Profiling Protein Prenylation Using a Quantitative Chemical Proteomics Approach

A thesis submitted to Imperial College London in candidature for the degree of Doctor of Philosophy of Imperial College

Elisabeth Maria Storck Saha

Department of Chemistry Imperial College London Exhibition Road London SW7 2AZ

September 2016 Declaration of Originality

I hereby declare that this thesis is my own work and reports the results of my own original research. Information or results originating from the work of others or in collaboration with others has been acknowledged in the text and references.

Elisabeth Storck Saha September 2016

The copyright of this thesis rests with the author and is made available under a Creative Commons Attribution Non-Commercial No Derivatives licence. Researchers are free to copy, distribute or transmit the thesis on the condition that they attribute it, that they do not use it for commercial purposes and that they do not alter, transform or build upon it. For any reuse or redistribution, researchers must make clear to others the licence terms of this work.

2 Abstract

Prenylation, the attachment of a farnesyl or geranylgeranyl isoprenoid to a C-terminal cysteine, is a post-translational modification which modulates localisation and function of key proteins such as members of the RAS superfamily, nuclear lamins and heterotrimeric G-proteins. In humans, three enzymes catalyse the prenylation reaction: farnesyl transferase, geranylgeranyl transferase type 1 and RAB geranylgeranyl transferase.

Substantial effort has gone into developing therapeutics that disrupt prenylation as a means of targeting RAS-driven cancers, yet these approaches have failed to deliver in clinical trials. This is largely due to the fact that the dynamics of prenylation in response to prenyl transferase inhibitors is more complex than originally envisaged and includes changes in prenylation state and altered substrate recognition by the different transferase enzymes. Emerging research shows that prenylation also plays a role in cardiovascular and neurodegenerative diseases, Hutchinson-Gilford progeria syndrome (HGPS), choroideremia and viral infections.

Advances in the study of protein prenylation are hampered by the lack of inherent handles that allow for isolation and identification of the prenylated proteome. Novel approaches for profiling the prenylated proteome, its dynamics, and effects of varied inhibitors of the prenyl transferases are imperative for explaining past failures and to direct future studies.

This thesis presents a novel set of isoprenoid analogues that can selectively label prenylated proteins in cells. The application of these probes in combination with quantitative proteomic techniques enabled the description of the complex dynamics between different isoprenoid substrates and transferases in response to prenyl transferase inhibitors, and to validate several novel prenylated substrates. We describe attempts to profile prenylation in cells derived from HGPS patients, and explore potential targets of farnesyl transferase inhibition in pulmonary arterial cells. We envisage that the robust methodology presented herein will be widely applicable to the study of prenylation in the context of both health and disease.

Dedication

This thesis is dedicated to my mother, Karin, and mother-in-law, Chhaya, without whose incredible support this thesis would not have seen the light of day.

Acknowledgements

First and foremost I would like to thank my supervisors, Professor Ed Tate and Dr Beata Wojciak-Stothard. Ed – thanks for your unrelenting support and enthusiasm throughout the years I have spent in your lab. I will be forever grateful to you for taking me in as a stray back in 2009 and mentoring me though my undergraduate, masters and PhD years. Thanks for continually providing exciting project opportunities and for inspiring great science. Beata – it’s been an absolutely pleasure to work with you and your group. Thanks for teaching me the ropes in the biology lab and for your support and advice throughout this project.

Many thanks go to Dr Julia Morales-Sanfrutos and Dr Remi Serwa whose ideas and technical help have been vital in carrying this project forward. Thanks are also due Professor Riki Eggert for her incredible support during the final stages of thesis writing.

Thanks to Dr Goska Broncel and Dr Megan Wright for providing capture reagents. Thanks to Dr Andrew Bottrill at Leicester University proteomics facility for processing and analysing proteomics samples during the early stages of this project. Thanks to Dr Lisa Haigh at Imperial College mass spectrometry service for analysing small molecules and proteomics samples. Thanks to Drs Remi Serwa, Goska Broncel, James Clulow and Tom Charlton for running LC-MS/MS samples. Thanks to Drs Will Heal, Victor Goncalves and Jennie Hutton for advice and support in the synthesis lab. Thanks to Dr Lucie Duluc for providing cell samples. Thanks to Dr Paulina Ciepla for conducting labelling studies in Zebrafish.

Thanks to Dr Anna Barnard, Dr Jennie Hutton, Dr Julia Morales-Sanfrutos, Dr Charlotte Sutherell and Scott Lovell for proofreading this thesis.

Thanks to the lunch-club crowd who kept me sane during my PhD – Julia, Jenny, Jennie, Gee, Yunyun, Megan, Manue, Goska and Soo Mei. Thanks to all member of the Tate and Wojciak- Stothard lab for making my time in the lab so much fun.

Thanks to my husband Suj for his support and encouragement throughout my PhD years and for putting up with my grumpiness during the writing of this thesis. Thanks to my family, especially my mother Karin and mother-in-law Chhaya without whose incredible help with child-minding and cooking this thesis would never have been completed. Finally, no thanks go to my daughter Indira as her arrival in no way facilitated the progress of this PhD! However, she did make it all the more worthwhile, taught me just how much one can achieve with very little sleep and certainly filled my life with more love and fun than I could ever have imagined!

Acknowledgement of collaboration

Dr Julia Morales-Sanfrutos provided YnMyr-tagged HeLa lysate (Chapter 3), conducted studies to identify YnF-modified peptides in EA.hy926 cells (Chapter 3), prepared Tipifarnib- treated YnF-labelled samples in EA.hy926 cells (Chapter 4) and processed HGPS lysates for proteomics analysis (Chapter 5).

Dr Remigiusz Serwa provided prenyl probes (Chapter 2).

Dr Lisa Haigh (Imperial College mass spectrometry service) performed mass spectrometry analysis of small molecules and proteomics samples (Chapters 2-5).

Dr Megan Wright provided capture reagent AzTB and YnTB (Chapters 2-5) and Dr Goska Broncel provided capture reagent AzRTB (Chapter 3).

Dr Andrew Bottrill (University of Leicester, UK) analysed LC-MS/MS samples in initial proteomics studies (Chapter 3).

Professor Herbert Waldmann (Max Planck Institute of Molecular Physiology, Dortmund, Germany) provided RABGGTase inhibitor Bon-15 (Chapter 4).

Dr Lucie Duluc provided Tipifarnib-treated HPASMC and HPAEC lysates (Chapter 5).

Dr Paulina Ciepla conducted labelling studies in zebrafish (Chapter 5).

Contents

Publications & Presentations Arising From This Thesis ...... 11

Abbreviations ...... 13

Chapter 1: Introduction ...... 16

1.1. Protein Prenylation...... 16 1.2. The Prenyl Transferases ...... 21 1.2.1. FTase and GGTase-1 ...... 21 1.2.2. RABGGTase ...... 26 1.3. The Role of Prenylation in Disease ...... 28 1.3.1. Cancer ...... 29 1.3.2. Hutchinson-Gilford Progeria Syndrome ...... 30 1.3.3. Cardiovascular Disease ...... 32 1.3.4. Choroideremia ...... 34 1.3.5. Hepatitis Delta and Hepatitis C Viral Infection ...... 34 1.3.6. Other Diseases ...... 35 1.4. Chemical Proteomics ...... 36 1.4.1. Prenylation Probes ...... 37 1.4.2. Mass Spectrometry for Proteomics ...... 42 1.4.3. Quantitative Proteomics ...... 44 1.5. Project Objectives ...... 48 Chapter 2: Novel Probes for Profiling Protein Prenylation ...... 50

2.1. Introduction ...... 50 2.2. Synthesis of an Initial Panel of Azido and Propargyl Ether Prenyl Probes ...... 51 2.3. Initial Labelling Studies with Azido and Propargyl Ether Probes ...... 52 2.4. Effect of Statin Treatment on Labelling Efficiency of Propargyl Ether Probes ...... 56 2.5. An Expanded Library of Farnesyl Probes ...... 57 2.6. Evaluation of Shorter YnC14 Probe ...... 59 2.7. Development of Geranylgeranyl Probe YnGG ...... 60 2.8. Labelling by YnF and YnGG is Time and Concentration Dependent ...... 62 2.9. Labelling by YnF and YnGG is Sensitive to Natural Substrate Competition ...... 64 2.10. Conclusions ...... 66 Chapter 3: Development and Optimisation of Proteomics Platform ...... 67

3.1. Introduction ...... 67 3.2. Optimising Sample Processing: Does Reduction/Alkylation Affect Labelling? ...... 67 3.3. Initial Proteomics Evaluation of YnF, AzC15 and AzC20 ...... 69 3.4. Initial Comparison of YnF and YnGG ...... 73 3.5. Optimising Labelling Conditions for Proteomics Analysis ...... 76 3.6. Quantitative Comparison of Probe Preference ...... 84 3.7. Development of Novel Capture Reagents to Identify Probe-Modified Peptides ..... 86 3.8. Proof of Concept: Detection of Myristoylated Peptides in HeLa Cells with Novel Capture Reagents ...... 90 3.9. Optimising the Post-CuAAC Precipitation Protocol for AzRB ...... 92 3.10. Applying AzRB for the Detection of Prenylated Peptides ...... 93 3.11. Conclusions ...... 96 Chapter 4: De Novo Identification of Prenylated Proteins and Prenyl Transferase Inhibitor Dynamics ...... 98

4.1. Introduction ...... 98 4.2. Defining the Prenylome ...... 98 4.3. Profiling the Cellular Targets of Prenyl Transferase Inhibition ...... 101 4.3.1. In-gel Studies ...... 102 4.3.2. Initial Proteomic Evaluation of Inhibitor Response ...... 109 4.4. Evaluating an In-Cell Dose Response to Prenyl Transferase Inhibition ...... 114 4.5. Cellular Roles of Putative Novel Farnesylated Substrates ...... 119 4.6. Conclusions ...... 122 Chapter 5: Applications in Disease Models and In Vivo ...... 124

5.1. Introduction ...... 124 5.2. In-gel Labelling Studies in HPASMCs and HPAECs ...... 125 5.2.1. Optimising Labelling of Prenylated Proteins in HPASMCs and HPAECs ...... 125 5.2.2. Attempts to Detect Farnesylated and Geranylgeranylated RHOB ...... 127 5.3. Proteomics Studies in HPAECs and HPSMCs ...... 128 5.3.1. Hypoxia-treated Cells ...... 128 5.3.2. The Effect of Tipifarnib on YnF and YnGG Labelling in HPASMCs and HPAECs ...... 130 5.4. Labelling Studies in HGPS Patient Cells ...... 133 5.4.1. Optimising Labelling in HGPS and Healthy Fibroblasts ...... 134 5.4.2. Assessing the Effect of Tipifarnib on YnF and YnGG Labelling in HGPS Cells ...... 135 5.5. Quantifying Prenylation in HGPS Patient Cells by Proteomics ...... 136

5.5.1. Comparison of Prenylation in HGPS versus Healthy Fibroblasts...... 137 5.5.2. Response to Tipifarnib ...... 139 5.6. Proof-of-Concept: In vivo Labelling of Zebrafish ...... 142 5.7. Conclusions ...... 144 5.7.1. Technical Challenges to Address ...... 144 Chapter 6: Conclusions and Future Work ...... 147

6.1. Novel Probes for Metabolic Labelling of Prenylated Proteins ...... 147 6.1.1. Future Work ...... 147 6.2. Development of a Proteomics Workflow to Identify and Quantify Prenylation ...... 148 6.2.1. Future Work ...... 149 6.3. Inhibition Studies ...... 149 6.3.1. Future Work ...... 150 6.4. Prenylation in Disease Models and In Vivo ...... 150 6.4.1. Future Work ...... 151 Chapter 7: Experimental Methods ...... 152

7.1. Biological methods ...... 152 7.1.1. Reagents ...... 152 7.1.2. Cell culture ...... 152 7.1.3. General procedure for preparation of cell lysate tagged with YnF or YnGG . 153 7.1.4. Preparation of spike-in SILAC samples for concentration-gradient or time- course labelling studies in EA.hy926 cells ...... 153 7.1.5. Preparation of isoprenoid competition SILAC samples in EA.hy926 cells ..... 153 7.1.6. Preparation of spike-in SILAC samples for inhibitor evaluation in EA.hy926 cells 153 7.1.7. Preparation of hypoxia-treated HPAEC spike-in SILAC samples ...... 154 7.1.8. Preparation of Tipifarnib-treated HPAEC and HPASMC proteomics samples ...... 154 7.1.9. Preparation of HGPS and healthy fibroblast spike-in SILAC samples ...... 154 7.1.10. Click chemistry (CuAAC) general protocol ...... 155 7.1.11. SDS-PAGE and in-gel fluorescence imaging ...... 155 7.1.12. Protein enrichment for Western blot analysis ...... 155 7.1.13. Immunoprecipitation and on-bead CuAAC for detection of RHOB ...... 156 7.1.14. Protein enrichment and on-bead tryptic digest for proteomic analysis ...... 156 7.1.15. Stage-tip purification of peptides ...... 157 7.1.16. On-stagetip dimethyl labelling of peptides ...... 158 7.2. Proteomics analysis ...... 158

