m Z o Z Z Z un d d (% g % 1 fis Ci Q o\ n /S o 'H Z Z Z Z Z ssi u-i 'H d d s ll§ Q 'H o\ q q Q Tf es Q Q SDH Z Ov (N f ' Tf Z en d Z Z iiii ip I saaad o\ Cl \o N m \e m q q Tf vq Ov 1 LZ iiii qSflDd c/3 Ii Ov Ov m 'O ». -H- . ,-1 00 c c I vo vo o LZ ov 00 IsS iiii . s IP! (o=aM) c». m» w) ^c ^ c c m 1 « sflD d Z v o t ' v o rH rH »S| I r H t - l t H r«^ (M rv| C4 C e - en z^iiS iSi iiii Il CQ v^ VI Tf q OO O v 3 B p o 2 O Sen CS2 M 2 glli CL ii iiii

vo vo vo Bidaq 3 Q ov q —^ ^ I Z Ov es oo' 00 z z z z 0 n N A m (7 c q vq un BX3H3 O O ON 00 VO en CS d Tf Tf 00 es 1 O n 1—^ O N O n i - H Z en es es en iiii Pi O ON d -1 TT vo B)U3d 3 O n vo ON es en oo v4 en I en en vo ^ VI i s iiii Iîl 2 î? t ON es en Q .H Q Q Q os BJPÎ 3 d Tt vd Tf Z es Z Z Z >z5 iiii - a 00 03 g m z ' ^ en' Û oq On en ON sflDd Z es es oô VD iiiiS iiii *(^) îAV O es vo es ^ siduiBS vo vo vo NO vo VO vo 3 vo ICfl d d d d d o o o o o o d d d d Iffi ?1|S 8 g C § -' PL u;3uojo o 2 00 o Xj;uno[) is iii 3 IIP: « Æ g »- CL sism i I 11 _ .'2-3 •’i I l\b g I S; 'î "O h ègg " i §l§2 = •- -B* -o 8 im fi üll Ç u o o C lue S „ ail B g s s I W « o " f i Ci îi III SsêÊg îtéUllîiï' f | | S lill r I •B £> g j OS H II Ils II

11 h 1-1 c 4 m ^ I/) vo I' U Oi l§ Table 2.11: Congener specific PCBs and PBDEs in fish and vegetable oils rich in omega -3 fatty acids: lipid adjusted (ng/g, ppb)

oxlliAerdls fishoils Miand\^daHeoils \ŒtaHecils i^'glipid sm S02 SOB SM SK SQ5 S07 SOB S09 SIO SU S12 S13 S14 S15 S16 S17 SI8 S19 sao S21 VIKBN PCB 18 NDND 02 0.6 15 0.7 02 ND NDND ND 02 02 ND ND ND ND ND ND ND ND 3.3 PCB28 ND M) 0.8 0.9 1.1 0.8 ND ND ND ND ND ND ND ND 1 ND ND ND ND ND ND 12 PCB31 ND ND 1.8 1.4 5.7 5.4 ND ND ND ND ND ND ND NDND ND ND ND ND ND ND 166 sumtri 0 0 2 8 2 9 83 6 9 02 0 0 0 0 0 2 0 2 0 1 0 0 0 0 0 0 319 PCB52 1.6 ND 2 8 27 1.9 1 24 ND 1.8 ND ND ND 3 ND 25 ND ND ND ND ND ND 47.9 PCB74 05 0.7 21 4 23 3.3 13 ND 0.3 ND ND ND 0.8 ND ND ND ND ND ND ND ND 31.3 sutntetra 2 2 0 7 4 9 6 7 4 2 4 3 3 8 0 21 0 0 0 3 S 0 25 0 0 0 0 0 0 792 PCB95 35 3.6 11.3 10.7 9.7 102 4.1 ND 4.4 ND 0.3 0.6 4.6 ND 42 ND ND ND ND ND ND 91.8 PCB99 5 4.8 7.4 15.7 9.4 82 66 ND 3 3.7 12 1 3.9 ND 15 4.9 ND ND ND ND ND 75.6 PCB 101 67 7 128 226 14 152 9.3 ND 4.6 63 1.6 1.9 7.1 ND 2 8 4 ND ND ND ND ND 144.6 PCB 105 5.4 4.3 85 14 84 102 7.4 ND 4.1 5.3 25 45 5 ND 15 2 ND ND ND ND ND 43.9 PCB 110 45 5.7 10 183 10.3 11.7 7.3 ND 4 5.9 1.7 2 55 ND 21 1.6 ND ND ND ND ND 118 PCB 118 13.9 115 19.6 39.6 21.3 229 19.4 ND 86 11.8 61 7 125 ND 35 11 ND ND ND ND ND 1583 sunpEita 39 369 695 1209 "ai 784 5 t l 0 286 329 B 4 17.1 385 0 154 136 0 0 0 0 0 6323 PCB128^m 5.8 63 5.3 122 7.1 65 86 ND 21 26 2 29 32 ND 22 02 ND ND ND ND ND 53.8 PCB 132 32 3.6 4.1 6 4.4 4.8 42 ND 2 1.3 2 23 3 ND 21 3 ND ND ND NDND 60.9 PCB 138 272 31.1 21.8 568 287 272 39.3 ND 10.6 7.6 5.9 7.6 10.4 ND 7.6 7.1 ND ND ND ND ND 2229 PCB 149 63 7.8 10.6 167 112 13 9.8 ND 4.1 2 25 26 5.4 ND 3L7 1.4 ND ND ND ND ND 1587 PCB 153 31.6 355 21.6 621 31.9 31.1 45 ND 11.8 86 11.3 10.9 13.9 ND 9.6 65 ND ND ND NDND 329.3 PCB 156 3.7 32 23 62 35 3.8 4.6 ND 25 12 25 3.9 22 ND 12 1.6 ND ND ND NDND 186 PCB 163 10.3 111 9.3 183 6 5.7 63 ND 1.3 ND 21 1.6 2 ND 15 ND ND ND ND ND ND 925 PCB 167 22 4.1 11 5.3 1 1.3 2 8 ND ND 15 02 0.6 0.3 ND 1 02 ND ND ND NDND 11.3 sunhraa 903 1027 79 1837 969 933 1205 0 344 248 286 325 404 0 288 20 0 0 0 0 0 948 PCB 17) 4.6 55 21 81 3.9 3 9 7.8 ND 05 0.3 12 1.8 1 ND 05 ND ND ND ND ND ND 34 PCB 177 ND 0.7 23 0.9 3.1 26 12 ND 38 ND 0.3 0.6 NDNDNDND NDND ND ND ND 41.3 PCB 180 11 11.6 52 17.1 89 9 18 ND 1.3 0.8 3 45 22 ND 3 ND ND ND ND NDND 61.9 PCB 183 3.3 3.6 ND 4.9 2 8 26 4.6 ND 1 0.3 1.8 ND 1.3 ND ND ND ND ND ND ND ND 238 PCB 187 87 10 9 126 84 82 126 ND 3.3 1 1.8 11 35 ND ND 0.7 ND ND ND NDND 715 sunhE{Éa 27.6 314 186 436 27.1 264 44 0 99 25 81 81 8 0 3 S 07 0 0 0 0 0 2254 PCB 191 ND ND ND ND ND ND ND ND ND ND ND ND ND NDND ND ND ND ND ND ND ND PCB 196 ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND NDND ND PCB 199 1.6 2 05 24 0.8 1.1 25 ND ND ND ND ND ND ND ND NDNDND ND NDND 27 sunoda 16 2 05 2 4 0 8 Ll 2 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 7 sumFCBs 1607 1717 1754 3601 2104 2103 2251 0 75 603 501 579 909 0 512 3M 0 0 0 0 0 1919.6 EŒ28 0.3 0.3 05 1.3 05 0.4 0.4 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0,0 H:E47 10.9 127 9.7 263 127 9.9 14.8 0.8 0.8 12 1.9 1.3 1.9 0.3 1.4 0.9 0.0 0.0 0.0 0.0 0.0 BŒ100 1.9 27 15 4.1 2 8 21 29 0.0 0.0 0.0 0.3 0.3 0.3 0.0 02 0.0 0.0 0.0 0.0 0.0 0.0 h :e 99 0.7 2 8 22 12 3.4 24 25 0.0 0.0 02 0.4 02 05 0.0 0.3 02 0.0 0.0 0.0 0.0 0.0 H E 151 1.0 12 0.4 1.3 0.6 05 1.4 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 H E 153 0.0 0.4 02 0.0 0.4 0.3 0.3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 EŒ183 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 suttiBCE 149 200 146 342 203 156 223 08 0 8 14 26 19 27 0 3 19 11 oo 00 0 0 0 0 0 0 HB 0.0 0.0 9.1 29 7.3 4.9 0.0 0.0 3.4 02 0.0 0.0 0.8 0.0 42 0.0 0.0 0.0 0.0 0.0 0.0 a-KH 0.0 0.0 4.0 0.9 42 1.9 0.0 0.0 05 0.0 0.0 0.0 05 0.0 1.0 0.4 1.0 0.0 0.0 0.0 0.0 &KH 0.0 0.0 5.1 0.9 12 1.3 0.0 0.0 0.0 0.0 0.0 0.0 0.3 0.2 0.7 0.0 15 5.9 1.3 0.6 0.3 simHCK 0 0 0 0 9.0 18 5 4 3l2 0 0 0 0 0 5 0 0 0 0 0 0 0 8 0 2 17 0 4 2 5 5 9 13 0 6 0 3 pp-ETE 564 567 44.9 1063 620 505 75.3 1.6 145 105 10.3 6 3 20.8 123 14.4 486 05 0.7 05 0.3 0.0 o,p^DCD 82 84 67 85 7.1 83 9.6 0.0 0.0 5.0 4.7 0.0 55 4.3 5.8 1.3 0.0 0.0 0.0 0.0 0.0 pp-EDT 64 9.7 3.1 72 29 0.0 9.1 0.0 0.0 0.0 7.1 0.0 0.0 0.0 1.8 5.1 0.0 1.3 0.7 0.0 0.0 ppCCD 44.1 55.1 329 73.7 368 387 652 1.3 10.2 11.4 7.6 1.9 127 27 11.7 102 0.6 0.8 05 0.7 0.0 pp-DDT 227 260 120 283 7.9 7.4 342 1.9 5.6 45 63 7.1 81 22 62 77.8 0.0 5.4 23 0.0 0.0 sunEDTs 137.7 1S&9 99.6 m i 1167 1019 isa3 48 303 314 359 154 47.1 214 m 1429 11 8 2 3 9 10 0 0

99 Figure 2.4 Correlation between concentrations of PCBs versus DDT in omega-3 rich dietary fish oil and vegetable oil supplements

PCB vs DDT

250.0 y = 0.6122X+ 10.977 R2 = 0.7799^^ 200.0

150.0 DDT 100.0

50.0

0.0 0.0 50.0 100.0 150.0 200.0 250.0 300.0 350.0 400.0 PCBs

Figure 2.5 Correlation between concentrations of PBDEs versus PCBs in omega-3 rich dietary fish oil and vegetable oil supplements

PBDE vs PCB

400.0 y = 9.7265X + 20.25 350.0 Ff = 0 .9 3 9 j^ ^ 300.0

250.0

150.0 100.0 50.0 0.0 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 ______PBDEs______

100 Figure 2.6 PCB profiles in omega-3 rich dietary fish oil supplements, ng/g lipid

□ PCB 18 □ PCB 28 □ PCB 31 □ PCB 52 a PCB 74 □ PCB 95 □ PCB 99 □ PCB 101 a PCB 105 aP C B IIO □ PCB118 □ PCB 128/174 □ PCB 132 ■ PCB 138 □ PCB 149 ■ PCB 153 □ PCB 156 □ PCB 163 □ PCB 167 □ PCB 170 Sample □ PCB 177 □ PCB 180 □ PCB 183 □ PCB 187 a PCB 194 IÛ " y Ü □ PCB 196 □ PCB 199 PCB congener

101 Figure 2.7. OC pesticides in omega-3 rich dietary oils

□ HOB

0a-HCH 250.0 □ g-HCH

□ sum HCHs 200.0 ■ p,p-DDE

150.0 □ o.p-DDD ng/g lipid Qo.p-DDT

100.0 □ p,p-DDD

Hp.p-DDT 50.0 □ sum DDTs

p-DDT o,p-DDT p.p-DDE

g-HCH 00 pesticides oil samples HOB

Figure 2.8 PBDE congeners in omega-3 rich dietary oils

□ BDE28

OBDE47

□ BDE100 35.0 □B D E99 30.0 IBDE154

OBDE153

ng/g HBDE183

□ sum B DE

oil sample

% ; IP : g g I s s a “ "

PBDEcongener

102 2.5. Principal Components Time Trend Analysis Of Organochlorine Contaminants Detected In Omega- 3 Rich Fish And Vegetable Oil Dietary Supplements from UK outlets

2.5.1. Abstract

To examine contaminant variability over time, the fish oil and vegetable oil dietary supplement data obtained in the preceding chapter 2.4, were compared on a congener specific basis with similar or the same brands as those analysed in a study conducted eight years ago. The majority of samples were obtained from the same outlets and suppliers in late 1994/early 1995 (Jacobs et al., 1997; 1998) and also in late 2001/early 2002 (Jacobs et al., 2002c). The eight year time trend differences in selected organochlorine contaminant level data sets and patterns for these dietary supplements are examined using principal component analysis (PCA) (SIMCA-P v. 10.0), as the datasets are multivariate (n=22:old samples; n=21 recent samples; 13 measured PCBs and organochlorine contaminants) and there is a large variation within some of the variables. OC contaminant data for omega-3 rich dietary oils, reported in Jacobs et al. (1997, 2002), were as follows: polychlorinated biphenyls (PCBs) 52, 101, 118, 138, 149, 153, 170, 180, HCB, alpha-HCH, gamma-HCH, p,pD D E and p,pT>DT. Sample descriptions are provided in Chapter 2.4., and while the samples cover the leading brands available on the London/UK market, both studies were not comprehensive surveys of all brands available. The data were converted to ng/g where necessary. A 3 component PCA model was constructed (R^X=0.76, Q^=0.44). While the recent oil samples (S1-S21), particularly the fish oils (S8-S13), showed less contamination compared with past reports (oldl-old21), the cod liver oils (S1-S7) had the greatest levels of contaminants compared to the other supplement groups. Fish and vegetable oil mixtures had lower levels (S14-S17) than the whole fish body oils (S8-S13), while no PCBs were detected in the vegetable oils (S18-S21) as observed previously (old22), but HCHs were detected. With most PCBs detected, a clear reduction in PCB contaminants were observed, but still at similar magnitudes, except for PCB 52 (a high detection limit of 18pg/l for the old samples) and the pesticides HCHs, and DDT, which were dissimilar in pattern. A part explanation for the comparative increase in DDT in the recent samples may be the poor recoveries for DDT for the

103 old samples (Jacobs et al., 1997). The difference in the HCH patterns may be partly due to the greater proportion of vegetable oils analysed in the recent study, which were not subjected to the same treatment processes as observed in the fish oils. In terms of risk assessment, the potential contribution to the human diet of OC contaminants from omega-3 fatty acid dietary supplements still increases if sourced from fish, particularly cod liver, and further investigation of contaminants such as DDT is needed.

2.5.2. Introduction Fish oil is a by-product of the global fish meal manufacturing industry. It is rich in the long chain omega-3 fatty acids eicosapentaenoic acid (EPA) and decosahexaenoic acid (DHA). Omega-3 fatty acids are essential to the human diet, and are sold for dietary purposes due to their health benefits. The National Diet and Nutritional Surveys (UK) report that 17% of the women respondents took dietary supplements, commonly fish oil (Gregory et al., 1990), and 20% of the children and teenage respondents took dietary supplements (Gregory et al., 2000). Specific vegetable oils are good sources of the short chain form a-linolenic acid (ALNA), the precursor to long chain form utilised in human metabolism as discussed in the previous chapter (2.4).

2.5.3. Methods 2.53.1. Sample Sources Samples were purchased from retail and practitioner suppliers of the UK market in late 1994/early 1995 and late 2001/early 2002. Samples were chosen to cover the products available to the UK consumer, including neighbourhood pharmacy; national high street chain pharmacy; whole food retailers and dietary supplement companies. The samples cover the leading brands available on the London/UK market, and where not a comprehensive survey of all brands available.

2.5.S.2. Multivariate Analytical Procedures Two data sets were combined. The old samples data set consisted of 22 dietary supplements rich in omega-3 fatty acids, including nine cod liver oil supplements, ten whole body fish oil based supplements, two probable cod liver oils and one vegetable oil (Jacobs et al., 1997; 1998).

104 The recent samples data set was as for the previous chapter (2.4): 21 dietary supplements rich in omega-3 fatty acids, seven cod liver oil supplements, six whole body fish oil based supplements, four vegetable and fish oil combination supplements and four vegetable oils. See Table 2.12 for the types of oils and respective codes, and Table 2.13 for the raw data. Principal components analysis (PCA) is a projection method that helps turn data into information by giving an overview of similarity, dissimilarity and outliers, by characterising the groups graphically, displaying the multidimensional data plotted in K-dimensions, thereby reducing it to a highly visual 2D plane. The software programme used was Umetrics SIMCA (v.lO).

2.5.4. Results Generally, the recent oil samples (S1-S21), particularly the fish oils (S8-S13), showed less contamination compared with past reports (oldl-old21) See Figures 2,8, 2.9, and 2.10. The fish liver oils had the greatest levels of contaminants compared to the other supplement groups, see Figures 2.9 and 2.10, and sub-plot Figure 2.11 (cod liver oil shown). Fish and vegetable oil combinations (S14-S17) had lower contaminant levels than the whole fish body oils, see Figures 2.9, and 2.12 (recent samples S9-S13 shown), while no PCBs were detected in the vegetable oils (S18-821), as observed previously (old22), but HCHs were detected. See Figures 2.9, 2.10 and 2.13.With most PCBs detected, a clear reduction in PCB contaminants were observed, but still at similar magnitudes (Figure 2.8), except for PCB 52 (which had a high detection limit of 18mg/l for the old samples) and the pesticides HCHs, and DDT, which were dissimilar in pattern, see Figures 2.8, 2.9 and 2.10. Sample 01d9, a salmon oil and one of the most contaminated supplements, is clustered with the most contaminated fish oil supplements (top right - figure 2.10), but the present day equivalent from the same manufacturer, S8, is clustered with the least contaminated supplements (left centre of ellipse - figure 2.10), sub-plot figure 2.14. The comparative increase in DDT in the recent samples may be due to the poor recoveries for DDT for the old samples (Jacobs et al., 1997). The difference in the HCH patterns may be partly due to the greater proportion of vegetable oils analysed in the recent study, which were not subjected to the same treatment processes as

105 observed in the fish oils, indeed many of the recent fish oil supplements showed evidence of thermal processing. The primary variables (Table 2.13, Figure 2.8), loadings (Figure 2.9), scores (Figure 2.10) and score sub-plots (Figure 2.11-2.14) are shown below. Table 2.12. Abbreviations of omega-3 rich oils analysed

FO Fish oil CL Cod liver oil SO MLC Marine lipid concentrate HL [alibi, ver oi' CFO Fish oil concentrate UK Unknown fish oil RF Redfish oil (CL) SE S andeel oi LO Linseed oil FBO Fish body oil FBC Fish body concentrate EPO/TFO Evening primrose lil/ti fisii oil EPO/MFO Evening primrose oil/marine fish oil FOplus Fish oil + garlic/lecithin/vitamins/ minerals FSO CPLO Cold pressed linseed oil FSOplus Flax seed oil + sunflower seed oil/sesame seed oil/ medium chain triglycerides/rice bran & rice germ oils /oat bran &oat germ oils Table 2.13. Variable data for Principle Component Analysis of OC contaminants in omega-3 ricb dietary supplements sampled in 1994 and 2002

Primary ID Type PCB52,_._ PCBJ01 PCB.118. RCB12&. ;B138 PC# 1 4 9 . PCB 153 PCB 170 PC B :18Q HCB a-HCH.. g-HCH.....p.p’-DDE . p.p'DDT. : old1 FO 0 16.5 26.4 9.9 83.6 17.6 57.2 6.6 24.2 0 0 0 49.5 81.4 old 2 FO 0 11 15.4 6.6 53.9 0 40.7 11 13.2 0 0 0 33 15.4 old 3 CL 0 0 0 0 6.6 0 0 0 0 0 0 0 2.2 0 old 4 FO 0 0 0 0 26.4 11 26.4 0 23.1 2.2 6.6 0 68.2 36.3 old 5 FO 0 0 0 0 0 0 0 0 0 0 0 0 2.2 0 old 6 CL 0 0 5.5 0 31.9 0 35.2 3.3 7.7 19.8 102.3 0 64.9 0 old 7 CL 0 0 0 0 29.7 0 26.4 15.4 12.1 2.2 0 0 24.2 13.2 old 8 CL 0 31.9 130.9 12.1 121 19.8 77 13.2 14.3 23.1 9.9 3.3 61.6 20.9 old 9 SO 0 171.6 66 0 67.1 56.1 0 12.1 48.4 50.6 8.8 9.9 95.7 0 old 10 CL 0 0 41.8 18.7 47.3 0 62.7 11 34.1 0 0 0 30.8 7.7 old 11 MLC 0 0 9.9 0 0 0 0 0 0 0 0 0 4.4 0 old 12 HL 0 28.6 128.7 11 134.2 0 86.9 14.3 31.9 0 0 0 3.3 0 old 13 CL 0 0 0 0 13.2 0 0 0 0 11 13.2 0 66 34.1 old 14 CL 0 0 0 0 0 0 0 0 6.6 0 0 0 0 0 old 15 CL 0 24.2 127.6 8.8 66 19.8 82.5 15.4 18.7 0 0 0 38.5 0 o ld ie MLC 0 0 0 0 0 0 0 0 0 15.4 39.6 3.3 12.1 7.7 old 17 CFO 0 20.9 33 0 67.1 57.2 77 12.1 14.3 23.1 36.3 22 61.6 33 o ld ie FO 0 0 39.6 0 17.6 0 0 14.3 0 26.4 19.8 0 36.3 31.9 old 19 UK 0 56.1 83.6 0 71.5 48.4 88 23.1 15.4 8.8 15.4 12.1 15.4 0 old 20 RF 0 19.8 86.9 0 35.2 28.6 0 34.1 12.1 11 28.6 0 9.9 0 old 21 SE 0 23.1 33 0 48.4 45.1 61.6 0 13.2 0 0 14.3 0 0 old22 LO 0 0 0 0 0 0 0 0 0 0 0 13 0 0 SI CL 1.6 6.7 13.9 5.8 27.2 6.3 31.6 4.6 11 0 0 0 56.4 22.7 82 CL 0.0 7.0 11.5 6.3 31.1 7.8 35.5 5.5 11.6 0 0 0 56.7 26 S3 CL 2.8 12.8 19.6 5.3 21.8 10.6 24.6 2.1 5.2 9.1 4 5.1 44.9 12 S4 CL 2.7 22.6 39.6 12.2 56.8 16.7 62.1 8.1 17.1 2.9 0.9 0.9 106.3 28.3 S5 CL 1.9 14.0 21.3 7.1 28.7 11.2 34.9 3.9 8.9 7.3 4.2 1.2 62 7.9 S6 CL 1.0 15.2 22.9 6.5 27.2 13.0 31.1 3.9 9.0 4.9 1.9 1.3 50.5 7.4 S7 CL 2.4 9.3 19.4 8.6 39.3 9.8 45.0 7.8 18.0 0 0 0 75.3 34.2 S8 SO 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 0 1.6 1.9 S9 MLC 1.8 4.6 8.6 2.1 10.6 4.1 11.8 0.5 1.3 3.4 0.5 0 14.5 5.6 S10 FBO 0.0 6.3 11.8 2.6 7.6 2.0 8.6 0.3 0.8 0.2 0 0 10.5 4.5 S11 FBO 0.0 1.6 6.1 2.0 5.9 2.5 11.3 1.2 3.0 0 0 0 10.3 6.3 SI 2 FBC 0.0 1.9 7.0 2.9 7.6 2.6 10.9 1.8 4.5 0 0 0 6.3 7.1 SI 3 FO 3.0 7.1 12.5 3.2 10.4 5.4 13.9 1.0 2.2 0.8 0.5 0.3 20.8 8.1 SI 4 EPO/TFO 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 0.2 12.3 2.2 SI 5 EPO/MFO 2.5 2.8 3.5 2.2 7.6 3.7 9.6 0.5 3.0 4.2 1 0.7 14.4 6.2 SI 6 FOplus 0.0 4.0 1.1 0.2 7.1 1.4 6.5 0 0 0 0.4 0 48.6 77.8 SI 7 EPO/MFO 0 0 0 0 0 0 0 0 0 0 1 1.5 0.5 0 SI 8 LO 0 0 0 0 0 0 0 0 0 0 0 5.9 0.7 5.4 SI 9 FSO 0 0 0 0 0 0 0 0 0 0 0 1.3 0.5 2.3 S20 CPLO 0 0 0 0 0 0 0 0 0 0 0 0.6 0.3 0 S21 FSOplus 0 0 0 0 0 0 0 0 0 0 0 0.3 0 0 UJ6 Figure 2.8 Primary variables plot: Principal component analysis of selected OC contaminants in omega 3 rich supplements Ellipse rings indicate poorly predicted contaminants, and are discussed in section 2.5.4. H R2V)^3](cum) Q2V)^3](cum) H2=estimate ot goodness of tit 02= estimate of goodness of prediction

Var D (Primary)

Figure 2.9 Loadings scatter plot, 2 components: Principal component analysis of selected OC contaminants in omega 3 rich supplements p[Component 1 ]/p[Component 2] The clustering and positions of the specific OC contaminants in the plot space indicate the closeness of association between the contaminants and are examined in relation to the scores scatter plot ( Figure 2.10 below) to identify the dominant contaminants detected in the study samples. For greater clarity score scatter sub-plots for supplements of interest are given in Figures 2.11-2.14.

0 50

0 .40 "g-HCH

0,30

0 ,20

P[2 ] 0.10

0.00 —PUBmjTpcB^ PCB 180 -0 1 0 "P^®.g^)DT -0 ,20

-0 30

pMl

Time trend analysis: 00 contaminants in omega 3 rich supplem ents Figure 2.10 Score scatter plot: all sam ples M2(PCA-X), t[Comp, 1]/t[Comp, 2] Coloured according to Obs ID (Type)

■ o ld 16 • o l d 19 • old 6 • o l d 2 - • old 1 3 • old 21 • S 3

A S 1 3 • 8 4 Si 2 • o l d 15 Aoljglfl Ijtold 1 2 - •old 10 •S?

Obs ID (Type) Figure 2.11

Cod liver oil samples Score scatter sub plot (PC/.-)Q, # qi_ t[Comp. 1]/t[Comp. 2] # pp Coloured according to Obs ID (Type) #

4 -

• o ld 1 •old 6 •old 2 •o ld 13 •o ld •S 3 CM - 9 ^ Hold 7 • 8 4 •o ld 15 • o i • 87

-4 t

_ J _ l —liU J —L I ^ —I _1 UD O Ü Ü OGD CO3 ÜÜ Obs ID (Type)

Figure 2.12

Fish & vegetable oil combinations & whole fish body oils, * EPO/MFO (recent samples) Score scatter sub plot. A EPO/TFO M2 (PCA-X), t[Comp. 1]/t[Comp. 2] ■ FBC Coloured according to Obs ID (Type) • FBO FO ■ FO plus SO

2 -

A S 1 4 ■ S 16 A S 13

-3 -

2 p

Obs ID (Type)

108 Figure 2.13

Omega 3 rich vegetable oil supplements $ CPLO Score scatter sub plot. M2 (PCA-)^, FSO t[Comp. 1]/t[Comp. 2] ^ FSOplus Coloured according to Obs ID (Type) LO

4 - 2

- 2 - -4L

Cbs ID (Type)

Figure 2.14

Salmon oil sample Score scatter sub plot.M2 (PCA-X), t[Comp. 1]/t[Comp. 2] Coloured according to Obs ID (Type)

O O CO

Obs ID (T yp e)

2.5.5. Discussion. The results in good agreement with the levels of PCBs recently reported for fish oil supplements (Fernandes et ah, 2002; Fromberg, 2001; Food Safety Authority of Ireland, 2002), and PCB data are highly comparable on a brand sample specific basis, and are in the same order of magnitude (same brand, different lot no.s) (Fernandes et al., 2002).

109 e.g. SI =92.1 ng/g, 81.7 ng/g DEFRA2002 84 = 201.9 ng/g, 196.4 ng/g DEFRA2002 (ICES 7 Marker PCB’s: 28, 52, 101,118,138,153, 180) National extrapolations for levels and time trends of PCB contaminants in fish oils available in the UK market could be suggested on the basis of these studies in addition to other recent reports (e.g. Fernandes et al., 2002; Fromberg, 2001; Food Safety Authority of Ireland, 2002) but not for the other target contaminants, due to the small sample size and lack of available comparative data. On the whole, the recent fish oil samples showed a trend for slightly lower OC POPs contamination compared with the old samples and past reports, but this may just be due to dilution. Both this study and the UK FSA (Fernandes, et al., 2002) found that the fish oils available to the consumer were often diluted (usually with vegetable oils) compared with the pre-2000 fish oils analyses in the public domain. While new contaminants, such as polybrominated diphenyl ethers, PBDEs, were detected, improved sourcing, processing (e.g. deodorisation) and nutritional profile development of ‘optimum’ essential fatty acid balances (S. 8, 14-17, 21) could be improving supplement quality for the consumer.

2.5.6. Risk assessment Fish and fish oil consumption are associated with numerous health benefits due to the long chain omega-3 essential fatty acids, but the potential contribution to the human diet of POPs from omega-3 fatty acids food sources increases if sourced from fish, particularly cod liver. The possible contribution to dietary intakes of POPs from cod liver oils currently available in UK and international markets could clearly still be significant for high consumers. The vegetable oils studied here presented little or no POP contamination, whilst still being good sources of omega 3 fatty acids, but according to current nutritional knowledge, do not provide an effective metabolic source of DHA for the average consumer, as discussed in section 2.4. Although preliminary, this analysis strongly supports the need for ongoing monitoring of POPs and further investigation of DDT isomer levels and ratios, in fish oils designated both as nutritional supplements for consumers and as animal feed ingredients. This analysis also illustrates the value of multivariate analytical techniques for assessing contamination patterns and trends.

110 Further, in the public interest, newer POPs such as PBDEs need to be included in national and international POPs contaminant monitoring programmes. An adequate risk benefit analysis of fish and especially cod liver oil supplements would need to consider all nutritional sources and currently available supplement formulations of short and long chain omega 3 fatty acids, together with reliable information on human variability in synthesis, metabolism and utilisation of the long chain forms, at different times of life, and in relation to other dietary fat intake as well as knowledge of critical windows of developmental exposure to POPs, and critical moments of nutritional requirement of the foetus of DHA particularly for brain and retinal development.

Ill C hapter 3.

Homology Modelling of the Nuclear Receptors

112 Chapter 3. Homology Modelling of the Nuclear Receptors

3.1. Abstract

This chapter describes how the three-dimensional structures of the ligand-binding domain (LED) of several interrelated human nuclear (NRs) and steroid hormone receptors (SHRs) were generated by homology modelling. The receptors modelled are the oestrogen receptor p (hERP), the pregnane-X- receptor (PXR), the Ah receptor (AhR) and the constitutive androstane receptor (CAR). They were produced by homology modelling from the human oestrogen receptor a

(hERa) crystallographic coordinates (Brzozowski et al., 1997) as a template together with the amino acid sequences for hERp (Mosselman et al., 1996), PXR (Lehmann et al., 1998), AhR (Burbach et al., 1992) and CAR (Forman et al., 1998; Baes et al., 1994) respectively. The selective endogenous ligand, in each case, was docked interactively within the putative ligand binding site using the position of oestradiol in hERa as a guide, and the total energy was calculated. In each receptor model a number of different ligands known to fit closely within the ligand binding site were interactively docked and binding interactions noted. Specific binding interactions included combinations of hydrogen bonding and hydrophobic contacts with key amino acid sidechains, which varied depending on the nature of the ligand and receptor concerned. A homology model of the human peroxisome proliferator activated receptor a (PPARa) was also produced using the human PPARy (hPPARy) LED crystallographic coordinates summarised in

Lewis and Lake, 1998, as a template, together with the amino acid sequence for hPPARa (Lewis and Lake, 1998; Nolte et al., 1998). It is intended that the models will provide a useful tool in unravelling the complexity in the physiologic response to xenobiotics by examining the ligand binding interactions and differences between the nuclear and steroid hormone receptors activation or inactivation by their ligands. This work supports the QSARs presented in chapter 5.

3.2. Introduction As reviewed in chapter 1, nuclear receptors are a large protein superfamily that is involved in a wide range of physiological functions including development, reproduction, differentiation and homeostasis. They are important drug targets for intervention in

113 disease processes. Exogenous compounds that target these receptors can therefore disrupt both normal and abnormal functioning of these key metabolic pathways. While environmental hormone mimics contribute to detrimental health effects by activating certain receptors and disturbing normal function, there are therapeutic uses from both dietary and pharmacological treatment for abnormal functioning of the hormone pathways and hormone dependent diseases. NRs and SHRs are regulated by hormones and chemicals that can mimic hormones (Honkakoshi and Negishi, 2000). After a hormone is produced, circulating and intracellular binding proteins regulate the hormone’s bioavailability. Then the hormone triggers action by binding to a specific cellular receptor by docking into a ligand binding domain (LED) which is a hydrophobic pocket, and binding with specific amino acids (Weigel, 1996). An important requirement for homeostasis is the detoxicifation and removal of endogenous hormones and xenobiotic compounds with biological activity. This crucial metabolic role is conducted by the cytochrome P450 (CYP) enzyme superfamily. The induction of specific CYPs via the adaptive increase of CYP gene expression commonly utilises the nuclear receptor pathway, where exposure to xenobiotics and drugs activates specific members of the nuclear receptor superfamily which in turn bind to their cognate DNA elements and stimulate the CYP target gene transcription (Denison and Whitlock, 1995; Kliewer et al., 1999; Negishi and Honakoski, 2000; Xie and Evans, 2001). Knowledge of nuclear hormone-receptor activation and action upon the regulation of gene expression can aid the understanding of the progression of certain diseases, and facilitate the design of drugs with improved efficacy and fewer side effects. The super family is broadly divisible into three subclasses: the type I receptors for steroid hormones, including progestins (PR), estrogens (ER), androgens (AR), glucocrticoids (GR) and mineralocorticoids (MR); the type II receptors for thyroid hormone (TR), vitamin D (VDR), 9-cis retinoic acid (RXRs), dX\-trans retinoic acid (RARs), PPAR and finally the orphan class, for which cognate ligands have not yet been characterised, such as the PXR, CAR, COUP-TFs, HNF4, Rev Erb (McKenna et al., 1999; Nuclear Receptors Nomenclature Committee, 1999; Neve and Ingelman-Sundberg, 2000; Cottone et al., 2001; Willson et al., 2003). As described in Chapter 1, the currently accepted theory of steroid hormone binding suggests that in the absence of the hormone, each receptor is associated with certain ‘chaperone’ proteins (Weigel, 1996). Binding of the steroid hormone with the

114 receptor protein causes a conformational change. This molecular switch results in the removal of the heat shock complex and allows the receptors to dimerise. Then binding to a hormone response element (HRE) on DNA occurs, to produce a complex that can trigger or suppress the transcription of a selected set of genes (Weigel, 1996; Alberts et al., 1997) See Figure 1.1. So far, 48 nuclear receptors have been identified in the human genome (Willson and Kliewer, 2000), but most of these are ‘orphan receptors’, in that they are awaiting the recognition of specific ligands and functions, and it is likely that more receptors will be discovered in the future. Each type of receptor has the potential to regulate a distinct endocrine signalling pathway, of which we only have a rudimentary knowledge. Members of this receptor family are related to each other in terms of their amino acid sequence and their function within cells. They therefore have structural features in common. These include a central highly conserved DNA binding domain (DBD) that targets the receptor to specific DNA sequences, termed hormone response elements (HREs). This domain contains 8 cysteines, which form a pair of tetra coordinate binding sites for zinc atoms. When the zinc atoms allow folding of the protein, an a-helix is placed into the major groove of the DNA double helix. The amino acids on this a-helix enable the receptor to recognise the DNA in a sequence specific fashion. A terminal portion of this receptor (COOH) includes the ligand binding domain (LBD), which interacts directly with the hormone. This part of the receptor is larger and more complex than the DNA-binding domain. It is composed of three layers of a-helices forming a pocket. Embedded within this pocket is a hormone dependent transcriptional activation domain, and this is where ligands are transported prior to binding (Weigel, 1996). In essence the LBD acts as a molecular switch that recruits co-activator proteins and activates the transcription of target genes when flipped into the active conformation by hormone binding. From a refinement of the characterisation of the role of the

3’untranslated region (3’UTR) of hERa mRNA, the existence of another level in the control of the expression of the ligand-activated transcription factor hER in addition to transcriptional regulation has been determined (Kenealy et al., 2000). Due to the general similarities in the structure and function of members of the steroid hormone family, it is likely that elements that influence the stability of mRNAs of other steroid hormones receptors should also be found in their 3’UTR (Kenealy et al., 2000). The members of this family also have dimérisation receptor partners in common, particularly with the ubiquitous 9-cis-TQtmoic acid receptor (RXR) (Kliewer et al., 1999).

115 The ligand activated nuclear receptors CAR, PXR, and PPAR bind to their cognate DNA elements as heterodimers with RXR and thus activate the transcription of their CYP2B, CYP3A, and CYP4A genes (Figure 3.1). This process of gene activation then leads to enhanced metabolism of the compound of exposure (Kliewer et al., 1999; Willson and Kliewer, 2000; Xie and Evans, 2001).

Figure 3.1. P450 Transcription factors and examples of classes of known activation compounds

P450 Transcription Factors

CYP1A f Steroids, Xenobiotics Xenobiotic metabolism

CYP2B Steroids, Xenobiotics Xenobiotic, Steroid metabolism

Steroids, Xenobiotics CYP3A Xenobiotic, Steroid metabolism

Fatty acids, FIbrates CYP4A Fatty acid metabolism

However, there are further complications as there may be competition between the receptors for RXR, as well as reduced RXR availability and activation in response to stress-signaling, triggered from a variety of environmental stimuli (Lee et al., 2000a). As RXR is a cofactor that is essential to transcriptional activity, this suggests that stress- signaling will also indirectly affect all the receptors that dimerise with RXR, disabling gene activation even though ligand binding has occurred. Several receptors may be affected by the lack of availability of the dimérisation partner, including the ERs, PXR, CAR, the PPARs and other receptors. This will affect their ability to trigger or suppress gene transcription. NRs are subject to cross talk interactions with other nuclear receptors, nuclear proteins, drug metabolizing enzymes, such as UGTs (Sugatani et al., 2001), the transporter P-glycoprotein, P-gp (Synold et al., 2001) and with a broad range of other

16 intracellular signaling pathways (Kliewer et al, 1999). There may even be a cascade effect, where metabolites produced through the activities of one receptor are specific signaling molecules (and ligands) to modulate the next receptor, along the chain of a nuclear receptor intercommunication web. Dependency upon interactions with other nuclear proteins or cofactors that are important tissue/cell specific mediators of nuclear receptor function introduces further regulation of the members of the NR family. These may differ between receptors, or receptors may hold certain proteins or receptors in common with each other, such that part of the mechanism of action may be ascribable to competition between the receptor signalling pathways from the co activators or co repressors. This has been reported for the ER and AhR, for example (Nguyen et al., 1999). Table 3.1 presents the activation compounds and ligands for the selected receptors under study, further ligands for PXR are described in Chapter 4.

3.2.1. The human oestrogen receptor a (hERa) and oestrogen receptor^ (hER/3) The oestrogen receptors are known to exist as two subtypes, each one encoded by a separate gene. These are ERa (Brzozowski et al., 1997), and the recently discovered

ER(3 (Mosselman et al., 1996) and its isoforms, of which a spliced isoform, ER|3/2 appears to be equally expressed in animal model tissue density studies (Petersen et al.,

1998). The classical ERa subtype and ERp receptors and isoforms apparently evolutionarily diverged over 450 million years ago, suggesting that although they have evolved in parallel, this ancient duplication was to facilitate unique roles in vertebrate physiology and reproduction (Kelley and Thackray, 1999). The ERs differ in tissue distribution and relative ligand binding affinities for both endogenous and exogenous ligands (Table 1.1, figure 1.4) (Kuiper et al., 1997; 1998), which may help explain the selective action of oestrogens and androgens in different tissues (Figure 1.7) (Hall and McDonnell, 1999; Fuqua et al., 2000).

117 Table 3.1. Selected receptors and their ligands, as reported in the literature

Receptor Activation compounds References ERp 5a androstane did and 3P androstane diol, Kuiper et al., 1997; 1998; 17 P oestradiol, coumestrol, genistein, daidzein, Gaido, 1999; Gustafsson, trans-nonachlor, endosulphan, o,p’ DDT 2000

PXR Rifampicin, RU 486, SR 12813, androstanol, Lehmann et al., 1998; Moore coumestrol, PB, TCPOBOP, pregnenolone 16a- et al., 2000a; 2000b; carbonitrile (PCN), hyperforin (active constituent Takeshita et al., 2001; Landes of St Johns Wort), 17 P oestradiol, P pregnane- et al., 2003; Nallani et al., 3,20,dione, vitamin E, kava kava, bisphenol A, 2003; Owsley and Chiang, paclitaxel, guggulsterone, trans nonachlor, 2003; Raucy, 2003;Wyde et triclosan, DDE (in rPXR), enterolactone al., 2003, and Chapter 4.

CAR Androstenol, androstanol, clotrimazole, Moore et al., 2000b; Maglich TCPOBOP (in mCAR), DDE (in rCAR) PB, 5p et al., 2002; Honkakoshi et pregnane-3,20,dione, PCN al., 2003; Wyde et al., 2003

AhR Halogenated aromatic hydrocarbons (HAHs), e.g. Gillner et al.,, 1993; Denison polychlorinated dibenzo-p-dioxins (dioxins), and Whitlock, 1995; Sinai dibenzofurans, and biphenyls (PCBs), polycyclic and Bend, 1997; Sinai and aromatic hydrocarbons (PAHs) e.g. Bend, 1997; Denison et al., benzo(a)pyrene, 3-methylcholanthrene, P- 1998; Heath-Pagliuso et al., naphthaflavone, benzoflavones, carbaryl 1998; Phelan et al., 1998; diaminotoluene, omeprazole, brevetoxin, indole Schaldach et al., 1999; carbinols (found in cruciferous vegetables) Adachi et al., 2001; Denison endogenous ligands e.g. bilirubin, biliverdin, et al., 2002; Gambone et al., water soluble metabolites of tryptophan, 2002; Song et al., 2002 tryptamine, indole acetic acid, 2-(l’H-indole- 3’carbonyl)-thiazole-4-carboxylic acid methyl ester, retinoids, antioxidants, Lipoxin A4, arachidonic acid, prostaglandins, indirubin, indigo

PPARy Fatty acids, eicosanoids, prostaglandins, Johnson et al., 1997; Memon metabolites of arachidonic acid and linoleic acid, et al., 2000; Bar-Tana, 2001; antidiabetic thiazolidinedione (TZD) drugs Chawla et al., 2001; Nixon et including NSAIDs al., 2003; Willson et al., 2003

PPARa Long chain polyunsaturated fatty acids Memon et al., 2000; Zomer et (LCPUFA), eicosanoids, fibrates, phthalate ester al., 2000; Bar-Tana, 2001; plasticisers, pristinic acid, phytanic acid Cronet et al., 2003; Willson et al., 2003; Xu et al., 2003

3.2.2. The Pregnane-X-receptor (PXR) The Pregnane X Receptor (PXR), recently isolated and it’s sequence published by Kliewer and co-workers (Lehmann et al., 1998; Kliewer et al., 1998), and later Blumberg and co-workers (Blumberg et al., 1998) is involved in activating the expression of several P450 detoxifying enzymes, including CYP3A4 in the adult and CYP3A7 in the foetus, in response to xenobiotics and steroids (Pascussi et al., 1999). CYP3A4 is the

118 major human hepatic P450, and is involved in the metabolism of over 60% of drugs in clinical use (Maurel, 1996). PXR is highly divergent between species, with great differences in PXR activation profiles due to differences in the LBD (Jones et al., 2000b). Variation in ligand binding domain is consistent with in vivo species differences in response to inducers, see figure 3.2. Figure 3.2. PXR class similarities with different species and human CAR (after Moore et al., 2000b)

NRl/ Subfamily

human PXR

rhesus PXR pig PXR dog PXR rabbit PXR mouse PXR rat PXR frog BXRa frog BXRP fish PXR chicken CXR human CAR

The major site of PXR expression is in the liver hepatocytes and the gastrointestinal tissues, but it is also expressed in both normal and neoplastic breast tissue. Indeed a statistically inverse relationship between the level of PXR mRNA expression and ER status has been observed by ligand binding analysis (Dotzlaw et al., 1999). PXR can be activated by a variety of chemically distinct ligands (Table 3.1), in a species dependent manner (Jones et al., 2000b; Moore et al., 2000b). Ligands include endogenous hormones such as pregnenolone, and progesterone and their synthetic derivatives such as pregnenolone 16a-carbonitrile (PCN), trans-nonachlor, rifampicin, dexamethasone, corticosterone, spironolactone, phénobarbital, and hyperforin (the active constituent of St John’s wort) (Moore et al., 2000a). The mechanism of action is described further in Chapter 4.

119 3.2.3. The Constitutive Androstane Receptor {CAR) The Constitutive Androstane Receptor (CAR) is a member of the same nuclear receptor subfamily as PXR, sharing around 40% amino acid identity in their LBDs, as shown in figure 3.2, with 70% similarity between hCAR and rodent CAR LBD regions, although marked differences in ligand binding, (based upon in vitro radioligand binding studies) between PXR and CAR have been reported (Moore et al., 2000b), as shown in Figure 3.3. In a pattern similar to that of the ERs, based upon phylogenetic analyses, it has been suggested that PXR and CAR are closely related to each other (Moore et al., 2000b). CAR is also present largely in the liver (it has also been detected by one group in the intestine, kidneys, lungs, heart, and muscle) (Baes et al., 1994), but not by Negishi’s group (Negishi et al., 2003) (Figure 1.7). It interacts with and is inhibited by two endogenous testosterone metabolites, androstanol and androstenol, via a mechanism that involves a widely expressed nuclear receptor coactivator, SRC-1 (Forman et al., 1998). The hierarchy of ligand activation differs between the receptors as well as for receptors isolated from different species, and in many instances, as identified in CAR, molecules that were previously regarded as metabolic intermediates are in fact “intracrine” signalling molecules within tightly coupled metabolic pathways for altering gene expression. Unlike most nuclear receptors, including PXR and ER, the steroidal ligand for CAR inhibits receptor-dependent gene transcription by way of a ligand-independent recruitment of transcriptional co-activators (Forman et al., 1998). CAR functions in a manner opposite to that of the conventional nuclear receptor pathways and can be considered a 'repressed ' nuclear receptor in the presence of androstane metabolites (Tzameli and Moore, 2001). In cell based reporter gene assays, exogenously expressed CAR enters the nucleus and regulates the expression of target genes (Baes et al., 1994; Forman et al., 1998), it is not present in the nucleus but is sequestered in the cytoplasm, unlike the other receptors modelled and discussed here. There are significant sex differences in plasma androstane levels and it has been recently implicated as a transcriptional regulator of the gene governing the steroid hydroxylase CYP2B after binding with its cognate DNA response elements as a heterodimer with RXR (Forman et al., 1998). There appear to be additional mechanisms for the regulation of CAR activity, including phosphorylation by phénobarbital (PB). The effects of PB on CYP2B expression are blocked by the phosphatase inhibitor okadaic acid (Kawamoto et al., 1999)

120 suggesting that déphosphorylation of CAR, rather than direct ligand binding, is involved in its translocation into the nucleus. Bilirubin also activates CAR, but does via binding in the CAR LBD (Honkakoshi et al., 2003). This receptor is a coordinate regulator of hepatic gene expression in defense against chemical toxicity, and it also suggests a new area of androgen physiology whose significance is unknown as yet (Blumberg and Evans, 1998). Negishi has presented a model of CAR where a cytoplasmic CAR retention protein (CCRP) bound with hsp90 (a cofactor held in common with the ERs, chapter 1, and AhR, figure 3.3 and chapter 6), folds over to enclose the CAR. It appears that potent drugs can trigger translocation of CAR from this protein bundle, but they are not necessarily direct activators (Negishi et al., 2003). A model of CAR can aid in the design of synthetic ligands that can help investigate the relevance of CAR to human metabolism and health.

Figure 3.3. Comparison of selected radioligand binding in CAR and PXR (after Moore et al. 2000b)

Oh PXR □ hCAR □

3.2.4. The Aryl hydrocarbon Receptor (AhR) The Ah receptor is a member of the Per-Arnt-Sim family of nuclear regulatory basic helix-loop-helix proteins (Swanson and Bradfield, 1993), and has been detected in

121 nearly all vertebrate groups examined (Hahn, 1998). The Ah receptor binds to 2,3,7,8 TCDD and other structurally similar PAHs to activate the cognate xenobiotic response element of the CYP lA and IBl genes (Swanson and Bradfield, 1993; Denison and Whitlock, 1995; Hahn, 1998). However, the AhR is the only member of that regulatory family known to bind to a ligand prior to heterodimerisation and bind to DNA in upstream regulatory regions of target genes. Predominantly found in hepatocytes, but also in breast cancer cells (Nguyen et al., 1999), the AhR regulates the expression of a number of genes, including cytochrome P450 lA l, 1A2, IBl, glutathione S transferase M (GSTM), DT-diaphorase, UGT and aldehyde dehydrogenase in a ligand-dependent manner (Nebert et al., 1981). AhR is also up regulated during cell division and is expressed in a specific spatial and temporal pattern in the developing foetus in vivo (Abbott et al., 1999). The best- characterised high affinity AhR ligands include a variety of ubiquitous lipophilic environmental contaminants in the polyhalogenated aromatic hydrocarbon family including dioxins, furans, coplanar biphenyls and polycyclic aromatic hydrocarbons (Denison and Whitlock, 1995). Other lower affinity ligands can be found endogenously, e.g. biliverdin (Sinai and Bend, 1997; Phelan et al., 1998), and in the diet (Gillner et al., 1993; Heath-Pagliuso et al., 1998). Synthetic retinoids and pesticides have also been reported to activate the AhR pathway (Denison et al., 1998; Gambone et al., 2002). Exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), the most potent AhR ligand known, results in a wide variety of species- and tissue-specific toxic and biological responses (van den Berg et al., 1998). They are associated with the disruption of almost every hormone system that has been examined and responses to activation include developmental and reproductive toxicity. Animals treated with 2,3,7,8 TCDD have developed abnormalities in several organs, including the thyroid, thymus, lung and liver, immune and endocrine function. Wasting, lethality and induction of gene expression have also been shown to be AhR dependent (Femandez-Salugero et al., 1995; Schmidt et al., 1996; Femadez-Salugero et al., 1997; Abbott et al., 1999; Diliberto et al., 2003). Figure 3.3 illustrates the cellular events of AhR activation. Within the cytosol of the cell, the AhR is associated with a heterodimeric transporter protein partner, termed the Aryl Hydrocarbon Receptor Nuclear Transporter protein (ARNT). Recently Yao and Bradfield (2003) have proposed an expanded model of AhR signalling, suggesting that receptor folding involves hsp90 co-chaperones Stilp, Cpr7p, and that trans-activation requires several complex subunits. From their

122 observations, Bradfield’s group also report that the toxicity of dioxin must be due to a battery of gene changes, not just those for CYP lAl and 1A2, and that most novel dioxin responsive genes are down regulated in AhR null mice. The unliganded AhR may also act through other mechanisms by being phosphorylated to key regulatory proteins such as HSP90, p37, AIP, XAP2, src, rel, and Rb (Bimbaum, 2000). AhR levels have been reported to rise with increased oxidative stress activity. Many antioxidants are weak AhR ligands (Table 3.1.), which via the AhR down regulate oxidation by specific oxidative signalling (Yao and Bradfield, 2003).

Figure 3.3. TCDD (dioxin) molecules and Ah receptors: cellular events (After Wæm, 1995) 1 . A emwAA nnd pemalnoas 2. IrbSM Ihû in fn© iha ceil momamno. cyWpJftyrv, thara nt»^ mcaplora. MomuOy Ihey sea ünftjHl la two nioiatda# a. Tho dc4cn mO&Kul# dock»' pftti am Ah -mWek M e s h a p e in such a way lhal lha hspdO moiecuiea are mlweand, cflC mam'brano

CYTOPLASM 4. The Ah recepier now o iw s iha npclm » pf tho ceiL I Irking up on taule with a pfooein denoted Ami. Thi» com- binadon cl mo4ecul©s. %» epecTk; ponky* ol iho OMAchiainsu Ono such pohkyi flomns part al a scquomcB repu- lasng tho aclrKTty d I he CVP 1A1, 6L T h o A h r a o c p te r c a n « ü e S m d 10 DMA M puencee feguelkig the ncdv.ty ol other b can ihus bYluancc the sfynlhests oS som a Avanty lypos protein m tfeoifee af: least- apart from oyiechrome P450 1A1. A s >ief. KV» tfOhW knanfm tiich ol-Siesa prnlW ig 2; Of am msponsfcfa lor lha tank: o f e d s o f dbouriB-. H 5. The gye oopios of tisall RcpukÉtonr k tha loan ol Yrgssongor AffA'. sequence tef rroBsonnor ANA Outside lha nucleus. Ihese OOpiO* C Y P l A l portefit Hie «.yfrthests of en « n z y m i çdtod cycochnxno P450 1A1- A dtoHln Intato Ihus rosuhs in Npliei levds of ihi» enzyme in ihe ceA NUCLEUS

C Y F 1 A 1 #

protein rrioieeii ee _—, # 0

cyxKhicmo PASO lAI rreSàenfler ANA

3.2.5. The human peroxisome proliferator activated receptors (PPAR) The Peroxisome Proliferator-Activated Receptors are a family of orphan receptors with fundamental roles in regulating energy balance (Johnson et al., 1997; Memon et al., 2000; Uauy et al., 2000; Xie and Evans, 2001). A number of prevalent

123 metabolic disorders such as obesity, atherosclerosis and type 2 diabetes are associated with a shift in this balance. The Peroxisome Proliferator Activated Receptors are activated by xenobiotics, which elicit increases in the number and size of peroxisomes when administered to rodents (Kliewer et ah, 1999), and also induce hepatocellular carcinoma development by a non-genotoxic mechanism (Waxman, 1999). There are three known closely related receptors: PPARa (3/6 and y, found in the liver, kidney, heart, haematopoietic and adipose tissue, but having different expression patterns. PPARa is found in liver, kidney, heart and muscle, PPAR6 is expressed in nearly all tissues and PPARy is expressed in fat cells, the large intestine, and monocyte lineage cells (Johnson et al., 1997). They each play key roles in lipid metabolism and homeostasis; PPARa is responsible for CYP4A induction; peroxisomal enzyme induction and hepatic peroxisome proliferation. PPARa has a central role in hepatogenesis, and

PPARy, a central regulatory role in adipogenesis (Waxman, 1999; Memon et al., 2000; Chawla et al., 2001; Willson et al., 2003).

