An Investigation into Changes to Trace Metals and Metabolic Profiling in the Diabetic Retina

A thesis submitted to the University of Manchester for the degree of Doctor of Philosophy in the Faculty of Biology, Medicine and Health

2017

Sandra Callagy

School of Biological Sciences

Division of Evolution and Genomic Sciences

1

Contents

List of Figures ...... 8

List of Tables ...... 10

List of Abbreviations ...... 11

Abstract ...... 14

Declaration ...... 15

Disclaimer ...... 15

Copyright statement ...... 16

Dedication ...... 17

Acknowledgements ...... 18

About the Author ...... 19

Chapter 1 | Introduction ...... 20

1.1 Introduction ...... 21

1.2 Overview of Glucose and Insulin Modulation ...... 21

1.3 Introduction to Diabetes Mellitus ...... 23

1.3.1 Diagnosis ...... 25

1.3.2 Cost of Diabetes ...... 26

1.3.3 Global Prevalence ...... 26

1.3.4 Treatments for Diabetes ...... 27

1.3.5 Diabetic Complications ...... 27

1.4 Diabetic Retinopathy: Introduction and Pathogenesis ...... 31

1.4.1 Introduction ...... 31

1.4.2 Introduction to Diabetic Retinopathy ...... 33

1.4.3 Pathogenesis of Diabetic Retinopathy ...... 33

1.4.4 Changes to the Retinal Vasculature during Diabetic Retinopathy ...... 36

1.5 Diabetic Retinopathy: Metabolic and Inflammatory Changes ...... 37

1.5.1 The Polyol Pathway ...... 38

1.5.2 Hexosamine pathway ...... 38

2

1.5.3 AGEs ...... 41

1.5.4 Lipid mediators ...... 43

1.5.5 Inflammatory Molecular Mediators of Diabetic Retinopathy ...... 44

1.5.6 Oxidative Stress ...... 45

1.5.7 Transition Metals ...... 46

1.6 Current treatments for Diabetic Retinopathy ...... 47

1.6.1 Preventative Therapy ...... 47

1.6.2 Treatments for Diabetic Macular Oedema ...... 47

1.6.3 Treatment for Proliferative Diabetic Retinopathy ...... 48

1.6.4 Prospective Therapeutics ...... 49

1.7 Rodent models of Diabetic Retinopathy ...... 50

1.7.1 Chemically-Induced Diabetic Models ...... 50

1.7.2 Spontaneous Models of Diabetes ...... 51

1.7.3 Proliferative Diabetic Retinopathy Models ...... 51

1.8 Summary and Aims ...... 53

Chapter 2 | Overview and validation of the in vivo Experiments used in this Study .... 54

2.1 Introduction ...... 55

2.2 Methods ...... 56

2.2.1 In vivo Methods ...... 56

2.2.2 Tissue Extraction and Analysis ...... 58

2.2.3 Statistical Analysis ...... 60

2.3 Results ...... 61

2.3.1 Characteristics of in vivo Studies ...... 61

2.3.2 Changes in the mRNA expression of Molecular Biomarkers in 12-Week Diabetic Retinas ...... 61

2.3.3 Changes in Retinal Biomarker mRNA Expression after 16 Weeks of Diabetes ...... 61

2.4 Discussion...... 65

2.4.1 Altered mRNA expression of Established Biomarkers Confirmed the Molecular Pathology after 12 and 16 weeks of STZ-induced Diabetes ...... 65

2.4.2 Gat3 and Nppa were changed in rat retinas after 12 weeks but not 16 weeks of STZ- induced diabetes ...... 66

3

2.4.3 Additional biomarkers Txnip and Vegfa were also changed in rat retinas after 16 weeks of STZ-induced diabetes ...... 67

2.4.4 Overall Conclusions ...... 67

Chapter 3 | Characterising Trace Metals in Human and Rat Retinas and the Effects of Diabetes ...... 69

3.1 Introduction ...... 70

3.2 Methods ...... 74

3.2.1 Donor Retinas...... 74

3.2.2 Dissection of Rat Retinas ...... 74

3.2.3 Tissue Digestion ...... 74

3.2.4 ICP-MS ...... 75

3.2.5 Statistical Analysis ...... 75

3.3 Results ...... 77

3.3.1 Donor Retinas...... 77

3.3.2 Effect of Diabetes on Retinal Trace Metal Content ...... 77

3.3.3 The Influence of Sex on Retinal Trace Metal Concentrations ...... 80

3.3.4 Effects of Age on Retinal Trace Metal Content ...... 82

3.3.5 The Effect of Smoking on Retinal Metals ...... 83

3.3.6 The Effect of Prescription Medication on Retinal Trace Metals ...... 84

3.3.7 Changes in Retinal Trace Metal Content in Rats with STZ-induced Diabetes ...... 86

3.3.8 Analysis of Copper Homeostatic Protein-encoding Genes by RT-qPCR analysis ... 90

3.3.9 The Effect of TETA-treatment on Expression of Established Biomarkers in STZ- Induced Diabetic Retinopathy ...... 91

3.4 Discussion...... 96

3.4.1 Quantification of Retinal Trace Metals ...... 96

3.4.2 The Effect of Diabetes on Retinal Trace Metal Concentrations ...... 98

3.4.3 Diabetes Evokes Molecular Changes in Expression of Genes Corresponding to Copper Homeostatic Mechanisms ...... 99

3.4.4 The Effect of TETA-treatment on Diabetes-Related Biomarkers ...... 104

3.4.5 Secondary Outcomes of Trace Metal Analysis in Human Retina ...... 105

3.5.6 Limitations ...... 107

3.5.7 Conclusions ...... 108

4

Chapter 4 | Polar Metabolite Profiling of the Rat Retina during Hyperglycaemia ...... 109

4.1 Introduction ...... 110

4.2 Methods ...... 113

4.2.1 Tissue Extraction ...... 113

4.2.2 Derivatisation ...... 113

4.2.3 GC-MS Analysis ...... 114

4.2.4 Data analysis ...... 114

4.2.5 Statistical analysis ...... 114

4.3 Results ...... 116

4.3.1 Gas Chromatography ...... 116

4.3.2 Principle Component Analysis (PCA) ...... 116

4.3.3 Overall Changes to Metabolites ...... 120

4.3.4 Sugars and polyols ...... 120

4.4.5 Amino Acids...... 120

4.3.6 Fatty acids and glycerol ...... 121

4.3.7 Others ...... 121

4.4 Discussion...... 124

4.4.1 Established Metabolite Changes Occurred in all Experiments ...... 124

4.4.2 Diabetes Increases Retinal Glucose, Glycolysis and Tricarboxylic Acid (TCA) Cycle Intermediates ...... 126

4.4.3 Polyols are Consistently Increased in the Diabetic Retina except for Scyllo-Inositol ...... 126

4.4.4 Amino Acid Changes ...... 128

4.4.5 Diabetes Increased Relative Retinal Concentrations of Lipid Mediators ...... 130

4.4.6 Limitations of study ...... 133

4.4.7 Conclusion ...... 134

Chapter 5 | Mapping Lipid Changes in the Rat Retina during Hyperglycaemia ...... 135

5.1 Introduction ...... 136

5.2 Methods ...... 142

5.2.1 Animal Studies...... 142

5.2.2 LC-MS/MS ...... 142

5

5.2.3 Peak picking, alignment and annotation ...... 143

5.2.4 Metabolite identification and data reduction for comparative statistical analysis .... 143

5.4.5 Statistical analysis ...... 143

5.3 Results ...... 146

5.3.1 PCA analysis ...... 146

5.3.2 Verification of Identifications ...... 146

5.3.3 Changes to Lipids in Diabetic Retinas ...... 150

5.3.4 Phosphatidylcholines and Phosphatidylethanolamines ...... 150

5.3.5 Phosphatidylglycerols ...... 153

5.3.6 Phosphatidylinositols ...... 153

5.3.7 Cholesterol esters ...... 153

5.3.8 Fatty acids ...... 153

5.3.9 Sphingomyelins and Ceramides...... 156

5.3.10 Mono-, Di- and Triacylglycerols ...... 156

5.4 Discussion...... 159

5.4.1 The Extent of Lipid Identification and its Limitations ...... 159

5.4.2 Diabetes Induces Substantial Changes in Glycerophospholipids ...... 160

5.4.3 Diabetes Leads to an Overall Increase in PCs in the Rat Retina ...... 160

5.4.4 Ether-linked PEs and Ester-linked PEs were Changed in an Inverse Manner during Diabetic Retinopathy ...... 161

5.4.5 Diabetes Induced an Overall Downregulation of Phosphatidylglycerols ...... 161

5.4.6 Phosphatidylinositols were Substantially Perturbed in Diabetic Retinopathy ...... 162

5.4.7 Diabetes Alters Retinal Ceramide Composition ...... 163

5.4.8 Triacylglycerols but not Di- or Monoacylglycerols are affected by Diabetes ...... 164

5.5.9 Diabetes Causes a Distinct Change in Retinal Omega-6 and Omega-3 Fatty Acid Content ...... 165

5.5.10 Conclusions and Future Studies ...... 166

5.5.11 Overall Conclusion ...... 167

Chapter 6 | Overall Conclusions and Future Studies ...... 169

6.1 Overall study results ...... 170

6.2 Potential interrelationships between copper metabolism, polar metabolites and non-polar metabolites...... 170

6

6.2.1 Copper overload in diabetic retinas made affect energy production ...... 170

6.2.2. Copper Overload in Diabetic Retinas May Affect Lipid Metabolism ...... 171

6.2.3 Dysregulated Polar Metabolites may Affect the Lipid Profile ...... 171

6.3 Limitations and future work ...... 172

6.3.1. Limitations of animal model ...... 172

6.3.2 Limitations and potential future work with respect to copper and copper metabolism...... 173

6.3.3 Limitations and future work with respect to polar and non-polar metabolites ...... 174

6.6 Overall Conclusions ...... 175

References ...... 176

Appendices ...... 210

Appendix I: List of Primer used in rt-qPCR analysis ...... 211

Appendix II: Medical records of Human donors ...... 213

Appendix III: List of all Identified Polar Metabolites ...... 222

Appendix IV: List of all Identified Lipids ...... 225

Word count: 47,418

7

List of Figures

Figure 1.1 | Overview of glycolysis and the TCA cycle ...... 22 Figure 1.2 | Overview of metabolic pathways modulated by insulin signalling ...... 24 Figure 1.3 | An overview of the effect of diabetes across the body ...... 29 Figure 1.4 | Layout of the retina ...... 32 Figure 1.5 | Visual impairment from diabetic retinopathy ...... 33 Figure 1.6 | Fundus images of the progressive stages of retinopathy ...... 35 Figure 1.7 | The Polyol Pathway and AGE production...... 39 Figure 1.8 | The Hexosamine Pathway ...... 40 Figure 1.9 | The process of AGE production ...... 42 Figure 2.1 | Transcriptomic changes induced after 12 weeks of STZ-induced diabetes (Study 1) ...... 63 Figure 2.2 | Changes to biomarkers after 16 weeks of STZ-induced diabetes as assessed by RT-qPCR (Study 2) ...... 64 Figure 3.1 | Diagram of the Agilent 7700 ICP-MS ...... 71 Figure 3.2 | Scatterplot of copper measurements in non-diabetic and diabetic donor retinas...... 80 Figure 3.3 | The Relationship between Age and Retinal Cadmium Concentration ...... 82 Figure 3.4 | The Effect of Smoking on Retinal Cadmium Concentration ...... 83 Figure 3.5 | The Effect of Prescription Drugs on Retinal Trace Metals...... 85 Figure 3.6 | Copper Changes in Diabetic Rat Retina and Kidney Cortex ...... 89 Figure 3.7 | Changes in the Expression of Genes that Encode Proteins Involved in Copper Homeostasis in Rat Retinas after 12 weeks of Diabetes ...... 92 Figure 3.8 | Changes in the Expression of Genes that Encode Proteins Involved in Copper Homeostasis in Rat Retinas after 16 weeks of Diabetes ...... 93 Figure 3.9 | The Effects of TETA-treatment compared with untreated diabetes on the Gene Expression of Copper Transporters after 16 weeks of STZ-induced diabetes ...... 94 Figure 3.10 | The Effect of TETA-treatment on biomarkers of STZ-induced Diabetic Retinopathy ...... 95 Figure 4.1 | Diagram of GC-MS components ...... 111 Figure 4.2 | Example of a Gas Chromatogram from Experiment A ...... 117 Figure 4.3 | An example of the use of PCA analysis to determine group separation ...... 118 Figure 4.4 | Final PCA plots for all GC-MS Experiments ...... 119 Figure 4.5 | Changes to Components of the Glycolysis and Polyol Pathways ...... 123 Figure 4.6 | Overview of metabolic changes that occur in the STZ diabetic rat retina compared with non-diabetic ...... 132 Figure 5.1 | Main Lipid Class Structures Discussed in this Chapter...... 138 Figure 5.2 | A Schematic Illustration of an Orbitrap Velos Mass Analyser ...... 140 Figure 5.3 | Example of liquid chromatogram output (QC sample) of rat retinal lipids ...... 147 Figure 5.4 | PCA plot for LC-MS experiment ...... 148 Figure 5.5 | Feature Map of Lipid Identifications...... 149 Figure 5.6 | The Effect of Diabetes on Phosphatidylcholines...... 151

8

Figure 5.7 Changes to phosphatidylethanolamines in diabetic rat retinas ...... 152 Figure 5.8 | Changes in phosphatidylglycerols in diabetic rat retinas ...... 154 Figure 5.9 | Mapping phosphatidylinositols changes in diabetic rat retinas ...... 155 Figure 5.10 | Changes to ceramides and sugar-bound ceramides in diabetic retinas ...... 157 Figure 5.11 | Triacylglycerol changes in the diabetic rat retina ...... 158

9

List of Tables

Table 1.1 | The recommended measures of glycaemic control, blood pressure and plasma lipids in T1DM and T2DM ...... 26 Table 1.2 | Current therapies for hyperglycaemia ...... 28 Table 2.1 | Criteria for primer design ...... 60 Table 2.2 | Summarisation of starting weights, final weights and final blood-glucose values for all STZ experiments ...... 62 Table 3.1 | A Summary of Donor information for ICP-MS analysis ...... 78 Table 3.2 | The Effect of Diabetes on Retinal Metal Content ...... 79 Table 3.3 | Comparison of Retinal Metals in Human Males versus Females ...... 81 Table 3.4 | Quantification of Trace Metals in Non-diabetic and Diabetic in Rat Retinas ...... 87 Table 3.5 | The Effect of TETA Treatment on Trace Metals in Diabetic Retinas ...... 88 Table 3.6 | Comparison of Previously Published Data on Retinal Metal Quantification ...... 97 Table 4.1 | Description of Animal Experiments used for GC-MS analysis ...... 113 Table 4.2 | Polar Metabolite Changes in Diabetic Rat Retinas Relative to Non-Diabetic ...... 122 Table 4.3 | Summary of the reproducibility of metabolite changes ...... 125 Table 5.1 | Annotated script in R for positive mode data reduction, using XCMS & CAMERA...... 144 Table 5.2 | Characteristic fragments used to infer molecular weights...... 145

10

List of Abbreviations

3-DG 3-deoxyglucosone ActB Actin B AGE Advanced glycation end products Akt AKT8 virus oncogene cellular homolog ALE Advanced lipoxidation end-products ATOX1 Antioxidant 1 copper chaperone ATP Adenosine triphosphate ATP7A Atpase Copper Transporting Alpha ATP7B Atpase Copper Transporting Beta BCAA Branched chain amino acid BCAA Branched-chain amino acid C1inh C1 esterase inhibitor Ca Calcium Cco Cytochrome c oxidase CCS Copper Chaperone for Superoxide Dismutase Cd Cadmium CE Cholesterol ester Chi3l1 Chitinase 3 like 1 CI Confidence interval CML Nε-(carboxymethyl) lysine CoA Coenzyme A Commd1 Copper metabolism domain containing 1 COX-17 Cytochrome c oxidase chaperone Cp Caeruloplasmin Ctr1 Copper transporter 1 Ctr2 Copper transporter 2 Cu Copper DG Diacylglycerol DHA Docosahexaenoic acid DM Diabetes mellitus DMO Diabetic macular oedema Dmt1 Divalent metal transporter 1 DPP4 Dipeptidyl peptidase-4 DR Diabetic retinopathy ETR Eye tissue repository FA Fatty acid FAD Flavin adenine dinucleotide FADH Reduced flavin adenine dinucleotide

11

Fe Iron GABA GABA transporter 3 GABA Gamma-aminobutyric acid GAPDH Glyceraldehyde 3-phosphate dehydrogenase Gat3 GABA transporter type 3 Gbp2 Guanylate-binding protein family 2 GC-MS Gas chromatography mass spectrometry GDM Gestational diabetes mellitus GLP-1 Glucagon-like peptide-1 GLUT GSH Glutathione GSH-Px Glutathione peroxidise

HbA1C Glycated haemoglobin HCD Higher energy collisional dissociation Hex2Cer 2 Hexose-bound ceramide HexCer Hexose-bound ceramide HIF-1α Hypoxia-inducible factor 1-alpha Hspb1 Heat shock factor binding protein 1 iBRB Inner blood-retinal barrier ICAM-1 Intercellular adhesion molecule-1 JAK Janus kinase K Potassium LC-MS Liquid chromatography mass spectrometry LDL Low density lipoprotein Lgals3 Lectin, galactosidase-binding, soluble, 3 LRP1 Lipoprotein receptor-related protein 1 MGO Methylglyoxal MMP Matrix metalloproteinase Mn Manganese MS/MS Tandem mass spectrometry MT Metallothionein mTOR Mammalian target of rapamycin Na Sodium NAD Nicotinamide adenine dinucleotide NADH Reduced nicotinamide adenine dinucleotide Ndc Nucleoporin 1 NMDA N-methyl-D-aspartate Nppa Natriuretic peptide a PC Phosphatidylcholine PCA Principle component analysis

12

PDGF Platelet-derived growth factor PE Phosphatidylethanolamine PG Phosphatidylglycerol PI Phosphatidylinositol PKC Protein kinase C PPAR Peroxisome proliferator-activated receptor QC Quality control RAGE Receptor for AGE RPE Retinal pigment epithelium RT-qPCR Reverse transcriptase quantitative polymerase chain reaction Sco1 Cytochrome c oxidase assembly protein Se Selenium SM Sphingomyelin SOD Superoxide dismutase STAT Signal Transducer and Activator of Transcription STZ Streptozotocin T1DM Type 1 diabetes mellitus T2DM Type 2 diabetes mellitus Tbp Tata-box binding protein TCA Tricarboxylic acid TETA Triethylenetetramine TG Triacylglycerol Timp Tissue inhibitor of metalloproteinases TMS Trimethylsilyl Txnip Thioredoxin-interacting protein UK United kingdom VEGF Vascular endothelial growth factor VEGFR Vascular endothelial growth factor receptor VLCPUFA Very-long chain polyunsaturated fatty acid VLDL Very-low density lipoprotein WHO World health organisation Zn Zinc

13

Abstract

The University of Manchester

Sandra Callagy

Doctor of Philosophy

2017

An Investigation into Changes to Trace Metals and Metabolic Profiling in the Diabetic Retina

Diabetes mellitus currently affects over 422 million people globally and over 80% of patients with diabetes will develop diabetic retinopathy. Patients with diabetic retinopathy initially develop background retinopathy, which does not cause significant deterioration of visual function; however, background retinopathy may progress and lead to proliferative diabetic retinopathy and diabetic macular oedema, both of which cause severe visual dysfunction if left untreated. Current therapies for diabetic retinopathy include invasive intravitreal injections and laser photocoagulation; however these treatments only attenuate the progression of proliferative diabetic retinopathy and diabetic macular oedema. Aside from prevention by maintaining good blood glucose and blood pressure control, there are currently no treatments to prevent progression to late-stage diabetic retinopathy and new innovations in the field have not significantly progressed. For this reason, we have used untargeted –omics approaches to identify previously unknown pathological pathways in diabetes.

In this thesis, I have analysed a range of trace metals in donor retinas and found that total copper was increased in diabetic retinas compared with non-diabetic. This result was replicated in streptozotocin-induced diabetic rat retinas and further evidenced by upregulation of metallothioneins and caeruloplasmin in diabetic rat retinas compared with non-diabetic. Treatment with the copper chelator triethylenetetramine modulated these changes, the downstream effects of which require further study. This is the first description, to our knowledge, of dysregulated copper homeostasis in the diabetic retina.

I have also mapped metabolic changes in streptozotocin-induced diabetic rat retinas and found previously undescribed metabolite changes such as diabetes-induced downregulation of scyllo-inositol. This coincided with substantial changes to retinal lipids during diabetes and changes to individual lipids were consistent within their respective class. I have also found a pattern whereby regardless of the extent of change to a lipid class in diabetes, lipids containing docosahexaenoic acid (22:6 carbon chain) were consistently downregulated. This is thought to be the first study to describe this range of metabolite changes in the diabetic retina but also the first study to describe this range of metabolite analysis concomitantly within the same tissue sample.

The data from this study provides new insights into metallomic and metabolic dysfunction in diabetic retinopathy and shown that these data are reproducible. We suggest that there is plenty of scope for further research to investigate mechanisms behind copper dysregulation, how this affects pathogenesis of diabetic retinopathy along with new insights into dysregulated metabolic pathways.

14

Declaration

I declare that this thesis is entirely my own work, with the exception of one clearly described figure, and that it has not been previously submitted as an exercise for a degree at this or any other University. I give my permission to the library to lend or copy this thesis.

Disclaimer All of the work described in this thesis was carried out by me except for the following:

 Chapter 2: The body weights and blood glucose values for Studies 3, 4 and 5 described in Table 2.2 were recorded by Nina Kureishy and Anne White, Sarah Al-Adham and Emad Hindi  Chapter 3: For Figure 3.7 B, the kidney cortex was dissected by Nina Kureishy and processed and analysed by Stephanie Church  Chapter 5: The R script described in Table 5.1 was developed by Stephanie Church and Paul Begley

Sandra Callagy 29 September 2017

15

Copyright statement

i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and s/he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. iii. The ownership of certain Copyright, patents, designs, trademarks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=24420), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.library.manchester.ac.uk/about/regulations/) and in The University’s policy on Presentation of Theses

16

Dedication

This work is dedicated to my mother Paula Callagy. Type 1 diabetes is no easy feat and I will never forget the day I learned that diabetic retinopathy existed. I did this project for you and I hope that something in here will one day help to prevent the progression of diabetic retinopathy.

In memory of Toby 10 Aug 2002–23 Mar 2017

17

Acknowledgements

First and foremost I would like to thank my supervisor Paul Bishop for giving me this opportunity and having faith in me when I lost faith in myself. Your ability to stay calm and collected when I worked myself up is something I have tried to take onboard myself. Thank you also to my supervisor Garth Cooper, my walking encyclopaedia. You have no idea how inspiring you are and thank you for all the support, especially in recent months. Thank you also to my supervisor Natalie Gardiner for being so caring and supportive and even helping me when you didn’t have to. Thank you for also being a very good friend. I feel so lucky to work with an amazing trio who each are so well-renowned in their respective fields and wonderful people.

Thank you to Fight for Sight who funded this study, without which would have been impossible.

A special thanks to Paul Begley and Stephanie Church for their vast array of technical knowledge and support, who taught me all about mass spectrometry and life beyond the lab. My favourite time during this process was working with you two. Thank you also to Nicole Brace for her support with donor retinas as part of the Manchester Eye Tissue Repository and for also being a dear friend. Thank you to Anne White for taking me under your wing and for showing me a variety of techniques in the lab, especially the RT-qPCR. Thank you also to Sarah Al-Adham, Nina Kureishy and Emad Hindi for their cooperation with in vivo work and to the rest of the AV Hill crowd and CADET both past and present for your support over the years. Thank you also to the technicians at the BSF who provided a fantastic service throughout this project.

Special consideration must be made for all of my colleagues at Helios Medical Communications. Thank you for taking me on in the midst of a thesis and for believing in me for the last year. Thank you to Andrew, Simon, Emma B, Ioana, Emma M, Charlie, Sim, Nicola, Fi, Jim and Nick for being so supportive during my write-up, tolerating my sleep-deprived self and of course for improving my grammar and writing skills as I wrote my thesis. Thank you most of all to Amelia Markey for being so supportive, especially these last few months.

Thank you to my family Paula, Martin, Louise, Michelle, Toby and Milo for their unwavering support. Moving to Manchester wasn’t the easiest thing I’ve done so thank you also to Emma Lynn, Kelly-Jo Lally, Deirbhle Lally, Erin McDonnell, Micheal Devaney and Anthony Cullina for continuing to stand by me and understanding that I couldn’t always be there when you needed me. Thank you also to my Manchester friends for the support over the years especially Nicole and Sarah (double thank yous), Amy Watkins, my ARVO friend Andrew, my writing buddy Bede, Dalia, Ray and Rob, and everyone else who has provided support over the years.

Last but not least I would like to acknowledge the sacrifices made for research. Thank you to those who donate tissue in the name of research. Thank you also to the rats who took part in this study. We as researchers must always acknowledge that they are not tools but living beings capable of emotion and we must always treat then with the utmost love and care.

18

About the Author

Sandra completed a Bachelor of Science at the National University of Ireland Galway from 2007–2011 where she studied pharmacology. Following this, she worked at Medtronic Cardiovascular for a year where she worked with drug-eluting stents. During this time she piqued an interest in the effect of diabetes on the central nervous system and then went on to receive a Master of Science at Trinity College Dublin where her project focused on a mouse model of prion disease. While studying prion disease, she came across studies describing the role of copper in neurological diseases.

For the last 12 months Sandra has been an Associate Medical Writer at Helios Medical Communications. Here she provides support to a range of clients from the pharmaceutical industry on publication planning and implementation, competitor intelligence, internal educational slide decks and strategic consulting in the fields of glaucoma, diabetic retinopathy and epilepsy.

19

Chapter 1 | Introduction

20

1.1 Introduction Diabetes mellitus (DM) is a disease characterised by chronic hyperglycaemia caused by insufficiency of the secretion and/or function of insulin and the propensity to chronic organ damage, otherwise known as the diabetic complications1–4. This condition leads to impaired glucose, protein and lipid metabolism5. There are four major categories of DM; type 1 (T1DM), type 2 (T2DM), gestational (GDM) and secondary diabetes. In all forms, the ability of insulin to transport glucose into cells is lost or impaired. Insulin replenishment does not fully abrogate diabetic complications induced by hyperglycaemia, which are fuelled by underlying mitochondrial dysfunction, oxidative stress and inflammation. These include the macrovascular complications, such as accelerated atherosclerosis, diabetic cardiomyopathy and vasculopathy, as well as the microvascular complications, namely neuropathy, nephropathy and retinopathy. Given the large increase in prevalence of diabetes in recent decades, there a clear need for improved therapies through which these complications may be ameliorated. To achieve this, further understanding of the molecular mechanisms behind these complications is essential, such as in the area of retinopathy, which remains poorly understood.

1.2 Overview of Glucose Metabolism and Insulin Modulation To understand the molecular pathology of diabetes, the normal physiology of glucose metabolism and regulation by insulin must first be briefly described. During normal respiration, glucose is taken up and utilised as a main source of energy. Glucose is taken up by glucose transporters GLUT1–4, with isoforms varying by tissue. GLUT-1 and GLUT-3 are found in brain and retina, GLUT-3 in liver, kidney, intestine and pancreatic β cells and GLUT-4, the most insulin sensitive glucose transporter, is found in adipose tissue, skeletal and cardiac muscle6. Upon entry into the cell, glucose is metabolised by glycolysis to produce pyruvate. Pyruvate is transported into mitochondria and decarboxylated to produce acetyl-CoA, which then enters the tricarboxylic acid (TCA) cycle, also known as Krebs cycle and the citric acid cycle as shown in Figure 1.17. During this process, nicotinamide adenine dinucleotide (NAD+) is reduced from NAD+ to NADH during the conversion of isocitrate to α-ketoglutarate and from the conversion of malate to oxaloacetate by dehydrogenation. Similarly, flavin adenine dinucleotide (FAD) is reduced during the dehydrogenation of succinate to fumarate. Oxidation of glutamate, malate, and pyruvate also produces NADH from the reduction of NAD+ by dehydrogenases8. The NADH and FADH2 produced are then used during oxidative phosphorylation to produce energy in the form of adenosine triphosphate (ATP) by donating electrons to the electron transport chain containing the complexes 1–4. The release of energy from electrons as they are passed through the complexes provides that required for ATP production. NADH enters the oxidative phosphorylation cycle through complex I and succinate enters through complex II. β-oxidation of fatty acyls in the mitochondrion provides electrons for complex I or complex II by producing acetyl-CoA that is metabolised through the TCA cycle7,8. Electrons are passed onto complexes III and IV (also known as cytochrome c oxidase), where with the help of transition metal

21

Figure 1.1 | Overview of glycolysis and the TCA cycle Upon entry into the cell, glucose is metabolised by glycolysis to eventually produce pyruvate. Pyruvate is shuttled into mitochondria and converted to acetyl-CoA that enters the TCA cycle. Succinate and by-products of the TCA cycle (not shown) NADH and FADH are utilised in oxidative phosphorylation to produce energy in the form of ATP. Polyols can also be produced by glucose metabolism by way of the sorbitol pathway where sorbitol subsequently produces the monosaccharide fructose, or from conversion of glucose-6-phosphate to the inositols. Fructose 1,6-bisphosphate may also be diverted from glycolysis to produce free glycerol that can then be used to create triacylglycerols; a glycerol molecule bound to three fatty acyl chains. Amino acids can also be utilised in energy production as shown by entering the TCA cycle at different points. FADH, reduced flavin adenine dinucleotide; NADH, reduced nicotinamide adenine dinucleotide; TCA, tricarboxylic acid.

22

cofactors, the final electron acceptor, an oxygen atom, is reduced to water7,8. Superoxide and free radicals that cause oxidative stress are produced in the process of oxidative phosphorylation but these are controlled by superoxide dismutase (SOD) under normal physiological conditions.

During diabetes, patients suffer from loss of insulin or loss of insulin function. Insulin has many functions; one of which is the maintenance of proper blood glucose concentrations by aiding the trafficking of GLUT-4 to the plasma membrane to facilitate glucose uptake (Figure 1.2). Whilst the other GLUT proteins are insulin-independent, insulin is also required to stimulate glycogen synthesis whereby glucose-6-phosphate is diverted from glycolysis to produce glucose-1-phosphate and subsequently glycogen; a polysaccharide created from multiple glucose molecules. Glycogen can then be utilised as a source of glucose during fasting to maintain energy homeostasis. Insulin also stimulates growth by stimulating protein and lipid synthesis5,7,9. In the absence of insulin, the lack of glycogen synthesis leads to increased glucose utilisation and disrupts the balance between protein and lipid synthesis and degradation causing proteolysis and lipolysis. The lack of GLUT-4 trafficking to the plasma membrane decreases glucose uptake into skeletal muscle, a major utiliser of glucose, and increases circulating blood glucose10.

1.3 Introduction to Diabetes Mellitus According to the World Health Organisation (WHO), 422 million people are currently diagnosed with diabetes mellitus (DM)11. As stated DM presents in one of four forms; T1DM, T2DM, GDM and secondary diabetes. T1DM is classified by insulin deficiency that can only be treated by insulin therapy. This is absolutely required by the patient to prevent hyperglycaemia and usually given via the parenteral route. T1DM is derived from an autoimmune response, antibodies for which can be detected prior to symptomatic presentation. This autoantibody response is thought to derive from three components; genetic susceptibility, a temporally-important trigger and a driving antigen12.

T2DM is associated with insulin resistance whereby despite continued insulin production, its signalling is ineffective. This can be affected at many levels of the signalling cascade5,13. Risk factors of T2DM include genetic susceptibility, aging, obesity and sedentary lifestyle choices. However T2DM is also associated with reduced insulin secretion and a loss of a sufficient insulin response over time that may be caused by an increase in the pancreatic β-islet cell apoptosis to neogenesis ratio14. Increased pancreatic β-cell apoptosis is caused to some extent by aggregation of islet amyloid peptide into amylin nanofibrils, leading to hydrogen peroxide production from cellular oxygen in the presence of transition metals15,16. Oxidative stress has also been implicated in mediating insulin resistance in vitro17. The metabolic defect in T2DM frequently progresses into insulin dependence termed “secondary failure”. This form of diabetes is strongly associated with obesity and ageing and contributes to ~90% of the total diabetic population18.

23

Figure 1.2 | Overview of metabolic pathways modulated by insulin signalling Insulin signalling mediates trafficking of the glucose transporter GLUT-4 to the plasma membrane in skeletal muscle tissue via numerous mechanisms as shown. Insulin also induces lipid, glycogen and protein synthesis and encourages cell growth and proliferation. Glycogen synthesis ensures that glucose is stored and can be quickly utilised upon fasting to ensure constant energy production. During diabetes where insulin signalling does not occur or is impaired, the lack of glucose stores leads to protein and lipid degradation to maintain energy production. Adapted from Boucher et al 2014 and Saltiel and Kahn 20015,9. Akt, AKT8 virus oncogene cellular homolog; C3G, Guanine nucleotide-releasing factor 2; CAP, c-Cbl-associated protein; Cbl, cellular homologue of Cas NS-1 oncogene; Crk, CT10 sarcoma oncogene cellular homolog; Foxo, forkhead box class O; GLUT-4, glucose transporter 4; GSK3, glycogen synthase kinase; mTORC1, Mammalian target of rapamycin complex 1; P, phosphate group; PKC, protein kinase C; PIP2, phosphatidylinositol 3,4-bisphosphate; PDK1, 3-phosphoinositide- dependent protein kinase; PIP3, phosphatidylinositol 3,4,5-trisphosphate; TC10, GTP-binding protein TC10; TSC, tuberous sclerosis complex.

24

GDM has become increasingly common where insulin insensitivity occurs during pregnancy and affects approximately 5% of all pregnancies in the UK. GDM can cause hypoglycaemia, respiratory problems, and increased body weight leading to delivery complications in the newborn baby. The mother has increased risk of developing preeclampsia which may cause premature delivery19. According to the Diabetes UK 2016 Fact Sheet, women who have had GDM are seven times more likely to develop T2DM later in life than those who have never had GDM. Weight gain during pregnancy is an added risk factor and there is a 40% increase in the likelihood of developing T2DM for every 1 kg gained20. Diabetes can also occur secondary to other pathologies such as pancreatic diseases, hormonal disorders, or genetic syndromes21.

1.3.1 Diagnosis DM is a clinically and genetically heterogeneous group of disorders that have one common feature that is abnormally high levels of circulating glucose, either because of insulin deficiency or to resistance of the body’s cells to the action of insulin. Different forms of DM, as described above, have different causes and may present in different manners, therefore it is essential to highlight diagnostic criteria and circumstances of diagnosis. T1DM patients present with elevated blood glucose conentrations and/or polydipsia, polyphagia, polyuria, fatigue, slow healing wounds and blurred vision22. Uncontrolled weight loss may also occur depending on the extent of diabetic ketoacidosis. Whilst some T2DM patients are diagnosed with unexpected hyperglycaemia during a routine check, many are not diagnosed until the presentation of diabetic complications. Risk factors may therefore be employed for screening of DM. These include age, obesity, familial history of DM in a first or second degree relative, race, such as Native American, African-American, Latino, Asian-American, and Pacific Islander, history of gestational diabetes and/or changes associated with insulin resistance such as dyslipidemia, hypertension, vascular disease and polycystic ovary syndrome4,23.

Diabetes can be identified by urine or blood tests however a definitive diagnosis requires demonstration of a fasting plasma glucose level of over 7.0 mmol/L (126mg/dL) or a two hour oral glucose tolerance test of over 11.1 mmol/L (200mg/dl)1. This is the threshold at which microvascular damage occurs21. The recommendations for adults with DM, aside from gestational DM are summarised in Table 1.1. HbA1c, or glycated haemoglobin, levels reflect levels of glycaemia over the preceding 120-day period, the measurement of which has recently been approved by the American Diabetes Association and the WHO as a diagnostic criterion for diabetes at a threshold level of (≥48 mmol/mol (≥6.5%)24. The threshold for gestational diabetes diagnosis is notably less, at a fasting plasma glucose result of over 5.8 mmol/L (105 mg/dL) or a two-hour oral glucose tolerance test result of over 9.2 mmol/L (165 mg/dL)21.

25

Glycaemic Control

HbA1C <7.0%

Preprandial plasma glucose 90-130 mg/dl (5.0-7.2 mmol/L)

Postprandial plasma glucose <180 mg/dl (<10.0 mmol/L)

Blood pressure <130/80 mmHg

Lipids

Low density Lipoproteins <10 mg/dl (<2.6 mmol/L)

Triglycerides <150 mg/dl (<1.7 mmol/L)

High-density Lipoproteins >40 mg/dl (>1.1 mmol/L)

Table 1.1 | The recommended measures of glycaemic control, blood pressure and plasma lipids in T1DM and T2DM Adapted from the American Diabetes Association23

1.3.2 Cost of Diabetes Diabetes has become a major financial burden for patients, families and health services globally and this is set to rise with the expected increase in diagnoses. Approximately 10% of the National Health Service budget is spent on diabetes, not accounting for lost working days. The total cost associated with diabetes in the UK currently stands at £23.7 billion inclusive of direct and indirect costs. This is predicted to rise to £39.8 billion by 203525.

1.3.3 Global Prevalence Vascular diseases have been viewed in the past as "Western" or "first-world" ailments; however, rising urban populations, longer life expectancies and westernisation of Asian-Pacific countries has caused a surge in diabetes diagnoses. It is thought that by 2025, 60% of the global diabetic population will be Asian-Pacific26. Currently 80% of diabetes-related deaths occur in developing countries; primary care for these patients, by way of consistent check-ups as advised by the 2009 Executive Summary of Standards of Medical Care in Diabetes23, may not be accessible to these populations leading to increased mortality from diabetic complications. Diagnosis of diabetes in children is most common in Europe but this is due to a higher prevalence of T1DM. Undiagnosed diabetes is also a major issue globally but is more endemic in lower income countries; for example 90% of total diabetics in Sub-Saharan Africa are thought to be undiagnosed22. In the UK, there are approximately 3.5 million people who have been diagnosed with diabetes and a further estimated 1.1 million in the population who have undiagnosed diabetes27.

26

1.3.4 Treatments for Diabetes There are numerous treatments for diabetes, each with their own limitations, as shown in Table 1.2. Insulin-dependent patients can only be treated by parenteral injection of insulin or an insulin analogue whereas T2DM patients tend to be treated by diet and oral hypoglycaemic drugs, primarily monotherapy with metformin, followed by polytherapy when initial efficacy is lost28, further increasing the cost of treatment. As well as glucose lowering, controlling hypertension according to the United Kingdom Prospective Diabetes Study delays the development of diabetic complications such as diabetic retinopathy29. The ACCORD30 and FIELD31 studies have also shown that lowering circulating lipids using fenofibrate and statins reduce the requirement for laser photocoagulation to treat diabetic retinopathy. Whilst the current treatment regimen focuses on glucose lowering, this suggests that also reducing blood pressure and circulating lipids is also beneficial for diabetic patients and shows that diabetes is not a one-dimensional disorder.

1.3.5 Diabetic Complications As previously mentioned, 80% of global diabetic mortality now occurs in developing countries but whilst hyperglycaemia or drug induced-hypoglycaemia may be fatal, diabetic complications account for increasing mortality rates and severely reduce quality of life. An overview of the complications that can arise from diabetes are shown in Figure 1.3. Long-term DM induces diabetic vascular disease whereby changes in the vasoconstrictor-vasodilator balance is dysregulated, leading to altered blood flow in affected areas32. Given that diabetes also can increase the risk of atherogenic dyslipidemia33, it can be argued that diabetes is to some extent a vascular disease, especially when considering the range of complications that are discussed below.

Cardiovascular Complications Diabetes causes numerous changes to the cardiovascular system, including elevating plasma concentrations of very-low-density lipoproteins and fatty acids. Patients have a higher risk of hypertension and thrombosis leading to an increased risk of stroke or myocardial infarction33. Diabetes induces macrovascular damage by decreasing nitric oxide production, increasing the expression of adhesion molecules and modulating platelet anti-coagulation functions leading to deficits in wound healing34. Platelet aggregation, fibrinolytic activity and pro-thrombotic factors are increased in diabetic patients leading to hypercoagulability35. Such vascular injury may not only increase the likelihood of ischaemia and myocardial infarction, due to coronary thrombus and platelet/fibrin microemboli but may also induce tissue damage in affected areas through prolonged impairment of blood supply.

27

Drug Class Example Mechanism of Action Limitation

Amylin Agonists Pramlintide Slows gastric emptying and Initial hypoglycaemia and inhibits glucagon nausea37. Cannot be used production depending on as monotherapy plasma glucose concentrations36

Biguanides Metformin Increases non-oxidative Associated with decreased glucose metabolism in the cognitive performance. May intestine, increasing lactate be improved with calcium which can be used for and B12 supplements39 gluconeogenesis38

DPP4 Inhibitors Sitagliptin Inhibit DPP4, an enzyme Mixed outcomes on heart that breaks down the failure, potentially incretins GLP-1 and GIP, increased risk of which increase insulin pancreatitis but generally secretion and inhibit well-tolerated41,42 glucagon in response to dietary stimuli40

GLP-1 Modulators Exenatide Binds GLP-1 receptors on Possible but unconfirmed pancreatic β islet cells to increases the risk of increase insulin secretion. pancreatitis and pancreatic DPP4 inhibitors reduce cancer as well as circulating GLP-1 potentially thyroid degradation43 cancer44,45

Insulin/ Insulin Glargine Used to mimic the effect of Links to pancreatic Analogues naturally secreted insulin cancer46–48

Sodium glucose Dapaglifozin Increases excretion of Increased frequency of co-transporter-2 glucose by decreasing urogenital tract infections inhibitor renal reabsorption42 and efficacy is reduced in patients with impaired renal function42

Sulfonylureas Glibenclamide Increases insulin secretion Higher mortality rate than 50 by closing energy-sensitive metformin use . K+ channels on pancreatic β Associated with increased islet cells49 risk of pancreatic cancer51. Older generations are cardiotoxic

Thiazolidinediones Rosiglitazone PPAR-γ agonist. Cause Significant risk of migration of adipocytes myocardial infarction and from peritoneum to increased risk of subcutaneous space cardiovascular mortality53,54 reducing secretion of free fatty acids and inflammatory cytokines49. Prevents amyloid-induced apoptosis of β-islet cells52

α-glucosidase Acarbose Slows degradation of Low efficacy and inhibitors polysaccharides and gastrointestinal effects disaccharides to glucose lower compliance49. Cannot and delays absorption49 be used as monotherapy Table 1.2 | Current therapies for hyperglycaemia Listed are different classes of drugs used to treat high plasma glucose levels. DPP4, Dipeptidyl peptidase-4; GLP, glucagon-like peptide

28

Figure 1.3 | An overview of the effect of diabetes across the body Shown is an overview of diabetes-associated microvascular (retinopathy, nephropathy, and neuropathy) and macrovascular (cerebrovascular, coronary artery, and peripheral arterial diseases) complications and their long-term outcomes. TIA, transient ischemic attack. Image taken from Van den Born et al 201655.

29

Diabetic cardiac disorders may come in the form of accelerated coronary artery disease, diabetic cardiomyopathy and cardiac autonomic neuropathy and are responsible for 65% of deaths in diabetic patients33. Prevention or treatment of cardiac complications is essential for improving mortality rates amongst the diabetic population. Coronary artery disease can drastically increase the risk of ischaemia and myocardial infarction. Conversely, diabetic cardiomyopathy can occur in the absence of coronary artery disease, hypertension or valvular heart disease and presents with a loss of myocardial contractility followed by myocardial fibrosis56. Cardiac autonomic neuropathy is a potential result of neuropathy of the autonomic nervous system. Given the well-known effects of both the sympathetic and parasympathetic systems on heart rate and contractility, as well as on smooth muscle contraction, it is inevitable that damage to neurons of either system may induce a pathological heart condition as well as chronic hypertension57. It also demonstrates how the presence of one complication may enhance the severity of another; in this case, neuropathy may enhance cardiac dysfunction. The concomitant circulating lipid abnormalities and inflammation that occurs with diabetes alongside an increased risk of hypertension in diabetic patients, leads to arterial stiffness, endothelial dysfunction and atherosclerosis that predisposes patients with diabetic to an increased risk of stroke. In the Lausanne Stroke Registry between 1983 and 2002, patients with diabetes had a higher prevalence of subcortical infarction and lower relative prevalence of intracerebral haemorrhage compared with non-diabetics indicative of a pathological role of diabetes in mediating occurrence of stroke58,59

Diabetic Neuropathy Diabetic neuropathy presents in many forms including the aforementioned autonomic neuropathy, polyneuropathy and focal neuropathy. Autonomic neuropathy may contribute to cardiac complications, gastrointestinal and genitourinary dysfunction along with central effects of dizziness, poor balance and nausea60,61. The dizziness and balance deficits, caused by orthostatic hypotension, leaves the patient at increased risk of falls and injury, the severity of which may be worsened by impaired wound healing. Distal symmetric polyneuropathy is more common, occurring in 50% of diabetic patients. The most common symptoms are paraesthesias, allodynia or numbness of the foot. Patients experiencing numbness or sensory loss are more at risk of unknowingly damaging their feet61. This risk is increased when coupled with autonomic neuropathy because of changes in orthostatic pressure leading to dizziness and balance problems and therefore increased falls. Patients experiencing allodynia may experience increased disability owing to chronic pain and a high risk of developing depression62 further emphasising the deficit in quality of life.

Diabetic Nephropathy Diabetic nephropathy is a debilitating complication of microvascular origin and occurs in four phases; microalbuminuria, macroalbuminuria, nephrotic syndrome and eventually chronic renal failure, requiring dialysis33. The presence of underlying hypertension can exacerbate nephropathy. The pathological features of diabetic nephropathy such as thickening of the

30

glomerular basement membrane, glomerular and renal-cell hypertrophy and tubulointerstitial fibrosis63 are of interest given the similar features of diabetic retinopathy that will be discussed in detail in the following section. Given the relationship between each of these complications and how they intertwine, the discovery of a single treatment that targets each of these could drastically improve not only mortality but quality of life amongst this patient group.

1.4 Diabetic Retinopathy: Introduction and Pathogenesis

1.4.1 Introduction The retina is a specialised light-sensitive tissue at the back of the eye that is responsible for transducing photons into nerve impulses. It consists of a mix of different neurons and glia distributed through distinct layers as illustrated in Figure 1.4. Light passes through non-light sensitive segments until it reaches the back of the eye. There, the outer segments of photoreceptors capture photons using photopigments, triggering a change in membrane potential and subsequent nerve transduction. These photopigments are expended after exposure to light and are replenished by the retinal pigment epithelium. Photoreceptors transduce the signal to bipolar cells and horizontal cells, the interactions of which are responsible for luminance contrast sensitivity. Bipolar cells signal to ganglion cells of the ganglion cell layer and this signalling can be modulated by amacrine cells. Ganglion cells transduce the signal to the optic nerve where it is relayed to the visual processing centres of the brain64. The central retinal artery divides into four major branches which lie in the nerve fibre and ganglion cell layers. The photoreceptors are supplied by the choroidal vessels and receive the majority of total ocular blood flow because of the high metabolic requirements of these cells65.

The neuronal and endothelial cells of the retina are supported by glial cells; namely astrocytes, Mueller cells and microglia. Glial cells have roles in immunity, angiogenesis, by producing VEGF, and neuroprotection66,67. They also regulate retinal blood flow at the neurovascular unit Astrocyte cell bodies are restricted to the nerve fibre layer and provide support for ganglion cells and endothelial cells in this region of the retina and support blood retinal barrier (BRB) integrity. Astrocytes are also responsible for gamma-aminobutyric acid (GABA) and glutamate uptake as well as replenishing neuronal glutamine68. Mueller cells span the width of the retina and provide antioxidant support to photoreceptors, take up shed outer segments of cones, and maintain BRB integrity. Mueller cells also have roles in GABA and glutamate regulation66,69. Microglia are found in both the inner and outer plexiform layers and like the previously discussed glia, provide support for maintaining retinal homeostasis by mediating communication between other retinal neurons and glia to clear debris and maintain the microenvironment. They also monitor the environment owing to having receptors for numerous cytokines, chemokine, complement components and antibodies, which can lead to activation of an immune reaction. Retinal neurons and Mueller cells secrete numerous inhibitory mediators to keep microglia in a screening state but this may be prevented by changes in the microenvironment leading to a pro-

31

inflammatory state. The above-mentioned roles of the various glia will be most integral at the neurovascular unit where the vasculature, neurons and glia meet. Glial signalling contributes to retinal blood flow changes in response to neuronal requirements and BRB tight junction integrity70.

Figure 1.4 | Layout of the retina Schematic drawing of the cellular components of the retina: glia and neurons. The different cell types are situated in a standard large mammalian retina. Note the interactions between the cells and blood vessels (BV). Amacrine cells (A), astrocytes in green (AS), bipolar cells (B), cones (C), ganglion cells (G), horizontal cells (H), Müller cells in blue (M), microglia in red (Mi), rods (R), cones (C). Note the location of the different layers of the retina (from the most internal to the outer layers): optic nerve (ON), nerve fibre layer (NFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL), outer plexiform layer (OPL), outer nuclear layer (ONL), outer segment layer (OS), pigment epithelium (PE), choroid (Ch). Image from Vecino et al 2016 under Creative Commons Licence66

32

1.4.2 Introduction to Diabetic Retinopathy The 2016 WHO report on diabetes has stated that diabetic retinopathy was responsible for 1.9% of moderate or severe visual impairment and 2.6% of blindness globally11. It occurs in most T1DM patients and 80% of T2DM patients after 20 years of diabetes71. This condition is progressive and patients can experience scotomas, patches missing from vision as illustrated in Figure 1.5, blurred or fluctuating vision and overall vision loss. Patients may also experience deficits in colour discrimination72 and contrast sensitivity as well as delayed adaption to darkness73. Risk factors for diabetic retinopathy include the duration of diabetes, control of hyperglycaemia, hypertension, obesity, increased apolipoprotein B and chronic inflammation74. Prolonged diabetic retinopathy can lead to diabetic macular oedema, neovascular glaucoma and proliferative diabetic retinopathy, where new blood vessels proliferate into the vitreous. When this occurs, movement of the head can lead to traction of vitreal proteins with these vessels causing haemorrhage and sudden vision loss.

Figure 1.5 | Visual impairment from diabetic retinopathy Normal vision (A) and a simulation of vision from a patient with diabetic retinopathy (B) are shown. Note the loss of contrast, sharpness and colour discrimination as well as large areas of dark vision. 1.4.3 Pathogenesis of Diabetic Retinopathy

Background Retinopathy The severity of diabetic retinopathy is associated with disease progression and occurs gradually in stages (summarised in Figure 1.6). Diabetic retinopathy occurs initially as background retinopathy where microaneurysms begin to form as shown in Figure 1.6(A). This formation occurs due to increased venous blood flow, followed by small haemorrhages75. Patients at this stage have altered sensitivity to contrast at all spatial frequencies76 which is not present in insulin dependent or non-insulin dependent patients without retinopathy. The retinas of patients

33

at this stage of retinopathy also have increased intercellular adhesion molecule-1 (ICAM-1) immunoreactivity in the retinal vascular endothelium compared with non-diabetic controls, which implies an early induction of leukostasis77.

Moderate-Severe Non-Proliferative Retinopathy Upon disease progression, background retinopathy develops into moderate and then severe non-proliferative diabetic retinopathy. At this stage, development of soft exudates, also known as cotton-wool spots, is evident due to capillary occlusion, limiting retinal circulation, as evident in Figure 1.6(B) compared with Figure 1.6(A). These are axoplasmic debris from neuronal infarction which arises due to arteriolar constriction. Constriction occurs by a combination of factors including thickening of the arterial wall75, occlusions by microthrombi, which increase with disease progression78, and leukostasis. This stage is also associated with advanced apoptosis of vascular cells78,79. Upon the loss of retinal microvasculature, the retina becomes ischaemic because of a lack of oxygen flow to the metabolically demanding neural retina. The resultant hypoxia leads to increased vascular endothelial growth factor (VEGF) expression, which is implicated in proliferative retinopathy. Changes to the structure of the vasculature can also be visible at this stage including venous beading and tortuosity.

Proliferative Retinopathy Proliferative diabetic retinopathy occurs during the late-stage of the condition. During this phase, there is compensatory new blood vessel growth, as shown in Figure 1.6(C). This occurs due to hypoxia-induced VEGF production which encourages angiogenesis. These new vessels exacerbate pathology by growing into the vitreous and haemorrhaging. These vessels cause vitreal traction causing fibrosis which may lead to retinal detachment and sudden vision loss71; therefore it is crucial to diagnose and take preventative measures during early background diabetic retinopathy to prevent severe vision loss.

Diabetic Macular Oedema Diabetic macular oedema is a main cause of vision loss in diabetes and is caused by fluid accumulation in the retinas leading to retinal thickening of the retina with and hard exudates deposits at, or within 500 μm of the macula, as shown in Figure 1.6(D). The oedema may occur focally because of microaneurysm leakage, accompanied by hard exudate deposition, or diffusively by leakage caused by the progressive breakdown of the blood retinal barrier80. The macula contains the highest density of cone-type photoreceptors and so is responsible for high- resolution vision, damage to which this area leads to profound vision loss. Risk factors for the development of diabetic macular oedema include poor blood glucose control, congestive heart failure, renal failure, albuminuria, anaemia, due to decreased erythropoietin production, and hyperlipidaemia81. The incidence of diabetic macular oedema is up to 7.9% and 12.8% in T1DM and T2DM patients respectively82.

