The Identification and Development of Small Molecule Inhibitors of Amyloid Β Aggregation

The Identification and Development of Small Molecule Inhibitors of Amyloid β Aggregation

Súil Collins Downing College University of Cambridge September 2017

Supervised by Professor David R. Spring

A dissertation submitted to the University of Cambridge for the degree of Doctor of Philosophy

Abstract

The Identification and Development of Small Molecule Inhibitors of Amyloid β Aggregation

Súil Collins

Amyloid β (1-42) (Aβ42) is a seminal neuropathic agent in Alzheimer’s disease (AD), a multifaceted neurodegenerative disorder for which no preventative measures or disease modifying therapies currently exist. Aggregation of this peptide plays a key role in the synaptic dysfunction and neuronal death associated with the disease. Perturbing the aggregation process, therefore, represents a key strategy for the development of new AD therapeutics. A variety of issues with current screening methods, including lack of reproducibility, high reagent consumption and spectral interference from the test molecules, can limit efforts to identify new small molecule inhibitors. Furthermore, the lack of robust, time- and cost-efficient methods for screening compounds in cellular or in vivo models limits the throughput with which seemingly active small molecules can be validated and prioritised. Herein, this thesis describes efforts to overcome such limitations through the development of a unified in vitro to in vivo assay system, in which hits identified in the ‘nanoFLIM’ microfluidic-based assay can quickly be tested in cellular and whole organism disease models.

The assay platform designed relies on the use of an amyloid aggregation fluorescence lifetime sensor. Aβ42 aggregation is monitored by changes in the fluorescence lifetime of an attached fluorophore, which is significantly quenched upon amyloid formation. To take advantage of the benefits associated with miniaturisation, an in vitro microfluidic platform was employed. A microfluidic chip capable of trapping 110 precisely ordered droplets was designed, allowing for increased sample size and greatly lowering reagent consumption relative to conventional assay formats. Optimisation of the lifetime sensor technique permitted real-time compound screening in SH-SY5Y neuroblastoma cells, as well as in disease model Caenorhabditis elegans (C. elegans). To demonstrate the potential of this assay, a selection of novel chemical libraries developed in the Spring research group was screened, resulting in the identification of a key library of interest. The inhibitory activity of the lead compound from this collection was validated using a variety of biophysical tests, and was also shown to suppress amyloid aggregation in the live cell fluorescence lifetime sensor assay, as well as in whole organism disease model C. elegans.

Whilst assay development was underway, additional screening of structurally diverse chemical libraries was performed using a conventional Thioflavin T spectroscopic assay. Such work identified another molecular scaffold capable of exerting a strong inhibitory effect against Aβ42 aggregation. A selection of analogues was synthesised to improve the in vivo profile of this library, giving rise to a second lead inhibitory compound. The activity of this compound was subsequently validated in biophysical and cellular tests, and was also tested in disease model Drosophila melanogaster.

The aggregation of Aβ42 lies at the root of Alzheimer’s disease. In light of the relatively few drug candidates in clinical trials for this disorder, the development of improved translational screening approaches and continued screening of novel chemical libraries is necessary to identify new potential therapeutics. In this study, an in vitro to in vivo fluorescence lifetime imaging assay has been established. Using this assay system and conventional screening approaches, two Aβ42 aggregation inhibitors have been identified and validated. These represent promising candidates for the development of new AD therapeutic agents, or for use as molecular probes to further dissect the mechanisms underlying this devastating disease.

Declaration

This dissertation is submitted in fulfilment of the requirements for the degree of Doctor of Philosophy. It describes work carried out in the Department of Chemistry, the Department of Biochemistry and the Department of Chemical Engineering and Biotechnology, at the University of Cambridge between June 2014 and September 2017, under the supervision of Professor David Spring. The work presented has not been submitted for any other degree. It does not exceed the prescribed word limit for the Physics and Chemistry Degree Committee.

Signed, Date:

______

Súil Collins

Downing College, Cambridge

iii

To my family, with all my love and appreciation

Acknowledgements

First and foremost, I would like to thank Prof David Spring for inviting me to join his research group and for giving me the freedom to develop a project that I feel so strongly about. I truly appreciate it. I would like to express my sincerest gratitude to Dr Gabriele Kaminski Schierle, her constant stream of ideas and unrivalled enthusiasm has been inspiring. Many thanks to Prof Florian Hollfelder, his guidance and constructive critique have kept me, and my project, on track. Prof Clemens Kaminski and Dr Damian Crowther also deserve special thanks, for taking the time to introduce me to microscopy and Drosophila studies, respectively. I greatly enjoyed all this work. I am very grateful to the BBSRC and the Cambridge Trust, for their generous funding and for giving me the opportunity to participate in such an exciting interdisciplinary programme.

I have been incredibly lucky to work in three inspiring research groups, each filled with exceptional individuals, who freely gave their time and expertise to help me and my research. Thank you to all the Spring group members, past and present, for sharing your chemistry knowledge and for all the fun and laughter we’ve shared, both in the lab, and out. Sean, Steve, Sarah, Joe, Josie, Tommy, Twigg and Warren deserve special acknowledgement for many helpful discussions and day-to-day moral support. Thanks to the members of the Laser Analytics and Molecular Neuroscience Groups for their kindness, encouragement and invaluable input. I am indebted to Claire, Colin and Amberley for helping me to get to grips with the biological experiments, and to Romain and Weiyue for sharing their boundless FLIM knowledge with me. Thanks to everyone in the Hollfelder group, for their continuous help and advice, and for being such great friends. I would also like to express my deepest appreciation to Liisa, for welcoming me to the group with a smile, and to Fabrice, for his instrumental help in getting my project up and running.

I am extremely grateful to the many wonderful friends, both old and new, who have made the last few years so special. Thanks to the basketball club for helping to keep me sane after long hours in the lab. Molly, Michelle and Tati – three-time Varsity champs – thank you for never accepting my attempts to retire. Thanks to Sylwia and Timo, my Cambridge family; Nicole, Sara and Alice for always being there for me with a smile and a hug; Yuteng, for letting me follow him to Cambridge; Eilíse, a breath of fresh Irish air when I needed one; Rita, for making my internship such an enjoyable one; Tom Cotton, my favourite Hot Numbers date; and Kerri, my partner-in-crime from early undergraduate years and hopefully for many years to come. To those who I have neither the space nor word-count to mention, I thank you for your patience, support and encouragement.

vii

I reserve special thanks for David, who came into my life during a period when my research was being especially uncooperative. Thank you for being so exceptionally tolerant and patient. Compassionate from miles away, never passive in the things you say. You make every day happier.

I will finish with thanks to my family, for always setting the bar high and giving me the space and encouragement to rise to the challenge. Much love and thanks to both my grannies, for providing inspirational examples of strong and determined women. Thanks to Aingeal, our surrogate auntie and wonderful friend. Leanbh, Eibh, and Tom, thank you for your support at every step of the journey. Your motivational messages, random jokes and Christmas countdown reminders have kept me smiling. Mariusz, Gar and Doc, thank you for providing family support, when I was otherwise distracted. Mum and dad, you deserve more thanks than I can ever put in words. Thanks to Dad for giving me my first science kit, and to mum for taking away all the dangerous pieces. Your constant love and unfaltering support have filled me with the hope, joy, and the confidence to be the best that I can be. I am eternally grateful to you both.

A final thanks to my new baby niece, who has kindly waited to make her debut until after I’ve submitted. We’re ready for you now, Dotty.

viii

Abbreviations

3D three-dimensional °C degrees Celsius Å Ångström(s) μ micro λ wavelength τ fluorescence lifetime Aβ amyloid β peptide

Aβ40 amyloid β peptide 40 residues in length

Aβ42 amyloid β peptide 42 residues in length

Aβ42-488 synthetic Aβ42 labelled with Hilyte™ Fluor 488 AD Alzheimer’s disease ADME absorption, distribution, metabolism, and excretion AFM atomic force microscopy APP amyloid precursor protein approx. approximately a.u. arbitrary units BACE beta-secretase BBB blood brain barrier (c)LogD (calculated) octanol-water distribution coefficient (c)LogP (calculated) octanol-water partition coefficient CNS central nervous system CR congo red COX cyclooxygenase Da Dalton(s) DMSO dimethyl sulfoxide DOS diversity-oriented synthesis EGCG (-)-Epigallocatechin-3-gallate EPPS 4-(2-Hydroxyethyl)-1-piperazinepropanesulfonic acid ESI electrospray ionisation equiv equivalent(s)

FC flow cytometry FLIM fluorescence lifetime imaging microscopy FRET Förster resonance energy transfer FSC forward scatter GFP green fluorescent protein h hour(s) HBA hydrogen bond acceptors HBD hydrogen bond donors hFTAA heptameric formyl thiophene acetic acid HFIP 1,1,1,3,3,3-hexafluro-2-propanol HRMS high resolution mass spectrometry

IC50 half maximal inhibitory concentration IDP intrinsically disordered protein IR infrared J coupling constant K18-488 recombinant Tau K18 construct with 10% Alex Fluor® 488 labelling LCMS liquid chromatography-mass spectrometry LCO luminescent conjugated oligothiophenes M molar m milli or meter(s) m.p. melting point min minute(s) MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide) mol mole(s) mol% mole percent n nano ns not significant or nanosecond NaPi sodium phosphate buffer NFT neurofibrillary tangle NMR nuclear magnetic resonance p pico PBS phosphate buffered saline

PDMS polydimethylsiloxane PFO 1H,1H,2H,2H-perflurooctanol P-gp P-glycoprotein PK pharmacokinetic PTFE polytetrafluoroethylene

Rf retention factor rpm revolutions per minute Ro5 rule of five rt room temperature SAR structure-activity relationship SBDD structure-based drug design SDS sodium dodecyl sulfate sec second(s) SEM standard error of the mean SIM structure illuminated microscopy SSC side scatter t time TCSPC time-correlated single photon counting TEM transmission electron microscopy TFA trifluoroacetic acid TG-FLIM time-gated FLIM ThT thioflavin T TLC thin layer chromatography tPSA topological polar surface area WT wild type

Contents

Abstract ...... i Declaration ...... iii Acknowledgements...... vii Abbreviations ...... ix Chapter 1 Introduction ...... 1 1.1 Historical Context ...... 1 1.1.1 A Brief History of Dementia ...... 1 1.1.2 Development of the Amyloid Cascade Hypothesis ...... 2 1.2 Alzheimer’s Disease ...... 3 1.3 Amyloid β ...... 5 1.3.1 The Amyloid Cascade Hypothesis ...... 7 1.3.2 Structure of Aβ ...... 8 1.3.3 Aggregation Pathway and Kinetics ...... 10 1.3.4 Aβ and Oxidative Stress in Alzheimer’s Disease ...... 12 1.4 Tau ...... 13 1.5 Monitoring the Aggregation of Aβ ...... 14 1.5.1 Microscopy ...... 14 1.5.2 Spectroscopic Techniques ...... 15 1.5.3 Separation Techniques ...... 17 1.5.4 Mass Spectrometry...... 17 1.5.5 Immunodetection ...... 18 1.5.6 Functional Screens ...... 18 1.6 Perturbing the Process of Amyloid β Aggregation ...... 19 1.6.1 Small Molecule Aβ Aggregation Inhibitors ...... 21 1.6.2 Aβ Binding and Inhibitory Mechanisms ...... 25 1.6.3 Past Failures and Future Direction for the Development of Aβ Aggregation Modifiers ...... 27 1.7 The Drug Discovery Process ...... 28 1.7.1 Early Drug Discovery and Compound Screening ...... 28 1.7.2 Physicochemical Properties ...... 30 1.8 Fluorescence Spectroscopy ...... 31 1.8.1 Basic Principles ...... 31

1.8.2 Fluorescence Lifetime Imaging Microscopy (FLIM) ...... 33 1.8.3 Fluorescence Techniques for the Study of Amyloid Formation ...... 36 1.9 Microfluidics ...... 40 1.9.1 Droplet-Based Microfluidics...... 40 1.9.2 Microfluidics for the Study of Amyloid Aggregation ...... 41 1.9.3 Compound Screening using Microfluidic Technology ...... 42 1.10 Project Aims and Thesis Structure ...... 42 Chapter 2 The Agarose Cage Microdroplet Assay ...... 45 2.1 Introduction ...... 45 2.2 Primary Objectives ...... 47 2.3 Results and Discussion ...... 48 2.3.1 Overview ...... 48 2.3.2 Optimisation of the Microdroplet Cage Assay ...... 48 2.3.3 Small Molecule Screening with the Microdroplet Cage Assay ...... 59 2.3.4 Tau Investigations ...... 61 2.3.5 FLIM Studies ...... 66 2.4 Summary and Conclusions ...... 70 Chapter 3 Development of a Fluorescence Lifetime Sensor-Based Microfluidic Assay for Monitoring Amyloid β Aggregation ...... 71 3.1 Introduction ...... 71 3.2 Primary Objectives ...... 72 3.3 Results and Discussion ...... 73 3.3.1 Microfluidic Chip Design ...... 73 3.3.2 Optimisation and Testing the Capabilities of the NanoFLIM ...... 76 3.3.3 Validation with Known Inhibitors ...... 87 3.4 Summary and Conclusions ...... 95 Chapter 4 NanoFLIM Compound Screening and Hit Validation ...... 99 4.1 Introduction ...... 99 4.2 Results and Discussion ...... 99 4.2.1 Library Screening ...... 99 4.2.2 Hit Library ...... 100 4.2.3 Hit Validation ...... 106 4.2.4 Analysis of Rescuing Effects of MJ Library Test Subset ...... 110

4.3 Summary and Conclusions ...... 111 Chapter 5 Cellular and In vivo Screening of Aβ Aggregation Inhibitors with the Fluorescence Lifetime Sensor ...... 115 5.1 Introduction ...... 115 5.1.1 Cellular Models for Monitoring Aβ Aggregation ...... 115 5.1.2 Caenorhabditis elegans as Disease Model Organisms ...... 116 5.1.3 Alzheimer’s disease model Caenorhabditis elegans ...... 117 5.2 Primary Objectives ...... 118 5.3 Results and Discussion...... 118 5.3.1 Live Cell Studies ...... 118 5.3.2 Caenorhabditis elegans Studies ...... 124 5.4 Summary and Conclusions ...... 128 Chapter 6 Inhibitory Activity of Polyphenol Derivatives ...... 131 6.1 Introduction ...... 131 6.1.1 Structure-Activity Relationship ...... 134 6.1.2 Molecular Binding Studies ...... 135 6.1.3 Antioxidant Polyphenols for Alzheimer’s Disease Treatment ...... 136 6.2 Primary Objectives ...... 137 6.3 Results and Discussion...... 137 6.3.1 Initial Screening ...... 137 6.3.2 Hit Validation ...... 144 6.3.3 Library Expansion and SAR ...... 147 6.3.4 Antioxidant Activity ...... 154 6.4 Summary and Conclusions ...... 156 Chapter 7 Compound Screening in Alzheimer’s Disease Model Drosophila Melanogaster ...... 157 7.1 Introduction ...... 157 7.1.1 Drosophila Melanogaster as Disease Model Organisms ...... 157 7.1.2 Alzheimer’s Disease Drosophila Models...... 157 7.1.3 The UAS-Gal4 System ...... 158 7.1.4 Drug Screening in Drosophila ...... 159 7.2 Primary Objectives ...... 162 7.3 Results and Discussion...... 162 7.3.1 Overview ...... 162

7.3.2 Rescue of the Rough Eye Phenotype ...... 163 7.3.3 Rescue from Aβ-induced Oxidative stress ...... 165 7.3.4 Longevity Studies ...... 170 7.4 Summary and Conclusions ...... 173 Chapter 8 Conclusions and Future Outlook ...... 175 8.1. An Overview ...... 175 8.2 The NanoFLIM ...... 176 8.3 Cellular and Whole Organism Model Testing ...... 178 8.3.1 Cellular Lifetime Sensor Assay ...... 179 8.3.2 Caenorhabditis Elegans Lifetime Sensor Assay ...... 179 8.3.2 Disease Model Drosophila Melanogaster ...... 180 8.4 Identification and Development of Hit Inhibitory Compounds ...... 181 8.4.1 MJ040 and MJ040X ...... 181 8.4.2 Heteroaromatic Chalcones ...... 183 8.4.3 Pharmacokinetic Considerations ...... 184 8.5 Implications for Future Alzheimer’s Disease Research ...... 185 Chapter 9 Experimental Procedures ...... 187 9.1 Reagents ...... 187 9.2 Peptide preparation ...... 187 9.2.1 Amyloid β ...... 187 9.2.2 Tau ...... 188 9.3 Microfluidics ...... 188 9.3.1 Chip Fabrication ...... 188 9.3.2 Agarose Cage Droplet Assay ...... 190 9.3.3 The nanoFLIM ...... 192 9.4 Microscopy ...... 192 9.4.1 Fluorescence Lifetime Imaging Microscopy (FLIM) ...... 192 9.4.2 Atomic Force Microscopy (AFM) ...... 193 9.4.3 Transmission Electron Microscopy (TEM) ...... 194 9.5 Miscellaneous in vitro Techniques ...... 194 9.5.1 Flow Cytometry ...... 194 9.5.2 Spectroscopic Assays ...... 195 9.5.3 Dot blot analysis ...... 195 9.6 Biological work ...... 196

9.6.1 Cell Studies ...... 196 9.6.2 Caenorhabditis elegans ...... 198 9.6.3 Drosophila Melanogaster ...... 201 9.8 Chemical Synthesis ...... 204 9.8.1 General Information and Materials ...... 204 9.8.2 Experimental...... 205 References ...... 217 Appendices ...... 239 Appendix I: Libraries Screened with the NanoFLIM ...... 241 Appendix II: MJ Library ...... 243 Appendix III: C. elegans Construct Map ...... 246 Appendix IV: Statistical Breakdown of C. elegans Data ...... 247 Appendix V: Methoxychalcones Library ...... 251 Appendix VI: Computer Aided Design of the Two-Part Shearing Chip ...... 254 Appendix VII: NMR Spectra ...... 255

Chapter 1 Introduction

In this chapter, an overview of the properties of amyloid β (Aβ) is presented, including details of its relationship with Alzheimer’s disease (AD), its aggregation behaviour, current strategies for screening for aggregation inhibitors and their potential therapeutic applications. A brief introduction into the drug discovery process and previous failures encountered with the development of AD therapies is given. The technologies employed in this research are described, with an outline of their fundamental properties and select examples of how they have previously been implemented for the study of amyloid aggregation. The chapter concludes with a summary of the primary aims and structure of the thesis.

1.1 Historical Context 1.1.1 A Brief History of Dementia

Many reports on the history of dementia begin with an account of Alois Alzheimer presenting his description of a ‘peculiar disease of the cerebral cortex’ at a conference of German psychiatrists in 1906.1 However, the history of dementia research is far richer than this, spanning from 2000 AD, where ancient Egyptians were the first to associate memory disorders with aging, to the present day, where the pathological mechanisms underlying neurodegenerative diseases continue to puzzle researchers.2 The link between cognitive decline and aging was reliably recorded in as early as the 7th century BC, when Pythagoras described regression of mental capacities during ‘senium’, the final two of his Five Stages of Life.3 Plato and Aristotle (384-322 BC) also commented on the occurrence of mental failure in aged individuals,3 and Celsius (30 BC- 50 AD) was the first to use the term ‘dementia’ within a medical context, in his listings of mental conditions.4 Aretaeus of Cappadocia (2nd century AD) gave a distinction between acute and chronic neurological disorder in his discussions on leresis – the ‘misfortune of old age’ that begins with aging and dies with the person.5 The idea that chronic neurological problems are associated with aging was further advanced by the Roman physician Galen [150-200 AD].3 ‘Morosis’ was the term he used to describe dementia, and old age was listed as a situation in which it arises. Notably, these early observations all incorrectly suggested that dementia was an inevitable feature of aging.4

Despite an interlude in the reports of dementia during medieval times, attributed to the decreased status of doctors, the increased status of the church and the prevalence of more pressing health concerns like plague,2,3 the inevitability of decrepitude upon aging was still a commonplace idea. Literary descriptions from Chaucer (“with old folk, save dotage, is namore”), Shakespeare (“An old man is twice a child”) and others highlighted this belief.4 From the end of the 16th century, reinvested interest in neurological maladies saw gradual refinement of the broad concepts of dementia. Scrutiny of afflicted brains, by scientists such as Philip Barrough (1560-1590), lead to the identification of underlying abnormalities, such as inflammation and reduction in mass.6 During this time, the term ‘dementia’ (démance) as we understand it to mean today, gradually entered the lexicon of medical literature. This is often credited to Philippe Pinel (1745-1826) in his descriptions of “the abolition of the thinking faculty” in his revolutionary book “Treatise on Insanity”.7 Senile dementia was recognised as a medical entity by William Cullen (1710-1790), and was defined as “decay of perception and memory in old age”.3 Enlightening reports from Jean Etienne Esquirol (1772-1840) further classified a variety of neurological disorders, providing succinct descriptions of the stages of cognitive decline that are still apt today.4 The end of the 19th and beginning of the 20th century saw tremendous advances in dementia research. Samuel Wilks (1824-1911) was the first to recognize that atrophy was responsible for the observed decrease in brain weight, and was an identifiable feature of senile dementia.8 In the early 1890s, Paul Blocq and Gheorghe Marinesco were the first to describe the accumulation of an unidentified substance in microscopic plaques as a novel pathological feature in the brain of an elderly patients.3,9 A similar pathology was then observed by Emil Redlich in two cases of senile cerebral atrophy associated with memory defects and mental confusion,4 observations made possible with the advance of histological staining methods. In 1903, Max Bielschowsky improved the silver stain, providing a means to sharply stain intracellular components and allow for the detailed visualisation of neurons.10 This technique was later used by Alois Alzheimer to identify pathological features defining the disease that would later bear his name.1

1.1.2 Development of the Amyloid Cascade Hypothesis

In the years following the initial observations of amyloid plaques in the brains of dementia patients, much controversy existed over the source, constitution and pathogenicity of these features.3 Progress came in 1928 with the discovery that plaques stained with the Congo Red dye displayed green birefringence. This is a characteristic indicator of the presence of “amyloid”,11 a term first used by Virchow to describe pathological starch-like aggregated substances.12

Improvements in staining techniques and advances in electron microscopy during the 1950’s provided insight into the fibrillary nature of the amyloid core of the plaques.3 It was not until the 1980’s, however, that the significance of the constituent amyloid was brought to light.13 In 1984, Glenner and Wong isolated and partially sequenced a ~4kD ‘amyloid-β protein (AβP)’ from meningeal vessels of an Alzheimer’s disease (AD) patient.14 They also provided the first insight into the potential role of genetic factors in AD, upon their report that the same AβP protein found in AD patients could also be isolated from Down’s Syndrome patients.15 This was significant, as people with Down’s Syndrome have an extra chromosome 21 (trisomy 21) and are known to be predisposed to getting an AD-like disease by the age of 40. Glenner’s observation, therefore, indicated a “genetic defect in Alzheimer’s disease … localized on chromosome 21”.15 The following year, amyloid plaque cores from AD patients were purified in a collaboration between Masters and Bayreuther, revealing the presence of a ~4kD protein that they named A4, of similar composition to the cerebrovascular AβP.16 Bayreuther quickly determined that A4 was a fragment of a larger neuronal protein, amyloid precursor protein (APP), encoded by the APP gene on chromosome 21.17

Together these results highlighted the significance of the renamed amyloid β (Aβ) peptide in AD and the next decade saw a flurry of activity attempting to understand the links between formation, aggregation and deposition of Aβ and the disease state. Of particular importance was the raft of genetic linkage analyses investigating familial and sporadic forms of the disease. Whilst some results were misinterpreted and premature conclusions drawn,18 strong evidence indicated that mutations in the APP gene were a significant risk factor, thereby highlighting a seminal role of Aβ in AD pathogenesis. Hardy and Higgins finally clarified this assessment in 1992, with the publication of the ‘Amyloid Cascade Hypothesis’, which states that “deposition of amyloid β protein (AβP), the main component of the (β) plaques, is the causative agent of Alzheimer's pathology and that the neurofibrillary tangles, cell loss, vascular damage, and dementia follow as a direct result of this deposition.”19 (See section 1.3.1 for details.)

1.2 Alzheimer’s Disease

Alzheimer’s disease is a debilitating neurodegenerative disease, characterised by a deterioration of memory and subsequent decline of other cognitive functions.20 It is the most prevalent form of dementia, afflicting 48 million people worldwide.21 It is the only disease in the top 10 causes of

3 death in the USA for which no preventative measures or disease modifying therapies exist, and current therapeutics only treat the symptoms.21,22 AD falls into the class of amyloidosis disorders, which are characterised by an accumulation of misfolded amyloid protein. Correct protein conformation is fundamental for proper cellular functioning and failure to achieve it can give rise to a host of protein misfolding disorders, including Parkinson’s, Huntington’s and prion diseases.23 The pathological hallmarks of AD are the presence of cerebral plaques containing Aβ and phosphorylated Tau tangles, as well as severe neuronal loss and disruption of synaptic function (Figure 1.1).20 Hereditary AD cases are rare and sporadic AD accounts for over 97% of cases, with a mean age of onset 80 years.21,22

Figure 1.1: Pathological hallmarks of Alzheimer’s disease. A comparison of the neurons and general brain size in normal individuals (left) and Alzheimer’s disease patients (right) is shown. The hallmark pathognomonic features of AD are extracellular deposits of Aβ plaques and intracellular neurofibrillary tangles, composed of phosphorylated Tau. Neuronal and synaptic loss are also characteristic features of the disease.

AD represents the largest unmet medical need in neurology, and although great advances in recent years have highlighted many opportunities for clinical intervention, no disease-modifying agents have been approved for clinical use, yet.24 Difficulties have arisen from the incomplete understanding of disease pathology, limitations in current animal models, complexities in elucidating potential drugs mechanism of action and difficulties in translating in vitro to in vivo activity, among other factors.24,25 Disease-modifying strategies rely on targeting processes

4 critically important to the disease state, however at this point, the significance of the complex interplay of molecular mechanisms and AD progression is not clearly understood. Given this limited understanding, it is difficult to distinguish if a pathological observation plays a causative role, is a neutral by-product of the pathology or is a remnant of an unsuccessful attempt at repair.26 Many pathological processes have been targeted, with some therapeutic candidates reaching various stages of clinical trials (Figure 1.2).24,26 Of particular importance are strategies aimed at Aβ, which has been experimentally shown to play a crucial role in AD pathogenies.27 Other strategies have been directed towards Tau pathology, metabolic dysfunction, inflammatory responses and targeting well-known genetic risk factors, such as the APOE4 genotype.26,28

1.3 Amyloid β

Aβ is a 4 kDa amphipathic polypeptide, prone to self-assembly and fibril formation.29,30 Aggregated Aβ is a cardinal feature of AD and the characteristic plaques found in the brains of AD patients are primarily composed of Aβ fibrils. The peptide originates from the amyloid precursor protein (APP), which is cleaved sequentially by the proteolytic enzymes β- and γ-secretase.20 Figure 1.2 shows the section of the APP protein from which Aβ is excised.31 α-secretase functions in a non- amyloidogenic pathway, cleaving within the Aβ sequence. In the amyloidogenic pathway, cleavage by β-secretase (BACE 1) forms the N terminus, and γ-secretase activity within the transmembrane region determines the length of the peptide. The resulting species range in size from 38-43 residues, but the predominant amyloid forms are Aβ40 and Aβ42, which differ in length (40 or 42 residues respectively) and propensity to aggregate. Aβ40 is the most abundant form, but the more

29 aggregation prone Aβ42 is enriched in the brains of AD patients. The ratio of Aβ40 : Aβ42 in the cerebrospinal fluid (CSF) is used to differentially diagnose AD from other forms of dementia.32 Inheritable forms of AD are caused by autosomal dominant mutations in APP gene or presenlin 1 and 2 (components of γ-secretase), which result in increased secretion of Aβ or a higher ratio or

33 secreted Aβ42 relative to Aβ40.

Figure 1.2: Section of the amyloid precursor protein (top numbering) and the Aβ peptide (lower numbering. Several forms of Aβ are formed through sequential proteolysis of the amyloid precursor protein (APP) by α, β and γ secretases. Two main isoforms are found in the brain – Aβ40 and Aβ42. Adapted from the review by Rauk.34

Despite the vast amount of evidence indicating a fundamental role of Aβ aggregation in AD onset and progression, there is still limited knowledge of its structure and the molecular processes underlying its involvement. Given Aβ’s intrinsically disordered nature, low solubility and high aggregation propensity, its physical characteristics are poorly understood and conflicting descriptions have been reported in the literature.30 What is known is that fibrillar Aβ shows a characteristic amyloid diffraction pattern, which reflects the β-sheet peptide secondary structure arranged perpendicular to the long axis of the fibril.35 Mechanical rigidity is provided by intermolecular peptide bonds in the underlying β-sheet architecture,36 and the resulting fibrils can adopt a range of supramolecular structures.37 As the self-assembly process is spontaneous, and as the peptide simultaneously assemblies into fibrils of different morphologies once a critical concentration of monomers is present, understanding the structure at the atomic-level becomes complicated.38,39 Furthermore, during the aggregation pathway, heterogeneous ensembles of oligomeric Aβ exist. The structures of these species also remains elusive.

Given the complexity of the disease, it is difficult to comprehend how oligomers manifest neurotoxicity. Various mechanisms have been reported and often indicate that oligomeric Aβ induces toxicity by interacting with cellular membranes and with membrane receptors that activate various signalling pathways.40–43 The use of specific antibodies and sophisticated structure determining techniques have aided efforts to characterise distinct species and shed light on the underlying structural motifs, but knowledge is still limited. Oligomer formation is not committed to a single pathway and they display polymorphism when formed under different experimental conditions in terms of epitope accessibility, toxicity and biochemical properites.44 Understanding the structure and mechanisms through which these species exert their toxic effect is of paramount significance, not only for delineating the processes driving disease pathogenesis, but also to facilitate the development of new therapeutic agents.

1.3.1 The Amyloid Cascade Hypothesis

The amyloid cascade hypothesis posits that Aβ is the primary causative agent of neuronal toxicity in AD.19,45 As the most physically distinct example of the aggregated structures in AD, Aβ plaques were first identified as key suspects in disease induction. It was originally postulated that the deposition of these extracellular species leads to inflammation, synaptic dysfunction, neuronal loss, Tau pathology and eventually dementia. However, the linearity of this cascade came under review when the significance of Aβ oligomers was recognised.46–48 Biophysical and animal modelling studies support the idea that oligomeric species are responsible for Aβ-associated toxicity, and the Aβ cascade hypothesis was therefore modified, from that in which amyloid plaques were the central disease-driving features, to instead acknowledge the role of soluble oligomers as the main pathogenic agents.45

Figure 1.3: Aβ-targeting strategies for the development of AD therapies Sequential cleavage of amyloid precursor protein (APP) by the proteolytic enzymes β-secretase and γ-secretase gives rise to extracellular monomeric Aβ. The monomers aggregate to form oligomeric species, fibrils and eventually the Aβ plaques that are a characteristic feature of the disease. Potential Aβ-directed therapeutic strategies include; a, b) targeting secretases to reduce Aβ production; c) activation of degrading enzyme, such as plasmin and neprilysin, to lower Aβ concentrations; d) inhibiting Aβ aggregation with the use of small molecules or chaperones; e ,f) use of anti-Aβ antibodies to sequester monomeric and oligomeric Aβ or activate microglia to clear amyloid deposits. Adapted from Pangalos et al.49

Given the central role of Aβ in the disease, many drug discovery initiatives have focused on the development of Aβ-based therapeutics. During a 10-year period of AD clinical trials (2002-2012), 48% of the disease-modifying therapies were targeted towards the peptide.50 Strategies have focused on decreasing Aβ production (targeting secretases), increasing clearance (activating degrading enzymes), inhibiting aggregation (small molecules and chaperones) and more recently through targeting downstream Aβ signalling pathways (Figure 1.3).26,51 Efforts to inhibit the production of Aβ by inhibiting γ-secretase have been limited given its essential proteolytic activities.26,51 β-secretase displays less substrate promiscuity, but strategies to target this enzyme have also been stunted, in this case by the large size of its catalytic pocket.26,51 Tactics aimed at inhibiting Aβ production or increasing Aβ clearance have been hampered by conceptual setbacks as studies have indicated that a certain amount Aβ is necessary to play a neuroprotective role in the normal brain.52 Inhibiting the aggregation process is an appealing strategy, as this would only target the pathogenic species without drastically affecting the endogenous (and functional) Aβ levels, potentially avoiding mechanism based toxicity. Aβ immunotherapy represents another promising strategy for disrupting the toxicity associated with the peptide.53,54 Interest in the field originated from the studies of Seubert and colleagues, where vaccination of APP transgenic disease model mice with aggregated Aβ42 prevented the formation of Aβ-containing plaques and AD-associated neuropathological changes.55 Since this study was published, a variety of immunotherapeutic agents have entered clinical trials, with the aim of promoting Aβ clearance and reducing the amyloid load.24

1.3.2 Structure of Aβ

Achieving a molecular level understanding of the structure of Aβ is incredibly important to unravel how the peptide functions in AD and to highlight targets for potential therapeutic intervention. Characterising the structural properties at the atomic level, however, is difficult. Aβ is an intrinsically disordered protein (IDP) and conformational fluctuations can give rise to a host of folded or partially folded states.56 Aggregated Aβ exists as a heterogeneous ensemble of polymorphic species, ranging from transient intermediary oligomers to well-ordered fibrils.57–59

Neither Aβ40 nor Aβ42 monomers crystallise and thus no x-ray structures of these entities exist. Under controlled conditions, oligomeric and fibrillar species are amenable to sophisticated structural characterisation techniques. Two- and three- dimensional nuclear magnetic resonance (NMR) spectroscopy and x-ray crystallography, for example, have been used to trap transient oligomeric intermediates under specific conditions, thereby providing insight into oligomeric

8 structural features.60–62 However, given the ambiguity over which species are responsible for toxicity, continued investigations are required.

In comparison to monomers, structural elucidation of fibrillar Aβ fibrils has shown more success.

Solid state NMR (ssNMR) studies have indicated that Aβ40 fibrils adopt a β-loop-β conformation with residues 1-10 fibrils structurally disordered, and 12-24 and 30-40 adopting β strand conformations (Figure 1.4a).63–65 The two β-strands are linked by a flexible loop region, which folds the two strands together to form a parallel β-sheet. The double layered sheet has a hydrophobic and a hydrophilic face, which is further stabilised by the formation of a salt bridge between

63 residues D23 and K28. Early ssNMR studies probing the structural features of fibrillar Aβ42, suggested that these fibrils also contain this β-arch motif61,66 More recent work, however, has indicated that fibrillar Aβ42 adopts more of an ‘S’ shape, with multiple parallel β-sheet segments (Figure 1.4b).60,67,68 Such work has provided the current understanding of the structure, in which

67,68 two S shaped Aβ42 chains meet around a twofold axis. Residues 1-14 are loosely ordered and residues 15-42 form four short β-strands, which stack along the fibril into parallel in-register sheets. Each stack of these dimeric units form a thread, two of which intertwine to give a two stranded protofibril.69 Elucidation of this structure offers new considerations for studying disease state. The observation that residues 22-23 are solvent exposed, for example, may aid efforts to understand the pathogenicity of mutations associated with early onset AD (e.g. Arctic E22G, Iowa E22Δ, Osaka D23N). Furthermore, an improved understanding of the molecular contacts may serve to highlight potential targets for drug development strategies.

Figure 1.4: Model structures of Aβ40 and Aβ42 determined using sold state NMR spectroscopy. a) Proposed structure of Aβ40 fibril. Residues 1–10 are believed to be unstructured and residues 11–40 adopt a β-turn- β fold. Dashed green lines indicate side chain packing. The turn conformation is stabilized by hydrophobic interactions (blue residues) and by a salt bridge between Asp23 and Lys28 (dashed pink line). Adapted from references 61. b) Ribbon diagram of the AB42 fibril core. Residue 15-42 of two symmetrical Aβ42 molecules are shown (yellow and orange). Unstructured residues 1-14 are omitted for clarity. β-stands are indicated by arrows. Positive and negative charged surface areas are shown in blue and red, respectively. Green indicates the presence of polar residues and all hydrophobic residues are in white. Pink labelling denotes residues associated with known familial Alzheimer’s mutations. Taken from reference 67.

1.3.3 Aggregation Pathway and Kinetics

Chemical kinetics – the measurement and analysis of the rates of chemical reactions – has significantly contributed to the fundamental understanding of the molecular mechanisms underlying amyloid formation.70 Despite the sigmoidal appearance of its kinetic profile, the aggregation of Aβ does not follow the typical linear processes associated with nucleation dependent polymerisation (Figure 1.5a). Amyloid formation is a complex multistep process, driven by a range of intricately connected microscopic reactions.56,71 Primary nucleation, whereby initial aggregated nuclei are formed from monomeric peptide, and elongation are generally accepted steps in the aggregation process. Some research groups have also suggested additional aggregation processes, such as secondary nucleation, where the surface of existing fibrils serve to catalyse the formation of new aggregated species, and fragmentation, where the breakage of existing fibrils exposes more free ends for elongation, thereby increasing the total number of elongating fibrils (figure 1.5b).71,72 Distinct aggregation processes occur simultaneously, albeit at different rates, so in contrast to the simplistic view given by the primary nucleation explanation, no single stage in the kinetic profile can be attributed to a specific aggregation event.71

Figure 1.5: The process of Aβ aggregation. a) General kinetic aggregation profile obtained using ensemble measurements. A sigmoidal curve, characteristic of nucleated polymerisation, is observed. This suggests that following an initial lag phase of monomer nucleation, an accelerated assembly phase occurs resulting in the depletion of monomers and formation of fibrillar structures.73 The aggregation process, however, is more complicated than this and consists of various microscopic aggregation events occurring simultaneously. b) Microscopic processes reported to underlie amyloid formation. Once primary nucleation occurs, fibril length can increase or decrease by elongation or dissociation. Secondary nucleation events driven by fibril catalysed monomer nucleation or the exposure of fibril ends by means of fibril fragmentation have also been reported as distinct events in the Aβ aggregation process. Adapted from Cohen et al.74

Dissecting aggregation behaviour into the component microscopic steps is of paramount importance for understanding the molecular processes involved in producing the pathogenic species and for probing the mechanisms through which aggregation modifiers exert their effect.56,70 Obstacles that initially restricted efforts to extract quantitative kinetic information from aggregation studies included difficulties in defining the rate laws for the complex self-assembly process and also the irreproducibility of experimental data. The development of protocols for scrupulous peptide preparation has allowed for the generation of highly reproducible kinetic data,75 and the formulation of new theoretical models has provided a means to relate macroscopic kinetic measurements with microscopic mechanisms. Such models have been employed to deduce differences in nucleation behaviour of the primary isoforms of Aβ,76 quantitatively assess the effect of extrinsic factors on the aggregation process,77 and to investigate the mechanism of action of small molecule inhibitors,78,79 chaperones, 80,81 and inhibitory ions, 82 as discussed in Section 1.6.2.1.

1.3.4 Aβ and Oxidative Stress in Alzheimer’s Disease

The term oxidative stress refers to the uncontrolled production of highly reactive oxygen species (ROS). These species can oxidise proteins, lipid and DNA, resulting in the impairment of many cellular functions and dysregulation of processes such as ion homeostasis and energy metabolism.83,84 Oxidative stress has been implicated in the development of AD, as well as a variety of other neurodegenerative disorders.83–85 The exact role of upregulated ROS production and disease onset and progression, however, is not yet clear. Various studies have indicated a link between Aβ aggregation and oxidative damage.85,86 For example, identification of abnormally high oxidatively damaged cellular components have been observed in Aβ dense regions of AD patients brains.86 Some research groups have also suggested an ‘Aβ induced oxidative stress hypothesis’, which places the majority of the causative effect of increased cellular oxidative stress upon

87,88 oligomeric Aβ42. Although the underlying processes linking Aβ pathology to oxidative damage are unknown, it has been shown that during the process of aggregation, Aβ can form hydrogen peroxide (H2O2) in a process that uses oxygen and is greatly accelerated in the prescence of the

2+ 3+ 89 redox active metals Cu and Fe (Figure 1.6). The resulting H2O2 is freely permeable to cross membranes and can generate reactive hydroxyl radicals, by means of Fenton chemistry, resulting in lipid peroxidation and damaging a host of other proteins.89 It has even been proposed that the neuronal cell loss in AD brains may be correlated with the excessive generation of such free radicals.83

Figure 1.6: Aβ aggregation and oxidative stress in AD. Aβ aggregation is associated with the reduction of 3+ redox active metals, such as Fe , with concomitant generation of hydrogen peroxide (H2O2). H2O2 is an oxidant stressor, the presence of which has many detrimental biological implications. This includes calcium dysregulation, which in turn can induce excitotoxicity and further production of reactive oxygen species (ROS). The free radical •OH, generated via the Fenton reaction, can also induce great oxidative damage and has been implicated as a causative agent in the neuronal atropy observed in AD. Therapeutic strageties that can target different steps in the pathway are highligeted in red. Inhibition of aggregation and metal chelation may reduce the generation of H2O2, ROS and other downstream effects. Memantine is used clinically to inhibit excitotoxicty by targeting the N-methyl-D-aspartate receptors, a glutamate receptor and ion channel found in nerve cells. Antioxidants could be used to dimish the reactiviy of toxic ROS. 83,90

1.4 Tau

Although there is a vast amount of evidence indicating that Aβ aggregation plays a central role in AD onset and progression, there are still many arguments which dispute its sole involvement.91,92 One of the main arguments against the amyloid cascade hypothesis is the discrepancy between the temporal and spatial Aβ plaque deposition with synaptic loss and clinical symptoms.93 There is, in fact, a better correlation between neuronal death and the presence of aggregated Tau structures.94 Tau is the primary microtubule associated protein in neuronal axons. In AD patients, the hyperphosphorylation and misfolding of this protein results in the formation of toxic neurofibrillary tangles (NFTs).26 Molecules capable of supressing the formation of NFTs, by preventing phosphorylation or aggregation for example, hold potential for the development of disease-modifying therapeutics (Figure 1.7). Various compounds capable of inhibiting the aggregation process have been identified and are progressing through clinical trials.95

Figure 1.7: Tau pathology potential strategies for therapeutic intervention. In a normal brain tau proteins are used to stabilize the microtubules, thereby supporting axonal transport. In AD, Tau becomes hyperphosphorylated and detaches from the microtubules. Soluble tau aggregates into pathological soluble tau oligomers, ultimately forming insoluble neurofibrillary tangles (NFT). Development of kinase inhibitors, to inhibit tau phosphorylation, and aggregation inhibitors represent promising tau-targeting therapeutic strategies. Adapted from the review by Citron.26

1.5 Monitoring the Aggregation of Aβ

Elucidating the molecular mechanisms guiding aggregation is essential for unravelling the relationship between aggregate structure and toxic properties, and also designing therapeutic and diagnostic strategies.56 There are a variety of assay formats for monitoring the self-assembly of amyloidogenic proteins and the effect that small molecule modifiers can have on the process. Although many powerful tools exist for probing amyloid aggregation, studying the real-time formation of different aggregated structures simultaneously remains difficult, given the complexity of the multistep process and the transient nature of the intermediary species. Furthermore, monitoring Aβ aggregation for the identification of inhibitors is further limited by the need to use high-throughput techniques, which often provide low-resolution results. Accurate analysis, therefore, requires a combination of orthogonal techniques. A number of such techniques, with respect to their applicability for potential drug screening shall be discussed, and a brief summary of the advantages and limitations outlined.

1.5.1 Microscopy

Scanning microscopy techniques, such as atomic force microscopy (AFM) and transmission electron microscopy (TEM), have been employed to probe the formation of amyloids for many years, and have provided great insight on the fine structural feature of the aggregated species.96–

98 These techniques benefit from the requirement of small sample sizes, high resolution and no need for modification of the peptide. They are, however, expensive to perform and generally involve elaborate or invasive preparation protocols, thereby limiting their applicability for real time measurements. Attempts to use fluorescence microscopy for the interrogation of amyloid structure were originally precluded by the sub-diffraction limit size of these species. The advent of super resolution microscopic techniques, such as stochastic optical reconstruction microscopy (STORM), stimulated emission depletion (STED) and structured illumination microscopy (SIM), has circumvented this issue and to date have permitted the imaging of individual fibrils revealing structural diversity and a means to dynamically study aggregation kinetics.99–103

1.5.2 Spectroscopic Techniques

Various spectroscopic techniques have been established to probe Aβ aggregation. Circular dichroism (CD) spectroscopy can be used to monitor changes in secondary structure,104–106 and electron paramagnetic resonance (EPR) and fluorescence anisotropy have been employed to detect the formation of early oligomeric structures.107–110 Dye binding assays remain the most commonly employed method and Thioflavin T (ThT) and Congo Red (CR) are the most classically used markers (Figure 1.8). CR has been used an amyloid marker since the beginning of the 1920’s, exhibiting a characteristic apple-green birefringence upon amyloid binding.111 It is primarily employed as a qualitative test in histological staining, and is not well suited to real time assays due to its reported ability to interfere with the self-assembly process.44,112,113

ThT is the most frequently used amyloid-specific dye for in vitro aggregation assays. It can bind to structures rich in β-sheets, resulting in a measurable shift and an enhancement in its fluorescent signal that correlates with increasing amyloid formation.114,115 Despite a vast array of experimental and computational studies investigating how it binds amyloidogenic species, the exact mechanisms of ThT-Aβ interactions are not fully understood.113,116–119 It is generally believed that the molecule intercalates within grooves between the solvent exposed side chains running parallel to the Aβ fibril axis. This causes rotational immobilisation of the central bond adjoining the benzothiazole and aniline ring, giving rise to an increased quantum yield. 115,120,121 The primary advantage of this assay is the relative simplicity and amenability to high throughput screening. ThT fluorescence assays have been employed in many successful primary screens for identifying aggregation inhibitors and in chemical kinetic studies.71,75,122 Issues with the technique arise from reproducibility and specificity issues.25,123 ThT is regarded as fibril specific, and as such it cannot

15 detect the formation of intermediary species lacking well-defined β-sheet structure. This is disadvantageous given the toxic nature of the soluble oligomers.124

Figure 1.8: Amyloid binding dyes and the spectroscopic properties of ThT exploited to monitor Aβ42 aggregation. a) Structure of Thioflavin T (ThT), Congo Red (CR) and heptameric formyl thiophene acetic acid (h-FTAA), a luminescent conjugated oligothiophene. b) Change in the emission fluorescence spectrum of ThT upon binding to Aβ42 fibrils. Assays are typically performed at an excitation wavelength of 440 nm and the emission recorded at 488 nm. c) Monitoring the Aβ42 fibril formation process by recording the relative changes in the fluorescence intensity as monomeric peptide is incubated over time.

Other shortcomings of dye-binding assays arise from the inherent properties of small molecules, such as intrinsic fluorescence, signal quenching or competitive binding with the dye or peptide.121,125,126 Many screening libraries are overpopulated with compounds that are autofluorescent.127 If these compounds are active in the spectral range interrogated by the assay, the resulting overlap can interfere with the assay readout. Competitive binding of exogenous compound along the fibrillary structure, or indeed interaction with ThT itself, can also drastically affect the assay performance.121,126

Recently, new amyloid specific dyes, either capable of detecting structurally distinct species or which have distinct absorbance profiles compared to ThT, have been added to the analysis toolkit. Luminescent conjugated oligothiophenes (LCOs; figure 1.8), for example, can detect oligomeric structures that occupy the early lag phase of the Aβ aggregation pathway, thereby giving a better representation of the modifying effects of small molecules that act at the early stages of the aggregation processes.128–130 These thiophene dyes represent sensitive and complementary tools for interrogating the aggregation process.

1.5.3 Separation Techniques

Techniques that provide insight into amyloid aggregation processes by separating aggregated species on the basis of size or mass include size exclusion chromatography (gel filtration), polyacrylamide gel electrophoresis (PAGE) and filter retardation (FR) assays.131 FR assays separate aggregates based on their size, by filtering a sample through a membrane trap with well-defined pore sizes.132,133 Aggregates above a certain cut off size remain on the membrane and are detected with antibodies or through the use of fluorescently labelled Aβ (Figure 1.9a).133 FR assays have been employed to measure the effects of extrinsic factors that affect ThT read out, such as human serum albumin and certain nanoparticles,133 and have proved useful in the detection of inhibitory molecules against the aggregation of the amyloidogenic huntingtin protein (HTT).134

1.5.4 Mass Spectrometry

The use of mass spectrometry and the hybrid technique ion mobility for studying amyloid aggregation has grown in popularity in recent times, and has been employed to detect aggregates in ex-vivo samples, to probe conformational conversions and to identify small molecule inhibitors.135–139 With respect to the screening of chemical libraries, this has proven to be a particularly advantageous technique, as it provides insight into the mechanisms of action of modulatory compounds, distinguishing between true, non-specific and colloidal inhibitors (Figure 1.9b).138,140 Furthermore, it has high sensitivity, the peptide does not require labelling and the assay is amenable to high-throughput screening.

1.5.5 Immunodetection

Immunodetection relies on the use of antibodies that are specific for an Aβ isoform or a specific conformation of an isoform.141–143 These can be employed with dot blot tests or in enzyme linked immunosorbent assays (ELISAs) to evaluate the concentration of a species of interest at a particular time point (Figure 1.9c). Conformation-specific antibodies can target soluble oligomers, fibrillary oligomers, protofibrils and fibrils.6,144–146 The A11 antibody, for example, specifically detects toxic oligomeric species and has been widely used to evaluate oligomerisation following incubation with small molecule inhibitors.144 These techniques are generally simple to perform, are quite specific and can be run with a relatively high throughput.131 However, immunoassays are often limited by the quality of the antibodies, which can be quite variable among batches.147 Also, the availability of epitopes is a factor that could interfere with the concentration to signal ratio, thereby compromising the ability to get a quantitative read out. For example, if the number of exposed epitopes is not well characterised, the assay cannot distinguish between the presence of a large number of small oligomers vs a small number of large oligomers.141,148

1.5.6 Functional Screens

Various functional assays that screen for aggregation inhibitors have been reported.140,149–151 These commonly employ fusion proteins, most often Aβ fused to green fluorescent protein (GFP). The misfolding of the attached Aβ causes the entire fusion protein to misfold, thereby preventing the GFP from adopting its fluorescently active structure (Figure 1.9d). Aggregation inhibitor discovery is facilitated through the addition of compounds that block aggregation and allow GFP to fold into its native conformation for detection. The assay can be performed in a high throughput manner using microwell plates and microscopy or flow cytometry, with hits easily identified by measuring the fluorescence output.149,151 Restrictions arise with compounds that have overlapping emission spectra with GFP, similar to the limitation in the ThT fluorescence assay. Furthermore, false negative results arise when compounds are unable to enter or produce toxic effects within cells. However, this may be viewed as advantageous, to filter out compound libraries for those that display unfavourable toxicity or low biological membrane permeability. Another recently developed functional assay employs the use of a β-lactamase tripartite fusion construct that associates antibiotic resistance with the inhibition of protein aggregation.140

Figure 1.9: Screening platforms for identifying Aβ aggregation inhibitors. a) Filter retardation assay. An Aβ sample is placed on a membrane with pores of a defined size. When a vacuum is applied monomeric species are pulled through the membrane. Larger aggregates are trapped and their concentration is quantified using immunological dot blots or fluorescence intensity readings. b) Mass spectrometry assay. Typical results of an nanoelectrospray ionisation mass spectrometry assay are shown. This technique enables the study of conformational changes in protein complexes with aggregation. Ligand binding can be observed and used to determine the mechanism through which potential inhibitors interact with the peptide. c) Immunodetection. Aβ-specific antibodies are immobilised on a surface and a sample of Aβ is added. Following the removal of unbound species, a secondary (detection) antibody that is specific for a certain Aβ isoform, or for a particular conformation, is added. The reporter tag is used to report on the concentration of captured species. d) Functional Aβ42-GFP assay. Aggregation of Aβ42 causes the entire fusion protein to misfold. The addition of molecules capable of inhibiting Aβ42 aggregation allows GFP to fold into its native fluorescent conformation. a), b) and d) adapted from references 133, 138, and 149 respectively.

1.6 Perturbing the Process of Amyloid β Aggregation

The identification and/or development of external agents or endogenous factors that can perturb the process of amyloid self-assembly holds great potential for the design of novel therapeutics agents or chemical probes to investigate the molecular mechanism underlying amyloidogenic diseases.26,28 Given the complexity of amyloid assembly processes, there are various stages for possible intervention. These include sequestering monomeric peptide, stabilizing native folded structures by ligand binding, direct oligomer formation into ‘off-pathway’ species, enhancing

19 fibrillisation or preventing the required conformational transitions by direct binding.28,152 At this point, it remains unclear as to which are the most effective strategies, as evidenced by the fact that no Aβ aggregation modifier has successfully progressed to the clinic for the treatment of AD.24,50 Since it has been established that oligomers are the most toxic species,45,58 attention has been mainly been directed at targeting the formation of these small entities.153–155 This is not, however, an easy task, as structural and functional characterisation of oligomeric aggregates is complicated by their heterogeneity in terms of mass, structure, stability and toxicity.58,156 The transient intermediates and oligomers that precede fibril formation adopt a large number of conformations, which are in rapid exchange with each other and also with monomers and fibrils.157,158

Figure 1.10: Strategies for interrupting amyloid aggregation. Small molecules can act at different stages of the aggregation process, inhibiting the formation of oligomeric or fibrillar structure (shown above), inducing the formation of off-target structures or disaggregating preformed fibrils. β-sheet breakers and peptidomimetics have been designed to inhibit aggregation by directly interacting with the central hydrophobic core of monomeric or fibrillar Aβ. The β-sheet breaker that is shown caps fibrils, preventing further elongation. Chaperones display inhibitory activity by acting at various stages of the aggregation process. The chaperones shown bind to the fibril surface, thereby blocking secondary nucleation. Antibodies have been developed to inhibit aggregation by binding to monomeric Aβ or to enhance clearance of preformed plaques. This can be achieved through binding to the amyloid and triggering either a phagocytic response by microglia, or by resolving the plaques directly, as shown in this figure.

A variety of ‘inhibitors’ capable of ameliorating the toxic effects of Aβ aggregation have been identified or designed (Figure 1.10). Peptidomimetic inhibitors and β-sheet breakers have been developed to target the hydrophobic core of the peptide, inhibiting the self-assembly process by ‘breaking’ the β-sheet forming ability.159,160 Cyclisation, strategic substitutions and the incorporation of unnatural residues has circumvented issues associated with poor bioavailability, and continued research in this field is providing promising results.161–164 The importance of chaperones in mediating Aβ aggregation has recently come to light and represent another worthy

20 avenue for drug development.81 These factors can function at different stages of the self-assembly process, preventing elongation disaggregating existing fibrils, for example.70,80,165,166 Aβ-targeted immunotherapy involves the use of different plaque-burden reducing strategies, which take advantage of the profound selectivity of antibodies.26,53 One such tactic includes inhibiting Aβ aggregation by means of binding to and sequestering monomeric peptide. Whilst this strategy has shown promise in animal AD models, the results in human trials have been disappointing. This is attributed to the poor in vivo efficacy, and the induction of adverse immune responses.54 Within the context of this project, small molecule modifiers are of particular interest, given their higher stability, blood brain barrier permeability and immunological tolerance than the aforementioned molecules.

1.6.1 Small molecule Aβ Aggregation Inhibitors

A great variety of small-molecule (organic, low molecular weight, chemical entities)167 inhibitors of amyloid formation have been discovered in attempts to develop anti-amyloid therapeutics. Natural products have provided a rich source of inhibitory compounds and chemical screening libraries have also identified promising leads.149,168–170 Technological development in mechanistic studies are ongoing and have started to provide insight into to the binding mechanism and function of different inhibitory classes. 70,171,172 Some general features of small molecule inhibitors have been elucidated. These include the observation that many small molecule inhibitors redirect the self-assembly process rather than inhibit it completely,59 that various independent remodelling pathways are possible,173 and notably, that relatively small changes in compound structure can encipher drastic changes in how compounds function in perturbing the amyloid aggregation process.44,173 This highlights the importance of continued compound screening and structure-activity relationship (SAR) generation to identify important pharmacophores and guide future drug development campaigns. A selection of key inhibitory compounds is discussed in detail below (Figure 1.11).

Figure 1.11: Selection of small molecule inhibitors

1.6.1.1. Select Aβ42 Aggregation Modulators

(-) Epigallocatechin Gallate (EGCG) is the most abundant antioxidant in green tea and has been shown to induce beneficial biological effects, including reducing the toxic effects of amyloid aggregation by modulating the self-assembly process.174,175 In the presence of EGCG, small non- toxic and non-seeding oligomers are formed from Aβ42 monomers in vitro. The compound can also remodel mature Aβ42 fibrils and toxic oligomers into smaller nontoxic aggregates with loss of β- sheet content.174,176 The mechanism of action, probed by NMR studies and other techniques, suggests that a combination of hydrophobic interaction and Schiff base formation between oxidized EGCG and the peptide is responsible for its effects on the aggregation process.175,176Curcumin is a component of the spice turmeric, and shares structural similarity with the organic dyes first identified as amyloid binders.112,177,178 It has been reported to bind both Aβ oligomers and fibrils, thereby inhibiting oligomerisation and promoting the deposition of non- toxic fibrillar structures.178–181 Some controversy exists over the extent of its inhibitory activity, as its intrinsic fluorescent properties can interfere with traditional spectroscopic screening techniques and other screening platforms have shown non-specific colloidal activity.121,138

Resveratrol’s ability to provide neuroprotective effects in AD models is attributed to a variety of factors, including its ability to abrogate Aβ aggregation-induced toxicity.182,183 The compound can directly interact with the peptide to inhibit fibrillisation and remodel or disaggregate preformed fibrils or toxic oligomeric species.182,184,185 Morin and myricetin belong to the flavonoid family of inhibitors, the activity of which depends greatly on the number and positioning of the hydroxyl groups.186–189 This family will be further discussed in Chapter 6. Brazilin was identified as an Aβ inhibitor through virtual screening of a natural product library, using compound-docking

190 simulations and incorporating 36 previously reported inhibitory compounds. It reduces Aβ42 cytotoxicity by inhibiting fibrillation, remodelling mature fibrils and redirecting monomers to unstructured aggregates.190

Tramisprosate is a small ionic compound that has been shown to bind to Aβ42 and maintain it in a

191,192 non-fibrillar form, thereby decreasing amyloid load and Aβ42 aggregation-induced toxicity. Although preclinical trials showed promising effects,193 the compound was unsuccessful in phase III trials, a result attributed to poor target engagement and limited understanding of its mechanism of action.92 EPPS (4-(2-hydroxylethyl)-1-piperazinepropanrsulphonic acid) shares a sulfonic acid group with tramisprosate, and has recently been shown to disaggregate Aβ oligomers and plaques.194 Its activity has been verified in vivo, where it was shown to rescue hippocampus- dependent behavioural deficits and other severe AD-like phenotypes in mouse models. Bexarotene was originally identified as a potential AD therapeutic through recognition of its ability to promote Aβ degradation by means of increasing ApoE production.195 Since this discovery, subsequent work has highlighted an alternative mechanisms of action of this repurposed drug in AD treatment, through the inhibition of Aβ aggregation.78,196 Chemical kinetics-based studies have shown that the compound functions by inhibiting the primary nucleation step, and a delay in the formation of toxic Aβ species was observed upon compound treatment in neuroblastoma cells and in a Caenorhabditis elegans model of Aβ-induced toxicity.78

Inositol is a simple polyol with nine naturally occurring stereoisomers, which differ in the arrangement of their hydroxyl groups.197 Inositol displays stereochemistry-specific inhibitory effects on Aβ fibril inhibition, with scyllo-, myo-, and epi-, but not chiro-inositol, capable of

198 modifying the Aβ42 aggregation process. This highlights the stringent structural and functional group considerations that need to be taken into account when designing and optimising inhibitors. Scyllo-inositol, with all equatorial hydroxyl groups, has been identified as a promising therapeutic candidate for the treatment of Alzheimer's disease. This compound functions by stabilizing non-

199 toxic oligomeric complex of Aβ42 and has reached phase II clinical trials.

1.6.1.2 Rational Design of Inhibitors

In general terms, structure based drug design (SBDD) is the process of developing drug candidates by exploiting structural knowledge of the target.200 NMR or X-ray crystallography are typically required to determine the molecular structure of the target macromolecule, which then facilitates the selection or construction of chemical libraries for High Throughput Screening (HTS) or the design of ligands to fit well-defined binding sites. Performing SBDD with IDPs, such as Aβ, is generally difficult, as they lack stable secondary or tertiary conformation, thereby precluding attempts to design small molecules complementary in size, shape and charge to the target in which they are intended to bind.201 As a result of the heterogeneous nature of oligomers and prefibrillar Aβ, experimental determination of the molecular structures of these species for targeted drug development remains challenging.202

In spite of the challenges presented by the intrinsically disordered nature of Aβ, considerable progress has been made with the rational design of new aggregation inhibitors. The approaches that have been employed rely on the experimental generation of SAR data using concise screening libraries, computational analysis and virtual screening, or fragment and homology-based methods. Several published studies have highlighted critical pharmacophoreic features or robust SAR observations, which can provide structural template to design new inhibitors, incorporating physiologically relevant motifs and optimised pharmacokinetic features.168,203,204 For example, in a SAR based drug development strategy, Arai et al. tested a selection of synthetic cyclic analogous of residues 16-20 of Aβ.205 Through analysis of the associated inhibitory activity, a pharmacophore involving specific amino acid side chains was identified and used to guide the synthesis of an active non-peptidic, small-molecule inhibitor. In another study, the known atomic structure of a portion of Aβ, which had been recrystallized with a ligand dye, was used as a template to computationally screen 18,000 compounds.204 A subset of likely binders was identified, and these compounds were shown diminish Aβ induced cell toxicity when tested experimentally. The SAR data generated gave insight into the mechanism of action (increased fibre stability and decreased fibril toxicity), which will be employed for further inhibitor development.204

The Vendruscolo research group has significantly contributed to the development of protocols for rationally developing amyloid inhibitor drugs.78,206 In a quasi-SBDD approach, structural characterisation of ligand binding domains of other proteins known to interact with the Aβ aggregation inhibitor bexarotene was used to assemble a small library of potential Aβ binders.79 These compounds were tested for inhibitory activity against the aggregation of Aβ, and were shown to perturb distinct microscopic steps in the self- assembly pathway.79 In an alternative

24 approach, molecular fragments extracted from previously reported amyloid aggregation inhibitors were employed in screens of existing libraries, to identify compounds rich in desirable IDP binding features.206 This approach was facilitated by open access and commercial drug databases, and the information mining allowed for the generation of small libraries of potential inhibitors against the aggregation of Aβ, Tau and α-synuclein – an amyloidogenic protein associated with Parkinson’s disease.