Figure 3.4. The mechanism for transcriptional regulation of the PPAR system (adapted from Uauy et al., 2000)

Tlvia^olidioe^lon^s

Essential faav acid» Saturated fats ànd m«teboliic«

i' Co-activ.ttçra

PPAR J

PPRE

Key: PPAR is a nuclear protein, PPRE, peroxisomal proliferator response element in the DNA; RXR, retinoic acid receptor; ?, unknown potential activators or inhibitors

124 PPAR function may be affected by the nuclear availability of RXR, as it is involved in heterodimers with other receptors, notably the oestrogen receptors and thyroid receptor.

PPARa regulates key steps in lipid and fibrate metabolism. It is the molecular target for naturally occurring plant fatty acids (pristinic acid and phytanic acid) present at physiological concentrations (Zomer et al., 2000), long chain polyunsaturated fatty acids (LCPUFA), eicosanoids (Uauy et al., 2000; Chawla et al., 2001), and peroxisome proliferators, which include drugs such as the fibrates, (used widely to lower high triglyceride levels, a risk factor in coronary heart disease), and synthetic chemicals such as the phthalate ester plasticisers, and pesticides (Waxman, 1999). PPARy ligands include fatty acids, prostaglandins and the antidiabetic thiazolidinedione (TZD) drugs (Willson et al., 2003; Memon et al., 2000). Pristinic acid and phytanic acid are branch chained fatty acids obtained through the diet from the chlorophyll in plants. Present at micromolar concentrations in healthy individuals, they can accumulate in a variety of inherited disorders. Potent binding of pristinic acid and phytanic acid in the LED of PPARa (Zomer et al., 2000) indicates a primary mechanism for metabolising these dietary fatty acids. The LED consists of 13 a-helices and a small 4-stranded p sheet forming a hydrophobic ligand binding pocket with a volume at least twice that of other receptors (Cronet et al., 2003). The PPARs have a far larger ligand-binding pocket than the receptors so far discussed, as indicated in Table 3.2, (Watkins et al., 2001), and there are differences in the shape of each PPAR ligand-binding pocket (Cronet et al., 2003; Xu et al., 2003), giving broad ligand specificity on a structural basis. Rosiglitazone occupied a fraction of the available LED space in PPARa, and less than that in PPARy, particularly the rosiglitazone TZD head group, and thus comparatively reduced selectivity was observed. This has been observed for different ligands in the PPAR family, and is a clear descriptor for PPAR selectivity. There may be expression of a dominant-negative inhibitory human PPARa variant found in some individuals (Roberts et al., 1998) and human polymorphisms. It is possible that PPARy and PPARô, which is highly expressed in multiple human tissues, may be transactivated and consequently perturbed by a subset of peroxisome proliferating compounds, affecting the PPAR metabolic pathways, to elicit a pathophysiological response.

125 Another factor to be considered is modulation through cross talk between PPAR and other nuclear receptors/signalling molecules. For example, thyroid hormone suppresses hepatic peroxisome proliferation responses and exhibits inhibitory cross-talk with PPARa, due in part to competition between the thyroid receptor and PPAR for their common heterodimerization partner RXR (Miyamoto et al., 2003). Indeed, all the receptors discussed here are interlinked, not least by their requirement for RXR as the heterodimerisation partner (See figure 3.1), except AhR (see figure 3.2).

3.3. Molecular modelling methods 3.3.1. Introduction The steroid hormone receptor super family shares a conserved primary sequence and it is likely that their three-dimensional (3-D) stmctures are similar, so it is possible to model one member of the family from another. In order to understand the role of ligand in receptor activation, homology molecular models of the ERp, hPXR, AhR and hCAR LBDs have been generated from the determinants of a ligand-bound hERa 3-D X-ray crystallographic structure (Brzozowski et al., 1997). Similarly, for the investigation of the role of ligand in hPPARa activation, a molecular model of the hPPARa ligand-bound hormone-binding domain (HBD) has been generated from a ligand bound hPPARy crystal structure (Nolte et al., 1998; Lewis and Lake, 1998b). Knowledge of molecular mechanisms action for key proteins is built upon scientific models, and X-ray crystal structures are considered to represent the most reliable model for describing the 3-D conformation of a protein in a solid state (Lewis, 1996). While crystal structures are reliable starting points for investigations of the molecular mechanisms of receptors, as with other proteins, they are not necessarily the physiologically relevant shape that occurs in reality. The crystal structure reveals a composite X ray image of the receptor, in the solid state, usually containing water molecules, at a single moment in time, not in a dynamic situation. However, when examined in conjunction with experimentally derived data which is dynamic (in vitro and in vivo biological data) mechanistic models of activation can be refined and made more relevant to the human situation. Homology modelling can facilitate deeper insights into substrate selectivity and improve our understanding of receptor mediated metabolism.

126 Using natural and synthetic ligands as chemical tools, the nature of receptor activation can be examined by assessing the structural mechanisms computationally, indicating highly probable modes and mechanisms of binding, together with the key amino acids involved. This can aid in the discovery of new hormone signalling pathways and cross-talk, and provide receptor specific insight into various disease scenarios, which in the case of hPPARa for example, would include the regulation and perturbation of lipid and TZD drug metabolism. It can also improve the prediction of the metabolic fate of new chemicals.

3.3.2 Materials and methods 3.3.2.1.Receptor alignments The alignment was determined based upon previous alignments in the literature (Brzozowski et al., 1997). Domains of human ERa, retinoic acid receptor (RARa) and retinoid X receptor (RXRa), show conserved residues, these were used as the basis for modelling the HBD of hERp, hPXR, AhR and CAR. The respective sequences were adjusted in relationship to the other proteins in order to match theoretical helices of the receptors to known helices of hERa (figure 3.5). Similarly, domains of human PPAR and retinoid X receptor (RXRa) show conserved residues, and these were used as the basis for modelling the hormone-binding domain (HBD) of the hPPARa. The sequence was adjusted in relationship to the other proteins in order to match theoretical helices of hPPARa to known helices of hPPARy. The ligand binding domain sequences are shown in figure 3.6.

3.3.2.2. Building in silico models The homology models of hERp (figure 3.7), hPXR (figure 3.8), and the hPXR crystal structure (figure 3.9) for comparison, CAR (figure 3.10), and AhR (figures 3.11 and 3.12) were generated from the existing hERa structure (figure 3.2) using SYBYL biopolymer software (Tripos associates, St Louis, Missouri) implemented on a Silicon Graphics Indigo^ IMPACT 10000 Unix workstation. Loops were chosen from the Brookhaven Protein Databank to incur minimum interference with the core model. After energy minimisation with the Tripos force field, the endogenous moiety was docked into the structure. Some adjustment was needed in order for hydrogen bonding to occur between ligand and receptor. Finally, the structures were energy minimised again to

127 achieve self-consistency. Similarly, the alignment and helices shown in figure 3.6 were used to generate the homology model of hPPARa from the existing hPPARy structure, shown in figure 3.13.

3.3.2.3. Range of ligand probes

A number of compounds, with known receptor specific high, medium and low activation were individually docked in each receptor. For ERp these included the hormones 17 P oestradiol, androstanes, phytoestrogens and synthetic chemicals such as triclosan. For hPXR these included pregnenolone, coumestrol, PCN (a low affinity hPXR ligand, but high affinity rPXR ligand), progesterone, corticosterone, rifampicin, enterolactone, PCB118 , PCB 153, and triclosan. For the AhR these included 2,3,7,8 TCDD, and PCB 126. With the AhR, little adjustment was needed in order for hydrogen bonding to occur. For the hCAR these included the testosterone metabolites androstenol and androstanol, and clotrimazole and TCPOBOP (both high affinity mouse CAR ligands).For both PPARa and PPARy these included fatty acids, thiazolidinediones (TZDs) and other drugs, and lipophilic chemicals.

128 Figure 3.5. Alignment Between Steroid Hormone and Retinoic Acid Receptors LBDs 298 345 h E R a SKKNSLALSLTADQMVSAL.LD.AE•PPILYSEYDPTR.PFSEASMMGL AhR draefqrqlhwalnpdsaqgvdeahgppqaavyytpdqlppenasfmer MER TKKNSPALSLTADQMVSALLDAEPPM lYSEYDPSRP...... FSEASMMGL LIGAND h E R b Ri/RELLLDALSPEQLVLTLLEAEPPH _____ VLISRPSAP ...... FTEASMMMS hPX R EDPATWSQIMKDRVPMKISLQLRGEDG..SIWNYQPPSK. SDGKEIIPLLPH CAR IRTLLGAHTRHMGTMFEQFVQFRPPAH..LFIHHQPLPT.. . . LAPVLP. . LVTH R xR NEVESTS.SANEDMPVERILEAELAVEPKTETYVEANMG. . . . LNPSSPNDPVTN RAR GSPDSYELSPQLEELITKVSKAHQETFPSLCQLGKYTTNSSADHRVQLDLGIiWDK BINDING 3 4 6 3 9 5 h E R a LTNLADRELV HMINVTAKRVP GFVDLTLHDQVH. LLECAWLE ILMIGLVWRS AhR CFRCRLRCLL DNSS...... GF LAMNFQGRLKYLHG...... QNKK MER LTNLADRELV HMINWAKRVP GFGDLNLHDQ VHLLECAWLE ILMIGLVWRS DOMAIN h E R b LTKLADKELV HMISWAKKIP GFVELSLFDQ VRLLESCWME VLMMGLMWRS hPX R LADVSTYMFK GVINFAKVIS YFRDLPIEDQ ISLLKGATFE MCILRFNT.. CAR FADINTFMVL QVIKFTKDLP VFRSLPIEDQ ISLLKGAAVE ICHIVLNT.. RxR ICQAADKQLF TLVEWAKRIP HFSELPLDDQ VILLRAGWNE LLIASFSHRS RAR FSELATKCII KIVEFAKRLP GFTGLSIADQ ITLLKAACLD ILMLRICTRY 3 9 6 4 4 5 h E R a MEHPVKLLF APNLLLDRNQG KC. . VEGMVEIFDMLLATSSRFRMMNLQGEE AhR •GKDGALLP PQLALFAIATP LQ. . PPSILEIRTKNFIFRTKHK.LDFTPIG MER ME.HPGKLLF APNLLLDRNQG K C .. VEGMVEIFDMLLATSSRFRMMNLQGEE LIGAND h E R b ID.HPGKLIF APDLVLDRDEG KC. . VEGILEIFDMLLATTSRFRELKLQHKE hPX R M F.DTETGTWECGRLAYCFED. PNGGFQKLLLDPLMKFHCMLK.K.LQLHKEE CAR TFCLQTQNFL. CGPLRYTIEDG ARVGFQVEFLELLFHFHGTLR.K. LQLQEPE RxR lAVKDGILLA TG.LHVHRNSA HSAGVGAIFDRVLTELVSKMRD..MQMDKTE RAR TPEQDTMTFS DG.LTLNRTQM HNAGFGPLTDLVFAFAGQLLP. . . LEMDDTE 4 4 6 4 9 5 h E R a FVCLKSIILL NSGVYTFLSS TLKSLEEKDH IHRVLDKITD TLIHLMAKAG AhR CDAKGQLILG YTEVELCTRG S ...... MER FVCLKSIILL NSGVYTFLSS TLKSLEEKDH IHRVLDKITD TLIHLMAKAG BINDING h E R b YLCVKAMILL NSSMYPLVTA TQDADS.SRK LAHLLNAVTD ALVWVIAKSG hPX R YVLMQAISLF S.PDRPGW Q ...... RSV VDQLQERFAL TLKAYIECS. CAR YVLLAAMALF S . PDRPGVTQ ...... RDE IDQLQEEMAL TLQSY IK G Q . RxR LGCLRAIVLF N .P ...... DSKGLSNPAE VEALREKVYA SLEAYCKHK. RAR TGLLSAICLI C.G ...... DRMDLEEPEK VDKLQEPLLE ALRLYARRR. 4 9 6 5 4 5 h E R a LTLQQQHQRL AQLLLILSHI RHMSNKGMEH LYSMKCKNVV PLYDLLLEML AhR ..GYFIHAAD IILHCAESHI R.M IKTG.E. . SGMTVFRLL AKHSRVJRWVQ MER LTLQQQHRRL A Q L L L IL S H I RHMSNKGMEH LYNMKCKNVV PLYDLLLEML DOMAIN h E R b ISSQQQSMRL ANLLMLLSH*/ RHASNKGMEH LLNMKCKNVV PVYDLLLEML hPX R RPYPAHRFLFLKIMAVL.TELRSINAQQTQQLLRIQDSHPFA.TPLMQELFSST CAR QRRPRDRFLYAKLLGLL.AELRSINEAYGYQIQHIQG.LSAM.MPLLQEICS.. RxR . . YPEQPGRF AKLLLRLPAL RSIGLKCLEH LFFFKLIGDT PIDTFLMEML RAR ..RPSQPYMF PRMLMKITDL RGISAKGAER AITLKMEIPG PMPPLIREML 5 4 6 5 9 7 h E R a DAHRLHAPTS RGGASVEETD QSHLATAGST SSHSLQKYYI TG.EAEGFPATV AhR SRWRWVQ MER DAHRLHAPAS RMGVPPEEPS QTQLATTSST SAHSLQTYYI PP.EAEGFPNTI h E R b NAHVLRG..C KSSITGSECS PAEDSKSKEG SQNP hPX R DGCAR RxR EAPHQMT...... RAR ENPEMFEDDS SQPGPHPNAS . SEDEVPGGQ GKGGLKSPA...... hera=human oestrogen receptor alpha;ahr=aryl hydrocarbon receptor; mer=mouse oestrogen receptor alpha; cer=chicken oestrogen receptor alpha; herb=human oestrogen receptor beta; hpxr=human pregnane X receptor; car=constitutive androstane receptor; rxr=retinoid X receptor; rar=retinoic acid receptor. A=alanine; C=cysteine; D=aspartate; E=glutamate; F=phenylalanine; G=glycine; H=histidine; I=isoleucine; K=lysine; L=leucine; M=methionine; N=asparagine; P=proline; Q=glutamine; S=serine; T=threonine; V=valine; W=tryptophan; Y=tyrosine. The colouring system shown here for illustrative purposes gives the following chemical information: Green = cyclic; ; cyan = aliphatic hydrophobic; red = acidic; blue = basic; magenta = amides The numbering system refers to the human ERa crystal structure (Brzozowski et al., 1997).

129 Figure 3.6. A section of the multiple sequence alignment of PPAR and RXR sequences from human, rat and mouse.

101 150 hrxr VSSSEDIKPP LGLNGVLKVP AHPSGNMASF TKHICAICGD RSSGKHYGVY mppar ASTDESPGSA LNIE...... CRICGD KASGYHYGVH rppar TSTDESPGNA LNIE...... CRICGD KASGYHYGVH hppar GSVDESPSGA LNIE...... CRICGD KASGYHYGVH DNA hpparg KPHEEPSNSL MAIE...... CRVCGD KASGFHYGVH hppard MGCDGASCGS LNME...... CRVCGD KASGFHYGVH 151 200 BINDING hrxr SCEGCKGFFK RTVRKDLTYT.CRDNKDCLID KRQRNRCQYC RYQKCLAMGM mppar ACEGCKGFFR RTIRLKLVYDKC.D.RSCKIQ KKNRNKCQYC RFHKCLSVGM rppar ACEGCKGFFR RTIRLKLAYDKC.D.RSCKIQ KKNRNKCQYC RFHKCLSVGM hppar ACEGCKGFFR RTIRLKLVYDKC.D.RSCKIQ KKNRNKCQYC RFHKCLSVGM DOMAIN hpparg ACEGCKGFFR RTIRLKLIYDRC.D.LNCRIH KKSRNKCQYC RFQKCLAVGM hppard ACEGCKGFFR RTIRMKLEYEKC.E.RSCKIQ KKNRNKCQYC RFQKCLALGM

201 hrxr KREAVQEERQ RGK.DR.... mppar SHNAIRFGRM P .RSEKAKLK AEILTCEHDL KDSETADLKS LGKRIHEAYL rppar SHNAIRFGRM P .RSEKAKLK AEILTCEHDL KDSETADLKS LAKRIHEAYL hppar SHNAIRFGRM P .RSEKAKLK AEILTCEHDI EDSETADLKS LAKRIYEAYL hpparg SHNAIRFGRI A .QAEKEKLL AEISSDIDQL .NPESADLRA LAKHLYDSYI hppard SHNAIRFGRM P .EAEKRKLV AGLTANEGSQ YNPQVADLKA FSKHIYNAYL

220 I < helix 1 > hrxr ...NENEVE. .STSSANE.D MPVERILEAE .LAV.EPKTETYVEANMGL.N mppar KNFNMNKVKA RVILAGKTSN NPPFVIHDME TLCMAE.KTLVAKMVANGV.E rppar KNFNMNKVKA RVILAGKTSN NPPFVIHDME TLCMAE.KTLVAKMVANGV.E hppar KNFNMNKVKA RVILSGKASN NPPFVIHDME TLCMAE.KTLVAKLVANGI.Q LIGAND hpparg KSFPLTKAKV RAILTGKTTD KSPFVIYDMN SLMMGE.DKIKFKHITPLQEQ hppard KNFNMTKKKA RSILTGKASH TAPFVIHDIE TLWQAE.KGLVWKQLVNGP.P

260 < helix 3 290 < helix4 X hrxr PSSPNDPVTN ICQAADKQLF T .LVEWAKRI PHFSELPLDD QVILLRAGWN BINDING mppar DKEAEVRFFH CCQCMSVETV TELTEFAKAI PGFANLDLND QVTLLKYGVY rppar NKEAEVRFFH CCQCMSVETV TELTEFAKAI PGFANLDLND QVTLLKYGVY hppar NKEVEVRIFH CCQCTSVETV TELTEFAKAI PAFANLDLND QVTLLKYGVY hpparg SKEVAIRIFQ GCQFRSVEAV QEITEYAKSI PGFVNLDLND QVTLLKYGVH hppard YKEISVHVFY RCQCTTVETV RELTEFAKSI PSFSSLFLND QVTLLKYGVH DOMAIN

The amino acid designations as the same as indicated in Figure 3.5. The numbering system refers to the human PPARy crystal structure (Nolte et al., 1998)

130 Table 3.2. Comparison of the volumes of nuclear receptor LBD cavities of known structure, (Adapted from Watkins et al., 2001 and Williams et al., 2003 indicated by *)

Receptor Ligand binding cavity volume (A" ) ERa 476 LXRp* 830 PXR 1,150 PPARy 1,619 Progesterone receptor 557 Vit D receptor 871

3.4. Results and Discussion Table 3.3 shows the results of the modelling of the receptors, including the energy before and after ligand binding, and indicates the main amino acid contacts between the ligands and receptors. Figures 3.2 to 3.8 show the LBDs of the in silico homology receptor models. Table 3.3. Results for homology models of selected receptors with known ligand interactions

Receptor Energy before Ligand Energy after Main contacts between ligand binding binding and receptor Kcal/mol Kcal/mol (ERa alignment) (200 iterations) (200 iterations) ERP -567.195 n p oestradiol -611.802 H bonding‘s with: Histidine, coumestrol -626.966 Arginine, Glutamate

PXR -558.873 pregnenalone -605.979 H bonding with: Glutamines dexemethasone -602.867" Glutamate, Serine, Histidine

CAR -542.440 androstane 3P -603.932 H bonding with: Arginine, lip diol-17 -one Aspartate, Gutamine, Histidine. Tyrosine may bond with longer structures or if androstane is moved further into the LBD Hydrophoblic contacts: Leucine, Valine, Cysteine, Methionine

AhR -334.982 2,3,7,8 TCDD -364.120 H bonding with: Arginines PCB 126 -368.301 7Ü-7Ü stacking between Phenylalanines PPARa -792.255 arachidonic acid -848.050 Ion bonding with: Histidine, bezafibrate -836.033 Lysine, Cysteines H bonding with: Serine 400 iterations H bonding: Hydrogen bonding Note: The values given in this table are relative and not absolute. They are included to demonstrate the stable low energy conformations of the proteins.

131 3.4.1. Promiscuity of ligands; cross talk of receptors Xenobiotics may act on some but not all of the receptors and their isoforms in the tissues of these organs, or act to different affinities, as methoxychlor and its analogue DDT do in ERa and ERp (Gao et al., 1999). Taking the ERs as a specific example, ERP is dependent on pure agonists for the activation of transcription from its target promoters, while ERa can be activated by agonists, partial agonists (such as tamoxifen, which is used in the treatment of ER+ breast cancer) and ligand independent mechanisms.

3.4.2. Evidence of ‘cross-talk’ Not only do receptors often have ligands in common (although with different binding affinities), there is also a great deal of ‘cross-talk’ and ‘ligand promiscuity’. For example, forms of the endogenous oestrogenic (oestradiol) and androgenic (androstanes) hormones are both ligands for ERp. Oestradiol is less potent in ERP than in ERa, whilst the natural ligands for ERp may actually be androgens: 5a androstane diol and Sp androstane diol (Gustafsson, 2000). The organochlorine pesticides trans-nonachlor and chlordane are known to activate both known oestrogen receptors (ER), and the pregnane X receptor (PXR), but with different affinities (Kuiper et al., 1997; Barkhem et al., 1998; Waxman, 1999; Jones et al., 2000; Brady et al., 2002). As these receptors are present in different ratios in different cell types and tissues, the response on a cellular, tissue and systemic level may be quantitatively very different, and may vary over time. Receptor modulation has been seen with lactation, when a form of ERp has been observed to increase in the rat mammary glands (Gaido et al., 1999; Gustafsson, 2000), and in breast tissue hyperplasias, where a frequent mutation in the ERa gene shows increased sensitivity to oestrogen compared with wild type ERa, by affecting the border of the hinge and hormone binding domains (HBD) (Fuqua, et al., 2000). There is evidence that isoforms of different receptors modulate each other at a functional level, attempting to retain a balance. The modulation is aided at low (Hall and

McDonnell, 1999), and high hormone levels (Petersen et al., 1998) by different ERP isoforms - this may enable a tissue to govern its own responsiveness to oestradiol, oestradiol metabolites and related hormones such as progestins, and oestrogen mimics or endocrine disrupting compounds (EDCs). This speculation is supported by what appears to be an emerging pattern in nuclear receptor signaling, as similar balancing acts have been observed in the a and P forms of the human glucocorticoid receptor, the progesterone

132 receptor, and now also between PXR and CAR (Watkins et al., 2001; Xie and Evans, 2001; Maglich et al., 2002; Honkakoshi et al., 2003). Phytoestrogens appear to have a greater affinity for ERp than ERa (Kuiper et al.,

1997), but they have an ERa selective efficacy, while the hydroxylated metabolites of methoxychlor appear to be an ERa specific agonist, and ERp antagonist (Gaido et al., 1999), but with about the same affinity for both isoforms (Kuiper et al., 1997). Figure 3.7. Homology model of human oestrogen receptorP (left) next to human oestrogen receptora crystal structure (right), hoth with oestradiol docked in the LBD

17 P oestradiol in the respective LBDs of ERa and ERp. The pink ribbon denotes the protein backbone of the receptors, and all the amino acids for both the model and the crystal structure are illustrated in this figure.

3.4.3. The Pregnane-X-Receptor (PXR) Examples of hydrogen bonding and hydrophobic contacts with key amino acid sidechains for pregnenolone, coumestrol and PCN were Gin 521, Gin 524 and Glu 404. The steroids, progesterone and corticosterone, displayed hydrogen bonding with the same amino acids, plus Ser 350. His 388 may well have a significant role to play with certain ligands such as tamoxifen, as it can both accept and donate hydrogen bonds, as well as change orientation to facilitate binding. Thus five key polar residues, including glutamine, glutamate, serine and histidine, are important for ligand binding in this hPXR model, although not necessarily all of them for every ligand. The hPXR crystal structure was published recently, both

133 alone and complexed with the high affinity ligand SR 12813, and here the five key polar residues critical for establishing the activation of PXR in the LBD were Ser 208, Ser 247, His 407, Arg 410 and Gin 285. The ligand-binding pocket is highly flexible and may be able to expand the number of possible hydrophobic contacts by enlarging via a pore (Watkins et al., 2001). Two of these amino acids. Gin 285 and His 407, have been shown by Ostberg et al, 2002, to be essential for the species specific activation of PXR in mouse/human PXR receptor studies. Mutating the Glutamine residue altered the activation of the receptor from rifampicin, the positive control for human PXR ,to PCN, the positive control used for rodent PXR species. The Histidine residue, on the other hand, appears to be important for the basal level of activation of PXR (Ostberg et al.,

2002). Comparisons between the homology model presented herein, and the recently published crystal structure of PXR (Watkins et al., 2001) and pharmacophore (Ekins and Erickson, 2002) compare favourably (figure 3.3 and 3.4), indicating that the use of homology modelling for receptors, as with enzymes, is a useful tool when crystal structure data are not available. Figure 3.8. Molecular homology model of PXR with pregnenolone docked in the LBD plotted from the crystal structure coordinates of ERa.

Figure 3.9. Crystal structure of hPXR with pregnenolone docked in the LBD.

m Figure 3.10. a) Molecular homology model of CAR with androstanol docked in the LBD, plotted from the crystal structure coordinates of hERa, left and b), close up photograph of the CAR LBD model, right. For all figures the ribbon denotes the protein backbone of the respective receptor, and hydrogen bonding of the ligands with key amino acids is indicated by the dashed blue lines.

c) PB docked in the LBD and d) methoxychlor, plotted from the crystal structure coordinates of hERa

e) TCPOBOP docked in the LBD, plotted from the crystal structure coordinates of hERa, compared with f) TCPOBOP in the LBD of the PXR crystal structure

135 3.4.4. The Constitutive Androstane Receptor (CAR) Figures 3.10a shows the molecular homology model of CAR plotted from the crystal structure coordinates of ERa. Close-up pictures of the following compounds docked in the homology model of the CAR LBD, include 3.10b) androstanol; 3.10c) phénobarbital (PB), which probably acts through a phosphorylation mechanism; 3.10d) the agonist methoxychlor, and 3.10e) TCPOBOP compared with 3.10f) TCPOBOP in the LBD of the PXR crystal structure. This model was built before the PXR crystal structure coordinates were made available, and subsequently a homology model of CAR based upon the PXR crystal structure has been built and published also (Maglich et al., 2003). The CAR LBD appears to be far less promiscuous, and thus less accommodating for ligands than the PXR LBD. CAR has chaperone proteins in common with the ERs, so models of CAR based upon the crystal structure of ERa and PXR may be able to indicate different aspects in the puzzle of CAR activation. There is clear evidence of cross talk between PXR and CAR and marked pharmacological differences for the same ligands in the different receptors, as is being observed between the ERs. Thus clotrimazole, for example, is a PXR activator, but a potent deactivator in CAR (Moore et al., 2000). CAR has also been observed to be transactivated through the CYP3A4 xenobiotic response element that serves as the PXR/RXR binding site (Sueyoshi et al., 1999) and has been observed to be sequestered in the intact liver, in the absence of activators (Kawamoto et al., 1999). A ligand that affects androstanol and androstenol levels will also affect CAR and possibly the steroid hydroxylase CYP2B. CAR also has PB like PCB ligands in common with AhR. Both CAR and PXR m RNAs are also markedly reduced by the cytokine interleukin-6 (IL-6), but this has not been observed for the AhR or glucocorticoid receptor (GR) (Pascussi et al., 2000b).

3.4.5. The Ah Receptor (AhR) 2,3,7,8 TCDD (figure 3.11) and 3,3’,4,4’,5 PCB (lUPAC 126) (figure 3.12), both high affinity ligands, were docked in the LBD. Differences in binding were observed. The HOMO energy appears to describe an important facet of the ligand binding process, which could involve a n-n stacking interaction (see figure 3.11) between benzene rings on the PCB molecule and one or more aromatic amino acid residues (eg. phenylalanine)

136 in the AhR ligand binding site itself. The ligand-binding site preferentially accepts relatively planar aromatic molecules within a specific rectangular envelope. Hydrogen bonding does not appear to be as significant as it is with 2,3,7,8 TCDD. For AhR ligands, QSAR studies indicate the HOMO energy to be an important descriptor with molecular planarity, together with the overall rectangularity as measured by the length/width ratio, planarity, length and energy of the highest occupied MO (Lewis and Jacobs, 1999). These descriptors have also been described for synthetic retinoids, with structural similarities to 2,3,7,8 TCDD and the coplanar PCBs, for the retinoic acid receptor (RAR) and the AhR/ARNT pathway (Gambone et al., 2002). Figures 3.11 and 3.12 illustrate the good steric complementarity for TCDD and closely related compounds such as PCB 126. The LBD of the AhR appears to also be able to accommodate many other recently identified dietary ligands, suggesting a dynamic plasticity in the LBD that may be similar to PXR, in vitro and in vivo. Another AhR homology model has been published recently, a theoretical model for the mouse AhR based on available crystal structures of homologous PAS family members (Procopio et al., 2002). This model is based on three PAS proteins (bacterial photoactive yellow protein, PYP, human potassium channel, HERG, and the heme binding domain of bacterial FixL protein) with very different activation and function as well as low sequence similarity, although there are some highly conserved structural characteristics to that of the AhR. These proteins do not appear to have cross talk interactions with SHRs (and chaperone proteins) to the same degree that the AhR does. Hsp90 for example regulates AhR activity in vivo, and Ah responsiveness is partly dependent upon cellular ER content through a mechanism that involves Hsp90 (Caruso et al., 1999; Petrulis and Perdew, 2002). RIP40 also interacts with both receptors (Kumar et al., 1999). At the time the AhR model based upon the ERa crystal structure was built (Jacobs and Lewis, 2000a) it provided an explanation of the likely reason for the high affinity binding of TCDD in the LBD, and continues to provide information on likely binding parameters in the LBD, which can be utilised further in molecular in vitro mutagenesis studies and ligand binding studies, in support QSAR analyses.

137 Figure 3.11. Molecular homology model of AhR with 2,3,7,8 TCDD docked in the LBD (hydrogen bonding indicated by the yellow dashed lines), plotted from the crystal structure coordinates of ERa.

Figure 3.12. Molecular homology model of AhR with PCB 126 docked in the LBD (showing 7 1-71 stacking), plotted from the crystal structure coordinates of ERa.

3.4.6. The Human Peroxisome Proliferator Activated Receptor (PPAR)

Rosiglitazone, a potent ligand in PPARy but not PPARa, occupied a fraction of the available LBD space in PPARa, less than that in PPARy, particularly the rosiglitazone TZD head group. Headgroup interactions in PPARa are different to those in

PPARy because of a His to Tyr mutation. Thus comparatively reduced selectivity was

138 observed in silico as reported experimentally for different ligands in the PPAR family, and is a clear descriptor for PPAR selectivity. Figure 3.13. a) Homology model of human PPARa with rosiglitazone docked in the LBD, and b) A closer view of rosiglitazone docked in the ligand binding pocket of hPPARa showing binding with key amino acid residues a) b)

3.5. Conclusion The nuclear receptors modelled display a spectrum of ligand specificities, ranging from the highly specific, as seen in CAR which binds 5a-androstan-3a-ol

(androstanol) but not 5a-androstan-3(3-ol (Moore et al., 2000b) and seen in PPAR selectivity, to the highly non-specific, such as hPXR, which is very flexible, and can bind with a large number of wide ranging molecules, from rifampicin to steroidal structures. They also display a spectrum of binding modes within the LBD, from hydrogen bonding with variable key amino acids (as observed in all the receptors) to tt-tc stacking, as seen in the AhR, and also binding outside the LBD, as seen with antagonists such as tamoxifen in ERa. It is also likely that other amino acids play key roles in ligand binding, but do not bind directly, and this is an area that could be usefully explored by using the crystal structures and models further. For instance, with ERa, where key ligand binding occurs with His524, Arg 394 and Glu353 (Figure 1.3), it appears that mutation of L536 alters the coupling between ligand binding, transcription activation, and receptor conformation (Zhao et al., 2003). The receptor models can be used to explore modes of binding of different ligands in the LBD, and conduct Quantitative Structure Activity Relationships in conjunction

139 with experimental ligand binding data from both the literature and in house in vitro systems. The identification and in silico assessment of the different ligands both in isolation and within the receptor models may add to a better knowledge of their specificity and may help explain the selective action of steroids and xenobiotics in different tissues. This information can be used to explore novel development compounds for therapeutic intervention in various functions that involve the PPARs, ERs, AhR, CAR and PXR, and their associated GYP enzymes, together with other chemicals of interest, such as industrial chemicals and dietary compounds. Since this work was conducted, the crystal structures of human ER(3 (Pike et al., 1999), PXR (Watkins et al., 2001), and PPARa (Cronet et al., 2001; Xu et al., 2002) have become available, and they compare favourably with the models described herein, as illustrated in figure 3.8 and 3.9, supporting the value of homology receptor protein models where crystal structures have not yet been determined. The models and crystal structure described here are utilised further in chapter 5.

140 C hapter 4

In Vitro Biological Parameter Estimation In Nuclear Receptors

141 Chapter 4 Experimental Case Study Activation of PXR by Dietary Ligands: In Vitro Transient Transfection Assays

4.1. Introduction: PXR activation

The nuclear receptors are a family of ligand-activated transcription factors that includes the steroid, retinoid, and thyroid hormone receptors. In 1998, a new member of this family was reported that is activated by nearly all of the structurally diverse compounds that induce CYP3A gene expression (Kliewer et al., 1998; Lehmann et al., 1998; Bertilsson et al., 1998; Blumberg et al., 1998). Based on its activation by natural and synthetic C21 steroids, the new receptor was named the pregnane X receptor by Kliewer and co-workers (Kliewer et al., 1998) or Steroid X Receptor (SXR) by Evans and colleagues (Blumberg et al., 1998). PXR binds as a heterodimer with the 9-cis retinoid X receptor a (RXRa) to DNA response elements composed of two copies of the consensus sequence AGTTCA organized as either a direct repeat with a three nucleotide spacer (DR3) or an everted repeat with a six nucleotide spacer (ER6). DR3 and ER6 response elements have been identified in the regulatory regions of rat, rabbit, and human CYP3A genes. The binding of ligands to PXR causes a conformational change in the receptor’s ligand binding domain (LBD) and results in the transcriptional activation of CYP3A4 and other PXR target genes (Kliewer et al., 1998; Lehmann et al., 1998; Bertilsson et al., 1998; Blumberg et al., 1998). Mutation of four key residues in the LBD that contact the ligand, and are divergent among species, can shift the ligand specifity from mouse to human PXR (Watkins et al., 2001). That very few key residues in NR LBDs may need to be mutated to affect the mechanism of activation has been observed for other receptors also, e.g. such as the AR (Rabenoelina et al., 2002) and ERa (Zhao et al., 2003). The sequence differences in the LBD of PXR account for most of the variation in CYP3A induction across species. Compounds that are known to differentially induce CYP3A expression in rat or rabbit or human activate only the corresponding PXR (Jones et al., 2000b; Xie et al., 2000).

142 As discussed in Chapter 1, PXR upregulates the body’s main chemical defense system in response to challenge primarily by inducing CYP3A4, an enzyme that is involved in the metabolism of >60% of all marketed drugs, including chemicals as diverse as macrolide antibiotics, imidazole antimycotics, benzodiazepines, and steroids (Maurel, 1996; Waxman, 1999; Moore et al., 2000b). Notably, transcription of the CYP3A4 gene is increased in response to an extensive array of drugs and other xenobiotics, see figure 4.1. The inducibility of CYP3A4 expression, coupled with the broad substrate specificity of this enzyme, represents the basis for an important class of drug-drug, phytochemical-drug-contaminant interactions. Humans are exposed to many ligands. In the pharmaceutical industry, until recently, it has been very difficult to accurately predict the clinical induction potential of drug candidate molecules because of the tremendous species differences seen in the induction of CYP3A enzymes. Thus, induction studies have generally been performed in primary human hepatocytes, which are difficult to obtain and vary considerably in quality. PXR transient transfection assays are utilised by the pharmaceutical industry for the prediction of the CYP3A induction potential of drug candidates. These assays also have applications for the analysis of dietary and endocrine disrupting compounds. There is increasing evidence that potent inducers of CYP3A can be present in the human diet, from food constituents, food contaminants and over-the- counter herbal remedies and vitamin supplements. These can be as potent as pharmaceutical compounds, as demonstrated by the effect of hyperforin (St John's Wort), which is ten-fold more potent than the prototypical PXR ligand, rifampicin (Moore et al., 2000b). Favourable comparisons of PXR activation have been reported between human hepatocytes and PXR transient transfection assays (Luo et al., 2002). Thus the recent development of high capacity assays to predict the CYP3A4 induction potential of drug candidates gives an early indication of the need for interaction studies and is considered to be a cost effective decision making tool to expedite the drug discovery process, particularly in the in-licensing arena where limited or incomplete pharmacology and toxicology information is available (Nolan et al., 2003). The PXR transient transfection assay is a tool that can result in safer medicines, particularly with the current prevalence of polypharmacy, but can also be part of a tier testing approach to assess endocrine disruption and dietary compound interactions mediated through PXR and CYP 3 A4.

143 Here, the cell-based reporter assay is described and the results are discussed. The data generated from this assay were used to construct and test a PXR QSAR model (Chapter 5). Figure 4.1. PXR: Mode of action

C om pound A

I ___ _ -'X Transcription RNA poly II Ù------CYP3A4 mRNA

Transiatlon Promoter CYP3A4 gene

Cotiipoimd B Cotnpniinrl H o ru ^ D CcNTHtnufHl C c -OH Increased Drug/Ptrytochemical/ Xenobiotic Metabolism —Interactions; drug-drug; Phytochemlcai (diet) - drug - contaminant

4.2. Materials and Methods: 4.2.1. Selection of chemicals Chemicals were selected to be representative of likely dietary and drug exposure as well as including some experimental NCEs, to yield a good spread of activity, by reviewing the literature (please see the previous chapters) and searching directories of commercial and phytochemical dictionaries (Merck 1996; Harboume et al., 1999) together with in-house GSK data libraries displaying chemical structures. The compounds finally tested were a representative selection of ligands showing weak to high affinity for binding with PXR, many of these compounds are also ligands for other NRs. Ideally a larger data set would be generated, given a longer timescale and sufficient funding.

4.2.2. Ceil Based Transfection Assay

This assay profiles compounds for their efficacy and potency on PXR in a cellular context, and is considered a straightforward format for studying PXR activation in different species. The cell-based transfection assay developed at GlaxoWellcome, now GSK, to measure PXR activity is performed in two host cell lines, HuH7, human hepatoma, and CV-1, African Green monkey kidney cells. The assay has been validated in two species, rat and human. The HuH7 cell assay uses transient over expression of the

144 full-length receptor and a corresponding reporter plasmid, which contains a PXR CYP3A4 responsive element in front of the minimal thymidine kinase (tk) promoter. The CV-1 cells use a truncated version of the receptor and plasmid. Both cell lines were utilised in this project, both for training and comparative purposes. The transfection method varied in terms of transfection reagent, cell type, reporter gene constructs and drug, as described below, but a standard scheme was followed, as established at GSK for the prediction of CYP3A4 induction potential in humans. All cell work was carried out in a Class II biological safety cabinet under aseptic conditions in a dedicated cell culture laboratory at GSK, Ware.

Cells were seeded onto 96 well plates at a density of 20,000 cells per well. After 24 hours the cells were transfected with full length hPXR (PSG5-PXR) for HuH7 cells, or the chimera (UAStkSPAP) for CV-1 cells, the reporter gene construct (CYP3AtkSPAP) and pCHllO containing p-galatosidase to correct for transfection efficiency. The following day the cells were dosed with test compound at lOuM to O.OluM. After 24 hours the culture medium was assayed for SPAP activity and the cell lysate assayed for Pgal activity. Activation of PXR by a compound was expressed as a percentage of that achieved with lOuM positive control, rifampicin (Rif) (EC 50 ca luM).

When a full dose response curve was achieved, the EC 50 value and thus potency can be calculated. If a full dose response is not achieved, the EC 50 value is reported as ‘>10uM’. Table 4.1. Interpretation of PXR transient transfection assay results

% M ax EC50 uM Interpretation

0 % >10 No response <30% >10 Weak response 30-70% >10 Moderate response >70% >10 PXR activation at high concentrations only >70% <10 PXR activation with some potency

4.2.2.1 Preparation o f nuclear receptor constructs The PXR chimeric constructs required for the CV-1 cell transfections were prepared and supplied in advance by Linda Moore, Systems Research, GSK, RTP (Kliewer et al., 1998; Moore et al., 2000b). Briefly, the GAL4-chimera expression constructs (pSGSStratagene) were made to contain the translation initiation sequence (Kozak sequence) and amino acids 1-76 of the glucocorticoid receptor fused to amino

145 acids 1-147 of the yeast transcription factor GAL4 DNA binding domain. The cDNA’s encoding LBD s of the nuclear receptors were amplified by polymerase chain reaction (PGR) and inserted C-terminal to the GAL4 DNA binding domain to generate the final chimeric nuclear receptor plasmids. The cDNA was transformed into competent cells (DH5a) using DNA prepared in Qiagen Maxi-Prep kits. Preparation of pSG5 human PXR:pSG5-fl hPXR-ATG The HuH7 cells were transfected with the full-length coding region of human PXR with a modified start codon, obtained using hybridisation technology and provided ready prepared. The expression plasmid pSG5-hPXR-ATG was generated by PGR amplification of cDNA encoding amino acids 1-434 of hPXR and inserted into the pSG5 expression vector (Stratagene). Preparation of reporter plasmid : Five copies of a GAL4 DNA binding element were inserted N-terminal to the minimal thymidine kinase promotor from the herpes simplex virus, followed by SPAP in the pGI12 expression vector, using Qiagen Maxi-Prep kits and provided ready prepared. Preparation of internal control plasmid: P-galactosidase (pGHllO from Pharmacia) was also prepared using Qiagen Maxi-Prep kits and provided as above. Preparation of pBluescript plasmid: pBluescript (Stratagene) was also prepared using Qiagen Maxi-Prep kits. Preparation of coactivator: The human SRG-1 p i40 was cloned into the pSG5 vector at the Bglll-Bglll site with a kozak sequence just before the ATG and was also prepared using Qiagen Maxi-Prep kits and provided as above.

4.2.2.2. PXR Species differences: Rat versus Human This transient cotransfection assay has a number of other uses. By transfecting different species specific PXR plasmids, species specific PXR and GYP 3A induction liabilities may be determined. Therefore, in GSK, rat PXR is routinely used as an assay for testing the induction potential of experimental drug compounds to predict GYP3A23 induction, and a variety of other species have been tested, these show divergent responses compared to induction in cell lines transfected with human PXR.

Pregnenolone 16a carbonitrile (PGN) is used as the positive control (EG50 ca luM ), at a top dose of lOuM in rat studies.

146 4.2.3. Materials 4.2.3.1. Chemicals and supplies Hyperforin, genistein, and coumestrol were purchased from Apin Chemicals (Abingdon, Oxon, UK). GW compounds and other phytochemicals were kindly donated by GSK. Lovastatin, lithocholic acid and 7 dehydrocholesterol (Vitamin D 3) were originally purchased from Sigma (Sigma-Aldrich Company Ltd. Poole, Dorset, UK), and subsequently were kindly donated by GSK. Triclosan (5-chloro-2-2(2,4- dichlorophenoxy)phenol), trans-nonachlor, PCB 118 and PCB 153 were purchased from LGC Promochem (Hatfield, Herts, UK).

4.2.3.2. Cotransfection assays 4.2.3.2.1.Cell maintenance reagents DME High Glucose HEPES Medium (Invitrogen), Foetal Bovine Serum (Hyclone), 6mL aliquot of 2mM L-Glutamine (Gibco), 6mL aliquot of Penicillin- Streptomycin (Gibco), DMEM/F-12 Medium without phenol red (Gibco), Foetal bovine serum Dextran Treated/Charcoal Stripped (Hyclone), HuH 7 Cells, CV-1 cells, 225cm^ cell culture flasks (Costar), 96 well plates for cells (Falcon), Phosphate Buffered Saline, 0.25% Trypsin EDTA. Trypan Blue Dye.

4.2.3.2.2. Cell media Maintenance Medium, Experimental Medium, 2X Recovery Medium, IX Drugging Medium, SPAP Substrate buffer, B-Gal Substrate buffer.

4.2.3.2.3. Nuclear receptor constructs for transfection Nuclear receptor constructs were very kindly provided by Linda Moore from GSK, RTP as per section 4.2.2.1. Additional reagents required included Lipofectamine Plus, Opti-MEM-1 reduced Serum Medium, 96 well deep-well blocks (Beckman), Triton X-100, Diethanolamine, NaCl, MgCLbHiO, pNPP (Sigma), ONPG (Sigma), Na 2HP04, NaH2P04.

147 4.2.4. Methods for media preparation 1. Maintenance Medium DME High Glucose HEPES medium, fetal bovine serum, one aliquot of 2mM L- glutamine and one aliquot of penicillin-streptomycin were warmed to 37°C. Each item was sprayed with 70% ethanol and placed in a Class II cabinet. 500mL DME High Glucose HEPES medium, 50 mL fetal bovine serum, 5 mL of 2mM L-glutamine and 5 mL of penicillin-streptomycin were then filtered through a 0.22pm filter system and gently mixed by swirling, to produce a medium ready for use. The medium shelf-life was 10 days when stored at +2°C to 8°C. 2. Experimental Medium DMEM/F12, Dextran treated/Charcoal Stripped foetal bovine serum, one aliquot of 2mM L-glutamine were warmed to 37°C. Each item was sprayed with 70% ethanol and placed in a Class II cabinet. 500mL DME High Glucose HEPES medium, 50 mL fetal bovine serum, 2 mL of 2mM L-glutamine were then filtered through a 0.22pm filter system and gently mixed by swirling, to produce a medium ready for use. The medium shelf-life was 10 days when stored at +2°C to 8°C. 3 .2X Recovery Medium DME High Glucose HEPES medium, delipidated heat inactivated foetal bovine serum and two aliquots of 2mM L-glutamine. Each item was sprayed with 70% ethanol and placed in a Class II cabinet. 500mL DME High Glucose HEPES medium, lOOmL delipidated fetal bovine serum, lOmL of 2mM L-glutamine were then filtered through a 0.22pm filter system and gently mixed by swirling, to produce a medium ready for use. The medium shelf life was 10 days when stored at +2°C to 8°C. 4. IX Drugging Medium DME High Glucose HEPES medium, delipidated heat inactivated foetal bovine serum, one aliquot of 2mM L-glutamine and one aliquot of penicillin-streptomycin were warmed to 37°C. Each item was sprayed with 70% ethanol and placed in a Class II cabinet. 500mL DME High Glucose HEPES medium, 50 mL delipidated fetal bovine serum, 5 mL of 2mM L-glutamine, 5 mL of penicillin-streptomycin were then filtered through a 0.22pm filter system and gently mixed by swirling, to produce a medium ready for use. The medium shelf life was 10 days when stored at +2°C to 8°C.

148 5. SPAP Substrate buffer The SPAP Substrate buffer was made as follows: 16.36g NaCl and 0.102g MgCl 26H20 were dissolved in 800mL MilliQ water, then 105.5 mL Diethanolamine was added. The pH was adjusted to 9.85 using concentrated hydrochloric acid, and the final volume adjusted to lOOOmL with MilliQ water. The buffer stock was kept for up to two months at room temperature. 6. B-Gal Substrate buffer The B-Gal Substrate buffer was made as follows: 68.4mL IM Na 2HP04 and 31.6 mL IM NaH 2P04 were added to 900 mL MilliQ water. The buffer stock was kept for up to two months at room temperature.

4.2.5. Cell culture and maintenance Both cell types were stored frozen in maintenance medium or 90% foetal bovine serum (FBS) plus 10% dimethylsulfoxide (DMSO), at a density of 10^ cells/mL, in liquid nitrogen. To begin culturing the cells, a ImL vial of cells was thawed by gently agitating at 37°C before transferring the cell suspension to a 75cm^ vented flask (Nunclon™, Nunc brand products), followed by the gradual addition of 25mL of maintenance medium. Sudden dilution of DMSO can cause osmotic shock and greatly reduce cell

viability. The flask was incubated at 37°C in an atmosphere of 5%C02 at >95% humidity for 6-24 hours to allow the cells to adhere firmly to the flask base. The following day, the medium was aspirated to limit exposure to residual DMSO from the freezing medium, and replaced with 25mL fresh maintenance medium. The cells were examined under a light microscope to establish percent confluence of the culture and were routinely split at 70-90% confluence and grown in 225 cm^ vented flasks (Nunclon™, Nunc brand products). A typical split ratio was 1:10 every Monday and Friday in maintenance medium for passaging and a ratio of 1:5 in experimental medium, three days prior to the start of an investigation. Figure 4.2 shows the morphology of the CV-1 cells, while figure 4.3 shows the morphology of the HuH7 cells. To passage the cells, the existing media was aspirated and the cells washed with 5ml of phosphate buffered saline (PBS, Dulbecco’s) to remove serum containing medium. The PBS was aspirated and 2 ml of warmed trypsin-EDTA was added and tilted gently to keep it flowing over the cells. The flask was also tapped to facilitate cell detachment. Once detached, 8 ml of warmed Maintenance media was added (See Figure

149 4.4) and mixed thoroughly to ensure all the cells were washed off from the base to achieve an overall trypsin neutralised single cell suspension. The trypsin inhibitor in the fetal bovine serum terminated the proteolytic activity of the trypsin. An aliquot of 0.5/1ml of cells was pipetted in a new sterile flask and 50ml of Maintenance media was added. The date, passage number and the flask split ratio was recorded on the new flask. The flask was returned to the incubator, first spraying with 70% ethanol to prevent contamination. The cells in the new flask would be ready to passage again in 3-5 days, depending on the split ratio and cell line. A higher ratio was used for HuH7 cells of 2-2.5, compared to 1 in CV-1 cells. The flasks were placed in the incubator and kept at 37°C with 5% C02 in air atmosphere in DME high glucose HEPES medium supplemented with 10% fetal bovine serum and 2 mM L-glutamine (Maintenance Medium), until the cells were -75% confluent. See figures 4.2 and 4.3 for CV-1 and HuH7 cell morphology, respectively, in maintenance media. Three days prior to plating for transfection, experimental cells were harvested with phenol red-free 0.25% trypsin/2 mM EDTA and collect in phenol red-free DMEM- F12/15 mM HEPES medium (Invitrogen) supplemented with 10% charcoal/dextran treated FBS (HyClone, Logan, Utah) and 2 mM L-glutamine (Experimental Medium). Figure 4.2. The morphology of CV-1 ceils in maintenance culture

'0 i

Figure 4.3. The morphology of HuH7 cells in maintenance culture

150 Figure 4.4. The morphology of trypsinised HuH7 cells

Figure 4.5. The morphology of HuH7 cells in a well (96 well plate) ready for transfection

, ' A <

v ? P :| •: ^ ■ i:..

Figure 4.6. The morphology of transfected HuH7 cells

Î.

■ï * •

151 4.2.6. Methods: Daily Experimental procedure

4.2.6.1. Day One- Cell plating After aspirating the experimental media, the cells were washed with 5ml of PBS, which was then aspirated and 2ml of warmed trypsin was added to detach the cells into a single cell suspension. 8ml of experimental media was added and mixed well. -Cell number A clean dry heamocytometer was used for cell counting, lOul of cell suspension was applied to one end, and the cells in each quadrant were examined under a light microscope, (XIO Phase contrast), and then counted using a grid system under increased magnification. The four numbers were averaged and multiplied by 10,000 to ascertain the number of cells in 1 ml of suspension. As 20,000 cells were required for each well, this number was further divided by 200,000 to derive the dilution factor. This multiple was used to deliver the final volume and ascertain the appropriate volume of experimental media to be added. lOOul of the cell suspension was pipetted into each well of a 96 well plate (Nunclon™, Nunc brand products), which were set up in triplicate for each experiment. Example calculation: Quadrant 1-93 cells Quadrant 2-85 cells Quadrant 3-104 cells Quadrant 4-94 cells Average = 94 cells X 10,000 = 940,000 cells/ml Dilution factor to seed at 200,000 cells per ml = 4.7 Therefore, take 1ml cells and add 3.7 mis media Therefore, take 20ml cells and add 74 mis media. The plates were then incubated for 24 hours at 37^C/5%C02 in humidified chambers to prevent evaporation affecting the small volumes in the plates. Figure 4.5 shows a well in a 96 well plate, plated HuH7 cells ready for transfection. If more than one flask was used, all flasks were pooled prior to counting cells to ensure consistency.

152 42.6.2. Day Two - Transfection All transfection calculations were performed using an Excel template, according to the transfection mixes required for the respective cell lines shown in tables 4.2 and 4.3, and according to the number of plates being set up per weekly experiment.