34

Figure 1.6 | Fundus images of the progressive stages of retinopathy Background retinopathy where microaneurysms can be identified as indicated by the black arrows (A). Pre-proliferative retinopathy where there is a clear increase in microaneurysms and soft exudate depositions, also known as cotton-wool spots are identifiable by the blue arrows. Other signs of pre-proliferative diabetic retinopathy include intraretinal microvasculature abnormalities and venous beading and tortuosity (B). Proliferative retinopathy with extensive neovascularisation is visible at the optic disc (white arrows) (C). Hard exudate deposits (purple arrow) by the macula indicative of vascular leakage of lipids indicative of diabetic macular oedema, a leading cause of vision loss during diabetic retinopathy (D). Images from Comm Eye Health Vol. 24 No. 75 201183 accessed at http://www.cehjournal.org/article/diabetic-retinopathy- dr-management-and-referral/

35

1.4.4 Changes to the Retinal Vasculature during Diabetic Retinopathy

Blood-Retinal Barrier Breakdown Blood flow to the retina is protected by two BRBs, an outer (oBRB) and inner BRB (iBRB), that are formed by tight junctions between adjacent retinal pigment epithelial (RPE) or endothelial cells respectively. Some of the earliest changes to occur in the retinal vasculature in DM are breakdown of the BRBs and subsequent changes to vascular permeability. When this breakdown occurs, excess liquid, which can increase with hypertension, cannot be drained leading to macular oedema. During diabetic macular oedema, leakage from the microvasculature into the macula leads to swelling and deposition of lipids from the plasma causing subsequent loss of central vision71. Balance between hydrostatic pressure and oncotic pressure is essential in preventing oedema, especially given the lack of a retinal lymph network, and is disrupted during BRB breakdown84. iBRB breakdown in the streptozotocin (STZ) rat model of diabetes is accompanied by increased nitric oxide formation, nitric oxide synthase activity, lipid peroxides and VEGF expression85. Loss of tight junctions between endothelial cells leads to increased vascular permeability. This is associated with decreased occludin expression in rat endothelial cells after 3 months of STZ- induced diabetes86 and is coupled with down-regulation of natriuretic peptide receptor C, a regulator of vascular electrolyte concentrations and fluid balance, expression in the same model implicating numerous molecular mechanisms in iBRB pathology87. However, macromolecule leakage has been found to occur more greatly from the oBRB than iBRB and leakage from the oBRB may have more substantial contribution to occurrence of diabetic macular oedema than iBRB leakage88. oBRB breakdown has been linked to protein kinase C ζ overactivation and concurrent cone degeneration89

Thickening of the Vascular Basement Membrane The vascular basement membrane is a thin membrane that surrounds the neurovascular unit. Major components include collagen IV, fibronectin, laminin and heparan sulphate proteoglycans. It provides a semi permeable barrier and its main function is in cell adherence90.Thickening of the basement membrane, a distinguishable histological feature in diabetic retinopathy, therefore may interrupt critical maintenance signalling and may contribute to vascular leakage that can lead to diabetic macular oedema91. Shear stress increases endothelial cell expression of basement membrane components to prevent endothelial tears by thickening vessel walls and increasing vascular integrity. Whilst initially a protective mechanism, the increased production of collagen IV, laminin and fibronectin84 thickens the basement membrane and becomes pathogenic. Therefore, hypertension, which causes endothelial tears, can exacerbate this effect. The degree of thickening coincides with the duration of diabetes and mostly consists of collagen IV which is surrounded by a thinner ring of fibronectin and tenascin, the latter of which is not present in non-diabetic basement membranes92. Extracellular accumulation of advanced glycation end-products (AGEs) also increases basement membrane

36

thickness. Basement membrane thickening is also associated with increased utilisation of the polyol pathway and protein kinase C (PKC) activation90,93.

Pericyte and Capillary Loss and Neovascularisation Pericyte loss is a characteristic feature of diabetic retinopathy and has been shown in numerous studies of the STZ-induced diabetic model. Hammes et al 199794 showed an increase in the endothelium-pericyte ratio and increased acellular capillaries at 6 months. This was accompanied by endothelial cell loss at 32 weeks, and 8 and 9 months in three studies95–97. A key lesion in diabetic retinopathy progression is the loss of vascular cells. Pericyte loss specifically appears to mediate progression and is of particular interest given their proximity to endothelial cells, and their role in modulating angiogenesis and architecture, barrier integrity and capillary blood flow. The loss of endothelial cells can lead to a hypoxic environment, which triggers neovascularisation. There are several interactive mediators between pericytes and endothelial cells including platelet-derived growth factor-B (PDGF-B) and VEGF. Whilst PDGF-B mediates pericyte proliferation and migration, this is inhibited during angiogenesis by VEGF-2 receptor-mediated VEGF signalling to allow the pericyte to surround the vascular sprout and aid stability and growth. Conversely in hypoxic conditions, VEGF directly induces pericyte proliferation and migration98.

This hypoxic environment induced by capillary loss drives consequent angiogenesis in proliferative diabetic retinopathy. This stimulates the production of weak, new blood vessels on the venous side of the capillary circulation75. These vessels penetrate the inner limiting membrane into the vitreous cavity of the eye where they may haemorrhage71.Production of fibrous connective tissue can cause tractional retinal detachment leading to sudden vision loss84. New blood vessels can also penetrate the stroma of the iris. The consequent fibrosis may block the drainage of the aqueous humour, increasing intraocular pressure and leading to diabetic neovascular glaucoma, another source of sight loss. VEGF expression is increased in hypoxic conditions99 and this increased expression has been implicated in the pathology of proliferative retinopathy. This may possibly be due to a loss of the balance between VEGF and pigment-epithelium-derived factor, which opposes neovascularisation, during disease progression71. Nppa mRNA expression (natriuretic peptide precursor type A) is reduced in rats after 3 months of STZ-induced diabetes, the normal function of which is to counteract VEGF induced angiogenesis and permeability100. This indicates an early attenuation in VEGF inhibition in the STZ model prior to capillary loss.

1.5 Diabetic Retinopathy: Metabolic and Inflammatory Changes Diabetes is a disease that impairs the metabolism of glucose and so metabolic pathways are affected in diabetic retinopathy. Under normal circumstances, glucose is metabolised by the glycolysis pathway where glucose is phosphorylated by hexokinase followed by a series of reactions leading to the production of pyruvate. This is the substrate for the Krebs cycle in which NADH is produced and used in oxidative phosphorylation for energy production. During

37

diabetes, there is an increase in the flux of glucose into the retina mediated by increased GLUT-1 at the blood retinal barrier as shown in diabetic donor eyes compared with non-diabetic101. This increase in glucose concentrations saturates the glycolysis pathway and causes glucose to be shuttled down the polyol or hexosamine pathways instead leading to the production of AGEs.

1.5.1 The Polyol Pathway The polyol pathway has been implicated extensively in diabetes due to sorbitol accumulation96,102,103. In this pathway, as shown in Figure 1.7, glucose is reduced by aldose reductase to produce sorbitol. Excessive use of this pathway can lead to oxidative stress, AGE and PKC production; all of which may exacerbate pathology in the microvasculature104. Oxidation of sorbitol to fructose can lead to further AGE production. Sorbitol can also increase expression of inflammatory genes and those involved in extracellular matrix remodelling and cell adhesion, and may also have a role in increasing polymerization of the actin cytoskeleton. In a 6-month STZ rat model, sorbinil, an aldose reductase inhibitor, modulated the expression of numerous immune mediated markers including antioxidants, anti-inflammatory agents, and markers of the IFN-γ pathway103. It should be noted that the consumption of NADPH may be responsible for these effects. Increased NADH production from the TCA cycle can inhibit complex III of the electron transport chain, increasing free radical intermediates of coenzyme Q and increasing superoxide production104 and therefore increase oxidative stress. NADH production is increased in the polyol pathway and so increased flux through the polyol pathway likely contributes to increased superoxide production during hyperglycaemia.

1.5.2 Hexosamine pathway Glucose metabolism is also modified in diabetes by increased utilisation of the hexosamine pathway. In this case, fructose-6-phosphate is diverted from the normal glycolysis pathway by glutamine: fructose-6-phosphate aminotransferase as illustrated in Figure 1.8. The products of this pathway, UDP-N-acetylglucosamine and UDP-N-acetylgalactosamine add glucosyl side chains to both proteins and lipids105 aiding proteoglycan synthesis thereby altering the functions and interactions of these macromolecules. GFAT may enhance hyperglycaemia-induced TGF-α-mediated smooth muscle cell proliferation106, matrix production107, and transcription of plasminogen activator inhibitor-1, a serine protease inhibitor108. SP1 is a transcription factor that regulates plasminogen activator inhibitor-1 transcription. SP1 activity can be enhanced by N-acetylglucosamine to increase plasminogen activator inhibitor-1108. Plasminogen activator inhibitor-1 has been shown to mediate increased endothelial cell migration and neovascularisation during retinopathy109, further emphasising a mediatory role of the hexosamine pathway during diabetic retinopathy. This pathway indicates a direct relationship between hyperglycaemia and AGE production.

38

Figure 1.7 | The Polyol Pathway and AGE production Excessive glucose metabolism causes a shift in equilibrium to the polyol pathway initiated by the rate-limiting enzyme aldose reductase. The reduction of glucose to sorbitol requires the oxidation of NADH to NAD+. The increased use of this pathway can deplete NADPH therefore limiting glutathione regeneration and leading to enhanced oxidative stress. The loss of NAD+ due to conversion to NADH inhibits GAPDH production. GAPDH production during glycolysis acts as a critical checkpoint and inhibition may lead to PKC and AGE production110. This is due to an increase in triose phosphates which can lead to diacylglycerol and subsequent PKC production. It may also lead to methylglyoxal production which is one of the precursors for AGEs104. Fructose production from sorbitol causes frutosylation of various proteins leading to subsequent production of 3-deoxyglucosone; another precursor for AGE products. Abbreviations: GAPDH, glyceraldehyde-3-phosphate dehydrogenase; AGE, advanced glycation end-products; DG, diacylglycerol; PKC, protein kinase C.

39

Figure 1.8 | The Hexosamine Pathway Glucose metabolism is diverted from the normal glycolysis pathway at the point of fructose-6- phosphate, which is normally metabolised by phosphofructokinase-1 to yield fructose 1,6- bisphosphate. However in the hexokinase pathway, fructose-6-phosphate is metabolised by GFAT and subsequently leads to UDP-N-acetylglucosamine and UDP-N-acetylgalactosamine production which can glycate proteins and lipids and lead to AGE accumulation. UDP-N- acetylglucosamine activity also increases PAI-1 activity105. PAI-1, plasminogen activator inhibitor-1; GFAT, glutamine:fructose-6-phosphate aminotransferase.

40

1.5.3 AGEs Whilst only the metabolic abnormalities that lead to AGE production have been discussed thus far, the consequences of their production play further roles in pathology. AGEs are formed as shown in Figure 1.9. Whilst excessive glucose strongly enhances AGE formation, the α- oxaloaldehydes, glyoxal, methylglyoxal (MGO) and 3-deoxyglucosone (3-DG) elicit higher rates of AGE production. These precursors are formed from glucose metabolism and lead to enhanced intracellular AGE production111. AGE is an umbrella term for a variety of moieties as each precursor can produce a different AGE; glyoxal produces carboxymethyl-arginine and MGO forms Nε-(carboxyethyl) lysine and arginine-hydroimidazole112. Not only does the precursor determine the AGE formed, but the amino acid initially bound can also shape the final product. For example, 3-DG can produce pyrraline or imidazolone-type AGE depending on the amino acid bound, in this case lysine and arginine respectively. It is also important to mention Nε-(carboxymethyl) lysine (CML) as it is generally included among the AGEs. However, whilst the former AGEs are formed under non-oxidative conditions, CML is produced by oxidative cleavage making it a glycoxidation product as opposed to a glycation product111. The heterogeneity of AGE products makes these difficult to detect and target.

Products of both oxidative and non-oxidative AGE formation have been detected in the diabetic retinal vasculature113. An in vitro study by Padayatti et al 2001114 showed increased MGO in retinal endothelial cells exposed to a high glucose environment, indicating the connection between AGE production and hyperglycaemia. The relevance of AGE production is due to intracellular and extracellular accumulation in diabetic vasculature95. AGE accumulation can modify proteins by inducing cross-linking or changing their tertiary structure thereby decreasing protease degradation, modifying enzymatic activity or disrupting signalling pathways112. CML is a component of the thickened basement membrane that occurs in diabetic retinopathy, thereby mediating retinal pathology in diabetes111. AGE accumulation induces oxidative stress, apoptosis115 and carbonyl stress97. AGE adducts elevate ICAM-1 thereby increasing leukostasis116, a mediator of capillary closure. Low concentrations of AGE have been shown to reduce pericyte proliferation but increase that of endothelial cells indicating a role in pericyte loss in retinopathy. Conversely higher concentrations can be toxic to endothelial cells in vitro117 which could suggest that progressive accumulation of AGE mediates endothelial cell loss with disease progression.

RAGE, the main AGE receptor, is found in endothelial cells and Mueller cells and co-localizes with CML-type AGE during STZ-induced diabetes94. RAGE can also bind numerous inflammatory mediators112 including S100B, a marker of inflammation in neuronal tissues118; high mobility group box 1 protein, which is known to interact with other inflammatory mediators119, and amyloid-β. The latter may be significant given that induction of diabetes in a mouse model of accelerated aging increased cerebral amyloid-β and memory deficits compared with non-diabetic mice120. This signifies a key inflammatory role of RAGE signalling, which is emphasised by induction of mitogen-activated protein (MAP) kinase and Janus kinase/Signal

41

Transducer and Activator of Transcription (JAK/STAT) phosphorylation and subsequent Nf-κB activation112. Nf-κB has a known role in oxidative stress and has been shown to be involved in diabetic cardiac dysfunction, where it produces mitochondrial free radicals and reduces mitochondrial integrity.

Lipid metabolism is often dysregulated during diabetes leading to the production of advanced-lipoxidation end-products (ALEs). Haemoglobin levels of FDP-lysine, an acrolein- derived ALE, are correlated with severity of retinopathy which may be explained by the retinal environment which is rich in lipids112 and oxidative reactions caused by photoreceptor transduction121. The lipophilic nature of ALEs can perturb plasma membrane integrity and inflammatory signalling by bleb formation122.

Figure 1.9 | The process of AGE production AGEs are produced as a result of non-enzymatic-glycation of the amino group of proteins, lipids, or DNA, by sugar or a sugar fragmentation product leading to the production of a Schiff base. This can slowly rearrange to form a protein-bound Amadori product, also known as an Amadori adduct. Both the Amadori product and Schiff base undergo oxidation reactions and dehydrations to form free-radical intermediates that irreversibly produce AGEs in a reaction that is mediated by transition metals, notably iron and copper112,123. Schiff base adducts may directly produce AGEs by a slower process involving oxidation and dehydration of the Schiff base. AGE production exceeds first order kinetics during diabetes leading to excessive accumulation.

42

1.5.4 Lipid mediators The retina has a unique lipid profile and contains the highest ratio of docosahexaenoic acid (DHA)-containing polyunsaturated fatty acids (PUFA) and very long-chain PUFAs (VLCPUFA) that have chains comprising up to 38 carbons. Therefore, it is unsurprising that lipids may play a major role in diabetic retinopathy. Retinal fatty acids are obtained from the circulation, de novo lipogenesis or from remodelling PUFAs, from glucose or dietary fatty acids, by desaturation and elongation reactions124. Insulin induces elongase activity and so the lack of insulin signalling that occurs in diabetes causes changes to fatty acid composition owing to reduced remodelling125. Changes to retinal fatty acid composition impacts the balance between a pro-inflammatory and anti-inflammatory environment owing to the pro-inflammatory effects of n-6 chains on endothelial cells that leads to increased intercellular adhesion molecule (ICAM)-1 and vascular cell adhesion molecule (VCAM)-1 proteins, an effect that was notably observed in retinal but not umbilical endothelial cells126, and the anti-inflammatory effects of n-3 fatty acids such as DHA which has been shown to prevent the induction of both ICAM-1 and VCAM-1127.

Catabolism of the major lipid classes such as glycerophospholipids, the main components in cell membranes produces small lipid mediators such as oxidised bioactive lipids, prostaglandins and endocannabinoids. The oxidation of n-6 PUFAs can lead to the production of thromboxanes and prostaglandins124. Prostaglandins are well-known for their roles in inflammation but also regulate blood flow. Phospholipases release arachidonic acid, a 20 carbon-length chain with four double bonds (20:4 fatty acyl chain), from phospholipids, which is then cleaved by cyclooxygenase enzymes to produce the eicosanoids. Prostaglandins are pro-angiogenic and cyclooxygenase-2 may regulate angiogenesis by interacting with VEGF128 which is further emphasised by the correlation between prostaglandin E2 and VEGF protein levels in the vitreous of patients with proliferative diabetic retinopathy129. Phospholipase activity is increased in bovine retinal endothelial cells during high glucose and fluctuating glucose conditions leading to increased prostaglandin E2 production, indicating a more inflammatory environment during hyperglycaemia. Of the cyclooxygenase enzymes only cyclooxygenase-2 is up-regulated in diabetic retinal microvessels compared with non-diabetic, and phospholipase activity in the microvasculature has been shown to be essential for TNF-α, ICAM-1 and VEGF upregulation in vivo130. Similarly, increased PKC activity during diabetic retinopathy increases phospholipase activity and may increase the production of endocannabinoids. Of the endocannabinoids, N- arachidonoylethanolamine, better known as anandamide, is increased in ocular tissues of patients with diabetic retinopathy compared with those without retinopathy131. Anandamide is a precursor of prostamides upon oxidation with cyclooxygenase-2132 and it is thought that endocannabinoids may have a role in diabetic retinopathy133. Lipid classes and lipid signalling is an area that has remained largely unexplored in diabetic retinopathy and may provide new insights and therapeutic targets.

43

1.5.5 Inflammatory Molecular Mediators of Diabetic Retinopathy Increased production of pro-inflammatory small lipid mediators only accounts for some of the contributions to a pro-inflammatory environment during diabetic retinopathy. Numerous immune mediators have been implicated in diabetic retinopathy causing a pro-inflammatory environment that induces apoptosis of pericytes and endothelial cells. The numerous pathways activated during diabetic retinopathy have made it difficult to isolate a key mechanism behind the induction of apoptosis and inflammation. Implicated mediators are discussed here with the aim of clarifying the interactions between different responses during diabetic retinopathy.

Cytokines and apoptosis induction Inflammatory cytokines are key immune mediators that can become dysregulated in the absence of an infectious pathogen and create a damaging microenvironment. Numerous mediators associated with cytokines are increased in the STZ diabetic rat model including A2m87, a cytokine transporter induced by interleukins expressed in blood vessel walls and induced by interleukins and JAK3 and STAT3, the signalling of which is associated with the common cytokine receptor γ chain100,134. Interleukin-1β (IL-1β) expression is induced by hyperglycaemia in has been bovine retinal endothelial cells, and not in Mueller cells or astrocytes, in a PKC dependent manner. IL-1β signalling causes downstream activation of Nf- κB-induced transcription leading to subsequent oxidative stress and CCL2 production; a chemoattractant that recruits inflammatory cells to the site of expression135. Similarly expression of CHI3L1, CCR5 and CD44, which encourage chemotaxis, leukocyte infiltration and trafficking respectively, is increased in the retina after 3 months of STZ-induced diabetes100. CD44 also aids signalling of the TGF-β type 1 receptor, activin receptor–like kinase 5 (ALK-5). ALK-5 signalling can induce apoptosis by activating caspase-3. This agrees with the finding of increased caspase-3 activity in rat retinas after 3 months of STZ-induced diabetes100. Conversely, expression of HSPB1, an inhibitor of caspase-dependent and -independent apoptosis, is also increased indicating a counter-active mechanism to the adverse effects of retinopathy that becomes overridden during disease progression100. Bax levels, indicative of increased apoptosis, are also increased in diabetes. This may be because of lower levels of IGF-1 which has pro-survival properties136 indicating an overall increase in pro-apoptotic factors during diabetic retinopathy.

PKC Activation PKC activation is a hallmark trait of diabetic retinopathy. Of the numerous isoforms, PKC-α, -β, -δ and -ε are activated in the diabetic retina and modulate endothelial cell enzymatic activity, pericyte growth and Nf-κB activation. PKC-δ in particular is up-regulated during hyperglycaemia, specifically in pericytes, and is associated with the appearance of acellular capillaries by inhibiting PDGF-B and inducing p38α MAPK 3. Hyperglycaemia-induced PKC activation is associated with decreased retinal blood flow137 and may account for early vascular changes138. PKC expression may also be induced by oxidative stress139. Diazepam binding inhibitor (Dbi) is an insulin regulated transport molecule that regulates lipid metabolism and modulates retinal

44

neurotransmission by interacting with GABA and is secreted in response to PKC. Protein levels of Dbi are increased in the retina after 3 months of STZ-induced diabetes73.

Complement Activation Complement activation has also been associated with diabetic retinopathy. Inhibitors of complement, CD55 and CD59, are down-regulated during hyperglycaemia in vitro by direct glycation140. Retinal sections from human diabetic donors have shown increased levels of the C3 and C5b-9 components of complement and the membrane attack complex compared to non-diabetic donors. There was no change in C1q and C4 levels between diabetic and non- diabetic donors indicating complement activation via the alternative pathway. Decreased levels of CD55 and CD59 have also been shown in the STZ rat model after 10 weeks of diabetes141. Given that the membrane attack complex induces a pore in the target cell leading to cell lysis, induction of complement can lead to inflammation due to the release of normally intracellular damage-associated molecular pattern molecules, leukostasis owing to inflammatory insult, and thrombosis given the procoagulant nature of the pathway and also because of narrowing of vessels during leukostasis. Considering these effects have been stated already, complement may have a role in inducing these effects or worsening existing pathology.

1.5.6 Oxidative Stress The excessive production of free radicals during oxidative stress enhances AGE production, as expected by the role of free radical intermediates in AGE formation as shown in Figure 1.8. Free radical production is increased in diabetes because of the utilisation of glucose by alternative metabolic pathways as described as well as increased oxidative phosphorylation leading to alterations in mitochondrial function and increased superoxide production. This can also lead to increased hydroxyl radical production via the Fenton and Haber-Weiss reactions142 and increased AGE production143.

Cells have endogenous mechanisms of countering oxidative stress by the use of mediators known as antioxidants. The most well known of these are the superoxide dismutases (SODs) consisting of SOD1, an intracellular copper-containing SOD, SOD2, a mitochondrial manganese-containing SOD, and SOD3, an extracellular copper-containing SOD. SOD1 and SOD2 are thought to mitigate a large extent of superoxide radicals whereas SOD3 protects the extracellular matrix and is highly expressed in blood vessels144,145. Glutathione peroxidises (GSH-Px) are also imperative to the protection of the cell from oxidative stress and where the

SODs protect against superoxide formation, GSH-Pxs protect from H2O2 that is produced from antioxidant action, such as that from the SODs in a reaction that requires NADPH. Four GSH-

Pxs are known including the cellular GSH-Px-1, which is ubiquitous and reduces H2O2 and fatty acid peroxides, GSH-Px-2 that is only found in gastrointestinal epithelial cells, GSH-Px-3, an extracellular antioxidant and GSH-Px-4, which is a membrane-bound antioxidant that reduces esterified lipids. Nonenzymatic antioxidants such as vitamins C and E, glutathione (GSH) which

45

aids production of GSH-Px and is replenished by NADPH, flavanoids and beta-carotene also contribute to mediating oxidative stress144.

There is substantial evidence for the presence of oxidative stress during diabetes. Superoxide concentrations are significantly increased in both retinal Mueller cells and bovine retinal endothelial cells incubated with high glucose media reducing cell viability, and in the retinas of diabetic rats146. Superoxide dismutase (SOD) activity, which is important in lowering superoxide levels, is itself decreased when exposed to excess glucose in vitro due to inactivation of the enzyme by glycation147. In retina, SOD2 has been found to decrease in diabetic rat retinas compared with non-diabetic where SOD1 was unchanged148. Elevated reactive oxygen species increases AGE and ALE formation and therefore oxidative stress in a feed-forward manner. Non-oxidatively derived AGEs, such as 3-DG and MGO, may induce carbonyl stress as opposed to oxidative stress, where modifications to proteins and lipids are caused by carbon groups rather than reactions with oxygen atoms. Neutralising reactive carbonyls can prevent formation of protein cross-links149. Metal-catalysis of ascorbate oxidation can lead to AGE formation150 which can be prevented by metal chelating agents97. This reduces the formation of reactive carbonyl species149 and indicates a relationship between both oxidative and carbonyl stress and transition metals. Studies have established that circulating SOD, GSH, GSH-Px, beta-carotene and antioxidant vitamins C and E are decreased in patients with diabetes. However, the results of randomised-controlled clinical trials that have investigated supplementation with antioxidants have been inconclusive151. This may be because of excessive oxidative stress and warrants further basic research on oxidative stress mechanisms in diabetic retinopathy.

1.5.7 Transition Metals There is increasing interest in the role of transition metals in the progression of diabetic complications given their redox reactivity and role in propagating AGE production152,153. There is increasing interest in investigating the roles of the transition metals zinc, iron and copper (Cu) in diabetic complications142,154,155. Zinc supplementation is thought to have protective effects in diabetic retinopathy and cardiomyopathy156–159. Iron accumulation has been demonstrate in rodent models of diabetic nephropathy and pathology is further aggravated during iron supplementation160,161. Administration of the iron chelator deferiprone has also been shown to alleviate inflammation in an STZ model of diabetic nephropathy.

There currently is vast evidence for the role of dysregulated copper (Cu) homeostasis in diabetic complications142. Cu is essential for enzymatic activity such as cytochrome-c oxidase- mediated electron transport in mitochondria, copper/zinc-requiring SOD for free radical neutralisation, and caeruloplasmin ferroxidase activity to name a few. As such, Cu is essential for oxidative phosphorylation as a component of cytochrome c oxidase because it induces a proton gradient during oxidative phosphorylation and reduces oxygen to water162. Cu occurs

46

physiologically in two forms, Cu(I) and Cu(II), and catalyses the Haber-Weiss reaction, which uses Fenton chemistry163, to produce toxic radicals in a cycling reaction described as:

●- Cu(II) + O2 → Cu(I) + O2

- ● Cu(I) + H2O2 → Cu(II) + HO + HO

This produces reactive oxygen species formation and hydroxyl production that can become pathological when these are not appropriately neutralised. AGE-mediated modification of extracellular matrix proteins, such as collagen, induces cross-linking and structural changes including increased copper-binding site formation164. Increased copper binding may enhance oxidative stress at the site of AGE accumulation due to free radical production but can also enhance glycation in a concentration-dependent manner165 indicating a feed-forward reaction between AGE formation and oxidative stress. Dysregulated Cu homeostasis has been implicated in rodent models of both diabetic cardiomyopathy and nephropathy whereby Cu is deficient in cardiac tissue166 and accumulates in the kidney cortex (unpublished data). Both copper balance and inflammatory mediators are normalised by Cu chelation166–169. To date, modifications to transition metal regulation during diabetic retinopathy have not been extensively evaluated.

1.6 Current treatments for Diabetic Retinopathy

1.6.1 Preventative Therapy Diabetic retinopathy is still a major source of clinical need. Currently the only prevention is maintenance of blood glucose and pressure control71. Maintenance of tight glycaemic and blood pressure control reduced the risk of retinopathy progression by 54%170 and 34% respectively29. Reduced hyperglycaemia evidently leads to a decrease in glycation modifications of macromolecules and therefore AGE production. As mentioned previously, high blood pressure can enhance the production of extracellular proteins and may have a role in the thickening of the retinal basement membrane. Treatments at present are only symptomatic and do not treat the root cause of damage. They are also only effective in late-stage disease, upon the occurrence of diabetic macular oedema and proliferative retinopathy.

1.6.2 Treatments for Diabetic Macular Oedema Diabetic macular oedema is a major complication of diabetic retinopathy that causes profound vision loss. Focal laser coagulation is currently the main treatment for diabetic macular oedema and was demonstrated by the Early Treatment Diabetic Retinopathy Study to reduce moderate vision loss by over 50%171. Whilst laser treatment reduces the risk of further vision loss, it lacks restorative qualities and may cause subretinal fibrosis, atrophy of the pigment epithelium and risk of sub-retinal neovascularization172.

47

Anti-VEGF therapy is a newer therapy for diabetic macular oedema. Drugs of this class currently on the market include pegaptanib, ranibizumab, bevacizumab and aflibercept all of which have been shown to improve visual acuity when injected intravitreally173. Ranibizumab has been shown to be more efficacious in improving visual acuity during diabetic macular oedema than focal laser photocoagulation. A serious adverse effect of treatment is endophthalmitis, infection of the intraocular space, which occurs in <1% of patients treated174. This is caused by the introduction of bacteria into the eye during the intravitreal injection procedure. Aflibercept has also been shown to improve visual acuity compared with laser treatment173.

Corticosteroids are also beneficial during diabetic macular oedema due to their anti- inflammatory properties. By targeting inflammation, intravitreal injection of triamcinolone acetonide can improve the integrity of tight junctions and improve visual acuity. These effects are short lasting and triamcinolone must be re-injected every three to six months. In this context laser therapy is considered superior due to the longer duration of therapeutic effect175. Triamcinolone treatment increases the risk of increasing ocular pressure and could lead to glaucoma176. Further disadvantages are the risks of infection and cataracts177. The use of an intravitreal implant of fluocinolone acetonide implant improved visual acuity without the need for reinjection and was shown to be more effective at reducing macular oedema compared with focal photocoagulation-treated patients or those with no treatment. However there was a high rate of raised intraocular pressure178.

1.6.3 Treatment for Proliferative Diabetic Retinopathy Scatter panretinal photocoagulation is generally used in the treatment of proliferative retinopathy as opposed to the focal photocoagulation that is used for macular oedema. By targeting evenly spaced laser burns onto the outer photoreceptors and retinal pigment epithelium, this method allows the choroid to supply more oxygen to the inner retina as the highly metabolically active photoreceptors are destroyed. By decreasing retinal hypoxia, it also inhibits pathogenic growth factor production. Whilst this method reduces the risk of severe vision loss from proliferative diabetic retinopathy, it can enhance diabetic macular oedema as well as induce peripheral visual-field defects, reduced night vision, diminished colour vision, and decreased contrast sensitivity given the widespread destruction of viable photoreceptors179.

The anti-VEGF drugs pegaptanib, ranibizumab, bevacizumab and aflibercept, previously mentioned as therapeutics for diabetic macular oedema have also been tested for effect in proliferative diabetic retinopathy. Pegaptanib treatment causes regression of neovascularisation, but in 3 out of 8 patients with regression, reappearance occurred 52 weeks post the last pegaptanib injection180. Ranibizumab significantly reduces diabetic retinopathy progression in patients with non-proliferative retinopathy181. In patients with proliferative retinopathy, ranibizumab improved visual acuity and significantly reduced the risk of recurrence compared with saline control182. Bevaciumab has not been approved for ocular use but

48

intravitreal injection has been associated regression of retinal neovascularisation and improvements in vitreous haemorrhage and macular thickness; however, this response decreased after 24 months183. Aflibercept is currently in phase II/III trials for active proliferative diabetic retinopathy with a primary completion date due at the end of 2014. Anti-VEGF therapy has an associated risk of fibrosis and scarring when used for treatment of proliferative diabetic retinopathy. This is due to the loss of balance between VEGF and connective tissue growth factor184.

1.6.4 Prospective Therapeutics Research into therapeutics for diabetic retinopathy is active with numerous targets under investigation. Pigment epithelium-derived factor is an anti-angiogenic agent that counters the effects of VEGF secretion. Pigment epithelium-derived factor levels are reduced in patients with diabetic retinopathy and intravitreal administration may serve to reduce pathological traits. Research in this area is ongoing to improve stability and half-life and reduce peptide size185. From the described changes to the polyol pathway, aldose reductase would appear to be an ideal target for drug treatment; however, whilst inhibitors have been effective in preclinical studies, toxicity in clinical trials due to deficits in aldehyde detoxification have negated the use of this target to date186. Inhibiting PKC-β using ruboxistaurin has shown clinical benefit in patients by reducing sustained moderate visual loss; however, effects in patients after 2 years of treatment were only borderline significant187. This has led to an FDA request for more data further indicating a clinical benefit prior to approval. Targeting the renin-angiotensin system may also induce benefits due to its role in pericyte migration, metalloproteinase-induced remodelling of the extracellular matrix, fibrosis and angiogenesis and its upregulation during hypoxic angiogenesis; however, studies to date have not provided strong evidence of a therapeutic benefit177. Another prospective therapy is the use of peroxisome proliferator-activated receptor (PPAR) receptor agonists. PPARα receptors are down-regulated in numerous rodent models during diabetes and STZ-induced diabetic PPARα knock-out mice had fewer pericytes and more acellular capillaries than diabetic wild-type mice. The FIELD and ACCORD Eye trials showed that diabetic patients with retinopathy receiving fenofibrate, a PPARα agonist, had a reduced requirement for laser photocoagulation and attenuation of disease progression over 4 years compared with those receiving simvastatin respectively188.

Other novel targets include somatostatin analogues, antioxidants and carbonic anhydrase inhibitors. Somatostatin analogues prevent neovascularisation by a direct pro-apoptotic effect on endothelial cells189. Despite the role of oxidative stress described in diabetic retinopathy so far, antioxidant therapy has not been extensively investigated. The HOPE study and the MICRO-HOPE sub-study indicated a lack of efficacy in preventing diabetic microvascular complications but only focused on diabetic nephropathy190 . The study by Mayer-David et al (1998)191 failed to find a protective effect of vitamin C, vitamin E or beta-carotene in diabetic retinopathy. However, this study did not assess serum levels of these antioxidants and a limitation was a lack of specificity of supplement use prior to commencing the study. A 5-year

49

follow-up study of the effects of antioxidant supplementation on the progression of diabetic retinopathy indicated that whilst best-corrected visual acuity was unchanged, progression of diabetic retinopathy was attenuated in patients receiving supplementation compared with those who did not192. Carbonic anhydrases are responsible for the conversion of carbon dioxide to bicarbonate and protons and are used as anti-glaucoma medication because of their intraocular pressure-lowering qualities. Carbonic anhydrase inhibition has been found to decrease vasopermeability193, improve retinal blood flow194and increase oxygen tension in the retina195 and therefore has the potential to reduce hypoxia. Whilst there are many therapies in development, it has also been demonstrated here that many therapies fail during clinical trials either due to loss of efficacy or safety issues. Approved therapeutics such as VEGF inhibitors and corticosteroids only target late stage retinopathy, at which stage the microvasculature is already severely damaged and these agents do not restore ischaemic areas.

1.7 Rodent models of Diabetic Retinopathy Whilst cell-specific research can be done by using retinal vasculature cells in culture, to study the pathology of diabetic retinopathy an in vivo model must be used. Generally, rodents are used due to the low cost of maintenance and their common use throughout the literature eases the comparability between different studies. Most models involve the induction of diabetes to induce diabetic retinopathy. Diabetes can be induced chemically or spontaneously by genetic or diet alterations. However, proliferative diabetic retinopathy does not occur in these models.

1.7.1 Chemically-Induced Diabetic Models The leading chemical inducers of diabetes are alloxan and STZ. Both inducers are toxic glucose analogues that cause pancreatic β-islet cell necrosis; however, the mechanisms of cell death mediation differ and results in differences in pathology and levels of variation in the two models.

Alloxan-Induced Model of Diabetes Alloxan is a pyrimidine derivative that inhibits glucokinase to reduce β-cell sensing of glucose and induces reactive oxygen species formation leading to cell necrosis196. The use of alloxan however is controversial due to inter-individual variation, even within the same species,197 in recovery from hyperglycaemia due to reversal of inhibition of glucokinase by dithiols. A short half-life also means that it should be administered quickly by intravenous injection as it may be metabolised within minutes in plasma. After two months of hyperglycaemia, rat retinas have increased monocyte and granulocyte accumulation and capillary loss which is exacerbated with disease progression198. There is also an increase in TUNEL positive cells after 8 months and an increase in acellular capillaries after 18 months199. The study by Kowluru et al (2001)200 found pericyte ghosts, acellular capillaries, and thickened retinal capillary basement membrane after 12 months of alloxan-induced diabetic rats which did not exhibit signs of microaneurysm, microvascular abnormalities, haemorrhage or neovascularisation.

50

Streptozotocin-Induced Model of Diabetes STZ is a glucose analogue with a toxic nitrosurea group from the bacteria Streptomyces grisceus201. It is a DNA alkylator that enters the cell via the physiological glucose pathway and produces insulin deficits over time by inducing necrosis of pancreatic β cells196; however, it is not representative of the T-cell mediated autoimmune response that occurs in T1DM. Upon the onset of diabetes, increases in iBRB permeability can be seen from 2 weeks of diabetes85,202. Also observed within the first month of hyperglycaemia is increased neural apoptosis203, decreased retinal blood flow204 and increased leukostasis and ICAM mRNA expression205. There is also an increase in inflammatory mediators from 3 months73,135 as well as increased AGE immunoreactivity95 and increased vascular permeability206. Histological traits such as increased acellular capillaries and decreased pericytes and endothelial cells can be seen at 32 weeks97 as well as increased AGE and RAGE staining111. The main argument against the use of the STZ model is that it is toxic to both liver and kidneys at high doses207. However, STZ does not directly induce toxicity in either the developing or mature retina208,209. Another argument against the STZ model is that it is only representative of T1DM and is not representative of natural diabetes induction, however in terms of retinopathy, given that in both T1DM and T2DM the complication arises from lack of effective insulin, it may be acceptable for use depending on the study to be done.

1.7.2 Spontaneous Models of Diabetes Spontaneous models of diabetes also exist and are more so representative of T2DM. These include Zucker diabetic fatty rats, WBN/Kob rats, Goto-Kakizaki rats and spontaneously diabetic Torii rats and their use may depend on the traits being investigated. However, the BB rat model is representative of T1DM and contains the immunological component of the clinical disease. Whilst STZ can induce diabetes within 5-6 days in mice210, spontaneous models take longer and are more variable in the time to develop diabetes and the time course to onset and progression of complications. Zucker diabetic fatty rats may take 6-7 weeks to become hyperglycaemic whereas WBN/Kob rats may take 9 months to develop hyperglycaemia. As retinopathy occurs with chronic hyperglycaemia, these models may require long-term care for pathological retinopathy to occur. Transgenic mouse models that spontaneously become hyperglycaemic may also be used such as db/db mice which represent T2DM due to deficiencies in the leptin receptor, and the Ins2Akita mouse model which represents T1DM due to a point mutation in the insulin-2 gene. These are reviewed in detail by Robinson et al 2012211. Whilst transgenic models may represent the histological traits, they are not representative of natural disease progression.

1.7.3 Proliferative Diabetic Retinopathy Models The previously mentioned models are not representative of proliferative diabetic retinopathy; therefore, models have been created for the study of this particular stage of disease. The oxygen-induced retinopathy model is currently the most used to date of preretinal neovascularisation (the growth of new blood vessels into the vitreous). This involves exposing a

51

rodent pup to hyperoxia, generally 75% oxygen. Mice are exposed between post-natal day 7 and 12 and rats can be exposed from post-natal day 5 to day 10212. This mainly affects capillaries adjacent to arterioles, which may die from oxygen toxicity. The retina subsequently becomes hypoxic upon exposure to normal oxygen levels due to hypoxaemic capillary loss. This leads to the formation of tortuous arteries, similar to those described in human eyes by Ashton (1953)75, as well as neovascularisation into the vitreous213. Another proliferative diabetic retinopathy model is the Akimba mouse, which overexpresses a human form of VEGF that induces retinal haemorrhaging and neovascularisation214. The galactosemia model is useful to observe neovascular changes in dogs but not rats. This is thought to be due to the shorter life span of rats that is not long enough to allow proliferative retinopathy to develop. Galactosemic rats develop acellular capillaries and pericyte ghosts after 15 months215. However these models do not account for the metabolic deficits and so are not suitable for etiology studies211.

52

1.8 Summary and Aims The predicted increase in diabetes diagnoses globally will increase the prevalence of diabetic retinopathy, the treatments for which only target late-stage disease by attenuating further progression. There are currently no treatments for the visual deficits that occur in early diabetic retinopathy, only prevention by good blood glucose and blood pressure control170,29. There is a clear unmet need to prevent the development of diabetic retinopathy both from an economic perspective and from a patient quality of life perspective. Endothelial cell death and pericyte loss are well established pathogenic markers of diabetic retinopathy that leads to hypoxia, neuronal loss and vision loss71,75,216. Whilst it has been established that the polyol pathway is changed in diabetic retina leading to the production of AGEs, there are few studies that have investigated the overall metabolic defects observed in diabetic retinopathy concomitantly in the tissue. Furthermore, despite the established role of oxidative stress in the pathogenesis of diabetic retinopathy, diabetes-induced changes to redox reactive metals have not yet been explored. By improving our understanding of the underlying changes to metal and metabolic regulation, we may provide new insights into the progression of diabetic retinopathy and potentially provide new therapeutic targets to attenuate endothelial cell loss and disease progression. In this thesis we aimed to investigate the following:

 To establish that the time points of the STZ model used in this study had a pathology that corresponded to that observed in other studies in the literature by assessing multiple markers of early diabetic retinopathy by reverse transcription polymerase chain reaction (RT-qPCR)  To investigate whether dysregulation of trace metals occurs in the diabetic retina by using a non-targeted approach to analyse detectable metals in diabetic and non- diabetic donor human retinas and the STZ rat model  To perform untargeted metabolomics to investigate changes to sugars and amino acids in the STZ rat retina  To perform untargeted metabolomics to map the main lipid classes in the rat retina and discover trends in lipid class changes that occur in diabetic retinas

53

Chapter 2 | Overview and validation of the in vivo Experiments used in this Study

54

2.1 Introduction The incidence of diabetic retinopathy is expected to increase with the rising prevalence of diabetes globally217. Despite this observation, there have been few novel advances in the field and early diabetic retinopathy is still untreatable. Basic research is still required to elucidate novel changes in early diabetic retinopathy to produce new targets for treatment.

This study aimed to use an untargeted approach to look for novel therapeutic targets in diabetic retinopathy by investigating pathways that have been under-studied in diabetic retinopathy. In particular, changes in the levels of physiological metals, and low molecular-weight metabolites including sugars, amino acids, lipids and other miscellaneous metabolites were measured in cases and matched controls. Untargeted approaches are advantageous because they provide large amounts of data from a single sample. By doing so, changes to a range of pathways can be monitored simultaneously. However, untargeted methods are limited in that they are not all-encompassing; components of the analysis that do not conform to the method of detection are missed in the analysis, thereby leading to missing data that may be potentially relevant. Another limitation is that substrates of higher abundance in the tissue sample will dominate the analysis and so substrate that are less abundant but may be significantly different between groups may be missed. In addition, targeted transcriptomic experiments were undertaken, and some of these are described in this chapter. The complexity of the retinal structure, with numerous neuronal cell types, glia and vascular cells, with the added complexity of some immune privilege and the inner blood-retinal barrier, makes in vitro analysis of the retina in diabetes challenging. Therefore, to establish the effect of reduced insulin signalling on the retina as a whole, in vivo studies were undertaken.

The streptozotocin (STZ)-treated diabetic rat model was chosen after a literature search revealed that it is one of the most commonly used in vivo model for the study of diabetic retinopathy, alongside the practical implications of obtaining larger retinas from a rat model rather than a mouse model. This is especially important when performing untargeted analyses of low-abundance substrates, where the rat retina is more suitable, due to its larger comparative size. Disadvantages of the STZ model include liver and kidney toxicity207 and that it does not produce diabetes naturally; however, STZ does not directly cause retinal toxicity other than by hyperglycaemia208,209.

Studies have shown that visual function becomes impaired by 12 weeks of STZ-induced diabetes in the rat. Deterioration of visual acuity commences 3 weeks post-STZ induction of diabetes. There is a reduction in contrast sensitivity by ~9 weeks and this coincides with the onset of cataract formation216. Electroretinogram analyses have shown that rod dysfunction occurs from ~4 weeks of STZ-induced diabetes216. Another study demonstrated changes in rod- mediated signs in electroretinograms at 7 weeks post-diabetes onset and similar changes were observed in diabetic human subjects without overt diabetic retinopathy218.

55

Capillary degeneration occurs from approximately 6 months after induction of STZ- diabetes84,95,97. The above studies suggest that retinal changes occur prior to demonstrable degenerative changes in the vasculature Therefore, in this thesis, as the rat retina was studied after 12 and 16 weeks of confirmed hyperglycaemia, early changes were analysed that occurred prior to the retinal microvascular changes of diabetes. The objective of work described in this chapter was to provide an overview of the different in vivo experiments used in this thesis and to validate these experiments by assessing retinas from both time-points for the presence of established biomarkers, thereby allowing comparison of novel data with that in the literature.

2.2 Methods

2.2.1 In vivo Methods

Animal Husbandry All animal experiments were conducted in accordance with the UK Home Office regulations for the care and use of laboratory animals, the UK Animals (Scientific Procedures) Act (1986), and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Rats were kept in a 12:12 h light:dark cycle and had access to standard laboratory chow and water ad libitum. At the end of the trial rats were terminally anaesthetised with isoflurane followed by decapitation to confirm death.

Induction of Diabetes Rats were randomly allocated into non-diabetic and diabetic groups and were fasted overnight. Thereafter, an intraperitoneal injection of 55 mg/kg STZ (Sigma-Aldrich S0130) in 0.9% w/v saline was administered. In Study 2, non-diabetic rats were administered an intraperitoneal injection of 0.9% w/v saline. A tail prick was done three days post-STZ injection to check blood glucose using an Accu-chek® Aviva Blood Glucose Meter System (Accu-chek®, Roche) with Accu-chek® Aviva Blood Glucose Test Strips (Accu-chek®). Hyperglycaemia was defined as a blood-glucose value greater than 15 mmol/L. At the end of the trial blood glucose was measured. Whole blood was diluted 1:2 in heparin-saline (1:20 dilution), so that it could be measured with the Accu-chek® Aviva Blood Glucose Meter System, which has a maximum detectable concentration of 33 mmol/L.

Study 1 Non-diabetic (N=10) and untreated-diabetic (N=18) rats were maintained for 12 weeks. Rats were supplemented with half an insulin implant (LinShin, Scarborough, Canada) inserted subcutaneously with a release rate of 2U/24 hours/implant if their health deteriorated by pre- defined measures: a loss of ≥20% of starting body weight; signs of sickness behaviour such as grimacing or lack of grooming; signs of unusual levels of lethargy and lack of awareness when handled without any sign of improvement. The retinas from this study were used in RT-PCR experiments to characterise the model and to assess changes in copper transporters gene expression (described in Chapter 3).

56

Study 2 Non-diabetic (N=12), untreated diabetic (N=13) and triethylenetetramine (TETA)-treated diabetic (N=9) were maintained for 16 weeks. Diabetic animals were divided equally, according to weight and weight change post-STZ injection, into untreated and TETA-treated groups. TETA treatment commenced one-day post-confirmation of hyperglycaemia. The TETA group were administered 86 µg/ml of triethylenetetramine disuccinate based on average water consumption of 350 ml/day, which equated to a target dose of 30 mg/day as described in a previous study 219. Water consumption was measured weekly and the individual dose titrated in accordance to this. After confirmation of diabetes, rats were housed in pairs separated by a cage-divider (Techniplast, Kettering, UK). This allowed individual access to water bottles without full social isolation, thereby reducing stress compared with conventional single housing. Untreated and drug-treated diabetic rat water consumption was measured weekly. At 10 weeks’ post disease onset, all diabetic rats received maintenance insulin via half an insulin implant inserted subcutaneously with a release rate of 2 U/24 hours/implant regardless of their condition to prevent any deterioration during the final six weeks of the study. The retinas from this study were used in RT-qPCR characterisation at the 16-week time-point, as discussed later in this chapter, for assessing changes to copper transporter gene expression (described in Chapter 3).

Study 3 Rats were allocated into one of four groups; non-diabetic (N=12); untreated-diabetic (N=9); diabetic TETA-prevention (Tp) treated (N=9); and diabetic TETA-intervention (Ti) treated (N=9). All groups were maintained until 16 weeks after disease onset. TETA was administered as described in Study 2. In the Tp group, TETA treatment commenced one-day post-confirmation of hyperglycaemia. TETA was administered as an intervention treatment to the Ti group, 9 weeks after confirmation of hyperglycaemia. Rats were doubly housed and separated by cage dividers as in Study 2. Water consumption of untreated and drug-treated diabetic rats was measured weekly. Rats were supplemented with half a subcutaneous insulin implant only if their condition required insulin maintenance. The retinas from this study were used for trace metal analysis (described in Chapter 3) and analysis of polar metabolites (Chapter 4).

Study 4 Non-diabetic (N=12) and untreated-diabetic (N=10) rats were maintained for 12 weeks from disease onset. Diabetic rats received a subcutaneous half implant of insulin as described in Study 1 if their health deteriorated to predefined levels. These rats were doubly caged and not separated by cage dividers. This study was used to replicate the results of the effects of hyperglycaemia on polar metabolites (Chapter 4), and to characterise retinal lipid changes in diabetic retinas (Chapter 5).

Study 5 Non-diabetic (N=10) and untreated diabetic (N=7) rats were maintained for 16 weeks. Diabetic rats were implanted with insulin pellets at week 10 as described in Study 2. These rats were

57

doubly housed and not separated by cage dividers, and retinas were used to repeat the 16- week measure of polar metabolites (Chapter 4).

2.2.2 Tissue Extraction and Analysis

Retinal Dissection Eyes were enucleated using blunt curved forceps and then washed in phosphate-buffered saline. They were then dissected on a petri dish in PBS. Titanium instruments were used to prevent any contamination of the tissue from metals that were to be analysed. An incision was made on the edge of the cornea. This was achieved by holding the tissue with fine forceps visualise the isolation of the retinas and The tissue was held using fine forceps (1401021W SD Healthcare, Irlam, UK) and an incision was made using vannas fine scissors (1403001 SD Healthcare, Irlam, UK). The tissue was cut along the perimeter of the cornea until it was dissected free and removed. The lens was then carefully extracted intact. Tools were washed in PBS after this point to remove lens contaminants. The retina was gently peeled from the sclera and the optic nerve was then cut allowing removal of the retina, which was then washed in PBS. All dissected retinas were frozen over dry ice and stored at -80° C until processing, with the exception of those used for RNA analysis. These were incubated at 4°C overnight in RNAlater™ (Sigma-Aldrich R0901-500ML) after which they were stored at -80°C. Owing to the quick processing required to prevent the breakdown of substrates to be analysed, the retina was removed and quickly frozen; therefore, contamination with the RPE was likely.

RNA Extraction Retinas were homogenised in 500 μL of TRI Reagent (Applied Biosystems AM9738) using a Qiagen Tissuelyser II (Qiagen, Manchester, UK) for two 4-minute runs at 25 Hz, after which they were incubated at room temperature for 5 minutes. 50 μL of 1-bromo-3-chloropropane (BCP) (Sigma-Aldrich B9673-200ML) was added to each sample, and these were then shaken vigorously by hand and incubated at room temperature for 10 minutes. After this, samples were centrifuged at 16,800 x g at 4°C for 10 minutes. The resulting aqueous phase was then transferred to a fresh tube with care being taken not to disturb the DNA-containing interphase. An equal volume of 70% (v/v) ethanol was added to the aqueous phase, immediately mixed and the solution was then loaded onto a Qiagen Mini column (Qiagen RNeasy mini kit 74104). At this point, the Qiagen RNeasy mini kit protocol was followed, according to manufacturer’s instructions. The yield of RNA obtained was determined using a Nanodrop 2000 spectrophotometer (Thermo Scientific, Runcorn, UK), and samples were stored at -80°C until analysis. cDNA Synthesis cDNA synthesis was done from the extracted RNA using a Transcriptor High Fidelity cDNA Synthesis Kit v6.0 (Roche Diagnostics Limited 05091284001). Using the RNA yields measured using the Nanodrop spectrophotometer, 1 μg of RNA was added to the appropriate amount of Molecular-grade water (BP2819-1 Fischer Scientific, Loughborough, UK). 1 μL of Anchored-

58

oligo(dT)18 Primer was added to this to make a total volume of 11.4 μL. The samples were incubated at 65°C for 10 minutes in a thermal block cycler with a heated lid and immediately cooled on ice. Samples were briefly centrifuged before further use. The reverse-transcriptase mix was made according to the kit protocol and mixed into the RNA solution by pipetting. The samples were briefly centrifuged and incubated at 50°C in a thermal block cycler with a heated lid for 30 minutes. The reaction was inactivated by heating to 85°C for 5 minutes. The samples were then cooled on ice and stored at -20°C. This final concentration of cDNA produced was ~50 ng/μL. Some of this was diluted to a concentration of 5 ng/μL and stored in aliquots. A standard curve was done for two separate pools of non-diabetic and diabetic samples to assess the quality of cDNA.

RT-qPCR 3.2 μL of 5 ng/μL cDNA was added to a Master Mix solution in a 96-well plate containing LightCycler® 480 SYBR Green I Master (04887352001, Roche, Burgess Hill, UK), the forward and reverse primers (each at 500nM) and PCR-grade water, to give a final volume of 43.4µL/well. Primer sequences were designed and checked for specificity using the National Center for Biotechnology Information Basic Local Alignment Search Tool (NCBI BLAST). The criteria for primer design are shown in Table 2.1 and all primers that were used in this thesis are described in Appendix I. No-template controls containing nuclease-free water instead of a cDNA sample were included in triplicate for each assay as a negative control. The 96-well plate was sealed with a non-heat-resistant seal and spun in a centrifuge at 872 x g for 1 minute. Using an electronic pipette, the cDNA was transferred to a 384 -well plate (Roche) in triplicate. The plate was sealed using a non-heat-resistant seal and centrifuged at 872 x g for 2 minutes. This seal was removed and after ensuring no air bubbles were present, a heat-resistant seal was used to tightly seal the plate. Amplification was carried out on a Roche Light Cycler® 480 (LC480) using the following thermal profile: The enzyme was activated by a 5-minute hot-start period at 95 °C; amplification of the template DNA was then carried out through 50 cycles of denaturation (95 °C for 10 seconds) and primer annealing/extension (60 °C for 30 seconds). A fluorescence reading (465–510 nm) was taken at the end of each cycle to quantify the reaction in real time. Changes to gene expression in diabetic compared with non-diabetic were quantified relative to reference genes. The reference genes used were: Tbp, the gene for TATA-Box Binding Protein; Ndc1, nucleoporin 1; and Actb, actin B. These reference genes were used because they remained unchanged in diabetic retinas relative to non-diabetic controls and produced products over a range of cycles. The mean threshold obtained for the reference genes was used to normalise the genes of interest. Each experiment was repeated three times.