7.2.1. LC-MS/MS analysis ...... 158 7.2.2. Proteomics data processing in Maxquant ...... 159 7.2.3. Data analysis of initial proteomics comparison of YnF and YnGG probes .... 159 7.2.4. Spike-in SILAC concentration-gradient or time-course labelling studies proteomics data analysis ...... 160 7.2.5. Identification of YnMyr-modified peptides ...... 160 7.2.6. Isoprenoid competition proteomics data analysis ...... 160 7.2.7. Initial evaluation of prenyl transferase inhibitors (FTI-277, GGTI-2133 and Bon-15) data analysis ...... 160 7.2.8. Prenyl transferase inhibitor (FTI-277, Tipifarnib, Manumycin A and GGTI-2133) dose response data analysis ...... 161 7.2.9. Hypoxia-treated HPAEC data analysis ...... 161 7.2.10. Tipifarnib-treated HPAEC and HPSMC data analysis ...... 162 7.2.11. HGPS versus healthy fibroblast data analysis ...... 162 7.3. Organic Synthesis ...... 162 Bibliography ...... 183

Appendix A: List of electronic files ...... 202

Appendix B: Unedited Western Blot Images ...... 203

Appendix C: Spectra of YnF-AzRB Modified Peptides ...... 204

Appendix D: NMR Spectra of YnF ...... 208

Appendix E: NMR Spectra of YnGG ...... 209

Appendix F: Reprint Permission (Figure 9, Section 1.3.3) ...... 210

Publications & Presentations Arising From This Thesis

Publications

Wright, M.H., Paape, D., Storck, E.M., Serwa, R.A., Smith, D.F., Tate, E.W., Global analysis of protein N-myristoylation and exploration of N-myristoyltransferase as a drug target in the neglected human pathogen Leishmania donovani, Chem Biol, 2015, 22(3), 342-354

Tate, E.W., Kalesh, K.A., Lanyon-Hogg, T., Storck, E.M., Thinon E., Global profiling of protein lipidation using chemical proteomic technologies, Curr Opin Chem Biol, 2015, 24, 48-57

Storck, E.M., Serwa, R.A., Tate, E.W., Chemical proteomics: a powerful tool for exploring protein lipidation, Biochem Soc Trans, 2013, 41(1), 56-61

Storck, E.M., Wojciak-Stothard, B., Rho GTPases in pulmonary vascular dysfunction, Vasc Pharmacol, 2013, 58(3), 202-210

Oral Presentations

Storck Saha, E.M. et al., “Quantitative Profiling of Protein Prenylation Using a Chemical Proteomics Approach”, Chemical Biology and Molecular Medicine Student Symposium, Cancer Research UK Cambridge Institute, University of Cambridge, 13 Oct 2014

Storck Saha, E.M. et al., “Quantitative Profiling of Protein Prenylation Using a Chemical Proteomics Approach”, EMBO Chemical Biology, EMBL, Heidelberg, Germany, 20-23 Aug 2014*

Storck Saha, E.M. et al., “Exploring Protein Prenylation Using Quantitative Chemical Proteomics”, Department of Chemistry Postgraduate Symposium, Imperial College London, 1 Jul 2014

Storck Saha, E.M. et al., “Chemical Tools to Explore Protein Prenylation”, British Heart Foundation 4-year PhD Symposium, University of Manchester, 11 Apr 2014

Storck, E.M. et al., “Chemical Tools to Explore Protein Prenylation”, COST Spring Meeting, Berlin, 8-9 Apr 2013

Poster Presentations

Storck Saha, E.M. et al., “Quantitative Profiling of Protein Prenylation”, RSC Chemical Biology and Bio-Organic Group Postgraduate Symposium, University of Warwick, 1 Apr 2014*

Storck, E.M. et al., “Chemicals Tools for Exploring Protein Prenylation”, Department of Chemistry Postgraduate Symposium, Imperial College London, 2 Jul 2013**

Storck, E.M. et al., “Chemical Tools to Study Rho GTPase Prenylation in Pulmonary Hypertension”, RSC Postgraduate Biological & Medicinal Chemistry Symposium, University of Cambridge, 14 Dec 2012

Storck, E.M. et al., “Chemical Tools to Study Protein Prenylation”, PTMs in Cell Signalling, Copenhagen Bioscience Conferences, Copenhagen, 2-5 Dec 2012

Storck, E.M., et al., “Targeting Rho GTPase Prenylation in Pulmonary Hypertension”, BHF Centre Annual Symposium, Imperial College London, 25 Oct 2012

Storck, E.M. et al., “Targeting Rho GTPase Prenylation in Pulmonary Hypertension”, 2nd Symposium on Chemical Biology for Drug Discovery, AstraZeneca, Macclesfield, 20-21 Mar 2012

Prize awarded for best (*) or runner-up (**) presentation/poster

Abbreviations

AGOH – Anilinogeraniol

Az – Azide

AzKB - Azido-biotin capture reagent with lysine cleavage site

AzK+RB - Azido-trimethyllysine-biotin capture reagent with arginine cleavage site

AzRB – Azido-biotin capture reagent with arginine cleavage site

AzRTB - Azido-TAMRA-biotin capture reagent with arginine cleavage site

AzTB – Azido-TAMRA-biotin capture reagent

B-GPP – Biotin-geranyl pyrophosphate

CuAAC – Copper(I)-catalysed azide-alkyne cycloaddition

DCM – Dichloromethane

DML – Dimethyl labelling

DMSO – Dimethyl sulfoxide

EDTA – Ethylenediaminetetraacetic acid

EtOAc – Ethyl acetate

EtOH - Ethanol

Et2O – Diethyl ether

ESI – Electrospray ionisation

FBS – Foetal bovine serum

FOH – Farnesol

FPP – Farnesyl pyrophosphate

FPPS – Farnesyl pyrophosphate synthase

FTase – Farnesyl transferase

FTI – Farnesyl transferase inhibitor

GGOH – Geranylgeraniol

GGPP – Geranylgeranyl pyrophosphate

GGPPS – Geranylgeranyl pyrophosphate synthase

GGTase-1 – Geranylgeranyl transferase-1

GGTI – Geranylgeranyl transferase-1 inhibitor

HBV – Hepatitis B virus

HCV – Hepatitis C virus

HDV – Hepatitis delta virus

HGPS – Hutchinson-Gilford progeria syndrome

HMG-CoA – 3-Hydroxy-3-methyl-glutaryl-coenzyme A

ICMT – Isoprenylcysteine carboxyl methyltransferase

IR – Infrared spectroscopy iTRAQ – Isobaric Tag for Relative and Absolute Quantitation

LC – Liquid chromatography

MeOH - Methanol

MS – Mass spectrometry

MS/MS – Tandem mass spectrometry

MTO – Multiple turnover m/z – Mass-to-charge ratio

NMR – Nuclear magnetic resonance

Pe – Propargyl ether

PH – Pulmonary hypertension

PP - Pyrophosphate

PPTs – Pyridinium p-toluenesulfonate

PTI – Prenyl transferase inhibitor

PTM – Post-translational modification

PVDF – Polyvinylidene difluoride

14 12 14 12 R0K0 – N4 C6-arginine and N2 C6-lysine containing SILAC medium

15 13 15 13 R10K8 – N4 C6-arginine and N2 C6-lysine containing SILAC medium

RABGGTase – RAB geranylgeranyl transferase/Geranylgeranyl transferase-2

RCE1 – RAS-converting CAAX endopeptidase 1

REP1/2 – RAB escort protein 1/2

SDS – Sodium dodecyl sulfate

SDS-PAGE – Sodium dodecyl sulfate polyacrylamide gel electrophoresis

SILAC – Stable isotope labelling with amino acids in cell culture

STO – Single turnover

TAMRA – 5-Carboxytetramethylrhodamine

TBTA - Tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine

TCEP - Tris(2-carboxyethyl)phosphine

THF – Tetrahydrofuran

THP – Tetrahydropyranyl

TLC – Thin layer chromatography

TMT – Tandem mass tag

TOF – Time-of-flight

WB – Western blot

Yn – Alkynyl

YnF – Alkynyl-farnesol

YnGG – Alkynyl-geranylgeraniol

YnTB – Alkynyl-TAMRA-biotin capture reagent

Chapter 1: Introduction

1.1. Protein Prenylation

A myriad of post-translational modifications (PTMs) serve to expand the complexity of the proteomic repertoire beyond the simple sequences encoded by the genome. Whereas the human genome contains approximately 20,000 protein-encoding genes1, the resultant proteome is several-fold more complex due to alternative splicing of mRNA and post- translational modifications of proteins. PTMs regulate a myriad of vital cellular functions such as signalling, protein-protein interactions, protein stability and localisation. The importance of PTMs is apparent when one considers that in higher eukaryotes up to 5% of the genome encodes for proteins that mediate these modification events2. The PTM umbrella term encompass a wide range of processing steps that may occur during or after protein translation on the ribosome including proteolytic maturation of the protein sequence, disulfide bond formation or the covalent attachment of functional groups to specific sites on amino acid side chains or the polypeptide backbone. A wide range of modifying groups are employed, from small chemical groups such as methyl or phosphate groups, to lipids, carbohydrates and even small proteins like ubiquitin.

PTMs are either reversible or irreversible, depending on their function. For example, phosphorylation is typically involved in rapid cellular signalling cascades and is thus highly dynamic and tightly regulated by a large number of kinases and phosphatases. In contrast, many of the lipid PTMs that mediate protein localisation within the cell are irreversible. Lipid PTMs encompass a large class of modifications3; the structures of selected lipid PTMs are shown in Figure 1 and include N-, S- and O-acylation of protein N-termini and sidechains4, 5, carboxyl-terminal O-cholesteroylation6 and S-prenylation7.

S-prenylation involves the attachment of a farnesyl (C15) or geranylgeranyl (C20) isoprenoid to one or more cysteine residues near the C-terminus via a thio-ether bond7. This process is catalysed by 3 dedicated prenyl transferase enzymes. Farnesyl transferase (FTase) and geranylgeranyltransferase-1 (GGTase-1) prenylate a so-called CAAX motif where C is the prenylated cysteine, A is an aliphatic amino acid and X is the amino acid that determines specificity for FTase versus GGTase-1. Proteins prenylated by the CAAX prenyl transferases include members of the RAS superfamily8, the γ subunit of heterotrimeric G proteins9 and nuclear lamins10. Geranylgeranyltransferase-2 (GGTase-2/RABGGTase) attaches one or two geranylgeranyl chains to a variety of cysteine-containing C-terminal motifs (CC, CXC, CXXX or CAAX) on RAB proteins11, 12.

Figure 1. Examples of lipid post-translational modifications. The lipid modification is depicted in black and the protein sequence in red. Farnesylation was first described in a fungal mating factor from Rhodosporidium toruloides13. The first evidence of prenylation in mammalian cells was presented in 1984 when a study suggested that by-products of the mevalonate pathway could be incorporated post- translationally onto proteins in Swiss 3T3 cells and, although not conclusively shown, the authors suggested that these may be isoprenoid chains14. Radiolabelling studies conducted by the same group showed that lamin B15 was one of the proteins modified in this manner and it was soon confirmed that the modification was in fact a farnesyl group10. The discovery that mammalian RAS was also farnesylated generated a widespread interest in studying protein prenylation8, 16. In 1990, two separate studies showed that geranylgeranyl isoprenoids could also be incorporated onto proteins in HeLa and Chinese hamster ovary (CHO) cells17, 18.

The attachment of the isoprenoid lipid is the first of three steps that constitute prenylation (Figure 2). In mammalian cells, CAAX proteins undergo further processing by RAS-converting CAAX endopeptidase 1 (RCE1), which cleaves the C-terminal AAX tripeptide19, 20. The carboxyl group of the cysteine is subsequently methylated by isoprenylcysteine carboxyl methyltransferase (ICMT)21. The majority of RAB proteins containing a CAAX motif also undergo proteolysis and carboxylmethylation11. RAB proteins with a CXC motif, but not CC, are carboxyl methylated22. Recent studies have identified an esterase that can cleave the methyl group from RHOA, suggesting that this modification is dynamic23. The post-prenylation processing steps are thought to further increase the hydrophobicity of the prenylated proteins, mediating its interaction with membranes. The loss of post-prenylation processing has a more pronounced effect on the localisation of farnesylated proteins than geranylgeranylated proteins, possibly as the longer isoprenoid is already more hydrophobic24, 25.