Table 4.2. hPXR transient transfection assay: transfection mix for CV-1 cells Mixl Amount per well

Nuclear receptor chimera 2ng* UAS5tkSPAP 8ng pCHllO 25ng SRC-1 p.l40 7ng pBluescript Add until total DNA /well=80ng OptiMEM-1 lOul Mix 2 Lipofectamine 0.7ul OptiMEM-1 lOul ^amount depends on the receptor’s basal activity but is normally 2ng/well Table 4.3. hPXR transient transfection assay: transfection mix for HuH7 cells

Mixl Amount per well

PXR DNA 5ng CYP3AtkSPAP 12ng pCHllO 25ng pBluescript Add until total DNA /well=65ng OptiMEM-1 lOul Lipofectamine Plus 0.7ul M ix 2 Lipofectamine 0.5ul OptiMEM-1 lOul

153 Lipofectamine Plus™ Transfection method for HuH 7 cells. Invitrogen’s Lipofectamine™ reagent (Life Technologies, Inc.) is a cationic lipid reagent that allows DNA introduction into the cells by forming complexes with the negatively charged DNA. For HuH7 cells only. Plus™ reagent is used in conjunction with Lipofectamine™ reagent to enhance transfection in adherent cell lines. The protocol suggested by the manufacturers was followed. On one occasion when there was insufficient Plus™ reagent available, the transfection was not facilitated and the experiment was unsuccessful. For HuH7 cells, the first transfection mix was mixed well and incubated for 15 minutes at room temperature, before dispensing the appropriate amount of mix 2 into mix 1, again mixing well and incubating at room temperature for a further 15 minutes. The media was aspirated from the cells, and replaced with lOOul of transfection mix before incubating for 3 hours. Figure 4.6 shows the morphology of transfected HuH7 cells.

CV-1 transfection methods CV-1 cell transfection mixes were prepared according to Table 4.2, and the first mix mixed gently and incubated at room temperature for 30 minutes then combined with the DNA mixture. The cells were meanwhile aspirated and washed once with 100|il of Opti

MEM-1 prior to the addition of lOOpl ml of diluted transfection mix to each well.

Treatment of cells post transfection After transfection treatment, the cells were then returned to the incubator for 3 hours (HuH7 cells ) or for 6 to 18 hours (CV-1 Cells), after which the transfection process was halted by aspirating the transfection mix, and replacing it with lOOul Opti-MEM. The test compounds were prepared as for HuH7 cells and applied to the cells after removal of the Opti-MEM. Washing CV-1 cells at the appropriate stages with Opti-MEM was a key element for successful transfection of the cells, HuH7 cells did not require these steps.

154 4.2.63. Day Three- Compound treatment of transfected cells Drug Delivery Protocol. Drug dilutions at 1:1000, were prepared under appropriate conditions, from lOmM drug stocks (in DMSO) in IX Drugging medium. Many of the phytochemicals tested, particularly hyperforin, are sensitive to light, therefore the preparation and dilution of these compounds was conducted in reduced light, and the compound stocks were stored at -20°C, in darkness. The dilutions of experimental compound stocks and vehicle were prepared in the Drugging media. 1:2 serial dilutions were carried out in drugging medium containing 0.1% DMSO, such that the concetrations ranged from lOuM to O.OluM. Rifampicin was included as the positive control compound for all plates. Triplicate plates were used for each experiment. The test compounds were serially diluted across the 96 deep well block (Figure 4.3). After aspirating the transfection mix from the wells, 100 [ll of diluted drug was dispensed to each well. For standards. Drugging medium containing 0.1% DMSO was added to wells A12-D12 as the vehicle control and lOuM rifampicin to wells E12-H12 as the positive control. The plates were incubated for a further 18-24 hours prior to the determination of reporter activity.

4.2.6.4. Day Four- Alkaline phosphatase (SPAP) assay After overnight incubation in the presence of test compounds, the supernatant was sampled and assayed for alkaline phosphatase activity as follows: 50 pi of medium from each well of treated plates was transferred to daughter plates to assay for SPAP as an internal control for transfection efficiency. The transfer was conducted using a biohit multichannel pipettor when first learning the techniques, but later a Platemate or Zymark Rapidplate robot was utilised for this step. Occasionally problems with the robots were encountered with release into the daughter plates, particularly when they were first installed. This affected the reliability of the results, and the data were not used. A 50ul sample of media used for drug dilution was used as a blank. Then p- nitrophenyl phosphate (186mg/100mL) was prepared in SPAP buffer (1 M diethanolamine/0.28 M sodium chloride/0.5 mM magnesium chloride at pH 9.85). 200 pi of the substrate was added to each well of the daughter plates and the blank, the plates incubated at room temperature for approximately 30 minutes until a yellow colour developed, when the absorbance could be read at 405 nM using a UV plate reader.

155 0.125% Triton X-100 and ONPG (1 lO/lOOmL) were dissolved in p-galactosidase buffer to lyse the cells in the mother plates. 200uI7well was dispensed to each mother plate and P-galactosidase blank plate using a hand pipettor or later, a multidrop. The plates were incubated at 37°C for up to an hour until a yellow colour developed. Basal values for the substrate buffers SPAP and B-Gal were determined from the blank plates. All the plates were read at 405 nM in 2 matching files, one for SPAP plates and one for B-Gal plates. The results were considered acceptable when the B-Gal values were consistently 0.4 across the plate and the SPAP positive control rifampicins were 0.8 or above, with the rest of the plate showing a dose response. The quality of the data improved with greater operator experience in cell aspiration, repair of tools (e.g. the biohit multihit pipettor) use of viable plasmids, fresh reagent and drugs, and carrying out the drug dosing in darkened conditions,- this was particularly important for most of the Phytochemicals tested. The experimental procedure and plate layout is summarised in figure 4.7, and figure 4.8 illustrates the PXR transfection assay diagrammatically

156 Figure 4.7. A diagrammatic summary of the hPXR transfection experimental procedure (Adapted from diagrams by Charlotte Ashby)

Day One - Plating /T m Day Two - Transfection

M1 I I I I I I ■ ^+UAStkSPAP Hu uH H m7 ? g I I I I I I I I I I I +pBluescript +B-Gal +SRC-1 +CYP3AtkSPAP +Gal 4 hPXR +Gal4 hPXR +pBluescript Day Three - Drugging +plus reagent A B C D F G H I J K Serial drug dilutions across plate Serial drug dilutions across plate ->->

Serial drug dilutions across plate i n Serial drug dilutions across plate -> -> 5 ^ r Serial drug dilutions across plate ->-> ->-> 1 Serial drug dilutions across plate 1 Serial drug dilutions across plate n i Rif Serial drug dilutions across plate lOuM Rif dose response curve " 1

Day Four - Assay

SPAP should show a dose + SPAP buffer concentration -30 minutes SPAP effect across ■ the plate

B-Gal + B-Gal buffer B-Gal should -45 minutes be consistent m across the plate

157 Inducer

-iüh> cell y

RXFM# P X R %

rxr^ K pxr -•-ffilt" ‘m]T?... Reparler Gene Reporter Plasmid

Figure 4.8. Diagram illustrating an overview of the hPXR Transfection Assay

4.2.7. Data Analysis Measurements were read using a UV plate reader (SpectraMax). Data were copied from the SoftMax Pro files into Excel, and then analysed using ORPH.xla and

RoboSage. ORPH.xla is a data reduction programme developed in house by Mark Lutz (GW RTP). Curves were fitted from these data and visually displayed using Robosage. Data were manipulated as triplicate values and expressed as percent of maximum or fold activation, by the following formulas Where Y = % Max = (fold activation of unknown/positive control fold activation average) x 100 And, fold activation = ((SPAP-SPAP substrate blank average)/B-Gal-B-Gal substrate blank average))-(DMSO vehicle average)/DMSO vehicle average.

And, Y2 = an adjustment factor. Using the robosage package, the equation used is a modified version of Michaelis -

Menton: Y = ((Vmax*x)/(K+x))-Y2

158 The plate layout also included quadruplicate wells for determination of basal activity from the vehicle itself, and quadruplicate wells for determination of the positive standard control, Rifampicin. Where ECso’s for agonists could be plotted (if there was a clear plateau in the dose response curve), the formula was Response = (%max*[S]ECso + [S] ) To correct for transfection efficiencies, SPAP activity was normalised. The normalised SPAP activity was expressed as a percentage of maximum induction by the positive control (rifampicin) and dose-response curves were plotted. By plotting fold activation vs. log [compound concentration] and using non-linear least squares curve fitting, curves with Vmax values (and EC 50 where possible) were generated using Robosage. The data were normalised to 100% rifampicin, and in all experiments the overall trend of the data was unchanged by the normalisation procedure. The results were acceptable when the overall fold activation across all plates was above 4.0 for HuH7 cells (fold activation for CV-1 cells was always above 4.0), and when the Rifampicin curve reached a plateau and had a % Max induction of between 80-120%. This was to ensure results were standardised between operators, with a 20% margin from the standard value being acceptable. The EC 50 value could only be quoted if a clear plateau was observed. In most cases a curve was generated, but not EC50 values, as this involved many more repetitions of the assays with incremental increases in the doses of drugs, and this became impractical given cell toxicity and solubility issues that can occur at higher concentrations, as well as time and financial constraints.

4.3. Compounds tested The full range of compounds tested in both cell lines can be found in Appendix 1. Initially, compounds were selected to include both a broad range of likely dietary ligands, drugs and experimental compounds, and drugs of known activation, to learn the operator procedure in the laboratory. Selection was based upon structural features (See selected structures in Table 4.4 and figures in Chapters 1, 3 and 5) and a literature review (Chapter 1 and Chapter 3, table 3.1). Proprietary GSK compounds were also tested, to explore the binding properties of specific structural series, such as those seen with SR compounds. However, while PXR is of strong pharmaceutical interest, because it has evolved to deal

159 principally with phytochemicals and other dietary compounds, testing with dietary compounds is a reasonable approach to examine hPXR activation in humans. To develop a better picture of cross talk, many of the compounds selected were ligands to differing degrees with associated receptors, particularly the ER (e.g. endogenous hormones and synthetic hormones: progesterone, pregnenolone 16a carbonitrile (PCN), 5(3pregnane

3,20 dione, 17a ethinyl oestradiol, 17(3 oestradiol, experimental/drug hormones: ICI 182780, tamoxifen and its metabolite, hydroxylated tamoxifen, SRI 2813 and in house GSK analogues, clotrimazole and dexamethasone, phytoestrogens: flavonoids, genistein, daidzein, coumestrol, quercetin and resveratrol; parent lignan: secoisolariciresinol and the metabolite enterolactone, equol, a daidzein product; organochlorines: PCB 118, PCB153, trans-nonachlor), the VDR (vitamin D 3 and the metabolite hydroxy vitamin D 3 ), LXR (the bile acid lithocholic acid), PPAR ( rosiglitazone and troglitazone). Other compounds were selected on the basis of possible self-medication, such as hyperforin, a known PXR inducer (Moore et al., 2000b), herbal hepatoprotectants such as silybin, also a selective ER(3 agonist (Seidlova-Wuttke et al., 2002). Many other compounds originating from the diet have been shown to be induce CYP 3A4 and/or SHR ligands, so the test set is limited, and does not present a spectrum of dietary ligand activity for PXR. Additional dietary ligands of interest include bergamottin in grapefruit juice, which inhibits CYP3A4 induction, (He et al., 1998; loannides, 2002; Plant and Gibson, 2003), glycyrrhizin, found in licorice, which is also a ligand for GR (Zhou et al., 2003; Paolini et al., 1999), kava kava (Raucy, 2003), vitamin E (a-tocopherol) and also dietary borne POPs not included here that are also of health concerns such as dioxins, DDT, DDE and PBDEs (Chapter 2). This is a realistic approach to examining drug- chemical interactions for drug development and, in total; this information can add a useful assessment component to a comprehensive risk assessment, for suspected EDCs (see Chapter 1). Table 4.4 gives the selected chemical structures of the following classes of ligands, which were of particular interest:

160 Table 4.4 Selected chemical structures of compounds tested in the PXR transient transfection assay and included in the PXR QSAR, (Chapter 5.4)

Endogenous hormones Drugs Phytochemicals POPs and additional and experimental phytochemicals hormones

OH OH PCB118 5Ppregnane3,20-dione Genistein (2,3',4,4'pentachloro hiphenyl) Positive control: Rifampicin HC OH PCB153 .OH (2,2'4,4',53'hexachloro H,C hiphenyl) Cl Cl H,C- HO I HO SR12813 Daidzein 17a ethinyl oestradiol OH Bacteriocide: Triclosan .OH

CH, OH HO' OH Clotrimazole OH Progesterone Quercetin

Bile acid: Pesticide: Trans- Lithocholic acid nonachlor

Rosiglitazone Silyhin

Vitamin D3 OH Lignan: HO. Secoisolariciresinol

CH, Dexamethasone Hyperforin

HO.

OH Tamoxifen Coumestrol Lignan metabolite: Enterolactone

ICI 182780 OH Daidzein metabolite: HO' Equol HO' HO' 'OH HO. trans-resveratrol

‘OH

161 4.3.1 Flavonoids Flavonoids are a diverse group of phytochemicals, sharing a common structural diphenylpyran skeleton, with two benzene rings linked through a heterocyclic pyran or pyrone ring (Hertog and Hollman, 1996). They are produced by various plants, including medicinal herbs, legumes, seeds and grains. They generally have high biological activity, and are considered to be beneficial, due to their antioxidant properties and ability to modulate several enzyme and receptor pathways. In some cases, however certain flavonoids have been shown to have mutagenic activity or interfere with essential biochemical pathways (as reviewed in Zhou et al., 2003), particularly if consumed in very large quantities. The isoflavones genistein and daizein, coumestrol are reviewed in chapter 1, so are not included here. Quercetin, is a flavonoid that is very common in onions (28.4-48.6 mg/lOOg fresh weight), and widely available in the diet from many other vegetables and fruits but to a lesser degree, e.g. apples (2.1-7.2 mg/lOOg fresh weight), red wine (0.4-1.6 mg/lOOg fresh weight), lettuce (0.3-3.0 mg/lOOg fresh weight) and black tea (1.7-2.5 mg/lOOg fresh weight), (Hertog and Hollman, 1996). Quercetin is believed to have a potent inhibitory effect upon estrogen sulphotransferase (EST), which determines free oestrogen levels in normal breast epithelial cells (Otake et al., 2000). It was found to suppress COX-2 promoter activity in colon cancer cells (Mutoh et al., 2000), and did not induce CYPlAl in Hep G2 cells (Allen et al., 2001). Equol, is an endogenous metabolite of daidzein, produced by around 36% of human subjects in a study reported by Rowlands et al., 2000 (in Pietinen et al., 2001), and is known to be a potent phytoestrogen activating the ER.

4.3.2. Herbs Hyperforin, the active ingredient of St Johns Wort {Hypericum perforatum), is one of the most commonly used self medicated herbal medicines for mild depression in the US and Europe (Diirr et al., 2000; Moore 2000b). However, because hyperforin is a potent inducer of CYP 3A4 and various other CYP enzymes, and the efflux transporter P-gp, there has been concern about the interactions with clinically important drugs. Serious and even fatal drug interactions have been reported. Silymarin, the active ingredient of Milk Thistle (Silybum marianum), is a flavonoid blend believed to have potent free radical scavenging activity, with prevention of glutathione oxidation and depletion, together with membrane stabilizing effects and

162 inhibition of arachidonic acid (AA) metabolism (Saliou et al., 1998). The main constituent of silymarin, the isomer silybinin, has been reported to inhibit the formation of leukotrienes via the 5-lipoxygenase pathway at clinical doses and so have cytoprotective qualities. Leukotrienes have deleterious effects by promoting inflammatory reactions (Dehmlow et al., 1996). Other effects include increased protein synthesis by activation of mRNA polymerase and the selective inhibition of NF- kB, a key regulator in inflammatory and immune reactions (Saliou et al., 1998). Further, it is reported to be an antiproliferative agent (Scambia et al., 1996), and be a potent skin cancer preventive agent (Zi and Agarwal, 1999).

4.3.3. Lignans and polyphenols The plant lignans matairesinol, secoisolariciresinol, pinoresinol and lariciresinol are present in foods such as rye, tea, linseed (flax), also a rich source of a-linolenic acid (see chapter 2, Tou and Thompson, 1999; Liggins et al., 2000; Xia et al., 2001), vegetables and berries. They are metabolized to enterolactone (Heinonen et al., 2001) and enterodiol, by intestinal bacteria, and matairesinol is the precursor to an important antiviral and anticancer agent, podophillotoxin (Xia et al., 2001). Enterolactone is the colonically fermented metabolite of lignans such as secoisolariciresinol and maitaresinol (Adlercreutz and Mazur, 1997; Heinonen et al., 2001; Adlercreutz, 2003) and is clinically important as high circulating serum and urinary enterolactone levels are associated with a significantly reduced risk of breast cancer in Finland (Pietinen et al. 2001) and a reduction in prostate cancer (Xia et al., 2001), Intestinal microflora patterns differ regionally, depending upon the dietary intake, and gut exposure in utero and whilst being breast fed. Resveratrol (3,5,4’-trihydroxystilbene) is a polyphenol manufactured by grapevines in response to pest attack (Jeandet et al., 1995) (much the same as glyceollin production in stressed soya; Jacobs 1998). It is of medicinal interest particularly because it appears to be a major component of the ‘French Paradox’ where moderate red wine consumption has cardioprotective activities, and because it has chemopreventative potential via cell signaling (Hsieh and Wu, 1999; Nielsen et al., 2000). It is also reported to be a strong ER agonist in ER transfected MCF-7 cells, and a weak and partial ER agonist in a rat uterotrophic assay and a yeast hERa transcription assay (Ashby et al., 1999a). It has also been reported to have ER antagonist activity in the presence of oestrogen in vitro, in MCF-7 cells (Damianaki et al., 2000; Basly et al., 2000), and in vivo

163 (Henry and Witt, 2002), and has been observed to modulate constitutive CYP IBl in MCF-7 cells (Chang et al., 2000) and CYPlAl in Hep G2 cells (Allen et al., 2001), and suppress COX-2 promoter activity in colon cancer cells as part of a chemopreventive action (Mutoh et al,, 2000). As with other flavonoids, resveratrol also modulates the AA metabolic cascade and is reported to affect prostaglandin synthesis (Moreno, 2003). The active metabolite is probably a glucuronide conjugate (Kuhnle et al., 2000), and in further experimentation, it would be interesting and physiologically relevant to use this form as a test compound, as metabolites frequently behave differently to the parent compounds (Nolan et al., 2003).

4.3.4 Persistent Organic Pollutants (POPs) Polychlorinated biphenyls (PCB 118 and PCB 153) The persistent organic pollutant nature of PCBs has already been introduced in previous chapters. PCBs 118 (2,3’,4,4’,5-Pentachlorobiphenyl) and PCB 153 (2,2’,4,4’,5,5’- Hexachlorobiphenyl) were selected as test compounds, because they consistently feature highly in food contamination studies (see Chapter 2), whilst also being detected in human body compartments such as maternal serum and fetal cord blood (Covaci et al., 2002b), and liver and adipose tissue for example (lida, Hirakawa, Matsueda, Nagayama, and Nagata, 1999; Maruyama et al., 2003; Covaci, 2002) at the greatest magnitudes, compared to many other PCBs and PCB metabolites (Chu et al., 2003). They are both known to be weakly oestrogenic in hydroxylated forms, and certain hydroxylated PCBs have recently been shown to mimic binding in the crystal structure of human oestrogen sulphotransferase (hEST) (Shevtsov et al., 2003). PCB 118 is a non-ortho PCB, PCB 153 is phénobarbital like (ten Tusscher and Koppe, 2003), and one of the ICEs 7 marker PCBs. Trans nonachlor is an organo-chlorine pesticide that has been previously shown to induce PXR (Lehmann et al., 1998), and was therefore included in the test set for both structural information for the QSAR and as a QA measure, to ascertain that the data produced was also reproduced by another laboratory. Triclosan ( also known under the registered trademark of Trgasan DP 300’) is an extensively used broad spectrum antimicrobial (Hanioka et al., 1996; Heath et al., 1999; Latch et al., 2003) present in cosmetics, soaps, shampoos and cleaning materials, food packaging materials, oral healthcare products such as toothpastes as well as shoes, carpets and fabrics, for example. It acts by inhibiting bacterial fatty acid synthesis (Heath et al.

164 1999). Triclosan is not banned or restricted in use as with the other OCs in the test sets, as the traditional toxicity tests indicate a low toxic effect, with an acute oral LD 50 of approximately 4000mg/kg in rat and mouse (Lyman and Furia, 1969; Jinno et al., 1997), and it is considered safe (Bhargava and Leonard, 1996; Jones et al., 2000). However, very little is known about enzyme and receptor induction by triclosan. In an in vivo rat study, triclosan has been shown to induce CYP2B1/2 together with a slight increase in CYP3A, suggesting that triclosan is a Phenobarbital-type inducer. Whilst its chemical structure closely resembles known non-steroidal oestrogens (e.g. DES, bisphenol A) and anti-oestrogens (e.g. TCDD), an in vivo fish study has suggested that triclosan is potentially weakly androgenic (Foran et al., 2000). Triclosan is being increasingly viewed as an environmental pollutant. It is frequently detected as a contaminant of rivers and lakes, and appears to be responsible for producing a dioxin compound ( 2,8 -dichlorodibenzo-p-dioxin) by an intramolecular photochemical substitution reaction (Latch et al., 2003), and a methylated form (Lindstrom et al., 2002; Erikson, 2002).

4.3.5. Essential Nutrients: Vitamins Vitamins that are hormonal in structure are known to be SHR ligands, for instance vitamin E is weakly oestrogenic, and has been reported to be a PXR ligand (Landes et al., 2003), Retinioids, such as Vitamin A, are the natural ligands for RAR and

RXR and Vitamin D 3 , particularly the hydroxylated form, is the endogenous ligand for the closely related VDR. la, 25-dihydroxy vitamin D 3 (OH Vit D3 ) is the hydroxylated form that Vitamin D 3 is metabolized to in vivo. OH Vit D3 has been shown to induce CYP3A4 in human intestinal Caco-2 cells, but not via PXR, as Caco-2 cells reportedly do not express PXR (Thummel et al., 2001). OH Vit D 3 has also been reported to induce P-gP and NADPH cytochrome P450 reductase (Schmiedlin-Ren et al., 1997), and to have antiproliferative activity in the same cell line (Scaglione-Sewell et al., 2000). Taken together, it is suggested by Thummel et al (2001) that OH Vit D 3 and VDR induce intestinal CYP3A expression by binding of the activated VDR-RXR heterodimer to the CYP3A PXR response element and promoting gene transcription. This hypothetical model may not be appropriate for liver CYP3A induction and PXR expression, as investigated by the hPXR transient transfection assay in HuH7 and CV-1 cells.

165 4.4. Results

Data for a diverse set of over 33 compounds and metabolites, many selected on the basis of likely dietary exposure, together with compounds of pharmacological interest, to ensure a broad spectrum of activity, were derived experimentally. Raw data are shown in Appendix 2, the selected structures of the test compounds are shown in table 4.4. Many of the compounds were first tested in CV-1 cells, both for training purposes and for comparison with HuH-7 cells. On the whole, both HuH7 and CV-1 cells showed similar results with the known inducers, however HuH7 cells were ultimately the cell line of choice, and only the data generated by using HuH7 cells were used for QSAR purposes, for two reasons. First, the HuH7 cells are human (while the CV-1 cell line originates from a monkey kidney cell line), they contained the full length PXR and may contain endogenous human transporters (i.e. P-gp), so are more biologically relevant to humans than CV-1 cells. Second, a substantial amount of in-house data on % max induction for hPXR was available for the HuH7 cells, but this was not the case for CV- I’s. These data bases were useful both for comparative purposes and for expanding the model, testing it for predictive qualities. Inducers were considered in three potency groups, low (<30%), medium (30- 70%) and high (>70%), see table 4.1. A good distribution of induction data was necessary for building an adaptive, robust model of PXR induction (Chapter 5). When testing a compound, only the induction of CYP3A via PXR is reported from this assay. Further induction studies would be required to eliminate other pathways such as CAR inducing CYP2B, AhR inducing CYPIA, cross-talk with these receptors and others such as HNF- 4, COUP-TF, GR, LXR, and FXR, for example.

4.4.1. The test compounds The test compounds included drugs with potent activity such as rifampicin, the positive control, (figure 4.9), experimental antioestrogens such as ICI 182780 and SR 12813, (figure 4.10), and phytochemicals with high activity such as hyperforin, (figure 4.11). Phytochemicals with low activity (trans resveratrol, quercetin and silybin) are shown in figures 4.12, 4.13, 4.14, plant lignans (secoisolariciresinol), and mammalian lignan metabolites (enterolactone>equol) with low to medium activity are shown in figure 4.15. Phytoestrogens compared with the endogenous oestrogen, 17(3 oestradiol.

166 with low to medium activity (coumestrol>daidzein>genistein) are shown in figure 4.16. Organochlorine contaminants including a pesticide, a pollutant and a bacteriocide, (trans nonachlor

167 Figure 4.9 shows standard curves for the positive control rifampicin in both a ) CV-1 and b) HuH7 cell lines

CVlPXRtkSPAP HUH7PXRtkSPAP (a) (b) rifam picin rifampicin -2.61 116.49 113.88 7.9 98.8 106.7

150 -I 150 T ■= 100 : I s 50 - 50 -

■f.Q)OE-Ofr 1.00E-07 1.00E-06 1.O0E-I 1 .OOE-08 1 .OOE-07 1 OOE-06 1 OOE-05 concentration [M] concentratlon[M] mni compound set D 18 july 2002 ______mnj IB 31.10.02

Figure 4.10 shows representative curves for potent experimental antioestrogens tested at a top concentration of lOuM

HUH7PXRtkSPAP HUH7PXRtkSPAP ICI 182780 S R 12813 -2.57 125.91 123.34 7.04 159.61 166.65

1 150 q 250 5 100 j 200 .£ Q 50 : 150

E -Q 50 -50 :- 1.00E- 1.00E- 1 OOE- -50 08 07 06 05 1.00E-08 1.00E-07 1.00E-06 1.00E-05 concentration [M] concentration [M] compound set A MNJ 12 July 02 mnj 12 July 2002

168 Figure 4.11 shows representative curves for hyperforin at top concentrations of a) luM, b) 3uM and c) 5uM in CV-1 cells. Hyperforin is cytotoxic at lOuM concentrations (NB. As the top dose here is not lOuM, the x axis giving the concentration [Ml is incorrect and should instead be interpreted as the respective serial dilutions of the top dose as given in the respective titles.)

CVlPXRtkSPAP CVlPXRtkSPAP CVlPXRtkSPAP hyperforin 1uM (b) hyperforin 3uM (c) hyperforin 5uM 6.27138.99 145.26 -9.54 167.12 167.68 -2.46 149.79 147.33

200 150 100

-50

1.00E-08 1.00E-07 1.00E-06 1.00E-05 1.00E-08 1.ODE-07 1.00E-06 1.00E-05

, concentration concentration [M] mnj compound set B 18July 200: mnj compound set mnj compound set B 18July 2002 Figure 4.12 shows representative curves for trans resveratrol at top concentrations of a)10uM, b) 20uM in CV-1 cells (NB. As for fig 4,11, the top dose here is not lOuM, the x axis giving the concentration [M] is incorrect and should instead be interpreted as the respective serial dilutions of the top dose given in the title).

CVlPXRtkSPAP CVlPXRtkSPAP (a) (b) trans resveratrol trans resveratrol 20uM -9.81 18.52 8.71 .5.14 13.72 8.58

100 100 Ô c -oS 50 50 11 0 IÎ ■ ...... I IÎ -5000E— 1-ÆQE-— ItOGE— ItOOE- -SÙOQE— ItOQE----liQOB 1-.OOE- 08 07 06 05 08 07 06 05 concentration [M] concentration [M] mnj compouod.set D 18 july 200 12002.

Figure 4.13 shows representative curves for the phytochemical silybin at a top concentration of lOuiVI, in a) CV-1 and b) HuH7 cells

CVlPXRtkSPAP HUHTPXRtkSPAP (a) silybin (b) siiybin -10.78 23.4 12.62 -7.26 -1.85 -9.11

c 100 100 80 60 40 0 IE 20 !i 0 -20 1.00E- 1.00E- i.bOE- 1.00E- 1.00E- 1.00E- 1.00E- 08 07 06 05 08 07 06 05 concentration [M] concentration [M] _ranjjaimpQuacLsfit_B_iaJuiy_2002 _____ mnj18.10.02B

169 Figure 4.14 shows representative curves for phytochemicals and metabolites at top concentrations of lOuM

HUHZPXRtkSPAP HUH7PXRtkSPAP secoisolariciresinol enterolactone -0.46 21.03 20.57

p ' g ■is 60 a s 0 i|l ” E.a 20 1.00E- 1.00E- 1.00E- 1.00E- 00E-- 1T00E- r.OOE- 08 07 06 05 08 07 06 05 concentration [M] concentration [M] ftniiaiQ.OPA- mnj18.10.Q2A HUH7PXRtkSPAP HUHTPXRtkSPAP quercetin equol 14.02 12.48 26.5

100 -

60 1 40 ° 20 0 « E«-C-C h 1.00E- 1.00E- 1.00E- 1.00E- 08 07 06 05 -SP.OOE- 1.00E- EOOE 1.00E 08 07 06 05 concentration [M] mnj18.10.02B concentration [M] mnj18.10.02A

The plant lignans (secoisolariciresinol), and mammalian lignan metabolites (enterolactone>equol) with medium to low activity.

Figure 4.15 shows representative curves for the phytoestrogens tested, compared to the endogenous oestrogen 1?P oestradiol all tested at top concentrations of lOuM Medium to low activity (coumestrol>daidzein>genistein)

HUHTPXRtkSPAP CV-1 hum an chimera coumestrol 17 beta oestradiol -9.72 28.96 19.24 -0.47 51.29 50.82 .2.,^ 100 80 60 40 20 0 i -20 1.00E- 1.00E- 1.00E- 1.00E- 08 07 06 05 -^QE-08 1.00E:0Z 1.00E-06 1.00E-05 concentration [M] concentration [M] mnj18.10.02B mni B 15.11.01

HUH7PXRtkSPAP HUH7PXRtkSPAP daidzein g e n is te in -4 .2 5 5 .7 8 1.53

?! -m-i ShObE l.OQE- l.OOE- 1.ÛDE- = -08 1.00E-07 1.00E-06 1 .GOE-06 08 07 06 05 If concentration [M] concentration [M] mni18.10.02A

170 Figure 4.16 shows representative curves for the organochlorines PCB 118, trans nonachlor and triclosan at top concentrations of lOuM, in HuH7 cells Medium to low activators at lOuM top dose (trans nonachlor

HUHTPXRtkSPAP HUHTPXRtkSPAP HUHTPXRtkSPAP trans-nonachlor PCB118 triclosan 13.88 44.07 57.95 -4.6 20.43 15.83 8.26 41.75 50.01

.2 _ 100

£ “ 3 40 20 OU i. n l l iri -20 1.00E- 1.00E .OOE l.OOE 08 1.00E-07 1.00E-06 1.00E-05 1.00E-08 1.00E-07 1.00E-06 l.OOE-05 concentration [M] concentration [M] concentration [M] mnj18.10.02B mnj18.10.02A mnj18.10.02B

Figure 4.17 shows representative incremental increases in the curves for triclosan at top concentrations of a) 10 and b) 20uM, in HuH7 cells and c) 30uM in CV-1 cells

HUHTPXRtkSPAP HUHTPXRtkSPAP CVlPXRtkSPAP (^) triclosan (b) triclosan 20uM triclosan 30uM - 5.69 70.63 64.94 -4.24 36.47 32.23 -10.95 97.16 86.21 100 80 g 3= 100

-50.d6E- 1:G0E- IrWE-'^EOOE 1.00E l.OOE 1.00E- 1.00E- 1 OOE 08 1.00E 07 l .OOE 06 1 OOE-05 08 07 06 05 06 concentration [M] concentration [M] compound set A MNJ 12 July 02 mnj 12 July 2002 concentration [M] mnj compound set D 18 julv 2002 (NB. As for fig 4.11 and 4.12, where the top dose is not lOuM, the x axis giving the concentration [M] is incorrect and should instead be interpreted as the respective serial dilutions of the top dose.)

Figure 4.18 shows representative curves for hydroxylated compounds compared to their unhydroxylated parent compounds

CVlPXRtkSPAP CVI PXRtkSPAP OH vitamin D3 vitaminDS -8.55 139.92 131.37 -8.04 26.69 18.65

ll™ 40 20 0 f 1.00E- 1.00E- 1.00E- 20 07 06 05 -4.00t-08 1 OOE 07 1.00E-06 1 OOE-05 concentration [M] concentration [M] mnj com pound set B 18July 2002 mnj compnund.set B Iftluly 2002------

HUHTPXRtkSPAP HUHTPXRtkSPAP tamoxifen 40H tamoxifen -4.14 52.9 48.76 10.45 77.53 87.98

0 U 11 20 1 OOE- 1.00E l.OOE 1.00E 1 OOE l.OOE l.OOE 1 OOE 08 07 06 05 08 07 06 05 concentration [M] concentration [M] compound set A MNJ 12 July 02______171 mnji 8.10.028 4.5. Discussion Several points of interest arose from these studies, as follows;

4.5.1 Lignan potency Lignans present in the human diets can be metabolised to weak-moderate affinity ligands for PXR as seen with the secoisolariciresinol metabolite enterolactone, as shown in figure 4,14. Common food sources such as rye, whole grains, flaxseed (Tou and Thompson, 1999) and berries provide the parent lignans from which enterolactone is fermented by colonic bacteria. Populations who consume large amounts of rye, such as Scandinavian and Eastern Europeans, may be stimulating moderate induction of PXR, in addition to the chemoprotective effects documented in the literature (Tou and Thompson, 1999; Pool-Zobel et al., 2000).

4.5.2 Potency of OH compounds: tamoxifen, vitamin D Hydroxylated Vitamin D 3 was significantly more potent than non-hydroxylated vitamin D 3 . (Figure 4.18). This has not been observed for PXR activation in the human colon adenocarcinoma cell line (CaCo-2) (Schmiedlin-Ren et al., 2001), but has been observed in other human cell lines, and as suggested by the results of the AhR QSAR in chapter 5, CaCo-2 cells may not be as appropriate models for ligand activation or ligand binding studies as Human liver cell lines such as HuH7 and Hep G2. A trend was also observed for the hydroxylated versus non hydroxylated tamoxifen. (See Figure 4.18). The fact that hydroxylated compounds tested were more potent than the non hydroxylated parent compounds implies that this is likely for other hydroxylated compounds, and this is explored further in chapter 5. The implications that hydroxylated forms are more potent than the parent compounds could be significant when considering synthetic endocrine disrupting chemicals in food. For instance, the PCBs delivered in the diet are metabolized to hydroxy forms which are also known to be more potent in ER screening assays than parent PCBs.

4.5.3. Persistent organic pollutants Polychlorinated biphenyls 118 and 153 are weak ligands, (Figure 4.12). However when metabolized into the hydroxylated forms, they are likely to have greater affinity.

172 The pesticide trans-nonachlor is a medium affinity ligand, (Figure 4.16), as is triclosan, (Figure 4.17), a commonly used bactericide.

4.5.4. Differential effects of increasing dosage between compounds The following compounds were tested above the lOuM top dose, and showed a subsequent increase in% max induction: triclosan (figure 4.17, top dose range 10-30uM), trans-resveratrol (Fig. 4.12, top dose range 10-20uM), trans nonachlor (figure 4.16, top dose range 10-20uM). This was not the case with other compounds, such as hyperforin which was tested at different ranges before reaching cytotoxicity (figure 4.11), and similar induction profiles were observed, not incremental increases. The effect of the host cell is the variability in this assay, and this may help explain why, for many compounds, increasing the dosage did not affect the % max induction. Taking hyperforin as an example, the % max induction remained similar at luM, 3uM and 5uM concentrations and there does not appear to be an increase in potency with increase in dosage below that which is cytotoxic (lOuM), (figure 4.11). This is a pattern that was seen with other phytochemicals and drugs, both in this study and other GSK studies, both at RTP and Ware (unpublished data, Gail Nolan, Linda Moore). The explanation for this effect is not known, but similar effects have been observed in other receptor assays. In one report using the Yeast Oestrogen Screen (YES), a yeast reporter gene assay which assesses oestrogenic activity via the human ERa (Andersen et al., 1999; Miller et al., 2001), the reduced aqueous solubility of certain compounds (in this instance o,p’-DDT) was found to be a limiting factor for the compound to get into the cell from the plate (Rajapakse et al., 2001), which when not controlled for, may give misleading results. Using DMSO to dissolve the compound, as for the studies presented here, overcame this effect. It is unlikely that lack of dissolution of the test compounds in the vehicle (DMSO) is responsible for the effects seen with varied doses of hyperforin. It is more likely that that the involvement of relevant transporters and cofactors (such as other key transporters) play a role, so the transporters for specific compounds may become saturated, preventing further PXR activation, as the compounds can no longer reach the receptor. One could speculate that this is the case for the test compounds hyperforin, trans resveratrol, trans nonachlor, but not for triclosan, where PXR activation was observed in a concentration-dependent manner. That transporter proteins, such as P-gp, MDR, OATP play significant roles in PXR activation is well documented in the literature and briefly reviewed below.

173 4.5.5. The role of transporters and other cofactors There are many types of cofactors that have a role to play in removing xenobiotics from cells. These are cell surface transporters, transcriptions factors and other ancilliary proteins. The following examples of transporters are of therapeutic and metabolic interest, with significant and variable effects upon xenobiotic and drug metabolism.

4.5.5.I. Transporters: P-Glycoprotein (P-gp) Discovered over twenty five years ago, P-gp is a surface phosphoglycoprotein encoded by the human multi-drug resistant genes (MDRl and MDR3). Regulated with CYP3A4, by PXR, it is an ATP-binding cassette (ABC) energy dependent membrane bound transporter involved in the extracellular efflux of many endogenous and xenobiotic compounds (and phospholipids), so functioning as a biological barrier (Kim, 2002; Lin and Yamazaki, 2003). It is found on the luminal membrane of the small intestine and blood brain barrier, in the apical membranes of excretory organs such as liver and kidney and in the brush border membrane of placental trophoblasts that form the matemo-foetal barrier. As a consequence of this expression pattern, P-gp plays a key role in both limiting intestinal absorption and CNS penetration, during initial digestive exposure to xenobiotics and excreting drugs and other xenobiotics for elimination, as well as preventing them from being reabsorbed from the intestine following bile secretion into the intestine for excretion. P-gp can be modulated by genetic polymorphism (Xie et al., 2001; Fischer, 2002; Lin and Yamazaki, 2003), gene induction, or drug-drug interactions, with consequent effects upon the pharmacokinetics and target organ specificity of P-gp substrates. P-gp is also associated with the development of inflammatory bowel disease (Fischer, 2002). The sister of P-gp (SP-gp) is the major hepatic bile salt export pump (BSEP), and bile acid ligands have been shown to induce BSEP in human hepatocyte cultures, in parallel with the bile acid /famysyl receptor FXR (Schuetz et al., 2001), and MRP2 is similarly up regulated by FXR and PXR (Takikawa, 2002). As many drug interactions involve CYP3A4 and P-gp, the relative contribution of P-gp and CYP3A4 to overall drug interactions needs to be quantitatively estimated, to understand the underlying mechanism of drug interactions, or when using in vitro PXR/CYP3A4 screening assays. It is not as yet possible to do this with current PXR

174 transient transfection assays, and consequently this is an ongoing area of pharmaceutical research. P-gp may therefore be a confounder of the PXR transient transfection assay assessment/screening procedure for many compounds. Multi-drug resistance associated protein (MRPl and MRP2) are also excretory proteins. Like P-gp, they are members of the ATP transporter superfamily. MRPl is a transporter of organic anions in normal cells, and confers resistance to anti cancer agents in tumor cells (Leslie et al., 2003). It is upregulated by a PXR ligand, dexamethasone, in rat hepatocytes (Courtois et al., 1999). MRP2 is a drug efflux pump that mediates the transport of several conjugated compounds across the apical membrane of the hepatocyte into the bile caniculi (Takikawa, 2002), and is governed by three receptors: PXR, CAR and FXR, as part of its multivariate signaling pathway (Takikawa, 2002). All three receptors converge upon a common response element in the 5’-flanking region of this gene (Kast et al., 2002). Many flavonoids induce MRPl (Leslie et al., 2003), and it is thus possible that this transporter may have been induced by the selected flavonoids run in the PXR HuH7 host transient transfection assay. Organic anion transporting protein (CAT?) belong to another large family of transport proteins. They are principally involved in the hepatocellular uptake of endogenous compounds as well as a variety of structurally divergent drugs (Kim, 2002), but can be found in brain (OATP-A, OATP-1,2), liver (OATP-B. OATP-C, OATP-8, OATP-1,2,3,4), kidney (OATP-B, OATP-1), small intestine (OATP-3), lung (OATP-B). They are capable of mediating rifampicin transport into the hepatocyte, and may modulate PXR function (Tirona et al., 2003), indeed they have many substrates in common with PXR ligands, and other NRs such as bile acids, thyroid hormones, ostradiol, eicosanoids and estrone-3-sulphate (Kim, 2002).

4.S.5.2. Receptor complements in cell lines TaqMan® studies conducted at GSK by K Swales for HepG2 cells (Swales, 2002) and C Ashby for CV-1 cells (Ashby, 2002) have characterized major receptors involved in induction for both cell lines, identifying the relative receptor abundance in cells, differences between the cell lines, as well as the potential cross talk facilitation and modulation. For the human liver derived cell line, HepG2, characterization of the receptor compliment has been identified by measuring mRNA levels using human probes and primers in TaqMan® analysis. Swales reports that TaqMan® data suggest that in human GYP 3A regulation, the relative abundance of PXR heterodimerization partners

175 were more important than the absolute level of PXR. In terms of absolute mRNA changes, the ratios of RXRa, RXRp and COUP-TF were the major factors influencing variability in response to xenobiotics. For the primate CV-1 cell line, the receptors shown to increase after transfection were; hPXR, CAR, GR, RXRa, RXRp, COUP-TF and P-Actin, and to decrease after transfection were; RXRy, HNF-4, and the housekeeping gene GADPH (involved in glycolysis, suggesting the cells were down regulating the glycolysis cycle). In the presence of test compounds, the receptor compliments altered in different ways in the transfected cells. Dexamethasone treatment stimulates some co-regulation between CAR and RXRa and RXRp, with PCN coordinated regulation was seen for CAR and hPXR, and PB induced COUP-TF. Ashby also reports a competition effect between COUP-TF and hPXR with rifampicin treatment (the positive control). The CYP3A and associated proteins phenotype of HuH7 cells, in comparison with human adult and fetal livers, has been reported recently (Phillips et al., 2003). The key receptor variables with respect to regulatory pathways were the ratios of COUP-TFl and RXRa to PXR. As with the other cell lines discussed above, CAR is probably also key in the receptor cross talk in response to xenobiotics for this cell line also, a strong correlation between CAR and PXR (and CYP2B6) expression has been reported for human liver (Chang et al., 2003). Transiently transfected HepG2 cells have been compared with human hepatocytes, with similar inductive properties of CYP3A4 mRNA observed for prototypical drug inducers, lOuM rifampicin (699±307% of control levels), lOOuM phenytoin (707 ±188% of control levels), ImM PB (536±207 % of control levels) and lOOuM omeprazole ( 404±8% of control levels), and the phytochemical kava kava

(386±185 % of control levels) but not flavonoids sueh as quercetin, resveratrol and curcumin, or phytochemieals such as grapeseed extract, ginseng and silymarin. Quercetin was observed to produce some CYP3A4 mRNA accumulation, but in the HepG2 reporter gene assay, this flavonoid produced neglible to low CYP3A4, here suggesting that CYP3A4 is induced by mechanisms that may not involve PXR (Raucy, 2003), unlike the prototypical inducers and kava kava.. The low degree of activation reported by Raucy, 2003 supports the low activation of PXR data by lOuM quercetin reported here, fig. 4.10, although it is reasonable that many phytochemieals will induce PXR, considering the evolutionary long term exposure of the intestine and liver to phytochemieals via the diet.

176 Another important consideration to consider physiologically is the cyclical variability of CYP3A4 and CYP3A7 in women, probably reflecting the physiological requirement to eliminate the localized increased endogenous steroid exposure (Sarkar et al., 2003), activating NRs such as PXR to induce CYP3A, as well as the ERs and PR, as part of the menstrual cycle.

4.5.S.3. Example of other factors, Albumin The main serum protein, albumin, binds a wide variety of lipophilic chemicals, but with low affinity. In a high capacity but low affinity binding capacity, it maintains osmotic homeostasis. Ligands include steroids, phytochemieals and xenobiotics that bind to the NRs. As a major carrier, albumin is a regulator of the access of these compounds to receptors. Albumin also functions as a sink for xenobiotics by reducing the binding of xenobiotics to NRs and other proteins (Baker, 2002).

4.6. Conclusions The data presented here extend the range of published PXR ligands to include phytochemieals and phytochemical metabolites such as lignans, PCBs, and the common bactericide triclosan. With respect to drug metabolism and pharmacological implications for drug-drug interactions, in addition to the interactions already reported to be of concern as, for example, seen with hyperforin, the contraceptive pill, idinavir and cylosporin (Moore et al., 2000a), for the first time to my knowledge, triclosan is also of similar concern. Particularly as it is used so widely with a high likelihood of oral ingestion from food contact materials and toothpaste directly and cleaning materials indirectly. It has been reported recently that Vitamin E activates gene expression via PXR at the same concentrations as rifampicin, and the authors suggest clinical trials are needed to identify the dosages necessary for analogous effects to hyperforin to be observed (Landes et al., 2003). A further consideration, the hydroxylated forms of vitamins and drugs are likely to be more potent than the parent compound, and so these should be assessed in future PXR screening, in addition to the parent compounds. This also applies to PCBs, the hydroxylated forms of which are known to be more potent ER ligands (for the oestrogenic PCBs). Finally it must also be remembered that using a single receptor test in vitro system does not necessarily reflect the in vivo situation, and should not be taken as such.

177 Large discrepancies noted for oestrogenicity in ER assays have been discussed in the literature (Ashby, 1999a; Miller et al., 2001). However, this does not appear to be the case for the PXR transient transfection assay, which has been found to correlate well with CYP3A4 expression in human hepatocytes (Luo et al., 2002). Cell lines for assays need to be closely evaluated with each other, to obtain the cell line that and assay that is the closest approximation to the human in vivo state. Multivariate tools can be used to discriminate between the cell lines and systems, for ascertaining the most reliable and complementary method to assess the PXR induction potential of drugs and other xenobiotics, for example the QSAR presented in chapter 5.4. From a clinical perspective, the data also illustrate the value of testing plant chemicals present in the diets of different populations, as diets rich in lignans, as with Scandinavian diets, or diets rich in phytoestrogens, as with Asian diets, may induce greater intestinal levels of PXR compared with UK diets. This also applies to contaminants, high consumers of Baltic herring, for example with have concomitantly higher levels of PCBs which as well as a broad range of health implications, may also increase PXR levels and CYP3A4 induetion to promote xenobiotic metabolism. This is likely therefore for other dietary contaminants such as organochlorine pesticides (and PCBs), as well as contact contaminants such as the widely used bacteriocide triclosan. With the latter in mind, these results suggest it may be possible to identify and develop safer analogues that retain bacteriocide activity, but that do not activate PXR and induce CYP3A4, or have other undesirable qualities such as environmental persistence and pollution.

178 C hapter 5

Quantitative Structure Activity Relationships (QSARs)

of Nuclear Receptors

179 C hapter 5

Quantitative Structure Activity Relationships (QSARs) of Nuclear Receptors

5. Abstract A variety of quantitative structural activity relationships (QSARs) for the xenobiotic and steroid hormone receptors human PXR, AhR and ERa using multivariate techniques, specific descriptors, and biological data sets are generated and compared. These QSARs have utilized in-house and experimentally generated data sets, collaborative (unpublished) and literature sourced data sets in a variety of ways, ranging from both chemical class specific and broad data sets (drugs, pollutants, phytochemieals, hormones) in hPXR and hERa to the flame retardant polybrominated diphenyl ethers (PBDEs) biological data derived from a range of different species specific cell lines in the AhR. VolSurf descriptors were generated for each of the compound structures for which biological data were available and subsequently used to develop PLS regression based models; With the ERa model: (n=30), the Root Mean Sqaure Error of Estimation (RMSEE) =1.3 for Log relative binding affinity (RBA), and R" = 0.62. Two subsequent validation sets with a similar range of compounds demonstrated the applicability of the model (n=30, n=60. Root mean Sqaure Error of Prediction, RMSEP = 1.6). With the hPXR model: (n=33, RMSEE = 22.37, R^ = 0.83) for % maximum induction relative to rifampicin. This model was successfully internally validated using the randomization test With the AhR model: (n=12, RMSEE = 15.71), 2,3,7,8 TCDD was an outlier compared to the PBDEs, and major differences were apparent between the different cell line EC50s (n=7) reflecting PBDE ligand potencies and cell line EROD values (n=7) reflecting P450 induction. These QSARs are supported by utilizing the crystal structures of hERa, hPXR and a homology model of AhR (based on the ERa crystal structure), to examine the mode of binding of representative compounds in the ligand binding domain (LBD) of

180 the receptor crystal structures and model. Key commonalities and differences between the receptors LBDs and between the ligands have been observed. Analyses of the Volsurf variable contributions together with observed in silico modes of ligand binding in the respective receptors allow additional insights into receptor activation on a compound specific basis as well as ligand promiscuity and ligand-receptor crosstalk. This is of relevance to drug design in, for example, the treatment of ERa breast cancers, consideration of drug-xenobiotic interactions, and for a better understanding of the mechanisms of pollutant toxicity and endocrine disruption.

5.1. Introduction This chapter reports the QSARs for three NRs, the ERa, PXR and AhR and shows how a particular type of multivariate analysis, namely principal component analysis (PCA), and partial least squares regression (PLS), were used to model the data sets, and how these methods can be adapted to suit the available data and could be used for conducting comparative work with more NRs, to uncover the crosstalk signaling pathways between NRs and SHRs. For this chapter, it was originally intended to utilize the NR models and crystal structures, as described in chapter 4, for interactive docking studies with selected ligands, and to then use the measurements taken of the ligand specific distances for binding to occur in the LBD, as part of the descriptor data. It was intended to be used with the biological data generated in chapter 3 for the three-dimensional (3-D) Quantitative Structure Activity Relationships (QSARs) receptor studies part of this project. However, I did not have access to the work station to conduct this work, for a prolonged period of time, due to both hardware and software problems. This gave insufficient time to dock all the compounds of interest in the NR data sets, but a reasonable spread of activity data was available from the studies conducted in Chapter 4 to establish relationships between the biological data (x’s) and the chemical descriptors (y’s), in a PXR QSAR model, that could be tested further for predictive purposes. Many other forms of computational descriptors were available in-house at GSK, so these were generated from the chemical structures. Two additional QSARs were also generated, one concerned with ERa ligand binding, and thus oestrogenicity, and the other with brominated flame retardant aetivation of the AhR and P450 induction. The in silico models and crystal structures were again utilised to examine selected ligands of interest in the LBDs of the NRs, when the work station later became available for use.

181 The methodology for the QSAR for PXR ligands was originally intended to be the classical regression technique, multiple linear regression (MLR) (Jacobs, 1998). However this was subsequently changed to take advantage of the latest chemometric resources available within the GSK Biomet department. Multivariate statistical computational approaches, including the chemical structure descriptors, have been developed recently to deal with scientific and technological data sets. Whilst PLS was originally developed as part of chemometric toolsets in chemistry and engineering, it is now increasingly applied in the field of drug discovery for an early optimization of ADME (absorption, distribution, metabolism and excretion) properties. PLS regression utilizes the two block predictive model to model the structural relationship between two matrices, X and Y, and so provides further information than that of traditional regression techniques. In situations where the X-matrix contains less objects than variables, MLR is not appropriate, as it is designed to deal with situations in which the number of objects (N) is at least three times larger than the number of variables (K). Whilst this can be overcome by using stepwise MLR (Jacobs, 1998; Lewis and Jacobs, 1999), there is still a high probability to obtain relationships just by chance (Eriksson et al., 1995; Eriksson et al., 1997). Further, MLR assumes that the X-variables are "independent" and not correlated, but even at the outset, relationships between the variables for all the studies in this chapter are known, as for example with structural classes between flavonoids, PCBs, PBDEs, PB-like compounds and so on. PLS models the structure of X and Y, giving richer results than the traditional MLR. PLS can examine strongly correlated data, many X variables and model several response variables simultaneously (Wold et al., 2001). These approaches are just as useful for examining compounds in other contexts, such as environmental pollutants and dietary composition, and are useful for elucidating descriptor-based relationships of ligands and receptors. Chapter 2 used PCA techniques to compare data sets, in this chapter PCA and PLS are utilised to interpret and model NR and SHR activation. These receptor models can help understand the molecular mechanisms of ligand binding activation, and in doing so, improve prediction and identification of potential ligands. For these studies, the biological data was correlated to the 3-D molecular interaction fields of the molecular structures of the compounds and transformed into descriptors using Volsurf and theoretical physicochemical packages. Molecular descriptors were correlated to the biological data by a discriminant and PLS procedure. The NR activation QSAR models

182 produced can be used for exploring molecular receptor-mediated signaling pathways, for improved understanding and prediction purposes, for example in the design of lead optimization libraries, in examining xenobiotic and drug metabolism, and examining drug interactions and efflux substrates (e.g, with CYP3A4 or P-gp, see chapter 4). These models can aid both the prioritization of metabolically stable compounds in medicinal chemistry and drug development, and be used in public health risk assessment for predicting whether the compounds are likely to be endocrine disrupters. Several receptor QSARs were run in a variety of ways governed by the shape the data took. As a first step, a literature based QSAR study of utilizing in vitro ER alpha ligand binding data provided a training ground for using multivariate tools and a variety of different descriptors including Volsurf descriptors. Volsurf descriptors were used for the next QSAR study of hPXR, and then for a study of a suspected environmental pollutant class of AhR ligands, the brominated flame retardants (PBDEs), tested in a variety of different cell lines. By using the same computational descriptors for each QSAR process, the differences and similarities between specific descriptors significant to each QSAR model could be identified. Collectively, this information gives an added dimension to the understanding of NR mechanisms of ligand binding.

5.2. Methods 5.2.1. Collection of data for QSAR analyses Apart from the QSAR for PXR ligands, using experimental data generated myself, the other QSAR studies presented here were essentially QSARs of ‘opportunity’, in that the biological data sets utilised were governed by availability in the literature, GSK in-house data sets and from collaboration with another researcher. Biological data for a large and diverse set of oestrogenic compounds were obtained from the literature for a ERa QSAR (section 5.3), from the literature and from Dr Guosheng Chen (University of Guelph, Canada) for the AhR PBDE QSAR (section 5.5).