59

Primer design criteria

Length: 18–24 bases

Guanine-cytosine content: 40–60%

Melting temperature (Tm): >60ºC ± 3ºC

Product length: 80–200 base pairs ideally <250 base pairs

Avoid adenine and thymine in the first and last position

Must span exon-intron boundaries where possible

Ensure that 5’ and 3’ primers are not complimentary to each other

Table 2.1 | Criteria for primer design These criteria were used to obtain the best melting temperature for the PCR method used. A balance of guanine-cytosine and adenine-thymine was required as guanine and cytosine have higher melting points than adenine and thymine. The presence of adenine or thymine in the first or last position was avoided to prevent the primer ends from binding and forming a loop. Primers were required to span an intron-exon boundary to prevent amplification of any residual contaminating DNA. 2.2.3 Statistical Analysis Statistical analyses were performed using GraphPad Prism 6 for Windows and are presented as mean ± 95% confidence interval (CI) of the mean in all cases. In experiments 1, 4 and 5, body weight and blood-glucose measurements were analysed by an unpaired t-test. In experiments 2 and 3, these parameters were assessed by one-way ANOVA because hyperglycaemia needed to be determined in these animals alongside an effect of TETA on body weight and blood- glucose values. RT-qPCR analyses were assessed by t-test with Welch’s correction to determine differences between non-diabetic and diabetic rats. This assumes that both groups of data are sampled from Gaussian populations, but does not assume those two populations have the same standard deviation.

60

2.3 Results

2.3.1 Characteristics of in vivo Studies The results for the weights and blood glucose in all the in vivo studies are summarised in Table 2.2. Overall, diabetic rats failed to gain as much weight as non-diabetic rats and in all studies, the difference in mean final body-weight values between groups was statistically significant (p<0.0001). In Study 2, the starting weight of the diabetic group was significantly higher than non-diabetic (p<0.05); however, this did not affect the outcome, and the untreated-diabetic group gained less weight than the non-diabetic group. In all studies, mean blood glucose was significantly elevated in diabetic compared with non-diabetic groups (p<0.0001). TETA was administered in Studies 2 and 3, either as Tp or Ti. In both studies, TETA treatment did not affect final weight or blood- glucose values (p>0.05 in all experiments).

2.3.2 Changes in the mRNA expression of Molecular Biomarkers in 12-Week Diabetic Retinas Markers that have previously been shown to display altered mRNA expression after 3-months of STZ-induced diabetic retinopathy220 were measured in retinas from Study 1 (12 week study) by RT-qPCR including: Chi3l1, Gat3, Gbp2, Hspb1, Jak3, Lgals3, Nppa, Timp and C1inh. All markers were significantly changed when analysed by t-test with Welch’s correction in diabetic rat retinas compared with non-diabetic as shown in Figure 2.1. All biomarkers were upregulated in diabetic retinas compared with non-diabetic (p<0.0001) with the exceptions of Gat3 and Nppa, which were both downregulated in diabetes (p=0.011 and p=0.01 respectively), consistent with the report of Freeman et al (2010)220.

2.3.3 Changes in Retinal Biomarker mRNA Expression after 16 Weeks of Diabetes This analysis was repeated in retinas from Study 2 where diabetes was maintained for 16 weeks. The same markers were assessed by RT-qPCR with the addition of Txnip and Vegfa (Figure 2.2). In this study, the mRNA expression of all biomarkers was changed in diabetes, except that of Gat3 and Nppa. The biomarkers, Chi3l1, Gbp2, Hspb1, Jak3, Lgals3, Timp and C1inh were all significantly increased in (untreated) diabetic retinas compared with non-diabetic (p<0.0001 for all). mRNA expression of Txnip was significantly increased in diabetic retinas (p<0.03) and expression of Vegfa was significantly lower in diabetic retinas (p=0.005).

61

Non-diabetic Diabetic Tp Ti

Study 1 Start weight (g) 391.5 ± 9.3 390.4 ± 6.9 – –

Final weight (g) 608.9 ± 14.1 396.6 ± 14.8*** – –

Blood glucose 7.4 ± 0.3 33.9 ± 1.3*** – – (mmol/L)

Study 2 Start weight (g) 344.1 ± 3.0 362.4 ± 4.8* 359.30 ± 6.4 –

Final weight (g) 604.7 ± 12.5 432.5 ± 14.0*** 427.4 ± 12.5*** –

Blood glucose 9.9 ± 0.5 29.2 ± 1.3*** 28.3 ± 1.2*** – (mmol/L)

Study 3 Start weight (g) 364.6 ± 14.1 376.8 ± 19.0 381.8 ± 21.4 374.3 ± 53.6

Final weight (g) 582.0 ± 39.9 397.1 ±27.9*** 408.9 ± 33.0*** 376.1 ± 62.3***

Blood glucose 9.1 ± 1.4 33.8 ± 5.0*** 31.6 ± 4.0*** 32.3 ± 5.3*** (mmol/L)

Study 4 Start weight (g) 332.7 ± 14.4 338.1 ± 14.1 – –

Final weight (g) 556.8 ± 33.0 391.6 ± 40.9*** – –

Blood glucose 10.1 ± 1.7 30.9 ± 4.7*** – – (mmol/L)

Study 5 Start weight (g) 341.1 ± 40.16 354.4 ± 22.94 – –

Final weight (g) 606.0 ± 69.69 421.9 ± 40.65*** – –

Blood glucose 6.4 ± 0.9 30.6 ± 5.3*** – – (mmol/L) Table 2.2 | Summarisation of starting weights, final weights and final blood-glucose values for all STZ experiments An overview is provided of the five STZ experiments from which retinal tissue was obtained and used in studies described in the following chapters. Mean values and standard deviation (SD) are presented for body weight and blood-glucose measurements. Maintenance insulin (2 U/24 hours) was provided for Studies 3 and 5. Diabetic rats were compared with non-diabetic using an unpaired two-tailed t-test with the exception of experiments 2 and 3 where an ANOVA analysis followed by Tukey’s multiple comparisons test with adjusted p-values was done do to multiple groups for comparison.* where p<0.05 and *** where p<0.0001.

62

A

B

Figure 2.1 | Transcriptomic changes induced after 12 weeks of STZ-induced diabetes (Study 1) The gene expression of numerous biomarkers, relative to housekeeping genes Tbp, Ndc1 and Actb, was compared by RT-qPCR in diabetic (N=18) and non-diabetic (N=10) rat retinas from Study 1. All the biomarkers assessed were significantly changed in diabetic relative to non- diabetic rat retinas after 12 weeks of diabetes when assessed by t-test with Welch’s correction. Data are presented as mean (95% confidence interval) where *p<0.05, ***p<0.001, ****p<0.0001.

63

A

B

Figure 2.2 | Changes to biomarkers after 16 weeks of STZ-induced diabetes as assessed by RT-qPCR (Study 2) Gene expression of biomarkers relative to housekeeping genes Tbp, Ndc1 and Actb, was assessed by RT-qPCR in diabetic (N=13) and non-diabetic (N=12) rat retinas from Study 2. All the biomarkers assessed were significantly changed in retinas with the exceptions of Gat3 and Nppa after 16 weeks of STZ-induced diabetes compared with non-diabetic retinas when assessed by t-test with Welch’s correction. Data are presented as mean (95% confidence interval) where *p<0.05, ***p<0.001, ****p<0.0001.

64

2.4 Discussion In this chapter, an overview and initial analyses was provided of the animal experiments that will be referred to in more detail in subsequent chapters. In all experiments, diabetic rats were maintained for 12 or 16 weeks. These time-points were validated in our studies by assessing the presence of changes in levels of biomarkers that have been described previously in the literature, as discussed below.

2.4.1 Altered mRNA expression of Established Biomarkers Confirmed the Molecular Pathology after 12 and 16 weeks of STZ-induced Diabetes The two time-points were validated by detection of established biomarkers of early diabetic retinopathy in the STZ rat, as previously reported by Freeman et al (2010)220. These markers were chosen because in this study, a five-step validation process was performed to ensure the specificity of the reported biomarkers across eight different STZ experiments of 1 and 3 month duration. The reproducibility shown within the Freeman et al (2010) study suggested that these markers were reliably changed in several separate animal studies and indicated that the retinas in the presented study followed a similar profile of those previously published. Pathology was defined in Freeman et al (2010) at 3 months by an increase in permeability and caspase-3 activity100,206,220. Whilst using only these markers to validate that these animals had early diabetic retinopathy is limiting, utilising the markers limited the amount of extra animals that would have been required to complete the Evans blue assay for assessing permeability and TUNEL staining to analyse apoptosis that have been completed previously202,221,222. In the current study, significant changes in the mRNA of all biomarkers tested were found with the exception of Gat3 and Nppa in the 16-week Study 2. Chi3l1, Gbp2, Hspb1, Jak3, Lgals3, Timp and C1inh were all significantly increased in diabetic retinas compared with non-diabetic at both 12 and 16 weeks.

Chi3l1 is an inflammatory marker that is associated with endothelial dysfunction, and plays roles including promotion of chemotaxis, cell attachment and migration of vascular smooth muscle cells. Moreover, plasma levels of Chi3l1 correlate with insulin resistance in diabetic patients223.

Little is known about the GBP2 protein or the guanylate-binding protein family (GBP) except that they are membrane-bound GTPases that are induced by interferon-γ in cultured endothelial cells 224,225. Interferon-γ is reportedly increased in retinas of STZ-diabetic rats226; therefore it is likely that this could initiate the enhanced Gbp2 mRNA levels observed in the current thesis.

Hspb1 is a member of the heat shock protein family which are induced by thermal, oxidative, hemodynamic, osmotic and hypoxic stresses227. The HSP25 protein is reportedly upregulated in diabetic retinas after 10 weeks of STZ-induced diabetes and is overexpressed in the ganglion cell layer alongside markers of oxidative stress and apoptosis228.

65

Jak3 (janus kinase 3) expression is induced by cytokines and leads to downstream STAT signalling 229. JAK/STAT signalling is also induced by hypoxia in microvascular endothelial cells and may play a role in retinal angiogenesis and platelet aggregation230,231 .

Lgals3 (lectin, galactosidase-binding, soluble, 3), also known as Gal3 (galectin 3), is an AGE- receptor that activates VEGFR2 and prevents its internalisation232. Knockout of galectin 3 in C57/BL6 mice reportedly prevents inner blood-retinal barrier breakdown and diabetes-induced increases in VEGF after 3 weeks of STZ-induced diabetes and improves the angiogenic response in the OIR-mouse model of proliferative retinopathy221. Gal3 is usually found in microglia located in the ganglion cell layer but upon induction of inflammation with lipopolysaccharide, is upregulated in microglia in the ganglion cell layer, inner nuclear layer and outer plexiform layer233.

TIMP-1, the corresponding protein of Timp1 is a tissue inhibitor of metalloproteinases that inhibits matrix metalloproteinases (MMPs), with some selectively for MMP-9234, produced by astrocytes and the RPE235,236. Alteration of the TIMP-1/MMP ratio modulates retinal neovascularisation and increased expression of TIMP-1 increases VEGF-induced angiogenesis in retina237.

C1inh (C1 esterase inhibitor) encodes a protein that inhibits the complement cascade, which has been found to be activated in diabetic human donor retinas and the vitreous of living diabetic patients with proliferative retinopathy238,239. C1INH is found in choriocapillaries, photoreceptor cell bodies and axons, throughout the inner nuclear layer, and in the lumen of large vessels240. Whilst intravitreal administration of C1INH protein reduces retinal vascular permeability in the STZ-rat after 2 weeks of diabetes, it is likely that it was increased in this present study and that by Freeman et al., (2010)220 in response to activation of the complement cascade; increases in C1INH may be beneficial in that it inactivates the complement cascade.

2.4.2 Gat3 and Nppa were changed in rat retinas after 12 weeks but not 16 weeks of STZ-induced diabetes The mRNA expression of Gat3 and Nppa were significantly decreased after 12 weeks of diabetes in Study 2, but not after 16-weeks of diabetes (Study 3). However, in Study 3 there was a non-significant trend towards decrease in expression levels in the diabetic rats. Freeman et al (2010)220 reported that these were decreased in their 12 week experiments. Gat3 encodes a GABA transporter and is expressed in the inner plexiform layer on retinal Mueller cells and is thought to be responsible for neuronal and glial uptake of GABA241. Bipolar cells in rat retinas have increased sensitivity to GABA after ~8 weeks of STZ-induced diabetes242. Together, these changes indicate dysregulation of GABA signalling in diabetic retinas. Nppa encodes natriuretic peptide precursor type A, which has a role in counteracting VEGF-induced angiogenesis and permeability100. The NPPA protein is localised to the ganglion cell and inner nuclear layers, and has been shown to prevent N-methyl-D-aspartate (NMDA)-induced dopamine reduction in rat retinas, consistent with a role in neuroprotection.

66

2.4.3 Additional biomarkers Txnip and Vegfa were also changed in rat retinas after 16 weeks of STZ-induced diabetes Here, levels of Txnip, the gene encoding thioredoxin-interacting protein and Vegfa were assessed only in the 16-week study. Txnip was chosen in light of the study by Bixler et al (2011) where its mRNA expression was increased in the rat retina after 3 months of STZ-induced diabetes but was not normalised with insulin treatment87. TXNIP has a role modulating angiogenesis by reducing VEGF production and function, independent of thioredoxin243. For these reasons, along with its role in oxidative-stress modulation, it was assessed for an effect of TETA-treatment in the next chapter. Upregulation of TXNIP is not specific to cell type and in diabetic retinopathy it is associated with endothelial cell dysfunction, pericyte loss, neuronal cell death and glial activation53. Vegfa mRNA expression was also studied due to its well- established role in diabetic retinopathy. In this study, Vegfa mRNA expression was decreased after 16-weeks’ diabetes, consistent with the report by Brucklacher et al (2008) where Vegfa mRNA expression was decreased in the rat retina after 3 months’ STZ diabetes100. Whilst other studies have shown an increase in retinal Vegfa mRNA expression and protein after 2 weeks and 6 months’ STZ-diabetes respectively 85,244, the study by Brucklacher et al (2008) indicated a decrease in Vegfa mRNA expression in multiple animal experiments at a time-point closer to that used in this study100. VEGF expression is increased in glial cells after six months of STZ diabetes245. It is possible that VEGF is subject to numerous changes throughout diabetic retinopathy and may be enhanced in more progressed disease.

2.4.4 Overall Conclusions In this chapter, I have provided an overview of the animal experiments, which will be described in detail in the following chapters. The aim of this chapter was to provide a foundation of information to establish that the model used is representative of that employed in previously- published studies. All of the experiments in diabetic rats consistently led to failure to gain weight in diabetic animals and higher circulating blood glucose. Data from 12- and 16-week experiments largely agreed with molecular information already published in the literature thereby validating the model and adding weight to the data presented in the rest of this thesis. The biomarkers described here cover a limited range of inflammatory pathways including the complement pathway, cytokine signalling, and oxidative stress, but would benefit further from studying more markers of oxidative stress such as 4-hydroxynonenal. It would also be of benefit to study some of these inflammatory pathways in detail and identify the cells in which these changes occur to better understand the changes that occur in diabetic retinopathy. Functional assays such as electroretinograms would also be valuable in characterising the pathology of diabetic retinopathy at this time point. Whilst there were subtle differences between experiments such as the use of maintenance insulin and cage dividers, the biomarker panel was tested in studies that captured these differences and overall similar results were found. This indicates that these studies provide a reproducible level of retinopathy. This has also provided a useful panel of inflammatory mediators that can be used to assess the effects of investigative

67

therapies. In the next chapter, the use of a novel therapeutic approach to target early diabetic retinopathy is investigated, in part by examining whether these biomarkers are normalised by the intervention.

68

Chapter 3 | Characterising Trace Metals in Human and Rat Retinas and the Effects of Diabetes

69

3.1 Introduction

Whilst a role for metal dysregulation in diabetic complications has been investigated in recent years, this remains largely unexplored in the retina. Changes to retinal iron, zinc and copper have been implicated in age-related macular degeneration and aging246 and several reviews suggest a role for different metal alterations in diabetic complications247–250.

Metals can be detected by multiple mechanisms including inductively coupled plasma mass spectrometry (ICP-MS), synchrotron X-ray fluorescence (SXRF) and particle-induced X-ray emission (PIXE). SXRF and PIXE can localise metals within a tissue section by emitting a beam and detecting the x-ray emission from the irradiated element. Compared with ICP-MS, these methods are time consuming and do not detect as large a range of metals as the former and have lower sensitivity251. For this reason, ICP-MS was used to screen an array of metals for an effect of diabetes in the retina.

For elemental analysis by ICP-MS, the biological material in a sample is degraded to leave only the total elemental composition. During the process of ICP-MS, the atoms present are ionised by a high temperature plasma discharge to form positively charged ions. The ability to measure an element depends on its ionisation potential, which is the energy required to remove an electron from the neutral atom. The lower the ionisation potential the more easily detected the element will be. Another factor is the ionisation efficiency, or the proportion of ions formed to the number of electrons used. Therefore, the most easily detected elements are those with a low ionisation energy and a high ionisation efficiency such as sodium, potassium and calcium252 whilst others with a high ionisation potential such as fluorine cannot be detected using this method.

The plasma discharge is created from ionised argon. A radio-frequency generator applies radio- frequency power to a load coil, which surrounds the torch (Figure 3.1), creating a current within the coil. This creates an electromagnetic field at the top of the torch253. Argon gas flows through the torch and a high-voltage spark is applied causing ionisation of argon atoms. These ions collide with inert argon atoms to form a chain reaction and create an inductively coupled plasma discharge. The continued electromagnetic field sustains the discharge and the sample is introduced into this plasma as an aerosol by a sample injector254. Through the different zones of the plasma torch the sample is dried, vaporised, atomised, and ionised, at the end of which it is transformed into a gas254. Upon analysis, the sample exists as excited atoms and ions, representing the elemental composition of the sample.

The development of an Octopole Reaction System has helped to increase collision frequency and reduce interferences. It is located before the analysing quadrupole and is infused with a collision gas, such as helium. A radio-frequency field focuses the ions into the collision gas to improve collision efficiency and breaks up polyatomic ions. The helium itself is inert and so does not produce any extra interference. Collisions between helium atoms and large interfering ions

70

Figure 3.1 | Diagram of the Agilent 7700 ICP-MS Upon injection, the sample enters the spray chamber as an aerosol through the nebuliser. Large particulate matter adheres to the side of the spray chamber whereas small particles are introduced to the torch by a vacuum. Within the torch, samples are dried, vaporised, atomised and ionised. After passing through the sampling cone, focusing lenses direct the ions to the Octopole Reaction System (ORS) that in this case acts as a collision cell, slowing down polyatomic ions by collisions with helium atoms. From here, the ions move to the quadrupole. From the quadrupole, a single mass is released to the detector. Changes to the voltage of the quadrupole define the mass that hits the detector. (Adapted from: Agilent Technologies 2012252

71

will slow down the latter compared with smaller analytes allowing for specific analyte discrimination due to energy discrimination, whereby slower ions are ignored252,255. This development improves the sensitivity of detection by reducing the signal-to-noise ratio.

Metals have important physiological roles in retinal function. Sodium, potassium and calcium are essential for neuronal transmission and neurotransmitter release. Calcium also has a role in regulating melatonin synthesis256. Light damage and numerous ocular diseases, including diabetic retinopathy induce oxidative stress. Copper and zinc are essential for the efficient function of the SOD1 and SOD3, as is manganese for that of SOD2257. Both SOD1 and SOD2 protect against oxidative damage in the retina; therefore zinc, copper and manganese are essential for maintaining the protective function of the respective antioxidants in the retina258,259.

Iron, zinc and copper are of particular interest in retinal dysfunction. This is due in part to the redox activity of these metals and the potential to induce oxidative stress upon their dysregulated transport. X-ray fluorescence microscopy has shown that iron is highly localised to the outer limiting membrane and/or photoreceptor outer segments, whereas zinc is distributed throughout the outer limiting membrane, the outer plexiform layer and the inner nuclear layer and copper was localised to photoreceptor outer segments, and both the inner and outer plexiform layers251. Isomerisation of all-trans-retinyl ester to 11-cis-retinol in the retinal pigment epithelium (RPE) is essential for visual function and the iron containing enzyme RPE65 is the required for this process260,261. Iron overload causes thickening of the central RPE, thinning of the central ganglion cell layer262, increased expression of markers of oxidative stress in photoreceptors and thinning of the outer nuclear layer263. As well as being important for SOD1 activity, zinc is also essential for visual processing; specifically the conversion of 11-cis-retinol to 11-cis-retinal and recovering activity, an enzyme involved in rhodopsin cycling246. Copper has an essential role in mitochondrial energy production, the disruption of which can cause phototoxicity264 and modified rhodopsin function246. Decreased electroretinogram activity in Menkes disease patients and decreased melanin production, mitochondrial swelling and reduced cytochrome c oxidase (CcO) activity in a mouse model of Menkes disease have been demonstrated265,266; this highlights the effects upon retinal physiology of severe copper deficiency. Patients with Wilson’s disease, where there is retinal copper overload, have prolonged P100 latencies when visual evoked potentials are analysed and this is improved with copper chelation therapy267. This implies that copper overload can cause reduced visual function. The evidence for copper dysregulation in other diabetic tissues, suggests a potential change in diabetes248. The above-mentioned transition metals have essential roles in phototransduction and so dysregulated homeostasis of any of them may lead to visual dysfunction. Their redox activity may also lead to oxidative stress and AGE production, which is emphasised by the metal-chelating activity of AGE-inhibitors such as pyridoxamine, which correlates with therapeutic potency249. Therefore, the question of whether metals are dysregulated in diabetic retinas may be important in understanding the underlying pathology in retinopathy.

72

The Manchester Eye Tissue Repository (ETR) was established in 2014 with the goal of aiding ophthalmology research by providing a bank of numerous ocular tissues. Upon removal of the cornea for transplant, the rest of the eye is dissected and sample of tissues across the whole eye are taken and banked including the neurosensory retina, the vitreous, sclera and the optic nerve to name a few.

The objective of this study was to utilise and optimise ICP-MS to investigate the metal content in retinas and to assess whether levels of the metals analysed are modulated by diabetes. The first component of this study was a comparison of detectable metals in diabetic and case control human donor retinas from the Manchester ETR. Other factors that were investigated included the effects of gender, age, smoking history and medication on retinal metal content. These data from donor retinas were then compared with a commonly used experimental rodent model, the STZ rat model and the effect of oral administration of the copper chelator TETA were investigated in this model. In addition, the expression of genes that encode copper transporter binding proteins and biomarkers for the immune response in diabetes were investigated in this rat model.

73

3.2 Methods

3.2.1 Donor Retinas

Post-mortem donor human neurosensory retinas were obtained from the Manchester Eye Bank. In all cases, prior consent for use of the ocular tissue had been obtained by the NHS, and this research adhered to the Human Tissue Act (2004). Diabetic donors were ascertained from medical records. A record of diagnosed diabetes or use of anti-hyperglycaemic medication was used to select donor tissue from diabetic subjects. Age-matched non-diabetic cases were selected based on a lack of a diagnosis of diabetes or use of anti-diabetic medication. Whilst there is a prevalence of undiagnosed prediabetes of 16% in the UK population aged 50-69 years and 26% in the >70 year population268: for the purposes of this study all those donors who were not confirmed to have diabetes were considered to be non-diabetic. Neurosensory retinas were dissected from the eyecup and frozen as described in McHarg et al 2015269. They were catalogued and stored at -80°C at the Manchester Eye Tissue Repository (ETR) until required for use.

3.2.2 Dissection of Rat Retinas

For details on animal husbandry, induction of diabetes, TETA-treatment, rat retina dissection and RT-qPCR methodology, please refer to Chapter 2, Methods: 2.2.1 In vivo Methods and 2.2.2 Tissue Extraction and Analysis. Titanium instruments were used to prevent any contamination of the tissue from metals that were to be analysed. Rat retinas were processed similarly to human-donor retinas; however, due to the small sample size the whole rat retina was analysed.

3.2.3 Tissue Digestion

Trace Metal Grade concentrated nitric acid (A509 Trace Metal Grade; Fisher, Loughborough, UK) containing (5% v/v) Agilent Internal Standard mixture (5183-4681; Agilent Technologies, Cheadle, UK) was used in this study. This internal standard mixture contained 10 ppm each of

6 lithium , scandium, germanium, yttrium, indium, terbium and bismuth in 5% HNO3. This internal- standard-containing acid was also used at appropriate dilutions to provide rinse and calibration solutions, at 2% (v/v) final nitric acid. Calibration curves were obtained by using appropriate dilutions of Environmental Calibration Standard (Agilent 5183-4688). This gave final concentrations of 0, 50, 100, 200, 500, 1000, 2000, 5000 and 10000 μg/L for sodium (Na), magnesium (Mg), potassium (K), calcium (Ca) and iron (Fe) and for the rest of the elements analysed: 0, 0.5, 1, 2, 5, 10, 20, 50 and 100 μg/L.

Whole human retinas were cut into 50 mg samples, placed in tubes and then accurately weighed prior to drying in tubes for 16 hours in an Eppendorf Concentrator Plus. Whole rat retinas were weighed and dried in the same way for 2 hours. The tube lids were punctured to prevent pressure build up, and 200 μL of standard-containing nitric acid was then added. The

74

tubes were then inserted into a “Dri-block” heater that was initially at room temperature. Additionally, tubes of standard-containing acid without sample were processed in each batch to provide “digestion blanks”. The temperature was set to 60 °C and the block switched on. After 30 minutes, the temperature was increased to 100 °C, and digestion continued for a further 210 minutes. After digestion, the tubes were allowed to cool for 30 minutes and quickly spun in a centrifuge at 2400 x g for 15 seconds. 100 μL of each digestion solution was then added to a 15 ml Falcon tube (Greiner) containing 5 ml LC-MS grade water (Sigma-Aldrich, Gillingham, UK), to produce solutions for analysis at a final nitric acid concentration of 2% (v/v).

3.2.4 ICP-MS

Tissue-metal concentrations were measured using an Agilent 7700x ICP-MS spectrometer with a multi-element method including all elements present in the calibration solution. This multi- element method was used as per manufacturers’ instructions.

Multiple batches were run because of the high number of samples; as such there was an equal number of diabetic and non-diabetic samples run in each batch. For each analytical batch, multi-element calibration was performed using serial dilutions of the calibration standard. An intermediate concentration from this calibration series was used as a periodic quality control (QC) sample throughout each analytical batch. Instrument and digestion blanks were also interspersed through each set of randomised samples. Signals derived from non-essential elements were identified by comparison of blank, calibration and sample concentrations, and eliminated prior to reporting.

3.2.5 Statistical Analysis

Calibration curves and blanks were assessed to ensure accurate quantification and check for contamination. Data were corrected for sample weight, dilution and molecular weight. The mean of both retinas was calculated for each donor. The data expressed in units of mg/mol or μg/mol depending on the concentration of each relevant metal, were entered into GraphPad Prism 6 for Windows (GraphPad Software Inc.). The data were log transformed to equivalent log10 values and analysed using an unpaired t-test with Welch’s correction (allowing for potential heteroscedasticity). Welch’s correction was utilised on the basis that variability could be different between groups because of a lack of information on the type or severity of donor diabetes. Means (± 95% confidence interval (CI)) were subsequently back-transformed to reflect the actual concentrations of each element. This analytical approach was used for all comparative outcomes in human donor tissue, and linear regression analysis was used to assess potential correlations with age. The same analysis was done for metal quantification in rat retina for consistency.

The data derived from rat retinas were assessed by a-priori specified comparisons whereby a t- test with Welch’s correction was done to evaluate changes in metal content in untreated diabetic rat retinas compared with age-matched non-diabetic rat retinas. Effects of TETA-

75

treatment were then compared with untreated-diabetic data using unpaired t-tests. The unpaired t-test is used because it is expected that an equal amount of variability in blood glucose values will be present between groups. In this study, TETA was only administered to diabetic rats and so TETA-treated groups were compared only with the non-treated diabetic group. Where there was a change due to (or associated with) disease, further analysis was done to test for the effect of treatment. For the effect of TETA treatment on biomarker expression, a t-test was done to compare TETA-treated and untreated-diabetic groups where an effect of diabetes was observed.

76

3.3 Results

3.3.1 Donor Retinas

A total of 200 retinas were analysed from 100 patients. The mean of both retinas was calculated in every instance to provide a value per donor. The average age of all donors was 73.2 years (range 55–89 years) and 50.5% of donors were female. Donor demographics by subgroup are shown in Table 3.1. Whilst post-mortem delay should not affect trace metal concentrations, it was similar between groups (non-diabetic, 43.2 ± 12.6; diabetic, 40.0 ± 10.3). Na, Mg, K, Ca, manganese (Mn), copper (Cu), Fe, zinc (Zn), selenium (Se) and cadmium (Cd) were all reliably detectable in human retinal samples at concentrations that were quantifiable using this ICP-MS methodology.

3.3.2 Effect of Diabetes on Retinal Trace Metal Content

The results for the analysis of human retinal tissue are shown in Table 3.2. Copper was the only metal that was significantly changed in diabetic retinas (mean (± 95% CI), 133.1 (128.5–137.7) µmol/kg) compared with non-diabetic (mean (± 95% CI), 126.3 (121.1–131.5) µmol/kg) (t (88.02)=2.051, p=0.043) as shown in Figure 3.2. There was no change to Na, Mg, Mn, Ca, Fe, Zn, Se or Cd compared with non-diabetic.

Excluded data points

The medical records for one donor noted signs of suspected diabetes including elevated HbA1C and high serum triglycerides, and recommended diagnostic testing for diabetes. However, the medical notes did not confirm whether the donor was actually found to have diabetes: this case did not meet our criteria for inclusion in the diabetic group, and so this donor’s samples were excluded from the analysis. In the analysis of iron content, two samples were removed from the diabetic group because they were greater than two standard deviations from the mean. One patient had previous received photocoagulation for haemorrhage and the other had haemorrhage noted by the ETR in the assessment of the retina. In all other cases, if the average measurement for a donor was greater than two standard deviations from the mean for a given metal, this was removed from the analysis.

77

Mean Post-mortem Mean age, Age range, post-mortem Group Gender Donors (N) delay range, years (SD) years delay, hours hours (SD)

Male 23 70.8±7.2 59-85 46.9±16.1 31–90

Non-diabetic Female 25 74.0±8.3 55-83 39.9±6.8 28–48 Total 48 72.5±7.9 55–85 43.3±12.4 28–90

Male 26 72.0±7.5 59-85 40.6±12.8 31–76

Diabetic Female 26 75.3± 9.7 56-89 38.4± 5.9 28–56 Total 52 73.6±8.7 56–89 40.0±10.3 28–76

Table 3.1 | A Summary of Donor information for ICP-MS analysis Information was obtained from donor records. Diabetes was defined as a recorded diagnosis of diabetes or documented use of anti-hyperglycaemic medications. Diet-controlled diabetes and a diagnosis of pre-diabetes were also regarded as consistent with a diabetes diagnosis. Males and females were age-matched as best possible and post-mortem time was not significantly different between groups. Data are presented as mean ± standard deviation.

78

Reference Element Concentration Unit Non-diabetic n Diabetic n p-value Isotope

Na (mmol/kg dry-wt) 23Na 1320 (1266–1375) 45 1360 (1292–1429) 51 NS

Mg (mmol/kg dry-wt) 24Mg 27.2 (26.45–28.02) 44 28.2 (27.3–29.0) 51 NS

K (mmol/kg dry-wt) 39K 223.4 (203.3–243.5) 43 239.8 (222.3–257.3) 52 NS

Ca (mmol/kg dry-wt) 44Ca 16.7 (15.7–17.7) 44 16.0 (15.1–16.81) 50 NS

Mn (μmol/kg dry-wt) 55Mn 16.1 (15.22–17.05) 45 16.1 (15.0–17.2) 52 NS

Cu (μmol/kg dry-wt) 63Cu 126.3 (121.1–131.5) 44 133.1 (128.5–137.7) 50 0.043

Fe (μmol/kg dry-wt) 56Fe 1419 (1355–1484) 45 1372 (1282–1462) 51 NS

Zn (μmol/kg dry-wt) 66Zn 1852 (1765–1938) 46 1892 (1793–1992) 52 NS

Se (μmol/kg dry-wt) 78Se 14.6 (13.57–15.6) 44 15.0 (14.0–15.9) 51 NS

Cd (μmol/kg dry-wt) 111Cd 7.9 (6.4–9.4) 45 8.2 (7.0–9.4) 52 NS Table 3.2 | The Effect of Diabetes on Retinal Metal Content The range of detected metals and the isotope used to measure them in non-diabetic (N=47) and diabetic donor retinas (N=52) are shown. Donor measurements were removed from the analysis if they were over two standard deviations from the mean and therefore statistically different to the group. Data are presented as mean (± 95% CI). NS, where p-value not significant with p=0.05 cut-off.

79

Figure 3.2 | Scatterplot of copper measurements in non-diabetic and diabetic donor retinas The shift towards increased copper in diabetic retinas (n=50) compared with age-matched non-diabetic retinas (n=44) is illustrated. Data are presented as mean (±95% CI).

3.3.3 The Influence of Sex on Retinal Trace Metal Concentrations

Trace metal concentrations were analysed in males and females to assess whether sex influenced retinal metal content. The non-diabetic donors were separated by gender and analysed by t-test with Welch’s correction. The results are summarised in Table 3.3. None of the metals analysed were significantly altered by sex.

80

Reference Element Concentration Unit Male n Female n P-value Isotope

Na (mmol/kg dry-wt) 23Na 1369 (1246–1491) 23 1346 (1274–1417) 25 NS

Mg (mmol/kg dry-wt) 24Mg 28.0 (26.5–29.6) 23 27.3 (26.3–28.2) 25 NS

K (mmol/kg dry-wt) 39K 219.3 (186.3–252.3) 23 213.5 (181.3–245.7) 25 NS

Ca (mmol/kg dry-wt) 44Ca 17.4 (16.1–18.7) 22 16.61 (15.0–18.2) 25 NS

Mn (μmol/kg dry-wt) 55Mn 17.56 (15.2–20.0) 23 16.02 (15.0–17.1) 25 NS

Cu (μmol/kg dry-wt) 63Cu 136.1 (122.2–150) 23 125.7 (119–132.3) 25 NS

Fe (μmol/kg dry-wt) 56Fe 1469 (1376–1562) 23 1386 (1297–1476) 25 NS

Zn (μmol/kg dry-wt) 66Zn 1911 (1763–2060) 23 1787 (1695–1878) 25 NS

Se (μmol/kg dry-wt) 78Se 15.2 (13.7–16.8) 23 15.44 (13.3–17.6) 25 NS

Cd (μmol/kg dry-wt) 111Cd 6.945 (5.3–8.6) 23 10.7 (7.1–14.3) 24 NS

Table 3.3 | Comparison of Retinal Metals in Human Males versus Females A secondary analysis of retina metals was done investigating the effect of gender. Males and females were compared for each metal using a t-test with Welch’s correction. Data are presented as mean (95% CI). NS where the p-value was not significant

81

3.3.4 Effects of Age on Retinal Trace Metal Content

We also determined the correlation between metal concentrations and age using linear regression. Donors were between 55 and 89 years of age, so a limited age-range of 34 years was analysed. Only non-diabetic donors were included in this subgroup analysis. Age did not affect metal concentrations with the exception of cadmium. Cadmium was positively correlated with age (p=0.006) as illustrated in Figure 3.3 but none of the other metals changed with age.

Figure 3.3 | The Relationship between Age and Retinal Cadmium Concentration Retinal cadmium levels increased with age in subjects aged between 59–89 years, as determined by linear regression

82

3.3.5 The Effect of Smoking on Retinal Metals

Medical records were inspected for information on smoking history. In the majority of cases, smoking history was not identified and so these donors were excluded from the analysis. We assessed the effect of smoking history only on samples that had a record of smoking habits (past and current). Diabetic and non-diabetic samples were pooled for smoking information to increase numbers. In the dataset, 14 non-smokers were identified from the medical records and 19 smokers/former smokers. Smoking did not influence concentrations of Na, Mg, K, Ca, Fe, Cu, Zn or Se; however, there was a substantial increase in cadmium levels in reported smokers (n=19) (mean (± 95% CI), 10.9 (8.9–12.9) µmol/kg) compared with that in non-smokers (mean (± 95% CI), 4.6 (2.6–6.5 µmol/kg) (n=14) as assessed by t-test with Welch’s Correction (t (29.43)=4.804, p<0.0001) and shown in Figure 3.4.

Figure 3.4 | The Effect of Smoking on Retinal Cadmium Concentration Smokers (n=19) had higher cadmium concentration than non-smokers (n=14). Data are presented as mean (± 95% CI)

83

3.3.6 The Effect of Prescription Medication on Retinal Trace Metals

The drugs selected for analysis were the ACE Inhibitors, β-adrenergic agonists, β-blockers, Ca+ channel blockers, metformin, insulin, ferrous fumarate, proton pump inhibitors (PPI)( lansoprazole and omeprazole) and statins (atorvastatin, pravastatin and simvastatin). These drugs were selected because they were prescribed to ≥ 8 donors according to records from Manchester ETR, with the exception of insulin which was chosen due to the diabetes aspect of this study. Diabetic and non-diabetic groups were pooled in this analysis, except in the case of anti-diabetic drug use where only diabetics were analysed.

In most cases prescription drugs did not have an effect on the retinal trace metals assessed. However, sodium was increased in the retinas of donors who had required statin treatment (mean (± 95% CI), 1519 (1394–1645) mmol/kg) (n=18) compared with those with no record of statin use (mean (± 95% CI), 1321 (1215–1427) mmol/kg) (n=31), (t (39.58)=2.51, p=0.016). This result is illustrated in Figure 3.5 (A). None of the other metals analysed were altered in donors that had been prescribed statins. Retinal magnesium was lower in donors with a history of PPI use (mean (± 95% CI), 26.54 (25.2–27.7) mmol/kg) (t (38.27)=2.105, p=0.042) compared with donors without a history of use (mean (± 95% CI), 28.28 (27.2–29.4)5.22 mmol/kg) as shown in Figure 3.5 (B). Diabetic donors had been prescribed insulin had lower concentrations of retinal sodium (mean (± 95% CI), 1114 (841.6) mmol/kg) (t (8.635)=2.893, p=0.019) compared with diabetic donors who had not been prescribed insulin (mean (± 95% CI), 1468 (1364–1572) mmol/kg) as shown in Figure 3.5 (C).

84

Figure 3.5 | The Effect of Prescription Drugs on Retinal Trace Metals Of the drugs analysed, only the statins, proton pump inhibitors and insulin showed evidence of altering retinal metal concentrations. Statin users (n=18) had increased retinal sodium compared to non-users (n=31) (A). Donors who had been prescribed proton pump inhibitors (n=18) had lower retinal magnesium content than those who had not (n=31) (B). Sodium was decreased in insulin-treated diabetic donors compared to diabetics who were not treated with insulin (n=24) (C). Data are presented as mean (± 95% CI)

85

3.3.7 Changes in Retinal Trace Metal Content in Rats with STZ-induced Diabetes

Retinas were analysed from the rats in Study 3, as described in Chapter 2. All the metals that were detected in human donor samples were also detected in the rat retinas with the exception of cadmium. Metals were quantified as described for the human retinas. This sample-set included two TETA-treatment groups; a preventative treatment and an intervention treatment.

In the rat retina, metal levels were unchanged. Copper trended toward an increase in the untreated diabetic rat retinas (mean (± 95% CI), 78.82 (72.8–84.9) µmol/kg) compared with non- diabetic retinas (mean (± 95% CI) 68.73 (58.2–79.3) µmol/kg) as assessed by t-test with Welch’s correction but this was not significantly (t (12.81)=2.048, p=0.062). None of the other detected metals were significantly changed in diabetic rat retinas compared with non-diabetic as shown in Table 3.4.

TETA-treatment, administered as either a preventative treatment or an intervention, did not change total copper content compared with untreated diabetes (p=0.41 and p=0.67 respectively). Nor did TETA alter any of the other measured metals compared with untreated diabetic retinas. The mean of measured metals in each TETA-treated groups and p-values obtained from an unpaired t-test compared with non-diabetic are shown in Table 3.5 for reference. Therefore, because the TETA-treated retinas were not statistically different from untreated diabetic retinas, these groups were merged to compare to the non-diabetic rat retinas and improve statistical power. Upon merging the diabetic groups, the copper levels were higher than in the non-diabetic retinas (mean (± 95% CI) 80.8 (77.7–83.8) µmol/kg) (t (10.34)=2.543, p=0.029). This comparison is illustrated in Figure 3.6 (A). In the human retina, there was a 5.4% increase in copper levels in diabetes and a similar result was obtained for the rat retinas with a 14.8% increase.

A similar analysis in wet kidney cortex showed a 13-fold increase in copper concentration in STZ diabetic kidney compared with non-diabetic in the renal cortex (samples from Study 2) where there is a 13-fold increase in the diabetic rats (t (16.33) t=11.78, p<0.0001) that is largely normalised by preventative TETA treatment (t (18.95) =7.39, p<0.0001) (Figure 3.6 (B)).

86

Reference ND vs D Element Concentration Unit Non-diabetic (ND) Diabetic (D) Isotope p-value

Na (mmol/kg dry-wt) 23Na 1724 (1531-1917) 1633 (1360-1906) NS

Mg (mmol/kg dry-wt) 24Mg 40.95 (37.35-44.56) 38.28 (33.28-43.28) NS

K (mmol/kg dry-wt) 39K 325.3 (289.5-361.1) 335.8 (294.9-376.8) NS

Ca (mmol/kg dry-wt) 44Ca 15.27 (12.94-17.59) 13.62 (8.36-18.88) NS

Mn (μmol/kg dry-wt) 55Mn 20.82 (19.01-22.63) 23.5 (19.26-27.74) NS

Cu (μmol/kg dry-wt) 63Cu 68.73 (58.2-79.26) 78.82 (72.76-89.12) NS (0.062)

Fe (μmol/kg dry-wt) 56Fe 599.1 (517.7-680.4) 609.0 (509.0-708.9) NS

Zn (μmol/kg dry-wt) 66Zn 1033 (953.10-1112) 1082 (912.2-1251) NS

Se (μmol/kg dry-wt) 78Se 9.409 (8.360-10.46) 10.55 (8.924-12.17) NS Table 3.4 | Quantification of Trace Metals in Non-diabetic and Diabetic in Rat Retinas Changes were assessed by a priori specified comparisons whereby diabetic retinas were compared with non-diabetic prior to being compared with TETA- treated diabetic retinas. Copper was the only element analysed that neared a p-value of <0.05. Copper values in both TETA groups were then compared with untreated diabetic retinas and were not statistically different (p=0.4092 for Tp and p=0.6709 for Ti). Data are mean (± 95% CI).

87

Reference D vs Tp p- D vs Ti p- Element Concentration Unit Diabetic (D) TETA-Prevention TETA-Intervention Isotope value value

Na (mmol/kg dry-wt) 23Na 1633 (1360-1906) 1380 (1134-1626) 0.155 1574 (1387-1761) 0.801

Mg (mmol/kg dry-wt) 24Mg 38.28 (33.28-43.28) 37.25 (34.93-39.56) 0.786 36.62 (33.10-40.15) 0.597

K (mmol/kg dry-wt) 39K 335.8 (294.9-376.8) 305.0 (278.5-331.5) 0.284 302.1 (275.2-329) 0.164

Ca (mmol/kg dry-wt) 44Ca 13.62 (8.36-18.88) 11.70 (8.483-14.92) 0.567 11.9 0(8.92-14.88) 0.689

Mn (μmol/kg dry-wt) 55Mn 23.5 (19.26-27.74) 26.28 (19.70-32.87) 0.415 22.11 (17.06-27.15) 0.581

Cu (μmol/kg dry-wt) 63Cu 78.82 (72.76-89.12) 82.10 (75.09-89.12) 0.409 80.28 (74.42-86.14) 0.674

Fe (μmol/kg dry-wt) 56Fe 609.0 (509.0-708.9) 740.6 (552.3-928.9) 0.149 728.8 (590.8-866.7) 0.114

Zn (μmol/kg dry-wt) 66Zn 1082 (912.2-1251) 1142 (1028-1256) 0.404 1057 (950.3-1164) 0.866

Se (μmol/kg dry-wt) 78Se 10.55 (8.924-12.17) 10.88 (9.474-12.28) 0.671 10.04 (8.395-11.68) 0.595 Table 3.5 | The Effect of TETA Treatment on Trace Metals in Diabetic Retinas Administration of TETA as a preventative (Tp) or intervention (Ti) treatment did not affect any of the trace metals analysed in diabetic retinas. Data are presented as mean (± 95% CI). Both Tp and Ti groups were compared with untreated diabetic using an unpaired t-test.

88

Figure 3.6 | Copper Changes in Diabetic Rat Retina and Kidney Cortex In 16-week diabetic rats when results from untreated- and TETA-treated rats were merged, there was a 14.8% increase in mean copper concentration compared with non-diabetic rats (p=0.029). (A). Copper content of the renal cortex is shown from a previous 16-week study (B). Copper content in diabetic renal cortex (n=12) was significantly higher than in controls (n=12), and this elevation was normalised by TETA-treatment (n=9). Data are mean (± 95% CI).

89

3.3.8 Analysis of Copper Homeostatic Protein-encoding Genes by RT-qPCR analysis

Analysis of the metal content of human and rat retinas demonstrated an increase in total copper levels in diabetes. Therefore, the effects of diabetes on the expression of genes involved in copper transport pathways and key copper containing enzymes was investigated in rat retinas following diabetes. The mRNA expression of Mt1, Mt2, Cp, Sod2, Ctr1, Ctr2, Atp7a, Atp7b, Atox1, Ccs, Cox-17, Sco1, Dmt1 and Commd1 was compared in non-diabetic rats and rats after 12 weeks of Diabetes by two tailed unpaired t-tests; results are shown in Figure 3.7. 12 weeks of untreated diabetes resulted in substantial increases in the gene expression of the metallothioneins Mt1 and Mt2 and caeruloplasmin (Cp) (t (46)=8.80, p<0.0001), t (46) =7.31, p<0.0001 and (t (46)=9.35, p<0.0001) respectively). Increased mRNA expression was also observed of Sod2 (t (45)=3.972, p=0.003), Ctr2 (t (45)=2.36, p=0.023) and Dmt1 (t (46)=2.07, p=0.044) expression.

The same genes (with the exception of Commd1) were analysed in 16-week retinas from Study 2 that included a TETA prevention group. These were initially assessed for an effect of diabetes for comparison with the 12-week study. The results obtained are shown in Figure 3.8. Both Mt1 and Mt2 mRNA expression were up-regulated in the 16-week study (t(38.43) =7.44, p<0.0001) and (t(38.93)=16.82, p<0.0001) respectively. Similar to the 12-week study, Cp and Dmt1 were also up-regulated in diabetic retinas compared with control, (t(33.68)=13.03, p<0.0001) and (t(54.73)=2.451, p=0.0175) respectively. Ctr2 and Sod2 expression contrasted with the 12-week experiment in that both remained unchanged in diabetes compared with control, (t(51.56)=1.298, p=0.2002) and (t(54.62)=0.7713, p=0.4439).

Treatment with the copper chelator TETA was then analysed for an effect on these transporters in the diabetic tissue matrix, as shown in Figure 3.9. It was found that TETA treatment significantly up-regulated Mt2 expression compared with the untreated diabetic group as analysed by an unpaired t-test (t(50)=2.096, p=0.0412). In contrast, Mt1 expression remained unchanged (t(52)=1.153, p=0.2541). TETA treatment also significantly reduced Cp expression compared to untreated diabetics (t(52)=3.194, p=0.0024). Atp7a expression was significantly decreased in TETA-treated retinas compared with untreated diabetic (t(54)=2.664, p=0.0102). Atp7b expression trended downwards (t(54)=1.796, p=0.0781). The fold-change indicated that expression was decreased by approximately 20.5%. Both Ctr2 and Sco1 expression trended toward the 0.05 threshold (t(52)=1.813, p=0.0756) and (t(50)=1.911, p=0.0617) respectively. However, both genes were only decreased by approximately 15% and 16% respectively. Dmt1 expression was not changed with TETA treatment compared with untreated diabetics (t(54)=0.692, p=0.4919) with a fold change of 7%. The gene expression of Ctr1, Ccs, Cox-17, Atox1 and Sod2 remained unchanged.

90

3.3.9 The Effect of TETA-treatment on Expression of Established Biomarkers in STZ-Induced Diabetic Retinopathy In Chapter 2, changes in expression of a panel of established biomarkers of diabetes in the rat were used to validate the STZ model. Whilst TETA treatment did not affect diabetes-induced copper accumulation in the retina, the increase in metallothioneins and partial normalisation of Cp mRNA expression indicated a potential therapeutic benefit. In the 16-week analysis, the biomarkers were analysed by a priori specified t-tests. This addressed the question of whether the expected disease effect was present. Biomarkers that were changed as shown by a p-value of <0.05 were then further assessed by a second t-test comparing TETA-treated and untreated diabetic rat retinas. Expression of Chi3l1, Gbp2, Hspb1, JAK3, Lgals3, Timp1, C1nh, Vegfa and Txnip were analysed for an effect from TETA; the results are shown in Figure 3.10. Most of the biomarkers were unaffected by TETA-treatment; however, Gbp2 and Txnip were further elevated in the TETA group (t (33)=2.35, p=0.0248, t (29)=3.45, p=0.0017) compared with untreated diabetes.

91

A

B

Figure 3.7 | Changes in the Expression of Genes that Encode Proteins Involved in Copper Homeostasis in Rat Retinas after 12 weeks of Diabetes The results for all the genes assessed after 12 weeks of Diabetes are shown in the table (A). 12 weeks of hyperglycaemia significantly increased the mRNA expression of Mt1, Mt2, Sod2, Cp, Ctr2 and Dmt1 but did not induce expression changes in the other copper transporters analysed. The statistically significant changes are illustrated graphically in B. Results shown are pools of three replicates comparing control (n=6) and diabetic (n=10) groups from Study 1. Data were analysed by two-tailed unpaired t-test and are presented as mean (95% CI). * Where p<0.05, *** where p<0.001, **** where p<0.0001

92

A

B

Figure 3.8 | Changes in the Expression of Genes that Encode Proteins Involved in Copper Homeostasis in Rat Retinas after 16 weeks of Diabetes Results for all genes are shown in A. The genes that were significantly changed are shown in graphical format in B. Mt1, Mt2 and Cp mRNA expressions were greatly increased after 16 weeks of hyperglycaemia compared with non-diabetics. Results shown are pools of three replicates gene comparing non-diabetic (n=12) and untreated diabetic (n=13). Data were analysed using a t-test with Welch’s correction and are presented as mean (± 95% CI).

93

A

B

Figure 3.9 | The Effects of TETA-treatment compared with untreated diabetes on the Gene Expression of Copper Transporters after 16 weeks of STZ-induced diabetes The results of the analysis of genes involved in copper homeostasis in TETA-treated diabetic retinas compared with untreated diabetic retinas are shown in (A). A graphical comparison of untreated and TETA-treated copper transport genes that were significantly changed in untreated diabetes compared with control (Mt1, Mt2, and Cp) for comparison with the previous figure is shown in (B). It also illustrates the change in Atp7a during TETA-treatment compared with untreated diabetes and genes that were close passing the significance threshold of p<0.05 (Atp7b, Ctr2 and Sco1). Results shown are pools of three replicates gene comparing untreated diabetic (n=13) and TETA-prevention treated diabetic (n=9). Data were analysed by t-test and are presented as mean (95% CI).

94

A

B

Figure 3.10 | The Effect of TETA-treatment on biomarkers of STZ-induced Diabetic Retinopathy The biomarker panel from Chapter 2 was assessed to test for an effect of TETA-treatment. Gene expression of biomarkers that were significantly changed in hyperglycaemia by t-test in the previous chapter were selected for analysis for an effect of TETA-treatment. The data presented are from the same dataset in Chapter 2. Presented here are three replicates for each gene with the exception of Vegfa and Txnip where two replicates were completed. Diabetics (n=13) and TETA-prevention treatment (n=9) were compared by t-test and are presented as mean (95% CI). All genes analysed are shown in (A) and are graphically illustrated in (B).

95

3.4 Discussion

To our knowledge, this is the first study that has investigated this range of trace metals in the retina and the first to thoroughly investigate the effect of diabetes on a range of retinal trace metals. The primary objective of this ICP-MS-based analysis of human retinal tissue was to determine the effect of diabetes on metal levels in human retina.

Here, total tissue-content of sodium, potassium, calcium, magnesium, manganese, copper, iron, zinc, cadmium and selenium was analysed to investigate concentrations of these trace metals in the human retina and how these are affected by diabetes.

Copper was the only measured metal that changed in clinical diabetes. Importantly, this finding was replicated in the results of a second study in rats with 16-weeks diabetes wherein, again, all other trace metals did not differ between diabetic and non-diabetic groups.