Figure 2. Prenylation is mediated by FTase, GGTase-1 or RABGGTase. Farnesylation (A) and geranylgeranylation (B) of CAAX-motif proteins is mediated by FTase and GGTase-1 respectively. After attachment of the isoprenoid, the C-terminal –AAX is removed by RCE1 and the C-terminal carboxylate is methylated by ICMT. Geranylgeranylation of RAB proteins (C) is mediated by RABGGTase and REP1/2, which transfers one or two geranylgeranyl groups to a variety of cysteine-containing C-terminal sequences. RABs with a C-terminal CXXX motif will undergo RCE1-mediated cleavage of the –XXX group. Some RABs are carboxyl methylated by ICMT. Constitutive knock-out of FTase in mice causes embryonic lethality early in development with embryos displaying gross malformations26. In contrast, conditional knock-out of FTase in post- natal mice results in viable animals that develop into adulthood, although deficiency of FTase causes delayed wound healing and delayed maturation of erythroid precursor cells26. Further studies found that conditional knock-out of FTase and/or GGTase-1 in mouse hepatocytes leads to severe hepatocellular disease27. Fibroblasts from GGTase-1 knock-out mice show impaired proliferation and defective actin cytoskeleton, presumably due to loss of function of geranylgeranylated RHO GTPases28. RCE1 or ICMT knock-out in mice also causes embryonic lethality29, 30. ICMT knock-out mice died at an earlier stage of gestation suggesting a more severe phenotype, which could be due to the importance of carboxyl methylation for the function of RAB proteins. Furthermore, knock-out of ICMT had a more pronounced effect on the localisation of prenylated RHO GTPases than RCE1 knock-out31.

Many prenylated proteins contain a secondary mechanism for membrane targeting, suggesting that prenylation on its own is may not be sufficient to confer a stable association with membranes. For example, HRAS, NRAS and KRAS4A are palmitoylated and KRAS4B contains a polybasic region upstream of the prenylated cysteine32. The attachment of the palmitate group further increases the hydrophobicity of the protein, and the polybasic region promotes electrostatic interactions with negatively charged headgroups of membrane lipids (Figure 3). The palmitoylation event is reversible and functions to regulate association/dissociation from the plasma membrane32.

Figure 3. Prenylated proteins commonly require a secondary membrane targeting mechanism. Proteins (e.g. HRAS) may be palmitoylated on cysteine residues upstream of the prenylation site. Poly-basic regions in proteins such as KRAS4B enhance membrane binding by interacting with negatively charged headgroups of membrane lipids. Many RAB proteins are doubly geranylgeranylated to enhance membrane affinity. In mammalian cells isoprenoids are synthesised as intermediates of the mevalonate pathway (Figure 4), which is the biosynthetic pathway that produces cholesterol33, 34. Pharmacological inhibition of the mevalonate pathway by HMG-CoA reductase inhibitors (statins) results in the depletion of cellular isoprenoids, indirectly inhibiting prenylation. Although statin therapy was developed as a means to reduce cholesterol levels, it has become clear that some of the beneficial effects of statin therapy can be attributed to inhibition of prenylated proteins35, 36. Isoprenoids can also be depleted by inhibition of the farnesyl pyrophosphate synthase (FPPS) and geranylgeranyl pyrophosphate synthase (GGPPS) enzymes. Extensive effort has gone into the development of therapies that directly inhibit the prenyl transferase enzymes, driven by the observation that prenylation is required for RAS function37. RAS is an oncoprotein and activating RAS mutations are present in >30% of cancers, and thus represent an important therapeutic target. Many potent prenyl transferase inhibitors (PTIs) have been developed and assessed in clinical trials, however with limited success. This failure has been mainly attributed to alternative prenylation of RAS proteins by GGTase-1 upon inhibition of FTase, enabling RAS to remain functional37. However, an incomplete picture of all the cellular substrates that undergo prenylation limits our understanding of the effects of prenyl transferase inhibition.

Figure 4. Synthesis of farnesyl and geranylgeranyl pyrophosphate via the mevalonate pathway. Enzyme names are shown in red and substrate/product names in blue.

1.2. The Prenyl Transferases

1.2.1. FTase and GGTase-1

Structure and catalytic mechanism

Mammalian FTase was first isolated from rat brain in 199028 and GGTase-1 shortly thereafter from bovine brain38, 39. Both enzymes are heterodimers, sharing a common 48 kDa α-subunit but each possessing a unique β-subunit of 46 kDa and 43 kDa respectively38, 40, 41. FTase and GGTase-1 are metallo-enzymes requiring a zinc cation for catalytic activity42, 43. FTase, but not GGTase-1, activity additionally depends on the presence of Mg2+ ions44.

Extensive crystallographic data are available for FTase and GGTase-145-53, which show that the two enzymes have near identical topology and any structural convergences are mainly contained within solvent-exposed loops51. Both subunits of the heterodimers consist of mainly α-helical elements. The shared α-subunit forms seven successive coiled-coils, resulting in a crescent shaped domain that partly envelops the β subunit45, 46. The β-subunit forms 12 α-helices that fold into an α-α barrel with a central cavity. One end of the barrel is blocked by a loop while the other end is open and solvent accessible45, 51. The catalytic zinc ion is coordinated to Cys299β (Cys271β in GGTase-1), Asp297β (Asp296β in GGTase-1) and His362β (His321β in GGTase-1), positioning it at the top end of the barrel51.

Structures of the isoprenoid binding pocket of FTase and GGTase-1 are depicted in Figure 5. The depth of the β-subunit barrel forms a “molecular ruler” that determines the enzyme’s preference for FPP or GGPP as a substrate45, 46. The second and third isoprene units of FPP and GGPP occupy identical positions in FTase and GGTase-1 respectively. The first isoprene unit of GGPP is slightly shifted in GGTase-1 compared to the binding of this unit of FPP in FTase. All isoprene units of FPP and the first three units of GGPP form a straight axis, while the terminal isoprene of GGPP sits at a right angle in relation to this axis. The depth of the barrel cavity is dictated by Trp102β in FTase and Thr49β in GGTase-146, 51. The fourth isoprene unit of GGPP occupies the space in GGTase-1 that in FTase is filled by the bulkier Trp102β, forming part of the basis for FPP selectivity. Although GGPP can bind to FTase, it cannot be utilised as a substrate42, 54. Mutation of Trp102β to a Thr residue changes the isoprenoid specificity of FTase, resulting in a preference for GGPP over FPP51, 53.

Figure 5. Isoprenoid binding pocket of FTase (A) and GGTase-1 (B). The depth of the FTase binding pocket is limited by residues Trp102β and Tyr365β, resulting in selectivity for FPP. The corresponding residues in GGTase-1 (Thr49β and Phe 324β) are smaller, allowing GGTase-1 to accommodate GGPP in the isoprenoid binding pocket. Images generated in PyMOL (v. 1.3r1) using PDB structures 1N4P (GGTase-1) and 2F0Y (FTase). Structural and kinetic studies support an ordered reaction mechanism for prenylation in which the enzyme first binds the prenyl pyrophosphate, generating the binding site for the CAAX protein substrate50. Although the CAAX peptide can bind directly to the apo-enzyme, the result is a non-productive conformation from which isoprenoid binding and product formation cannot follow. The cysteine thiol of the CAAX peptide co-ordinates to the zinc cation, forming a ternary complex55, which activates the thiol for attack on C1 of the isoprenoid. The reaction is preceded by a rotation of the first two isoprene units to bring the prenyl pyrophosphate in proximity of the cysteine thiol50. In FTase, Lys164α (Lys164α in GGTase-1) and Tyr300β (Tyr272β in GGTase-1) are crucial for catalytic activity50, 56, 57. Tyr300β stabilises the negative charge of the transition state by interacting with the oxygen atom between C1 and the α- phosphate, and Lys164α forms a hydrogen bond with the α-phosphate50. The magnesium ion coordinates the diphosphate moiety and Asp352β50. Sequence alignment and mutational studies suggest the Lys311β residue in GGTase-1 (Asp325β in FTase) replaces the Mg2+ required by FTase58, 59. The reaction pathway provides further information about the isoprenoid selectivity of FTase. X-ray crystallographic studies have shown that FTase can accommodate GGPP within its isoprenoid binding site; however, the conformation adopted by GGPP prohibits the necessary rearrangement to bring the isoprenoid into contact with the cysteine of the substrate peptide and thus the prenyl transfer reaction cannot occur54.

The rate limiting step of the reaction mechanism is release of the prenylated substrate. Before product release, a new isoprenoid molecule binds leading to displacement of the isoprenoid of the prenylated protein to a new binding site, the so-called “exit groove”50, 51. In the newly adopted conformation the thiol is no longer co-ordinated to the zinc ion50, 51. Full displacement of the prenylated protein from the prenyl transferase enzyme requires the binding of a new substrate molecule51, 60. The enzyme mechanism is unique in that the product release step helps to dictate the isoprenoid preference for each enzyme. In GGTase-1, FPP cannot displace the prenylated peptide product from the enzyme51. Thus, although FPP can sterically be accommodated within the GGTase-1 binding pocket, there is no opportunity for FPP to bind during the reaction cycle. The catalytic cycle of FTase, which is analogous to the reaction cycle of GGTase-1, and the key interactions of the transition state for the two enzymes are depicted in Figure 6.

Figure 6. (A) The catalytic cycle of FTase-mediated farnesylation. FPP and peptide substrate bind in the active site of FTase in an ordered mechanism. Following isoprenoid transfer to the substrate peptide, a new FPP molecule binds, displacing the isoprenoid of the prenylated substrate to an exit groove. The product release is mediated by binding of a new peptide substrate. Geranylgeranylation by GGTase-1 follows an analogous catalytic cycle. The key interactions stabilising the transition state of the isoprenoid transfer reaction in FTase (B) and GGTase-1 (C). The catalytic zinc coordinates a Cys, Asp and His residue of the enzyme and the cysteine thiol of the peptide substrate, activating the thiol for attack on C1 of the isoprenoid. The pyrophosphate moiety interacts with a Tyr and Lys residue and a magnesium ion in FTase and a Tyr and two Lys residues in GGTase-1 to stabilise the negative charge of the transition state. The peptide substrate is shown in red, the isoprenoid in black, the catalytic zinc ion in blue and the enzyme residue interactions in green.

Comparison between rat and human FTase structures show that all residues within the FPP and peptide binding sites are the same and adopt near identical conformations. Non- conserved residues lie in the disordered termini or in loops connecting helices in the α-subunit49. In crystal structures of the apo-enzyme45, 46 the nine C-terminal amino acids of the β-subunit occupy the peptide binding site of the adjacent protein molecule in the crystal lattice. In fact, the binding of the nona-peptide results in a conformational change that would occlude FPP binding. The implications of this binding and any functional consequences are unclear, and is not present in structures of the enzyme:FPP complex46. Interestingly, this nona-peptide contains some of the non-conserved regions in rat versus human FTase. Whether this observation is an artefact of crystallisation that is not applicable in vivo remains unanswered.

Peptide substrate selectivity

Traditionally the “X” of the CA1A2X motif was thought to dictate the preference for FTase or GGTase-161, 62, with FTase preferring Met, Ser, Ala or Gln and GGTase-1 preferring Leu63. However, more recent studies have shown that the selectivity criteria are more complex and synergistically depend on the whole -AAX motif64-66. The prenyl pyrophosphate makes substantial contact with the CAAX peptide substrate, and this interaction is considered to in part affect the selectivity of the peptide substrate50.

The zinc ion, in addition to its role in catalysis, greatly enhances the binding of the CAAX peptide substrate and is required for correct binding in the enzyme active site43, 48. The structure of the zinc-depleted FTase:FPP:peptide complex indicates that zinc has no impact on FPP positioning or the overall conformation of the enzyme. However, the binding of the CAAX peptide changes markedly in the absence of zinc. Rather than the extended conformation observed in the presence of zinc, the CAAX peptide forms a β turn which repositions the Cys and A1 residues, leading to a 9.2Å displacement of the cysteine thiol in comparison to the zinc-containing enzyme.

In structural studies of both enzymes with non-reactive isoprenoid analogues and model CAAX peptides the carboxylate of the X residue coordinates to Gln167α, anchoring the peptide at the bottom of the binding site47, 48, 51. Further interactions with the carboxylate are mediated through water solvent molecules coordinated to His149β (His121β in GGTase-1), Glu198β (Glu169β in GGTase-1) and Arg202β (Arg173β in GGTase-1). Arg202β/Arg173β also forms a direct hydrogen bond with the backbone carbonyl oxygen of the A2 residue. The X-residue makes extensive van der Waals interactions with surrounding amino acids in the enzyme pocket. The residues forming the cavity containing the X residue are different in FTase and GGTase-1, which likely dictates the divergent preferences at the X-position 48, 51. For example, modelling studies show that in addition to Met, FTase can accommodate Ser, Ala, Gln, but not

Leu due to steric clashes with amino acid sidechains. Furthermore, residues such as Glu and Lys cannot be accommodated due to their charge48. GGTase-1 can accommodate Met and Phe, which are weak substrates of this enzyme. Additionally, Gly and Ala can fit, but do not fill, the X-residue cavity of GGTase-151. The residues surrounding the X-residue are however conserved between human and rat FTase, indicating that these two enzymes will have similar peptide substrate preferences48. This is an important aspect to consider as many in vitro studies are conducted with rat FTase.

The A1 residue of the substrate peptide is solvent-exposed, pointing away from the enzyme:isoprenoid complex in both FTase and GGTase-1. The A2 residue, on the other hand, makes extensive van der Waals interactions with the enzyme:isoprenoid complex47, 48, 51. This may account for the relaxed selectivity at the A1 residue and preference for hydrophobic

67 residues at the A2 position . The enzyme residues that make contact with A2 are conserved across FTases from different species, but are different in GGTase-1. In FTase, A2 interacts with Tyr361β, Trp102β, Trp106β and the second and third isoprene units of FPP47, 48. In GGTase-1 these interactions are made by Phe53β, Leu320β and the fourth isoprene unit of GGPP51.