5.2.2. Generation of Smiles As the descriptor programmes were written to read molecular structures as SMILES strings, all the chemical structures of the compounds used to build the models were first transformed into SMILES line notation code, in-house at GSK, using the Daylight toolkit. This was done in ISIS, by utilising Iris, accessed through the Mantis

183 GSK intranet site, and sending the structure to ISIS as a sketch file. Using a GSK networked personal Pentium computer, the structures were built individually, then transferred, and the SMILES notation strings were retrieved and cut and pasted into the raw datasets built for PLS analyses.

5.2.3 Descriptors Descriptors for QSARs vary according to the model being produced, and the purposes for which it has been designed. Historically, basic physico-chemical descriptors were first developed, and are still utilised, both in chemometrics (Olsson and Oprea, 2001), environmental modelling (Calmari, 2002), and computational biology or bioinformatics (Fielden et al., 2002). A classie example is the octanol-air/water coefficient utilised extensively in environmental toxicology, such as surface/air partitioning and marine food web modelling for example (Escher and Hermens, 2002; Chen et al., 2002). This first approach and other similar parameters have been applied as empirical correlations between bioconcentration or solubility factors such as log Kow, in environmental QSARs, or lipophilicity, such as Log P, in pharmacokinetic QSARs (Van de Waterbeemd et al., 2001), or other properties or descriptors, particularly hydrogen bonding (Raevsky et al., 2002) for protein-ligand interactions (Eldridge et al., 1997). Theoretical descriptors range from simple atom counts and geometric atom pair descriptors (Sheridan et al., 1996) to complex quantum mechanical calculations (Olsson and Oprea, 2001; Soffer et al., 2001). In drug discovery and development, several models based on different physicochemical descriptors have been proposed for building predictive pharmacokinetic models. These are generally based around molecular weight (MW), lipophilicity, the number of hydrogen bond donor/acceptors and groups of compounds (Lipinski et al., 1997; Raevsky et al., 2002; Lewis and Dickins, 2003). Lipinski’s ‘rule of 5’ method (Lipinski et al., 1997) is commonly used for screening drugs and NCEs for solubility and permeability. The rule states that if a compound satisfies any two of the following rules, it will probably be poorly absorbed in the gastro intestinal tract (GIT), the exception being if it is also a substrate for biological transporters: 1. MW is greater than 500, 2. Number of hydrogen bond donors is greater than 5 (expressed as the sum of OHs and NHs), 3. Number of hydrogen acceptors are more than 10 (expressed as the sum of Os and Ns), and 4. Clog P is over 5.0 (or Mlog P >4.15) (Lipinski et al., 1997). While a compound needs to be absorbed to activate the NRs studied here; it is the activation of the receptors that is the parameter of interest here.

184 Lipinski’s ‘rule of 5’ does not apply to several compound classes, ineluding some antibiotics, vitamins, antiseptics/fungals, (and cardiac glycosides). As many of the compounds in the data sets used for the QSARS in this chapter include antibiotics, vitamins and an antifungal, and a better estimate of H-bond strength is required for quantitative predictions, a more appropriate choice and range of descriptors were used. The descriptors selected here included the basic physicochemical descriptors, Abraham descriptors, GSK in-house descriptors produced from GSK in-house models (Modi, S, unpublished models and descriptors) and VolSurf descriptors. Selected examples of the calculated descriptor data sets are given in Appendix 3. The logarithm of the octanol-water partition coefficient (ClogP) was calculated with the use of ClogP for Windows software (Biobyte version 4.7.1. Claremont, CA, USA). The experimental octanol-water coefficient was also obtained from the Biobyte program data base.

5.2.3.1 Abraham descriptors The ‘chemically intuitive’ Abraham descriptors are designed to correlate solute properties such as solubility, partitioning, blood-brain distribution, cell permeability and intestinal absorption with a set of five molecular descriptors, based on a solvation equation that correlates these properties. These parameters can also be useful for examining ligand-receptor H bonding interactions. Based on a theoretical cavity model of solute-solvent interactions, the descriptors are: A- solute H bond aeidity, B- solute H bond basicity, E- excess molar refraction (this models dispersion force interactions arising from the increased polarisability of 7C- and n- electrons, S- solute polarity/polarisability, V- McGowan characteristic molar volume (Abraham et al., 1994; Abraham et al., 2002).

5.2.3.2 GSK in-house descriptors A number of useful descriptors have been produced within GSK by Sandeep Modi, such as structural alerts that signal likelihood to be metabolized by transporters, such as Pgp. As these descriptors are proprietary, further specific information cannot be given in this thesis, only the general findings.

185 S.2.3.3 Volsurf descriptors VolSurf is a computational procedure to produce and explore the physicochemical property space of a molecule starting from 3-dimensional (3-D) interaction energy grid maps. The information present in 3-D grid maps is compressed into a few quantitative 2-D numerical descriptors using image analysis software. Each 3- D map is considered as 3-D image, but the image compression process is made by adding chemical knowledge. The pharmacokinetic properties of a compound depend on a variety of 3-D physicochemical parameters, mediated by surface properties such as shape, van der Waals forces, electrostatics, hydrogen bonding and hydrophobicity. They therefore require a multivariate description. VolSurf descriptors quantitatively charaeterize size, shape, polarity, hydrophobicity and the balance between them. VolSurf descriptors are both fast to calculate and are independent of alignment of molecules (Cruciani et al., 2000a; 2000b). Volsurf was developed for structure-property relationships, but is also useful for structure-activity relationships and molecular diversity based on surface properties. VolSurf builds on the descriptors that are routinely used in medicinal chemistry (MW, number of hydrogen bond donors/acceptors and lipophilicity), by using the GRID forcefield (Goodford, 1985; Bobbyer et al., 1989; Kastenholz et al., 2000; Molecular Discovery Ltd), which is able to calculate energetically favourable interaction sites around a molecule to produce 3-D molecular interaction fields (MIF) onto a 3-D grid. It uses a potential based on the total energy of interaction including dispersion forces (the sum of Lennard-Jones, H bonding and electrostatic terms) between a target molecule and a probe which can be either an atom or a group. By moving probes (such as a water molecule) over the surface of the target molecule, GRID generates a property distribution of attractive and repulsive forces between the probe and the target compound. Thus, the information on the characterisation of the putative polar and hydrophobic interaction sites around target molecules, contained within 3-D molecular fields, is related to the interacting molecular partners, which here are the key amino acids for binding in the LBD of the NR’s studied. Such information is more specific and more appropriate than that generated by single or 2-D computed molecular descriptors. The water probe simulates the solvation-desolvation processes, while the hydrophobic probe (termed ‘dry’) and H bonding carbonyl probe (termed ‘O’) are used to simulate interactions between polar headgroups and the dry hydrophobic core of biological membranes. The overall energy of the hydrophobic probe is computed at each grid point

186 as Eentropy + £" u ~ ^HB, whcrc ^entropy IS the ideal eiitropic component of the hydrophobic effect in an aqueous environment, E u measures the induction and dispersion interactions occurring between any pair of molecules and Em measures the H-bonding interactions between water molecules and polar groups on the target surface. Although 3- D-QSAR models can be obtained from these 3-D maps (CoMFA (Cramer et al., 1998) or GOLPE (Baroni et al., 1993) procedures), they are hard to interpret. Further difficulties arise for alignment and molecular flexibility, although programmes are becoming available that have improved flexibility (Hindle et al., 2002). Using image analysis software, VolSurf is a specialised tool designed to automatically convert these 3-D molecular fields into physico-chemieally relevant molecular descriptors and to link experimental observations with molecular structures, thereby only extracting the relevant information for the investigation in question (Cruciani et al., 2000a; 2000b). Volsurf generates 28 descriptors from the ‘water’ probe and 16 descriptors from the ‘dry’ probe. The first four parameters describe the size and the shape of the molecule, descriptors 5-12 indicate polar interaction sites at 8 different energy levels and descriptors 21-28 calculate the coneentration of polar interactions on the molecular surface. The "integy moment" (13-20) describes the distance of the centre of mass to the barycentre of polar interaction sites, at a given energy level. If the integy moment is high, there is a clear separation between polar and hydrophobic interaction sites. If the integy moment is small, the polar moieties are either close to the centre of mass or they are at opposite ends of the molecule and the resulting barycentre is close to the centre of the molecule. Descriptors 29-36 indicate interactions with the hydrophobic probe at 8 different energy levels, which have been adapted to the energy range of the DRY probe. Descriptors 37-44 "hydrophobic integy moment" describe the distance of the molecular centre of mass and the barycentre of hydrophobic interaction regions. In summary, the 3-D molecular structure is translated into physicochemically meaningful descriptors for the test compounds, without the need for alignment data from docking studies in the LBDs of the NR crystal structures or homology models (Chapter 3). Size, shape, hydrogen-bonding and hydrophobicity are quantitatively differentiated within a series of molecules and the resulting collinearity of descriptors is analysed using suitable multivariate statistics, such as PCA and PLS, as described in the next section.

187 5.2.4. Statistical tools The statistical tools used were PCA, the multivariate analysis technique that deals with the X-variables only and PLS regression, a statistical technique that explains one or more dependent variables (Y’s) in terms of a number of explanatory variables (predictors, X’s). PCA and PLS are well documented in the literature, the following overview is taken from the literature, Umetrics training manuals and GSK in-house modelling guidelines (Eriksson et al., 2001a; Luscombe, 2003). PCA and PLS were implemented in the SIMCA-P v.10.0 software (Umetrics AB, Umea, Sweden).

5.2.4.I. Principal Component Analysis (PCA) PCA summarizes the information contained within initial data into a form that may be more easily overviewed and used. It is useful for understanding the distribution of the objects, and exploring similarity, dissimilarity and why there are differences. PCA can highlight the variables that contain similar information, or the variables that contain completely independent information. PCA is utilised in this thesis in section 2.4 (where the contaminants in the oils sampled at two time points seven to eight years apart) and with the PBDE/AhR study using a number of species specific cell lines, as presented in this chapter, section 5.4. In PCA, the original multidimensional space defined by the chemical descriptors is contracted into a few descriptive dimensions. The dimensions are termed principal components (PC), and they represent the main variation in the data. Each PC may be displayed graphically and analysed separately, and its meaning may often be interpreted according to simple fundamental biological characteristics, such as lipophilicity. The PCA loading matrix (P) contains information about the variables. It contains a few vectors (PCs), which are obtained as linear combinations of the original X- variables. The score matrix (T), contains information about the objects and each object is described in terms of their projections onto the PCs, instead of the original variables.

X = TP’ + E

Information absent from these matrices is the "unexplained X-variance" in a residual matrix (E), which has just the same dimensionality as the original X-matrix. PCs are extracted in decreasing order according to how much of the variance they explain. Furthermore, as each PC is orthogonal to each other, there is no correlation between the information contained in different PCs. The dimensionality of the model, i.e. how many

188 PCs should be extracted, is decided according to how much information is explained by each PC. Usually the first four of five PCs explain more than 90% of the X-variance. Further information is obtained from comparative analysis of the plotted matrices.

The scores plot represents the relative position of the objects in the two- dimensional (2D) or three-dimensional (3-D) space of the principal components. They are useful to identify clusters of objects and single objects that behave in a peculiar way (outliers). The position of the objects in the scores plot relative to the loadings plot may serve to explain the clusters. The first PCs try to explain the maximum amount of variation and to distinguish between clusters of objects. In essence they are a composite of the distinctive features of the objects in these clusters. The loadings plot represents the original variables in the 2D or 3-D space of the PCs, which originally are obtained as linear combinations of the original X-variables. In this plot one can see which variables are influential, and how the variables are correlated. The loading of a single variable indicates how much this variable participates in defining the PC (the squares of the loadings indicate their percentage in the PC). Variables contributing very little to the PCs have small loading values and are plotted in the centre of the plot; the variables contributing the most are plotted around the borders of the plot. Negatively correlated variables are positioned on opposite sides of the plot origin, in diagonally opposed quadrants. The distance to the origin indicates the stronger impact that a variable has on the model. Loading plots are a test of the "homogeneity" of the contributions of the X-variables to the model. When there are more than one single field group of variables, the loadings can be plotted to highlight the variables belonging to one of them, to show how each field contributes to the whole description of the objects, as seen in chapter 2.4.

5.2.4.2. Partial Least Squares regression techniques (PLS) Partial Least Squares (PLS) is the regression method than can deal with the kind of X-matrices used in 3-D-QSAR, by decomposing the X-matrix as the product of two smaller matrices, much like PCA does. The PLS algorithm originally proposed by Wold is well documented in the literature, so this is not presented in detail here (Eriksson et al., 2001a; Wold and Ericksson, 1995). However, an overview is given to show how the PLS techniques were used and how the models were validated in this chapter. PLS assumes that there are a few ‘prineipal properties’ of the molecule whieh are directly related to the Y response, which here is the receptor-specific biological response

189 data. Different linear combinations of all the descriptors, called ‘components’, are made, in turn, to describe each principal property. By evaluating the regression coefficients of the PLS model in conjunction with the variable influence on the prediction, the relative importance of each descriptor can be identified. The loading matrix (P) contains information about the variables. It contains a few vectors (termed Latent Variables, LVs) which are obtained as linear combinations of the original X-variables. LVs are constructed in an analogous way to PCs in PCA, but differ in that PCA obtains the PCs that represent at best the structure of the X-matrix, while PLS obtains the LVs under two constraints, they have to represent the structure of the X-matrix and Y-matrix and they have to maximize the covariance between the X's and the Y's. The score matrix (T) contains information about the objects. Each object is described in terms of the LVs, instead of the original variables. The X-loadings (P) are calculated at the end of each iteration, to best approximate the X-matrix. They update the X-matrix, but are not directly involved in the model building. Both the LVs and PCs share some important properties as they are both extracted in decreasing order according to how much of the variance is explained. The first LV always contains more information than the second, the second more than the third and so on. Also each LV is orthogonal to each other, and there is no correlation between the information contained in different LVs. As with PCA, the scores represents objects in the space of X-scores (T) against the Y-scores (U), showing the correlation between the X's and the Y's obtained in the model for each one of the LVs. The plot of the first LV contains the main relationship between activities and structural descriptors. As for PCA, the best way to understand and diagnose the information from PLS is to plot the matrices obtained and examine all of the available plots with care and in conjunction with each other. Variable influence on the projections (VIP) An independent measure of the relative importance of the x variables is calculated as the VIP. They are derived from the PLS weights and take into account the fraction of variance explained in each model dimension. When examined in conjunction with the coefficients plot, information about the major contributing descriptors to the model is discernable.

5.2.4.3. Model Validity The conceptual model underlying 3-D-QSAR is that the large numbers of variables included in the X-matrix encapsulate the dominating effects due to changes in

190 structure over the actual molecules. Model validation is a crucial step in the development of QSARs. However when attempting to define the relationship between X and Y, a large number of variables in the X-matrix have no relationship with the activity and introduce only noise in the description of the molecules. Like any other experimental variable, the Y variables eontain error. The models will try to fit the Y’s as much as possible, but if the fitting is overworked, the model will also explain the noise. Overfitted models seem to work well on trainig data, but they often prove to be less useful to predict the Y’s of objects not included in the available data set (training set). In PLS, selecting the correct dimensionality is critical to the validity of the model, if too many LVs are included, the model will be over-fitted. To prevent this, without destroying the structure of the X variables, a conservative variable selection is undertaken. If there are too many variables, the models are more weighted to explaining the X-structure, and the Y's do not fit well. Sufficient variables are needed to give the X- structure, with too few variables the models are not stabilized. The Y's may fit well, but their predictive ability is doubtful, even if the internal validation indicates otherwise and shows promising results. The models presented here were therefore tested for predictive ability, taking into account how many LVs to include. There are four tools for model validation as follows:

5.2.4.3.1.External validation using a designed validation set Validating a QSAR with external data, while the most demanding, is considered the best method of validation. The validation should be performed by interpolation and the training set should be diverse, while the test set should ideally be representative of the training set, so that the predicted values can be compared to the actual observed values to give a reliable estimate of the predictive variance (Q^).

5.2.4.3.2. Splitting of parent data set into training and validation sets Producing large experimental data sets generated via in vitro test systems can be both time consuming and costly, and relatively large NR data sets that can be split into training and validation sets are not common in the literature. However it was possible to conduct this evaluation for the ERa QSAR, section 5.3, as this is arguably the most studied SHR receptor published. With the external prediction approach, the objects in the original data set are split up into two groups from the very beginning of the analysis. The first one, the learning set, will be used to build the PLS model. The other, the prediction

191 set, will be used to compare their experimental Y-values with the predictions made by the PLS model. This technique tests the predictive ability, but the results depend critically upon how many and which objects are assigned to each group. Also, the data sets in QSARs often contain too few objects and it is not possible to remove objects from the analysis without a loss of information. This form of validation, requiring large amounts of relevant biological data, was implemented for the ERa QSAR based on a large literature data set, with random selection between the test set and training set of compounds within representative structural classes (section 5.3).

5.2A.3.3. The randomization test For smaller biological experimental data sets it is not always possible to validate with another data set, because of lack of availability of further data. In this case, with PLS it is possible and preferable to internally validate the model. Repetitive randomizations of the response data (Y) of N compounds in the training set into an array of reordered variables is the basis of a probabilistic approach. In each cycle the resulting arrangement of random integers is used to reorder the Y data, with the X data intact, and then a data analysis is executed on the scrambled data to produce R“ and Q^ estimates. If these values are lower than for the original data, one can be confident in the relevance of the original QSAR model. Precise estimates can be produced from comparisons of the histograms of R^ and Q^ values derived from hundreds of cycles, but this can be very time consuming to conduct. Wold and Ericksson, 1995, consider that around ten trials can be sufficient for identifying the key features of the R" and Q^ histograms. The randomization test is an internal validation tool that is built into the multivariate software package used for these studies (SIMCA v 10. Umetrics AB, Sweden), and was conducted for the PXR QSAR model, section 5.4.

5.2.4.3.4. Cross validation Cross-validation (CV) differs from the previous method as it has a predictive basis, and is considered a reliable way to test the predictive significance (Wold and Eriksson., 2001). CV entails dividing the data into a certain number of groups, then constructing a number of parallel models from reduced data (with one of the groups deleted). The differences between the predictive Y values and the model Y values are calculated for the deleted data. This process is repeated until each group has been omitted once. The sum of the squares of these differences is computed and collected from all the

192 models to form the predictive residual sum of squares (PRESS), which estimates the predictive ability of the model (Wold and Eriksson. 2001). The PRESS is also expressed as (the cross validated R^), which is (1-PRESS/SS) where SS is the sum of the squares of Y corrected for the mean. SDEP: Standard Deviation of Errors of Prediction.

PRBSS

Çf: Predictive correlation coefficient

I j y - r f S " ' H X - 7 Ÿ

Y : Experimental value Y' : Predicted value y : Average value N : Number of objects

The CV technique performs an "internal validation" of the model and obtains an estimation of the predictive ability without the help of external data sets. This is particularly important in QSAR studies, where the number of objects available is usually small, and one cannot afford to remove objects from the learning data set. This method was therefore utilised for the PXR QSAR, section 5.4. There is no agreed method on how to build the reduced groups and on the criterion to decide how many objects to keep for CV, although the objects should be deleted once only. The appropriate approaches used depend upon the peculiarities of the QSAR data sets and the ways in which the objects are clustered. The approaches can include leaving one or several out, where models are built keeping one (or several) object(s) at a time out of the analysis, and repeating the procedure until all the objects are kept out once. Leaving several out is considered preferable both on a theoretical and practical basis (Wold and Eriksson., 1995), as it is less likely to over-predict the model, thus giving a much better, but more conservative estimation of the real predictive ability; the uncertainty of future predictions may be numerically worse but the model is more reliable. To obtain stable results, the procedure is repeated many times, and the Standard Deviation Error of Prediction (SDEP) is computed to give an estimate of the dispersion of the SDEP values obtained from different runs, and thus the reproducibility between the reduced models (Eriksson et al., 2001a). This internal validation approach was used for

193 the PXR QSAR. Later, test sets of in-house proprietary data were also used to explore the predictive nature of the PXR model, but the test sets were not representative of the training set, as they consisted of drugs and NCEs, most of which did not belong to the structure classes in the training set, which was dominated by dietary compounds, hormones and drugs. The number of principal components is determined with cross-validation. R^X, is used to assess which variables are well explained, and those which are not. R^X, is the fraction of the sum of the squares (SS) of all the X’s explained by the current component. R^ increases with increased number of principal components. However, if the model increases too much in complexity, the validity can be reduced. To determine the appropriate number of principal components to include in a model, a cross-validation (CV) method is used to simulate the predictive power of a PC model. QSAR models are conventionally evaluated by considering the R^, the ‘explained variance’ or ‘goodness of fit’ and s, the residual standard deviation (RSD). The predictive residual SD’s are termed the standard error of prediction (SDEP). R^ can range between 0 and 1, where 1 is perfection, with 100% of the response (Y) data explained, and 0 indicating no explanatory power. A high R^ and low RSD are among the requirements for valid models, together with a reliable estimate of Q^, the ‘predicted variance’ or ‘goodness of prediction’. With the models shown in this chapter, it is preferable for Q^ to be greater than 0.5, and a very good score is greater than 0.9. While R^ must always be larger than Q^, and a high R^ value and high Q^ value is good, the difference between R^ and Q^ should not be too large (Wold and Ericksson, 1995).

5. 2.4.4. Predictive models A good model is able to calculate Y values that correspond to the experimental ones, even for molecules not included in building the model. These models can be used to calculate reliable estimations of Y values for new molecules, prior to their availability, and so be predictive. The predictive ability of a model is attributed to the existence of a "true" relationship between the measured X properties (physico-chemical and field interaction energies) and the Y (biological response) property measurements. Three practical examples using these techniques are shown in the following sections. The interpretation of VolSurf parameters will be explained and the pharmacokinetic relevance of these descriptors will be shown together with correlations

194 with biological activity parameters. In every model described below, values of R and Q from CV are quoted. Section 5.4, ‘Building a QSAR for PXR’, is explored further with two GSK proprietary data sets. An additional computational programme, Spotfire (www.spotfire.com) was utilised to obtain further graphical information on the outlier descriptor activity observed when two external proprietary training sets were tested in the PXR model. In the end, as with all regression models, even the best constructed 3-D-QSAR model is limited by the series it was obtained from. However even if the models are rough and are not able to explain to a full extent the activity values, they are extremely valuable to identify the structural features that contribute most to the activity. These are important in the context of the ligand-receptor mechanisms of interaction, and suggest how new chemical entities (NCEs) might be designed to stimulate or remove binding potential, and how a test compound might be predicted to behave at the molecular ligand-LBD interface. As the same VolSurf parameters are used in the QSARs (sections 5.3, 5.4 and 5.5), the models can be compared with each other more easily than with those in the scientific literature, so giving a new perspective in receptor signaling behaviour.

5.2.5. Molecular modelling These models are also compared to the ligands visualized in silico in the crystal structures and homology models of the respective receptors. The ERa crystal structure, the PXR crystal structure and the AhR homology model as described in chapter 3 were refined using molecular mechanics to perform energy minimization of the entire 3-D structure, to achieve a low energy, stable optimized geometry. Each structure was minimized to produce a minimum energy geometry for each molecule (test compound) docked in the respective NR LBDs. The 3-D structures of compounds of interest were produced for docking into the relevant receptor using the build and edit/sketch option in Sybyl (Tripos Associates, St Louis, MO). The putative binding site of each receptor was located and the endogenous ligand or high potency ligands were fitted into the respective sites and became the template to position other compounds. Hydrogen bonds were displayed to ensure the binding orientation was consistent with their formation, and modes of binding were examined. While one cannot (as yet) visualize the dynamic activity of NRs and ligand binding in real time, these models are a close approximation tool. The techniques used

195 can be applied to many available data sets for different NR biological data to generate structure activity information and to predict a chemical’s potency as a ligand for a given hierarchy of receptors, as well as for the individual receptor.

196 5.3. Investigation of structure-activity relationships using PLS regression and VoiSurf descriptors for a large diverse set of natural, synthetic and environmental compounds in Estrogen receptor a.

5.3.1. Abstract Relative binding affinity (RBA) data for a large and diverse set of oestrogenic compounds were obtained from the literature and utilised to build a semi-quantitative partial least squares (PLS) regression model for the ligand binding properties of estrogen receptor a (ERa) ligands. The biological data ranged over 7 orders of magnitudes (Fang et al., 2001). VolSurf descriptors were generated for each of the compound structures and subsequently used to develop a PLS regression based model (n=30, RMSEE = 1.3) for Log RBA. Two subsequent validation sets with a similar range of compounds demonstrated the applicability and generalisability of the model (n=30, n=60, RMSEP = 1.6) to allow reliable classification of new compounds into low/medium/high compartments for prescreening purposes. Further analyses of VolSurf variable contributions (particularly hydrogen bonding, capacity factors, hydrophilic-lipophilic balance, hydrophobic regions at eight energy levels, globularity, rugosity, and amphiphilic and integy moments) will allow additional insights into ER ligand activation on a compound specific basis which could be of relevance to drug design in the treatment of ERa dependent cancers.

5.3.2. Introduction The physiological roles of endogenous oestrogen (17|3 oestradiol) are of critical importance in regulating development, growth and maintenance of reproductive tissues as well as in normal processes of numerous tissues including: cardiovascular, neural and skeletal cell functions (reviewed in chapter 1). Oestrogens are involved with many signaling pathways, but the major receptor mediated pathway is through the Estrogen Receptors (ERs). The ERs are ligand regulated transcription factors whose activity as activators or repressors of gene transcription depends on the availability and nature of the ligand with which they are bound, as well as the nature of the coregulator proteins and receptors. As the ERs play a role in bone density maintenance, regulation of blood-lipid profiles and brain function, understanding structural features necessary for a chemical to

197 bind to the ERa is a target for the discovery of new drugs for treating or regulating a variety of hormone-related conditions. It is also essential for understanding oestrogenic endocrine disrupting effects. Many natural and synthetic chemicals have been shown to be capable of eliciting or antagonising oestrogenic responses in various species, through similar mechanisms, although the physiological outcomes can differ. Fang and co-workers (Fang et al., 2001; Blair et al., 2000) recently reported satisfactory sized data sets of experimental binding data for ERa, and the latter report was utilised for the present PLS study. There is a number of other data sets of appropriate size in the literature for conducting similar QSARs (e.g. Miller et al., 2001), and future studies would consider these also, particularly with respect to tier testing and assay assessment for chemical risk assessment and relevance to human health. Recently a number of other ERa QSAR studies based on small (Jacobs and Lewis, 1999), medium (Sadler et al., 1998) and large data sets, such as those generated by high throughput screening (HTS) (Gao et al., 1999; Hong et al., 2002), have been published. The model presented here supports the information generated by other studies, and also adds to the understanding of ERa ligands in relation to the AhR and PXR (section 5.4), which can be used to explore further cross-talk relationships with other receptors.

5.3.3. Methods 5.33.1. Source of biological data RBA data, conducted in the same laboratory and obtained from the scientific literature (Fang et al., 2001), was split into three data sets, the training set and two validation sets, each containing a representative spread of hormones, phytochemieals, pharmaceutical compounds and environmental compounds.

5.3.3.2. Generation of VolSurf descriptors. As for section 5.2.2.2. The VolSurf descriptors generated are shown for the QSAR data test set 2, in Appendix 3.Table ii.

5.3.3.3. Statistical tools PLS was utilised and it was assumed that there are a few ‘principal properties’ of the molecule which are directly related to the Y response, which in this model are the RBAs reported by Fang et al., (2001). The basic physico-chemical features of the molecule such as acidity, basicity, lipophilicity, size, polarity, etc. are the principle

198 properties. Different linear combinations of all the descriptors called ‘components’ were made, in turn, to describe each principal property. The contribution that a descriptor makes to each component will vary depending on how relevant that particular descriptor is to the principal property. By evaluating the regression coefficients of the PLS model in conjunction with the variable influence on the prediction, the relative importance of each descriptor was identified.

5.3.3.4. Molecular modelling

The ERa crystal structure as described in chapter 3 was refined using molecular mechanics to perform energy minimization of the entire 3-D structure, to achieve a low energy, stable optimized geometry. Each minimization took 100 iterations to produce a minimum energy geometry for each molecule (test compound) docked in the LBD. The 3-D structures of each compound were produced using the build and edit/sketch option in Sybyl (Tripos Associates, St Louis, MO). The putative binding site of the hERa was located and ivpoestradiol was fitted into the site. This ligand binding pocket is well defined within a loop of peptide containing complementary amino acids for binding interaction, as with agonists and antagonists. Combined with a predominance of leucine residues, the site is in an optimum position for dimerization leading to DNA interaction (Jacobs and Lewis, 1999). Hydrogen bonds were displayed to ensure the binding orientation was consistent with their formation. The fitted l?Poestradiol molecule became the template to position other compounds.

5.3.4. Results and Discussion The PLS regression based model (n=30, RMSEE = 1.3, R^ = 0.62) for the Log RBA training set is shown in figure 5.1. Two subsequent validation sets with a similar range of compounds demonstrated the applicability and generalisability of the model (n=30, n=57, RMSEP = 1.6). The test plots are shown in figures 5.2 and 5.3. Whilst the R^ value is not significant, this model demonstrates a clear trend and the compounds may be classified into suggested low and high compartments as indicated by the dashed (green) lines. Figures 5.2 and 5.3 show the test sets, and indicates selected chalconids, polychlorinated biphenyls, triphenylethylenes, and diphenolalkanes to be outliers. Figures 5.4 and 5.5 show the standardized coefficients plot and the variable importance plot for the ER a model.

199 Selected compounds were modelled in the ERa crystal structure, as shown in figures 1.5, 5.6, 7.7 and 5,8. Figure 5.1. Training set for ERa VolSurf model

■mm Training set for Volsurf Model : ER -alpha y=1"x+9.762e-008 ^lethylstilbest. ^ ^ R 2 = 0 .6 1 7 9 ^thynyiestrag*hyd■thvnvfc Al7-i>eta-eistradi

Adroloxifern^ I AE s r o n e m stra n o l Üclomiphene

ABppben o B 1

; .APi|phen

APhenolred làdich orodipheny

r AM-ethylphenol Arutin A P -e th y lp h e n o y ;

^ -ethylphenol &

- v YPred[1)(Log RBA)

RMSEE ^ 1.3 3047

200 Figure 5.2. Test set 1 for ERa VolSurf model

Test set 1

ydroxyEstradi

▲Droloxifene

tranol

ABalphahydroxy ANalo

APhloretm AAurin A3Methylestnol DodecymjJi^nol enolphthalein

ABenzylparaben A4Hydroxychalcon AP rune tin uivlDarabetABalaMéthenylp lenol aoerf A6Hyc|A0l^h*myipl»en A2_4Dichlorobiph :

A- YPredPS[1](Log RBA)

RMSEP • = 1.6 9 9 0 5

Figure 5.3. Test set 2 for ERa VolSurf model

I est set 2

AHexestrol-meso

17-Deo) v -estrad 6-arrethyl he i V A H exestrol

AMeso-ü

A4-ethywai0 3,6^^rihydi^ A 5(10) # 4 A2,2.4.4-tetra AHFTg^rnonohyd A^,4',5Vte 4R?-Di#Pox\^

4’-trih^wESdimethyl eth

A4-t-octy^-fîiJpylhexylF|

A 7 ,3 ',# 4 2'm£:thylbis(4 A4,4’-dihydroxyb , ,, A2,4-dihydroxybe A triphen

4.4'-suffonyldi ï^-hy^foxy'llavéi^JsfeS(:§^l3^^ A3,3’,5,5' tetra diphenolic acid . . A^'butylphè,4-b«nzyloxyphen I A2-chloroA4nHtihDro -2-metn A7-hvdroxvflaAoHlA7-hydroxyf lavèpKnolphtha lin

-1 0

YPredPS[1](Log RBA)

RMSEP = 1 . 61186

201 5.3.4.I. Key VolSuif variables The key variables were: Ratio Volume/Surface (R), a measure of molecular wrinkled surface or rugosity; Hydrophobic regions (D1 - D8); Hydrogen Bonding (HE 1-8); Integy Moments (Iwl-IwS) which measure the imbalance between the centre of mass of a molecules and the position of the hydrophilic regions around them; Polarizability (POL) (Miller, 1990); Amphiphilic Moment (A), which is a vector pointing from the centre of the hydrophobic domain to the centre of the hydrophilic domain. The vector length is proportional to the strength of the amphiphilic moment, and it may determine the ability of a compound to penetrate a membrane; Hydrophilic-Lipophilic balance (HLl, HL2), the ratio between the hydrophilic regions measured at -3 and -4 kcal/mol and the hydrophobic regions measured at -0.6 and -0.8 kcal/mol. The balance towards HLl indicates that hydrophilicity dominates in the molecule; Capacity factors (Cwl-Cw8) which represent the ratio between the hydrophilic regions and the molecular surface; Molecular Globularity (G) which is 1.0 for perfectly spherical molecules. It assumes values greater than 1.0 for spheroidal molecules and is also related to molecular flexibility.

202 Figure 5.4

Standardised Coefficients Plot for Model ER -alpha

0.100

0.080

0.060

0.040 3 0.020

I 0.000

-0.020 ■

-0.060-

g g g g ;§DIQÛQo Sq QDQQQQQq Sqd QQQQ^^<; !ee

Var ID (Primary)

Figure 5.5

Variable Importance Plot for model ER -alpha

9: 1.20

111 r?

L U L U $ L U Var ID (Primary)

203 S.3.4.2. Molecular modelling: Comparison with selected compounds docked in the E R a crystal structure npoestradiol and selected test compounds, representing the different classes of ligands in the QSAR, formed three hydrogen bonds with Glu 353, His 524 and Arg 394 in agreement with other studies (Waller et al, 1996; Métivier et al., 2002). The compounds modeled included hydroxylated PCB 153, hydroxylated PCB 118, coumestrol, daidzein, genistein, 4-hydroxytamoxifen, ICI 182780, PBDE 47 (2,2’,4,4’PBDE) and 17P oestradiol, using the template provided by the crystal structure complex of E2 bound in the hERa (Jacobs, 1998). Key binding interaction is observed, the A-ring phenolic hydroxyl makes direct hydrogen bonds to the carboxylate of Glu 353, the guanidinium group of Arg 394 and a water molecule. The D-ring’s 17P hydroxyl makes a single hydrogen bond with His 524. The A-ring is sandwiched between the sidechains of hydrophobic residues on its a and p-faces, and the D-ring has non-polar contact with hydrophobic residues. Where bonding occurs, OH charges on the phenyl ring may be significant. The critical presence of an unhindered phenolic OH group in a para position for ER binding has been noted in other ER models (Métivier et al., 2002) and for other ER bioassays such as the yeast oestrogen screening assay (Miller et al., 2001). The other main interactions are believed to be non-polar. Analogues of oestradiol with a 3-position moiety that is only capable of acting as a hydrogen bond acceptor, and not a donor, have poor binding affinities. This may be due to electrostatic repulsion between the 3-position heteroatom of the ligand and the carboxylate of the Glu 353. The hydrogen bond accepting feature of His 524 appears to be quite flexible when comparing agonist and antagonist structures, whereas Glu 353 forms a salt bridge with Arg 394, defining its location (Gillner et al., 1998). A key structural difference between phytoestrogens and oestrogens is that the 6 Â biophore (a distance descriptor) present in endogenous oestrogens and xenoestrogens (related to the A ring region in 17poestradiol) is not present in phytoestrogens. They appear to be less conformationally flexible due to a planar aromatic structure, unlike many of the other ligand classes included in the ER QSAR study (Jacobs, 1998; Jacobs and Lewis, 1999), as one can see when comparing the anti-oestrogens: tamoxifen and ICI182780, shown in figures 5.6 and 5.7, compared with figure 5.8, showing coumestrol in the ERa LBD. This may increase the likelihood of phytoestrogens being outliers in this model.

204 The distances for binding with the amino acid residues in the binding site were within the range expected for the formation of the hydrogen bonds (2 to 3Â). Receptor flexibility or the presence of a bridging water molecule might account for the ability of the ER to accommodate ligands with a variation in the D-ring and differing oxygen-oxygen distances, as observed previously and discussed in Chapter 1 (shown in figures 1.5, 1.6 and 1.7), further illustrations are given in figures 5.6, 5.7 and 5.8. Ligand binding modulates the hydrogen bonding networkwhich provides the interpeptide communication in the hER between ligand binding and DNA activation. When an agonist molecule completes the electron transport ‘circuit’, the resulting effect may disrupt the existing framework of hydrogen bonds to free the DNA zinc finger DNA binding domain. In the crystal structure complex, a number of ligand classes have been observed to bind, including anti-oestrogens tamoxifen and ICI 182780 (figures 5.6, 5.7), and phytoestrogens coumestrol (figure 5.8), daidzein and genistein (figures 1.5, 1.6), together with PCBs and PBDEs, as shown in figure 1.7 (Jacobs, 1998). The anti-oestrogens tamoxifen and ICI 182780 bind with additional amino acids outside the core LBD, as identified for the endogenous ligand 17|3 oestradiol see (figure 1.3), and discussed in Chapter 3 (see figure 3.7 ) and (Jacobs, 1998). Chlorinated compounds, such as the PCBs, which were also outliers in this model, may have an H bonding capacity in a similar manner to that reported for chlorinated pesticides such as hexachlorohexanes (HCH), and p ,p ’ DDT, they contribute to the overall H bond basicity (Abraham et al., 2002a) and are also known to be weakly oestrogenic. Figure 5. 6. Tamoxifen docked in the ERa LBD As with the previous model and crystal structure NR figures, the ribbon denotes the protein backbone of the receptor, key amino acids required for binding are indicated in blue and red, and hydrogen bonding is indicated by dashed yellow lines.

205 Figure 5.7. ICI 182780 docked in the ERa LBD

Figure 5.8. Coumestrol docked in the ERa LBD

5.3.5. Discussion In combination with molecular structural data, this study has generated theoretical data for the production of relationships (QSARs) which examine and estimate potency differences between these compounds from a large and varied data set. Molecular dynamics simulation is a useful tool to investigate the mechanism behind the conformational energy changes following ligand binding which lead to hER activation, particularly when compared to in vitro estimations of the ligand binding affinity. The results of this investigation support and expand homology modelling and structure activity relationship (SAR) studies for oestrogenic compounds available in the literature (Hong et al., 2002), which utilize Insight software packages (Molecular Simulations Inc) and the Sybyl software package (Tripos Associates Inc) with MLR,

206 respectively. A previous ERa QSAR study with a small relative binding data set utilised quantum mechanical computer calculations obtained by docking ligands in the LBD of the ERa crystal structure. In this study the key parameters were hydrogen bonding, ionization potential, AE (where AE is the E l u m o -E h o m o , for E h o m o the relevant xenobiotic characteristics include ionization potential, ease of oxidation and nucleophilic reactivity, for E l u m o these are reduction potential, ease of reduction and electrophilic activity) and minimized free energy. (Jacobs, 1998; Jacobs and Lewis, 1999; Soffer et al., 2001). There are several QSAR and structure-binding affinity studies of the ER in the literature utilising different 2D and 3-D descriptors (Suzuki et al., 2001; Gillner et al., 1998; Sadler et al., 1998), as in vitro ligand binding data, using different test systems, has been generated in many different laboratories for some years (Waller et al., 1996; Fang et al., 2001; Miller et al 2001; Suzuki et al., 2001; Hong et al., 2002). As it becomes available, biologically derived data on novel ERa selective ligands can be evaluated further using such QSAR models, particularly when investigating the molecular mechanism of activation in ERa compared to ER(3, as observed for the furans, also a known AhR ligand class (Mortensen et al., 2001). The key Volsurf descriptors generated for this QSAR support the descriptor information reported for other ER QSAR studies, and are also compared with the VolSurf descriptors generated in the PXR QSAR (section 5.4) and the AhR QSAR (section 5.5) in Table 5.3 and section 5.6. In the context of whole biological systems, ERa QSARs need to consider self regulation of the ER family between ERa, ERp, ERy and mutant ERs, and associated proteins, together with exposure, half-lives and body burdens, metabolism and excretion of different classes of compounds, on a gender specific basis. Further development of such models, based upon sequence homology, site-directed mutagenesis and crystallographic data, in combination with multivariate data handling techniques and descriptor tools (e.g. VolSurf), in relation to experimental and observed data, can provide a useful tool in an area of considerable public concern. It could be invaluable for providing a scientifically sound perspective on the potential risk to human health from environmental and dietary exposure to oestrogenic compounds and other endocrine disrupters, and be of potential relevance in the design of novel therapeutic antioestrogens for oestrogen related cancer treatments. These results are being utilised further in ongoing comparative QSAR receptor studies examining crosstalk and ligand promiscuity.

207 5.4. Building a QSAR for the Pregnane-X-Receptor (PXR) 5.4.1. Abstract There is increasing evidence that potent inducers of CYP3A can be present in the human diet, from food constituents, food contaminants and over-the- counter herbal remedies and vitamin supplements. These can be as potent as pharmaceutical compounds, as demonstrated by the effect of hyperforin (St John’s Wort), which is ten-fold more potent than the prototypical pregnane X receptor (PXR) ligand, rifampicin. PXR transient transfection assays are utilised by the pharmaceutical industry for the prediction of the CYP3A induction potential of drug candidates. These assays also have applications for the analysis and risk assessment of dietary and endocrine disrupting compounds. Data for a diverse set of 33 compounds and metabolites, including drugs (tamoxifen), bile acids (lithocholic acid), hormones (progesterone), phytochemicals (silybin), lignans (enterolactone), bacteriocides (triclosan), and persistent organic pollutants (polychlorinated biphenyls) were derived experimentally. The PXR activation ranged from 1.3 % (trans resveratrol) to 152 % (ICI 182780) and 184 % (GSK SR analogue) when compared to lOuM rifampicin. VolSurf descriptors were generated for each of the compound structures and subsequently used to develop a PLS regression based model (n=33, RMSEE = 22.37, R2 = 0.83) for % maximum induction relative to rifampicin. This model was successfully validated internally using the randomization test.

5.4.2. Introduction An important requirement for homeostasis is the detoxication and removal of endogenous hormones and xenobiotic compounds with biological activity. The Pregnane X Receptor (PXR) is involved in activating the expression of several P450 heme containing enzymes, including CYP3A4 in the adult and CYP3A7 in the foetus, in response to xenobiotics and steroids. While this provides a route for excess steroids to be eliminated from the body, it also means that drug-drug-xenobiotic-phytochemical interactions are likely, as CYP3A4 is the major human hepatic P450 responsible for the oxidative metabolism of over 60% of drugs in clinical use. Chapter 4 and figure 4.1, summarize the literature on the mode of PXR activation. In this study the in vitro data generated in chapter 3 were developed to generate PXR QSAR models using VolSurf descriptors, which was then supported and illustrated by docking ligands in the PXR

208 crystal structure, in various orientations as indicated by the orientations of SR12813 in the crystal structure (Watkins et al., 2001). This approach has allowed a better understanding of PXR ligands.

5.4.3. Methods Biological data were generated to yield a spread of activity for hormones, phytochemicals, pharmaceutical compounds and environmental compounds.

5.43.1. Generation and collation of biological data using an in vitro transient transfection assay The transient transfection assay, developed in the GSK Nuclear Hormone Receptor Group at RTP (Moore et al., 2000b), was the in vitro procedure used to generate the biological data in the form of the % maximum induction in relation to the positive control Rifampicin (Rif) (lOuM). The generation of these data and typical dose- response curves are described in Chapter 4. The data set for building the VolSurf Model is shown in table 5.1. Only reproducible data were used, the number of experiments, mean values and standard errors are indicated. Where only one experiment is indicated, these data were supported by similar values observed using CV-1 cells for the same PXR transient transfection assay, or reported by another operator using HuH7 cells. Two further data sets generated from the same assay, using the same HuH7 cells, meeting all the control criteria identified in chapter 3 (i.e. acceptable fold induction, positive control Rif values within the 80-120% range for % mean induction, relative to Rif), and conducted within the last two years, were mined and collated from GSK in- house data banks, for later exploration as test sets. These data included GSK validation sets, e.g. table 5.2, and data on a large number of proprietary experimental compounds. For proprietary reasons the structure and identities of these compounds cannot be reproduced here, however. Selection was based only on the reliability of each experiment, and availability of data. Selection was not based on structure or exposure as with the original data set. Test set one was generated principally by Gail Nolan at the GSK site in Ware, UK (n= 67). Test set two was generated under the supervision of Linda Moore at the GSK site in Research Triangle Park, North Carolina, US (n= 149). The test set data (selected data shown overleaf) was collated similarly to that for the training set, as shown in figures 5.1 and 5.2, with mean values taken where several

209 experimental data sets were available for certain compounds, particularly those that constituted validation sets.

5.4.3.2. Generation of descriptors Physico-chemical and VolSurf descriptors were generated as described in section 5.2.2.2.

Table 5.1. Summary of the biological data for PXR QSAR model

Primary m ax ID N am e induction SE Top dose N 1 5 beta pregnane 3,20 dione 42.88 3.06 lOuM 4 2 17 alpha ethinyl oestradiol 30.95 0.55 lOuM 2 3 17beta oestradiol 24.49 3.66 lOuM 5 4 4hydroxytamoxifen 84.17 12.66 lOuM 4 5 Clotrimazole 96.17 2.54 lOuM 3 6 c o u m estro l 7.49 2.99 lOuM 4 7 daidzein 14.48 2.52 lOuM 2 8 dexamethasone 24.21 7.44 lOuM 4 9 enterolactone 35.61 17.09 lOuM 2 10 eq u ol 9.95 4.44 lOuM 2 11 genistein 8.09 2.83 lOuM 3 12 hyperforin 112.78 22.1 luM 4 13 101182780 152.05 19.44 lOuM 3 14 lithocholic acid 11.07 lOuM 1 15 hydroxyvitaminD3 55.90 5.75 lOuM 4 16 PC B118 26.63 2.07 lOuM 2 17 PCB 153 16.93 2.07 lOuM 2 18 pregnenolone 16 alpha carbonitrile 7.45 7.45 lOuM 2 19 Progesterone 15.73 2.14 lOuM 4 20 quercetin 11.78 0.7 lOuM 2 21 rosiglitazone 37.00 lOuM 1 22 secoisolariciresinol 20.03 1 lOuM 2 23 silyb in 41.30 36.04 lOuM 2 24 S R 12813 136.32 9.6 lOuM 5 25 tam oxifen 74.60 21.7 lOuM 2 26 trans-resveratrol 1.34 1.34 lOuM 3 27 trans-nonachlor 53.82 5.47 lOuM 8 28 triclosan 46.18 4.45 lOuM 5 29 vitam in D3 1.61 1.61 lOuM 3 30 GSK SR analoguel 137.40 lOuM 1 31 GSK SR analogue2 184.30 lOuM 1 32 GSK SR analogue3 130.10 lOuM 1 33 GSK SR analogue4 122.70 lOuM 1

5.4.3.3. Statistical tools Partial Least Squares Regression (PLS) analytical techniques were used as described in section 5.2.2.3. PLS assumes that there are a few ‘principal properties’ of the molecule which are directly related to the biological (Y) response, which in this model is the % mean induction relative to the positive control, Rif. Different linear combinations of all the descriptors called ‘components’ are made, in turn, to describe each principal

210 property. As with the ERa QSAR study, by evaluating the regression coefficients of the PLS model in conjunction with the variable influence on the prediction, the relative importance of each descriptor can be identified. Table 5.2. Summary of selected biological data used as part of a later test set (set 1) for the PXR QSAR model

Compound %MAX SEN

Om eprazole 45.14 3.02 5 Troieandomycin 82.42 4.63 6 Dexamethasone 21.96 13.98 3 Phenytoin 19.50 3.50 2 S R 12813 145.70 18.90 4 Troglitazone 111.75 10.50 5 Troglltazone metabolite 74.65 5.35 2 Phénobarbital 9.90 9.90 2 Testosterone 0.00 0.00 2 Estradiol 38.60 19.20 2 Progesterone 15.45 8.25 2 Clofibrate 14.31 10.01 2 PON 8.64 1.04 2 RU486 80.00 13.80 4 Mevinolin 230.63 41.25 5 Primidone 7.62 2.62 2 Felbamate 2.10 2.10 2 Carbamazepine 12.96 5.26 2 Clotrimazole 67.72 13.92 2 Taxol 49.61 12.68 4 Hydrocortisone 6.10 3.66 4 Prednisone 0.00 0.00 2 Lansoprazole 27.25 9.22 4 Cyclosporine 0.00 0.00 2

5.43.4. Molecular modelling: in silico investigations of ligand binding in PXR. Ligands of interest were selected as described in section 4.3 and were initially investigated in the homology model of PXR as described in Chapter 3, and then the PXR crystal structure when it became available (I gratefully acknowledge the help of Sandeep Modi, GSK in obtaining the PXR crystal structure coordinates). Examples of the compounds, first built in Sybyl (see Chapter 3) are shown, docked in the ligand binding domain (LBD) of the PXR crystal structure below (Figures 5.9, 5.10) using SYBYL biopolymer software from (Tripos associates, St Louis, Missouri) on a Silicon Graphics

Indigo2 IMPACT 10000 Unix workstation. Compounds were docked in the PXR LBD using SR12813 as the template in the orientation shown in figure 5.14.

211 5.4.4. Results The models and validation sets are shown in figures 5.9 to 5.12, overleaf. The physico-chemical and Volsurf models compared well (figures 5.9a and 5.10), but with a better value, of 0.8, the Volsurf model was validated and tested with further GSK proprietary data sets. However, the coefficients and variable importance plots for the physico-chemical model are given (see figures 5. 9b and 5.9c), showing that the key positively correlated descriptors were MW, the number of heavy non-H atoms in the molecule, CMR (a function related to size), and Vx, relating to McGowan characteristic volume. As VolSurf was not able to generate descriptors for the larger compounds in the test sets, but physico-chemical descriptors could be generated, this allowed a direct physico-chemical descriptor comparison with the model generated for the outliers (figure 5. 26), and this is further elaborated on in the discussion in section 5.4.5. The VolSurf PXR model demonstrates a clear trend and the compounds may be classified into low, medium and high affinity ligands. As highlighted by the red arrows, vitamin D3 (29) was an outlier, but the hydroxylated form was a medium affinity ligand (Figure 5.10). Figure 5.11 shows the successful 3-component internal validation via the randomization test. Figures 5.14 to 5.21 are examples of high, medium and low affinity ligands docked in the LBD of PXR. Differences in binding with key amino acids are apparent for the larger high affinity compounds. Figure 5.14, shows SR12813 docked in the PXR crystal structure in one of the crystallization orientations a) docked in the LBD, showing the whole protein ribbon backbone of the PXR crystal structure and b) close up view of the LBD showing hydrogen bonding with key amino acids. Figures 5.15 to 5.21 show rifampicin, hyperforin, RU486, ICI 182780, enterolactone, PCB 153, and triclosan, docked in the LBD of the PXR crystal structure.

212 Figure 5.9a Physico-chemical PXR model

METPXR_testset1 .M5 (PLS), physchem model reduced YPred[Comp. 1](YVarmean % max induction)/YVar(YVar mean % max Induction)

A.31 y=1*x-1.354e-006 R 2 =0.7696 o 1 5 0 -

A33 A12 g 100- A4 25 coC < D & 5 0 - I A28 >- A3 A18 ^14

10 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190

YPred[1](mean % max induction)

RMSEE = 25.0617

Figure 5.9b Physico-chemical coefficients plot PXR model

METPXR_testset1.M5 (PLS), physchem mode! reduced CoeffCS[Comp. 1](YVar mean % max induction)

0.140 0.120 0.100

0.060

0.020 0.000 -0.020 0.040 -0.060

Var D (Primary)

213 Figure 5.9c Physico-chemical variable importance plot PXR model

M ETPXRjestsetl,M 5 (PLS), physchem modal reduced VIP[Comp, 1]

1 20

1.00

I 0.80

0.60

0.40

0.20

0.00

Var ID (Primary)

Figure 5.10 Volsurf PXR model

METPXR_testset1a.M3 (PLS), volsurf model reduced YPred[Comp. 3](YVar mean % max induction)/YVar(YVar mean % max induction)

A 31

150 - A30 g 100 AS

A 25

50 -

A19

10 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190

YPred[3](mean % max induction)

RMSEE = 22.3724

214 Figure 5.11 Internai validation: Randomisation test for Volsurf PXR model

Internal validation of the VolSurf PXR model provides an estimation of the predictive ability without the help of external data sets, as described in the Cross Validation section 5.2.4.3.4. Here the approach taken involved building models where several objects at a time were kept out of the analysis, and the procedure was repeated until all the objects were kept out once to yield a conservative estimation of the real predictive ability. 25 permutations were conducted. To obtain stable results, the procedure was repeated many times, and the Standard Deviation Error of Prediction (SDEP) was computed to give an estimate of the dispersion of the SDEP values obtained from different runs, and thus the reproducibility between the reduced models. The intercepts of R“ and shown below indicate successful cross validation of the PXR VolSurf model.

Randomisation test R2 Q 2 Intercepts: R2 = (0.0, 0,1 95), Q2 = (0,0, -0.330)

1.00

0 80

0.60

0.40

0.20

0.00

0.20 ia

0.00 0.10 0.20 0.30 0.40 0.50 0 60 0.70 0 80 0.90 1.00

25 permutations 3 components

215 Figures 5,12 and 5.13 below show the variable importance plot and the standardised coefficients plot for the PXR model. Fig 5.12 PXR Variable Importance Plot

1.40

1.20

1.00

g 0.80

^ 0.60

0.40

0.20

0.00 i 9 I 9 S 9 Fig 5.13 Var ID (FYlmary)

PXR Coefficients plot

0 10

0.00

0.10

•0 .2 0

9 9 9 9 9*^ $

V a r ID (Prim ary)

Figure 5.14. The PXR crystal structure with a) SR12813 docked in the LBD, showing the whole protein ribbon backbone of the PXR crystal structure and b) a close up of the LBD showing hydrogen bonding (dashed yellow lines) with key amino acids. (lOuM SRI2813 % max induction 136% relative to lOuM rifampicin). a)

216 Figure 5.15. The PXR crystal structure with a) pregnane dione docked in the LBD and b) rifampicin docked in the LBD. (% max induction 98,8%) a) b)

Figure 5.16. The PXR crystal structure with hyperforin docked in the LBD. (luM hyperforin % max induction 152% relative to lOuM rifampicin).