1.6.2 Aβ Binding and Inhibitory Mechanisms

Computational simulation methods can circumvent the challenges posed by the heterogeneity of Aβ structures and the lack of molecular-level understanding of small molecule binding sites, and have been employed to provide atomic level description of the inhibitor-peptide interaction.158,172 Molecular dynamics have played a key role in deciphering the binding interactions of various aggregation inhibitors and amyloid binding dyes.188,207–210 Molecular docking has also been employed to characterise the interaction of Aβ and small molecules, providing insight into inhibitory mechanisms.211,212 Although increasingly useful, these techniques still suffer limitation and thus far only some of the physical reality of compound binding can be perceived and/or rendered by modern computer-based techniques.172,213

An armoury of mechanistic information can be extracted from the aforementioned amyloid aggregation screening strategies (Section 1.5). Different biophysical techniques have been employed to elucidate mechanistic pathways used by small molecules to perturb the aggregation process, independent of how they may interact with the peptide at the molecule level. Understanding these processes is incredibly important, not only for pharmaceutical implications, but also for unravelling the mechanisms underlying amyloid aggregation as a whole. Insight into the mechanistic action of aggregation modulators hold potential for illuminating the steps of the self-assembly process. For example, in a comprehensive study by Necula et al., a combination of light scattering, ThT fluorescence, and TEM imaging and immuno dot blots were used to reveal that inhibitors can function to inhibit the processes of oligomerisation, fibrillation or both. This highlighted that oligomers are not necessarily obligate intermediates on the aggregation pathway, an insight that had not been explicitly revealed by any other study to date.44

1.6.2.1 Chemical kinetics

Chemical kinetic studies aim to connect macroscopic measurements with microscopic mechanisms, through the use of powerful kinetic models. Such studies have been employed to shed light on the processes guiding Aβ self-assembly and the effect that aggregation-modulators can have on these processes, and have made extensive use of ThT fluorescence assays.70,214 In general terms, this is achieved by dissecting aggregation behaviour into the component steps and making quantitative predictions about how the overall kinetic profile would change if specific processes were inhibited (Figure 1.21).70,215 Supressing primary nucleation, for example, would be expected to result in a longer lag-time in the characteristic profile. Aggregation in the presence of test compounds can then be compared with predicted profiles, to determine the microscopic steps affected. This technique requires reproducible experimental data, which has been facilitated by the publication of rigorous peptide preparation protocols,75 and complementary experimental analysis is also required to determine if the modulators function in multiple processes. This strategy has been employed with inhibitory small molecules, chaperones and ions, allowing their effect at a molecular level to be translated to observable differences in macroscopic behaviour.78– 82 For example, the chaperone Brichos was found to inhibit the secondary nucleation pathway by capping the catalytic sites on fibril surfaces.80 In contrast, zinc ions have been shown to suppress the aggregation process by exclusively inhibiting the elongation process.82

Figure 1.12 Predicted macroscopic kinetics upon perturbation of the specific microscopic processes in Aβ aggregation. a) Aggregation modifiers can interfere with specific microscopic steps within the global reaction process. Inhibition of primary nucleation, elongation and secondary nucleation are shown. b) In each graph the purple line represents the evolution of fibril mass over time for a control Aβ sample. Model simulations are performed to predict how changes in the microscopic steps in the aggregation process would affect the observed macroscopic profiles. Through analysis of global kinetic profiles with different

26 protein and inhibitor concentration, the specific processes targeted by an inhibitory molecule can be determined. The graphs above show the predicted changes in global aggregation profiles when the individual molecular steps in panel a) are specifically inhibited. Adapted from Arosio et al.215

1.6.3 Past Failures and Future Direction for the Development of Aβ Aggregation Modifiers

The amyloid hypothesis has significantly influenced research and development strategies for the design of AD therapeutics.92 Although recent times have seen advances in our understanding of the molecular mechanisms underlying Aβ aggregation, as well as in the development of analytic tools to profile the self-assembly process and screen for molecular inhibitors, no therapeutics aimed at perturbing Aβ structure or function have been approved for clinical use. In fact, no disease-modifying AD therapeutics at all have come to market.24 Various publications have reviewed the failures of previous clinical trials targeting Aβ and have evaluated the current landscape of Aβ-focused clinical studies.24,50,92,216 In an analysis of potential AD therapeutics in trials from 2002-2012, of the 53% that were classified as disease modifying (as opposed to symptomatic), 48% were directed against Aβ.50

Deficits in the understanding of how potential drugs function at a molecular level are often held responsible for their disappointing activity in human trials, which is further complicated by the lack of mechanistic links between the formation and deposition of aggregates and the resulting toxicity.92 Such failures have called some to question the ‘amyloidocentric’ therapeutic approach, suggesting that drug development should focus exclusively on other pathogenic elements of the disease.91,217,218 Prominent researchers in the field, however, have emphasized that the failure of hypothesis testing must be distinguished from failures of the hypothesis.92,216,219 The clinical results obtained, thus far, may not support the amyloid cascade hypothesis, but they do not refute it either. Failures may be attributed to technical limitations, rather than a result of following a fundamentally misguided approach.216 For example, there were large gaps in both data and knowledge in the tramiprosate clinical trials. Early investigations were performed with very high drug and peptide concentrations, that aren’t considered physiologically relevant and little dose- response data was reported.25,92 Also as target engagement wasn’t monitored it is difficult to define if the lack of cognitive enhancement was a due to limited compound activity or the failures of inhibited Aβ aggregation to bring about the expected result.92 In this regard, it is noteworthy that in the aforementioned 10 year clinical trial analysis, no disease-modifying compounds at all were approved for the clinical.50 Incredibly high attrition rates were observed for all therapies, with rates of 72% in phase I, 92% in phase II and 98% in phase III.50

A better understanding of the aetiology and pathological mechanisms of disease are likely to contribute to the development of more efficacious therapies. It is believed that clinical trials often take place too late in the disease process, and earlier intervention is necessary to provide a more significant impact on treating amyloidogenic disorders.25 Targeting the formation of the initial oligomers represents a promising strategy. This tactic requires treatment sooner in the disease progression, thereby highlighting the need for prodromal biomarkers for diagnosis.24 Improved understanding of all the contributing factors of AD may guide the development of more holistic approaches to treating the disease. Poor recognition of the many factors which together bring about pathological phenotypes limits current therapeutic strategies and the efficacy of combination drug therapies should be studied.25 Targeting different steps of the aggregation process may even provide cumulative disease-modifying effects.

1.7 The Drug Discovery Process 1.7.1 Early Drug Discovery and Compound Screening

The development of an approved drug from the initial idea to its entry into the market is a long and expensive process, taking approximately 12-15 years and costing over $1 billion.220 The early drug discovery process usually starts with the identification and validation of a target, then passes through subsequent stages of compound screening, hit identification and optimisation via medicinal chemistry approaches, secondary screening and then in vivo analysis to determine safety and efficacy (Figure 1.13).220 At this point, a final candidate has been sufficiently characterised to start the preclinical investigation stage, before finally entering clinical trials.

Figure 1.13: Early drug discovery process. Medium or high-throughput screening is performed to identify compounds of interest, which are then optimised through strategic SAR investigations and iterative rounds of chemical synthesis and rescreening. Hits are validated in secondary assays to confirm activity in orthogonal assay formats and in cellular models. Mechanistic and binding studies are performed to gain a better understanding of hits’ mode of action. Lead compounds are next tested in whole organisms. Primary in vivo screening filters out compounds that induce toxic effects or that cannot reach their targets due to instability or inability to penetrate anatomical barriers. Disease model animals are then employed to probe pharmacological activity. Given promising results in early toxicology, pharmacokinetic and pharmacodynamics screening the lead compound will enter the preclinical testing stage. Adapted from reference 220.

The success of a drug screening campaign relies on the constituent compounds of the libraries used.221 Whilst HTS screening strategies have represented a valuable starting point for many drug discovery programs in the past, it is increasingly acknowledged that the use of classical compound libraries is providing fewer and fewer hits. Many of the commercial and pharmaceutical libraries employed contain structurally and functionally similar compounds, often sharing very similar regions of chemical space.221,222 In this respect, it is often said that all the low hanging fruit is depleted and that new and chartered regions of chemical space must be explored to identify hits for illusive targets.222,223 Diversity-oriented and biologically-oriented synthesis represent promising strategies for improving the structural diversity of chemical libraries and have shown promise in targeting previously intractable disease related processes.221,224

1.7.2 Physicochemical Properties

The chemical structure of a compound has a direct influence on its physical, chemical, and biological properties. The potential of a small molecule to act as an effective therapeutic depends on a combination of these physicochemical properties, which can be modified to optimise the drugs pharmacokinetic (PK) profile - namely the absorption, distribution, metabolism and excretion (ADME) of the compound.225 Most small molecule therapeutics are designed to adhere to the criteria set forward by Lipinski.226 According to Lipinski’s rule of five (Ro5), active drugs are those in which the molecular weight is <500 Da, the logP value is <5, there are less than 5 hydrogen bond donors (HBDs) and less than 10 hydrogen bond acceptors (HBAs). Lipinski’s Ro5 has been well validated and adopted by the scientific community, however revaluations have been deemed necessary for the development of CNS drugs.227

The brain is protected by a complex multicellular system, known as the blood brain barrier (BBB).228,229 This sophisticated barrier consists of a layer of high-density endothelial cells at the interface between the blood capillaries of the brain tissue and the brain. It restricts the passage of substances from the bloodstream, thereby protecting neural tissue from variations in blood composition and potentially harmful xenobiotics and toxins. Existence of the BBB presents a unique challenge for CNS drug development, as designing drugs to infiltrate this physical and enzymatic barrier can prove difficult. What’s more, no satisfactory in vitro models of the BBB are available, complicating attempts to determine pharmacokinetics of hit compounds.227 It is estimated that 98% of systematically administered small molecules, and almost all biologic therapeutics, are unable to pass the BBB. 227 As such, CNS drugs have significantly higher attrition rates than non-CNS drugs in clinical trials. 230 Despite these unfavourable statistics, the BBB is not impenetrable and CNS active therapeutics have been approved.227,231 Analysis of the features of these marketed drugs, which mostly cross the BBB via transcellular passive diffusion, has guided efforts to characterise the physicochemical properties required for optimal brain exposure. 227,232,233

Table 1.1: Optimal physicochemical parameters for CNS exposure.233

Property Desirable range cLogP ≤ 3 cLogD ≤ 2 Molecular weight ≤ 360 Topological polar surface area 40 < tPSA ≤ 90 Hydrogen bond donors ≤ 0.5 pKa ≤ 8

Several calculated fundamental physiochemical properties are widely recognised as key parameters in drug design.227,231,232 Lipophilicity, which is expressed as the calculated logarithm of the octanol/water partition coefficient, is considered one of the most important parameters. It has important implications for in vitro potency and the total brain level achieved. cLogD is a related measure, the partition coefficient calculated at pH 7.4, and is a more physiologically relevant parameter for investigating drug portioning in an in vivo environment.231 Hydrogen bonding capacity can also significantly affect CNS drug activity, with increased numbers of HBDs and HBAs associated with lower permeability. Lowering the number of these moieties is a frequent strategy for improving brain exposure.227 Polar surface area, a measure of the surface area (Å2) over all polar atoms including attached hydrogens, is used as a surrogate measure of H-bonding capacity and molecular polarity. Size, ionization properties, and molecular flexibility are other factors observed to influence transport of compound across the BBB.231 An overview of the desirable feature for BBB permeability is shown in table 1.1. Generally, CNS drugs are smaller, more lipophilic, have a small polar surface area (PSA) and fewer H bonds donors than non-CNS drugs.227 However, these parameters are not set in stone and applying hard physicochemical cut-offs in the selection of screening libraries is not compulsory. This may, in fact, restrict deign space, resulting in important hits being missed. Similarly, focusing on a single property is not an appropriate strategy, as it may not align multiple attributes at once.233 Multiple physicochemical properties must be intricately balanced in drug optimisation strategies to improve on deficits in one property without causing negative impact on another.

1.8 Fluorescence Spectroscopy 1.8.1 Basic Principles

Fluorescence is described as the process of excitation of an atom or molecule by absorption of a photon and its de-excitation by emission of another photon.234 The processes that occur between the absorption and emission of light are often illustrated by the Jablonski diagram (Figure 1.14a).

The ground state and lowest energy singlet excited state are denoted as S0 and S1 respectively, and the closely spaced levels within each electronic state represent vibrational energy levels. A fluorophore is excited by absorption of an excitation photon (represented by the energy hνA) leading to a transition to a higher electronic energy state (S0 → S1). The fluorophore then relaxes

(S1 → S0) by emitting another photon (energy hνF). The emitted photon has a lower energy (and hence longer wavelength) than the absorbed photon, due to loss of energy to the environment, as shown in the characteristic absorbance and emission spectra (Figure 1.14b). This phenomenon

31 is called ‘Stokes shift’ after the scientist who first reported it.235 Fluorescence spectroscopy takes advantage of the Stokes shift to selectively collect the fluorescence from a sample (a population of fluorophores) by spectral separation (Figure 1.14c) .

The radiative relaxation pathway (i.e. the one leading to fluorescence) for the excited electrons is in competition with non-radiative pathways, such as molecular collisions, intersystem crossing and Förster resonance energy transfer (FRET).234,236 The interplay between these factors contribute to the quantum yield and the fluorescence lifetime of the fluorophore.234 The quantum yield is the ratio of the number of photons emitted to the number of photons absorbed, and represents the probability of the fluorophore to lose its excitation by fluorescence (radiative decay). The fluorescence lifetime (τ) is the average time that a fluorophore remains in the excited state, and is defined by the equation 1.1, where kr is the fluorescence emission rate (radiative) and knr is the non-radiative decay rate.

1 τ = 푘푟 + 푘푛푟 Fluorescence Lifetime (1.1)

These fluorescence properties, as well as other factors, can be experimentally exploited to study biological systems and pathways in fine detail.236–239

Figure 1.14: Spectral properties of fluorescence molecules a) Modified Jablonski diagram. Each arrow within the diagram represents a transition from one energy state to another. When light of the correct energy is absorbed it can raise a fluorophore from the ground state (S0) to an excited singlet state (S1). Excitation to any of the multiple closely spaced vibrational energy levels is possible. The emission of a photon (fluorescence) returns the molecule to the ground state. Fluorescence occurs within the ps to ns time scale, and is always Stoke shifted to a longer wavelength due to loss of energy to the environment through non-radiative processes. Intersystem crossing occurs when a fluorophore undergoes a transition towards its triplet state, resulting in the emission of a photon by phosphorescence. b) Relative indication of the wavelength positions of the absorption, fluorescence and phosphorescence spectra. Wavelength (λ) is inversely proportional to energy. c) A sample is excited with a spectrally selected light. The fluorescence emission is Stokes shifted and is separated from the excitation light with an emission filter.

1.8.2 Fluorescence Lifetime Imaging Microscopy (FLIM) 1.8.2.1 Basic Principles

Fluorescence lifetime is defined as the average time that a fluorophore remains in an excited state before returning to the ground state by emitting a photon, and can range from picoseconds to hundreds of nanoseconds.234,240,241 Environmental composition and a variety of other processes, such as quenching, charge transfer, solvation dynamics and FRET can affect the lifetime of a fluorophore.242 Variation in fluorescence lifetime with changes in environmental condition, protein structure or interactions with a substrate can be exploited to monitor reaction processes, providing absolute measurements along the molecular pathways.240,241 In this respect, lifetime

33 measurements are advantageous over steady state fluorescence techniques, where data are usually relative or average. Additionally, FLIM is mostly unaffected by inner filter effects and independent of concentration.240 The technique is particularly useful in situations where local probe concentrations can’t be controlled, as in the cellular environment.

There are two principle methods for measuring fluorescence lifetime: frequency-domain and time-domain measurements.234,240 In frequency-domain measurements, fluorophores are excited with a modulated light source. The fluorescence emission is demodulated and phase shifted with respect to the excitation beam, which directly depends on (and can be used to calculate) the fluorescence lifetime. In time-domain measurements, fluorophores are excited with short light pulses and the arrival times of individual photons are recorded to measure the fluorescence intensity decay. Such measurements can be made using time-correlated single photon counting (TCSPC) or fast-gated image intensifiers. Multidimensional TCSPC delivers the highest time resolution, as well as the best lifetime accuracy and photon efficientcy.240

1.8.2.2 Time-Correlated Single Photon Counting FLIM

TCSPC FLIM is based on the measurement of photon arrival times after a laser pulse and is achieved by repetitive raster scanning of a sample to build up two-dimensional image. Here, image contrast is given by the measured lifetime at every pixel and visualization is achieved using false colour coding.240 TCSPC measurements are performed by periodic and pulsed excitation of a fluorescent sample. A short pulse of light is used to excite the sample and the arrival time of a photon following the laser pulse is detected (Figure 1.15).243,244 Using accurate pulsed excitations sources together with extremely sensitive photon counting detectors and an advanced set of electronics, TCSPC set-ups can determine the time difference between excitation and emission with picosecond resolution. Data are collected over multiple cycles of excitation and emission. It is necessary to collect less than one photon per 100 excitation pulses, to avoid potential bias of the lifetime measurement by photon pile up.245,246 This process is caused, in part, by the dead time in the detector and electronics after a photon event. If there are more than one photons per cycle, those arriving after the first will fall into the dead time and will not usually be registered (Figure 1.15d). This can lead to an over representation of early arriving photons in the histogram, thereby skewing the final result.245 The repetitive and precisely timed detection of single photons allows for the reconstruction of the fluorescence decay profile from many single events collected over many cycles. A distribution of photon arrival time after excitation over the coordinates of the whole field of view is built up, generating the final time-resolved image.240

Figure 1.15: Time-resolved florescence measurement with TCSPC. a) A sample is excited by a laser pulse of low intensity, such that the arrival time of the individual photons can be measured with picosecond or nanosecond resolution. b) Start-stop times for each of the individual photons are recorded, with the time of image acquisition determining the number of photon arrival times recorded. c) A histogram of these times is built up and is representative of the fluorescence decay characteristics, thereby allowing calculation of the lifetime. d) If a photon arrives within the dead time after the first photon event, it will not be registered and will be omitted from the histogram. This is known as photon pile-up. As only the first photon to arrive is recorded, this results in an over representation of photons with short arrival times in the calculation of the fluorescence lifetime, highlighting the need for a low probability of registering more than one photon per cycle. Figure adapted from reference 245.

A general TCSPC set up is shown in figure 1.16.246 A point in the fluorescent sample is excited with a pulsed laser and individual fluorescence photons are detected and converted into electronic signals by a sensitive detector and a constant fraction discriminator (CFD). Photon arrival is time correlated with a reference signal from the excitation pulse using a time-to-amplitude converter (TAC). This functions as a ramp generator, which starts charging upon a ‘start’ trigger and is stopped when a ‘stop’ pulse is received. The resulting voltage is proportional to the time difference between the two signals and is read out using an analogue-to-digital converter (ADC). The arrival times of hundreds to thousands of individual photons are stored within the TCSPC and are plotted as a histogram, which represents the probability density of fluorescence emission. The fluorescence decay parameters may be extracted from this histogram, if sufficient number of photons have been acquired. TCSPC provides very high photon efficiency and temporal resolution,

35 but is limited by the long acquisition time (minutes) due to single pixel scanning. Photostability can also affect fluorescence lifetime measurements. Issues with photobleaching, whereby fluorophores no longer fluoresce due irreversible chemical reactions upon repeated excitation, can preclude lifetime measurements by limiting the number of detectable photons.240 Measuring fluorescence lifetime requires a high photon count, which requires many excitation-emission cycles for the generation of representative data. This can lead to relatively fast photobleaching, especially in samples with low concentrations of fluorophores.

Figure 1.16: TCSPC Set up. The laser pulse used to irradiate the sample and the detection of the photon are converted to electronic signals by the constant fraction discriminator (CFD). These signals act as a ‘start’ and ‘stop’ triggers, to generate a voltage proportional to the time difference in the time-to-amplitude converter (TAC). The voltage generated is converted to a digital recording of arrival times by an analogue-to-digital converter (ADC). Storage of hundreds to thousands of arrival times results in the formation of a histogram which reflects the fluorescent decay.246

1.8.3 Fluorescence Techniques for the Study of Amyloid Formation

Fluorescence based spectroscopic techniques have provided incredible insight into amyloid self- assembly processes in vitro, as well as in cellular and whole organism models.247 Fluorescently

36 active amyloid binding dyes have seen extensive use for probing aggregation kinetics and staining biological samples, as previously discussed in section 1.5.2.114,248,249 Super resolution microscopy has allowed for the dynamic investigation of aggregation at the individual fibril level.99,101,103,250 The discovery of an amyloid specific intrinsic fluorescence signature in the visible range has provided a non-destructive platform to study self-assembly,251 and prompted study into the development of a fluorescence lifetime sensor for monitoring the aggregation process.252–255 Given their relevance to this study, these two techniques shall be discussed in further detail.

1.8.3.1 Intrinsic Fluorescence

Proteins can display intrinsic fluorescence as a result of their constitute aromatic amino acids, which can absorb and fluoresce in the ultraviolet range (250-400 nm).256 The use of intrinsic fluorescence measurements for studying protein structure and function can be advantageous over the use of fluorescent fusion protein (such as GFP) or synthetic fluorophore labels, when little or no structural protein modifications are required. Alterations in the fluorescence decay of tyrosine residues, for example, have been used to characterise Aβ aggregation events and can detect the formation of aggregated species much earlier than conventional dye binding experiments.257 Limitations with such measurements arise when mutant forms of the protein are necessary, as the non-native substitutions can have a deleterious effect on the aggregation process as a whole, or on the final species formed.258 Although this technique is not generally applicable to proteins lacking UV active amino acids, intrinsic fluorescence has been reported for proteinaceous structures high in β-sheet content – such as amyloids.251

It has been shown that upon aggregation, amyloidogenic proteins develop an intrinsic fluorescence, the emergence of which can be measured as a function of time and correlated with the extent of amyloid content.251,259 The characteristic fluorescence properties have been experimentally determined (excitation λ > 360 nm, emission λ > 400 nm, lifetime 1-3 ns),251,259 and the similar spectral signatures between amyloid species suggest that the observed intrinsic fluorescence is a generic property of species with the characteristic β-sheet core.259 State-of-the- art ab initio molecular dynamic simulations and fluorescence spectroscopy were used to elucidate the origin of this phenomenon.260 It was found that proton transfer across the dense network of hydrogen bonds that stabilise the β-sheet structure is directly coupled with the intrinsic fluorescence. Fluorescence is observed even in the absence of classical π conjugated system, and the emission originates from the presence of fibrillary species, thereby providing a label free method for probing the fibril assembly process.260

The use of intrinsic fluorescence measurements to faithfully report on peptide aggregation has been validated through comparison with established assay formats and direct visualisation techniques.251,259 The system also provides a method for screening small molecules for inhibitory activity against the aggregation process, through monitoring their effect on the emerging fluorescence. Furthermore, it is not limited by the formation of a probe specific binding site, as in the case with ThT ad CR, so represents a more robust assay format. Indeed, it has been shown to detect an incorrectly assigned ThT fluorescence false positive inhibitor (lacmoid) of α-syn aggregation.259 The technique does, however, suffer from a low quantum yield. Monitoring an enhancement in the fluorescence intensity signal with self-assembly cannot be applied in a biological context, due to competing intrinsic fluorescence of other proteins.251 The great potential of this intrinsic fluorescence phenomenon has, however, prompted the development of a related probe that can be applied in a cellular environment and in vivo, hereafter called the ‘lifetime sensor’.

1.8.3.2 The Amyloid Aggregation Lifetime Sensor

The amyloid aggregation lifetime sensor relies on the use of partially labelled monomeric peptide, whereby the fluorescence lifetime of a covalently attached fluorophore is used to report on the overall aggregation state of the peptide. A time dependent decrease in fluorescence lifetime of the dye is observed upon amyloid formation.252–254 Fluorescent fusion proteins or synthetic dyes have both been validated as effective reporters in this format.253,261 Comprehensive biophysical studies have been performed to confirm that decreasing the fluorophore lifetime of the attached fluorophore correlates with increasing peptide aggregation and the system has successfully been employed to quantify the aggregation kinetics of amyloidogenic proteins in cells and living systems.252,253

The fluorescence lifetime change upon amyloid aggregation was initially attributed to FRET – a distant dependent interaction between two fluorophores, where the emission band of one (the donor) overlaps with the absorption of the other (the receptor).252 Non-radiative transfer of energy from donor to acceptor, results in quenching of the donor fluorophore and an associated decrease in its fluorescence lifetime. The labelled peptides were originally designed so that there was a spectral overlap of the emission spectrum of the fluorophore and the amyloid-specific excitation spectrum. It was believed that the new energy level imparted by the β-sheet high amyloid structure was acting as an acceptor for the extrinsic dye via hetero-FRET processes.252 More recent work, however, has shown that the mechanism of the lifetime sensor is in fact based

38 on fluorescence self-quenching of the reporter dye (Figure 1.17).255 Amyloid aggregation brings dye molecules closer together, thereby increasing the local concentration and permitting increased self-quenching of the dye molecule and resulting in an associated decrease in fluorescence lifetime. As such, the sensor is sensitive to fibril packing and reports on the density and other morphological features of the developing amyloid clusters. This is of great importance as structurally distinct aggregates are formed under different environmental conditions and are believed to have different implications for the disease state.144,262

Figure 1.17: Use of the fluorescence lifetime sensor for monitoring Aβ aggregation. The lifetime sensor reports on Aβ aggregation by measuring the fluorescence lifetime of a covalently attached reporter dye. As the peptide aggregates, the dye molecules are brought close together, and this increased local concentration results in increased self-quenching. The concomitant reduction in the fluorescence lifetime of the dye can be used to report on the degree of peptide aggregation.255

Self-quenching of neighbouring dyes has previously been employed for monitoring aggregation, by means of measuring the overall decrease in fluorescence intensity in a time-dependent manner.124,263 However, the lifetime sensor described is advantageous over this and other techniques in many respects. The picosecond time-resolved fluorescence measurement is less prone to concentration and intensity artefacts.255 In steady state techniques (including conventional ThT fluorescence assays) relative or average results are typically obtained by the measurement of spectra. Such results can be difficult to compare technical and biological repeats obtained independently. In contrast, the lifetime sensor gives absolute values, which are consistent under the same experimental conditions and permit data comparison from separate experiments.255 Furthermore, this sensor has potential for high throughput studies, so is favourable over highly specialised single molecule techniques, which are difficult to implement in a high-throughput manner. This is of special interest for the efficient screening of compound libraries for aggregation inhibitors. As the lifetime values along the aggregation pathway are indicative of the extent of aggregation, morphological properties and the density of the resulting species, such efforts may lead not only to the identification of inhibitors, but also compounds that

39 can modify the aggregation process in other ways. Compounds that can induce the formation of certain species, diffuse structures or ‘off target’ compact oligomers for example, could easily be picked up in such screens, providing benefit over morphological specific detection techniques. Additionally, fluorescence lifetime imaging is a non-invasive technique, and can be employed to monitor aggregation in whole organism models, thereby providing a unique capability for monitoring and comparing Aβ aggregation in vitro and in vivo.

1.9 Microfluidics

Microfluidics is the science and technology of systems that process or manipulate small amounts of fluid (10-6 to 10-12 litres) in networks of sub millimetre scale channels.264 The use of microfluidic techniques has grown exponentially since it first applications in the early 1990’s and the miniaturisation associated with these systems allow for high resolution and sensitivity in a variety of analytical techniques.264 Within this context, the use of microdroplets has opened a realm of possibilities for carrying out large scale investigations of integrated biological processes, without compromising the precise control or quantitative results achieved in comparable macroscale studies.265,266

1.9.1 Droplet-Based Microfluidics

Water-in-oil droplets are essentially a miniature version of the classical test tube, in which every monodisperse droplet represents a distinct compartmentalised assay. Figure 1.18 shows two typical droplet making geometries, where carefully controlled flows of immiscible liquids can form droplets in kHz frequencies.266 Physical advantages are manifold, and include high-throughput capability, large sample sizes, accelerated reaction times and reduced cost and reagent consumption – which is particularly advantageous when working with limited materials, such as primary cells, patient samples or precious compounds of short supply.265,266 Microfluidic droplets have nano-, pico- or even femtoliter capacity,267 which can be easily tuned to achieve optimised conditions for a given reaction or encapsulation and analysis of specific entities.268 In addition to these conceptual advantages, there also exists a toolbox of unit operations that can be used to manipulate droplets in a sophisticated, yet highly controllable manner. These range from short term incubation and storage, to fusion, sorting, splitting and detection.266 Through the integration of these operations, and exploiting the physical characteristics of the microsystems, complex biochemical processes can be assessed in a cohesive work flow. Directed evolution,

40 transcriptomics and drug screening represent just a few of the uses for microfluidic droplets, with constant developments in the field continuously highlighting novel applications.269–271

Figure 1.18: Microdroplet making geometries. a) T-junction. Droplets are generated when the perpendicular flow of the dispersed phase (aqueous) is sheared by the continuous phase (oil). b) Flow focusing. Shearing of the dispersed phase (aqueous) with the flow of the continuous phase (oil) from two directions results in the formation of droplets. Adapted from Shembekar et al.272

1.9.2 Microfluidics for the Study of Amyloid Aggregation

Various microfluidic techniques have been employed to provide novel structural or mechanistic insight into the amyloid formation process or have shown potential for diagnostic or drug screening approaches. Such methods are coupled with a variety of online characterisation techniques, which typically involve fluorescence readings,273–275 but have also made use of less conventional protocols.276,277 A selection of examples are given below.

Bulk ThT fluorescence assays are limited by potential dye interference in the in-situ format and by the laborious nature and requirement of large sample volumes of the conventional ex-situ format. To circumvent these issues, a continuous and automated microfluidic device has been implemented for ThT monitoring, providing improved time resolution and greatly reduced reagent consumption.273 A droplet based format has been employed to study the consequences of interfacial effects on Aβ40 aggregation. The presence of an air-, solid- or oil- interface results in an acceleration in the rate of amyloid formation, which becomes even more pronounced in the microdroplet format, due to the increased surface to volume ratio. It was found, however, that the use of the fluorinated surfactants can completely abrogate the surface adsorption and associated accelerating effects, which may prove useful for studying kinetics in an interfacial independent manner.278 In another study, the formation of droplets encapsulated in laterally organised lipid membranes was used to provide early insight into how the peptide associates with heterogeneous biomimetic membranes.279

1.9.3 Compound Screening using Microfluidic Technology

Given the aforementioned advantages, drug screening in the microfluidic format represents a promising economical and efficient method for interrogating large compound libraries. There are still, however, certain limitations that must be overcome before such techniques can employed for large scale screening.266,272 Generating high numbers of microdroplets containing individual distinct chemicals remains challenging, as it involves changing the composition of the aqueous phase for each compound.280 This can be achieved in parallel or sequentially, and autosamplers and liquid handling robots have been developed to produce these on-demand droplets.267,280 At this point, the systems reported provide relatively limited throughput.272 Additionally, there must be methods put in place to identify active compounds from droplet populations. Sequential droplet formation allows characterisation of a droplets composition based on its position within a droplet train.267 Alternatively, barcoding can be performed to include a distinct identifier with each compound. A variety of such indexing techniques, including DNA-primer and optical barcodes, have been employed in different microfluidic applications.281–283 The drug screening microdroplet platform developed by Brouzes et al. successfully incorporates a selection of these important factors, and was used to assess inhibitory activity by quantitatively measuring the viability of U937 cells incubated in droplets with small molecules of interest. Using the system, an

IC50 value for an inhibitory compound was measured by encoding drug dilutions through addition of different concentrations of a fluorescent dye. The data obtained in microdroplets correlated very well with analogous microwell plate experiments. Following qualitative primary screens, to initially identify activity at the target, subsequent dose response analysis can also be performed with microfluidic approaches. This is somewhat easier than the initial screens, as advanced gradient generation techniques have been developed.284,285

1.10 Project Aims and Thesis Structure

The primary objective of this project was to develop a unified screening platform for the identification, validation and development of small molecule inhibitors against the aggregation of

Aβ42. This was achieved through the design of an in vitro to in vivo, fluorescence lifetime-based aggregation sensor, which was employed to screen an array of compound libraries developed by the Spring research group.

In Chapter 2, the development of an ‘agarose cage’ microdroplet assay is discussed. Through such work, a microfluidic screening platform was optimised to a stage where the aggregation of Aβ42, in the presence of exogenous compounds, could be monitored. The throughput and sensitivity of the assay was, however, limited and it was not believed to be sufficiently advantageous over current techniques for large scale screening. Early FLIM investigations with the agarose droplets were used to guide the design of an alternative lifetime sensor assay.

Chapter 3 outlines the development of the ‘nanoFLIM’ assay platform, a medium-throughput system, employing microfluidic techniques and fluorescence lifetime imaging to monitor Aβ42 aggregation with high sensitivity and sampling size. The capabilities of the assay were tested and the system validated with a selection of known inhibitors.

In Chapter 4, a selection of novel chemical libraries (that were developed using diversity-oriented synthesis or rationally targeted drug discovery strategies) were screened using the nanoFLIM. A total of 445 compounds were screened and variety of structurally distinct compounds were shown to display inhibitory activity. One library of particular interest was focused on and the anti- aggregation activity of a small selection of constituent compounds studied in detail. Through such work a lead compound, MJ040, was identified and validated using a host of biophysical techniques.

Chapter 5 outlines the work performed to adapt the Aβ42 aggregation lifetime sensor for application in live cells and in disease model C. elegans. In this work, the optimised lead compound MJ040X was further validated and shown to inhibit Aβ aggregation in both cellular studies and in the whole organism disease model

Chapter 6 gives an account of the screening campaign carried out using conventional ThT fluorescence assays. Three libraries of interest were identified and are discussed. A library of heterocyclic chalcone derivatives was synthetically expanded to generate structure-activity relationship data, and resulted in the development of another hit compound, SCO17.

In Chapter 7, efforts to screen the inhibitory activity of the hit compounds in Drosophila melanogaster AD models are discussed. Three experimental formats were employed to probe if the compounds could rescue flies from Aβ aggregation induced pathogenicity, however no rescuing effects were observed. The advantages and limitations of the assay formats and result obtained are reviewed.

Chapter 8 provides a summary of work performed and the conclusions drawn. An outlook is given for possible future directions of the work and implications thereof.

Chapter 2 The Agarose Cage Microdroplet Assay

2.1 Introduction

The field of microfluidics has drastically grown in popularity in recent years, with the associated miniaturisation allowing for high resolution and sensitivity in a variety of analytical techniques.264,286 The assay developed in this project was designed to take advantage of the benefits associated with the use of microdroplets, by using this format to monitor the aggregation of Aβ42 and subsequently investigate the effects that small molecules can have on this process.265,266 An overview of the advantages associated with the use of microfluidic techniques, and their relevance to the study of amyloid aggregation in this work, is given in table 2.1.286

Table 2.1: Overview of the advantages of microdroplet techniques and their relevance for the study of Aβ42 aggregation.

Advantage of microfluidic Relevant to Comment techniques Aβ42 studies

Low sample volume Reduced quantities of expensive and laboriously-  requirements prepared Aβ42 required.

High number of repeats per Avoids prevailing issue of irreproducibility in monitoring  sample Aβ42 aggregation.

Improved analytical performance when Use of flow cytometry allows for the measurement of  compared to macroscale the Aβ42 aggregation state on the millisecond scale analogues

Reduced instrumental  No specific advantage in this work footprints/low unit cost

Potential to exploit shorter heat and mass transfer Shorter reaction times  times due to high surface-to-volume ratios at the microscale

Facile integration of  Not exploited in this work functional components

High-throughput  Not exploited in this work experimentation

Mimic conditions similar to  Not exploited in this work that of a single cell

Compartmentalizing Isolating peptide species within individual microvessels  biological reactions to probe early steps in aggregation process

This assay strategy employed centres on the use of agarose microdroplets, which can physically trap developing protein aggregates.287 A schematic of the process is shown in figure 2.1. Partially labelled Aβ42 monomers are incubated in low percentage molten agarose microdroplets under conditions conducive to amyloid aggregation. Upon cooling the low melting point liquid agarose solidifies and acts as hydrogel mesh cage with pores of defined size.288 This mesh can selectively retain aggregates bigger than a certain threshold, but allows for the removal of unbound monomeric species. Using flow cytometry (FC) and fluorescence lifetime imaging microscopy (FLIM) the resulting beads can be analysed, with the quantity or fluorescence lifetime of the retained fluorescent structures providing insight into the extent of peptide aggregation that has occurred.

Figure 2.1: Schematic representation of the proposed agarose microdroplet assay. i) Partially labelled Aβ42 monomers are encapsulated in a low percentage liquid agarose microdroplet. ii) Incubation at 37 °C allows Aβ42 aggregation to occur. iii) Cooling causes the agarose to solidify resulting in the formation of bead ‘cages’ with defined pore sizes. Monomers and aggregated Aβ42 species smaller than the agarose pore size are washed away through the porous matrix, however larger oligomeric or fibrillar structures are retained. iv) The contents of the resulting agarose beads can be optically interrogated using flow cytometry and fluorescent lifetime imaging microscopy to inform on the quantity of the aggregation state of the peptide within.

It was initially proposed that the microdroplet based assay would display a variety of advantages over currently used techniques. First, the assay avoids the use of ThT and other external probe dyes, therefore circumventing the previously mentioned related issues (Section 1.5.2).121. The

46 selective retention of the aggregates in the agarose is a crucial factor, as this allows the removal of unbound monomers. Without this feature, there would be nothing to distinguish the difference in fluorescence intensity of the same concentration of free monomeric or aggregate incorporated dye labels and consequently a distribution over time would not be achieved. Additionally, the pore size can be reproducibly controlled by varying the percentage of agarose, giving a readily tuneable parameter for more detailed investigations into the effects that any compounds of interest may have.288,289

Combining the hydrogel mesh structure with the microdroplet format avoids issues associated with weak diffuse bulk fluorescent signals and also allows the use of flow cytometry (FC) analysis. This laser based technique provides an efficient means to screen huge droplet populations according to a variety of parameters, in a high-throughput manner.290 Finally, it was proposed that

FLIM could also be utilised to monitor the Aβ42 self-assembly process, by means of an amyloid aggregation lifetime sensor, representing a huge benefit of this technique.255,258 FLIM is a non- invasive technique, and therefore can be employed to monitor aggregation events in live cells and in vivo.252 It was envisaged that this could allow for a direct comparison of the aggregation kinetics observed in the agarose droplets with those seen dynamically in live cells. This option is not possible by any other method to date, due to the general requirement of cell fixation for immunological or dye staining, and other dye specificity and background fluorescence related issues (discussed in Section 5.1).

2.2 Primary Objectives

The aim of this project was to develop a novel agarose cage microdroplet assay to screen compound libraries for inhibitory activity against the aggregation of Aβ42. Two assay protocols were developed, in which microdroplet production of a Aβ42 and agarose solution was performed either before or after incubation under conditions conducive to peptide aggregation. The associated kinetic profiles of aggregation of each format was measured and the applicability of monitoring the modulatory effects of small molecule inhibitors was investigated. The use of the agarose microdroplet format for monitoring the aggregation of Tau was also investigated, due to the increasing interest in Tau aggregation inhibitors for potential neurodegenerative disease therapeutics.26

2.3 Results and Discussion 2.3.1 Overview

The work carried out in this project can be separated into several distinct stages. Establishment and optimisation of the general microdroplet assay protocol was first performed. This included systematic modifications of the basic assay parameters until reproducible Aβ42 aggregation profiles were observed. Through such work two assay protocols were identified, the droplet and the bulk methods. Once this was accomplished, the assay was validated by testing the effects of known inhibitory molecules. The results here were compared to those achieved in a conventional ThT fluorescence assay, highlighting the benefits of the detection methods of this novel assay.

Given initial promising results attained when monitoring the self-assembly of Aβ42, work investigating the aggregation of Tau using the same format was initiated. In parallel with both Aβ42 and Tau assay optimisation and validation stages, FLIM studies were undertaken. Initial issues experienced have been addressed and the assay format has been developed to a stage where reliable aggregation kinetics can be observed. Early investigations using the fluorescence lifetime sensor provided promising results, which have guided the design of an improved FLIM assay strategy (Chapter 3).

2.3.2 Optimisation of the Microdroplet Cage Assay 2.3.2.1 Outline of the Assay Protocol

A schematic representation of the assay is illustrated in figure 2.1, but a more detailed overview shall be given again for clarity. i) Microdroplets containing Aβ42 monomers partially labelled with

Hilyte™ Fluor 488 labelled (hereafter named Aβ42-488) and low melting liquid agarose are formed and ii) incubated under conditions conducive to amyloid aggregation. iii) After a specified period of time the droplets are cooled on ice, causing the agarose to solidify and form ‘cage-like’ structures of defined pore size. Removal of the oil layer and washing of the beads results in the selective removal of unincorporated monomers, whilst retaining larger fluorescently labelled oligomeric or fibrillar structures. iv) FC can then be used to analyse the quantity of fluorescently labelled structure within the beads, and thereby monitor the extent of peptide aggregation over time (Figure 2.2). Briefly, this is achieved by ‘gating’ a specific region on the forward scatter (FSC) and side scatter (SCC) dot plot (Figure 2.2a), with these parameters corresponding to droplet size and granularity respectively. The average fluorescence intensity of this subgroup can be calculated at each time point (Figure 2.2b, d), and can be directly correlated to the quantity of fluorescent aggregates within the ‘cage’ structure.

Figure 2.2: Interpretation of flow cytometry readout. The forward scatter (FSC) and side scatter (SCC) readout, corresponding to droplet size and granularity respectively, are used to generate a two-dimensional dot plot. A graphical boundary (blue hexagon) known as a ‘gate’ can segregate the droplet population of interest. The fluorescence intensity of the beads within this subgroup can be represented as a histogram for subsequent interpretation. a, c) Droplet populations incubated for 0 h or 8 h respectively. b, d) Mean fluorescence intensity (BLuFl1) values of the populations gated in the dot plots. e) An overlay histogram of the 0 h and 8 h samples, showing a distinct separation in fluorescence over time, correlated to the difference in concentration of fluorescently labelled Aβ42 aggregates within the bead populations.

2.3.2.2 Proposed Workflow

The initial experimental protocol followed was defined by a previous student (Dr Jing Yan)291 who worked on the development of this assay. In my hands, and in experiments done together with Dr Liisa van Vliet (Hollfelder group), these results were not reproducible, so it was concluded that the protocol would need to be redeveloped and improved at various steps. Table 2.2 provides a summary of the parameters investigated, the associated issues, and the adjustments made to the assay protocol to remedy these problems.

Table 2.2: Optimisation of the agarose cage droplet assay protocol. The difficulties experienced at different steps in the protocol are listed and briefly outlined. The adjustments made in the assay to address these issues are described.

Step in protocol Issue Remedy

Adapted Aβ42 preparation; Great variability in the length of lag Sequential TFA and i) Peptide preparation phase and aggregation kinetics HFIP treatment, then between technical repeats 30 min centrifugation

Too low – aggregation process very slow and dynamic range too small to reliably monitor aggregation process ii) Peptide concentration Too high – expensive in terms of monetary cost and lab hours required to prepare the peptide 2.5 μM with 10% labelling density - Compromise between rate of Too low – dynamic range too small to aggregation, dynamic range and reliably monitor aggregation process peptide cost

iii) Peptide labelling Too high – perturbs the aggregation density process (steric and electrostatic interference) and expensive (labelled Aβ42 10-fold higher in price than unlabelled Aβ42)

Too high – solidification of agarose 1.5% agarose during droplet making process - Captures a high proportion of iv) Percentage agarose aggregates but stays liquid Too low – pores in agarose cage too during droplet preparation if large to capture aggregates solution is maintained at 37 °C

Variations in droplet size and Agarose and peptide solution concentration of peptide within the v) Peptide and agarose premixed immediately before droplets when the aqueous peptide mixing droplet preparation solution and viscous agarose solution (2-inlet device) are mixed on-chip (3-inlet device)

Droplet with different diameters Too small – very slow aggregation (Ø) were used for the droplet kinetics, heterogeneity in size due to and bulk preparation methods increased occurrence of junction (discussed in Section 2.3.2.2) blockage in the microfluidic device v) Droplet size Droplets: Ø = 80 μm for Too large – suboptimal FC analysis, due increased aggregation rate and to reduced suction and detection rates, dynamic range giving smaller test populations Bulk: Ø = 50 μm for optimised FC analysis

Step in protocol Issue Remedy

Merging and/or droplet shrinkage due Humid air-tight container to dehydration observed overtime Gentle shaking (350 rpm)

vi) Incubation - Changes in FC fluorescence intensity - Droplet populations remain reading with changes in droplet size monodisperse for at least 12 h during incubation of incubation

Room temperature collection Room temperature – differences in the

aggregation state of Aβ42 in first and - Increased flow rates to last droplets accelerate droplet formation

vii) Droplet collection - Multiple flow focusing devices On ice – causes agarose droplets to used simultaneously for large solidify. Incomplete re-melting with samples of droplets (facilitated incubation at 37 °C restricts movement with use of the multi-syringe and reduces aggregation rates. neMESYS pump)

Centrifugation, pipetting or vortexing Centrifugation – 5 min, 7.4 rpm

Too vigorous – aggregates broken to - Expels monomers, but a high form structures that are smaller than viii) Bead washing concentration of aggregated pore size species remains

- Enhances consistency between Too gentle – monomeric peptide not samples and technical repeats expelled from the agarose cages

Finding the optimal conditions for droplet incubation proved to be one of the most problematic aspects faced during assay development. Issues arose in attempts to keep the droplets monodisperse over the long time course needed to monitor the aggregation process. Initially, nothing was added to the droplet/oil mixtures, and the samples were just incubated in an Eppendorf bench top incubator shaker (37 °C). FC analysis showed an increase in fluorescence over time, indicating increased aggregation within the droplets. Bright field images of the beads at the various time points revealed, however, that this incubation method resulted in droplet shrinkage and coagulation after prolonged periods (Figure 2.3a, b). Further control experiments using the large fluorescently labelled polymer FITC-dextran showed an increase in fluorescence intensity per unit volume with smaller droplets (Figure 2.3c), questioning whether the previous increase in florescence observed was actually correlated to increased peptide aggregation.

Figure 2.3: Variation in droplet size upon incubation and associated differences in fluorescence intensity. a) A monodisperse population of droplets are produced using strictly controlled flow rates. Ø = 50 μm b) After 6 hours of incubation the droplet population is no longer homogeneous, due to droplet shrinking and merging. This population cannot be accurately analysed using flow cytometry, as the droplets are not all contained within the same gating region. c) Fluorescence intensity per unit volume of FITC-dextran within differently sized agarose droplets. An increase in intensity is observed with decreasing droplet size, indicating that droplet dehydration and shrinkage affects the fluorescence intensity of the agarose droplets during the incubation process.

Various incubation methods were tested to overcome the issues of droplet dehydration and shrinkage. It was concluded that incubating the tightly sealed tubes in very high humidity conditions, with controlled gentle shaking, was the best method to prevent droplet dehydration. The size and homogeneous nature of these droplets were maintained for at least 12 h and any changes in fluorescence identified by FC analysis could therefore be directly attributed to changes in the quantity of fluorescent structures within the bead.

2.3.2.2 Droplet and Bulk Preparation

While the issues associated with the incubation methods were being addressed, an alternative method of droplet preparation was used to test other aspects of the system. These methods will henceforth be referred to as the initial ‘Droplet’ and alternative ‘Bulk’ preparations (Figure 2.4).

In the previously outlined droplet preparation method, the agarose and peptide solutions are mixed to give the 1.5% w/v agarose and 2.5 µM Aβ42-488 (10% labelled) solution. This sample is taken into the syringe and the droplets formed immediately. The population of droplets are split into a number of aliquots and are incubated. At given time points the aliquots are cooled on ice to stop the aggregation process, then demulsified (moved from the oil solution into an aqueous one) and analysed (Figure 2.4a). In the bulk preparation the agarose and peptide solutions are mixed to give the same starting solution. In this method, however, the mixture is incubated in bulk and droplets are formed at each specified time point. The droplets are then cooled and demulsified directly. This method avoids the issues of droplet shrinkage or merging associated with the incubation process (Figure 2.4b).

Figure 2.4: Methods of agarose droplet preparation. a) Droplet preparation: The agarose and peptide solutions are mixed and the droplets formed immediately. The sample is aliquoted and incubated as populations of droplets maintained on an aqueous layer above the oil. At a given time point an aliquot sample is cooled, demulsified and washed prior to FC analysis. b) Bulk preparation: The agarose and peptide solutions are mixed, but the sample is incubated in bulk. Droplets are formed at specified time points and are cooled, demulsified and washed straight away.

A driving factor for the development of this microdroplet based assay was the belief that the encapsulation of the peptide into a micrometre-scale droplet would result in an accelerated aggregation process relative to that experienced in bulk solution. Increased self-assembly kinetics were postulated to arise from higher local concentrations within the droplet, as a result of the high degree of surface adsorption to the droplet boundary and increased surface to volume ratio compared to that experienced in bulk. Development of the ‘Droplet’ and ‘Bulk’ preparation methods have disproved this belief, with the observation that agarose promoted Aβ42-488 aggregation is slower in the droplet format to that detected in bulk solution (Figure 2.5).

600 Bulk Droplet 400

200

Fluorescence(a.u.) 0 0 5 10 15 20 25 Time (h)

Figure 2.5: Aβ42 aggregation kinetics with droplet and bulk preparation protocols. Comparison of the increase in fluorescence intensity (corresponding to increased aggregation) achieved if the peptide is incubated in a bulk agarose solution or in droplets. Accelerated initial rates of aggregation are observed in the bulk format. Data acquired using FC. Plot shows mean ± SEM, n = 3. Conditions: 2.5 µM Aβ42-488, ø = 50 μm.

Primary nucleation of Aβ42 is a stochastic event, representing one of the principle factors

292 determining the length of the lag phase observed in the typical Aβ42 aggregation profile. Once sufficient nucleated structures have assembled an aggregation cascade ensues, with subsequent rapid propagation of aggregation processes. The kinetics observed between various bulk phase samples can be expected to follow similar profiles, with the individual primary nucleation events quickly overwhelmed by secondary processes.75 In the droplet format, however, the stochastic primary nucleation events are limited to individual droplets with the subsequent propagating cascade confined within the microvessel boundary. Consequently, it can be expected that whilst

54 individual droplets may show an overall accelerated kinetic profile than that observed in the bulk, the total population of microdroplets will follow a slower process. This is dictated primarily by the time required for the stochastic early nucleation events to occur, rather than the accelerated elongation methods. This interpretation is supported by a study by Knowles et al. which investigated the spatial propagation of the hormone insulin in microdroplets.293 Here the microfluidic format allowed for the detection of single primary nucleation events, in which a significant dependence of lag time on system size was observed. To explore this relationship within the context of the agarose droplets, the aggregation kinetics of Aβ42-488 within 20, 30, 50 and 80 µm droplets were investigated (Figure 2.6).

Figure 2.6: Influence of droplet size on the rate of Aβ42 aggregation a) Comparison of the aggregation kinetics of Aβ42-488 (2.5 µM), in terms of fluorescence intensity, with the use of the droplet preparation, if the peptide is incubated in 20, 30, 50 or 80 µm droplets. An enhancement in initial aggregation rate with increasing droplet size observed. b) Plot of initial reaction rate from panel a) versus droplet volume (pL), fitted with a linear regression.

A significant difference in the initial rates of aggregation was observed between the droplet samples, with aggregation in the 80 µm droplets following a greatly accelerated profile relative to that of the smaller samples. Figure 2.7b shows a plot of increasing rate with increasing system volume, indicating a correlation between the length of the lag phase and the system size. To further investigate the low rate of droplet aggregation in relation to its dependence on the stochastic generation of nuclei, unlabelled fibrillar seeds were introduced into the starting sample (Figure 2.7). It is known that amyloid seeds can act as nuclei to speed up the initial rates of amyloid aggregation.294 It was postulated that if the delayed aggregation profile observed in the droplet format could be accelerated with the addition of these artificial nuclei, then the slow kinetics previously observed could be attributed to the long lag time in the miniaturised system. It was found that the addition of 1% amyloid seeds does increase the speed and extent of aggregation,

55 indicating that the there is a correlation between the length of the lag phase and the volume of the system size.

500 A 42-488 A -488 400 42 + 1% A 42 seeds

300

200

100

Fluorescence(a.u) 0 0 2 4 6 8 10 Time (h)

Figure 2.7: Aβ42 aggregation kinetics with the addition of Aβ42 seeds in the droplet format. Addition of 1% unlabelled fibrillar seeds accelerates the initial rates of aggregation and overall extent of aggregation than in the droplet format. This supports the belief that the there is a correlation between the length of the lag phase and the volume of the system size. Conditions: 2.5 µM Aβ42-488, 25 nM Aβ42 seeds (0.25 μM total Aβ42, estimated 10 monomers per seed, ø = 50 μm.

Together these observations suggest a possible application of the droplet system, which is otherwise restricted by the lack of sensitivity in ensemble measurements.101 Aggregation experiments in the droplet format may represent an effective method to investigate individual nucleation events, where the ability of small molecules to act at this step in the aggregation pathway could be tested. Alternatively, the results from the seeded experiment highlight how this format could be used for detecting the presence of very low concentrations of amyloid seed in a biological sample. Due to the low concentration of soluble aggregated Aβ42 species in AD patient’s blood and cerebrospinal fluid, the development of diagnostic tests that can detect these disease biomarkers has been limited.295 Monitoring minority aggregation events in enclosed droplets, physically separate from the majority of negligibly slow aggregating peptide, could provide a digital-like effect in detecting seeds. Observation of increased rates of aggregation in just a few individual droplets containing the biological sample would be indicative of the presence of the

Aβ42 seed biomarkers. This potential application is currently being researched by other members of the Hollfelder group.

2.3.2.3 Validation of Agarose-Induced Aggregation

As a consequence of the increased rates of aggregation, subsequent assay optimisation and validation work was carried out using the bulk preparation method. A final concentration of 2.5

µM Aβ42-488, with 10% labelling, was used as this gives a suitable dynamic range, but is also efficient in terms of cost and the quantity of peptide and test molecule needed (Figure 2.8). Using a total peptide concentration that is as low as possible is important for investigating processes comparable to those experienced in vivo, where typically the peptide exists in nanomolar quantities.296 This low concentration requirement represents an advantage over currently used screening techniques employing synthetic Aβ42, where at least 10 μM is generally necessary.

Figure 2.8: Aβ42 aggregation kinetics with increasing peptide concentrations in the bulk format. a) Kinetic profiles of Aβ42 aggregation using the bulk preparation method. Increasing peptide concentrations provides a higher dynamic range. 2.5 µM Aβ42-488 was used for further studies as it gives an acceptable range whilst minimising the quantity of peptide and test molecule needed. b) FC data showing increase in fluorescence density observed after 3 h bulk incubations of 2.5 µM, 5 µM or 10 µM Aβ42-488 respectively. Conditions: 1, 2.5, 5, 10 µM Aβ42-488, ø = 50 μm.

The effect of adding Aβ42 amyloid seeds to the bulk preparation method was also investigated, to ensure that a typical accelerated response indicative of normal aggregation process was observed.294 The introduction of seeds was shown to increase the early aggregation rates in the agarose solution in a comparable manner to that observed in conventional buffer solutions or in the droplet method, albeit to a lesser extent (Figure 2.9).

3000 A 42-488

A 42-488 + 1% A seeds 2000 42

1000

Fluorescence(a.u) 0 0 1 2 3 4 5 6 7 Time (h)

Figure 2.9: Aβ42 aggregation kinetics with the addition of Aβ42 seeds in the bulk format. Addition of 1% unlabelled seeds results in an acceleration of the initial aggregation rate. Plot shows mean ± SEM, n = 3. Conditions: 2.5 µM Aβ42-488, 0.25 µM Aβ42 seeds, ø = 50 μm.

To determine if the incubation in agarose truly gives rise to biologically relevant fibrillar structures, the aggregation process in agarose was also monitored using conventional ThT fluorescence analysis. The characteristic enhancement of ThT fluorescence signal upon peptide aggregation was observed in the agarose solution, indicating normal ThT sensitive amyloid species are formed under these conditions (Figure 2.10). The initial rate of the process appears accelerated compared to that observed without the presence of agarose. This was expected as a result of the effects of volume exclusion and macromolecular crowding in the hydrogel format.297 The dispersity of the read-out is attributed to light-scattering effects with the use of agarose.

100 A 42 A 42 + agarose

0 0 4 8 12 16 Normalised Fluorescence Normalised (%) Time (h)

Figure 2.10: ThT fluorescence assay monitoring Aβ42 aggregation in an agarose solution. Aβ42 aggregation profiles obtained when the peptide is incubated with or without agarose. An accelerated initial rate is observed in agarose, potentially due to the effects of macromolecular crowding. Light scattering effects of the agarose is believed to be responsible for the disperse readout. Conditions: 10 µM Aβ42, 20 µM ThT, 1.5% w/v agarose.

The reproducible aggregation profiles obtained under various conditions and the promising results observed using comparative solution phase experiments, suggested that agarose was a suitable medium for monitoring the aggregation process. Consequently, this prompted the initiation of small molecule screening efforts to further validate the assay format.

2.3.3 Small Molecule Screening with the Microdroplet Cage Assay 2.3.3.1 (-)-Epigallocatechin-3-gallate

Initial work with known inhibitory small molecules focused on the use of the green tea extract EGCG a compound which has previously reached Phase III clinical trials for its anti-aggregation effects.174,298 Incubation with this polyphenol at a 1:10 peptide: inhibitor molar ratio showed complete inhibition of aggregation, in both the droplet and bulk method (Figure 2.11). It is worth nothing that droplets of different diameters are used in the two assay formats. 80 μm droplets were used in the droplet preparation, to take advantage of the higher aggregation kinetics and larger dynamic range with this size. 50 μm droplets were used in the bulk preparation, as these are best suited to the available flow cytometer and generate a higher sample size from the same quantity of stock solution.

Figure 2.11: Inhibitory effects of EGCG on Aβ42 aggregation. a) Bulk preparation, ø = 50 μm. b) Droplet preparation, ø = 80 μm. In both formats, the addition of EGCC drastically inhibits the aggregation process. Plots show mean ± SEM, n = 3, measured using FC. Conditions: 2.5 µM Aβ42-488, 25 µM EGCG.

To test the scope of the assay system, the bulk droplet platform was tested with a variety of concentrations of EGCG (Figure 2.12a). The aggregation profiles show the expected behaviour

59 upon changing total inhibitor concentration. The initial rates of peptide aggregation were used to generate an IC50 value – that is the concentration needed to inhibit the aggregation process by

50%. Figure 2.12b shows IC50 curve, from which an IC50 value of 0.16 ± 0.1 µM was obtained (Figure 2.12b).

Figure 2.12: Inhibitory effect of an EGCG concentration gradient on Aβ42 aggregation. a) Aggregation profiles of Aβ42-488 with a concentration gradient of EGCG, acquired using FC. b) EGCG IC50 curve. Curve plotted using the initial rate of aggregation at each EGCG concentration from panel a). An IC50 value of 0.16 ± 0.1 µM was extracted. Conditions: 2.5 µM Aβ42-488, 1.5% w/v agarose, bulk preparation, ø = 50 μm. IC50 푦 1 values were calculated on GraphPad® using the equation: = 푥 푦푚푎푥 1+ 퐼퐶50

2.3.3.2 Comparison of Microdroplet and ThT Assay Results

To investigate how comparable the results obtained with the microdroplet assay are with those acquired using the conventional ThT fluorescence assay, three known inhibitors were tested in each assay format (Figure 2.13). The overall Aβ42 aggregation profile was similar in both, as were the results for EGCG and curcumin. Differences arose, however, in the results obtained with the inhibitor morin. This polyphenol was shown to completely inhibit the formation of ThT sensitive aggregated structures at a 10:1 inhibitor: peptide molar ratio. In contrast, only partial inhibition was observed in the microdroplet assay. The method by which morin exerts its inhibitory action may be used to explain the differences in these results. A study by Lemkul et al. has revealed that morin redirects the normal aggregation process pathway, producing off-pathway structures.188 Whilst these structures are not detected with ThT, it is believed that they can grow large enough to be captured in the droplet cages. This gives an early indication of the potential applicability of the microdroplet assay, providing a secondary means to test the activity of potential aggregation inhibitors.

Figure 2.13: Comparison of inhibitor screening in ThT fluorescence and microdroplet cage assay.

Aggregation profiles in the presence of the small molecules are shown relative to that of Aβ42 alone, which is set at 100%. A difference in the inhibitory activity of morin is identified between the screening methods. Complete inhibition is indicated in the ThT assay, whereas only partial aggregation is observed in the microdroplet format. This highlights how the different assays are capable of detecting aggregation modifiers that function by different mechanisms. Conditions: ThT Assay – 10 µM Aβ42, 20 µM ThT, 100 µM inhibitor.

Microdroplet Assay – 2.5 µM Aβ42-488, 25 µM inhibitor, ø = 50 μm.

Following a long period of optimisation, the assay is now believed to be ready for use. The droplet method provides a means to probe early nucleation events and the specific effects that small molecule modifiers have on this stage of the aggregation process. The bulk preparation method used is more labour intensive than the original droplet method, limiting the assay throughput. However, this preparation has been shown to generate reproducible aggregation profiles, detect the inhibitory effects of small molecules and allow for the calculation of IC50 values.

2.3.4 Tau Investigations

Following the initial optimisation of the cage microdroplet assay with Aβ42, its applicability for probing the aggregation of Tau was next investigated. Unlike membrane derived Aβ42, which can form fibrils without any external additives, cytosolic Tau cannot readily aggregate in vitro without the addition of anionic agents.299 Consequently, most Tau chemical screens include the use of heparin to initiate aggregation.300 The effects observed in such studies, however, may not give a true representation of how the small molecules can modify the actual processes in vivo, as recent work has shown that heparin can act a structural component of amyloid growing fibril, rather than just an aggregation inducer.301 It was postulated that if the addition of agarose resulted in a sufficiently increased aggregation rate, this could avoid the use of external anionic additives, giving it a pronounced advantage over currently used assay techniques.

To investigate the aggregation propensity of Tau in the agarose format, the protocols developed for the Aβ42 aggregation studies were followed exactly. For this, a K18 Tau construct 10% labelled with Alexa Fluor® 488 (hereafter referred to as K18-488) was incubated in a 1.5% w/v agarose solution, employing either the droplet or bulk preparation methods. Gratifyingly, an increase in the quantity of fluorescent aggregates was observed over time (Figure 2.14). Again, an accelerated kinetic profile was observed in the bulk format and such events were observed in samples with as low as 2.5 µM total protein concentration. Most impressively, it was found that no anionic additives were needed to initiate the aggregation process.

9000 Bulk 7500 Droplets

6000

4500

3000

1500

Fluorescence(a.u) 0 0 2 4 6 8 10 Time (h)

Figure 2.14: Tau aggregation kinetics. Kinetic profiles shown for the aggregation process in the droplet and bulk preparation methods. The addition of agarose (1.5% w/v) in these preparations accelerates the aggregation to such an extent that anionic additives are not necessary to initiate the self-assembly events. Conditions: 10 µM K18-488, bulk preparation ø = 50 μm, droplet preparation ø = 80 μm.