The mechanism of the change in the copper content of diabetic rat retinas was further investigated by analysing changes to copper transporter mRNA expression that occur in diabetes in two separate studies, after 12 and 16 weeks of hyperglycaemia. It was observed that after both 12 and 16 weeks of untreated-diabetes, there was a dramatic increase in both Mt1 and Mt2 gene expression: the proteins encoded by these genes contain both copper and zinc atoms. There was also a large increase in Cp gene expression. The Cp-encoded copper- protein, caeruloplasmin, is secreted from tissues such as the liver, where it acts as a means of cellular copper export into the blood. Whether caeruloplasmin is also secreted from the retina into the bloodstream is not entirely clear.

Here, TETA-mediated copper-chelation therapy did not lower the diabetes-induced increase in total retinal copper; however, it did decrease retinal Cp mRNA expression compared with untreated diabetes. These data therefore provide evidence for a selective dysregulation of the copper homeostasis in diabetes and alterations following TETA treatment.

3.4.1 Quantification of Retinal Trace Metals

The ICP-MS-based approach utilised allows for quantification of all detectable metals in a given sample. Values for the concentrations of some of the metals analysed here have been reported previously. A comparison of the values obtained in this study compared with previous publications is shown in Table 3.6. Notably, the study by Eckhert et al 1983270 assessed concentrations of copper, calcium, iron and zinc in human eye tissues using atomic absorption spectrophotometry. The values reported are substantially less than the respective values obtained in the current study, with the exception of iron. This difference is most likely attributable to technological advancement and increased sensitivity of the new ICP-MS equipment used in this study. A more recent study found iron levels similar to those obtained in this study271. The study by Erie et al (2009)272 analysed copper and zinc values in human retinas and obtained values that align with the data obtained in this study. Another study has

96

observed a mean cadmium value of 9.12 µmol/kg in human retina273, which is similar to that observed in this study.

This study has shown that trace metals in rat retinas tended to be proportionally lower than in humans with the exceptions of magnesium and manganese. Ugarte et al., (2012) reported levels of zinc, iron, copper and selenium in rat retinal tissue using ICP-MS251. The values obtained, shown in Table 3.6, are consistently lower than those observed in this study: the reasons for this are unclear but are likely because of increased sensitivity of the ICP-MS system used in this study, thereby having an increased ability to detect metals in a small amount of tissue such as a dried rat retina. The relative proportions of these four metal are however similar between the studies.

Human

Copper Iron Zinc Calcium Cadmium Study (µmol/kg) (µmol/kg) (µmol/kg) (mmol/kg) (µmol/kg)

This Study 126.3 1419 1852 16.7 7.9

Eckbert (1983)270 95 2100 57.6 8.75 –

Hahn (2006)271 – 1826.8 – – –

Erie (2005)273 – – – – 9.1

Erie (2009)272 142.9 – 1853.6 – –

Rat

Copper Iron Zinc Selenium Study (µmol/kg) (µmol/kg) (µmol/kg) (µmol/kg)

This study 68.7 599.1 1033 9.4

Ugarte (2012)251 28.57 358.9 484.9 6.41

Table 3.6 | Comparison of Previously Published Data on Retinal Metal Quantification Values for previous studies that quantified metals are shown. In some cases, an approximate µmol/kg value was calculated from the µg/kg value provided in the study for comparison with the data obtained in this study. All values shown were obtained from dry tissue.

97

3.4.2 The Effect of Diabetes on Retinal Trace Metal Concentrations

In this study, a moderate increase in total copper concentration was observed in both donor human and 16-week STZ rat retinas. Copper was the only metal altered in both clinical disease and the rat model of diabetes. To our knowledge, this is the first study to investigate copper content in diabetic retinas compared with non-diabetic controls. Consistent with our findings, a study by Kokavec et al (2016)274 reported an increase in total copper content in the vitreous of 12 living diabetic patients compared with 15 non-diabetic case controls. Kokavec et al measured sodium, potassium, chloride, calcium, magnesium, copper, zinc and selenium, out of which copper and iron were the only metals changed. The decrease in iron in the diabetic vitreous compared with non-diabetic contrasts with our study. However, whilst the vitreous and retina are adjacent in the eye they are very different in that the retina is a highly cellular structure whereas the vitreous is virtually acellular, so it would not be surprising if they regulated metals in different ways.

In this study, data have also been presented showing a large increase in copper content in diabetic rat kidney cortex compared with control cortex. In this tissue, TETA-treatment decreased copper content to close to control levels. Previous studies have indicated a potential therapeutic role of TETA-treatment in STZ-induced diabetic nephropathy. Nine weeks of TETA intervention treatment in a 16-week study normalised D-amino oxidase, a H2O2 producer, epoxide hydrolase-1, glutathione S-transferase α3 and aquaporin-1, a water channel, compared with 16 weeks of untreated diabetes in rats as well as amelioration of albuminuria, an indicator of kidney function275. TETA also normalised diabetes-evoked upregulation of electron-transfer flavoproteins and downregulation of acetyl-coA acetyltransferase. Another study has shown that TETA normalised collagen content in 16-week STZ rat kidneys and partially normalised levels of CML and cathepsin L, a collagen-degrading lysosomal proteinase that acts downstream of MMP-2276. Therefore, normalisation of copper content may produce benefits in tissues affected by diabetes.

TETA did not have any demonstrable effect on the overall copper levels in the diabetic retina. This could be because the retinal copper increase was not large enough to be affected by chelation treatment. A copper balance study has shown that while TETA-treatment decreased copper content in diabetic subjects that had increased urinary copper excretion, copper balance remained unchanged in non-diabetic controls167. This indicates that in non-diabetic subjects with normal physiological copper homeostasis, hypocupraemia is not induced. It may be that a therapeutic effect by TETA requires a certain threshold of copper dysregulation and as this study has shown, the copper change in retina is mild in comparison to that observed in kidney. The kidney contains the highest concentration of copper in the body277. Copper is filtered and reabsorbed in the kidney but in cases of copper overload, it is excreted in the urine278. Another point to note is that TETA is a Cu(II)-specific chelator167,248,279. As the ICP-MS measures total copper, it is possible that Cu(I) increases in the retina (probably in intracellular compartments) whereas TETA will only chelate extracellular Cu(II). This however is speculation and further

98

investigation is required to determine the valency and localisation of the increased copper in the diabetic retina.

Another factor for consideration is whether TETA can cross the blood-retinal barrier and, if so, how effectively. TETA is a charge-deficient analogue of spermidine that is similarly metabolised by spermine-N1-acetyltransferase to produce mono- and diacetylated derivatives (SSAT1)280,281. Approximately 3–4% of putrescine, the precursor to spermidine, crosses the blood-brain barrier mostly by passive diffusion, but with some active transport (33%)282.

The hydrophilic nature of TETA-disuccinate283 implies a possible limitation to its ability to successfully undergo passive diffusion through cell membranes, which are hydrophobic. Preliminary data was acquired during this study using a targeted liquid-chromatography mass spectrometry approach to identify TETA and its mono- and diacetylated derivatives in the retinas diabetic rats after they were administered 86 µg/ml of TETA disuccinate in drinking water. This assay requires validation and optimisation to determine concentrations of TETA and metabolites; however, the preliminary evidence obtained appeared to indicate that TETA and its metabolites were present in the retinas of TETA-treated rats.

It is becoming increasingly apparent that copper has a role in diabetic pathology, especially given its role in oxidative stress and AGE accumulation. Here, it has been shown that copper is dysregulated in the diabetic retina. TETA treatment did not change the diabetes-induced increase in retinal copper. The effects of TETA on molecular mediators are described in the next section.

3.4.3 Diabetes Evokes Molecular Changes in Expression of Genes Corresponding to Copper Homeostatic Mechanisms

Whilst the ICP-MS data indicated an increase in total copper content in both rat and human retina, the gene expression studies provided evidence of changes in copper metabolism in the diabetic retina. The data obtained showed a dramatic increase in metallothioneins, Mt1 and Mt2, and Cp mRNA expression after both 12 and 16 weeks of diabetes in the STZ rat model. Ctr2 mRNA expression changed in the 12-week study but not the 16-week study but there was no clear effect of diabetes overall. A diagram of the overall changes to copper transporters in diabetes is shown in Figure 3.11. The implications of these changes will be discussed in the sections below.

Metallothioneins Metallothioneins protect the cell from the damaging effect of free copper in the cytosol. Whilst metallothioneins preferentially bind zinc under normal circumstances, copper binds with higher affinity: therefore, when intracellular copper is in excess, copper is preferentially bound.

This study has shown that mRNA expression of both metallothioneins Mt1 and Mt2 are upregulated after 12 and 16 weeks of diabetes. It has been observed that metallothioneins are

99

increased in diabetic kidneys after 6 months of diabetes alongside increased copper content284. Similarly, during copper deficiency in the left ventricle of the heart after 16 weeks of Diabetes, both Mt1 AND Mt2 gene expressions are downregulated166.

This study has also shown that TETA treatment induces a further increase in Mt2 mRNA expression. This was not observed in Mt1 expression despite the observation that both genes have been shown to demonstrate increased expression upon copper binding to the Mtf-1 transcription factor285. Further research would be required to investigate the preferential upregulation of Mt2 during TETA treatment. This effect can be deemed a protective mechanism because metallothioneins sequester free copper preventing oxidative damage286.

Metallothioneins also protect against NMDA-induced toxicity, namely loss of ganglion cells in mouse retinas. Pre-treatment of retinas with ZnSO4 increased the protective effect of metallothioneins against NMDA toxicity287. NMDA receptor activation causes a calcium influx that induces reactive oxygen species. NMDA receptors are found in photoreceptors, horizontal, bipolar and ganglion cells288. Activation of NMDA receptors may have a role in driving the pathogenesis of diabetic retinopathy. The use of memantine, an uncompetitive NMDA receptor antagonist, prevented visual function and ganglion cell loss, blood-retinal barrier breakdown and diabetes-induced VEGF protein increases without affecting VEGF levels in treated non-diabetic animals289. VEGF expression and blood-retinal barrier integrity were also reduced diabetic rats were treated with brimodine, a selective α2-adrenergic receptor agonist that has been shown to prevent the increase in intracellular Ca2+ induced by glutamate in cultured retinal ganglion cells290,291. This suggests that NMDA receptor activation may have a pathogenic role in diabetic retinopathy and the TETA-induced increase in Mt2 expression may be protective.

Metallothioneins have been implicated in angiogenesis, where expression was increased during oxygen-induced retinopathy and it was found that knockout of Mt1 and Mt2 reduced the expression of hypoxia-inducible factor 1-alpha (HIF-1α ) and VEGF 292. Interestingly, use of the copper chelator tetrathiomolybdate decreased neovascularisation and VEGF protein in this model293. However, in this study Vegfa mRNA expression was downregulated in the diabetic retina whilst metallothioneins were upregulated. Furthermore, TETA-induced upregulation of metallothionein expression did not influence VEGF expression in this study. The protein expression of VEGF should be analysed to confirm this result, but it is possible that copper chelation may not restore decreased VEGF during early retinopathy but may inhibit retinal neovascularisation.

Caeruloplasmin This study has also observed substantial increases in Cp expression. Caeruloplasmin is a copper-containing ferroxidase that plays a role in modulating iron homeostasis. The protein contains up to six bound copper atoms that are incorporated early in the course of its biosynthesis294. In this study, Cp expression was increased consistently in both 12- and 16- week trials. TETA-treatment partially normalised Cp expression. Martin et al (2005)295 showed

100

that Cu(II) upregulated HIF-1α protein. It was also shown that caeruloplasmin gene expression is regulated by HIF-1αdependent promoter activation. Given the hypoxia that occurs during diabetic retinopathy as discussed in Chapter 1, copper may increase Cp expression during diabetes through this pathway. Treatment with copper chelator tetraethylenepentamine inhibited hypoxia-induced activation HIF-1α296 and so the effect observed with TETA treatment may occur in a similar manner.

Given the lack of effect of TETA treatment on total retinal copper content in this study, the partial normalisation of Cp expression is indicative of a potential therapeutic effect in the retina. Caeruloplasmin has important roles in iron homeostasis, namely in the movement of iron from cells to plasma and the oxidation of Fe2+ to Fe3+ that is essential for binding to transferrin297. Caeruloplasmin expression can also be modulated by dysregulated iron homeostasis298. Given this role of caeruloplasmin in iron modulation, it may be of interest to study the effects of TETA on iron transporters. Whilst total iron content was not changed in diabetic human or rat retinas, the charge state was not assessed nor was the influence of diabetes and copper chelation on iron transporters such as transferrin, hephaestin, and ferritin. A detailed review of iron homeostasis is provided by Song and Dunaief (2013)250.

ATP7A and ATP7B In this study, it was found that Atp7a mRNA expression was unaffected by both 12 and 16 weeks of diabetes. However, TETA treatment reduced Atp7a expression compared to both untreated diabetes and control groups. Interestingly, in the study by Zhang et al (2014), where copper content in the left ventricle of the heart in 16-week STZ rats was downregulated, there was no change to Atp7a expression in untreated diabetes; however, it was upregulated with TETA treatment166. This was also reflected at the protein level. Taken together with the current study, these findings indicate that TETA appears to exert an effect on Atp7a mRNA expression in an inverse direction to that of the copper change.

Whilst Atp7b mRNA expression did not change significantly in either study in diabetic rats, there was a 20% increase after both 12 and 16 weeks of hyperglycaemia, suggesting the possibility of an effect that the studies were not sufficiently powered to detect. In the 16-week study, TETA treatment decreased Atp7b expression by approximately 21% thereby returning it to the levels found in non-diabetic rats. Zhang et al (2014)166 found no change to the mRNA expression of Atp7b during diabetes-induced cardiac copper deficiency. However, the level of protein was significantly reduced and not ameliorated by TETA treatment. Given the trends seen in Atp7b expression in this study, future studies could investigate the influence of diabetes and TETA treatment on ATP7B protein expression and localisation.

The copper-transporting ATPases under normal circumstances are found in the Golgi complex where they incorporate copper in copper-requiring proteins. ATP7B is known to incorporate copper into caeruloplasmin protein; however, when ATP7B is deficient, ATP7A can compensate and instead incorporate copper into caeruloplasmin299,300. ATP7A expression was

101

downregulated to a similar extent as caeruloplasmin, and ATP7B overall follows a similar trend to caeruloplasmin expression. Therefore, it is possible that the partial normalisation of caeruloplasmin expression by TETA could be driven by modulating ATP7A and ATP7B expression. Interferon-γ increases protein expression of ATP7A in microglia in a copper- dependent manner301. NMDA receptor activation may also affect ATP7A and ATP7B localisation, and copper dysregulation may have an effect on NMDA-induced toxicity302,303. Interferon-γ is increased in the rat retina after 5 months of Diabetes226. TETA modulation of ATP7A and ATP7B expression may have an effect on this pathway and may warrant further exploration.

DMT1 In this study, it was observed that Dmt1 mRNA expression was increased in diabetes in both 12- and 16-week experiments. DMT1, divalent-metal transporter 1, has an important role in cellular iron transport 304 and can also compensate for deficient Ctr1 expression305. DMT1 is expressed throughout the retina, in rod bipolar cell bodies, rod bipolar cell axon termini, horizontal cell bodies, and photoreceptor inner segments306. Cu(I) uptake has been shown to inhibit DMT1-mediated uptake of Fe2+ and so diabetes may modulate copper and iron uptake via altered DMT1 expression307.

Other Copper transporters The majority of copper transporters analysed remained unchanged. This includes the expression of the main importer for copper into the cell Ctr1. Despite observed changes to Ctr2 gene expression, Ctr1 expression remained tightly controlled. This is unsurprising given that other groups have shown that Ctr1 expression is unchanged following exposure to excess copper concentrations308,309; however, trafficking of the protein to the plasma membrane is affected310 and total Ctr1 protein is decreased305.

The current study has shown conflicting results in the effect of diabetes on Ctr2 gene expression. The 12-week experiment indicated a small increase in Ctr2 expression in diabetic compared with control rat retinas. Conversely, the 16-week experiment indicated a trend toward a decrease in diabetic rat retinas compared with non-diabetic rat retinas that was further enhanced by TETA-treatment. Neither of these 16-week comparisons was statistically significantly. These conflicting results suggest that changes observed in Ctr2 are inconclusive and overall Ctr2 expression is unaffected.

Ccs expression also remained unchanged here. A previous study has also shown that during copper deficiency Ccs gene expression remained unchanged; however the protein levels were modulated311. Conversely the study by Zhang et al (2014)166 showed a decrease in both mRNA and protein expression in the copper-deficient cardiac left ventricle following induction of diabetes. Here, the change in copper content in the retina was less than that observed in the left ventricle in the heart166, where there was a marked deficiency of copper; thus, the smaller increase measured in retina in the current study may not have been sufficient to induce a

102

Figure 3.11 | Summary of Copper Transporter mRNA Expression Changes in 16-week Diabetic Rat Retina This diagram summarises changes to intracellular copper transporters during diabetes. The genes in green, Mt1, Mt2, Dmt1 and Cp were upregulated in the retina after 12 and 16 months of diabetes. The genes labelled in grey remained statistically unchanged and those in white were not measured. The specific increase in metallothionein expression is indicative of a mild increase in intracellular copper due the copper storage properties of metallothioneins. Excess copper may be exported as caeruloplasmin, the product of Cp. DMT1 is responsible for a proportion of copper import and so a change in diabetes may influence intracellular concentrations of Cu(I) and Fe2+. Adapted from (Leary et al., 2009, Wee et al., 2013, Polishchuk and Lutsenko, 2013)312–314

103

change in Ccs expression. Atox1 and Cox17 expression also remained unchanged in this study however; there is very little information in the literature regarding changes to the expression of these genes during copper dysregulation.

Overall, these results appear to indicate that copper content is increased in the diabetic retina. The changes observed in copper transporters and other proteins/genes of cell copper homeostasis in this study, appear to be generally consistent with those reported in the literature with copper accumulation in other organs and scenarios, and are consistent with the data from ICP-MS identifying copper accumulation in the diabetic retina.

3.4.4 The Effect of TETA-treatment on Diabetes-Related Biomarkers

The previous chapter showed changes to known diabetes-related biomarkers in rat retina after 16 weeks of diabetes. Here, the effect of copper chelation treatment on the gene expression of these biomarkers is shown. The expression of the majority of genes analysed remained unchanged. Expression of Gbp2 and Txnip were statistically different following TETA-treatment and Lgals3 and C1inh expression had a ≥20% fold change (1.20 fold change reported) compared with untreated diabetes indicating a trend toward a difference in mean.

Little is known about Gbp2 except that it is induced by interferon-γ315. The current study has shown that TETA-treatment further increased the diabetes-induced upregulation of Gbp2 expression. This suggests that the interferon-γ pathway may be changed in diabetes; indeed it has been reported that interferon-γ is upregulated in retina after 5 months of diabetes226. If TETA-treatment further upregulated interferon-γ, this could be a protective effect. Interferon-γ has been shown to protect against cuprizone-induced demyelination and glutamate-receptor- induced excitotoxicity316,317. Interferon-γ has also been shown to protect against excitotoxicity317. Glutamatergic signalling is required during phototransduction as is evident by the release of glutamate as a neurotransmitter by both photoreceptors and bipolar cells. NMDA receptors are found in photoreceptors, horizontal, bipolar and ganglion cells288. NMDA may have a role in driving the pathogenesis of diabetic retinopathy. The use of memantine, an uncompetitive NMDA receptor antagonist, prevented visual function and ganglion cell loss, blood-retinal barrier breakdown and diabetes-induced VEGF protein increases without affecting VEGF levels in treated non-diabetic animals289. VEGF expression and blood-retinal barrier integrity were also reduced diabetic rats were treated with brimonidine, a selective α2-adrenergic receptor agonist that has been shown to prevent the increase in intracellular Ca2+ induced by glutamate in cultured retinal ganglion cells290,291. This suggests that NMDA receptor activation may have a pathogenic role in diabetic retinopathy. Interestingly, copper is released from neurons in response to stimulation of the NMDA receptor302. As mentioned previously, Zheng et al., (2010) found that interferon-γ modulates ATP7A trafficking in a copper-dependent manner and that there was increased copper accumulation in interferon-γ-treated BV-2 cells due to increased Ctr1 mRNA expression301. These studies together implicate a potential relationship between copper homeostasis, NMDA receptor signalling and interferon-γ. TETA-induced upregulation of

104

Gbp2 expression suggests that copper chelation may modulate the interferon-γ pathway and suggests a previously unknown pathway by which TETA may exert an effect that warrants further investigation.

Txnip, the gene that encodes thioredoxin-interacting protein, was selected as an extra biomarker of interest because the study by Bixler et al (2001) showed that insulin treatment did not normalise Txnip expression. Insulin treatment trended toward further increasing TXNIP expression87. The current study also found that Txnip expression increased in diabetic retinas compared with control and was further increased by TETA. TXNIP is best known for the function after which it was named, that is inhibiting the endogenous antioxidant thioredoxin318,319. This would suggest that increased TXNIP expression increases oxidative stress and therefore would worsen the redox state in the diabetic retina. However, the review by Singh (2013) highlights numerous other functions of TXNIP including apoptosis induction, NMDA-mediated cell death, VEGFR2 internalization and inflammation. Interestingly, recent studies have indicated roles in fructose transport321, lipid metabolism322,323 and a role in mediating lipid-induced inhibition of glucose uptake324. Given the trend toward further upregulation of Txnip expression in insulin- treated diabetic rat retinas, the effects observed with TETA may not have toxic implications. Importantly, the study by Hansen et al (2006)325 showed that copper does not oxidise thioredoxins 1 or 2 and does not activate ASK1 or cause cell death. Therefore, the TETA- induced increase in Txnip expression is likely not caused by an increase in free intracellular copper.

3.4.5 Secondary Outcomes of Trace Metal Analysis in Human Retina

Whilst the main objective of this study was to assess the effect of diabetes on retinal trace metal content, the data also allowed for an exploratory study of the effect of gender, aging, smoking and prescription medication. An important factor in these analyses is that the data were not optimised for these outcomes and therefore are limited, subject to confounding variables and may be underpowered to detect effects.

Trace Metal Content is not Influenced by Gender None of the metals analysed were significantly different in concentration in males compared with females. This concurs with a previous study on retinal metal concentrations326, and also with studies performed in the vitreous and lens that showed no differences in males compared with females274,327. The data in this study did not show a significant difference in cadmium in females compared with males. This disagrees with a previous study that showed an increase in females compared with males328; however, this study did not take smoking history into account and this may have influenced the data. Cadmium, whilst naturally occurring, differs from the other metals examined in this study in that it does not have a biological function and therefore is purely toxic if sufficient levels are reached329. The tight regulation of the other trace metals is unsurprising given the important biological roles of these elements. These results demonstrate that while the human data is gender balanced, and so gender is not a confounding factor.

105

Smoking Changes Retinal Cadmium Content The effect of smoking on retinal trace metals was investigated. Former smokers were grouped with smokers. This is justified by the absence of biological mechanisms for cadmium excretion from tissues. These data are limited due to the mixed population and low numbers. A further limitation is the lack of information acquired regarding smoking duration and quantity. In this study it was found that most all metals were unchanged in smokers and former smokers, except for cadmium which showed an increased concentration. Previous studies have increased cadmium in the retina, lens and plasma of smokers273,330,331. In a previous study, cadmium accumulation was highest in the RPE, which would adversely affect retinal health328. Upon entry into the cell, cadmium induces oxidative stress and binds metallothioneins, thereby causing increased intracellular concentrations of copper and zinc. Intracellular cadmium can also cause release of intracellular calcium stores leading to the activation of many downstream signalling pathways332. Increasing concentrations of cadmium decrease the activities of SOD and glutathione thereby reducing the capacity of the cell’s endogenous antioxidants333. Copper has been indicated in previous studies as being changed in the lens of smokers, however there is conflicting data as to whether it increases or decreases in concentration272,331,334,335. Copper was not affected by smoking in this study and it is possible that these other studies did not account for diabetes as a potential confounding factor. Overall, the data shown here indicate an increase in cadmium in the retina of smoking donors. This agrees with data in the literature and therefore supports the novel data in this study.

Aging only Affects Cadmium Content In this study, the total content of most metals analysed remained tightly controlled with age. This is inconsistent with a previous study where zinc and copper have been shown to decrease in the retina with age326; however, it is important to note that age ranged from 1.5–87 years in that study. In this current study, the age range was 59-90 years of age and so the data presented here only relates to retinal metal content in older age. Whilst age was controlled for in this study when comparing non-diabetic with diabetic donors, these data indicate that age was not an additional variable in this dataset and highlights the specificity of the change in copper in diabetes. Cadmium was an exception however, and accumulated in retinas with increasing age. This is unsurprising given the lack of cadmium elimination mechanisms and renal reabsorption336. Retinal cadmium has also been shown to increase with age328. This study however was undertaken in donors between ages of 1.5 and 87 years. Therefore, the current study shows that this effect continues in an older range of retinas.

The Effect of Prescription Medication on Retinal Trace Metal Content The use of prescription medication was assessed for an effect on trace metals in retina. The effects of ACE Inhibitors, β-adrenergic agonists, β-blockers, Ca+ channel blockers, metformin, insulin, ferrous fumarate, PPIs and statins were analysed. Overall, there was little demonstrable effect of these drugs on retinal trace metals.

106

However, there were some notable changes. Donors who had been prescribed statins had increased retinal sodium. This may be related to the observation that statin use results in systemic sodium retention by decreasing excretion337. Statins are usually prescribed to reduce low-density cholesterol. However, patients with high cholesterol are likely to also have hypertension338 that is often associated with a high-sodium diet.

Magnesium content was decreased in donors with a medical history of PPI use. PPIs are thought to reduce magnesium absorption and in rare cases induce hypomagnesemia339–342. Therefore, the reduction in PPI-treated donors compared with non-PPI-treated donors could be due to lowered circulating magnesium in this population.

The last effect of prescription drugs observed was a decrease in sodium in retinas from insulin- prescribed diabetic donors compared with those who had not been treated with insulin. Whilst insulin increases sodium retention343, this analysis compared insulin-treated diabetics with those that did not require insulin and therefore were T2DM. This effect may therefore be due to the higher level of sodium intake that has been observed in patients with T2DM344. The observations of increased sodium in statin users, decreased magnesium in donors who had been prescribed PPIs, and the increased sodium in non-insulin dependent donors are most likely reflective effects of uptake via peripheral tissues that is reflected in the retina rather than via a direct effect in retina. Whilst these effects are most likely not detrimental to retinal health, they give an indication that the dataset in this study have changes that are compatible with previous literature, so the increase in copper content in diabetic retinas compared with non- diabetic retinas, is not likely to be an artefact.

3.5.6 Limitations

The data obtained in this study are limited by several factors. As previously mentioned, the medical records were not sufficiently detailed to provide information on diabetes type and duration and diabetes type was inferred from donors who had received oral hypoglycaemic agents. A history of drug use was not provided for all patients, which may introduce unknown confounding factors. Importantly, 1 in 4 diabetics, approximately 1 million people, in England are not aware they have diabetes345. As previously stated, the prevalence of undiagnosed pre-diabetes is 16% in the UK population aged 50-69 years and 26% in the >70 year population268. Therefore, it is likely that a portion of donors studied here who had not been diagnosed with diabetes in fact had diabetes or pre-diabetes. The data for smoking history similarly lacked sufficient detail and so subtle changes to other metals may have occurred and were not observed. The data for all results are representative of the whole retina and thus subtle changes to particular metals in specific cell types may be hidden. Another limitation is that total copper is measured and the valency of the copper was not assessed. In addition, it is unknown whether the changes in copper levels were intracellular and/or extracellular although the former is probable (because most copper is intracellular). Another limitation of this study is that the whole retina was measure and so it is unclear in which cells copper was accumulated. It

107

would be of interest to assess the difference in diabetic and non-diabetic tissue sections as was done by Ugarte et al 2012251. The role of astrocytes in maintaining ion and copper homeostasis in the brain as well as acting as metal depots and as regulators for the distribution of essential metals to other types of brain cells make them a likely candidate to accumulate copper to mitigate neuronal damage from copper overload346.

3.5.7 Conclusions

There results obtained demonstrate a mild increase in total retinal copper in diabetes; this change was specific to copper. This is the first study to demonstrate these results in human retina. While the change to total copper was not significant in the STZ rat, the changes to metallothioneins and caeruloplasmin suggest that copper dysregulation occurs in the rat retina. This emphasises that the STZ model may replicate changes seen in clinical disease and that copper dysregulation may be a useful biomarker of disease. Copper homeostasis in diabetes, or indeed in disorders other than Wilson’s or Menkes diseases, is an area that is not very well studied and merits further research. Here, a new dysregulated pathway in diabetic retinopathy has been demonstrated. So far, a therapeutic effect by chelation using TETA has not been shown in the retina, but this requires further study such as by administration via intravitreal injection and use in a longer duration of diabetes, whereby more copper accumulation could occur, may influence endothelial cell loss.

108

Chapter 4 | Polar Metabolite Profiling of the Rat Retina during Hyperglycaemia

109

4.1 Introduction Diabetes is a metabolic disorder whereby a lack of functional insulin signalling leads to decreased gluconeogenesis and therefore increased circulating glucose. Insulin is essential for energy homeostasis and has important roles in aiding glucose transport across cell membranes, mediating glycolysis, glycogen synthesis, attenuating lipolysis and increasing amino acid and protein synthesis347. Therefore, disruption of insulin function causes an array of metabolic dysfunction across the metabolome. It is well established that hyperglycaemia leads to increased polyol pathway flux, AGE formation, PKC activation and hexosamine pathway flux95,102,105,348.

Gas chromatography mass spectrometry (GC-MS) is a useful tool for metabolite analysis. This is because retention times can be calculated to provide a retention time index for each metabolite detected. The use of internal standards to normalise the output means that an accurate retention index can be reproduced across experiments and even across different laboratories. Retention indices can be compared with online metabolite spectra libraries, such as the NIST library, for putative metabolite identification. A matching combination of spectra at the correct retention index is indicative of a metabolite that can be identified confidently. Identifications can then be confirmed by running a standard solution of the proposed metabolite and assessing whether the behaviour of the standard reflects the putatively identified metabolite; therefore, GC-MS is a method by which reliable identifications can be made. The components of a GC-MS are illustrated in Figure 4.1 (A).

Gas chromatography uses mobile and stationary phases to separate metabolites. The mobile phase is the carrier gas and the stationary phase is the inner lining of the capillary column, which is illustrated in Figure 4.1 (B). A small quantity of the sample is injected into the gas chromatograph where individual components are separated according to volatility. Very volatile components pass through the capillary column quickly, and arrive in the mass spectrometer earlier than less volatile components due to temperature-dependent variations in metabolite binding to the stationary phase. For a certain temperature range, these molecules spend some time stationary, and some time moving with the carrier gas349. The temperature of the oven increases over time to produce an analysis over a wide range of volatility.

Whilst polar groups make metabolites water soluble, they also make the metabolite involatile. For this reason, derivatisation of metabolites is done to increase their volatility and thermal stability and therefore the metabolites are capable of running through the gas chromatograph. Silylation, performed using trimethylsilyl (TMS), is one of the most common methods of derivatisation for metabolite analysis. Trimethylsilyl is not a naturally occurring side chain and so identified metabolites with a silyl chain group can be confirmed as being identified from the sample set as opposed to contamination. During this process, a TMS group replaces the active hydrogen of a hydroxyl, thiol, amine, or carboxylic acid group with a preference of alcohols > phenols > carboxylic acids = amines > amides350. Prior to silylation, oximation of the

110

Figure 4.1 | Diagram of GC-MS components An overview of the GC-MS system is illustrated in A. The sample is injected in the inlet and immediately vaporised, after which the carrier gas carries the metabolic components through the gas chromatogram. After this, the electron beam is focused and directed toward the detector and displayed on a processing computer. The components of the capillary column are shown in B.

111

sample sugars is done to minimise trimethylsilyl ethers produced and reduce the number of peaks obtained for sugars and polyols such as glucose and sorbitol. Without oximation, these metabolites produce multiple forms and create complicated chromatograms where instead after modification each sugar will have a maximum of two peaks351.

After chromatographic separation, the mass spectrometer ionises and fragments the components as they arrive from the gas chromatograph, and separates the fragments. The fragments formed by any given metabolite reflect the strengths of the bonds between its atoms, so a characteristic mass spectrum is recorded which gives information about the chemical structure of the metabolite. Often, this is sufficient to allow the metabolite to be identified however as mentioned, retention index and comparison with a standard solution of the metabolite assures correct identification.

Innovations in the area of diabetic retinopathy (DR) have been limited in the last decade suggesting a need for further understanding of the mechanisms behind the disease to uncover new therapeutic targets. Despite being a metabolic disease, it is only in recent years that untargeted metabolic profiling has been attempted in the field of diabetic retinopathy. This has been done using vitreous and plasma from human DR patients352,353, but untargeted full-scale metabolomics has not yet been completed in the retina of either donor human retinas or in an animal model of diabetes; partly owing to the small quantities of retinal tissue in rodents. A method has been developed that can detect a range of metabolites in the rat retina that was previously difficult due to the small retinal tissue yield. This method separates polar and non- polar metabolites that are run on separate experimental set-ups. This chapter focuses on polar metabolites such as sugars, polyols and amino acids. Whilst it has been established that there are changes to the sorbitol pathway102,348, glycolysis353,354 and amino acids355, these changes have not been observed in unison in a single tissue sample. Confirmation of these established changes concomitantly will help establish the global metabolite changes in the diabetic retina.

There is still a significant gap in the knowledge of metabolite dysregulation in diabetes, with only a few recent studies having investigated changes to glycolysis intermediates. There is a need to investigate these changes concomitantly with other global metabolite changes in the diabetic retina. This will provide new insights and novel targets that may benefit patients with early diabetic retinopathy. The objective of this chapter was to utilise GC-MS to provide a reproducible dataset of mapped polar metabolites in non-diabetic and diabetic rat retinas. This overview should provide insight into carbohydrate and amino acid changes in diabetic retinas and, when coupled with the data regarding lipids in the next chapter, an overview of the overall metabolic profile during diabetic retinopathy.

112

4.2 Methods The GC-MS method described below was completed using retinas from three separate studies, all of which were described in Chapter 2. The first experiment completed used retinas from 16- week STZ Study 3. The second was done with retinas taken from the 12-week Study 4, and the third from the 16-week Study 5. For simplicity, these will be referred to as GC-MS Experiments A, B and C respectively in this chapter. These are summarised in Table 4.1

Experiment A Experiment B Experiment C

Experiment number 3 5 4

Duration of diabetes, 16 16 12 weeks

Table 4.1 | Description of Animal Experiments used for GC-MS analysis For details on start and final body weights and blood glucose levels, please refer to Chapter 2: results, Table 2.2.

4.2.1 Tissue Extraction Whole rat retinas were prepared using a Folch-style extraction using a TissueLyser bead homogeniser (Qiagen, Manchester, UK). Each sample was extracted in 800 µL of 50:50 (v/v)

13 methanol:chloroform, to which a solution of the labelled internal standards ( C6-D-fructose, d10- leucine, d4-Citric acid, d4-Succinic acid, d5-L-tryptophan, d7-L-alanine, d-35-stearic acid, d5- benzoic acid, and d5-) (Cambridge Isotopes, Tewksbury, Massachusetts, USA) in methanol had been added. Tissue was homogenised for 10 min at 25 Hz with a single 3-mm tungsten carbide bead. Phase separation was achieved by addition of 400µL water followed by brief mixing by vortex and centrifugation (2400 x g, 5 min). The chloroform layer was removed and retained for lipid analysis in Chapter 5. Great care was taken to ensure the full removal of excess chloroform from the methanol phase. Extraction blanks were prepared by including tubes without tissue samples in each batch.

From the methanol:water supernatant, 400-μL aliquots were transferred to pre-labelled tubes. Quality non-diabetic replicates (QC) were prepared by pooling equal amounts of extract (200 μL) from each sample. After brief mixing, 400-μl aliquots of the QC pool and each sample were dried (30 °C, 16–18 h) in a Speedvac centrifugal concentrator (Thermo Scientific SPD331DDA). Dried residues were then stored at 4 °C until derivatisation for GC–MS analysis.

4.2.2 Derivatisation The method used was developed in-house and described by Patassini et al (2016)356. Briefly, this method used methoximation and trimethylsilylation to generate a profile of polar small- molecule metabolites such as amino acids, simple organic acids and monosaccharides, and is

113

applied here to provide comparative data in a case-non-diabetic experimental design. Dried residues were reconstituted in 30-μL methoxylamine hydrochloride solution (Sigma-Aldrich, Gillingham, UK) (20mg/ml in dry pyridine (Acros Organics, Geel, Belgium)) and heated at 80 °C for 20 min in sealed tubes. After this, 30 μL of N-methyl-N-(trimethylsilyl) trifluoroacetamide (MSTFA) (Acros Organics) was added and heated again at 80 °C for a further 20 min. Finally, 10 μL of a retention-time marker solution (nine n-alkanes covering the range C12-C32 dissolved at 10mg/ml in 1:1 hexane:pyridine) was added and the solutions transferred to autosampler vials for GC–MS analysis.

4.2.3 GC-MS Analysis Chromatography was done using an Agilent/J&W DB17-MS column (Agilent Part No. 122-4732 30 m × 0.25 mm × 0.25 μm) with a 3 m × 0.25 mm retention gap, and helium carrier at a constant flow of 1.4 ml/min. The oven temperature progressively increased from 50 °C to 300 °C at 10 °C /min. 1 μL injections were performed in Pulsed-Splitless mode using an MPS2 autosampler (Gerstel; Germany), and a 7890A Gas Chromatograph with Split/Splitless inlet (Agilent, Santa Clara, USA). Column effluent was analysed using a Pegasus HT time-of-flight mass spectrometer (LECO, Stockport, UK), acquiring 10 spectra over the mass range of 45–800 Da.

This study was performed in three separate STZ-rat experiments. In each experiment, diabetic and non-diabetic groups were randomised and run with QC samples placed consistently throughout the sequence. Extraction blanks were inspected visually to confirm absence of carryover and contamination.

4.2.4 Data analysis Data were prepared using the ‘Reference Compare’ method within ChromaTOF 4.5 (LECO). Databases that were employed included the NIST Mass Spectral Reference Library (NIST08/2008; National Institute of Standards and Technology/Environmental Protection Agency/National Institutes of Health Spectral 262 Library; NIST, Gaithersburg, MD, USA); and a library of approximately 90 definitive standards developed at the CADET laboratory. Pooled QC samples were used to create a reference table of metabolites across all the samples. By comparing several QC samples, it was possible to identify metabolites that could be reliably reported and exclude features which could not.

Due to the length of time elapsed between the first experiment and the following second and third experiments, two separate reference tables were created. Repeating the process by which the reference table is created shows the metabolites that are reliably identified between separate experiments and increases confidence in identifications.

4.2.5 Statistical analysis Raw data were entered into SIMCA-P+ version 12.0.1.0 (Umetrics AB) to assess variability and class separation. This can be used to find potential outliers within each class. After this, the data

114

were transferred to GraphPad prism version 7.00 (GraphPad Software Inc.). Row means and coefficient of variance (CV) were assessed and any metabolite that was found to have >25% CV in the QC group was removed from the dataset. Multiple t-tests were then done between non-diabetic and diabetic without assuming a consistent SD. Discovery was determined using the Original FDR method of Benjamini and Hochberg, with Q=10%. Fold changes were also calculated between groups.

115

4.3 Results

4.3.1 Gas Chromatography The first assessment of the quality of the run was done by viewing the chromatograph in QC samples to ensure that a good chromatograph was produced across the experiment. An example of a chromatograph from a QC sample in Experiment A is shown in Figure 4.2. The gas chromatogram shows the retention time on the x-axis that is used alongside the m/z obtained from the mass spectrometry detection to identify metabolites. The peaks obtained in the example shown are sharp and narrow indicating high resolution. A wide peak, a broad peak or a peak that does not return to baseline would be indicative of a poor signal-to-noise ratio.

4.3.2 Principle Component Analysis (PCA) The PCA plot is a descriptive technique that summarises the multivariate structure of the dataset. It is a matrix of observations and samples that accounts for sources of variation across the dataset by indicating correlations between observations. PCA analysis indicates the quality of the experiment. This analysis can help to identify whether the conditions set for the experiment influence the data obtained such as the effect of diabetes on detectable metabolites and can also be used to identify outliers as illustrated in Figure 4.3. In this example in which Experiment C was analysed, two samples in the diabetic group lie with the non-diabetics. Upon further investigation, these rats were found to have reverted from diabetes to normoglycaemia upon reaching the end of the experiment. This plot also indicated potential problems with samples 4 and 32. Whilst the in vivo study data appeared to be in range with the rest of the group, it appeared that sample 4 had a noticeably small signal across all metabolites and so this sample was excluded from further analysis. Sample 32 was investigated but remained included because there was no technical issue that may have caused this variation and could be because of biological variation. This indicates how PCA analysis is useful in assessing the quality of the data across all metabolites and samples. Figure 4.4 shows the final PCA plots for each Experiment Completed on the GC-MS. In all three experiments, a clear distinction is observed between non-diabetic and diabetic groups as shown in Figure 4.4.

116

Figure 4.2 | Example of a Gas Chromatogram from Experiment A Peaks were identified according to mass and retention time using the NIST library and a library developed at the CADET laboratory at the University of Manchester. Only the most abundant metabolites, (those with the tallest peaks) are labelled. The y-axis refers to the abundance of that particular metabolite.

117

Figure 4.3 | An example of the use of PCA analysis to determine group separation All the samples processed from Experiment C are shown in A. This includes two samples from rats in the diabetic group (samples labelled 18 and 29) that were normoglycaemic upon sacrifice. They are included here to illustrate how this is reflected in the metabolic profile. This PCA analysis also highlighted variability in samples 4 and 32. Upon further investigation, it was found that sample 4 had noticeably smaller peaks, likely due to poor injection. The removal of sample 4, but not 32, as shown in B, shows a clear separation between diabetic and non- diabetic groups. This increases confidence in the likelihood that changes seen in this study are due to the effect of diabetes.

118

Figure 4.4 | Final PCA plots for all GC-MS Experiments PCA analysis indicates a separation of diabetic and non-diabetic groups in all three experiments. Here the final PCA plots are presented with technical outliers excluded. The QCs are generally close and between the diabetic and non-diabetic groups. Arguably, Experiment A has the best class separation and QC reproducibility (A), followed by Experiment B (B) and Experiment C (C). None were excluded from Experiment A (non-diabetic, n=12; diabetic, n=8), one non-diabetic rat was removed after PCA analysis, because it was vastly different from the rest of the samples and two reverted rats were removed from the diabetic group in Experiment B (non-diabetic, n=9; diabetic, n=7), and three samples were removed from Experiment C as discussed in the previous figure (non-diabetic, n=12; diabetic, n=9),. D, diabetic; ND, non- diabetic; QC, quality control.

119

4.3.3 Overall Changes to Metabolites In Experiment A, 127 metabolite spectra were identified including unidentifiable groups of spectra. Any metabolite with a >20% coefficient of variation (CV) in the QC group was removed from final analyses. Upon removal of preliminary identifications with a high CV, those with multiple peaks and unknown metabolite spectra, the final number of identifications was 51 metabolites. A full list of identifications is available in Appendix III. Of 51 metabolites, 28 remained unchanged, 14 were significantly upregulated and 10 were significantly downregulated in diabetic retinas compared with non-diabetic. All changed metabolites are shown in Table 4.2.

4.3.4 Sugars and polyols Of the upregulated metabolites, sorbitol increased in diabetes to the greatest extent and was increased over 6-fold compared with non-diabetic retinas in all three experiments. This was followed by glucose (>5-fold) and fructose (>5-fold). The results for glycolysis and polyol pathways are illustrated in Figure 4.5. Similar results were found Experiments B and C as summarised in Table 4.5. In all three experiments these were statistically significant. The polyols threitol, and maltitol and xylitol were also upregulated 3-fold, 2-fold and just under 2-fold respectively. Of these three, only threitol was identified in all experients and was increased approximately 3-fold in all experiments in diabetic retinas compared with non-diabetic. Maltitol and xylitol were not detected in Experiments B and C. Of the inositols, scyllo-inositol was consistently downregulated in each experiment: 70% in Experiment A, 56% in Experiment B and 63% in Experiment C. Conversely, myo-inositol was not changed in any experiment.

Components of glycolysis, and the citric acid cycle were increased including fructose-bis- phosphate (56%), glucose-6-phosphate (39%) and malic acid (24%). Fructose-bis-phosphate and glucose-6-phosphate were not detected in Experiments B and C. Malic acid was not significantly changed in Experiments B and C and was only 1% and 15% different from non- diabetic respectively. Citrate was downregulated in Experiment A by 26% but was upregulated in Experiments B and C (44% and 28% respectively).

4.4.5 Amino Acids Of the amino acids identified, only the branch chain amino acids (BCAAs) were upregulated in diabetes. Leucine and isoleucine were upregulated in Experiment A, by 45% and 40% respectively. Valine was identified in the dataset but the small signal led to poor reproducibility across the QC group and so the <20% CV criteria was not met. Both leucine and isoleucine were upregulated to similar extents in Experiments B and C compared with Experiment A; however, the change in isoleucine was not statistically significant in Experiments B and C.

The remaining detected amino acids that changed in diabetic retinas compared with non- diabetic were downregulated. Threonine was consistenly downregulated in diabetes compared with non-diabetics. In both 16-week experiments (A and B), it was downregulated by over 30%.

120

In the 12-week experment C, it was only downregulated 15% but was still significant. Methionine and tryptophan were only detected in Experiment A and were downregulated by approximately 40% and 50% respectively. Phenylalanine and tyrosine were detected in all three experiments and were significantly downregulated in both 16 week experiments in diabetics compared with non-diabetics. Phenylalanine was decreased by approximately 50% in Experiments A and B and only 34% in Experirment C. Tyrosine was decreased by 71% in Experiment A and 50% in Experiments B and C.

4.3.6 Fatty acids and glycerol Myristic, palmitic and nonanoic acid were all detected in rat retina with variable results. Myristic acid was significantly upregulated by 26% in Experiment A and increased by 29% in Experiment B but this was not significant. In Experiment C, it also remained unchanged. Palmitic and nonanoic acid were significantly decreased in Experiment A; however, both were increased in Experiments B and C. Glycerol, a component of triglycerides, was increased in both Experiment A (24%) and B (2-fold). Glycerol was increased by 25% in the 12 week Experiment C but this was not significant.

4.3.7 Others Beta-hydroxybutyric acid was significantly increased in all three experiments. This increase was most prouncounced in Experiment A where it was increased >6-fold, followed by Experiment C where it was increased >4-fold and Experiment B where a >3-fold increase was observed. L-threonic acid was downregulated by 70% in Experiments A and C but remained unchanged in Experiment B.

121

Table 4.2 | Polar Metabolite Changes in Diabetic Rat Retinas Relative to Non-Diabetic List of metabolites from Experiment A (16 weeks of diabetes) found to be changed in diabetes relative to non-diabetic samples by multiple t-tests with a 10% FDR arranged from the highest increase to the most decreased. The changes from Experiment A were compared with Experiments B (16 weeks of diabetes) and C (12 weeks of diabetes) and most of the same changes occurred. All fold changes were statistically significant unless stated otherwise by NS. Not all of the metabolites identified in Experiment A were identified in Experiments B and C, as marked by NR. Most of the changed metabolites were a definite identification. D, definite; C, confident; NR, not reported; NS, not significant; P, putative.

122

Figure 4.5 | Changes to Components of the Glycolysis and Polyol Pathways Glucose is metabolised by the glycolysis pathway or, upon saturation of this pathway, may instead be metabolised to fructose and sorbitol by the polyol pathway. Here it is demonstrated that glucose is high in diabetic rat retinas compared with non-diabetic. Experiment A demonstrated a concomitant increase in glucose-6-phosphate and fructose 1,6-bisphosphate. In the same samples, a substantial increase in fructose and sorbitol was observed. All metabolites presented were taken from Experiment A and were statistically significant as assessed by multiple t-tests with a 10% FDR.

123

4.4 Discussion This study has shown an array of metabolites that change in the STZ diabetic rat retina compared with non-diabetic. Sorbitol was the metabolite with the greatest magnitude of change in all three experiments. Glucose, fructose, beta-hydroxybutyric acid and threitol were the following four metabolites with the largest increases compared with non-diabetic, though the ranking of these varied in each experiment. Metabolites that changed in all three experiments and those that significantly changed only at 16 weeks of diabetes are summarised Table 4.3. These include the polyol threitol, the BCAA leucine and scyllo-inositol. Experiments A and B were 16-week samples whereas Experiment C contained 12-week diabetic samples. Metabolite changes that were common to only the 16-week experiments included glycerol and the amino acids threonine, phenylalanine and tyrosine, though it should be noted that the fold changes for glycerol and tyrosine in the 12 experiment were similar to 16-week data. The conflicting changes observed in fatty acids across experiments are easily explained as a limitation of the method utilised. The dual-extraction of polar and non-polar metabolites potentially led to some contamination of the polar phase with fatty acids. Therefore, the results obtained are likely to be unreliable and will not be discussed further. It has also shown that the majority of these changes were reproducible in two subsequent experiments. In all three experiments, there was clear separation between the non-diabetic and diabetic groups. Each QC is indicative of the reproducibility of the chromatograph throughout the experimental run and therefore the distribution of the QCs indicates the quality of the run.

4.4.1 Established Metabolite Changes Occurred in all Experiments The method utilised in this study measures a vast range of metabolites and so it is imperative to establish the validity of the dataset. This can be done by assessing the literature for any changes previously observed. One of the key traits of metabolic changes in STZ rat retinas is an increase in glucose, sorbitol and fructose relative to control357,358. Increased retinal BCAAs and decreased retinal threonine have been demonstrated previously after 5 days of STZ- induced diabetes355. Regarding the BCAAs, an increase in leucine was observed across all three experiments in this dataset. Isoleucine was significantly increased in one experiment but trended toward an increase in the remaining two experiments indicating a similar profile. Valine concentrations were too small to be reliably measured. Threonine was consistently downregulated in the data presented. The confirmation of similar changes in previous studies validates the data obtained for less-studied metabolites examined in this dataset355,357,358.

124

Table 4.3 | Summary of the reproducibility of metabolite changes Changes that were common to all three experiments were statistically significant after analysis by multiple t-test with a 10% FDR that is equivalent to a p-value of 0.1. The metabolites common to both 16-week experiments were not statistically significant at 12 weeks however in some cases, such as with glycerol, the fold change was similar to both 16 week studies.

125

4.4.2 Diabetes Increases Retinal Glucose, Glycolysis and Tricarboxylic Acid (TCA) Cycle Intermediates As stated, it has been established that glucose content increases in diabetic retinas compared with non-diabetic in the STZ model. This suggests that a concomitant increase in pathway intermediates may occur. Most metabolites in this pathway are at concentrations too low to be detected using this method. However, in Experiment A, glucose-6-phosphate and fructose-bis-phosphate were detected and increased in diabetic rat retinas compared with non- diabetic. Glucose-6-phosphate has only been measured once previously in the diabetic retina and was found to be unchanged compared with non-diabetic359; however, these rats were only maintained for 20 days post-STZ. Fructose 1,6 bis-phosphate has not extensively investigated but the study by Sas et al (2016)360 showed that fructose bisphosphate and hexose-6-phosphate were increased in 24-week db/db mouse retinas compared with non- diabetic mice. The lack of studies investigating these intermediates is probably due lack of sensitivity, as these were not detected in the Experiments B and C.

Of the TCA cycle intermediates detected, only citrate and malate were changed in diabetes. The changes to citrate contrasted between experiments and therefore the results are deemed inconclusive. Malate was increased in Experiment A and trended the same way in Experiment B. The study by Sas et al., (2016) investigated changes to the TCA cycle by measuring metabolic flux and found that citrate, α-ketoglutarate and malate were increased in the diabetic rat retina compared with non-diabetic360. This agrees with the results obtained in this study, though α-ketoglutarate was not reliably detected across the QC sample set.

Previous studies have observed increased lactate concentrations in diabetic rat retinas compared with non-diabetic and a change to the lactate/pyruvate ratio359,361,362. The data presented here consistently showed no change in lactate or pyruvate in diabetic retinas relative to non-diabetic. The GC-MS method utilised provides only a relative comparison between case and non-case rather than a concentration and so the lactate/pyruvate ratio could not be calculated. It is highly unlikely that this ratio would have been different between groups given the lack of change to both lactate and pyruvate. The discrepancy between the results presented here and previous studies is likely due to the duration of diabetes. None of the aforementioned studies had diabetes for longer than 4 weeks. Salceda et al (1998) showed that retinal lactate and pyruvate increased after 7 and 14 days respectively of STZ-induced diabetes but had normalised by 45 and 20 days respectively363. This suggests that lactate may have fluctuated over the course of the trial but was normalised by the end of the study. Overall, this study has shown a trend toward a general increase in glycolysis and TCA cycle intermediate metabolites.

4.4.3 Polyols are Consistently Increased in the Diabetic Retina except for Scyllo-Inositol During glycolysis pathway overload, glucose may be metabolised by alternative pathways such as the aldose reductase mediated polyol pathway. In all three experiments presented, a

126

significant increase was observed in both fructose and sorbitol. Sorbitol formation occurs in the retina and is not transported by the bloodstream357,358. Sorbitol and fructose accumulation have also been observed in lens, kidney, aorta, diaphragm, erythrocytes and sciatic nerve in the STZ rat357,358,364.

Treatment with an aldose reductase or a sorbitol dehydrogenase inhibitor reduces retinal sorbitol and fructose respectively361,365. Aldose reductase expression was increased in human retinas in patients that were diabetic for over 14 years366, the type of diabetes was not determined, and sorbitol accumulation occurs in human retinas exposed to high glucose levels indicating increased aldose reductase activity 96. However, there are no data available regarding direct measurement of sorbitol content in human diabetic retinas compared with non-diabetic. Aldose reductase also produces the polyol xylitol from xylose, a reaction that is inhibited by sorbinil367. In the presented study, xylitol was increased in diabetic rat retinas compared with non-diabetic. This suggests greater aldose reductase activity as has been shown in human retinas ex vivo, in this case increasing xylitol metabolism.