W'4

Figure 5.17. The PXR crystal structure with a) 4-hydroxytamoxifen docked in the LBD. (% max induction 84% relative to rifampicin) b) RU846 docked in the LBD. a) b)

217 Figure 5.18. The PXR crystal structure with phénobarbital docked in the LBD.

Figure 5.19. The PXR crystal structure with a) secoisoresorciresinol (% max induction 20% relative to rifampicin) and b) enterolactone docked in the LBD. (% max induction 36 % relative to rifampicin) a)

Figure 5.20. The PXR crystal structure with a) PCB 153 and PCB 118 docked in the LBD. (% max induction 16% relative to rifampicin) and b) PCB 153 in a different orientation, in the LBD a) b)

218 Figure 5.21, The PXR crystal structure with dexamethasone docked in the LBD. (% max induction 24% relative to rifampicin)

Figure 5.22. The PXR crystal structure with triclosan docked in the LBD. (% max induction 46% relative to rifampicin)

Figure 5.23. The PXR crystal structure with RU 486, SR12813, rifampicin and the other compounds all docked in the LBD, showing the structural space in the LBD cavity of PXR, and hydrogen bonding with key amino acids.

219 Results of training sets with GSK compounds Figures 5.24 and 5.25 show the test set plots. Both sets had a large number of outliers in the lower sections of the plots (e.g. approximate outliers within the red triangles). A blue line is inserted through both plots to give a visual indication of the optimum regression line one would expect to see in a robust test set, as seen for the PXR Volsurf model and internal validation (Figures 5.10 and 5.11), and as shown in figure 5.26, where the outliers from test set 1 (below) are remodelled. As shown, the model could account for approximately 50% of the compounds in the test sets collated from RTP and Ware, but could not account for the subsets. These outliers were removed and modelled separately, and although the R“ was low at 0.4, a trend in the remodelled test set 1 outlier data was apparent, as shown in figure 5. 26.

Figure 5.24. Physico-chemical model: test set 1 (data collated from RTP)

METPXR_testset1 .MS (PLS), physchem model reduced, PS-METPXRJralnset YPredPS[Gomp. 1](YVar mean % max induction)/YVarPS(YVar mean % max induction)

300

A 152 A 70 E 200

A 138^ 50

100 182 A 52 ' A 1 4 2 ^ 1 % A 53, |?2Aia6iaMa^/

116 37. A 183 66-

0 100 200 300

YPredPS[1](mean % max induction)

RMSEP = 9 4 . 9 5 3 2

220 Figure 5.25. Physico-chemical PXR model: Test set 2 (general outliers indicated within the red triangles) (Ware data set)

METPXR_testset1.M5 (PLS), physchem model reduced, PS-PennyvalidationPXRtestsetG 1202 YPredPS[Comp. 1](YVar mean % max induction)/YVarPS(YVar mean % max induction)

A22

,217 220 A 222.

I I 160 E

i ’ A 213 A 200 II 100 A 203 I

A 245 40 A 252 A 1 9 3 20 - A2a6L_ !9A230A*30B43-

10 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 210 220 230 240 250 260

YPredPS[1](mean % max induction)

RMSEP - 63.2238

Figure 5.26. Physico-chemical PXR outliers from test set 1, replotted

outliers from rtpset predicted using METPXR_testset1_M3.fVi2 (PLS), reduced version of Ml YPred[Comp. 1 ](YVar mean % max induction)/YVar(YVar mean % max induction)

y=1*x+2.053e-006 A 37 R2=0.4421 A 22

40 - A 43 A 35 I I A2 E A 42

i A 39

A 17 A 28 AiS

A 3

0 10 20 30 40 50 YPred(1](mean % max induction)

RMSEE = 11.845

221 Figure 5.27. Variable importance plot for the Physico-chemical PXR outliers from test set 1

outliers from rtpset predicted using M E TP XR_tests et 1 _M 3. M 2 (PLS), Untitled CoeffCS[Com p 1](YVarmean % max induction)

.1 0.120 -- 0.100 2 0.080 -

E 0.020 « 0.000 E -0 .0 2 0

0.060 2

0.080 2 0.1 00 ■' 0.1 20 £

Var ID (Primary)

5.4.5. Discussion The key variables for the original PXR QSAR model were: Amide responsive regions (WAM8 and WAMl): the molecular envelope accessible by an amide probe. The volume of this envelope varies with the level of interaction energies between the amide (NH) probe and the solute molecule. Due to the lone pair of electrons on the nitrogen, the amide probe provides specific information that good acceptor abilities increase the potency of a high affinity PXR ligands. WAM8 accounts for hydrogen bond acceptor regions, WAMl for dispersion forces. These descriptors are closely related to W1-W8 (resulting from a water probe), where W1 accounts for polarisability and dispersion forces, and W8 for polar and hydrogen bond donor-acceptor regions. Local Interaction Energy Minima Distance (D23): the distances between the best three local minima of interaction energies between a water probe and the target molecule, as figuratively illustrated in figure 5.28. Figure 5. 14b, showing SRI2813 docked in the PXR LBD, illustrates the optimum triangulation shape for hydrogen bonding with three amino acids in the PXR LBD.

222 Figure 5.28. Simple representation of the Local Interaction Energy Minima Distance (D23)

Key: Each side of the triangle represents the distance between the three best local minima of interaction molecule energies represented by the ellipses.

Hydrogen Bonding descriptors (HE 1-8): represent details about the hydrogen bonding capabilities of the ligands. Hydrogen (H) bonding is a common feature in NR studies, and also features in transporter studies, e.g. Pgp substrate studies (Penzotti et al.,

2002). Abraham et al (2002a) have examined H bonding with respect to hexachlorohexanes, and suggest that for this class of compounds the hydrogen bond basicity of polychlorinated compounds, even those with no functional H bond group, can be explained by the presence of several non-interacting aliphatic chlorosubstituents that each contribute to the overall hydrogen bond basicity (B). While this finding can be applied to the ligand binding activity of such chlorinated compounds in the ERa and AhR binding domains, the principle is also applicable to the PXR QSAR study with respect to the dominance of the D23 descriptors which represent the distances between the best three local minima of interaction energies when a water probe interacts with a target molecule. The less important, but relevant, variables included Molecular Surface (S), Molecular Volume (V), Polarisability (POL), Hydrophobic Integy Moments (ID 1-IDS), Capacity Factors (Cwl-CwS), Amphiphilic Moment (A) and Molecular Globularity (G). This model was and is being tested further, utilising in house GSK proprietary % mean induction PXR data, generated from the same assay procedure and HuH7 cell line.

223 Two test data sets were collected, one collated from data generated and posted by operators at the GSK RTP site in North Carolina, US; the other from validation data generated at the GSK Ware site in the UK. In both cases only data generated within the last two years was utilised, and only data from experimental runs reporting acceptable values for the positive control rifampicin (within 80-120%) and compound data that compared favourably with in house validated data sets. The compounds in these data sets cannot be identified by name or structure due to commercial confidentiality. The models generated in both cases had a large number of outliers in the lower quadrant of the model, see figure 5.11. These models are based on the Abraham and other physicochemical descriptors, as some of the compounds were too large to generate VolSurf descriptors. A preliminary analysis showed that the compounds with relatively large MW (larger than rifampicin), and fatty acid-like in structure, with extended molecular shapes, were the outliers. When removed from the model and remodelled separately these outliers did show a linear trend, but the correlation was not significant (R^=0.4), figure 5.26. This result could indicate that a bimodal form of ligand binding may occur in the PXR LBD, such that this group of structurally similar compounds binds with an amino acid in another part of the LBD, as opposed to the amino acids that are required for binding as shown in the PXR crystal structure figures (5.27). The following physicochemical properties dominate the outliers model (n=31, taken from the RTP test set, n=149), in descending order: Positive associations: Calculated log P (clogp), trifluoromethylbenzene descriptor (proprietary to GSK), flexibility (related to the number of rotatable bonds to total bonds), Andrew’s binding energy (the functional group contributions to drug- receptor interactions, (Andrews et al., 1984), Number of heavy atoms in the molecule, MW, CMR (a size descriptor), Vx (a molecular descriptor relating to McGowan characteristic volume), two in-house (and proprietary) Pgp descriptors, and the count of negatively ionisable groups in a molecule. Although not dominating the outliers coefficients plot, this last descriptor contrasts with the positively charged descriptors observed in the original PXR physico-chemical model (figure 5.9a). Negative associations were apparent for amine groups (figure 5.27). The descriptors identified differed to those identified for the physicochemical model for the original test. The differences were observed particularly for the amine descriptor and BetaH, a molecular descriptor relating to H-bond basicity and the count of the number of positively ionisable/charged groups in a molecule, which all held a positive association in the PXR

224 physicochemical model (figure 5.9c); while phenolic OH groups and Alpha, a molecular descriptor relating to H-bond acidity were both negatively associated. Similar descriptors were also observed and are shown in both plots (figure 5.9c) and figure 5.27, respectively. Other possibilities may be the involvement of transporters, e.g. Pgp as discussed in chapter 4, or something else. If transporters are part of this interaction, the compound would be more likely to have organic acid functions (A. Ayrton pers comm.), but this was not observed when the test set 1 model was screened using Spotfire (v 5.1. www.spotfire.com). the outliers tended to be more basic as identified by the pale blue boxes in figure 5.29.

Figure 5.29. Spotfire plot, PXR test set 2: acid-base class

Scatter Plot

■ ■

50 100 150 M3.YPredPS[3](mean % max induction) Color by b33e_dass ago ■ 1 D2 H4 B |5 ■ 6 □?

5.4.6. Conclusions This investigation has allowed a better understanding of PXR ligands and the following may be concluded: 1. Compounds with high affinity for PXR are likely to have increased hydrogen bond acceptor and donor abilities, have an optimum distance between the three local minima of the interaction energies to fit within the main LBD pocket (e.g. SR 12813 and SR analogues) and may accept binding with additional amino acids in the LBD, with

225 increased surface and volume (e.g. hyperforin, 40H tamoxifen) and/or enlarge the binding pocket further to accommodate (e.g. rifampicin and large fatty acid-like structures). Furthermore, the hydroxylated forms of compounds are likely to be more potent than the parent compound (e.g. enterolactone, OH vitamin D 3,4 OH tamoxifen). Lignans present in the human diets can be metabolised to weak-medium affinity ligands for PXR. Common food sources such as rye, whole grains, flaxseed and berries provide the parent lignans from which enterolactone is fermented by colonic bacteria. Relatively high intestinal levels of enterolactone are likely to be present in populations with high dietary intakes of these foods (Heinonen et al., 2001), for example Northern and Eastern European populations. 2. Persistent organic pollutants, such as PCB118 and PCB158 are weak ligands, but may well become more potent when hydroxylated as observed in the ER. The pesticide trans-nonachlor is a medium affinity ligand, as is triclosan, a commonly used bacteriocide. From the test set studies conducted with in-house GSK NCEs, it is apparent that there may be another form of binding occurring in PXR for compounds with long expanded side-chains. This was observed particularly for NCEs designed to be PPAR ligands, and therefore fatty acid like in structure. The possibility of a bimodal form of ligand binding in PXR is supported by the compound docking studies reported here and PXR crystal structure studies, upon which they are based, both indicate an expandable LBD. While descriptors relating to ClogP, MW, and others, featured in both the original physicochemical coefficients plot and the later outliers plot, certain physicochemical descriptors differed between the PXR model and outliers trend model, particularly the amine descriptor, and positively and negative charged groups. Being mainly fatty acid like in structure, these outliers also have a greater amount of hydrophilic regions per surface unit. That different classes of compounds bind differently within the same receptor is known for other NRs, such as the ER for instance. These results are being utilised in ongoing comparative QSAR receptor studies examining crosstalk and ligand promiscuity. With sufficient resources, future QSAR studies of PXR could ideally use statistical experimental design to generate well balanced training and validation data sets of representative PXR ligands.

226 5.5. Multivariate (QSAR) Modelling Of Polybrominated Diphenyl Ethers Activity In Different Species Cell Lines And Their Capacity To Induce CYPIA By The Ah Receptor Mediated Pathway

5.5.1. Introduction The widely used flame retardants polybrominated diphenyl ethers (PBDEs) are persistent and ubiquitous organic pollutants (POPs) that biomagnify and may have endocrine disrupting effects similar those observed for other halogenated compounds, such as polychlorinated and polybrominated biphenyls (PCBs and PBBs), polychlorinated dibenzo-p-dioxins (PCDDs) and polychlorinated dibenzofurans (PCDFs). Concern with the risks to human health, particularly among infants, is increasing due to observations of increasing PBDE concentrations in human breast milk (Meironyté Guvenius et al., 2003), although available data suggest that current levels of PBDEs are an order of magnitude lower than those of PCBs (Meironyté Guvenius et al., 2003). Levels are reported to be lower in European populations (Meironyté Guvenius et al., 2001), compared to those reported for California, where body burdens of BDE 47 were 3-10 times higher, ranging between 5 and 510 ng/g lipid (Petreas et al., 2003). Exposure occurs occupationally (Eriksson et al., 2001b; Rahman et al., 2003), environmentally (Boon et al., 2002) and in the diet (Bocio et al., 2003; Damerud et al., 2001), as elaborated upon in section 2.2. While PBDEs are in extensive production and use, existing data on the receptor mediated toxicology of PBDEs are very limited, but there are indications of toxicity via the estrogen, thyroid and Ah receptors (AhR) and oestrogen sulphotransferase inhibition (Bergman, 2000). In vivo evidence suggests that PBDEs interfere with the thyroid hormone system, via upregulation of uridinediphosphate- glucuronosyltransferase (UDPGT) by 3 to 4 fold. The acute toxicity of mixtures appears to induce porphyrinicity, and increase hepatic enzymes such as ethoxy- and pentoxy- resorufm-D-deethylase (EROD and PROD) levels between 10-40 fold (Zhou et al., 2001). BDE 47 and BDE 99 are reported to cause learning and memory developmental defects, as observed with PCBs (Eriksson et al. 2001b), and so PBDEs are considered a potential risk for neurobehavioural development in humans (Branchi et al., 2003). In cell signaling, an in vitro mechanism of toxicity has been observed by way of increasing Ca^^ oscillations stimulating greater release of arachidonic acid (AA), (Branchi et al., 2003). As AA is a ligand for the PPARs, and is implicated in autoimmune dysfunction.

227 the health impacts of PBDE exposure could be further implicated in the destabilizing of NR and PPAR molecular pathways, through multiple mechanisms. Multivariate techniques are useful for data analyses of selected compounds, tested in a broad battery of test systems where there is a large variation within some of the variables. Here the AhR QSAR data are suitable for activity estimation, and they can provide warnings/alerts about possible toxic properties of the test compounds. QSARs can be used, therefore, for decision support in early phases of product development to regulatory decision-making such as in risk assessment and risk classification. This study generates a principal component analysis (PCA) and quantitative structural activity relationship (QSAR) for the AhR, using multivariate techniques, specific descriptors, and biological data sets. This QSAR utilized experimentally generated PBDE data sets (Chen et al., 2001), together with an unpublished data set from the same laboratory, for twelve PBDEs capacities to bind and activate the AhR in rat and human hepatocytes, human intestinal cells and to induce CYPIA, which here was assayed as 7-ethoxyresorufin-O-deethylase (EROD) activity in cells from rainbow trout, chick, rat and human. VolSurf and physico-chemical descriptors were generated for each of the compound structures for which biological data were available and subsequently used to develop a partial least squares regression based model. 2,3,7,8-TCDD was an outlier compared to the PBDEs, and major differences were apparent between the different cell line ECso’s (n=7) reflecting PBDE ligand potencies in AhR, and cell line EROD % values (n=7) reflecting CYPIA induction. These multivariate analyses are supported by utilising a homology model of AhR

(based on the ERa crystal structure), see chapter 3, to examine the mode of binding of representative compounds in the ligand-binding domain (LBD) of the receptor model. Key commonalities and differences between the ligands when bound in the LBD have been observed and are discussed in chapter 3.

5.5.2. Methods

5.5.2.7. Generation of PBDE biological data using in vitro assays. The following PBDEs (lUPAC No’s 28, 47, 66, 77, 85, 99, 100,119, 126, 153, 154, 183) were tested relative to 2,3,7,8 TCDD in the following cell lines: RTL-Wl: rainbow trout hepatoma; CEH: primary chick embryo hepatocytes; PRH: primary rat hepatocytes; H41IE; rat hepatoma; Caco-2; human adenocarcinoma; Hep G2; human

228 hepatoma, from which EC50’s and EROD%, representing the highest EROD activities from individual congeners compared with the activity of the positive control InM TCDD. Methodologies are described in Chen et al. (2001b).

5.522. Statistical analysis for the evaluation and interpretation of the data To enable the evaluation and interpretation of these multivariate data PCA and PLS multivariate techniques (Umetrics, SIMCA-P v.10.0) and VolSurf descriptors were used, as described in section 5. 2.

5.5.2.3. Utilization of a homology model of AhR based on the ERa crystal structure

The homology model of AhR based on the ERa crystal structure was used for docking studies of PBDE, TCDD and PCB ligands in the hAhR model, using SYBYL biopolymer software and visualised on a Silicon graphics Indigo^ IMPACT 10000 Unix work station (Tripos Assocs. St Louis, MO), see Chapter 3.

5.5.3. Results and Discussion

The data sets were analyzed using PCA and PLS techniques. Initial analysis suggested the presence of an outlier, (2,3,7,8-TCDD), which was subsequently excluded in the 3-component model shown. In this way the PLS model R2X value improved from 0.4 to 0.77 (n=12, RMSEE=15.71), with a good fit (R2Y=0.71) but poor predictability (Q2=0.16) as seen in the PLS primary variables plot (fig 5.30). The summary plot of the model fit, (fig 5.31) and the loadings plot (fig 5.32) show the success of the model fit for each of the Y responses. Specific descriptors were positively correlated with the biological responses in the top left box, and negatively correlated with the bottom right box.

229 Figure 5. 30. Primary variables plot

Primary variables plot M3 (PLS), ■ R2VY[3](cum) R2=estimate of goodness of fit H Q2VY[3](cum) Q2=estimate of goodness of prediction

0.60 -

Var ID (Primary)

Figure 5. 31. Loadings scatter plot

Loadings scatter plot: PBDE cell line biological profiles. M3 (PLS), w*c[Comp 1]/w*c[Comp. 2]

12 0.50 Wolsurf-IDivolsurf-D ■ H4IIE ECSOnM 0.40 iHep-G2EC5*rTOHEC5pnM . ■ CBH EC5Ô nM

0.30 ■ PRH AVolsurF.H@K EROD iC aco EG MBU-WI ECSOnM 0,20 ■ CBW^WSSa EROD , Caco p - 0,10 -HL1 «RTL-Wl mOD% AVolsurf-Bnlnl AVol s AVolsurf-WAMS 0.00 iVolsurf-W/WG

- 0.10 AVolsurt-Bnii2 AVolsurf-W6 - 0,20

AVolsurf-Hpi -0.30 AVolsurf-Cw6 AVolsurf-Emin3 AVolsurf-HBS

0.00 0,20 w*c[1]

230 Taking the best-predicted biological parameter, the human hepatocyte cell line Hep G2 EC50, the following Volsurf variable importance (fig 3) and coefficients (fig 4) plots were derived. Figure 5.32. Variable Importance plot

Variable importance plot.MS (PLS), VIP[Comp. 3]

1 1 1 1 2 % 1> s. Var ID (Primary)

Figure 5.33. Coefficients plot: Hep G2 EC50

Coefficients plot.M3 (PLS), CoeflCS[Comp. 3](YVar Hep-G2 ECSOnM)

c 0.30 - o 0 . 20 -

0. 10 - 9 Q.

(3 "D CM CO CM CD p - CD c C c Q Q _ i CO m m Û X O < < X XX (D S t: ■t: ' îi 1 1 3 3 3 3 3 3 SD. =j m m C/) cn O o SÙ 3 3 3 JÛ o o s n 3 (A « m o o 3 OO o CO > > o o > > > > s n SD. (I) > > OO o o o > > > Q. > > > > > Var ID (Primary)

23 The key variable was the local interaction energy minima (Emin3) which had a large negative correlation. This represents the energy of interaction in kcal/mol of the best three local minima of interaction energies between a water probe and the target molecule. Positive correlations were observed for: Local interaction energy minima distances (D12), these are shape descriptors for the distances between the best three local minima of interaction energies when a water probe interacts with a target molecule; Hydrophobic integy moments (ID7 and IDS) which measure the imbalance between the centre of mass of a molecule and the position of the hydrophobic regions around them, and hydrogen bonding (HB7) capabilities of the ligands. Less important negatively correlated variables were; capacity factors (Cw5 and Cw6), which represent the amount of hydrophilic regions per surface unit, and amide responsive regions (WAM7), suggesting that good acceptor abilities may not increase the potency of high affinity AhR ligands. WAM7 accounts for hydrogen bond acceptor regions, WAMl for dispersion forces. These descriptors are closely related to the hydrophilic regions (W5 and W6) where, in the Volsurf hydrophilic scale, W1 accounts for polarisability and dispersion forces, and W8 for polar and hydrogen bond donor-acceptor regions. Both in vitro and in silico docking studies suggest different modes of binding in the AhR for the planar TCDD compared to other ligands, and this is reflected in the key VolSurf variables observed from the QSAR exercise. 2,3,7,8-TCDD appears to bind in the AhR more potently by hydrogen bonding (Chapter 3, section 3.4 and figure 3.11), while other less potent ligands such as PCBs might ligand bind through n-n stacking (figure 3.12) or just one hydrogen bond with an arginine residue, as suggested from docking PBDE 47, a non-planar compound in the LBD (figure 5.34). Both possible modes of binding with specific key amino acid residues triggering different conformational shapes of the receptor-ligand binding sites and energies to that of 2,3,7,8- TCDD (Jacobs et al., 2003), that have also been termed electronic elasticity-toxicity relationships for the PCDD congeners (Asatryan et al., 2002). The importance of the electrostatic effects at some substituent positions of halogenated PCDD and PCB congeners has been shown in other QSAR studies using MLR (Lewis and Jacobs, 1999) and PLS (Oprea et al., 2001).

232 Figure 5.34. 2,2’4,4’ TBDE (PBDE 47) shown in two different perspectives, docked in the LBD of the AhR in silico model based upon the ERa crystal structure.

Hydrogen bonding with the oxygen (in red) is indicated by the dashed orange line.

In this study, the only outlier was 2,3,7,8, TCDD, which does exhibit different structural characteristics from the rest of the PBDE series. In PLS models it has been often observed that the outlier molecule has different vertex occupancy compared to the remaining majority. These structural differences are of real importance for ligand- receptor interactions, so such outliers have been termed ‘artificial’, since they are likely to have similar and additional binding mechanisms, but are detected as belonging to a different cluster (Oprea et al., 2001).

233 5.5.4. Conclusions There are many considerations for cell lines of choice when conducting in vitro assay screening, including tissue, ease of maintenance and use, receptor and transporters expression. In this study, the human Hep G2 cell line was the best for predicting receptor binding and CYPlAl induction for PBDEs. Analyses of the Volsurf variable contributions together with observed in silico modes of ligand binding in the AhR model allow additional insights into receptor activation on a compound and chemical family specific basis, as well as ligand promiscuity and ligand-receptor cross talk when compared to other receptor QSAR studies. This is of relevance in the consideration of AhR and CYPIA mediated xenobiotic species-specific interactions, and a better understanding of the molecular mechanisms of AhR mediated PBDE pollutant toxicity and endocrine disruption. It would also be useful to conduct similar QSARs with the PBDE data generated from, for example, ER activation assays (Meerts et al., 2001), particularly in view of the known cross talk between the ER and AhR (Chapters 1 and 3).

5.6. Overall discussion and conclusions QSARs are used extensively by the pharmaceutical industry, for prediction of drug metabolism, but for many applications the rules are not very well informed by the biology of the processes involved, and it is not possible to make effective literature comparisons between studies of receptors for example, because diffferent drug properties (descriptors) are used from one study to another. Apart from the basic physico-chemical descriptors, this makes it hard to directly compare the results for different processes and thus uncover possible models for biological processes, such as NR cross talk. Here 1 have made a start at examining NR metabolism by attempting a thorough review of the molecular to epidemiological biology of the processes involved, and then utilized the same body of descriptors, VolSurf descriptors, to compare a number of biological and physicochemical processes. The methodology was successfully validated using data from the literature, in house, and by conducting internal validations. There are key differences in descriptors for the different receptors, with respect to different ligands. Sometimes the compound combinations might have additive effects (Silva et al., 2002) in defined NR systems and sometimes they might negate each other in a specific NR system, but have effects on other receptors, that are part of the communication system. A good example of this can be seen in the varied literature

234 reports of ligand interactions between the ER and AhR, the steering role of PXR and CAR (Honkakoski et al., 2003) and other interactions discussed in chapters 1 and 3. The key Volsurf descriptors for each receptor study in this chapter are summarized in descending order of significance, in Table 5.3. The Volsurf chemical information is distributed differently for each receptor studied, and key differences between receptor-ligand activation demonstrate the complexity and concerted defense mechanisms available at the molecular level. The varieties of receptor-ligand binding interaction clearance strategies provide an intricate back-up homeostatic system to clear tissues of xenobiotics that disturb the homeostatic equilibrium. Similar studies in the same and other NRs, including AhR, are currently underway, which can add to this information, and so give indications of which compounds activate which receptors, and so inform as to which receptor-mediated pathways are likely to be stimulated by a compound or given class of compounds. For all the receptors studied, positive correlations were observed for hydrogen bonding (HBl-8), and negative correlations were observed for capacity factors (Cwl-8), which describe the amount of hydrophilic regions per surface unit. Apart from these commonalities, the QSARs presented here suggest that for the ERa, rugosity (R), hydrophobic regions (D1-D8), polarisability (POL), amphilic moment (A) and globularity (G) were positively correlated with ligand binding, while this was not the case for PXR, where the most important descriptor was WAM8. There are many other QSAR analytical rule-based techniques that these studies could be incorporated into, such as feature trees (Rarey and Dixon, 1998), decision trees (Bursi et al., 2001) hierarchical cluster analysis (Suzuki et al., 2001), chemical global positioning mapping devices (Oprea et al., 2002) and neural networks (Bayada et al., 1999; Vendrame et al., 1999; Huuskonen, 2003). Ideally one should also incorporate the water molecules present in the LBD when conducting 3-D QSARs, as water present in the LBD has been recognized to play a major role in ligand-protein interactions (Pastor et al., 1997).

235 Table 5.3. Summary of VolSurf descriptor data generated from the NR QSAR studies described in this chapter.

Key VolSurf Other key Typical Biological Cell line Test Biological da d Descriptors* known LBD parameter ligands Reference 1 NR Correlation descriptors features of 1 +ve -ve ligands 5.2 ERa R, D1-D8, wl-8, H bonding 6Â Uterine ER Rat uterine Broad Fang et al., 2001 HBl-8, Cwl-Cw8, biopbore/pbeno ligand binding cytosol ER range of POL, A, HLLHL2 lie A ring competitive chemical G structure binding classes assay 5.3 PXR WAM8- IDl-8, H bonding 1.H bond % maximum HUH7 Broad Chapter 3 WAMl, HB7, HBl- acceptors induction transient range of D23, S, G, 2, Cwl- 2. an optimum relative to transfection chemical V,POL Cw8, distance rifampicin classes A between 3 local minima of interaction energies 5.4 AhR D12, ID8, Emin3, H bonding 1. Planar PAH AbR ligand HepG2 PBDEs Chen et al., 20011 HB7 Cwl-8, K-n stacking structure binding and Chen, HB6, HBl, 2. Non-planar unpublished data WAM7 PAH

* Symbols are explained in the text and in Appendix 3 Table I: Summary of VolSurf descriptors

There are also many other descriptor methods available that relate well with the Volsurf and physico-chemical descriptors used here, such as the minimum topological difference method (MTD), which also aims at describing steric, hydrophobic, electrostatic and hydrogen bonding characteristics of the binding site regions in the vicinity of the ligand atom positions (Kurunczi et al, 2002), and molecular fingerprinting descriptors (Bayada et al., 1999). By looking at NR QSARs comparatively, together with homology modelling, crystal structures and multiple molecular dynamics simulations for different NRs as described in this thesis and other related NRs as reported in the literature, e.g. the AR (Marhefka et al., 2001) further insights and greater interpretability into receptor-ligand dynamics can be achieved.

236 C hapter 6

Nuclear Receptors and Ligand-Gene Interactions Implications For Human Health

237 Chapter 6 Nuclear Receptors and Ligand-Gene Interactions Implications For Human Health

‘The flaws of nature may turn out to be as significant as the laws o f nature for our understanding of true realityd John Barrow (2003)

6.1. Nuclear receptors and human health

6.1.1. Summary The studies presented in the preceding chapters examine molecular gene-ligand interactions with some of the many receptors involved with development and xenobiotic metabolism. The activation of nuclear receptors by metabolites underlies most of the genetic responses controlling fat, glucose, cholesterol, bile acid and xenobiotic/endobiotic metabolism. Steroid, heme, and xenobiotic metabolism are controlled by a network of interrelated but unique nuclear receptors. The ER, AhR, PXR and CAR receptors, (Figure 6.1. gives an overview) mediate the steroidogenic pathway shown in figure 6.2. PFARa regulates the catabolism of fatty acids and impacts upon several steps of the reverse cholesterol pathway, and several more receptors, such as LXR, FXR and HNF-4, are part of the endobiotic bile regulatory pathways. The combined effects of the ligands for the nuclear receptors are vast, influencing virtually every fundamental biological process, from development and homeostasis, to proliferation and differentiation. The supply of ligands is a further regulatory route for the receptors. The cognate ligand for each receptor is usually dependent on the supply of a specific dietary component for its synthesis, e.g. iodine and tyrosine for thyroid hormone, cholesterol for the VDR, GR and receptor mediated steroidogenic pathway, fatty acids for the PPARs, and vitamin A, as the ligand for all the trans-VQtinoic acid receptors (including the receptor heterodimerization partner RXRa). The receptors studied in this thesis all have diverse modes of binding and activation mechanisms, but are part of an integrated, highly elaborate and complex system that has adaptively evolved in response to challenges to the endocrine system, breakdown of heme metabolites and exposure to xenobiotics and foods. Figure 6.3 gives an indication of the complexity of the communication web between the receptors studied here and selected cofactors reviewed from the literature. The adaptive nature of this system allows for flexibility and has given rise to the ‘evolution of novelty’ in hormone-

238 receptor systems (Thornton, 2001), to ensure optimum endobiotic management and

xenobiotic response. Figure 6.1 An overview of xenobiotics as modulators of nuclear receptor function (Adapted from Carpenter et al., 2002)

Xenobiotics

Hormone dependent action Hormone independent action

Binding to hormone receptor agonist/antagonist Binding to xenobiotic receptors

( e.g. ER, AR, PR, PXR, CAR) (e.g. AhR, PXR, CAR)

Binding to plasma transport proteins (eg PgP) Binding to DNA

Alteration of biosynthesis/metabolism Generation of ROS

Tor i availability of level of hormone &/or receptor Regulation of P450s Oxidative damage

Mutations

DNA adducts Stimulation or inhibition of Cellular effects N ecrosis proliferation

Cell cycle arrest

A popstosis

Figure 6.2. The key known receptors and P450s involved in the steroidogenic pathway

-► Fatty acids & eicosanoids Acetyl CoA PPAR RXR Iso p ren es I RXR RXR' B-carotene Squalene -4- LXR All-frans-reti noic RAR 24(5)25 epoxycholesterol acid I FXR ► Cholesterol*— VDR

chenodoxycholic acid CYP1 A1

Pregnenolone *- GR, PXR, 1 CAR

Progesterone ^ GR, P)^R, CAR, PR C Y P1 liAAi l

I CYP2l)bik Corticosteroids, aldosterone

Sex steroids: CAR, AR -► Androstendione testosterone* AR CYP|9A1 CYP1|A1

Oestrone oestradiol,^ ___ E R s

239 TCDD, PAH, PCB, indoles VEGF I Ah gene battery GLUT I Bax p27 \

bHLH-PAS proteins HIF-1a ipoptosis

ARNT A poptosis? AhR Xenobiotic m etabolism

isp9( NF kB

ER a arrest P hase II ERp Soul c IE . Phase I RB P450S

PXR RXR PPARS CAR Nrf2 GR P P A % f PPAR h

Retinoids HO-1 m etals Steroid horm ones, glucocorticoids, bile acids, drugs, NQ01 PCBs, pesticides, phytoestrogens, phytochemicals Fatty acids

Figure 6.3. Functionally significant cross-talk pathways between nuclear receptors and selected cofactors (Adapted from Carpenter et al., 2002) Abbreviations not included in the text: Glut-1, glucose transporter 1; HIFa, Hypoxia Inducible Factora;Nrf2, nuclear factor drythrod 2-related factor 2; RB, retinoblastoma protein; VEGF, vascular endothelial growth factor

6.1.2. The development of steroid receptors, an evolutionary perspective The oestrogen receptor (ER) was the starting point for this study as not only was the crystal structure available for homology modelling purposes, but it was also one of the most studied of all the NRs. Indeed, Joseph Thornton (2001) has argued convincingly that all the steroid receptors have evolved in vertebrates from an ancestral ER. Based on phylogenetic analyses, sequence reconstruction and knowledge of functional attributes of ancestral proteins, he has shown that the first steroid receptor was an oestrogen receptor, followed by a progesterone receptor. He suggests that NRs have developed through ligand exploitation and serial genome expansions. The full complement of present day mammalian steroid receptors evolved from these ancient receptors by two large-scale genomic expansions, before and after the advent of jawed vertebrates about 450 million years ago (Sharman and Holland 1998, cited in Thornton, 2001). Specific regulation of physiological processes by androgens and corticoids emerged after these events. Thornton suggests that the terminal ligand is the first for which a receptor evolves;

240 selection for this hormone also selects for the synthesis of intermediates despite the absence of receptors, and duplicated receptors then evolve affinity for these substances. This might even be a pre-emptive strategy for vertebrates in combating plant defenses (Wynne-Edwards, 2001). Whilst originally developing to facilitate an increasing complex endocrine system in vertebrates, nuclear receptors have also been required to respond to a wide variety of xenobiotic challenges. They have a well developed role as the sensory component of the defense system that includes regulation of Phase I and Phase II enzymes and transporter proteins in response to xenobiotics. While exposure to the newer pollutants is discussed in this thesis, polyaromatic hydrocarbons (PAHs) exposure has always occurred, as PAHs are the most abundant form of carbon known (Mautner et al., 1997; Hecht, 2002). With UV irradiation, PAHs in ice have been shown to produce ketones, alcohols and quinone- or squalene-like structures (Bernstein et al., 1999). The Ah receptor, a member of a different family to that of the ER, has developmental and adaptive functions, it has similarly undergone duplication and diversification in vertebrates, resulting in three, possibly more, members of an AhR gene family, namely AhRl, AhR2 and AhR repressor (Karchner et al., 2002). The adaptation to binding dioxins and related chemicals was probably a vertebrate innovation (Hahn, 2002a), that may have been part of the response to PAH exposure, co-evolving with the ER, such that ER signaling can be modulated by association with activated AhR (Ohtake et al., 2003), but also ER and other NRs (Jana et al., 1999) may display activation by AhR ligands. Hydrogen bonding properties of the ligands have been a key part of the evolutionary development of NRs. Hydrogen bonding has long been recognized as an important type of molecular interaction that is ubiquitous in nature, being found in systems ranging from water to cellular proteins and DNA (Pimentel and McClellan, 1960; Del Bene, 1998). The variability in forms of hydrogen bonding (Abraham et al., 2002a; 2002b) give rise to an even more elaborate communication system, where it may only be a difference between a few key residues in the ligand binding domain (LBD) that produces species variability and within-species receptor differences in response.

241 6.1.3. Receptor-ligand interactions Receptor-ligand interactions have been examined for selected receptors in Chapters 3 and 5. This work adds to and supports the literature on ligand-NR interactions. Ligands bind with, activate, or suppress receptors in a wide variety of ways, but conformational change that leads to DNA interaction is not always induced. For instance, recent work by Henry and Gasiewicz (Henry and Gasiewicz, 2003) shows that agonists, but not antagonists (such as dietary and synthetic flavones), cause a conformational change to the DRE-binding conformation in vitro. As mentioned in Chapters 3 and 4, the mutation of one key residue in the AR LBD ablates the ligand binding of testosterone (Rabenoelina et al., 2002), while the mutation of four specific LBD residues (or two different residues in mouse and human each) causes a species specificity change in PXR (Watkins et al., 2001; Ôstberg et al., 2002).

6.1.4. Role of Protein and Enzyme Coregulators Molecular mechanisms of ligand-receptor binding and gene activation include a large number of modulatory protein transcription factors and transporters as part of the receptor activation process. There is an intricate network of these cofactors embedded into the NR communication web, each one having functionally significant inter-related roles with specific receptors, sometimes several receptors, and other proteins. Many cofactors act in concert with each receptor, to facilitate transport, activation or repression, and understanding of the significance of cofactors has increased dramatically in the last two or three years. Common intermediaries are shared in receptor and other signaling pathways. Some of the cofactors known to have functionally significant roles for the receptors, particularly for PXR, have been identified in Chapters 3 to 5. Receptors are also known to regulate cofactors, thus PXR and CAR not only co-regulate xenobiotic metabolism, but also drug transport (Staudinger et al., 2003). As previously discussed, the key cofactors that are part of the cross-talk pathways between the AhR, ERs, PXR, CAR and PPARs, shown in figure 6.3, emphasizes the influence that cofactors and receptors exert upon each other to maintain equilibrium. Although not discussed further in the context of this thesis, as with the cofactors mentioned below, repressor proteins are also likely to be important cofactors in receptor mediated steroid and xenobiotic metabolism.

242 6.1.4.1. Receptor-receptor and receptor-enzyme interactions Receptor-receptor and receptor-enzyme interactions constitute a very large field of study, and this thesis has attempted to describe some of the interactions between the ERs, AhR, PXR, CAR, the PPARs (to a lesser degree) and the P450’s these receptors induce. Here three further examples are given of the many varied signaling pathways between related NRs. The AhR and ER have a well established intricate relationship that may be further modulated or regulated by COUP-TFl. COUP-TF, known to be part of the CYP3A signaling pathway (Swales, 2002), has been shown by Klinge and colleagues to regulate AhR action in vitro by direct competitive DNA binding and protein-protein interactions (Klinge et al., 2000). Also part of the CYP3A signaling pathway, Drocourt et al (2002) have recently shown in primary human hepatocytes that, together with PXR and CAR, the VDR has a regulatory (but not primary) role, controlling the basal and inducible expression of CYP3A4, CYP2B6 and CYP2C9 (which is primarily a GR responsive gene). In doing so, Drocourt and co-workers suggest that the expression of VDR controlled genes might be affected by xenobiotics, such as rifampicin, through the PXR and CAR pathway. In vivo studies in an immature gonadotrophin-primed female rat model have shown that in ovaries (an oestrogen responsive tissue), COX-2 is induced the morning after ovulation, but greatly inhibited by TCDD via AhR signaling (Mizuyachi et al., 2002). GR was also observed to be down-regulated. Blockage or reduction of COX-2 expression is known to impair ovulation, and there appears to be a feedback loop in the ovary between AhR signaling and reduced COX-2 expression that may depend upon a reduced Leutinising Hormone (LH) surge due to the stimulation of corticoids, via GR autoregulation, and the inhibition of gonadotrophin surges (Mizuyachi et al., 2002).

6.1.4.2. Heterodimerization Partner: R X R a Belonging to the steroid hormone nuclear receptor superfamily, ubiquitous RXRa is the necessary heterodimerization partner for many receptors in order to bind to DNA, including the thyroid hormone receptor, ERs, PXR, CAR (Kakizaki et al., 2002), LXR, FXR, PPARa (Cai et al., 2002) and PPARy (Dubuquoy et al., 2002). The crystal structure data of the PPARy/RXRa heterodimer show the asymmetric conserved heterodimerization interfaces of both receptors (Gampe Jr et al., 2000). RXRa dimerizes through a 40 amino acid sub-region within the LBD known as the identity box. Mutation

243 of two important determinants (Alanine 416 and Arginine 421) within this box has been shown to impair the actions of receptor dimerization partners (Lee et ah, 2000). RXRa is established as a heterodimeric integrator of multiple physiological processes in the liver and is a regulatory component of cholesterol (Edwards and Ericsson, 1999; Russell, 1999; Song and Liao, 2000), fatty acid (Edwards et al., 2002), bile acid (Wang et al., 1999; Makishima et al., 1999), steroid and xenobiotic metabolism and homeostasis (Wan et al., 2000). Suppression of RXRa has a concomitant effect upon the heterodimerization

partner. For example, LXR is reported to inhibit PPARa signaling in this way, in the

nutritional regulation of fatty acid metabolism and PPARa has a counter inhibitory action repressing LXR/RXR binding through the sterol regulatory element binding protein -Ic (SREBP-lc) (Ide et al., 2003; Yoshikawa et al., 2003).

RXRa also has cofactors in common with the NRs, for instance Weibel et al (1999), have shown a competitive element between RIP 140 and SRC-1 in binding to heterodimers of a novel orphan receptor, OR-1 with RXR. RIP 140 is also implicated in the potentiation of EDCs in yeast ER in vitro studies (Sheeler et al., 2000). SRC-1 RXR phosphorylation can also be induced through stress pathway activation (Lee et al., 2000), which would reduce RXR heterodimerization availability for other receptor partners.

6.1.4.3. Heterodimerization Partner: ARNT The heterodimerization partner of the AhR is ARNT (Elbi et al., 2002) and has been described in Chapter 3. However ARNT is also a potent coactivator of ER dependent transcription that requires the oestradiol activated LBD of ERa or ER(3 (Brunnberg et al., 2003), but unlike the AhR does not also interact directly with orphan receptor COUP-TF (Klinge et al., 2000).

6.1.4.4. Chaperone proteins Heat shock proteins such as hsp90, discussed in Chapter 1, can be shared as a chaperone factor, as seen with the ER and AhR but AhR activation is possible without the presence of hsp90 (Phelan et al., 1998).

244 6.I.4.5. Selected examples o f other activators and cellular responses Chromatin remodelling complexes have been noted to be coactivators of receptors, thus Brahma/SWI2-related Gene 1 Protein (BRG-1) is part of the functional signaling pathway of the AhR/ARNT complex that mediates CYPlAl transcription (Wang and Hankinson, 2002). GR associated proteins are known to interact with PXR and CAR. Glucocorticoid receptor interacting protein-1 (GRIPl), for instance mediates ligand -independent nuclear translocation and activation of CAR in a mouse model (Min et al., 2002). A CAR and PXR response element has been identified in the promoter of the human inducible nitric oxide synthase gene (iNOS). Since iNOS-induced production of nitric oxide is known to influence inflammation and apopotosis, CAR and PXR regulation of iNOS activity is probably part of the modulation effect of steroids and xenobiotics on these cellular processes (Toell et al., 2002). Cross-talk involving stress signals and AhR mediated gene transcription may determine physiologic responses to xenobiotics via CYPlAl or CYP1A2 induction. Oxidative stress is known to cause a down regulation of the AhR-regulated P450’s CYPlAl and 1A2 (Carpenter et al 2002), but whilst associated with a function of the AhR, mitochondrial reactive oxygen production can also produce these effects independently of CYP lA l and 1A2 (Senft et al., 2002). One of the key sources of oxidative stress is the P450 system, which itself is regulated by these receptors.

6.1.4.6. CYP17A1/CYP19A1 (Aromatase) and cofactors Aromatase (highlighted in Chapter 1) is part of the steroidogenic P450 signaling pathway and is expressed to high degree in or near ER positive breast cancer tumour sites (Brodie et al., 1999; Liao and Dickson, 2002). Aromatase is also stimulated through other biological pathways. For instance elevated levels of the cyclooxygenases COX-1 and COX-2 appear to increase aromatase activity, and thereby increase oestradiol biosynthesis (Richards et al., 2002), as well as increasing prostaglandin secretion, which, by providing an appropriate supply of ligands, may increase activation of PPARs in local tissues.

245 6.1.5. Considerations and conclusions NRs are dynamic protein structures and will adapt in different ways to given situations. Thus the in silico models, the in vitro host cell systems and even the receptor null in vivo studies cannot give precise information on the real life situation: At best, they can only suggest possible molecular mechanisms of activation, predict cellular mechanisms in vitro and indicate the possible physiological pathways, in vivo. Other receptors, signaling proteins and transporters will expand their metabolic activities to fulfill the role required of the missing receptor, to adapt and optimize the receptor signaling pathways in living systems. However, as described in this thesis, the utilisation of in silico and in vitro tools can help tease out our understanding of how the receptors function as part of hormone dynamics and xenobiotic metabolism. Applying these techniques to an extended range of receptors, and new crystal structures as they become available is a primary avenue of future research. There are three key strands running through this thesis that such work can be applied to. These include applications for the pharmaceutical industry, applications for environmental health and dietary exposure to chemical contaminants, and screening for potential EDCs. A brief overview of possible applications and further research arising from this thesis follows.

6.2. Future work

6.2.1. Future research with respect to contaminants in the aquatic food chain

As discussed in Chapter 2, high persistent organic pollutants (POPs) levels are known to occur in fish, when compared to other food groups. The difference in these levels is species and fat content dependent. Therefore, certain oily fish are known to be a potential health risk, and dietary restrictions and maximum limits have been introduced in the European Union (EU). In recent years, farmed fish products have become increasingly available to the European consumer, and the farmed fish industry, particularly salmonid, has grown significantly in Europe. While there is extensive research documenting the bioaccumulation of POPs in the marine environment, relatively little data have been available on contamination pathways in aquaculture systems, especially for farmed salmon. Raw aquaculture feed materials can greatly influence the fish body burden of POPs, and be significant sources

246 of POPs in the feed. High POP levels may cause negative health effects in consumers of such fish, as well as in the fish. The Baltic sea fish caught in Finland and Sweden have specific EU market exceptions, which are due to be revised by the EU at the end of 2006. However, with reports in 2003 that there are no spawning cod in the Baltic (only fish less than two years old), together with a similar situation for the cod in the Atlantic, the consumption of salmonids from the Baltic (which remain relatively plentiful), for communities bordering the Baltic, is likely to increase. Due to the combination of the contamination situation in the Baltic Sea and a greater fat content than non-oily fish, Baltic salmonids have high reported levels of dioxins and PCBs. Ideally, the reduction of contaminant body burdens in Baltic fish need to be assessed and developed (as described in Gobas et al., 1999, for example). The data available on POPs and other contaminants in oily fish and fish products entering the European markets need to be improved. Risk assessment-based new methods for pre-screening important POP chemicals can be developed, based on in silico modelling techniques combining NR crystal structures and compound QSARs. This information could then be part of consumer information strategies with respect to the risks and benefits associated with eating (oily) farmed or wild fish.

6.2.1.1. Future research: A European Union policy perspective Due to the concern of human health risks, the EU has introduced maximum limits (MLs) for dioxins. The Baltic region is of special concern; Sweden and Finland have obtained exceptions from the ML on Baltic fish intended for consumption within respective countries until 2006. A condition for this exception is that these countries effectively inform the consumers about the possible risk, and give recommendations for restrictions on intake. Other Baltic countries are likely to apply for similar exceptions once they become members of the EU (in 2004). Information on consumers’ eating habits (with particular reference to girls and women of child bearing ages), and risk attitudes regarding oily fish, could also provide a base for information strategies. Modelling techniques for dietary intake (including probabilistic modelling) and multivariate techniques for contaminant analysis can be applied to the analysis of the biomagnification of chemical contaminants between fish feed and fish as food. The collection of epidemiological data collected for POPs breast milk studies, and meta-analyses of data from the literature, would address vulnerable population groups.

247 Further studies could include selective contaminant assessment of convenience and processed fish products, compared with non-processed raw materials, to provide the means to assess contamination pathways in convenience foods and catering systems with respect to changing lifestyles. Such data would contribute to the EU overview of the maximum permitted levels of specific food contaminants. Multivariate analyses of time- related differences, as shown in the fish oil study in Chapter 2, section 2.5, would be appropriate monitoring and evaluation tools. This would indicate whether the overall reduction of at least 25% in dietary intake of POPs by 2006, in the EU, is likely to be achieved with current measures. The Commission has a written opinion in Council directive 2375/2001 that an overall reduction of at least 25% should be achieved by 2006. The further development of relatively inexpensive pre-screening computer techniques for determining receptor mediated endocrine disruption can provide a means to reduce animal usage in chemical screening. Such further research would offer a balanced methodological approach to assessing undesirable POPs components arising in the food chain that pose a continuous evolving challenge to food chain safety, by providing evidence from dietary intervention, epidemiological and toxicological studies concerning the benefits and risks of seafood for human health and well being.

6.2.1.2. Potential Impact Public health policy and the risk assessment of chemicals can only effectively deal with the issue of POPs and endocrine disruption in humans and wildlife by considering the whole endocrine system and not limit consideration to the oestrogenic, androgenic and thyroid systems, since regulation of these substances may not address issues related to effects on the function of other endocrine hormone and xenobiotic clearance systems at the molecular level, such as the AhR, PXR and CAR and enzyme mediated processes such as steroidogenesis. A more integrated systemic approach in future research, rather than only a single receptor focus, will give a more accurate picture of real-life mechanisms of action of EDC and NCE toxicity. This is more likely to satisfy future research needs as identified in the Community Strategy for Endocrine Disrupters which states that “7n order to support a rapid development of test methods, a focused research effort is required into the mechanisms of action of the endocrine system and the range of effects, including the role of hormones at key stages in life cycles. This should also

248 facilitate the development of useful models both for estimation of exposures and for biological testing strategies. ” The POPs of primary concern discussed here include PCBs, organochlorine pesticides, PBDEs, dioxins and furans, but this list is by no means exclusive. POPs and endocrine disruption are issues of global importance. The major source for human POPs exposure is dietary, particularly via oily fish, so an international assessment of POPs concentrations in the farmed salmonid food chain, and the development of both broad mechanistic and biomagnification toxicity assessment using in silico tools, would facilitate EU Public Health Policy aimed at improving the quality of life of the European consumer and protecting future generations.

6.2.2. In silico molecular modelling and QSARs: Further research Further research in the area of EDCs could also utilize the in silico models and crystal structures, and in silico modelling techniques in the development of predictive screening tool for improving the human risk assessment of POPs in the diet, as currently undertaken in drug discovery and development. Further development of the in silico receptors modelling would provide an effective pre-screening tool for the chemicals of concern. The multivariate analytical techniques described here are being increasing used to model biological systems. They are being developed and applied further from the field of chemistry, where PLS and chemical descriptors are now standard NCE discovery tools, to model cellular processes. QSARs such as those described in Chapter 5 show how chemical data can be generated in vitro (Chapter 4), or efficiently captured from the literature or in-house data sets to aid chemical screening. In the field of pharmaceutical research, human pharmacokinetic factors are modeled in various ways, including the symCYP software program. The information from NR QSARs could be added to these models to improve predictability. The models and crystal structures provide a very useful prescreening prediction tool; the information generated can flag potential unwanted interactions that can be explored further in biological systems. For instance, examination of a PBDE molecule in the ERa crystal structure LBD (Fig 1.7) suggested weak oestrogenicity, as indicated by weak hydrogen bonding in the LBD, which was later observed from in vitro studies conducted by Meerts et al., 2001. As more and more biological pathways are mapped out, in silico models will be increasingly applied. The value of in silico modelling is well known in drug development

249 and advances in this area can be effectively and usefully applied to improve our understanding of the molecular mechanisms of action in the biosynthetic pathways. This information contributes to improved holistic models for our understanding of nutrient- xenobiotic-hormone-receptor gene interactions. In relation to common human multifactoral diseases, such as cardiovascular disease and hormone-related cancers, these interactions have strong links with both nutrient, anti-nutrient and food contaminant intake, in addition to genetic predispositions. There are also very serious concerns surrounding foetal, infant and child development with respect to the interactions between POPs and various receptors. Under such circumstances, this information can be used to better inform public health and environmental policy, to better protect public health.

6.2.3. Applications: Tier testing Various endocrine disruptor screening programmes are currently being instigated in response to public concern on the effects of endocrine disrupters on animal and human health (Ashby, 2001; Baker, 2001; Cooper and Kavlock, 1997) and a tiered approach is being recommended and debated in the literature. Thresholds of toxicological concern (Schmidt, 1999; Cheeseman et al., 1999; Kroes et al., 2000) such as dioxin (Mackie et al., 2003), together with mixtures (Payne et al., 2001; Carpenter et al., 2002; Silva et al., 2002) and low dose assessment concerns (Mauras et al., 1989; Brucker-Davis et al., 2001; Welshons et al., 2003) particularly during foetal and childhood development (Bigsby et al., 1999; Straessen et al., 2001; Jacobson et al., 2002; Jacobson and Jacobson, 2002; ten Tusscher and Koppe, 2003) are all issues of concern. The QSAR methods presented in chapter 5 can be utilised for assessing the relative merits of receptor screening assays, as conducted for the AhR and flame retardants in section 5.4. They are also being used for integrating testing of receptor- specific activation in vitro assays for endocrine disruption, such as those for the TR (DeVito et al., 1999), the AhR (Hahn, 2002b), the ER (e.g. Yamasaki et al., 2003). They are also being used in mixture effects (Escher and Hermens, 2002), biotransformation and toxicity studies (Soffer et al., 2001), particularly in environmental toxicity studies. In addition to the AhR (Schneder et al., 1995; Chen, 2001b; Denison et al., 2002; Hahn, 2002b; Nagy et al., 2002), the ER, AR and TR (Bigsby et al., 1999), further receptors, including orphan receptors, also need to be part of a comprehensive screening process. Thus the most biologically relevant GR (Wang et al., 2003; El-Sankary et al., 2000), PXR, CAR activation (Moore et al., 2000b) and CYP3A4 induction assays (Pascussi et

250 al., 2000a; El-Sankary et al., 2001; Drocourt et al., 2002; Gibson et al., 2002; Plant and Gibson, 2003) could be part of a comprehensive assessment in a receptor/enzyme induction tier-testing scheme, in addition to those currently in use.