The effect of known inhibitory molecules were tested next, again using the bulk preparation method to take advantage of the faster aggregation kinetics and larger dynamic range. After 6 hours of incubation with a 10:1 inhibitor: protein molar ratio, neither EGCG nor curcumin were shown to exert the significant inhibitory effects reported in the literature (data not shown).300,302 The effects of myricetin, an even more potent Tau inhibitor, was next investigated (Figure 2.15).300 At early stages little difference in either the rate of aggregation or the total fluorescent intensity was observed. Whilst the aggregation profiles did begin to separate at later time points, the result at the early stages contrasts with previously reported data, where complete inhibition has been observed.303 These results called into question what aggregation events are actually occurring in the agarose.

15000 2.5 M K18-488 25 M Myricetin

10000

5000

Fluorescence(a.u) 0 0 4 8 12 16 20 24 Time (h)

Figure 2.15: Inhibition of Tau aggregation with myricetin. Previous investigations with this inhibitor show complete aggregation within this time frame. This questions the aggregation process occurring within the agarose. Conditions: 2.5 µM K18-488, 25 µM myricetin, bulk preparation, ø = 50 μm.

To further examine the observed Tau self-assembly processes, the aggregation profile with and without the presence of agarose were investigated spectroscopically. Using a ThT fluorescence assay, it was shown that when K18 is incubated with agarose and heparin, the formation of conventional ThT sensitive Tau aggregates are formed (Figure 2.16). Without the heparin additive in the agarose solution, however, there appears to be no aggregation. A similar result was obtained when testing the effects of agarose on K18 aggregation using the luminescent conjugated oligothiophene (LCO) ligand hFTAA. A reproducible increase in fluorescence, indicative of standard amyloid aggregation, was observed when the Tau fragment is incubated with heparin, or with heparin and agarose. Again, however, no increase in detectable aggregates was observed in the agarose solution without the addition of heparin (Figure 2.17).

K18 100 K18, heparin K18, agarose K18, agarose, heparin K18, agarose, myricetin 50

Normalised

Fluorescence (%) 0 0 2 4 6 8 Time (h)

Figure 2.16: Tau ThT fluorescence aggregation profiles. Aggregation is observed when incubation is carried out with the presence of heparin or alternatively with agarose and heparin. Without the heparin additive no aggregation is observed, either in agarose or in normal buffer solution. Complete inhibition of aggregation is observed with myricetin in the conventional K18, heparin and ThT assay. Conditions: 10 µM K18-488, 20 µM ThT, ~ 40 µM heparin, 1.5% w/v agarose, 50 µM myricetin.

K18 100 K18, heparin K18, agarose K18, heparin, agarose

Normalised

Fluorescence(%) 0 0 2 4 6 8 10 Time (h)

Figure 2.17: Tau LCO fluorescence aggregation profiles. An increase in LCO fluorescence signal (indicative of peptide aggregation) is observed when incubation is carried out with the presence of heparin or alternatively with agarose and heparin. Incubation in agarose without the addition of heparin, does not result in the formation of LCO sensitive aggregates. Conditions: 10 µM K18-488, 0.5 µM hFTAA, 40 µM heparin, 1.5% w/v agarose.

The combined ThT and LCO fluorescence assay results call into question the processes underlying the increase in fluorescence observed in the Tau microdroplet experiments. If the aggregates formed in agarose are not detectable by either amyloid specific ligand, is the aggregation observed representative of the true aggregation process? If not, then the use of such a format would not be applicable for screening attempts to identify inhibitors to target the process in vivo. In an

64 attempt to better understand the morphology of the structures formed in the agarose conditions, the structures were microscopically examined. Such studies were limited by the presence of the agarose. AFM and TEM scanning microscopy techniques, for example, could not be used. Structure illuminated microscopy (SIM) was instead employed to examine the structures formed (Figure 2.18). 20 μm droplets were used, as acquiring images above a sample thickness of 10 μm is challenging with this technique and it was desired to image at least half way through the bead. Only small aggregated structures and no fibrils could be identified following overnight incubation in agarose. This may have been caused in part by the high labelling density. 100% labelled Tau was needed, as the fluorescence readout was not high enough with a lower labelling percentage. We have previously observed this high a density to slow down the aggregation process with Aβ42. To overcome this, longer incubation times were trialled, however this resulted in droplet shrinkage or distortion, so it was difficult to determine if fibrillar structures did eventually form. Due to the variety of issues experienced, no follow up work was carried out with this microscopy technique.

Figure 2.18: Structure illuminated microscopy image of an agarose bead containing aggregated fluorescently labelled Tau. Spots representing small aggregated structures are observed, however no fibrils could be identified. Conditions: 2.5 µM K18-488, bead size = 20 µm, 12 h bulk incubation, conventional 1 µm widefield stack.

As the aggregation events observed within the agarose could not been shown to be a true representation of the self assembly process in vivo, until the morphology of the resulting aggregates can be better understood, this assay format is believed to have limited applicability for screening small molecules for Tau anti-aggregation activity.

2.3.5 FLIM Studies

As the assay protocol was being optimised, the use of fluorescence lifetime imaging for monitoring the aggregation process within the droplets was explored. This technique makes use of the previously discussed fluorescence lifetime sensor (Section 1.8.3.2). Briefly, Aβ42 self-assembly is monitored by changes in the fluorescence lifetime of a covalently attached reporter dye, which decreases upon peptide aggregation as a result of increased self-quenching.255

A variety of practical issues were addressed in designing how the fluorescence lifetime change in the agarose droplets could be monitored overtime. Namely, how best to form, trap and wash the droplets. Initially, after the set incubation time, washed agarose beads were placed in silicon wells on glass coverslips, where they sunk to the bottom of the sample and were imaged directly. Difficulties here stemmed from issues with acquiring data at early time points. For FC analysis, all of the unbound monomeric structures are washed out of the agarose ‘cage’. Whilst this protocol gives ideal FC conditions, fluorescence lifetime imaging of the resulting beads is problematic. Without any monomers there is no early lifetime values within the microdroplet format. More importantly though, by washing the monomers and small oligomeric structures out of the beads at each stage, only low lifetime aggregated structures remain. As fluorescence lifetime imaging is concentration independent, the values obtained for washed beads at each time point will only represent that of the captured aggregates. This is therefore not indicative of the quantity of aggregated species or overall degree of aggregation the peptide has undergone at a given time. This precluded attempts to monitor a characteristic lifetime decrease from the initiation of the aggregation process and to measure the effect that different small molecule inhibitors can have on the aggregation rate. This is displayed in figure 2.19 where the increased concentration of aggregated Tau structures in the droplets over time is not reflected in the fluorescence lifetime values obtained.

Figure 2.19: FLIM analysis of washed beads containing aggregated Tau. a) Average fluorescence lifetime values at specified time points (n = 3). A large initial drop is observed between the K18-488 monomeric value and the fluorescence lifetime when the first aggregates are seen (4 h). A further significant decrease in lifetime is not observed with the appearance of an increasing number of aggregates at later time points. Washing of the beads removes the monomers so the average fluorescence lifetime values observed are not a true representative of the mixture of aggregates and non-aggregated peptide at each time point. Plot shows mean ± SEM, n = 3. b) Confocal images of 4 h and 10 h washed beads. An increase in fluorescent aggregated structures can be observed overtime. c) FLIM image of a 1.5% agarose solution of monomers on a coverslip. Mean fluorescence lifetime 3402 ± 2 ps, n = 3. d, e, f) Fluorescence lifetime images of washed beads following 4, 6 or 10 h incubation. An increase in the number of florescent aggregates is observed, however the higher degree of aggregation is not reflected as statistically significant in the fluorescent lifetime values obtained. 2881 ±59 ps, 2840 ± 58 ps and 2805 ± 35 ps respectively. Conditions: 2.5 µM K18- 488, 1.5% w/v agarose, bulk preparation, ø = 50 μm., scale bar = 20 μm.

In an attempt to overcome these issues an alternative post-incubation protocol was investigated. Here the cooled beads were kept in the oil phase, thereby retaining the unbound monomers, for imaging. As the aqueous droplets float on the fluorinated oil, imaging in the well format was no longer suitable. Instead a method to trap these beads within a microfluidic device was investigated. First, trapping was attempted by using the device shown in figure 2.20 (designed by Hans Kleine- Bruggeney, Hollfelder lab).

Figure 2.20: FLIM analysis of trapped unwashed beads containing aggregated Aβ42. a) A portion of the microfluidic trapping device. The beads flow into the serpentine channel containing evenly distributed traps. Scale bar = 250 μm. b, c) Brightfield images of trapped 80 μm beads, which were formed after bulk incubation of Aβ-488 for 3 h and 24 h respectively. Scale bar = 50 μm. d) FLIM image of a 1.5% agarose solution of monomers on a coverslip. Mean fluorescence lifetime 2909 ± 17 ps, n = 3. Scale bar = 10 μm. e, f) FLIM images of the 3 h and 24 h incubated beads above. Mean lifetime 2814 ps and 1595 ps. Conditions:

2.5 μM Aβ42-488, bulk preparation, ø = 80 μm.

By omitting the washing step and trapping immediately, fluorescence lifetime values at earlier time points could be better obtained. Trapping efficiency using this device, however, was not optimal. Also, longer acquisition times of zoomed in beads were needed. This limited potential assay efficiency, slowing attempts to image many droplets for a single condition. So, whilst these early FLIM images of unwashed trapped agarose beads showed promise, given these issues and uncertainty of the structures observed following the Tau aggregation in agarose, the necessity of agarose in the FLIM setup was re-evaluated.

To explore if aqueous droplets would provide a better result with the lifetime sensor, incubation and subsequent droplet formation in buffer alone was trialled. A variety of trapping devices were tested and figure 2.21 shows an early trapping attempt. The lack of agarose in these aqueous droplets means they are more susceptible to merging and optimisation of the trapping device was clearly required. However, even in this early test, a steadier decrease in fluorescence lifetime with time could be observed, than when using the agarose platform. Furthermore, it was proposed

68 that with improvement in the design of the microfluidic chip, that droplets could be formed and trapped in line. This would provide a means to monitor the full aggregation profile in multiple distinct droplets, and the development of this system will be discussed in Chapter 3.

Figure 2.21: Trapping and FLIM analysis of aqueous Aβ42 droplets. a) Bright field images of trapped 80 μm aqueous Aβ42-488 droplets, which were formed following 0, 2 4 and 24 h incubation in the bulk preparation method, but without agarose. As the droplets do not contain agarose they are more susceptible to merging with overfilling of the device, high injection pressures or just due to close proximity overtime. b) Corresponding FLIM images, with average lifetimes of 3244 ± 4 ps, 2709 ± 50 ps, 2740 ± 4 ps and 2189 ps (Mean ± SEM).

2.4 Summary and Conclusions

This chapter outlines the experimental work performed to develop an ‘agarose cage’ microdroplet based assay for the identification of small molecules capable of disrupting the aggregation of Aβ42. A principle driving force for the development of this assay was the belief that increased aggregation within the miniaturised microdroplet system would result in a high throughput assay system for inhibitor identification. In-depth investigations into various aspects of the procedures, however, have revealed quite the opposite. It has been shown that the reduction in droplet size is associated with a decrease in the rate of aggregation, therefore calling for a complete re- evaluation of the proposed value of the assay format.

The droplet format, whilst potentially telling terms of nucleation kinetics, suffers from issues in droplet storage for the lengths required to monitor the entire aggregation process at low peptide concentrations. Higher concentrations can be employed for investigating individual compounds of interest, but this is not feasible for larger library screens. The bulk format generates reproducible aggregation profiles at low peptide concentrations, can detect the inhibitory effects of small molecules and allow for the calculation of IC50 values. Unfortunately, it is quite laborious and necessitates manual droplet formation over the entire time course of the experiment. This limits the number of compounds that can be screened within a certain period of time, and this loss in the throughput is of great detriment to the overall value of the system.

It was hoped that using the fluorescence lifetime imaging to monitor the aggregation process in the agarose droplets would provide a novel benefit of the system, in terms of adapting the technique to generate comparable aggregation data in live cells. Difficulty in washing, trapping and imaging the droplets over prolonged periods hindered attempts to develop this screening format. Conclusions drawn from such efforts did, however, guide the development of a different fluorescence lifetime sensor technique, using aqueous samples of labelled Aβ42. This was found to be easier to implement and more advantageous and shall be discussed in more detail in Chapter 3.

Through this work, two variations of the agarose cage droplet assay have been developed and can be employed to screen small molecule libraries for inhibitory activity. It is believed, however, that through the optimisation of the system to provide reproducible aggregation kinetics, the assay does not meet the criteria we had initially set for it to provide significant advantage over currently available techniques. As such, further investigations into the agarose assay development and compound screening were not deemed necessary at this point.

Chapter 3 Development of a Fluorescence Lifetime Sensor-Based Microfluidic Assay for Monitoring Amyloid β Aggregation

3.1 Introduction

Time-resolved fluorescence spectroscopy is a powerful analysis tool in fundamental physics as well as in the life sciences. The use of FLIM techniques for monitoring the aggregation of amyloidogenic proteins has greatly grown in popularity in recent times.253,254,304–306 Development of the fluorescence lifetime sensor, in which changes in the fluorescence lifetime of partially labelled peptide are observed upon amyloid formation, has been particularly advantageous. This concentration independent technique is non-invasive, so can be used to dynamically probe the aggregation process in live cells and in vivo disease model organism, as well as in vitro. In this project, the lifetime sensor has been coupled with microfluidic techniques to develop a medium- throughput assay to screen small molecule libraries for inhibitory activity against the aggregation of Aβ42. The lifetime-based assay developed – the nanoFLIM – employs droplet-on-demand technology and a newly developed microfluidic device, in which 110 precisely ordered droplets (fitting in a 1 cm2 chip) can be trapped and imaged concurrently over several hours.267 A general schematic is shown in figure 3.1. Droplets containing fluorescently labelled monomeric peptide and test compounds are formed and filled into a trapping chip in a user defined sequence. Fluorescence lifetime imaging is carried out at specified time points, to report on the aggregation state of the peptide within. The assay format benefits from high sampling size and low reagent consumption, and permits the induction of a shear force on the droplets to agitate the contents in a controlled manner if necessary. Overall, the assay provides an easy-to-implement method to assess the inhibitory activity of small molecule libraries, and shows various advantages over currently used techniques.

Figure 3.1: Schematic illustration of the nanoFLIM assay workflow. Using droplet-on-demand technology (Mitos Dropix/Dolomite267) droplets containing monomeric peptide together with compounds of interest are formed and filled into a droplet array chip in a precisely known order. Fluorescence lifetime imaging is used to monitor the aggregation kinetics of the peptide (which is partially labelled with a reporter dye) and the effect that different extrinsic factors or inhibitory compounds can have on the process. The filled microfluidic chip contains 110 droplets, with 18 nL of aqueous solution per droplet. The fluorescence intensity of monomeric or aggregated peptide droplets are similar. However, the fluorescence lifetime sensor provides a quantitative measure of the aggregation state of the peptide, which are either monomeric (blue, lifetime 3200 ps) or aggregated (yellow, lifetime 2500 ps).

3.2 Primary Objectives

The primary objective of this project was to design and validate a lifetime-sensor based assay incorporating microfluidic techniques, to screen compound libraries for inhibitory activity against the aggregation of Aβ42. A new microfluidic design was implemented, the parameters optimised and the capabilities of the system tested. A selection of known inhibitors was used to validate the potential of the system for detecting small molecule modifiers of Aβ42 aggregation and to probe the strengths of the system compared to the conventional ThT fluorescence assay.

3.3 Results and Discussion

3.3.1 Microfluidic Chip Design

Guided by the results obtained using the amyloid aggregation lifetime sensor with the agarose droplets in Chapter 2, an improved general protocol for droplet preparation was proposed. The strategy involved forming and directly filling peptide containing aqueous droplets into a trapping device. The inclusion of agarose was not deemed necessary in this assay format as monomeric species did not need to be washed out. The first microfluidic chip tested contained channels of 80 μm in width, separated by small pillars (Figure 3.2a). Using precisely controlled flow rates, droplets of uniform size were formed and filled into the device. Over time, however, the droplets began to move and merge within the device as there were no physical boundaries to hold them in place (Figure 3.2b). This prevented time resolved measurements of the aggregation process within equally sized droplets being taken.

Figure 3.2: Formation and trapping of aqueous droplets in-line. a) t = 0 h, carefully controlled flow rates permitted the device to be filled with equally sized aqueous droplets. Using this system only one condition (small molecule, extrinsic factor, etc.) could be tested at once. b) t = 2 h, the lack of physical separation resulted in droplet movement and joining overtime, even under quiescent conditions at 4 °C. Scale = 200 μm.

To circumvent the issues of droplet movement and merging, a new microfluidic chip was designed and implemented. Dr Fabrice Gielen (Hollfelder Group, University of Cambridge) designed this new shearing trapping chip, whereby 110 grids sit along a serpentine channel (Figure 3.3a). Droplets are formed externally and are filled in the chip in a slow flow of oil (Figure 3.3b, c). Due to the lower density of the aqueous droplets relative to the fluorinated oil, the droplets float upwards and are trapped in the first empty grid they pass under. Subsequent droplets pass under

73 the trapped droplets until they reach an empty grid, where they then fill in order. The device can easily be emptied and reused by inverting the chip and flowing the droplets out. No contamination was observed after repeated use.

Figure 3.3: The shearing trapping chip. a) Filled trapping chip. 11 rows of 10 square grids sit upon a serpentine channel, which has an inlet at the bottom left corner and an outlet at the top right. Droplets fill into the grids in the order in which they are formed and flow into the device. Each droplet contains 18 nL of aqueous solution and the flow of oil under the sample can be used to generate a shear force on the trapped droplets, causing their contents to rotate Scale = 500 μm. b) 3D schematic of a profile view of the chip showing the square traps sitting upon the oil channel. Droplets are carried in a flow of oil, which enters the device through tubing that feeds directly into the channel. c) 3D schematic showing chip filling. i) Droplets of a known sequence fill into the serpentine channel. ii) The droplets travel until they reach an empty grid, where they float upwards and are trapped and can be held for several hours for imaging. iii) Subsequent droplets move under the previous trapped droplets until vi) they reach an empty grid and also become trapped.

To generate precisely ordered droplets of equal size and defined composition, droplet-on-demand technology was employed. The Mitos Dropix is a robotic platform for droplet sequence production and sampling, developed by Dr Fabrice Gielen, Dr Liisa van Vliet and Dr Xize Nie (Drop-Tech, Cambridge, UK), and distributed by Dolomite (Royston, UK).267 The system contains 24 bottomless wells partially immersed in a container of oil (carrier phase) (Figure 3.4a). These wells are filled with aqueous samples, and sit on the surface of the oil. Polytetrafluoroethylene (PTFE) tubing runs from a syringe pump through the robot into the oil phase, where it is hooked upwards and aligned under the aqueous sample (Figure 3.4b). Droplets are formed when the tubing is moved between

74 the sample and the carrier phase under negative pressure, as applied by the syringe pump. The size of the droplets is controlled by the residence time in the aqueous sample, and the time spent in the oil dictates the distance between droplets. As the droplets are formed they are pulled through the tubing directly into the microfluidic chip where they are trapped for imaging in the order in which they were formed (see figure 3.1 for full assay overview).

Figure 3.4: The Dropix droplet-on-demand platform. a) Schematic of the Dropix. Aqueous samples containing the peptide and test compounds are filled into bottomless tubes that are partially submerged in oil. Tubing is aligned under the wells and can be raised into the samples to form droplets by the movement of the robot arm, as controlled by the computer software. b) Generation of droplets. Negative pressure is applied by continuous withdrawal using a syringe pump. The droplets are formed when the tubing is moved between the sample and the carrier phase. The length of time the tubing is held in the aqueous or oil phase determines the droplet volume and space between droplets respectively. The tubing can alternate between samples to create a precisely ordered droplet array. In the nanoFLIM the tubing feeds directly into the trapping device, where the user defined droplet sequence is filled in the order in which it was generated. Settings: 18 nL aqueous phase, 32 nL oil phase (0.1% picosurf in HFE7500), 200 μm diameter tubing, 2 μL min-1 flow rate. Adapted from Gielen et al.267

To initially test if peptide aggregation could be monitored within the shearing trapping chip, the peptide sample was spiked with 1% preformed Aβ42 seeds (10% total peptide concentration, assuming 10 monomeric units per seed). These pre-aggregated templates act as nuclei to speed up the process, facilitating efficient assay testing and optimisation. 110 droplets were formed from a 10 μL aqueous peptide sample, exemplifying the vast sampling size that can be generated with very low reagent consumption. A time dependent decrease in fluorescence lifetime was observed, indicative of peptide aggregation (Figure 3.5). It is worth noting that the use of 20% labelling density in this experiment was deemed too low to generate reliable fluorescence lifetime images of all the droplets within the device in one image, so a subset was focused on (n = 30).

Optimisation of the labelling density, chip preparation protocol, and use of a more appropriate objective allowed for imaging of the entire chip at once in later experiments. Also, 15-20 minutes preparation time is typically required to form and fill all 110 droplets, so aggregation profiles miss any measurements before this time. If earlier time points are necessary, fewer droplets can be formed and filled into the device within a few minutes. This was not necessary when un-seeded

Aβ42 samples were used in later experiments, as the lag phase exceeded the preparation and filling time.

Figure 3.5: Monitoring Aβ42 aggregation in trapped droplets using the nanoFLIM. a) Aggregation profile of Aβ42-488 with the addition of 1% seeds. Plot shows mean ± SEM, n = 30. b) FLIM images of select time points of 30 droplets trapped in the device. A time dependent drop in fluorescence lifetime in association with peptide aggregation was observed. Scale = 500 μm. Conditions: 10 μM Aβ42, 20% labelled with Hilyte™ Fluor 488, 1% Aβ42 seeds.

3.3.2 Optimisation and Testing the Capabilities of the NanoFLIM 3.3.2.1 Labelling Density

The first step needed to optimise the assay format for improved fluorescence lifetime measurements was to identify the optimal labelling density. Too high a labelling density has been reported to perturb the aggregation process,307 whereas too low a dye concentration does not provide a high enough photon count to generate representative fluorescence lifetime data.246 To probe this issue, a peptide labelling density gradient of droplets within the microfluidic chip was tested. Figure 3.6a shows the filled device, where each row of droplets contains a different percentage of labelled Aβ42 (10-100%). The change in fluorescence lifetime was measured over time, to give the representative aggregation profiles at each labelling percentage (Figure 3.6b, 10% labelling data omitted due to the low photon count).

Fluorescence self-quenching of the covalently attached reporter dye is the principle mechanism governing the amyloid aggregation lifetime sensor.255 As the peptide aggregates, the dye molecules are brought closer together into a structured array. This increases the local concentration of the fluorophore, thereby inducing self-quenching and giving rise to the observed drop in fluorescence lifetime. The lifetime sensor is highly sensitive to both fibril packing density and dye concentration.255 Increasing the peptide labelling density minimises the distance between adjacent dye molecules in the aggregated structures, resulting in a greater extent of self- quenching and a larger dynamic range in lifetime upon aggregation. Interestingly, above a labelling density of 50%, the difference between the fluorescence lifetime of the monomeric and aggregated species begins to decrease (Figure 3.6b). Very high dye concentrations may prevent the formation of dense aggregates, due to steric and electrostatic interference from the attached fluorophores. The potential increase in dynamic range with increased labelling ratio is therefore restricted by how densely the resulting aggregates can pack, which is in turn limited by the number of dye molecules. These factors balance to give the largest fluorescence lifetime difference between monomeric and aggregated species at 50% labelling. The similarity in final fluorescence lifetime values observed for different percentage labelling above and below 50% labelling, is attributed to a balance between the density of the resulting structure and the number of dye molecules. For example, a more diffuse structure with an 70% ratio could potentially give the same final lifetime as a dense structure with a 40% labelling ratio.

The kinetics of aggregation with increasing labelling densities were examined by normalising the aggregation profiles (Figure 3.6c). Here the data from 90% and 100% labelled peptide have been omitted, as no aggregation was observed in the time measured. The profiles of 30-70% essentially overlap, showing negligible differences in the aggregation rate. The initiation of aggregation with the 20% labelled Aβ42 appeared to be delayed in comparison to the intermediate labelling ratios.

In contrast, 80% labelled Aβ42 was seen to aggregate at a slightly higher rate. It is possible that at 20% labelling, the dye concentration within initial small aggregates is too low to be substantially affected by self-quenching, and the presence of aggregates aren’t detected until they have grown in size and density. At 80% labelling, the high concentration of dye molecules with small aggregates may allow for the detection of very early aggregated species.

50% labelled peptide provided the optimal aggregation profile, where the fluorescence intensity was high enough to generate consistent fluorescence lifetime data and a large dynamic range in

77 lifetime between monomeric and aggregated species was observed. High labelling densities have previously been reported to influence the morphology of resultant amyloid aggregates.255 To ensure that the aggregates formed following incubation with 50% labelled Aβ42 were comparable to those observed with unlabelled peptide, AFM analysis was performed (Figure 3.7). The fibrillary species formed using the partially labelled or unlabelled Aβ42 were believed to be sufficiently similar in structure to validate the use of this labelling density for screening small molecules with the nanoFLIM.

Figure 3.6: Effect of labelling density on Aβ42 aggregation kinetics. a) Labelling density gradient in microfluidic chip. Each row contains 10 droplets of increasingly labelled Aβ42 (10 μM, 10-100% labelled). Scale 500 μm. b) Aggregation profiles of Aβ42 at different labelling densities. 50% labelled shows the largest dynamic range in fluorescence lifetime change. The photon count at 10% is too low to obtain reliable

78 fluorescence lifetime values. c) Normalised aggregation profiles of Aβ42 at different labelling densities. The data from 90% and 100% labelled peptide have been omitted as these showed essentially no aggregation in the time frame monitored. The profiles of 30-70% labelled peptide overlap, showing the aggregation rate is largely unaffected with these changes in labelling ratios. 20% labelled Aβ42 appears to aggregate at a slightly slower rate, whilst 80% labelled Aβ42 aggregates at a slightly higher rate. Plot shows mean ± SEM, n = 10.

Figure 3.7: AFM images showing the effect of labelling density on Aβ42 fibril morphology. a, b) Unlabelled and 50% labelled Aβ42 fibrils, respectively, following 7 days of incubation. Scale bar = 600 nm. c, d) Zoomed in images of unlabelled and labelled Aβ42 showing similarities in fine structure. Given the resemblance, a 50% labelling density was deemed acceptable for screening the compound collections and follow up IC50 calculations. Scale bar = 100 nm. Conditions: 10 μM Aβ42, unlabelled or 50% labelled with Hilyte™ Fluor488 at the N terminal, incubated at 37 °C with shaking (350 rpm) for 7 days.

3.3.2.2 Temperature Dependence

To test the capabilities of the fluorescence lifetime sensor microfluidic device, various parameters were altered to examine if the ensuing aggregation data conformed with the expected results.

The rate of Aβ42 aggregation depends on temperature and an acceleration in fibril formation is generally observed at higher values.308,309 The microfluidic chip is housed within an easily- adjusted, temperature controlled chamber. Reducing the temperature from 37 °C, which was used for all other experiments, to 25 °C resulted in a large drop in the rate of aggregation (Figure 3.8). At the cooler temperature, the length of the lag phase greatly increased and lasted approximately 120 mins, double that of the 60 min observed at 37 °C. Species formed at each temperature setting

79 reach the same final fluorescence lifetime level, indicating that there is no drastic difference in the morphology of the aggregates formed, just the rate in which the aggregation process proceeds. The option to vary temperature with the nanoFLIM may be of use in probing the aggregation kinetics of amyloidogenic peptides that aggregate either very quickly or very slowly.

4000 25 C 37 C 3500

3000

2500

2000

Fluorescence(ps) Lifetime 0 100 200 300 Time (min)

Figure 3.8: Effect of temperature on the Aβ42 aggregation profile. Reducing the temperature slows down the rate of aggregation, with a greatly lengthened lag time observed at 25 °C. This may prove useful when monitoring the aggregation kinetics of very fast aggregating peptides. Conditions: 10 μM Aβ42, 50% labelled. Plot show mean ± SEM, n = 18.

3.3.2.3 pH Dependence

310–313 It is well reported that the process of Aβ42 self-assembly is highly sensitive to changes in pH. There is, however, still some debate over the morphology, physical properties and function of the aggregates formed. To assess the ability of the nanoFLIM to monitor Aβ42 aggregation kinetics under different pH conditions, the trapping device was filled with a pH gradient of droplets (pH 5- 9). Differences in the rates of aggregation and in the final fluorescence lifetime value were observed (Figure 3.9a). Normalising the aggregation curves provided a means to investigate the rate of Aβ42 aggregation under different pH conditions (Figure 3.9b). The rate of aggregation is most accelerated at pH 6 – a value close to the peptides isoelectric point of pH 5.5,30 thus suggesting an important role of charge variation in the Aβ42 self-assembly process. This result

313 agrees with a previous study by Su and Chang, in which Aβ42 was observed to aggregate fastest at pH 5-6. In this range, large and complex fibrils were formed within minutes. Similar rates for the self-assembly process were observed at pH 7 and 8. At pH 9 no aggregation was seen. It is worth noting that the final fluorescence lifetime reached with each pH varies slightly. The changes in the morphology of the Aβ42 structures formed under the different conditions can be used to rationalise this observation. As discussed, the fluorescence lifetime of the reporter dye molecules is related to the overall aggregation state of the peptide, with higher density structures resulting in increased quenching.255 It has been suggested that low pH (pH 2-3) induces oligomerisation of

80 the peptide, whereas more neutral conditions (pH 7.4) are conducive to fibril formation.44 Such species would differ in β-sheet content and overall architecture. The reporter dye would therefore encounter varying levels of quenching when in different structures, giving rise to the observed changes in the final fluorescence lifetime. Formation of the complex and densely packed structures at pH 5-6, reported by Su and Chang,313 may explain the especially low final fluorescence lifetime observed with the nanoFLIM at pH 6.

Figure 3.9: Effect of pH on the Aβ42 aggregation profile monitored using the nanoFLIM. a) A pH gradient (pH 5-9) of droplets containing Aβ42 were filled into the microfluidic chip and the aggregation monitored over time. No aggregation was observed at pH 9. The level at which the fluorescence lifetime plateaus varies slightly between conditions, suggesting the formation of morphologically distinct species at different pH values. b) Normalised aggregation profiles highlighting the slight differences in aggregation rates of the peptide at different pH values. Plots show mean ± SEM, n = 9. Conditions: 20 μM Aβ42, 50% labelled, TRIS buffer acidified with HCl.

The effects of pH change on the aggregation profiles observed using the lifetime sensor were compared to those acquired using a conventional ThT fluorescence assay (Figure 3.10a). To rule out pH dependent changes in the spectroscopic properties or binding ability of ThT, a control experiment was performed using fibrils performed at pH 7.4 and re-suspended in buffers at pH 4, 5, 6, 7, 8, and 9 (Figure 3.10b). Only at pH 9 was a significant change in fluorescence intensity observed with the same fibrils, suggesting altered ThT binding efficiency at basic pH, a result which has been observed previously.314 This indicated that under all the other pH conditions measured, observed changes in the ThT fluorescence kinetic profiles were due to changes in the aggregation process in terms of reaction rate or species formed, not due to the spectroscopic properties of the dye. Again, no aggregation was observed at pH 9. At pH 5 and 6, no lag phase was observed and a relatively fast rate of self-assembly was seen under both conditions. Aggregation occurred more slowly at pH 7 and 8, with a lag phase of 3-4 h. The variations in aggregation rates observed were all generally in agreement with those seen with the nanoFLIM.

Figure 3.10: Effect of pH on the Aβ42 aggregation kinetics monitored by ThT fluorescence. a) ThT fluorescence assay employing different pH buffers. The peptide aggregates at similar rates at pH 5 and 6, and at pH 7 and 8. No aggregation is observed at pH 9. b) ThT fluorescence intensity of preformed aggregates (formed at pH 7.4) suspended in various pH buffers. The fluorescence intensity is constant at pH 4-8, but a significantly increased intensity is observed at pH 9. Conditions: a) 10 μM Aβ42, 20 μM ThT, n = 3. b) 10 μM Aβ42 fibrils (total monomer concentration), 20 μM ThT, 440 ex, 488 em, n = 5. Statistical analysis performed using a one-way ANOVA with Dunnett’s multiple comparisons post-test, using pH 7.4 as the column control; ***p<0.001.

Extracting information about the morphology of the species formed under different conditions is difficult using the ThT fluorescence assay. The plateau of the ThT fluorescence curve provides a rough assessment of the fibril mass content following an aggregation experiment.315 A similar concentration of ThT sensitive aggregated species was observed at pH 5 and 6 (Figure 3.10). This was lower than that observed at pH 7 and 8. In this assay format it is difficult to deduce if this lower plateau is indicative of a reduced concentration of aggregates, or the formation of structures which are not as sensitive to ThT binding. In this respect, the fluorescence lifetime sensor provides added benefit compared to the ThT assay, as it provides insight into morphological changes of final species under different condition, as well as informing on the aggregation rate.

3.3.2.4 Shear Force Effects

Mechanical agitation is commonly employed in assays monitoring Aβ42 self-assembly, both to speed up the aggregation process and to provide homogeneous solutions for acquiring reproducible kinetic data. A novel attribute of the microfluidic assay system is the ability to induce a shear force in the device. As the oil flows under the trapped droplets, a shear stress develops at the intersection between the two immiscible fluid layers. This causes the droplets to rotate in an even and reproducible manner, but remain trapped in their individual grids for continuous monitoring (Figure 3.11a). This is not believed to be possible with the use of any other microfluidic trapping device, where a continuous oil flow would most likely result in droplet movement or

82 merging, or cannot be precisely controlled. Increasing the rate of oil flow under the samples accelerates the rate of peptide aggregation (Figure 3.11b). A 1.75-fold increase in aggregation rate was observed in flowing oil through the device at 6 μL min-1. The functionality may prove useful for probing the self-assembly of amyloidogenic proteins that aggregate at a slower rate than Aβ42 – Parkinson’s disease related protein α-synuclein, for example.

Figure 3.11: Effect of shearing on Aβ42 aggregation kinetics. a) Single droplet within a trap. Green arrows indicate the direction of the movement within the droplet when the oil is flowed beneath. Scale = 50 μm. b) The effect of shearing on the aggregation process. Increasing the oil flow rate under the trapped droplets results in an increased rate of aggregation. A 1.75-fold increase in aggregation rate is observed when the oil -1 is flowed at 6 μL min throughout the whole experiment. Conditions: 40 μM Aβ42, 50% labelled, Plot show mean ± SEM, n = 20-30.

Agitated conditions have been proposed to accelerate the rate of aggregation by increasing monomeric collisions or by amplifying fibril fragmentation.316–318 These effects increase the rate of nucleation, or exposes more free ends for accelerated fibril elongation, respectively. The factors involved with increasing the rate of aggregation in the droplet format have not yet been extensively studied, however similar effects are most likely at play. Examining the effects of small molecules inhibitors at different shearing rates, may provide information regarding which steps of the aggregation process they target. Furthermore, the capacity to monitor aggregation under shearing conditions provides a novel opportunity to mimic convection in the brain. Aβ42 is transported by bulk flow of interstitial fluids in the brain,319 and this movement may play an important, but underappreciated role in the aggregation process in vivo. The shearing functionality of the nanoFLIM provides an easy method to examine such effects.

3.3.2.5 Interfacial Effects

Factors which are not often taken into account when designing and analysing the output from amyloid aggregation assays are surface and interfacial effects.123 The aggregation process in vitro encounters diverse interfaces, from container walls to air and biological membranes. Surfaces have a significant impact on the rates of aggregation of amyloidogenic proteins and can either accelerate or retard fibril formation depending on the peptides affinity for the given surface.320,321 Adsorption at an interface can change physical and chemical properties of a protein, under the influence of many factors. These include electrostatic interactions, surface roughness and curvature and hydrophobic-hydrophilic interfaces (HHIs).314,322 Of particular importance are HHIs, as various studies have shown that lipid membranes mediate amyloid aggregation in model systems, with human or artificial membranes accelerating the self-assembly process.323,324 Understanding how this interaction affects amyloid formation is necessary to dissect pathogenic mechanisms and design therapeutic strategies. Consequently, recognizing and standardising interfacial effects in in vitro assays is of great importance, for generating reproducible data and targeting biologically relevant aggregation processes.

In most in vitro assays, the predominant interfacial effects encountered are at the air-water interface (AWI), the presence of which generally enhances the rate of formation of amyloid aggregates.314,321,325,326 Aβ displays amphipathic organization. The amino-terminal residues 1-28 are largely polar, derived from the extracellular domain, whereas the 29-42 residues are hydrophobic in accordance with their transmembrane localisation with APP.327 This confers surfactant like properties on the peptide, as evidenced by a significant reduction in surface tension of an aqueous sample when this peptide is added, which implies surface adsorption.327 The link between Aβ’s surface activity and how AWI accelerate aggregation is not entirely known, but is potentially due to higher local concentrations at the interface.325,328 The peptide accumulates in a specific orientation along the surface.329 The lipophilic moieties are oriented towards the hydrophobic phase and lipophobic moieties towards the hydrophilic phase, causing the peptide to display biophysical and kinetic behaviour different from that in the bulk solution. The increased local concentration and ordered nature induces a higher propensity to aggregate, favouring β- sheet formation.328,329 Removal of the AWI greatly diminishes the rate of aggregation of various amyloidogenic peptides, whereas the introduction of additional HHIs in the form of beads or stirrers accelerates the process.326,330 It is unclear if the structures formed at the AWI are more or

84 less representative of structures formed in vivo, relative to those formed in bulk solutions without the accelerating interface.

As lipid membranes are believed responsible for mediating interfacial aggregation in vivo, it is expected that the oil-water interface (OWI) in the microfluidic assay, represents an equally useful, if not better, system for modelling aggregation at the HHI. The droplets have a higher surface area to volume ratio than samples in a normal 96 well plate format, and their smaller size allows faster peptide diffusion to the interface. The droplet interface is also more simple than that encountered in the plate format, and the presence of only one surface component introduces less variability into the assay. Changing the percentage of surfactant in the carrier phase represents a facile method to controllably modify the properties of the interface and monitor the associated changes in aggregation rate, and would be facilitated by the chip design, where the trapping of the droplets in individual grids allows for the use of very low surfactant concentrations.278 Such investigations have not yet been performed.

3.3.2.6 Advantages of Miniaturisation

A clear advantage of the microfluidic system is the ability to perform many repeats with minimal peptide consumption. Aβ42 is highly aggregation prone, so to obtain representative aggregation kinetics, the peptide must be scrupulously prepared before use to remove any preformed aggregates.75,331 If the means for such sophisticated preparation methods are not available, then the presence of very low concentrations of seeded structures can greatly skew data. Performing multiple repeats is therefore necessary to get real insight into the underlying aggregation process. Using the droplet-on-demand technology, 110 droplets, with a volume of 18 nL each, can be generated from a 10 μL sample. These are trapped and individually monitored over time, generating precise kinetic data. The need for many replicates in monitoring Aβ42 aggregation is highlighted in figure 3.12, which shows the aggregation profiles of a row of 10 droplets formed from the same 10 μL of stock peptide solution. As clearly shown, there is a great discrepancy between the initiation of aggregation in the first droplet and the last. Guided by such data it was deemed necessary to use at least 5 droplets per condition in subsequent compound screening, to detect subtleties in the inhibitory activity of different compounds. This requires approximately 496 ng of peptide per condition. For comparison, 22.6 μg of peptide would be necessary for five repeats using a 96 well plate ThT fluorescence format at the same concentration. This 50-fold

85 reduction in peptide consumption is particularly important for large scale compound screening, both in terms of cost and time efficiency.

Figure 3.12 also highlights another potential application of the system, a means to investigate early aggregation events and the effect that exogenous compounds can have on this process. The length of the lag phase reflects the energy barrier that must be overcome in order for the otherwise unfolded Aβ to nucleate and adopt a conformation rich in β pleated sheets, for subsequent oligomerisation and elongation. Once sufficient nucleated structures have assembled a rapid aggregation cascade ensues. In ensemble techniques, the individual primary nucleation events are quickly overwhelmed by secondary processes.75 In the droplet format, however, the stochastic primary nucleation events are limited to individual droplets with the subsequent propagating cascade confined within the microvessel boundary. Consequently, the microfluidic format gives a better representation of the rate at which aggregation is initiated, rather than just the accelerated elongation. This may be of use in investigating at which step small molecules function in the complex aggregation pathways by means of kinetic analysis (Section 1.6.2.1).

Figure 3.12: Monitoring the stochastic nature of Aβ42 aggregation with the nanoFLIM. a) 10 droplets formed from the same 10 μL stock solution of peptide. Initiation of aggregation within the peptides occurs stochastically and the fluorescence lifetime of individual droplets can be seen to change at different time points. b) Aggregation profiles of the 10 droplets. A great difference in the length of lag phase can be observed, highlighting the need for multiple repeats when monitoring the kinetics of Aβ42 aggregation. Conditions: 20 μM Aβ42-488, 20% labelled.

3.3.3 Validation with Known Inhibitors

Once a variety of extrinsic factors had been tested to ensure that the nanoFLIM was capable of providing realistic aggregation data, a series of known inhibitory small molecules were screened to validate the ability of the system to detect aggregation modifiers. Nine compounds known to display varying degrees of inhibitory activity were employed. These were EGCG, curcumin, scyllo- inositol, myo-inositol, morin, myricetin, o-vanillin, resveratrol and EPPS.44,174,178,194,332,333 10 droplets of each of the compounds were formed and filled into rows into the device (Figure 3.13a).

10 μM Aβ42-488 at a 50% labelling density with 50 μM of each compound was tested. The change in fluorescence lifetime over time in association with peptide aggregation was measured, providing distinct aggregation profiles for each of the inhibitors (Figure 3.13b).

Figure 3.13: Validation of the nanoFLIM with known inhibitory small molecules. a) Structures of the known inhibitory molecules used to validate the system. b) Filled droplet array chip following 150 min incubation. Each row contains 10 identical droplets containing 10 μM Aβ42488 (50% labelled) with 50 μM of a small molecule known to modulate the aggregation process. Scale 500 μm. c) Aβ42 aggregation profiles obtained through FLIM imaging of the filled chip in b). Plot shows mean ± SEM, n = 10.

As expected, a selection of the compounds appeared to completely prevent (EGCG, myricetin,

EPPS) or delay (myo-inositol, resveratrol) the change in fluorescence lifetime associated with Aβ42 aggregation. Interestingly, the aggregation profiles observed with a few of the compounds suggest an increased aggregation rate relative to the peptide alone control (scyllo-inositol, morin, o- vanillin). To further investigate this observation, ThT fluorescence assays and TEM imaging was performed (Figure 3.14a, b). It has been reported that some aggregation modifiers function by redirecting monomeric peptide from an inert fibril forming pathway, to one that results in the generation of toxic oligomers instead.144 Dot blot assays were also performed to investigate the potential toxicity of the species formed following treatment with the inhibitory compounds. The toxic oligomer specific A11 antibody and the control 6E10 antibody (which detects all Aβ forms) were used (Figure 3.14c).144 An overview of the results observed are given in table 3.1

Table 3.1 Overview of the observed activity of known small molecule aggregation inhibitors. Comparison of the inhibitory compounds’ modulatory effects on the Aβ42 aggregation process, as monitored using TEM, ThT fluorescence, nanoFLIM imaging and immunoassay dot blots. Ticks () are used to denote a positive A11 antibody result from Aβ42 samples incubated with the text compound, indicative of the presence of toxic oligomeric species. Crosses () denote no A11 sensitivity.

Compound TEM ThT Fluorescence nanoFLIM A11

Small clumped Strong inhibition Strong inhibition EGCG structures 

No inhibition High density of Curcumin Inhibition - Reduced final curly fibrils  fluorescence lifetime

scyllo-inositol Few small knotted No inhibition No inhibition fibrillar species - Accelerated aggregation 

No inhibition Moderate inhibition myo-inositol Long stringy fibrils - No lag phase - Lengthened lag phase 

No inhibition

High concentration - Accelerated aggregation Morin Moderate inhibition of small aggregates - Reduced final 

fluorescence lifetime

High concentration Myricetin Strong inhibition Strong inhibition of long thin fibrils 

Amorphous Moderate inhibition No inhibition o-vanillin aggregates - Reduced plateau - Accelerated aggregation 

Small and large Moderate inhibition amorphous Moderate inhibition Resveratrol - No lag phase  aggregates

Few clustered EPPS No inhibition Strong Inhibition aggregates 

Figure 3.14: Comparison of the inhibitory compounds’ modulatory effects on the Aβ42 aggregation process, as monitored using TEM, ThT fluorescence, nanoFLIM imaging and immunoassay dot blots. a) TEM image of untreated Aβ42 (10 μM), showing large clumps of short fibrillary species. Scale = 500 nm. A zoomed in image shows individual fibrils of approximately 10-20 nm in width. Scale = 50 nm). b) TEM images and ThT fluorescence and nanoFLIM profiles showing the effect of compound treatment. TEM was performed using Aβ42 (10 μM) following 7-day incubation in the presence of the small molecule inhibitors (50 μM) (Scale = 500 nm). The black lines in the aggregation profiles represent the Aβ42 control, which is either unlabelled or 50% labelled depending on the assay format. Varying degrees of inhibitory activity can be observed with the ThT assay, whilst the nanoFLIM indicates that some small molecules also are capable

91 of speeding up the aggregation process relative to that of Aβ42 alone. c) Immunoassay dot blots investigating the potential formation of toxic species, using oligomer-specific antibody A11 and control 6E10. Positive results for oligomeric species are observed in the Aβ42 samples treated with curcumin, scyllo-inositol, morin, myricetin, o-vanillin and resveratrol. No A11 sensitive species are evident following incubation in the presence of EGCG, myo inositol or EPPS. A positive result is seen for all samples with the 6E10 antibody, which detects all Aβ species.

As described, the results acquired for a selection of the inhibitory compounds differ in the ThT fluorescence and nanoFLIM assay. To investigate potential causes of these differences, the spectroscopic properties of the compounds were analysed. Many of the inhibitors employed are polyphenolic in nature, namely EGCG, curcumin, morin, myricetin and resveratrol. Such compounds are characterised by conjugated ring systems, in which the occurrences of π→π* electronic transitions often result in spectroscopic activity. If this activity falls into the range measured for the ThT assay it can bias the read out.121,126

Absorbance spectra (320-600 nm) of the compounds alone (50 μM) were recorded and compared to that of preformed Aβ42 fibrils (10 μM) with ThT (20 μM) (Figure 3.15a). Broad absorbance bands were observed for curcumin, myricetin, o-vanillin and morin. Of these, the peak for curcumin overlapped and dominated the ThT absorption at 440 nm, the excitation wavelength used in the ThT aggregation assays. This result has been previously reported.121 The absorption band of myricetin also overlapped with the ThT peak, as did that of morin but to a lesser degree. This implies that these compounds can interfere with the ThT assay through inner filter effects – by absorbing incident radiation intended for ThT excitation and/or reabsorbing ThT emitted radiation. This would result in a decrease in the dyes emission quantum yield, effectively quenching the ThT fluorescence and interfering with the real-time aggregation assay read out. To further investigate if this quenching was affecting the results from the ThT assay, emission spectra with fibrillary Aβ42 with ThT and each of the compounds were acquired (Figure 3.15b, c). As expected, significant quenching of the ThT-Aβ42 fibril signal (normalised to 100%) was observed upon the addition of increasing concentrations of curcumin, myricetin, o-vanillin and morin. This quenching is believed to be responsible for the difference in inhibitory activity observed between the two aggregation assays for curcumin and morin, and may also contribute to the results observed with o-vanillin.

Figure 3.15: Investigations into the spectroscopic activity of known inhibitory compounds. a) Absorbance spectra of the inhibitory compounds (50 μM), relative to that of ThT (20 μM) and preformed Aβ42 fibrils (10 μM) The absorbance band of curcumin drastically dominates over the Aβ42-ThT sample. The absorbance bands of myricetin and morin also overlap, but to a lesser degree. b) Normalised emission spectra of aggregated Aβ42 and ThT samples in the presence of 50 μM of each of the known inhibitory compounds. The Aβ42 fibril-ThT spectrum is normalised to 100% and the intensity of all other spectra shown relative to this. Addition of EGCG, curcumin, resveratrol, myricetin or o-vanillin has a profound effect on the fluorescence spectra, relative to the control. scyllo-inositol, myo-inositol, morin and EPPS have little effect on the emission spectra. c) Detailed emission spectra of the aggregated Aβ and ThT samples with increasing concentrations of the inhibitory compounds.

Interestingly, resveratrol and EGCG were not seen to be spectroscopically active at the excitation wavelength used in the ThT fluorescence assay (Figure 3.15a). The fluorescence intensity in the emission scan, however, is still dampened with increasing compounds concentrations (Figure 3.15b, c). This result was previously observed for resveratrol in a similar study investigating bias in ThT amyloid aggregation studies,121 and indicates that the compounds interfere with the ThT

93 fluorescence assay read-out by another mechanism, potentially through binding site competition with ThT or through direct interaction with ThT itself. The nanoFLIM also detected aggregation modifying effect, independent of ThT interaction, with the use of both resveratrol and EGCG. Moderate aggregation is seen with resveratrol and the final fluorescence lifetime observed is higher than that observed with the peptide alone, possibly suggesting the formation of more diffuse structures. This idea is supported by TEM analysis, where amorphous blobs are observed. In the case of EGCG, although the spectroscopic properties of the compound interfere with the ThT fluorescence assay, the compound was still validated as an inhibitor through analysis with the nanoFLIM and with TEM imaging. The ThT interference may, however, result in an over-estimation of the compounds inhibitory activity through conventional ThT fluorescence assays. Generating

IC50 values using the fluorescence sensor represents an alternative approach for assessing the activity of hit inhibitory compounds, in which the intrinsic properties of the test molecules are less likely to bias the result.

At this point, it is unclear why the ThT assay does not detect a change in Aβ42 aggregation propensity with scyllo-inositol, even though very few fibrillary structures are observed by TEM. The compound has been proposed to function by coating the surface of Aβ protofibrils thereby disrupting their lateral stacking into amyloid fibrils.334 These structures may be high in β-sheet character, thereby sensitive to ThT interaction, however this has not been further investigated. Gratifyingly, the nanoFLIM detects that the compound modifies the aggregation profile, quickly resulting in the formation of species which are morphologically distinct from the typical Aβ42 fibrils formed.

A strong inhibitory effect was oberved with myricetin and moderate inhibitory activity shown with myo-inositol using the nanoFLIM, however, high densities of fibrillar species were in fact observed in both samples through TEM investigations. A plausible explanation may be that these compounds delay the initial nucleation, thereby extending the lag phase beyond the time measured in this experiement. Monitoring the aggregation process with the nanoFLIM for a longer time may shed some time on this discrepancy between TEM and lifetime sensor results.

Finally, there is a stark contrast between the ThT fluorescence and lifetime sensor aggregation profiles of EPPS, which cannot be rationalised in terms of ThT fluorescence or binding interference. EPPS has been reported to successfully disaggregate preformed fibrils in AD mouse models, leading to a significant reduction in hippocampus-dependent behavioural deficits and brain inflammation.194 According to the data shown here, ThT fluorescence screening would not have picked up on the activity of this confirmed aggregation modifier. However, the nanoFLIM has

94 detected that the compound can perturb the aggregation process. This highlights an advantage of the system over the traditional ThT technique, even in the case of spectroscopically inactive compounds.

3.4 Summary and Conclusions

The process of Aβ42 aggregation is affected by a multitude of environmental factors, including buffer pH, temperature, solvent composition, agitation and the presence of interfaces and macromolecular components.77,219 In this work, a microfluidic assay system based on the use of an amyloid aggregation lifetime sensor has been designed. This nanoFLIM platform provides a simple means to carry out systemic variations of extrinsic conditions and to investigate their effect on Aβ42 aggregation propensity, with or without the presence of inhibitory small molecules.

The lifetime sensor requires the use of fluorescently labelled monomeric peptide. Aggregation of the peptide positions the dye molecules in close proximity, permitting inter dye interaction and resulting in fluorescence self-quenching between neighbouring peptides.255 A time dependent decrease in fluorescence lifetime of the reporter dye can be monitored in association with peptide aggregation. It was observed that increasing the percentage of peptide labelling increases, to a certain extent, the dynamic range between fluorescence lifetimes of the monomeric and aggregated peptide. 50% labelling was identified as the optimal labelling density for the assay platform, providing a large dynamic range, high photon count and reproducible aggregation kinetics, without greatly affecting the morphology of the structures formed.

The ability of the lifetime sensor to report on the aggregation tendency of Aβ42 as a function of temperature and pH was also investigated. An expected increase in aggregation rate was temperature increase (25 to 37 °C) was observed.309 Negligible self-assembly was observed at pH 9, and the highest rate of aggregation was seen to occur at pH 6. This agrees with previous studies employing alternative spectroscopic and electron microscopy techniques to probe pH sensitivity.44,313 The lifetime sensor detected that the species formed under different pH conditions differed in morphology, as indicated by variations in final fluorescence lifetime value achieved. It has been reported that structures formed within acidic environments are most toxic, and are more representative of the species formed in acidic lysosomal vesicles in vivo (further

313 discussed in Chapter 6). Therefore the ability of the lifetime sensor to report on Aβ42 aggregation at different pH has practical implications, as it can be used to monitor how small molecule inhibitors function in different conditions that mimic physiologically relevant environments.

Biological membranes and convection in the brain may influence Aβ aggregation in vivo. Paying insufficient attention to surface and interfacial effects has previously been reported as one of the factors responsible for the poor reproducibility of in vitro aggregation screens.123 Such effects are not usually well-controlled, with the amount and composition of solid or air interfaces not consistently maintained. The oil-water interface in the microdroplets provides a relevant but simple model of the hydrophilic-hydrophobic interfaces experienced in live systems, and is less susceptible to variations between experimental repeats. There is also potential to examine the effects of controlled changes in the properties of the interfaces experiences, by modifying the concentration of surfactant in the carrier phase.278 Inducing sheer force on the peptide sample is relevant for studying the effects of convection on the aggregation process, and the microfluidic device developed provides an easy and well controlled means to achieve this. Increased oil flow within the device permits increased shearing rates, which was observed to increase the rate of aggregation.

The aggregation profiles observed using the nanoFLIM were compared with those obtained with conventional ThT fluorescence assays. The results were largely in agreement for the extrinsic factors, with the results from the nanoFLIM benefitting from more controlled aggregation conditions and greatly reduced reagent consumption. The lifetime sensor may also provide better insight into morphological differences in the species formed under different conditions. The ThT fluorescence assay shows a typical sigmoidal evolution of fluorescence signal as aggregates are formed. Whilst the plateau is often used to estimate the fibrillary mass content within a sample, the relationship between ThT intensity and structural properties of the resulting amyloid is not well documented, and the effects are often overlooked by normalising the emission date. Different morphologies do, however, result in different levels of fluorescence intensity emission, as shown by Lindberg et al. in a comparison of the emission of concentrated matched Aβ aggregates formed under different incubation conditions.120 As fluorescence intensity is generally expressed in arbitrary units, ThT fluorescence data is necessarily normalised to compare aggregation profiles acquired in different experimental repeats, and the potential to deduce structural information is therefore lost. The lifetime sensor provides a more quantitative measure, an exact end point fluorescence lifetime, which reports on differences in aggregate architecture. This allows for comparison between samples from different experiments and highlights potential differences in species morphology when the peptide is incubated under various conditions.

The possibility of monitoring the inhibitory activity of small molecules was tested using a small selection of known inhibitory compounds. The results obtained were compared to those observed with conventional ThT fluorescence assays and TEM imaging. The insights into Aβ42 aggregation

96 provided by the assay systems were largely in agreement, however some differences were observed and highlighted advantages of the FLIM system compared to the ThT fluorescence format. Reduced assay interference by the intrinsic properties of the test compounds was the primary benefit observed. Limitations of the newly developed assay were also noted, specifically the detection of false positive results (myricetin and myo-inositol) which were shown to be incorrectly assigned as inhibitory by subsequent TEM analysis. Longer screening times may have avoided such issues.

Overall, the assay developed permits real time probing of amyloid self-assembly, in an easy to implement and relatively fast manner. It enables inhibitor screening with very low reagent consumption, and circumvents the issues of ThT fluorescence interference and binding site competition, which often restrict conventional spectroscopic assay. In this work, only Aβ42 was used, but the assay format could be applied for monitoring the aggregation of other amyloidogenic proteins, for the study of other neurodegenerative disease. To understand the features that contribute to the onset and progression of amyloid related disorders, it is crucial to investigate all the factors that modulate amyloid formation. Understanding at a molecular level may reveal important information for treatment strategies. The nanoFLIM assay allows for easy alteration of environmental conditions, to mimic relevant aggregation pathways and to probe the associated amyloidogenic polymorphisms, with or without the presence of small molecules modifiers.

Chapter 4 NanoFLIM Compound Screening and Hit Validation

4.1 Introduction

This chapter outlines the screening campaign of an array of small molecule libraries using the newly developed nanoFLIM assay, with the aim of identifying compounds capable of perturbing the process of Aβ42 aggregation. A key library of interest was identified, and the inhibitory activity of a subset of its constituent compounds was studied in detail to evaluate the reliability of this screening platform and its advantages over conventional assay formats. Additional analysis of the lead compound, by means of various biophysical tests and cell-based screens, was performed to further validate its inhibitory activity.

4.2 Results and Discussion 4.2.1 Library Screening

Once the newly developed nanoFLIM screening system had been optimised and validated with a selection of known inhibitors (Chapter 3), a screening campaign with a collection of novel chemical libraries commenced. The libraries were synthesized by current and previous members of the Spring research group and ranged in size from 7–111 compounds. The libraries were either prepared in accordance with the diversity-oriented synthesis (DOS) approach, for screening campaigns with specific biological targets or to probe unchartered regions of chemical space with the strategic incorporation of pharmacologically relevant motifs. A total of 445 compounds from 9 libraries were screened. A full list of the libraries, with their hit rates and associated literature references (if previously published) is presented in Appendix I. Through such work a 13% total hit rate (i.e. compounds displaying >30% inhibitory activity at 50 μM) was observed. Figure 4.1 shows the results from a representative screen, in which 9 compounds from a cinchophen library were screened.335 9 droplets per compound at a single concentration are shown in the rows, and these were imaged to give single start point and end point fluorescence lifetime values. Of these compounds, MJ040 (row 3) displayed the strongest inhibitory activity, completely preventing aggregation of the peptide for the time course of this experiment. To test the reliability of the assay system, and assess the inhibitory activity of the constituent compounds, this cinchophen

99 library was further studied. It is referred to as the” MJ library”, after Dr Matej Janeček, the researcher who synthesized it.

Figure 4.1: Example of nanoFLIM screening approach for hit identification. Droplets (18 nL) containing partially labelled peptide and the library compounds (n = 9) were filled in rows in the microfluidic device. Start and end point fluorescence lifetime measurements were taken to identify compounds capable of perturbing Aβ42 aggregation. MJ040 (row 3) is seen to exhibit a strong inhibitory effect is this assay. Scale = 500 μm. Conditions: 10 μM Aβ42-488, 50% labelled, 10 μM compound.

4.2.2 Hit Library

Leading on from the original start and end point readings for initial inhibitor identification, the nanoFLIM was next employed to monitor the aggregation kinetics of compounds of interest by taking fluorescence lifetime measurements throughout the course of the experiment. The characteristic aggregation profiles generated were then compared with those obtained using a conventional ThT assay. Similar to the previously described investigations with the known inhibitors (Section 3.3.3), some striking differences were observed between the results attained in the two assay formats. A subset of the particularly interesting library components (MJ001, MJ014, MJ018, MJ036, MJ040, MJ041, MJ042; Figure 4.2) were studied in further detail to investigate the observed effects. Figure 4.3 shows a comparison of the nanoFLIM and ThT fluorescence assay aggregation profiles, as well as TEM images showing the Aβ42 species formed following seven days of incubation with the chosen compounds. Inhibition of Aβ42 aggregation with the addition of MJ001 is clearly seen with the nanoFLIM and TEM imaging. The ThT fluorescence assay, however, indicates little inhibition of the self-assembly process. The ThT fluorescence assay shows an unusual aggregation profile with MJ014 treatment, where the curve peaks at the same level as the control but quickly reduces again. Increased aggregation kinetics are seen with this compound in the nanoFLIM assay. Irregularly long fibrils are observed with TEM imaging. Differences in the morphology of these structures may account for differences in ThT

100 sensitivity compared to control fibrils, resulting in the unusual aggregation profile observed in that assay format.

A strong inhibitory effect is observed with MJ036 in the ThT fluorescence assay, however aggregation is detected with the nanoFLIM. TEM analysis shows the presence of clusters of aggregates, with seemingly knotted fibrillary structures. A profound inhibitory effect is observed with MJ040 in all three assay formats. The lifetime sensor indicates complete inhibition with MJ041, but a reduced inhibitory effect is seen with the ThT fluorescence assay. The formation of few small aggregates are observed with TEM imaging. Finally, the ThT fluorescence signal of MJ042 exceeds that of ThT and the peptide alone, indicating no inhibitory effect. Conversely, no aggregation is seen to occur with the nanoFLIM assay. TEM imaging does not show the presence of any fibrillar species, however a high density of small aggregated structures is evident. Dot blots with the oligomer-specific A11 antibody were performed to investigate the potential toxicity of the species observed in the TEM imaging. This revealed no A11 sensitive structures with MJ014 and MJ018, trace amounts with MJ001, MJ040 and MJ041, and a higher density of immunoreactivity species with MJ036 and MJ042 (Figure 4.3c). From this analysis, it appears as though the ThT assay incorrectly identified one false positive (MJ036) and three false negatives (MJ001, MJ041 and MJ042) and provides an unusual kinetic profile for MJ014, which does not reflect on the extent of fibril formation observed with TEM imaging. The full MJ library and associated ThT and nanoFLIM aggregation profiles are shown in Appendix II.

Figure 4.2: MJ library test subset. Structures of the subset of compounds from the MJ library used to compare the detection of inhibitory activity using the nanoFLIM or ThT fluorescence assay.

101

102

Figure 4.3: Comparison of the inhibitory compounds’ modulatory effects on the Aβ42 aggregation process, as monitored with TEM, ThT fluorescence, the nanoFLIM and immunoassay dot blots. a) TEM image of untreated Aβ42 (10 μM), showing large clumps of short fibrillary species. Scale = 500 nm. A zoomed in image shows individual fibrils of approximately 10-20 nm in width. Scale = 50 nm). b) TEM images and ThT fluorescence and nanoFLIM profiles showing the effect of compound treatment on the aggregation of Aβ42. TEM was performed using Aβ42 (10 μM) following 7-day incubation in the presence of the small molecule inhibitors (50 μM) (Scale = 500 nm). The black lines in the aggregation profiles represent the Aβ42 control, which is either unlabelled or 50% labelled depending on the assay format. Varying degrees of inhibitory effect can be observed with the compounds. c) Immunoassay dot blots with control antibody 6E10, which detects all Aβ species, and with toxic oligomer-specific A11 antibody. From this combined analysis it appears as though the ThT assay incorrectly identified one false positive (MJ036) and three false negatives (MJ001, MJ041 and MJ042), and does not reflect in the increased rate of fibril formation with MJ014 treatment.