Threitol was putatively identified. Despite testing a standard solution of threitol to confirm the identity, threitol has the same mass and retention index as erythitol, another product of ascorbic acid metabolism, and these cannot be identified separately using this method. This metabolite is referred to in this study as threitol however; the conclusions that can be drawn from the data are limited. Threitol was significantly elevated and this was easily detected and consistent in all three experiments. Oxidation of L-ascorbic acid produces threose368,369 and reduction of threose by aldose reductase produces threitol370. This change to threitol was accompanied by a decrease in L-threonic acid in diabetic compared with non-diabetic rat retinas. L-threonic acid results from oxidative metabolism of ascorbic acid369,371. The metabolism of ascorbic acid is increased in diabetes, however the metabolic pathway requires further elucidation to increase the understanding of the relationship of different metabolite intermediates372. The varying results for ascorbic acid catabolism intermediates may be due to the oxidative state of the tissue, as in high peroxide conditions, threonic acid is produced and non-oxidative ascorbic acid degradation produced threitol371. More research would be required to comprehend the results obtained and the experiment is limited by potential misidentification.

Maltitol was also increased in diabetic rat retina compared with non-diabetic; however, the associated metabolic pathway has not been explored in the literature. This is likely to be metabolised differently to the aforementioned polyols because it is structurally different in that it is a disaccharide. These data indicate that there may be increased aldose reductase activity, leading to increased polyol formation96. It is likely that this increased activity is not exclusive to sorbitol production, but that other pathways may be implicated, such as that which produces maltitol, and these could have unknown toxic consequences.

Unlike the previously discussed polyols that were all increased in diabetes, scyllo-inositol was consistently downregulated. Myo-inositol, the most abundant inositol isomer in tissues,

127

remained unchanged in all experiments. It was found that myo-inositol remained unchanged in diabetic retinas357,361,365. This appears to be the first study to show a specific, reproducible downregulation of scyllo-inositol in the diabetic rat retina. Myo-inositol can be converted to scyllo-inositol via the formation of an intermediate myo-inosose-2 that was not detected in this study. This intermediate is over 10-fold lower than scyllo-inositol in numerous tissues and therefore it is likely that the concentrations in rat retina are too small for detection using this GC-MS. The CNS contains much higher levels of inositol than peripheral tissues and this may be indicative of an important role within the neuronal tissue. Myo-inositol content is approximately 20-fold higher in rat cerebellar cortex and optic nerve than scyllo-inositol373. Both isomers are decreased in peripheral nervous tissue during diabetes indicating a different metabolic footprint in peripheral and central tissues364. Inositols are required for inositol phosphate and diacylglycerol signalling that leads to intracellular calcium release and PKC activation. The inositols have roles in osmolarity homeostasis, restoring Na+-K+-ATPase activity during diabetic neuropathy and translocation of glucose transporter GLUT4 to the plasma membrane in skeletal muscle374–376. Scyllo-inositol specifically has been implicated as a potential therapeutic for Alzheimer’s disease. Administration of scyllo-inositol to TgCRND8 mice inhibits amyloid-β peptide oligomerisation and fibrillisation and improved blood vessel tortuosity that occurs in the model; a pathology that also occurs in human diabetic retinopathy377. Scyllo-inositol has not previously been assessed in diabetic retinopathy and the discovery that it is highly downregulated in a reproducible manner is a novel finding.

4.4.4 Amino Acid Changes Of the amino acids detected and changed in this study, all were downregulated with the exception of the BCAAs. BCAAs differ from other amino acids due to their structure whereby they have aliphatic side chains with a branch consisting of a central carbon bound to three or more other carbons. Unlike other amino acids that are catabolised in the liver, BCAAs are catabolised in skeletal muscle by branched-chain amino acid aminotransferase to produce branched-chain keto acids. Increased retinal BCAAs, namely valine, and decreased tyrosine and phenylalanine were described previously by Nishimura and Kuriyama, (1985)378. This study has also shown an increase in the branched chain amino acids leucine and isoleucine. Leucine was significantly changed in all three experiments and isoleucine was only significant in Experiment A, but the fold changes in Experiments B and C were similar to that of Experiment A indicating reproducibility. Threonine, tyrosine and phenylalanine were all consistently downregulated as will be discussed.

BCAAs are Consistently Upregulated in the Diabetic Retina BCAAs were consistently upregulated in the retina in this study. This effect has been observed previously in the STZ-rat retina355,379 This effect is not exclusive to the retina, as they have also been shown to be upregulated in the plasma of STZ-diabetic rats compared with non- diabetic355,380. Increased concentration of plasma BCAAs are associated with insulin resistance, thought to be due to downstream phosphorylation of IRS-1 and IRS-2 in a mammalian target of

128

rapamycin (mTOR)-mediated reaction381,382. Leucine is ketogenic and so is further metabolised to produce aceto-acetate CoA and subsequently acetyl CoA as shown in Figure 4.6. Isoleucine and valine are both ketogenic and glucogenic and are similarly metabolised to produce acetyl CoA, but in addition produce succinyl CoA, a component of the TCA cycle7. Increased plasma BCAAs suggests reduced catabolism in skeletal muscle. Muscle atrophy is an effect of the STZ model and reduced muscle synthesis and increased degradation is associated with high circulating concentration of BCAAs383. Impaired muscle catabolism of BCAAs is due to reduced branched-chain amino acid aminotransferase activity. This is restored by administration of adiponectin, an adipocytokine that is downregulated in diabetics and has roles in glucose and fatty acid metabolism384,385. The direct effects of increased BCAAs in retina are unknown, but increased levels in the circulation may result in competitive inhibition of the transport across the blood-retinal barrier of other large, neutral amino acids such as tyrosine and phenylalanine386.

Glucogenic Amino Acids Methionine and Threonine were Downregulated in the Diabetic Retina It has previously been demonstrated a decrease in threonine in the diabetic rat retina355, as was observed in this study, but no change to methionine. Methionine was more difficult to detect than threonine and was only detected in Experiment A in this study. A previous study has shown that methionine supplementation in the alloxan model of diabetes in rats restored endogenous antioxidants in liver and kidney387, the effect being comparable to that of insulin treatment, but did not reduce oxidative stress markers, with the exception of thiols. The effect on endogenous antioxidants is thought to be due to the donation of sulfur from methionine to glutathione. Neural tissue however was not assessed, so the effect in retina is currently unknown, but the decrease in methionine observed in this study suggests that methionine supplementation could be beneficial in diabetic retinas. Methionine may be decreased due to increased catabolism for utilisation as a fuel source to produce succinyl-CoA, an intermediate in the TCA cycle, as shown in Figure 4.6. Homocysteine is produced as an intermediate in this pathway and it should be noted that homocysteine is increased in the plasma of diabetic individuals with retinopathy compared with diabetics without retinopathy7,388.

Threonine was more easily detected and was downregulated in diabetic retinas compared with non-diabetic across all three experiments. Threonine is an essential component of proteins for allowing interaction with serine/threonine kinases to induce protein phosphorylation by glycosylation, deficiency of which may affect protein function or trafficking. For example, O- glycosylation of threonine residues is essential for the prevention of Ctr1 cleavage in the cytosol, prior to delivery to the plasma membrane 389,390. This indicates a role for O-glycosylation and threonine in copper homeostasis; a pathway that was earlier shown to be dysregulated in diabetic retinas. Serine/threonine kinases have numerous functions, for example mTOR, is a serine/threonine protein kinase, the activation of which phosphorylates downstream signalling mediators such as p70 ribosomal S6 protein kinase (p70S6K), leading to modulation of mRNA translation. mTOR inhibition attenuates STZ-induced VEGFR2 upregulation and activation of

129

CASPASE-3 apoptosis pathways391. Increased utilisation of the hexosamine pathway and therefore increased production of UDP-GlcNac, O-glycosylation and threonine modification may explain why it was decreased in diabetes. However, it is likely that this is due to threonine catabolism. Threonine is glucogenic and is catabolised to produce both pyruvate and succinyl- CoA 7 as shown in Figure 4.4.4.

Diabetes Caused a Marked Reduction in Neurotransmitter Precursor Amino Acids Tyrosine, tryptophan and phenylalanine were also decreased in diabetes. Tyrosine and phenylalanine were detected in all three experiments but were both only significantly decreased at 16 weeks. Conversely, tryptophan was only detected in Experiment A due to low tissue concentrations. Tyrosine is a non-essential amino acid and can be produced endogenously from phenylalanine by phenylalanine hydroxylase. Both phenylalanine and tyrosine are essential for catecholamine synthesis; that is synthesis of adrenaline, noradrenaline and dopamine7. Retinal tyrosine downregulation was previously described by Fernstrom et al (1984)386 after 2–3 weeks of diabetes. In that study, both tyrosine and DOPA were downregulated in diabetic retinas but tyrosine remained unchanged in the serum. Retinal dopamine and tyrosine hydroxylase activity also remained unchanged in diabetes. Conversely, retinal dopamine was decreased from 3 weeks of diabetes in the study by Nishimura and Kuriyama (1985)378. Retinal dopaminergic signalling is essential for light-adapted visual function, acuity, and contrast sensitivity392. Decreased tyrosine content may be a contributing factor for the changes in visual acuity and contrast sensitivity in early diabetic retinopathy.

Tryptophan is a precursor for serotonin, produced by hydroxylation of tryptophan in a reaction dependent on BH4. Serotonin in turn is a substrate for melatonin that is produced in the pineal gland and retina7. Melatonin and its metabolites have antioxidant properties and have been proposed to also chelate copper393,394. Melatonin is downregulated in the retinas of STZ-induced diabetic rats after 3 days and supplementation of melatonin reduces markers of oxidative stress, VEGF, pigment epithelium-derived factor and HIF-1α levels395,396. It is likely that decreased retinal tryptophan may lead to downregulation of retinal serotonin and therefore melatonin.

Tyrosine, phenylalanine and tryptophan are all large neutral amino acids that were found in this study to be downregulated in retina. Fernstrom et al (1986)386 found that supplementation with BCAAs trended toward decreasing serum levels of tyrosine, phenylalanine and tryptophan and significantly reduced tyrosine uptake into the brain and retina. This suggests that the increased circulation of BCAAs discussed previously may have decreased the uptake of tyrosine, phenylalanine and tryptophan into the retina, which may in turn affect retinal function as discussed.

4.4.5 Diabetes Increased Relative Retinal Concentrations of Lipid Mediators In this study, there was a consistent upregulation of beta-hydroxybutyric acid and glycerol in diabetic retinas compared with non-diabetic. Βeta-hydroxybutyric acid is a ketone body derived

130

from acetoacetate, the intermediate metabolite between ketogenic amino acids and acetyl-CoA. During fasting, the liver accumulates fatty acids from degraded adipose tissue. Metabolism shifts from glycolysis to gluconeogenesis so acetyl-CoA is not utilised for the TCA cycle, but instead is metabolised to produce ketone bodies such as beta-hydroxybutyric acid. This is a reversible reaction that occurs in all cells, with the exception of those without mitochondria such as erythrocytes, when ketone bodies can be used to provide energy for peripheral tissues and so beta-hydroxybutyric acid can be utilised to increase acetyl-CoA and therefore increase TCA cycle flux. When the production of ketone bodies outweighs the rate of metabolism, this induces a rise in serum concentrations and increased delivery these metabolites to the retina7. Beta- hydroxybutyric acid is the ligand for the GPR109A receptor that is localised to the RPE397. This receptor is upregulated in diabetic mouse and human retinas and has a role in suppression of

131

Figure 4.6 | Overview of metabolic changes that occur in the STZ diabetic rat retina compared with non-diabetic An overview is shown of metabolites that were changed in the STZ diabetic rat retina and their roles in glycolysis and the TCA cycle. Changes to metabolites are indicated by the fold- change on the colour scale. Metabolites in grey were not identified.

132

interleukin-6 and chemokine ligand-2 mediated inflammation398. Whilst the main effect of increased beta-hydroxybutyric acid production is maintaining energy production during failed utilisation of glucose, it is interesting to note the potential alternative effects of beta- hydroxybutyric acid upregulation.

Glycerol was also increased in all three experiments and was significantly changed in both 16- week experiments. Glycerol is an essential component of triacylglycerides and is released during hydrolysis of these lipids. This suggests that in diabetes, triacylglycerides are metabolised to produce free glycerol and is a subject of investigation in the next chapter. Glycerol can be phosphorylated by glycerol kinase to produce glycerol-phosphate and further oxidised to produce dihydroxyacetone phosphate and utilised for glycolysis (Figure 4.6). This reaction has been shown in retina, as has triacylglycerol production during normoxic conditions399. This suggests that changes to the retinal lipid profile may be of interest in elucidating changes to the whole metabolome during diabetic retinopathy and this is described in Chapter 5.

4.4.6 Limitations of study The data obtained from these experiments while reproducible, have certain limitations. Derivatisation methods can never encapsulate the entire range of metabolites. Notably, in this study, there were difficulties obtaining discernible information for glutamine and glutamate because of the formation of pyroglutamic acid. This artefact is produced by cyclisation reaction involving the alpha amino group and, respectively, the -CONH2 and -COOH side groups of glutamine and glutamic acid400. Arginine does not form TMS derivatives and so cannot be detected using this method, instead trifluoroacetic acid derivatisation has been used for this pathway401,402. However, that method would not be useful for analysing glucose and polyols and it is important to note that no method is perfect for assessing every metabolite due to variable chemistry. In this experiment, a time-of-flight mass spectrometer was used that has limited capacity for mass determination. An Orbitrap mass spectrometer such as that which is used in the next chapter can differentiate masses to a higher degree of accuracy and therefore may discern more metabolites. Another limitation is that the Pegasus HT, the mass spectrometer used, has a range of four orders of magnitude. The substantial upregulation of glucose, sorbitol and fructose within the samples restricts the detection limit of metabolites that are at much lower concentrations. An experimental model that induces milder hyperglycaemia may be more suitable for the detection of these metabolites however; this may not induce the same amount of metabolic disturbance. The experimental replicates were not perfect given that one study was in 12-week rats however, most of the changes observed at 16 weeks occurred at 12 weeks. It should also be noted that 16-week rats in Experiment B received subcutaneously implanted insulin maintenance. Insulin release was not at a level that attenuated hyperglycaemia but may have had an effect on other mediators such as amino acid transport or ascorbic acid metabolism and introduced another variable to that study set. This study is limited because it does not differentiate between cell types. Astrocytes and neurons differ in that neurons

133

preferentially produce energy by oxidative phosphorylation whereas astrocytes favour glycolytic metabolism of glucose68. With this in mind, it would be of interest to investigate the changes to metabolic pathways in individual cell types to further elucidate how these metabolic pathways are affected by glial and neuronal dysfunction induced by diabetes. Lastly, whilst there was a definite increase in flux observed in the glycolysis pathway, few mediators from the hexosamine and pentose phosphate pathways were reliably identified and so these pathways may need to be measured in a different manner to study how increased glycolytic flux affects these alternative pathways.

4.4.7 Conclusion Overall, the changes observed in this study closely match those observed in a study of plasma from DR patients compared with non-DR diabetic patients. Glycolic metabolites including sorbitol, glucose and glucose-6-phosphate were observed in DR patients, which were also increased in STZ rat retinas in this study353. Whilst the in vivo experimental studies were not identical, the aim was to provide a general dataset of changes that occur during hyperglycaemia. The methodology used provides a relative comparison and changes to metabolites were not quantified. The focus of this work was on correct metabolite identification by removing metabolites that were not reproducibly identified in the quality control samples and by ensuring reasonably reliable identifications. Overall there were changes to numerous metabolites from different pathways including increases to glycolysis mediators indicating intracellular accumulation of glucose, markers of ketosis indicative of improper utilisation of intracellular glucose and changes to the amino acid profile that may emphasise how systemic changes such as accumulation of BCAAs, may affect vision by limiting the transport of neurotransmitter precursors. Novel changes were also observed including a decrease in scyllo-inositol but not myo-inositol in diabetic retinas. This is indicative of a change to a specific isoform that may produce as of yet unknown downstream signalling consequences. The basic functions of scyllo-inositol should be elucidated to determine the consequences of this change.

134

Chapter 5 | Mapping Lipid Changes in the Rat Retina during Hyperglycaemia

135

5.1 Introduction In the previous chapter, the effect of diabetes on polar metabolites including sugars, polyols and amino acids in the retina was investigated. The extraction process used also allowed the isolation of lipids from the same retinas for analysis by GC-MS. It is well established that diabetes induces abnormalities in plasma lipid profiles including elevated levels of triacylglycerols (TGs) and cholesterol/cholesterol esters (CEs). Dyslipidemia is a common consequence of diabetes but varies between type 1 and type 2 patients; type 2 diabetics present with high circulating TG levels and decreased high-density lipoprotein (HDL) cholesterol whereas type 1 diabetics may present with elevated TGs, but HDL cholesterol levels tend to be normal or even high if the patient has good glycaemic control403,404. In T2DM, the period prior to development of hyperglycaemia is characterised by risk factors for cardiovascular disease including possible obesity, insulin resistance, hypertension and dyslipidemia403. Elevations in TGs and CEs occur because of a lack of functional insulin signalling increasing lipolysis and excess fatty acids are produced and released storage depots; they are then transported to the liver where they increase lipid-transport molecules such as very low-density lipoprotein (VLDL) and apoprotein B-containing lipoproteins, thereby increasing the content of TGs and cholesterol in the circulation405,406.

With relation to diabetic retinopathy, elevated serum lipids increase the likelihood of hard exudate deposition in the diabetic retina407. Elevated serum cholesterol levels also predict the severity of retinal hard exudate accumulation408. Lipid-lowering treatment may also prevent or ameliorate vision loss from diabetic macular oedema409. Whilst these studies have shown the correlation of circulating lipids with progression of diabetic retinopathy, relatively little is known about the effect of diabetes on endogenous lipids within the retina. Therefore, the goal of the studies reported in this chapter was to elucidate changes in the lipid profile of retina using the STZ-rat model of diabetes and to shed further light on how lipid-based mechanisms might contribute to diabetic retinal damage.

Variation in lipid class structures, the main classes of which that are discussed in this chapter are shown in Figure 5.1, equates in part to variable polarity. For example, TGs have three fatty- acyl chains; therefore, they are expected to be less polar than glycerophospholipids such as the phosphatidylcholines (PCs), which only have two fatty-acyl chains. Both the phosphate and the glycerol groups are polar, as observed by the identification of glycerol in the previous chapter, which described polar metabolites in diabetic retinas. Free fatty acids and single-chain- containing glycerophospholipids, or lyso-phospholipids, are also more polar than the aforementioned lipid classes, because the polarity caused by the phosphate and glycerol groups is more proportionally balanced by fewer fatty acid chains. The length of these chains compared with hydroxyl groups and the issue of polarity means that these cannot be assessed using the TMS-derivative GC-MS method from the previous chapter. Therefore, a different method was utilised for this analysis whereby lipids were separated using liquid

136

chromatography (LC) based on changing polarity throughout the run. This was done by using a mobile phase comprising ramps containing varying ratios of water:methanol.

An untargeted lipidomics approach, as followed here, allows for the analysis of hundreds to thousands of individual lipid species that may be valuable to assess perturbations of lipid metabolism in disease. Similar to the method described in the previous chapter, lipid characterisation is dependent on ionisation efficiency and fragmentation patterns. A high resolution is required to differentiate mass (typically to 0.01 decimal place) to allow the correct identification of lipids and provide confidence in the results obtained. However, even with this degree of mass accuracy, it may not be possible to differentiate certain lipids.

To accurately assess lipids within a class, the ability to differentiate between carbon chains is required. Chains fragment in particular patterns that can be identified using higher energy collisional dissociation (HCD)410. Combining the data obtained from the accurately measured mass of the intact molecule and chain fragments following HCD can allow the precise identification of specific lipids. Furthermore, by analysing both positive and negative modes different adducts are introduced which allows the further discrimination between lipids of the

137

Figure 5.1 | Main Lipid Class Structures Discussed in this Chapter. An example of each of the main lipid class structures is shown with the main structural differences highlighted in red. Whilst only the triacylglycerols are represented in the above diagram, the acylglycerols also comprise the monoacylglycerols and diacylglycerols, which have only one and two fatty acyl chains respectively as opposed to the three shown for the triacylglycerol example. Cer, ceramide; Galcer, galactoceramide; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PI, phosphatidylinositol; PG, phosphatidylglycerol; TG, triacylglycerol. 138

same or very similar mass. For example, a phosphatidylcholine may not produce a distinctive pattern in positive mode and identification is reliant on a library search; however, in negative mode, the loss of both a formate and methyl adduct is a distinctive characteristic of a PC thereby confirming the identification411.

The high-resolution detection required for such analyses can be performed using an Orbitrap mass analyser as shown in Figure 5.2410. Upon exit from the liquid chromatogram, samples undergo electrospray ionisation. During this process, the liquid samples, dissolved in volatile solvent, pass into the capillary which has a high voltage (positive or negative, 3-5kV) applied to the tip causing the formation of a charged droplet. The droplet is forced out of the capillary as a fine aerosol spray of charged droplets. The fine spray is created because of increased surface tension from charged particles repelling one another, eventually causing them to separate into smaller droplets until charged ions are formed. The ions are transferred to the S-lens via an ion transfer tube using radio frequency. The S-lens has plates that decrease in orifice size to focus the ion beam. The ion beam is then focused by multipoles to the ion trap. The pressure decreases from atmospheric pressure in the electrospray ionisation chamber, gradually to <1x10-5 Torr in the octopole chamber, drawing the ions into the octopole and reducing the transport of neutral particles, thereby reducing noise410. The gate lens controls the rate of ion entry into the ion trap. All ions enter the high-pressure cell and are retained until transfer to the low-pressure cell to be ejected into the detection system. Ions are then focused via a multipole into the C-trap. Here, the ions are stored in the bent quadrupole

139

Figure 5.2 | A Schematic Illustration of an Orbitrap Velos Mass Analyser Samples are passed from the liquid chromatogram to the electrospray ion source, where the analytes are ionised and the solvent evaporated before entry into the mass spectrometer. They are then transferred to the S-lens via the ion transfer tube. The S-lens focuses ions into a tight beam by using radio frequency. The ions are then transmitted from the S-lens to the ion trap through the multipoles. This process allows any remaining neutral atoms or ions of opposite polarity to be lost in the vacuum. The gate lens controls the number of ions entering the ion trap, within which ions are retained and then ejected through a multipole to the C-trap. The C-Trap ejects ions into the Orbitrap in short pulses, or the higher energy collisional dissociation (HCD) cell for MS/MS analysis. Within the Orbitrap, the centrifugal force generated by the velocity of the ions forces them into orbit around the electrode. These oscillations are converted into a frequency that is assigned to a specific m/z, producing a mass spectrum. Image was adapted from the LTQ Orbitrap Velos™ Hardware Manual from Agilent Technologies410. 140

of the C-trap410,412,413. When the rf is ramped down and a high voltage is applied across the trap, the charged ions are released into the Orbitrap analyser413. Within the analyser, the ions are trapped by a strong electrostatic field that pushes them towards the central electrode, which they circulate due to centrifugal force. The oscillations are transformed into frequencies that can then be designated a specific m/z, thereby producing a mass spectrum413. Where tandem mass spectrometry (MS/MS) experiments are done, prior to entry into the Orbitrap, the ions are first transported into from the C-trap into the HCD cell. The voltages applied increase the kinetic energy of ions and therefore increase the number of collisions with the gas in the cell causing fragmentation of the ions. These fragments are transferred back to the C-trap and shuttled into the Orbitrap as previously described. This results in high-resolution mass spectrometry and when combined with MS/MS is ideal for fatty acid chain identification.

This chapter describes the use of liquid-chromatography mass spectrometry (LC-MS) with Orbitrap technology and MS/MS to produce a map of lipids and their classes in the rat retina and how they change in STZ-induced diabetes. It provides novel insights into the perturbation of lipid metabolism in the retina during diabetes and highlights how individual lipid classes are affected by diabetic retinopathy.

141

5.2 Methods

5.2.1 Animal Studies This experiment used retinal tissue from the 12-week STZ Study 4 that is described in Chapter 2 (and in the preceding chapter the GC-MS analysis of the same retinas is referred to as GC- MS Experiment C).

5.2.2 LC-MS/MS Tissue was processed as described in Chapter 3, to the point of chloroform extraction. 100 µL of the extracted chloroform was aliquoted each for positive and negative run samples. Another 100 µL from each sample was pooled to create a quality control (QC) solution. 100 µL of this pool was aliquoted for each QC. All samples and QCs were dried at 30 °C for 1 hour at ramp 1 followed by 1.5 hours at ramp 5 in a Speedvac centrifugal concentrator (Thermo Scientific SPD331DDA). Samples were stored at 4ºC until required. Samples and QCs were reconstituted in 200 µL of 80:20 methanol:chloroform for positive mode analysis and 120 µL for negative mode analysis when required.

For both positive and negative mode analysis, a conditioning sequence of 10 QC injections was followed by the experimental samples in randomised order, interspersed with further QC injections at every fourth injection. Following this sequence, further injections of the QC pool were done to conduct low and high energy HCD MS/MS.

Chromatographic separations were performed on a Hypersil GOLD column (100 x 2.1mm, 1.9 µm; ThermoFisher Scientific, Runcorn, UK) operating at a column temperature of 50°C. Two solvents were applied (solvent A comprising 5mM ammonium formate in water, and solvent B comprising of 5mM ammonium formate in MeOH) at a flow rate of 400μl/min. Solvent A was held at 80% for 1 minute, then a linear gradient to 80% solvent B was applied over 7 minutes. Next linear gradients were applied to increase solvent B to 90% over 17 minutes, then to 95% solvent B over 15 minutes. Finally, solvent B was increased in a linear gradient to 100% over 5 minutes and held at 100% for a further 60 minutes (eluting neutral lipids such as triglycerides, diglycerides, cholesterol esters, etc.). The column was then re-equilibrated at 80% solvent A for 5 minutes. The column eluent was introduced into the Orbitrap Velos mass spectrometer and full-scan profiling data acquired calibrated to 3 ppm, at a mass resolution of 30,000, and m/z=400m, in accordance with the manufacturers’ recommended procedures. The mass range scanned was from 100-2000 Da and this step is referred to as MS1.

For data-dependent MS/MS, two QC samples were used. Full scan MS spectra were acquired (mass range 100-2000, resolution 30000 as previously). The four most intense ions were selected and fragmented in the HCD collision cell (normalised collision energy of 30% and 60%) and product ions were detected in the Orbitrap.

142

5.2.3 Peak picking, alignment and annotation XCMS (October 2013 version), using its CentWave algorithm, and CAMERA (October 2013 version) were utilised for initial data reduction, running under R (version 3.0.3)414–416. The procedure processes the experiment as a batch, aligning features across samples and QCs using a script shown in Table 5.1. This includes additions listed in Table 5.2.

5.2.4 Metabolite identification and data reduction for comparative statistical analysis The annotated output from CAMERA was inspected manually, initially inspecting the dataset derived from positive ion mode followed by that derived from negative ion mode. To allow for ions generated by adduct formation, we searched masses against the LipidMAPS database (at 0.01 Da tolerance). Candidate identifications were obtained from both positive and negative modes. These data were then merged to compile a list of all candidate identities and to confirm the identification of lipids that are detected in both modes. The merged data made it possible in many cases to specify lipid class, the sum-total fatty acid chain length, and the total number of double bonds in the attached chains. Characteristic fragments also allowed the definition of the fatty acid chains in some of the TGs and DGs. Information on positions of double bonds within chains or isomeric forms cannot be identified with this method. Lipid class and chain information could be identified using the original mass obtained in low-energy positive mode and high- energy negative mode respectively in the MS/MS data.

The two-dimensional feature maps previously described417 were informative, and used to confirm that identifications in a particular lipid class were internally consistent with each other.

5.4.5 Statistical analysis Data were analysed as described in Chapter 4. Briefly, base peaks were loaded into SIMCA-P+ (version 12.0.1.0 Umetrics AB) to assess data quality, variability and class separation. Data were further analysed using GraphPad prism (version 7.00 GraphPad Software Inc.). QC means and %CV were assessed and lipids were removed from the dataset if they were found to have >25% CV in the QC group418. Multiple t-tests were then done between non-diabetic and diabetic groups without assuming a consistent standard deviation. Discovery was determined using the Original false discovery rate (FDR) method of Benjamini and Hochberg419, with Q=10%. This method corrects for false discoveries by penalising the p-value to produce a corresponding adjusted p-value, or “q-value”. Fold-changes were also calculated and compared between groups.

143

library(Hmisc)

library(xcms)

library(CAMERA)

xset <- xcmsSet(method="centWave",ppm=3,pe m/z peak finding using “centwave” algorithm, akwidth=c(5,20), snthresh=25, prefilter=c(3,10000)) with 3ppm tolerance on mass, acceptable LC

peak widths 5-20 s, a S/N threshold of 25, and default prefilter setting. These settings reflect the grp <- group(xset,bw=1,minfrac=0.5,minsamp=1 expected performance of the hardware (Orbitrap ,mzwid=0.005) calibrated to 3 ppm, Hypersil GOLD column). an <-xsAnnotate(grp) Initial rough grouping (mass tolerance 0.005 Da), test for feature presence in “minfrac” of batch anF<- groupFWHM(an) samples (50% as set)

Refined grouping based on chromatographic anI <-findIsotopes(anF, mzabs=0.02) peak width similarity

Assignment of peaks into isotope clusters nIC <-groupCorr(anI, cor_eic_th=0.90) Rigorous grouping, based on correlated changes

in feature intensity between samples

Annotation of potential adducts, neutral losses file <- system.file('rules/extended_addusts_paul and neutral additions within each group, using _pos.csv',package="CAMERA") customised lists rules <- read.csv(file) (“extended_adducts_paul_pos.csv”) in place of

anFA <-findAdducts(anIC, CAMERA’s default rule set polarity="positive", rules=rules) Assemble the output list and write to a comma- peaklist<-getPeaklist(anFA) delimited text file write.csv(peaklist,'kidneycortexnewrules. csv')

Table 5.1 | Annotated script in R for positive mode data reduction, using XCMS & CAMERA. The negative mode script was similar but used a lower S/N threshold to reflect cleaner, lower intensity data, and appropriate substitutions in the “file” and “FindAdducts” commands to access negative mode rule tables

144

Neutral Losses (positive mode) Mass loss (Da) Class Defining Fragments Phosphatidylethanolamine 141.019 Phosphatidylserine 185.009 InositolPhosphate 260.029 Fatty acid chains (primarily for TG and DG) 12:0 199.177 14:0 227.205 16:0 255.233 16:1 253.218 18:0 283.261 18:1 281.246 18:2 279.232 18:3 277.217 20:0 311.289 20:1 309.274 20:2 307.260 20:3 305.245 20:4 303.231 20:5 301.216 22:0 339.319 22:1 337.304 22:2 335.290 22:3 333.275 22:4 331.261 22:5 329.246 22:6 327.232 Water loss (Sphingomyelins and Ceramides) H20 18.0153 (H2O)2 36.0306 Fragments Choline Headgroup 184.073 Ceramide 264.268 Coenzyme Q 197.081 Cholesterol Ester 369.351 Negative Mode [M-CH3]- 15.0235 [M-H+HCOOH]- 44.998194 Table 5.2 | Characteristic fragments used to infer molecular weights. These assignments were useful in interpreting lipid spectra in MS1 and in confirming identifications in MS/MS. In some cases these were incorporated into custom rule tables for CAMERA, but the fatty acid losses were applied manually. In negative mode, the loss of both methyl and formate groups was indicative of a phosphatidylcholine moiety.

145

5.3 Results In total, 190 individual lipid molecules were identified after removal of metabolite candidates with poor reproducibility across the QC group. These identifications were from several classes including various glycerol phospholipids as will be discussed below; these were sphingomyelins, ceramides, free fatty acids and cholesterol esters, a cholesterol molecule with a fatty acyl chain bound. A representative example of a chromatogram obtained from a QC sample is presented in Figure 5.3. This shows the lipid classes identified by the methodology applied in this study,

5.3.1 PCA analysis PCA analysis indicated a strong separation of classes with tight QCs as shown in Figure 5.4. This shows reproducibility across the QC group with a slight drift that similarly occurs across the sample groups. Diabetic and non-diabetic groups were clearly separated with no overlap indicative of an effect of diabetes across the metabolites identified. This plot indicated that the difference between classes was the main source of variation.

5.3.2 Verification of Identifications MS/MS analysis indicated whether the appropriate chain had been identified by fragmenting the lipid. Where the fragmentation characteristics were not identified during LC-MS, obtaining the fragmentation patterns during MS/MS was informative of lipid class as described in the methods. After this a features map containing all lipids within their respective classes was created. Lipids of a particular class produce similar behaviour on a feature map and should be clustered when arranged by retention time and the inferred mass. The feature map for the lipids obtained in this experiment is shown in Figure 5.5 (A). The red arrow highlights a lipid molecule that was likely misidentified at this point. Within each particular class, additional carbon groups cause a consistent increment in mass and similarly, each double bond reduces the retention time on the chromatogram. When different lipids within a class are graphed according to inferred mass and retention time, in particular lipid classes, such as the phosphatidylethanolamines (PEs) for example, they should align parallel to those with a different number of carbon chains, and similarly should align in the opposite dimension according to the number of double bonds. The assessment of all identities within a lipid class produced a scatterplot such as the example shown for PEs in Figure 5.5 (B).

146

Figure 5.3 | Example of liquid chromatogram output (QC sample) of rat retinal lipids The chromatogram shows the relative abundance of lipids at different retention times. Lipid classes are determined by mass and the retention time. This chromatogram shows the retention times at which different classes may be identified.

147

Figure 5.4 | PCA plot for LC-MS experiment This plot shows the spread of the data by rat number with red symbols indicating the non- diabetic (ND) group and green indicating the diabetic (D) group. Quality control QCs samples are indicated by the + symbol. This plot indicates definite class separation with QC values between the two groups as expected. Class is the primary source of variation as component 1 and component 2 separate in this manner. QC values lie between the diabetic (N=10) and non- diabetic (N=12) groups as expected. D, diabetic; ND, non-diabetic; QC, quality control.

148

Figure 5.5 | Feature Map of Lipid Identifications. This graph indicates where different lipid classes lie with regards to retention time and mass (A). Any identity that falls out of place with the class may be considered an outlier or to be of questionable identity. The feature map shown in B is specific to one class, in this case the PEs. The lines joining identifications with the same chain length (dotted blue) or double bonds (dotted red) should be parallel if all identifications are correct as the mass and retention time will change by a similar magnitude with the loss or addition of double bonds in lipids with the same sum-total chain length. DG, diacylglycerol; FA, fatty acid; HexCer, hexose-bound ceramide; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PI, phosphatidylinositol; PG, phosphatidylglycerol; SM, sphingomyelin; TG, triacylglycerol.

149

5.3.3 Changes to Lipids in Diabetic Retinas A full list of individual lipid identifications is supplied in Appendix IV. Of the 190 identified lipids, 95 had a ≥20% fold-change in diabetic retinas compared with non-diabetic and of these 86 were significantly changed as assessed by multiple t-tests with FDR protection. The effects of diabetes on the various lipid classes are described below.

5.3.4 Phosphatidylcholines and Phosphatidylethanolamines Phosphatidylcholines (PC) and PEs were the most abundant two lipid classes in rat retinas, with 22 different lipids identified in each class. Eight identified PCs were significantly upregulated and two were downregulated in diabetic retinas relative to non-diabetic. The proportion of lipids changed and the fold-changes for these individual lipids are shown in Figure 5.6. Whilst most lipids were upregulated, it should be noted that PC(22:6_22:6) was downregulated by 47.7%. PC(14:0_22:6) was also downregulated (24.5%). Additionally, six lysoPCs, or PCs with a single fatty-acyl chain were identified. One of these, lysoPC(22:6_0:0) was downregulated in diabetic rat retinas compared with non-diabetic. Other PC variants were also identified; the alkyl ether PCs (PC(O)) and plasmalogens or vinyl ether PCs (PC(P)). These PCs are structurally linked to the glycerol phosphate on one chain by an ether group on the first carbon as opposed to the ester linkage on the previously described PCs. This method could not discriminate PC(O)s from PC(P)s and so they are named as “PC(P-chain) or PC(O-chain)”. Three ether-linked PCs were identified of which one was upregulated by 40% and one was downregulated by 20%; PC(P-14:0_20:4)/PC(O-14:1_20:4) and PC(P-34:0)/PC(O-34:1) respectively, both q<0.001.

Of the PEs, six were upregulated and three were downregulated in diabetic rat retinas relative to non-diabetic controls. The proportion of lipids changed and the fold-changes for individual lipid chains are shown in Figure 5.7. Of six identified lysoPEs, one, lysoPE(22:5_0:0) was upregulated in diabetic compared with non-diabetic rat retinas by 122% (q<0.0001). Fifteen ether-linked PEs (PE(P) or PE(O)) were identified, of which six were changed. One of these, PE(O-34:3)/PE(P-34:2) was upregulated in diabetic rat retinas relative to the non-diabetic group. The remaining changed ether PEs were downregulated as shown in Figure 5.7. In both ester-linked and ether-linked lipid groups, the lipids with longer chain lengths tended to be downregulated in diabetic retinas compared with non-diabetic.

150

Figure 5.6 | The Effect of Diabetes on Phosphatidylcholines. LC-MS analysis of diabetic rat retinas (N=10) relative to non-diabetic controls (N=12) showed that nearly half of the lipids in the class were changed. Most of the changed lipids were upregulated in the diabetic samples except for PC(14:0_22:6), PC(22:6_22:6), LysoPC(22:6) and the ether-linked PC(34). All lipids were statistically significant as assessed by multiple t- tests, penalised with a 10% FDR. All lipids shown were q<0.05. For ease of viewing, the fold changes are coloured green to red by fold change as defined in the key to the left. Please note that the pie chart does not include lyso PCs or ether-linked PCs due to the small number of identifications.

151

Figure 5.7 Changes to phosphatidylethanolamines in diabetic rat retinas Approximately 41% of ester-linked PEs were changed in diabetic rat retinas compared with non- diabetic controls. Most of these were upregulated, as was the only changed lysoPE. Conversely, most of the ether-linked PEs were downregulated in diabetic rat retinas (N=10) relative to non-diabetic (N=12). All lipids were statistically significant as assessed by multiple t- tests, penalised with a 10% FDR. All lipids shown were q<0.05. For ease of viewing, the fold changes are coloured green to red by fold change as defined in the key to the left. Please note that the pie charts do not include lysoPEs due to the small number of changes (1 of 6).

152

5.3.5 Phosphatidylglycerols Phosphatidylglycerols (PG) were much less abundant than the PCs and PEs. Six lipids were identified in this class of which five were significantly changed with a >20% fold-change. All of these were downregulated except for PG(36:2), which was upregulated 55.4% (q<0.05). The proportion of lipids changed in diabetic compared with non-diabetic rat retinas and the individual changes are shown in Figure 5.8.

5.3.6 Phosphatidylinositols Twelve phosphatidylinositols (PI) were identified in the current study; of these, four were upregulated and six were downregulated in diabetic rat retinas relative to non-diabetic retinas (Figure 5.9). When the fold-changes were assessed by total chain length, it was apparent that lipids with longer chain lengths were preferentially downregulated with the largest magnitude change observed in PI(44:12).

5.3.7 Cholesterol esters In this study in rat retina, cholesterol was identified and three members of the cholesterol ester class were identified; these were the cholesterol esters with 18:2, 20:4 and 22:6 fatty acyl chains. Cholesterol itself remained unchanged in the diabetic rat retina compared with non-diabetic controls (q>0.1). Of the three cholesterol esters, two were significantly changed in diabetes. The 18:2 cholesterol ester was upregulated 2-fold (q<0.0001) and the 22:6 cholesterol ester was downregulated by ~50% in diabetic retinas relative to non-diabetic controls (q<0.01).

5.3.8 Fatty acids The free fatty acids identified were as follows: 16:0 (palmitic acid), 18:0 (stearic acid), 18:1 (oleic acid), 20:4 (arachidonic acid), 22:6 (docosahexaenoic acid (DHA)), 32:6 and 34:6 (very-long chain polyunsaturated fatty acids (VLCPUFA)). None of the free fatty acids were changed in diabetic rat retinas compared with non-diabetic controls.

153

Figure 5.8 | Changes in phosphatidylglycerols in diabetic rat retinas Over 80% of PGs were changed in diabetic rat retinas (N=10) compared with non-diabetic (N=12). All but one were downregulated as shown. All lipids were statistically significant as assessed by multiple t-tests, penalised with a 10% FDR. All lipids shown in the table had q<0.05 values. For ease of viewing, the fold changes are coloured green to red by fold change as defined in the key to the left.

154

Figure 5.9 | Mapping phosphatidylinositols changes in diabetic rat retinas Identified PIs were compared between diabetic (N=10) and non-diabetic controls (N=12) rat retinas. Most PIs were changed in diabetes and the majority of those changed were downregulated. Given the mix of upregulated and downregulated PIs, these were assessed for a pattern by chain length and double bond number. Upregulated PIs tended to have shorter chain lengths than downregulated PIs. All lipids were statistically significant as assessed by multiple t-tests, penalised with a 10% FDR. All lipids shown had values of q<0.05. For ease of viewing, the fold changes are coloured green to red by fold change as defined in the key to the left.

155

5.3.9 Sphingomyelins and Ceramides Nine sphingomyelins were identified and most remained unchanged. One sphingomyelin, SM(d18:1_22:0), was significantly downregulated by 21.4% in diabetic rat retinas compared with non-diabetic (q<0.05). Eleven ceramides were identified as well as nine ceramides bound to a 6-carbon sugar, or hexose ceramides (HexCer), of which two had two 6-carbon sugars bound, being 2-hexose ceramides (Hex2Cer). Of the unbound ceramides, three were changed in diabetic retinas relative to non-diabetic controls; one was upregulated by 32% (Cer(d36:2) and two were downregulated by 27 and 25% (Cer(d18:1_20:0) and Cer(d18:1_22:0) respectively) as shown in Figure 5.10 (A). Of the sugar-bound ceramides, all but one were upregulated between 57–174% (Figure 5.10 (B)).

5.3.10 Mono-, Di- and Triacylglycerols Of the mono-, di- and triacylglycerol classes, eight, five and 29 members were identified respectively. The mono- and diglycerides were mostly unchanged except for one diacylglycerol, DG(18:0_22:4_0:0) that was upregulated by 21% (q<0.05). Conversely, the triacylglycerols were mostly changed as shown in Figure 5.11. Of 29 triacylglycerols measured, seven were upregulated and 15 were downregulated in diabetic rat retinas compared with non-diabetic controls. Of the upregulated lipids, these tended to be polyunsaturated lipids with total chain lengths between 52 and 54 carbons and all contained at least one 18:2 chain with the possible exception of both identifications of TG(54:5) that had undetermined chains. The VLCPUFA TG(22:6_22:6_22:6) had the greatest magnitude of downregulation (72.9%, q<0.00001).

156

Figure 5.10 | Changes to ceramides and sugar-bound ceramides in diabetic retinas Of the identified ceramides, only three changed in diabetes (A). Of these three, two were downregulated and one was upregulated. Of the sugar-bound, or hexose ceramides, all but one were upregulated in diabetic rat retinas (N=10) compared with non-diabetic controls (N=12) including ceramides with two 6-carbon sugar entities bound. All lipids were statistically significant as assessed by multiple t-tests, penalised with a 10% FDR. All lipids shown were q<0.05. For ease of viewing, the fold changes are coloured green to red by fold change as defined in the key to the left.

157

Figure 5.11 | Triacylglycerol changes in the diabetic rat retina Over 75% (22/29) of identified TGs in the rat retina were affected by diabetes (N=10) compared with non-diabetic retinas (N=12). The majority of TGs were downregulated (>50%) and 24% were upregulated. When assessed by chain length, all were downregulated except the majority of those with a chain length between 52–56.TGs with a chain length >56 were again downregulated in diabetic retinas relative to mom-diabetic. All lipids were statistically significant as assessed by multiple t-tests, penalised with a 10% FDR. All lipids shown were q<0.05. For ease of viewing, the fold changes are coloured green to red by fold change as defined in the key to the left.

158

5.4 Discussion In this chapter, a novel method application has been presented whereby multiple specific members of the main lipid classes were identified in rat retina and compared between diabetic and non-diabetic control tissue. The identified lipids were structurally unmodified upon analysis and so it is expected that they have been observed in their endogenous forms. To our knowledge, this is the first study to describe changes in unmodified lipids at this level of detail across this range of classes in diabetic rat retinas. Most previous studies have focused only on the changes to levels of fatty acids in the diabetic rat retina and the effects of fatty acid supplementation 420–422. In the current study, the effect of diabetes on the abundance of lipids with (mostly) identified fatty acid chains was explored across different lipid classes. The measurement of lipids from the major lipid classes is an under-researched area. The study by Lydic et al., (2009) was, to our knowledge, the first to detect changes in the levels of individual lipid molecules of the PC and PE classes in diabetic retina423. The lack of chromatographic separation obtained by the methodology employed in that study, however, limited the range of detectable lipids, because as observed in the feature map shown in this study, chromatographic separation meant that lipids of similar mass were measured at different times thereby reducing background noise. The current study is the first to study these classes, among others, concurrently in a single sample using a methodology sufficient to allow separation and identification of large numbers of the members of each lipid class.

In normal physiological circumstances, insulin inhibits hormone-sensitive lipase thereby inhibiting degradation of stored triacylglycerols in adipose tissue7,424; therefore, it is logical that during hypoinsulinemia there is a shift to increasing lipid degradation leading to a ketogenic state. In the brain, lipoprotein receptor-related protein 1 (LRP1) regulates insulin signalling and glucose uptake and its expression is increased by insulin and suppressed by hyperglycaemia425. LRP1 is also an apoE receptor, a lipid transporter, that in retina is synthesised by Mueller glia and taken up by retinal ganglion cells426,427.Therefore a substantial change in lipid content in retina is perhaps unsurprising given the close relationship between circulating insulin and lipid metabolism.

5.4.1 The Extent of Lipid Identification and its Limitations A variety of lipids were identified in rat retina including PCs, PEs, PIs, PGs, mono-, di-, and triacylglycerols, cholesterol esters, fatty acids, sphingomyelins and ceramides as well as miscellaneous other lipids. In most cases, these were identified to the level of individual fatty acid chains. However, the position of each fatty acid chain could not be identified and so the naming convention of x:y_x:y has been used throughout, as opposed to x:y/x:y which is indicative of a known fatty acid moiety at a known position428. The exact position and geometry of double bonds also could not be identified and so this information was excluded from the nomenclature of identifications. Like the GC-MS analysis presented in Chapter 4, the range of detection limits the ability to identify less abundant molecules when highly abundant molecules

159

are present. For this reason, this study did not identify small eicosanoids or prostaglandins as phospholipids account for 90–95% of the rat retina429.

5.4.2 Diabetes Induces Substantial Changes in Glycerophospholipids Glycerophospholipids comprised the majority of lipid classes observed in this study. These lipids all contain a hydrophilic head, containing the phosphate group and the functional head group, such as choline in PCs and ethanolamine in PEs, and a hydrophobic tail comprised of up to two fatty acid chains. A glycerophospholipid with one fatty acid chain is referred to as a lyso phospholipid of the respective class. Their dual polarity structure makes them ideal cell membrane components and this is their main function; however, they may also serve as components of intracellular signalling pathways and have roles including protein trafficking in the cell membrane7. Here the changes to the PCs, PEs, PGs and PIs are described in diabetic retinas as observed in the presented study; their structures, for reference can be found in Figure 5.1.

5.4.3 Diabetes Leads to an Overall Increase in PCs in the Rat Retina PCs and PEs constitute major components of the lipid bilayer in the plasma membrane and therefore contribute to the structural integrity of cells. Of the PCs that were changed in diabetic retinas compared with non-diabetic controls, all were upregulated in diabetic rat retinas relative to non-diabetic controls with the exception of two PCs containing 22:6 fatty acid chains. Similarly the lysoPC(22:6_0:0) was also downregulated. Of the ether PCs, one with unidentified chains was downregulated and one containing a 20:4 chain was upregulated in diabetic rat retinas compared with non-diabetic controls. Unfortunately, the chains of most of the changed PCs could not be identified; this is likely attributable to the detection of similar masses being at approximately the same elution time thereby limiting the detection of these lipids at that particular time. This produces difficulties in drawing specific conclusions about the changes observed in the PC group. Saturated PCs comprise the main structural element of biological membranes, as their structure allows for the formation of an even bilayer. Unsaturated PCs increase the fluidity of the bilayer430. The PCs with identifiable chains that were changed in this study tended to have an unsaturated chain and therefore may affect fluid dynamics of cell membranes. Of those with unidentified chains, half had at least five double bonds in total, suggesting that these could possibly have had two polyunsaturated fatty acid chains.

PCs are synthesised initially by phosphorylation of dietary choline to phosphocholine, followed by the addition of a cytidine monophosphate group, donated by cytidine triphosphate, to create cytidine diphosphocholine that can then be added to a diacylglycerol to form phosphatidylcholine431. Alternatively, PEs can be methylated three times to produce PCs432. PCs can be metabolised to produce phosphatidic acid and are also a source of eicosanoids, such as leukotrienes and prostaglandins, by removal of the 20:4 chain, arachidonic acid, from the PC after which it is metabolised to produce eicosanoids433. Phospholipases C and D have been shown to hydrolyse PCs to release the lipid messenger, diacylglycerol (DG), and fatty

160

acids in synaptic endings during iron-induced oxidative stress434. PCs have also been implicated in activating PPAR receptors in hepatocytes, and compete with Insulin Receptor Substrate 2 (IRS2), an insulin effector that is attenuated in diabetes, for the PC binding site in the PC transporter phosphatidylcholine transfer protein435.

5.4.4 Ether-linked PEs and Ester-linked PEs were Changed in an Inverse Manner during Diabetic Retinopathy PEs were changed similarly to PCs, since; most PEs were upregulated in diabetic compared to non-diabetic rat retinas. The exceptions to this were PE(16:0_16:0), PE(20:4_22:6) and PE(22:6_22:6). The ether-linked PEs had a different pattern of changes to the ester-linked PEs and were mostly downregulated in diabetic rat retinas relative to non-diabetic controls. PEs are derived from dietary ethanolamine, by a process in which they are phosphorylated in the presence of ATP to produce phosphoethanolamine. After conversion to cytidine diphosphoethanolamine, this molecule combines with DG to produce phosphatidylethanolamine. As stated, PEs are plasma membrane components and whereas PCs will produce a flat bilayer, the presence of PEs produce kinks to bend the membrane436. PEs also have roles in the production of endocannabinoids437, providing ethanolamine for PI-anchoring to signalling proteins438, and possibly autophagy439. Specifically in retina, PEs in the lipid bilayer of photoreceptors interact with all-trans-retinal to form N-retinylidene-PE, whose function is to prevent potential toxic aldehyde interactions with all-trans-retinal. N-retinylidene- PE also augments the activity of lipid transporter ABCA4 in photoreceptors, mutations in the ABCA4 gene cause Stargardt disease440.

Ether-linked PEs, namely plasmalogens or vinyl-ether PEs, given the “P” prefix in Figure 5.7, have been shown to be protective against oxidative damage from free radicals in the rat brain and this may protect other lipid entities from oxidative damage441,442. As most of the ether-linked PEs were downregulated in diabetes compared with non-diabetic controls, it is possible that this was due to interactions with free radicals causing plasmalogen degradation. The presence of ether-linked PEs also increases rigidity and thickness of the plasma membrane. Cell plasma membranes contain ether-linked PE-rich areas which associate with certain proteins, so the decrease in ether-linked PEs in diabetic compared with non-diabetic retinas may impact on the protein content of cell membranes and/or protein trafficking to plasma membranes443. It has been shown that in plasmalogen deficiency, the concentration of ester-linked PEs increases as an adaptive measure444. This agrees with the PE data observed in the present study whereby there was a trend for upregulation of ester-linked PEs while there was a concomitant overall decrease in ether-linked PEs.

5.4.5 Diabetes Induced an Overall Downregulation of Phosphatidylglycerols In the presented study, the total number of PGs identified was much fewer than that of PCs and PEs. Unlike the previously described phospholipids, however, all but one of the identified PGs

161

were changed in diabetes. Of these, all were downregulated in diabetic retinas compared with non-diabetic controls with the exception of one PG. The individual chains of this lipid were not identified and so it is difficult to postulate why this particular lipid behaved differently from the rest of the class. Of the downregulated lipids with identified chains, these contained: 16:0, palmitic acid; 18:0, stearic acid; 20:4, arachidonic acid and 22:6, docosahexaenoic acid.

Much less is known about PGs compared with PCs and PEs. PGs are formed from cytidine disphosphate DG and glycerol-3-phosphate in mitochondrial inner membranes. The product is then dephosphorylated to phosphatidylglycerol445. PGs are best known as cardiolipin precursors. Cardiolipins are located almost exclusively in the mitochondria and are essential for mitochondrial function by interacting with inner mitochondrial membrane proteins and having roles in maintaining mitochondrial membrane fluidity and protein transport446,447. During diabetes, cardiolipins are susceptible to peroxidation and depletion in cardiac tissue447. In the retina, diabetes causes morphological changes in endothelial cell mitochondria so they become enlarged and show partial cristolysis, the latter being indicative of compromised oxidative phosphorylation448,449. Whilst cardiolipins were not detected in this study, it is possible that the downregulated PGs in diabetic retinas were metabolised to restore depleted cardiolipins. Cardiolipin content in diabetic retinas and mitochondrial dysfunction would need to be explored further to elucidate these changes in PG levels.