6.3. Conclusion The nuclear receptors discussed in this thesis act as steroid, metabolic and toxicological sensors that have evolved to provide a fast, highly adaptive, response system to environmental changes by inducing the appropriate metabolic pathways that will provide the optimum protection from overexposure. For the xenobiotic receptor sensors, these compounds can be present at high concentrations (|iM range), unlike the classical SHRs (e.g. ERs), where the endocrine signaling molecules are generated in minute quantities (nM range). Xenobiotic NRs have generally evolved in response to intermediates from metabolic pathways, phytochemicals and metabolites from dietary origin. Now, however, humans are chronically exposed to higher levels of ligands for these receptors, due to a westernized lifestyle characterized by high fat and processed food intakes and exposure to drugs and xenobiotics. Multiple contaminant or POPs exposure is also evolutionarily a relatively recent phenomenon. In response, the altered signaling by the nuclear receptors contributes to the pathogenesis of many common metabolic and hormone dependent diseases, such as obesity, insulin resistance, type 2 diabetes, hyperlipidemia, athersclerosis, gall bladder and liver disease, endometriosis and hormone dependent cancers. The role of these receptors in mediating these diseases provides a potential therapeutic target for their management (e.g. the statin and glitazone drugs for PPAR mediation and antioestrogens such as tamoxifen to inhibit ER+ breast cancers). Perturbing the receptor network can also result in developmental dysfunction consequences for the foetus and young child that may be irreversible. The nuclear receptors studied here appear to underpin the high incidence of many diseases to differing degrees, and receptor perturbation can have serious implications for human health which cannot be explained by genetic factors and the ‘rules of nature’ alone. Environmental and dietary factors together with the ‘flaws of nature’ (Barrow, 2003), are integral to the multidimensional puzzle, and there is much to be learned. Further research, particularly in the pharmaceutical industry, is developing and expanding NR and P450 based therapeutic treatment strategies. With improved understanding of the signaling pathways of receptor-ligand-gene interactions, public

251 health policy decisions can also be better informed. This work has demonstrated the importance of NRs to human health, and the application of this understanding to drug discovery and public health policy that benefits future generations.

252 7. Bibliography

Abbott, BD, Schmid, JE, Pitt, JA, Buckalew, AR, Wood, CR, Held, GA, and Diliberto, JJ. Adverse reproductive outcomes in the transgenic Ah receptor-deficient mouse. Toxicology and Applied Pharmacology 155,62-70.1999.

Abbott, DH, Dumesic, DA, Franks, S. Developemntal origin of polycystic ovary syndrome - a hypothesis. Journal of Endocrinology 174,1-5. 2002.

Abraham, MH, Chadha, HS, and Mitchell, R. Hydrogen bonding. 33 Factors that influence the distribution of solutes between blood and brain. Journal of Pharmacological Science 83,1257-1268. 1994.

Abraham, MH, Enomoto, K, Clarke, ED, and Sexton, G. Hydrogen bond basicity of the chlorogroup; Hexachlorohexanes as strong hydrogen bond bases. Journal of Organic Chemistry 67, 4782-4786. 2002a.

Abraham MH, Ibrahim A, Zissimos AM, Zhao YH, Comer J, and Reynolds DP. Application of hydrogen bonding calculations in property based drug design. Drug Discovery Today 7,1056-1063. 2002b.

Adachi, J, Mori, Y, Matsui, S, Takigami, H, Fujino, J, Kitagawa, H, Miller, CA, Kato, T, Sacki, K, and Matsuda, T. Indirubin and indigo are potent aryl hydrocarbon receptor ligands present in human urine. The Journal of Biological Chemistry 276, 31475- 31478.2001.

Adami, H-O, Bergstrom, R, Mohner, M, Zatonski, W, Storm, H, Ekbom, A, Tretli, S, Teppo, L, Zeigler, H, Rahu, M, Gurevicius, R, and Stengrevics, A. Testicular cancer in nine northern European countries. International Journal of Cancer 59, 33-38. 1994.

Adlercreutz, H and Mazur, W. Phyto-oestrogens and western diseases. Annals of Medicine 29, 95-120. 1997.

Adlercreutz, H. Phytoestrogens and breast cancer. Journal of Steroid Biochemistry and Molecular Biology 1803,1-6. 2003.

Ahlborg, UG, Lipworth, L, Titus-Emstoff, L, Hsieh, C-C, Hanberg, A, Baron, J, Trichopolous, DM, and Adami, H-O. Organochlorine compounds in relation to breast cancer, endometrial cancer, and endometriosis: An assessment of the biological and epidemiological evidence. Critical Reviews in Toxicology 25,463-531. 1995.

Akutsu, K, Obana, H, Okihashi, M, Kitgawa, M, Nakazawa, H, Matsuki, Y, Makino, T, Oda, H, and Hori, S. GC/MS analysis of polybrominated diphenyl ethers in fish collected from the Inland Sea of Seto, Japan. Chemosphere 44, 1325-1333. 2001.

Alberts, B, Bray, D, Johnson, A, Lewis, J, Raff, M, Roberts, K, and Walter, P. Essential Cell Biology, Garland publishing. New York and London, 1997.

253 Aldercreutz, H. Phytoestrogens and human health. British Toxicological Society Annual Congress, 19-22 April 1998. University of Surrey. Programme and abstracts. 45. 1998.

Allchin, CR, Morris, S, Bennet, M, Law, RJ, and Russell, I. PBDE residues in cormorant livers from England, UK. Organohalogen compounds 47,190-194. 2000.

Allen, SW, Mueller, L, Williams, SN, Quattrochi, LC, and Raucy, J. The use of high- volume screening procedure to assess the effects of dietary flavonoids on human cyplAl expression. Drug Metabolism and Disposition 29,1074-1079. 2001.

Andersen, HR, Andersson, AM, Arnold, SF, Autrup, H, Barfoed, M, Beresford, NA, Bjerregaard, P, Christiansen, LB, Gissel, B, Hummel, R, J0rgensen, EB, Korsgaard, B, Le Guevel, R, Leffers, H, McLachlan, J, Moller, A, Nielsen, JB, Olea, N, Oles- Karasko, A, Pakdel, F Pedersen KL, Perez, P, Skakkebaek, NE, Sonnenschein, C, Soto, AM, Sumpter, JP, Thorpe, SM, and Grandjean, P. Comparison of short-term estrogenicity for identification of hormone-disrupting chemicals. Environmental Health Perspectives 107, 89-107.1999.

Andersson, A-M and Skakkebaek, NE. Exposure to exogenous estrogens in food: possible impact on human development and health. European Journal of Endocrinology 140, 477-485. 1999.

Andrews, Craik, and Martin. Functional group contributions to drug-receptor interactions. Journal of Medicinal Chemistry 27(12), 1648-1657. 1984.

Anon. Hormone studies flawed, researcher warns. Chemistry and Industry October, 784. 1999.

Arnold, SF, Klotz, DM, Collins, BM, Vonier, PM, Guillette Jr, LJ, and McLachlan, JA. Synergistic activation of estrogen receptor with combinations of environmental chemicals. Science 272,1489-1492.1996.

Arnold, SF, Vonier, PM, Collins, BM, Klotz, DM, Guillette Jr, LJ, and McLachlan, JR. In vitro synergistic interaction of alligator and human estrogen receptors with combinations of environmental chemicals. Environmental Health Perspectives 105, 615-632. 1997.

Asatryan, RS, Mailyan, NS, Khachatryan, L, and Dellinger, B. Electronic elasticity-toxicity relationships for polychlorinated dibenzo-p-dioxin congeners. Chemosphere 48, 227- 236. 2002.

Ashby, C. Industrial Report, GlaxoSmithKline. 2002.

Ashby, J, Tinwell, H, Pennie, W, Brooks, AN, Lefevre, PA, Beresford, N, and Sumpter, JP. Partial and weak oestrogenicity of the red wine constituent resveratrol: consideration of its superagonist activity in MCF-7 cells and its suggested cardiovascular protective effects. Journal of Applied Toxicology 19, 39-45. 1999a.

Ashby, J, Tinwell, H, Soames, A, and Foster, J. Induction of hyperplasia and increased DNA content in the uterus of immature rats exposed to coumestrol. Environmental Health Perspectives 107, 819-822. 1999b.

254 Ashby, J. Testing for endocrine disruption post EDSTAC: extrapolation from low dose rodent effects to humans. Toxicology Letters 120, 233-242. 2001.

Asplund, L, Athanasiadou, M, Sjodin, A, Bergman, A, and Boijeson, H. Organohalogen substances in muscle, egg and blood from healthy Baltic salmon (Salmo Salar) and Baltic salmon that produced offspring with the M74 syndrome. Ambio 28, 67-76. 1999.

Atuma, SS, Aune, M, Bergh, A, Wicklund-Glyn, A, Damelund, PO, Larsson, L, Olsson, M, and Sandstrom, O. Polychlorinated biphenyls in salmon {Salmo salar) from the Swedish east coast. Organohalogen Compounds 39,153-156. 1998.

Auger, J, Kunstmann, JM, Czyglik, F, and Jouannet, P. Decline in semen quality among fertile men in Paris during the past 20 years. New England Journal of Medicine 332, 281-285. 1995.

Baes, M, Gulick, T, Choi, H-S, Martinoli, MG, Simha, D, and Moore, DD. A new orphan member of the nuclear hormone receptor superfamily that interacts with a subset of retinoic acid response elements. Molecular and Cellular Biology 14,1544-1552. 1994.

Baker, ME. Beyond carrier proteins Albumin, steroid hormones and the origin of vertebrates. Journal of Endocrinology 175,121-127. 2002.

Baker, VA. Endocrine disrupters - testing strategies to assess human hazard. Toxicology in vitro 15,413-419. 2001.

Barkhem, T, Carlsson, B, Nilsson, Y, Enmark, E, Gustafsson, J-Â, and Nilsson, S. Differential response of estrogen receptor a and estrogen receptor to partial estrogen agonists/antagonists. Molecular Pharmacology 54, 105-112. 1998.

Baroni, M, Cruciani, G, Riganelli, D, Valigi, R, and dementi, S. Generating Optimal Linear PLS Estimation (GOLPE): an advanced chemometric tool for handling 3D- QSAR problems. Quantitative Structure-Activity Relationships(12), 9-20. 1993.

Barrow, J. Glitch. New Scientist, 44-45. 7-6-2003.

Bar-Tana, J. Peroxisome proliferator-activated receptor gamma (PPARy) activation and its consequences in humans. Toxicology Letters 120, 9-19. 2001.

Basly, J-P, Marre-Foumier, F, Le Bail, J-C, Habrioux, G, and Chulia, AJ. Estrogenic/Antiestrogenic and scavenging properties of (E)- and (Z)-reservatrol. Life Sciences 66,769-777. 2000.

Bayada, DM, Hamersma, H, and van Geerestein, VJ. Molecular diversity and representativity in chemical databases. J Chem Inf Comput Sci 39, 1-10. 1999.

Bell, JG. Current aspects of lipid nutrition in fish farming. Black KD, and Pickering AD Eds. Biology of farmed fish. 114-115. 1998. Sheffield, UK, Sheffield Academic Press.

Bergman, Â. Brominated flame retardants-a burning issue. Organohalogen Compounds 47, 36-40. 2000.

255 Bergqvist, P-A, Hjelt, M, and Rappe, C. Bioaccumulation of PCDDs and PCDFs in farmed salmon with natural herring oil added to the feed. In: Water Pollution Research Report 24: Transport of organic micropollutants in estuaries, marine and brackish waters. Proc. of a workshop "Organic Micropollutants in the Aquatic environment" 4-5 October 1990. ( Wahlberg C, Minderhovd A., and Angeletti G. eds). Commission of the European communities, Directorate-General for Science, Research and Development, Brussels, pp. 111-119. 1990.

Bernstein, L and Press, ME. Does Estrogen Receptor expression in normal breast tissue predict breast cancer risk ? Journal of the National Cancer Institute 90, 5-7. 1998.

Bernstein, MP, Sandford, SC, Allamandrola, LJ, Gillete, JB, Clemett, SJ, and Zare, RN. UV irradiation of polycyclic aromatic hydrocarbons in ices: production of alcohols, quinones, and esters. Science 283,1135-1138. 1999.

Bertilsson, G, Heidrich, J, Svensson, K, Asman, M, Jenderberg, L, Sydow-Backman, M, Ohlsson, R, Postlind, H, Blomquist, P, and Berkenstam, A. Identification of a human nuclear receptor defines a new signalling pathway for CYP3A induction. Proc Natl Acad Sci USA 95,12208-12213. 1998.

Bhargava, HN and Leonard, PA. Triclosan: Applications and safety. Am J Infect Control 24, 209-218. 1996.

Bigsby, R, Chapin, RE, Daston, GP, Davis, BJ, Gorski, J, Gray, E, Howdenshell, KL, Zoeller, TR, and vom Saal, FS. Evaluating the effects of endocrine disrupters on endocrine function during development. Environmental Health Perspectives 107, 613- 618. 1999.

Bingham, SA, Atkinson, C, Liggins, J, Bluck, L, and Coward, A. Phyto-oestrogens : where are we now ? British Journal of Nutrition 79, 393-406. 1998.

Bimbaum, LS and Tuomisto, J. Non-carcinogenic effects of TCDD in animals. Food Additives and Contaminants 17, 275-288. 2000.

Bimbaum, LS. Health effects of Dioxins: People are animals and vice versa ! Organohalogen Compounds, 101-103. 2000.

Blair, RM, Fang, H, Branham, WS, Hass, BS, Dial, SL, Moland, CL, Tong, W, Shi, L, Perkins, R, and Sheehan, DM. The estrogen receptor relative binding affinities of 188 natural and xenochemicals: stmctural diversity of ligands. Toxicological Sciences 54, 138-153. 2000.

Blumberg, B and Evans, RM. Orphan nuclear receptors - new ligands and new possibilities. Genes and Development 12, 3149-3155. 1998.

Blumberg, B, Sabbagh, W, Juguilon, H, Bolado Jr, J, van Meter, CM, Ong, ES, and Evans, RM. SXR, a novel steroid and xenobiotic sensing nuclear receptor. Genes and Development 12, 3195-3205.1998.

256 Bobbyer, DNA, Goodford, PJ, and McWhinnie, PM. New hydrogen-bond potential for use in determining enegetically favourable binding sites of molecules of known structure. Journal of Medicinal Chemistry 32,1083-1094. 1989.

Bocio, A, Llobet, JM, Domingo, JL, Corbella, J, Teixido, A, and Casas, C. Polybrominated diphenyl ethers (PBDEs) in foodstuffs: human exposure through diet. Journal of Agriculture and Food Chemistry 51, 3191-3195. 2003.

BOCM. Pauls Fish Feed Group. Fulmar salmon feed. Rising to the challenge. Renfrew Mill, Renfrew, Scotland. 1998.

Boettger-Tong, H, Murthy, L, Chiappetta, C, Kirkland, JL, Goodwin, B, Adlercreutz, H, Stancel, GM, and Makela, S. A case of a laboratory animal feed with high estrogenic activity and its impact on in vivo responses to exogenously administered estrogens. Environmental Health Perspectives 106, 369-373. 1998.

Boon, JP, Lewis, WE, Tjoen-A-Choy, MR, Allchin, CR, Law, RJ, De Boer, J, ten Hallers- Tjabbes, CC, and Zegers, BN. Levels of polybrominated diphenyl (PBDE) flame retardants in animals represnting different trophic levels of the North Sea food web. Environmental Science and Technology 36,4025-4032. 2002.

Bradbury, RB and White, DE. Estrogens and related substances in plants. Vitamins and Hormones 12, 207-233. 1954.

Bradlow, HL, Davis, D, Sepkovic, DW, Tiwari, R, and Osbourne, MP. Role of the estrogen receptor in the action of organochlorine pesticides on estrogen metabolism in human breast cancer cell lines. Science of the Total Environment 208, 9-14.1997.

Bradlow, HL, Davis, DL, Lin, G, Sepovic, D, and Tiwari, R. Effects of pesticides on the ratio of 16 alphahydroxyestreone/2hydroxyestrone: a biologic marker of breast cancer risk.. Environmental Health Perspectives 103,147-150. 1995.

Brady , H, Doubleday, M, Gayo-Fung, LM, Hickman, M, Khammungkhune, S, Kois, A, Lipps, S, Pierce, S, Richard, N, Shevlin, G, Sutherland, MK, Anderson, DW, Bhagwat, SS, and Stein, B. Differential response of estrogen receptors a and b to SP500263, a novel potent selective estrogen receptor modulator. Mol Pharmacology 3, 562-568. 2002.

Branchi I, Capone F, Alieva E, and Costa LG Polybrominated diphenyl ethers: Neurobehavioral effects following developmental exposure. NeuroToxicology 24,449- 462. 2003

Britt, KL and Finday, JK. Estrogen activity in the ovary revisited. Journal of Endocrinology 175,269-276. 2002.

Brodie, A, Lu, Q, and Long, B. Aromatase and its inhibitors. Journal of Steroid Biochemistry and Molecular Biology 69, 205-210. 1999.

Brucker-Davis, F, Thayer, K, and Colbom, T. Significant effects of mild endogenous hormonal changes in humans: considerations for low-dose testing. Environmental Health Perspectives 109, 21-26. 2001.

257 Brunnberg S, Pettersson K, Rydin E, Matthews J, Hanberg A, and Pongratz I The basic helix-loop-helix-PAS protein ARNT functions as a potent coactivator of estrogen receptor-dependent transcription. Proc Natl Acad Sci. USA 100, 6517. 2003

Brzozowski, AM, Pike, AC, Dauter, Z, Hubbard, RE, Bonn, T, Engstrom, O, Ohman, L, Greene, GL, Gustafsson, J-Â, and Carlqvist, M. Molecular basis of agonism and antagonism in the oestrogen receptor. Nature 389,753-758.1997.

Buist SCN, Cherrington NJ, and Klaassen CD . Endocrine Regulation of Rat Organic Anion Transporters. Drug Metabolism and Disposition 31,559. 2003

Burbach, KM, Poland, A, and Bradfield, CA. Cloning of the Ah-Receptor cDNA Reveals A Distinctive Ligand-Activated Transcription Factor. Proc Natl Acad Sci USA 89, 815- 819. 1992.

Burow, ME, Tang, Y, Collins-Borow, BM, K, jewski, S, Reed, JC, McLachlan, JA, and Beckman, BS. Effects of environmental estrogens on tumor necrosis factor -mediated apoptosis in MCF-7 cells. Carcinogenesis 20, 2057-2061. 1999.

Bursi, R, de Gooyer, ME, Grootenhuis, A, Jacobs, PL, Van der Louw, J, and Leysen, D. (Q)SAR study on the metabolic stability of steroidal androgens. Journal of Molecular Graphics and Modelling 19, 552-556. 2001.

Cai, Y, Konishi, T, Han, G, Campwala, KH, French, SW, and Wan , Y-J Y. The role of hepatocyte RXRa in xenobiotic-sensing nuclear receptor-mediated pathways. European Journal of Pharmaceutical Sciences 15, 89-96. 2002.

Calmari, D. Assessment of persistent and bioaccumulating chemicals in the aquatic environment. Toxicology 181-182,183-286. 2002.

Carpenter, DO, Arcaro, K, and Spink, DC. Understanding the human health effects of chemical mixtures. Environmental Health Perspectives 110, 25-42. 2002.

Caruso, JA, Laird, DW, and Batist, G. Role of HSP90 in mediating Cross-Talk between the estrogen receptor and the Ah receptor signal transduction pathways. Biochemical and Biophysical Research Communications 58,1395-1403. 1999.

Cassidy, A and Faughnan, M. Phyto-oestrogens through the life cycle. Proceedings of the Nutrition Society 59,489-496. 2000.

Cassidy, A and Milligan, S. How significant are environmental oestrogens to women? Climeractic 1, 229-242. 1998.

Cassidy, A. Physiological effects of phyto-oestrogens in relation to cancer and other human health risks. Proceedings of the Nutrition Society 55, 399-417. 1996.

Chang, TKH, Bandiera, SM, and Chen, J. Constitutive andostane receptor and pregnane X receptor gene expression in human liver: interindividual variability and correlation with CYP2B6 mRNA levels. Drug Metabolism and Disposition 31,7-10. 2003.

258 Chang, TKH, Lee, WBK, and Ko, HH. Trans-resveratrol modulates the catalytic activity and mRNA expression of the procarcinogen-activating human cytochrome P450 IBl. Can J Physiol Pharmacol 78, 874-881. 2000.

Chawla, A, Repa, JJ, Evans, RM, and Mangelsdorf, DJ. Nuclear receptors and lipid physiology opening the X-files. Science 294,1866-1870. 2001.

Cheeseman, MA, Machuga, EJ, and Bailey, AB. A tiered approach to threshold of regulation. Food and Chemical Toxicology 37, 387-412. 1999.

Chen, G., Konstantinov, A. D., Chittim, B. G., Joyce, E. M., Bols, N. C., and Bunce, N. J. Synthesis of polybrominated diphenyl ethers and their capacity to induce CYPl A by the Ah receptor mediated pathway. Environmental Science and Technology 35, 3749- 3756. 2001.

Chen, J, Xue, X, Schramm, K-W, Quan, X, Yang, F, and Kettrup, A. Quantitative structure-property relationships for octanol-air partition coefficients of polychlorinated biphenyls. Chemosphere 48,535-544. 2002.

Choi, H-S, Chung, M, Tzameli, I, Simha, D, Lee, Y-K, Seol, W, and Moore, DD. Differential transactivation by two isoforms of the orphan nuclear hormone receptor CAR. Journal of Biological Chemistry 272, 23565-23571. 1997.

Chu, S, Covaci, A, Jacobs, W, Haraguchi, K, and Schepens, P. Distribution of methyl sulphone metabolites of polychlorinated biphenyls and /?,/?-DDE in human tissues. Environmental Health Perspectives 111 (9), 1223-1227. 2003.

Colburn, T, Dumanoski, D, and Myers, JP. Our Stolen future: are we threatening our fertility, intelligence, and survival? - a scientific detective story. New York, Dutton, 1999.

Colburn, T, vom Saal, FS, and Soto, AM. Developmental effects of endocrine disrupting chemicals in wildlife and humans. Environmental Health Perspectives 101, 378-384. 1993.

Collett, GP, Betts, AM, Johnson, MI, Pulimood, AB, Cook, S, Neal, DE, and Robson, CN. Peroxisome proliferator-activated receptor a is an androgen-responsive gene in human prostate and is highly expressed in prostatic adenocarcinoma. Clinical Cancer Research 6, 3241-3248. 2000.

Cooper, RL and Kavlock, RJ. Endocrine disrupters and reproductive development: a weight-of-evidence overviews. Journal of Endocrinology 152, 159-166. 1997.

Cottone, E, Orso, F, Biglia, N, Sismondi, P, and De Bortoli, M. Role of coactivators and corepressors in steroid and nuclear receptor signalling: potential markers of tumor growth and drug sensitivity. The International Journal of Biological Markers 16, 151- 166. 2001.

Courtois, A, Payen, L, Guillouzo, A, and Fardel, O. Up-regulation of multidrug resistance- associated protein 2 (MRP2) expression in rat hepatocytes by dexamethasone protein. FEBS Letters 459, 381-385. 1999.

259 Covaci, A, De Boer, J, Ryan, JJ, Voerspoels, S, and Schepens, P. Determination of polybrominated diphenyl ethers and polychlorinated biphenyls in human adipose tissue by large volume injection - narrow bore capillary Gas Chromatography/Electron Impact -Low Resolution Mass Spectrometry. Analytical Chemistry 74,790-798. 2002a.

Covaci, A, Jorens, P, Jacquemyn, Y, and Schepens, P. Distribution of PCBs and organochlorine pesticides in umbilical cord and maternal serum. The Science of the Total Environment 298,45-53. 2002b.

Covaci, A, Ryan, JJ, and Schepens, P. Patterns of PCBs and PCDD/PCDFs in contaminated chicken and pork following a Belgian food contamination. Chemosphere [(in press)]. 2002c.

Covaci, A. Environmental Analysis and biomonitoring of persistent organohalogenated pollutants. PhD . 2002. University of Antwerp.

Crain, DA, Rooney, AA, Orlando, EF, and Guillette, LJ Jr. Endocrine disrupting contaminants and hormone dynamics: Lessons from wildlife. In Environmental Endocrine disrupters, an evolutionary perspective. (Eds Guillette L Jr and Crain DA.) Taylor and Francis, New York. 2000.

Cramer, RD, Patterson, DE, and Bunce, JD. Comparative Molecular Field Analysis (CoMFA). Effect of Shape on Binding of Steroids to Carrier Proteins. Journal of the American Chemical Society 110, 5959-5967. 1998.

Cronet, P, Petersen, JEW, Folmer, R, Blomberg, N, Sjoblom, K, Karlsson, U, Lindstedt, E- L, and Bamberg, K. Structure of the PPAR a and (3- ligand biding domain in complex with AZ 242; ligand selectivity and agonist activation in the PPAR family. Structure 9, 699-706. 2001.

Cronet, P, Peterson, JF, Folmer, R, Blomberg, Sjoblom, K, Karlson, U, Lindstedt, E, and Bamberg, K. Structure of the PPAR alpha and gamma ligand binding domain in complex with AZ242; ligand selectivity and agonist activity in the PPAR family. Structure 9, 699-706. 2003.

Cruciani, G, Crivori, P, Carrupt, P-A, and Testa. Molecular fields in quantitative structure- permeation relationships: the VolSurf approach. Journal of Molecular Structure (Theochem) 503,17-30. 2000a.

Cruciani, G, Pastor, M, and Guba, W. VolSurf: a new tool for the pharmacokinetic optimization of lead compounds. European Journal of Pharmaceutical Sciences 11, S29-S39. 2000b.

Damianaki, A, Bakogeorgou, E, Kampa, M, Notas, G, Hatzoglou, A, Panagiotou, S, Gemetzi, C, Kouroumalis, E, Martin, P-M, and Castanas , E. Potent inhibitory action of red wine polyphenols on human breast cancer cells. Journal of cellular biochemistry 78, 429-441. 2000.

260 Damerud, PO, Eriksen, GS, Johannesson, T, Larsen, PB, and Viluksela, M. Polybrominated Diphenyl Ethers: Occurrence, Dietary Exposure, and Toxicology. Environmental Health Perspectives 109,49-68. 2001.

Davidson, NE and Yager, JD. Pesticides and breast cancer: Fact or fad? Journal of the National Institute 89,1743-1744. 1997.

Davis, DL, Bradlow, HL, Wolff, M, Woodruff, T, Hoel, DG, and Anton-Culver, H. Medical hypothesis: Xenoestrogens as preventable causes of breast cancer. Environmental Health Perspectives 101, 372-377. 1993.

Davis, DL, Telang, NT, Osborne, MP, and Bradlow, HL. Medical hypothesis: bifunctional genetic-hormonal pathways to breast cancer. Environmental Health Perspectives 105, 571-578. 1997.

de Boer, J, Stronck, CJN, van der Valk, F, Wester, PG, and Daudt, MJM. Method for the analysis of non-ortho substituted chlorobiphenyls in fish and marine mammals. Chemosphere 25-7,1277-1283.1992.

de Boer, J, Wester, PG, Klamer, HJC, Lewis, WE, and Boon, JPL. Do flame retardants threaten ocean life? Nature 394, 28-29. 1998.

de Boer, J. First worldwide interlaboratory study on PBDEs in biota and sediments. Organohalogen Compounds 45, 118-121. 2000.

de Longeril, M, Salen, P, Martin, J-L, Monjaud, I, Delaye, J, and Mamelle, N. Mediterranean diet, traditional risk factors, and the rate of cardiovascular complications after myocardial infarction. Final report of the Lyon Heart Study. Circulation 99,779- 785.1999.

DeBruin, LS and Josephy, PD. Perspectives on the chemical etiology of breast cancer. Environmental Health Perspectives 110, 119-128. 2002.

Dehmlow, C, Murawski, N, and de Groot, H. Scavenging of reactive oxygen species and inhibition of arachidonic acid metabolism by silibinin in human cells. Life Sciences 58, 1591-1600. 1996.

Del Bene, JE. Hydrogen Bonding: 1. von Ragué Schleyer, P, Allinger, NL, Clark, T, Gasteiger, J, Kollman, PA, Schaefer, HE, and Schreiner, PR. Encylopedia of Computational Chemistry E-L. 2,1263-1271. 1998. Chichester, UK, John Wiley and Sons.

Denison, MS and Whitlock, JP. Xenobiotic-inducible transcription of cytochrome P450 genes. The Journal of Biological Chemistry 270,18175-18178. 1995.

Denison, MS, Pandini, A, Nagy, SR, Baldwin, EP, and Bonati, L. Ligand binding and activation of the Ah receptor. Chemico-Biological Interactions 141, 3-24. 2002.

Denison, MS, Phelan, D, Winter, GM, and Ziccardi, MH. Carbaryl, a carbamate insecticide is a ligand for the hepatic Ah (dioxin) Receptor. Toxicology and Applied Pharmacology 152,406-414. 1998.

26 1 DeVito, M, Biegel, L, Brouwer, A, Brown, S, Brucker-Davis, F, Cheek, AO, Christensen, R, Colbom, T, Cooke, P, Crissman, J, Crofton, K, Doerge, D, Gray, E, Hauser, P, Hurley, P, Kohn, M, Lazar, J, McMaster, S, McClain, M, McConnell, E, Meir, C, Miller, R, Tietge, J, and Tyl, R. Screening methods for thyroid hormone dismptors. Environmetal Health Perspectives 107,407-415, 1999.

Diliberto, JJ, Abbot, BD, and Bimbaum, LS. Use of AhR knockout (AhR-/-) mice to investigate the role of the Ah receptor on the disposition of TCDD. Organohalogen Compounds 49,121-123. 2000.

Dotzlaw, H, Leygue, E, Watson, P, and Murphy, LC. The human orphan receptor PXR messenger RNA is expressed in both normal and neoplastic breast tissue. Clinical Cancer Research 5,2103-2107. 1999.

Drocourt, L, Ourlin, J-C, Pascussi, J-M, Maurel, P, and Vilarem, M-J. Expression of CYP3A4, CYP2B6, and CYP2C9 is regulatied by the vitamin D receptor pathway in primary human hepatocytes. The Joumal of Biological Chemistry 277,25125-25132. 2002.

Dubuquoy, L, Dharancy, S, Nutten, S, Pettersson, S, Auwerx, J, and Desreumaux, P. Role of peroxisome prolifeator-activated receptor y and retinoid X receptor heterodimer in hepatogastroenterological diseases. Lancet 360,1410-1418. 2002.

Durr, D, Stieger, B, Kullak-Ublick, GA, Rentsch, KM, Steinert, HC, Meier, PJ, and Fattinger, K. St John's Wort induces intestinal P-glycoprotein/MDRl and intestinal and

hepatic CYP3A4. Clinical Pharmacology and Therapeutics 6 8 , 598-604. 2000.

Edwards, PA and Ericsson, J. Sterols and isoprenoids: signalling molecules derived from

the cholesterol biosynthetic pathway. Ann Rev Biochem 6 8 , 157-185. 1999.

Edwards, PA, Kast, HR, and Anisfeld, AM. BAREing it all: the adoption of LXR and FXr and their roles in lipid homeostasis. Joumal of Lipid Research 43, 2-12. 2002.

Ekins, S and Erickson, JA. A pharmacophore for human pregnane X receptor ligands. Dmg Metabolism and Disposition 30, 96-99. 2002.

Elbi, C, Misteli, T, and Hager, GL. Recmitment of dioxin receptor to active transcription sites. Molecular Biology of the Cell 13, 2001-2015. 2002.

Eldridge, MD, Murray, CW, Auton, TR, Paolini, GV, and Mee, RP. empirical scoring functions: I. The development of a fast empirical scoring function to estimate the binding affinity of ligands in receptor complexes. Joumal of Computer Aided Molecular Design 11,425-445. 1997.

El-Sankary, W, Gibson, GG, Ayrton, A, and Plant, N. Use of a reporter gene assay to predict and rank the potency and efficacy of CYP3A4 inducers. Dmg Metabolism and Disposition 29, 1499-1504. 2001.

262 El-Sankary, W, Plant, NJ, Gibson, GG, and Moore, DJ. Regulation of the CYP3A4 gene by hydrocortisone and xenobiotics: role of the glucocorticoid and pregnane X receptors. Drug Metabolism and Disposition 28,493-496. 2000.

Emken, EA, Adlof ,R0, and Gulley, RM. Dietary linoleic acid influences desaturation and acylation of deuterium-labeled linoleic and linolenic acids in young adult males. Biochim Biophys Acta 1213, 277-288. 1994.

Emken, EA. Stable isotope approaches, applications, and issues related to polyunsaturated fatty acid metabolism studies. Lipids 36(9), 965-973. 2001.

Erikson, BE. Methyl triclosan found in Swiss lakes. Environmental Science and Technology, 228A-230A. 2002.

Eriksson, L, Hermens, JLM, Johansson, E, Verhaar, HJM, and Wold, S. Multivariate analysis of aquatic toxicity data with PLS. Aquatic Sciences 57, 217-241. 1995.

Eriksson, L, Johansson, E, Kettaneh-Wold, N, and Wold, S. Multi-and Megavariate Data Analysis Principles and Applications. 2001a. Umetrics Academy.

Eriksson, L, Johansson, E, Muller, M, and Wold, S. Cluster-based design in environmental QSAR. Quant Struct-Act Relat 16, 383-390.1997.

Eriksson, P, Jakobsson, E, and Frediksson, A. Brominated flame retardants: A novel class of developmental neurotoxicants in our environment ? Environmental Health Perspectives 109 (9), 903-908. 2001b.

Escher, BI and Hermens, JLM. Modes of action in ecotoxicology: their role in body burdens, species sensitivity, QSAR, and mixture effects. Environ Sci Techol 36,4201- 4217. 2002.

European Commission, Health and Consumer Protection Directorate-General. Assessment of dietary intake of dioxins and related PCBs by the population of EU member states. Report on tasks for scientific cooperation. Report of experts participating in Task 3.2.5, 7. http://europa.eu.int/comm/food/fs/sc/scf/index en.html. 2000.

Falandysz, J. Polychlorinated biphenyl concentrations in cod-liver oil: Evidence of a steady state condition of these compounds in the Baltic area oils and levels noted in Atlantic oils. Arch Environ Contam Toxicol 27, 266-271. 1994.

Falandysz, J, Tanabe, S, and Tatsukawa, R. Most toxic and highly bioaccumulative PCB congeners in cod-liver oil of Baltic origin processed in Poland during the 1970s and 1980s, their TEQ-values and possible intake. Science of the Total Environment 145, 207-212. 1994.

Falck, F Jr, Ricci, A Jr, Wolff, MS, Godbold, J, and Deckers, P. Pesticides and polychlorinated biphenyl residues in human breast lipids and their relation to breast cancer. Archives Environmental Health 47,143-146. 1992.

Fang, H, Tong, W, Shi, LM, Blair, R, Perkins, R, Branham, W, Hass, BS, Xie, Q, Dial, SL, Moland, CL, and Sheeman, DM. Struture-Activity relationships for a large diverse set

263 of natural, synthetic, and environmental estrogens. Chem Res Toxicol 14, 280-294. 2001.

FAO Fishery Information, Data and Statistics Unit. Aquaculture Production Statistics 1987-1996. Fisheries Circular. No. 815, Rev. 10. Rome, FAO. 1998.

FAO Globefish, Fisheries Industries Division. Available: http://www.globefish.org. 2000.

Faustman, EM, Silbemagel, SM, Fenske, RA, Burbacher, TM, and Ponce, RA. Mechanisms underlying children’s susceptibility to environmetal toxicants. Environ Health Perspect 108,13-21. 2000.

Femadez-Salugero, P, Ward, JM, Sunberg, JP, and Gonzalez, FJ. Lesions of arylhydrocarbon receptor deficient mice. Vet Pathol 34, 605-614.1997.

Fernandes, A, Rose, M, and White, S. Dioxins (PCDD/Fs) and PCBs in fish oil dietary supplements and medicinal products. DEFRA Report FDOO/101. Pp 1-32. 2002.

Femandez-Salugero, P, Pinneau, T, Hilbert, DM, McPhail, T, Lee, SST, Kimura, S, Nebert, DW, Rudikoff, S, Ward, JM, and Gonzalez, FJ. Immune system impairment and hepatic fibrosis in mice lacking the dioxin binding Ah Receptor. Science 268,722-726. 1995.

Ferrario, J and Byrne, C. The concentration and distribution of 2,3,7,8-dibenzo-p-dioxins/- furans in chickens. Chemosphere 40,221-224. 2000.

Ferrario, J, Byrne, C, and Dupuy, A. Background Contamination By Coplanar PCBs in Trace Level High Resolution Gas Chromatography/ High Resolution Mass Spectrometry (HRMS/MRGC) Analytical Procedures. Chemosphere 34, 2351-2466. 1997.

Ferrario, J, Byrne, C, McDaniel, D, Dupuy, A, and Harless, R. Determination of 2,3,7,8- Chlorine-Substituted Dibenzo-p-dioxins and -furans at the Parts per Trillion Level in United States Beef Fat using High- Resolution Gas Chromatography/ High Resolution

Mass Spectrometry. Analytical Chemistry 6 8 , 647-652. 1996.

Ferrario, J, McDaniel, D, and Byrne, C. The Isomer Distribution and Congener Profile of Polychlorinated Dibenzo-p-dioxins (PCDDs) in Ball Clay from the Mississippi Embayment (Sledge Mississippi). Organohalogen Compounds 40, 95-99.1999.

Fielden, MR, Matthews, JB, Fertuck, KC, Halgren, RG, and Zacharewski, TR. In Silico approaches to mechanistic and predictive toxicology: an introduction to bioinformatics for toxicologists. Critical Reviews in Toxicology 32,67-112. 2002.

Fischer, M. Genetically determined differences in P-glycoprotein function: implications for disease risk. Toxicology 181-182, 299-303. 2002.

Food Safety Authority of Ireland. Investigation on PCDDs/PCDFs and several PCBs in fish liver oil capsules. 2002. Available from http//:www.fsai.ie/industry/Dioxins3.htm

264 Food Safety Authority of Ireland. Investigation on PCDDs/PCDFs and several PCBs in fish samples (Salmon and trout). 2003. Available from http//:www.fsai.ie/industry/Dioxins3.htm

Food Standards Agency. Food Standards Agency UK- dioxins and PCBs in the UK diet: 1997 total diet study samples. Food Surveillance Information Sheet 4/00, FSA, London.UK. 2000.

Foran, CM, Bennett, ER, and Benson, WH. Developmental evaluation of a potential non steroidal estrogen: triclosan. Marine Environmental Research 50,153-156. 2000.

Forman, BM, Tzameli, I, Choi, H-S, Chen, J, Simha, D, Seol, W, Evans, RM, and Moore, DD. Androstane metabolites bind to and deactivate the nuclear receptor CAR-p. Nature 395,612-615. 1998.

Forster, C, Makela, S, Warri, A, Kietz, S, Becker, D, Hultenby, K, Warner, M, and Gustafsson, J-A. Involvement of estrogen receptor P in terminal differentiation of mammary gland epithelium. Proceedings of the National Academy of Sciences 99, 15578-15583. 2002.

Forster, J. Aquaculture chickens, salmon- a case study. World Aquaculture 30, 33-70. 1999.

Franks, S, Gharani, N, and Gilling-Smith, C. Polycystic ovary syndrome: evidence for a primary disorder of ovarian steroidogenesis. Joumal of Steroid Biochemistry and Molecular Biology 69, 269-272. 1999.

Fujimoto, J, Hirose, R, Sakaguchi, H, and Tamya, T. Clinical significance of expression of estrogen receptor receptor a and p mRNAs in ovarian cancers. Oncology 58, 334-341.

2 0 0 0 a.

Fujimoto, J, Sakaguchi, H, Aoki, I, and Tamaya, T. Steroid receptors and metastatic potential in endometrial cancers. European Joumal of Cancer 36, S33. 2000b.

Fuqua, SAW, Wiltschke, C, Zhang, QX, Borg, A, Castles, CG, Friedrichs, WE, Hopp, T, Hilsenbeck, S, Mohsin, S, O'Connell, P, and Allred , DC. A hypersensitive estrogen receptor alpha mutation in premalignant breast lesions. Cancer Research 60,4026- 4029. 2000.

Gaido, KW, Leonard, LS, Maness, SC, Hall, JM, McDonnell, DP, Saville, B, and Safe, S. Differential interaction of the methoxychlor metabolite 2,2-Bis-(p-hydroxyphenyl)- 1,1,1 -Tricloroethane with estrogen receptors a and p. Endocrinology 140, 5746-5753. 1999.

Gambone, CJ, Hutcheson, JM, Gabriel, JL, Beard, RL, Chandraaratna RAS, and Soprano, RD. Unique property of some synthetic retinoids: Activation of the Aryl hydrocarbon receptor Pathway. Molecular Pharmacology 61, 334-342. 2002.

Gampe Jr, RT, Montana, VG, Lambert, M, Miller, AB, Bledsoe, RK, Milbum, MV, Kliewer, SA, Willson, TM, and Xu, HE. Asymmetry in the PPARy/RXRa crystal

265 structure reveals the molecular basis of heterodimerization among nuclear receptors. Molecular Cell 5, 545-555. 2000.

Gao, H, Williams, C, Labute, P, and Bajorath, J. Binary quantitative structure - activity relationship (QSAR) analysis of estrogen receptor ligands. J Chem Inf Comput Sci 39, 164-168. 1999.

Gard, P. Human Endocrinology. Taylor and Francis London. 2001.

Gibson, GG, El-Sankary, W, and Plant, NJ. Receptor-dependant regulation of the CYP3A4 gene. Toxicology 181-182,199-202. 2002.

Gillner, M, Bergman, J, Ambilleau, C, Ferustrom, B, and Gustafsson, J-Â. Interactions of indolo [3-2-6] carbazoles and related polycyclic aroma hydrocarbons with specific binding sites for 2,3,7,8-tetrachlorodobenzo-p-dioxin in rat liver. Molecular Pharmacology 28, 336-345. 1993.

Gillner, M, Carlsson, P, and Greenidge, P. Applications of molecular mechanics to prediction of binding affinities of estrogen agonists to the human Estrogen alpha receptor using it's X- ray crystallographic structure. Organohalogen Compounds 37, 261-263. 1998.

Giwercman, A, Carlsen, E, Keiding, N, and Skakkebaek, NE. Giwercman A, Carlsen E, Keiding N, and Skakkebaek NE. Evidence for the increasing incidence of abnormalities of the human testis: A review. Environmental Health Perspectives 102, 56-71. 1993.

Gobas, EAPC, Wilcockson, JB, Russell, RW, and Haffner, GD. Mechanism of biomagnification in fish under laboratory and field conditions. Environmental Science and Technology 33,133-141. 1999.

Goodford, PJ. Computational procedure for determining enegetically favourable binding sites on biologically important macromolecules. Joumal of Medicinal Chemistry 28, 849-857. 1985.

Gray Jr, EL, Kelce, WR, Wiese, T, Tyl, R, Gaido, K, Cook, J, Klinefelter, G, Desaulniers, D, Wilson, E, Zacharewski, T, Waller, C, Foster, P, Laskey, J, Reel, J, Giesy, J, Laws, S, McLachlan, J, Breslin, W, Cooper, R, Di Gulio, R, Johnson, R, Purdy, R, Mihaich, E, Safe, S, Sonnenschein, C, Welshons, W, Miller, R, McMaster, S, and Colbom, T. Endocrine screening methods workshop report: detection of estrogenic and androgenic hormonal and antihormonal activity for chemicals that act via receptor or steroidogenic enzyme mechanisms. Reproductive Toxicology 11,719-750. 1997.

Green, BJ and Rudel, RA. Environmental pollutants and breast cancer. Environmental

Health Perspectives 111 ( 8 ), 1007-1019. 2003.

Gregory, J, Foster, K, Tyler, H, and Wiseman, M. The Dietary and Nutritional Survey of British Adults. Ministry of Agriculture Fisheries and Food, Dept of Health. HMSO. London. 1990.

Gress, SR., Lang, DH., Tomsho, RJ., Sharkey, NA., and Corwin, RL. High fat diets reduce bone biomechanical properties in aged rats. FASEB J. A16 A625. 2002. cited in

266 Corwin, RL. Effects of dietary fats on bone health in advanced age. Prostaglandins,

Leukotrienes and Essential Fatty Acids 6 8 , 379-386. 2002.

Guillette, U and McLachlan, JA. Ecosystems and embryos - is there a scientific basis for concern? Animal Reproduction Science 42,13-24. 1996.

Gustafsson, J-A. Estrogen receptor - a new dimension in estrogen mechanism of action. Joumal of Endocrinology 163, 379-383. 1999.

Gustafsson, J-A. The Role of the ER(3 in oestrogen mediated carcinogenesis. Toxicol Letters 116,1. 2000.

Hahn, ME, Woodin, BR, Stegeman, JJ, and Tillitt, DE. Aryl hydrocarbon receptor function in early vertebrates: inducibility of cytochrome P450 lA in agnathan and elasmobranch fish. Comparative Biochemistry and Physiology Part C 120, 67-75. 1998.

Hahn, ME. The aryl hydrocarbon receptor: a comparative perspective. Comparative Biochemistry and Physiology Part C 121, 23-53. 1998.

Hahn, ME. Aryl hydrocarbon receptors: diversity and evolution. Chemico-Biological Interactions 141,131-160. 2002a.

Hahn, ME. Biomarkers and bioassays for detecting dioxin-like compounds in the marine environment. The Science of the Total Environment 289,49-69. 2002b.

Hall, JM and McDonnell, DP. The Estrogen receptor p-isoform (ERp) of the human estrogen receptor modulates ER a transcriptional activity and is a key regulator of the cellular response to estrogens and antiestrogens. Endocrinology 140, 5566-5578. 1999.

Hamilton, RJ and Rice, RD. Eds. Fish Oil: Technology, Nutrition and Marketing. PJ Bames and Associates. 1995.

Hanioka, N, Omae, E, Nishimura, T, Jinno, H, Onodera, S, Yoda, R, and Ando, M. Interaction of 2,4,4'-trichloro-2'-hydroxydiphenyl ether with microsomal cytochrome P450-dependent monooxygenases in rat liver. Chemosphere 33, 265-276. 1996.

Harboume, JB, Baxter, H, Moss GP, and ed. Phytochemical Dictionary. (2nd edition). 1999. Taylor and Francis Ltd, London UK.

Harris, CA, Henttu, P, Parker, MG, and Sumpter, JP. The estrogenic activity of phthalate esters in vitro. Environmental Health Perspectives 105, 802-811. 1997.

Hayes, TB, Collins, A, Lee, M, Mendoza, M, Noriega, N, Stuart, AA, and Vonk, A. Hermaphroditic, demasculinized frogs after exposure to the herbicide atrazine at low

ecologically relevant doses. Proc Natl Acad Sci USA 99 ( 8 ), 5476-5480. 2002.

He, K, Lyer, KR, Hayes, RN, Sinz, MW, Woolf, TF, and Hollenberg, PF. Inactivation of cytochrome P450 3A4 by Bergamottin, a component of grapefruit juice. Chem Res Toxicol 11,252-259. 1998.

267 Heath, RJ, Rubin, RJ, Holland, DR, Zhang, E, Snow, ME, and Rock, CO. Mechanisms of triclosan inhibition of bacterial fatty acid synthesis. The Joumal of Biological Chemistry 274, 1110-1114. 1999.

Heath-Pagliuso, S, Rogers, WJ, Tullis, K, Seidel, SD, Cenijn, PH, Brouwer, A, and Denison, MS. Activation of the Ah receptor by tryptophan and tryptophan metabolites. Biochemistry 37,11508-11515. 1998.

Hecht, J. And God said let there be smog. New Scientist 175,14. 2002.

Heinonen, S, Nurmi, T, Liukkonen, K, Poutanen, K, Wahala, K, Deyama, T, Nishibe, S, and Adlercreutz, H. In vitro metabolism of plant lignans: new precursors of mammelian lignans enterolactone and enterodiol. Joumal of Agriculture and Food Chemistry 49, 3178-3186.2001.

Henry, EC and Gasiewicz, TA. Agonist but Not Antagonist Ligands Induce Conformational Change in the Mouse Aryl Hydrocarbon Receptor as Detected by Partial Proteolysis. Mol Pharmacol 63,392. 2003

Henry, LA and Witt, DM. Resveratrol: Phytoestrogen effects on reproductive physiology and behavior in female rats. Hormones and Behaviour 41, 220-228. 2002.

Hertog, MGL and Hollman, PCH. Potential health effects of the dietary flavonol quercetin. European Joumal of Clinical Nutrition 50,63-71. 1996.

Hess, RA, Bunick, D, Lee, K-H, Bahr, J, Taylor, JA, Korach, KS, and Lubahn, DB. A role for oestrogens in the male reproductive system. Nature 390, 509-512. 1997.

Hillman, GG, Forman, JD, Kucuk, O, Yudelev, M, Maughan, RL, Rubio, J, Layer, A, Tekyi-Mensah, S, Abrams, J, and Sarker, FH. Genistein potentiates the radiation effect on prostate carcinoma cells. Clinical Cancer Research 7, 382-390. 2001.

Hindle, SA, Rarey, M, Buning, C, and Lengauer, T. Flexible docking under pharmacophore type constraints. Joumal of Computer Aided Molecular Design 16, 129-149. 2002.

Hong, H, Tong, W, Fang, H, Shi, L, Xie, Q, Wu, J, Perkins, R, Walker, JD, Branham, W, and Sheehan, DM. Prediction of estrogen receptor binding for 58,000 chemicals using an integrated system of a tree-based model with stmctural alerts. Environmental Health Perspectives 110, 29-36. 2002.

Honkakoshi, P and Negishi, M. Regulation of cytochrome P450 (CYP) genes by nuclear receptors. Biochemistry Joumal 347, 321-337. 2000.

Honkakoshi, P, Sueyoshi, T, and Negishi, M. Dmg-activated nuclear receptors CAR and PXR. Annals of Medicine 35,172-182. 2003.

Honkakoshi, P, Zelko, I, Sueyoshi, T, and Negishi, M. The nuclear orphan receptor CAR- retinoid X receptor heterodimer activates the phenobarbital-responsive enhancer nodule of the CYP2B gene. Molecular and Cellular Biology 18, 5652-5658. 1998.

268 Honkakoski, P, Sueyoshi, T, and Negishi, M. Drug-activated nuclear receptors CAR and PXR. Annals of Medicine 35,172-182. 2003.

Howdeshell, KL, Hotchkiss, AK, Thayer, KA, Vandenbergh, JG, and vom Saal, FS. Exposure to bisphenol a advances puberty. Nature 401,763-765. 1999.

Hsieh, T-C and Wu, JM. Differential effects on growth, cell cycle arrest, and induction of apoptosis by resveratrol in human prostate cancer cell lines. Experimental Cell Research 249,109-115. 1999.

Hu, FB, Stamfer, MJ, and Manson, JE. Dietary intake of -linolenic acid and risk of fatal ischemic heart disease among women. American Joumal of Clinical Nutrition 69, 890- 897. 1999.

Hunter, DJ and Kelsey, KT. Pesticide residues and breast cancer: The harvest of a silent spring. /. Joumal of the National Cancer Institute 85, 598-599. 1993.

Hunter, DJ, Hankinson, SE, Laden, F, Colditz, GA, Manson, JE, Willet, WC, Speizer, FE, and Wolff, MS. Plasma organochlorine levels and the risk of breast cancer. New England Joumal of Medicine 337,1253-1258. 1997.

Huuskonen, J. QSAR modelling with the electrotopological state indices: predicting the toxicity of organic chemicals. Chemosphere 50, 949-953. 2003.

Ide, T, Shimano H, Yoshikawa, T, Yahagi, N, Amemiya-Kudo, M, Matsuzaka, T, Nakakuki, M, Yatoh, S, lizuka, Y, Tomita, S, Ohashi, K, Takahashi, A, Sone, H, Gotoda, T, Osuga, Ji, Ishibashi, S, and Yamada, N. Cross-Talk between Peroxisome Proliferator-Activated Receptor (PPAR) {alpha} and Liver X Receptor (LXR) in Nutritional Regulation of Fatty Acid Metabolism. II. LXRs Suppress Lipid Degradation Gene Promoters through Inhibition of PPAR Signaling. Molecular Endocrinolology 17, 55-1267. 2003 lida, T, Hirakawa, H, Matsueda, T, Nagayama, J, and Nagata, T. Polychlorinated dibenzo- p-dioxins and related compounds: correlations of levels in human tissues and in blood. Chemosphere 38, 2767-2774.1999. loannides, C. The CYP3 family, in C. loannides, (Ed.), Cytochromes P450: metabolic and toxicological aspects. CRC Press, Inc., Boca Raton, FL. pp 241-270. Biochemical and Biophysical Research Communications 260, 377-381. 1996. loannides, C. Pharmacokinetic interactions between herbal remedies and medicinal dmgs. Xenobiotica 32,451-478. 2002.

Jackson, LJ, Carpenter, SR, Manchester-Neesvig, JB, and Stow, CA. PCB congeners in Lake Michigan Coho (Oncorhynchus kisutch) and Chinook (Oncorhynchus tsawytscha) salmon. Environmental Science and Techology 35, 856-862. 2001.

269 Jacobs, M, Covaci, A, and Schepens, P. Investigation of selected persistent organic pollutants in farmed atlantic salmon (Salmo salar), salmon aquaculture feed, and fish oil components of the feed. Environmental Science and Techology 36, 2797-2805.

2 0 0 2 b.

Jacobs, M, Covaci, A, and Schepens, P. Investigations of polybrominated diphenyl ethers in Scottish and European farmed Atlantic salmon (Salmo salar), salmon aquaculture feed and fish oils. Organohalogen Compounds 51, 239-242. 2001a.

Jacobs, MN, Covaci, A, Ghoerghe, A and Schepens, P. Investigation of selected PCBs, PBDEs and organochlorine pesticides in omega-3 fatty acid rich dietary fish oil and vegetable oil supplements. Organohalogen Compounds, 57,225-229. 2002c

Jacobs, M, Covaci, A, Schepens, P, and Millstone, E. Investigation of selected PCB s and organochlorine pesticides in farmed atlantic salmon (Salmo salar), salmon aquaculture feed and fish oil components of the feed. Organohalogen Compounds 51, 314-318.