As previously discussed (Section 1.5.2) many libraries contain small molecule which are in themselves spectroscopically active in the range probed with the ThT assay.127 This activity can mask or mimic inhibitory activity, thereby biasing the ThT read out and giving rise to false positives of negatives.125. To investigate the potential spectroscopic activity of the test subset, absorbance and emission spectra were obtained. Figure 4.4a shows the absorbance spectrum, where broad absorbance bands overlapping with that of the ThT and preformed Aβ42 sample are evident with each of the compounds tested. This indicates potential interference of the ThT fluorescence assay readout by inner filter effects (as discussed in section 3.3.3). Emission spectra were acquired using preformed Aβ42 fibrils and ThT samples (normalised to 100% intensity), and the changes in fluorescence intensity or shape of this control spectrum with the addition of increasing

103 concentrations of the small molecules were investigated (Figure 4.4c). The emission spectrum for 440 nm excitation (the wavelength used in the ThT assay) shows that both MJ036 and MJ040 effectively quench ThT fluorescence (Figure 4.4b,c). MJ014 does not appear to significantly change the extent of ThT fluorescence, indicating that the unusual ThT aggregation profiles observed with this compound is not due to its spectroscopic properties. MJ001, MJ018, MJ041 and MJ042 exhibit higher fluorescence than the Aβ42 and ThT alone. This suggests that the intrinsic fluorescence of the compounds biases the ThT assay readout potentially masking their inhibitory activity and wrongly classifying them as non-inhibitors. The nanoFLIM on the other hand, picks up the inhibitory activity of these compounds, which was confirmed by TEM analysis.

Figure 4.4: Investigations into the spectroscopic properties of the MJ library. a) Absorbance spectra of the select compounds (50 μM), relative to that of ThT (20 μM) and preformed Aβ42 fibrils (10μM). The absorbance bands of all the compounds overlap with that of the peptide-ThT sample. b) Normalised emission spectra of Aβ42 fibrils (10μM) and ThT samples (20μM) in the presence of 50 μM of the MJ library test compounds. The Aβ42 fibril-ThT spectrum is normalised to 100% and the intensity of all other spectra shown relative to this. Addition of all compounds, except MJ014, has profound effects on the fluorescence spectra, relative to the control. c) Detailed emission spectra of the aggregated Aβ42 and ThT samples with increasing concentrations of the inhibitory compounds. To investigate the potential false positive classification of MJ036 in the ThT assay, AFM was performed to track peptide aggregation using a fluorescence independent technique (Figure 4.5).

104

In the Aβ42 control sample, short fibrillar structures were detected following three days of incubation. By day seven, long clumped fibrils were observed. In the MJ040 sample, no aggregated species could be identified after three days after incubation. At day seven, only small aggregated species were present. With MJ036 incubation, many small aggregates could be seen after three days of incubation. After seven days, larger aggregates were present. The long fibres observed were interrupted with large clumps of peptide. This may have served to block ThT binding sites, further contributing to the false positive result for inhibitory activity detected by the ThT fluorescence assay.

Figure 4.5: Monitoring the inhibitory activity of MJ036 and MJ040 with AFM. AFM images of Aβ42 species formed in the absence or presence of a fivefold excess of the test compounds following 3 (upper panel) or 7 day (lower panels) incubation. Untreated sample; small aggregates could be observed at day 3 (i) and long amyloid fibrils were present at day 7 (iv). MJ040 treated; no structures were observed at day 3 (ii), but some small species were evident at day 7 (v). MJ036 treated; at day 3 many aggregates could be observed (iii). Longer fibrillar species with clumps along the structure are detected at day 7 (vi). Images were acquired with tapping mode AFM in air. i),ii),iii) scale bar = 100 nm. v),v),vi) scale bar = 600 nm.

The toxic oligomer specific A11 antibody was again used to investigate the toxicity of the species formed in a time dependent manner (Figure 4.6). The control antibody, which detects all Aβ forms, showed a positive result at all days of the measurement for untreated, MJ040 and MJ036 treated samples. The oligomer specific A11, did not detect the formation of toxic species in the MJ040 sample, suggesting that the structures observed by the scanning microscopy techniques were innocuous aggregates. A11-sensitive structures were, however, detected in the MJ036 treated sample. Given the presence of aggregated structures with AFM and TEM imaging, as well as the

105 evidence of toxic oligomeric species using immunoassay dot blots, the compound MJ036 was no longer studied as a potential inhibitor. Attention was instead focused on the more active compounds.

Figure 4.6: Time course immunoassay to detect the formation of toxic Aβ42 oligomers. Dot blots following 1,3 or 5-day incubation with MJ036 or MJ040 to investigate the potential formation of toxic species, using oligomer-specific antibody A11. Oligomeric structures were present following 1,3 and 5 days of incubation in the absence of any compounds or in the presence of MJ036. No A11 sensitive species were detected following incubation with MJ040. Positive readouts were obtained for all samples using the 6E10 control antibody, which detects all Aβ species.

4.2.3 Hit Validation

Due to the small sample size of the MJ library, it was difficult to generate any SAR data to guide synthetic optimisation strategies. Consequently, the most active compound from the original screen, MJ040, was brought through to subsequent inhibitor validation stages without any synthetic attempts of improvement. The IC50 value was calculated using the nanoFLIM by filling the device with a concentration gradient of droplets and monitoring the characteristic aggregation profiles (Figure 4.7a). Normalised initial rates of aggregation were used to plot the IC50 curve and an IC50 value of 3.8 ± 0.9 μM was calculated (Figure 4.7b). A similar calculation was also performed using a ThT fluorescence assay, to compare the value obtained with the new assay format and that conventionally used. A value of 5.2 ± 1.4 μM was calculated in this format (Figure 4.7c). The value obtained with the nanoFLIM is believed to be more accurate than that calculated with the ThT assay, given the likelihood that the fluorescence quenching ability of compound interferes with the ThT assay read out. In any case, the IC50 of the compound falls in the same range as that calculated for the phase III clinical trials inhibitor EGCG (IC50 = 6.4 ± 0.7 μM, Figure 4.7d), thereby confirming that the compound displays therapeutically relevant, inhibitory activity.

106

Figure 4.7: IC50 calculations of MJ040 and EGCG. a) Aβ42 aggregation profiles obtained with the nanoFLIM by filling the microfluidic device with a concentration gradient of MJ040. Plot shows mean fluorescence lifetime, errors bar omitted for clarity. Dotted lines show linear fit for each concentration to calculate the initial rate of aggregation. b) IC50 calculation using the initial rates obtained for each concentration with the fluorescence lifetime sensor. IC50 = 3.8 ± 0.9 μM. c) IC50 calculation of MJ040 using the initial rates obtained for a concentration gradient of MJ040 in a ThT fluorescence assay; IC50 = 5.2 ± 1.4 μM. d) EGCG IC50 curve using initial aggregation rates obtained using a ThT fluorescence assay; IC50 = 6.4 ± 0.7 μM. Conditions: nanoFLIM - 10 μM Aβ42-488, 20% labelled, compound = 0.5 – 50 μM, n = 7. ThT fluorescence assay - 10 μM Aβ42, 20 μM ThT, compound = 0.5 – 50 μM, n = 3. IC50 values were calculated on GraphPad using the 푦 1 equation: = 푥 푦푚푎푥 1+ 퐼퐶50

To test the ability of the hit compound to rescue cells from Aβ42 induced toxicity, MTT viability assays were carried out using SH-SY5Y human neuroblastoma derived cells. An MTT assay is a colorimetric test for assessing cell metabolic activity, by means of measuring the ability of NAD(P)H-dependent cellular oxidoreductase enzymes to reduce the added MTT tetrazolium dye.336 Two approaches were taken to monitor the ability of MJ040 to rescue cells from the toxic effects of Aβ42; 1) monomeric Aβ42 was directly added to the cells, with or without the compound;

2) Aβ42 species that had been previously been incubated in conditions conducive to aggregation, again with or without the hit compound, were added to the cells. First, 500 nM monomeric Aβ42 and increasing concentrations of MJ040 were added to cells and incubated for 48 h. Here, the compound was not capable of significantly rescuing the cells at any of the concentrations tested (Figure 4.8). In a bid to improve cellular activity, the carboxylic acid of the compounds was masked as a methyl ester, to generate the prodrug MJ040X (Figure 4.9a). It was believed that this uncharged compound would be more capable of permeating the cell membrane, where it would

107 then be hydrolysed by cellular esterases. The prodrug displayed poor inhibitory activity in vitro, as indicated with a ThT fluorescence assay, suggesting that the carboxylic acid is necessary for interaction with the peptide (Figure 4.9b). In the monomeric MTT assay, however, cell treatment with MJ040X resulted in a significant increase in cell survival (Figure 4.8). This suggests that the lack of activity observed for the original compound is caused, at least in part, by poor cellular uptake as a consequence of the anionic group. It is worth noting that the cells were necessarily pre-incubated in the compound for 1 h prior to the addition of the monomeric peptide to observe a rescuing effect.

100 ns * ** ** 80

% Cell Survival Cell %

M M  42   A 50 nM 50 nM 500 nM 2.5 500 nM 2.5

MJ040 MJ040X

Figure 4.8: Monitoring the rescuing effect of MJ040 and MJ040X from monomeric Aβ42 induced toxicity. Rescuing effect of MJ040 and MJ040X on the cytotoxicity induced by the addition of Aβ42 monomers. Cells were incubated in the compound for 1 h prior to the addition of Aβ42 monomers (500 nM). Following 48 h incubation the cytotoxicity was assessed using an MTT cell viability assay. Anionic MJ040 was unable to rescue Aβ42 induced cell death, but prodrug MJ040X could significantly increase cell survival – a result attributed to its increased cell permeability. Cell survival was calculated using an MTT cell viability test, using SH-SH5Y cells. The viability of untreated cells was set as 100%. Error bars represent SEM, n = 4, statistical analysis performed by one-way ANOVA with Dunnett’s multiple comparison post-test; *p<0.05; **p<0.01.

108

Figure 4.9: Synthesis and analysis of in vitro inhibitory effects of MJ040X. a) Synthetic scheme for MJ040X synthesis. b) ThT fluorescence assay showing the lack of in vitro inhibitory activity of MJ040X compared to MJ040. This highlights the importance of the carboxylic acid for anti-aggregation activity. Conditions: 10 μM Aβ42, 20 μM ThT, 50 μM compound, n = 3.

To further explore the idea that the low cellular activity of MJ040 results from its low permeability, the viability assay was repeated using preformed Aβ42 aggregates (Figure 4.10). Here, the peptide was allowed to aggregate for 24 h at 10 μM, with or without MJ040 or MJ040X. The resulting samples were then diluted, added to the cells and incubated for a further 48 h. The preformed

Aβ42 aggregates were shown to induce greater cell death than the addition of monomers alone (61 ± 2% survival vs 69 ± 1% survival for aggregates and monomers, respectively). MJ040 was capable of reducing the formation of toxic aggregates enough during the in vitro pre-incubation that the addition of these treated peptide species to the cells was significantly less toxic than with untreated peptide aggregates. Incubation with MJ040X, however, had no significant effect on preventing the formation of toxic aggregates. No increase in cell survival was observed when the peptide was incubated in the presence of this compound, which is believed to be a result of its poor in vitro inhibitory activity.

109

100 ns * 80 ns ns 60

% Cell Survival Cell %

42 M M    A 50 nM 2.5 50 nM 500 nM 500 nM 2.5

MJ040 MJ040X

Figure 4.10: Monitoring the rescuing effect of MJ040 and MJ040X from pre-aggregated Aβ42 induced toxicity. Rescuing effect of MJ040 and MJ040X on the cytotoxicity induced by the addition of pre- aggregated Aβ42 species. Aβ42 (10 μM) was pre-incubated for 24 h with or without the compounds, then diluted (500 nM) and added to SH-SY5Y cells. After 48 h treatment, cytotoxicity was evaluated using an MTT assay. The species formed in the presence in the active drug MJ040 induced less cell death than the untreated peptide. MJ040X has poor in vitro inhibitory activity and did not rescue the cells from pre- aggregated Aβ42 induced cytotoxicity. Cell survival was calculated using an MTT cell viability test, using SH- SH5Y cells. The viability of untreated cells was set as 100%. Error bars represent SEM, n = 4, statistical analysis performed using one-way ANOVA with Dunnett’s multiple comparison post-test; *p<0.05.

4.2.4 Analysis of Rescuing Effects of MJ Library Test Subset

The rescuing effects of the other compounds of the MJ library test subset were investigated by means of MTT cell viability tests, after the lead compound MJ040 had been selected and carried forward for further biological testing (Chapter 5). This analysis, albeit in need of additional repeats and more in-depth experimental work, revealed that two more compounds exert significant protective effects (Figure 4.11). The addition of MJ014 and MJ042 were shown to significantly increase cell survival when monomeric Aβ42 is added to the cells. MJ014 promotes the formation of fibrils, as evidenced by nanoFLIM and TEM analysis. Compounds capable of providing protective effects through promoting the formation of stable non-toxic fibrils have previously been observed, and this activity is worth further examination for potential therapeutic application.204 The unusual aggregation profile of the ThT fluorescence assay with the compound hints at an abnormal aggregation process. The nanoFLIM, however, provides a clearer view of the increased aggregation kinetics, highlighting an advantage of this assay system. The rescuing effects observed with MJ042 is unexpected, due to the observation of A11-sensitive species in the dot blot assay.

110

Given the unusual dose-activity relationship observed, further investigations are necessary to confirm this initial result.

100 *** *** *

% Cell Survival Cell % 60

 42 A MJ001 MJ014 MJ018 MJ036 MJ040 MJ041 MJ042

Figure 4.11: Monitoring the rescuing effect of the full test subset of the MJ from monomeric Aβ42 induced toxicity. Rescuing effect of increasing concentrations of MJ001, MJ014, MJ018, MJ036, MJ040, MJ041 and MJ042 on the cytotoxicity induced by the addition of monomeric Aβ42 species. Cells were incubated in the compound for 6 h prior to the addition of Aβ42 monomers (250 nM). Bars for each condition represent the addition of 50 nM, 500 nM or 2.5 μM of the compound, respectively. Following 48 h incubation the cytotoxicity was assessed using an MTT cell viability assay. MJ014 (2.5 μM), MJ040 (500 nM) and MJ042 (500 nM) were seen to significantly increase the cell viability following exposure to Aβ42. Cell survival was calculated using an MTT cell viability test, using SH-SH5Y cells. The viability of untreated cells was set as 100%. Error bars represent SEM, n = 4, two technical repeats, statistical analysis performed by one-way ANOVA with Dunnett’s multiple comparison post-test; *p<0.05, **p<0.01.

4.3 Summary and Conclusions

In this chapter, the newly-developed nanoFLIM was employed to screen a selection of chemical libraries for inhibitory activity against the aggregation of Aβ42. In this first round of library screening, 445 compounds were tested, with at least 5 repeats per compound. An important advantage of this system is that only 217 μg of Aβ42 was required for the whole screening campaign. In contrast, the general 96 well plate format would require 11.2 mg of peptide to carry out such analysis. Working with Aβ42 is very expensive, in term of monetary cost and lab hours required to meticulously prepare monomeric peptide samples. The microfluidic system described benefits from greatly reduced running costs. A 10-fold reduction in the peptide price alone was achieved – calculated as £222 with the nanoFLIM and £2397 in a 96 well plate format, using commercially available labelled and unlabelled synthetic Aβ42. The requirement for reduced quantities of compound (360 pg per droplet) is also advantageous, especially for screening novel libraries composed to small samples of precious compounds.

111

Through this screening campaign, one particular library of interest was identified and focused on in detail. This ‘MJ library’ consists of 20 compounds, which were shown to display varying levels of inhibitory activity. Comparison of the aggregation profiles obtained using the nanoFLIM with those obtained with ThT fluorescence screening, highlighted some striking differences between the results achieved in both assays. Specifically, the fluorescence lifetime sensor was less susceptible to the detection of false positive and negative results than the conventional ThT assay, which was shown to be more prone to bias by intrinsic spectroscopic properties of the test library. Table 4.1 shows a summary of the results obtained with the different screening formats. One ThT false positive and three false negatives within the seven-member hit library test subset were identified using the nanoFLIM. An aggregation accelerating compound that provided protective effects in SH-SY5Y cells was also identified through the use of fluorescence lifetime measurements. It is suspected that further analysis of the other compounds of the cinchophen may reveal more incorrectly assigned aggregation inhibitory or accelerating compounds.

Table 4.1: Summary of the modulatory effects of the MJ library subset on the Aβ42 aggregation process, as monitored using TEM, ThT fluorescence, nanoFLIM imaging, immunoassay dot blots and cell viability tests. Ticks () are used to denote a positive A11 antibody result from Aβ42 samples incubated with the text compound, indicative of the presence of toxic oligomeric species. Crosses () denote no A11 sensitivity.

Compound TEM ThT Fluorescence nanoFLIM A11 Increased cell viability

Very few short MJ001 No inhibition Strong inhibition fibrils  

Many long Moderate inhibition Accelerated MJ014 fibrils Unusual profile aggregation  

Very few MJ018 Little inhibition Moderate inhibition fibrillar clumps  

Large clumps MJ036 Strong inhibition Moderate inhibition of fibrils  

MJ040 Very few small Strong inhibition Strong inhibition aggregates  

Very few small MJ041 Moderate inhibition Strong inhibition aggregates  

Many small MJ041 No inhibition Strong inhibition aggregates  

112

An IC50 value in the low micromolar range was calculated for the lead compound, MJ040, which was further validated using immunoassay dot blots and AFM. The compound was also brought forward for testing in neuroblastoma cells. A lack of cellular activity was initially observed, however, this was improved upon by chemical modification of MJ040 to give prodrug MJ040X. The carboxylic acid of this original structure is masked as a methyl ester in this modified structure, permitting easier access across the cell membrane. Intracellular hydrolysis of MJ040X is believed to release the active compound within the cell. MTT cell viability assays were used to confirm that the compound can rescue SHSY5Y cell populations from Aβ42 aggregation induced toxicity.

The next step, for both hit validation and assay expansion, was testing the applicability of the lifetime sensor for monitoring peptide aggregation in live cells. In vivo Aβ aggregation is an incredibly complicated and poorly understood process, the intricacies of which are often missed with in vitro screening techniques. Investigating if and how hit compounds function within cells in real time will provide insight into the suitability of the compounds as potential drugs, and can act to guide future drug design strategies. The development of this assay protocol will be discussed in Chapter 5.

113

114

Chapter 5 Cellular and In vivo Screening of Aβ Aggregation Inhibitors with the Fluorescence Lifetime Sensor

5.1 Introduction 5.1.1 Cellular Models for Monitoring Aβ Aggregation

The activity of small molecule Aβ42 inhibitors identified in vitro does not always correlate well with results obtained in cellular models and in vivo, due to the role that extrinsic factors, such as pH, metal ions and biological membranes, play on the aggregation process.219 This necessitates robust cellular assays to verify inhibitory activity of hit compounds in biologically relevant systems. There is, however, a shortage of real-time methods to monitor the self-assembly of Aβ in live cells. Aggregates in fixed cells can be imaged through immunofluorescence or with the use of amyloid reporting dyes such as ThT or Congo red.337–339 Issues in immunofluorescence studies arise from cross reactivity of Aβ specific antibodies, which often also recognise full length amyloid precursor protein or its derivatives.340,341 Furthermore, the elaborate preparation protocols necessary can influence Aβ immunoreactivity or provide suboptimal visualisation, and also can’t be applied dynamically.342,343 Staining with ThT is limited because Aβ exists as a mixture of metastable species, often lacking the β-sheet structure required for detection.344 Potentially toxic oligomers aren’t easily identified and the low fluorescence readout from small aggregates is masked by non- specific background fluorescence.345 Congo red also has restricted use for in vivo studies. It lacks specificity, has been observed to induce a mild inhibitory effect on the aggregation process and displays toxic effects in in vivo samples.346,347 Additionally, these spectroscopic techniques only give end point readings and provide little insight into how the aggregation process is progressing in real time.

Until recent developments by the Kaminski research group (which shall be discussed in further detail in section 5.2.1) fluorescence imaging of dynamic Aβ aggregation processes in living cells was not feasible.253,254 A selection of studies had previously shown the uptake of labelled Aβ species, and tracked their movement throughout the cell.124,348 Whilst useful in terms of localisation studies, these have limited ability to inform on the underlying self-assembly processes and the extent of aggregated species at a particular time. Current hit validation in cellular systems therefore relies almost exclusively on cell viability tests. These tests provide a format to quickly

115 elucidate if a compound or the specific Aβ species it generates are toxic to cells, but such work sheds little light on how the compounds are actually acting within the cellular environment.

Studying cellular system at physiologically relevant peptide concentrations presents another challenge for investigating in vivo aggregation processes. It is estimated that the concentration of Aβ in human cerebrospinal fluid is in the low nanomolar range.296,349,350 The aforementioned techniques are not sensitive in this range, further emphasising the need for improved screening techniques to resolve how aggregation modifiers function in live cells.

5.1.2 Caenorhabditis elegans as Disease Model Organisms

Research into the molecular and developmental biology of Caenorhabditis elegans was initiated by Sydney Brenner in 1974, and this multicellular organism has since been used to gain a better understanding of a great variety of biological processes and disease states.351 The advantages of using these sophisticated organisms as disease models are manifold.352,353 They reproduce rapidly and prolifically, and each self-fertilising hermaphrodite has approximately 300 progeny. In comparison to mice models, which have a 9-week generation time, it takes only 3 days for C. elegans to transition from egg to hermaphrodite adult. They are easily and inexpensively cultured in the laboratory and benefit from transparent bodies, allowing their internal structures and reporter dyes to be easily visualised. Furthermore, various fundamental cellular processes are conserved are exist between these nematodes and higher organisms.354 C. elegans homologues have been identified for 60-80% of human genes.355,356

Figure 5.1: Simplified schematic of the anatomy of an adult Caenorhabditis elegans hermaphrodite. C. elegans are free-living non-parasitic nematodes, which were first introduced as a model organism in the early 1970’s. These transparent roundworms are 1-1.3 mm in length and have a life span of 2-3 weeks. Hatching animals pass through four larval stages (L1-L4), each punctuated by a molt, to give rise to an adult hermaphrodite, consisting of 959 somatic cells.

116

Although there are many physiological differences between nematodes and mammals, C. elegans still represent a promising model for screening potential human therapeutics, and have been confirmed to predict human activity of CNS drugs such as nicotine and anaesthetics.353 Compared to cellular work, the use of worm models avoids issues with isolated drug-target interactions, instead providing the opportunity to study the implications of the drug effect in context with a whole system. Due to their short fluorescence lifetime and experimental flexibility, these models provide ample opportunity to test the effect of chronic exposure of molecules of interest. Alternatively, drug treatment can be initiated at different stages in the C. elegans development. This allows for life span-resolved characterisation of drug administration.

5.1.3 Alzheimer’s disease model Caenorhabditis elegans

Of the hermaphroditic C. elegans’ 959 somatic cells, 302 are neurons. These drive complex neuromuscular behaviours and their pattern of connectivity has been completely mapped out.357 Their short life span of 2-3 weeks makes them particularly well suited to the study of age-related disorders. As such, they have been extensively employed in probing neurodegenerative diseases.358–361 This is generally achieved by using models which emulate specific aspects of a disease mechanism and can be achieved by: i) knock out (mutant) or knock down (RNAi) of a gene of interest; ii) monitoring a C. elegans process that resembles a human disease mechanism; iii) expressing a human gene to induce a disease-related phenotype.352

The relevance of C. elegans in AD research is well-founded.358,362 The first genetic understanding of the disease came through the use of these organisms with the identification of sel-12.363,364 This is a homologue of human presenilin-1 and mutations in this protein were later shown to be associated with early-onset familial AD.365 To date, no Aβ-like peptide has been detected in C.

366 elegans, which has facilitated the use of C. elegans models to probe Aβ42-mediated dysfunction through the site-specific expression of human Aβ.366–368 While they do express an APP orthologue (Amyloid Precursor-like-1, AP-1), this lacks β-secretase sites and the nematode genome also does not encode for a β-secretase orthologue. As such, the effects of the expression of transgenic human-Aβ and the influence that inhibitory compounds or aggregation modifiers can be examined in isolation from APP processing.366 The aggregation of the Aβ peptide in disease model C. elegans is commonly monitored through staining and counting of amyloid inclusions in fixed worms or by measuring fitness biomarkers, such as number of body bends per units of time. The formation of amyloid deposits in the body muscle is associated with progressive paralysis, and the anti-

117 aggregation activity of small molecules has been probed by testing their ability to delay this phenotype.78,79,369

5.2 Primary Objectives

The primary aim of this work was to evaluate the capability of the lifetime sensor to probe the effects of inhibitor treatment on Aβ aggregation in live cell and whole organism AD models. To achieve this, the previously employed lifetime sensor protocol was adapted to monitor Aβ42 self- assembly in neuroblastoma cells and to observe how exposure to aggregation inhibitors can influence this process. The application of an Aβ-expressing C. elegans disease model was also investigated to monitor changes in fluorescence lifetime in association with peptide aggregation, and to test the ability of the hit compound MJ040X to inhibit this aggregation process.

5.3 Results and Discussion 5.3.1 Live Cell Studies

Previous studies within the Kaminski group have explored the internalisation and subsequent

253 aggregation of partially labelled Aβ42 using the amyloid aggregation lifetime sensor. In such work it was found that extracellularly added Aβ is taken into SH-SY5Y neuroblastoma cells endocytically, where fusion with lysosomes ensues. Aβ aggregation occcurs within these late endosomes and can be measured by means of fluorescence lifetime imaging. In the initial work, a difference in aggregation rate was observed between Aβ40 and the more aggregation-prone Aβ42, highlighting the sensitivity of the sensor in live cells. It was proposed that the non-invasive FLIM technique could be optimised to monitor the effects of small compounds on the Aβ42 aggregation process within live cells in real-time. A general schematic of the proposed assay scheme is shown in figure 5.2.

118

Figure 5.2: Schematic representation of the proposed cellular fluorescence lifetime sensor assay. i) Partially labelled Aβ42-488 is added to a sample of SH-SH5Y neuroblastoma cells, with or without the compound of interest. The peptide is kept at a low enough concentration so that aggregation does not occur in the extracellular medium. ii) Over time the monomeric peptide is taken into the cells by endocytosis. iii) In untreated control cells, the peptide self-assembles within late endosomal compartments. A decrease in fluorescence lifetime of the attached reporter dye is observed in association with the formation of amyloid aggregates. iv) It was proposed that if cells were treated with an active inhibitor, the aggregation process would be supressed, and the fluorescence lifetime would remain significantly higher than the untreated Aβ control.

Initial work aimed to minimise the total peptide concentration and peptide labelling density employed in the original studies (500 nM, 100% labelled). It was found that using 250 nM of Aβ42, at a labelling density of 50% HilyteFluor™488, was sufficient to monitor the uptake and aggregation of the peptide. A drop in fluorescence lifetime from 3291 ± 9 ps in the extracellular medium at 0 h, to 2926 ± 26 ps inside the cells following 24 h incubation, was observed (Figure 5.3). Previous studies that have tracked labelled Aβ internalisation using confocal microscopy and then quantified the formation of aggregates using agarose gel electrophoresis, have only been able to detect aggregation following the addition of 1 μM of Aβ to the extracellular medium.370

The detection of Aβ42 self-assembly at 250 nM in the cellular FLIM technique already highlights an advantage of this screening platform, as the sensitivity is more closely approaching that needed to detect physiologically relevant concentrations.

To verify if the fluorescence lifetime change detected with the untreated cells was truly associated with aggregation, and not an artefact of the fluorescence measurement, a control was carried out using ammonium chloride. This endosome-lysosome system acidification inhibitor has been noted by other members of the group to suppress cellular aggregation of internalised amyloidogenic

119 proteins.371 Addition of 10 mM ammonium chloride to the extracellular media reduced the drop in fluorescence lifetime observed after 24 h when the peptide was added alone (3041 ± 18 ps), suggesting that aggregation after internalisation could be prevented with the addition of exogenous compounds (Figure 5.3a).

Figure 5.3: Use of the lifetime sensor to report on the aggregation of extrinsically added Aβ42 in live cells. a) Bar diagram displaying the mean fluorescence lifetime values of the reporter dye in Aβ42-488 (250 nM, + - 50% labelled) treated cells with or without the presence of ammonium chloride (NH4 Cl , 10 mM). This weak base was used as a positive control, as it prevents the process of Aβ aggregation by blocking the acidification of the lysosomal compartments, where the aggregation of internalised Aβ has been shown to occur. A dramatic drop in fluorescence lifetime is observed in the control cells following 24 h incubation with the + - labelled peptide. The fluorescence lifetime of cells exposed to NH4 Cl remains significantly higher, indicating that the aggregation process has been suppressed. b) The fluorescence lifetime image of 50% labelled Aβ42- 488 monomers (250 μM) in cell culture medium, measured in a silicon well on a glass coverslide. Mean value = 3291 ± 9 ps. c) Confocal image and d) corresponding FLIM image of SH-SY5Y cells 24 h after the addition of Aβ42-488 (250 nM, 50% labelled) to the extracellular medium. After the incubation period, the cells were washed with cell culture medium and the fluorescence lifetime of the internalised Aβ42-488 was measured. Mean value = 2929 ± 26 ps. Scale bar = 20 μm. Plot shows mean ± SEM, n = 23–31 cells per condition, data analysed using a two-way ANOVA with a Bonferroni post-test; **p<0.01.

To test if small molecule inhibitors could affect the aggregation process in the cellular format, the known inhibitor EGCG was first employed. 2.5 μM of the compound was added to the extracellular medium at the same time as the Aβ42-488. This resulted in a final fluorescence lifetime value at 24

120 h similar to that obtained with the Aβ42-488 only cells (2912 ± 24 ps), suggesting no inhibitory effect. To further probe if inhibition could be achieved with EGCG, the cells were pre-incubated with the compound for 1 h prior to the addition of Aβ42. It was reasoned that if the uptake of Aβ42 was a receptor mediated process, that the peptide would be endocytosed faster than an equilibrium of the small molecule across the cell membrane had been established.253,370 Pre- treatment of the drug would therefore allow the compound more time to permeate into the cells, so that it was already present and able to exert its inhibitory activity when the peptide is internalised. It was observed that treatment of the cells with EGCG prior to the addition of Aβ42 significantly supressed the formation of aggregates with the cells (Figure 5.4). This inhibitory activity was observed at equimolar concentrations and was maintained for 48 h post Aβ42 exposure.

Figure 5.4: Monitoring the inhibitory activity of EGCG on Aβ42 aggregation in live cells using the lifetime sensor. a) Incubation of cells with the EGCG significantly diminishes the drop in fluorescence lifetime associated with peptide aggregation, following 12, 24 or 48 h incubation. b) Confocal and corresponding FLIM image of EGCG (2.5 μM) and Aβ42-488 (250 nM) treated cells following 24 h incubation. The drugged cells were incubated with the compound for 1 h prior to the addition of the labelled peptide. An average fluorescence lifetime of 3141 ± 6 ps was measured. Scale bar = 10 μm. Plot show mean ± SEM, n = 18-31 cells per condition, data analysed using a two-way ANOVA with a Bonferroni post-test; **p<0.01, ***p<0.001.

121

A drop in fluorescence lifetime after 24 h is observed even with the addition of the ammonium chloride or pre-treatment with EGCG (Figure 5.3a, 5.4a). This may indicate that even with the presence of inhibitory factors, a degree of aggregation still occurs. Alternatively, this decrease may be associated with changes in the spectroscopic properties of the dye as a result of changes in the local molecular environment. The fluorescence lifetime of a fluorophore can be affected by changes in pH, viscosity, refractive index, presence of ions and other factors within the cell.372 This sensitivity to changes in the microenvironment may account for the difference in fluorescence lifetime in monomeric Aβ42-488 species in the extracellular media (3291 ± 9 ps) and within the

+ - ‘inhibited’ late endosomal compartments (3041 ± 18 ps and 3031 ± 27 ps for NH4 Cl and EGCG treated cells respectively). It is worth noting that when attached to the peptide the fluorescence lifetime of the HilyteFLuor™ 488 dye is insensitive to changes in pH, as shown in previous studies employing this technique.253,254 The acidity of the late endosomal organelles is therefore not contributing to the initial changes in fluorescence lifetime observed in this case, however any combination of other extrinsic factors may be responsible for the observed change.

3200 *** *** *** ns * * 3000

2800

2600

Fluorescence(ps) Lifetime  42 A EGCG Aspirin MJ040 Myricetin Bexarotene Scyllo-inositol

Figure 5.5: Anti-aggregation activity of a selection of known inhibitors aggregation in live cells. Bar diagram displaying the mean fluorescence lifetime values obtained following 24 h treatment with the known small molecule inhibitors (bexarotene, EGCG, myricetin, scyllo-inositol) and a non-inhibitor (aspirin). The activity of hit compound MJ040 was tested to compare its cellular inhibitory activity against the known inhibitory molecules. Conditions: 250 nM Aβ42-488, 50% labelled, 2.5 μM compound, 24 h incubation. Plot show mean ± SEM, n = 18-30 cells per condition, data analysed using a one-way ANOVA with Dunnett’s multiple comparison post-test; *p<0.05, ***p<0.001.

122

The cellular FLIM assay system was validated with a small selection of known inhibitors (Bexarotene,78 myricetin44,189 and scyllo inositol332) and a non-inhibitor (aspirin138)(Figure 5.5). Such work showed that the aggregation process could be suppressed in the live cells with the addition of active inhibitory compounds, but was not a general response of adding any small molecule. The activity of the hit compound MJ040 was also tested. After 24 h an inhibitory effect was observed, however this compound did not achieve as high levels of activity as the other known inhibitors (3034 ± 32 ps for MJ040 vs 3102 ± 31 ps for EGCG). As discussed in section 4.2.3, this was attributed to the anionic nature of the compound. Gratifyingly, treatment with the methyl ester prodrug MJ040X afforded a stronger inhibitory activity (Figure 5.6).

Figure 5.6: The inhibitory effect of hit compound MJ040 and prodrug MJ040X on Aβ42 aggregation in live cells. a) Bar diagram displaying the mean fluorescence lifetime values obtained following 48 h treatment with MJ040 or MJ040X. The prodrug MJ040X shows a greater effect on reducing peptide aggregation, which is believed to be a consequence of its improved cell permeability. b) Confocal images and c) corresponding FLIM images of cells that were either untreated or pre-incubated with MJ040 or MJ040X for 1 h, prior to the addition of Aβ42-488. Mean average fluorescence lifetime of 2873 ± 19 ps, 2980 ± 21 ps, 3050 ± 17 ps for the control, MJ040 treated and MJ040X treated cells respectively. Scale = 20 μm. Conditions: 250 nM Aβ42-488, 50% labelled, 2.5 μM compound, 48 h incubation. Plot shows mean ± SEM, n = 50-70 cells per condition, data analysed using a two-way ANOVA with a Bonferroni post-test; *p<0.05, ***p<0.001.

123

Through the optimisation of the previously reported lifetime sensor Aβ uptake technique, an assay system for validating the effect of small molecule inhibitors in live cells was developed. Pre- incubation of the cells with the test compounds was deemed necessary to observe and inhibitory effect. By adding this step, supressed peptide aggregation that was maintained for 48 h post Aβ42- 488 exposure could be observed. The assay gives an early indication of cell permeability, which can be correlated with bioavailability – an important pharmacokinetic consideration for drug development.227 It also provides a means to test if strategic chemical modifications improve cellular activity, as evidenced by the increased activity of prodrug MJ040X compared to MJ040, the original hit compound. The lifetime sensor technique represents a robust means to obtain information about the self-assembly of Aβ42 and aggregation state at a particular time point, far beyond that achievable by mere localization of labelled protein or fixed cell aggregate staining techniques.

5.3.2 Caenorhabditis elegans Studies

The final stage in designing the unified lifetime sensor screening platform was the development of a protocol to monitor the inhibitory effects of hit compounds in a whole organism AD model.

For this, a Pmyo3:: GFP::Aβ42 C. elegans construct was employed. These worms were supplied by Dr. Della David at the German Centre for Neurodegenerative Diseases (DZNE) Tübingen. In this model system, the myo-3 promoter (Pmyo-3) is used to drive expression of GFP-Aβ42 (construct map in Appendix III). The muscle myosin gene myo-3 is turned on in post-mitotic embryonic body- wall muscle and as such fluorescently labelled Aβ42 is easily identified along the periphery of the worms (Figure 5.7). Similar to the HilyteFLuor™488 fluorophore dye previously employed, GFP can act as reporter to inform on the aggregation state of attached amyloidogenic peptides. Previous work using a C. elegans Parkinson’s disease model, where α-synuclein attached to yellow fluorescent protein (AS-YFP model) was expressed in the body muscle, showed an age dependent decrease in fluorescence lifetime in association with protein aggregation.252 It was hoped that a similar result could be achieved with this new Aβ42-GFP construct, and that any rescuing effects of the hit compound MJ040X could be monitored in a quantitative manner.

124

Figure 5.7: Alzheimer’s disease model Aβ42-GFP C. elegans. a) The worms express Aβ42-GFP in the myo3 body-wall muscle. Scale = 50 μm. b) Brightfield image showing the anterior ~30% of the worm. Only this section is imaged to avoid interference from the auto-fluorescence of the gut. Scale = 20 μm. c) Florescence lifetime image corresponding to panel b). A characteristic lifetime is observed for the GFP attached to non- aggregated Aβ42 on day one of adulthood (2769 ± 7 ps). Scale = 20 μm.

The worms were cultivated on a diet of an OP50 E. coli and were maintained at 20 °C according to widely used protocols.373 Compound treatment was achieved by adding solutions of the compound or vehicle control (10 μM, 0.05% DMSO) to the bacterially-lawned agar plates and allowing the solvent to dry off before adding synchronised worms. To avoid potential toxicity issues due to excessive drug absorption in relatively permeable larvae, the worms were initially treated with MJ040X from the first day of adulthood. The drugged growth media plates also contained 5-Fluoro-2′-deoxyuridine (FUDR, 75 μM), a thymidylate synthase inhibitor that prevents the growth of further offspring, thereby maintaining all the worms at the same age. Imaging was carried out at day 3, 6, 9, 12 and 15 of adulthood, and only the head region was focused on to avoid interference from the auto-fluorescence of the gut. Two technical repeats of each time point were performed, with 8-10 worms imaged per condition.

The GFP reporter showed a starting fluorescence lifetime in day 1 adult worms of 2769 ± 7 ps (Figure 5.8). This is within the range previously observed for GFP, which has been reported to vary between 2.5 and 3.3 ns as a function of the refractive index of its environment.374–376 A significant drop in fluorescence lifetime was observed at day 12 in the untreated worms (2698 ± 11 ps), signifying appreciable peptide aggregation. A comparable decrease was not observed in the M040X treated worms, where the fluorescence lifetime remained close to the initial starting value (2754 ± 10 ps). By day 15, however, a lower fluorescence lifetime was observed in both the treated worms and the untreated controls, indicative of the formation of aggregated species in both sample sets (Figure 5.8). This result suggests that at this concentration (0.7 μM) the compound can delay, but not completely inhibit the aggregation process in this AD whole organism model.

125

Figure 5.8: Use of the lifetime sensor to report on Aβ42 aggregation in a GFP-Aβ42 C. elegans disease model. a) Brightfield and b) corresponding fluorescence lifetime images of day 1 control (2769 ± 7 ps), day 12 control (2693 ± 11 ps) and day 12 MJ040X treated (2754 ± 10 ps) adult worms, expressing GFP-Aβ42 in the myo3 body muscle. A decrease in fluorescence lifetime is indicative of peptide aggregation and is observed at day 12 in the control worms. At day 12 in the treated worms, a significant change in fluorescence lifetime is not observed suggesting appreciable aggregation has not yet occurred. c) Bar diagram displaying the mean fluorescence lifetime values at each time point measured, following the initiation of drug treatment at day 1 of adulthood. A delay in Aβ42 aggregation with MJ040X treatment is observed. Scale = 20 μm. Plots show mean ± SEM, 2 repeats, n = 16-20, statistical analysis performed using a two-way ANOVA with a Bonferroni post-test; **p<0.01.

126

To further investigate the in vivo anti-aggregation effects of MJ040X, a population of worms were exposed to the compound from the L1 larval stage (Figure 5.9). Using this protocol, the fluorescence lifetime of the treated worms remained at the high initial starting values until day 15 (2769 ± 9 ps), the final day of the c. elegans lifespan measured. This observation suggests that MJ040X is capable of completely inhibiting the aggregation process when administered early enough in the development of the C. elegans. A breakdown of the full statistical analysis in given in Appendix IV.

Control Drugged at adulthood Drugged at larval stage

2800 ***

2760

2720

2680

2640

Fluorescence(ps) Lifetime Day 1 Day 6 Day 15

Figure 5.9: MJ040X treatment from larval stage completely suppresses Aβ42 aggregations in C. elegans disease model. Bar diagram displaying a comparison of the mean fluorescence lifetime values observed when the C. elegans were treated from day 1 of adulthood or from larval stage. Drug treatment from larval stage prevents aggregation until day 15, the final day of adulthood. Plots show mean ± SEM, 1 repeat, n = 16-20, statistical analysis was performed using a two-way ANOVA with a Bonferroni post-test; ***p<0.001.

The mechanism of drug action, drug uptake pathway and the effective drug concentration achieved within the worm are factors that must be taken into account when rationalising the difference in inhibitory effect when drug treatment is initiated at the larval stage or in early adulthood. As the GFP-Aβ42 is expressed from the embryonic stage, it is possible that during the 2 days taken to reach adulthood from synchronisation, the formation of aggregated nuclei has already begun. Whilst the expected concentration of these species would be too low to alter the overall average fluorescence lifetime value measured, such species could trigger the aggregation process in adult worms even in the presence of an inhibitory compound. This suggests that MJ040X functions, at least in part, at the primary nucleation phase, as treatment before any amyloid seeds have formed completely inhibits the aggregation process. The observation that late treatment delays the appearance of aggregates indicates that the compound may also exert an

127 inhibitory effect on the later stages on the aggregation process, but more experimental work is necessary to probe this idea.

C. elegans have robust physical and enzymatic defences that prevent exogenously applied drugs accumulating to effective concentration in the tissue.377 It has been shown that only 2% of pharmacologically active compounds can induce a robust phenotype in C. elegans when screened with a concentration of 25 μM.377 The stronger inhibitory activity following larval treatment may also be a result of increased accumulation of the drug in these worms. Drug uptake in C. elegans is achieved by three mechanisms; ingestion, uptake through the skin and uptake via exposed sensory neuronal endings.352 Given the increased permeability of the larvae, addition of the drug at this early stage may result in increased diffusion across the external cuticle, thereby amplifying the total concentration of drug that accumulates within the worm. Dose-activity relationships and investigations into the total concentration of accumulated drugs following treatment at difference stages could be examined in future work using HPLC analysis.377

5.4 Summary and Conclusions

The development of sensitive, robust and time-efficient methods to test the inhibitory activity of anti-aggregation hit compounds is imperative for validating biological activity, filtering and prioritising hits, and getting early insights into compound bioavailability. Current methods for achieving this in cellular and whole organism AD models are relatively scarce. Detecting the formation of aggregates in cells has previously been limited to fixed cell immunofluorescence and staining techniques, or through fluorescence localisation and subsequent invasive protocols to extract and quantify the aggregates formed.370 By adapting the FLIM protocol previously reported by the Kaminski group,253 a dynamic assay format capable of monitoring Aβ aggregation and the effect of inhibitory small molecule on this process was developed. Here, the sensitivity of spectroscopic characteristics of fluorophores in response to nanoenvironmental changes upon peptide aggregation are exploited, where self- quenching lowers fluorescence lifetime.255

Using this lifetime sensor technique, aggregation in live cells was observed with as low as 250 nM externally added Aβ42-488. It is estimated that the concentration of Aβ in human cerebrospinal fluid is in the low nanomolar range.296,349,350 Whilst the assay protocol developed here uses mid- nanomolar concentrations of the peptide, this is the lowest value reported in the literature that can provide a quantitative measurement of amyloid aggregation. The cellular FLIM screen was validated with a selection of known inhibitors and used to confirm the inhibitory activity of the hit

128 compound MJ040X. It is believed that the technique is amenable to high throughput drug screening. The use of microtitre plates would greatly accelerate the assay output. Furthermore, this live cell format could be used as the first step in the compound screening protocol, to identify compounds which may not be detected in an in vitro assay format due to the requirement of other cellular factors to function.

The relevance of measuring the aggregation of internalised Aβ is becoming increasingly clear with the growing evidence that Aβ accumulates in vesicles in neurons before the hallmark extracellular plaques are formed.338,378,379 The mechanism of toxicity of these intracellular Aβ species is unknown, but various links between concentration of internal and external Aβ pools suggest pathogenic interactions.338 An ‘inside out’ hypothesis even suggests that the internalisation of Aβ results in a high enough effective concentration to cause aggregation, which then results in cell death. This idea posits that the remaining intraneuronal Aβ species, after neuronal apoptosis, serve as seeds for the formation of amyloid plaques.338,370,380 A functional connection between the internal and external Aβ species, however, has not yet been experimentally validated. Tracing the kinetics and dynamics of Aβ aggregation and the effects that exogenously added molecules have on the process in live cells provides insight that cannot be gained from in vitro experiments, and may guide further study into the role of intracellular Aβ in the disease.

To explore the potential of the lifetime sensor for monitoring anti-Aβ aggregation activity of small molecules in whole organism system, a C. elegans model of Aβ42-mediated dysfunction was used.

The transgenic species employed expresses GFP-Aβ42 in body muscle, and a decrease in fluorescence lifetime of the GFP reporter was observed in association with amyloid formation over time. This model does not recapitulate all pathological aspects of the disease, but its simplicity is advantageous in terms of excluding confounding factors when probing the function of the small molecules of interest. Treatment with hit compound MJ040X from day 1 of adulthood resulted in a delay in peptide aggregation, whereas treatment from larval stage resulted in the complete suppression of aggregation for the entire lifespan. This implies that the compound can function, at least in part, at the primary nucleation stage of the aggregation process.

An important aspect which has not yet been studied in this work, but definitely requires further investigation, is the effect of the compound on fitness abnormalities associated with the peptide aggregation in the worms. Expression of amyloidogenic peptides in the neurodegenerative disease model worms is often associated with age related paralysis. Measuring the reduction in frequency in body bends with the aging of such worms is a useful parameter for monitoring the aggregation process and has been employed in confirming the activity of a host of potential drug

129

78,206,366 candidates. As the phenotype of the GFP:Aβ42 worm model used in this study has not yet been completely characterised, the rescuing effect of the compound on Aβ42 induced paralysis will be tested in a more widely validated Aβ42 worm model. This behavioural analysis is underway in the David research group (DZNE, Tübingen, Germany).

It is worth noting that a relatively small change in fluorescence lifetime was detected for the overall aggregation process in the C. elegans model. The use of different constructs in future studies may provide a larger dynamic range for monitoring the change in fluorescence lifetime and the effects of inhibitory small molecules. Changing the reporter fluorescent protein, for example, could be a promising strategy. The use of YFP in the aforementioned α-syn fusion protein variant showed a greater change in fluorescence lifetime upon amyloid formation,252 so the substitution of GFP for this reporter may be an avenue worth further exploration. Changing the promoter in the worm construct may also provide a larger dynamic range between non- aggregated and aggregated amyloid states.

A great advantage of the use of C. elegans, which has not been extensively discussed, is their conveniently small size. They are 1-1.3 mm in length, so it is possible to maintain approximately 100 worms per single well in a 96 well plate. As such, they are amenable to cost effective, high throughput whole organism compound screens.381,382 It is anticipated that the lifetime sensor protocol developed here could be applied to such high throughput screening strategies, which would greatly accelerate the speed and ease of the current whole organism screening approach. A HTS lifetime sensor screen would further benefit from the use of an alternative FLIM measurement techniques. TCSPC imaging (used in this study) takes 100-300 sec per worm, which greatly limited the total number of worms that could be imaged in this work due to time restraints on microscope usage. Alternatively, time-gated FLIM (TG-FLIM) takes only seconds to acquire an image. The use of this FLIM platform would therefore allow for the acquisition of a far greater sample size, which would more reliably reflect the population mean and potentially allow for a finer distinction between the in vivo activity of closely related inhibitors. High-throughput automated TG-FLIM is currently being developed in the Kaminski lab.

Overall, a single microscopy technique has been adapted from an in vitro test, to probe the activity of Aβ42 aggregation inhibitors in live cells and in a C. elegans disease model. This is not possible using any other technique. The lifetime sensor protocol described provides quantitative measurements of the aggregation process and the inhibitory action of small molecules in these live organisms, and may be employed to expedite drug discovery strategies for the development of new AD therapeutics.

130

Chapter 6 Inhibitory Activity of Polyphenol Derivatives

6.1 Introduction

Polyphenols, both natural and synthetic, represent a rich source of potential therapeutic agents for a wide range of human diseases.383–385 These compounds contain at least two phenol rings and can be largely classified into one of two groups, the flavonoids and non-flavonoids (Figure 6.1). Within these groups many subdivisions exist, and a broad spectrum of biological activities are associated with variations in the core structure.

Figure 6.1: A selection of flavonoid and non-flavonoid polyphenols. The basic flavonoid structure consists of aromatic rings A and B connected by the pyran ring C. Flavones, flavonols, flavanones, flavonols, isoflavones, aurones and chalcones all fall into this class or are derivatives thereof. Benzoic and cinnamic acid derivatives, stilbenes and curcuminoids are all examples of non-flavonoid polyphenols.

Polyphenols have shown great potential for the development of AD therapeutics and a variety of flavonoid and non-flavonoid inhibitors have been tested as AD candidate therapies in clinical

131 trials.383,386–388 Many polyphenols display Aβ anti-aggregation effects and can reduce the amyloid burden associated with AD.389,390 Due to their metal-chelating and antioxidant activity, polyphenolic compounds have also shown promise for reducing the oxidative damage associated with the disease, as well as other neuroprotective effects.333,391,392 Figure 6.2 shows a selection of previously identified Aβ aggregation inhibitory polyphenols, some of which have already been described in Section 1.6.1.1. EGCG, a gallic acid substituted flavanol and the primary polyphenolic component of green tea, shows powerful antioxidising and anti-aggregation activity with many amyloidogenic proteins.393–396 Curcumin displays inhibitory activity against the self-assembly of Aβ and α-synuclein in various assay formats and has been shown to reduce amyloid load when injected peripherally in Aβ-related disease mouse model.178,181,397 Characterisation of its binding interaction with Aβ has also guided many other drug design strategies.203 The stilbenoid resveratrol has excited great interest in recent years, not only because of its proposed anti- aggregation activity, but also due to its ability to ameliorate the damage caused by oxidative stress.389,392 The inhibitory activity of nordihydroguaiaretic acid (NDGA) has more recently been acknowledged, and the details of how this dihydro-stilbene interrupts the aggregation process is still under review.398,399 Kaempferol, myricetin, morin, quercetin and fisetin are all flavonol inhibitors, which display varied levels of activity associated with the quantity and positioning of their hydroxyl group substituents.333,389 It is worth noting that although the flavonoids derivatives are all believed to exert their modularity effect on the Aβ42 aggregation process by directly binding with the peptide, the residues targeted and binding interactions made by various polyphenol scaffolds differ.168 Consequently structure-activity relationship (SAR) or modes of inhibitory action cannot be assumed similar for different scaffolds.

132

Figure 6.2: Polyphenolic Aβ aggregation inhibitors. EGCG inhibits the aggregation of many amyloid peptides, including Aβ, α-synuclein and huntingtin. Curcumin is a powerful antioxidant and anti-aggregation inhibitor. NDGA inhibits Aβ aggregation, and has been proposed to disaggregate pre-formed fibrils. Resveratrol can inhibit the aggregation process and scavenge reactive oxygen species. (+) Taxifolan, a catechol-type flavonoid, is believed to inhibit aggregation in its oxidised form, by means of a site specific interaction with lysine residues in the peptide. The inhibitory activity of the flavonols kaempferol, myricetin, morin, quercetin and fisetin depends on the quantity and positioning of the hydroxyl groups. Flavanols (+)- catechin and (-)-epicatechin display milder anti-aggregation activity than their flavonol counterparts. EGCG: (-)-epigallocatechin gallate, NDGA: Nordihydroguaiaretic acid.

133

6.1.1 Structure-Activity Relationship

Many SAR studies investigating the Aβ aggregation inhibitory activity of flavonoid libraries have been undertaken to examine the structural requirements necessary for activity. Such work has revealed that complex interplay between the number and positioning of hydroxyl groups, electron withdrawing ability and substitution patterns of functional groups, planarity, and the length and flexibility of the linkers joining the phenolic rings.189,333,389 In general, isoflavonoids exert the mildest inhibitory effect, as C2 positioning of the B-ring seems essential for appreciable activity.400 Planarity of the inhibitor also appears important for maximising interaction with the peptide. This can be observed through a comparison of the activity of flavonols quercetin and morin with flavanols catechin and epicatechin, where the former exert a greater inhibitory effect.389 Rotation of the hydroxyphenyl ring in the flavonols allows both rings to be positioned in the same plane. The flavanols, however, contain chiral centres that diminish the molecular planarity and consequently the inhibitory activity. The absence of the keto group and associated loss of potential interactions in these compounds may also play a role in the reduced activity observed.390

When investigating the anti-amyloidogenic effects of hydroxyl substitution patterns of several commercially available fisetin analogues, Akaishi et al. noted that the positioning of the hydroxyl groups on such compounds can result in either an inhibitory or acceleratory effect on Aβ42 fibrillogenesis. They observed that 3’,4’-dihydroxyl groups, but not 3- or 7- hydroxyl groups are necessary for inhibitory action.333 It was, in fact, observed that flavonols lacking the 3’,4’- dihydroxyl groups and instead containing either none or just one B-ring hydroxyl group results in the enhancement of peptide aggregation. In such examples, 3,3’,7-trihydroxyflavone enhanced fibril formation to the highest extent. This highlights the complexities associated with rationally designing Aβ flavonoid inhibitors. It is further worth noting here that inhibitory activity of the penta-hydroxylated quercetin was higher than that of fisetin, which has four hydroxyl groups. Myricetin, the hexa-substituted analogue showed the highest activity, indicating a positive correlation between the number of hydroxyl groups and inhibitory activity. This finding has been supported by many studies, whereby increased hydroxylation is associated with increased anti- aggregation activity.189,401,402

The inhibitor curcumin has been subject to extensive SAR studies, which have shown strict requirements for inhibitory activity.203,403 Such work has focused on three prominent features, namely the number of aromatic rings, substitution patterns and the length and flexibility of the linker. To establish the contribution of each of these features a series of curcumin analogues with systematic variations were tested.203 It was found that two co-planar aromatic groups are essential

134 for appreciable anti-aggregation activity. To achieve strong inhibitory activity these must also be joined by a linker containing minimal freely rotatable carbon centres, with an optimal length of 9- 16 Å. Such stringent requirements suggest that this compound targets two binding sites within the peptide, and that concurrent interaction of the compounds at both sites is needed for significant activity. Hydroxyl groups were again observed to be critical for inhibition, with modification of the constituent 3-methoxy 4-hydroxy substitution pattern to a 3,4 dimethoxy substitution resulting in a loss of activity. 3,4 dihydroxy substitution, however, is tolerated, as is seen in the inhibitory NGDA and resveratrol.

6.1.2 Molecular Binding Studies

Whilst various studies have been carried out to inform on the structural requirements necessary for anti-aggregation activity, there is a lack of knowledge regarding how polyphenol compounds exert their inhibitory effect at the molecular level. Elucidating the interaction sites of inhibitors with Aβ is of paramount importance for characterising key binding motifs, guiding future drug design strategies and understanding the underlying mechanisms of action.

In vitro experimental techniques for such investigations are still limited. Solid state NMR has been employed, but only for a selection of inhibitors, where binding to small localized amino acid regions has generally been observed.176,187,404 With regards to the flavanols, NMR studies have indicated that there are different binding mechanisms for catechols (benzenediol isomers containing vicinal hydroxyl group) and non-catechols. Catechols, such as (+) taxifolan, have been shown to function through auto-oxidation to form o-quinone. This functionality then interacts with the Lys16 or Lys28 residues of Aβ to inhibit the self-assembly process.405 Conversely, the non- catechol flavonols do not undergo such auto-oxidation reactions, and are instead believed to make hydrophobic interactions with the peptide in either its monomeric or aggregated form.189 These findings emphasise the fact that a ‘one size fits all strategy’ cannot be employed in designing and modifying inhibitory polyphenols.

Additional experimental techniques that have been employed to elucidate inhibitory mechanisms of action include combinations of spectrophotometry, spectrofluorimetry, AFM, SDS gel electrophoresis and surface plasmon resonance. A selection of these techniques were used in studies aimed at deducing how myricetin functions.406 Such work indicated that the compound exerts its inhibitory activity by reversibly binding to fibrils rather than monomers. NMR studies, on the other hand, have indicated that myricetin binds preferentially to monomers.187 This highlights

135 that there are still inherent limitations in deriving mechanistic information from in vitro experimental data at this point. Various technique should, therefore, be employed when investigating the binding mechanisms of hit compounds.

6.1.3 Antioxidant Polyphenols for Alzheimer’s Disease Treatment

Oxidative stress is an imbalance between the production of harmful free radicals and the body’s ability to neutralise these reactive species by means of various antioxidising strategies. As discussed in Section 1.3.4, extensive evidence suggests a prominent role of oxidative stress in AD.83,89 The uncontrolled production of reactive oxygen species (ROS) may not necessarily be an initiating event in the onset of the disease, but ROS are strongly believed to exacerbate the progression of neurodegeneration. As such, the development of compounds that are able to scavenge excess ROS and provide protection against oxidative damage are highly sought after.386,407

An antioxidant is an endogenous or exogenous compound that “when present in low concentrations compared to that of an oxidizable substrate, significantly delays or inhibits the oxidation of that substrate”.408 Phenolics compose a major class of plant derived antioxidants.169,409 Within this class catechins, stilbenes, curcuminoids and chalcones have all demonstrated significant antioxidant ability, and have provided protection against neurotoxic insult in many AD models.169,407 Antioxidants can function directly, by means of quenching ROS activity, or indirectly by preventing the formation of ROS or promoting endogenous antioxidant capacity.407,410 Some natural flavonoids exert their antioxidising effect through a combination of pathways. Green tea derived antioxidants, including EGCG and other catechins for example, have been suggested to transfer electrons to ROS-induced radical sites on DNA to prevent oxidative

DNA modifications, chelate metal ions to reduce H2O2 production via the Fenton reaction and also to suppress the propagatory chain reaction of lipid peroxidation.169

As the polyphenolics display such pleotropic antioxidant behaviour, it is difficult to make general correlations between structural features and indirect antioxidising activity. With regards to structure-ROS scavenging ability, however, a large number of polyphenols of different classes (flavonols, flavones, chalcones, etc.) have been evaluated and some broad SAR conclusions drawn.409 In such work, differences in radical quenching ability were attributed to positional differences in hydroxylation and methoxylation. Ortho dihydroxyl groups were identified as the most important structural feature for high activity in all tested phenolic compounds. The

136 importance of a central double bond was also noted, a finding that has been supported by in other biological assays with flavonoid derivatives.389,411

EGCG, resveratrol and curcumin, among other phytochemicals, have been employed in clinical trials to test the activity of antioxidants as potential AD therapeutics.386 Such studies have provided conflicting results, quite possibly due to metabolic instability and poor bioavailability of the compounds tested. A number of small participant studies has provided promising initial results, but much larger, randomised, placebo controlled investigations are required before any conclusive results can be drawn.386,407 These early insights, however, inspire enthusiasm that with continued strategic medicinal chemistry approaches therapeutically active AD targeted antioxidants can be developed.

6.2 Primary Objectives

The aim of this project was to identify and develop small molecule inhibitors of the process of Aβ42 aggregation. A selection of polyphenolic compound libraries, synthesised by previous members of the Spring research group, was screened using ThT fluorescence assays. Three libraries of interest were identified, with a heteroaromatic chalcone derivative library seen to display the most promising inhibitory activity. Hit validation, by means of biophysical and cell-based techniques, was used to investigate the therapeutic potential of lead compounds.

6.3 Results and Discussion 6.3.1 Initial Screening

The flavonoid libraries screening in this work were synthesised by previous Spring group members, Drs Anthony Sum and Max Sum, to expand on the promising activities of previously reported polyphenolic derivative, to investigate the biological relevance of increased structural, appendage and linkage diversity, and to probe uncharted regions of chemical space. Figure 6.3 shows representative molecules from a selection of the compound collections screened. Three libraries of interest were identified: triazole-bridged flavonoid dimers, phenyl-linked biaurones and heteroaromatic chalcones derivatives.

137

Figure 6.3: Representative compounds from the flavonoid derivative libraries screened. The libraries were synthesized by members of the Spring research group to expand on the scope of scaffold and linkage diversity observed in natural biologically active polyphenolic derivatives.

6.3.1.1 Triazole Linked Biflavonoids

Synthesis of the triazole linked biflavonoid library was inspired by previously published data, whereby biflavonoids were shown to exert stronger biological activity than their mono-flavonoid counterparts.412 Triazole linkers were used to join the biflavonoid units, to take advantage of the proven biological relevance and metabolic stability of these motifs.413 A modular synthetic strategy was employed, resulting in a small but structurally diverse library containing 15 members that incorporated 6 different structural subclasses of flavonoids.414 Inhibitory activity was assessed using a ThT fluorescence assay (Section 1.5.2). In this assay format the plateau of the Aβ42 aggregation profile reflects fibril mass content,206,405,415 and the percentage inhibition achieved by compounds is expressed as the relative change in final ThT fluorescence value. Of the 15 compounds synthesised, five were very poorly soluble, five overly fluorescent and two inactive.

The compounds that displayed inhibitory activity against the aggregation of Aβ42 are shown in figure 6.4. Moderate activity, comparable to the aforementioned flavonol inhibitor morin, was

138 observed for compounds 6.1 and 6.2. Whilst these initial results were interesting, the issues in solubility observed in many library members discouraged any continued investigations and it was instead decided to focus on alternative libraries.