CAAX proteins with a polybasic region upstream of the CAAX motif (e.g. KKKSKTKCVIM from KRAS4B) make further interactions with FTase, possibly accounting for the increased affinity of the KRAS substrate compared to that of HRAS and NRAS, which both lack polybasic regions48. It has been proposed that GGTase-1 makes similar interactions with KRAS4B, which could explain the ability of GGTase-1 to accept KRAS as a substrate68.

A large number of in vitro studies have been conducted with model peptides to decipher the sequence selectivity of the prenyl transferases. Enzymatic studies with CAAX peptides show that substrates can be prenylated with either single- or multiple turnover (STO or MTO) reactivity66, 69, 70. STO substrates are prenylated, but the peptide product does not dissociate from the enzyme, effectively inhibiting the prenyl transferase. However, upon addition of a MTO substrate, prenylated STO substrates can be released, suggesting that these types of sequences may in fact be prenylated in vivo69. Hougland et al. screened a large library of dansyl-TKCXXX peptides to assess their prenylation by FTase, identifying 106 MTO substrate sequences and a further 130 STO substrate sequences70. A separate study showed that the hydrophobicity of the X residue affects enzyme reactivity, but not binding69. The steady-state rate constant for FTase decreased with increasing hydrophobicity or volume of the X residue, while the opposite was true for GGTase-169.

Deciphering substrate selectivity based on CAAX peptides has limitations as further selectivity is conferred by the sequence preceding the CAAX box. For example, Dansyl-GCVIL65 was

25 found to be good MTO substrate for FTase while in a different study Dansyl-TKCVIL69 only demonstrated STO substrate activity. Proteins with identical CAAX-motifs can also exhibit different substrate selectivity. Both RHOF and RAC1 harbour a C-terminal CLLL sequence, but whereas RHOF is a dual substrate for both FTase and GGTase-1, RAC1 is only a substrate for GGTase-131. Furthermore, a Dansyl-GCLLL peptide was found to be a poor substrate for FTase in enzyme assay65.

1.2.2. RABGGTase

As for the CAAX prenyl transferases, RABGGTase structure and function have been the subject of intense research since the discovery of the enzyme in the early 1990s11, 12. Similarly to the CAAX prenyl transferases, RABGGTase forms a heterodimer consisting of an α and β subunit of 60 and 38 kDa, respectively61, 71. The α-subunit of rat RABGGTase consists of three distinct domains: a helical domain, an immunoglobulin (Ig)-like domain and leucine-rich repeat (LRR) domain72. The helical domain is structurally similar to the α-subunit of FTase and GGTase-1, consisting of 15 α-helices that fold into a crescent-shaped structure. The Ig-like domain is inserted between helices α11 and α12, forming a β sandwich. The LRR domain forms a right-handed superhelix composed of the C-terminal sequence of the subunit. The function of the Ig-like and LRR domains are unclear, although both domain types are typically involved in mediating interactions with other proteins and cellular components73. These domains are not conserved across different species and their deletion do not alter the catalytic function of RABGGTase73, 74. Analogous to FTase and GGTase-1, the β-subunit forms an α-α barrel composed of 12 α-helices, which is closed off at one end. The cavity of the barrel is lined with hydrophobic residues and the outside is partially enveloped by the α-subunit. The catalytic zinc ion is positioned at the open end of the barrel, co-ordinated to residues Asp238β, Cys240β, His290β and His2α.

The isoprenoid binding pocket of RABGGTase is deeper and wider than the corresponding binding site in FTase and GGTase-172, 74. In RABGGTase the bulky Trp102β which restricts the binding of GGPP to FTase is replaced by the smaller Ser48β at the bottom of the cavity. The isoprenoid chain makes extensive hydrophobic interactions with aromatic residues lining the barrel and the pyrophosphate moiety binds in a hydrophilic region close the catalytic zinc. The β-phosphate forms hydrogen bonds with Lys105α and Lys235β and the α-phosphate forms hydrogen bonds with Arg232β and a water molecule74.

In RABGGTase, Gly108α replaces the Gln167α residue that anchors the C-terminal carboxylate group of the substrate peptide to the bottom of the binding site in FTase and GGTase-172. As a result, substrate peptides do not bind in a well-defined conformation, which may explain the ability of RABGGTase to prenylate a variety of C-terminal sequences. The

26 flexible binding of the substrate peptide have rendered attempts to obtain structural information about RABGGTase:peptide complexes unfruitful74. In contrast to FTase, RABGGTase does not appear to have an “exit groove” where the isoprenoid of the prenylated peptide can bind74. In x-ray structures of RABGGTase with a doubly geranylgeranylated RAB7 peptide, only the isoprenoid bound to the active site can be identified, indicating that the second geranylgeranyl group does not adopt a defined conformation74.

The mechanism of prenylation by RABGGTase is distinct from that of the CAAX prenyl transferases. Prenylation of RABs by RABGGTase requires the accessory proteins RAB escort protein 1 or 2 (REP1/REP2)75, 76. RAB proteins are unable to bind RABGGTase independently of REP and unlike the CAAX prenyl transferase enzymes, RABGGTase is unable to prenylate short peptide substrates. Structures of a REP1:RAB7 complex shows that REP1 binds the globular domain of RAB7 through its RAB-binding platform (RBP)77. Additionally, the C-terminal region of RAB7 interacts with the C-terminus binding region (CBR) and the prenyl tail is bound by Domain II of REP1. Formation of the REP1:RAB7 complex is dependent on the RBP interaction. Mutations in the CBR domain show that this interaction is not necessary for REP1:RAB7 binding, but is required for prenylation, presumably by stabilising the disordered RAB C-terminus to enhance interaction with the active site of RABGGTase77.

The order of events in the RABGGTase prenylation reaction is unclear. The classic model, which is outlined in Figure 7, supports a mechanism where newly synthesised RABs bind to its associated REP11. The RAB:REP complex subsequently binds to RABGGTase which is loaded with a GGPP molecule. Inhibition studies indicate that in the case of doubly prenylated RABs the N-terminal cysteine is modified first78. After the first prenylation event the attached isoprenoid is moved to a lipid binding cavity in REP and a second GGPP molecule binds to RABGGTase without dissociation of the RAB:REP complex11, 79. After the second prenylation step the prenylated RAB:REP complex is dislodged from RABGGTase by binding of a new GGPP substrate molecule80. REP acts as a chaperone for the geranylgeranylated RAB protein, escorting it to its final membrane location. RAB proteins that contain two cysteines within their prenylation motif require double prenylation for correct targeting within the cell81, 82. In an alternate model, REP first binds RABGGTase, and RAB binds to the pre-formed REP:RABGGTase complex. The exact order of events appears to be dominated by the available concentration of each component of the complex. The prenylation reaction proceeds via a similar mechanism independent of the order of formation of the RAB:REP:RABGGTase complex.

Figure 7. The classic model for RAB prenylation by RABGGTase/REP. Newly synthesised RAB proteins bind to REP1 or 2, and the complex subsequently associates with GGPP-bound RABGGTase. One or two GG moieties are transferred to the RAB protein in successive order. The RAB:REP complex subsequently dissociates from RABGGTase, and REP escorts RAB to its membrane position. Traditionally, RAB proteins have been viewed as exclusively geranylgeranylated rather than farnesylated. Binding studies with FPP and GGPP indicate that although RABGGTase can bind FPP, it does so with less affinity than GGPP. Depending on the assay employed, the affinity of GGPP over FPP is estimated at 10-100-fold74, 83. In vitro, RABGGTase is able to use both efficiently as substrates for prenylation of RAB proteins74, but only geranylgeranylation is typically observed in cellular systems. Recent evidence suggests that the short isoform of RAB28, which has a C-terminal CAAX motif, is farnesylated, but that this modification event may be mediated by FTase84. Other CAAX-RABs such as RAB8 have been shown to act as dual substrates for both RABGGTase and GGTase-1 in vitro, although in vivo this prenylation is predominantly mediated by RABGGTase only85.

1.3. The Role of Prenylation in Disease

Prenylation affects the cellular localisation and function of a large number of proteins, and thus has important implications for cellular function both in health and disease7, 37, 86. Several prenylated proteins, such as oncogenic RAS, play a key role in cancer initiation and progression. Other diseases, such as Hutchinson-Gilford progeria syndrome (HGPS) and choroideremia, result from mutations that cause aberrant prenylation of selected proteins. This section briefly covers the role of prenylation in a number of pathologies, including cancer, HGPS, cardiovascular disease, choroideremia, viral and parasitic infections and neurodegenerative disease.

1.3.1. Cancer

The interest in targeting prenylation as a treatment for cancer stems from the importance of RAS in tumour initiation and progression, coupled to the discovery that prenylation is essential for the correct function of RAS37. The RAS oncogene is mutated in >30% of cancers, and its mutation incidence has been reported to be as high as 80-92% in pancreatic adenocarcinomas87-89. RAS mutations are also highly prevalent in colon and thyroid carcinomas, adenocarcinomas of the lung and acute myeloid leukemia87. RAS is a small GTPase protein, which cycles between an active GTP-bound form and an inactive GDP-bound form. Transforming mutations result in loss of GTPase activity, leading to constitutive activation of RAS90. In humans, RAS is encoded by three genes (NRAS, HRAS and KRAS) producing four protein isoforms: NRAS, HRAS, KRAS4A and KRAS4B. Differences between the RAS proteins are mainly contained within the C-terminal hypervariable region (HVR) which mediates cellular trafficking and localisation91. The two KRAS isoforms are produced by alternative splicing of the fourth exon, which encodes the HVR. As a result, KRAS4A harbours a palmitoylation site which is not present in KRAS4B.

RAS has been a notoriously “undruggable” target, and to date no RAS inhibitor is available in the clinic90. The importance of farnesylation to RAS subcellular localisation and function has generated nearly 30 years of research focused on the development of farnesyl transferase inhibitors (FTIs) to target cancer37. However, despite exciting results in preclinical and animal studies, more than 70 clinical trials have failed to show robust results in the clinic37, 92. The failure of FTIs is multi-faceted, but has mainly been attributed to the alternate geranylgeranylation of KRAS and NRAS by GGTase-1 in the presence of FTIs37, 93, 94. In contrast, HRAS cannot be geranylgeranylated by GGTase-195.

In addition to RAS, numerous other prenylated proteins play a role in cancer37. RAS homologue enriched in brain (RHEB) is an exclusively farnesylated protein which is commonly over-expressed in various cancers, including those of the bladder, lung, liver, breast and prostate96, 97. RHEB is localised to endomembranes and stimulates the protein kinase mTOR, a positive regulator of cellular growth and proliferation, either by direct binding or by binding to an mTOR suppressor protein. In agreement with the requirement of farnesylation for RHEB activity, it was found that lymphoma cell lines with high RHEB expression were more sensitive to FTI treatment98. Several centromere associated proteins (CENP-E and CENP-F) also require farnesylation for correct function. Pharmacological inhibition of CENP-E and CENP-F farnesylation results in defects in chromosome alignment and segregation, which may explain the delay in G2/M progression observed in cells treated with FTIs99, 100.

Several geranylgeranylated proteins such as the RHO GTPases are also likely to play an important role in cancer101. RHO GTPases such as RHOA and RAC1 are key regulators of cellular mechanisms such as proliferation, migration, cell adhesion and invasion, all of which have important implications for cancer progression. Although mutations of the RHO GTPases are rare, RHOA and RAC1 have been found to be up-regulated in numerous types of cancer102. Inhibition of GGTase-1 was found to limit the invasive phenotype of human colon cancer cells, which was attributed to the loss of RHOA geranylgeranylation and membrane association103. Ral GTPases (RALA and RALB) are also geranylgeranylated, and act downstream of RAS signalling to regulate cancer cell survival and proliferation.104 Migration and invasion enhancer 1 (MIEN1) was recently described as a novel geranylgeranylated protein105, which mediates cancer migration and invasion. MIEN1 is highly expressed in breast, prostate and oral cancer tissue, with increased expression levels correlated with poor clinical outcomes106-109.

1.3.2. Hutchinson-Gilford Progeria Syndrome

Hutchinson-Gilford progeria syndrome (HGPS) is an extremely rare genetic condition with a striking phenotype of accelerated ageing leading to patient death during early teenage years, generally as a result of cardiovascular pathologies110, 111. In most cases HGPS is caused by a de novo point mutation (GGC to GGT, G608G) in the LMNA gene112, 113. This silent mutation introduces a cryptic splice site leading to alternative splicing of exon 11 and ultimately the deletion of a 50 amino-acid stretch near the C-terminus of the prelamin A/C (LMNA) protein112, 113. This mutant form of LMNA is termed progerin.

LMNA has a C-terminal prenylation motif (CSIM) which undergoes farnesylation by FTase followed by RCE1 proteolysis and ICMT carboxyl methylation114 (Figure 8). Farnesylation of LMNA targets the protein to the inner nuclear membrane where it undergoes proteolysis by ZMPSTE24. ZMPSTE24 processing removes the 15 C-terminal amino-acid residues, including the farnesylated cysteine methyl ester, yielding mature lamin A115. Lamin A is an intermediate filament protein that forms a key component of the nuclear lamina which is located underneath the inner nuclear membrane and provides structural support to the nucleus116, 117. Progerin, the truncated version of LMNA, harbours the prenylation motif but not the ZMPSTE24 cleavage site. As a result, progerin remains farnesylated and accumulates at the nuclear rim, causing nuclear envelope defects such as blebbing and thickening of the nuclear lamina118. The differences in LMNA and progerin processing are highlighted in Figure 8 below.