2 0 0 1 b.

Jacobs, M. Persistent organic pollutants in farmed fish - issues and concerns Salmon has plenty of health benefits but problems with POPs must be addressed. Aquavision, 43- 47. 2002.

Jacobs, MN and Lewis, DEV. Homology modelling of the aryl hydrocarbon receptor (AhR) ligand binding domain from the human oestrogen receptora (hERa) crystal structure. Toxicology Letters 116 (Supplement 1), 59. 2000a.

Jacobs, MN and Lewis, DEV. Steroid hormone receptors and dietary ligands: a selected review. Proceedings of the Nutrition Society 61,105-122. 2002.

Jacobs, MN and Lewis, DFV. A QSAR study of organochlorine and isoflavonoid compounds ligand binding affinity to the human oestrogen receptor. Organohalogen Compounds 41, 517-520. 1999.

Jacobs, MN and Lewis, DFV. Homology modelling of the human peroxisome proliferator activated receptora (PPARa) ligand binding domain from the human PPARy crystal structure. Toxicology Letters 116, (Supplement 1), 59. 2000b.

Jacobs, MN, Covaci, A, Ghoerghe, A, and Schepens, P. Investigation of selected PCBs, PBDEs and organochlorine pesticides in omega-3 fatty acid rich dietary fish oil and vegetable oil supplements. Organohalogen Compounds 57, 225-229. 2002c.

Jacobs, MN, Dickins, M, and Lewis, DFV. Homology modelling of the nuclear receptors: human oestrogen receptor P (hERP), the human pregnane-X-receptor (PXR), the Ah receptor (AhR) and the constitutive androstane receptor (CAR) ligand binding domains from the human oestrogen receptor a (hERa)crystal structure, and the human peroxisome proliferator activated receptor a (PPARa) ligand binding domain from the human PPARy crystal structure. Joumal of Steroid Biochemistry and Molecular Biology 84, 117-132. 2003a.

270 Jacobs, MN, Chen, G, Luscombe, C, Bunce, NJ, Multivariate (QSAR) Modelling of Polybrominated Diphenyl Ethers activity in different species cell lines and their capacity to induce CYPl A by the Ah Receptor Mediated Pathway. Organohalogen Compounds 61, 96-99. 2003.

Jacobs, MN, Ferrario, J, and Byrne, C. Investigation of PCDDs, PCDFs and selected coplanar PCBs in Scottish farmed Atlantic salmon. Organohalogen Compounds 47, 338-341. 2000.

Jacobs, MN, Ferrario, J, and Byrne, C. Investigation of polychlorinated dibenzo-p-dioxins, dibenzo-p-furans and selected coplanar biphenyls in Scottish farmed Atlantic salmon {Salmo salar). Chemosphere 46,183-191. 2002a.

Jacobs, MN, Johnston, PA, Santillo, D, and Wyatt, CL. Organo-chlorine pesticide and PCB residues in fish oil dietary supplements: comparison with industrial grade oils. Chemosphere 37,1709-1731. 1998.

Jacobs, MN, Starkey, S, and Hoffman, R. Combination effects of different ratios of PCB 153 and the phytoestrogen genistein upon DNA synthesis in an MCF-7 breast cancer cell line. Organohalogen Compounds 42,73-77. 1999.

Jacobs, MN, Wyatt, CL, Santillo, D, and French, M. Organo-chlorine pesticide and PCB residues in pharmaceutical and industrial grade fish oils. International Joumal of

Environment and Pollution 8 ,74-93. 1997.

Jacobs, MN. Molecular modelling of oestrogenic compounds in the human oestrogen receptor and investigation of the activity of oestrogenic compounds in breast cancer cell lines. MSc Thesis, University of Surrey. 1998.

Jacobs, MN. Principal Components Time Trend Analysis Of Organochlorine Contaminants Detected In Omega- 3 Rich Fish And Vegetable Oil Dietary Supplements from UK outlets. Toxicology Letters, in press. 2003.

Jacobson, JL and Jacobson, SW. Association of prenatal exposure to an environment contaminant with intellectual function in children. Clinical Toxicology 40,467-475. 2002.

Jacobson, JL, Janisse, J, Baneijee, M, Jester, J, Jacobson, SW, and Ager, JW. A benchmark dose analysis of prenatal exposure to polychlorinated biphenyls. Environmental Health Perspectives 110, 393-398. 2002.

Jana, NR, Sarkar, S, Ishizuka, M, Yonemoto, J, Tohyama, C, and Sone, H. Cross-talk between 2,3,7,8-Tetrachlorodibenzo-p -dioxin and testosterone signal transduction pathways in LNCaP prostate cancer cells. Biochemical and Biophysical Research Communications 256,462-468. 1999.

Jeandet, P, Bessis, R, Sbaghi, M, and Meunier, P. Production of the phytoalexin resveratrol by grapes as a response to Botrytis attack under natural conditions. Joumal of Phytopathology 143, 135-139. 1995.

271 Jinno, H, Hanioka, N, Onodera, S, Nishimura, T, and Ando, M. Irgason ® DP 300 (5- chloro-2-(2,4-dichlorophenoxy)- phenol) induces cytochrome P450s and inhibits haem biosynthesis in rat hepatocytes cultured on Marigel. Xenobiotica 27,681-692. 1997.

Johnson, TE, Holloway, MK, Vogel, R, Rutledge, SJ, Perkins, JJ, Rodan, GA, and Scmidt, A. Structural requirements and cell type specifity for ligand activation of peroxisome proliferator activated receptors. Joumal of Steroid Biochemistry and Molecular Biology 63,1-8. 1997.

Jones, RD, Jampani, HB, Newman, JK, and Lee, AS. Triclosan: A review of effectiveness and safety in health care setting. Am J Infect Control, 184-196. 2000.

Jones, SA, Moore, LB, Shenk, JL, Wisely, GB, Hamilton, GA, McKee, DD, Tomkinson, NCO, LeCluyse, EL, Lambert, MH, Wilson, TM, Kliewer, SA, and Moore, JT. The pregnane X receptor: a promiscuous xenobiotic receptor that has diverged during evolution. Molecular Endocrinology 14, 27-39. 2000.

Kakizaki, S, Karami, S, and Negishi, M. Retinoic acids repress constitutive active receptor-mediated induction by l,4-bis[2-(3,5-dichlorpyridyloxy)]benzene of the CYP2B10 gene in mouse primary hepatocytes. Dmg Metabolism and Disposition 30, 208-211.2002.

Karchner, SI, Franks, DG, Powell, WH, and Hahn, ME. Regulatory interactions among three members of the vertebrate aryl hydrocarbon receptor family: AHR repressor; AHRl and ANHR2. Joumal of Biological Chemistry 277, 6949-6959. 2002.

Kast, HR, Goodwin, B, Tarr, PT, Jones, SA, Anisfeld, AM, Stoltz, CM, Tontonoz, P, Kliewer, SA, Willson, TM, and Edwards, PA. Regulation of multidmg resistance- associated protein 2 (ABCC2) by the nuclear receptors pregnane X receptor, famesoid X-activated receptor, and constitutive androstane receptor. The Joumal of Biological Chemistry 277, 2908-2915. 2002.

Kastenholz, MA, Pastor, M, Cmciani, G, Haaksma, EE, and Fox, T. GRID/CPCA: A new computatonal tool to design selective ligands. Joumal of Medicinal Chemistry 43, 3033-3044. 2000.

Kawamoto, T, Seuyoshi, T, Zelko, I, Moore, R, Washbum, K, and Neigashi, M. Phenobarbital-responsive nuclear translocation of the receptor CAR in induction of the CYP2B gene. Molecular Cell Biology 19, 6318-6322. 1999.

Kelce, RW, Stone, RC, Laws, CS, Gray, EL, Kemppainen, AJ, and Wilson, ME. Persistent DDT metabolite /?,/?-DDE is a potent androgen receptor antagonist. Nature 373, 581- 585. 1995.

Kelley, ST and Thackray, VG. Phylogenetic analyses reveal ancient duplication of estrogen receptor isoforms. Joumal of Molecular Evolution 49, 609-614. 1999.

Kenealy, M-R, Flouriot, G, Sonntag-Buck, V, Dandekar, T, Brand, H, and Gannon, F. The 3'untranslated region of the human estrogen receptor a gene mediates rapid messenger ribonucleic acid tumover. Endocrinology 141, 2085-2113. 2000.

272 Kim, RB. Transporters and xenobiotic disposition. Toxicology 181-182,291-297. 2002.

Kiviranta, H, Vartiainen, T, and Tuomisto, J. Polychlorinated dibenzo-p-dioxins, dibenzofurans, and biphenyls in fishermen in Finland. Environmental Health Perspectives 110, 355-361. 2002.

Kiviranta, H, Vartianinen, T, Verta, M, Tuomisto, JT, and Tuomisto, J. High fish specific dioxin concentrations in Finland. Lancet 355, 1883-1885. 2000.

Klein, KO, Baron, J, Colli, MJ, McDonnell, DP, and Cutler Jr, GB. Estrogen levels in childhood determined by an ultrasensitive recombinant cell bioassay. Joumal of Clinical Investigation 94, 2475-2480. 1994.

Kliewer, SA, Lehmann, JM, and Willson, TM. Orphan nuclear receptors: shifting endocrinology into reverse. Science 284,757-760.1999.

Kliewer, SA, Moore, JT, Wade, L, Staudinger, JL, Watson, MA, Jones, SA, McKee, DD, Oliver, BB, Willson, TM, Zetterstrom, RH, Perlmann, T, and Lehmann, JM. An orphan nuclear receptor activated by pregnanes defines a novel steroid signalling pathway. Cell 92,73-82. 1998.

Klinge, CM, Kaur, K, and Swanson, HI. The Aryl hydrocarbon receptor interacts with estrogen receptor alpha and orphan receptors COUP-TFl and ERRot 1. Archives of Biochemistry and Biophysics 373 (1), 163-174. 2000.

Kreiger, N, Wolff, MS, Hiatt, RA, Rivera, M, Vogelman, J, and Orentreich, N. Breast cancer and semm organochlorines: a prospective study among white, black and Asian

women. Joumal of the National Cancer Institute 8 6 , 589-599. 1994.

Kroes, R, Galli, C, Munro, I, Schilter, B, Tran, L-A, Walker, R, and Wurtzen, G. Threshold of toxicological concem for chemical substances present in the diet: a practical tool for assessing the need for toxicity testing. Food and Chemical Toxicology 38,255-312. 2000.

Krstevska-Konstantinova, M, Charlier, C, Craen, M, Du Caju, M, Heinrichs, C, de Beafort, C, Plomteux, G, and Bourguignon, JP. Sexual precocity after immigration from developing countries to Belgium: evidence of previous exposure to organochlorine pesticides. Human Reproduction 16,1020-1026. 2001.

Kuhnle, G, Spencer, JPE, Chowrimootoo, G, Schroeter, H, Debnam, ES, Srai, KS, Rice- Evans, C, and Hahn, U. Resveratrol is absorbed in the small intestine as resveratrol glucuronide. Biochemical and Biophysical Research Communications 272, 212-217. 2000.

Kuiper, GCJM, Carlsson, B, Grandien, K, Enmark, E, Haggblad, J, Nilsson, S, and Gustafsson, J-Â. Comparison of the ligand binding specificity and transcript tissue distribution of estrogen receptors alpha and beta. Endocrinology 138, 863-870. 1997.

Kuiper, GCJM, Lemmen, JG, Carlsson, B, Corton, JC, Safe, SH, van der Saag, PT, van der Burg, B, and Gustaffson, J-A. Interaction of estrogenic chemicals and phytoestrogens

with estrogen receptor 6 . Endocrinology, 4252-4263. 1998.

273 Kumar, MB, Tarpey, RW, and Perdew, GH. Differential recruitment of coactivator RIP 140 by Ah and Estrogen receptors. Absence of a role of LXXLL motifs. The Joumal of Biological Chemistry 274, 22155-22164. 1999.

Kumnczi, L, Olah, M, Oprea, TI, Bologa, C, and Simon, Z. MTD-PLS: A PLS-based variant of the MTD Method 2. Mapping Ligand-Receptor Interactions. Enzymatic acetic acid esters hydrolysis. J Chem Inf Comput Sci 42, 841-846. 2002.

Kutz, FW, Bames, DG, Bottimore, DP, Greim, H, and Bretthauer, EW. The Intemational Toxicity Equivalency factor (I-TEF) method of risk assessment of complex mixtures of dioxins and related compounds. Chemosphere 20,751-757. 1990.

Lamartinière, CA, Moore, J Holland MB, and Bames, S. Neonatal genistein chemoprevents mammary carcinogenesis. Proceedings of the Society of Experimental Biology and Medicine 208,120-123. 1995.

Landes, N, Pfluger, P, Kluth, D, Birringer, M, Ruhl, R, Bol, G-F, Glatt, H, and Brigelius- Flohé, R. Vitamin E activates gene expression via the prennane X receptor. Biochemical Pharmacology 65, 269-273. 2003.

Latch, DE, Packer, JL, Amold, WA, and McNeill, K. Photochemical conversion of triclosan to 2,8-dichlorodibenzo-p-dioxin in aqueous solution. Joumal of Photochemistry and Photobiology A: Chemistry 158, 63-66. 2003.

Lau, K-M, Mok, SC, and Ho, S-M. Expression of human estrogen receptor a and (3, progesterone receptor, and androgen receptor mRNA in normal and malignant ovarian epithelial cells. Proc Natl Acad Sci USA 96, 5722-5727. 1999.

Lee, S-K, Lee, B, and Lee, JW. Mutations in retinoid X receptor that impair heterodimerization with specific nuclear hormone receptor. Joumal of Biological Chemistry 275, 33522-33526. 2000a.

Lee, H-Y, Suh, Y-A, Robinson, MJ, Clifford, JL, Hong, WK, Woodgett, JR, Cobbs, MH, Mangelsdorf, DJ, and Kurie, JM. Stress pathway activation induces phosphorylation of retinoid X receptor. Joumal of Biological Chemistry 275, 32193-32199. 2000b.

Lehmann, JM, McKee, DD, Watson, MA, Willson, TM, Moore, JT, and Kliewer, SA. The human orphan nuclear receptor PXR is activated by compounds that regulate CYP3A4 gene expression and cause dmg interactions. Joumal of Clinical Investigation 102, 1016-1023. 1998.

Leslie, EM, Deeley, RG, and Cole, SPC. Bioflavonoid stimulation of glutathione transport by the 190-kDa multidmg resistance protein 1 (MRPl). Dmg Metabolism and Disposition 31,11-15. 2003.

Lewis, DFV and Dickins, M. Baseline lipophilicity relationships in human cytochromes P450 associated with dmg metabolism. Dmg Metabolism Reviews 35, 1-18. 2003.

Lewis, DFV and Jacobs, MN. A QSAR study of some PCBs ligand-binding affinity to the cytosolic Ah Receptor. Organohalogen Compounds 41, 537-540. 1999.

274 Lewis, DFV and Lake, B. Molecular modelling of the rat peroxisome proliferator-activated receptor a (rPPARa) by homology with the human retinoic acid X receptor a (hRARa) and investigation of peroxisome proliferator binding interactions: QSARs. Toxicology in vitro 12,619-632. 1998.

Lewis, DFV. Cytochrome P450. 1996. London, UK, Taylor and Francis.

Liao, DJ and Dickson, RB. Roles of androgens in development, growth and carcinogenesis of the mammary gland. Journal of Steroid Biochemistry and Molecular Biology 80, 175-189. 2002.

Liem, AKD and Theelen, RMC. Dioxins: Chemical analysis, exposure and risk assessment. [PhD thesis.] National Institute of Public Health and the Environment, Bilthoven, The Netherlands. 1997.

Liem, AKD, Furst, P, and Rappe, C. Exposure of populations to dioxins and related compounds. Food Additives and Contaminants 17, 241-259. 2000.

Liggins, J, Grimwood, R, and Bingham, SA. Extraction and quantification of lignan phytoestrogens in food and human samples. Analytical Biochemistry 287,102-109. 2000.

Lin, JH and Yamazaki, M. Role of P-Glycoprotein in pharmacokinetics. Clinical Implications. Drug Disposition 42(1), 59-98. 2003.

Lindstrom, A, Buerge, IJ, Poiger, T, Bergqvist, P-A, Muller, MD, and Buser, H-R. Occurrence and environmental behaviour of the bactericide triclosan and its methyl derivative in surface waters and in wastewater. Environmental Science and Techolnology 36, 2322-2329. 2002.

Lipinski, CA, Lombardo, F, Dominy, BW, and Feeny, PJ. Experimental and Computational Approaches to Estimate Solubility and Permeability in Drug Discovery and Development Setting. Advanced Drug Delivery Reviews 23, 3-25. 1997.

Longnecker, MP, Rogan, WJ, and Lucier, G. The human health effects of DDT (Dichlorodiphenyltrichloroethane) and PCBS (Polychlorinated biphenyls) and an overview of organochlorines in public health. Annu Rev Public Health 18, 211-244. 1997.

Luo, G, Cunningham, M, Kim, S, Bum, T, Lin, J, Sinz, M, Hamilton, G, Rizzo, C, Jolley, S, Gilbert, D, Downey, A, Mudra, D, Graham, R, Carroll, K, Xie, J, Madan, A, Parkinson, A, Christ, D, Selling, B, Lecluyse, E, and Gan, L-S. CYP3A4 induction by drugs: correlation between a pregnane X receptor reporter gene assay and CYP3A4 expression in human hepatocytes. Drug Metabolism and Disposition 30,795-804. 2002.

Luo, X, Ikeda, Y, Lala, D, Rice, D, Wong, M, and Parker, KL. Steroidogenic factor-1 (SF- 1) is essential for endocrine development and function. Journal of Steroid Biochemistry and Molecular Biology 69,13-18. 1999.

275 Luscombe, C. The application of statistical tools to development of in-silico models for drug discovery: Modelling guideline. GSK Intranet, 1-20. 2003.

Lyman, FL and Furia, T. Toxicology of 2,4,4 -trichloro-2 -hydroxydiphenyl ether. Industrial Medicine and Surgery 38, 64-71. 1969.

Mac, MJ, Nicholson, LW, and McCauley, CA. The Progressive Fish Culturist 41, 210-211. 1979.

Mackie, D, Liu, J, Loh, Y-S, and Thomas, V. No evidence of dioxin cancer threshold. Environmental Health Perspectives 111 (9), 1145-1147. 2003.

Maglich, JM, Parks, DJ, Moore, LB, Collins, JL, Goodwin, B, Billin, AN, Stoltz, CA, Kliewer, SA, Lambert, MH, Willson, TM, and Moore, JT. Identification of a novel human Constitutive Androstane Receptor (CAR) agonist and its use in the identification of CAR target genes. Journal of Biological Chemistry 278,17277-17283. 2003

Maglich, JM, Stolz, CM, Goodwin, B, Hawkins-Brown, D, Moore, JT, and Kliewer, SA. Nuclear pregnane X receptor and constitutive androstane receptor regulate overlapping but distinct sets of genes involved in xenobiotic detoxification. Molecular Pharmacology 62, 638-646. 2002.

Makishima, M, Okamoto , AY, Repa, JJ, Tu, H, Learned, MR, Luk, A, Hull, MV, Lustig, KD, and Mangelsdorf, DJ. Identification of a nuclear receptor for bile acids. Science 284,1362-1368. 1999.

Manchester-Neesvig, JB, Valters, K, and Sonzogni, WC. Comparison of polybrominated diphenyl ethers (PBDEs) and polychlorinated biphenyls (PCBs) in Lake Michigan

salmoids. Environmental Science and Technology 35 ( 6 ), 1072-1077. 2001.

Marhefka, CA, Moore II, BM, Bishop, TC, Kirkovsky, L, Mukheijee, A, Dalton, JT, and Miller, DD. Homology modelling using multiple molecular dynamics simulations and docking studies of the human androgen receptor ligand binding domain bound to testosterone and nonsteroidal ligands. Journal of Medicinal Chemistry 44,1729-1740. 2001.

Martucci, CP and Fishman, J. P450 enzymes of estrogen metabolism. Pharmacology Therapy 57, 237-257. 1993.

Maruyama, W, Yoshida, K, Tanaka, T, and Nakanishi, J. Simulation of dioxin accumulation in human tissues and analysis of reproductive risk. Chemosphere, in press, 1-13. 2003.

Masahiko, N and Honkakoski, P. Induction of drug metabolism by nuclear receptor CAR: molecular mechanisms and implications for drug research. European Journal of Pharmaceutical Sciences 11, 259-264. 2000.

Masuyama, H, Hiramatus, Y, Kunitomi, M, Kudo, T, and MacDonald, PN. Endocrine disrupting chemicals, phthalic acid and nonylphenol activate pregnane X receptor- mediated transcription. Molecular Pharmacology 14,421-428. 2000.

276 Mato, Y, Isobe, T, Takada, H, Kanehiro, H, Ohtake, C, and Kaminuma, T. Plastic resin pellets as a transport medium for toxic chemicals in the marine environment. Environmental Science and Techology 35, 318-324. 2001.

Mathews, JB, Twomey, K, and Zacharewski, TR. In vitro and in vivo interactions of bisphenol A and its metabolite, bisphenol a glucuronide, with estrogen receptors a and (3. Chemical Research in Toxicology 14, 149-157. 2001.

Mauras, N, Rogol, AD, and Veldhuis, JD. Specific, time-dependent actions of low-dose ethinyl estradiol administration on the episodic release of growth hormone, follice- stimulating hormone, and luteinizing hormone in prepubertal girls with Turner’s syndrome. Journal of Clinical Endocrinology and Metabolism 69,1053-1058. 1989.

Maurel, P. The CYP3 family, in C. lonnides, (Ed.), Cytochromes P450: metabolic and toxicological aspects. CRC Press, Inc., Boca Raton, FL. pp 241-270. 1996.

Mautner, MN, Conner, AJ, Killham, K, and Deamer, DW. Biological potential of extraterrestrial materials 2. Microbial and plant responses to nutrients in the murchison carbonaceous meteorite. ICARUS 129, 245-253.1997.

McKenna, NJ, Xu, J, Nawaz, Z, Tsai, SY, Tsai, M-J, and OMalley, BW. Nuclear receptor coactivators: multiple enzymes, multiple complexes, multiple functions. Journal of Steroid Biochemistry and Molecular Biology 69, 3-12. 1999.

McLachlan, JA, Burow, M, Chiang, T-C, and Fang, Li S. Gene imprinting in developmental toxicology: a possible interface between physiology and pathology. Toxicology Letters 120,161-164. 2001.

McLachlan, JA. Synergistic effect of environmental oestrogens: report withdrawn. Science 277,463-464. 1997.

Meerts, I ATM, Letcher, RJ, Hoving, S, Marsh, G, Bergman, A, Lemmen, JG, van der Berg, B, and Brower, A. In vitro estrogenicity of polybrominated diphenyl ethers, hydroxylated PBDEs and polybrominated bisphenol A compounds. Environmental Health Perspectives 109, 399-407. 2001.

Mehrabi, MR, Steiner, GE, Dellinger, C, Kofler, A, Schaufler, K, Tamaddon, F, Plesch, K, Ekmekcioglu, C Maurer G, Glogar, HD, and Thalhammer, T. The arylhydrocarbon receptor (AhR), but not the AhR-nuclear translocator (ARNT), is increased in hearts of patients with cardiomyopathy. Virchows Arch 441,481-489. 2002.

Meironyté Guvenius, D, Aronsson, A, Ekman-Ordeberg, G, Bergman, A and Noren K. Human prenatal and post natal exposure to polybrominated diphenyl ethers, polychlorinated biphenyls, polychlorobiphenylols and pentachlorophenol. Environmental Health Perspectives 111 (9), 1235-1241. 2003.

Meironyté Guvenius, D, Bergman, A, and Noren, K. Polybrominated diphenyl ethers in Swedish human liver and adipose tissue. Arch Environ Contam Toxicol 40, 564-570. 2001.

277 Meironyté, D, Noren, K, and Bergman, A. Analysis of polybrominated diphenyl ethers in Swedish human milk. A time-related trend study, 1972-1997. Journal of Toxicology and Environmental Health Part A 58, 329-341.1999.

Memon, RA, Tecott, LH, Nonogaki, K, Beigneux, A, Moser, AH, Grunfeld, C, and Feingold, KR. Up-regulation of peroxisome proliferator-activated receptor a (PPARa) and PPAR Messenger ribonucleic acid expression in the liver in murine obesity: Troglitazone induces expression of PPAR responsive adipose tissue-specific genes in the liver of obese diabetic mice. Endocrinology 141,4021-4031. 2000.

Métivier, R, Stark, A, Flouriot, G, Hiibner, MR, Brand, H, Penot, G, Manu, D, Denger, S, Reid, G, Kos, M, Russell, RB, Kah, O, Pakdel, F, and Gannon, F. A dynamic structural model for estrogen receptor-a activation by ligands, emphasising the role of interactions between distant A and E domains. Molecular Cell 10,1019-1032. 2002.

Miller, D, Wheals, BB, Beresford, N, and Sumpter, JP. Estrogenic activity of phenolic additives determined by an m vitro yeast bioassay. Environmental Health Perspectives 109,133-138. 2001.

Miller, KJ. Additvity methods in molecular polaiizability. J Am Chem Soc 112, 8533-8542. 1990.

Milligan, SR, Balasubramanian, AV, and Kalita, JC. Relative potency of xenobiotic estrogens in an acute in vivo mammalian assay. Environmental Health Perspectives 106, 23-26. 1998.

Milligan, SR, Kalita, JS, Heyerick, A, Rong, H, De Coomsan, L, and De Keukeleire, D. Identification of a potent phytoestrogen in hops {Humulus lupulus L.) and beer. Journal of Clinical Endocrinology and Metabolism 84, 2249-2252. 1999.

Millward, DJ and Griffin, BA. Relative rates of long chain coversion of ^^C a-linolenic acid in adult men fed high a-linolenic acid or high linoleic acid diets. Draft paper. Project N02008, available from the UK Food Standards Agency, 2002.

Min, G, Kemper, JK, and Kemper, B. Glucocorticoid receptor interacting protein-1 (GRIP) mediates ligand-independent nuclear translocation and activation of constitutive androstane receptor (CAR) in vivo. Manuscript M200051200. Journal of Biological Chemistry papers in press . 8-5-2002.

Ministry of Agriculture Fisheries and Food / Department of Health. Dioxins and polychlorinated biphenyls in farmed trout. Food Safety Information Bulletin 94, MAFF, London. UK. 1998.

Ministry of Agriculture Fisheries and Food. Dioxins and Polychlorinated Biphenyls in Fish Oil Dietary Supplements and Licensed Medicines. Food Surveillance Information Sheet 106, MAFF, London. UK. 1997.

Ministry of Agriculture Fisheries and Food. Dioxins and Polychlorinated Biphenyls in Marine fish, fish and fish products. Food Safety Information Sheet 184, MAFF, London. UK. 1999.

278 Ministry of Agriculture Fisheries and Food. Polychlorinated biphenyls in food - UK dietary intakes Food Surveillance Information Sheet 89, MAFF, London. UK. 1996.

Miyamoto, M, Keneko, A, Kakizawa, T, Yajima, H, Kamijo, K, Sekine, R, Hiramatsu, K, Nishii, Y, Hashimoto, T, and Hashizume, K. Inhibition of peroxisome proliferator signaling pathways by thyroid hormone receptor - Competitive binding to the response element. J Biol Chem 272,7752-7758. 2003.

Miyamoto, T, Keneko, A, Kakizawa, T, Yajima, H, Kamijo, K, Sekine, R, Hiramatsu, K, Nishii, Y, Hashimoto, T, and Hashizume, K. Inhibition of peroxisome proliferator signaling pathways by thyroid hormone-receptor-competitive binding to the response element. Journal of Biological Chemistry 272,7752-7758. 1997.

Mizuyachi, K, Son, D-S, Rozman, KK, and Terranova, PR Alteration in ovarian gene expression in response to 2,3,7,8-tetrachlorodibenzo-p-dioxin: redultion of cyclooxygenase-2 in the blockage of ovulation. Reproduction Toxicology 16, 299-307. 2002.

Molecular Discovery Ltd. GRID version 19. Molecular Discovery Ltd. 20a, Berkeley Street, Mayfair, London,UK.

Moore, LB, Goodwin, B, Jones, SA, Wisely, GB, Willson, TM, Collins, JL, and Kliewer, SA. St. John’s Wort induces hepatic drug metabolism through activation of the pregnane X receptor. Proceedings of the New York Academy of Sciences USA 97, 7500-7502. 2000a.

Moore, LB, Parks, DJ, Jones, SA, Bledsoe, RK, Cousler, TG, Stimmel, JB, Goodwin, B, Liddle, C, Blanchard, SG, Willson, TM, Collins , JL, and Kliewer, SA. Orphan nuclear receptors constitutive antrostane receptor and pregnane X receptor share xenobiotic and steroid ligands. Journal of Biological Chemistry 275,15122-15127. 2000b.

Moreno, JJ. Reservatrol modulaties arachidonic acid release, prostaglandin synthesis, and 3T6 fibroblast growth. The Journal of Pharmacology and Experimental Therapeutics 294, 339-346. 2003.

Mortensen, DS, Rodriguez, AL, Carlson, KE, Sun, J, Katzenellenbogen, BS, and Katzenellenbogen, JA. Synthesis and biological evaluation of a novel series of furans: ligands selective for estrogen receptor a. Journal of Medicinal Chemistry 44, 3838- 3848. 2001.

Mosselman, S, Polman, J, and Dijkema, R. ER^: identification and characterization of a novel estrogen receptor. FEBS Letters 392,49-53. 1996.

Mueller, E, Sarraf, P, Tontonoz, P, Evans, RM, Martin, KJ, Zhang, M, Fletcher, C, Singer, S, and Spiegelman, BM. Terminal differentiation of human breast cancer through PPAR gamma. Molecular Cell 1,465-470. 1998.

Mutoh, M, Takahashi, M, Fukuda, K, Matsushima-Hibiya, Y, Mutoh, H, Sugimura, T, and Wakkabayashi, K. Suppression of cyclooxygenase-2 promoter-dependent

279 transcriptional activity in colon cancer cells by chemopreventive agents with a resorcin- type structure. Carcinogenesis 21, 959-963. 2000.

Nagy, SR, Sanborn, JR, Hammock, BD, and Denison, MS. Development of a green fluorescent protein-based cell bioassay for the rapid and inexpensive detection and characterization of Ah receptor agonists. Toxicological Sciences 65, 200-210. 2002,

Nallani SC, Goodwin B, Maglich JM, Buckley DJ, Buckley AR, and Desai PB. Induction of Cytochrome P450 3A by Paclitaxel in Mice: Pivotal Role of the Nuclear Xenobiotic Receptor, Pregnane X Receptor. Drug Metabolism and Disposition 31,681. 2003.

Naylor, RL, Goldburg, RJ, Beveridge, MCM, Clay, J, Folke, C, Kautsky N, Lubchenco J, Mooney H, Primavera JH, and Troell M. Effect of aquaculture on world fish supplies. Nature 405,1017-1024. 2000.

Nebert, DW, Eisen, HJ, Negishi, M, Lang, MA, Hjelmeland, LM, and Okey, AB. Genetic mechanisms controlling the induction of polysubstrate monooxygenase (P-450) activities. Ann Rev Pharmacol Toxicol 21,431-462. 1981.

Negishi, M and Honakoski, P. Induction of drug metabolism by nuclear receptor CAR: molecular mechanisms and implications for drug research. European Journal of Pharmalogical Science 11,259-264. 2000.

Negishi, M„ Moore R, and Seuyoshi T. Transient humanisation of nuclear receptor CAR in mouse liver. British Toxicology Society Annual Congress, Abstracts ,51. 2003.

Nelson, KG, Takahashi, T, Bossert, NL, Walmer, DK, and McLachlan, JA. Epidermal growth factor replaces estrogen in the stimulation of female genital tract growth and

differentiation. Proc Natl Acad Sci USA 8 8 , 21-25. 1991.

Neve EPA and Ingelman-Sundberg M. Molecular Basis for the Transport of Cytochrome P450 2E1 to the Plasma Membrane. Journal of Biological Chemistry 275,17130- 17135. 2000.

Newbold, RR, Padilla Banks, E, Bullock, B, and Jefferson, WN. Uterine Adenocarcinoma in Mice Treated Neonatally with Genistein. Cancer Research 61,4325-4328. 2001.

Nguyen, TA, Hoivik, D, Lee, J-E, and Safe, S. Interactions of nuclear receptor coactivator/corepressor proteins with the aryl hydrocarbon receptor complex. Archives of Biochemistry and Biophysics 367, 250-257. 1999.

Nielsen, M, Ruch, RJ, and Vang, O. Resveratrol reverses tumor-promotor-induced inhibition of gap-junctional intercellular communication. Biochemical and Biophysical Research Communications 275, 804-809. 2000.

Nixon, JB, Kamitani, H, Back, SJ, and Eling, TE. Evaluation of eicosanoids and NSAIDs as PPAR[gamma] ligands in colorectal carcinoma cells. Prostaglandins, Leukotrienes

and Essential Fatty Acids 6 8 , 323-330. 2003

Nolan, GT, Hood, SR, and Clarke, SE. The application of the PXR transient transfection assay in preclinical drug development. Drug Metabolism Reviews 35(1), 72. 2003.

280 Nolle, RT, Wisely, GB, Westin, S, Cobb, JE, Lambert, MH, Kurokawa, R, Rosenfeld, MG, Willson, TM, Glass, CK, and Milbum, MV. Nolle RT, Wisely GB, Westin S, Cobb JE, Lambert MH, Kurokawa R, Rosenfeld MG, Willson TM, Glass CK, Milbum MV. Ligand binding and co-activator assembly of the peroxisome proliferator-activated receptor-gamma. Nature 395, 137-143.1998.

North Sea Task Force. North Sea Quality Status Report Oslo and Paris Commissions/International Council for exploration of the Sea. London. 1993.

Nuclear Receptors Nomenclature Committee. A unified nomenclature system for the nuclear receptor family. Cell 97, 161-163. 1999.

Nutreco. Social and Environmental Report 2002. The Netherlands, Nutreco Holding N.V. 2003.

Ohtake, F, Take-yama, K, Matsumoto, T, Kitagawa, H, Yamamoto, Y, Nohara, K, Tohyama, C, Krust, A, Mimura, J, Chambon, P, Yanagisawa, J, Fuji-Kuriyama, and Kato. Modulation of oestrogen receptor signalling by association with the activated dioxin receptor. Nature 423,545-550.2003.

Olsson, T and Oprea, TI. Chemoinformatics: A tool for decision-makers in dmg discovery. Current Opinion in Dmg Discovery and Development 4, 308-313. 2001.

Omoto, Y, Kobayashi, S, Inoue, S, Ogawa, S, Toyama, T, Yamashita, H, Muramatsu, M, Gustafsson, J-A, and Iwase, H. Evaluation of oestrogen receptor p wild-type and variant protein expression, and relationship with clinicopathological factors in breast cancers. European Journal of Cancer 38, 380-386. 2002.

Oprea, TI, Kumnczi, L, Olah, M, and Simon, Z. MTD-PLS: a PLS-based variant of the MTS method, a 3D-QSAR analysis of receptor affinities for a series of halogenated dibenzoxin and biphenyl derivatives. SAR and QSAR in Environmental Research 12, 75-92. 2001.

Oprea, TI, Zamora, I, and Ungell, A-L. Pharmacokinetically based mapping device for chemical space navigation. Journal of Combinatorial Chemistry 4, 258-266. 2002.

Ostberg, T, Bertilsson, G, Jedeburg, L, Berkenstam, A, and Uppenberg, J. Identification of residues in the PXR ligand binding domain critical for species specific and constitutive activation. European Journal of Biochemistry 269,4896-4904. 2002.

Otake, Y, Nolan, AL, Walle, K, and Walle, T. Quercetin and resveratrol potently reduce estrogen sulfotransferase activity in normal human mammary epithelial cells. Journal of Steroid Biochemistry and Molecular Biology 73, 265-270. 2000.

Owsley, E and Chiang, JYL. Gugglesterone antagonizes famesoid X receptor induction of bile salt pump but activates pregnane X receptor to inhibit cholesterol 7a-hydroxylase gene. Biochemical and Biophysical Research Communications 304, 191-195. 2003.

281 Paech, K, Webb, P, Kuiper, GGJM, Nilsson, S, Gustafsson, J-A, Kushner, PJ, and Scanlan, TS, Differential ligand activation of estrogen receptors ER and ER at API sites. Science 277,1508-1510. 1997.

Paolini, M, Barillari, J, Broccoli, M, Pozzetti, L, Perocco, P, and Cantelli-Forti, G. Effect of liquorice and glycyrrhizin on rat liver carcinogen metabolizing enzymes. Cancer Letters 145,35-42.1999.

Parker, MG. Steroid and related receptors. Current Opinion in Cell Biology 5,499-504. 1993.

Parsley, K, Wright, C, and Thorpe, S. Report FD 97/66, CSL Food Science Laboratory, MAFF, London.UK. 1999.

Pascussi, J-M, Drocourt, L, Fabre, J-M, Maurel, P, and Vilarem, M-J. Dexamethasone induces pregnane X receptor and retinoid X receptora - expression in human hepatocytes: synergistic increase of CYP3A4 induction by pregnane X receptor activators. 58,361-372. 2000a.

Pascussi, J-M, Gerbal-Chaloin, S, Pichard-Garcia, L, Daujat, M, Fabre, J-M, Maurel, P, and

Vilarem, M-J. Interleukin- 6 negatively regulates the expression of pregnane X receptor and constitutively activated receptor in primary human hepatocytes. Biochemical and Biophysical Research Communications 274, 707-713. 2000b.

Pascussi, J-M, Jounaidi, Y, Drocourt, L, Domergue, J, Balabaud, C, Maurel, P, and Vilarem, M-J. Evidence for the presence of a functional Pregnane X Receptor response element in the CYP3A7 promoter gene. Biochemical and Biophysical Research Communications 260, 377-381. 1999.

Pastor, M, Cruciani, G, and Watson, KA. A strategy for the incorporation of water molecules present in a ligand binding site into a three-dimensional quantitative structure-activity relationship analysis. Journal of Medicinal Chemistry 40,4089-4102. 1997.

Pawlosky, RJ, Hibbeln, JR, Novotny, JA, and Salem, N, Jr. Physiological compartmental analysis of a-linolenic acid metabolism in adult humans. Journal of Lipid Research 42, 1257. 2001

Payne, J, Scholze, M, and Kortenkamp, A. Mixtures of four organoehlorines enhance human breast cancer cell proliferation. Environmental Health Perspectives 109, 391- 397.2001.

Penzotti, JE, Lamb, ML, Evensen, E, and Grootenhuis, PDJ. A computational ensemble pharmacophore model for identifying substrates of P-glycoprotein. Journal of Medicinal Chemistry 45,1737-1740. 2002.

Petersen, DN, Tkalcevic, GT, Koza-Taylor, PH, Turi, TG, and Brown, TA. Identification

of estrogen receptor beta ( 2 ), a functional variant of estrogen receptor beta expressed in normal rat tissues. Endocrinology 139, 1082-1092. 1998.

282 Petreas, M, She, J, Winkler, P, Visita, P, McKinney, M, Reynolds, P, Smith, D, Gilliss, D, Hurley, S, Jeffrey, S, and Mahoney, ME. Levels of PCDD/PCDFs, PCBs and OC pesticides in breast adipose tissue of women enrolled in a California breast cancer study. Organohalogen Compounds 38, 37-40. 1998.

Petreas, MJ, She, J, Brown, PR, Winkler, J, Windham, G, Rogers, E, Zhao, GR, Bhatia, R, and Charles, MJ. High body burdens of 2,2’,4,4’ - tetrabromo diphenyl ether (BDE-47) in California Women. Environmental Health Perspectives 111 (9), 1175-1179. 2003.

Petrulis, JR and Perdew, GH. The role of chaperone proteins in the aryl hydrocarbon receptor core complex. Chemico-Biological Interactions 141, 25-40. 2002.

Phelan, DM, Brackney, WR, and Denison, MS. The Ah receptor can bind ligand in the absence of receptor-associated heat-shock protein 90. Archives of Biochemistry and Biophysics 353,47-54. 1998b.

Phelan, D, Winter, GM, Lam, JC, and Dennison, MS. Activation of the Ah receptor signal transduction pathway by bilirubin and biliverdin. Archives of Biochemistry and Biophysics 357, 155-163. 1998a.

Phillips, A, Hood, S. R, Gibson, G, and Plant, NJ. CYP3A and nuclear receptor, messenger RNA expression in liver and HuH7 cells. Drug Metabolism Reviews, 35 supplement 1, 83. 2003.

Pietinen, P, Stumpf, K, Mannisto, S, Kataja, Vuusitupa, M, and Adlercreutz, H. Serum enterolactone and risk of breast cancer: a case-control study in eastern Finland. Cancer Epidemiology, Biomarkers and Prevention 10, 339-344. 2001.

Pike, ACW, Brzozowski, AM, Hubbard, RE, Bonn, T, Thorsell, A-G, Engstrom, O, Ljunggren, J, Gustafsson, J-Â, and Carlquist, M. Structure of the ligand-binding domain of oestrogen receptor beta in the presence of a partial agonist and a full agonist. The EMBO Journal 18,4608-4618. 1999.

Pimentel, GC and McClellan, AL. The Hydrogen Bond. Pauling, L. A Series of Chemistry Books. 1960. New York, US, W.H. Freeman and Company.

Plant, NJ and Gibson, GG. Evaluation of the toxicological relevance of CYP3A4 induction. Current Opinion in Drug Discovery and Development 6(1), 50-56. 2003.

Poellinger, L. Mechanistic aspects-the dioxin (aryl hydrocarbon) receptor. Food Additives and Contaminants 17,261-266. 2000.

Pool-Zobel, BL, Adlercreutz, H, Glei, M, Liegiebl, U, Sittlingon, J, Rowland, I, Wahala, K, and Rechkemmer, G. Isoflavonoids and lignans have different potentials to modulate oxidative genetic damage in human colon cells. Carcinogenesis 21, 1247-1252. 2000.

Porta, M, Malats, N, Jariod, M, Grimault, JO, Rifa, J, Carrato, A, Guamer, L, Salas, A, Santiago-Silva, M, Vorominas, JM, Andreu, M, and Real, FX. Serum concentrations of organochlorine compounds and K-ras mutations in exocrine pancreatic cancer. The Lancet 354, 2125-2129. 1999.

283 Procopio, M, Lahm, A, Tramonato, A, Bonati, L, and Pitea, D. A model for recognition of polychlorinated dibenzo-p-dioxins by the aryl hydrocarbon receptor. European Journal of Biochemistry 269, 13-18. 2002.

Rabenoelina, F, Nirdé, P, Servant, N, Sultan, C, Terouanne, B, and Auzou, G. A single amino acid substitition in the ligand binding domain of the human androgen receptor modulates the receptor transactivation capacity. P I28. 15th International Symposium of the Journal of Steroid Biochemistry and Molecular Biology, Munich 17-20 May 2002. 2002.

Raevsky, OA, Schaper, K-J, Artursson, P, and McFarland, JW. A novel approach for prediction of intestinal absorption of drugs in human based on hydrogen bond descriptors and structural similiarity. Quant Struct-Act Relat 20,402-413. 2002.

Rahman, F, Langford, K, Scrimshaw, MD, and Lester, JN. Polybrominated diphenyl ether (PBDE) flame retardants. The Science of the Total Environment 275,1-17. 2003.

Rajapakse, N, Ong, D, and Kortenkamp, A. Defining the impact of weakly estrogenic chemicals on the action of steroidal estrogens. Toxieological Sciences 60, 296-304. 2001.

Ramamoorthy, K, Wang, K, Chen, I-C, Safe, S, Norris, JD, McDonnell, DP, Gaido, KW, Bocchinfuso, WP, and Korach, KS. Potency of combined estrogenic pesticides. Science 275,405-406.1997.

Rappe, C, Bergqvist, P-A, and Kjeller, LO. Levels, Trends And Patterns Of PCDDs And PCDFs in Scandinavian Environmental Samples. Chemosphere 18, 651-658. 1989.

Rarey, M and Dixon, JS. Feature trees: a new molecular similarity measure based on tree matching. Journal of Computer Aided Molecular Design 12,471-490. 1998.

Raucy, JL. Regulation of CYP3A4 Expression in Human Hepatocytes by Pharmaceuticals and Natural Products. Drug Metabolism and Disposition 31, 533-539. 2003

Richards, JA, Petrel, TA, and Brueggemeier, RW. Signaling pathways regulating aromatase and cycooxygenases in normal and malignant breast cells. Journal of Steroid Biochemistry and Molecular Biology 80, 203-212. 2002.

Richtoff, J, Rylander, L, Jonsson, BAG, Akesson, H, Hagmar, L, Nilsson-Ehle, P, Stridsberg, M, and Giwercman, A. Serum levels of 2,2',4,4',5,5'-hexachlorobiphenyl (CB-153) in relation to markers of reproductive function in young males from the general Swedish population. Environmental Health Perspectives 111(4), 409-413. 2003.

RIVO Fisheries Institute of Holland. Report-C034/00. 2000.

Roberts, RA, James, NH, Woodyatt, NJ, Macdonald, N, and Tugwood, JD. Evidence for the suppression of apoptosis by the peroxisome proliferator activated receptor alpha (PPAR alpha). Carcinogenesis 19,43-48. 1998.

284 Royce, WF. Introduction to the Practice of Fishery Science. Academic Press Ltd. London, UK and San Diego, USA. 1996.

Rubin, GL, Zhao, Y, Kalus, AM, and Simpson, ER. Peroxisome proliferator-activated receptor g ligands inhibit estrogen biosynthesis in human breast adipose tissue: possible implications for breast cancer therapy. Cancer Research 60,1604-1608. 2000.

Russell, DW. Nuclear orphan receptors control cholesterol catabolism. Cell 97, 539-542. 1999.

Sadler, BR, Cho, SJ, Ishaq, KS, Chae, K, and Korach, KS. Three-dimensional quantitative structure-activity relationship study of nonsteroidal estrogen receptor ligands using the comparative molecular field analysis/cross-validated r^- guided region selection approach. Journal of Medicinal Chemistry 41, 2261-2267. 1998.

Safe, SH. Environmental and dietary estrogens and human health: is there a problem ? Environmental Health Perspectives 103, 346-351. 1995.

Safe, S, Connor, K, Ramamoorthy, K, Gaido, K, and Maness, S. Human exposure to endocrine-active chemicals: hazard assessment problems. Regulatory Toxicology and Pharmacology 26, 52-58. 1997.

Safe, S. Xenoestrogens and breast cancer. New England Journal of Medicine 337,1303- 1304. 1997a.

Safe, SH. Is there an association between exposure to environmental estrogens and breast cancer? Environmental Health Perspectives 105, 675-678. 1997b.

Saito, H, Kotani, Y, Keriko, JM, Xue, C, Taki, K, Ishihara, K, Ueda, T, and Miyata, S. High levels of n-3 polyunsaturated fatty acids in Euphausia pacifica and its role as a source of docosahexaenoic and icosapentaenoic acids for higher trophic levels. Marine Chemistry 78, 9-28. 2002.

Saliou, C, Rihn, B, Cillard, J, Okamoto, T, and Packer, L. Selective inhibition of NF- kB activation by the flavonoid hepatoprotector silymarin in HepG2 Evidence for different activating pathways. FEBS Letters 440, 8-12. 1998.

Sarkar, MA, Vadlamuri, V, Ghosh, S, and Glover, DD. Expression and Cyclic Variability of CYP3A4 and CYP3A7 Isoforms in Human Endometrium and Cervix During the Menstrual Cycle. Drug Metabolism and Disposition 31, 1-6. 2003

Sauerwald, TU, Hachey, DL, Jensen, CL, Chen, H, Anderson, RE, and Heird, WC. Effect of dietary -linolenic acid intake on incorporation of DHA and AA into plasma phospholipids of term infants. Lipids 31, S131-S135. 1996.

285 Scaglione-Sewell, A, Bissonnette, M, Skarosi, S, Abraham, C, and Brasitus, A. Vitamin D 3

analog induces a G1-phase arrest in CaCo- cells by inhibiting Cdk2 and Cdk 6 : roles of cyclin E, p21 wafl, and p27 Kipl. Endocrinology 141,3931-3939. 2000.

Scambia, G, De Vincenzo, R, Ranelletti, FO, Panici, PB, Ferrandina, G, D’Agostino, G, Fattorossi, A, Bombardelli, E, and Mancuso, S. Antiproliferative effect of silybin on gynaecological malignancies: synergism with cisplatin and doxorubicin. European Journal of Cancer 32A, 877-882. 1996.

Schaldach, CM, Riby, J, and Bjeldanes, LF. Lipoxin A4 : A new class of ligand for the Ah receptor. Biochemistry 38,7594-7600. 1999.

Schepens, PJC, Covaci, A, Jorens, PG, Hens, L, Scharpe, S, and van Larebeke, N. Suprising findings following a belgium food contamination with polychlorobiphenyls and dioxins. Environmental Health Perspectives 109,101-103. 2001.

Schmidt, CW. Answering the endocrine test questions. Environmental Health Perspectives 107, A458-A460. 1999.

Schmidt, JV, Su, G H-I, Reddy, JK, Simon, MC, and Bradfield, CA. Characterisation of a murine A/iR null allele: Involvement of the Ah receptor in hepatic growth and development. Proc Natl Acad Sci USA 93, 6731-6736. 1996.

Schmiedlin-Ren, P, Thummel, KE, Fisher, JM, Paine, MF, and Watkins, PB. Induction of

CYP3 A4 by 1 a, 25-dihydroxy vitamin D 3 is human cell line specific and is unlikely to involve pregnane X receptor. Drug Metabolism and Disposition 29,1446-1453. 2001.

Schmiedlin-Ren, P, Thummel, KE, Fisher, JM, Paine, MF, Lown, KS, and Watkins, PB. Expression of enzymatically active CYP3A4 by Caco-2 cells grown on extracellular

matric-coated permeable supports in the presence of la, 25-dihydroxyvitamin D 3 . Molecular Pharmacology 51,741-754.1997.

Schneder, UA, Brown, MM, Logan, RA, Millar, LC, and Bunce, NJ. Screening assay for dioxin-like compounds based on competitve binding to the murine hepatic Ah receptor. 1. Assay development. Environmental Science and Technology 29, 2595-2602. 1995.

Schuetz, EG, Brimer, C, and Schuetz, JD. Environmental xenobiotics and the antihormones cyproterone acetate and spironolactone use the nuclear hormone pregnenolone X receptor to activate the CYP3A23 hormone response element. Molecular Pharmacology 54,1113-1117. 1998.

Schuetz, EG, Strom, S, Yasuda, K, Lecureur, V, Assem, M, Brimer, C, Lamba, J, Kim, RB, Ramachandran, V, Komoroski, BJ, Venkataramanan, R, Cai, H, Sinai, CJ, Gonzalez, FJ, and Schuetz, JD. Disrupted bile acid homeostasis reveals an unexpected interaction among nuclear hormone receptors, transporters and cytochrome P450. The Journal of Biological Chemistry 276, 39411-39418. 2001.

Scientific Committee on Food. Opinion of the SCF on the risk assessment of dioxins and dioxin like PCBs in Food. European Commission, Health and Consumer Protection Directorate-General. Available from

http://europa.eu.int/comm/food/fs/sc/scf/index en.html. 2 0 0 0 a.

286 Scientific Committee on Food. Opinion of the Scientific Committee on animal nutrition on the dioxin contamination of feeding stuffs and their contribution to the contamination of food of animal origin. European Commission Health and Consumer Protection Directorate-General. 6 November 2000. Available from http://europa.eu.int/comm/food/fs/sc/scf/index en.html. 2000b.

Seidlova-Wuttke, D, Jarry, H, Christoffel, V, and Wuttke, W. Estrogen receptor P -selective action of Silybum Marianum. 'Recent Advances in Steroid Biochemistry and Molecular Biology' Abstracts of lectures and poster presentations. Journal of Steroid Biochemistry and Molecular Biology 15th International Symposium , 105-P. 2002. Munich, Germany.

Senft, AP, Dalton, TP, Nebert, DW, Genter, MB, Puga, A, Hutchison, RJ, Kerzee, JK, Uno, S, and Shertzer, HG. Mitochondrial reactive oxygen production is dependent on the aromatic hydrocarbon receptor. Free Radical Biology and Medicine 33,1268-1278, 2002.

Sharpe, RM and Skakkebaek, NE. Are oestrogens in involved in falling sperm counts and disorders of the reproductive tract ? Lancet 341, 1392-1395. 1993.

Sharpe, RM, Atanaessova, N, McKinnell, C, Parte, P, Turner, KJ, Fisher, JS, Kerr, JB, Groome, NP, Macpherson, S, Millar, MR, and Sanders, PTK. Abnormalities in functional development of the Sertoli cells in rates treated neonatally with diethylstilbestrol: A possible role for estrogens in Sertoli cell development. Biology of Reproduction 59,1084-1094. 1998.

Sharpe, RM. Do males rely on female hormones ? Nature 390,447-448. 1997.

Sharpe, RM. Hormones and testis development and the possible adverse effects of environmental chemicals. Toxicology Letters 120, 221-232. 2001.

Sheeler, CQ, Dudley, MW, and Khan, SA. Environmental estrogens induce transcriptionally active estrogen receptor dimers in yeast: Activity potentiated by the coactivator RIP 140. Environmental Health Perspectives 108,97-103. 2000.

Shen, K and Novak, RF. DDT stimulates c-erbB2, c-met, and STATS tyrosine phosphorylation, Grb2-sos association, MAPK phosphorylation, and proliferation of human breast epithelial cells. Biochemical and Biophysical Research Communications 231,17-21. 1997.

Shen, K and Novak, RF. Differential effects of aroclors and DDT on growth factor gene expression and receptor tyrosine kinase activity in human breast epithelial cells. Advances in Experimental Medicine and Biology 407, 295-302. 1997.

Sher, T, Yi, H-F, McBride, OW, and Gonzalez, FJ. cDNA-cloning, chromosomal mapping and functional characterisation of the human peroxisome proliferatior activated receptor. Biochemistry 32, 5598-5604. 1993.