Figure 6.4: Inhibitory activity of the triazole-bridged flavonoid dimers against the aggregation of Aβ42. Percentage inhibition of Aβ42 aggregation achieved by three of the triazole-linked flavonoid dimers relative to that of Aβ42 alone, as measured by means of ThT fluorescence assay. 100% represents complete inhibition of aggregation (complete reduction in ThT fluorescence) and 0% shows no inhibition. Both 6.1 and 6.2 display higher inhibitory activity than the previously published inhibitor morin. Conditions: Aβ42 10 μM, ThT 20 μM, compound 50 μM, mean ± SEM, n = 3. Figure adapted from Sum et al. 414

139

6.3.1.2 Phenyl Linked Biaruones

The second library that showed promising inhibitory activity was the phenyl linked biaurones (Figure 6.5).416 Aurones constitute a subclass of flavonoids, which rarely occur in nature, but have been shown to display promising biological activity in a range of human diseases.417 Radiolabelled aurones have previously been employed for in vivo plaque labelling,418 but until recently their use as aggregation inhibitors had not been reported.419 The aforementioned study, whereby biflavonoids displayed greater inhibitory effects than monoflavonoids,412 suggested the presence of two distinct binding pockets in the peptide, thereby prompting optimisim for success with this biaurone library. Again the activity of this library was assessed by monitoring the final ThT fluorescence value (representative of fibril mass content) in the presence of each compound, and

IC50 value calculation was performed by taking the slope of the ThT aggregation curve to represent the aggregation rate.405 Of the 17 compounds screened, 3 were poorly soluble, but the rest displayed a range of inhibitory activities (Figure 6.5).

The most potent compound, 6.15 was calculated to have an IC50 of 15.8 ± 3.7 μM. This is comparable to that calculated for the reference inhibitor EGCG (6.4 ± 0.7 μM) and is therefore a promising starting value for an early hit compound. The extraction of SAR information was difficult, owing to the small sample size of the library. However, a general trend was observed with regards to the effect of the substitution pattern around the phenyl linker. Para-substituted phenyl linkers were consistently associated with higher levels of inhibition than meta-substituted linkers, and the activity of the diphenyl linkers were in between these two extremes. This is evident when comparing compounds 6.8 vs 6.15 and 6.11. It is worth noting that an equivalent library of biflavones – that is one in which the substitution and linking patterns were analogous but a core flavone unit was used – showed negligible inhibitory activity. This indicates strict requirements in the size or planarity of the heteroaromatic core structure for the particular binding site targeted here. A similar result was also previously found in the development of amyloid imaging probes, whereby radiolabelled aurones were found to bind to Aβ42 with a greater affinity than their flavone analogues.418

140

Figure 6.5: Inhibitory activity of the phenyl linked biaurones against the aggregation of Aβ42. a) Aβ42 aggregtion profiles with or without biaurones, as monitored by ThT fluorescence. The curve for Aβ42 (blue) represents the time course of Aβ42 aggregation in the absence of any small molecules. Modified aggregation profiles are evident with the addition of the compounds. b) Percentage inibition of Aβ42 aggregation observed with the biaurones and known inhibitors morin, myricetin and EGCG. The values are calculated by the saturation phase of the aggregation profiles shown in panel a). 100% represents complete inhibition of aggregation (complete reduction of ThT fluorescence) and 0% shows no inhibition. c) Stucture of the active biaurones assayed. Conditions: Aβ42 10 μM, ThT 20 μM, compound 50 μM, mean ± SEM, n = 3. Figure adapted from Sum et al.416

As a result of the small sample set, it was also difficult to make definitive comments about how ring substitution influences the inhibitory activity of this scaffold. Given the previously discussed enhancement in inhibitory activity with the incorporation of hydroxyl groups, it is believed that the biaurone library should be elaborated to investigate the effects of adding such funtionalities. Ideally monomeric units could be tested individually to quickly probe the effects of different

141 substituion on the aromatic rings. Such monomeric units of the active biaurones were not available at the time of initial screening. The effects of combining active monomers, either symmetrically or not, may then provide some insight into the modes of binding interaction. Systematic variation of the substitution patterns could be used to identify key binding motifs and lead to development of even more potent inhibitors.

6.3.1.3 Chalcone Derivatives

The third library of interest identified as potential Aβ aggreation inhibitors consisted of methoxylated and heteroaromatic chalcone derivatives (Figure 6.6). Chalcones are precursors in the biosynthesis of flavonoids and are characterised by an open chain α,β unsaturated ketone linking the A- and B-rings. These compounds display a broad spectrum of activity, predominantly anti-inflammatory, anti-infective, anti-cancer and antioxidative effects.411,420,421 Whilst the activity of chalcones and their derivatives as potential therapeutic for AD has been briefly reported, such studies have focused on their radical scavenging ability and antioxidative properties.422,423 Investigation into their Aβ anti-aggregation activity has been relatively limited.

Figure 6.6 shows a selection of the active chalcone derivatives identified in initial screening attempts with the library. Inhibitory activity was again assessed using a ThT fluorescence assay and the percentage inhibition achieved by compounds is expressed as the relative change in final ThT fluorescence value.206,405,415 It is evident that a variety of heteroatoms give pronounced inhibitory activity at the B-ring position, as has been observed with chalcones in other biological

424,425 assays. Approximate IC50 values for aggregation inhibition were obtained for several of the compounds using the initial rate of Aβ42 aggregation in the presence of 5, 10, 25 and 50 μM inhibitor. These IC50 values cannot be considered to be quantitatively accurate, as they are derived from small data sets; however, such values were deemed sufficient to make relative comparisons of inhibitory action between library members. The resulting IC50 graphs, showing normalised rate as a function of compound concentration, and the associated IC50 values for the four most active compounds are shown in figure 6.7. Through such work TJS285 (6.25) was identified as the most active inhibitor at this stage in the study, and was taken forward for further hit validation studies.

142

Figure 6.6: Inhibitory activity of the chalcone derivative library against the aggregation of Aβ42. a) Aβ42 aggregtion profiles in the absence or presence of active chalcone derivatives, as monitored by ThT fluorescence. The curve for Aβ42 (blue) represents the time course of Aβ42 aggregation in the absence of any small molecules. Modified aggregation profiles are evident observed with the addition of the compounds. b) Percentage inibition of Aβ42 aggregation with the selection of the biaurones and known inhibitors morin, myricetin and EGCG. The values are calculated by the saturation phase of the aggregation profiles shown in panel a). 100% represents complete inhibition of aggregation and 0% shows no inhibition. c) Stucture of a selection of the chalcone derivatives assayed. Conditions: Aβ42 10 μM, ThT 20 μM, compound 50 μM, mean ± SEM, n = 3.

143

Figure 6.7: IC50 graphs used to calculate approximate IC50 values for select chalcone derivative inhibitors. The slope of the ThT fluorescence aggregation curve at the growth phase was taken to represent the aggregation rate. For each compound the normalised rate, compared to untreated Aβ42, was obtained at a range of inhibitor concentrations. These values were plotted to give the graphs shown. TJS285 was shown to exert the strongest inhibitory activity against the aggregation of Aβ42 and was carried forward for further 푦 1 hit validation investigations. IC50 values were calculated on GraphPad using the equation: = 푥 푦푚푎푥 1+ 퐼퐶50

6.3.2 Hit Validation

To validate the anti-aggregation activity of TJS285, and to investigate the morphology of the Aβ42 species formed in its presence, TEM images of the Aβ42 species formed following incubation with the compound were taken and immunoassay dot blots with the same samples prepared. TEM images of TJS285-treated samples revealed the presence of very few individual fibrils, which did not appear to clump like the control Aβ42 (Figure 6.8). Furthermore, the treated samples did not give a strongly positive result when tested with the toxic oligomer-sensitive A11 antibody (Figure 6.8). This suggests that the compound does not redirect the aggregation pathway to one in which toxic oligomeric structures are formed.

144

Figure 6.8: Investigations into the morphology of Aβ42 species formed in the presence of TJS285. a, b, c) TEM images of Aβ42 aggregates formed after 7 days of incubation with and without the addition of TJS285. a) Dense clusters of fibrils are observed in the Aβ42 control sample. b) With TJS285 treatment, very few fibrils are observed. c) Zoomed in section of image b) showing finer structure of fibrils formed in the presence of TJS285. Scale bar = 500 μm. d) Time course dot blots examining the potential formation of toxic species, using oligomer specific A11 antibody. Oligomeric structures were present following, 1,3 and 5 days of incubation in control samples. Low quantities of A11 sensitive species were detected following incubation with the TJS285. Positive readouts were obtained for all samples using the 6E10 control antibody, which detects all Aβ species. Conditions: 20 μM Aβ42, 100 μM TJS285, 1-7 day incubation.

To assess the ability of the hit compound to inhibit the Aβ42aggregation process in cells, the previously developed cellular lifetime sensor assay was used (Section 5.3.1). After 24 h, the fluorescence lifetime of internalised labelled Aβ42 was seen to drop from the high monomeric (3352 ± 11 ps) to a lower value associated with peptide aggregation (2926 ± 26 ps), in control cells. A high fluorescence lifetime (3120 ± 27 ps), indicative of supressed amyloid formation, was maintained in the TJS285 treated cells. This inhibitory activity in the live cells was statistically significant and comparable to that of the known inhibitory small molecules used to validate the assay platform (Figure 6.9).

145

3200 *** *** *** *** * 3000

2800

2600

Fluorescence Lifetime (ps) Lifetime Fluorescence  42 A EGCG TJS285 Myricetin Bexarotene Scyllo-inositol

Figure 6.9: Cellular lifetime sensor assay investigating the inhibitory activity of TJS285 against the aggregation of Aβ42. Bar diagram displaying the mean fluorescence lifetime values of Aβ42-488 (250 nM, 50% labelled) following 24 h incubation with known inhibitors or hit compound TJS285 (2.5 μM). Values of 2926 ± 26 ps and 3120 ± 27 ps were observed for the Aβ42 control and the TJS285 treated sample respectively. Plot shows mean ± SEM, n = 18-30, data analysed using a one-way ANOVA with Dunnett’s multiple comparison post-test; *p>0.5, ***p>0.001.

Whilst the fluorescence lifetime value obtained for TJS285 was promising, a reduction in the number of viable cells was seen in the samples treated with the compound. To further investigate this observation, MTT cell viability assays were carried out. Increasing concentrations of the compound were added to SH-SY5Y cells and a dose-dependent increase in toxicity was observed (Figure 6.10a). The cytotoxicity of TJS285 was not entirely unexpected as an α,β unsaturated ketone functionality is present in the molecule; the β-carbon atom of this Michael acceptor would be expected to be the preferred site of attack for soft nucleophiles such as the amino acid residues of proteins, resulting in covalent modification of these structures.426,427 In many chalcone drug development studies, the presence of this reactive α,β unsaturated carbonyl group has raised concerns about the about long term safety effects of this compound class.428 In vitro mutagenic and clastogenic effects have been observed with various chalcones. In some cases, however, this unwanted reactivity is not evident in closely related analgoues.420 For this reason, a selection of other compounds form the chalcone derivative library were also tested with cell viability screens (Figure 6.10b). It was found that the majority of compounds did not result in a significant loss in cell survival, as TJS285 did. This suggested that alternative hits could be identified through strategic expansion of the test library.

146

Figure 6.10: Cell viability assays showing the effect of the chalcone derivative library compounds on cell survival. a) The addition of increasing concentrations of TJS285 results in a dose-dependent reduction in cell survival, indicating that the compound has cytotoxic effects. b) A statistically significant decrease in cell viability is not observed upon addition of the majority of the other compounds in the chalcone derivative library (2.5 μM), suggesting that a non-cytotoxic hit compound can be developed. Cell survival was calculated using an MTT cell viability test, using SH-SH5Y cells that were incubated with the test compound for 48 h prior to viability measurement. Control indicates the viability of untreated cells, and is set as 100%. Statistical analysis performed using a one-way ANOVA with Dunnett’s multiple comparison post-test; **p>0.1, ***p>0.001.

6.3.3 Library Expansion and SAR

Following the observation that the lead inhibitory compound TJS285 causes cell death, whereas other compounds from the chalcone derivative library do not, attention was turned back to the library to investigate if any conclusive SAR could be generated to guide the development of more suitable lead compounds. Initial comparison of the activity observed with the heteroaromatic chalcone derivatives indicated some general patterns, but it was deemed necessary to synthesize additional compounds for a more comprehensive SAR overview. Compounds were formed by means of Claisen-Schmidt adol condensation of an aldehyde and ketone by base catalysis (Scheme 6.1), and were chosen to incorporate moieties which could potentially increase activity or provide insight into the features necessary for activity. Table 6.1 shows the experimental details of all the compounds synthesised in this work. Only two methoxylated compounds (6.35 and 6.36) were synthesized, and further investigations into this library were not undertaken, as such structures have already been subject of various studies for antioxidant and anti-inflammatory activity.429 The full methoxy chalcone library and their associated Aβ42 inhibitory activities are shown in Appendix V for reference. Table 6.2 shows the inhibitory activity of the entire heteroaromatic chalcone library, with newly synthesised compounds labelled in green.

147

Scheme 6.1: Synthesis of chalcones via the Claisen-Schmidt aldol condensation.

Table 6.1: Experimental details for the compounds synthesised in this work.

Compound Structure Experimental details Yield

52% 6.25 (TJS285) EtOH, piperidine, reflux, 48 h

EtOH, piperidine, reflux, 48 h 31% 6.26 (SC017)

EtOH, piperidine, reflux, 48 h 49% 6.27

EtOH, piperidine, reflux, 24 h 73% 6.28

EtOH, KOH, rt, 48h 52% 6.29

EtOH, KOH, rt, 48 h 30% 6.30

EtOH, piperidine, reflux, 48 h 50% 6.31

6.32 EtOH, piperidine, reflux, 48 h 40%

148

Compound Structure Experimental details Yield

EtOH, piperidine, reflux, 72 h 17% 6.33

EtOH, KOH, rt, 72 h 27% 6.34

EtOH, KOH, rt, 48 h 75% 6.35

EtOH, KOH, rt, 72 h 26% 6.36

Hydrogenation of TJS285 60% 6.37 MeOH, 2 mol% PdO2, H2, rt, 24 h

149

Table 6.2: Percentage inhibition of Aβ42 aggregation achieved by the heteroaromatic chalcone library. Inhibition was calculated from the plateau of the ThT fluorescence curve in the presence of the compound, relative to that of Aβ alone, and is shown as mean ± SEM from three independent assays. Compounds with a single percentage inhibition value were only screened once. * 6.37 contains the same substitution pattern as hit compound TJS285, but the central double bond was reduced, resulting in a drop in activity. Compounds newly synthesized for this work are coloured in green.

Number B-ring R2 R3 R4 R5 R6 % Inhibition 6.38 H H Br H H 25

*6.37 OMe H OMe H OH 37 ± 2 6.39 H H OMe H OH 59 ± 3 6.40 H OMe OH H H 71 ± 4 6.26 H H OH H OH 85 ± 2

6.25 OMe H OMe H OH 88 ± 3

6.41 H H Br H H 0 6.27 H H OMe H OH 29 ± 1 6.42 H O-CH2-O H OH 31 ± 4 6.28 OMe H OMe H OH 36 ± 2 6.21 OMe H H H OH 57 ± 2 6.43 H OMe OH H H 67 ± 5

6.44 H O-CH2-O H OH 43 ± 4 6.45 H H H H OCHCH 44 6.46 H Br H H OH 52 ± 4 6.20 H H H H OH 52 ± 5

6.23 H OMe OH H H 75 ± 2

6.29 H H H H OH 39 ± 4

6.30 OMe H OMe H OH 55 ± 5 6.18 H H OH H OH 66 ± 6

6.17 H O-CH2-O H OH 73 ± 3

6.47 H H H H OH 18 6.31 H H H H Br 22 ± 4 6.48 H H H Br H 27 6.32 H H OMe H OH 49 ± 7 6.19 H OMe OH H H 61 ± 3

6.33 OMe H OMe H OH 58 ± 3 6.22 H H OH H OH 62 ± 11

6.49 H H OMe H OH 66 ± 11

6.34 H H H H OH 0 6.50 H Br H H OH 12 6.51 H H OMe H OH 29 6.16 H OMe OH H H 31 ± 6 6.52 H O-CH2-O H OH 44 ± 2

150

Each of the heteroaromatic rings tested at the B-ring position displayed some level of Aβ42 anti- aggregation activity. A general trend in increased activity was observed in the order: pyridine < N- methylated pyrrole < pyrrole < furan < N-methylated indole < quinoline ≈ indole (Figure 6.11). Indoles at the B-ring resulted in the highest inhibitory activity. N-methylation of indole resulted in a slight drop in activity, as did N-methylation of pyrrole, although the sample size for comparison here was much smaller. It is worth noting that the quinoline derivatives showed similar, if not better, activity than the indole chalcone derivatives (compare 6.40 vs 6.23). However, the quinolines were considerably more difficult to handle. These were more susceptible to oxidation or degradation during purification and attempts to make certain analogues for comparison were unsuccessful. Reduction of the linking olefin resulted in diminished anti-aggregation activity (compare 6.37 vs 6.25).

Correlating A-ring functionalisation with inhibitory activity was somewhat difficult, as the same substitution in analogues differing only in the heteroaromatic B-ring did not always result in the same pattern of activity. For example, the 6’-hydroxylated pyridine chalcone 6.34 displayed no activity, but incorporation of the dioxolane ring (6.52) gave rise to a 44% inhibitory effect. The inclusion of this dioxolane ring in the quinoline derivatives, however, resulted in diminished activity relative to the 6’-hydroxylated quinoline alone (compounds 6.20 and 6.44). Although the synthesis of more compounds would be needed for systematic analysis of the inhibitory effects achieved with substitution at each position around the A-ring, some general trends were still observed (Figure 6.11). Bromine substitution was generally associated with low activity. Similarly, 6 hydroxylation did not confer high activity. This may be due to steric hindrance at this position preventing the group from forming interactions with the peptide. Compounds that were dihydroxylated at the A-ring displayed higher inhibitory activity than compounds with either one hydroxyl group, or one hydroxyl and one methoxy group.

151

Figure 6.11: Aβ42 anti-aggregation structure-activity relationship of the heteroaromatic library. Substitution of the A-ring with bromine or a hydroxyl group at the 6’ position alone contributed little to inhibitory activity. Dihydroxylation of the A-ring results in a higher activity than substitution with one hydroxyl and one methoxy group. Reduction of the double greatly diminishes inhibitory activity. A variety of heteroaromatics are tolerated at the B-ring position. Of these, quinoline or indole provides the highest anti- aggregation effects.

Following library expansion, a second generation hit compound SC017 (6.26), an indole chalcone derivative, was identified. Due to the versatility and ease of functionalisation of their “privileged” structure, indoles are accepted as biologically relevant pharmacophores and have represented a valuable starting point in various medicinal chemistry campaigns.430,431 Indole chalcones linked at the C3 position have not previously been reported for use as Aβ aggregation inhibitors. They have, however, previously shown anti-inflammatory activity through inhibition of the COX1 and COX2

432 enzymes, with IC50 values in the low micromolar range. The effects of the library discussed here in anti-inflammatory assays have not been investigated yet, but this activity may be promising in terms of developing a multifunctional AD therapeutic, as the activation of inflammatory pathways is observed with AD pathogenesis.433,434 Indole-chalcone derivatives linked at the C5 have been shown to display Aβ binding activity in the development of radiolabelled imaging probes.435 However, the binding mechanisms here have not been extensively studied, providing little evidence to help deduce the interaction mechanisms of the heteroaromatic hits identified in this work.

152

The Aβ42 anti-aggregation activity of SC017 was evaluated using ThT fluorescence assays.

Quantitative IC50 calculations showed that SC017 has a similar inhibitory activity to TJS285 (4.4 ± 0.4 μM and 4.5 ± 0.4 μM respectively; Figure 6.12). This value is below that observed for EGCG (6.4 ± 0.7 μM), a phase III clinical candidate, thereby indicating that the compound displays inhibitory activity in a therapeutically relevant range.

Figure 6.12: IC50 graphs used to calculate IC50 values for EGCG, TJS285 and SC017. The slope of the ThT aggregation curve was taken to represent the aggregation rate. For each compound the normalised rate at a range of inhibitor concentrations was plotted to form the graphs shown. IC50 values were calculated on 푦 1 GraphPad using the equation: = 푥 푦푚푎푥 1+ 퐼퐶50

The inhibitory activity of SC017 was also tested in SH-SY5Y cells. The compound was capable of preventing the aggregation process in the cellular lifetime sensor assay, and a fluorescence high lifetime indicative of minimal peptide aggregation was observed in the cells for at least 48 h post incubation with Aβ42-488 and the compound (control 2873 ± 19 ps, SC017 treated 3017 ± 14 ps; ***p>0.001). Gratifyingly, no decrease in cell survival was observed with the compound in this assay format or with subsequent MTT cell viability assays. As in Chapter 4, two protocols were employed to monitor the ability of SC017 to rescue cells from the toxic effects of Aβ42 – using

153 either Aβ42 monomers or preformed aggregates (Figure 6.13). First, 500 nM monomeric Aβ42 and increasing concentrations of the compound were added to cells and incubated for 48 h. Addition of the compound resulted in a significant increase in cell survival relative to the addition of monomers alone. In the second approach, the peptide was incubated in conditions conducive to aggregation for 24 h with or without SC017. These samples were then added to the cells and incubated for a further 48 h. The Aβ42 species that existed following in vitro pre-incubation with

SC017 were significantly less toxic to the cells that the untreated Aβ42 aggregates.

Figure 6.13: Cell viability assays showing the rescuing effect of SC017 from Aβ42-induced toxicity. a) Aβ42 monomers (500 nM) and increasing concentrations of SC017 were incubated with cells for 48 h. SC017 shows a significant protective effect against the toxic effects of the Aβ42. b) Aβ42 monomers were incubated with increasing concentrations of SC017 for 24 h under conditions conducive to aggregation. The resulting species were incubated with cells for 48 h. The Aβ42 species formed in the presence of SC017 in vitro are significantly less toxic than the aggregated Aβ42 control. Cell survival was calculated using an MTT cell viability test, using SH-SH5Y cells. The viability of untreated cells was set as 100%. Plots show mean ± SEM, n = 4, data analysed using a one-way ANOVA with Dunnett’s multiple comparison post-test; *p>0.5, **p>0.01, ***p>0.001.

6.3.4 Antioxidant Activity

A study by Bayati et al. examining the neuroprotective activity of chalcone derivatives, in terms of the restoration of cell viability in response to the addition of oxidant stressors, found that the chalcones are capable of providing a rescuing effect.422 Guided by this observation, the antioxidising capabilities of the hit compounds were tested. Intracellular ROS accumulation was monitored using fluorescent probe CM-H2DCFDA, which forms the fluorescent compound dichlorofluorescein (DCF) upon oxidation. SH-SH5Y cells were incubated with increasing

154 concentrations of TJS285 and SC017, as well as the control antioxidant EGCG, for 12 h prior to the addition of the oxidation-sensitive dye. Figure 6.14 shows the percentage fluorescence under the different treatment conditions, relative to that of the control cells, where increasing fluorescence intensity is indicative of increased ROS accumulation. It is evident that all three compounds significantly reduce the amount of ROS within the cellular environment, indicating antioxidant activity.

100 * *** *** *** *** *** *** *** *** 50

% Intracellular ROS Intracellular % 0

M M M    50 nM 50 nM 50 nM Control 500 nM2.5 500 nM2.5 500 nM2.5 EGCG TJS285 SC017

Figure 6.14: Effects on cellular ROS accumulation with the addition of potential antioxidants. ROS detected my monitoring the fluorescence of the CMH2DCFDA dye. Incubation of cells with TJS285 and SC017 for 12 h results in a reduced concentration of intracellular ROS species relative to untreated control cells. X-axis show increasing concentrations of each compound. Control indicates the ROS signal observed for untreated cells and is set as 100%. The values are corrected to discount fluorescence not attributed to the dye. 10 mM H2O2 was added to a population of cells as a control, 137 ± 2% intracellular ROS. Plot shows mean ± SEM, n = 4 -8 for each cell type in two independent experiments. Data analysed using a one-way ANOVA with a Dunnett’s multiple comparison post-test; *p>0.5, ***p>0.001.

Time has not permitted for further investigations into this ROS diminishing effect. Planned future studies include determining if pre-treatment with SC017 is capable of reducing the extent of cell death and ROS build-up when cells are subjected to an oxidant stressor, such as H2O2. Additionally, investigating the accumulation of ROS with the addition of Aβ42 species, and how compound treatment effects this may also provide insight into how the compound affords the rescuing effect observed in the previous cell viability studies.

155

6.4 Summary and Conclusions

In this work a variety of flavonoid-based chemical libraries were screened for inhibitory activity against the aggregation of Aβ42 using a ThT fluorescence assay. Three libraries exhibiting varying degrees of inhibitory activity were identified: (1) triazole-bridged flavonoid dimers, (2) phenyl- linked biaurones, and (3) heteroaromatic chalcone derivatives. The heteroaromatic chalcone derivatives displayed the greatest activity, and were chosen as the main library for subsequent hit validation, library expansion and testing in cells.

Evaluation of the activity of the first hit compound TJS285 using TEM imaging and immunoassay dot blots gave initial promising in vitro results. Tests with live cells, however, uncovered a dose dependent cytotoxicity associated with the compound. The electrophilic reactivity of the α, β unsaturated ketone present in TJS285 is believed to be responsible for the observed toxicity, as this moiety has raised concerns in many drug development studies. It is generally difficult to forecast if the presence of such a Michael acceptor will produce toxic effects in cellular assays until experimentally tested, and it has been observed that similar compounds within a library containing this feature can result in greatly contrasting cellular effects.428 Expansion of the heteroaromatic chalcone library resulted in the identification of a second generation hit, SC017, which showed similar in vitro anti-aggregation activity as TJS285 and greatly improved cellular activity. Initial results with both compounds also suggested that they display antioxidant activity, which may prove useful for the development of multifunctional AD therapeutics. Although more expansive library modifications would likely result in the development of even more active compounds, due to time constraints it was decided to test the hit compounds in whole organism disease models. Drosophila melanogaster were selected for these studies, given the availability of these organisms, the flexibility in experimental design and their short propagation time and lifespan relative to other AD model species. This work is discussed in Chapter 7.

156

Chapter 7 Compound Screening in Alzheimer’s Disease Model Drosophila Melanogaster

7.1 Introduction 7.1.1 Drosophila Melanogaster as Disease Model Organisms

Drosophila melanogaster are widely used as invertebrate models for studying a range of biological processes and diseases.436–439 They benefit from a host of practical features, including short generation time and lifespan, ease of handling, a compact genome and simplicity in genetic manipulation.440,441 The use of Drosophila for modelling human disorders relies on the assumption that fundamental aspects of cell biology are conserved throughout evolution in higher organisms. This is supported by studies which have analysed the Drosophila genome, revealing that approximately 75% of human disease-related genes have homologues in these flies.442,443 The short lifetime and conserved ageing mechanisms make Drosophila particularly well-suited as model organisms for the study of age-related neurodegenerative diseases.444–446 Furthermore, there is a reasonable similarity between the central nervous systems of humans and flies, with both consisting of neurons and glia and utilizing the same neurotransmitters.447 As such, their use in modelling neurodegenerative diseases, including Alzheimer’s disease (AD), has greatly increased in popularity in the past 10-15 years.440,447

7.1.2 Alzheimer’s Disease Drosophila Models

A variety of Drosophila disease models have been developed in attempts to unravel the molecular mechanism underlying the pathogenesis of AD.440,441,448,449 Essential steps in the disease pathway have been reconstructed in such models, and the resulting phenotypes characterised. These include models that express human Aβ40 and Aβ42, human wild type or mutant Tau, human APP and processing enzymes, BACE and presenilin.441,447 Within the context of this study, transgenic flies expressing Aβ isoforms represent the most useful models. They serve to mimic the extracellular distribution of this pathogenic species, and provide a platform to screen the in vivo effects of small molecule aggregation inhibitors. Neuronal expression of Aβ42 has been shown to cause neuronal death, locomotion defects and reduced lifespan.450 These features can be monitored to probe the effects of treatment with exogenous drugs or inclusion of other genetic

157 modifications. Additionally, the expression of different Aβ isoforms display varying degrees of toxicity, allowing the severity of the resulting phenotype to be tuned appropriately for the desired study.450,451 The location of Aβ expression, as well as the strain of Aβ expressed, can be reliably controlled using the UAS-Gal4 system.

7.1.3 The UAS-Gal4 System

The UAS-Gal4 system is commonly employed for controlling the spatially restricted expression of effector genes in disease model Drosophila.452,453 This bipartite system consists of the yeast transcriptional activator Gal4 and its target upstream activating sequence (UAS). The Gal4 activator gene is fused to an endogenous tissue specific promoter sequence and the UAS is coupled to a target transgene – Aβ in the case of the AD models discussed (Figure 7.1). As the target gene is separated from its transcriptional activator, target genes are not expressed until fly strains are appropriately crossed. This ensures that the parent lines are viable and that the target gene is only turned on in the progeny, where the phenotypic consequences of its expression can be studied.

In this work, two tissue-specific Gal4 lines were used – the retinal specific glass multimer reporter (GMR) and the pan-neuronal driver elavc155. Aβ expression under the control of GMR results in an easily visualised rough eye phenotype, if peptide aggregation occurs. The consequences of neuronal Aβ expression directed by elavc115 can be investigated microscopically or with locomotive and longevity tests. Three Aβ transgenes were investigated with this system; Aβ42, arctic Aβ42 and tandem Aβ42-Aβ42. These peptides vary in terms of aggregation propensity and the degree of toxicity they induce in the developing fly populations. The arctic mutation, which causes early- onset familial AD (FAD) in humans, encodes a Glu22Gly amino acid substitution in the Aβ42 sequence. The resulting peptide has a greatly increased aggregation propensity than wild-type

Aβ42 and the expression of the arctic mutant is therefore more toxic to flies than the wildtype

450 peptide. The tandem Aβ42-Aβ42 construct employed consists of two copies of the Aβ42 monomer, tethered with a 12 amino acid linker. The higher effective concentration of Aβ42 and reduced entropic costs of self-assembly greatly speed up the aggregation rate, giving rise to an even more aggregation-prone and toxic species.451

158

Figure 7.1: Schematic of the UAS-Gal4 system. In the driver line, a Gal4 coding sequence is inserted downstream of a tissue specific endogenous promoter. The pan-neuronal elavc115 driver construct, present on the X chromosome, is shown here. The Gal4 protein is a transcriptional activator that can bind specifically to an upstream activating sequence (UAS) in the transgenic construct, in this case on the second chromosome. Binding activates the expression of a downstream target sequence, arctic Aβ42 in this schematic. Crossing the female and male flies gives a progeny with both the driver line and UAS line, resulting in the expression of the target gene only in the tissue specified by the endogenous promoter.

7.1.4 Drug Screening in Drosophila

Given the ease of handling and flexibility in experimental design in working with Drosophila, these species represent promising models for whole organism screening of compound libraries for hit identification and validation. Drosophila have successfully been employed to validate in vivo activity of lead compounds and to identify potential therapeutic compounds for a variety of diseases, including cancer and neurodegenerative disorders.436,437 Different assay formats can be used to provide insight into if and how small molecules function in whole organisms. In terms of AD research, these predominantly involve monitoring locomotor deficits, lifespan, morphological aberrations, accumulation of amyloid plaques and vacuolation of the brain.447,449,450,454 Locomotion studies inform on the fitness of treated fly populations by monitoring their activity levels, through the measurement of their climbing capability. Longevity of a population provides a robust estimate of their general health, so a quantitative comparison of survival rates with different drug treatments can provide information on the disease-rescuing in vivo activity of the small molecules. Direct imaging can be performed to assess the morphology of well characterised features, or those that have been manipulated to be diagnostic of a particular process. For example, the appearance

159 of a rough eye phenotype in developing flies has been used to indicate the aggregation propensity of retinally-expressed amyloidogenic proteins in AD models.450 The presence of protein inclusions or brain vacuolisation can also be assessed microscopically in fixed Drosophila samples using specific dyes or antibody staining.

Using Drosophila for drug screening has, of course, its limitations. When investigating complex processes and multifaceted human diseases, only certain aspects can be modelled and results from drug treatment may, therefore, be misinterpreted or overestimated.436 Variations in the fly metabolic systems can also give rise to differences in the toxicity of small molecules between species, but generally a strong correlation has been observed in toxicity between Drosophila and higher organisms.455 Despite these issues, these organisms have still provided many successful results using various screening strategies.436,437 Validation of hit compounds in Drosophila models has been carried out to filter high throughput screening (HTS) hits, for those which show greatest in vivo activity.170,456 Alternatively, primary assays can be performed using these genetically tractable organisms.457,458 Such efforts result in the identification of compounds that already display desirable features, such as oral bioavailability, metabolic stability and low toxicity, a benefit that cannot be effectively mimicked in vitro or in cell cultures screening approaches.

This is especially advantageous in terms of screening compounds that have CNS targets. A major hindrance in the development of neurologically active drugs is the permeability of the blood brain barrier (BBB) (Section 1.7.2).228 The Drosophila BBB is a compound structure, and is tightly controlled with cellular and non-cellular layers and various transporters to maintain brain homeostasis and protect from xenobiotic substances.459,460 Whilst there are quite obvious differences between the BBB of vertebrate and invertebrates, increasing evidence suggests that certain features are conserved.460 Importantly, Drosophila have a P-glycoprotein homologue (Mdr65),461 a drug efflux pump held responsible for the low CNS accumulation and activity of many drugs in humans. Testing the ability of small molecules to penetrate and exert a function within the Drosophila brain can, therefore, be indicative of potential activity in higher organisms. This belief is supported by studies in which Drosophila have been shown to respond to humans CNS active drugs.439,462,463

In addition to compound screening for hit identification or validation, Drosophila models have also been employed to test the efficacy and dose limiting toxicity of known drugs,437,464 or to synthetically optimise lead drugs to achieve maximal brain activity.465 They also represent a promising method to test drug combinations, for the design of combinatorial therapies. Using this whole organism model provides a means to investigate synergistic effects, taking into account

160 drug interactions and activity on a range of potential targets. This can highlight unforeseen benefits or disadvantages of drug combinations, or may result in a lower effective dosing.466 Furthermore, they serve as potential models for investigating chemical genetics – that is identifying pathways affected by current therapeutics.436,464,467 This can provide a better understanding of disease pathway and facilitate the design of optimised derivatives

7.1.5 Modelling the Relationship between Aβ Aggregation and Oxidative Stress in Drosophila

Oxidative stress reflects an imbalance between the production of free radicals and the ability of the body to counteract or detoxify their harmful effects, and repair the resulting damage. A plethora of evidence indicates that brain tissue in AD patients is exposed to oxidative stress during the course of the disease. Such evidence is manifested through high levels of oxidised proteins, oxidative modifications of DNA and elevated levels of lipid peroxidation end products.468–470 Various strategies have been investigated to interrupt the vicious cycle of oxidative stress and neurodegeneration in AD, (Section 1.3.4, Figure 1.6). Of these, the development of antioxidants has been extensively studied.407,471,472 In contrast, targeting Aβ aggregation represents an uncommon but intriguing approach. An ‘Aβ-induced oxidative stress hypothesis’ has previously been suggested, whereby increased cellular oxidative stress in the disease is attributed to the

87,88 presence of oligomeric Aβ42. Investigating how Aβ-aggreation inhibitors affect the accumulation of ROS and subsequent oxidative damage is, therefore, an avenue worth exploration.

In terms of Drosophila studies, it has been shown that the expression of Aβ42 alters the expression of genes associated with the oxidative stress response.90 In a study by Rival et al., Drosophila populations were challenged with conditions conducive to increased oxidative stress, through the

90 addition of H2O2 to their food source. Flies expressing Aβ42 were found to be more senstive to the harmful effects of oxidative stress than wild type flies, highlighting a link between this stress response and Aβ-induced toxicity. This study also showed that the co-expression of the iron binding protein ferritin or the administration of the metal chelator clioquinol suppressed Aβ toxicity, further supporting this relationship and indicating that the effects can be rescued with targeted small molecules. In subsequent work, younger flies expressing aggregation-prone Aβ species (Aβ42 and arctic Aβ42, 2 days old) were found to respond to oxidative stressors in a similar manner to old Aβ40 expressing flies (20 days old), indicating that aging makes the interaction with oxidative stressors more powerful.446 This provides additional evidenece that the aggregation of Aβ plays a key role in mediating the oxidative stress response, and how the damage from oxidative

161 stress is more intense in older flies – perhaps an observation that could be extended to older humans. Such work has prompted studies to invesitgate if Aβ aggregation inhibitors can provide rescuing effects from the oxidative stress associated with AD.

7.2 Primary Objectives

The primary aim of this work was to investigate the in vivo activity of the previously identified small molecule aggregation inhibitors in AD model Drosophila melanogaster. For this, Aβ- expressing AD model strains were cultured and examined in three experimental set ups. The ability of the small molecules to alleviate the rough eye phenotype, to rescue flies from oxidative stress-induced damage and also to prolong the longevity of the disease model Drosophila was investigated.

7.3 Results and Discussion 7.3.1 Overview

The ability of the lead compounds to alleviate the toxicity associated with Aβ42 aggregation in AD Drosophila models was tested using three different experimental formats. This work was performed together with Dr Liisa van Vliet, of the Hollfelder group. The AD fly models were supplied by Dr Damian Crowther, who also helped to plan and oversee the experimental work.

First investigated was the effect of drug treatment on improving an Aβ-induced rough eye phenotype. This developmental abnormality of the insect’s compound eye is caused by the ectopic expression of Aβ42 in neuronal cells of the eye, as driven by the GMR-Gal4 genetic driver. Potential rescuing effects of lead compounds were determined by qualitative assessment of the severity of

115 the phenotype. Next, an elav -Gal4 driver line was used to induce pan-neuronal arctic Aβ42 expression, and the resulting fly populations were maintained under conditions conducive to oxidative damage. Various studies have linked Aβ aggregation with the oxidative stress associated with AD progression.85 This experimental format was used to examine the effect of Aβ aggregation inhibitors on reducing oxidative damage caused by adding H2O2 to the flies food. This was measured in terms of survival rates in these oxidising conditions.

Finally, a longevity study was undertaken to monitor changes in the length of lifespan in treated and untreated AD fly models expressing neuronal arctic Aβ in non-oxidising conditions. These

162 disease model flies have greatly reduced lifespans compared to wild type flies. It was hoped that delaying or preventing amyloid aggregation would result in increased survival rates. Initial studies were carried out with MJ040 (Chapter 4), TJS285 (Chapter 6), and EGCG as a control. Following optimisation of the compounds throughout the course of the project, later tests also included MJ040X (Chapter 4) and SC017 (Chapter 6).

7.3.2 Rescue of the Rough Eye Phenotype

The GMR-Gal4 driver was employed to ectopically express Aβ strains in the developing eyes, giving an easy visualised phenotype depending on the aggregation propensity of the peptide isoform used. In practice, this was achieved by crossing the appropriate virgin female Gal4 line with the males carrying the UAS-effector fusion. The resulting F1 population carry both constructs and consequently express the required Aβ transgene in a spatially restricted manner (Figure 7.1) The compound eye of Drosophila consists of well-ordered hexagonal arrays of ommatidia, which are equally sized and regularly spaced.437,473 Any subtle defects that disrupt the defined lattice structure are clearly evident, giving the ‘rough eye’ appearance. Aggregation of Aβ is associated with neuronal death during development, which in turn results in failures as the ommatidia develop. This leaves holes that disrupt the regular lattice, thereby distorting the delicate pattern of the eye.449,450 It was hoped that drug treatment of the developing flies would be able to rescue this distortion in an easily detectable manner.

Initially flies expressing a single Aβ42 transgene were used, but the rough eye phenotype was very weak, so it was hard to compare potentially subtle difference between treated and untreated flies (data not shown). In an attempt to generate a more robust phenotype, three strains expressing

Aβ42 species with increasing degrees of aggregation propensity were employed. Single arctic Aβ42 transgene, double arctic Aβ42 transgene and tandem Aβ42-Aβ42 were used. The effect of raised temperature (25 to 29 °C) on the development and adult life of flies was also investigated, as GMR expression is heat dependent and displays increased expression levels at higher temperatures. The rough eye phenotype was clearly more evident in these Drosophila strains (Figure 7.2). In contrast to the symmetrical organization observed in the control flies (w1118), subtle misalignments could be seen in the single arctic Aβ42 expressing flies maintained at 25 °C. This distortion was further exacerbated with development at 29 °C. In the double arctic Aβ42 expressing flies, significantly disturbed ommatidia were observed. Expression of the covalently tethered Aβ42 monomers in the tandem species resulted in severe eye malformation, depigmentation and the presence of necrotic spots. Again, more extreme deformations were evident in flies maintained at 29 °C.

163

Figure 7.2: Identifying the optimal fly line for the rough eye rescue experiment. Well-ordered hexagonal arrays of ommatidia are observed in the compound eye of wild type flies at 25 °C. With low arctic Aβ42 expression (single transgene) a subtle loss in symmetry is evident. Increasing depigmentation, indentations and distortion of the ommatidia is seen with high arctic Aβ42 expression (double transgene). A severe rough eye phenotype, with eye malformation and necrotic spots, is visible with the expression of tandem Aβ42- Aβ42. The same pattern of increasing rough eye severity is observed at 29 °C, with more severe deformation observed.

Guided by these result, it was decided to carry out further rough eye experiments on flies expressing either low arctic Aβ42 or tandem Aβ42-Aβ42, maintained at 29 °C. These models were chosen to investigate the rescuing effect of drugs on both a mild and severe phenotype, in case the effects were more easily distinguished at one extreme. Practically, the experiment was carried out by crossing the Aβ-transgenic males and virgin Gal4 driver line females overnight, then transferring equal numbers of mated females to individual drugged vials. High (50 μM) and low (12.5 μ M) concentrations of MJ040, TJS285 and control EGCG were tested. To treat the vials, 500 μL of each compound solution was added to a vial containing cornmeal food. The surface of the food was gently punctuated with a pipette tip and the food allowed to dry overnight in air. A sprinkle of dried yeast was added to the tube to encourage egg laying before the females were introduced. After 2 days the females were removed and the larvae allowed to develop. The resulting phenotypes observed for the high drug concentrations are shown in figure 7.3. Mild deformation of the ommatidial pattern was evident in the low arctic Aβ control flies, as previously observed. The differences in treated flies, if any, were difficult to quantify. Examination of the entire progeny of each population did suggest that on average the TJS285 treated flies displayed a milder rough eye phenotype. Proving this qualitative assessment, however, was not possible. A severe rough eye phenotype was seen in all populations expressing tandem Aβ42-Aβ42, and no drug treatments appeared to provide any rescuing effect. It was concluded that the qualitative nature

164 of the rough eye phenotype would preclude definitive conclusions if the drug effects were subtle, and no further investigations with this model were performed.

Figure 7.3: Rescue of the rough eye phenotype with high concentration drug treatment. Tandem Aβ42- Aβ42 expression resulted in a very severe rough eye phenotype, which was not rescued by any drug treatment. Necrosis and depigmentation could be observed under all conditions, even with EGCG which was included as a control inhibitor. The severity of low arctic Aβ42 induced eye deformation was much milder. Qualitatively, there did appear to be a slight improvement in the appearance of flies treated with TJS285, as shown here the size of the eye is slightly larger. However, any difference between treated and untreated flies was too difficult to definitively prove.

7.3.3 Rescue from Aβ-induced Oxidative stress

Guided by the aformentioned study by Ott et al., in which young flies expressing aggregation prone Aβ species (Aβ42 and arctic Aβ42) were found to be more susceptible to oxidative damage than age-matched control flies,446 it was next decided to investigate if drug treatment could interupt the potential role that aggregated Aβ42 species play in mediating the oxidative stress response. It was hypothesised that if a small molecule inhibitor could prevent the aggregation of

Aβ42, an associated decrease in the generation of ROS could be expected, resulting in a longer lifespan than in untreated Aβ model control. To test this, the effect of maintaining fly populations on food supplemented with the oxidative stressor H2O2, with or without the hit small molecule inhibitors, was investigated. Survival curves were then employed to compare the survival of drug- and vehicle-treated flies (Figure 7.4).

c115 The pan-neuronal elav driver was used in conjunction with the arctic Aβ42 transgene, to generate a robust phenotype in clinically relevant tissue. w1118 flies were used as the healthy

165 control. These flies survived for longer in the oxidising conditions, as expected due to the increased sensitivity of Aβ AD models to oxidative stress.446 Drug treatment with all but one of the compounds of interest (MJ040) resulted in a significant reduction in lifespan of the control flies maintained on H2O2. A possible explanation for this observation is that the drug concentration (50 μM) may have been too high. Cell viability tests with the compounds previously indicated toxic effects above a certain concentration (Section 4.2.4, 6.3.2). Interestingly, a detrimental effect of drug treatment was not evident with the arctic Aβ flies. Here, the difference in the length of survival in treated and untreated flies was statistically insignificant for all the compounds. Two possible reasons were initially used to rationalise this observation. First, increased oxidative stress associated with the arctic Aβ phenotype may have been so strong that it induced death before the drug accumulated to toxic levels. Alternatively, the drug treatments may have provided a slight Aβ associated rescuing effect, but this alleviation of oxidative stress was counterbalanced by toxic effects of the high drug concentration.

166

Figure 7.4: Monitoring fly survival in highly oxidising conditions to assess the oxidative stress-rescuing ability of small molecules inhibitors. Survival curves of 3-day old wild type flies and arctic Aβ42 AD model flies maintained on agar supplemented with H2O2, an oxidant stressor. Treatment with all the inhibitors, except MJ040, resulted in a statistically significant reduction in lifespan (relative to the untreated control) in the wild type flies. None of the small molecules significantly affected the lifespan of the AD model flies. The plots show the median lifespan of the flies treated, from time of exposure to H2O2. The control flies’ food was treated with a solution of 0.05% DMSO in H2O, [Drug] = 50 μM, n = 47-50, statistical significant assessed using log-rank test; *p<0.05; ***p<0.001.

To further probe these possibilities, the experiment was performed again using reduced compound concentrations. Only SC017 was tested due to constraints on time and resources, and this compound had displayed the greatest rescuing effect in the MTT cell viability assays. Use of ethanol as the drug solubilising solvent was another modification incorporated into the experimental design. The compound was dissolved in minimal ethanol (95%) prior to dilution in water for treatment of the food. Previously, DMSO had been used for this step, however it was believed that this solvent was causing a slight rescuing effect through unknown mechanisms. This effect may have biased the result, so ethanol was substituted to avoid this potential issue. Figure

167

7.5 shows the survival curves of AD model flies maintained on H2O2 agar, following treatment with 1 μM, 5 μM or 20 μM of SC017. Treatment at any concentration did not statistically affect the lifespan of the flies relative to the untreated controls, suggesting that a dose dependent toxicity of the drug was not the only factor governing the differences in survival rates with the type flies.

Figure 7.5: Monitoring fly survival in highly oxidising conditions with treatment using low concentrations of SC017. a, b, c) Survival curves of 3-day old arctic Aβ42 AD model flies treated with increasing concentrations of SC017 and maintained on agar supplemented with H2O2. No significant effect on lifespan was observed with any drug concentration. d) Plot showing median lifespan of the flies treated, from time of exposure to H2O2, with 0, 1, 5, 20 μM of SCO17. The control flies’ food was treated with a solution of 0.05% EtOH in H2O, n = 38-59, statistical significant assessed using log-rank test.

From this combined work, it appears as though the toxicity of the H2O2 oxidising conditions on the wild type flies is exacerbated with the addition of the inhibitory compounds. This may be explained in terms of the oxidising capabilities of the inhibitors. For example, EGCG is well acknowledged to be a strong antioxidant and has previously been shown to increase the lifespan of healthy Drosophila by inducing the expression of antioxidant enzymes.474,475 An increasing body of data, however, suggests that some of the biological activity of EGCG is due to an induction of oxidative stress.476–478 Figure 7.6 shows how the compound can act as either a radical scavenger or a pro- oxidant, resulting in the formation of ROS.479 This pro-oxidant activity may be responsible for the diminished survival curves upon compound treatment of wild type flies with EGCG, as well as with TJS285 and SC017, which were also shown to display antioxidant activity in preliminary studies

168 investigating their ability to reduce ROS accumulation in SH-SY5Y cells (Section. 6.4.4). This idea is supported by a study by Elbling et al., in which no protection against H2O2-induced oxidative stress was observed with EGCG treatment in human cell lines.480 In fact, increasing concentrations of

EGCG resulted in increased H2O2 in solution, thereby enhancing oxidative and genotoxic effects. This agrees with the result observed in this study, whereby EGCG administration increased toxicity in wild-type flies in the highly oxidising conditions.

Figure 7.6: Anti-oxidant and pro-oxidant effects of EGCG. a) ROS scavenging ability of EGCG to form innocuous substances, or those that can be enzymatically degraded. b) Formation of ROS 479 with aid of EGCG in the presence of H2O2. SOD: Superoxide dismutase.

In contrast with the wild type Drosophila results, the survival curves of AD model flies do not seem to be affected by the addition of the drugs. EGCG has been proposed to exert it’s inhibitory effect by binding to unfolded Aβ.174 It is possible that this binding reduces the ROS generating ability of the compound, by sequestering it in a bound state with the Aβ. The other flavonoid derivatives are speculated to interact with Aβ in a similar manner, through comparison if their structure with known inhibitors. This may indicate that the compounds are not strong enough to rescue the flies from the H2O2 stressor, but are capable of reaching and binding to the Aβ target, which is

169 promising in terms of their BBB permeability. Pharmacokinetic studies would be necessary to further investigate this idea.

7.3.4 Longevity Studies

The longevity of a fly population provides a robust estimate of it general health, allowing for a quantitative comparison of the positive or negatives effects of a genetic modification or drug treatment. Expression of Aβ42 and arctic Aβ42 results in age-dependent and dose-dependent neurodegeneration. Such models have reduced longevity and locomotor deficits, with the arctic mutation showing accelerated phenotypic deficiencies.450 To investigate if the hit small molecule inhibitors could alleviate neuronal Aβ42 induced toxicity, the lifespan of flies maintained on drugged food was measured.

As in the oxidative stress experiment, the pan-neuronal elavc115 driver was used with the arctic

c115 Aβ42 transgene. After mating of the elav gal4 virgin females with the UAS-arcAβ42 males, the fertilised females were transferred to jars containing drugged cornmeal food. The offspring were allowed to develop in the drugged medium and were transferred to fresh drugged vials on the day of eclosion. Females were maintained in group of 10, in treated vials, and were counted and moved to freshly treated food every 2-3 days, to maintain a steady drug level throughout the entirety of the experiment. Control flies (the offspring of crossing w1118 flies with the Gal4 driver line) were also maintained under the same conditions. The survival curves of the flies for each of the conditions are shown in figure 7.7. As expected the expression of arctic Aβ42 causes a marked reduction in the lifespan of the flies, with the untreated AD models dying on average 28 days sooner than their wild type counterparts. With the wild type flies, EGCG treatment resulted in an increase in lifespan, TJS285 a decrease in lifespan and the rest of the compounds had no statistically significant effect. None of the compounds affected the length of survival of the AD model flies.

170

Figure 7.7: The effect of drug treatment on Drosophila longevity. Survival curves of wild type and arctic Aβ42 expressing model Drosophila maintained on drugged food tubes. EGCG treatment of wild type flies significantly increases lifespan, whereas TJS285 treatment leads to a reduction in survival time. None of the compounds have a significant effect in the lifespan of the AD flies. The control flies’ food was treated with a solution of 0.05% DMSO in H2O, n = 47-50, statistical significant assessed using log-rank test; *p<0.05; **p<0.01.

Of particular interest in these results is that EGCG provides a protective effect in the wild type flies, but no effect in the AD models. EGCG has reached clinical trials for anti-Aβ aggregation AD treatment, so it is unclear why no protective effect is observed here. The dose used may have been too low to accumulate within the brain at a high enough concentration to exert a strong effect against the highly aggregation prone Aβ peptide, and the positive effect in the healthy flies just a result of in increased expression of antioxidant enzymes. The difference in toxicity observed with TJS285 treatment in wild type and AD model flies is also interesting. It was originally

171 suspected that inherent compound toxicity (as observed in cell viability cells, Section 6.3.2) may be balanced by the beneficial effects of the compounds in relation to its ability to inhibit the aggregation of Aβ. A lower concentration of the compound was tested in the AD flies (5 μM) to investigate if an anti Aβ-aggregation protective effect could be observed before toxic concentrations were reached (Figure 7.8a). A significantly shorter lifespan with this lower concentration was observed. It is difficult to rationalise this dose relationship at this point, as the experiment has only been performed at this lower concentration once with a different solubilising solvent than the previous experiments (EtOH vs DMSO). More repeats with higher and lower dosing, as well as pharmacokinetic studies on BBB penetration, are needed to provide more insight. A lower concentration gradient of SC017 was also tested in the AD model flies (Figure 7.8b). Interestingly, lower drug concentrations resulted in significant reductions in median lifespan with this compound also. This is particularly unexpected as no effect with high concentration treatment was observed in the wild-type flies, suggesting little inherent toxicity at this dose or below More experimental work is again needed to rationalise this observation.

Figure 7.8: The effect of low drug concentrations on the longevity of arctic Aβ AD model Drosophila. a) Survival curve showing treatment with 5μM TJS285 compared to control flies maintained on 0.05% EtOH (drug vehicle solution). b) Median lifespan of AD model flies treated with 1, 5 or 10 μM SC017. Lower concentrations show significantly shortened lifespans. c, d, e) Corresponding survival curves for treatment with 1, 5 or 10 μM SC017. The control flies’ food was treated with a solution of 0.05% EtOH in H2O, n = 37- 56, 1 independent repeat, statistical significant assessed using log-rank test; *p<0.05; **p<0.01.

172

7.4 Summary and Conclusions

Many basic biological, physiological, and neurological properties are conserved between mammals and Drosophila melanogaster. The degree of conservation in molecular pathways between flies and humans has led to the discovery of many pathological mechanisms in disease, and the development of Drosophila disease models has permitted the screening of potential therapeutics for a variety of disorders in a quick and reliable manner.436,437 In this work, three screening methods were employed to investigate the potential in vivo activity of the lead inhibitory compounds previously identified. Experiments aimed at monitoring the extent that drug treatment could rescue the rough eye phenotype were inconclusive, as a result of the qualitative nature of the measurement. Monitoring the effects of the drugs on improving the survival statistics of fly populations in highly oxidising conditions showed no rescuing effects. As this experimental format had not yet been validated as a reliable indicator of in vivo activity, the results obtained do not disprove potential anti-aggregation activity, of for that matter antioxidant effects, for the compounds tested.

Longevity was used as a surrogate marker of Aβ toxicity and comparison of survival rates between treated and untreated flies was used to inform on in vivo drug activity. The longevity studies suggested that MJ040, MJ040X and SC017 show little toxicity with wild type flies. TJS285 was seen to exert a significantly toxic effect, in agreement with result observed in cell studies (Section 6.3.2). Treatment with any of the compounds at 50 μM did not improve the survival statistics of the AD model flies. As no pharmacokinetic tests were performed it is difficult to tell if the lack of activity is attributed to pharmacokinetic or pharmacodynamics issues. It is unclear how orally bioavailable and metabolically stable the compounds are, and if they can cross the blood brain barrier. Confirming the presence of the drug in the flies’ brains would be necessary to probe these pharmacokinetic considerations, and associated issues could be improved by increasing the dose or administering the drug by microinjection into the haemolymph (the flies’ blood orthologue that bathes the body and brain). It is worth noting that there was no effective positive control in this experiment, so it is hard to deduce if the results observed were issues related to the compounds or due to deficiencies in the experimental set up or fly models used. EGCG was included for this role, but did show the desired rescuing effect. Further validation with small molecules known to exert the protective effects in AD fly models would therefore be required to validate the fly models used and the sensitivity of the screens employed. Congo red or curcumin could be used, as they have previously shown anti-aggregation activity and rescuing effects with these AD models flies.181,450 Future inclusion of alternative screening approaches, such as monitoring deficits in

173 locomotive activity or the of the presence of amyloid plaques in the brain, may also provide a better insight into the in vivo activity of the compounds tested.

174

Chapter 8 Conclusions and Future Outlook

8.1. An Overview

Alzheimer’s disease is the most prevalent form of dementia and its incidence is increasing at an alarming rate.21 The aging population has never before borne such an impact on the global total, and the social and economic pressures of Alzheimer’s disease is of growing concern. The molecular underpinnings of AD are not well understood, but many lines of evidence indicate that the

27 aggregation of Aβ42 underlies the neuronal dysfunction and death observed in AD. Therefore, identification of molecules that can perturb the process of Aβ42 aggregation holds great potential for the development of disease-targeting AD therapeutics.24,26

In this work a medium-throughput microfluidic assay – the nanoFLIM – was designed to screen compound libraries for inhibitory activity against the aggregation of Aβ42, under various extrinsic factors if necessary, with high sampling sizes and minimal volume requirements (Chapter 3). The assay relies on the use of a fluorescence lifetime sensor and a newly developed microfluidic chip, in which 110 microreactors can be monitored simultaneously. Once validated with a collection of known inhibitory molecules, the nanoFLIM was used to screen compound libraries developed by previous members of the Spring research group (Chapter 4). This resulted in the identification of a selection of structurally diverse compounds displaying a range of inhibitory activity. One key library of interest was focused on to probe the reliability of the nanoFLIM system and investigate the activity of the constituent compounds. The inhibitory activity of the lead compound from this library, MJ040, was validated using a host of biophysical screening approaches. Tuning of the fluorescence lifetime sensor technique enabled Aβ42 aggregation, and the effect that inhibitory compounds can have on the process, to be monitored in live neuroblastoma cells (Chapter 5). Such work further validated the inhibitory activity of the optimised hit, MJ040X – a prodrug of MJ040, in which the carboxylic acid was masked as a methyl ester to improve cell permeability. The technique was also adapted for compound screening in whole organism disease models (Chapter 5). A Caenorhabditis elegans AD model expressing GFP-Aβ was employed and with the use of the fluorescence lifetime sensor, it was shown that treatment of the worms with MJ040X supressed the rate of Aβ aggregation.

175

Development of the nanoFLIM was guided by the research investigating the use of ‘agarose cage’ microdroplets in a novel Aβ42 aggregation screening platform (Chapter 2). This assay system was optimised to a stage where it could be implemented to monitor the aggregation of the peptide in the presence of exogenous small molecules. However, it was not believed sufficiently advantageous over current assays systems to warrant further development as a screening platform. The system may prove useful for other applications and is under investigation for such uses by other members of the Hollfelder group. In parallel with nanoFLIM optimisation and validation, a selection of small molecule libraries from the Spring research group were also screened using the conventional ThT fluorescence assay format (Chapter 6). Here, three flavonoid derivative libraries were shown to display inhibitory activity. Of these, a library of heterocyclic chalcone derivatives was deemed the most promising and an indole chalcone hit TJS285, was identified. The library was also expanded to increase SAR data, resulting in the identification of another hit chalcone derivative, SC017. The activity of these compounds, as well as MJ040X, was tested in AD model Drosophila melanogaster (Chapter 7). Although no rescuing effect was observed, such work revealed some unexpected results, which will be used to guide future Drosophila drug testing experiments.

8.2 The NanoFLIM

The nanoFLIM was developed to combine the sensitivity of the fluorescence lifetime sensor with the benefits associated with miniaturisation in microfluidic systems. A new microfluidic chip capable of trapping 110 precisely ordered droplets was employed and FLIM was used to monitor the self-assembly of the peptide within. From a mechanical point of view, this system lowers reagent consumption, allows for greatly increased sampling size and also provides a means to introduce a measurable shearing force on the trapped droplet microreactors. For the aim of probing peptide aggregation, this device provides great sensitivity and temporal resolution.

Measuring the fluorescence lifetime change of encapsulated Aβ42 was confirmed to represent a reliable method of monitoring peptide aggregation and the modulatory effects of potential inhibitors Due to the stochastic nature of Aβ42 primary nucleation, multiple replicates are generally required to gain reproducible kinetics. The nanoFLIM permits up to 110 replicates from 10 μL of stock solution, with a working volume of 18 nL per droplet. Furthermore, encapsulation of the peptide in individual droplets gives a more realistic insight into early aggregation events than ensemble techniques. This may provide a means to escape bulk behaviour and investigate primary nucleation and the effect that inhibitory molecules may have on this process, in more

176 detail. The fluorescence lifetime sensor was shown to be more sensitive than conventional ThT assays, detecting the formation of species that are not recognised by this benzathiole dye. The nanoFLIM was also shown to be less susceptible to the misleading readouts that cause false positive and negative results than the ThT ‘gold standard’, when screening spectroscopically active small molecule libraries.

In the pilot nanoFLIM screening campaign, 445 compounds were screened, with at least 5 repeats per compound. The minimised volume (10 μL for 5 repeats in the microfluidic format, compared to 500 μL in the comparable ThT fluorescence 96 well plate format) resulted in a drastic reduction in the amount of peptide required – 217 μg of Aβ42 in the nanoFLIM assay and 11.2 mg in the ThT fluorescence assay. The nanoFLIM is also greatly advantageous in terms of cost efficiency and a 10-fold reduction in the peptide price alone was achievable – calculated as £222 in the nanoFLIM and £2397 in a 96 well plate format. Furthermore, lower compound requirements (360 pg per droplet) makes the technique especially beneficial for precious compound libraries, in which the small molecules are in short supply.

Of the screened libraries, one particular library of interest was focused on in detail. The inhibitory activity of a subset of seven compounds from this ‘MJ’ cinchophen library was investigated using various different screening techniques – ThT fluorescence, TEM, immunological dot blot assay and cell viability assays (Figure 8.1). Through such work, the nanoFLIM detected one ThT assay false positive (MJ036), three ThT identified false negatives (MJ001, MJ041, MJ042) and one pro- aggregating compound (MJ014), which was later shown to provide a rescuing effect in cells, potentially by accelerating the formation of non-toxic Aβ42 fibrils. The lead compound from the nanoFLIM screening campaign, MJ040, was carried through to subsequent cellular and in vivo testing.

177

ThT  ThT  ThT  NanoFLIM  NanoFLIM  NanoFLIM  TEM  TEM  TEM   Cell rescue  Cell rescue Cell rescue 

ThT  ThT  ThT  ThT  NanoFLIM  NanoFLIM  NanoFLIM  NanoFLIM  TEM  TEM  TEM  TEM  Cell rescue  Cell rescue  Cell rescue  Cell rescue 

Figure 8.1: Structure and inhibitory activity of MJ library subset. Ticks denote inhibitory activity and crosses indicate that no inhibitory activity was observed.

8.3 Cellular and Whole Organism Model Testing

In vivo Aβ aggregation is an incredibly complicated and poorly understood process, the intricacies of which are often missed with in vitro screening techniques.481 Tuning the in vitro lifetime sensor protocol has provided a means to probe amyloid aggregation and the inhibitory effect of exogenous small molecules in live neuroblastoma cells and in disease model C. elegans. Furthermore, Aβ-related Drosophila melanogaster AD models were also employed to probe the activity of the hit compounds in an alternative whole organism system.