5.4.6 Phosphatidylinositols were Substantially Perturbed in Diabetic Retinopathy A large proportion of PIs (10 out of 12) were changed in diabetic retinas; there was an increase in shorter chain length PIs and a decrease in longer chain length PIs compared with non-diabetic retinas. Two of four upregulated longer chain length PIs contained 20:4 arachidonic acid chains, one of the others had unidentified chains and the other contained an 18:2 linoleic chain that is an omega-6 fatty acid along with a 22:6 docosahexaenoic acid chain that is omega-3. All downregulated lipids with an identified fatty acid chain contained a 22:6 chain.

PIs are formed when cytidine diphosphate DG reacts with inositol via CDP-diacylglycerol inositol phosphatidyltransferase, also referred to as phosphatidylinositol synthase, producing a glycerol backbone with two fatty acid chains and a myo-inositol head group. Phosphorylation of multiple hydroxyl groups in this head group leads to the production of PI derivatives that have roles in signal transduction and membrane structure and function450. In the previous chapter, it was found that glucose and sorbitol were substantially upregulated and scyllo-inositol, but not myo-inositol, was greatly downregulated in diabetic retinas relative to non-diabetic controls. In this chapter, an overall trend toward decreased PIs has been observed. The study by Nakamura et al (1992)451 suggested that sorbitol accumulation induced by high glucose may inhibit phosphatidylinositol synthase, thereby limiting PI synthesis. Whilst the previous chapter observed a specific downregulation of scyllo-inositol and not myo-inositol in diabetic retinas

162

compared with non-diabetic controls, it should be noted that scyllo-inositol is not incorporated into PIs452. An increased demand for myo-inositol may cause scyllo-inositol to be changed to myo-inositol to prevent myo-inositol deficiency. More work would be required studying the synthetic pathway of the PIs in diabetic tissue to determine the relationship between the types of inositol, PI synthesis and the effects of dysregulated glucose. This could determine whether downregulation of PIs in diabetic retinas is due to lack of synthesis or increased degradation.

PI derivatives are involved in numerous signalling processes such as the PI3K/AKT pathway that plays roles in the signalling of numerous mediators including PDGF, insulin, insulin-like growth factor-1, and nerve growth factor. Insulin uses this pathway to promote protein synthesis450. This pathway also mediates angiogenesis, proliferation and permeability in endothelial cells, all of which are implicated in diabetic retinopathy453–455. Insulin has been shown to reduce apoptosis in retinal neurons via this pathway456. Specifically in retina, PI3K concentration increases upon exposure to light in rod photoreceptors and seems to have an essential role in rod viability457. This pathway is involved in rhodopsin-mediated phototransduction458. This suggests that deficient PI concentrations may have downstream effects on phototransduction, retinal neuronal viability and endothelial cell growth.

5.4.7 Diabetes Alters Retinal Ceramide Composition In this study, only 3 of 10 ceramides were changed of which 2 were downregulated in diabetes; however there was a substantive increase in the number of ceramides bound to a six-carbon sugar. These ceramides are likely glucosyl or galactosylceramides, but our method cannot differentiate the two and so they are named here as hexosylceramides (HexCers). The HexCer group also contains the Hex2Cers whereby a mass was detected for two hexose molecules bound to a ceramide. Given the substantive increase in glucose observed in Chapter 4, an increase in sugar bound ceramides is likely due to excess intracellular sugars.

Ceramides are synthesised via multiple pathways including a multi-step pathway from serine and palmitoyl-CoA, a pathway that can be activated by oxidised low-density lipoprotein and cannabinoids459–461. Ceramides may also be synthesised from hydrolysis of sphingomyelin that produces ceramides and phosphocholines, a pathway that is induced by oxidative stress461,462. Complex sphingolipids may also be broken down to produce ceramide461. The results obtained in the present study showed that changed ceramides were mostly decreased relative to non-diabetic concentrations (2 of 3 significantly different changed ceramides). This trend agrees with the study by Fox e al (2006), where total ceramide concentration was significantly decreased in rat retina from 4 weeks of diabetes compared with non-diabetic controls463. Only one of the nine identified SMs were changed in the present study, indicating that the SMs overall were not affected by diabetes. This was also the case in the study by Fox et al (2006), where SMs were not significantly changed in diabetic retinas compared with non-diabetic controls463.

163

Ceramides can be modified by glucosylceramide synthase and galactosylceramide synthase to produce glucosylceramide and galactosylceramide respectively, also known as cerebrosides464. In the study by Fox et al (2006), glucosylceramides were also increased 4-weeks post STZ-induced diabetes463. The aforementioned study also localised glucosylceramide synthase to the neuronal plexiform layers and the outer segments of the photoreceptors but the expression was not changed between diabetic and non-diabetic controls. This study, however, did not measure enzyme activity that may be changed in diabetes. It has been shown previously in a model of spinal cord injury that lactosylceramide production is upregulated during TNF-α signalling that is thought to stimulate astrocyte proliferation and therefore astrogliosis465. Whilst the presented study could not determine that the Hex2Cers observed were lactosylceramide, this may give an indication of the effects of increased Hex2Cers.

5.4.8 Triacylglycerols but not Di- or Monoacylglycerols are affected by Diabetes In this study a substantive change was observed in the TGs, but not the members of the overall DG and MG classes. When assessed by total chain length, the shorter chain length TGs were decreased and the longer chain length (52-56 total carbons) increased in diabetic compared with non-diabetic rat retinas, but then the longest chain length TGs (above 60 total carbons) were decreased A similar trend was observed in the diabetic sciatic nerve although TGs with chains beyond 60 carbons were not reported364. Unpublished data from CADET has shown similar data in STZ rat liver but not kidney but again did not have identifications with a chain length >60 carbons. This suggests that the changes observed in retinal TGs are probably not specific to the retina but reflect the systemic TG modification with the exception of those containing the very-long chain polyunsaturated fatty acid (VLCPUFA). Whilst the majority of the changed TGs were downregulated, upon inspection of the upregulated TGs with identified acyl chains, a pattern emerged whereby they all contained at least one 18:2, linoleic, chain. On further inspection, none of the downregulated TGs contained a linoleic acid chain and the largest magnitudes of upregulation occurred in TG(18:2_18:2_18:2) and TG(18:2_18:2_18:3), the lipids with the most linoleic acid. Only one 22:6 or DHA-containing lipid was upregulated and this lipid also contained an 18:2 linoleic chain. Otherwise DHA-containing lipids were consistently downregulated, occurring most starkly in the TG(22:6_22:6_22:6) chain.

One previous study to our knowledge has investigated TG localisation in retinal layers466. Whilst that study only observed TGs in connective tissue outside the sclera, the tissue processed in the presented study was only extracted and washed retina and so the presented study shows TG content in the retina where a previous study suggested it was not present466. As the latter study used human eyes it is possible that there is species variation as shown in Fliesler and Anderson, (1983)429 or that the concentrations were too low for detection.

Upon ingestion of TGs, they are broken down in the duodenum. Hydrolysis of TGs by hormone- sensitive lipase and adipose triacylglycerol lipase leads to the production of DGs and MGs, after

164

which MGs are metabolised to free fatty acids and glycerol by monoacylglycerol lipase467. Upon entry into the enterocytes, lipid components migrate to the endoplasmic reticulum where fatty acids are converted to fatty-acyl derivatives by fatty-acyl-CoA synthetase. These fatty-acyl moieties and MGs are converted back to TGs by the enzyme complex TG synthase. The current study showed a substantial change in members of the TG class in the retina during diabetes but no change in any measured MGs or free fatty acids. This is likely because DG, MG and free fatty acid contents in the retina are low (and previously thought negligible), with free fatty acids comprising 1% of total retinal lipids429.

The very-low-density lipoprotein receptor (VLDLR) enables the cleavage of long-chain fatty acids on TGs by lipase for fatty acid β-oxidation in the retina. Deficiency of this receptor results in reduced energy production. This receptor is highly expressed in photoreceptors and when knocked out, it increases the expression of free fatty acid receptor 1 (Ffar1). Ffar1 modulates Glut1 expression, and reduces it when there are increased circulating fatty acids468. It is highly likely that these receptors are modified in diabetes given the well-established increase in circulating lipids and triacylglycerols; therefore it is possible that diabetes-induced changes to the expression of these receptors may modify retinal TG content, amongst other lipids, leading to this downregulation of TGs.

5.5.9 Diabetes Causes a Distinct Change in Retinal Omega-6 and Omega-3 Fatty Acid Content An overall trend that was observed across all lipid classes was downregulation of 22:6 chain containing (or DHA-containing), lipids across all classes in diabetic rat retinas compared with non-diabetic controls. This was most apparent in the lipids with only DHA fatty acids as they were consistently the most downregulated in diabetes. Notably, there was a trend towards an increase in 18:2 chain-containing, or linoleic acid-containing, lipids across all classes. As DHA is a well-known omega-3 fatty acid and linoleic acid is an omega-6 fatty acid, this suggests a change in the omega-6 to omega-3 ratio.

Whilst most fatty acids are derived from the diet, some de novo synthesis does occur in an insulin-induced reaction process involving acetyl-CoA carboxylase and fatty acid synthase to generate 16:0 fatty acid chains (palmitate)469. Two 18-carbon fatty acids are required in the diet as they cannot be synthesised de novo; that is, they are nutritionally essential. These are the 18:2 omega-6 fatty acid, linoleic acid, and the 18:3 α-linolenic omega-3 fatty acid470. These fatty acids are lengthened through an elongation process, which occurs in the endoplasmic reticulum. Elongation occurs by the successive addition of two-carbon units from malonyl-CoA to fatty-acyl-CoA by enzymes termed elongases. The elongase family consists of the six elongases, which are responsible for elongation of very long chain fatty acids proteins, abbreviated as Elovl1–Elovl6. Elovl1 and Elovl6 catalyse the elongation of saturated and monounsaturated fatty acids; Elovl2 and Elovl5 convert 18:2 linoleic acid and 18:3 α-linoleic

165

acid to 20:4 arachadonic acid and 22:6 DHA. Elovl4 elongates a range of fatty acids and Elovl-3 is present only in brown adipose tissue469.

In retina, Elovl4 is the most abundantly expressed elongase. A previous study found the Elovl4 mRNA expression is the most downregulated elongase in diabetic rat retinas471. Elovl4 deficiency caused by ELOVL4 mutation leads to Stargardt-like macular dystrophy and retinal degeneration472. Whilst Elovl4 is not involved in DHA synthesis 473, it is involved in the synthesis of VLCPUFAs, specifically the synthesis of fatty acids with >26-carbon chains471, of which only the free fatty acids were putatively identified in this study (Appendix IV). Elovl2 and Elovl6 mRNA expression was also downregulated in diabetic retinas compared with non-diabetic controls. Elovl5 transcription was unchanged in that study; however the data from the diabetic group was highly variable471. As previously mentioned, Elovl2 and Elovl5 are responsible for the production of the 22:6 carbon chain from the 18:2 and 18:3 chains and may have a role in the downregulation of DHA observed471. This observation is strengthened by the observed retention of the 18:2 chains in the present study. More research is required to determine the changes of elongases in the diabetic retina and whether they specifically are driving forces behind lipid changes or if these changes occur solely (or mainly) in response to increased circulating glucose.

It has been shown that DHA protects photoreceptors against oxidative stress-induced upregulation of ceramides by enhancing ceramide glucosylation474. It is possible that DHA-containing lipids are broken down to free the 22:6 carbon chain for protection against oxidative stress leading to upregulation of hexose-bound ceramides, such as those observed in the current study. Docosanoids, the oxidised lipids produced by oxygenase-catalysed modification of DHA, are known to possess anti-inflammatory and pro-survival properties124. The downregulation of omega-3 chain-containing lipids suggests that supplementation of omega-3 fatty acids may restore some of the lipids that were downregulated in diabetic retinas relative to non-diabetic controls. Supplementation with omega-3 fatty acids must be done carefully however, because overdosing in STZ rats caused acceleration of diabetic retinopathy as observed by increased capillary occlusion compared with diabetic rats without omega-3 supplementation475. However when supplemented with a recommended the 10% caloric intake, omega-3 fatty acid supplementation may be beneficial in inflammatory diseases124.

5.5.10 Conclusions and Future Studies Whilst the changes to individual lipid classes have been described in this work, in reality these modifications are all intertwined. For example, PCs constituted the class with the largest number of identified lipids observed in rat retinas. These can be broken down by phospholipase D to produce DG and choline. Alternatively, phospholipase D may also exchange the choline head group for an inositol molecule thereby producing a PI from a PC. Subsequent metabolism of PIs by phospholipase C produces inositol phosphates and DG476. The production of DG leads to protein kinase C (PKC) production that has been implicated as a therapeutic target in diabetic

166

retinopathy as described in Chapter 1. LysoPC, produced by the release of a fatty acid chain via

476 catalysed by phospholipase A2, potentiates the production of PKC from DG . Therefore, it is important to observe the effects of diabetes on multiple lipid classes, as was done in the current study, to gain insights into the concurrent outcomes in these classes and their relationships.

This study, however, is limited in certain aspects. Firstly, the small volume of tissue obtained from a single rat retina may limit the yield of identifiable lipids due to low concentrations. Whilst nearly 200 lipids were identified from a single rat retina, both retinas from the same animal, when processed as a single sample, could increase the number of identities and may aid the identification of cardiolipins. This would require an initial pilot study in non-diabetic rats prior to the commencement of a diabetes study. It must be stated however that this method using an Orbitrap Velos obtained a surprisingly abundant output of novel and insightful data.

Secondly, lipid identifications were dependent on the library stored on the LipidMAPS database. As discussed in the previous section, the retina is capable of producing VLCPUFAs due to specific expression of Elovl4 that is not found elsewhere in the body. This means that lipids containing VLCPUFAs are not recorded in the LipidMAPS library and therefore not detected in this study. Further investigation would be required to elucidate the classes that these VLCPUFAs are incorporated into and how they are affected by diabetes.

The level of identification obtained allows for identification of the chain length and number of double bonds in the fatty-acyl chains. This level of identification, however, does not allow confirmation of the double bond position in the chain. Therefore, in cases such as the 18:3 chain where two physiological isomers are known to exist, of which one is an omega-6 fatty acid and the other an omega-3, the exact isomer present cannot be confirmed, and a mixture of both is a distinct possibility. This makes interpretation of some of the data difficult; however in most cases, including DHA where only isomer is known to exist, clear interpretation is possible.

Lastly, it would be beneficial to investigate how changed lipids are affected by induction of normoglycaemia using insulin implants. In clinical diabetes, diagnosed patients will have some level of blood glucose control that contrasts with this study wherein subjects had consistently elevated blood glucose. It would be of interest to investigate which lipids are normalised in insulin-treated rats to better reflect the lipid profile of patients with well-controlled diabetes. If no trends were found or negligible amounts of lipids were changed this would still be telling, suggesting that diabetes-induced perturbation of lipid metabolism is completely modulated by (lack of) circulating insulin. In addition, comparison between retinal changes in diabetes causes mainly by lack of insulin, as in the current study, and those in animals with predominant insulin resistance, may also be informative.

5.5.11 Overall Conclusion This study has produced two important novel datasets that provide insights into rat retinal lipid composition and how this composition is changed by uncontrolled diabetes. A detailed map of

167

lipids from the main lipid classes, mainly consisting of glycerophospholipids and sphingomyelins has been mapped to the level of identifiable fatty-acyl chains in most lipid identifications. Unfortunately, we could not identify the position of the fatty acid chain in the respective backbone; nor could we determine the position(s) of double bonds in the carbon chains (where applicable). This information would provide even more confidence in identifications and information about the lipid profiles; however, despite lacking this information, substantial changes to the lipid profile were still observed in diabetic retinas. Despite these limitations, this is the most detailed description of lipid identifications in rat retina (to our knowledge) compared with the current literature423,463,477. Furthermore, whilst quantified concentrations of these lipids could not be calculated, the changes to these lipids in diabetes have been effectively mapped with many more identifications than currently present in the literature related to the diabetic rat retina 423,463. Little is known currently about the roles of the individual lipid signalling functions in the retina, but this is an exciting area of unexplored research that may potentially produce new targets relevant to the prevention and/or treatment of diabetic retinopathy, and further to elucidate the functional role of lipids in health and disease.

168

Chapter 6 | Overall Conclusions and Future Studies

169

6.1 Overall study results In this thesis, untargeted mass spectrometry approaches were utilised to investigate changes to retinal metal homeostasis, along with polar and non-polar metabolites in diabetic retinas. This has resulted in the generation of large amounts of data that could provide a basis for future studies. Notable outcomes from the research include the selective increase in copper levels in the human diabetic retinas compared with non-diabetic retinas that was corroborated in the rat retina with changes to the mRNA expression of copper-related mediators. The identification of major perturbations in both polar and non-polar metabolites in STZ rat retinas compared with non-diabetic retinas. Of the metabolite changes, downregulation of scyllo-inositol in diabetic retinas relative to non-diabetic consistently in three separate in vivo experiments stood out as a completely novel observation alongside most of the lipid discoveries; a dataset that is completely novel. The range of metabolites identified and discovered to be changed during diabetic retinopathy within a single sample, including observing both polar and non-polar metabolites from Study 2, has not been previously achieved and provides insight into concomitant changes to metabolites. Whilst in depth discussion were provided earlier, here we summarise how the observed changes to metal homeostasis and metabolites may be interrelated, limitations of the work completed and future work that could be undertaken.

6.2 Potential interrelationships between copper metabolism, polar metabolites and non-polar metabolites.

6.2.1 Copper overload in diabetic retinas made affect energy production The first major finding of this study was the observation that copper is increased in human donor diabetic retinas whilst other measured transition metals remained unchanged. Copper has a major role in energy production, being a major component of complex IV of the electron transfer chain, cytochrome c oxidase478. Products of glycolysis and β-oxidation enter the mitochondrion and undergo oxidative phosphorylation. Oxidation of glutamate, malate, and pyruvate produces NADH from the reduction of NAD+ by dehydrogenases. NADH then enter the oxidative phosphorylation cycle through complex I. Succinate, a product of the TCA cycle, enters the cycle through complex II. β-oxidation of fatty acyl CoAs in the mitochondrion provides electrons for complex I or complex II by producing acetyl-CoA that is metabolised through the TCA cycle7,8. The active-site of cytochrome c oxidase contains three copper atoms, two of which are in the CuA subunit and the other in the CuB subunit and catalyses electron transfer from cytochrome c, a protein that transfers electrons between complex III to cytochrome c oxidase, to oxygen to produce water and energy in the form of ATP479. This suggests that copper has an important role in the utilisation of metabolites for energy. An in vitro study in liver mitochondria has shown that exposure to increasing concentrations of Cu(II) reduces the activity of complex I, II and IV, cytochrome c oxidase and reducing the ATP/ADP ratio in a dose- dependent manner. Increasing concentrations of copper also increased lipid peroxidation,

170

decreased mitochondrial glutathione and negatively affected mitochondrial membrane potential and integrity480. Overall, this suggests that intracellular copper accumulation may affect the outcome of glycolysis and the TCA cycle by downregulation of the activity of components of the electron transport chain. Changes to mitochondrial structure were observed in the retinas of STZ rats that had poor glycaemic control for 12 months. Another study indicated that there was little change to activity of the electron transport chain components; however, diabetic rats where only maintained for 1.5 months and this effect may occur after a longer duration of diabetes481.

6.2.2. Copper Overload in Diabetic Retinas May Affect Lipid Metabolism Whilst the role of copper in respiration is well established, a lesser known role for copper is as a modulator of lipid metabolism. Copper deficiency may be a contributing factor in cardiovascular disease, non-alcoholic fatty-liver disease and increased plasma cholesterol482. The mechanisms by which copper modulates lipid metabolism are only in the early stages of research; however, a recent study showed copper is a regulator of lipolysis in a manner dependent on cyclic- adenosine monophosphate (cAMP) signalling483. In the aforementioned study, additional of copper to adipocyte culture media increased free glycerol and non-esterified fatty acyls; the opposite effect was observed when the media was instead treated with a copper chelator. Additionally, copper overload in the liver of ATP7B-/- mice was observed alongside decreased triglyceride staining and conversely in white adipose tissue, decreased copper content compared with wild type liver was associated with a lower concentration of free glycerol, indicative of a lower rate of lipolysis. This was attributed to copper-induced inhibition of phosphodiesterase PDE3B. PDE3B, which contains a metal-binding domain, is activated by down-stream insulin signalling after which it degrades cAMP and deactivates protein kinase A leading to inhibition of hormone-sensitive lipase, inhibiting lipid catabolism484. In this thesis, an increase in copper was observed in both diabetic donor retinas and STZ rat retinas using ICP- MS. Our results in metabolite changes indicated an increase in glycerol using GC-MS and overall downregulation of TGs. A 20% fold change was observed in mRNA expression of Atp7b in both 12- and 16-week studies but these were both not statistically significant; however, together these results correlate with those observed in the literature but more work is required to conclude how copper mediates lipid metabolism. To further elucidate this relationship, it would be worthwhile to investigate the activity of PDE3B in STZ rat retinas to confirm if this pathway is associated with the pathogenesis diabetic retinopathy.

Interestingly, while copper deficiency leads to reduced lipolysis, this effect is exacerbated by a high sucrose diet as opposed to a starch-rich diet whereby there is an increase in fructose485– 487. Given the increase in fructose in diabetic retinas compared with non-diabetic, it would be of interest to research the relationship between copper and fructose accumulation and lipolysis.

6.2.3 Dysregulated Polar Metabolites May Affect the Lipid Profile A novel outcome of this study was the observation that scyllo-inositol was consistently downregulated in diabetic retinas in three separate STZ experiments relative to non-diabetic

171

retinas. Concomitantly, the PI lipid class was mostly downregulated. Little is currently known about the function of scyllo-inositol. The incorporation of scyllo-inositol into PIs has been described in plant cells but not mammalian cells and so whether these observations are related is as of yet unknown488.

In this study, fructose was upregulated consistently in three separate comparisons of diabetic rat retinas with non-diabetic. Conversely in the described lipidomic experiment, DHA-containing lipids regardless of class were downregulated in diabetic rat retinas relative to non-diabetic. A recent study489 has described an interesting relationship between fructose and DHA whereby a group of rats were fed a fructose-rich diet and another group received fructose and DHA for six weeks. The fructose-fed rats had significantly increased serum triglycerides, insulin, insulin resistance, blood glucose and cognitive dysfunction, all of which was significantly reduced by DHA with the exception of blood glucose values. Fructose also induced changes in DNA methylation and gene networks including those associated with extracellular matrix modulation which trended toward being normalised by DHA supplementation. DHA supplementation has been shown to improve rod function in STZ rat retinas and ganglion cell function in normal rats490,491. This may not be limited to just DHA because a mixed supplement of DHA and eicosapentaenoic acid also improved visual function in diabetic mice422. Given the opposing changes in fructose in diabetic retinas and DHA content observed in this thesis, it would be of interest to investigate the relationship between fructose and DHA in diabetic retinopathy and the mechanisms by which DHA may reverse the effects of fructose accumulation.

6.3 Limitations and future work

6.3.1. Limitations of animal model Firstly, the data were only obtained in one model of diabetes, i.e. the STZ rat model. It would be desirable to replicate these data in another model of diabetes. Whilst copper overload was observed in both donor human retinas and the STZ rat retina, it would be informative to investigate a model of type 2 diabetes such as the Zucker diabetic fatty rat model.

Another limitation was the duration of the model used. Whilst molecular changes and visual dysfunction do occur at 12 weeks of STZ-induced diabetes100,216,492, capillary loss only becomes apparent at approximately 32 weeks of STZ diabetes97.

The model is also limited in that the rats were constantly hyperglycaemic and the extent of increased blood glucose levels (approximately 33 mmol/L) observed in these rats are generally only observed on rare occasions in diabetic patients. This may produce an unrealistic reproduction of clinical diabetes whereby patients will only reach this level of blood glucose elevation on rare occasions and blood glucose is generally controlled with anti-hyperglycaemic medication and kept under 5 mmol/L.

172

6.3.2 Limitations and potential future work with respect to copper and copper metabolism. In this thesis, an overall insight into changes to metal homeostasis in diabetic retinopathy was provided; however, this study had some limitations:

 The valency of copper (Cu(I) or Cu(II)) was not determined  The observed increase in copper was not determined to be intracellular or extracellular and so requires further investigation  There may have been specific changes to copper and copper homeostatic mechanisms in the vasculature that were not identified by global analysis of retinas  Transcriptomics data requires further research by investigating changes to the corresponding proteins  The ability of TETA to cross the blood retinal barrier at sufficient concentrations to provide a therapeutic effect after oral administration needs to be investigated

It would be of interest to assess changes to copper homeostasis during diabetic retinopathy with a more advanced pathology. It has been previously observed that vascular-specific changes may be lost within the neural retina such as that observed with the transforming growth factor-β103. The observable difference in the extent of copper change in the kidney compared with retina suggests that perhaps the extent of changes to copper homeostasis in vascular tissue could be masked by the abundance of neuronal tissue. Copper therefore could be measured in isolated vasculature to assess whether copper accumulation occurs majorly in the vasculature as opposed to the neural retina. It is possible to isolate vascular endothelial cells a well-established technique to observe the effects of TETA exclusively in the retinal vasculature such as that described by Matsubara et al 2001493.

It would be of interest to use synchrotron X-ray fluorescence microscopy, as was done in rat retina in the study by Ugarte et al 2012251 to study how the localisation of metals may change and provide insight into the possible cell types where diabetes-induced copper accumulation occurs. Furthermore, studying the effects of hyperglycaemia on copper homeostasis and metallothioneins on endothelial cells, neurons, Mueller cells, astrocytes, and microglia in vitro would be beneficial in understanding the change in copper observed in this study. This would be possible using fluorescence microscopy and a fluorescent copper probe alongside quantification using ICP-MS as was done in the study by Ogra et al 2016494.

Using tissue from the current time points utilised, Atp7a, Atp7b proteins should be assessed because the fold changes in diabetes and TETA treatment indicated potential dysregulation that could be better represented at the protein level and these may also be implicated during a longer course of diabetes. The effect of TETA treatment on ceruloplasmin mRNA expression has also indicated that TETA treatment may be more efficacious than anticipated. This would need to be confirmed at the protein level and the effect of TETA on iron transporters is another

173

potential new lead given the role of ceruloplasmin in iron reduction. Whilst this could currently be done with remaining harvested tissue, the effect of a longer duration of diabetes on copper homeostatic mechanisms would also be of interest to explore if the effect of copper accumulation increases with time and if copper chelation prevents capillary loss.

Were the effects of TETA to be further explored, it would be imperative to detect its presence in retina to ascertain if any effects observed are because of a direct effect in retina or because of the effects of copper chelation elsewhere. A targeted method is currently being developed at the CADET laboratory whereby TETA and its metabolites N1-acetyltriethylenetetramine and diacetyltriethylenetetramine may be observed in retinal tissue; indicative of TETA metabolism in retina. This method is currently in the very early stages of development and very early preliminary work has produced some promising results.

6.3.3 Limitations and future work with respect to polar and non-polar metabolites Untargeted methods provide encompasses a variety of metabolites; however, metabolite identifications depend on the breadth of shared libraries. This limitation is apparent when performing the first untargeted metabolomics experiments in retina; it is likely that metabolites not commonly found in other tissues may not have been recognised because they do not yet appear in metabolite libraries. This was especially evident in lipid identification when the free fatty acyls of VLCPUFAS were identified by recognising that the mass was equivalent to an extended carbon chain of a fatty acyl. It is likely that complex phospholipids containing VLCPUFA chains, such as a PC containing a 36:6 chain, were not identified. Another problem that becomes apparent when discussing the data is the lack of information surrounding many identifiable metabolites. Individual lipids and sugars observed in this study as being dysregulated during early diabetic retinopathy may have as of yet unknown implications in disease progression, for example scyllo-inositol. It will likely require years of research before the function implications of some of these findings can be elucidated. For this reason, changes to the lipid classes have been described as a whole. The CADET laboratory have recently began working on a method for absolute quantification of polar metabolite changes and so it would be possible in the very near future to quantify the concentrations of changed metabolites presented in this study in retina.

The polar metabolite analysis showed a clear pattern where glycolysis intermediates were upregulated. This study is limited in that we did not investigate metabolic flux; this can be done in vivo354 and in vitro495 using labelled metabolites and can be measured using nuclear magnetic resonance methodology. The activity of glycolytic enzymes in vitro to demonstrate at which point the pathway becomes dysregulated should also be measured in vitro using different retinal cell types under hyperglycaemic conditions to identify which cells are responsible for the increased glycolysis observed in the presented study. This should be coupled with investigations into the alternative metabolic pathways such as the sorbitol to localise where

174

sorbitol accumulation occurs but also to elucidate how glucose overload impacts the hexosamine and pentose phosphate pathways and compare how each cell types copes during hyperglycaemia.

It would be worthwhile to repeat metabolic profiling in a study comprising of non-diabetic, diabetic and diabetic corrected with insulin groups, to ideally model the clinical presentation of patients receiving treatment for diabetes with islet cell depletion. This may highlight any metabolic pathways that are not corrected by normoglycaemia and may be more representative of the clinical diabetic population. Lastly, in order to get the full metabolic picture of diabetes, it would be imperative to complete metabolic profiling studies in most tissues affected by diabetes. Our laboratory has previously used this approach on sciatic nerve364, kidney, liver, left ventricle and brain (unpublished data) using tissue from different studies. Were this analysis completed in tissues from the same rats, concomitant concentrations of different metabolite could be compared and correlated across different tissues and provide an overall insight into whole-system metabolic dysregulation in diabetes.

6.6 Overall Conclusions In this thesis, untargeted -omics approaches were used to elucidate previously unknown changes in diabetes. With these methods, the metabolic profile of rat retinas was mapped and those affected by chronic diabetes described. These data showed the extent of perturbation of metabolic pathways during diabetic retinopathy. Where the focus of previous studies has been on the sorbitol pathway and omega-3 fatty acids, this study has shown the variety of dysregulated sugars, amino acids and polyols alongside the extent of changes to lipids that may affect downstream mediators. Untargeted analysis of metals in diabetic retinas were described and quantified and indicated mismanagement of retinal copper in both donor retinas and in the STZ model of diabetes, where this was further evidenced by change in the mRNA expression of copper transporters. Whilst copper chelation with TETA modulated the mRNA expression of genes with copper homeostatic roles, this was not reflected in the mRNA expression of inflammatory mediators; however analysing markers of oxidative stress would provide further insight into the therapeutic potential of TETA treatment. We suggest that there is plenty of scope for further research to investigate mechanisms behind copper dysregulation, how this affects pathogenesis of diabetic retinopathy along with new insights into dysregulated metabolic pathways.

175

References

1. World Health Organization & International Diabetes Federation. Definition and diagnosis of diabetes mellitus and intermediate hyperglycaemia. Report of a WHO/IDF consultation. (2006).

2. Geraldes, P. et al. Activation of PKC-delta and SHP-1 by hyperglycemia causes vascular cell apoptosis and diabetic retinopathy. Nat. Med. 15, 1298–1306 (2009).

3. Geraldes, P. & King, G. L. Activation of Protein Kinase C Isoforms and Its Impact on Diabetic Complications. Circ. Res. 106, 1319–1331 (2010).

4. Harris, M. Chapter 1 Summary. Diabetes in America, 2nd Edition (National Diabetes Data Group of the National Institute of Diabetes and Digestive and Kidney Diseases, 2013).

5. Saltiel, A. R. & Kahn, C. R. Insulin signalling and the regulation of glucose and lipid metabolism. Nature 414, 799–806 (2001).

6. Knott, R. M. et al. A model system for the study of human retinal angiogenesis: activation of monocytes and endothelial cells and the association with the expression of the monocarboxylate transporter type 1 (MCT-1). Diabetologia 42, 870–877 (1999).

7. Harvey, R. A. & Ferrier, D. R. Lippincott’s Illustrated Reviews: Biochemistry Fifth Edition. Wolters Kluwer, Lippincott Williams & Wilkins (2011). doi:10.1016/0307-4412(87)90018- 5

8. Sivitz, W. I. & Yorek, M. A. Mitochondrial dysfunction in diabetes: from molecular mechanisms to functional significance and therapeutic opportunities. Antioxid. Redox Signal. 12, 537–77 (2010).

9. Boucher, J., Kleinridders, A. & Ronald Kahn, C. Insulin receptor signaling in normal and insulin-resistant states. Cold Spring Harb. Perspect. Biol. 6, (2014).

10. Huang, S. & Czech, M. P. The GLUT4 Glucose Transporter. Cell Metab. 5, 237–252 (2007).

11. World Health Organization. Global Report on Diabetes. Isbn 978, 88 (2016).

12. Knip, M. et al. Environmental Triggers and Determinants of Type 1 Diabetes. Diabetes 54, (2005).

13. Pessin, J. E. & Saltiel, A. R. Signaling pathways in insulin action : molecular targets of insulin resistance. J. Clin. Invest. 106, 165–169 (2000).

14. Leonardi, O., Mints, G. & Hussain, M. A. Beta-cell apoptosis in the pathogenesis of human type 2 diabetes mellitus. Eur. J. Endocrinol. 149, 99–102 (2003).

15. Masad, A. et al. Copper-mediated formation of hydrogen peroxide from the amylin peptide: A novel mechanism for degeneration of islet cells in type-2 diabetes mellitus?

176

FEBS Lett. 581, 3489–3493 (2007).

16. Keller, A. et al. Influence of hydrophobicity on the surface-catalyzed assembly of the islet amyloid polypeptide. ACS Nano 5, 2770–2778 (2011).

17. Bloch-Damti, A. & Bashan, N. Proposed mechanisms for the induction of insulin resistance by oxidative stress. Antioxid. Redox. Signal. 7, 1553–1567 (2005).

18. DeFronzo, R. et al. Type 2 diabetes mellitus. Nat. Rev. Dis. Prim. 1, 15019 (2015).

19. National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). Definition & Facts of Gestational Diabetes | NIDDK. Available at: https://www.niddk.nih.gov/health- information/diabetes/overview/what-is-diabetes/gestational/definition-facts. (Accessed: 8th January 2018)

20. Diabetes UK. State of the Nation 2016 (England): time to take control of diabetes. (2016).

21. Harris, M. in Diabetes in America, 2nd Edition (ed. Harris C. C. Stern M. P., Edward J. Boyko E. J. Reiber, G. E. Bennett P. H., M. I. C.) (National Diabetes Data Group of the National Institute of Diabetes and Digestive and Kidney Diseases, 2013).

22. International Diabetes Federation. IDF Diabetes Atlas, 6th edn. 160 (2013).

23. American Diabetes Association. Executive summary: standards of medical care in diabetes--2009. Diabetes Care 32 Suppl 1, S6–12 (2009).

24. Florkowski, C. HbA as a Diagnostic Test for Diabetes Mellitus - Reviewing the Evidence. Clin. Biochem. Rev. 34, 75–83 (2013).

25. Hex, N., Bartlett, C., Wright, D., Taylor, M. & Varley, D. Estimating the current and future costs of Type 1 and Type 2 diabetes in the UK, including direct health costs and indirect societal and productivity costs. Diabet Med 29, 855–862 (2012).

26. Chan, J. C. et al. Diabetes in Asia: epidemiology, risk factors, and pathophysiology. JAMA 301, 2129–2140 (2009).

27. Diabetes UK. Diabetes UK: key facts and stats. (2016).

28. Bell, D. S. Type 2 diabetes mellitus: what is the optimal treatment regimen? Am J Med 116 Suppl, 23S–29S (2004).

29. UK Prospective Diabetes Study Group. Tight blood pressure control and risk of macrovascular and microvascular complications in type 2 diabetes: UKPDS 38. UK Prospective Diabetes Study Group. BMJ 317, 703–713 (1998).

30. Wright, A. D. & Dodson, P. M. Medical management of diabetic retinopathy: fenofibrate and ACCORD Eye studies. Eye 25, 843–849 (2011).

31. Keech, a C. et al. Effect of fenofibrate on the need for laser treatment for diabetic retinopathy (FIELD study): a randomised controlled trial. Lancet 370, 1687–1697 (2007).

177

32. Creager, M. A., Lüscher, T. F., Cosentino, F. & Beckman, J. A. Diabetes and Vascular Disease: Pathophysiology, Clinical Consequences, and Medical Therapy: Part I. Circulation 108, 1527–1532 (2003).

33. Grundy, S. M. et al. Diabetes and cardiovascular disease: a statement for healthcare professionals from the American Heart Association. Circulation 100, 1134–1146 (1999).

34. Kolluru, G. K., Bir, S. C. & Kevil, C. G. Endothelial dysfunction and diabetes: effects on angiogenesis, vascular remodeling, and wound healing. Int. J. Vasc. Med. 2012, 918267 (2012).

35. Carr, M. E. Diabetes mellitus: a hypercoagulable state. J. Diabetes Complications 15, 44–54 (2001).

36. Nathan, D. M. et al. Management of hyperglycemia in type 2 diabetes: A consensus algorithm for the initiation and adjustment of therapy: a consensus statement from the American Diabetes Association and the European Association for the Study of Diabetes. Diabetes Care 29, 1963–1972 (2006).

37. Younk, L. M., Mikeladze, M. & Davis, S. N. Pramlintide and the treatment of diabetes: a review of the data since its introduction. Expert Opin. Pharmacother. 12, 1439–1451 (2011).

38. Bailey, C. J. Biguanides and Niddm. Diabetes Care 15, 755–772 (1992).

39. Moore, E. M. et al. Increased risk of cognitive impairment in patients with diabetes is associated with metformin. Diabetes Care 36, 2981–2987 (2013).

40. Marín-Peñalver, J. J., Martín-Timón, I., Sevillano-Collantes, C. & Del Cañizo-Gómez, F. J. Update on the treatment of type 2 diabetes mellitus. World J. Diabetes 7, 354–95 (2016).

41. Secrest, M. H., Udell, J. A. & Filion, K. B. The cardiovascular safety trials of DPP-4 inhibitors, GLP-1 agonists, and SGLT2 inhibitors. Trends in Cardiovasc. Med. 27, 194– 202 (2017).

42. Marín-Peñalver, J. J., Martín-Timón, I., Sevillano-Collantes, C. & Del Cañizo-Gómez, F. J. Update on the treatment of type 2 diabetes mellitus. World J. Diabetes 7, 354–95 (2016).

43. Butler, P. C., Elashoff, M., Elashoff, R. & Gale, E. A. A critical analysis of the clinical use of incretin-based therapies: Are the GLP-1 therapies safe? Diabetes Care 36, 2118– 2125 (2013).

44. Cohen, D. Reports of pancreatitis are 20-30 times more likely with GLP-1 drugs, analysis finds. BMJ 346, f2607 (2013).

45. Kahn, S. E. Incretin therapy and islet pathology: a time for caution. Diabetes 62, 2178– 2180 (2013).

178

46. Hemkens, L. G. et al. Risk of malignancies in patients with diabetes treated with human insulin or insulin analogues: a cohort study. Diabetologia 52, 1732–1744 (2009).

47. Jonasson, J. M. et al. Insulin glargine use and short-term incidence of malignancies-a population-based follow-up study in Sweden. Diabetologia 52, 1745–1754 (2009).

48. Li, D., Yeung, S. C., Hassan, M. M., Konopleva, M. & Abbruzzese, J. L. Antidiabetic therapies affect risk of pancreatic cancer. Gastroenterology 137, 482–488 (2009).

49. Bell, D. S. Type 2 diabetes mellitus: what is the optimal treatment regimen? Am. J. Med. 116 Suppl, 23S–29S (2004).

50. Wheeler, S. et al. Mortality among veterans with type 2 diabetes initiating metformin, sulfonylurea or rosiglitazone monotherapy. Diabetologia 56, 1934–1943 (2013).

51. Bodmer, M., Becker, C., Meier, C., Jick, S. S. & Meier, C. R. Use of Antidiabetic Agents and the Risk of Pancreatic Cancer: A Case-Control Analysis. Am. J. Gastroenterol. 107, 620–626 (2012).

52. Consoli, A. & Formoso, G. Do thiazolidinediones still have a role in treatment of type 2 diabetes mellitus? Diabetes Obes. Metab. 15, 967–977 (2013).

53. Singh, S., Loke, Y. K. & Furberg, C. D. Long-term risk of cardiovascular events with rosiglitazone: a meta-analysis. JAMA 298, 1189–1195 (2007).

54. Nissen, S. E. & Wolski, K. Rosiglitazone Revisited An Updated Meta-analysis of Risk for Myocardial Infarction and Cardiovascular Mortality. Arch. Intern. Med. 170, 1191–1201 (2010).

55. van den Born, J. C. et al. Gasotransmitters in Vascular Complications of Diabetes. Diabetes 65, 331–45 (2016).

56. Pappachan, J. M., Varughese, G. I., Sriraman, R. & Arunagirinathan, G. Diabetic cardiomyopathy: Pathophysiology, diagnostic evaluation and management. World J. Diabetes 4, 177–189 (2013).

57. Rang, H. P., Ritter, J. M., Flower, R. J. & Henderson, G. Rang & Dale’s Pharmacology, 8th Edition. Pharmacology (2015).

58. Chen, R., Ovbiagele, B. & Feng, W. Diabetes and Stroke: Epidemiology, Pathophysiology, Pharmaceuticals and Outcomes. Am. J. Med. Sci. 351, 380–6 (2016).

59. Karapanayiotides, T., Piechowski-Jozwiak, B., van Melle, G., Bogousslavsky, J. & Devuyst, G. Stroke patterns, etiology, and prognosis in patients with diabetes mellitus. Neurology 62, 1558–62 (2004).

60. Duby, J., Campbell, R., Setter, S., White & Rasmussen, K. Diabetic neuropathy: an intensive review. Am. J. Heal. Pharm. 61, 160–173 (2004).

61. Singleton, J. R. & Smith, A. G. The diabetic neuropathies: practical and rational therapy. Semin. Neurol. 32, 196–203 (2012).

179

62. Haley, W. E., Turner, J. A. & Romano, J. M. Depression in chronic pain patients: relation to pain, activity, and sex differences. Pain 23, 337–343 (1985).

63. Arora, M. K. & Singh, U. K. Molecular mechanisms in the pathogenesis of diabetic nephropathy: an update. Vasc. Pharmacol 58, 259–271 (2013).

64. Purves, D. Neuroscience. (Sinauer Associates, 2001).

65. Zhang, H. R. Scanning Electron-Microscopic Study of Corrosion Casts on Retinal and Choroidal Angioarchitecture in Man and Animals. Prog. Retin. Eye Res. 13, 243–270 (1994).

66. Vecino, E., Rodriguez, F. D., Ruzafa, N., Pereiro, X. & Sharma, S. C. Glia–neuron interactions in the mammalian retina. Prog. Retin. Eye Res. 51, 1–40 (2016).

67. Stone, J. et al. Development of retinal vasculature is mediated by hypoxia-induced vascular endothelial growth factor (VEGF) expression by neuroglia. J. Neurosci. 15, 4738–47 (1995).

68. Bélanger, M., Allaman, I. & Magistretti, P. J. Brain Energy Metabolism: Focus on Astrocyte-Neuron Metabolic Cooperation. Cell Metab. 14, 724–738 (2011).

69. Reichenbach, A. & Bringmann, A. New functions of Müller cells. Glia 61, 651–678 (2013).

70. Duh, E. J., Sun, J. K. & Stitt, A. W. Diabetic retinopathy: current understanding, mechanisms, and treatment strategies. JCI insight 2, (2017).

71. Frank, R. N. Diabetic retinopathy. N. Engl. J. Med. 350, 48–58 (2004).

72. Fong, D. S., Barton, F. B., Bresnick, G. H. & Rtinopath, E. T. D. Impaired color vision associated with diabetic retinopathy: Early treatment diabetic retinopathy study report no. 15. Am. J. Ophthalmol. 128, 612–617 (1999).

73. VanGuilder, H. D. et al. Multi-modal proteomic analysis of retinal protein expression alterations in a rat model of diabetic retinopathy. PLoS One 6, e16271 (2011).

74. Ding, J. & Wong, T. Y. Current epidemiology of diabetic retinopathy and diabetic macular edema. Curr Diab Rep 12, 346–354 (2012).

75. Ashton, N. Arteriolar Involvement in Diabetic Retinopathy. Br J Ophthalmol 37, 282–292 (1953).

76. Sokol, S. et al. Contrast Sensitivity in Diabetics with and without Background Retinopathy. Arch. Ophthalmol. 103, 51–54 (1985).

77. McLeod, D. S., Lefer, D. J., Merges, C. & Lutty, G. A. Enhanced expression of intracellular adhesion molecule-1 and P-selectin in the diabetic human retina and choroid. Am. J. Pathol. 147, 642–653 (1995).

78. Boeri, D., Maiello, M. & Lorenzi, M. Increased prevalence of microthromboses in retinal capillaries of diabetic individuals. Diabetes 50, 1432–1439 (2001).

180

79. Mizutani, M., Kern, T. S. & Lorenzi, M. Accelerated death of retinal microvascular cells in human and experimental diabetic retinopathy. J. Clin. Invest. 97, 2883–2890 (1996).

80. Wu, L., Fernandez-Loaiza, P., Sauma, J., Hernandez-Bogantes, E. & Masis, M. Classification of diabetic retinopathy and diabetic macular edema. World J. Diabetes 4, 290–294 (2013).

81. Gardner, T. W. et al. Diabetic retinopathy: More than meets the eye. Surv. Ophthalmol. 47, S253–S262 (2002).

82. Lee, R., Wong, T. Y. & Sabanayagam, C. Epidemiology of diabetic retinopathy, diabetic macular edema and related vision loss. Eye Vis. (London, England) 2, 17 (2015).

83. Diabetic retinopathy (DR): management and referral. Community Eye Health Journal (2011). Available at: https://www.cehjournal.org/article/diabetic-retinopathy-dr- management-and-referral/. (Accessed: 30th August 2017)

84. Curtis, T. M., Gardiner, T. A. & Stitt, A. W. Microvascular lesions of diabetic retinopathy: clues towards understanding pathogenesis? Eye 23, 1496–1508 (2009).

85. El-Remessy, A. B. et al. Experimental diabetes causes breakdown of the blood-retina barrier by a mechanism involving tyrosine nitration and increases in expression of vascular endothelial growth factor and urokinase plasminogen activator receptor. Am. J. Pathol. 162, 1995–2004 (2003).

86. Antonetti, D. A. et al. Vascular permeability in experimental diabetes is associated with reduced endothelial occludin content: vascular endothelial growth factor decreases occludin in retinal endothelial cells. Penn State Retina Research Group. Diabetes 47, 1953–1959 (1998).

87. Bixler, G. V et al. Chronic insulin treatment of diabetes does not fully normalize alterations in the retinal transcriptome. BMC Med Genomics 4, 40 (2011).

88. Xu, H.-Z. & Le, Y.-Z. Significance of outer blood-retina barrier breakdown in diabetes and ischemia. Invest. Ophthalmol. Vis. Sci. 52, 2160–4 (2011).

89. Omri, S. et al. PKCζ Mediates Breakdown of Outer Blood-Retinal Barriers in Diabetic Retinopathy. PLoS One 8, e81600 (2013).

90. Roy, S., Ha, J., Trudeau, K. & Beglova, E. Vascular basement membrane thickening in diabetic retinopathy. Curr. Eye Res. 35, 1045–1056 (2010).

91. Dosso, A. A., Leuenberger, P. M. & Rungger-Brandle, E. Remodeling of retinal capillaries in the diabetic hypertensive rat. Invest. Ophthalmol. Vis. Sci. 40, 2405–2410 (1999).

92. To, M. et al. Diabetes-induced morphological, biomechanical, and compositional changes in ocular basement membranes. Exp. Eye Res. 116, 298–307 (2013).

93. Robison, W. G., Kador, P. F. & Kinoshita, J. H. Retinal capillaries: basement membrane

181

thickening by galactosemia prevented with aldose reductase inhibitor. Science 221, 1177–1179 (1983).

94. Hammes, H. P., Bartmann, A., Engel, L. & Wulfroth, P. Antioxidant treatment of experimental diabetic retinopathy in rats with nicanartine. Diabetologia 40, 629–634 (1997).

95. Stitt, A. W. et al. Advanced glycation end products (AGEs) co-localize with AGE receptors in the retinal vasculature of diabetic and of AGE-infused rats. Am. J. Pathol. 150, 523–531 (1997).

96. Dagher, Z. et al. Studies of rat and human retinas predict a role for the polyol pathway in human diabetic retinopathy. Diabetes 53, 2404–2411 (2004).

97. Bhatwadekar, A. et al. A new advanced glycation inhibitor, LR-90, prevents experimental diabetic retinopathy in rats. Br J Ophthalmol 92, 545–547 (2008).

98. Ribatti, D., Nico, B. & Crivellato, E. The role of pericytes in angiogenesis. Int. J. Dev. Biol. 55, 261–268 (2011).

99. Shweiki, D., Itin, A., Soffer, D. & Keshet, E. Vascular Endothelial Growth-Factor Induced by Hypoxia May Mediate Hypoxia-Initiated Angiogenesis. Nature 359, 843–845 (1992).

100. Brucklacher, R. M. et al. Whole genome assessment of the retinal response to diabetes reveals a progressive neurovascular inflammatory response. BMC Med Genomics 1, 26 (2008).

101. Kumagai, a K., Glasgow, B. J. & Pardridge, W. M. GLUT1 glucose transporter expression in the diabetic and nondiabetic human eye. Invest. Ophthalmol. Vis. Sci. 35, 2887–2894 (1994).

102. Asnaghi, V., Gerhardinger, C., Hoehn, T., Adeboje, A. & Lorenzi, M. A role for the polyol pathway in the early neuroretinal apoptosis and glial changes induced by diabetes in the rat. Diabetes 52, 506–511 (2003).

103. Gerhardinger, C., Dagher, Z., Sebastiani, P., Park, Y. S. & Lorenzi, M. The Transforming Growth Factor-beta Pathway Is a Common Target of Drugs That Prevent Experimental Diabetic Retinopathy. Diabetes 58, 1659–1667 (2009).

104. Brownlee, M. Biochemistry and molecular cell biology of diabetic complications. Nature 414, 813–820 (2001).

105. Nakamura, M. et al. Excessive hexosamines block the neuroprotective effect of insulin and induce apoptosis in retinal neurons. J. Biol. Chem. 276, 43748–43755 (2001).

106. Sayeski, P. P. & Kudlow, J. E. Glucose metabolism to glucosamine is necessary for glucose stimulation of transforming growth factor-alpha gene transcription. J. Biol. Chem. 271, 15237–15243 (1996).

107. Kolm-Litty, V., Sauer, U., Nerlich, A., Lehmann, R. & Schleicher, E. D. High glucose-

182

induced transforming growth factor beta1 production is mediated by the hexosamine pathway in porcine glomerular mesangial cells. J. Clin. Invest. 101, 160–169 (1998).

108. Du, X. L. et al. Hyperglycemia-induced mitochondrial superoxide overproduction activates the hexosamine pathway and induces plasminogen activator inhibitor-1 expression by increasing Sp1 glycosylation. Proc. Natl. Acad. Sci. U.S.A. 97, 12222– 12226 (2000).

109. Basu, A., Menicucci, G., Maestas, J., Das, A. & McGuire, P. Plasminogen activator inhibitor-1 (PAI-1) facilitates retinal angiogenesis in a model of oxygen-induced retinopathy. Invest. Ophthalmol. Vis. Sci. 50, 4974–4981 (2009).

110. Madsen-Bouterse, S., Mohammad, G. & Kowluru, R. A. Glyceraldehyde-3-phosphate dehydrogenase in retinal microvasculature: implications for the development and progression of diabetic retinopathy. Invest. Ophthalmol. Vis. Sci. 51, 1765–1772 (2010).

111. Hammes, H. P. et al. Differential accumulation of advanced glycation end products in the course of diabetic retinopathy. Diabetologia 42, 728–736 (1999).

112. Stitt, A. W. AGEs and diabetic retinopathy. Invest. Ophthalmol. Vis. Sci. 51, 4867–4874 (2010).

113. Lorenzi, M. & Gerhardinger, C. Early cellular and molecular changes induced by diabetes in the retina. Diabetologia 44, 791–804 (2001).

114. Padayatti, P. S., Jiang, C., Glomb, M. A., Uchida, K. & Nagaraj, R. H. High concentrations of glucose induce synthesis of argpyrimidine in retinal endothelial cells. Curr. Eye Res. 23, 106–115 (2001).

115. Chen, B. H., Jiang, D. Y. & Tang, L. S. Advanced glycation end-products induce apoptosis involving the signaling pathways of oxidative stress in bovine retinal pericytes. Life Sci. 79, 1040–1048 (2006).

116. Moore, T. C. et al. The role of advanced glycation end products in retinal microvascular leukostasis. Invest. Ophthalmol. Vis. Sci. 44, 4457–4464 (2003).

117. Chibber, R., Molinatti, P. A., Rosatto, N., Lambourne, B. & Kohner, E. M. Toxic action of advanced glycation end products on cultured retinal capillary pericytes and endothelial cells: relevance to diabetic retinopathy. Diabetologia 40, 156–164 (1997).

118. Rothermundt, M., Peters, M., Prehn, J. H. & Arolt, V. S100B in brain damage and neurodegeneration. Microsc. Res.Tech. 60, 614–632 (2003).

119. Sims, G. P., Rowe, D. C., Rietdijk, S. T., Herbst, R. & Coyle, A. J. HMGB1 and RAGE in inflammation and cancer. Annu. Rev. Immunol. 28, 367–388 (2010).

120. Mehla, J., Chauhan, B. C. & Chauhan, N. B. Experimental Induction of Type 2 Diabetes in Aging-Accelerated Mice Triggered Alzheimer-Like Pathology and Memory Deficits. J. Alzheimers Dis. 39, 145–162 (2014).