Figure 8. Processing of prelamin A/C and progerin. Prelamin A/C (LMMNA) and progerin both undergo farnesylation by FTase, followed by RCE1-mediated cleavage of the –AAX C-terminal sequence and carboxyl methylation by ICMT. Proteolytic cleavage of LMNA by ZMPSTE24 removes 15 amino acids from the C-terminus, yielding mature lamin A. Progerin cannot be processed by ZMPSTE24 due to the loss of the cleavage site, and thus retains the farnesylated cysteine. Blocking farnesylation of progerin leads to accumulation of the protein in the nucleoplasm instead of the nuclear rim119. Extensive research in mouse models and cell lines from HGPS patients indicated that inhibiting the farnesylation of progerin reduces the nuclear envelope defects associated with HGPS and ameliorates the disease phenotype118, 120-128. A clinical trial with the FTI Lonafarnib recently showed positive outcomes in a group of HGPS patients129, 130, resulting in extended life-span for the FTI-treated cohort, along with improvements in weight gain, cardiovascular parameters and bone density. However, no correlation between loss of prenylation and clinical response could be established.

Despite the huge excitement surrounding the potential therapeutic benefits of FTIs in treating HGPS, some key questions remain unanswered111. The striking effects of FTIs in improving nuclear morphology in model systems cannot directly be attributed to the inhibition of progerin farnesylation. For example, a knock-in mouse model harbouring a non-prenylated progerin protein (-CSIM to –SSIM C-terminal mutation) still produced a diseased phenotype, suggesting that the prenylation status of progerin may be irrelevant.131 Although the phenotype generated by the non-farnesylated progerin was milder, it was concurrently shown that cellular levels of the mutant protein were lower than those of farnesylated progerin. Similarly, it was shown that FTI treatment lowered the steady state levels of normal progerin. Taken together, it appears that progerin exerts toxic effects independent of its prenylation state, but loss of

31 farnesylation may reduce the severity of the disease phenotype by lowering the levels of progerin in cells. Remarkably, a further study from the same lab generated a very different result. The group generated a second knock-in mouse model where the C-terminal CAAX motif was mutated to CSM by deletion of the Ile residue. This mouse model was free of disease132. In both models the mutant progerin localised at the nuclear rim in a fashion indistinguishable from lamin A localisation.

The conflicting evidence pertaining to the role of farnesylated progerin in the pathology of HGPS requires further investigation. The positive outcomes of the recent clinical trial are encouraging, however; although FTIs were able to prolong the life-span of patients, it did not provide a cure for the condition. Furthermore, LMNA processing has also been identified as a sign of vascular ageing in healthy individuals133, highlighting the need to better understand the cellular implications of its farnesylation status.

1.3.3. Cardiovascular Disease

The main interest in the role of prenylation in cardiovascular disease stems from the fact that statin therapy yields beneficial effects that cannot simply be attributed to their cholesterol- lowering properties134, 135. As discussed in section 1.1 and outlined in Figure 4, isoprenoids are intermediates of cholesterol biosynthesis, and are depleted upon treatment with statins, resulting in indirect inhibition of protein prenylation136.

The main use of statins has been in the prevention and treatment of atherosclerosis which is a key underlying cause of cardiovascular disease and death worldwide. Atherosclerosis is characterised by the development of atherosclerotic lesions, which are depositions of lipids, cells and connective tissue, in the inner layer of the vessel walls. Thrombosis resulting from rupture of unstable atherosclerotic lesions is a leading cause of myocardial infarction and sudden death. Atherosclerosis is a highly complex and progressive disease, but key pathological features include de-regulated lipid metabolism and chronic inflammation.

RHO GTPases and their downstream effectors have been heavily implicated in the pathogenesis of atherosclerosis and other cardiovascular diseases. Loss of RHO prenylation in particular is thought to mediate some of the beneficial effects conferred by statins. The complex involvement of RHO and other small GTPases in cardiovascular disease has been covered in great detail in previous reviews137-139.

The RHO GTPases are also key regulators of the pulmonary vasculature, controlling vascular tone and remodelling. Their basal activity is important in foetal lung development and vascular adaptation to changes in oxygen levels, but their continuous activation in the neonatal or adult lung leads to the development of pulmonary hypertension (PH). PH is associated with elevated

32 pressures in the pulmonary circulation due to remodelling and constriction of the pulmonary vasculature. PH is classified into 5 different subclasses categorised according to the underlying cause of disease140.

The role of RHO GTPases in pulmonary vascular (dys-)function was reviewed as part of this PhD project141. RHOA is the best characterised of the small GTPases in the context of PH due to its activation of RHO kinase (ROCK)142. Vascular tone is controlled by the 20 kDa myosin light chain (MLC20), which in its active state is phosphorylated on Ser19. ROCK phosphorylates the myosin-binding subunit of MLC phosphatase (MLCP), which renders

MLCP inactive and results in sustained activation of MLC20. ROCK is also able to directly

141 phosphorylate, and hence activate, MLC20 .

Figure 9. Proposed role of small GTPases RHOA and RHOB in the development of pulmonary hypertension. Abbreviations: Akt, protein kinase B; LIMK, Lim-kinase; mDia, mammalian diaphanous; MLC20, 20 kDa myosin light-chain ; MLCP, myosin light chain phosphatase; NF-κB, nuclear factor kappa B; PDGF, platelet derived growth factor; PKA, cAMP-dependent protein kinase; PKG, cGMP-dependent protein kinase; ROCK, Rho kinase; SLK, Ste20-related kinase. Figure reproduced with permission from Storck et al.141 Recently the Wojciak-Stothard lab published a first account of the role for RHOB in the development of hypoxia-induced PH143. They found that hypoxia induces RHOB in both pulmonary arterial endothelial and smooth muscle cells independently of RHOA, and that

33 development of hypoxia-induced PH was attenuated in a RHOB null mice model. Furthermore, they showed that RHOB was able to stabilise hypoxia-inducible factor 1α (HIF-1α). HIF-1α is a transcription factor for endothelin-1, another key regulator of vascular tone144. Treatment with a FTI attenuated the effects of RHOB over-expression, suggesting that inhibition of RHOB prenylation may present a novel treatment for hypoxia-induced PH143. Further in vivo experiments in the Wojciak-Stothard lab indicated that treatment with the highly potent FTI Tipifarnib protected against development of hypoxia-induced PH by preventing several pathological mechanisms typically associated with the disease. Tipifarnib-treated animals showed improved endothelial function as well as decreased vascular remodelling and right heart hypertrophy (Duluc et al., manuscript in preparation).

1.3.4. Choroideremia

Choroideremia is an X-linked genetic disease caused by loss-of-function mutations in the CHM gene, which encodes the REP1 protein145, 146. Choroideremia, which affects approximately 1 in 50,000 people, causes degeneration of the choroid, retinal pigment epithelium and photoreceptors. Gradual loss of vision results in blindness by middle age in affected patients. Because REP1 is required for prenylation by RABGGTase, loss of REP1 results in the accumulation of un-prenylated RAB proteins. Although choroideremia patients retain a functional REP2 protein, it is unable to compensate for the loss of REP1 in the eye.

Analysis of the pool of un-prenylated proteins from choroideremia cell lines identified accumulations of RAB27A, suggesting that this protein was a key player in the pathogenesis of the disease147. RAB27A is abundant in the eye where it localises to the retinal pigment epithelium and choriocappillaris. Kinetic analysis of RAB27A prenylation revealed that although RAB27A binds REP1 and REP2 with similar affinity, the RAB27A:REP1 complex is a better substrate for RABGGTase148. Although RAB27A plays a key role in the development of choroideremia, it is likely that other un-prenylated RABs also contribute to the disease pathology. A recent study employed an in vitro prenylation assay to label un-prenylated RABs in REP1 knockdown cells with a biotin-isoprenoid analogue149. Their results suggested that REP2 can only partially rescue the effects of REP1 knock-down, and that several proteins remained un-prenylated, although the authors did not identify these. Global analysis to identify all the members of the RAB family that are affected by the loss of REP1 is required to gain a fuller understanding of the disease.

1.3.5. Hepatitis Delta and Hepatitis C Viral Infection

Hepatitis Delta virus (HDV) is a small RNA virus which causes severe viral hepatitis in humans, leading to cirrhosis of the liver. Although HDV only exists as a co-infection with Hepatitis B

34 virus (HBV) it is estimated to affect approximately 20 million people worldwide. There are currently no effective therapies. HDV consists of a RNA genome, large and small delta antigens and a lipid envelope which is embedded with HBV surface antigen proteins. The HBV proteins are required for exit and entry into cells, and thus crucial for the process of infection by HDV. The large delta antigen, which is essential for assembly of HDV virions and virus-like particles, has a C-terminal CAAX motif which is farnesylated by the host cell’s prenylation machinery150. This farnesylation step is required for the large delta antigen to interact with the HBV proteins, and thus presents a possible point of therapeutic intervention. Indeed, both in vitro and in vivo studies have shown that treatment with FTIs inhibits the production of HDV virions151, 152. The results of the first phase 2A clinical trial using the FTI Lonafarnib to target chronic HDV infection was recently published, showing significant reductions in viral load in the FTI treatment groups153. These promising initial results indicate that FTIs may provide a novel route to target HDV.

Hepatitis C virus (HCV) is another RNA virus which is dependent on host cell prenylation. HCV forms a replication complex which requires association with host cell membranes. HCV does not express any proteins with a prenylation motif, but assembly of the replication complex is inhibited by statins and GGTIs154. This effect has been explained by the requirement of HCV protein NS5A to associate with the geranylgeranylated host protein F-box and leucine-rich repeat protein 2 (FBL2)155.

1.3.6. Other Diseases

Protein prenylation is of interest in several other diseases in addition to those discussed in detail here, including tropical parasitic infections156, 157. Protozoan parasites such as Plasmodium falciparum158, Trypanosoma brucei159 and Leishmania species160, the causative agents of malaria, African sleeping sickness and Leishmaniasis respectively, all harbour prenyl transferase enzymes. Several studies have shown that FTIs can act as potent antimalarial agents, and that the parasite transferase enzyme is more sensitive to inhibition than the human counterpart157, 161, 162. However, resistance to FTIs have been reported in P. falciparum suggesting that the scope of targeting malaria by this mechanism may be limited163, although FTIs could find use in combination therapies.

The role of protein prenylation in neurodegenerative diseases is a further area of current research164, 165. Prenylated small GTPases such as RAS and RHO play a key role in regulating synaptic plasticity164 which is important for cognitive functions such as learning and memory. Impairments in synaptic plasticity contribute to neurodegenerative diseases such as Alzheimer’s disease166. A key pathological feature of Alzheimer’s disease is the deposition of amyloid-β peptide plaques in the brain of patients. Several studies have shown that statins

35 reduce the production of amyloid-β, and this effect has been attributed to reduction in protein prenylation rather than effects on cholesterol levels167. In line with these results it was found that geranylgeranyl pyrophosphate supplementation leads to increased levels of amyloid-β168. Furthermore, elevated cellular levels of FPP and GGPP with a concomitant increase in FPPS protein expression has been observed in Alzheimer’s patients169. A recent study indicated that FTase or GGTase-1 haplodeficiency in a mouse model of Alzheimer’s disease resulted in lower amyloid-β deposition in the brain and attenuated neuro-inflammation170. In addition, FTase, but not GGTase-1, haplodeficiency lead to increased clearance of amyloid-β. Interestingly, improved cognitive functions such as memory and learning were only observed in the FTase haplodeficient model170.

1.4. Chemical Proteomics

The study of protein lipidation is challenging due to the lack of inherent handles to isolate and detect proteins bearing these modifications. Furthermore, the often bulky and hydrophobic nature of the lipid groups present challenges in detecting these PTMs via mass spectrometry (MS). Traditional biochemical methods such as labelling with radioactive lipids or lipid precursors are, in addition to the inherent hazards of handling radioactive reagents, ridden with issues of poor sensitivity. The utility of radiolabelling is limited as it does not enable unbiased global profiling across the whole proteome. In recent years the use of chemical reporters has become well-established as a powerful method to circumvent some of these issues. The use of chemical probes to study protein lipidation has been extensively reviewed as part of this project171, 172, as well as by other groups173-177.

Lipid chemical probes rely on the derivatisation of the native lipid molecule with a small chemical ‘tag’. The relatively large size of the lipid modifications enables introduction of the tag without severely affecting substrate recognition by the cognate lipid transferase enzymes. The tags consist of small chemical groups such as an azide or alkyne, which consist of three nitrogen or two carbon atoms respectively. In contrast, direct introduction of a reporter group such as a fluorophore or affinity handle onto the lipid molecule would severely disrupt the molecular structure and in most cases prohibit its use by the native lipidation machinery.