Sheridan, RP, Miller, MD, Underwood, DJ, and Kearsley, SK. Chemical similarity using geometric atom pair descriptors. J Chem Inf Comput Sci 36,128-136. 1996.

287 Shevtsov, S, Petrotchenko, EV, Pedersen, L, and Negishi, M. Crystallographic analysis of a hydroxylated polychlorinated biphenyl (OH-PCB) bound to the catalytic estrogen binding site of human estrogen sulphotransferase. Environmental Health Perspectives 111(7), 884-888. 2003.

Shughrue, PJ and Merchenthaler, I. Estrogen is more than just a "sex hormone": Novel sites for estrogen action in the hippocampus and cerebral cortex. Frontiers in Neuroendocrinology 21, 95-101. 2000.

Silva, E, Rajapanske, N, and Kortenkamp, A. Something from "nothing"-eight weak estrogenic chemicals combined at concentrations below NOECs produce significant mixture Effects. Environ Sci Techol 36, 1751-1756. 2002.

Simpson, ER, Rubin, G, Clyne, C, Robertson, K, ODonnell, L, Jones, M, and Davis, S. The role of local estrogen biosynthesis in males and females. Trends in Environmental Medicine 11, 184-188. 2000.

Sinai, CJ and Bend, RJ. Aryl hydrocarbon receptor-dependent induction of CyplAl by bilirubin in mouse hepatatoma hepa lclc7 cells. Molecular Pharmacology 52, 590-599. 1997.

Soffer, AEMF, Boersma, MG, Vases, WHJ, Vervoot, J, Tyrakowska, B, Hermens, JLM, and Rietjens, IMCM. Computer-modelling-based QSAR for analysing experimental data on biotransformation and toxicity. Toxicology in vitro 15, 539-551. 2001.

Soil Association. UK Organic Aquaculture Standards Final Food Certification Scotland Organic Food Federation, 2000.

Solomon, GM and Weiss, PM. Chemical contaminants in breast milk: time trends and regional variability. Environmental Health Perspectives 110, A339-A347. 2002.

Song, C and Liao, S. Cholestenoic acid is a naturally occurring ligand for liver X Receptor a. Endocrinology 141,4180-4184. 2000.

Song, J, Clagett-Dame, M, Peterson, RE, Hahn, ME, Westler, WM, Sicinski, RR, and DeLuca, HE. A ligand for the aryl hydrocarbon receptor isolated from lung. Proceedings of the National Academy of Sciences 99,14694-14699. 2002.

Soto, AM, Sonnenschein, C, Chung, KL, Fernandez, MF, Olea, N, and Serrano, FO. The E- SCREEN assay as a tool to identify estrogens: an update on estrogenic environmental pollutants. Environmental Health Perspectives 103, 113-122. 1995.

Staudinger, JL, Madan, A, Carol, KM, and Parkinson, A. Regulation of Drug Transporter Gene Expression by Nuclear Receptors. Drug Metabolism and Disposition 31, 523. 2003.

Stevens, JT, Breckenridge, CB, Wetzel, LT, Gillis, JH, Luempert, LG, and Eldridge, LC. Hypothesis for mammary tumourigenesis in Sprague-Dawley rats exposed to certain triazine herbicides. Journal of Toxicology and Environmental Health 43, 139-153. 1994.

288 Straessen, JA, Nawrot, T, Hond, ED, Thijs, L, Fagard, R, Hoppenbrouwers, K, Koppen, G, Nelen, V, Schoeters, G, Vanderschueren, D, van Hecke, E, Verschaeve, L, Vlietinck, R, and Rods, HA. Renal function, cytogenic measurements, and sexual development in adolescents in relation to environmental pollutants: a feasibility study of biomarkers. Lancet 357,1660-1669. 2001.

Sueyoshi, T, Kawamoto, T, Zelko, I, Honakakoshi, P, and Neighashi, M. Phenobarbital- responsive enhancer module (PBREM) of human CYP2B6 gene and inducible transactivation by nuclear receptor CAR.. Journal of Biological Chemistry 274, 6043- 6046. 1999.

Sugatani, J, Kojima, H, Ueda, A, Kakizaki, S, Yoshinari, K, Gong, Q-H, Owens, IS, Negishi, M, and Sueyoshi, T. The phénobarbital response enhancer module in the human bilirubin UDP-glucuronosyltransferaseU G TlAl gene and regulation by the nuclear receptor CAR. Hepatology 33,1232-1238. 2001.

Suzuki, T, Ide, K, Ishida, M, and Shapiro, S. Classification of environmental estrogens by physiochemical properties using principal component analysis and hierarchial cluster analysis. J Chem Inf Comput Sci 41,718-726. 2001.

Svensson, B-G, Nilsson, A, Hansson, M, Rappe, C, Akesson, B, and Skerfving, S. Exposure to dioxins and dibenzofurans through the consumption of fish. New England Journal of Medicine 324, 8-12. 1991.

Swales, KE. An investigation of host cell effects on the xenobiotic induction of cytochrome P450 3A. University of Surrey PhD Thesis 2002.

Swanson, HI and Bradfield, CA. Swanson HI, Bradfield CA The AH-receptor: Genetics, structure and function. Pharmacogenetics 3, 213-230. 1993.

Synold, TW, Dussault, I, and Forman, BM. The orphan nuclear receptor SXR coordinately regulates drug metabolism and efflux. Nature Medicine 7, 584-590. 2001.

Takeshita, A, Koibuchi, N, Oka, J, Taguchi, M, Shishiba, Y, and Ozawa, Y. Bisphenol-A, an environmental estrogen, activates the human orphan nuclear receptor, steroid and xenobiotic receptor-mediated transcription. European Journal of Endocrinology 145, 513-517.2001.

Takikawa, H. Hepatobiliary transport of bile acids and organic anions. J Hepatobiliary Pancreat Surg 9,443-447. 2002. ten Tusscher, GW and Koppe, JG. Perinatal dioxin exposure and later effects-a review. Chemosphere. In Press, Corrected Proof. 2003.

289 Tetsuka, M, Anderson, RA, and Hillier, SG. Distribution of messenger RNA’s encoding oestrogen receptor isoforms in female reproductive tissues. Journal of endocrinology 152, O C 20.1997.

The Merck Index. (12th ed.). 1996. Whitehouse Station, NJ, Merck and Co Inc.

Thornton, JW. Evolution of vertebrate steroid receptors from an ancestral estrogen receptor by ligand exploitation and serial genome expansions. Proceedings of the National Academy of Sciences 98, 5671-5676. 2001.

Thornton, MJ, Taylor, AH, Mulligan, K, Al-Azzawi, F, Lyon, CC, ODriscoll, J, and Messenger, AG. Oestrogen receptor beta is the predominant oestrogen receptor in human scalp skin. Experimental Dermatology 12 (2), 181-190. 2003.

Thummel, KE, Brimer, C, Yasuda, K, Thottassery, J, Senn, T, Lin, Y, Ishizuka, H, Kharasch, E, Schuetz, J, and Schuetz, E. Transcriptional control of intestinal

cyctochrome P-4503A by la,25- Dihydroxy Vitamin D 3. Mol Pharmacology 60,1399- 1406. 2001.

Tielemans, E, van Kooij, R, Te Velde, ER, Burdoff, A, and Heederik, D. Pesticide exposure and decreased fertilisation rates in vitro. Lancet 354,484-485. 1999.

Tirona, RG, Leake, BF, Wolkoff, AW, and Kim, RB. Human organic anion trasporting polypeptide-C (SLC21A6) is a major determinant of rifampin-mediated pregnane X receptor activation. The Journal of Pharmacology and Experimental Therapeutics 304, 223-228. 2003.

Toell, A, Kroncke, K-D, Kleinert, H, and Carlberg, C. Orphan nuclear receptor binding site in the human inducible nitric oxide synthase promoter mediates responsiveness to steroid and xenobiotic ligands. Journal of Cellular Biochemistry 85, 72-82. 2002.

Tou, JCL and Thompson, LU. Exposure to flaxseed or its lignan component during different development stages influences rat mammary gland structures. Carcinogenesis 20,1831-1835. 1999.

Trouw Aquaculture. Fish Feed. Feeding for profit. Trouw Aquaculture, Wincham, North wich Cheshire, UK.

Tzameli, I and Moore, DD. Role reversal: new insights from new ligands for the xenobiotic receptor CAR. Trends in Endocrinology and Metabolism 12,7-10. 2001.

Uauy, R, Mena, P, and Rojas, C. Essential fatty acids in early life: structural and functional role. Proceedings of the Nutrition Society 59, 3-15. 2000.

UK Committee on Toxicity of Chemicals in Food. Consumer products and the Environment. Statement on the dietary exposure to dioxins and dioxin like PCBs August 2000 COT Statement 2000/03. 2000.

Van de Waterbeemd, H, Smith, DA, and Jones, BC. Lipophilicity in PK design: methyl, ethyl, futile. Journal of Computer Aided Molecular Design 15,273-286. 2001.

290 van den Berg, M, Bimbaum , LS, Bosveld, ATC, Brunstrom, B, Cook, P, Feeley, M, Giesy, JP, Hanberg, A, Hasegawa, R, Kennedy, SW, Kubiak, T, Larsen, JC, van Leeuwen, FXR, Liem, AKD, Nolle, C, Peterson, RE, Poellinger, L, Safe, S, Schrenk, D, Tillit, D, Tysklind, M, Younes, M, Waem, F, and Zacharewski, TR. Toxic equivalency factors (TEFs) for PCBs, PCDDs, PCDFs for humans and wildlife. Environmental Health Perspectives 106,775-792. 1998. van Leeuwen, FX and Younes, M. WHO revises the Tolerable Daily Intake (TDI) for dioxins. Organohalogen Compounds 38, 295-298. 1998.

Vendrame, R, Braga, RS, Takahata, Y, and Galvao, DS. Structure-activity relationship studies of carcinogenic activity of polycyclic aromatic hydrocarbons using calculated molecular descriptors with principal component analysis and neural network methods. J Chem Inf Comput Sci 39,1094-1104. 1999.

Veterinary Medicines Directorate. The Veterinary Medicines Directorate Annual Report on Surveillance for Veterinary Residues in 1997. MAFF, London. UK. 1998.

Vuorinen, PJ, Vartiainen, T, and Keinanen, M. Polychlorinated dibenzo-p- dioxins, dibenzofurans and biphenyls in salmon (Salmo salar ) from the southern Baltic sea and the arctic Tenjoki river. Vartiainen, T, Komulainen, and H, Eds. 7th Nordic Symposium on Organic Pollutants. Natural and Environmental Sciences 68.1997. Finland, Kupioi University Publications.

Waem, F. Studies on polychlorinated dibenzo-/?-dioxins and dibenzofurans with emphasis on their relative potency and interactions in the rat. PhD Thesis 1995 Karolinska Institute. Cited in Bernes, C. Persistent Organic Pollutants: A Swedish View of an International Problem Monitor 16., Sweden (First issue 1998). 1995. Sweden, Swedish Environment Agency.

Waller, CL, Oprea, TI, Chae, K, Park, HK, Korach, KS, Laws, SC, Wiese, TE, and Kelce, WR. Ligand based identification of environmental estrogens. Chem Res Toxicol 9, 1240-1248. 1996.

Wan, Y-J Y, An, D, Cal, Y, Repa, JJ, Chen , T H-P, Flores, M, Postic, C, Magnuson , MA, Chen , J, Chien, KR, French, S, Mangelsdorf, DJ, and Sueov, HM. Hepatocyte-specific mutation establishes retinoid X receptor a as a heterodimeric integrator of multiple physiological processes in the liver. Molecular and Cellular Biology 20,4436-4444. 2000.

Wang H, Faucette SR, Gilbert D, Jolley SL, Sueyoshi T, Negishi M, and LeCluyse EL. Glucocorticoid Receptor Enhancement of Pregnane X Receptor-Mediated CYP2B6 Regulation in Primary Human Hepatocytes. Drug Metabolism and Disposition, 31, 620- 630. 2003.

Wang, H, Chen, J, Hollister, K, Sowers, LC, and Forman, BM. Endogenous bile acids are ligands for the nuclear receptor FXR/BAR. Molecular Cell 3, 543-553. 1999.

Wang, S and Hankinson, O. Functional involvement of the Brahma/SWI2-related gene 1 protein in cytochrome P4501A1 transciption mediated by the aryl hydrocarbon receptor complex. Journal of Biological Chemistry 277,11821-11827. 2002.

291 Warner, M, Nilsson, S, and Gustafsson, J-A. The estrogen receptor family. Current Opinion in Obstetrics and Gynecology 11, 249-254. 1999.

Watkins, RE, Wisely, GB, Moore, LB, Collins, JL, Lambert, MH, Williams, SP, Willson, TM, Kliewer, SA, and Redinbo, MR. The human nuclear xenobiotic receptor PXR: structural determinants of directed promiscuity. Science 292, 2329-2333. 2001.

Waxman, DJ. P450 gene induction by structurally diverse xenochemicals: central role of nuclear receptors CAR, PXR and PPAR. Archives of Biochemistry and Biophysics 369,11-23. 1999.

Weigel, N. Steroid hormone receptors and their regulation by phosphorylation. Biochem J 319, 657-667. 1996.

Weihua, Z, Lathe, R, Warner, M, and Gustafsson, J-A. An endocrine pathway in the prostate, ER(3, AR, 5a-androstane-3p, 17p-diol, and CYPBl, regulates prostate growth. Proceedings of the National Academy of Sciences 99,13589-13594. 2002.

Welshons, WV, Thayer, KA, Judy, BM, Taylor, JA, Curran, EM, and vom Saal, FS. Large effects from small exposures. 1. Mechanisms for endocrine disrupting chemicals with estrogenic activity. Environmental Health Perspectives 111(8), 994-1006. 2003.

Whitten, PL and Patisaul, HB. Cross-species and interassay comparisons of phytoestrogen action. Environmental Health Perspectives 109, 5-20. 2001.

Wiebel, FF, Steffensen, KR, Treuter, E, Feltkamp, D, and Gustafsson, J-A. Ligand- independent coregulator recruitment by the triply activatable OR 1/Retinoid X Receptor - a nuclear receptor heterodimer. Molecular Endocrinology 13,1105-1118. 1999.

Wilkinson, PA, Ah-Sing, E, Fereday, A, Millward, DJ, and Griffin, BA. The importance of alpha-linolenic acid as a source of long chain n-3 polyunsaturated fatty acids and its influence on risk factors of cardiovascular disease. Proceedings of the Nutrition Society 5 9 ,16A. 1999.

Williams, SP, Bledsoe, RK, Collins, JL, Boggs, S, Lambert, MH, Miller, AB, Moore, JT, McKee, DD, Moore, L, Nichols, J, Parks, D, Watson, M, Wisely, B, and Willson, TM X-ray crystal structure of the liver X receptor ligand binding domain: Regulation by a histidine-tryptophan switch. Journal of Biological Chemistry. In press. Corrected proof. M302260200. 2003.

Willson, TM and Kliewer, SA. The Science of orphan nuclear receptors. Core findings, a gene - to -function approach to drug discovery, Odyssey 97,16-22. 2000.

Willson, TM, Brown, PJ, Stembach, DD, and Henke, BR. The PPARs: From orphan receptors to drug discovery. Journal of Medicinal Chemistry 43, 527-550. 2000.

Wise, PM, Dubai, DB, Wilson, ME, Ran, SW, and Bottner, M. Minireview: Neuroprotective effects of estrogen - new insights into mechanisms of action. Endocrinology 142, 969-973. 2001.

292 Wold, Svante and Ericksson, Lennart. Statistical validation of QSAR results, van de Waterbeemd, Han. Chemometric Methods in Molecular Design. 2, 309-318.1995. Weinheim, VCH. Methods and Principles in Medicinal Chemistry. Eds. Mannhold, R, Krogsgaard-Larsen, P, and Timmerman, H.

Wold, S, Sjostrom, M, and Eriksson, L. PLS Regression: a basic tool of chemometrics. Chemometrics and Intelligent Laboratory Systems 58,109-130. 2001.

Wolff, MS and Toniolo, PG. Environmental organochlorine exposure and potential etiologic factor in breast cancer. Environmental Health Perspectives 103, 141-145. 1995.

Wolff, MS, Toniolo, PG, Lee, EW, Rivera, M, and Dublin, N. Blood levels of organochlorine residues and risk of breast cancer. Journal of the National Cancer Institute 85, 648-652. 1993.

World Health Organisation. Brominated Diphenyl Ethers. Environmental Health Criteria 162. Geneva. Switzerland. WHO ISBN 92 4 1571624 4. 1994.

World Health Organisation. Food safety issues associated with products from aquaculture. Report of a joint FAO/N AC A/WHO study group. WHO Technical report series 883. World Health Organisation, Geneva. 1999.

Wyde, ME, Bartolucci, E, Ueda, A, Zhang, H, Yan, BF, Negishi, M, and You, L. The

environmental pollutant 1 ,1 -dichloro- 2 ,2 -bis (p-chlorophenyl)ethylene induces rat hepatic cytochromeP4502B and 3A expression through the constitutive androstane receptor and pregnane X receptor. Molecular Pharmacology 64 (2), 474-481. 2003.

Wynne-Edwards, KE. Evolutionary biology of plant defenses against herbivory and their predictive implications for endocrine distruptor susceptibility in vertebrates. Environmental Health Perspectives 109,443-448. 2001.

Xia, Z-Q, Costa, MA, Pélissier, HC, Davin, LB, and Lewis, NG. Secoisolariciresinol dehydrogenase purification, cloning, and functional expression, implications for human health protection. Journal of Biological Chemistry 276,12614-12623. 2001.

Xie, H-G, Kim, RB, Wood, AJJ, and Stein, CM. Molecular basis of ethnic differences in drug disposition and response. Annu Rev Pharmacol Toxicol 41, 815-850. 2001,

Xie, W and Evans, RM. Orphan nuclear receptors: the exotics of xenobiotics. Journal of Biological Chemistry 276,37739-37742. 2001.

Xie, W, Barwick, JL, Simon, CM, Pierce, AM, Safe, S, Blumberg, B, Guzelian, PS, and Evans, RM. Reciprocal activation of xenobiotic response genes by nuclear receptors SXR/PXR and CAR. Genes and Development 14, 3014-3023. 2000.

Xu, HE, Lambert, MH, Montana, VG, Parks, DJ, Blanchard, SG, Brown, PJ, Stembach, DD, Lehmann, JM, Wisely, GB, Willson, TM, Kliewer, SA, and Mibum, MV. Molecular recognition of fatty acids by peroxisome proliferator activated receptors. Mol Cell 3, 397-403. 2003a.

293 Xu, HE, Lambert, MH, Montana, VG, Plunket, KD, Moore, LB, Collins, JL, Oplinger, JA, Kliewer, SA, Gampe Jr, RT, McKee, DD, Moore, JT, and Willson, TM. Structural determinants of ligand binding specifricty between the peroxisome proliferator activated receptors. Proc Natl Acad Sci USA 98,13919-13924. 2003b.

Xu, HE, Stanley, TB, Montana, VG, Lambert, MH, Shearer, BG, Cobb, JE, McKee, DD, Galardi, CM, Plunket, KD, Nolte, RT, Parks, DJ, Moore, JT, Kliewer, SA, Willson, TM, and Stimmel, JB. Structural basis for antagonist-mediated recruitment of nuclear co-repressors by PPAR. Nature 415, 813-817. 2002.

Yamasaki K, Takeyoshi M, Yakabe Y, Sawaki M, and Takatsuki M. Comparison of the reporter gene assay for ER-alpha antagonists with the immature rat uterotrophic assay of 10 chemicals. Toxicology Letters 142,119-131. 2003.

Yao, G and Bradfield, CA. Visualising the regulatory network of Ah receptor signalling using a combined and experimental approach. British Toxicology Society Annual Congress, Abstracts , 54. 2003.

Yoshikawa, T, Ide, T, Shimano, H, Yahagi, N, Amemiya-Kudo, M, Matsuzaka, T, Yatoh, S, Kitamine, T, Okazaki, H, Tamura, Y, Sekiya, M, Takahashi, A, Hasty, AH, Sato, R, Sone, H, Osuga, J, Ishibashi, S, and Yamada, N. Cross-Talk between Peroxisome Proliferator-Activated Receptor (PPAR) a and Liver X Receptor (LXR) in Nutritional Regulation of Fatty Aeid Metabolism. I. PPARs Suppress Sterol Regulatory Element Binding Protein-Ic Promoter through Inhibition of LXR Signaling. Molecular Endocrinology 17,1240-1254. 2003

Zhao, C, Koide, A, Abrams, J, Deighton-Collins, S, Martinez, A, Schwartz, JA, Koide, S, and Skafar, DF. Mutation of L536 in the human estrogen receptor-alpha alters the coupling between ligand-binding, transcription activation, and receptor conformation. Journal of Biological Chemistry In press. Corrected proof. M303840200. 2003

Zhou, S, Gao, Y, Jiang, W, Huang, M, Xu, A, and Paxton, JW. Interactions of herbs with cytochrome P450. Drug Metabolism Reviews 35,35-98. 2003.

Zhou, T, Ross, DG, DeVito, MJ, and Crofton, KM. Effects of short term in vivo exposure to polybrominated diphenyl ethers on thyroid hormone and hepatic enzyme enzyme activities in weanling rats. Toxicological Sciences 61,76-82. 2001.

Zi, X and Agarwal, R. Modulation of mitogen-activated protein kinase activation and cell cycle regulators by the potent skin cancer preventive agent silymarin. Biochemical and Biophysical Research Communications 263,528-536. 1999.

Zomer, AWM, van der Burg, B, Jansen, GA, Wanders, JA, Poll-The, BT, and van der Saag, PT. Pristanic acid and phytanic acid: naturally ocurring ligands for the nuclear receptor peroxisome proliferator-activated receptor a. Journal of Lipid Research 41, 1801-1807. 2000.

294 7. Appendices

295 Appendix 1. Table i) Batch sample details for omega 3 rich dietary supplement samples, Chapter 2 Original Description Prep date Purchase date Batch no. Brand Sample code MJl Extra high 4. 03.02 28.12.01 311833 Seven Seas strength pure code liver oil MJ2 Pure Cod liver 4. 03.02 28.12.01 301952 Seven Seas oil MJ3 Efanatal 4. 03.02 28.12.01 15322 Nutricia MJ4 EYEQ 4. 03.02 28.12.01 50555 Equazen MJ5 Omega 3 ‘700’ 4. 03.02 28.12.01 18294 Solgar MJ6 Cod liver oil 4. 03.02 31.12.01 02401 Cytoplan MJ7 EPA 300 4. 03.02 18.12.01 00201 Cytoplan marine lipid concentrate MJ8 Linseed oil 4. 03.02 18.12.01 050701 Cytoplan MJ9 Fish oil 4. 03.02 18.12.01 8281 Cytoplan supplement MJIO Organic flax 4. 03.02 18.12.01 8248 Cytoplan seed oil M Jll High potency 4. 03.02 28.12.01 48609 Lamberts fish oil MJ12 Super strength 14.03.02 8.03.02 BA55 Boots concentrated fish oils lOOOmg MJ13 Mobility cod 14.03.02 8.03.02 IJJ Boots liver oil 200ml MJ14 Mobility cod 14.03.02 8.03.02 14673M Boots liver oil one a day MJ15 Mobility high 14.03.02 8.03.02 14945M Boots strength cod liver oil lOOOmg one a day MJ16 Cardioace 14.03.02 8.03.02 93876 Vitabiotics MJ17 Efamarine 14.03.02 8.03.02 800951 Efamarine 500mg MJl 8 PULSE pure 14.03.02 8.03.02 312517 Seven Seas fish oils 500mg MJ19 Neutrataste 14.03.02 8.03.02 312105 Seven Seas taste-free cod liver oil 500mg MJ20 Linseed oil 14.03.02 7.03.02 L140HC2 Biona organic MJ21 Ultimate 14.03.02 7.03.02 020115A Udo’s choice omega 3 vegetable oil blend

296 M I o S Q 1 s 1 Pü Qh f il w “ '2 W V) s î î KS <4-1 0 a O ûSP t S i g oo1 (N I Q fil bû bû bû o I î î i I « § I 1) cd

> I (SI es >> If,?a -r, c es = s J I D t|: ° - S' > 'O O O I il I ile' ^ ? ^ 4>>. ç I I 1 I I § II -O 1 20,2-S-o 9 ë S N o >. r^D T'M3 &l a tJ M ^'46-ë I U 'g es o a'S ■B -f I Z

i|eu G iil es < H ol•C I I ÆS il es 'Q «n »n E ç . K -§ ë g es < I > . o 5^ I O a g I a es |§ a I I ^ a d a s I P •| ^ I e I iH o B 00 I I bû (d il »o w % i l I | l è il jo I: a |i f t § t l III(u oo îll A i Ii ed .52 cfl « If 0 il i 3 00 I i ilîen f T) Voi 1 11 ! f i Il 11 e u , t es _ T3 o es es w ■13 III II ISîî î î en ^ Ii S o î & 1 en I I o "o I x> 1ê- I % s? o r o % o en ed 1 I Z Æ % A oo I o '2 ■ £ bû 2 lU e > ■ë S 1 g ë X3 % T3 o Æ O a ' CA O I S i ilî U bû 1 k

< ufî o P g g

ed > . & cd « c O rC) .52 -Q a A ^ if ë-g ed jU en en 11 o C a t I g - la al. #A .a 1 1= O cd s 11 K VO II ^ g ed Ü ! l O w - i -s i >».1 111 S' ed .ts O -S JS 73 o lit “I .ts O S ^ 2 es «4-H E-i II lî •X3if Æ îi a A 111 I T3 A

en la hJ o 1 â l b 11 -s i s PQ (Il 1 IF as 1 E S lîl fai! o »ri Q »o I .13 fS I > »nI f 1 3 o I I ’êd oo W bû 3 < 1 O I oI p bû > bû bû # S bû B Q o ^ B 3 o I O oo 8 O (U Î 8 ^ 'o a 0Û cn ê i < oo o ü K s .. S ' o I ii o oo § Q i s h I § 73 o â3 |i| CJ o cnI 00If - es cn 'o f! >n < T3 ;Z •p î |5 JS JS < 2 2 ë On 11 II Ph

I 4 . g s ■l |i . s Ü 1 1 § c bû o I O bû cu u . 2 B 1 t o I t l CJ il (U C\ bû 2 £o R

a î | s C cd I I I ■g o M s Ï I 1 = a i s s l i 1 I - p II s a 1^1 II 3 1 ~ a ■= :§ I § a g (S 8 a a cs «3 (s ^ B O 1§I III ed 2 ll I lii l ■ i i 111 . 1 1 li 3 2 HI I t II IÎII B cd 0) ed a cS u bû II ë = I JS 3 bû eu 113 bû P a p P 1 3 O es o II œ II en % CJ P a-a alt œ a CJ

(S 111

fS fall V3 ? o & <§“ S' 0Û

6Û 0Û Ë mI i I w ^1 M I 1£ W !>5 ^ I0 S| m cd '—' 0û2 -O EP M 60 »n op 1 o I Ü o lë ii t/3 Q <4-4 0 o . ? O II o 6Û I o m CL (U cd (N o pq f i CJN o o 1m M O COM r—I I(2 u

■M ^ cd TO B - 13 'C CL i Mpq lCL l l 1 11 . B II 1OX) •— I cd O c g I cd 0)q s -# s i Se "S 111 III&O cn I «.g 6 g o O (U 3 o ^ 3 =5 « N 2-cd A W O Vh Vh eg «a (J o Sli o JO T3 CL I 0Û 0 s o 'o il cn •§ t-T q !i > O if I U o I s I Ü

T3 0) IÎ I-I I S? J=1 CA 60 g . 6 #2 o cd T3 S S •a CL O Oû^ o il J | II II S o cd If •S O If ^ < N II If (N S no --H i II CLT3

cd IS 2 - S l q a TO 1 pa i l i l l ; g i l 11II

*c O c

I o (N i

T3 î en 60 6060 vd ON ON S: I i 60 I cd s § 3 ^ 60 eu ^

O ü

o en H cd 60 i q cd fî U o 1

I t § en co

!îi

1 -§ II 3 = 1 lîl M I ° a S g: M T3 "S

Û 0 tr DC ■O TO g g c 0 0 c o TO TO g Ü 0 0 *=Ü 3 O) O) 3 1Ü TO 0 0 TO C TO TO C s TO . TO c o 0 0 c O o o £ s z JC c c O) O) 0 0 1 !c 1 in 0 s E E B 3 1 0 jf TO TO 0 1 1 O) C CO) o TO c TO ! c 3 3 ic (0 o o O c 1 1 o Q. 0 . O to o 2 E E 2 o O) o o _D) DC I Q.I Q. ! c Ü Ü ic z : 0 3 C/3 fi 0) ID CO CO CDIDIDCD CD CDIDIDCO 00 ID CM CO CM CO O) CM Ti Tl" CD Tf CD CO CO O Tj- CO id OID q CO ID id q 'M' 00 q CD CD 00 q CM od id cd od T|- cd N od od Tj- ed 0) CM N CM N id CM od Tf I o ■n- o 00 'd' 00 o 0) CM CM Tf O) D) O) O) CM I I m o5 0 0 od I % OJ CM CM CM CM CM T- O T— O s % od od od m CO CD M 00 § O) O) O)O) 00 (0) ■S§ ■nf q q q q q CM CO CM CM ID CM id id ID 00 ID od Tl- Tl- od Tf O)ID 00 O 00 c o O) q 11 -1— CM ID É ^ od N i d ÿ i d g i d od 1 s K 00 ID s Tl- q Tf q Tr CM O R ID s id CO CM cd cd CM cd M c cd N o d 00 CM . CD CO o' ^ O |5 CM CM Tt c d CM Tt c d c d Tl- ID ID ^ S r — CO -4" Tf s 0 ) 0 ) ^ q g K CM N M cd cd M I— M CO cd McdO)T-IDcdcOMCD N-' ID CD <3) 2 2 I 3 3 3 3 3 3 3 3 3 3 3 3 3 II o o o o o o o o o Oo o o o o o o o o o o O O O O O e CO CM ID 00 M 00CD O)O)CD NCOCDCO N CM CO) ■«— M t T Tf ID M M CO CM 00 CO O) ID O) O)O) O CD q CO CM ID q _ CM O CM cd CD M id od Tl- id CM od ID od ID T|- ID ID CO CM CO S C CD CM CM r 00 ID M CM o ■o c

” s

I XXXIXXXIXXIXXXXXXXIXIIXXXXXXX

s I Ü a CO CO < < < CO m Ü Q< < Ü UJ o < CM O CM < CM o I o o CM o s O CD CD i i § So CM O s g§ § CD 8 i i s aCD i q q q q q CM CD Mi q CM CD Mi Mi CD CM CD Mi CD CD CD CD CD CD CD CM CM O q O CD q o o CD cd od CM CD Tt Tj- CM od cd od Tt CM od CM od od od CM od cd cd od CM O o Tt T— o 00 T— CO 00 T— 00 ■*“ 00 CO 00 g o o O ri 1 y t o t o C C c C C 0) 0 3 O 0 0 0 2 2 2 2 2 c c ü) « W XX X 2 2 2 2 0 0 0 o o o o o c c c i % E o O O EEEE co w to to 1 1 0 1 E 0 0 0 0 CD 0 0 0 0 ■J ‘1 BBB B 0 BB B N N N i l w w I o 0 0 0 E E E E EE ■gTO O O X XX XX 3 3 3 3 2 X § 2 2 3 3 c c JDJD JD 0 0 r - M. M O OOO OO O O 0 *0 0 0 0 c c CT CT i i o LU LU LU UJ CD CD I Tt Tt Tt Tt Tt ü ü ü o ü ü Q O Q û û TJ Ç ]0 O

00 00 CO CO Q. Q. t^ t'. in M- 00 00 in in Oi in in CO in m O) Oi in t^ 00 in 00 in CM CO CM CO O)O) CO CO CO CJ> o CD CD Oi Oi CM o CM o Tt o q CD CD G) O) id O)CDCD Tt CD Tt O) ai O) cd d d CM od CM CM od id od O) id CM G) oi CD CD O) CM Mi CM l< CM id Tt 00 00 CO o Tt CO o o CM Tt Tt 00 00 Tt O)O) Tt q q q 0 0 od 00 (M CM CM

8 oCO CO 8 8 8 8 CD CO t^ od 8 CO 8 8 CD 8 8 O) O) CO 5 9 5 CM CO q CM 8 q q Mi CM 8 CM CM 5 q CM 8 id 8 id CM 8 id id id id od id 00 od in od id id T t od r ; a> T t T-" O) G> T-T T-“ G) Tt 8 Tt od t'i T t in 8 in Mi id q o q m in in 00 in 00 in in q8 q m 00 in T t 8 8 q CD< CM CM CM cd T t T t CM cd cd cd cd t'i cd Mi t'i T t T t cd cd cd t'i t 'i Ti CM cd o d m T h»' t-T CM CM CM t'i CMCM CMCMT t CO t'i CM t'i T— CM T~ CM T-T T—- T t t ' i e d id t» M- t«- in O) O) O) Tt T t in Tt T t T t in in in in t' q T t q m G) Gi Gi O) in in Mi' id id id cd T t T t T t cd Mi CM cd t'i t 'i cd id CM cd id T t cd cd cd t'i cd t'i CM id IE :§ 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 ë § 3 3 3 3 3 3 3 3 o CD CD o CDCD o o CD CD CDCD CD CD CD o CD CD o CDCD CD CD CO m CO in CO T t q CO O) Tt CM q t-. o 00 O) T t CO CM in 00 q N G> CO in G) CD 00 q Tt T t t-i CD q CD Tt q T t O) in T t O) CD O) t'- q CD t^ 00 1'- CD T t 00 Gi q T t id od CD id CM id CD id CD id cd CD G)G) 8 O) CM O) CD CD CM od T t od m in CM T t CMCM CM O) CM CM Tt CM in T t CD CMCM CO in Rl 8 T t

XXXXIXXIXXXXIXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

Q CO CO CO < < < < < < < O Ü < Q CO < CO Q < < < Ü o CO CO CO CM CM CM CM CM CM CM CM CM m CM CM CM CM < CM < CM q o o o CD O O OOO OCD CM o CD O CD CD CD CD CD CD CM CD CDCD § 8 CD CD q q q q CM CM CM o CD t'i t'i t'i t'i Mi CD t'i t'i CM t'i CD t'i CD CD t'i t'i t'i CDCD CD CD CM CM CM o o CD q CD q CD CD CD CM CD q O CD CO CO CO od CM CM CM T t CM od T t CM CD CM od CM od od CM Tt CM od od od od CM cd cd cd o CD CM 00 CD o 00 00

CO CO CO CO 0 o o Q Q r-OOOOOOOO —O 00 N 00 t~ 00 M 00 I 00 00 00 00 00 00 CO CO >tCMCMCMCMCMCMCMCM m in COTt(D00a)0CD0 mOOOOOOOOOOOOOOOO q T - T — T— T— T— T - T — T - L CO CO CO 00 N O Ü u Ü U gllllllll ^ O Ü O Ü Ü Ü Ü Ü § £ O O 0_ CL Q. o_ Q- CL o o O o O) CD) O) CD) 0 0 0 0 0 0 0 0 0 0 0 0 0 >> > , > XD 0 0 0 0 0 0 0 0 I > > > > Q. o o o o Q. Q. Q. CD. 3 3 3 3 •t 0 0 0 0 TD ■o TD TD 0 0 0 0 I T3 E E EE -a 0 TJ 0 0 o 0 OT 0 0 0 0 0 3 § 5 5 5 ^ E < % .-tt •t; z ^ I Q 1 "CD X3 TD TD Û Û 0 0 0 0 oc DC OC S’ TD ■O "O TD ■D 0 0 0 0 g O) O) O) O) g g x: 0 0 0 0 -a •O TD TD JOI 1 c « 3 0 0 0 0 I TD > > > > $ Ü C 0 0 0 0 I a X: .C XI 3 "5 OT Ds > , > \ > , 0 o C 0 0 0 0 .5 o Î-I 2 EEE E - I c ■o T3 TD TD 0 c CC C 3 ÎI O3 O3 o3 o 1 0 Q. CD. CD. CD. fî EE EE o o o o 1 1 II II o CD CD CD !o 0 0 0 0 o CD) 0 Ic ic Q. Z E Q. X: Ii O) CD (O CM T- q N O) q CD ]- O) CO ID t' O) in S 5 CD IX ) CD CD CO IX) CD U) CD CO ID T-; q q CO CD 00 Tt O) IIID 00 I— Tt q q T-; id O) m in q q q Tt id q q q ^ ID id cd t'i Tt CM 00 CD) CM CD N Tt ID id od O) CM CD Mi Mi od cd t'i id T- CO CD) M ID ■t— Tt GO d O) O) 00 00 CM 1— O T- 00 00 00 O Tt 00 f - ■I— T t "M" 0 0 o5 O) od 8 CM

CO 8 CO CO CO CO s CD q CD CM CM CM CD q 8 od CO 00 CO O) CD) CO 00 00 m CO CD 00 Tt Tt T t q CM Tt CO • 8 8 8 8 id Tt CM 8 t 'i CM CM CM CM a CM CM 8 P: q P: id od Tt id in od od id ID - 00 00 o od CD od CD od od cd T t CD) T t od 5 T- q CM CM CM id 8 8 CD 8 00 in t ' i s ' t ' i s' 8 q in 8 q CM CM 8 8 in CM cd in CD cd in CM CD cd Tt Mi^ cd ■d Tt ^ cd cd CD CD CM CD cd CD CM CD od cd CO cd CO cd Tt Mi Mi Tt cd (R 5 M M>' CM CM CM m t ' i id CM •«— Tt M ID Tt O) in cd cd m cd 8 O)

Ü CO < “ < CO CO Ü < < < < CO CO CO < CO < < O CO < CM CM CM < CD CM CM CM CM CO CM CM CM CM o CM CM CM q CM q O O O CD o CD CD CD OCD o o O OCD o CD CD CD CD 8 CD CD O 8 8 CM o o q q t'i CD CD CD CD q o CD t ' i CD q t ' i CM t ' i CD t ' i t ' i CDCD CMCD M i CD CDCM M i CDCD CD CM G y— 7— CM CD O O CM q o o q 8 CM 00 CO 00 od CM od CM CD CM od T t CM od od cd od CM T t od CD T t od od o od CD od CM CD CD T- CD CM CD CD CM O O O O C _C Ç I0 f 0 ' > > c c C c c c c c S 5 CO 0 0 0 o o o o o o o o c c c O O 2 2 0 0 0 0 0 0 0 0 0 .t; 0 II c c c c o c o c Ü Ü Ü.g j 'o <"> <"N c C -TTC 0 0 0 0 0 0 0 0 0 0 0 i o XD iCD 3 3 § 0 0 O O O 0 0 0 0 Ü _ _ C C c c c c c c c c Ü CD CD 3 3 3 0 0 0 0 0 0 1 0 0 5 2 2 2 2 2 2 2 CD 0 0 0 Ï1 O'-•o 0 I ■OS’ 0 X3 3 c < 3 z R Q E cr 8 s 8 8

T3 "O c c T3 T3 2 2 c D) O) 0 ic ic 1 O) -2D) 1 I in CD O) CO CO 00 q q -r-; 00 00 Tt CO in O) od cd in Tt Tf in Tt od CD O Tt I— CM CM i ^ Tt Tt 00

8 in o o CD CO od od CO t'. O i 8 CM CN K t'i t^ Tt od 00 in in M. in o Tt T-; CM M- 0 0 00 in O ) -I— T— t 'i in q in in O t'i Tt cd "t 00 od CO CM ( i . CO ^ Tt in 9 CO in CD CD P: q CO in T— T— m T in 2 3 3 3 3 CD CD CD CD CD CD CD CD CD 1— 1— CM CO CM in CD CO in 1- CO O Tt Tt -1- CD 00 Tt t'- Tt 00 T— Tt CD CM ^ tsi CO in CD 1- Tt

t'.r'>i'>t'-M~t'-t^t'«t' IIIXIIIIX 33=>=)=)Z)3=)=) XXXXXXXXX

< Ü < CO CM CM CM CD CÛ CM CD CD CD CM CM O CD CD t'i CD t'i q q CM CM q t'i CDCD CM O CM Tt CM CD cd CM o 00 00 CM CD .CXIXIXI BB B Ü -g .C 0 0 0 0 2 XD XDXDXD 0 0 XD XD TD TD 1 1 1 1 CO CO CO - £ £ . 2 . 2 a Q O 'c ^ C C ' c c c 0 0 0 0 0 0 0 0 0 0 0 0 OOOOOO Ü ÜÜÜ CD Ü S3 S3 S3 i : i: > > > CM N un M I'. 05 M q co CO lO CO co co t'« M. 05 q q CO ço in co co 00 m q CM lO CO q CO O q CM q Tt O) t" Tt cd q CO O co ■M- CM q Tt 05 Tt cd CD <35 cô O) c> •d 05 un Tt cô co (d in Tt 05 in 05 CÔ cd cd o 05 t co CO 05 CM <35 Tt 05 05 I'. 05 CM o Tt 05 8 i g Tt U5 un 00 Tt CO q Tt Tt Tt Tt Tt co s CD CO O od q od cd 2 S cd t'i 8 1'. t'i 8 - cd cd CD un cd CM to CM ri 05 l'T 8 CM l'i CD CD S CD o tn cd un od cd in cd cd S 1 1 un m I oc o d CM l'T in q t'i t'i tn co co o od o q o d CD CM g cd cd od un od cd U) tn •l I û a 1 1 1 1 1 1 1 1 1 1 S 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 D 3 3 3 3 3 3 3 3 D 3 3 3 3 3 3 3 3 3 3 3 3 3 3 g CD CDCD CDO CD CD o CD CDCDO O O CD CD CDO CD CDCD CD CD CD CD CD O CO t'. CO 00 05 05 Tt co 05 Tt Tt q CO tn Tt CO CO 00 Tt 05 Tt CD CD CD CD tn 00 Tt I co q q «N CM CD tn CDCD Tt CO 05 q CO 00 co Tt CM CM 0 cô Tt CM CM CD cd ■M" CÔCD 00 od 05 05 cô CM t'i LO CM cô JS CD CO cô CM tn un 00 t'. un 00 t'. CM CM co il 73 C

1 î ü I ÜÜÜÜÜOOOÜOÜÜÜÜOOÜÜOÜÜÜOOÜOÜOÜOO

- ë CD C: m 00 CM co 0 0 CO 00 o CD 0 00 CM CM CD CD £2 CD CD 0 i i 00 CD 0 q 0 q q CD 0 q q 8 1 g q 8 <3 CD q g CDCDq q CD q q 0 q CMCM Mi t ' i CD CM CM CD CM CMCM I 0 CD Tt" Tt tn CMod U5 CM cd CM CD cô Tt Tt U5 od cd CM CD cô Tt Tt tn CM00 CD CD <35 CM CM od CD CD C35 CM CM CM 00

Æ 0 H 73 C ■5 ■5 ■5 ■5 ■5 ■5 CD CD CD r4 3 73 73 7) 73 73 73 c C CCC 0 (0 COCO COco CO m m 0 0 0 a iz C/5 U) tn 0) CO co co W V) W O) tn CO g I 73 O g O ü p I 1 <15 CD CD CDCD CD £ £ £ c C CC C C ü 0 0 0 0 0 0 E E C tn (/) V) tn

Tt S N N r» co o cd CD CD CD co t'i CÔ CÔ cô cd id CM CD CD CD t'i o id t" M t ' LO cd eno od LO LÔ LO LO t 'i co CM CM CM od o cd Tt Tt Tt LO od

2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 2 2 2 2 3 3 3 3 3 OO O OO OOOO OOOO OOO OOO OOOO 3 3 3 5 O O OO O OO co LO LOO 00 CO lO CO CO Tt CO CM CO Tt 1'- CO CD CD 00 t'. CD CM CD CO LO OO q !'« CD CM o CD CM q q t'. q q q o q LO co q CO t'. Tt t'; CO CD !'- CM 00 O CM Tt t'i cd CD d cd id CM CÔ Tt CM t'. Tt Tt cd co cd cd Tt t< cô od Tt d od t'i CD CD id CD t'i co 00 co t'. t'. O LO t' CO LO CO co CM CM CO co CD CO co CO Tt 00 CD CM CM CO CM CM CM

>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> ÜÜOÜÜOOOOÜOÜÜÜOOOÜOOÜÜÜÜÜÜÜÜÜÜÜÜÜ

O t'. 03 t'- 03 co m GO CO CO d § d d B O d d O d § d O o i i d OO d i O GO CM CM CM CM CM O O o o q q o q q o O o O q q q o O o q q o o q q o O o q q q o q q CM CM CM CM CM CM CM t< N MM t< CM CM CM CM CM CM CM o O o o o Tt Tt Tt Tt Tt Tt Tt Tt d cô Tt Tt d cô Tt Tt id cd od od od od id CM d cô CM CM CM CM CM CM CM CM CM CM 00 o CM CM

CM (TM CM (N (N

_ 8 s s s ^ S ° t'. CD - q , q. o ^ q CÔ q q ^ CM CM O) CM o CD Tt Tt oo od M co cd °° if> id i d o id ld id en CM od od I i i Tt cd cd

2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 OOO O OOO OO O 3 3 O o O O o O OOOO O O o o O OOO OOO CM Tt O 00 O CO CO CM en LO CM Tt CM en CD CD O o O 00 e n O O O co en OO CD O Tt CM t'. o Tt CM 00 CD CD CO CD CD en Tt CD Tt q cd co CD !'■ CM q cô q co cd cd Tt o cd od cd t'i CD cd CÔ cd cd CD cd cd Tt Tt CM m CD Tt N CM 00 CD en CO Tt 8 8 8 CO CO CM

> > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > ü O o o O ü ü ü o ü ü ü ü O ü ü ü ü ü ü ü ü ü O ü ü ü O O ü ü O ü 00 o 00 o

ü ü ü m m co ü co ü O Q Q O o OO CM o o m m o o CO o o CO CM CM CM O q q q o q q o q q q o q g q q o q g q q O o O CM CM N cd cd M t< N O q o q Tt in cd cô Tt cd cd od cd CM id cd CM cd id cd CM cd id 00 » 00 CM CM o O 00 00 o CD oo O CD oo

co CO Q Q CD CD 1 1 1 'S i i .9 .y .y = _c _ c _ ç _ _ _ 0 N N 0_ _0 _ 0 0 o o Tn 9 •■î= = Ica 1(0 C c c .C j c x : s0 s 0 s 0 c > ÏT >ÏT B '2- ’e- CD O ) .hz .h : . t 0 0 0 Z z Z Z Z 0 0 0 > > > co CO CO CO CO CO u ü ü ü O •§ i i. i, i. i, OOO s ë CL CL Q. CL CL 0_ CL CL CL CL CL 2 2 2 X X « W w w .g o

m uo O) co 0 5 05 CD M CM N CD 1'. CD M- CD CDCD CO co lO co co 00 00 o> Tt Tt LOCO O q Tt CO OO Tt CMq Tt N Tt co N Tt q q co t'- q q q co CM CÔCÔ Tt CDCÔ CD lÔ Tt CO CD LÔ CÔ co CÔ co oô in CM Tt CM B co CD B CM CD CM CM Tt CDCD t'. CM B

LO Tt Tt o oô q oô 8 co oô 8 M. 8 q q co CM K q co CM CM q co CM LO cô Tt 1'.“ cô o cô Tt Tt CD K cô o Tt K cô cd 0\ c ô LO cô o LO cq co co M LO co co en LÔ LÔ LÔ lÔ LÔ t'i co LÔLÔLÔ LÔ K LÔLÔ in o 00 00 00 o 00 cô o o co co CM co cô cô cô LÔ oô CD l Ô LÔ •M- LO oô LÔ LÔ LÔ oô

s s 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 23 lE 2 23 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 o o o o o o o o o o o o o o o o o q o o o o o o o o o o CM CMCM 2co 00 co LON. CMCM CD co co CD t' CMM CD LO Tt t'. co LO q q £2 o £2 o t'. o co co co q LO q co q co 05 q q 00 q co Tt 00 s oô s cô 05 oô co t'i co CD co t'i LÔ t'i co CMCD oô t'i 05 LÔ co cô LÔ q ■M" 00 LOCM co co t~ CM Tt co CD t' t' LO

OÜOÜÜOOÜOOOÜOOOÜÜOÜOÜÜÜOOÜÜÜOÜ

m co ü Q Q co o CM CM CM CM s o o q q q o q q o o q o q q s s q q q q s g g g q o o t'i t ' t'i t'i CM CM cd 1 CM o q o o CD oô LÔ CM LÔ 00 CÔ 00 Tt Tt LÔ CM oô LÔ LÔ LÔ CM CD Tt o É É s 05 §

o o o o o o o 0 0 0 0-=-=7=7=-=7=7=sz sz sz sz sz sz o o ü C5 ü u ü 0 0 0 0 0 0 0 COCO co co co co co 0 0 0 0 c c c c c c c c c c û Q Q Q QQQ ^>£^.2 CD0 0 0 o o o o o o o 0 0 0 c c c c c c c co •= = s? c c c ç c c ç 0 0 0 0 0 0 0 o o o EEE E E EE c c ç c C/5CD c/50 c/5 c/5 0 0 0 0 0 0 0 0 0 0 0 0 c c c c c c ° ° ÿ.y 1 ü 1.ÿ o H H ^ ^ 2 2 2 2 2 2 s H H Ht- t- H > 5 5 i> Appendix 3. Chapter 5. Table 1: Descriptions of the principle VolSurf descriptors used for the NR QSAR studies

VoISurf Symbol Description descriptor

Molecular Volume V Represents the water solvent excluded volume (in Â^), or the volume contained within the water accessible surface computed at 0.20 kcal/mol, as determined with the water probe interaction with a target molecule.

Molecular Surface S Represents the accessible surface (in Â^) traced out by a water probe interacting at 0.20 kcal/mol when it rolls over the target molecule.

Ratio R A measure of molecular wrinkled surface (rugosity). The smaller the ratio, the larger is the rugosity. Volume/Surface

Molecular G Represented by S/Scquiv., in which Sequiv. is the surface area of a sphere of volume V. Globularity is Globularity 1.0 for perfect spherical molecules, and assumes values greater than 1.0 for real spheroidal molecules. It is also related to molecular flexibility.

Hydrophilic Wl- Represent the molecular envelope accessible by solvent water molecules. The volume of this regions at eight W8 envelope varies with the level of interaction energies between water and the solute molecule. In energy levels general, W 1-4 account for polarisability and dispersion forces, while W 5-8 account for polar and hydrogen bond donor-acceptor regions.

Amide responsive WAMl- Represent the molecular envelope accessible by the amide probe. The volume of this envelope regions at eight WAM8 varies with the level of interaction energies between the amide (NH) probe and the molecule. Due to energy levels the lone pair of electrons on the nitrogen, the amide probe provides specific information about good acceptor abilities. In general, WAM 1-4 account for dispersion forces, while WAM 5-8 account for hydrogen bond acceptor regions. These descriptors are closely related to W1-W8.

Integy Moments Iwl- INTEraction enerGY (integy) moments express the imbalance between the centre of mass of a Iw8 molecule and the position of the hydrophilic regions around them. Integy moments calculated at eight different energy levels (-0.2 -0.5 -1.0 -2.0 -3.0 -4.0 -5.0 -6.0 kcal/mol) and are vectors pointing from the centre of mass to the centre of W1-W8 respectively. If the integy moment is high, there is a clear concentration of hydrated regions in only one part of the molecular surface. If the integy moment is small, the polar moieties are either close to the centre of mass or they are at opposite ends of the molecule and the resulting barycentre is close to the centre of the molecule.

Capacity factors Cwl- Represent the ratio between the hydrophilic regions and the molecular surface, that is the amount of Cw8 hydrophilic regions per surface unit. Tliey are calculated at eight different energy levels (-0.2 -0.5 - 1.0 -2.0 -3.0 -4.0 -5.0 -6.0 kcal/mol).

Hydrophobic D1-D8 Hydrophobic regions indicate interactions with the hydrophobic probe at 8 different energy levels, regions at eight which have been adapted to the energy range of the DRY probe (-0.2 -0.4 -0.6 -0.8 -1.0 -1.2 -1.4 - energy levels 1.6). They are defined when a DRY probe is interacting with a target molecule.

Hydrophobic IDl- Similar to the Integy moments, but calculated from DRY probe 3D interaction maps. They measure Integy moments ID8 the imbalance between the centre of mass of a molecule and the position of the hydrophobic regions around them.

Local Interaction Eminl Represent the energy of interaction in kcal/mol of the best three local minima of interaction energies Energy minima Emin3 between the water probe and the target molecule.

Local interaction dl2 dl3 Represent the distances between the best three local minima of interaction energies when a water Energy minima d23 probe interacts with a target molecule. distances

Hydrophilic- HLl- The ratio between the hydrophilic regions measured at -3 and -4 kcal/mol and the hydrophobic Lipophilic balance HL2 regions measured at -0.6 and -0.8 kcal/mol. The balance describes which effect dominates in the molecule, or if they are roughly equally balanced.

Amphiphilic A This is a vector pointing from the centre of the hydrophobic domain to the centre of the hydrophilic Moment domain. The vector length is proportional to the strength of the amphiphilic moment, and it may determine the ability of a compound to penetrate a membrane.

310 Critical Packing CP Describes a ratio between the hydrophobic and lipophilic part of a molecules. Conversely to HLb, parameter CP refers just to the molecular shape: CP = Volume(hydrophobic part)/[(Surface(hydrophilic part)*(length of the hydrophobic part)]. The hydrophobic calculation is made at -0.6 kcal, the hydrophilic calculation at -3.0 kcal. CP is a good parameter to predict molecular packing, as with micelle formation.

Polarizability POL An estimation of the average molecular polarisability, calculated following the additive method reported by K.J. Miller (Miller 1990), which uses the structure of the conpounds (extracted from the GRID files) outside any molecular field to compute the values such that it is independent of the number and type of probe used.

Hydrogen Bonding HBl-8 Represent details about the H bonding capabilities of the targets, which can be different depending upon the nature of the polar probe used. The water probe presents the optimal ability to establish H bonds with the target, either donating or accepting one or more Hs. If a different polar probe is used, the interaction may be less favorable, as this probe is unable to accept or donate Hs. This variable is obtained as the differences between the hydrophilic negative volumes (Wl-8) between the water and any other polar probe included for each negative energy level considered.

311