183

121. Du, Y., Veenstra, A., Palczewski, K. & Kern, T. S. Photoreceptor cells are major contributors to diabetes-induced oxidative stress and local inflammation in the retina. Proc. Natl. Acad. Sci. U.S.A. 110, 16586–16591 (2013).

122. Vistoli, G. et al. Advanced glycoxidation and lipoxidation end products (AGEs and ALEs): an overview of their mechanisms of formation. Free Radic. Res. 47, 3–27 (2013).

123. Loske, C. et al. Transition metal-mediated glycoxidation accelerates cross-linking of beta-amyloid peptide. Eur. J. Biochem. 267, 4171–4178 (2000).

124. Busik, J. V, Esselman, W. J. & Reid, G. E. Examining the role of lipid mediators in diabetic retinopathy. Clin. Lipidol. 7, 661–675 (2012).

125. Wang, Y. et al. Regulation of hepatic fatty acid elongase and desaturase expression in diabetes and obesity. J. Lipid Res. 47, 2028–41 (2006).

126. Chen, W., Jump, D. B., Grant, M. B., Esselman, W. J. & Busik, J. V. Dyslipidemia, but Not Hyperglycemia, Induces Inflammatory Adhesion Molecules in Human Retinal Vascular Endothelial Cells. Investig. Opthalmology Vis. Sci. 44, 5016 (2003).

127. Chen, W., Esselman, W. J., Jump, D. B. & Busik, J. V. Anti-inflammatory Effect of Docosahexaenoic Acid on Cytokine-Induced Adhesion Molecule Expression in Human Retinal Vascular Endothelial Cells. Investig. Opthalmology Vis. Sci. 46, 4342 (2005).

128. Wilkinson-Berka, J. L. Vasoactive factors and diabetic retinopathy: vascular endothelial growth factor, cycoloxygenase-2 and nitric oxide. Curr. Pharm. Des. 10, 3331–3348 (2004).

129. Schoenberger, S. D. et al. Increased prostaglandin E2 (PGE2) levels in proliferative diabetic retinopathy, and correlation with VEGF and inflammatory cytokines. Invest. Ophthalmol. Vis. Sci. 53, 5906–5911 (2012).

130. Lupo, G. et al. Role of phospholipases A2 in diabetic retinopathy: in vitro and in vivo studies. Biochem. Pharmacol. 86, 1603–1613 (2013).

131. Matias, I. et al. Changes in endocannabinoid and palmitoylethanolamide levels in eye tissues of patients with diabetic retinopathy and age-related macular degeneration. Prostaglandins Leukot. Essent. Fat. Acids 75, 413–418 (2006).

132. Yu, M., Ives, D. & Ramesha, C. Synthesis of prostaglandin E2 ethanolamide from anandamide by cyclooxygenase-2. J. Biol. Chem. 272, 21181–21186 (1997).

133. Behl, T., Kaur, I. & Kotwani, A. Role of endocannabinoids in the progression of diabetic retinopathy. Diabetes Metab Res Rev. 32, 251–259 (2016).

134. Suzuki, K. et al. Janus kinase 3 (Jak3) is essential for common cytokine receptor gamma chain (gamma(c))-dependent signaling: comparative analysis of gamma(c), Jak3, and gamma(c) and Jak3 double-deficient mice. Int. Immunol. 12, 123–32 (2000).

135. Liu, Y., Biarnes Costa, M. & Gerhardinger, C. IL-1beta is upregulated in the diabetic

184

retina and retinal vessels: cell-specific effect of high glucose and IL-1beta autostimulation. PLoS One 7, e36949 (2012).

136. Gerhardinger, C., McClure, K. D., Romeo, G., Podesta, F. & Lorenzi, M. IGF-I mRNA and signaling in the diabetic retina. Diabetes 50, 175–183 (2001).

137. Ishii, H. et al. Amelioration of vascular dysfunctions in diabetic rats by an oral PKC beta inhibitor. Science (80-. ). 272, 728–731 (1996).

138. Koya, D. & King, G. L. Protein kinase C activation and the development of diabetic complications. Diabetes 47, 859–866 (1998).

139. Curtis, T. M. & Scholfield, C. N. The role of lipids and protein kinase Cs in the pathogenesis of diabetic retinopathy. Diabetes Metab Res Rev 20, 28–43 (2004).

140. Acosta, J. et al. Molecular basis for a link between complement and the vascular complications of diabetes. Proc. Natl. Acad. Sci. U.S.A. 97, 5450–5455 (2000).

141. Zhang, A., Gerhardinger, C. & Lorenzi, M. Early complement activation and decreased levels of glycosylphosphatidylinositol-anchored complement inhibitors in human and experimental diabetic retinopathy. Diabetes 51, 3499–3504 (2002).

142. Cooper, G. J. S. Selective Divalent Copper Chelation for the Treatment of Diabetes Mellitus. Curr. Med. Chem. 19, 2828–2860 (2012).

143. Ahmed, M. U., Thorpe, S. R. & Baynes, J. W. Identification of N-Epsilon- Carboxymethyllysine as a Degradation Product of Fructoselysine in Glycated Protein. J. Biol. Chem. 261, 4889–4894 (1986).

144. Birben, E., Sahiner, U. M., Sackesen, C., Erzurum, S. & Kalayci, O. Oxidative stress and antioxidant defense. World Allergy Organ. J. 5, 9–19 (2012).

145. Sandstrom, J., Karlsson, K., Edlund, T. & Marklund, S. L. Heparin-affinity patterns and composition of extracellular superoxide dismutase in human plasma and tissues. Biochem. J. 294 ( Pt 3, 853–857 (1993).

146. Du, Y. P., Miller, C. M. & Kern, T. S. Hyperglycemia increases mitochondrial superoxide in retina and retinal cells. Free Radic. Biol. Med. 35, 1491–1499 (2003).

147. Wolff, S. P., Jiang, Z. Y. & Hunt, J. V. Protein Glycation and Oxidative Stress in Diabetes-Mellitus and Aging. Free Radic. Biol. Med. 10, 339–352 (1991).

148. Kowluru, R. A., Atasi, L. & Ho, Y.-S. Role of Mitochondrial Superoxide Dismutase in the Development of Diabetic Retinopathy. Investig. Opthalmology Vis. Sci. 47, 1594 (2006).

149. Figarola, J. L. et al. LR-90 a new advanced glycation endproduct inhibitor prevents progression of diabetic nephropathy in streptozotocin-diabetic rats. Diabetologia 46, 1140–1152 (2003).

150. Shangari, N., Chan, T. S., Chan, K., Wu, S. H. & O’Brien, P. J. Copper-catalyzed ascorbate oxidation results in glyoxal/AGE formation and cytotoxicity. Mol. Nutr. Food

185

Res. 51, 445–455 (2007).

151. Williams, M., Hogg, R. E. & Chakravarthy, U. Antioxidants and Diabetic Retinopathy. Curr. Diab. Rep. 13, 481–487 (2013).

152. Monnier, V. Transition metals redox: reviving an old plot for diabetic vascular disease. J. Clin. Invest. 107, 799–801 (2001).

153. Qian, M., Liu, M. & Eaton, J. W. Transition metals bind to glycated proteins forming redox active ‘glycochelates’: implications for the pathogenesis of certain diabetic complications. Biochem. Biophys. Res. Commun. 250, 385–9 (1998).

154. Miao, X. et al. Zinc and diabetic retinopathy. J Diabetes Res 2013, (2013).

155. Simó, R., Ciudin, A. & Hernández, C. Iron overload in diabetic retinopathy: A cause or a consequence of impaired mechanisms? Exp. Diabetes Res. 2010, (2010).

156. Grant, M. B. et al. Plasminogen activator inhibitor (PAI)-1 overexpression in retinal microvessels of PAI-1 transgenic mice. Invest. Ophthalmol. Vis. Sci. 41, 2296–302 (2000).

157. Moustafa, S. A. Zinc might protect oxidative changes in the retina and pancreas at the early stage of diabetic rats. Toxicol. Appl. Pharmacol. 201, 149–55 (2004).

158. Faure, P., Benhamou, P. Y., Perard, A., Halimi, S. & Roussel, A. M. Lipid peroxidation in insulin-dependent diabetic patients with early retina degenerative lesions: effects of an oral zinc supplementation. Eur. J. Clin. Nutr. 49, 282–8 (1995).

159. Kumar, S. D. et al. Zinc supplementation prevents cardiomyocyte apoptosis and congenital heart defects in embryos of diabetic mice. Free Radic. Biol. Med. 53, 1595– 1606 (2012).

160. Dominguez, J. H., Liu, Y. & Kelly, K. J. Renal iron overload in rats with diabetic nephropathy. Physiol. Rep. 3, e12654 (2015).

161. Gao, W., Li, X., Gao, Z. & Li, H. Iron Increases Diabetes-Induced Kidney Injury and Oxidative Stress in Rats. Biol. Trace Elem. Res. 160, 368–375 (2014).

162. Tapiero, H., Townsend, D. M. & Tew, K. D. Trace elements in human physiology and pathology. Copper. Biomed. Pharmacother. 57, 386–398 (2003).

163. Kehrer, J. P. The Haber-Weiss reaction and mechanisms of toxicity. Toxicology 149, 43– 50 (2000).

164. Saxena, A. K. et al. Protein aging by carboxymethylation of lysines generates sites for divalent metal and redox active copper binding: relevance to diseases of glycoxidative stress. Biochem. Biophys. Res. Commun. 260, 332–338 (1999).

165. Sajithlal, G. B., Chithra, P. & Chandrakasan, G. An in vitro study on the role of metal catalyzed oxidation in glycation and crosslinking of collagen. Mol. Cell Biochem. 194, 257–263 (1999).

186

166. Zhang, S. et al. Diabetic cardiomyopathy is associated with defective myocellular copper regulation and both defects are rectified by divalent copper chelation. Cardiovasc. Diabetol. 13, 18 (2014).

167. Cooper, G. J. et al. Demonstration of a hyperglycemia-driven pathogenic abnormality of copper homeostasis in diabetes and its reversibility by selective chelation: quantitative comparisons between the biology of copper and eight other nutritionally essential elements in norma. Diabetes 54, 1468–1476 (2005).

168. Lu, J. et al. Copper(II)-selective chelation improves function and antioxidant defences in cardiovascular tissues of rats as a model of diabetes: comparisons between triethylenetetramine and three less copper-selective transition-metal-targeted treatments. Diabetologia 53, 1217–1226 (2010).

169. Gong, D. et al. Quantitative proteomic profiling identifies new renal targets of copper(II)- selective chelation in the reversal of diabetic nephropathy in rats. Proteomics 9, 4309– 4320 (2009).

170. Diabetes Control and Complications Trial Research Group. Effect of Intensive Diabetes Treatment on the Development and Progression of Long-Term Complications in Adolescents with Insulin-Dependent Diabetes-Mellitus - Diabetes Control and Complications Trial. J. Pediatr. 125, 177–188 (1994).

171. Abu El-Asrar, A. M. Evolving strategies in the management of diabetic retinopathy. Middle East Afr J Ophthalmol 20, 273–282 (2013).

172. Lovestam-Adrian, M. & Agardh, E. Photocoagulation of diabetic macular oedema-- complications and visual outcome. Acta Ophthalmol Scand 78, 667–671 (2000).

173. Moradi, A. et al. Vascular endothelial growth factor trap-eye (Aflibercept) for the management of diabetic macular edema. World J. Diabetes 4, 303–309 (2013).

174. Krispel, C., Rodrigues, M., Xin, X. & Sodhi, A. Ranibizumab in diabetic macular edema. World J. Diabetes 4, 310–318 (2013).

175. Al Dhibi, H. A. & Arevalo, J. F. Clinical trials on corticosteroids for diabetic macular edema. World J. Diabetes 4, 295–302 (2013).

176. Romero-Aroca, P. Current status in diabetic macular edema treatments. World J. Diabetes 4, 165–169 (2013).

177. Simo, R. & Hernandez, C. Advances in the medical treatment of diabetic retinopathy. Diabetes Care 32, 1556–1562 (2009).

178. Pearson, P. A. et al. Fluocinolone acetonide intravitreal implant for diabetic macular edema: a 3-year multicenter, randomized, controlled clinical trial. Ophthalmology 118, 1580–1587 (2011).

179. Bressler, N. M., Beck, R. W. & Ferris 3rd, F. L. Panretinal photocoagulation for proliferative diabetic retinopathy. N. Engl. J. Med. 365, 1520–1526 (2011).

187

180. Adamis, A. P. et al. Changes in retinal neovascularization after pegaptanib (Macugen) therapy in diabetic individuals. Ophthalmology 113, 23–28 (2006).

181. Ip, M. S., Domalpally, A., Hopkins, J. J., Wong, P. & Ehrlich, J. S. Long-term effects of ranibizumab on diabetic retinopathy severity and progression. Arch. Ophthalmol. 130, 1145–1152 (2012).

182. Diabetic Retinopathy Clinical Research, N. Randomized clinical trial evaluating intravitreal ranibizumab or saline for vitreous hemorrhage from proliferative diabetic retinopathy. JAMA Ophthalmol 131, 283–293 (2013).

183. Arevalo, J. F. et al. Intravitreal Bevacizumab (Avastin) for Diabetic Retinopathy: The 2010 GLADAOF Lecture. J Ophthalmol 2011, 584238 (2011).

184. Kuiper, E. J. et al. The angio-fibrotic switch of VEGF and CTGF in proliferative diabetic retinopathy. PLoS One 3, (2008).

185. Liu, X., Chen, H. H. & Zhang, L. W. Potential therapeutic effects of pigment epithelium- derived factor for treatment of diabetic retinopathy. Int J Ophthalmol 6, 221–227 (2013).

186. Anil Kumar, P. & Bhanuprakash Reddy, G. Focus on molecules: aldose reductase. Exp. Eye Res. 85, 739–740 (2007).

187. Sheetz, M. J. et al. The effect of the oral PKC beta inhibitor ruboxistaurin on vision loss in two phase 3 studies. Invest. Ophthalmol. Vis. Sci. 54, 1750–1757 (2013).

188. Wright, A. D. & Dodson, P. M. Medical management of diabetic retinopathy: fenofibrate and ACCORD Eye studies. Eye 25, 843–849 (2011).

189. Davis, M. I., Wilson, S. H. & Grant, M. B. The therapeutic problem of proliferative diabetic retinopathy: targeting somatostatin receptors. Horm. Metab. Res. 33, 295–299 (2001).

190. Lonn, E. et al. Effects of vitamin E on cardiovascular and microvascular outcomes in high-risk patients with diabetes: results of the HOPE study and MICRO-HOPE substudy. Diabetes Care 25, 1919–1927 (2002).

191. Mayer-Davis, E. J. et al. Antioxidant nutrient intake and diabetic retinopathy: the San Luis Valley Diabetes Study. Ophthalmology 105, 2264–2270 (1998).

192. Garcia-Medina, J. J., Pinazo-Duran, M. D., Garcia-Medina, M., Zanon-Moreno, V. & Pons-Vazquez, S. A 5-year follow-up of antioxidant supplementation in type 2 diabetic retinopathy. Eur. J. Ophthalmol. 21, 637–643 (2011).

193. Gao, B.-B. et al. Extracellular carbonic anhydrase mediates hemorrhagic retinal and cerebral vascular permeability through prekallikrein activation. Nat. Med. 13, 181–188 (2007).

194. Reber, F., Gersch, U. & Funk, R. H. W. Blockers of carbonic anhydrase can cause increase of retinal capillary diameter, decrease of extracellular and increase of

188

intracellular pH in rat retinal organ culture. Graefes Arch. Clin. Exp. Ophthalmol. 241, 140–148 (2003).

195. Pedersen, D. B. et al. Carbonic anhydrase inhibition increases retinal oxygen tension and dilates retinal vessels. Graefes Arch. Clin. Exp. Ophthalmol. 243, 163–168 (2005).

196. Lenzen, S. The mechanisms of alloxan- and streptozotocin-induced diabetes. Diabetologia 51, 216–226 (2008).

197. Misra, M. & Aiman, U. Alloxan: an unpredictable drug for diabetes induction? Indian J Pharmacol 44, 538–539 (2012).

198. Schroder, S., Palinski, W. & Schmid-Schonbein, G. W. Activated monocytes and granulocytes, capillary nonperfusion, and neovascularization in diabetic retinopathy. Am. J. Pathol. 139, 81–100 (1991).

199. Kern, T. S. et al. Response of capillary cell death to aminoguanidine predicts the development of retinopathy: comparison of diabetes and galactosemia. Invest. Ophthalmol. Vis. Sci. 41, 3972–3978 (2000).

200. Kowluru, R. A., Tang, J. & Kern, T. S. Abnormalities of retinal metabolism in diabetes and experimental galactosemia. VII. Effect of long-term administration of antioxidants on the development of retinopathy. Diabetes 50, 1938–1942 (2001).

201. Wei, M. et al. The streptozotocin-diabetic rat as a model of the chronic complications of human diabetes. Hear. Lung Circ 12, 44–50 (2003).

202. Qaum, T. et al. VEGF-initiated blood-retinal barrier breakdown in early diabetes. Invest. Ophthalmol. Vis. Sci. 42, 2408–2413 (2001).

203. Barber, A. J. et al. Neural apoptosis in the retina during experimental and human diabetes. Early onset and effect of insulin. J. Clin. Invest. 102, 783–791 (1998).

204. Johnson, E. I., Dunlop, M. E. & Larkins, R. G. Increased vasodilatory prostaglandin production in the diabetic rat retinal vasculature. Curr. Eye Res. 18, 79–82 (1999).

205. Miyamoto, K. et al. Prevention of leukostasis and vascular leakage in streptozotocin- induced diabetic retinopathy via intercellular adhesion molecule-1 inhibition. Proc. Natl. Acad. Sci. U.S.A. 96, 10836–10841 (1999).

206. Antonetti, D. A. et al. Vascular permeability in experimental diabetes is associated with reduced endothelial occludin content - Vascular endothelial growth factor decreases occludin in retinal endothelial cells. Diabetes 47, 1953–1959 (1998).

207. Brouwers, B. et al. Phlorizin pretreatment reduces acute renal toxicity in a mouse model for diabetic nephropathy. J. Biol. Chem. 288, 27200–27207 (2013).

208. Mecklenburg, L. & Schraermeyer, U. An Overview on the Toxic Morphological Changes in the Retinal Pigment Epithelium after Systemic Compound Administration. Toxicol. Pathol. 35, 252–267 (2007).

189

209. Kermorvant-Duchemin, E. et al. Neonatal Hyperglycemia Inhibits Angiogenesis and Induces Inflammation and Neuronal Degeneration in the Retina. PLoS One 8, e79545 (2013).

210. Graham, M. L., Janecek, J. L., Kittredge, J. A., Hering, B. J. & Schuurman, H. J. The Streptozotocin-Induced Diabetic Nude Mouse Model: Differences between Animals from Different Sources. Comp. Med. 61, 356–360 (2011).

211. Robinson, R., Barathi, V. A., Chaurasia, S. S., Wong, T. Y. & Kern, T. S. Update on animal models of diabetic retinopathy: from molecular approaches to mice and higher mammals. Dis Model Mech 5, 444–456 (2012).

212. Ricci, B. Oxygen-induced retinopathy in the rat model. Doc Ophthalmol 74, 171–177 (1990).

213. Scott, A. & Fruttiger, M. Oxygen-induced retinopathy: a model for vascular pathology in the retina. Eye 24, 416–421 (2010).

214. Rakoczy, E. P. et al. Characterization of a Mouse Model of Hyperglycemia and Retinal Neovascularization. Am. J. Pathol. 177, 2659–2670 (2010).

215. Kern, T. S. & Engerman, R. L. Galactose-induced retinal microangiopathy in rats. Invest. Ophthalmol. Vis. Sci. 36, 490–496 (1995).

216. Aung, M. H., Kim, M. K., Olson, D. E., Thule, P. M. & Pardue, M. T. Early visual deficits in streptozotocin-induced diabetic long evans rats. Investig. Ophthalmol. Vis. Sci. 54, 1370–1377 (2013).

217. NCD Risk Factor Collaboration (NCD-RisC). Worldwide trends in blood pressure from 1975 to 2015: a pooled analysis of 1479 population-based measurement studies with 19·1 million participants. Lancet 389, 37–55 (2017).

218. Pardue, M. T. et al. Rodent Hyperglycemia-Induced Inner Retinal Deficits are Mirrored in Human Diabetes. Transl. Vis. Sci. Technol. 3, 6 (2014).

219. Zhang, L. et al. Protection of the heart by treatment with a divalent-copper-selective chelator reveals a novel mechanism underlying cardiomyopathy in diabetic rats. Cardiovasc Diabetol 12, 123 (2013).

220. Freeman, W. M. et al. A multistep validation process of biomarkers for preclinical drug development. Pharmacogenomics J. 10, 385–395 (2010).

221. Canning, P. et al. Inhibition of advanced glycation and absence of galectin-3 prevent blood-retinal barrier dysfunction during short-term diabetes. Exp. Eye Res. 2007, 51837 (2007).

222. Layton, C. J. et al. Monocarboxylate transporter expression remains unchanged during the development of diabetic retinal neuropathy in the rat. Invest. Ophthalmol. Vis. Sci. 46, 2878–2885 (2005).

190

223. Rathcke, C. N. & Vestergaard, H. YKL-40, a new inflammatory marker with relation to insulin resistance and with a role in endothelial dysfunction and atherosclerosis. Inflamm. Res. 55, 221–7 (2006).

224. Vestal, D. J., Gorbacheva, V. Y. & Sen, G. C. Different Subcellular Localizations for the Related Interferon-Induced GTPases, MuGBP-1 and MuGBP-2: Implications for Different Functions? J. Interf. Cytokine Res. 20, 991–1000 (2000).

225. Neun, R., Richter, M. F., Staeheli, P. & Schwemmle, M. GTPase properties of the interferon-induced human guanylate-binding protein 2. FEBS Lett. 390, 69–72 (1996).

226. Johnsen-Soriano, S. et al. IL-2 and IFN-gamma in the retina of diabetic rats. Graefe’s Arch. Clin. Exp. Ophthalmol. 248, 985–990 (2010).

227. Kim, H.-J., Hwang, N. R. & Lee, K.-J. Heat shock responses for understanding diseases of protein denaturation. Mol. Cells 23, 123–131 (2007).

228. Pinach, S. et al. Retinal heat shock protein 25 in early experimental diabetes. Acta Diabetol. 50, 579–585 (2013).

229. O’Shea, J. J., Pesu, M., Borie, D. C. & Changelian, P. S. A new modality for immunosuppression: targeting the JAK/STAT pathway. Nat. Rev. Drug Discov. 3, 555– 564 (2004).

230. Dudley, A. C., Thomas, D., Best, J. & Jenkins, A. A VEGF/JAK2/STAT5 axis may partially mediate endothelial cell tolerance to hypoxia. Biochem. J. 390, 427–436 (2005).

231. Tibbles, H. E. et al. Role of a JAK3-dependent Biochemical Signaling Pathway in Platelet Activation and Aggregation. J. Biol. Chem. 276, 17815–17822 (2001).

232. Markowska, A. I., Jefferies, K. C. & Panjwani, N. Galectin-3 protein modulates cell surface expression and activation of vascular endothelial Growth factor receptor 2 in human endothelial cells. J. Biol. Chem. 286, 29913–29921 (2011).

233. Bauer, P. M. et al. Inflamed In Vitro Retina: Cytotoxic Neuroinflammation and Galectin-3 Expression. PLoS One 11, e0161723 (2016).

234. Brew, K., Dinakarpandian, D. & Nagase, H. Tissue inhibitors of metalloproteinases: evolution, structure and function. Biochim. Biophys. Acta - Protein Struct. Mol. Enzymol. 1477, 267–283 (2000).

235. Agapova, O. A., Ricard, C. S., Salvador-Silva, M. & Hernandez, M. R. Expression of matrix metalloproteinases and tissue inhibitors of metalloproteinases in human optic nerve head astrocytes. Glia 33, 205–16 (2001).

236. Padgett, L., Lui, G., Werb, Z. & Lavail, M. Matrix Metalloproteinase-2 and Tissue Inhibitor of Metalloproteinase-1 in the Retinal Pigment Epithelium and Interphotoreceptor Matrix: Vectorial Secretion and Regulation. Exp. Eye Res. 64, 927–938 (1997).

237. Yamada, E. et al. TMP-1 promotes VEGF-induced neovascularization in the retina.

191

Histol. Histopathol. 16, 87–97 (2001).

238. Gerl, V. B. et al. Extensive deposits of complement C3d and C5b-9 in the choriocapillaris of eyes of patients with diabetic retinopathy. Investig. Ophthalmol. Vis. Sci. 43, 1104– 1108 (2002).

239. Gao, B. B. et al. Extracellular carbonic anhydrase mediates hemorrhagic retinal and cerebral vascular permeability through prekallikrein activation. Nat. Med. 13, 181–188 (2007).

240. Mullins, R. F. et al. Localization of complement 1 inhibitor (C1INH/SERPING1) in human eyes with age-related macular degeneration. Exp. Eye Res. 89, 767–73 (2009).

241. Hu, M., Bruun, a & Ehinger, B. Expression of GABA transporter subtypes (GAT1, GAT3) in the adult rabbit retina. Acta Ophthalmol. Scand. 77, 255–260 (1999).

242. Ramsey, D. J., Ripps, H. & Qian, H. Streptozotocin-induced diabetes modulates GABA receptor activity of rat retinal neurons. Exp. Eye Res. 85, 413–422 (2007).

243. Dunn, L. L. et al. A critical role for thioredoxin- Interacting protein in diabetes-related impairment of angiogenesis. Diabetes 63, 675–687 (2014).

244. Hammes, H. P., Lin, J., Bretzel, R. G., Brownlee, M. & Breier, G. Upregulation of the vascular endothelial growth factor/vascular endothelial growth factor receptor system in experimental background diabetic retinopathy of the rat. Diabetes 47, 401–406 (1998).

245. Sueishi, K. et al. Endothelial and glial cell interaction in diabetic retinopathy via the function of vascular endothelial growth factor (VEGF). Pol. J. Pharmacol. 48, 307–16

246. Ugarte, M., Osborne, N. N., Brown, L. A. & Bishop, P. N. Iron, zinc, and copper in retinal physiology and disease. Surv. Ophthalmol. 58, 585–609 (2013).

247. Levy, J., Gavin, J. R. & Sowers, J. R. Diabetes mellitus: A disease of abnormal cellular calcium metabolism? Am. J. Med. 96, 260–273 (1994).

248. Cooper, G. J. Therapeutic potential of copper chelation with triethylenetetramine in managing diabetes mellitus and Alzheimer’s disease. Drugs 71, 1281–1320 (2011).

249. Nagai, R., Murray, D. B., Metz, T. O. & Baynes, J. W. Chelation: A fundamental mechanism of action of AGE inhibitors, AGE breakers, and other inhibitors of diabetes complications. Diabetes 61, 549–559 (2012).

250. Song, D. & Dunaief, J. L. Retinal iron homeostasis in health and disease. Front. Aging Neurosci. 5, (2013).

251. Ugarte, M. et al. Concentration of various trace elements in the rat retina and their distribution in diff erent structures Concentration of various trace elements in the rat retina and their distribution in different structures. Metallomics 41140, 59–1245 (2012).

252. Agilent Technologies. Agilent 7700 Series ICP-MS Techniques and Operation Course Number R1777A Lecture Manual. (2012).

192

253. Thomas, R. A beginner’s guide to ICP-MS - Part III: The plasma source. Spectroscopy 16, 26–30 (2001).

254. Thomas, R. A beginner’s guide to ICP-MS Part I. Spectroscopy 16, (2001).

255. Thomas, R. A beginner’s guide to ICP--MS: Part IX -- Mass analyzers: collision/reaction cell technology. Spectroscopy 17, 42–49 (2002).

256. Zawilska, J. B. The role of calcium in the regulation of melatonin biosynthesis in the retina. Acta Neurobiol. Exp. (Wars). 52, 265–74 (1992).

257. Luk, E., Carroll, M., Baker, M. & Culotta, V. C. Manganese activation of superoxide dismutase 2 in Saccharomyces cerevisiae requires MTM1, a member of the mitochondrial carrier family. Proc. Natl. Acad. Sci. U. S. A. 100, 10353–10357 (2003).

258. Dong, A. et al. Superoxide dismutase 1 protects retinal cells from oxidative damage. J. Cell. Physiol. 208, 516–526 (2006).

259. Kasahara, E., Lin, L.-R., Ho, Y.-S. & Reddy, V. N. SOD2 protects against oxidation- induced apoptosis in mouse retinal pigment epithelium: implications for age-related macular degeneration. Invest. Ophthalmol. Vis. Sci. 46, 3426–3434 (2005).

260. Moiseyev, G. et al. RPE65 Is an Iron(II)-dependent Isomerohydrolase in the Retinoid Visual Cycle. J. Biol. Chem. 281, 2835–2840 (2005).

261. Moiseyev, G., Chen, Y., Takahashi, Y., Wu, B. X. & Ma, J.-X. RPE65 is the isomerohydrolase in the retinoid visual cycle. Proc. Natl. Acad. Sci. U. S. A. 102, 12413– 8 (2005).

262. Gnana-Prakasam, J. P. et al. Iron-mediated retinal degeneration in haemojuvelin- knockout mice. Biochem. J. 441, 599–608 (2012).

263. Rogers, B. S. et al. Differential sensitivity of cones to iron-mediated oxidative damage. Investig. Ophthalmol. Vis. Sci. 48, 438–445 (2007).

264. Jaiswal, M. et al. Impaired mitochondrial energy production causes light-induced photoreceptor degeneration independent of oxidative stress. PLoS Biol. 13, (2015).

265. Ferreira, R. C., Heckenlively, J. R., Menkes, J. H. & Bronwyn Bateman, J. Menkes disease: New ocular and electroretinographic findings. Ophthalmology 105, 1076–1078 (1998).

266. Mishima, K., Amemiya, T. & Takano, K. X-ray microanalysis of melanin granules of retinal pigment epithelium and choroid in hereditary copper deficient mice (macular mice). Exp. Eye Res. 68, 59–65 (1999).

267. Satishohandra, P. & Ravishankar Naik, K. Visual pathway abnormalities Wilson’s disease. J. Neurol. Sci. 176, 13–20 (2000).

268. Public Health England. NHS Diabetes Prevention Programme (NHS DPP) Non-diabetic hyperglycaemia. (2015).

193

269. McHarg, S., Brace, N., Bishop, P. N. & Clark, S. J. Enrichment of Bruch’s Membrane from Human Donor Eyes. J. Vis. Exp. (2015). doi:10.3791/53382

270. Eckhert, C. D. Elemental concentrations in ocular tissues of various species. Exp. Eye Res. 37, 639–647 (1983).

271. Hahn, P., Ying, G., Beard, J. & Dunaief, J. L. Iron levels in human retina: sex difference and increase with age. Neuroreport 17, 1803–6 (2006).

272. Erie, J. C., Good, J. A., Butz, J. A. & Pulido, J. S. Reduced Zinc and Copper in the Retinal Pigment Epithelium and Choroid in Age-related Macular Degeneration. Am. J. Ophthalmol. 147, 276–282.e1 (2009).

273. Erie, J. C. et al. Heavy Metal Concentrations in Human Eyes. Am. J. Ophthalmol. 139, 888–893 (2005).

274. Kokavec, J. et al. Biochemical analysis of the living human vitreous. Clin. Experiment. Ophthalmol. 44, 597–609 (2016).

275. Gong, D. et al. Quantitative proteomic profiling identifies new renal targets of copper(II)- selective chelation in the reversal of diabetic nephropathy in rats. Proteomics 9, 4309– 4320 (2009).

276. Brings, S. et al. Diabetes-induced alterations in tissue collagen and carboxymethyllysine in rat kidneys: Association with increased collagen-degrading proteinases and amelioration by Cu(II)-selective chelation. Biochim. Biophys. Acta - Mol. Basis Dis. 1852, 1610–1618 (2015).

277. Linder, M. C. et al. Copper transport. Am. J. Clin. Nutr. 67, 965S–971S (1998).

278. Wu, Z.-Y., Lin, M.-T., Murong, S.-X. & Wang, N. Molecular diagnosis and prophylactic therapy for presymptomatic Chinese patients with Wilson disease. Arch. Neurol. 60, 737–741 (2003).

279. Cooper, G. J. et al. Regeneration of the heart in diabetes by selective copper chelation. Diabetes 53, 2501–2508 (2004).

280. Hyvönen, M. T. et al. Metabolism of triethylenetetramine and 1,12-diamino-3,6,9- triazadodecaneby the spermidine/spermine-N1-acetyltransferase and thialysine acetyltransferase. Drug Metab. Dispos. 41, 30–32 (2013).

281. Lu, J. et al. Triethylenetetramine and metabolites: Levels in relation to copper and zinc excretion in urine of healthy volunteers and type 2 diabetic patients. Drug Metab. Dispos. 35, 221–227 (2007).

282. Shin, W. W., Fong, W. F., Pang, S. F. & Wong, P. C. Limited blood-brain barrier transport of polyamines. J. Neurochem. 44, 1056–9 (1985).

283. Wichmann, K. A. et al. Characterization of dicarboxylic salts of protonated triethylenetetramine useful for the treatment of copper-related pathologies. in Cryst

194

Growth Des 7, 1844–1850 (2007).

284. Cai, L., Chen, S., Evans, T., Cherian, M. G. & Chakrabarti, S. Endothelin-1-mediated alteration of metallothionein and trace metals in the liver and kidneys of chronically diabetic rats. Int. J. Exp. Diabetes Res. 3, 193–198 (2002).

285. Heuchel, R. et al. The transcription factor MTF-1 is essential for basal and heavy metal- induced metallothionein gene expression. EMBO J. 13, 2870–5 (1994).

286. Tapia, L., Gon, M. & Gon Alez, M. Metallothionein is crucial for safe intracellular copper storage and cell survival at normal and supra-physiological exposure levels. Biochem. J. 378, 617–624 (2004).

287. Suemori, S. et al. Metallothionein, an endogenous antioxidant, protects against retinal neuron damage in mice. Investig. Ophthalmol. Vis. Sci. 47, 3975–3982 (2006).

288. Shen, Y., Liu, X.-L. & Yang, X.-L. N-methyl-D-aspartate receptors in the retina. Mol. Neurobiol. 34, 163–179 (2006).

289. Kusari, J., Zhou, S., Padillo, E., Clarke, K. G. & Gil, D. W. Effect of memantine on neuroretinal function and retinal vascular changes of streptozotocin-induced diabetic rats. Investig. Ophthalmol. Vis. Sci. 48, 5152–5159 (2007).

290. Kusari, J., Zhou, S. X., Padillo, E., Clarke, K. G. & Gil, D. W. Inhibition of vitreoretinal VEGF elevation and blood-retinal barrier breakdown in streptozotocin-induced diabetic rats by brimonidine. Investig. Ophthalmol. Vis. Sci. 51, 1044–1051 (2010).

291. Baptiste, D. C. et al. Comparison of the neuroprotective effects of adrenoceptor drugs in retinal cell culture and intact retina. Investig. Ophthalmol. Vis. Sci. 43, 2666–2676 (2002).

292. Nakamura, S. et al. Role of metallothioneins 1 and 2 in ocular neovascularization. Invest. Ophthalmol. Vis. Sci. 55, 6851–6860 (2014).

293. Elner, S. G., Elner, V. M., Yoshida, A., Dick, R. D. & Brewer, G. J. Effects of tetrathiomolybdate in a mouse model of retinal neovascularization. Invest. Ophthalmol. Vis. Sci. 46, 299–303 (2005).

294. Sato, M. & Gitlin, J. D. Mechanisms of copper incorporation during the biosynthesis of human ceruloplasmin. J. Biol. Chem. 266, 5128–5134 (1991).

295. Martin, F. et al. Copper-dependent activation of hypoxia-inducible factor (HIF)-1: Implications for ceruloplasmin regulation. Blood 105, 4613–4619 (2005).

296. Feng, W., Ye, F., Xue, W., Zhou, Z. & Kang, Y. J. Copper regulation of hypoxia-inducible factor-1 activity. Mol. Pharmacol. 75, 174–182 (2009).

297. Osaki, S. Kinetic studies of ferrous ion oxidation with crystalline human ferroxidase (ceruloplasmin). J. Biol. Chem. 241, 5053–5059 (1966).

298. Mukhopadhyay, C. K., Attieh, Z. K. & Fox, P. L. Role of ceruloplasmin in cellular iron

195

uptake. Science 279, 714–7 (1998).

299. Lutsenko, S., Barnes, N. L., Bartee, M. Y. & Dmitriev, O. Y. Function and regulation of human copper-transporting ATPases. Physiol. Rev. 87, 1011–1046 (2007).

300. Barnes, N., Tsivkovskii, R., Tsivkovskaia, N. & Lutsenko, S. The copper-transporting ATPases, Menkes and Wilson disease proteins, have distinct roles in adult and developing cerebellum. J. Biol. Chem. 280, 9640–9645 (2005).

301. Zheng, Z. et al. Altered microglial copper homeostasis in a mouse model of Alzheimer’s disease. J. Neurochem. 114, 1630–1638 (2010).

302. Schlief, M. L., Craig, A. M. & Gitlin, J. D. NMDA receptor activation mediates copper homeostasis in hippocampal neurons. J. Neurosci. 25, 239–246 (2005).

303. Schlief, M. L., West, T., Craig, A. M., Holtzman, D. M. & Gitlin, J. D. Role of the Menkes copper-transporting ATPase in NMDA receptor-mediated neuronal toxicity. Proc. Natl. Acad. Sci. 103, 14919–14924 (2006).

304. Burdo, J. R. et al. Cellular distribution of iron in the brain of the Belgrade rat. Neuroscience 93, 1189–1196 (1999).

305. Lin, C., Zhang, Z., Wang, T., Chen, C. & James Kang, Y. Copper uptake by DMT1: a compensatory mechanism for CTR1 deficiency in human umbilical vein endothelial cells. Metallomics 7, 1285–9 (2015).

306. He, X. et al. Iron homeostasis and toxicity in retinal degeneration. Progress in Retinal and Eye Research 26, 649–673 (2007).

307. Arredondo, M., Muñoz, P., Mura, C. & Nùñez, M. DMT1, a physiologically relevant apical Cu1+ transporter of intestinal cells. Am. J. Physiol. Cell Physiol. 284, C1525–C1530 (2003).

308. Tennant, J., Stansfield, M., Yamaji, S., Srai, S. K. & Sharp, P. Effects of copper on the expression of metal transporters in human intestinal Caco-2 cells. FEBS Lett. 527, 239– 244 (2002).

309. Lin, C., Zhang, Z., Wang, T., Chen, C. & James Kang, Y. Copper uptake by DMT1: a compensatory mechanism for CTR1 deficiency in human umbilical vein endothelial cells. Metallomics 7, 1285–9 (2015).

310. Petris, M. J., Smith, K., Lee, J. & Thiele, D. J. Copper-stimulated endocytosis and degradation of the human copper transporter, hCtr1. J. Biol. Chem. 278, 9639–9646 (2003).

311. Prohaska, J. R., Broderius, M. & Brokate, B. Metallochaperone for Cu,Zn-superoxide dismutase (CCS) protein but not mRNA is higher in organs from copper-deficient mice and rats. Arch. Biochem. Biophys. 417, 227–234 (2003).

312. Leary, S. C., Winge, D. R. & Cobine, P. A. ‘Pulling the plug’ on cellular copper: The role

196

of mitochondria in copper export. Biochim. Biophys. Acta-Molecular Cell Res. 1793, 146–153 (2009).

313. Wee, N. K., Weinstein, D. C., Fraser, S. T. & Assinder, S. J. The mammalian copper transporters CTR1 and CTR2 and their roles in development and disease. Int J Biochem Cell Biol 45, 960–963 (2013).

314. Polishchuk, R. & Lutsenko, S. Golgi in copper homeostasis: a view from the membrane trafficking field. Histochem. Cell Biol. 140, 285–295 (2013).

315. Vestal, D. J., Gorbacheva, V. Y. & Sen, G. C. Different Subcellular Localizations for the Related Interferon-Induced GTPases, MuGBP-1 and MuGBP-2: Implications for Different Functions? J. Interf. Cytokine Res. 20, 991–1000 (2000).

316. Gao, X. et al. Interferon-gamma protects against cuprizone-induced demyelination. Mol. Cell. Neurosci. 16, 338–349 (2000).

317. Lee, J., Kim, S. J., Son, T. G., Chan, S. L. & Mattson, M. P. Interferon-gamma is up- regulated in the hippocampus in response to intermittent fasting and protects hippocampal neurons against excitotoxicity. J. Neurosci. Res. 83, 1552–1557 (2006).

318. Nishiyama, A. et al. Identification of thioredoxin-binding protein-2/vitamin D3 up- regulated protein 1 as a negative regulator of thioredoxin function and expression. J. Biol. Chem. 274, 21645–21650 (1999).

319. Patwari, P., Higgins, L. J., Chutkow, W. A., Yoshioka, J. & Lee, R. T. The interaction of thioredoxin with Txnip: Evidence for formation of a mixed disulfide by disulfide exchange. J. Biol. Chem. 281, 21884–21891 (2006).

320. Singh, L. P. Thioredoxin Interacting Protein (TXNIP) and Pathogenesis of Diabetic Retinopathy. J. Clin. Exp. Ophthalmol. 4, (2013).

321. Dotimas, J. R. et al. Diabetes regulates fructose absorption through thioredoxin- interacting protein. Elife 5, (2016).

322. Du, C. et al. Thioredoxin-interacting protein regulates lipid metabolism via Akt/mTOR pathway in diabetic kidney disease. Int. J. Biochem. Cell Biol. 79, 1–13 (2016).

323. Chen, J. et al. TXNIP regulates myocardial fatty acid oxidation via miR-33a signaling. Am. J. Physiol. - Hear. Circ. Physiol. 311, H64–H75 (2016).

324. Mandala, A. et al. Thioredoxin interacting protein mediates lipid-induced impairment of glucose uptake in skeletal muscle. Biochem. Biophys. Res. Commun. 479, 933–939 (2016).

325. Hansen, J. M., Zhang, H. & Jones, D. P. Differential oxidation of thioredoxin-1, thioredoxin-2, and glutathione by metal ions. Free Radic. Biol. Med. 40, 138–145 (2006).

326. Wills, N. K., Ramanujam, V. M., Kalariya, N., Lewis, J. R. & van Kuijk, F. J. Copper and zinc distribution in the human retina: relationship to cadmium accumulation, age, and

197

gender. Exp. Eye Res. 87, 80–88 (2008).

327. Langford-Smith, A. et al. Age and Smoking Related Changes in Metal Ion Levels in Human Lens: Implications for Cataract Formation. PLoS One 11, 1–16 (2016).

328. Wills, N. K. et al. Cadmium accumulation in the human retina: Effects of age, gender, and cellular toxicity. Exp. Eye Res. 86, 41–51 (2008).

329. Tchounwou, P. B., Yedjou, C. G., Patlolla, A. K. & Sutton, D. J. in Molecular, Clinical and Environmental Toxicology 101, 133–164 (2012).

330. Ellis, N. I., Lloyd, B., Lloyd, R. S. & Clayton, B. E. Selenium and vitamin E in relation to risk factors for coronary heart disease. J. Clin. Pathol. 37, 200–6 (1984).

331. Cekic, O. Effect of cigarette smoking on copper, lead, and cadmium accumulation in human lens. Br J Ophthalmol 82, 186–8 (1998).

332. Martelli, A., Rousselet, E., Dycke, C., Bouron, A. & Moulis, J.-M. Cadmium toxicity in animal cells by interference with essential metals. Biochimie 88, 1807–1814 (2006).

333. Aziz, R. et al. Impact assessment of cadmium toxicity and its bioavailability in human cell lines (Caco-2 and HL-7702). Biomed Res. Int. 2014, 839538 (2014).

334. Jung, J. S. et al. Chemical characteristics of long-range transboundary air pollutants from Asian continent observed at two remote sites and one urban site in Korea. Geochim. Cosmochim. Acta 73, A609–A609 (2009).

335. Kocyigit, A., Erel, O. & Gur, S. Effects of tobacco smoking on plasma selenium, zinc, copper and iron concentrations and related antioxidative enzyme activities. Clin. Biochem. 34, 629–633 (2001).

336. Satarug, S. & Moore, M. R. Adverse health effects of chronic exposure to low-level cadmium in foodstuffs and cigarette smoke. Env. Heal. Perspect 112, 1099–1103 (2004).

337. Paulsen, L., Holm, C., Bech, J. N., Starklint, J. & Pedersen, E. B. Effects of statins on renal sodium and water handling Acute and short-term effects of atorvastatin on renal haemodynamics, tubular function, vasoactive hormones, blood pressure and pulse rate in healthy, normocholesterolemic humans. Nephrol Dial Transpl. 23, 1556–1561 (2008).

338. Castelli, W. P. & Anderson, K. A population at risk: Prevalence of high cholesterol levels in hypertensive patients in the framingham study. Am. J. Med. 80, 23–32 (1986).

339. Cundy, T. & Dissanayake, A. Severe hypomagnesaemia in long-term users of proton- pump inhibitors. Clin. Endocrinol. (Oxf). 69, 338–341 (2008).

340. Sheen, E. & Triadafilopoulos, G. Adverse effects of long-term proton pump inhibitor therapy. Dig. Dis. Sci. 56, 931–950 (2011).

341. Toh, J. W. T., Ong, E. & Wilson, R. Hypomagnesaemia associated with long-term use of proton pump inhibitors. Gastroenterol. Rep. 3, 243–253 (2015).

198

342. William, J. H. & Danziger, J. Proton-pump inhibitor-induced hypomagnesemia: Current research and proposed mechanisms. World J. Nephrol. 5, 152–7 (2016).

343. DeFronzo, R. A. The effect of insulin on renal sodium metabolism - A review with clinical implications. Diabetologia 21, 165–171 (1981).

344. Provenzano, L. F., Stark, S., Steenkiste, A., Piraino, B. & Sevick, M. A. Dietary Sodium Intake in Type 2 Diabetes. Clin. Diabetes 32, 106–112 (2014).

345. Diabetes UK. New diabetes prevalence figures for England - Diabetes UK. (2016). Available at: https://www.diabetes.org.uk/About_us/News/New-diabetes-prevalence- figures-for-England/. (Accessed: 1st January 2017)

346. Scheiber, I. F. & Dringen, R. Copper-treatment increases the cellular GSH content and accelerates GSH export from cultured rat astrocytes. Neurosci. Lett. 498, 42–46 (2011).

347. Dimitriadis, G., Mitrou, P., Lambadiari, V., Maratou, E. & Raptis, S. A. Insulin effects in muscle and adipose tissue. Diabetes Res. Clin. Pract. 93, S52–S59 (2011).

348. Dagher, Z. et al. Studies of rat and human retinas predict a role for the polyol pathway in human diabetic retinopathy. Diabetes 53, 2404–2411 (2004).

349. Hites, R. a. Gas Chromatography Mass Spectrometry. Handb. Instrum. Tech. Anal. Chem. 609–626 (1997). doi:10.1016/B978-0-12-663971-1.50012-9

350. Villas-Bôas, S. G., Mas, S., Åkesson, M., Smedsgaard, J. & Nielsen, J. Mass spectrometry in metabolome analysis. Mass Spectrom. Rev. 24, 613–646 (2005).

351. Ruiz-Matute, A. I., Hernández-Hernández, O., Rodríguez-Sánchez, S., Sanz, M. L. & Martínez-Castro, I. Derivatization of carbohydrates for GC and GC-MS analyses. J. Chromatogr. B. Analyt. Technol .Biomed. Life Sci. 879, 1226–1240 (2011).

352. Paris, L. P. et al. Global metabolomics reveals metabolic dysregulation in ischemic retinopathy. Metabolomics 12, 1–10 (2016).

353. Chen, L. et al. Plasma metabonomic profiling of diabetic retinopathy. Diabetes 65, 1099– 1108 (2016).

354. Sas, K. M. et al. Tissue-specific metabolic reprogramming drives nutrient flux in diabetic complications. JCI Insight 1, (2016).

355. Frayser, R. & Buse, M. G. Branched Chain Amino Acid Metabolism in the Retina of Diabetic Rats. Diabetologia 14, 171–176 (1978).

356. Patassini, S. et al. Metabolite mapping reveals severe widespread perturbation of multiple metabolic processes in Huntington’s disease human brain. Biochim. Biophys. Acta - Mol. Basis Dis. 1862, 1650–1662 (2016).

357. Hutton, J. C., Schofield, P. J., Williams, F. & Hollows, F. C. Sorbitol Metabolism in the Retina: Accumulation of Pathway Intermediates in Streptozotocin Induced Diabetes in the Rat. Aust. J. Exp. Biol. Med. Sci. 52, 361–373 (1974).

199

358. Heath, H. & Hamlett, Y. C. The Sorbitol Pathway: Effect of Streptozotocin Induced Diabetes and the Feeding of a Sucrose-Rich Diet on Glucose, Sorbitol and Fructose in the Retina, Blood and Liver of Rats. Diabetologia 12, 43–46 (1976).

359. Heath, H., Kang, S. S. & Philippou, D. Glucose, glucose-6-phosphate, lactate and pyruvate content of the retina, blood and liver of streptozotocin-diabetic rats fed sucrose- or starch-rich diets. Diabetologia 11, 57–62 (1975).

360. Sas, K. M. et al. Tissue-specific metabolic reprogramming drives nutrient flux in diabetic complications. JCI Insight 1, (2016).

361. Tilton, R. G. et al. Inhibition of sorbitol dehydrogenase: Effects on vascular and neural dysfunction in streptozocin-induced diabetic rats. Diabetes 44, 234–242 (1995).

362. Poulsom, R., Mirrlees, D. J., Earl, D. C. N. & Heath, H. The effects of an aldose reductase inhibitor upon the sorbitol pathway, fructose-1-phosphate and lactate in the retina and nerve of streptozotocin-diabetic rats. Exp. Eye Res. 36, 751–760 (1983).

363. Salceda, R., Vilchis, C., Coffe, V. & Hernández-Muñoz, R. Changes in the redox state in the retina and brain during the onset of diabetes in rats. Neurochem. Res. 23, 893–897 (1998).

364. Freeman, O. J. et al. Metabolic dysfunction is restricted to the sciatic nerve in experimental diabetic neuropathy. Diabetes 65, (2016).

365. Cameron, N. E., Cotter, M. A., Basso, M. & Hohman, T. C. Comparison of the effects of inhibitors of aldose reductase and sorbitol dehydrogenase on neurovascular function, nerve conduction and tissue polyol pathway metabolites in streptozotocin-diabetic rats. Diabetologia 40, 271–281 (1997).

366. Vinores, S. A. et al. Aldose Reductase Expression in Human Diabetic Retina and Retinal Pigment Epithelium. Diabetes 37, (1988).

367. Kador, P. F. et al. Purified rat lens aldose reductase Polyol production in vitro and its inhibition by aldose reductase inhibitors. Biochem. J. 240, 233–237 (1986).

368. Lopez, M. G. & Feather, M. S. The Production of Threose as a degradation product from Ascorbic Acid. J. Carbohydr. Chem. 11, 799–806 (1992).

369. Shin, D. & Feather, M. The Degradation of L-Ascorbic Acid in Neutral Solutions Containing Oxygen. J. Carbohydr. Chem. 9, 461–469 (1990).

370. Devamanoharan, P. S. & Varma, S. D. Studies on L-threose as substrate for aldose reductase: A possible role in preventing protein glycation. Mol. Cell. Biochem. 159, 123– 127 (1996).

371. Simpson, G. L. W. & Ortwerth, B. J. The non-oxidative degradation of ascorbic acid at physiological conditions. Biochim. Biophys. Acta - Mol. Basis Dis. 1501, 12–24 (2000).

372. Som, S. et al. Ascorbic acid metabolism in diabetes mellitus. Metab. Clin. Exp. 30, 572–

200

577 (1981).

373. Sherman, W. R., Stewart, M. A., Kurien, M. M. & Goodwin, S. L. The measurement of myo-inositol, myo-inosose-2 and scyllo-inositol in mammalian tissues. Biochim. Biophys. Acta. 158, 197–205 (1968).

374. Yorek, M. A. et al. Reduced motor nerve conduction velocity and Na+-K+-ATPase activity in rats maintained on L-fucose diet: Reversal by myo-inositol supplementation. Diabetes 42, 1401–1406 (1993).

375. Fisher, S. K., Novak, J. E. & Agranoff, B. W. Inositol and higher inositol phosphates in neural tissues: Homeostasis, metabolism and functional significance. J. Neurochem. 82, 736–754 (2002).

376. Yamashita, Y., Yamaoka, M., Hasunuma, T., Ashida, H. & Yoshida, K.-I. Detection of Orally Administered Inositol Stereoisomers in Mouse Blood Plasma and Their Effects on Translocation of Glucose Transporter 4 in Skeletal Muscle Cells. J. Agric. Food Chem. 61, 4850–4854 (2013).

377. Dorr, A. et al. Amyloid-β-dependent compromise of microvascular structure and function in a model of Alzheimer’s disease. Brain 135, 3039–50 (2012).

378. Nishimura, C. & Kuriyama, K. Alterations in the Retinal Dopaminergic Neuronal System in Rats with Streptozotocin-Induced Diabetes. J. Neurochem. 45, 448–455 (1985).

379. Gowda, K., Zinnantif, W. J. & LaNoue, K. F. The influence of diabetes on glutamate metabolism in retinas. J. Neurochem. 117, 309–320 (2011).

380. Eizirik, D. L., Germano, C. M. & Migliorini, R. H. Dietetic supplementation with branched chain amino acids attenuates the severity of streptozotocin-induced diabetes in rats. Acta Diabetol. Lat. 25, 117–126 (1988).

381. Newgard, C. B. et al. A Branched-Chain Amino Acid-Related Metabolic Signature that Differentiates Obese and Lean Humans and Contributes to Insulin Resistance. Cell Metab. 9, 311–326 (2009).