Azides and alkynes have no intrinsic reactivity in the biological environment but can be activated in highly specific chemical reactions during downstream analysis to attach reporters such as fluorophores for rapid visualisation or affinity tags for enrichment purposes. The secondary reporter molecule can be attached using a number of bio-orthogonal chemical transformations including copper(I)-catalysed Huisgen 1,3-dipolar cycloaddition between and azide and alkyne (CuAAC)178, 179, strain-promoted cycloaddition180, Staudinger ligation

36 between an azide and triarylphosphine 181, 182 or a reverse electron-demand Diels-Alder reaction183, 184. This project employed CuAAC (Figure 10) as the conjugation method of choice due to the reliable nature of this reaction and the availability of suitable reagents and optimised protocols. Furthermore, the alkyne and azide moiety required on the chemical probe can be installed with relative synthetic ease. A disadvantage of CuAAC is the requirement of copper, which is highly toxic to biological systems, and thus the reaction can only be performed in cell lysates and/or fixed cells.

Figure 10. Mechanism of Cu(I)-catalysed Huisgen 1,3-dipolar cycloaddition between an azide and terminal alkyne. In recent years, chemical probes have been employed to study a variety of lipid modifications, including myristoylation185-188, palmitoylation177, 189-192, cholesteroylation193, 194 and prenylation (further discussed in section 1.4.1 below). Coupled with advances in liquid chromatography- mass spectrometry (LC-MS) techniques that allow analysis of highly complex biological samples, the use of these chemical reporters has enabled the rapid identification (and quantification) of hundreds of lipidated proteins within one experiment.

1.4.1. Prenylation Probes

Extensive work has been conducted on the development of chemical reporters to study protein prenylation. Kho et al. were the first to report the use of a chemical probe, an azide-tagged farnesyl analogue (AzC15, Table 1A), for metabolic labelling of farnesylated proteins in cell culture195. Since then, numerous chemical probes with alkyne or azide tags or biotin handles suited to profiling protein prenylation both in vitro149, 196-203 and in cell culture204-209 have been reported (Table 1). In addition, a number of fluorescent isoprenoid analogues have been developed for use in biochemical assays and imaging applications. These will not be covered in this account, but have recently been reviewed elsewhere210.

Table 1. Isoprenoid probes employed in previous metabolic labelling or in vitro re-prenylation studies. Entry Structure Key reported applications AzC15 & AzC20  Metabolic labelling in Jurkat, COS-7 and MCF-7 cell lines204  Metabolic labelling in HeLa, 3T3, Jurkat, DC2.4 and RAW264.7 cells206.

AzC10PP & AzC15PP A  Enzyme assay with yeast FTase196, 197 and E. Coli FTase199.

AzC20PP  Re-prenylation of statin treated HEK293A cell lysate with GGTase1 and RABGGTase. Re-prenylation of mouse tissue from REP1 knockout mice201.

 Metabolic labelling in cultured HeLa, 3T3, Jurkat, DC2.4, RAW264.7 and HEK293T cells206, 207.  Metabolic labelling in HeLa cells expressing Legionella pneumophila effector proteins205. B  Combined with 2D-gel analysis to quantify prenylation in response to FTIs209.  Combined with flow cytometry to quantify cellular levels of prenylated proteins208

B-GPP & B-FPP  Enzyme assay with rat FTase198, 203.

B-GPP C  Enzyme assay with recombinant FTase, GGTase-1 and RABGGTase and mutant FTase and GGTase-1200.  In vitro re-prenylation of statin-treated cell lysate with recombinant RABGGTase and subsequent identification of RAB proteins by LC-MS/MS200.  In vitro re-prenylation of choroideremia patient cells149.

D  Metabolic labelling in Jurkat cells206

A number of approaches are utilised to install the prenyl probes on target, as outlined in Figure 11. The simplest system involves prenylation of a substrate peptide or protein in vitro by recombinant prenyl transferases (Figure 11A). This approach requires the use of isoprenoid pyrophosphate probes, which are challenging to synthesise. Prenylation is monitored by various techniques; the most common method is to use a peptide substrate conjugated to environmentally sensitive fluorophores such as Dansyl or nitrobenzoxadiazole (NBD). Attachment of the prenyl group to the peptide increases the hydrophobic environment of the fluorophore, which in turn enhances its fluorescence, enabling a direct read-out of the prenyl transfer reaction.

Figure 11. Methods to install tagged isoprenoids on prenylated proteins. (A) Tagged isoprenoid pyrophosphates can be enzymatically incorporated in vitro onto proteins or peptides bearing a prenylation motif by recombinant prenyl transferases. (B) Tagged isoprenoid alcohols are taken up by cells in culture and converted to pyrophosphates through the isoprenoid rescue pathway and subsequently incorporated onto prenylated proteins by the endogenous prenylation machinery. (C) Endogenous prenylation is disrupted (typically by statin treatment) in metabolically active cells. Following lysis the pool of un-prenylated proteins are subjected to in vitro prenylation with the tagged isoprenoid pyrophosphate and recombinant prenyl transferase enzymes. (Note: the prenylation motif is generalised as CAAX, but the methods are equally applicable to RAB proteins with varying C-terminal prenylation motifs.) For deciphering the targets of prenylation in a cellular context, the Alexandrov lab pioneered the use of an in vitro re-prenylation assay200 (Figure 11C). The concept relies on disrupting cellular prenylation in the model system of interest, for example by treating cells in culture with a statin. Following extraction of the proteome, the proteins are subjected to prenylation in vitro with recombinant prenyl transferase enzymes and isoprenoid probes. This technique is particularly powerful when using model systems that are not amenable to metabolic labelling, for example

39 patient samples or animal models. A distinct disadvantage, however, is that any spatial or temporal control of the prenylation reaction conferred by the native biological environment is lost, possibly leading to non-physiological patterns of prenylation. Furthermore, the analysis relies on identifying proteins that were not prenylated, and thus information about prenylation in the native environment is only inferred.

Arguably the most relevant experimental workflow for monitoring prenylation is in the native cellular environment (Figure 11B). Cultured cells incorporate isoprenoid chemical reporters via their native prenylation machinery. Probes are generally administered to the cells in their free alcohol form, as the negatively charged pyrophosphates are unlikely to penetrate the cell membrane. Cells harbour an isoprenoid rescue pathway distinct from the mevalonate pathway that enables conversion of isoprenoid alcohols to their cognate pyrophosphates. This rescue pathway consists of a number of poorly characterised kinases which sequentially transfers a phosphate group to the isoprenoid alcohol and the resultant isoprenoid monophosphate to generate the isoprenoid pyrophosphate211-214. Certain cell types appear to be more efficient at incorporating isoprenoid alcohol probes, and this can probably be attributed to differential activity of this rescue pathway. As the name infers, metabolic labelling is only suitable to probe metabolically active biological systems, and is thus not applicable to the analysis of patient samples. The use of isoprenoid probes to label proteins in complex animal models such as mice is likely to be extremely challenging as the lipophilicity of the probes will limit bioavailability.

The most widely used isoprenoid probe for metabolic labelling applications is a propargyl ether farnesol derivative (PeC15, Table 1B), which has been applied in numerous studies by the Hang and Distefano labs. This probe acts as a dual substrate for both FTase and GGTase-1215. The Hang lab employed this probe for LC-MS profiling of prenylated proteins in RAW264.7 macrophages, which led to the discovery of the novel farnesylated protein zinc finger CCCH- type antiviral protein 1 (ZC3HAV1)207. In further studies they also applied PeC15 to identify prenylated Legionella pneumophila effector proteins expressed in HeLa cells205. The Distefano lab has developed a number of analysis workflows incorporating the PeC15 probe. For example, the probe was applied in combination with 2D-gel analysis to quantify changes in prenylation in response to prenyl transferase inhibitors209. Most recently, they reported a method to assess levels of prenylated proteins in cells by flow cytometry208. With this approach they showed that cellular levels of prenylated proteins were increased in model systems of defective autophagy by knockdown of autophagy-related genes BECN1 and ATG7.

A distinct disadvantage of the PeC15 probe is its lack of selectivity for farnesylated versus geranylgeranylated proteins. Proteomic analysis of the prenylated targets indicated that the probe is used by all three prenyl transferases207. The addition of the propargyl ether group

40 makes the chain equal in length to a geranylgeranyl moiety. Thus, it is likely that this probe may not be optimal for labelling farnesylated proteins. In support of this, in vitro kinetic studies indicated that the analogous probe based on geranyl pyrophosphate (PeC10PP), which is one isoprene unit shorter, was a much better substrate for yeast FTase199.

The Alexandrov and Waldmann groups have developed a biotin-geranylpyrophosphate (B-GPP, Table 1C) probe that can act as a substrate for RABGGTase. As discussed in section 1.2.2, the isoprenoid binding site of RABGGTase is less selective than the corresponding site in FTase and GGTase-1, and as such is more amenable to accepting derivatised isoprenoids. In a comprehensive study by Nguyen et al.200 the B-GPP probe was used in a re-prenylation assay with recombinant RABGGTase and REP1/2 to label RAB proteins. Subsequent enrichment and LC-MS analysis enabled the identification of 42 RAB proteins, which represents a substantial percentage of all cellular RABs. Although B-GPP was not a substrate for wild-type FTase and GGTase-1, mutation of selected amino residues within the isoprenoid binding site of these enzymes created mutant enzymes that could competently employ B-GPP as a substrate.

The B-GPP probe was utilised in a study of RAB mis-prenylation in choroideremia patient cells149. As discussed earlier, choroideremia is caused by accumulation of unprenylated RAB proteins due to defective REP1. Lymphoblasts from choroideremia patients were subjected to in vitro re-prenylation with B-GPP. In contrast to control cells, these lymphoblasts contained a pool of un-prenylated RABs. In further analysis using a block-and-release assay they could establish that different RABs were prenylated at different rates, presumably as a function of their affinity for the RABGGTase prenylation machinery.

As discussed, a particular strength of the re-prenylation assay is the ability to analyse samples from patients and animal models. Berry et al. employed an azido-tagged geranylgeranyl pyrophosphate analogue (AzC20PP, Table 1A) to develop an assay for labelling geranylgeranylated proteins in vitro201. The assay was applied to tissue lysate from a choroideremia mouse model to label un-prenylated RAB proteins. The alcohol version of the same azido-geranylgeranyl probe (AzC20) was shown to be incorporated metabolically into proteins in cultured cells204.

A recent study by the Distefano lab surveyed a number of prenyl probes in combination with different CAAX-peptides and found that the structure of the probe directly influenced the peptide substrate selectivity of the prenyl transferase216. This result is not surprising considering that the isoprenoid moiety forms extensive interactions with the peptide substrate in the enzyme binding sites of FTase and GGTase-1 (as discussed in section 1.2.1). This study highlights the care needed in designing prenyl probes, and suggests the need for a complement of prenyl probes to access the whole population of prenylated proteins.

In addition to the probes presented in Table 1, a large number of studies have explored the ability of the prenyl transferases to accept a variety of structurally modified isoprenoids. A selection of isoprenoid analogues that contain functional groups for bio-orthogonal ligation reactions (azide, alkyne or di-ene) or a biotin affinity handle are shown in Figure 12.

Figure 12. Previously reported isoprenoid analogues tested in vitro with recombinant prenyltransferase enzymes. The reported analogues harbour a variety of functional groups for bio-orthogonal conjugation reactions (azide or alkyne for CuAAC, di-ene for Diels-Alder reaction) or a biotin affinity handle. The analogues have been confirmed as FTase, GGTase-1 and/or RABGGTase substrates in vitro198, 199, 202, 203, 216. Despite the plethora of probes developed for probing protein prenylation, a number of issues persist. As such, there is scope for the development of novel probes and workflows to overcome some of these issues. This will be discussed in closer detail in Chapter 2.

1.4.2. Mass Spectrometry for Proteomics

Mass spectrometry (MS) is a powerful tool to gain unbiased information about protein expression, interactions and PTMs in different biological environments217. In contrast to traditional biochemical techniques such as Western blot analysis, MS enables analysis of the proteome on a global scale within a single experiment. Two general approaches are applied in the identification of proteins by MS: top-down or bottom-up analysis. In top-down proteomics218 whole proteins are purified (e.g. by excising bands from a SDS-PAGE gel) and subjected to MS analysis. Identification of the protein is based on analysis of the protein fragments generated

42 during the MS analysis. In bottom-up proteomics219 the protein sample is digested prior to analysis by liquid chromatography tandem MS (LC-MS/MS), typically by trypsin. Some type of separation or enrichment may be performed before or after the digestion event to reduce the complexity of the sample. In so-called “shot-gun proteomics” no separation is employed. Protein identification is subsequently performed by matching the identified peptides to a database of known protein sequences.