178

8.3.1 Cellular Lifetime Sensor Assay

The cellular lifetime assay was performed by adding partially labelled Aβ42 to the extracellular medium of SH-SY5Y neuroblastoma cells, then using non-invasive fluorescence lifetime imaging to follow the transport of the peptide from the medium into the cells. A drop in fluorescence lifetime indicative of peptide aggregation was observed over time, and the addition of active inhibitory compounds diminished the change in fluorescence lifetime, by means of supressing aggregation of Aβ42 within the cells. In implementing this approach, important realisations were made regarding the activity of original hits, MJ040 and TJS285. Despite comparable in vitro activity with known inhibitor EGCG (as indicated through IC50 value calculations), MJ040 exerted a significantly lower inhibitory effect in the cellular environment than EGCG. Cell permeability issues associated with the presence of the anionic charge on the molecule were deemed responsible, and masking of the carboxylic acid to form prodrug MJ040X resulted in a greater inhibitory effect in cells. TJS285 was shown to significantly inhibit Aβ42 aggregation in the cellular FLIM assay format. Observations when performing this study, however, indicated that the compound was having a detrimental effect on the overall survival rate of the cell population. This was supported by subsequent cell viability tests, which showed decreased cell survival with increased compound concentrations. Performing only viability tests here would not have revealed that, although cytotoxic, TJS285 does significantly inhibit the aggregation process. This observation, recognised in the cellular FLIM assay, prompted library expansion to find a non-toxic inhibitory analogue, the second generation lead compound SC017.

8.3.2 Caenorhabditis Elegans Lifetime Sensor Assay

The fluorescence lifetime sensor was also employed to monitor the aggregation of Aβ in disease model C. elegans. C. elegans expressing GFP-Aβ in the body muscle were used, and a time dependent decrease in the fluorescence lifetime of the GFP reporter dye was observed in association with Aβ aggregation. It was shown that treatment of the worms with MJ040X from the first day of adulthood supressed the rate of peptide aggregation, whereas treatment from the earliest larval stage inhibited the Aβ self-assembly completely. This technique can, therefore, provide early insight into a compound’s in vivo activity, and help to filter and prioritise hits for subsequent hit-to-lead strategies. It allows a direct comparison of Aβ aggregation propensity in vitro and in vivo, which is not possible by any other method. This format also shows potential for evaluating the effects of drug treatment from different development stages, providing easily

179 comparable absolute measurement of peptide aggregation at particular time points. Given the small size of these model organism, they are also amenable to HTS in the 96 well plate format. This could be used to obtain aggregation kinetics quickly and easily for a range of potential inhibitors at a range of concentrations, greatly facilitating hit prioritisation and determining dosing regimens for preclinical testing. Furthermore, there is great scope for starting HTS campaigns in these living systems, through which the complex interplay of features involved with the aggregation process would be better modelled than in simple in vitro systems.

It is worth noting that a relatively small change in fluorescence lifetime was detected for the overall aggregation process in this work. The use of different constructs in future studies may provide a larger dynamic range for monitoring the change in fluorescence lifetime and the modulatory effects of inhibitory small molecules. Changing the reporter fluorescent protein, for example, could be a promising strategy. Future work will also see the use of fitness tests to investigate the ability of MJ040X to rescue worms from the Aβ aggregation induced paralysis phenotype. This is currently underway in the David Research Group (DZNE, Tübingen, Germany).

8.3.2 Disease Model Drosophila Melanogaster

To test the activity of the hit compounds in a different whole organism model, Drosophila melanogaster AD models were used. The flies employed expressed aggregation prone Aβ isoforms in specified locations (retinal and neuronal tissue).450 Unfortunately, the hit compounds were not shown to effectuate rescuing results under any of the experimental platforms investigated. The qualitative nature of the rough eye phenotype precluded definitive conclusion with the treatment of flies expressing aggregation prone Aβ in retinal tissue. The oxidative stress experiment showed no rescuing effect, however the design of this experiment may have been flawed, as the control EGCG was unexpectedly seen to cause a significant reduction in the survival rates of healthy flies in the highly oxidising conditions employed. Statistically insignificant results following compound treatment in the experiments measuring longevity, a rough estimator of population health, were also disappointing. This does not necessarily mean that the compound is not active, or indeed that Aβ inhibition does not confer rescuing effects, as various issues may have contributed to the results obtained. Undesirable pharmacokinetic effects, such as low oral bioavailability, metabolic instability or poor BBB permeability, would need to be ruled out before such an assessment is made. This work did, however, show an unusual dose response relationship with SC017, in that lower drug concentrations significantly reduced longevity relative to the control, whereas higher

180 drug concentrations made no significant effect on survival rates. This unexpected observation will be further investigated in future work.

Another interesting point to note with regards to this work, is that MJ040X was seen to exert an effect in the C. elegans model but with the Drosophila studies, highlighting that different model systems can provide distinct insight into a compound’s inhibitory activity. As the methods used to monitor compound activity in both models differed, direct measurement of amyloid formation in the C. elegans versus measurement of survival rates in the Drosophila, the results obtained are not directly comparable. Results from the ongoing fitness tests with C. elegans and potential brain imaging of Drosophila would be better suited for comparing the activity of the compounds between species.

8.4 Identification and Development of Hit Inhibitory Compounds 8.4.1 MJ040 and MJ040X

The inhibitory activity of MJ040, identified using the nanoFLIM, was validated using TEM, AFM, ThT fluorescence assays and immunological dot blots. This compound was calculated to have an

IC50 of 5.2 ± 1.4 μM, which is in a similar range to that of phase III clinical trial candidate EGCG (IC50 = 6.4 ± 0.7 μM), a promising observation for future drug development strategies with the compound. Cellular FLIM studies and MTT viability tests suggested low inhibitory activity in cells, which was attributed to the presence of the carboxylic acid. This functional group is typically associated with poor brain exposure due to a combination of multiple untoward factors, including poor permeability, high plasma protein binding and P-gp recognition.232 The methyl ester prodrug MJ040X, displayed enhanced anti-aggregation activity in neuroblastoma cells and successfully supressed the rate of Aβ aggregation in disease model C. elegans (Figure 8.2). No rescuing effect was observed in AD model Drosophila. Due to the lack of pharmacokinetic or pharmacodynamic studies performed, it is not known the cause of this disappointing result. It may be due to the inability of the compounds to pass the BBB or survive Drosophila metabolic systems long enough to reach the Aβ target in high enough concentrations. Alternatively, the compound may have perturbed the aggregation process as desired, but this did not result in any improvement on the survival rates of these disease model flies. This is more unlikely as other aggregation inhibitors have been seen to lengthen the survival rates of these model organism, albeit with higher compound concentrations.181 Different screening techniques here, such as fly brain imaging, behavioural assays, and pharmacokinetic studies would be necessary to further probe these possibilities.

181

Figure 8.2: Structure of original hit MJ040 and optimised lead MJ040X.

Studies on the mechanism of action of this compound has not been rigorously performed, given the difficulty in preparing Aβ42 to the strictly homogenous monomer solutions required for such analysis and also due to the suspected complications arising from the spectroscopic activity of the compound. Observations that the methyl ester prodrug MJ040X does not exert an anti- aggregation effect in vitro, suggests that the carboxylate is necessary for interaction with the peptide. The C. elegans studies have provided early insight into the compounds mechanism of action. Such work has suggested that MJ040X may perturb primary nucleation, as evidenced by the increase in inhibitory activity level observed when the worms are treated from an earlier stage in their development.

It is worth noting that a selection of other compounds from the cinchophen library showed activity in modulating the Aβ42 self-assembly process, displaying both anti-aggregation and pro- aggregation effects depending on the functionalisation. Retrospective cell viability assays with the entire MJ subset, performed after MJ040 had been selected and carried on alone for further biological studies, showed that two other compounds effectuated greater rescuing effects from

Aβ42 toxicity than the lead compound. The strong protective effect of the aggregation-accelerating MJ014 is of particular interest, as bypassing the formation of toxic oligomers through the enhanced production of innocuous fibrils represents a promising therapeutic strategy. The activity of this compound will be further investigated in time. The library was not expanded during the course of this work and little SAR could be generated through the current small library. It is believed, however, that future synthesis could provide insight into the structural feature governing the activity of this core structure, which could, in turn, guide future hit development.

182

8.4.2 Heteroaromatic Chalcones

The indole chalcone TJS285, identified through conventional ThT fluorescence screening, displayed a potent inhibitory effect on Aβ42 aggregation in vitro with the use of various screening techniques. It was, however, observed to cause a dose-dependent toxicity in cell studies. Other members of the heteroaromatic chalcone library did not appear to cause the same degree of toxicity (as measured using MTT viability tests), so the library was extended with the aim of identifying analoguous compounds showing similar anti-aggregation activity without the associated toxicity. This was achieved through the synthesis of SC017, another indole chalcone derivative with an IC50 of 4.4 ± 0.4 μM (Figure 8.3). As this compound was found later in the research, it’s inhibitory activity has not yet been characterised in detail. It was shown, however, to supress Aβ42 aggregation in the cellular FLIM format and to increase cell survival rates when SH-

SY5Y cells were incubated with monomeric Aβ42.

Figure 8.3: Structure of original hit TJS285 and second generation hit SC017.

There is great scope for future investigation with this hit compound, and indeed the library as a whole. The synthesis of a collection of heterocyclic chalcones in this work contributed to the overall SAR data of the library. This was used to build some general guidelines, which can guide future library expansion (Figure 6.11). Leading on from this, investigations into the molecular binding interactions may provide more information to guide future compound development.

Polyphenols act at a variety of different binding sites for Aβ42, so at this point it is difficult to predict where the hit compounds are acting.482 NMR studies or molecular simulations could provide insight here.172,189

An interesting avenue worth further exploration is the antioxidising activity of SC017. Initial experiments have indicated that the compound displays antioxidant behaviour comparable to EGCG in the absence of any oxidant stressors. Follow up studies may provide insight into the extent of this activity and potential implications. Some well-defined SAR guidelines for flavonoid derivatives antioxidising activity have previously been reported,409,422 and the incorporation of key motifs may increase the antioxidant behaviour of specific members of this library. This is of

183 relevance as oxidative stress plays a significant role in AD, so the design of dual anti-aggregation and antioxidant compounds holds great potential for the development of multifunctional therapeutic agents. On this note, it may be worth exploring the anti-inflammatory activity of this library also. Heterocyclic chalcone derivatives have previously shown inhibitory activity against the inflammatory response-associated cyclooxygenase (COX) enzymes. As several studies have indicated that use of nonsteroidal anti-inflammatory drugs is associated with delayed onset or slowed cognitive decline in AD,483 inhibitory activity against COX enzymes may also contribute to the development of a multifaceted disease-modifying drug.

8.4.3 Pharmacokinetic Considerations

As discussed in Chapter 1 (Section 1.7.2), a major limiting factor in the development of CNS active drugs is the presence of the BBB, which prevents approximately 98% of systematically administered small molecules from entering the brain.227,228 The physicochemical properties of CNS-permeable drugs have been used to guide the development of new CNS therapeutics, defining optimal properties in terms of lipophilicity, weight, polar surface area and more.232 Table 8.1 shows how the properties of lead compounds MJ040X and SC017 compare to those generally associated with optimal brain exposure.

Table 8.1: Physicochemical parameters of lead compounds MJ040X and SC017 compared to those associated with optimal CNS exposure.233 cLogP and tPSA calculated using http://www.molinspiration.com software.

Property Desirable range MJ040X SC017 cLogP ≤ 3 4.6 3.2 Molecular weight ≤ 360 387 279 Topological polar surface 40 < TPSA ≤ 90 85 73 area (Å2) Hydrogen bond donors ≤ 0.5 0 3

Both compounds fit a selection, but not all of, the criteria. Through medicinal chemistry approaches there is scope to modify the compounds in such a way that they retain activity but better align with the optimal physicochemical parameters. Binding studies would aid such work, potentially highlighting what motifs are not necessary and could be removed or swapped with

184 more favourable bioisosteres.484 It is worth noting that the optimal physiochemical parameters described are guidelines and not definitive rules, and studies have indicated that the property space of potential CNS penetrating molecules is slightly bigger and more forgiving than that defined by analysis of current marketed drugs.232 Additionally, the parameters focus on the ability of small molecules to passively diffuse into the brain, whereas other factors such as active or facilitated transport may be at play.485 In strictly employing these guidelines, BBB permeability can therefore be underestimated.

To gain a better understanding of the PK profiles of the two compounds, they are currently being tested for their ability to penetrate the BBB in wildtype mice in the laboratory of Professor Paul Fraser (University of Toronto, Ontario, Canada). The compounds will be administered by oral gavage (direct administration to the stomach) and the concentration of the compounds within the brain will be determined by homogenisation and high performance liquid chromatography analysis at defined time points post exposure. The results obtained through this work will then guide the design of subsequent pharmacological studies in transgenic AD mouse models.

8.5 Implications for Future Alzheimer’s Disease Research

In the 19th century Esquirol wrote, “A man in a state of dementia is deprived of advantages which he formerly enjoyed; he was a rich man who has become poor”.3 Due to the lack of disease- treating measures treatments for AD, and indeed of other age-related neurodegenerative disorders, this description remains an inevitable truth today. Protein misfolding lies at the root of AD, and perturbing the process of amyloid formation is a promising conceptual framework for the development of disease-modifying AD drugs. Recent failure in clinical trials, however, suggest that current screening strategies are limited in their ability to provide such hits. The research described in this thesis was performed with the aims; 1) developing techniques to facilitate efficient compound screening and hit validation in vitro and in biologically relevant systems; 2) screening diverse small molecule libraries for the identification of new modulatory compounds; 3) validating and optimising hit compounds found.

The microfluidic assay designed serves to decrease the cost, peptide requirements and compound consumption associated with library screening. It allows for high sampling sizes and reduces false positive and negative detection, thereby reducing attrition rates and saving time and money. The unified fluorescence lifetime screening system developed bridges the gap that intrinsically exists between in vitro and in vivo approaches, permitting results from high-throughput in vitro

185 screening to be directly compared to physiologically relevant cellular and whole organism analysis. The nanoFLIM assay is unique in its fluorescence lifetime readout, which can also be dynamically applied in live cells. Previously there was no other single technique that could be used to measure amyloid aggregation in the three distinct relevant formats. It is believed that breadth of this system and the potential high throughput afforded by its use in screening, could see its implementation speeding up the rate at which hits are identified, validated in vivo and prioritised for future hit development strategies. Furthermore, there is the potential to start the compound screening process in the living systems, in which the cooperative effects of chaperones and other intracellular factors are better taken into consideration. With this assay protocol, there would be a reduced chance of missing active compounds that do not function in the simplified in vitro format, and also the opportunity to eliminate compounds with unfavourable drug-like properties early in the screening process. The efficiency afforded by the fluorescence lifetime sensor assay platform has already yielded a lead in MJ040X, which was shown to exert a strong inhibitory effect in live cells, and in disease model C. elegans. It is hoped that this comprehensive strategy will provide new opportunities for the identification of probe molecules to unravel the fundamental pathways underlying AD, and in the development of therapeutics to target the processes giving rise to this devastating disease.

186

Chapter 9 Experimental Procedures

9.1 Reagents

Aβ42 (>95 %) and Aβ42 Hilyte™ Fluor 488 (>95%) were purchased from Eurogentec Ltd as lyophilised powder. Tau-K18 was acquired from the Molecular Neuroscience Group (Department of Chemical Engineering and Biotechnology, University of Cambridge). hFTAA was acquired from the Nilsson group (Linkoping University, Sweden). 1H,1H,2H,2H-perflurooctanol was acquired from Alfa Aesar. ThT was purchased from AbCam. The compound screening libraries were obtained from the Spring Research group (Department of Chemistry, University of Cambridge). All other chemicals, unless otherwise stated, were purchased from Sigma Aldrich.

9.2 Peptide preparation 9.2.1 Amyloid β 9.2.1.1 Unlabelled Aβ

Lyophilised Aβ42 (1 mg) was dissolved in ice cold trifluroacetic acid (TFA, 500 μL). The solution was transferred to a 1.5 mL Eppendorf tube and sonicated in an ice-water bath for 60 s. The TFA was removed in a vacuum desiccator overnight. Ice cold 1,1,1,3,3,3-hexafluro-2-propanol (HFIP, 1 mL) was added to re-suspend the lyophilised peptide. The sample was sonicated for 60 s at 0 °C, then aliquoted into 10 µL, 20 µL or 50 µL portions. 5 µL aliquots were taken at the start, middle and end of the aliquoting process for mass spectrometry analysis. The HFIP was removed in a vacuum desiccator overnight and the lyophilised samples were stored at 80 °C until use. The concentration of the aliquots was determined using amino acid analysis (Department of Biochemistry, University of Cambridge).

The required concentration of unlabelled Aβ42 was prepared by dissolving the peptide solution in dimethyl sulfoxide (DMSO) (5% of total solvent volume), then adding sodium phosphate buffer (NaPi, 50 mM, pH 7.4). The solution was sonicated at 0 °C for 3 min, then centrifuged at 13,400 rpm at 0 °C for 30 min to remove possible preformed aggregates.

For the generation of Aβ42 seeds, 10 µM unlabelled Aβ42 was incubated for 72 h (37 °C, agitation 350 rpm). The resulting fibrils were sonicated for 3 min and were stored in aliquots at 20 °C prior to use.

187

9.2.1.2 Labelled Aβ

The fluorescent Aβ42 used in this study was labelled with Hilyte™ Fluor 488 (HF-488) at the N- terminus. The lyophilised peptide (0.1 mg) was dissolved in 1% NH4OH (200 µL) and sonicated for

60 s at 0 °C. The sample was aliquoted into 5 or 10 µL units, snap frozen in liquid N2 then stored at 80 °C. Amino acid analysis was carried out to determine the peptide concentration (Department of Biochemistry, University of Cambridge). The required concentration of labelled AB42 was prepared by adding NaPi (50 mM, pH 7.4) buffer to an aliqout in NH4OH that had been thawed on ice.

For studies with partially labelled peptide, each peptide was prepared as described then mixed at the appropriate ratios before each set of experiments. This stock was aliquoted into small units, then snap frozen and stored at 80 °C until use.

9.2.2 Tau

The K18 construct used for these studies was acquired from Ms Na Yu in the Molecular Neuroscience Group (University of Cambridge). It was prepared by the previously reported method.254 For the labelled peptide, the recombinant K18 construct was prepared by replacing the native cysteines at 291 and 322 with alanines, then replacing the isoleucine at 260 with a cysteine. The cysteine at position 260 was labelled with Alexa Flour® 488.

9.3 Microfluidics 9.3.1 Chip Fabrication 9.3.1.1 Flow Focusing device

The flow focusing device used in this study was designed by Dr Fabrice Gielen (Hollfelder Research Group, Department of Biochemistry), who also fabricated the silicon master using a soft lithographic technique.486 Figure 9.1 shows a computer aided design (CAD) of a flow focusing device with a 50 μm junction. Chips were created by pouring a mixed 1:10 w/w mixture of silicone elastomer curing agent and silicone elastomer base (Slygard, Dow Corning) into a petri dish containing the silicon master. The PDMS was degassed in a vacuum desiccator until there were no air bubbles observed (~1 h). The dish was cured at 65 °C overnight. The PDMS layer was cut and peeled from the master using a disposable scalpel. The inlet and outlet holes were punched using a biopsy punch with an inner diameter of 1 mm (Kai Medical). The device and glass slide base were

188 cleaned with adhesive tape, then placed in a low pressure oxygen plasma generator. When the chamber was evacuated to 0 mbar, a flow of oxygen was introduced and the PDMS layer and slide were exposed to the oxygen plasma for 12 s. After the plasma exposure, the PDMS layer was gently pressed onto the surface of the glass slide. To generate the hydrophobic channels within the device, a solution of 2% v/v Trichloro (1H,1H,2H,2H)perfluoroctylsilane in fluorinated oil HFE 7500 (3M Novec) was injected into each of the outlets. Overnight incubation at 65 °C afforded the finished device.

Figure 9.1: Computer aided design of the flow focusing device.

9.3.1.2 Shearing-trapping Chip

The shearing-trapping chip used with the nanoFLIM was designed by Dr Fabrice Gielen. In the initial work optimising the assay platform, the chip was synthesised using two PDMS layers that were individually formed using the same protocol as the flow focusing device (Section 9.3.3.1). One layer of the PDMS displayed the serpentine channel, and this was bonded with the channel facing upward on a thin glass coverslip (thickness 130 μm) using oxygen plasma. The other PDMS chip displayed the array of square grids. This was bonded to the first PDMS chip, with the grids aligned over the serpentine channel. This was achieved by exposing both PDMS layers to oxygen plasma, then placing 10 μL of methanol on to the channel-containing chip before inverting and affixing the grid-containing chip on top. The grids were quickly aligned over the channel using an

189 inverted microscope (SP98I, Brunel Microscope) and gently pressed together to bond. A CAD design for each PDMS layer is shown in Appendix VI.

For later experiments a silicon master mould was fabricated by MicroLiquid that allowed for the production of a single PDMS chip with both the serpentine channel and the grids.486 The depth of the first layer with a serpentine channel was 175 μm, while the depth of the square traps were 250 μm. PDMS replicates of these devices were prepared using the same protocol outlined for the flow focusing devices and were bonded to a thin glass coverslip (thickness 130 μm) using oxygen plasma.

Following plasma bonding, both the two-part and one-part trapping chips were silanized by pipetting a fresh solution of 2% Trichloro(1H,1H,2H,2H-perfluorooctyl)silane in HFE-7500 (3M) into an inlet. Subsequently, PTFE tubing (diameter: 200 μm) was manually inserted in designed entrance and exit channels and glued in place by curing PDMS over on a hot plate at 100 °C. This ensured air-tight connection as well as preserving the order of the droplets as they transited from tubing to chip.

Figure 9.2: Photograph of the trapping shearing chip. Inlet and outlet tubing is inserted directly into the serpentine channels at diagonal corners and glued in place by curing with PDMS.

9.3.2 Agarose Cage Droplet Assay

9.3.2.1 Sample Preparation

Microdroplet formation was achieved using a flow focusing device with channels of a width and height of 50 or 80 µm. The process was monitored with a 10× microscope objective by a Phantom MicroEX2 fast camera (Vision research) mounted in an inverted microscope (SP98I, Brunel Microscope). A neMESYS syringe infusion pump (Cetoni) was used to control the flow rate of the

190 sample injection. Glass syringes (100 µL for aqueous sample, 2.5 mL for oil; SGE) fitted with luer lock needles and polyethylene tubing were used to inject the samples into the device. An aqueous sample flow rate of 3.5 µL min-1 and oil flow rate 30 µL min-1 were used.

HFE 7500 oil (3M Novec) containing 1% w/w of the in house formulated surfactant AZ C14 (synthesised by Dr Anastasia Zinchenko) was used as the continuous phase. To prepare the sample, ultra-low melting point agarose (Type IX-A, Sigma Aldrich) (3% w/v in 50 mM NaPi) was heated to 90 °C for 5 min, then allowed to cool to 37 °C. A solution of 5 µM AB42-488 (10% labelled), with or without inhibitor, was quickly prepared on ice. A 1:1 v/v mixture of the agarose solution and peptide solution were mixed to give a final 1.5% agarose w/v and 2.5 µM AB-488 solution. The sample was taken into the glass syringe, where it was contained in the tubing with a backstop of HFE 7500, then injected into the flow focusing device immediately.

9.3.2.2 Microdroplet Production

Two methods for droplet preparation were used, the ‘Droplet’ or ‘Bulk’ preparation. In both cases the test compounds were added to the peptide solution prior to mixture with the agarose solution. The compounds were first dissolved in DMSO to make a 20 mM stock solution, then diluted with NaPi buffer (50 mM, pH 7.4) to obtain the required concentration before addition to the peptide solution. i) Droplet preparation: The peptide and agarose solution were mixed and the droplets formed immediately (100 µL, 1.5% w/v agarose). The droplets were collected on ice, then aliquoted into several Eppendorf tubes. The droplets were incubated (37 °C, agitation 350 rpm) in a high humidity environment. At the specified time points, individual aliquots were removed and cooled on ice for a minimum of 15 min before further manipulation. ii) Bulk preparation: The peptide and agarose solution (100 µL, 1.5% w/v agarose) was incubated (37 °C, agitation 350 rpm) as a bulk solution. 8-10 µL aliquots were removed at the specified time points to form droplets. The droplets were collected and cooled on ice for 15 min prior to demulsification.

9.3.2.3 Demulsification and Washing

NaPi buffer (200 µL, 50 mM, pH 7.4) and 1H,1H,2H,2H-perflurooctanol oil (PFO) (50 µL) were added to the cooled droplet mixture. The oil and surfactant layer coating the microdroplets was broken by pipetting the emulsion until the white layer at the interface had disappeared. The

191 sample was spun down, and the aqueous layer was carefully transferred to a new Eppendorf tube. The droplets in aqueous phase were centrifuged (5 min, 7.5 r.c.f.) then 150 µL of supernatant buffer was carefully removed. Fresh NaPi buffer (150 µL, 50 mM, pH 7.4) was added and the agarose beads were re-suspended by brief pipetting. The droplets were stored in buffer at 4 °C prior to analysis.

9.3.3 The nanoFLIM

Droplets were formed using droplet on demand technology, Dropix (Mitos/ Dolomite).267 Tubing from the microfluidic chip outlet was connected to a syringe pump (Chemyx Fusion 200) operating in withdrawal mode. The inlet tubing was inserted and clamped into the stainless steel hook of a Mitos Dropix. A gas-tight glass syringe (100 μL) was used to fill the device with HFE-7500 oil containing 0.1% picosurf surfactant and the chip was inspected to confirm the absence of air bubbles. 10 μL of partially labelled Aβ42 and test compounds was pipetted into the loading strip of the Dropix. The droplet sequence was programmed to obtain 18 nL droplets with 36 nL oil spacing between each drop. The typical flow rate for producing the droplets was 2 μL/min. Before reaching the device, the droplets were slowed down to 1 μL/min and the filling process was monitored with a bright-field camera of an inverted microscope. After completion of the filling process, the flow rate was stopped, unless specified in shearing experiments.

9.4 Microscopy 9.4.1 Fluorescence Lifetime Imaging Microscopy (FLIM)

9.4.1.1 Microscope Settings

Fluorescence lifetime imaging was carried out using an in-house built time-correlated single photon counting (TCSPC)-FLIM. A supercontinuum laser (SC390, Fianium) was used for excitation at 480 nm, at a repetition rate of 40 MHz. The excitation wavelength was selected by applying an acousto-optic tunable filter (AOTF). A band-pass filter (FF01-470/28-25) was used to improve wavelength selection.

Fluorescence emission from the sample passed through a band pass filter (FF01-525/39-25) before being sent to the TCSPC detector (SPC-830, Becker & Hickl GmbH). A time-correlated single photon counting (TCSPC) (SPC-830, Becker & Hickl GmbH) module was employed for the fluorescence lifetime detection. Each FLIM image has 256 pixel x 256 pixel. The TCSPC detection window was

192

25 ns, separated into 256 time bins. The photon detection rate for each pixel was kept below 1% of the laser repetition rate in order to avoid photon pile-up.

Air objectives (PlanApo 2x and PlanApo 40x, Olympus) were used for imaging the microfluidic chip and the C. elegans respectively. An oil objective (PlanApo 60x BFP1 C2, Olympus) was used for the imaging the cells. For the microfluidic chip, data acquisition times were 200-300 s. For imaging the cells and worm, acquisition times were 100-300 s.

9.4.1.2 Image Analysis

All data recorded on the FLIM-TCSPC system were analysed using the FLIM-fit software developed in Imperial College London.487 The fluorescence lifetime values were obtained by fitting the data with a mono-exponential decay model.

Individual droplets in the images of the microfluidic chip were segmented via the “segmentation” tool in the software. On the FLIM-fit interface, the “global fitting" was selected as “image-wise", and the “global mode" selected as “global binning". Using these settings, all of the pixels in the same segmented area were binned together, to form one fluorescence decay curve. The associated fluorescence lifetime was then analysed and assigned to the corresponding segment, thereby allowing fluorescence lifetimes of different segmented areas (droplets) within the image to be analysed separately.

In imaging cells and C. elegans, the “segmentation” tool and “global binning” protocol using image-wise analysis were again used. Segments that comprised of 2 or more cells were manually modified to separate individual cells. For the C. elegans, segmented regions within the middle of the body of the worms were manually removed as only fluorescence lifetime values from the periphery body muscle were required. For preparing thesis figures with the cells studies, the “global fitting" option was selected as “pixel-wise", and binning set to 7x7. This clearly shows the variation in fluorescence lifetime throughout an aggregate, rather than showing just an average ‘image-wise’ fluorescence lifetime value.

9.4.2 Atomic Force Microscopy (AFM)

The AFM slide was prepared by treating a freshly cleaved mica surfaces with 300 μL of potassium hydroxide for 30 min. The surface was rinsed with MilliQ water (300 μL × 3) then treated with 0.1% poly-lysine solution (300 μL) for a further 30 min. The surface was rinsed with MilliQ water (300

193

μL × 3) and the sample (100 μL) was added and allowed to settle for 30 min. The slide was then rinsed with MilliQ water (300 μL × 3) and allowed to dry in the air.

AFM images were acquired on a commercial system (Bioscope RESOLVE, Bruker) and Nanoscope software (Bruker). The instrument was operated in tapping mode in air using silicon cantilevers with a resonant frequency of 300 kHz, a spring constant of 40 N/m and a tip radius of 10 nm (RTESP, Bruker AXS, Cambridge, UK).

9.4.3 Transmission Electron Microscopy (TEM)

Samples were prepared for TEM by applying 10 μL of sample onto a carbon-coated copper grid (Agar scientific, Stansted, UK). The sample was incubated for 1 min and then blotted off with filter paper. The grid was rinsed twice with MilliQ water (10 μL, blotted off after 30 sec), followed by the addition of a 2% (w/v) uranyl acetate solution (10 μL) for 30 sec. The solution was blotted off with filter paper and the grid was air dried for 5 min before use. The samples were imaged at various magnifications using a Tecnai G2 80-200 kV transmission electron microscope at the Cambridge Advanced Imaging Centre (Multi-Imaging Unit in the Department of Physiology, Development and Neuroscience, University of Cambridge, UK), and were captured with a bottom mounted AMT digital camera.

9.5 Miscellaneous in vitro Techniques 9.5.1 Flow Cytometry

The fluorescence intensity of the agarose beads was measured using a BD FACScan (Becton Dickinson) equipped with a 488 nm laser. The BluFL1 channel (515-545 nm band pass filter) was used to detect the fluorescence intensity of the fluorophore label on the growing aggregates. Analysis of the results was carried out using Flowing software (Turko Centre of Biotechnology). The main population was defined by gating around beads of the correct size in the forward scatter (FSC) vs side scatter (SSC) dot plot. The same gate was used for each of a series of time points for a certain sample and the mean fluorescence value for the gated population was used.

194

9.5.2 Spectroscopic Assays

All spectra were collected on a Tecan plate reader (Tecan, Switzerland) at 37 °C. Aβ42 peptide (10 μM) and ThT (20 μM) were combined to a final volume of 100 μL in filtered sodium phosphate buffer (NaPi, 50 mM, pH 7.4). The samples were measured in a black non-binding 96 well plate (Greiner Bio-One, Switzerland) that was sealed with polyolefin acrylate film (Nunc™, Thermofisher). Samples was performed in triplicate, along with buffer only and 20 μM ThT only control samples. For screening the chemical libraries, 50 µM of the test compounds were added, to give final volumes of 100 μL. The compounds were first dissolved in DMSO to make a 20 mM stock solution, then diluted with NaPi buffer (50 mM, pH 7.4) to obtain the required concentration before addition to the 96 well plate. Fluorescence kinetics were measured at 5 min reading intervals, with 15 s shaking before each read to ensure sample homogeneity. The excitation and emission wavelengths were 440 and 480 nm respectively.

For samples containing agarose, 50 µL of a 3% w/v agarose solution at 37 °C was added to the prepared peptide solution (50 µL) immediately prior to fluorescence measurements.

For LCO studies, 0.5 µM hFTAA was used and the excitation and emission wavelengths were 435 and 535 nm respectively.

9.5.3 Dot blot analysis

20 μM Aβ42 and 100 μM compound were incubated at 37 °C. At specified time points, 10 μL aliquots were removed and stored at –20 °C until use. 5 μL samples were spotted onto a nitrocellulose membrane (Amersham Hybond ECL, GE Healthcare Life Sciences) and were allowed to dry for 1 h. The membranes were blocked with 5% non-fat milk in tris-buffered saline (TBS), then washed with TBS. The membranes were treated with the primary antibodies A11 (Invitrogen) and 6E10 (Invitrogen) for 12 h. After washing (0.01% Tween20 in TBS), anti-rabbit horseradish peroxidase (HRP) conjugated antibodies (ThermoScientific) were added. After 2 h, the membranes were washed (0.01% Tween20 in TBS). An enhanced chemiluminescent (ECL, Thermo Scientific Pierce™) substrate was added and the membranes incubated at room temperature for 3 min. The samples were exposed to X-ray film for 1-2 h.

195

9.6 Biological work 9.6.1 Cell Studies

9.6.1.1 Cell Culturing

SH-SY5Y human neuroblastoma derived cells (Sigma-Aldrich, Gillingham, UK) were grown in serum containing medium (SCM) consisting of 15% FBS, 1% non-essential amino acids (Sigma), 1% L- glutamine (Life Technologies, UK), 41.5% minimal essential medium (MEM, Sigma) and 41.5% nutrient mixture F-12 Ham (Sigma).

For the MTT viability tests, cellular FLIM assays and ROS accumulation tests the cells were cultured using a serum free medium (SFM), in which the FBS was replaced with 2% B27 complement (Life Technologies).

9.6.1.2 MTT Cell Viability Assay

General Protocol: SH-SY5Y cells (10,000 cells/well) were cultured in SFM in a black, clear bottom 96 well plate (Nunc™, Thermofisher) at 37 °C. After the specified incubation period, the medium was removed from the wells and replaced with 100 μL of fresh SFM. 10 μL of MTT (12 mM, Invitrogen) was added and the cells were incubated for 4 h at 37 °C. 85 μL of the medium was then removed from the wells, and 50 μL of DMSO was added and thoroughly mixed. Following 10 min incubation at 37 °C, the samples were mixed again and the absorbance at 540 nm measured, using a Tecan plate reader. All experiments performed in triplicate, unless otherwise stated. Statistical analysis was performed by one-way ANOVA, with Dunnett’s multiple comparison post-test using GraphPad Prism 5.0 software (GraphPad Software Inc. San Diego, CA, USA); *p<0.05; **p<0.01; ***p<0.001.

Compound only experiments: Cells were seeded overnight in SFM. The required concentration of test compounds (5 μL) were added to the SH-SY5Y cells and were incubated for 24-48 h. The compounds were first dissolved in DMSO to make a 20 mM stock solution, then diluted with NaPi buffer (50 mM, pH 7.4) to obtain the required concentration before addition to the cells. Following incubation, the media was removed and the cells were rinsed with minimum essential medium (MEM), and the cytotoxicity was evaluated using the general protocol.

Monomeric Aβ42 experiments: Test compounds (5 μL, prepared as above) were added to the SH- SY5Y cells and were incubated for 1 h. The media was removed and the cells were rinsed with minimum essential medium (MEM). Monomeric Aβ42 (500 nM) with or without the test compound

196 in SFM was added. Following 48 h incubation the cytotoxicity was assessed by means of MTT Viability assay using the general protocol.

Pre-aggregated Aβ42 experiments: Aβ42 (10 μM) was pre-incubated for 24 h with or without the test compounds (prepared as above). The samples were diluted with SFM (500 nM Aβ42) and added to MEM-rinsed SH-SY5Y cells. After 48 h treatment, cytotoxicity was evaluated using the general protocol.

9.6.1.3 Cellular Lifetime Sensor Assay

SH-SH5Y cells (30,000 cells/well) were plated in Lab-Tek II chambered coverglass plates (Nunc™, Thermo Fisher Scientific) in SCM, and were incubated for 24 h. The media was removed and fresh SCM containing the test compound of interest, or vehicle control, was added. The compounds were first dissolved in DMSO to make a 20 mM stock solution, then diluted with NaPi buffer (50 mM, pH 7.4) to obtain a 1 mM stock solution, which was further diluted in SCM to give the required concentration for addition to the cells. The cells were incubated at 37 °C for 1 h. The media was removed and the cells rinsed twice with MEM. SFM containing 50% labelled 250 nM Aβ42-488 and the test compound was added, and the cells then incubated for 12, 24 or 48 h. The cells were rinsed twice with MEM then imaged in a chamber at 37 °C and 5% CO2 on the microscope stage. 8-10 images were taken per condition, with 3-10 cells per image, depending on magnification. The FLIM data was analysed using the FLIMfit software.487 The images were segmented to estimate image-wise average fluorescence lifetime values for each cell. All experiments performed in triplicate, unless otherwise stated. Statistical analysis was performed by one-way ANOVA, with Dunnett’s multiple comparison post-test, or by two-way ANOVA with Bonferroni post-test, using GraphPad Prism 5.0 Software; *p<0.05; **p<0.01; ***p<0.001.

For 0 h controls, 50% labelled Aβ42-448 samples (250 nM) with test compounds in MEM were placed in silicon gaskets (Life Technologies) on a coverslip and were imaged immediately.

9.6.14 ROS Accumulation Tests

SH-SY5Y cells (10,000 cells/well) were cultured in SFM in a black, clear bottom 96 well plate (Nunc™, Thermofisher) at 37 °C. Test compounds were added (5 μL) and incubated for 12 h. The compounds were first dissolved in DMSO to make a 20 mM stock solution, then diluted with NaPi buffer (50 mM, pH 7.4) to obtain the required concentration before addition to the cells. Media

197 was removed and the cells rinsed with PBS, then incubated with PBS for 10 min at 37 ° C. The PBS was removed a freshly prepared solution of CMH2DCF-DA in PBS (5 μM, 100 μL) was added and the cells incubated for 45 min. The cells were rinsed and incubated in PBS for 10 min at 37 °C. Fluorescence was measured using a Tecan plate reader. The excitation and emission wavelengths were 492 and 523 nm respectively.

9.6.2 Caenorhabditis elegans 9.6.2.1 Buffers, growth medium and E. coli preparation

M9 Buffer

 3 g KH2PO4

 6 g Na2HPO4  5 g NaCl

 1 mL 1 M MgSO4

 H2O to 1 L

Synchronisation buffer

 2.75 mL H2O  1.25 mL 1 M NaOH  1 mL 4% NaClO

Nematode Growth Medium (NGM)  34 g agar  3 g NaCl  2.5 g peptone  1 mL cholesterol (5 g in 1 L 95% ethanol, not autoclaved)

 1 mL 1 M MgSO4

 1 mL 1 M CaCl2

 25 mL 1 M KPO4 pH 6

 H2O to 1 L

NGM plate preparation

A mixture of agar, NaCl and peptone in 1 L of H2O was autoclaved, and kept warm at 60 °C using a hot plate. Cholesterol, MgSO4, CaCl2 and KPO4 solutions were added. The medium was filled into

198 petri plates (110 x 15 nm, 7 mL), which were allowed to cool in a biosafety cabinet for 30 min. The plates were sealed with parafilm and stored upside down in an airtight container at 4 °C until use.

For synchronised adult plates, 5-fluoro-2’deoxy-uridine (FUDR; 75 μM, Alfa Aesar) was added following autoclave sterilisation and the plates prepared as described above.

OP50 E. coli culture preparation A plate containing LB agar (Lennox L agar, ThermoFisher Scientific) was streaked with a concentrated solution of E. coli strain OP50 and was incubated for 16 h at 37 °C. LB broth (10 mL; Gibco, ThermoFisher Scientific) was inoculated with a single colony of OP50 E. coli and incubated for 16 h at 37 °C. The culture was centrifuged at 10,000 rpm for 10 min, then the supernatant removed. The bacteria was re-suspended in LB broth. For standard seeding stocks, 5 mL of LB broth was added. For concentrated seeding stocks, 1 mL of LB broth was added.

30 μL of a standard seeding stock of OP50 was added to the centre of NGM plates, and was lawned out to avoid the edges of the dish. The plates were allowed to dry in the flow hood for 30 min, then were sealed with parafilm and stored upside down, in an airtight container, at room temperature for 24 h. The plates were then stored at 4 °C until use. For synchronised adult FUDR plates, 30 μL of a concentrated seeding stock was added and the same protocol followed.

8.6.2.2 Strains and Maintenance

A pSM{myo3::GFP::Aβ} strain was used (see construct map in Appendix III). Worms were maintained as described in reference 373. The worms were grown at 20 °C on NGM plates seeded with OP50 E.coli. Stocks were transferred to fresh plates every 2-3 days by chunking or washing with M9 buffer. In chunking, a sterile scalpel was used to transfer a small cube of NGM (~1 cm3) from a used plate to a fresh plate. For the M9 washing technique, sterile M9 (100 μL) was used to rinse a plate containing worms and was then transferred to a 15 mL falcon tube. The worms were centrifuged at 1600 rpm for 2 min at 20 °C. The supernatant was removed and worms washed with M9 and centrifuged twice more. The worms were re-suspended in minimal M9 then added to fresh plates. The buffer was allowed to dry then the plate incubated.

8.6.2.3 Synchronisation

Worms were synchronised as described by Fontrodona et al.488 Plates containing worms were washed with sterile M9 buffer (1 mL), following the protocol outlined above (8.6.2.2). Following

199 the final centrifugation step excess buffer was removed and freshly prepared synchronisation buffer (2 mL) was added. The solution was shaken for 6 min, then M9 buffer (8 mL) was added to stop the bleaching reaction. The sample was centrifuged at 1600 rpm. for 2 min at 20 °C. The supernatant was removed and eggs washed and centrifuged twice more. M9 buffer (1 mL) was added and to the eggs, and the sample was incubated with shaking (350 rpm.) for 16 h at 15 °C. The sample was centrifuged at 1500 rpm for 1 min and the supernatant removed. The synchronised worms were re-suspended in minimal M9 buffer and transferred to NGM plates seeded with OP50 E. coli. 2.5 days after synchronisation, adult worms were moved to concentrated OP50 seeded NGM plates containing 5’fluoro-2’deoxy-uridine (FUDR) to inhibit the growth of offspring.

8.6.2.4 Small Molecule Treatment

For treated plates, 10 μM MJ040X solution or vehicle control (0.05% DMS0; 500 μL) was added to the FUDR seeded plates and was allowed to dry over 3 h in a flow hood before the worms were added. The compounds were first dissolved in DMSO to make a 20 mM stock solution, then diluted with NaPi buffer (50 mM, pH 7.4) to obtain the required concentration before addition to the NGM plates. On day 8 of adulthood, the worms were transferred to newly drugged FUDR plates. For treatment from larval stage synchronised L1 worms were transferred directly to treated NGM plates.

8.6.2.5 Imaging

For imaging, the worms were mounted on 2% agarose pad containing NaN3 (10 μL, 40 mM) as an anaesthetic, on a glass microscope slide. The samples were covered with a glass cover slip and the edges sealed with nail polish. The slides were then mounted into an inverted confocal microscope attached to the time correlated single photon counting (TCSPC) device and imaged as previously described (Section 9.4.1). Only the anterior of the worms were imaged to avoid autofluorescence from the gut. The FLIM data were analysed using the FLIMfit software.487 The FLIM data were segmented to contain only the GFP-Aβ PMYO3 body muscle, and the combined fluorescence lifetime of each region combined to give the mean lifetime of each worm. The fluorescence lifetime values are presented as the mean average value with error bars referring to the standard error of the mean. Experiments with adult stage treatment were carried out in two independent repeats with the data presented showing the combined results from each time

200 point, n = 16-20. Experiments with larval stage treatment were carried out in one repeat, n = 16- 20. Statistical analysis was performed by two-way ANOVA with Bonferroni post-test, using GraphPad Prism 5.0 Software; *p<0.05; **p<0.01; ***p<0.001. A breakdown of the statistical analysis is given in Appendix IV.

9.6.3 Drosophila Melanogaster 9.6.3.1 Maintenance of Stocks

Drosophila cultures were maintained on standard fly medium containing cornmeal, yeast, glucose, agar, water and nipagin, in glass jars or vials. All stocks were maintained at 18 °C or 25 °C, on a 12- hour light/dark cycle at constant humidity (65%) in the Department of Genetics, University of Cambridge.

The wild type white1118 were employed for controls in all experiments. This is an inbred isogenic strain with white eyes due to a deletion in the sex-linked white gene

For the AD model flies, one of two tissue specific promoters were used. These were positioned upstream of the Gal4 gene, to ensure spatially restricted UAS-linked transgene expression. The retinal specific glass multimer reporter (GMR) was used for the rough eye experiments. The pan- neuronal elavc155 (embryonic lethal, abnormal vision) was used for the oxidative stress and longevity experiments. All UAS stocks contain a GAL4 binding site upstream of the Aβ transgene.

Three Aβ transgenes were employed, UAS-Aβ42, UAS-Arctic Aβ42 (arcAβ42) and UAS-Tandem Aβ42-

Aβ42. UAS-Aβ42 encodes the human Aβ42 peptide fused to a secretion signal peptide from the

450 Drosophila necrotic gene. Arctic Aβ42 strain is the arctic mutant form of Aβ42, which contains a Glu22Gly mutation that results in an increased rate of peptide aggregation and early-onset AD.450

Tandem Aβ42 - Aβ42 consists of 2 copies of the Aβ42 monomer, tethered with a 12 amino acid linker. This peptide has a very high aggregation propensity.451

All AD model strains were provided by Dr Damian Crowther, AstraZeneca Neuroscience.

9.6.3.2 Crossing and Experimental Set up

Males and females were sorted by anaesthetising on a dish of CO2, then using of a fine paintbrush for separation to avoid any structural damage. Virgin females were collected by Glynnis Johnson, Senior Research Technician, Department of Genetics, within 8 hours after eclosion at 25 °C. Flies were crossed by combining males and virgin females for 24-48 h. The males were then removed and the females allowed to lay eggs on the cornmeal food, which was either pre-treated with the

201 chosen drug or vehicle solution control. After 48 h the females were removed and the larvae allowed to develop (approx. 10 days). All adult flies emerging over a 24 h period were transferred to either drugged or untreated fresh tubes, where they were maintained for at least 24 h following eclosion. This ensured all females had mated at least once prior to the initiation of each experiment. Once-mated, the females were then sorted into treated vials (usually at a density of 10 flies per vials) and the males discarded.

Tubes were drugged by adding 100-200 μL of the appropriate compound solution and allowing the buffer to dry overnight. The compounds were first dissolved in DMSO to make a 20 mM stock solution, then diluted with NaPi buffer (50 mM, pH 7.4) to obtain the required concentration before addition to the tubes.

All experiments performed in triplicate, unless otherwise stated.

9.6.3.3. Rescue of Rough Eye Phenotype

Matched numbers of mated females (5-15) were added to drugged vials, which had been sprinkled with yeast to promote egg laying, After 48 h, the females were removed and larvae allowed to develop, with incubation at either 25 or 29 °C. For the single transgene arc Aβ42 flies, the entire progeny under each condition were examined to assess general trends in eye roughness. Three images of randomly picked flies were taken for each condition.

9.6.3.4. Rescue from Oxidative Stress

Mated female flies were sorted into groups of 10 (5-7 groups per condition) and were cultured at 25 °C throughout the experiment. The flies were maintained on drugged tubes for 3 days prior to movement to high oxidative stress inducing conditions. For this, 2% (w/v) agar supplemented with 5% (w/v) sucrose and 10% (v/v) hydrogen peroxide was used.446 The agar (4 mL per vial) was prepared 4 h before use, and drug solutions added once this had cooled (200 μL, after approx. 1 h). Post-exposure survival rates were monitored 4 times daily until no live flies remained. The survival proportions were calculated using a Kaplain-Meier plots and log-rank analysis using the GraphPad Prism 5.0 software. *p<0.05; **p<0.01; ***p<0.001.

202

9.6.3.5. Longevity Screens

Mated female flies were sorted into groups of 10 (5-7 groups per condition) and were cultured at 25 °C throughout the experiment. These were transferred to fresh, drugged food every 2-3 days. At the beginning of the experiment, deaths and censors were recorded only when flies were transferred to new food tubes. When an increased rate of fly death was observed, the flies were scored on a daily basis. Censored flies included those that had escaped, got stuck in the food or had been accidentally damaged during the transferring process. The flies were transferred by quickly tipping the contents of an old tube into a fresh one, with the aid of a plastic funnel to prevent any flies escaping. For the wild type flies, which were generally more active than the AD flies, CO2 was injected into the old tubes to anaesthetise the flies, thereby reducing fly motility before transferral. Data are presented as survival curves and the survival proportions were calculated using a Kaplain-Meier estimation by using a log-rank method with the GraphPad Prism 5.0 software. *p<0.05; **p<0.01; ***p<0.001.

203

9.8 Chemical Synthesis 9.8.1 General Information and Materials

All reagents and solvents were purchased from commercial sources and used without further purification unless otherwise stated. Melting points were measured using a Büchi B545 melting point apparatus and are uncorrected.

Reactions were monitored by thin layer chromatography (TLC) and/or liquid chromatography- mass spectrometry (LCMS). TLC analysis was performed on commercially prepared glass plates pre-coated with Merck silica gel GF254. Visualisation was by the quenching of ultraviolet (UV) fluorescence (λmax= 254 nm). Retention factors (Rf) are quoted to the nearest 0.01. LCMS analysis was performed on a Waters ACQUITY H-Class UPLC with an ESCi Multi-Mode Ionization Waters SQ Detector 2 spectrometer using MassLynx 4.1 software; LC system: solvent A: 2mM NH4OAc in water/acetonitrile (95:5); solvent B: acetonitrile; solvent C: 2% formic acid; gradient: A/B/C, 90:5:5-0:95:5 over 1 min at a flow tare of 0.6 mL/min.

IR spectra were recorded on a Perkin-Elmer Spectrum One (FT-IR) spectrophotometer. Flash column chromatography was performed using slurry packed Merck 9385 Keiselgel 60 silica gel (230-400 mesh) under a positive pressure of nitrogen.

High resolution mass spectrometry (HRMS) measurements were recorded on a Micromass Quadruple-Time of Flight (Q-ToF) mass spectrometer, Waters LCT Premier TOF mass spectrometer or a Vion IMS Q-ToF mass spectrometer.

Magnetic resonance spectra were processed using TopSpin v. 3.5 (Bruker). An aryl, quaternary, or two or more possible assignments were given when signals could not be distinguished by any means. Measured coupling constants are reported for mutually coupled signals; coupling constants are labelled apparent in the absence of an observed mutual coupling, or multiplet when none can be determined.

Proton magnetic resonance spectra were recorded using an internal deuterium lock (at 298 K unless stated otherwise) on Bruker DPX (400 MHz; 1H-13C DUL probe), Bruker Avance III HD (400 MHz; Smart probe), Bruker Avance III HD (500 MHz; Smart probe) and Bruker Avance III HD (500 MHz; DCH Cryoprobe) spectrometers. Proton assignments are supported by 1H-1H COSY,

1 13 1 13 H- C HSQC or H- C HMBC spectra, or by analogy. Chemical shifts (δH) are quoted in ppm to the nearest 0.01 ppm and are referenced to the residual non-deuterated solvent peak. Discernable coupling constants for mutually coupled protons are reported as measured values in Hertz,

204 rounded to the nearest 0.1 Hz. Data are reported as: chemical shift, number of nuclei, multiplicity (br, broad; s, singlet; d, doublet; t, triplet; q, quartet; m, multiplet; or a combination thereof), coupling constants and assignment.

Carbon magnetic resonance spectra were recorded using an internal deuterium lock (at 298 K unless stated otherwise) on Bruker DPX (101 MHz), Bruker Avance III HD (101 MHz) and Bruker Avance III HD (126 MHz) spectrometers with broadband proton decoupling. Carbon spectra assignments are supported by DEPT editing, 1H-13C HSQC or 1H-13C HMBC spectra, or by analogy.

Chemical shifts (δC) are quoted in ppm to the nearest 0.1 ppm and are referenced to the deuterated solvent peak. Data are reported as: chemical shift, number of nuclei (if not one), multiplicity (if not a singlet), coupling constants and assignment.

Compounds were named using ChemBio Draw 14.0 and may not be in agreement with IUPAC guidelines.

9.8.2 Experimental 9.8.2.1 General Procedures (GPs)

GP- 1: Synthesis of indole and pyrrole chalcones (GP-1) a) To a stirred solution of indole or pyrrole aldehyde (1.0 equiv) and the corresponding 2- hydroxyacetophenone (1-2 equiv) in absolute EtOH (20 mL) was added piperidine (1.0 equiv). The reaction mixture was heated at reflux with stirring for 24 - 72 h under a nitrogen atmosphere, until TLC analysis indicated complete consumption of starting material. The resulting mixture was allowed to cool to room temperature, poured into ice-water (20 mL) and then acidified to pH 3-4 with 3 M HCl. The resulting suspension was filtered under suction and the precipitate was washed with ice-water (2 × 20 mL) and suction-dried. The crude residue was purified by flash column chromatography over silica and/or recrystallization to afford the corresponding indole or pyrrole chalcones. b) To a stirred solution of indole or pyrrole aldehyde (1.0 equiv) and the corresponding 2- hydroxyacetophenone (1-2 equiv) in absolute EtOH (20 mL) was added piperidine (1.0 equiv). The reaction mixture was heated at reflux with stirring for 24–72 h under a nitrogen atmosphere, until TLC analysis indicated complete consumption of starting material. The resulting mixture was allowed to cool to room temperature, poured into ice-water (20 mL) and then acidified to pH 3-4 with 3 M HCl. The aqueous solution was extracted with CHCl3 (3 × 20 mL) and the combined

205 organic layer was washed with saturated NaHCO3 (2 × 50 mL), brine (2 × 50 mL), dried over anhydrous MgSO4, filtered and the solvent removed under reduced pressure. The crude residue was purified by flash column chromatography over silica and/or recrystallization to afford the corresponding indole or pyrrole chalcones.

GP-2: Synthesis of furan, pyridine and dimethoxyphenyl chalcones (GP-2)

To a stirred solution of KOH (5.0 equiv) in absolute EtOH (20 mL) cooled to 0 oC in an ice-bath was added a solution of the corresponding aldehyde (1.0 equiv) and acetophenone (1.2-2.0 equiv). The reaction mixture was stirred at 0 oC for 1 h and then at room temperature for 24-72 h under nitrogen, until TLC analysis indicated complete consumption of starting material. The mixture was poured into ice-water (20 mL) and acidified to pH 3-4 with 3 M HCl. The resulting suspension was filtered and the precipitate was washed with H2O (2 × 20 mL) and suction dried. The crude residue was purified by flash column chromatography over silica and/or recrystallization to afford the corresponding chalcones.

9.8.2.3 Individual Compounds

2-(3-bromophenyl)-6-nitroquinoline-4-carboxylic acid

According to the procedure described by Giardina et al.,489 5’nitroisatin (1.00 g, 5.21 mmol), 3’ bromoacetophenone (1.25 g, 6.25 mmol) and KOH (876 mg, 15.62 mmol) were dissolved in EtOH (80 mL). The reaction mixture was heated under reflux for 48 h, allowed to cool, then the solvent removed under reduced pressure. The residue was dissolved in H2O (100 mL) and washed with

Et2O (100 mL x 3). The aqueous layer was cooled to 0 °C, acidified to pH 2 with 3 M HCl, and the precipitate was filtered under suction. The crude residue was purified by column chromatography over silica (10% MeOH in CH2Cl2) to afford a light brown solid (110 mg, 0.42 mmol, 8%).

-1 TLC Rf = 0.12 (10% MeOH in CH2Cl2); IR νmax (neat)/cm : 2923 m (C-H), 1706 m (C=O), 1588 s (C=C), 1528 m (N=O), 1328 m (N=O), 1223 m, 1056 ; 1H NMR (500 MHz, DMSO): δ 7.56 (1H, d, J = 8.0 Hz, ArH12), 7.79 (1H, ddd, J = 8.0, 2.0, 0.9 Hz, ArH11/ArH13), 8.36 (1H, ddd, J = 8.0, 1.5, 0.9 Hz, ArH11/ ArH13), 8.38 (1H, d, J = 9.3 Hz, ArH7), 8.52- 8.57 (2H, m, ArH6, ArH15), 8.70 (1H, s, ArH16), 9.66

206

(1H, d, J = 2.5, ArH4); 13C NMR (126 MHz, DMSO): δ 121.6 (ArC16), 122.7 (ArC4), 123.0 (Cq), 123.7 (ArC6), 126.9 (ArC11/ArC13), 130.2 (ArC15), 131.4 (C12), 131.8 (ArC7), 133.7(ArC11/ArC13), 139.1 (Cq), 139.4 (Cq), 146 (Cq), 150.4 (Cq), 152.8 (Cq) 157.9 (Cq), 166.7 (C=O); HRMS (ESI) m/z =

+ 79 + 372.9810, [M+H] found, C16H10O4N2 B required 372.9818.

methyl 2-(3-bromophenyl)-6-nitroquinoline-4-carboxylate

To a mixture of 2-(3-bromophenyl)-6-nitroquinoline-4-carboxylic acid MJ040 (100 mg, 0.268 mmol) in MeOH (10 mL) was added conc. H2SO4. The mixture was heated under reflux for 24 h, then allowed to cool and neutralized with NaOH (25%). The solvent was reduced under vacuum.

EtOAc (20 mL) was added and the organic layer was washed with saturated NaHCO3 (2 × 20 mL) and brine (2 × 20 mL) then dried over anhydrous MgSO4. The crude residue was purified by flash column chromatography over silica with a 0 – 50% gradient EtOAc in Hexane to afford a pale orange solid MJ040X (71.1 mg, 0.182 mmol, 68% yield).

-1 TLC Rf = 0.83 (1:1 EtOAc/Hex); m.p. 164-165 °C; IR νmax (neat)/cm : 2962 m (C-H), 1724 s (C=O), 1530 s (N=O), 1501 m, 1436 s, 1342 (N=O), 1269 s, 1077 s; 1H NMR (500 MHz, DMSO): δ 4.06 (3H, s, -OCH3), 7.56 (1H, d, J = 7.9 Hz, ArH12), 7.79 (1H, ddd, J = 7.9, 1.9, 1.0 Hz, ArH11/ ArH13), 8.34 (1H, ddd, J = 1.0, 1.9, 7.9 Hz, ArH11/ ArH13), 8.36 (1H, dd, J = 9.2, 0.3 Hz, ArH7), 8.52 (1H, t, J = 1.8 Hz, ArH15), 8.54 (1H, d, J = 2.6, 9.2 Hz, ArH6), 8.69 (1H, s, ArH16), 9.53 (1H, d, J = 2.6, ArH4); 13C

NMR (126 MHz, DMSO): δ 53.4 (-OCH3), 121.6 (C16), 122.3 (C4), 122.6 (Cq), 122.7 (Cq), 123.9 (Cq), 126.9 (C11/C13), 130.2 (C15), 131.4 (C12), 131.8 (C7), 133.8 (C11/C13), 137.7(Cq) , 139.2 (Cq), 146.1 (Cq), 150.2 (Cq), 157.2 (Cq), 165.4 (C=O); HRMS (ESI) m/z = 386.9962 [M+H]+ found,

+ C17H12N2O4Br required 386.9980.

207

(E)-1-(2-hydroxy-4,6-dimethoxyphenyl)-3-(1H-indol-3-yl)prop-2-en-1-one (TJS285, 6.25)

A mixture of indole-3-carboxaldehyde (200 mg, 1.36 mmol), 2’hydroxyl- 4’6’dimethylacetophenone (322 mg, 2.72 mmol) and piperidine (145 mg, 1.36 mmol) in EtOH (20 mL) were reacted according to GP-1b. . The crude residue was purified by column chromatography (1:1 EtoOAc/Hex) to afford chalcone TJS285 (230 mg, 0.71 mmol, 52%) as a yellow solid.

-1 TLC Rf = 0.48 (1:1 EtOAc/Hex); m.p. 208-210 °C. IR νmax (neat)/cm 3178 w (O-H), 2936 w (C-H), 1606 s (C=O), 1581 s (C=C), 1536 s (C=C), 1475 m, 1434 s, 1349 s, 1273 m (C-O),1223 s, 1115 s,

1 1036 m (C-O). H NMR (500 MHz, CDCl3): δ 3.84 (3H, s, -OCH3), 3.99 (3H, s, -OCH3), 6.00 (1H, d, J = 2.4 Hz, H5), 6.13 (1H, d, J = 2.4 Hz, H3), 7.28-7.35 (2H, m, ArH4’, ArH6’), 7.42-7.48 (1H, m, ArH3’), 7.57 (1H, d, J = 2.8 Hz, ArH8’), 8.01–8.06, (1H, m, ArH5’), 8.10 (1H, d, J = 15.6 Hz, -CH=CHO-), 8.14

13 (1H, d, J = 15.6 Hz, -HC=CHO-), 8.53 (1H, br, s, NH), 14.78 (1H, s, OH); C NMR (126 MHz, CDCl3):

55.5 (-OCH3), 55.9 (-OCH3), 91.2 (ArC5), 93.9 (ArCH3), 106.3 (Cq), 111.9 (ArC3’), 115.2 (Cq), 120.7 (ArC5’), 121.6 (ArC4’/ArC6’), 123.3 (-HC=CHO-), 123.4 (ArC4’/ArC6’), 125.3 (Cq), 130.1 (ArC8), 137.0 (Cq), 137.3 (-CH=CHO-), 162.4 (Cq), 165.7 (Cq), 168.5 (Cq), 192.7 (C=O); HRMS (ESI) m/z =

+ + 324.1245 [M+H] found, C19H18O4 required 324.1236.

These spectroscopic data are in accordance with those reported in the literature.490

(E)-1-(2,4-dihydroxyphenyl)-3-(1H-indol-3-yl)prop-2-en-1-one (SC017, 6.26)

A mixture of indole-3-carboxaldehyde (200 mg, 1.36 mmol), 2,4 dihydroxyacetophenone (416 mg, 2.72 mmol) and piperidine (145 mg, 1.36 mmol) in EtOH (20 mL) was reacted according to GP-1a. The crude residue was purified by column chromatography (1:2 EtOAc/Hex) to afford chalcone SC017 (118.2 mg, 0.42 mmol, 31%) as a yellow solid.

208

-1 TLC Rf = 0.58 (2:1 EtOAc/Hex); m.p. 222-225 °C; IR νmax (neat)/cm : 3228 m (O-H), 1615 s ( C=O), 1543 m (C=O), 1491 m, 1443 m, 1273 s, 1232 s, 1204 s, 1121 s, 1036 m; 1H NMR (500 MHz, DMSO): δ 6.26 (1H, d, J = 2.5 Hz, ArH3), 6.42 (1H, dd, J = 8.8, 2.5 Hz, ArH5), 7.20–7.26 (2H, m, ArH4’, ArH6’), 7.45–7.51 (1H, m, ArH3’), 7.64 (1H, J = 15.4 Hz, -CH=CHO-), 8.08-8.12 (3H, m, ArH6, ArH5’, - CH=CHCO-), 8.13 (1H, d, J = 2.8 Hz, H8’), 10.55 (1H, s, OH), 11.95 (1H, s, NH), 13.88 (1H, s, OH); 13C

NMR (126 MHz, CDCl3): δ 102.7 (ArC3), 108.1 (ArC5), 112.6 (ArC3’), 113.0 (Cq), 113.0 (Cq), 113.9 (-CH=CHO-), 120.5 (ArC5’), 121.3 (ArC4’) 122.9 (ArC6’), 125.2 (Cq), 132.5 (ArC6), 133.6 (ArC8’), 137.6 (Cq), 139.0 (-CH=CHCO-), 164.6 (Cq), 165.7 (Cq) 191.5 (C=O); HRMS (ESI) m/z = 278.0823

- - [M-H] found, C16H12NO3 required 278.0817.