382. Lynch, C. J. & Adams, S. H. Branched-chain amino acids in metabolic signalling and insulin resistance. Nat. Rev. Endocrinol. 10, 723–736 (2014).

383. Rodriguez, T. et al. The increased skeletal muscle protein turnover of the streptozotocin diabetic rat is associated with high concentrations of branched-chain amino acids. Biochem. Mol. Med. 61, 87–94 (1997).

384. Lian, K. et al. Impaired adiponectin signaling contributes to disturbed catabolism of branched-chain amino acids in diabetic mice. Diabetes 64, 49–59 (2015).

385. Turer, A. T. & Scherer, P. E. Adiponectin: Mechanistic insights and clinical implications. Diabetologia 55, 2319–2326 (2012).

386. Fernstrom, M. H., Volk, E. A. & Fernstrom, J. D. In vivo inhibition of tyrosine uptake into

201

rat retina by large neutral but not acidic amino acids. Am. J. Physiol. 251, E393--E399 (1986).

387. Anuradha, C. V. & Selvam, R. Effect of oral methionine on tissue lipid peroxidation and antioxidants in alloxan-induced diabetic rats. J. Nutr. Biochem. 4, 212–217 (1993).

388. Brazionis, L., Rowley, K., Itsiopoulos, C., Harper, C. A. & O’Dea, K. Homocysteine and diabetic retinopathy. Diabetes Care 31, 50–56 (2008).

389. Maryon, E. B., Zhang, J., Jellison, J. W. & Kaplan, J. H. Human copper transporter 1 lacking O-linked glycosylation is proteolytically cleaved in a Rab9-positive endosomal compartment. J. Biol. Chem. 284, 28104–28114 (2009).

390. Maryon, E. B., Molloy, S. A. & Kaplan, J. H. O-linked glycosylation at threonine 27 protects the copper transporter hCTR1 from proteolytic cleavage in mammalian cells. J. Biol. Chem. 282, 20376–20387 (2007).

391. Dodd, K. M., Yang, J., Shen, M. H., Sampson, J. R. & Tee, A. R. mTORC1 drives HIF-1α and VEGF-A signalling via multiple mechanisms involving 4E-BP1, S6K1 and STAT3. Oncogene 34, 2239–50 (2015).

392. Jackson, C. R. et al. Retinal Dopamine Mediates Multiple Dimensions of Light-Adapted Vision. J. Neurosci. 32, 9359–9368 (2012).

393. Bonnefont-Rousselot, D. & Collin, F. Melatonin: action as antioxidant and potential applications in human disease and aging. Toxicology 278, 55–67 (2010).

394. Galano, A., Medina, M. E., Tan, D. X. & Reiter, R. J. Melatonin and its metabolites as copper chelating agents and their role in inhibiting oxidative stress: A physicochemical analysis. J. Pineal Res. 58, 107–116 (2015).

395. Buonfiglio, D. do C. et al. Early-stage retinal melatonin synthesis impairment in streptozotocin-induced diabetic wistar rats. Investig. Ophthalmol. Vis. Sci. 52, 7416– 7422 (2011).

396. Özdemir, G. et al. Melatonin prevents retinal oxidative stress and vascular changes in diabetic rats. Eye 28, 1020–1027 (2014).

397. Martin, P. M. et al. Expression and localization of GPR109A (PUMA-G/HM74A) mRNA and protein in mammalian retinal pigment epithelium. Mol. Vis. 15, 362–372 (2009).

398. Gambhir, D. et al. GPR109A as an anti-inflammatory receptor in retinal pigment epithelial cells and its relevance to diabetic retinopathy. Invest. Ophthalmol. Vis. Sci. 53, 2208–2217 (2012).

399. Giusto, N. M. & Bazán, N. G. Phospholipids and acylglycerols biosynthesis and 14CO2 production from [14C]glycerol in the bovine retina: The effects of incubation time, oxygen and glucose. Exp. Eye Res. 29, 155–168 (1979).

400. Guida, M., Salvatore, M. M. & Salvatore, F. A Strategy for GC/MS Quantification of Polar

202

Compounds via their Silylated Surrogates: Silylation and Quantification of Biological Amino Acids. J .Anal. Bioanal. Tech. 6, 263 (2015).

401. Darbre, A. & Islam, A. Gas-liquid chromatography of trifluoroacetylated amino acid methyl esters. Biochem J. 106, 923–5 (1968).

402. Halket, J. M. et al. Chemical derivatization and mass spectral libraries in metabolic profiling by GC/MS and LC/MS/MS. in J. Exp. Bot. 56, 219–243 (2005).

403. Schofield, J. D., Liu, Y., Rao-Balakrishna, P., Malik, R. A. & Soran, H. Diabetes Dyslipidemia. Diabetes Ther. 7, 203–19 (2016).

404. Dean, J. D. & Durrington, P. N. Treatment of Dyslipoproteinaemia in Diabetes Mellitus. Diabet. Med. 13, 297–312 (1996).

405. Ginsberg, H. N. Diabetic dyslipidemia: Basic mechanisms underlying the common hypertriglyceridemia and low HDL cholesterol levels. Diabetes 45, (1996).

406. Goldberg, I. J. Diabetic Dyslipidemia: Causes and Consequences. J. Clin. Endocrinol. Metab. 86, 965–971 (2001).

407. Chew, E. Y. et al. Association of elevated serum lipid levels with retinal hard exudate in diabetic retinopathy. Early Treatment Diabetic Retinopathy Study (ETDRS) Report 22. Arch. Ophthalmol. (Chicago, Ill. 1960) 114, 1079–84 (1996).

408. Klein, R., Klein, B. E., Moss, S. E. & Cruickshanks, K. J. The Wisconsin Epidemiologic Study of Diabetic Retinopathy. XV. The long-term incidence of macular edema. Ophthalmology 102, 7–16 (1995).

409. Miljanovic, B., Glynn, R. J., Nathan, D. M., Manson, J. E. & Schaumberg, D. a. A Prospective Study of Serum Lipids and Risk of Diabetic Macular Edema in Type 1 Diabetes. Diabetes 53, 2883–2892 (2004).

410. Thermo Fisher Scientific. LTQ Orbitrap VelosTM Hardware Manual. Thermo Fisher Scientific Inc. (2009).

411. Godzien, J. et al. Rapid and Reliable Identification of Phospholipids for Untargeted Metabolomics with LC-ESI-QTOF-MS/MS. J. Proteome Res. 14, 3204–3216 (2015).

412. Eliuk, S. & Makarov, A. Evolution of Orbitrap Mass Spectrometry Instrumentation. Annu. Rev. Anal. Chem. 8, 61–80 (2015).

413. Zubarev, R. A. & Makarov, A. Orbitrap mass spectrometry. Anal. Chem. 85, 5288–5296 (2013).

414. Smith, C. A., Want, E. J., O’Maille, G., Abagyan, R. & Siuzdak, G. XCMS: processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification. Anal. Chem. 78, 779–87 (2006).

415. Tautenhahn, R., Böttcher, C. & Neumann, S. Highly sensitive feature detection for high resolution LC/MS. BMC Bioinformatics 9, 504 (2008).

203

416. Kuhl, C., Tautenhahn, R., Böttcher, C., Larson, T. R. & Neumann, S. CAMERA: An integrated strategy for compound spectra extraction and annotation of liquid chromatography/mass spectrometry data sets. Anal. Chem. 84, 283–289 (2012).

417. Sandra, K., Pereira, A. D. S., Vanhoenacker, G., David, F. & Sandra, P. Comprehensive blood plasma lipidomics by liquid chromatography/quadrupole time-of-flight mass spectrometry. J. Chromatogr. A 1217, 4087–4099 (2010).

418. Contrepois, K., Jiang, L. & Snyder, M. Optimized Analytical Procedures for the Untargeted Metabolomic Profiling of Human Urine and Plasma by Combining Hydrophilic Interaction (HILIC) and Reverse-Phase Liquid Chromatography (RPLC)–Mass Spectrometry. Mol. Cell. Proteomics 14, 1684–1695 (2015).

419. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Series B Stat. Methodol. 57, 289– 300 (1995).

420. Moriguchi, T., Loewke, J., Garrison, M., Catalan, J. N. & Salem, N. Reversal of docosahexaenoic acid deficiency in the rat brain, retina, liver, and serum. J. Lipid Res. 42, 419–427 (2001).

421. Hegde, K. R. & Varma, S. D. Electron impact mass spectroscopic studies on mouse retinal fatty acids: effect of diabetes. Ophthalmic Res. 42, 9–14 (2009).

422. Sapieha, P. et al. Omega-3 polyunsaturated fatty acids preserve retinal function in type 2 diabetic mice. Nutr. Diabetes 2, e36 (2012).

423. Lydic, T. a, Renis, R., Busik, J. V & Reid, G. E. Analysis of Retina and Erythrocyte Glycerophospholipid Alterations in a Rat Model of Type 1 Diabetes. JALA Charlottesv. Va. 14, 383–399 (2009).

424. Vergès, B. Pathophysiology of diabetic dyslipidaemia: where are we? Diabetologia 58, 886–899 (2015).

425. Liu, C.-C. et al. Neuronal LRP1 regulates glucose metabolism and insulin signaling in the brain. J. Neurosci. 35, 5851–9 (2015).

426. Amaratunga, A. et al. Apolipoprotein E is synthesized in the retina by Müller glial cells, secreted into the vitreous, and rapidly transported into the optic nerve by retinal ganglion cells. J. Biol. Chem. 271, 5628–5632 (1996).

427. Bu, G. Apolipoprotein E and its receptors in Alzheimer’s disease: pathways, pathogenesis and therapy. Nat Rev Neurosci 10, 334–344 (2009).

428. Liebisch, G. et al. Shorthand notation for lipid structures derived from mass spectrometry. J. Lipid Res. 54, 1523–1530 (2013).

429. Fliesler, S. & Anderson, R. Chemistry and metabolism of lipids in the vertebrate retina. Prog. Lipid Res. 22, 79–131 (1983).

204

430. Lodish, H. et al. Molecular Cell Biology. 4th edition. New York: W. H. Freeman (2000). doi:10.1017/CBO9781107415324.004

431. Kennedy, E. P. & Weiss, S. B. The function of cytidine coenzymes in the biosynthesis of phospholipides. J. Biol. Chem. 222, 193–214 (1956).

432. Kent, C. Regulatory enzymes of phosphatidylcholine biosynthesis: A personal perspective. BBA - Molecular and Cell Biology of Lipids 1733, 53–66 (2005).

433. Kanno, K., Wu, M. K., Scapa, E. F., Roderick, S. L. & Cohen, D. E. Structure and function of phosphatidylcholine transfer protein (PC-TP)/StarD2. BBA - Molecular and Cell Biology of Lipids 1771, 654–662 (2007).

434. Mateos, M. V., Uranga, R. M., Salvador, G. A. & Giusto, N. M. Activation of phosphatidylcholine signalling during oxidative stress in synaptic endings. Neurochem. Int. 53, 199–206 (2008).

435. Shishova, E. Y. et al. Genetic ablation or chemical inhibition of phosphatidylcholine transfer protein attenuates diet-induced hepatic glucose production. Hepatology 54, 664–674 (2011).

436. McMahon, H. T. & Boucrot, E. Membrane curvature at a glance. J. Cell Sci. 128, 1065– 1070 (2015).

437. Hansen, H. S., Moesgaard, B., Hansen, H. H. & Petersen, G. N-acylethanolamines and precursor phospholipids - Relation to cell injury. Chem. Phys. Lipids 108, 135–150 (2000).

438. Menon, A. K. & Stevens, V. L. Phosphatidylethanolamine is the donor of the ethanolamine residue linking a glycosylphosphatidylinositol anchor to protein. J. Biol. Chem. 267, 15277–15280 (1992).

439. Vance, J. E. & Tasseva, G. Formation and function of phosphatidylserine and phosphatidylethanolamine in mammalian cells. BBA - Molecular and Cell Biology of Lipids 1831, 543–554 (2013).

440. Sparrow, J. R., Wu, Y., Kim, C. Y. & Zhou, J. Phospholipid meets all-trans-retinal: the making of RPE bisretinoids. J. Lipid Res. 51, 247–261 (2010).

441. Kuczynski, B. & Reo, N. V. Evidence that plasmalogen is protective against oxidative stress in the rat brain. Neurochem. Res. 31, 639–656 (2006).

442. Maeba, R., Sawada, Y., Shimasaki, H., Takahashi, I. & Ueta, N. Ethanolamine plasmalogens protect cholesterol-rich liposomal membranes from oxidation caused by free radicals. Chem. Phys. Lipids 120, 145–151 (2002).

443. Rog, T. & Koivuniemi, A. The biophysical properties of ethanolamine plasmalogens revealed by atomistic molecular dynamics simulations. Biochim. Biophys. Acta - Biomembr. 1858, 97–103 (2016).

205

444. Dorninger, F. et al. Homeostasis of phospholipids - The level of phosphatidylethanolamine tightly adapts to changes in ethanolamine plasmalogens. Biochim. Biophys. Acta 1851, 117–128 (2015).

445. Daum, G. & Vance, J. E. Import of lipids into mitochondria. Prog. Lipid Res. 36, 103–130 (1997).

446. Paradies, G., Paradies, V., De Benedictis, V., Ruggiero, F. M. & Petrosillo, G. Functional role of cardiolipin in mitochondrial bioenergetics. Biochimica et Biophysica Acta - Bioenergetics 1837, 408–417 (2014).

447. Ferreira, R. et al. Lipidomic characterization of streptozotocin-induced heart mitochondrial dysfunction. Mitochondrion 13, 762–771 (2013).

448. Zhong, Q. & Kowluru, R. A. Diabetic retinopathy and damage to mitochondrial structure and transport machinery. Investig. Ophthalmol. Vis. Sci. 52, 8739–8746 (2011).

449. Arismendi-Morillo, G. Electron microscopy morphology of the mitochondrial network in gliomas and their vascular microenvironment. Biochimica et Biophysica Acta - Bioenergetics 1807, 602–608 (2011).

450. Rajala, R. V. S. Phosphoinositide 3-kinase signaling in the vertebrate retina. J. Lipid Res. 51, 4–22 (2010).

451. Nakamura, J., Del Monte, M. A., Shewach, D., Lattimer, S. A. & Greene, D. A. Inhibition of phosphatidylinositol synthase by glucose in human retinal pigment epithelial cells. Am. J. Physiol. 262, E417-26 (1992).

452. Fenili, D., Brown, M., Rappaport, R. & McLaurin, J. Properties of scyllo-inositol as a therapeutic treatment of AD-like pathology. J. Mol. Med. 85, 603–611 (2007).

453. Jiang, B. H., Zheng, J. Z., Aoki, M. & Vogt, P. K. Phosphatidylinositol 3-kinase signaling mediates angiogenesis and expression of vascular endothelial growth factor in endothelial cells. Proc. Natl. Acad. Sci. U. S. A. 97, 1749–1753 (2000).

454. Gousseva, N., Kugathasan, K., Chesterman, C. N. & Khachigian, L. M. Early growth response factor-1 mediates insulin-inducible vascular endothelial cell proliferation and regrowth after injury. J. Cell. Biochem. 81, 523–534 (2001).

455. Lal, B. K., Varma, S., Pappas, P. J., Hobson, R. W. & Durán, W. N. VEGF Increases Permeability of the Endothelial Cell Monolayer by Activation of PKB/akt, Endothelial Nitric-Oxide Synthase, and MAP Kinase Pathways. Microvasc. Res. 62, 252–262 (2001).

456. Barber, A. J. et al. Insulin Rescues Retinal Neurons from Apoptosis by a Phosphatidylinositol 3-Kinase/Akt-mediated Mechanism That Reduces the Activation of Caspase-3. J. Biol. Chem. 276, 32814–32821 (2001).

457. He, F. et al. Phosphatidylinositol-3-phosphate is light-regulated and essential for survival in retinal rods. Sci. Rep. 6, 26978 (2016).

206

458. Rajala, A. et al. G-protein-coupled receptor rhodopsin regulates the phosphorylation of retinal insulin receptor. J. Biol. Chem. 282, 9865–9873 (2007).

459. Kitatani, K., Nemoto, M., Akiba, S. & Sato, T. Stimulation by de novo-synthesized ceramide of phospholipase A2-dependent cholesterol esterification promoted by the uptake of oxidized low-density lipoprotein in macrophages. Cell. Signal. 14, 695–701 (2002).

460. Gómez del Pulgar, T., Velasco, G., Sánchez, C., Haro, A. & Guzmán, M. De novo- synthesized ceramide is involved in cannabinoid-induced apoptosis. Biochem. J. 363, 183–8 (2002).

461. Kitatani, K., Idkowiak-Baldys, J. & Hannun, Y. A. The sphingolipid salvage pathway in ceramide metabolism and signaling. Cell. Signal. 20, 1010–1018 (2008).

462. Goldkorn, T. et al. H2O2 acts on cellular membranes to generate ceramide signaling and initiate apoptosis in tracheobronchial epithelial cells. J. Cell Sci. 111 ( Pt 2, 3209–20 (1998).

463. Fox, T. E. et al. Diabetes alters sphingolipid metabolism in the retina: A potential mechanism of cell death in diabetic retinopathy. Diabetes 55, 3573–3580 (2006).

464. Liu, Y. Y., Hill, R. A. & Li, Y. T. Ceramide Glycosylation Catalyzed by Glucosylceramide Synthase and Cancer Drug Resistance. Adv. Cancer Res. 117, 59–89 (2013).

465. Pannu, R., Singh, A. K. & Singh, I. A novel role of lactosylceramide in the regulation of tumor necrosis factor alpha-mediated proliferation of rat primary astrocytes. Implications for astrogliosis following neurotrauma. neurotrauma. J. Biol. Chem. 280, 13742–13751 (2005).

466. Zemski Berry, K. a, Gordon, W. C., Murphy, R. C. & Bazan, N. G. Spatial organization of lipids in the human retina and optic nerve by MALDI imaging mass spectrometry. J. Lipid Res. 55, 504–15 (2014).

467. Watt, M. J. & Steinberg, G. R. Regulation and function of triacylglycerol lipases in cellular metabolism. Biochem. J. 414, 313–25 (2008).

468. Joyal, J.-S. et al. Retinal lipid and glucose metabolism dictates angiogenesis through the lipid sensor Ffar1. Nat. Med. 22, 439–445 (2016).

469. Wang, Y. Tissue-specific, nutritional, and developmental regulation of rat fatty acid elongases. J. Lipid Res. 46, 706–715 (2005).

470. Chilton, F. H. et al. Diet-gene interactions and PUFA metabolism: A potential contributor to health disparities and human diseases. Nutrients 6, 1993–2022 (2014).

471. Tikhonenko, M. et al. Remodeling of Retinal Fatty Acids in an Animal Model of Diabetes A Decrease in Long-Chain Polyunsaturated Fatty Acids Is Associated With a Decrease in Fatty Acid Elongases Elovl2 and Elovl4. Diabetes 59, 219–227 (2010).

207

472. Bedell, M., Harkewicz, R., Wang, X. & Zhang, K. Focus on Molecules: ELOVL4. Exp. Eye Res. 90, 476–477 (2010).

473. Agbaga, M. P. et al. Role of Elovl4 protein in the biosynthesis of docosahexaenoic acid. in Advances in Experimental Medicine and Biology 664, 233–242 (2010).

474. German, O., Miranda, G., Abrahan, C. & Rotstein, N. Ceramide is a mediator of apoptosis in retina photoreceptors. Invest. Ophthalmol. Vis. Sci. 47, 1658–1668 (2006).

475. Hammes, H. P. et al. Acceleration of experimental diabetic retinopathy in the rat by omega-3 fatty acids. Diabetologia 39, 251–255 (1996).

476. Nishizuka, Y. Intracellular signaling by hydrolysis of phospholipids and activation of protein kinase C. Science (80-. ). 258, 607–614 (1992).

477. Ford, D. A. et al. Lipidomic analysis of the retina in a rat model of Smith-Lemli-Opitz syndrome: Alterations in docosahexaenoic acid content of phospholipid molecular species. J. Neurochem. 105, 1032–1047 (2008).

478. Wainio, W. W., Vander Wende, C. & Shimp~, N. F. Copper in Cytochrome c Oxidase*. J. Biol. Chem. 234, (1959).

479. Horn, D. & Barrientos, A. Mitochondrial copper metabolism and delivery to cytochrome C oxidase. IUBMB Life 60, 421–429 (2008).

480. Hosseini, M.-J., Shaki, F., Ghazi-Khansari, M. & Pourahmad, J. Toxicity of Copper on Isolated Liver Mitochondria: Impairment at Complexes I, II, and IV Leads to Increased ROS Production. Cell Biochem. Biophys. 70, 367–381 (2014).

481. Osorio-Paz, I., Uribe-Carvajal, S. & Salceda, R. In the early stages of diabetes, rat retinal mitochondria undergo mild uncoupling due to UCP2 activity. PLoS One 10, (2015).

482. Burkhead, J. L. & Lutsenko, S. The Role of Copper as a Modifier of Lipid Metabolism. Lipid Metab. 39–60 (2013). doi:10.5772/51819

483. Krishnamoorthy, L. et al. Copper regulates cyclic-AMP-dependent lipolysis. Nat. Chem. Biol. 12, 586–592 (2016).

484. Jeon, Y. H. et al. Phosphodiesterase: Overview of protein structures, potential therapeutic applications and recent progress in drug development. Cell. Mol. Life Sci. 62, 1198–1220 (2005).

485. Fields, M., Ferretti, R. J., Smith, J. C. & Reiser, S. Effect of copper deficiency on metabolism and mortality in rats fed sucrose or starch diets. J. Nutr. 113, 1335–45 (1983).

486. Fields, M., Ferretti, R. J., Smith, J. C. & Reiser, S. The interaction of type of dietary carbohydrates with copper deficiency. Am. J. Clin. Nutr. 39, 289–295 (1984).

487. Burkhead, J. L. & Lutsenko, S. The Role of Copper as a Modifier of Lipid Metabolism. Lipid Metab. 39–60 (2013). doi:10.5772/51819

208

488. Kinnard, R. L., Narasimhan, B., Pliska-Matyshak, G. & Murthy, P. P. N. Characterization of scyllo-inositol-containing phosphatidylinositol in plant cells. Biochem. Biophys. Res. Commun. 210, 549–555 (1995).

489. Meng, Q. et al. Systems Nutrigenomics Reveals Brain Gene Networks Linking Metabolic and Brain Disorders. EBioMedicine 7, 157–66 (2016).

490. Yee, P., Weymouth, A. E. A., Fletcher, E. L. EL & Vingrys, A. A. J. A role for omega-3 polyunsaturated fatty acid supplements in diabetic neuropathy. Invest. Ophthalmol. Vis. Sci. 51, 1755–64 (2010).

491. Nguyen, C. T. O., Vingrys, A. J. & Bui, B. V. Dietary omega-3 fatty acids and ganglion cell function. Investig. Ophthalmol. Vis. Sci. 49, 3586–3594 (2008).

492. Giebel, S. J., Menicucci, G., McGuire, P. G. & Das, A. Matrix metalloproteinases in early diabetic retinopathy and their role in alteration of the blood-retinal barrier. Lab. Investig. 85, 597–607 (2005).

493. Matsubara, T., Murata, T., Wu, G.-S., Barron, E. & Rao, N. A. Current Eye Research Isolation and culture of rat retinal microvessel endothelial cells using magnetic beads coated with antibodies to PECAM-1 Isolation and culture of rat retinal microvessel endothelial cells using magnetic beads coated with antibodies to PECAM-1. Curr. Eye Res. 0, 271–3683 (2001).

494. Ogra, Y. et al. Changes in intracellular copper concentration and copper-regulating gene expression after PC12 differentiation into neurons. Sci. Rep. 6, 33007 (2016).

495. Amaral, A. I., Alves, P. M. & Teixeira, A. P. in 107–144 (Humana Press, New York, NY, 2014). doi:10.1007/978-1-4939-1059-5_5

209

Appendices

210

Appendix I: List of Primer used in rt-qPCR analysis

Symbol Gene Forward Primer Sequence (5’–3’) Reverse Primer Sequence (5’–3’)

ActB Actin, beta GTCCACACCCGCCACCAGTTC GAAGACGGCCCGGGGAGCAT

Antioxidant 1 copper Atox1 AAAGCGGTCTCCTACCTTGG AGCTGGACTGAGCAGTTGGT chaperone

ATPase, Cu++ Atp7a transporting, alpha TGCAAGCAGGAGCTGGACAC TTCCACAGCCGGGTACAACC polypeptide

ATPase, Cu++ Atp7b transporting, beta AACGGCGTCCTAATCAAGG TGCCCGTTTTGTCAAACATA polypeptide

Cp Ceruloplasmin CTTCACAAACCGGAAGCAAA ATTGGCTGAATGCTGAGAGG

Chi3l1 Chitinase 3-Like 1 TTGCCCAGATAGCCCAACACC TCGGAAGAGGGGGCTGTGAT

C1nh Complement 1-inhibitor AGCCCAGATCTAGCCATACGGG TGGTTGGTGTTCTCCGCCAC

Copper Metabolism Murr1 (Murr1) Domain GTCACGGCACTCAACTCAAA TCCCGTCCACTTCTCCCAAA Containing 1

Ctr1 Copper Transporter 1 CAACACACCTGGAGAAATGG CGGGCTATCTTGAGTCCTTC

Ctr2 Copper Transporter 2 CTAGTGGGCTCACCATGCCG GCCATGCCTGTAGGGCTGTG

Copper-chaperone for Ccs CTGTGCACAAGACCCTGAAA CCATCTGGTTCTCCAACTGAA superoxide dismutase

Cytochrome c oxidase Cu Cox17 TCAGGAGAAGAAGCCTCTGAA TTCTCCTTTCTCAATGATGCAC chaperone

Cytochrome c oxidase Sco1 GAAGCTTGGGCCTGTTTCTTGG TGCCGCTGTTTCTCCAGCTTT subunit 1

Divalent Metal Dmt1 AACGAATAGGCTGGAGGAT TGTTGATGGAGCAGACGAGA Transporter 1

Gat3 GABA transporter 3 CTCTGGGGCAGTTCACGAGC TGTTGCATAGCCGATGCCTTCA

Guanylate nucleotide Gbp2 GCCACAGAGCCCGTGAAAAA TAACATGTGGATCTCCGAGGC binding protein 2

Heat shock 27kDa Hspb1 GGCGTGGTGGAGATCACTGG ACACCTGGAGGGAGCGTGTAT protein 1

Jak3 Janus kinase 3 CCATCTTCTGGTACGCCCCT TCATACAGCACCACGCCGAA

Lectin, galactoside- AGTGATACTGTTTGCGTTGGGCT Lgals3 ATGCCCTTGCCTGGAGGAGT binding, soluble, 3 T

Mt1 Metallothionein 1 AGTGACGAACAGTGCTGCTG TCGGTAGAAAACGGGGTTTA

Mt2 Metallothionein 2 GCGATCTCTCGTTGATCTCC GCATTTGCATTGTTTGCATT

Natriuretic peptide TCCAGGTGGTCTAGCAGGTTCTT Nppa TTCCTGGCCTTTTGGCTC precursor A G

Ndc1 Nucleoporin 1 TTCCCAAAGCATGGATTAGC CAGCCAGACATGGTAGAGCA

Superoxide dismutase 2 Sod2 ACCACGCGACCTACGTGAAC TGTAACATCTCCCTTGGCCAGC (Mn) (mitochondrial)

Tbp TATA-binding Protein AGAACAATCCAGACTAGCAGCA GGGAACTTCACATCACAGCTC

211

Tissue inhibitor of Timp1 CACAGGTTTCCGGTTCGCCT AAACGGCCCGCGATGAGAAA metalloproteinase 1

Thioredoxin-interacting Txnip GTCCCAGCCAGCCAACTCAA GGCAGACACTGGTGCCATTA protein

Vascular endothelial Vegfa CGGACGGGCCTCTGAAACC CCTGGGACCACTTGGCATGG growth factor A

212

Appendix II: Medical records of Human donors

PM Ocular disease or Medication/prescription drugs (noted when ETR SEX AGE Diagnosis Cause of Death Smoke Diabetes Additional Notes TIME surgery patient information is lacking

Diabetic donors 32 T2DM., AMD, ETR16 F 78 PMP – Yes – Type 2 AMD, retinopathy – – hrs retinopathy

56 Former ETR24 F 71 Diabetes Lung cancer Yes – NR No – – hrs smoker

45 Cataracts in one ETR57 F 73 T1DM – – Yes – Type 1 – Insulin, Losartan (blood pressure) hrs eye

Multi organ Former ETR70 F 75 28 T1DM Yes – Type 1 glasses – Insulin failure smoker

Diabetes, Rampril, furosemide, salbutamol, insulin ETR75 F 62 43 PMP – Yes – NR Glasses Nephropathy possible DR GuluoRx

Contacts & ETR76 F 77 44 Diabetes PMP – Yes – NR – – Glasses

ETR143 F 80 38 Diabetes – – Yes – NR No – Fentanyl and morphine(pain)

T2DM & Glaucoma (1992), ETR148 F 83 32 – – Yes – Type 2 Family history of diabetes – glaucoma Cataract (2013)

Simvastatin (cholesterol lowering), sodium ETR150 F 80 29 T2DM – – Yes – Type 2 Glasses Diabetic neuropathy bicarbonate (heartburn), metformin (T2DM), bumetanide (Heart failure), coracten (angina),

ETR187 F 77 39 T2DM – – Yes – Type 2 Glasses – –

Insulin– novofine needles, unistik 3 lancets, Former Cataracts, artificial furosemide (oedema), clopidogrel (blood clots) ETR224 F 84 34 T2DM – Yes – Type 2 – smoker lens 2013 bisoprolol (blood pressure) ferrous sulphate (iron) , losartan (blood pressure)

213

PM Ocular disease or Medication/prescription drugs (noted when ETR SEX AGE Diagnosis Cause of Death Smoke Diabetes Additional Notes TIME surgery patient information is lacking

Diet controlled diabetes since ETR232 F 58 42 T2DM, glaucoma NR – Yes – Type 2 Glaucoma Lyrica (epilepsy), ferrous fumarate (anaemia) 2005

Pre–diabetes Cataracts 5 years ETR233 F 83 33 and macular Pneumonia – Pre-diabetes ago, macular – Salbutamol, fluticasone, salmeterol (asthma) degeneration degeneration

Versatis (neuropathic pain), linagliptina T2DM and Leg ulcers, breathing problems (metporhphin – diabetes) , humulin m3 (insulin), ETR244 F 72 34 – – Yes – Type 2 Glaucoma glaucoma (chronic kidney disease) salbutamol (breathing), furosemide (oedema), fentanyl patch (pain)

CCF, T2DM & ETR252 F 70 40 Diabetes – Yes Cataracts – – CKD

Yes – Type 1 ETR261 F 83 44 T1DM PMP – Cataracts 2010 Diabetes diet controlled – since 2009

Right cataracts removed Sub 2015 – bilateral Simvastatin (cholesterol), ramipril ETR323 F 81 30 Pre-diabetes Yes Pre-diabetes (growth found in eye), left Haemorrhage cataracts (hypertension), awaiting removal

Furosemide (oedema), qvar (asthma), bisoprolol Former Yes – Type 2 Cataract surgery ETR326 F 86 45 T2DM – COPD, CCF (blood pressure), clopidogrel (antiplatelet), smoker since 2012 2013 ferrous fumarate (iron), colchicine (gout),

Pravastatin (cholesterol), amlodipine (high blood pressure), hypromellose (eye drops), nateglinide ETR330 F 61 32 T2DM – No Yes – Type 2 No – (T2DM, metformin (diabetes), ferrous fumarate (anaemia), candesartan (hypertension)

Allopurinol (gout), ramipril (hypertension), Yes – recently Vascular degenerative disease, atenolol (beta blocker), levodropropizine ETR333 F 62 38 T2DM PMP Yes – Type 2 Reading glasses started carotid stenosis (Cough), omeprazole (GERD), naproxen (arthiritis), simvastatin (cholesterol)

Hypertension, cerebrovascular ETR350 F 75 40 T2DM – No Yes – Type 2 – accident, gout, anaemia, chronic – renal failure

Cataracts surgery Family history of non-insulin Metformin (diabetes), hydroxocobalamin (Vit ETR371 F 89 48 T2DM – No Yes – Type 2 (2014) dependent diabetes B12), nifedipine(Coronary vasodilator)

214

PM Ocular disease or Medication/prescription drugs (noted when ETR SEX AGE Diagnosis Cause of Death Smoke Diabetes Additional Notes TIME surgery patient information is lacking

Yes – Type 2 ETR376 F 87 46 T2DM PMP Yes NR – Simvastatin (cholesterol), omeprazole (GERD) (2007)

Warfarin (anticoagluant), adizem (calcium Hyperplasia, CVA, left ETR378 F 89 43 T2DM Stroke No Yes – Type 2 NR channel blocker), lansoprazole (gerd), ventricular hypertrophy furosemide (oedema)

Omeprazole (GERD), sitagliptin, metformin ETR383 F 56 44 – – No Yes – NR NR – (T2DM), simvastatin (cholesterol), quinine sulfate (malaria), ferrous sulfate (anaemia)

Yes, diet Glaucoma: right Patient had regular diabetic eye ETR276 M 78 36 T2DM, glaucoma – – controlled Type eye 21 mmHg, left – clinics. 2 eye 22mmHg

Viscotears (eye drops), bumetanide (edema),levothyroxine (thyroid hormone), T2DM , dry eye Yes – Type 2 dry eyes syndrome ETR271 F 66 38 – – Diverticulosis (2011) Gliclazide, metformin (T2DM), pregabalin syndrome (2011) (2014) (neuropathic pain) glyceryl trinitrate (angina/heart failure)

visual acuity defect, type II Yes – Type 2, diabetes diagnosed, macular Multi organ Furosemide, larsoprazole, apixaban, solifanacin, ETR62 M 76 34 T2DM, DR – insulin Yes changes, photocoagulation DR failure linagliptin, statins dependent screening, central retinal occlusion

Diabetes – Diet Yes – Diet Leg ulcers, Excess alcohol ETR104 M 68 46 PMP – Reading glasses – controlled controlled intake, duodenal ulcer

Diabetes – diet Yes – Diet ETR90 M 62 34 Cancer – Glasses – – controlled Controlled

13 Former ETR47 M 70 Diabetes Cancer Yes – Type 2 No – – hrs smoker

Hypromellose (eye drops), lansoprazole, ETR132 M 84 49 Pre-diabetes PMP – Pre-diabetes Reading glasses Impaired glucose regulation tamsulosin

Diabetes, Severe retinopathy Received photocoagulation and ETR93 M 75 44 Stroke – Yes – severe DR and cataracts had haemorrhages,

215

PM Ocular disease or Medication/prescription drugs (noted when ETR SEX AGE Diagnosis Cause of Death Smoke Diabetes Additional Notes TIME surgery patient information is lacking

Nephrostomy, COPD, previous ETR122 M 62 36 T2DM – – Yes – Type 2 Reading glasses Gliclazide (T2DM), candesartan excess alcohol

Neutropenic Allopurinol (kidney stones), nifedipine (coronary ETR139 M 82 31 T2DM sepsis & ovarian – Yes – Type 2 Glasses – vasodilator), doxazosin, oxybutynin cancer

~34 Former ETR14 M 74 T2DM MI Yes – Type 2 No – – hrs smoker

Clopidogrel (antiplatelet), metformin, gliclazide ETR151 M 72 35 T2DM – – Yes – Type 2 Glasses – (T2DM), simvastatin (Cholesterol),

Ferrous fumarate (iron), salbutamol, beclomethasone (asthma), omeprazole ETR152 M 72 36 T2DM – – Yes– Type 2 No Diabetes since 2000 (stomach), Pioglitazone, gliclazide (T2DMM), simvastatin (cholesterol),

Right eye ETR162 M 66 36 T2DM Cancer No Yes – Type 2 – – stigmatism

Diabetic complications, bilateral ETR172 M 65 25 T2DM – – Yes – Type 2 No below knee amputation, – previous vascular surgery

45 ETR19 M 79 T2DM Not reported – Yes – Type 2 – – – hrs

67 Heart failure and Previous cataract ETR33 M 68 T2DM – Yes –Type 2 – – hrs renal disease surgery

76 Previous cataract Chronic Renal Failure via ETR34 M 69 T2DM Cardiac arrest Yes Yes –Type 2 Chronic Renal Failure via diabetes hrs surgery diabetes

Yes – Type 2, 42 Former Insulin dependent, leg ETR46 M 66 T2DM PMP insulin right eye cataract Insulin hrs smoker amputation from foot ulcer dependent

32 ETR48 M 59 T2DM – – Yes – Type 2 Cataracts – – hrs

216

PM Ocular disease or Medication/prescription drugs (noted when ETR SEX AGE Diagnosis Cause of Death Smoke Diabetes Additional Notes TIME surgery patient information is lacking

No retinopathy as of 11/14. ETR67 M 74 32 T2DM – – Yes – Type 2 cataract surgery COPD, Angina, bronchitis, – rheumatoid arthiritis

ETR72 M 59 55 T2DM Pneumonia – Yes – Type 2 No – –

DR, retinal Neuropathy, laser treatment T2DM and detachment, ETR113 M 81 25 Not reported – Yes – Type 2 detachment to correct rods, Insulin severe DR cataracts, diabetic Coronary angiography maculopathy

T2DM, Glaucoma (2013), ETR157 M 85 48 – – Yes – Type 2 Diabetes since 2013 Terbutaline (asthma), latanoprost (glaucoma) glaucoma, DR DR (2013/4)

Undiagnosed Secondary Amputated leg 2015, leg ulcers, ETR127 M 68 46 secondary – – undiagnosed Glasses Psychoactive drugs swollen foot diabetes diabetes

Aspirin, atorvastatin, sugar testing strips, Yes– cataract ETR194 M 71 45 T2DM Cardiac arrest – Yes – Type 2 dapagliflozin, doxazosin, gabapentin, isosorbide, surgery levothyroxine, metformin, novomix, perindopril

Aspirin, bisoprolol, dexamethasone, ETR206 M 78 46 T2DM – – Yes – Type 2 No CKD, MI omeprazole, pregabalin, simvastatin, tamsulosin

omeprazole (GERD), bisoprolol (blood ETR255 M 81 49 T2DM – NR Yes – Type 2 – – pressure), sitagliptin, gliclazide (diabetes), ramipril (hypertension), simvastatin (cholesterol)

Non–diabetic donors

ETR77 F 71 44 – PMP – – – – –

ETR87 F 73 48 – – – – – Alcohol Abuse –

ETR131 F 75 28 – PMP – – – – –

217

PM Ocular disease or Medication/prescription drugs (noted when ETR SEX AGE Diagnosis Cause of Death Smoke Diabetes Additional Notes TIME surgery patient information is lacking

Other – ETR88 F 62 46 – Ischaemic heart – – Reading glasses – – disease

Dry eyes, early ETR92 F 82 28 – Other – Cardiac – – cataracts, – – blepharitis

Respiratory ETR94 F 77 40 – Yes – Reading glasses Leg Ulcer – failure

Cataracts both ETR142 F 80 44 – – – No – – eyes

Hypoxic brain ETR147 F 80 47 – – No – 2 bottles wine/week – damage

ETR149 F 83 30 – Renal Failure – No No – –

Lifelong Glaucoma, tear ETR119 F 77 32 Glaucoma – – – – smoker drop operation

Dry eyes, early ETR92 F 82 28 – Other – Cardiac NR – cataracts, – – blepharitis

Metronidazole, chloramphenicol, cefaclor ETR240 F 55 48 – PMP No No Reading glasses Jaundice (infections)

simvastatin (cholesterol lowering) seretide ETR241 F 55 48 – PMP No No No – (asthma), monomil (angina), bisoprolol , ramipril,

Botox for dropping ETR214 F 82 48 – – – – – – eyelids

Intracranial ETR179 F 72 32 – – – No – – haemorrhage

218

PM Ocular disease or Medication/prescription drugs (noted when ETR SEX AGE Diagnosis Cause of Death Smoke Diabetes Additional Notes TIME surgery patient information is lacking

ETR219 F 71 39 – Cardiac Failure – No – – Lansoprazole, bisoprolol, salbutamol, warfarin

ETR129 F 78 38 – PMP – No Reading glasses – –

ETR199 F 79 43 – MOF – No – – –

Sucralfate (gastric ulcers), omeprazole (GERD), ferrous fumarate (anaemia), furosemide ETR200 F 79 42 – Cancer – – – – (oedema), carbocisteine (mucolytic), nifedipine (coronary vasodilator)

Simvastatin (cholesterol lowering) seretide ETR214 F 82 48 – – No No No – (asthma), monomil (angina), bisoprolol , ramipril, diltiazem (blood pressure)

20 cigarettes a amlodipine (blood pressure), levothyroxine day. Drop to (hypothyroidism), zopiclone (insomnia), ETR338 F 79 40 – Not reported – Cataracts Thyroidectomy 10 a day as tramadol (pain), omeprazole GERD), illness atorvastatin (cholesterol)

Furosemide, Allcal D3 (calcium supplement), ETR289 F 74 38 – Cancer – – – Omeprazole, ramipril, statin

Left eye central Left eye central retinal vein Left central retinal vein Dexamethasone, mirtazapine, urodeoxycholic ETR277 F 61 47 retinal vein Cancer – – occlusion, caused occlusion in 2015. acid occlusion poor vision.

ETR195 F 62 41 – – Yes No No – Inhalers, amlodipine

ETR243 F 75 36 – – No No Glasses – Amlodipine (Blood pressure)

39 Respiratory ETR51 F 78 – – – Cataract surgery – Atorvastatin –statins (cholesterol) hrs failure

219

PM Ocular disease or Medication/prescription drugs (noted when ETR SEX AGE Diagnosis Cause of Death Smoke Diabetes Additional Notes TIME surgery patient information is lacking

ETR115 M 74 36 – PMP No No No – –

Intracranial Salbutamol, aclidinium bromide. ETR137 M 68 46 – No No Reading glasses – haemorrhage seretide (asthma),

ETR161 M 66 42 – – No No No – –

ETR174 M 62 48 – Pneumonia – – – – –

GP mentions elevated HbA1c, Possible early Hypoxic brain Former Zoladex, omeprazole, citalopram, simvastatin, ETR197 M 68 31 Possibly early NR liver enzyme levels, triglycerides diabetes injury smoker bisoprolol, lisinopril and HDL– possible early T2DM

ETR218 M 59 32 – Sepsis – No Yes – glasses – Diclofenac, paracetamol

Colchicine (gout), lansoprazole (gastric acid), ETR225 M 68 33 – – – No Driving glasses – bisoprolol (hypertension),

Omeprazole, isosorbide mononitrate, 89 Former ETR27 M 84 – – – Cataracts – perindopril, bisoprolol, cetirizine, simvastatin, hrs smoker tamsulosin

Glycerol trinitrate, atorvastatin, ranolazine, 90 Former ETR28 M 79 – – – Yes– wore glasses' – bisoprolol, monomil, ramipril, furosemide, hrs smoker lansoprazole, atenolol,

Detailed GP notes, in 2015 GP Omeprazole, metodopramide, bisoprolol, Former Possibly – see ETR288 M 85 38 – – No suspected diabetes, booked furosemide, allopurinol, spironolactone, smoker notes more tests, patient then died. candesartan, nicorandil, morphine sulphate

70 No– Glasses for No insulin, Family history of ETR35 M 82 – PMP – – No insulin, Family history of diabetes hrs sight diabetes

47 ETR37 M 75 – Not reported – – No – – hrs

220

PM Ocular disease or Medication/prescription drugs (noted when ETR SEX AGE Diagnosis Cause of Death Smoke Diabetes Additional Notes TIME surgery patient information is lacking

Cataracts 44 ETR42 M 70 – – – – developing –not – – hrs diagnosed

40 ETR43 M 60 – PMP – – No – – hrs

49 Hypoxic brain ETR54 M 67 – No No – – – hrs damage

47 ETR59 M 72 – PMP – – Glasses Excessive alcohol consumption – hrs

ETR64 M 69 35 – – – – No – –

Lansoprazole, allopurinol, meloxicam, ETR66 M 76 34 – Cancer – – Yes, short–sighted' – omeprazole, dexamethasone, levothyroxine, loperamide, statins

Glasses for ETR80 M 74 53 – PMP – – – – distance

Intracranial ETR89 M 62 32 – – – – – – haemorrhage

ETR97 M 72 47 – PMP No – No – –

ETR153 M 71 48 – – – – – – –

47 Multi organ ETR55 M 65 – – – – – – hrs failure

CCF, congestive cardiac failure, CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; DR, diabetic retinopathy; GERD, gastroesophageal reflux disease; MI, myocardial infarction; T2DM, type 2 diabetes mellitus

221

Appendix III: List of all Identified Polar Metabolites Metabolites are listed in chromatographic order alongside their retention index (RI) (if identified from the CADET library) and the level of confidence in the identification. This order may be used as a guide of the expected order of metabolite identification. Metabolites were only included in this list if they were consistently identified across the QC group (less than 10% coefficient of variation).

Identification Confidence in Metabolite Identification Lactic acid C Alanine D Glycolic Acid P beta-Hydroxybutyric acid (RI 1204) P Ethanolamine (RI 1256) D Leucine (RI 1250) D Glycerol C Phosphoric acid P Isoleucine (RI 1322) D Glycine (RI 1331) D Serine C Threonine D Urea, N,N'-bis(trimethylsilyl)- D Benzoic acid C Succinic acid D Nonanoic acid P Beta-Alanine, (RI 1454) C Threitol (RI 1469.5) D Malic acid C Aspartic acid (RI 1545) D

222

L-Threonic acid P Methionine (RI 1599) D Methyl à-D-mannopyranoside, D Xylitol (RI 1652) D Creatinine C L-Glutamic acid C Pyroglutamic acid C Phenylalanine (RI 1763) D Glycerol-3-phosphate (RI 1826) D Sorbitol (RI 1830) D Fructose (RI 1837.4) D Phosphoric acid P Glucose (RI 1860) D Citric acid C Myristate C Scyllo-inositol (RI1934.6) D Myo-Inositol, (RI 1979) D Tyrosine, TMS2 (RI 2070) D Pantothenic acid (RI 2083) D Palmitic Acid (RI 2097) D Heptadecanoic acid P Stearic Acid (RI 2296.2) D Myo-Inositol phosphate P Glucose-6-phosphate Peak1 (RI 2330) D Tryptophan (RI 2403) D

223

Maltitol (RI 2637) D Uridine, (RI 2660) D Inosine, (RI 2755) D Fructose bis-phosphate (RI 2774) D Adenosine (RI 2805) D Cholesterol trimethylsilyl ether D C, confident; D, definite, P, putative.

224

Appendix IV: List of all Identified Lipids Lipids are listed by class and the level of confidence is noted. Standards were not run so none were identified definitely. The level of information obtained about the mass using the Orbitrap alongside the chain information from MS/MS ensure that identifications are highly confident.

ID Confidence Cholesterol Ester 18:2 Cholesteryl ester C 20:4 Cholesteryl ester C 22:6 Cholesteryl ester C Cholesterol D Coenzyme Q Coenzyme Q10 D Coenzyme Q9 C Fatty Acids FA-27:3 P FA-32:6 P FA-34:6 P FA-Arachidonic acid (20:4) C FA-DHA (22:6) C FA-Oleic acid (18:1) C FA-Palmitic acid (16:0) D FA-Stearic acid (18:0) D FMC FMC-5(d18:1_18:0) P FMC-5(d18:1_22:0) P FMC-6(d18:1_22:0(2-OH)) P Gangliosides NeuAcα2-Hex2-Cer(d34:1) P NeuAcα2-Hex2-Cer(d36:1) P Hexadecanedioic acid P Lauroyl diethanolamide P Phosphatidylglycerols PG(16:0) P PG(16:0_20:4) C PG(18:0_20:4) C PG(20:4_22:6) C PG(22:6_22:6) C PG(34:0) P PG(36:2) P Phosphatidylinositols PI(16:0_20:4) C PI(38:6) C PI(18:0_22:6) C

225

PI(18:2_20:4) C PI(18:2_20:4) C PI(20:4_17:0) C PI(20:4_22:6) C PI(22:6_18:2) C PI(36:0) P PI(38:5) C PI(40:7) P PI(44:12) P Phosphoserines PS(18:0_22:6) C Phosphatidylcholines LysoPC(16:0_0:0) C LysoPC(18:0_0:0) C LysoPC(18:1_0:0) C LysoPC(18:1_0:0) C LysoPC(20:4_0:0) C LysoPC(22:6_0:0) C PC(14:0_22:6) C PC(16:0_14:0) C PC(16:0_16:0) C PC(16:0_16:1) C PC(16:0_18:0) C PC(16:0_18:1) C PC(16:0_18:2) C PC(16:0_20:4) C PC(16:0_22:6) C PC(17:0_22:6) C PC(18:0_18:1) C PC(18:0_20:4) C PC(18:1_18:1) C PC(18:1_22:6) C PC(22:10) C PC(22:6_22:6) C PC(33:1) C PC(34:3) P PC(36:3) C PC(38:5) C PC(38:6) C PC(40:5) C PC(40:6) C PC(41:6) C PC(42:8) C PC(42:9) C

226

PC(P-14:0_20:4) or PC(O-14:1_20:4) C PC(P-34:0) or PC(O-34:1) C PC(P-36:4) or PC(O-36:5) P PC(14:0_22:6) C Phosphatidylethanolamines LysoPE(16:0_0:0) P LysoPE(18:0_0:0) P LysoPE(22:4_0:0) P LysoPE(22:6_0:0) C LysoPE(20:4_0:0) C LysoPE(0:0_22:5) P PE(16:0_16:0) C PE(16:0_18:1) C PE(16:0_20:4) C PE(16:0_22:4) C PE(16:1_22:6) C PE(17:0_22:6) C PE(18:0_18:1) C PE(18:0_20:4) C PE(18:0_22:4) C PE(18:0_22:6) C PE(18:0_24:4) C PE(18:3_22:6) C PE(18:3_22:6) C PE(20:3_18:0) C PE(20:3_22:6) C PE(20:4_18:1) C PE(20:4_22:6) C PE(22:4_22:6) C PE(22:6_18:0) C PE(22:6_18:1) C PE(22:6_18:2) C PE(22:6_20:2) C PE(22:6_22:6) C PE(36:0) C PE(38:6) P PE(40:6) P PE(O-34:1) or PE(P34:0) C PE(O-34:3) or PE(P-34:2) C PE(O-34:3) or PE(P-34:2) C PE(O-34:3) or PE(P-34:2) C PE(O-34:3) or PE(P-34:2) C PE(O-36:7) or PE(P-36:6) C PE(O-38:4) or PE(P-38:3) P

227

PE(O-38:5) or PE(P-38:4) C PE(O-38:5) or PE(P-38:4) P PE(O-38:5) or PE(P-38:4) C PE(O-38:6) or (PE(P38:5) C PE(O-40:3) or PE(P-40:4) C PE(O-40:4) or PE(P-40:3) C PE(O-40:6) or PE(P40:5) C PE(O-42:3) or PE(P-42:4) P PE(P-18:1_22:6) or PE(O-18:2_22:6) C PE(P34:1) or PE(O34:2) C Ceramides Hex2Cer(d34:1) C Hex2Cer(d36:1) C Cer(d42) C Cer(d42) C Cer(d18:1/16:0) C Cer(d18:1_18:0) C Cer(d36:2) C Cer(d18:1_20:0) C Cer(d18:1/22:0) C Cer(d41) C Cer(d18:1/24:0) C Cer(d18:1/24:1) C Cer(d34:2) C HexCer(d38:1) C HexCer(d42:2) C HexCerd(d38:2(OH)) C HexCer(d34:1) C HexCer(d36:1) C HexCer(d40:1) C HexCer(d42:1) C Sphingomyelins SM(d16:1_16:0) C SM(d16:1_18:1) C SM(d18:0_16:0) C SM(d18:0_18:0) C SM(d18:0_18:2) C SM(d18:1_16:0) C SM(d18:1_18:0) C SM(d18:1_20:0) C SM(d18:1_20:1) C SM(d18:1_22:0) C SM(d18:1_24:0) C Sphingosine C

228

Monoglyceriols MG(18:1) P MG(20:4) P MG(20:4) P MG(22:6) P MG(22:6) P MG(16:0) P MG(16:0) P MG(18:0) P MG(18:0) P Diacylglycerols DG(34:0) C DG(36:1) C DG(18:0_20:4) C DG(18:0_22:6) C DG(36:4) C DG(18:0_22:4) C Triacylglycerols TG(14:0_16:0_16:0) C TG(14:0_16:0_22:6) C TG(16:0_16:0_16:0) C TG(16:0_16:0_16:1) C TG(16:0_16:0_18:0) C TG(16:0_16:0_18:1) C TG(16:0_16:0_18:2) C TG(16:0_16:0_22:6) C TG(16:0_18:0_18:0) C TG(16:0_18:0_18:1) C TG(56:6) C TG(52:2) C TG(16:0_18:1_18:2) C TG(16:0_18:2_18:2) C TG(16:0_18:2_18:2) C TG(16:0_18:1_22:4) C TG(16:0_18:2_22:6) C TG(16:0_22:6_22:6) C TG(18:0_18:2_18:2) C TG(16:0_18:0_20:4) C TG(18:0_22:6_22:6) C TG(54:5) C TG(54:5) C TG(22:4_18:0_16:0) C TG(56:7) C TG(18:2_18:2_18:2) C

229

TG(18:2_18:2_18:3) C TG(16:0_18:1_22:4) C TG(16:0_20:4_22:6) C TG(19:0_20:0_20:4) C TG(22:6_20:4_18:0) C TG(22:6_22:6_22:6) C Other N-stearoyl taurine P Stearamide D 1,2-eicosanediol P Docosanamide P C, confident; D, definite, P, putative.

230