Technological advances in MS technology have greatly spurred the field of proteomics217. MS- based proteomics requires the use of high-resolution instrumentation to achieve sufficient mass resolution and accuracy. The MS instrument generates information about mass-to-charge ratio (m/z) of the analysed molecule, from which the molecular mass can be determined. Commonly used MS instruments for proteomics analysis include Time-of-Flight (TOF) and Orbitrap analysers. In a typical analysis set-up the protein or peptide sample of interest is first separated by LC. Upon elution from the LC column the sample is subjected to a very high voltage which induces formation of an aerosol of charged droplets that subsequently enter the MS detector. MS analysis relies on magnetic fields, and thus the ionisation step is crucial for detection by the MS instrument. In proteomics, electrospray ionisation (ESI) is most commonly used220. ESI is a soft ionisation technique that does not induce fragmentation, meaning the mass of the intact biomolecule can be assessed. In TOF instruments the m/z is related to the speed at which the ions arrive at the detector221. In Orbitrap instruments the ions oscillate in an ion trap; the frequency of the oscillations are related to the m/z of the molecule222.

In tandem MS/MS mode precursor ions of interest are selected by the MS instrument and subjected to further fragmentation. The associated spectra (termed MS2) generated contain the information necessary to decipher the amino acid sequence of the injected peptides or protein fragments. Protein identification is performed in dedicated software suites by matching the identified peptide masses to theoretical masses generated by in silico digest of database proteomes such as those published by UniProt (www.uniprot.org).

Proteomics analysis has been widely applied to studying PTMs, particularly for the identification of phosphorylation sites223, 224. However, the study of lipid PTMs by MS is challenging for a number of reasons. Firstly, due to the low levels of lipidated proteins they may escape detection in a background of more abundant protein species. Secondly, the lipidated peptides typically suffer from poor ionisation and fragmentation. As ionisation is required for detection by the MS instrument, molecular species with poor ionisation typically give rise to very low signal which may be undetectable above background. Fragmentation is necessary to enable confident identification of the modifying lipid. Thirdly, the hydrophobic nature of the lipidated peptides may lead to challenges in achieving efficient LC separation. For proteomics analysis, peptide

43 separation is typically performed by reverse-phase LC employing a gradient of acetonitrile in water as the eluent. Unless care is taken to optimise the LC gradient to account for the hydrophobicity of the lipidated peptides these may be poorly resolved, or in the worst case scenario, may even be retained on the reverse-phase column.

A number of studies report direct identification of prenylation by MS methods. Kassai and co-workers used a top-down approach to identify farnesylation of the Transducin γ-subunit in mouse retina. Although this study shows that prenylation events can be detected by MS, the top-down approach limits through-put and does not enable unbiased analysis of farnesylation across the proteome. A further study analysed the fragmentation pattern of the cysteine-farnesyl moiety, identifying fragment ions that may be used as a marker for farnesylated peptides225. Wotske et al. applied multidimensional LC to the analysis of farnesylated RAB proteins in a background of HeLa cell lysate226. Although they could confidently identify the modified C- terminal peptide of the three RAB proteins studied, no attempts were made to identify farnesylated peptides from a more complex sample, suggesting that their method may have limited scope in global profiling of prenylation. In a recent report, researchers used chemical oxidation of prenylated peptides as a means to enhance detection227. Oxidation by mCPBA introduced a mono-oxidised thio-ether bond as well as epoxidation of the isoprene double bonds. The oxidised thio-ether bond is more labile than the corresponding thio-ether, leading to release of a characteristic fragment ion from modified peptides. The oxidation of the double bonds increases hydrophilicity of the prenylated peptides, aiding LC separation. Although the researchers showed the applicability of this method in a peptide sample spiked with BSA, the applicability in a complex sample such as a cell lysate remains to be proven. Furthermore, the oxidation step produces isoprenoid chains with varying number of epoxidised double bonds. Although the authors highlighted this feature as an advantage in assignment of the fragment ions, this complexity may present problems in the analysis of more complex samples.

1.4.3. Quantitative Proteomics

A powerful capability in proteomics is the ability to quantify protein levels, for example to assess differential expression in diseased states, or changes in protein abundance as a result of external stimuli such as a drug treatment. Although absolute quantification (Figure 13A) is cumbersome and requires the generation of a standardized sample for each protein of interest, achieving relative quantification between different samples on a whole proteome level has been enabled by numerous techniques developed in recent years. Relative quantification can be done by using label free methods (Figure 13B) by comparing ion intensity or spectral counts228. Alternatively, several labelling technologies have been developed to achieve more accurate quantification results229. A summary of key quantitative workflows are summarised in Figure 13.

Figure 13. Summary of quantification methods in proteomics. Proteomics enables both absolute (A) and relative (B-F) quantification. Absolute quantification (A) requires addition of protein standard containing a known amount the protein(s) of interest. Relative quantification can be achieved though label-free methods (B) or incorporation of mass tags through metabolic (C&D) or chemical (E&F) labelling methods.

Metabolic Labelling by SILAC

Stable isotope labelling with amino acids in cell culture (SILAC) was pioneered by the group of Mathias Mann. SILAC labelling relies on the metabolic incorporation of isotopically labelled amino acids by cellular proteins230 (Figure 13C). This is commonly done in cultured cells, but can also be used in vivo to label mouse models231, 232. A limitation of this approach is the need for metabolic activity, which means it is not applicable to patient-derived samples. Furthermore, it may not be applicable to more sensitive cells such as primary cells as the dialysed serum required may be void of necessary growth promoting components. Also, to ensure complete incorporation of the isotopic labels, cells need to be grown for a number of passages prior to use in experiment, which may not be possible with certain cell types.

The isotopic labels employed are typically Lys or Arg amino acids with combinations of 2H, 13C and 15N labels. These labels are particularly compatible with protein digest by trypsin, which cleaves at the C-terminal side of Arg and Lys residues. Thus, this should ensure that all peptides contain one (or more depending on the presence of missed cleavages) labelled amino acids. Generally, two or three experimental conditions are compared in one experiment by employing a combination of “Light”, “Medium” and “Heavy”-labelled cells that are subjected to different treatments (e.g. drug treatment versus no treatment). Following cell lysis the protein samples are combined in known ratios, digested and subjected to LC-MS/MS analysis. In a duplex- SILAC experiment employing “Light” and “Heavy” cells, each peptide will generate two peaks in the MS1 spectrum separated by a mass shift induced by the isotopic label; a triplex-SILAC experiment will generate three peaks per peptide. The relative intensity of each peak is directly related to the relative abundance in each sample.

A variant of SILAC, termed spike-in SILAC or Super-SILAC has found widespread appeal in recent years233-235 (Figure 13D). In this workflow, a SILAC-labelled spike-in standard is generated and subsequently added to each experimental sample before digest. This method removes the constraint of needing to grow the experimental samples in special SILAC-medium, and enables the method to be used with tissue samples. The spike-in acts as an internal standard that enables relative quantification between all samples. This method is particularly appealing as an unlimited number of experimental conditions can be compared, in contrast to a maximum of three samples in traditional SILAC experiments. Optimisation of the SILAC spike- in may be required to ensure that it mirrors the protein expression profile of the experimental sample as quantification of a particular protein requires its presence in both the spike-in and the sample under investigation. Furthermore, for accurate quantification the proteins require similar levels of expression. This may for example be achieved by mixing a number of spike-in samples generated in different cell lines. The utility of this quantitative method in a chemical proteomics

46 workflow was recently explored by Thinon et al.186. In this study the spike-in SILAC method was employed in combination with a lipid-alkyne probe to quantify changes in myristoylation in response to lipid transferase inhibition.

Chemical Labelling Methods

In chemical labelling methods such as Isobaric Tag for Relative and Absolute Quantitation 236, 237 (iTRAQ), Tandem Mass Tags238 (TMT) and Dimethyl labelling239-241 (DML) a mass tag is appended to the peptide sample after digest (Figure 13E & F). Both iTRAQ and TMT employ isobaric tags that react with amines, and are thus added to each peptide N-terminus and any free lysine side-chains. The isobaric tags consist of three components: a chemically reactive group, a reporter group and a balance group. Although the reporter and balance groups are varied in different tags, their overall mass is equal and as a result the complexity of the MS1 spectrum is not increased. Further fragmentation of selected precursor ions leads to release of the reporter group, which can be identified and quantified in the MS2 spectrum. Optimisation of the isobaric mass tags has enabled 8-plex comparisons with iTRAQ labelling242 and 10-plex comparisons with TMT labelling243. A distinct disadvantage of the isobaric mass tags is the expense of the reagents. DML labelling has emerged as a cost-effective alternative in recent years. DML labelling also adds mass labels to the amine of N-terminal amino groups and lysine sidechains, but uses cheap and readily available reagents. Similarly to SILAC, quantification is performed in MS1 as the added mass tags generate peptide precursors with a distinct mass shift. Chemical labelling methods are particularly suited to quantification in samples that cannot be metabolically labelled, such as tissue samples or bodily fluids. However, as the addition of the mass label occurs downstream in the sample handling process compared to SILAC-labelled samples, these methods suffer from poorer reproducibility244.

1.5. Project Objectives

The over-riding aim of this PhD project was to develop a chemical proteomics platform to enable identification and quantification of prenylated proteins. As discussed, protein prenylation plays a key role in key cellular processes and has been implicated to play a role in the pathogenesis of numerous diseases. Despite extensive efforts over the past 30 years to target prenylation as treatment for cancer, including in excess of 70 clinical trials, this approach has not been translated into a successful therapeutic. Recent evidence indicates that prenyl transferase inhibitors may provide a novel treatment for diseases such as HGPS and HDV infection. To explain past failures and to direct future research in the area, it is imperative that we identify all of the cellular proteins that undergo prenylation. Furthermore, a better understanding of the significant cross-talk and dynamics between the prenyl transferases in response to inhibitors of the specific enzymes is required.

Chemical proteomics and advances in high resolution MS have transformed our ability to detect and quantify post-translational modifications. Although previous studies have reported probes for profiling prenylation several short-comings exist including poor selectivity for farnesylated versus geranylgeranylated proteins. As such, I sought to develop a novel workflow to identify and quantify prenylated proteins to gain a fuller understanding of prenylation in response to prenyl transferase inhibitors and in various disease states.

I hypothesised that isoprenoid analogues with an azide or alkyne functionality could serve as substrates for the prenyl transferase enzymes in metabolically active cells, enabling the incorporation of a chemical tag onto prenylated proteins. The chemical tag would facilitate attachment of reporter groups such as fluorophores or affinity tags to allow downstream analysis by gel-based methods and mass spectrometry to visualise and identify the prenylated proteome.

Specifically, in this PhD project I aimed to:

 Develop novel isoprenoid probes to enable selective labelling of farnesylated versus geranylgeranylated proteins by metabolic labelling in living cells. (Chapter 2)  Apply the probes in combination with high resolution MS technologies for global profiling of prenylated proteins. (Chapters 3 & 4)  Develop a quantitative MS platform to enable relative quantification of prenylated proteins. (Chapters 3 & 4)  Profile a variety of prenyl transferase inhibitors to assess their effect on the prenylation of individual protein targets and to explore any associated switches in prenylation status. (Chapters 4)

 Explore prenylation in a number of disease models including HGPS and hypoxia- induced PH. (Chapter 5)  Assess the utility of the probes in in vivo models. (Chapter 5)

Chapter 2: Novel Probes for Profiling Protein Prenylation

2.1. Introduction

In recent years a plethora of prenyl probes to study protein prenylation have been reported in the literature172, 245. These have been utilised in combination with various biochemical, imaging and mass spectrometric methods to investigate prenylation in enzyme assays and in cell culture. Despite these efforts, two main challenges have not been overcome. Firstly, currently reported probes show poor selectivity for farnesylated versus geranylgeranylated proteins. In fact, the most widely used probe, a propargyl ether derivative of farnesol developed by the Hang lab (PeC15, Figure 14), is used as a dual specificity probe for both farnesyl and geranylgeranyl transferase enzymes206. Secondly, current in-cell labelling technologies rely on statin treatment to deplete endogenous isoprenoids in order to facilitate sufficient incorporation of the unnatural isoprenoid substrate. Statin treatment is also used in traditional radiolabelling experiments to gain incorporation of exogenous radiolabelled isoprenoids246. However, statins have several cellular effects, including altered expression of numerous prenylated proteins35. Furthermore, it is likely that depletion of natural isoprenoids in combination with introduction of exogenous isoprenoids will result in altered prenylation patterns, akin to the effects seen upon treatment with prenyl transferase inhibitors. Thus, the question arises whether studies of prenylation in a system disrupted by statins are in fact physiologically relevant.

This chapter describes the development of novel prenyl probes to selectively label farnesylated and geranylgeranylated proteins. We argued that the differences in the binding pocket of the farnesyl and geranylgeranyl transferase enzymes warranted the development of two complimentary probes mimicking the natural substrates as closely as possible. The importance of structural similarity to the natural isoprenoids was indeed highlighted in a recent study which showed that isoprenoid analogues can have drastic effects on the selectivity of the prenyl transferase enzyme in respect to their preferred peptide substrate216. This result is not surprising, as the isoprenoid makes extensive contacts with the peptide substrate in the prenyl transferase active site.

A library of previously reported and novel prenyl probes was designed and synthesised (Figure 14). All probes were synthesised as isoprenoid alcohols rather than the pyrophosphate form required by the prenyl transferase enzymes. For studies in living cells, the alcohol form is preferred as the charged pyrophosphate is unlikely to permeate the cell membrane. Mammalian cells harbour a rescue pathway that converts farnesol and geranylgeraniol into the cognate pyrophosphate211, 212, 214. Previous labelling studies have demonstrated no difference in the labelling achieved with isoprenoid alcohols versus pyrophosphates195. In fact, it is likely that the

50 pyrophosphate is hydrolysed to the alcohol before uptake into the cell. Therefore, as our aim was to develop probes for studies in living cells it was deemed unnecessary to embark on the challenging synthesis of the pyrophosphate derivatives.