(E)-1-(2-Hydroxy-4-methoxyphenyl)-3-(1-methyl-1H-indol-3-yl)prop-2-en-1-one (6.27)

A mixture of N-methylindole-carboxaldehyde (200mg, 1.26 mmol), 2-hydroxy-4 methoxyacetophenone (208.8 mg, 1.26 mmol), and piperidine (107 mg, 1.26 mmol) in absolute EtOH (20 mL) was reacted according to GP-1a. The crude residue was purified by recrystallization from MeOH to afford sc048 (192 mg, 0.62 mmol, 49%) as a bright yellow solid.

-1 TLC Rf = 0.51 (1:1 EtAOc/Hex); m.p. 172-174 °C; IR νmax (neat)/cm : 2911 w, 1623 s (C=O), 1570 s (C=C), 1527 m (C=C), 1506 m (C=C), 1462 w, 1365 s, 1347m, 1261 s, 1203 s, 1129 s, 1072 s; 1H NMR

(500 MHz, CDCl3): δ 3.81 (3H, s, -NCH3), 3.86 (3H, s, -OCH3), 6.48 (1H, d, J = 2.5 Hz, ArH3), 6.51 (1H, dd, J = 2.5, 8.8 Hz, ArH5), 7.30-7.42 (3H, m, ArH4’, ArH5’ and ArH6’), 7.49 (1H, s, ArH8’), 7.58 (1H, d, J = 15.0 Hz, -CH=CHCO-), 7.88 (1H, d, J = 8.8 Hz, ArH6), 7.99-8.03 (1H, m, ArH-4’), 8.15 (1H, d, J

13 = 15.0 Hz, -CH=CHCO-), 13.86 (1H, s, OH); C NMR (126 MHz, CDCl3): δ 33.2 (-NCH3), 55.5 (-OCH3), 101.0 (ArC5), 107.3 (ArC3), 110.2 (ArC4’/ArC5’/ArC6’), 113.0 (Cq), 114.2 (Cq), 114.6 (-CH=CHCO-), 120.7 (ArC4’/ArC5’/ArC6’), 121.6 (ArC4’/ArC5’/ArC6’), 123.2 (ArC3’), 126.0 (ArCq), 130.8 (CH=CHCO-), 134.9 (ArC6), 138.3 (AcC8’), 165.5 (Cq), 166.4 (Cq), 192.0 (C=O); HRMS (ESI+) m/z =

+ + 308.1266 [M+H] found, C19H18O3N required 308.1281.

209

(E)-1-(2-hydroxy-4,6-dimethoxyphenyl)-3-(1-methyl-1H-indol-3-yl)prop-2-en-1-one (6.28)

A mixture of 1-methylindole-3-carbaldehyde (500 mg, 3.14 mmol), 2’hydroxy-4’6’ dimethoxylacetophenone (615 mg, 3.14 mmol) and piperidine (270 mg, 3.14 mmol) in EtOH (50 mL) was reacted according to GP-1a. The crude residue was purified by recrystallization from MeOH to afford chalcone sc010 (776 mg, 2.29 mmol, 73%) as a bright fluffy yellow solid.

-1 TLC Rf = 0.60 (1:1 EtOAc/Hex); m.p. 175 °C; IR νmax (neat)/cm : 1610 s (C=O), 1587 m (C=C), 1558

1 s (C=C), 1539 m (1439), 1373 s, 1276 s, 1214 s, 1158 m, 1038 m; H NMR (500 MHz, CDCl3): δ 3.84

(3H, s, -OCH3), 3.84 (3H, s, -NCH3), 3.98 (3H, s, -OCH3), 6.00 (1H, J = 2.4 Hz, ArH5), 6.12 (1H, d, J = 2.4, ArH3), 7.28–7.37 (2H, m, ArH4’, ArH6’), 7.37–7.40 (1H, m, ArH3’), 7.45 (1H, s, ArH8’), 8.03– 8.01 (1H, m, ArH5’), 8.06 (1H, d, J = 15.5 Hz, -CH=CHCO-), 8.12 (1H, d, J = 15.5 Hz, -CH=CHCO-),

13 14.85 (1H, s, -OH); C NMR (126 MHz, CDCl3): δ 33.3 (-NCH3), 55.5 (-OCH3), 55.9 (-OCH3), 91.2 (ArC5), 93.9 (ArC3), 106.3 (Cq), 110.1 (ArC3’), 113.7 (Cq), 120.8 (ArC5’), 121.4 (ArC4’/ArC6’), 122.3 (-CH=CHCO-), 123.1 (ArC4’/ArC6’), 126.1 (Cq), 134.7 (ArC8’), 137.1 (-CH=CHCO-), 138.4 (Cq), 162.4

+ + (Cq), 165.5 (Cq), 168.4 (Cq), 192.5 (C=O); HRMS (ESI) m/z = 338.1398 [M+H] found, C20H20NO4 required 338.1392.

(E)-3-(Furan-2-yl)-1-(2-hydroxyphenyl)prop-2-en-1-one (6.29)

A mixture of furan-2-carboxaldehyde (200 mg, 2.08 mmol), 2-hydroxyacetophenone (283 mg, 2.08 mmol) and KOH (583 mg, 10.4 mmol) in absolute EtOH (10 mL) was reacted according to GP-2. The crude residue was purified by recrystallization from MeOH to afford chalcone sc048 (232 mg, 1.08 mmol, 52%) as a bright yellow solid.

210

-1 TLC Rf = 0.72 (1:1 EtAOc/Hex); m.p. 114-116 °C; IR νmax (neat)/cm : 3127 w (C-H), 1637 m (C=O), 1574 s (C=C), 1552 s (C=C), 1472 s, 1336 s, 1258 s, 1210 s, 1158 s, 1078 m, 1014 s; 1H NMR (500

MHz, CDCl3): δ 6.54 (1H, dd, J = 3.4, 1.7 ArH3’), 6.78 (1H, d, J = 3.4 Hz, ArH4’), 6.94 (1H, ddd, J = 8.5, 7.2, 1.1 Hz, ArH5), 7.02 (1H, dd, J = 8.4, 1.1 Hz, Ar3), 7.49 (1H, ddd, J = 8.4, 7.2, 1.6 Hz, ArH4), 7.56 (1H, d, J = 15.2 Hz, -CH=CHCO-), 7.56 (1H, d, J = 1.7 Hz, ArH2’), 7.68 (1H, d, J = 15.2 Hz, -

13 CH=CHCO-), 7.92 (1H, dd, J = 8.2, 1.7 Hz, ArH6), 12.89 (1H, s, OH); C NMR (125 MHz, CDCl3): δ 112.8 (ArC3’), 117.1 (ArC2’), 117.6 (- CH=CHCO-), 118.5 (ArC3), 118.8 (ArC5), 120.0 (Cq), 129.6 (ArC6), 131.1 (-CH=CHCO-), 136.3 (ArC4), 145.4 (ArC4’), 151.5 (Cq), 163.5 (Cq), 193.3 (C=O);

- - HRMS (ESI) m/z = 213.0556 [M-H] found, C13H9O3 required 213.0557.

These spectroscopic data are in accordance with those reported in the literature.491

(E)-3-(furan-2-yl)-1-(2-hydroxy-4,6-dimethoxyphenyl)prop-2-en-1-one (6.30)

A mixture of 2-furaldehye (200 mg, 2.08 mmol), 2’hydroxy-4’6’ dimethoxylacetophenone (490 mg, 2.50 mmol) and KOH (117 mg, 2.08 mmol) in EtOH was reacted according to GP-2. The crude residue was purified by column chromatography (1:2 EtoOAc/Hex) to afford chalcone sc019 (173 mg, 0.62 mmol, 30%) as an orange/yellow solid.

-1 TLC Rf = 0.58 (1:1 EtOAc/Hex); m.p. 96 °C; IR νmax (neat)/cm : 3127 w (C-H), 1621 m (C=O), 1584

1 m (C=C), 1479 m, 1438 m, 1336 m, 1219 s, 1156 s, 1107 m, 1026 m; H NMR (400 MHz, CDCl3): δ

3.82 (3H, s, -OCH3), 3.90 (3H, s, -OCH3), 5.95 (1H, d, J = 2.5 Hz, ArH5), 6.09 (1H, d, J = 2.5 Hz, ArH3), 6.49 (1H, dd, J= 3.5,1.6 Hz, ArH3’), 6.66 (1H, d, J = 3.5 Hz, ArH2’), 7.51 (1H, d, J = 1.6 Hz, ArH4’), 7.57 (1H, d, J = 15.4 Hz -CH=CHCO-), 7.78 (1H, d, J = 15.4 Hz, -CH=CHCO-), 14.37 (1H, s,OH); 13C

NMR (100 MHz, CDCl3): δ 55.6 (-OCH3), 55.8 (-OCH3), 91.2 (ArC5), 93.7 (ArC3), 106.3 (ArCq), 112.5 (ArC3’), 115.5 (ArC2’), 125.0 (-CH=CHO-), 129.0 (ArC4’), 144.7 (-CH=CHO-), 152.2 (ArCq), 162.5 (ArCq), 166.2 (ArCq), 168.4 (ArCq), 192.1 (C=O); HRMS (ESI) m/z = 273.0769 [M-H]- found,

- C15H13O5 required 273.0768.

211

(E)-1-(2-Hydroxyphenyl)-3-(1-methyl-1H-pyrrol-2-yl)prop-2-en-1-one (6.31)

A mixture of N-methyl-2- pyrrolecarboxaldehyde (200 mg, 1.83 mmol), 2-hydroxyacetophenone (249 mg, 1.83 mmol) and piperidine (156 mg, 1.83 mmol) in absolute EtOH (20 mL) was reacted according to GP-1a. The crude residue was purified by recrystallization from MeOH to afford chalcone sc047 (207 mg, 0.92 mmol, 50%) as a bright orange solid.

-1 TLC Rf = 0.70 (1:1 EtAOc/Hex); m.p. 98-100 °C; IR νmax (neat)/cm : 2913 (C-H), 1671 w, 1625 s (C=O), 1578 w (C=C), 1548 s (C=C), 1477 s, 1337 m, 1259 m, 1248 m, 1203 s, 1154 s, 1061 s, 1025

1 s; H NMR (500 MHz, CDCl3): δ 3.81 (3H, s, -NCH3), 6.24-6.29 (1H, m, ArH3’), 6.87 (1H, t, J = 2.0 Hz, ArH4’), 6.90-6.95 (2H, m, ArH2’ and ArH5), 7.02 (1H, dd, J = 8.4, 1.0 Hz, ArH3), 7.40 (1H, d, J = 15.0 Hz, -CH=CHCO-), 7.44-7.48 (1H, m, ArH4), 7.89 (1H, dd, J = 8.4, 2.0 Hz, ArH6), 7.91 (1H, d, J = 15.0

13 Hz, -CH=CHCO-), 13.14 (1H, s, OH); C NMR (126 MHz, CDCl3): δ 34.4 (-NCH3), 110.2 (ArC3’), 113.3 (ArC2’/5), 114.5 (-CH=CHCO-), 118.5 (ArC3), 118.7 (ArC2’/5), 120.2 (Cq), 128.6 (ArC4’), 129.2 (ArC6), 130.2 (Cq), 132.8 (-CH=CHCO-), 135.8 (ArC4), 163.4 (Cq), 193.1 (C=O); HRMS (ESI) m/z =

+ + 228.1020 [M+H] found, C14H14O2N required 228.1019.

(E)-1-(2-hydroxy-4-methoxyphenyl)-3-(1-methyl-1H-pyrrol-2-yl)prop-2-en-1-one (6.32)

A mixture of N-methyl-2-pyrrolecarboxaldehyde (200 mg, 1.83 mmol) 2-hydroxy-4- methoxyacetophenone (306 mg, 1.83 mmol) and piperidine (156 mg, 1.83 mmol) in EtOH (20 mL) was reacted according to GP-1b. The crude residue was purified by column chromatography (1:1 EtoOAc/Hex) to afford chalcone sc035 (187 mg, 0.73 mmol, 40%) as a yellow solid.

-1 TLC Rf = 0.53 (1:1 EtOAc/Hex); m.p. 152 °C; IR νmax (neat)/cm : 3106 m (C-H), 2933 w, 1615 s (C=O), 1571 m (C=O), 1543 s (C=O), 1481 m, 1372 s, 1256 s, 1204 s, 1129 m, 1059 s, 1020 m; 1H NMR

(400 MHz, CDCl3): δ 3.77 (3H, s, -NCH3), 3.83 (3H, s, -OCH3), 6.23 (1H, dd, J = 4.0, 2.8 Hz, ArH3’),

212

6.44-6.49 (2H, m, ArH-3, ArH5), 6.81–6.84 (1H, m, ArH4’), 6.86 (1H, dd, J = 4.0, 1.2 Hz, ArH2’), 7.29 (1H, d, J = 15.1 Hz, -CH=CHCO-), 7.78 (1H, d, J = 8.7 Hz, ArH6), 7.68 (1H, d, J = 15.1, -CH=CHO-),

13 13.71 (1H, s, OH); C NMR (101 MHz, CDCl3): δ 34.4 (-NCH3), 55.6 (-OCH3), 101.0 (ArC3/ArC5), 107.5 (ArC3/ArC5), 110.0 (ArC3’), 112.7 (ArC2’), 114.2 (ArCq), 114.9 (-CH=CHCO), 128.1 (ArC4), 130.3 (ArCq), 130.9 (ArC6), 131.8 (-CH=CHCO-), 165.8 (ArCq), 166.5 (ArCq), 191.5 (C=O); HRMS

- - (ESI) m/z = 256.0978 [M-H] found, C15H14NO3 required 256.0979.

(E)-1-(2-hydroxy-4,6-dimethoxyphenyl)-3-(1-methyl-1H-pyrrol-2-yl)prop-2-en-1-one (6.33)

A mixture of N-methyl-2-pyrrolecarboxaldehyde (250 mg, 2.29 mmol) 2’hydroxy-4’6’ dimethoxylacetophenone (450 mg, 2.29 mmol) and piperidine (195 mg, 2.29 mmol) in EtOH (20 mL) was reacted according to GP-1a. The crude residue was purified by column chromatography (1:2 EtoOAc/Hex) to afford chalcone sc015 (48.9 mg, 0.40 mmol, 17%) as an orange/yellow solid.

-1 TLC Rf = 0.44 (1:1 EtOAc/Hex); m.p. 129–133 °C; IR νmax (neat)/cm : 2943 w, 1614 (C=O), 1589 m (C=C), 1546 s (C=O), 1479 s, 1438 s, 1338 s, 1268 s, 1216 s, 1156 s, 1114 m, 1054 s; 1H NMR (400

MHz, CDCl3): δ 3.76 (3H, s, -NCH3), 3.83 (3H, s, OCH3), 3.83 (3H, s, -OCH3), 5.95 (1H, d, J = 2.4 Hz, ArH5), 6.10 (1H, d, J = 2.4 Hz, ArH3), 6.20 (1H, dd, J = 3.9, 2.8 Hz, ArH3’), 6.67 (1H, dd, J = 3.9, 1.6 Hz, ArH2’), 6.79 (1H, dd, J = 2.8, 1.6 Hz, ArH4’), 7.69 (1H, d, J = 15.3 Hz, -CH=CHCO-), 7.81 (1H, d, J

13 = 15.3 Hz, -CH=CHCO-), 14.59 (1H, s, OH) ppm. C NMR (101 MHz, CDCl3): δ 34.5 (-NH3), 53.6 (-

OCH3), 55.8 (-OCH3), 91.2 (ArC5), 93.8 (ArC3), 106.0 (ArCq), 109.6 (ArC3’), 112.6 (ArC2’), 122.7 (- CH=CHCO-), 127.6 (ArCH’), 130.8 (-CH=CHCO-), 130.9 (ArCq), 163.3 (ArCq), 165.7 (ArCq), 168.3

+ + (ArCq), 190.0 (C=O) ppm; HRMS (ESI) m/z = 286.1082 [M+H] found, C16H16NO required 286.1085.

213

(E)-1-(2-Hydroxyphenyl)-3-(pyridin-2-yl)prop-2-en-1-one (6.34)

A mixture of 2-pyridinecarboxaldehyde (200 mg, 1.83 mmol), 2-hydroxyacetophenone (249 mg, 1.20 mmol) and KOH (512 mg, 9.15 mmol) in absolute EtOH (20 mL) was reacted according to GP-2. The crude residue was purified by recrystallization from MeOH to afford chalcone sc050 (111 mg, 0.49 mmol, 27%) as a bright yellow fluffy solid.

-1 TLC Rf = 0.47 (1:1 EtAOc/Hex); m.p. 108-110 °C. IR νmax (neat)/cm : 3048 w (C-H), 1639 m (C=O), 1577 s (C=C), 1560 s (C=C), 1483 s, 1433 s, 1337 s, 1274 s, 1208 s, 1147 s, 1016 s; 1H NMR (500

MHz, CDCl3): δ 6.95 (1H, ddd, J = 8.2, 7.3, 1.1 Hz, ArH5), 7.03 (1H, dd, J = 8.4, 1.1 Hz, ArH3), 7.3 (1H, ddd, J = 7.7, 4.8, 1.1 Hz, ArH3’), 7.47 – 7.53 (2H, m, ArH4, ArH5’), 7.76 (1H, td, J = 7.7, 1.7 Hz, ArH4’), 7.86 (1H, d, J = 15.0 Hz, -CH=CHO-), 8.04 (1H, dd, J = 8.2, 1.6 Hz, ArH6), 8.27 (1H, d, J = 15.0

13 Hz, -CH=CHCO-), 8.71 (1H, d, J = 4.8 Hz, ArH2), 12.75 (1H, s, OH) ppm; C NMR (126 MHz, CDCl3): δ 118.5 (ArC3), 118.9 (ArC5), 120.1 (Cq), 124.1 (-CH=CHO-), 124.7 (ArC3’), 125.9 (ArC5’), 130.3 (ArC6), 136.7 (ArC4), 137.0 (ArC4’), 143.2 (-CH=CHCO-), 150.3 (ArC2’), 152.8 (Cq), 163.6 (Cq), 194.1

- (C=O); LCMS (ESI+) m/z = 226.1 ([M+H]+, tr = 1.57 min); HRMS (ESI) m/z = 224.0715 [M-H] found,

C14H11NO2 required 224.0717.

These spectroscopic data are in accordance with those reported in the literature.492

(E)-3-(2,5-dimethoxyphenyl)-1-(2-hydroxy-4,6-dimethoxyphenyl)prop-2-en-1-one (6.35)

A mixture of 2,5-dimethoxybenzaldehyde (200 mg, 1.20 mmol), 2’hydroxy-4’6’ dimethoxylacetophenone (283 mg, 1.44 mmol) and KOH (810 mg, 14.4 mmol) in EtOH (20 mL) was reacted according to GP-2. The crude residue was purified by recrystallization from EtOH to afford chalcone sc023 (312 mg, 0.90 mmol, 75%) as a bright yellow solid.

214

-1 TLC Rf = 0.55 (1:1 EtOAc/Hex); m.p. 175 -178 °; IR νmax (neat)/cm : 2945 w, 1617 s (C=O), 1580 (C=C), 1561 s (C=C), 1494 s, 1346 s, 1262 m, 1212 s, 1160 s, 1110 s, 1039 s; 1H NMR (500 MHz,

CDCl3): δ 3.82 (3H, s, -OCH3), 3.84 (3H, s, -OCH3), 3.87 (3H, s, -OCH3), 3.91 (3H, s, -OCH3), 5.96 (1H, d, J = 2.4 Hz, ArH5), 6.11 (1H, d, J = 2.4 Hz, ArH3), 6.87 (1H, d, J = 9.0 Hz, ArH3’), 6.92 (1H, dd, J = 30.0, 9.0 Hz, ArH4’), 7.15 (1H, d, J = 3.0 Hz, ArH6’), 7.94 (1H, d, J = 15.8 Hz, -CH=CHCO-), 8.10 (1H,

13 d, J = 15.8 Hz, -CH=CHCO-), 14.37 (1H, s, OH); C NMR (101 MHz, CDCl3): δ 55.6 (-OCH3), 55.8 (-

OCH3), 55.8 (-OCH3), 56.1 (-OCH3), 91.2 (ArC5), 93.8 (ArC3), 106.4 (CqAr), 112.4 (ArC3’), 113.7 (ArC4’), 116.5 (-CH=CHCO-), 125.3 (ArCq), 128.2 (ArC6’), 137.6 (-CH=CHCO-), 153.2 (ArCq), 153.5 (ArCq), 162.5 (ArCq), 166.1 (ArCq), 168.4, (ArCq), 192.9 (C=O); HRMS (ESI) m/z = 345.1340 [M+H]+

+ found, C19H21O6 required 345.1338.

(E)-3-(2,4-dimethoxyphenyl)-1-(2-hydroxy-4,6-dimethoxyphenyl)prop-2-en-1-one (6.36)

A mixture of 2,4-dimethoxybenzaldehyde (200 mg, 1.20 mmol), 2’hydroxy-4’6’ dimethoxylacetophenone (283 mg, 1.44 mmol) and KOH (810 mg, 14.4 mmol) in EtOH (20 mL) was reacted according to GP-2. The crude residue was purified by recrystallization from EtOH to afford chalcone sc028 (107 mg, 0.31 mmol, 26%) as a yellow solid.

-1 TLC Rf = 0.49 (1:1 EtOAc/Hex); m.p. 127-130 °C; IR νmax (neat)/cm : 2945 w, 1604 (C=O), 1545 (C=C), 1503 (C=C), 1438 m, 1340 m, 1293 m, 1276 s, 1208 s, 1160 s, 1106 s, 1028 s; 1H NMR (400

MHz, CDCl3): δ 3.82 (3H, s, -OCH3), 3.89 (3H, s, -OCH3), 3.89 (3H, s, -OCH3), 3.90 (3H, s, -OCH3), 5.95 (1H, d, J = 2.4 Hz, ArH5), 6.09 (1H, d, J= 2.4 Hz, ArH3), 6.46 (1H, d, J = 2.4 Hz, ArH3’), 6.52 (1H, dd, J = 8.6, 2.4 Hz, ArH5’), 7.54 (1H, d, J = 8.6 Hz, ArH6’), 7.90 (1H, d, J = 15.8 Hz, -CH=CHCO-), 8.10

13 (1H, d, J = 15.8 Hz, -CH=CHCO-), 14.54 (1H, s, OH); C NMR (101 MHz, CDCl3): δ 55.5 (-OCH3), 55.5

(-OCH3), 55.6 (-OCH3), 55.7 (-OCH3), 91.2 (ArC5), 93.8 (ArC3), 98.4 (ArCH3’), 105.4 (ArC5’), 106.5 (ArCq), 117.8 (ArCq), 125.4 (-CH=CHCO-), 130.5 (ArC6’), 138.3 (-CH=CHO-), 160.2 (ArCq), 162.4 (ArCq), 162.8 (ArCq), 165.8, (ArCq), 168.3, (ArCq), 193.0 (C=O); HRMS (ESI) m/z = 367.1136

+ + [M+Na] found, C19H20O6Na required 367.1152.

215

1-(2-hydroxy-4,6-dimethoxyphenyl)-3-(1H-indol-3-yl)propan-1-one (6.37)

To a mixture of TJS285 (150 mg, 0.46 mmol) in dry MeOH (15 mL) under N2, was added platinum(IV) oxide (2 mol%). The reaction was stirred at rt. under a hydrogen atmosphere for 24 h. The resulting mixture was filtered under gravity and washed with MeOH (3 × 10 mL). The combined organic layers were concentrated in vacuo and the crude residue was purified by column chromatography (2:1 EtOAc/Hex) to afford sc032 (79.9 mg, 0.28 mmol, 60%) as a pale yellow solid.

-1 TLC Rf = 0.69 (1:1 EtOAc/Hex); m.p. 120–122 °C; IR νmax (neat)/cm : 3407 m (O-H), 2912 w, 1613 s

1 (C=O), 1589 (C=C), 1436 m, 1410 m, 1204 s, 1154 s, 1114 s, 1036 m; H NMR (400 MHz, CDCl3): δ

3.14 (2H, t, J = 7.4 Hz, CH2), 3.41 (2H, t, J = 7.4, CH2), 3.79 (3H, s, -OCH3), 3.89 (3H, s, -OCH3), 5.91 (1H, d, J = 2.3 Hz, ArH5), 6.07 (1H, d, J = 2.3 Hz, ArH3), 7.02 (1H, m, ArH8’), 7.12 (1H, td, J = 7.8, 1.1 Hz, H4’), 7.19 (1H, td, J = 7.8, 1.1 Hz, H5’), 7.36 (1H, d, J = 7.8 Hz, H6’), 7.64 (1H, d, J = 7.8 Hz, H3’),

13 7.93 (1H, s, br, NH), 14.09 (1H, s, OH); C NMR (101 MHz, CDCl3): δ 20.0 (CH2), 44.6 (CH2), 55.5 (-

OCH3), 55.6 (-OCH3), 90.8 (ArC5), 93.6 (ArC3), 105.9 (ArCq), 111.1 (ArC6’), 116 (ArCq), 118.9 (ArC3’), 119.2 (ArC4’), 121.4 (ArC8’), 121.9 (ArC5’), 127.5 (ArCq), 136.3 (ArC9), 162.0 (ArCq), 165.9

+ (ArCq), 167.6 (ArCq), 205.1 (C=O) ppm; HRMS (ESI) m/z = 324.3558 [M+H] found, C19H18NO4 required 324.3565.

216

References

(1) Strassnig, M.; Ganguli, M. Psychiatry 2005, 412, 30–33. (2) Boller, F.; Forbes, M. M. J. Neurol. Sci. 1998, 158 (2), 125–133. (3) Berchtold, N. C.; Cotman, C. W. Neurobiol. Aging 1998, 19 (3), 173–189. (4) Mahendra, B. Dementia; Springer: Dordrecht, 1987. (5) Parkin, T. G. Old age in the Roman world : A cultural and social history; JHU Press: Baltimore, 2003. (6) Kayed, R.; Head, E.; Sarsoza, F.; Saing, T.; Cotman, C. W.; Necula, M.; Margol, L.; Wu, J.; Breydo, L.; Thompson, J. L.; Rasool, S.; Gurlo, T.; Butler, P.; Glabe, C. G. Mol. Neurodegener. 2007, 2 (18). (7) Brodaty, H.; Tariot, P.; Kaye, J.; Chui, H.; Erkinjuntti, T. Evidence based Dementia Practice; Qizilbash, N., Schneide, L. S., Eds.; Wiley-Blackwell: New Jersey, 2002. (8) Wilks, S. Br. J. Psychiatry 1864, 10 (51), 381–392. (9) Buda, O.; Arsene, D.; Ceausu, M.; Dermengiu, D.; Curca, G. C. Neurology 2009, 72 (1), 88– 91. (10) De Carlos, J. A.; Borrell, J. Brain Res. Rev. 2007, 55 (1), 8–16. (11) Chambers, W. Br. J. Psychiatry 1928, 74 (306), 533. (12) Philip Schwartz. Amyloidosis; cause and manifestation of senile deterioration; Charles C Thomas Publisher: Springfield, 1970. (13) Schnabel, J. Nature 2011, 475, S12–S14. (14) Glenner, G. G.; Wong, C. W. Biochem. Biophys. Res. Commun. 1984, 120 (3), 885–890. (15) Glenner, G. G.; Wong, C. W. Biochem. Biophys. Res. Commun. 1984, 122 (3), 1131–1135. (16) Masters, C. L.; Simms, G.; Weinman, N. A.; Multhaup, G.; McDonald, B. L.; Beyreuther, K. Proc. Natl. Acad. Sci. U. S. A. 1985, 82 (12), 4245–4249. (17) Kang, J.; Lemaire, H.-G.; Unterbeck, A.; Salbaum, J. M.; Masters, C. L.; Grzeschik, K.-H.; Multhaup, G.; Beyreuther, K.; Müller-Hill, B. Nature 1987, 325 (6106), 733–736. (18) Hardy, J. FEBS J. 2017, 284 (7), 1040–1044. (19) John, H.; Gerald, H. Science, 1992, 256 (6), 184–185. (20) Querfurth, H. W.; LaFerla, F. M. N. Engl. J. Med. 2010, 362 (4), 329–344. (21) Alzheimer’s Association. Alzheimers Dement. 2017, 13 (2), 325–373. (22) Masters, C. L.; Bateman, R.; Blennow, K.; Rowe, C. C.; Sperling, R. A.; Cummings, J. L. Nat. Rev. Dis. Prim. 2015, 15056. (23) Dobson, C. M. Nature 2003, 426 (6968), 884–890. (24) Parsons, C. G.; Rammes, G. Expert Opin. Investig. Drugs 2017, 26 (5), 579–592. (25) Doig, A. J.; del Castillo-Frias, M. P.; Berthoumieu, O.; Tarus, B.; Nasica-Labouze, J.;

217

Sterpone, F.; Nguyen, P. H.; Hooper, N. M.; Faller, P.; Derreumaux, P. ACS Chem. Neurosci. 2017, 8 (7), 1435–1437. (26) Citron, M. Nat. Rev. Drug Discov. 2010, 9 (5), 387–398. (27) Selkoe, D. J.; Hardy, J. EMBO Mol. Med. 2016, 8 (6), 595–608. (28) Wolfe, M. S. Developing Therapeutics for Alzheimer ’ S Disease; Academic Press: San Diego, 2016. (29) Hamley, I. W. Chem. Rev. 2012, 112 (10), 5147–5192. (30) Rauk, A. Chem. Soc. Rev. 2009, 38 (9), 2698–2715. (31) Esler, W. P.; Wolfe, M. S. Science 2001, 293 (5534), 1449–1454. (32) Wiltfang, J.; Esselmann, H.; Bibl, M.; Hüll, M.; Hampel, H.; Kessler, H.; Frölich, L.; Schröder, J.; Peters, O.; Jessen, F.; Zimmermann, R.; Bruckmoser, R.; Kornhuber, J.; Lewczuk, P. J. Neurochem 2007, 101 (4), 1053–1059. (33) Tanzi, R. E.; Bertram, L. Cell 2005, 120 (4), 545–555. (34) Hane, F. T.; Lee, B. Y.; Petoyan, A.; Rauk, A.; Leonenko, Z. Biosens. Bioelectron. 2014, 54, 492–498. (35) Eanes, E. D.; Glenner, G. G. J. Histochem. Cytochem. 1968, 16 (11), 673–677. (36) Knowles, T. P.; Fitzpatrick, A. W.; Meehan, S.; Mott, H. R.; Vendruscolo, M.; Dobson, C. M.; Welland, M. E. Science. 2007, 318 (5858), 1900–1903. (37) Eisenberg, D.; Jucker, M. Cell 2012, 148 (6), 1188–1203. (38) Volpatti, L. R.; Vendruscolo, M.; Dobson, C. M.; Knowles, T. P. J. ACS Nano 2013, 7 (12), 10443–10448. (39) Toyama, B. H.; Weissman, J. S. Annu. Rev. Biochem. 2011, 80 (1), 557–585. (40) Sciacca, M. F. M.; Kotler, S. A.; Brender, J. R.; Chen, J.; Lee, D. K.; Ramamoorthy, A. Biophys. J. 2012, 103 (4), 702–710. (41) Kawarabayashi, T.; Shoji, M.; Younkin, L. H.; Wen-Lang, L.; Dickson, D. W.; Murakami, T.; Matsubara, E.; Abe, K.; Ashe, K. H.; Younkin, S. G. J. Neurosci. 2004, 24 (15), 3801–3809. (42) Talantova, M.; Sanz-Blasco, S.; Zhang, X.; Buczynski, M. W.; Solomon, J. P.; Michael, S.; Powers, E. T.; Kelly, J. W.; Roberts, A.; Tong, G.; Fang-Newmeyer, T.; Parker, J.; Holland, E. A.; Zhang, D.; Nakanishi, N.; Chen, H.-S. V.; Wolosker, H.; Wang, Y.; Parsons, L. H.; Ambasudhan, R.; Masliah, E.; Heinemann, S. F.; Pina-Crespo, J. C.; Lipton, S. A. Proc. Natl. Acad. Sci. U. S. A. 2013, 110 (27), 2518–2527. (43) Um, J. W.; Nygaard, H. B.; Heiss, J. K.; Kostylev, M. a; Stagi, M.; Vortmeyer, A.; Wisniewski, T.; Gunther, E. C.; Strittmatter, S. M. Nat. Neurosci. 2012, 15 (9), 1227–1235. (44) Necula, M.; Kayed, R.; Milton, S.; Glabe, C. G. J. Biol. Chem. 2007, 282, 10311–10324. (45) Hardy, J.; Selkoe, D. J. Science 2002, 297 (5580), 353–356. (46) Glabe, C. G. Neurobiol. Aging 2006, 27 (4), 570–575. (47) Kirkitadze, M. D.; Bitan, G.; Teplow, D. B. J. Neurosci. Res. 2002, 69 (5), 567–577. (48) Caughey, B.; Lansbury, P. T. Annu. Rev. Neurosci. 2003, 26 (1), 267–298.

218

(49) Pangalos, M. N.; Schechter, L. E.; Hurko, O. Nat. Rev. Drug Discov. 2007, 6 (7), 521–532. (50) Cummings, J. L.; Morstorf, T.; Zhong, K. Alzheimers. Res. Ther. 2014, 6 (4), 37–43. (51) Wolfe, M. S. Nat. Rev. Drug Discov. 2002, 1 (11), 859–866. (52) Giuffrida, M. L.; Caraci, F.; Nicoletti, F.; Rizzarelli, E.; Copani, A. J. Neurosci. 2009, 29 (34), 10582–10587. (53) Liu, Y.-H.; Giunta, B.; Zhou, H.-D.; Tan, J.; Wang, Y.-J. Nat. Rev. Neurol. 2012, 8 (8), 465– 469. (54) Liu, J.; Yang, B.; Ke, J.; Li, W.; Suen, W. C. Drugs and Aging 2016, 33 (10), 685–697. (55) Schenk, D.; Barbour, R.; Dunn, W.; Gordon, G.; Grajeda, H.; Guido, T.; Hu, K.; Huang, J.; Johnson-Wood, K.; Khan, K.; Kholodenko, D.; Lee, M.; Liao, Z.; Lieberburg, I.; Motter, R.; Mutter, L.; Soriano, F.; Shopp, G.; Vasquez, N.; Vandevert, C.; Walker, S.; Wogulis, M.; Yednock, T.; Games, D.; Seubert, P. Nature 1999, 400 (6740), 173–177. (56) Knowles, T. P. J.; Vendruscolo, M.; Dobson, C. M. Nat. Rev. Mol. Cell Biol. 2014, 15 (6), 384–396. (57) Eichner, T.; Radford, S. E. Mol. Cell 2011, 43 (1), 8–18. (58) Glabe, C. G. J. Biol. Chem. 2008, 283 (44), 29639–29643. (59) Kodali, R.; Wetzel, R. Curr. Opin. Struct. Biol. 2007, 17 (1), 48–57. (60) Xiao, Y.; Ma, B.; McElheny, D.; Parthasarathy, S.; Long, F.; Hoshi, M.; Nussinov, R.; Ishii, Y. Nat. Struct. Mol. Biol. 2015, 22 (6), 499–505. (61) Ahmed, M.; Davis, J.; Aucoin, D.; Sato, T.; Ahuja, S.; Aimoto, S.; Elliott, J. I.; Van Nostrand, W. E.; Smith, S. O. Nat. Struct. Mol. Biol. 2010, 17 (5), 561–567. (62) Pham, J. D.; Chim, N.; Goulding, C. W.; Nowick, J. S. J. Am. Chem. Soc. 2013, 135 (33), 12460–12467. (63) Petkova, A. T.; Ishii, Y.; Balbach, J. J.; Antzutkin, O. N.; Leapman, R. D.; Delaglio, F.; Tycko, R. Proc. Natl. Acad. Sci. U. S. A. 2002, 99 (26), 16742–16747. (64) Paravastu, A. K.; Leapman, R. D.; Yau, W.-M.; Tycko, R. Proc. Natl. Acad. Sci. U. S. A. 2008, 105 (47), 18349–18354. (65) Tycko, R. Q. Rev. Biophys. 2006, 39 (1), 1–55. (66) Luhrs, T.; Ritter, C.; Adrian, M.; Riek-loher, D.; Bohrmann, B.; Do, H.; Schubert, D.; Riek, R. Proc. Natl. Acad. Sci. U. S. A. 2005, 102 (48), 17342–17347. (67) Wälti, M. A.; Ravotti, F.; Arai, H.; Glabe, C. G.; Wall, J. S.; Böckmann, A.; Güntert, P.; Meier, B. H.; Riek, R. Proc. Natl. Acad. Sci. U. S. A. 2016, 113 (34), E4976–E4984. (68) Colvin, M. T.; Silvers, R.; Ni, Q. Z.; Can, T. V.; Sergeyev, I.; Rosay, M.; Donovan, K. J.; Michael, B.; Wall, J.; Linse, S.; Griffin, R. G. J. Am. Chem. Soc. 2016, 138 (30), 9663–9674. (69) Eisenberg, D. S.; Sawaya, M. R. Proc. Natl. Acad. Sci. U. S. A. 2016, 113 (34), 9398–9400. (70) Arosio, P.; Vendruscolo, M.; Dobson, C. M.; Knowles, T. P. J. Trends Pharmacol. Sci. 2014, 35 (3), 127–135. (71) Cohen, S. I.; Linse, S.; Luheshi, L. M.; Hellstrand, E.; White, D. A.; Rajah, L.; Otzen, D. E.;

219

Vendruscolo, M.; Dobson, C. M.; Knowles, T. P. J. Proc. Natl. Acad. Sci. 2013, 110 (24), 9758–9763. (72) Arosio, P.; Knowles, T. P. J.; Linse, S. Phys. Chem. Chem. Phys. 2015, 17 (12), 7606–7618. (73) Blake, Roberts E. Shorter, J. Nat. Struct. Mol. Biol. 2008, 15 (7196), 544–545. (74) Cohen, S. I. A.; Vendruscolo, M.; Welland, M. E.; Dobson, C. M.; Terentjev, E. M.; Knowles, T. P. J. J. Chem. Phys. 2011, 135 (6), 1–16. (75) Hellstrand, E.; Boland, B.; Walsh, D. M.; Linse, S. ACS Chem. Neurosci. 2010, 1 (1), 13–18. (76) Meisl, G.; Yang, X.; Hellstrand, E.; Frohm, B.; Kirkegaard, J. B.; Cohen, S. I. A.; Dobson, C. M.; Linse, S.; Knowles, T. P. J. Proc. Natl. Acad. Sci. U. S. A. 2014, 111 (26), 9384–9389. (77) Meisl, G.; Yang, X.; Frohm, B.; Knowles, T. P. J.; Linse, S. Sci. Rep. 2016, 6, 18728. (78) Habchi, J.; Arosio, P.; Perni, M.; Costa, A. R.; Yagi-Utsumi, M.; Joshi, P.; Chia, S. K. R.; Cohen, S. I. A.; Müller, M. B. .; Linse, S.; Nollen, E. A. A.; Dobson, C. M.; Knowles, T. P. J.; Vendruscolo, M. Sci. Adv. 2016, e1501244. (79) Habchi, J.; Chia, S.; Limbocker, R.; Mannini, B.; Ahn, M.; Perni, M.; Hansson, O.; Arosio, P.; Kumita, J. R.; Challa, P. K.; Cohen, S. I. A.; Linse, S.; Dobson, C. M.; Knowles, T. P. J.; Vendruscolo, M. Proc. Natl. Acad. Sci. U. S. A. 2016, 114 (2), e200–e208. (80) Cohen, S. I. A.; Arosio, P.; Presto, J.; Kurudenkandy, F. R.; Biverstål, H.; Dolfe, L.; Dunning, C.; Yang, X.; Frohm, B.; Vendruscolo, M.; Johansson, J.; Dobson, C. M.; Fisahn, A.; Knowles, T. P. J.; Linse, S. Nat. Struct. Mol. Biol. 2015, 22 (3), 207–213. (81) Arosio, P.; Michaels, T. C. T.; Linse, S.; Månsson, C.; Emanuelsson, C.; Presto, J.; Johansson, J.; Vendruscolo, M.; Dobson, C. M.; Knowles, T. P. J. Nat. Commun. 2016, 7 (10948). (82) Abelein, A.; Gräslund, A.; Danielsson, J. Proc. Natl. Acad. Sci. 2015, 112 (17), 5407–5412. (83) Barnham, K. J.; Masters, C. L.; Bush, A. I. Nat. Rev. Drug Discov. 2004, 3 (3), 205–214. (84) Andersen, J. K. Nat. Med. 2004, 10, S18–S25. (85) Mariani, E.; Polidori, M. C.; Cherubini, A.; Mecocci, P. J. Chromatogr. B Anal. Technol. Biomed. Life Sci. 2005, 827 (1), 65–75. (86) Butterfield, D. A.; Drake, J.; Pocernich. Chava; Castegna, A. Trends Mol. Med. 2001, 7 (12), 548–554. (87) Swomley, A. M.; Förster, S.; Keeney, J. T.; Triplett, J.; Zhang, Z.; Sultana, R.; Butterfield, D. A. Biochim. Biophys. Acta 2015, 1842 (8), 1248–1257. (88) Markesbery, W. R. Free Radic. Biol. Med. 1997, 23 (1), 134–147. (89) Mattson, M. P. Nature 2004, 431 (7004), 107–107. (90) Rival, T.; Page, R. M.; Chandraratna, D. S.; Sendall, T. J.; Ryder, E.; Liu, B.; Lewis, H.; Rosahl, T.; Hider, R.; Camargo, L. M.; Shearman, M. S.; Crowther, D. C.; Lomas, D. A. Mol. Dev. Neurosci. 2009, 29, 1335–1347. (91) Herrup, K. Nat. Neurosci. 2015, 18 (6), 794–799. (92) Karran, E.; Hardy, J. Ann. Neurol. 2014, 76 (2), 185–205.

220

(93) Terry, R. D.; Masliah, E.; Salmon, D. P.; Butters, N.; DeTeresa, R.; Hill, R.; Hansen, L. A.; Katzman, R. Ann. Neurol. 1991, 30 (4), 572–580. (94) Braak, E.; Braak, H. Neurobiol. Aging 1995, 16 (3), 271–278. (95) Bulic, B.; Pickhardt, M.; Mandelkow, E. J. Med. Chem. 2013, 56, 4135–4155. (96) Shirahama, T.; Cohen, A. S. J. Cell Biol. 1967, 33 (3), 679–708. (97) Goldsbury, C.; Baxa, U.; Simon, M. N.; Steven, A. C.; Engel, A.; Wall, J. S.; Aebi, U.; Müller, S. A. J. Struct. Biol. 2011, 173 (1), 1–13. (98) Harris, J. R. Subcell. Biochem. 2005, 38, 1–44. (99) Duim, W. C.; Chen, B.; Frydman, J.; Moerner, W. E. Chemphyschem 2011, 12 (13), 2387– 2390. (100) Mahou, P.; Curry, N.; Pinotsi, D.; Kaminski Schierle, G.; Kaminski, C. Proc. SPIE 2015, 9331. (101) Pinotsi, D.; Buell, A. K.; Galvagnion, C.; Dobson, C. M.; Kaminski Schierle, G. S.; Kaminski, C. F.; Christopher, M.; Schierle, G. S. K.; Kaminski, C. F.; Dobson, C. M.; Kaminski Schierle, G. S.; Kaminski, C. F. Nano Lett. 2014, 14 (1), 339–345. (102) Kaminski Schierle, G. S.; Van De Linde, S.; Erdelyi, M.; Esbjörner, E. K.; Klein, T.; Rees, E.; Bertoncini, C. W.; Dobson, C. M.; Sauer, M.; Kaminski, C. F. J. Am. Chem. Soc. 2011, 133 (33), 12902–12905. (103) Ries, J.; Udayar, V.; Soragni, A.; Hornemann, S.; Nilsson, K. P. R.; Riek, R.; Hock, C.; Ewers, H.; Aguzzi, A. A.; Rajendran, L. ACS Chem. Neurosci. 2013, 4 (7), 1057–1061. (104) Barrow, C. J.; Zagorski, M. G. Science 1991, 253 (5016), 179–182. (105) Juszczyk, P.; Kołodziejczyk, A. S.; Grzonka, Z. Acta Biochim. Pol. 2005, 52 (2), 425–431. (106) Bartolini, M.; Bertucci, C.; Bolognesi, M. L.; Cavalli, A.; Melchiorre, C.; Andrisano, V. ChemBioChem 2007, 8 (17), 2152–2161. (107) Matveeva, E. G.; Rudolph, A.; Moll, J. R.; Thompson, R. B. ACS Chem. Neurosci. 2012, 3 (11), 982–987. (108) Allsop, D.; Swanson, L.; Moore, S.; Davies, Y.; York, A.; El-Agnaf, O. M.; Soutar, I. Biochem. Biophys. Res. Commun. 2001, 285 (1), 58–63. (109) Sepkhanova, I.; Drescher, M.; Meeuwenoord, N. J.; Limpens, R. W. A. L.; Koning, R. I.; Filippov, D. V.; Huber, M. Appl. Magn. Reson. 2009, 36 (2), 209–222. (110) Ionuţ Iuraşcu, M.; Cozma, C.; Tomczyk, N.; Rontree, J.; Desor, M.; Drescher, M.; Przybylski, M. Anal. Bioanal. Chem. 2009, 395 (8), 2509–2519. (111) Howie, A. J.; Brewer, D. B. Micron 2009, 40 (3), 285–301. (112) Lorenzo, A.; Yankner, B. A. Proc. Natl. Acad. Sci. 1994, 91 (25), 12243–12247. (113) Groenning, M. J. Chem. Biol. 2010, 3 (1), 1–18. (114) LeVine, H. Protein Sci. 1993, 2 (3), 404–410. (115) Xue, C.; Lin, T. Y.; Chang, D.; Guo, Z. R. Soc. Open Sci. 2017, 4 (160696). (116) Biancalana, M.; Koide, S. Biochim. Biophys. Acta - Proteins Proteomics 2010, 1804 (7), 1405–1412.

221

(117) Wu, C.; Wang, Z.; Lei, H.; Duan, Y.; Bowers, M. T.; Shea, J. E. J. Mol. Biol. 2008, 384 (3), 718–729. (118) Kuang, G.; Murugan, N. A.; Tu, Y.; Nordberg, A.; Ågren, H. J. Phys. Chem. B 2015, 119 (35), 11560–11567. (119) Rodríguez-Rodríguez, C.; Rimola, A.; Rodríguez-Santiago, L.; Ugliengo, P.; Alvarez-Larena, A.; Gutiérrez-de-Terán, H.; Sodupe, M.; González-Duarte, P. Chem. Comm. 2010, 46 (7), 1156–1158. (120) Lindberg, D. J.; Wranne, M. S.; Gilbert Gatty, M.; Westerlund, F.; Esbjörner, E. K. Biochem. Biophys. Res. Commun. 2015, 458 (2), 418–423. (121) Hudson, S. A.; Ecroyd, H.; Kee, T. W.; Carver, J. FEBS J. 2009, 276 (20), 5960–5972. (122) Arosio, P.; Cukalevski, R.; Frohm, B.; Knowles, T. P. J.; Linse, S. J. Am. Chem. Soc. 2014, 136 (1), 219–225. (123) Giehm, L.; Otzen, D. E. Anal. Biochem. 2010, 400 (2), 270–281. (124) Garai, K.; Frieden, C. Proc. Natl. Acad. Sci. U. S. A. 2013, 110 (9), 3321–3326. (125) Zou, L.; Harkey, M. R.; Henderson, G. L. Phytomedicine 2002, 9 (3), 263–267. (126) Noormägi, A.; Primar, K.; Tõugu, V.; Palumaa, P. J. Pept. Sci. 2012, 18 (1), 59–64. (127) Simeonov, A.; Jadhav, A.; Thomas, C. J.; Wang, Y.; Huang, R.; Southall, N. T.; Shinn, P.; Smith, J.; Austin, C. P.; Auld, D. S.; Inglese, J. J. Med. Chem. 2008, 51 (8), 2363–2371. (128) Klingstedt, T.; Åslund, A.; Simon, R. A.; Johansson, L. B. G.; Mason, J. J.; Nyström, S.; Hammarström, P.; Nilsson, K. P. R. Org. Biomol. Chem. 2011, 9 (24), 8356–8370. (129) Klingstedt, T.; Nilsson, K. P. R. Biochem. Soc. Trans. 2012, 40 (4), 704–710. (130) Sigurdson, C. J.; Grathwohl, S.; Hof, P. R.; Aguzzi, A.; Hammarstro, P. ACS Chem. Biol. 2009, 4 (8), 673–684. (131) Bruggink, K. A.; Müller, M.; Kuiperij, H. B.; Verbeek, M. M. J. Alzheimers. Dis. 2012, 28 (4), 735–758. (132) Wanker, E. E.; Scherzinger, E.; Heiser, V.; Sittler, A.; Eickhoff, H.; Lehrach, H. Methods Enzymol. 1999, 309 (1983), 375–386. (133) Nasir, I.; Linse, S.; Cabaleiro-Lago, C. ACS Chem. Neurosci. 2015, 6 (8), 1436–1444. (134) Heiser, V.; Engemann, S.; Bröcker, W.; Dunkel, I.; Boeddrich, A.; Waelter, S.; Nordhoff, E.; Lurz, R.; Schugardt, N.; Rautenberg, S.; Herhaus, C.; Barnickel, G.; Böttcher, H.; Lehrach, H.; Wanker, E. E. Proc. Natl. Acad. Sci. U. S. A. 2002, 99, 16400–16406. (135) Grasso, G. Mass Spectrom. Rev. 2010, 30 (3), 347–365. (136) Bleiholder, C.; Dupuis, N. F.; Wyttenbach, T.; Bowers, M. T. Nat. Chem. 2011, 3 (2), 172– 177. (137) Zovo, K.; Helk, E.; Karafin, A.; Tougu, V.; Palumaa, P. Anal. Chem. 2010, 82 (20), 8558– 8565. (138) Young, L. M.; Saunders, J. C.; Mahood, R. a; Revill, C. H.; Foster, R. J.; Tu, L.-H.; Raleigh, D. P.; Radford, S. E.; Ashcroft, A. E. Nat. Chem. 2015, 7 (1), 73–81.

222

(139) Phillips, A. S.; Gomes, A. F.; Kalapothakis, J. M. D.; Gillam, J. E.; Gasparavicius, J.; Gozzo, F. C.; Kunath, T.; MacPhee, C.; Barran, P. E. Analyst 2015, 140 (9), 3070–3081. (140) Saunders, J. C.; Young, L. M.; Mahood, R. A.; Jackson, M. P.; Revill, C. H.; Foster, R. J.; Smith, D. A.; Ashcroft, A. E.; Brockwell, D. J.; Radford, S. E. Nat. Chem. Biol. 2015, 12 (2), 94–101. (141) Zhao, W.-Q.; Toolan, D.; Hepler, R. W.; Wolfe, A. L.; Yu, Y.; Price, E.; Uebele, V. N.; Schachter, J. B.; Reynolds, I. J.; Renger, J. J.; McCampbell, A.; Ray, W. J. J. Alzheimers. Dis. 2011, 25 (4), 655–669. (142) Hillen, H.; Barghorn, S.; Striebinger, A.; Labkovsky, B.; Muller, R.; Nimmrich, V.; Nolte, M. W.; Perez-Cruz, C.; van der Auwera, I.; Van Leuven, F.; van Gaalen, M.; Bespalov, A. Y.; Schoemaker, H.; Sullivan, J. P.; Ebert, U. J. Neurosci. 2010, 30 (31), 10369–10379. (143) Lambert, M. P.; Velasco, P. T.; Chang, L.; Viola, K. L.; Fernandez, S.; Lacor, P. N.; Khuon, D.; Gong, Y.; Bigio, E. H.; Shaw, P.; De Felice, F. G.; Krafft, G. A.; Klein, W. L. J. Neurochem. 2007, 100 (1), 23–35. (144) Kayed, R.; Head, E.; Thompson, J. L.; McIntire, T. M.; Milton, S. C.; Cotman, C. W.; Glabe, C. G. Science, 2003, 300 (5618), 486–489. (145) Lambert, M. P.; Viola, K. L.; Chromy, B. A.; Chang, L.; Morgan, T. E.; Yu, J.; Venton, D. L.; Krafft, G. A.; Finch, C. E.; Klein, W. L. J. Neurochem. 2001, 79 (3), 595–605. (146) O’Nuallain, B.; Wetzel, R. Proc. Natl. Acad. Sci. U. S. A. 2002, 99 (3), 1485–1490. (147) Bradbury, A.; Pluckthun, A. Nature 2015, 518 (7537), 27–29. (148) LeVine, H. Anal. Biochem. 2006, 356 (2), 265–272. (149) Kim, W.; Kim, Y.; Min, J.; Kim, D. J.; Chang, Y. T.; Hecht, M. H. ACS Chem. Biol. 2006, 1 (7), 461–469. (150) Villar-Pique, A.; Espargaro, A.; Sabate, R.; de Groot, N.; Ventura, S.; Villar-piqué, A.; Espargaró, A.; Sabaté, R.; Sdegroot, N.; Ventura, S. Microb. Cell Fact. 2012, 11 (55). (151) Lindberg, H.; Sandersjöö, L.; Meister, S.; Uhlén, M.; Löfblom, J.; Ståhl, S. Biotech Method 2012, 12 (1600364). (152) Härd, T.; Lendel, C. J. Mol. Biol. 2012, 421 (4–5), 441–465. (153) Goure, W. F.; Krafft, G. A.; Jerecic, J.; Hefti, F. Alzheimers. Res. Ther. 2014, 6 (4), 42. (154) Klein, W. L.; Krafft, G. A.; Finch, C. E. Trends Neurosci. 2001, 24, 219–224. (155) Young, L. M.; Ashcroft, A. E.; Radford, S. E. Curr. Opin. Chem. Biol. 2017, 39, 90–99. (156) Benilova, I.; Karran, E.; De Strooper, B. Nat. Neurosci. 2012, 15 (3), 349–357. (157) Cremades, N.; Cohen, S. I. A.; Deas, E.; Abramov, A. Y.; Chen, A. Y.; Orte, A.; Sandal, M.; Clarke, R. W.; Dunne, P.; Aprile, F. A.; Bertoncini, C. W.; Wood, N. W.; Knowles, T. P. J.; Dobson, C. M.; Klenerman, D. Cell 2012, 149 (5), 1048–1059. (158) Nguyen, P.; Derreumaux, P. Acc. Chem. Res. 2014, 47 (2), 603–611. (159) Ghanta, J.; Shen, C. L.; Kiessling, L. L.; Murphy, R. M. J. Biol. Chem. 1996, 271 (47), 29525– 29528. (160) Tjernberg, L. O.; Na, J.; Lindqvist, F.; Johansson, J.; Karlstro, A. R.; Thyberg, J. J. Biol. Chem.

223

1996, 271 (15), 8545–8548. (161) Turner, J. P.; Lutz-Rechtin, T.; Moore, K. A.; Rogers, L.; Bhave, O.; Moss, M. a; Servoss, S. L. ACS Chem. Neurosci. 2014, 5 (7), 552–558. (162) Arai, T.; Sasaki, D.; Araya, T.; Sato, T.; Sohma, Y.; Kanai, M. Chembiochem 2014, 15 (17), 2577–2583. (163) Kaffy, J.; Brinet, D.; Soulier, J.-L.; Khemtémourian, L.; Lequin, O.; Taverna, M.; Crousse, B.; Ongeri, S. Eur. J. Med. Chem. 2014, 86, 752–758. (164) Rajasekhar, K.; Suresh, S. N.; Manjithaya, R.; Govindaraju, T. Sci. Rep. 2015, 5, 8139–8138. (165) Mansson, C.; Arosio, P.; Hussein, R.; Kampinga, H. H.; Hashem, R. M.; Boelens, W. C.; Dobson, C. M.; Knowles, T. P. J.; Linse, S.; Emanuelsson, C. J. Biol. Chem. 2014, 289 (45), 31066–31076. (166) Gao, X.; Carroni, M.; Nussbaum-Krammer, C.; Mogk, A.; Nillegoda, N. B.; Szlachcic, A.; Guilbride, D. L.; Saibil, H. R.; Mayer, M. P.; Bukau, B. Mol. Cell 2015, 59 (5), 781–793. (167) Schreiber, S. L. Nat. Chem. Biol. 2005, 1 (2), 64–66. (168) Bu, X. Le; Rao, P. P. N.; Wang, Y. J. Mol. Neurobiol. 2016, 53 (6), 3565–3575. (169) Choi, D. Y.; Lee, Y. J.; Hong, J. T.; Lee, H. J. Brain Res. Bull. 2012, 87, 144–153. (170) McKoy, A. F.; Chen, J.; Schupbach, T.; Hecht, M. H. J. Biol. Chem. 2012, 287 (46), 38992– 39000. (171) Konrat, R. J. Magn. Reson. 2014, 241 (1), 74–85. (172) Lemkul, J. A.; Bevan, D. R. ACS Chem. Neurosci. 2012, 3 (11), 845–856. (173) Ladiwala, A. R. A.; Dordick, J. S.; Tessier, P. M. J. Biol. Chem. 2011, 286 (5), 3209–3218. (174) Ehrnhoefer, D. E.; Bieschke, J.; Boeddrich, A.; Herbst, M.; Masino, L.; Lurz, R.; Engemann, S.; Pastore, A.; Wanker, E. E. Nat. Struct. Mol. Biol. 2008, 15 (6), 558–566. (175) Palhano, F. L.; Lee, J.; Grimster, N. P.; Kelly, J. W. J. Am. Chem. Soc. 2013, 135 (20), 7503– 7510. (176) Lopez Del Amo, J. M.; Fink, U.; Dasari, M.; Grelle, G.; Wanker, E. E.; Bieschke, J.; Reif, B. J. Mol. Biol. 2012, 421 (4–5), 517–524. (177) Klunk, W. E.; Debnath, M. L.; Koros, A. M. C.; Pettegrew, J. W. Life Sci. 1998, 63 (20), 1807–1814. (178) Yang, F.; Lim, G. P.; Begum, A. N.; Ubeda, O. J.; Simmons, M. R.; Ambegaokar, S. S.; Chen, P. P.; Kayed, R.; Glabe, C. G.; Frautschy, S. A.; Cole, G. M. J. Biol. Chem. 2005, 280, 5892– 5901. (179) Yanagisawa, D.; Taguchi, H.; Yamamoto, A.; Shirai, N.; Hirao, K.; Tooyama, I. J. Alzheimers Dis. 2011, 24, 33–42. (180) Hamaguchi, T.; Ono, K.; Murase, A.; Yamada, M. Am. J. Pathol. 2009, 175 (6), 2557–2565. (181) Caesar, I.; Jonson, M.; Nilsson, K. P. R.; Thor, S.; Hammarstro, P. PLoS One 2012, 7 (2), e31424. (182) Ladiwala, A. R. a; Lin, J. C.; Bale, S. S.; Marcelino-Cruz, A. M.; Bhattacharya, M.; Dordick, J.

224

S.; Tessier, P. M. J. Biol. Chem. 2010, 285 (31), 24228–24237. (183) Jang, J.-H.; Surh, Y.-J. Free Radic. Biol. Med. 2003, 34 (8), 1100–1110. (184) Feng, Y.; Wang, X. P.; Yang, S. G.; Wang, Y. jiong; Zhang, X.; Du, X. T.; Sun, X. X.; Zhao, M.; Huang, L.; Liu, R. T. Neurotoxicology 2009, 30 (6), 986–995. (185) Ge, J. F.; Qiao, J. P.; Qi, C. C.; Wang, C. W.; Zhou, J. N. Neurochem. Int. 2012, 61 (7), 1192– 1201. (186) Porat, Y.; Abramowitz, A.; Gazit, E. Chem. Biol. Drug Des. 2006, 67 (1), 27–37. (187) Ono, K.; Li, L.; Takamura, Y.; Yoshiike, Y.; Zhu, L.; Han, F.; Mao, X.; Ikeda, T.; Takasaki, J. I.; Nishijo, H.; Takashima, A.; Teplow, D. B.; Zagorski, M. G.; Yamada, M. J. Biol. Chem. 2012, 287 (18), 14631–14643. (188) Lemkul, J. A.; Bevan, D. R. Biochemistry 2012, 51, 5990–6009. (189) Hanaki, M.; Murakami, K.; Akagi, K. I.; Irie, K. Bioorganic Med. Chem. 2016, 24 (2), 304– 313. (190) Du, W.-J.; Guo, J.-J.; Gao, M.; Hu, S.; Dong, X.; Han, Y.; Liu, F.-F.; Jiang, S.; Sun, Y. Sci. Rep. 2015, 5 (7992). (191) Gervais, F.; Paquette, J.; Morissette, C.; Krzywkowski, P.; Yu, M.; Azzi, M.; Lacombe, D.; Kong, X.; Aman, A.; Laurin, J.; Szarek, W. A.; Tremblay, P. Neurobiol. Aging 2007, 28 (4), 537–547. (192) Caltagirone, C.; Ferrannini, L.; Marchionni, N.; Nappi, G.; Scapagnini, G.; Trabucchi, M. Aging Clin. Exp. Res. 2012, 24 (6), 580–587. (193) Aisen, P. S.; Saumier, D.; Briand, R.; Laurin, J.; Gervais, F.; Tremblay, P.; Garceau, D. Neurology 2006, 67 (10), 1757–1763. (194) Kim, H. Y.; Kim, H. V.; Jo, S.; Lee, C. J.; Choi, S. Y.; Kim, D. J.; Kim, Y. Nat. Commun. 2015, 6, 8997–9008. (195) Cramer, P.; Cirrito, J.; Wesson, D. Science. 2012, 335 (6075), 1503–1506. (196) Tousi, B. Neuropsychiatr. Dis. Treat. 2015, 11, 311–315. (197) Fisher, S. K.; Novak, J. E.; Agranoff, B. W. J. Neurochem. 2002, 82 (4), 736–754. (198) McLaurin, J.; Golomb, R.; Jurewicz, A.; Antel, J. P.; Fraser, P. E. J. Biol. Chem. 2000, 275 (24), 18495–18502. (199) Salloway, S.; Sperling, R.; Keren, R.; Porsteinsson, A. P.; Van Dyck, C. H.; Tariot, P. N.; Gilman, S.; Arnold, D.; Abushakra, S.; Hernandez, C.; Crans, G.; Liang, E.; Quinn, G.; Bairu, M.; Pastrak, A.; Cedarbaum, J. M. Neurology 2011, 77 (13), 1253–1262. (200) Anderson, A. C. Chem. Biol. 2003, 10, 787–797. (201) Uversky, V. N. Expert Rev Proteomics 2010, 7 (4), 543–564. (202) Heller, G. T.; Aprile, F. A.; Vendruscolo, M. Cell. Mol. Life Sci. 2017, 74 (17), 1–19. (203) Reinke, A. A.; Gestwicki, J. E. Chem. Biol. Drug Des. 2007, 70 (3), 206–215. (204) Jiang, L.; Liu, C.; Leibly, D.; Landau, M.; Zhao, M.; Hughes, M. P.; Eisenberg, D. S. Elife 2013, No. 2, 1–27.

225

(205) Arai, T.; Araya, T.; Sasaki, D.; Taniguchi, A.; Sato, T.; Sohma, Y.; Kanai, M. Angew. Chemie - Int. Ed. 2014, 53 (31), 8236–8239. (206) Joshi, P.; Chia, S.; Habchi, J.; Knowles, T. P. J.; Dobson, C. M.; Vendruscolo, M. ACS Comb. Sci. 2016, 18 (3), 144–153. (207) Biancalana, M.; Makabe, K.; Koide, A.; Koide, S. J. Mol. Biol. 2009, 385 (4), 1052–1063. (208) Wang, S.-H.; Liu, F.-F.; Dong, X.-Y.; Sun, Y. J. Phys. Chem. B 2010, 114 (35), 11576–11583. (209) Skeby, K. K.; Sørensen, J.; Schiøtt, B. J. Am. Chem. Soc. 2013, 135 (40), 15114–15128. (210) Wu, C.; Wang, Z.; Lei, H.; Zhang, W.; Duan, Y. J. Am. Chem. Soc. 2007, 129 (5), 1225–1232. (211) Wang, Y.; Xia, Z.; Xu, J. R.; Wang, Y. X.; Hou, L. N.; Qiu, Y.; Chen, H. Z. Neuropharmacology 2012, 62 (2), 871–881. (212) Yang, Y.; Zhu, L.; Chen, X.; Zhang, H. J. Mol. Graph. Model. 2010, 29 (4), 538–545. (213) Lounnas, V.; Ritschel, T.; Kelder, J.; McGuire, R.; Bywater, R. P.; Foloppe, N. Comput. Struct. Biotechnol. J. 2013, 5 (6), e201302011. (214) Cohen, S. I. A.; Vendruscolo, M.; Dobson, C. M.; Knowles, T. P. J. J. Mol. Biol. 2012, 421, 160–171. (215) Arosio, P.; Meisl, G.; Andreasen, M.; Knowles, T. P. J. Proc. Natl. Acad. Sci. U. S. A. 2015, 112 (17), 5267–5268. (216) Toyn, J. Expert Rev. Clin. Pharmacol. 2015, 8 (3), 1–3. (217) Kepp, K. P. J. Alzheimer’s Dis. 2017, 55 (2), 447–457. (218) Drachman, D. A. Alzheimer’s Dement. 2014, 10 (3), 372–380. (219) Teplow, D. B. Alzheimers. Res. Ther. 2013, 5 (4), 39–51. (220) Hughes, J. P.; Rees, S. S.; Kalindjian, S. B.; Philpott, K. L. Br. J. Pharmacol. 2011, 162 (6), 1239–1249. (221) O’ Connor, C. J.; Beckmann, H. S. G.; Spring, D. R. Chem. Soc. Rev. 2012, 41 (12), 4444– 4456. (222) Galloway, W. R. J. D.; Isidro-Llobet, A.; Spring, D. R. Nat. Commun. 2010, 1 (6), 1–13. (223) Wells, J. A.; McClendon, C. L. Nature 2007, 450 (7172), 1001–1009. (224) Collins, S.; Bartlett, S.; Nie, F.; Sore, H. F.; Spring, D. R. Synth. 2016, 1457–1473. (225) Curatolo, W. Pharm. Sci. Technol. Today 1998, 1 (9), 387–393. (226) Franc, I.; Lipinski, A.; Feeney, P. J. Adv. Drug Deliv. Rev. 1997, 23, 3–25. (227) Rankovic, Z. J. Med. Chem. 2015, 58 (6), 2584–2608. (228) Pardridge, W. M. NeuroRX 2005, 2 (1), 3–14. (229) Banks, W. Adv. Drug Deliv. Rev. 2012, 64 (7), 629–639. (230) Kesselheim, A. S.; Hwang, T. J.; Franklin, J. M. Nat. Rev. Drug Discov. 2015, 14 (12), 815– 816. (231) Pajouhesh, H.; Lenz, G. R. NeuroRX 2005, 2 (4), 541–553.

226

(232) Rankovic, Z. J. Med. Chem. 2017, 60 (14), 5943–5954. (233) Wager, T. T.; Hou, X.; Verhoest, P. R.; Villalobos, A. ACS Chem. Neurosci. 2010, 1, 435– 449. (234) Lakowicz, J. R. Principles of fluorescence spectroscopy; 2006. (235) Stokes, G. G. Philos. Trans. R. Soc. London 1853, 143, 385–396. (236) Lavis, L. D.; Raines, R. T. ACS Chem. Biol. 2007, 3 (3), 142–155. (237) Toseland, C. P. J. Chem. Biol. 2013, 6 (3), 85–95. (238) Hawe, A.; Sutter, M.; Jiskoot, W. Pharm. Res. 2008, 25 (7), 1487–1499. (239) Giepmans, B. N. G.; Adams, S. R.; Ellisman, M. H.; Tsien, R. Y. Science. 2006, 312 (5771), 217–224. (240) Becker, W. J. Microsc. 2012, 247 (2), 119–136. (241) Suhling, K.; Hirvonen, L. M.; Levitt, J. A.; Chung, P.-H.; Tregidgo, C.; Le Marois, A.; Rusakov, D. A.; Zheng, K.; Ameer-Beg, S.; Poland, S.; Coelho, S.; Henderson, R.; Krstajic, N. Med. Photonics 2015, 27, 3–40. (242) Adília Lemos, M.; Sárniková, K.; Bot, F.; Anese, M.; Hungerford, G. Biosensors 2015, 5 (3), 367–397. (243) O’Connor, D. V.; Philips, D. Time-correlated single photon counting; Academic Press London, Ed.; London, 1984. (244) Becker, W. Advanced Time-Correlated Single Photon Counting Techniques, 1st ed.; Springer-Verlag, Ed.; Heidleberg, 2005. (245) Wahl, M. PicoQuant Technical Note. 2014, pp 1–14. (246) Becker, W. The BH TCSPC Handbook; 2010. (247) Kaminski, C. F.; Schierle, G. S. K. Neurophotonics 2016, 3 (4), 41807. (248) Burdick, D.; Soreghan, B.; Kwon, M.; Kosmoski, J.; Knauer, M.; Henschen, a; Yates, J.; Cotman, C.; Glabe, C. J. Biol. Chem. 1992, 267 (1), 546–554. (249) Jameson, L. P.; Smith, N. W.; Dzyuba, S. V. ACS Chem. Neurosci. 2012, 3 (11), 807–819. (250) Schierle, G. S. K.; Linde, S. Van De; Erdelyi, M.; Esbj, E. K.; Klein, T.; Rees, E.; Bertoncini, C. W.; Dobson, C. M.; Sauer, M.; Kaminski, C. F. J. Am. Chem. Soc. 2011, 133, 12902–12905. (251) Chan, F. T. S.; Kaminski Schierle, G. S.; Kumita, J. R.; Bertoncini, C. W.; Dobson, C. M.; Kaminski, C. F. Analyst 2013, 138 (7), 2156–2162. (252) Kaminski Schierle, G.; Bertoncini, C. W.; Chan, F. T. S.; Goot, A. T. Van Der; Schwedler, S.; Skepper, J.; Schlachter, S.; van Ham, T. J.; Esposito, A.; Kumita, J. R.; Nollen, E. A. A.; Dobson, C. M.; Kaminski, c. Chemphyschem 2011, 12 (3), 673–680. (253) Esbjörner, E. K.; Chan, F.; Rees, E.; Erdelyi, M.; Luheshi, L. M.; Bertoncini, C. W.; Kaminski, C. F.; Dobson, C. M.; Kaminski Schierle, G. S. Chem. Biol. 2014, 21 (6), 732–742. (254) Michel, C. H.; Kumar, S.; Pinotsi, D.; Tunnacliffe, A.; St. George-Hyslop, P.; Mandelkow, E. E.-M. E.-M.; Mandelkow, E. E.-M. E.-M.; Kaminski, C. F.; Kaminski Schierle, G. S.; St George-Hyslop, P.; Mandelkow, E. E.-M. E.-M.; Mandelkow, E. E.-M. E.-M.; Kaminski, C. F.;

227

Kaminski Schierle, G. S. J. Biol. Chem. 2014, 289 (2), 956–967. (255) Chen, W.; Young, L. J.; Lu, M.; Zaccone, A.; Strohl, F.; Yu, N.; Schierle, G. S. K.; Kaminski, C. F. Nano Lett. 2017, 17 (1), 143–149. (256) Teale, F. W.; Weber, G. Biochem. J. 1957, 65 (3), 476–482. (257) Rolinski, O. J.; Amaro, M.; Birch, D. J. S. Biosens. Bioelectron. 2010, 25 (10), 2249–2252. (258) Chan, F. T. S.; Pinotsi, D.; Kaminski Schierle, G. S.; Kaminski, C. F. In Bio-nanoimaging: Protein Misfolding and Aggregation; 2013; pp 147–155. (259) Pinotsi, D.; Buell, A. K.; Dobson, C. M.; Kaminski Schierle, G. S.; Kaminski, C. F. Chembiochem 2013, 14 (7), 846–850. (260) Pinotsi, D.; Grisanti, L.; Mahou, P.; Gebauer, R.; Kaminski, C. F.; Hassanali, A.; Kaminski Schierle, G. S. J. Am. Chem. Soc. 2016, 138 (9), 3046–3057. (261) van Ham, T. J.; Esposito, A.; Kumita, J. R.; Hsu, S. T. D.; Kaminski Schierle, G. S.; Kaminski, C. F.; Dobson, C. M.; Nollen, E. A. A.; Bertoncini, C. W. J. Mol. Biol. 2010, 395 (3), 627–642. (262) Kodali, R.; Wetzel, R. Curr. Opin. Struct. Biol. 2007, 17 (1), 48–57. (263) Quinn, S. D.; Dalgarno, P. a; Cameron, R. T.; Hedley, G. J.; Hacker, C.; Lucocq, J. M.; Baillie, G. S.; Samuel, I. D. W.; Penedo, J. C. Mol. Biosyst. 2014, 10 (1), 34–44. (264) Whitesides, G. M. Nature 2006, 442 (7101), 368–373. (265) Theberge, A. B.; Courtois, F.; Schaerli, Y.; Fischlechner, M.; Abell, C.; Hollfelder, F.; Huck, W. T. S. Angew. Chem. Int. Ed. Engl. 2010, 49 (34), 5846–5868. (266) Kintses, B.; van Vliet, L. D.; Devenish, S. R.; Hollfelder, F. Curr. Opin. Chem. Biol. 2010, 14 (5), 548–555. (267) Gielen, F.; Vliet, L. Van; Koprowski, B. T.; Devenish, S. R. A.; Fischlechner, M.; Edel, J. B.; Niu, X.; Andrew, J.; Hollfelder, F. Anal. Chem. 2013, 85 (9), 4761–4769. (268) Leng, J.; Salmon, J.-B. Lab Chip 2009, 9 (1), 24–34. (269) Sackmann, E. K.; Fulton, A. L.; Beebe, D. J. Nature 2014, 507 (7491), 181–189. (270) Zhang, Y.; Jiang, H. R. Anal. Chim. Acta 2016, 914, 7–16. (271) Prakadan, S. M.; Shalek, A. K.; Weitz, D. A. Nat. Rev. Genet. 2017, 18 (6), 345–361. (272) Shembekar, N.; Chaipan, C.; Utharala, R.; Merten, C. A. Lab Chip 2016, 16, 1314–1331. (273) Saar, K. L.; Yates, E. V.; Müller, T.; Saunier, S.; Dobson, C. M.; Knowles, T. P. J. Biophys. J. 2016, 110 (3), 555–560. (274) Horrocks, M. H.; Tosatto, L.; Dear, A. J.; Garcia, G. A.; Iljina, M.; Cremades, N.; Dalla Serra, M.; Knowles, T. P. J.; Dobson, C. M.; Klenerman, D. Anal. Chem. 2015, 87 (17), 8818–8826. (275) Brouzes, E.; Medkova, M.; Savenelli, N.; Marran, D.; Twardowski, M.; Hutchison, J. B.; Rothberg, J. M.; Link, D. R.; Perrimon, N.; Samuels, M. L. Proc. Natl. Acad. Sci. 2009, 106 (34), 14195–14200. (276) Mikkonen, S.; Jacksen, J.; Roeraade, J.; Thormann, W.; Emmer, A. Anal. Chem. 2016, 88 (20), 10044–10051. (277) Islam, K.; Jang, Y.-C.; Chand, R.; Jha, S. K.; Lee, H. H.; Kim, Y.-S. J. Nanosci. Nanotechnol.

228

2011, 11 (7), 5657–5662. (278) Meier, M.; Kennedy-Darling, J.; Choi, S. H.; Norstrom, E. M.; Sisodia, S. S.; Ismagilov, R. F. Angew. Chem. Int. Ed. Engl. 2009, 48 (8), 1487–1489. (279) Hamada, T.; Morita, M.; Kishimoto, Y.; Komatsu, Y.; Vestergaard, M.; Takagi, M. J. Phys. Chem. Lett. 2010, 1 (1), 170–173. (280) Clausell-Tormos, J.; Griffiths, A. D.; Merten, C. a. Lab Chip 2010, 10 (10), 1302–1307. (281) Zilionis, R.; Nainys, J.; Veres, A.; Savova, V.; Zemmour, D.; Klein, A. M.; Mazutis, L. Nat. Protoc. 2017, 12 (1), 44–73. (282) Shahi, P.; Kim, S. C.; Haliburton, J. R.; Gartner, Z. J.; Abate, A. R. Sci. Rep. 2017, 7 (44447). (283) Humar, M.; Seok, S.; Yun, H. Lab Chip 2017, 17, 2777–2784. (284) Miller, O. J.; El, A.; Mangeat, T.; Baret, J.; Frenz, L.; El, B.; Mayot, E.; Samuels, M. L.; Rooney, E. K.; Dieu, P.; Galvan, M.; Link, D. R.; Griffiths, A. D. Proc. Natl. Acad. Sci. 2012, 109 (2), 378–383. (285) Gielen, F.; Buryska, T.; Vliet, L. Van; Butz, M.; Damborsky, J.; Prokop, Z.; Hollfelder, F. Anal. Chem. 2015, 87 (87), 624–632. (286) Huebner, A.; Sharma, S.; Srisa-Art, M.; Hollfelder, F.; Edel, J. B.; Demello, A. J. Lab Chip 2008, 8 (8), 1244–1254. (287) Crowther, DC, Hollfelder, F., Gielen, F., van Vliet, L. , Yan, J., Fischlechner, M., Kaminski, C. and Kaminski, G. WO 2014132053 A1, 2014. (288) Pluen, A.; Netti, P. A.; Jain, R. K.; Berk, D. A. Biophys. J. 1999, 77 (1), 542–552. (289) Pernodet, N. Electrophoresis 1998, 19 (19), 1606–1610. (290) Herzenberg, L. A.; Parks, D.; Sahaf, B.; Perez, O.; Roederer, M.; Herzenberg, L. A. Clin. Chem. 2002, 48 (10), 1819–1827. (291) Yan, J. Analysis of Amyloid Aggregation on Water-based Microfluidics Platform. Ph.D. Dissertation, University of Cambridge, 2014. (292) Hortschansky, P.; Schroeckh, V.; Christopeit, T.; Zandomeneghi, G.; Fändrich, M. Protein Sci. 2005, 14, 1753–1759. (293) Knowles, T. P. J.; White, D. A.; Abate, A. R.; Agresti, J. J.; Cohen, S. I. A.; Sperling, R. A.; De Genst, E. J.; Dobson, C. M.; Weitz, D. A. Proc. Natl. Acad. Sci. 2011, 108 (36), 14746– 14751. (294) Harper, J. D.; Lansbury, P. T. Annu. Rev. Biochem. 1997, 66, 385–407. (295) Lih-Fen Lue; Guerra, A.; Walker, D. G. Neurol. Ther. 2017, 6 (1), 25–36. (296) Lue, L. F.; Kuo, Y. M.; Roher, A. E.; Brachova, L.; Shen, Y.; Sue, L.; Beach, T.; Kurth, J. H.; Rydel, R. E.; Rogers, J. Am. J. Pathol. 1999, 155 (3), 853–862. (297) Lee, C. F.; Bird, S.; Shaw, M.; Jean, L.; Vaux, D. J. J. Biol. Chem. 2012, 287 (45), 38006– 38019. (298) Sinha, S.; Du, Z.; Maiti, P.; Klärner, F.-G.; Schrader, T.; Wang, C.; Bitan, G. ACS Chem. Neurosci. 2012, 3 (6), 451–458.

229

(299) Nussbaum, J. M.; Seward, M. E.; Bloom, G. S. Prion 2013, 7 (1), 14–19. (300) Taniguchi, S.; Suzuki, N.; Masuda, M.; Hisanaga, S.; Iwatsubo, T.; Goedert, M.; Hasegawa, M. J. Biol. Chem. 2005, 280 (9), 7614–7623. (301) Nespovitaya, N.; Mahou, P.; Laine, R. F.; Schierle, G. S. K.; Kaminski, C. F. Chem. Commun. 2017, 47, 311–320. (302) Dolai, S.; Shi, W.; Corbo, C.; Sun, C.; Averick, S.; Obeysekera, D.; Farid, M.; Alonso, A.; Banerjee, P.; Raja, K. ACS Chem. Neurosci. 2011, 2 (12), 694–699. (303) Taniguchi, S.; Suzuki, N.; Masuda, M.; Hisanaga, S.; Iwatsubo, T.; Goedert, M.; Hasegawa, M. J. Biol. Chem. 2005, 280 (9), 7614–7623. (304) Bacskai, B. J.; Skoch, J.; Hickey, G. A.; Allen, R.; Hyman, B. T. J. Biomed. Opt. 2003, 8 (3), 368–375. (305) Anthony, N. R.; Mehta, A. K.; Lynn, D. G.; Berland, K. M. Soft Matter 2014, 10, 4162–4172. (306) Roberti, M. J.; Bertoncini, C. W.; Klement, R.; Jares-Erijman, E. A.; Jovin, T. M. Nat. Methods 2007, 4 (4), 345–351. (307) Amaro, M.; Wellbrock, T.; Birch, D. J. S.; Rolinski, O. J. Appl. Phys. Lett. 2014, 104 (63704). (308) Kusumoto, Y.; Lomakin, A.; Teplow, D. B.; Benedek, G. B. Proc. Natl. Acad. Sci. U. S. A. 1998, 95 (21), 12277–12282. (309) Tiiman, A.; Krishtal, J.; Palumaa, P.; Tõugu, V. AIP Adv. 2015, 5 (92401). (310) Fraser, P. E.; Nguyen, J. T.; Surewicz, W. K.; Kirschner, D. A. Biophys. J. 1991, 60 (5), 1190– 1201. (311) Snyder, S. W.; Ladror, U. S.; Wade, W. S.; Wang, G. T.; Barrett, L. W.; Matayoshi, E. D.; Huffaker, H. J.; Krafft, G. a; Holzman, T. F. Biophys. J. 1994, 67 (3), 1216–1228. (312) Wood, S. J.; Maleeff, B.; Hart, T.; Wetzel, R. J. Mol. Biol. 1996, 256 (5), 870–877. (313) Su, Y.; Chang, P. T. Brain Res. 2001, 893, 287–291. (314) Li, S.; Leblanc, R. M. J. Phys. Chem. B 2014, 118 (5), 1181–1188. (315) Murphy, R. M. Biochim. Biophys. Acta 2007, 1768 (8), 1923–1934. (316) Batzli, K. M.; Love, B. J. Mater. Sci. Eng. C 2015, 48, 359–364. (317) Hamilton-Brown, P.; Bekard, I.; Ducker, W. A.; Dunstan, D. E. J. Phys. Chem. B 2008, 112 (51), 16249–16252. (318) Dunstan, D. E.; Hamilton-Brown, P.; Asimakis, P.; Ducker, W.; Bertolini, J. Protein Eng. Des. Sel. 2009, 22 (12), 741–746. (319) Abbott, N. J. Neurochem. Int. 2004, 45 (4), 545–552. (320) Linse, S.; Lund, M. J. Am. Chem. Soc. 2014, 136, 11776−11782. (321) Morinaga, A.; Hasegawa, K.; Nomura, R.; Ookoshi, T.; Ozawa, D.; Goto, Y.; Yamada, M.; Naiki, H. Biochim. Biophys. Acta - Proteins Proteomics 2010, 1804 (4), 986–995. (322) Williams, T. L.; Serpell, L. C. FEBS J. 2011, 278 (20), 3905–3917. (323) Vestergaard, M.; Hamada, T.; Takagi, M. Biotechnol. Bioeng. 2008, 99 (4), 753–763.

230

(324) Morgado, I.; Garvey, M. Adv. Exp. Med. Biol. 2015, 855, 67–94. (325) Jean, L.; Lee, C. F.; Vaux, D. J. Biophys. J. 2012, 102 (5), 1154–1162. (326) Campioni, S.; Carret, G.; Jordens, S.; Riek, R. J. Am. Chem. Soc. 2013, 136, 2866−2875. (327) Soreghan, B.; Kosmoski, J.; Glabe, C. J. Biol. Chem. 1994, 269 (46), 28551–28554. (328) Chi, E. Y.; Ege, C.; Winans, A.; Majewski, J.; Wu, G.; Kjaer, K.; Lee, K. Y. C. Proteins Struct. Funct. Genet. 2008, 72 (1), 1–24. (329) Hoernke, M.; Falenski, J. A.; Schwieger, C.; Koksch, B.; Brezesinski, G. Langmuir 2011, 27 (23), 14218–14231. (330) Jean, L.; Lee, C. F.; Lee, C.; Shaw, M.; Vaux, D. J. FASEB J. 2010, 24, 309–317. (331) Walsh, D. M.; Thulin, E.; Minogue, A. M.; Gustavsson, N.; Pang, E.; Teplow, D. B.; Linse, S. FEBS J. 2009, 276 (5), 1266–1281. (332) Lai, A. Y.; McLaurin, J. Biochim. Biophys. Acta 2012, 1822 (10), 1629–1637. (333) Akaishi, T.; Morimoto, T.; Shibao, M.; Watanabe, S.; Sakai-kato, K. Neurosci. Lett. 2008, 444, 280–285. (334) Li, G.; Pomès, R. J. Phys. Chem. B 2013, 117 (22), 6603–6613. (335) Janeček, M.; Rossmann, M.; Sharma, P.; Emery, A.; Huggins, D. J.; Stockwell, S. R.; Stokes, J. E.; Tan, Y. S.; Almeida, E. G.; Hardwick, B.; Narvaez, A. J.; Hyvönen, M.; Spring, D. R.; McKenzie, G. J.; Venkitaraman, A. R. Sci. Rep. 2016, 6 (1), 28528. (336) Riss, T. L.; Niles, A. L.; Minor, L. Cell Viability Assays Assay Guidance Manual; 2004. (337) Zheng, L.; Cedazo-Minguez, A.; Hallbeck, M.; Jerhammar, F.; Marcusson, J.; Terman, A. Transl. Neurodegener. 2012, 13 (1), 19–25. (338) LaFerla, F. M.; Green, K. N.; Oddo, S. Nat. Rev. Neurosci. 2007, 8 (7), 499–509. (339) Maezawa, I.; Hong, H. S.; Liu, R.; Wu, C. Y.; Cheng, R. H.; Kung, M. P.; Kung, H. F.; Lam, K. S.; Oddo, S.; LaFerla, F. M.; Jin, L. W. J. Neurochem. 2008, 104 (2), 457–468. (340) Winton, M. J.; Lee, E. B.; Sun, E.; Wong, M. M.; Leight, S.; Zhang, B.; Trojanowski, J. Q.; Lee, V. M.-Y. J. Neurosci. 2011, 31 (21), 7691–7699. (341) Gouras, G. K.; Tampellini, D.; Takahashi, R. H.; Capetillo-Zarate, E. Acta Neuropathol. 2010, 119 (5), 523–541. (342) D’Andrea, M. R.; Reiser, P. A.; Polkovitch, D. A.; Gumula, N. A.; Branchide, B.; Hertzog, B. M.; Schmidheiser, D.; Belkowski, S.; Gastard, M. C.; Andrade-Gordon, P. Neurosci. Lett. 2003, 342 (1), 114–118. (343) Ohyagi, Y.; Tsuruta, Y.; Motomura, K.; Miyoshi, K.; Kikuchi, H.; Iwaki, T.; Taniwaki, T.; Kira, J. ichi. J. Neurosci. Methods 2007, 159 (1), 134–138. (344) Walsh, D. M.; Hartley, D. M.; Kusumoto, Y.; Fezoui, Y.; Condron, M. M.; Lomakin, A.; Benedek, G. B.; Selkoe, D. J.; Teplow, D. B.; Biol, D. B. J. J Biol Chem 1999, 274 (36), 25945–25952. (345) Shen, D.; Coleman, J.; Chan, E.; Nicholson, T. P.; Dai, L.; Sheppard, P. W.; Patton, W. F. Cell Biochem. Biophys. 2011, 60 (3), 173–185.

231

(346) Ando, Y.; Haraoka, K.; Terazaki, H.; Tanoue, Y.; Ishikawa, K.; Katsuragi, S.; Nakamura, M.; Sun, X.; Nakagawa, K.; Sasamoto, K.; Takesako, K.; Ishizaki, T.; Sasaki, Y.; Doh-ura, K. Lab. Invest. 2003, 83 (12), 1751–1759. (347) Khurana, R.; Uversky, V. N.; Nielsen, L.; Fink, A. L. J. Biol. Chem. 2001, 276 (25), 22715– 22721. (348) Johnson, R. D.; Schauerte, J. A.; Wisser, K. C.; Gafni, A.; Steel, D. G. PLoS One 2011, 6 (8), e23970. (349) Galasko, D.; Chang, L.; Motter, R.; Clark, C. M.; Kaye, J.; Knopman, D.; Thomas, R.; Kholodenko, D.; Schenk, D.; Lieberburg, I.; Miller, B.; Green, R.; Basherad, R.; Kertiles, L.; Boss, M. A.; Seubert, P. Arch. Neurol. 1998, 55 (7), 937. (350) Seubert, P.; Vigo-Pelfrey, C.; Esch, F.; Lee, M.; Dovey, H.; Davis, D.; Sinha, S.; Schlossmacher, M.; Whaley, J.; Swindlehurst, C. Nature 1992, 359 (6393), 325–327. (351) Brenner, S. Genetics 1974, 77 (1), 71–94. (352) Kaletta, T.; Hengartner, M. O. Nat. Rev. Drug Discov. 2006, 5 (5), 387–399. (353) Jones, A. K.; Buckingham, S. D.; Sattelle, D. B. Nat. Rev. Drug Discov. 2005, 4 (4), 321–330. (354) Consortium, T. C. elegans S. Scicence. 1998, 282 (3396), 2012–2018. (355) Lai, C.-H. Genome Res. 2000, 10 (5), 703–713. (356) Harris, T. W. Nucleic Acids Res. 2004, 32 (90001), 411D–417. (357) J. G. White, Eileen Southgate, J. N. Thomson, S. B. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 1976, 275 (938), 327–348. (358) Alexander, A. G.; Marfil, V.; Li, C. Front. Genet. 2014, 5, 1–21. (359) Dimitriadi, M.; Hart, A. C. Neurobiol. Dis. 2010, 40 (1), 4–11. (360) Nass, R.; Miller, D. M.; Blakely, R. D. Park. Relat. Disord. 2001, 7 (3), 185–191. (361) David, D. C.; Ollikainen, N.; Trinidad, J. C.; Cary, M. P.; Burlingame, A. L. PLoS Biol. 2010, 8 (8), e1000450. (362) Link, C. D. Exp. Gerontol. 2006, 41 (10), 1007–1013. (363) Sundaram, M.; Greenwald, I. Genetics 1993, 135 (3), 765–783. (364) Levitan, D.; Greenwald, I. Nature 1995, 377 (6547), 351–354. (365) Bruni, A. C. Funct. Neurol. 1998, 13 (3), 257–261. (366) McColl, G.; Roberts, B. R.; Gunn, A. P.; Perez, K. A.; Tew, D. J.; Masters, C. L.; Barnham, K. J.; Cherny, R. A.; Bush, A. I. J. Biol. Chem. 2009, 284 (34), 22697–22702. (367) Link, C. D. Proc. Natl. Acad. Sci. 1995, 92 (20), 9368–9372. (368) Link, C. D.; Johnson, C. J.; Fonte, V.; Paupard, M. C.; Hall, D. H.; Styren, S.; Mathis, C. A.; Klunk, W. E. Neurobiol. Aging 2001, 22 (2), 217–226. (369) McColl, G.; Roberts, B. R.; Pukala, T. L.; Kenche, V. B.; Roberts, C. M.; Link, C. D.; Ryan, T. M.; Masters, C. L.; Barnham, K. J.; Bush, A. I.; Cherny, R. A. Mol. Neurodegener. 2012, 7 (1), 57.

232

(370) Hu, X.; Crick, S. L.; Bu, G.; Frieden, C.; Pappu, R. V; Lee, J.-M. Proc. Natl. Acad. Sci. U. S. A. 2009, 106 (48), 20324–20329. (371) Misinzo, G.; Delputte, P. L.; Nauwynck, H. J. J. Virol. 2008, 82 (3), 1128–1135. (372) Nakabayashi, T.; Ohta, N. Anal. Sci. 2015, 31 (4), 275–285. (373) Stiernagle, T. Maintenance of C. elegans; The C. elegans Research Community, Ed.; WormBook, ed, 2006. (374) Pliss, A.; Zhao, L.; Ohulchanskyy, T. Y.; Qu, J.; Prasad, P. N. ACS Chem. Biol. 2012, 7 (8), 1385–1392. (375) Tregidgo, C.; Levitt, J. A.; Suhling, K. J. Biomed. Opt. 2015, 13 (3), 31218. (376) Striker, G.; Subramaniam, V.; Seidel, C. a. M.; Volkmer, A. J. Phys. Chem. B 1999, 103 (40), 8612–8617. (377) Burns, A. R.; Wallace, I. M.; Wildenhain, J.; Tyers, M.; Giaever, G.; Bader, G. D.; Nislow, C.; Cutler, S. R.; Roy, P. J. Nat. Chem. Biol. 2010, 6 (7), 549–557. (378) Wirths, O.; Multhaup, G.; Czech, C.; Blanchard, V.; Moussaoui, S.; Tremp, G.; Pradier, L.; Beyreuther, K.; Bayer, T. A. Neurosci. Lett. 2001, 306 (1–2), 116–120. (379) Gouras, G. K.; Tsai, J.; Naslund, J.; Vincent, B.; Edgar, M.; Checler, F.; Greenfield, J. P.; Haroutunian, V.; Buxbaum, J. D.; Xu, H.; Greengard, P.; Relkin, N. R. Am. J. Pathol. 2000, 156 (1), 15–20. (380) Gouras, G. K.; Willén, K.; Faideau, M. Neurodegener. Dis. 2014, 13, 142–146. (381) O’Reilly, L. P.; Luke, C. J.; Perlmutter, D. H.; Silverman, G. A.; Pak, S. C. Adv. Drug Deliv. Rev. 2014, 69–70, 247–253. (382) Schwendeman, A. R.; Shaham, S. PLoS One 2016, 11 (10), e0164595. (383) Bhullar, K. S.; Rupasinghe, H. P. V. Oxid. Med. Cell. Longev. 2013, No. 891748. (384) Isemura, M.; Saeki, K.; Kimura, T.; Hayakawa, S.; Minami, T.; Sazuka, M. Biofactors 2000, 13 (1–4), 81–85. (385) Daglia, M. Curr. Opin. Biotechnol. 2012, 23 (2), 174–181. (386) Mangialasche, F.; Solomon, A.; Winblad, B.; Mecocci, P.; Kivipelto, M. Lancet Neurol. 2010, 9 (7), 702–716. (387) Mazzanti, G.; Di Giacomo, S. Molecules 2016, 21 (9), 1243–1250. (388) Yamada, M.; Ono, K.; Hamaguchi, T. Natural Compounds as Therapeutic Agents for Amyloidogenic Diseases; Vassallo, N., Ed.; Springer, 2015; Vol. 863. (389) Ono, K.; Yoshiike, Y.; Takashima, A.; Hasegawa, K.; Yamada, M. J. Neurochem. 2003, 87, 172–181. (390) Lakey-beitia, J.; Berrocal, R.; Rao, K. S.; Durant, A. A. Mol. Neurobiol. 2015, 51, 466–479. (391) Cole, G. M.; Teter, B.; Frautschy, S. A. Adv. Exp. Med. Biol. 2007, 595, 197–212. (392) Villaflores, O. B.; Chen, Y. J.; Chen, C. P.; Yeh, J. M.; Wu, T. Y. Taiwan. J. Obstet. Gynecol. 2012, 51 (4), 515–525. (393) Rezai-Zadeh, K. J. Neurosci. 2005, 25 (38), 8807–8814.

233

(394) Wobst, H. J.; Sharma, A.; Diamond, M. I.; Wanker, E. E.; Bieschke, J. FEBS Lett. 2015, 589 (1), 77–83. (395) Bieschke, J.; Russ, J.; Friedrich, R. P.; Ehrnhoefer, D. E.; Wobst, H.; Neugebauer, K.; Wanker, E. E. Proc. Natl. Acad. Sci. 2010, 107 (17), 7710–7715. (396) Ehrnhoefer, D. E.; Duennwald, M.; Markovic, P.; Wacker, J. L.; Engemann, S.; Roark, M.; Legleiter, J.; Marsh, J. L.; Thompson, L. M.; Lindquist, S.; Muchowski, P. J.; Wanker, E. E. Hum. Mol. Genet. 2006, 15 (18), 2743–2751. (397) Singh, P. K.; Kotia, V.; Ghosh, D.; Mohite, G. M.; Kumar, A.; Maji, S. K. ACS Chem. Neurosci. 2013, 4, 393–407. (398) Ono, K.; Hasegawa, K.; Yoshiike, Y.; Takashima, A.; Yamada, M.; Naiki, H. J. Neurochem. 2002, 81 (3), 434–440. (399) Moss, M. A.; Varvel, N. H.; Nichols, M. R.; Reed, D. K.; Rosenberry, T. L. Mol. Pharmacol. 2004, 66 (3), 592–600. (400) Henry-Vitrac, C.; Berbille, H.; Mérillon, J. M.; Vitrac, X. Food Res. Int. 2010, 43 (8), 2176– 2178. (401) Kim, H.; Park, B. S.; Lee, K. G.; Cheol, Y. C.; Sung, S. J.; Kim, Y. H.; Lee, S. E. J. Agric. Food Chem. 2005, 53 (22), 8537–8541. (402) Shimmyo, Y.; Kihara, T.; Akaike, A.; Niidome, T.; Sugimoto, H. J. Neurosci. Res. 2008, 86 (2), 368–377. (403) Endo, H.; Nikaido, Y.; Nakadate, M.; Ise, S.; Konno, H. Bioorg. Med. Chem. Lett. 2014, 24 (24), 5621–5626. (404) Masuda, Y.; Fukuchi, M.; Yatagawa, T.; Tada, M.; Takeda, K.; Irie, K.; Akagi, K.; Monobe, Y.; Imazawa, T.; Takegoshi, K. Bioorg. Med. Chem. 2011, 19 (20), 5967–5974. (405) Sato, M.; Murakami, K.; Uno, M.; Nakagawa, Y.; Katayama, S.; Akagi, K.; Masuda, Y.; Takegoshi, K.; Irie, K. J Biol Chem 2013, 288 (32), 23212–23224. (406) Hirohata, M.; Hasegawa, K.; Tsutsumi-yasuhara, S.; Ohhashi, Y.; Ookoshi, T.; Ono, K.; Yamada, M.; Naiki, H. Biochemistry 2007, 46, 1888–1899. (407) Moosmann, B.; Behl, C. Expert Opin. Investig. Drugs 2002, 11 (10), 1407–1435. (408) Sies, H. Eur. J. Biochem. 1993, 215 (2), 213–219. (409) Cai, Y.-Z.; Mei Sun; Jie Xing; Luo, Q.; Corke, H. Life Sci. 2006, 78 (25), 2872–2888. (410) Sam, S. K.; Ji, Y. L.; Yoo, K. C.; Sun, S. S.; Ju, S. K.; Su, J. J.; Yong, N. H.; Kun, H. S.; Byung, H. H. Bioorganic Med. Chem. Lett. 2005, 15 (15), 3588–3591. (411) Singh, P.; Anand, A.; Kumar, V. Eur. J. Med. Chem. 2014, 85, 758–777. (412) Thapa, A.; Woo, E.-R.; Chi, E. Y.; Sharoar, G.; Jin, H.-G.; Shin, S. Y.; Park, I.-S. Biochemistry 2011, 50, 2445–2455. (413) Hou, J.; Liu, X.; Shen, J.; Zhao, G.; Wang, P. G. Expert Opin. Drug Discov. 2012, 7 (6), 489– 501. (414) Sum, T. H.; Sum, T. J.; Galloway, W. R. J. D.; Collins, S.; Twigg, D. G.; Hollfelder, F.; Spring, D. R. Molecules 2016, 21 (9).

234

(415) Masuda, M.; Suzuki, N.; Taniguchi, S.; Oikawa, T.; Nonaka, T.; Iwatsubo, T.; Hisanaga, S.; Goedert, M.; Hasegawa, M. Biochemistry 2006, 45 (19), 6085–6094. (416) Sum, T. H.; Sum, T. J.; Collins, S.; Galloway, W. R. J. D.; Twigg, D. G.; Hollfelder, F.; Spring, D. R. Org. Biomol. Chem. 2017, 15 (21). (417) Roussaki, M.; Costa Lima, S.; Kypreou, A.-M.; Kefalas, P.; Cordeiro da Silva, A.; Detsi, A. Int. J. Med. Chem. 2012, 196921. (418) Ono, M.; Maya, Y.; Haratake, M.; Ito, K.; Mori, H.; Nakayama, M. Biochem. Biophys. Res. Commun. 2007, 361 (1), 116–121. (419) Li, Y.; Qiang, X.; Luo, L.; Li, Y.; Xiao, G.; Tan, Z.; Deng, Y. Bioorganic Med. Chem. 2016, 24 (10), 2342–2351. (420) Ni, L.; Meng, C. Q.; Sikorski, J. A. Expert Opin. Ther. Patents 2004, 14 (12), 1669–1691. (421) Sahu, N. K.; S. Balbhadra, S.; Choudhary, J.; V. Kohli, D. Curr. Med. Chem. 2012, 19 (2), 209–225. (422) Bayati, S.; Yazdanparast, R.; Majd, S. S.; Oh, S. J. Appl. Toxicol. 2011, 31 (6), 545–553. (423) Kim, M. J.; Lee, Y. H.; Kwak, J.; Na, Y.; Yoon, H. G. BMB Rep 2011, 44 (11), 730–734. (424) Mathew, B.; Suresh, J.; Anbazghagan, S.; Paulraj, J.; Krishnan, G. K. Biomed. Prev. Nutr. 2014, 4 (3), 451–458. (425) Bandgar, B. P.; Gawande, S. S.; Bodade, R. G.; Gawande, N. M.; Khobragade, C. N. Bioorg. Med. Chem. 2009, 17 (24), 8168–8173. (426) Wondrousch, D.; Koleva, Y. K.; Madden, J. C.; Cronin, M. T. D.; Schu, G. 2010, 1576–1585. (427) Keiser, M. J.; Setola, V.; Irwin, J. J.; Laggner, C.; Abbas, A.; Hufeisen, S. J.; Jensen, N. H.; Kuijer, M. B.; Matos, R. C.; Thuy, B.; Whaley, R.; Glennon, R. A.; Hert, J.; Thomas, K. L. H.; Edwards, D.; Shoichet, B. K.; Roth, B. L. Nature 2009, 462 (7270), 175–181. (428) Rodrigues, T.; Reker, D.; Schneider, P.; Schneider, G. Nat. Chem. 2016, 8 (6), 531–541. (429) Bandgar, B. P.; Gawande, S. S.; Bodade, R. G.; Totre, J. V.; Khobragade, C. N. Bioorganic Med. Chem. 2010, 18 (3), 1364–1370. (430) Sravanthi, T. V.; Manju, S. L. Eur. J. Pharm. Sci. 2016, 91, 1–10. (431) de Sa Alves, F.; Barreiro, E.; Manssour Fraga, C. Mini-Reviews Med. Chem. 2009, 9 (7), 782–793. (432) Özdemir, A.; Altıntop, M. D.; Turan-Zitouni, G.; Çiftçi, G. A.; Ertorun, İ.; Alataş, Ö.; Kaplancıklı, Z. A. Eur. J. Med. Chem. 2015, 89, 304–309. (433) Wyss-coray, T.; Rogers, J. Cold Spring Harb. Perspect. Biol. 2012, 1–24. (434) Wyss-Coray, T. Nat. Med. 2006, 12 (9), 1005–1015. (435) Cui, M.; Ono, M.; Kimura, H.; Liu, B. L.; Saji, H. Bioorg. Med. Chem. Lett. 2011, 21 (3), 980– 982. (436) Pandey, U. B.; Nichols, C. D. Drug Deliv. 2011, 63 (2), 411–436. (437) Gonzalez, C. Nat. Rev. Cancer 2013, 13 (3), 172–183. (438) Laurijssens, B.; Aujard, F.; Rahman, A. Drug Discov. Today Technol. 2013, 10 (3), e319–

235

e327. (439) Nichols, C. D. Pharmacol. Ther. 2006, 112 (3), 677–700. (440) Crowther, D. C.; Page, R.; Chandraratna, D.; Lomas, D. A. Methods Enzymol. 2006, 412, 234–255. (441) Fernandez-funez, P.; Mena, L. De; Rincon-limas, D. E. Exp. Neurol. 2015, 274, 58–71. (442) Reiter, L. T.; Potocki, L.; Chien, S.; Gribskov, M.; Bier, E. Genome Res. 2001, 11 (6), 1114– 1125. (443) Chien, S. Nucleic Acids Res. 2002, 30 (1), 149–151. (444) Hulbert, A.; Pamplona, R.; Buffenstein, R.; Buttemer, W. Physiol. Rev. 2007, 87 (4), 1175– 1213. (445) Partridge, L.; Pletcher, S. D.; Mair, W. Mech. Ageing Dev. 2005, 126 (1), 35–41. (446) Ott, S.; Vishnivetskaya, A.; Malmendal, A.; Crowther, D. C. Neurobiol. Aging 2016, 41, 39– 52. (447) Prüßing, K.; Voigt, A.; Schulz, J. B. Mol. Neurodegener. 2013, 8 (35). (448) Yagi, Y.; Tomita, S.; Nakamura, M.; Suzuki, T. Mol Cell Biol Res Commun 2000, 4 (1), 43– 49. (449) Finelli, A.; Kelkar, A.; Song, H. J.; Yang, H.; Konsolaki, M. Mol. Cell. Neurosci. 2004, 26 (3), 365–375. (450) Crowther, D. C.; Kinghorn, K. J.; Miranda, E.; Page, R.; Curry, J. A.; Duthie, F. A. I.; Gubb, D. C.; Lomas, D. A. Neuroscience 2005, 132, 123–135. (451) Speretta, E.; Jahn, T. R.; Tartaglia, G. G.; Favrin, G.; Barros, T. P.; Imarisio, S.; Lomas, D. A.; Luheshi, L. M.; Crowther, D. C.; Dobson, C. M. J. Biol. Chem. 2012, 287 (24), 20748–20754. (452) Brand, A. H.; Perrimon, N. Development 1993, 118 (2), 401–415. (453) del Valle Rodríguez, A.; Didiano, D.; Desplan, C. Nat. Methods 2012, 9 (1), 47–55. (454) Wittmann, C. W.; Wszolek, M. F.; Shulman, J. M.; Salvaterra, P. M.; Lewis, J.; Hutton, M.; Feany, M. B. Science 2001, 293 (5530), 711–714. (455) Rand, M. D.; Kearney, A. L.; Dao, J.; Clason, T. Insect Biochem. Mol. Biol. 2010, 40 (11), 792–804. (456) Pollitt, S. K.; Pallos, J.; Shao, J.; Desai, U. A.; Ma, A. A. K.; Thompson, L. M.; Marsh, J. L.; Diamond, M. I. Neuron 2003, 40 (4), 685–694. (457) Chang, S.; Bray, S. M.; Li, Z.; Zarnescu, D. C.; He, C.; Jin, P.; Warren, S. T. Nat. Chem. Biol. 2008, 4 (4), 256–263. (458) Willoughby, L. F.; Schlosser, T.; Manning, S. A.; Parisot, J. P.; Street, I. P.; Richardson, H. E.; Humbert, P. O.; Brumby, A. M. Dis. Model. Mech. 2013, 6 (2), 521–529. (459) Stork, T.; Engelen, D.; Krudewig, A.; Silies, M.; Bainton, R. J.; Klämbt, C. J. Neurosci. 2008, 28 (3), 587–597. (460) Hindle, S. J.; Bainton, R. J. Front. Neurosci. 2014, 8 (414). (461) Mayer, F.; Mayer, N.; Chinn, L.; Pinsonneault, R. L.; Kroetz, D.; Bainton, R. J. J. Neurosci.

236

2009, 29 (11), 3538–3550. (462) McClung, C.; Hirsh, J. Curr. Biol. 1998, 8 (2), 109–112. (463) Andretic, R.; Kim, Y.-C.; Jones, F. S.; Han, K.-A.; Greenspan, R. J. Proc. Natl. Acad. Sci. U. S. A. 2008, 105 (51), 20392–20397. (464) Dar, A. C.; Das, T. K.; Shokat, K. M.; Cagan, R. L. Nature 2012, 486 (7401), 80–84. (465) Sarantseva, S. V.; Bolshakova, O. I.; Timoshenko, S. I.; Kolobov, A. A.; Schwarzman, A. L. Bull. Exp. Biol. Med. 2011, 150 (4), 429–431. (466) Edwards, A.; Gladstone, M.; Yoon, P.; Raben, D.; Frederick, B.; Su, T. T. Dis. Model. Mech. 2011, 4 (4), 496–503. (467) Gladstone, M.; Su, T. T. J. Genet. Genomics 2011, 38 (10), 497–504. (468) Aksenov, M. Y.; Aksenova, M. V.; Butterfield, D. A.; Geddes, J. W.; Markesbery, W. R. Neuroscience 2001, 103 (2), 373–383. (469) Montine, T. J.; Neely, M. D.; Quinn, J. F.; Beal, M. F.; Markesbery, W. R.; Roberts II, L. J.; Morrow, J. D. Free Radic. Biol. Med. 2002, 33 (5), 620–626. (470) Lovell, M. A.; Markesbery, W. R. Nucleic Acids Res. 2007, 35 (22), 7497–7504. (471) Mecocci, P.; Polidori, M. C. Biochim. Biophys. Acta - Mol. Basis Dis. 2012, 1822 (5), 631– 638. (472) Frank, B.; Gupta, S. Ann. Clin. Psychiatry 2005, 17 (4), 269–286. (473) Thomas, B. J.; Wassarman, D. A. Trends Genet. 1999, 15 (5), 184–190. (474) Wagner, A. E.; Piegholdt, S.; Rabe, D.; Baenas, N.; Schloesser, A.; Eggersdorfer, M.; Stocker, A.; Rimbach, G. Oncotarget 2015, 6 (31), 30568–30578. (475) Li, Y. M.; Chan, H. Y. E.; Huang, Y.; Chen, Z. Y. Mol. Nutr. Food Res. 2007, 51 (5), 546–554. (476) Lambert, J. D.; Kennett, M. J.; Sang, S.; Reuhl, K. R.; Ju, J.; Yang, C. S. Food Chem. Toxicol. 2010, 48 (1), 409–416. (477) Yu, F.; Sheng, J.; Xu, J.; An, X.; Hu, Q. Eur. Food Res. Technol. 2007, 225 (5–6), 843–848. (478) Hou, Z.; Sang, S.; You, H.; Lee, M. J.; Hong, J.; Chin, K. V.; Yang, C. S. Cancer Res. 2005, 65 (17), 8049–8056. (479) Forester, S. C.; Lambert, J. D. Mol Nutr Food Res 2013, 55 (6), 844–854. (480) Elbling, L.; Weiss, R. M.; Teufelhofer, O.; Uhl, M.; Knasmueller, S.; Schulte-Hermann, R.; Berger, W.; Micksche, M. FASEB J. 2005, 19 (7), 807–809. (481) Stewart, K.; Radford, S. E. Biophys. Rev. 2017, 9 (4), 405–419. (482) Das, S.; Stark, L.; Musgrave, I. F.; Pukala, T.; Smid, S. D. Food Funct. 2016, 7 (2), 1138– 1146. (483) Trepanier, C. H.; Milgram, N. W. J. Alzheimer’s Dis. 2010, 21 (4), 1089–1099. (484) Sutera, F. M.; De Caro, V.; Giannola, L. I. Expert Opin. Drug Deliv. 2017, 14 (1), 93–107. (485) Mikitsh, J. L.; Chacko, A. M. Perspect. Medicin. Chem. 2014, 6, 11–24. (486) Xia, Y.; Whitesides, G. M. Angew. Chemie Int. Ed. 1998, 37 (5), 550–575.

237

(487) Warren, S. C.; Margineanu, A.; Alibhai, D.; Kelly, D. J.; Talbot, C.; Alexandrov, Y.; Munro, I.; Katan, M.; Dunsby, C.; French, P. M. W. PLoS One 2013, 8 (8), e70687. (488) Fontrodona, L.; Villanueva, A.; Cerón, J. J. Vis. Exp. 2012, 64 (e4019). (489) Giardina, G. A. M.; Sarau, H. M.; Farina, C.; Medhurst, A. D.; Grugni, M.; Raveglia, L. F.; Schmidt, D. B.; Rigolio, R.; Luttmann, M.; Vecchietti, V.; Hay, D. W. P. J. Med. Chem. 1997, 40 (12), 1794–1807. (490) Venkatesan, P.; Sumathi, S. J. Heterocycl. Chem. 2010, 47 (1), 81–84. (491) Gopinath, V. S.; Rao, M.; Shivahare, R.; Vishwakarma, P.; Ghose, S.; Pradhan, A.; Hindupur, R.; Sarma, K. Das; Gupta, S.; Puri, S. K.; Launay, D.; Martin, D. Bioorganic Med. Chem. Lett. 2014, 24 (9), 2046–2052. (492) Kadayat, T.; Song, C.; Shin, S.; Magar, T.; Bist, G.; Shrestha, A.; Thapa, P.; Na, Y.; Kwon, Y.; Lee, E. Bioorg. Med. Chem. 2015, 23 (13), 3499–3512.

238

Appendices

239

240

Appendix I: Libraries Screened with the NanoFLIM

Number Number of Library Name Reference screened hits* Partially Saturated Bicyclic Partially Saturated Bicyclic Heteroaromatics as an sp3 -Enriched Fragment Collection 21 5 Heteroaromatics Fragments - Angew. Chem. Int. Ed. 2016, 55, 12479 –12483 Quinolone Natural Products Divergent Synthesis of Quinolone Natural Products from Pseudonocardia sp. CL38489, 7 0 derivatives - Eur. J. Org. Chem. 2016, 5799–5802

Unpublished DOS library 10 0 Unpublished

Diversity-Oriented Synthesis of Drug-Like Macrocyclic Scaffolds Using an Orthogonal Organo- and DOS drug like macrocycles 25 3 Metal Catalysis Strategy, - Angew. Chem. Int. Ed. 2014, 53, 13093 –13097

Concise Synthesis of Substituted Quinolizin-4-ones by Ring-Closing Metathesis Quinolizin-4-ones and - Eur. J. Org. Chem. 2014, 5767 pyrido[1,2-a]pyrimidin- 44 3 Concise synthesis of rare pyrido[1,2-a]pyrimidin- 2-ones and related nitrogen-rich bicyclic scaffolds 2-ones with a ring-junction nitrogen

- Org. Biomol. Chem., 2016, 14, 1031–1038 |

Paracyclophane analogues 28 0 Unpublished

Divergent and concise total syntheses of dihydrochalcones and 5-deoxyflavones recently isolated from Tacca species and Mimosa diplotricha Flavone + heterocyclic flavone - Tetrahedron 71 (2015) 4557e4564 derivatives 54 2 Divergent Total Syntheses of Flavonoid Natural Products Isolated from Rosa rugosa and

Citrus unshiu - Synlett 2016, 27, 1725–1727 Divergent synthesis of biflavonoids yields novel inhibitors of the aggregation of amyloid β Flavone dimers 0 30 (1–42),

- Org. Biomol. Chem., 2017, 15, 4554–4570

241

Number Number of Library Name Reference screened hits*

Chalcones and heteroaromatic 32 19 Unpublished chalcone derivatives

Triazole linked flavonoid Combinatorial Synthesis of Structurally Diverse Triazole-Bridged Flavonoid Dimers and 1 derivatives 31 Trimers

- Molecules 2016, 21, 1230; Divergent synthesis of biflavonoids yields novel inhibitors of the aggregation of amyloid β Aurones + biaurones 32 14 (1–42), - Org. Biomol. Chem., 2017, 15, 4554–4570 A strategy for the diversity-oriented synthesis of macrocyclic scaffolds using multidimensional coupling Build/Couple/Pair/Diversify - Nat. Chem. 2013, 5, p861 Macrocylic DOS library 111 10 A Multidimensional Diversity-Oriented Synthesis Strategy for Structurally Diverse and Complex Macrocycles - Angew. Chem. Int. Ed. 2016, 55, 11139 –11143

Allosteric modulation of AURKA kinase activity by a small-molecule inhibitor of its protein- Phenyl Quinoline library 20 9 protein interaction with TPX2 - Scientific Reports | 6:28528 | DOI: 10.1038/srep28528

242

Appendix II: MJ Library

243

Figure A1: The MJ cinchophen library. Compounds showing >30% inhibitory activity after 2.5 h of incubation in the nanoFLIM assay are denoted as hits.

4000 x x x x x x x x x 3500

3000

2500

2000

Fluorescence Lifetime (ps)Lifetime Fluorescence



A MJ001 MJ006 MJ014 MJ017 MJ018 MJ020 MJ022 MJ023 MJ028 MJ029 MJ031 MJ033 MJ036 MJ038 MJ039 MJ040 MJ041 MJ042 MJ043 MJ044

Figure A2: Start and end points fluorescence lifetime values of Aβ42488 incubated with each of the MJ library. Fluorescence lifetime was monitored with the nanoFLIM and values are given for each compound at t = 20 and t = 150. Compounds that showed >30% inhibitory activity after at 2.5 h incubation, in 3 independent repeats (n = 5 droplets per repeat), are marked with an ‘x’. Conditions: 10 μM Aβ42,488, 50% labelled, 50 μM compound.

244

Figure A3: Relative changes in the ThT fluorescence intensities with MJ library treatment. Fluorescence intensity at the plateau following 12 h of incubation with each of the compounds of the MJ library. The fluorescence in the untreated Aβ sample is normalised to 100%. Conditions: 10 μM Aβ42, 20 μM ThT, 50 μM compound.

Figure A4: Relative changes in the intensity of the ThT fluorescence emission spectra of preformed Aβ fibrils with the addition of each of the compounds of the MJ library. The fluorescence intensity of the untreated Aβ sample is normalised to 100%. Intensity measured at 488 nm emission, with 440 nm excitation. Conditions: 10 μM preformed Aβ42 fibrils, 20 μM ThT, 50 μM compound.

245

Appendix III: C. elegans Construct Map

Figure A5: pSM:myo3::GFP::Aβ graphic restriction map

246

Appendix IV: Statistical Breakdown of C. elegans Data

Table A3: Statistical analysis for figure 5.8. Statistical analysis of the difference in the mean fluorescence lifetime (ps) of untreated control GFP:Aβ worms and the MJ040X treated worms. P-values were computed in GraphPad Prism using the either the Bonferroni post-test or Dunnett’s test of multiple comparison. *p<0.05; **p<0.01, ***p<0.001. a)

Two-way ANOVA with Bonferroni post-test Day Control MJ040X Difference 95% CI of diff. t P value Summary Day 1 2769 2769 0 -43.45 to 43.45 0 P > 0.05 ns Day 3 2754 2747 -6.824 -51.53 to 37.89 0.4407 P > 0.05 ns Day 6 2745 2743 -2.163 -47.56 to 43.24 0.1375 P > 0.05 ns Day 9 2724 2745 21.49 -22.74 to 65.71 1.403 P > 0.05 ns Day 12 2698 2754 56.56 5.635 to 107.5 3.207 P<0.01 ** Day 15 2715 2714 -1.646 -42.37 to 39.08 0.1167 P > 0.05 ns

247 b)

One way ANOVA with Dunnett's Multiple Comparison Test

Mean Diff. q Significant? P < 0.05? Summary 95% CI of diff Control D1 vs Control D3 15.01 1.086 No ns -23.41 to 53.44

Control D1 vs Control D6 23.99 1.708 No ns -15.05 to 63.03

Control D1 vs Control D9 45.33 3.371 Yes ** 7.958 to 82.70

Control D1 vs Control D12 71.44 4.431 Yes *** 26.62 to 116.2

Control D1 vs Control D15 53.6 4.036 Yes *** 16.68 to 90.51

Control D1 vs MJ040X D1 0 0 No ns -37.87 to 37.87

Control D1 vs MJ040X D3 21.84 1.58 No ns -16.59 to 60.26

Control D1 vs MJ040X D6 26.16 1.892 No ns -12.27 to 64.58

Control D1 vs MJ040X D9 23.84 1.698 No ns -15.20 to 62.88

Control D1 vs MJ040X D12 14.88 1.106 No ns -22.49 to 52.25

Control D1 vs MJ040X D15 55.24 4.207 Yes *** 18.75 to 91.74

(D = day)

248

Table A4: Statistical analysis for figure 5.9. Statistical analysis of the difference in the mean fluorescence lifetime (ps) of the untreated control GFP:Aβ worms and the worms treated with MJ040X from larvae or from adulthood. P-values were computed in GraphPad Prism using either the Bonferroni post-test or Dunnett’s test of multiple comparisons. *p<0.05; **p<0.01, ***p<0.001. a) Two-way ANOVA with Bonferroni post-test

Control vs Drugged at larval stage Day Control Drugged at larval Difference 95% CI of diff. t P value Summary stage Day 1 2769 2763 -5.775 -42.30 to 30.75 0.423 P > 0.05 ns Day 6 2745 2774 29.09 -16.95 to 75.12 1.69 P > 0.05 ns Day 15 2715 2769 53.22 18.68 to 87.76 4.122 P<0.001 ***

Control vs Drugged at adulthood Day Control Drugged at Difference 95% CI of diff. t P value Summary adulthood Day 1 2769 2769 0 -35.44 to 35.44 0 P > 0.05 ns Day 6 2745 2743 -2.163 -39.19 to 34.87 0.1562 P > 0.05 ns Day 15 2715 2714 -1.646 -34.86 to 31.57 0.1326 P > 0.05 ns

249 b)

One-way ANOVA with Dunnett's Multiple Comparison Test

Mean Diff. q Significant? P < 0.05? Summary 95% CI of diff Control D1 vs Control D6 23.99 1.757 No ns -12.74 to 60.72

Control D1 vs Control D15 53.6 4.151 Yes *** 18.87 to 88.33

Control D1 vs Adult MJ040X D1 5.774 0.4229 No ns -30.96 to 42.50

Control D1 vs Adult MJ040X D6 -5.094 0.3017 No ns -50.52 to 40.33

Control D1 vs Adult MJ040X D15 0.3723 0.02811 No ns -35.26 to 36.01

Control D1 vs Larvae MJ040X D1 0 0 No ns -35.63 to 35.63

Control D1 vs Larvae MJ040X D6 26.16 1.946 No ns -9.997 to 62.31

Control D1 vs Larvae MJ040X D15 55.24 4.328 Yes *** 20.90 to 89.58

(D = day)

250

Appendix V: Methoxychalcones Library

251

Figure A6: Inhibitory activity of the methoxychalcone library. Percentage inhibition for each compounds against the aggregation of Aβ42 is displayed underneath its structure. Values calculated using the saturation phase of the ThT aggregation profiles, compared to the untreated Aβ42 control. 100% represents complete inhibition of aggregation and 0% indicates no inhibition. Conditions: Aβ42 10 μM, ThT 20 μM, compound 50 μM. n = 1 or 3. For n = 3, percentage inhibition shown as mean ± SEM.

252

Overly Fluorescent Compounds

Figure A7: Overly fluorescent methoxychalcones. Percentage inhibition against the aggregation of Aβ42 for these structures could not be accurately monitored by means of ThT fluorescence assays, as a result of their intrinsic spectroscopic properties.

253

Appendix VI: Computer Aided Design of the Two-Part Shearing Chip

254

Appendix VII: NMR Spectra 2-(3-bromophenyl)-6-nitroquinoline-4-carboxylic acid (MJ040)

255 methyl 2-(3-bromophenyl)-6-nitroquinoline-4-carboxylate (MJ040X)

256

(E)-1-(2-hydroxy-4,6-dimethoxyphenyl)-3-(1H-indol-3-yl)prop-2-en-1-one (TJS285, 6.25)

257

(E)-1-(2,4-dihydroxyphenyl)-3-(1H-indol-3-yl)prop-2-en-1-one (SC017, 6.26)

258

(E)-1-(2-Hydroxy-4-methoxyphenyl)-3-(1-methyl-1H-indol-3-yl)prop-2-en-1-one (6.27)

259

(E)-1-(2-hydroxy-4,6-dimethoxyphenyl)-3-(1-methyl-1H-indol-3-yl)prop-2-en-1-one (6.28)

260

(E)-3-(Furan-2-yl)-1-(2-hydroxyphenyl)prop-2-en-1-one (6.29)

261

(E)-3-(furan-2-yl)-1-(2-hydroxy-4,6-dimethoxyphenyl)prop-2-en-1-one (6.30)

262

(E)-1-(2-Hydroxyphenyl)-3-(1-methyl-1H-pyrrol-2-yl)prop-2-en-1-one (6.31)

263

(E)-1-(2-hydroxy-4-methoxyphenyl)-3-(1-methyl-1H-pyrrol-2-yl)prop-2-en-1-one (6.32)

264

(E)-1-(2-hydroxy-4,6-dimethoxyphenyl)-3-(1-methyl-1H-pyrrol-2-yl)prop-2-en-1-one (6.33)

265

(E)-1-(2-Hydroxyphenyl)-3-(pyridin-2-yl)prop-2-en-1-one (6.34)

266

(E)-3-(2,5-dimethoxyphenyl)-1-(2-hydroxy-4,6-dimethoxyphenyl)prop-2-en-1-one (6.35)

267

(E)-3-(2,4-dimethoxyphenyl)-1-(2-hydroxy-4,6-dimethoxyphenyl)prop-2-en-1-one (6.36)

268

1-(2-hydroxy-4,6-dimethoxyphenyl)-3-(1H-indol-3-yl)propan-1-one (6.37)

269