Figure 14. Library of prenyl probes. Structure of the propargyl ether (Pe), azido (Az) and alkyne (Yn)-containing isoprenoid analogues synthesised and/or tested and the natural isoprenoid alcohols prenol, geraniol, farnesol and geranylgeraniol. Probes denoted by (RS) were initially provided by Dr Remigiusz Serwa. Probes denoted by (*) have been previously reported, either in alcohol or pyrophosphate form. 2.2. Synthesis of an Initial Panel of Azido and Propargyl Ether Prenyl Probes

An initial selection of probes was synthesised based on previously reported azido and propargyl ether probes195-197, 199, 201, 204, 206. The library included structures based on C5, C10, C15 and C20 isoprenoids to investigate the effect of isoprenoid length on labelling efficiency, and to explore whether the shorter chain probes were able to enter into the biosynthetic pathway of the longer chain isoprenoids. The C5 prenyl and C20 geranylgeranyl probes AzC5, AzC20, PeC5 and PeC20 were synthesised as part of this project, whereas AzC10, AzC15, PeC10 and PeC15 were kindly provided by Dr Remigiusz Serwa.

The synthesis of the propargyl ether and azido probes was achieved using established synthetic methodologies196, 206 as outlined in Scheme 1. The commercially available isoprenoid alcohol precursors were protected as tetrahydropyranyl (THP) ethers in a facile overnight reaction. Subsequently, the allylic oxidation of the terminal methyl group was performed using selenium dioxide. Two different protocols were trialled for this reaction. Oxidation by Sharpless’ method247 with a catalytic amount of selenium dioxide and tert-butyl hydroperoxide gave the most reproducible results and required a smaller amount of the highly toxic selenium dioxide. Overall yields were however poor (14-31%), even after prolonged reaction times. Addition of further tert- butyl hydroperoxide did not improve yields. In the interest of time, no further attempt was made to optimise this reaction, instead the reaction was scaled to produce sufficient amounts of this key intermediate.

Scheme 1. Synthetic route to propargyl ether and azido probes. a) 3,4-dihydro-2H-pyran, pyridinium p-toluenesulfonate, DCM, 72% (1a), 83% (1b); b) i) SeO2, pyridine, EtOH, 80°C, ii) NaBH4, EtOH, 31%; c) SeO2, t-BuOOH, DCM, 0°C, 14%; d) NaH, 18-crown-6, propargyl bromide, THF, 0 – 50°C; e) (PhO)2PON3, DBU, toluene, 0°C, 38% (4a), 59% (4b); f) pyridinium p- toluenesulfonate, EtOH, 60°C, 73% (PeC5, 2 steps), 49% (PeC20, 2 steps), 49% (AzC5), 90% (AzC20) The allylic alcohol precursors 2a and 2b were converted to propargyl ethers by reaction with propargyl bromide under basic conditions206 or to azides by means of the azide-transfer reagent diphenylphosphoryl azide196 as outlined in Scheme 1. Subsequent cleavage of the THP group produced the final probes in moderate overall yields.

2.3. Initial Labelling Studies with Azido and Propargyl Ether Probes

The labelling efficiency of the synthesised panel of propargyl ether and azido probes was assessed by in-gel fluorescence as outlined in the workflow in Figure 15. Labelling experiments were performed in immortalised human umbilical vein endothelial cells (cell line EA.hy926)248, a cell line which rapidly proliferates in cheap and readily available growth medium. Importantly,

52 this cell line showed excellent incorporation of the prenyl probes, suggesting that these cells harbour an efficient prenyl rescue pathway able to convert the isoprenoid alcohols into pyrophosphates211, 212.

Figure 15. Schematic of in-cell labelling workflow. Cells are cultured in medium supplemented with prenyl alcohol probe (YnF shown as an example). The alcohol is converted into the pyrophosphate by an isoprenoid rescue pathway and incorporated onto prenylated proteins by cellular prenyl transferase enzymes. The cell lysate is subjected to CuAAC to append a capture reagent onto tagged proteins via a Huisgen 1,3-dipolar cycloaddition reaction between the alkyne of the prenyl probe and the azide of the capture reagent. In cases where an azido prenyl probe is fed to the cells a complementary alkyne capture reagent is utilised. Labelled proteins can be visualised by in-gel fluorescence after SDS-PAGE separation. The biotin affinity handle enables purification of labelled proteins for subsequent identification by immunoblotting or mass spectrometry. For initial labelling studies the EA.hy926 cells were cultured in medium supplemented with the azido or propargyl ether prenyl probes at a final concentration of 50 µM. Control cells were grown in the presence of DMSO vehicle only. Cells were lysed and subjected to CuAAC with an azido- or alkynyl capture reagent bearing a TAMRA fluorophore and biotin affinity handle (AzTB or YnTB, see Figure 16)193.

Figure 16. Structures of capture reagents AzTB and YnTB. To remove excess capture reagent the proteins were precipitated and subsequently separated by sodium-dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). In-gel fluorescence scanning of the gel enabled visualisation of the TAMRA-labelled proteins (Figure 17). The strongest labelling was achieved with AzC20, an azido-tagged analogue of geranylgeranyl alcohol. Very strong labelling was evident in the 20-25 kDa molecular weight region, consistent with labelling of small GTPases such as RHO and RAB proteins. There are approximately 70 RAB proteins in the human proteome, many of which incorporate two geranylgeranyl chains, accounting for the very strong signal achieved in this region. In previous reports, AzC20 shows a similar labelling pattern when fed to Jurkat, COS-7 and MCF-7 cell lines204. The pyrophosphate derivative of AzC20 has also previously been used to profile geranylgeranylated RAB proteins201. The analogous propargyl ether probe PeC20 also showed efficient labelling in the 20-25 kDa region. Both probes resulted in weak signal in the 10 kDa region, whereas AzC20 labelling also resulted in several higher molecular weight bands not present in the PeC20 sample.

Figure 17. Labelling by initial panel of azido and propargyl ether probes. EA.hy926 cells were cultured in medium supplemented with prenyl probes (50 μM) or vehicle (DMSO) for 16 hours. The protein lysate was subjected to CuAAC with AzTB or YnTB, separated by SDS-PAGE and visualised by in-gel fluorescence (top gel image). The gel was stained in Coomassie Blue to assess protein loading (bottom gel image).

AzC10, AzC15, PeC10 and PeC15 also labelled proteins in the 20-25 kDa region but the signal was much weaker than for the AzC20 and PeC20 probes. These probes are likely to predominantly label the pool of farnesylated GTPases such as RAS. In contrast to the very large number of RABs in the human proteome, there are only a handful of RAS isoforms and as such the labelling intensity in this region was as expected much weaker than for the geranylgeranyl probes. The labelling pattern of AzC15 and PeC15 are consistent with that previously observed in Jurkat, COS-1, COS-7 and MCF-7 cells195, 204, 206.

Both the azido and propargyl ether C10 and C15 probes labelled two bands in the 50 kDa region. It is unclear what proteins these bands represent. Nuclear lamins (prelamin-A/C, lamin B1 and lamin B2) are abundant proteins that are known to be farnesylated; however, these proteins have a molecular weight of 66-79 kDa and are thus unlikely to represent the observed bands.

Several bands in the 37 kDa region were labelled by all azido probes. However, the presence of these bands in the DMSO control sample suggests that this is due to background labelling by YnTB. This is consistent with the observation that alkyne capture reagents in general produce higher background labelling than azido reagents249. YnTB is more hydrophobic than AzTB, and may thus be more likely to form non-specific interactions with abundant hydrophobic proteins, rendering it difficult to remove during the precipitation step.

Overall, the azido probes appeared to label a larger set of proteins than the propargyl ether probes as judged by in-gel fluorescence. However, in addition to the higher background labelling observed with the azido probes, these probes suffered from several unfavourable features. Firstly, allylic azides undergo a 3,3-sigmatropic rearrangement and as a result the azido probes exist as two different isomers in dynamic equilibrium250. Figure 18 depicts the mechanism of the rearrangement using AzC15 as an example. Characterisation of the probes by nuclear magnetic resonance (NMR) clearly revealed the presence of both isomers of the synthesised azido probes. It is unclear whether both isomers can act as substrates for the prenyl transferase enzymes.

Figure 18. 1,3-Sigmatropic rearrangement of allylic azide250. Secondly, the azido probes exhibited poor long-term stability, even when stored at -20°C. Poor labelling was achieved when using aliquots of probe that had been stored for a number of months (data not shown). Regularly reiterating the synthesis of the probes throughout the

55 project would have proven highly time-consuming and expensive. Thus, I focused my work on generating more probes to overcome these issues.

2.4. Effect of Statin Treatment on Labelling Efficiency of Propargyl Ether Probes

Statins inhibit HMG-CoA reductase, the rate-limiting enzyme of the mevalonate pathway. As isoprenoids are intermediates of this biosynthetic pathway, they are depleted upon statin treatment. As discussed earlier, most metabolic labelling studies to date rely on pre-treatment with statins to enhance incorporation of prenyl probes or radiolabelled isoprenoids. We wanted to assess the effect of statin treatment on the labelling of prenylated proteins in EA.hy926 cells.

Cells were cultured in medium supplemented with Mevastatin overnight prior to addition of the propargyl ether probes. The resulting cell lysate was subjected to CuAAC with capture reagent AzTB and analysed by in-gel fluorescence (Figure 19). In agreement with previous studies in Jurkat cells206, statin pre-treatment greatly enhanced the labelling efficiency of PeC15. The same slight enhancement effect was not observed for the geranylgeranyl probe PeC20. Previous labelling studies with geranylgeranyl-azide probe AzC20 in COS-7 cells also showed that labelling was not improved by statin pre-treatment204. Thus, it appears that statins increase the incorporation of unnatural farnesyl, but not geranylgeranyl, analogues.

Figure 19. The effect of statin pre-treatment on labelling by propargyl ether probes. EA.hy926 cells were pre-treated with Mevastatin (10 μM, 19 hours) followed by incubation with prenyl probes (50 μM) or vehicle (DMSO) for 5 hours. The protein lysate was subjected to click reaction with AzTB or YnTB, separated by SDS-PAGE and visualized by in-gel fluorescence (top gel image). The gel was stained in Coomassie Blue to assess protein loading (bottom gel image). As my aim was to study prenylation in as physiologically relevant conditions as possible, it was decided that statins should not be used routinely in my labelling experiments. Statins have a

56 highly disruptive effect on cellular metabolism, and affect the levels of several prenylated proteins such as HRAS and RHOA251, 252 (see Figure 20B). However, incubating cells with isoprenoids or isoprenoid analogues on their own does not have an effect on RAS and RHO expression251, 253. To circumvent the need for statin treatment I sought to develop an improved farnesyl probe to enhance labelling efficiency.

2.5. An Expanded Library of Farnesyl Probes

As the azido and propargyl ether probes previously tested did not give optimal results, we expanded our library of farnesyl probes. Two more farnesyl probes termed AzF and YnF were obtained from Dr Remigiusz Serwa (see Figure 14 for structures). The pyrophosphate derivative of AzF was previously reported as an excellent substrate for FTase in vitro199, and as such I wanted to test its ability to label proteins in cells. Importantly, in contrast to the previously tested azido probes AzF did not contain an allylic azide. The analogous YnF was developed to provide an alkyne version of AzF.

AzC15, PeC15, AzF, YnF and AzC20 labelling in EA.hy926 cells was compared by in-gel fluorescence (Figure 20A). With statin pre-treatment, AzC15, PeC15 and YnF all showed comparable labelling efficiency. In contrast to the previous in vitro data199, AzF did not appear to be an efficient substrate for FTase. Encouragingly, YnF resulted in very strong signal even in the absence of statin pre-treatment. YnF labelled two protein bands in the 50 kDa region, which were only weakly labelled by the widely used PeC15 probe.

Figure 20. Comparison of farnesyl probe labelling efficiency. EA.hy926 cells were pre-treated with Mevastatin (10 μM) or vehicle (DMSO) for 24 hours followed by incubation with prenyl probes (25 μM) or vehicle (DMSO) for 24 hours. (A) The cell lysate was subjected to CuAAC with YnTB or AzTB, separated by SDS-PAGE and visualised by in-gel fluorescence. The protein loading was assessed by Coomassie Blue staining. (B) Labelled proteins were immobilised on NeutrAvidin beads. The purified proteins were subjected to Western blot analysis to detect HRAS and RHOA. Note that bands in the input blots have been cropped and repositioned from the original image for ease of comparison to the pulldown blots. Original images of the input blots are available in appendix B. To further validate the labelling efficiency of the probes, affinity purification followed by Western blot analysis was performed for a number of known prenylated proteins (Figure 20B). The biotin moiety of the capture reagent was exploited to bind labelled proteins to NeutrAvidin-coated beads. After washing to remove unlabelled proteins, the bound proteins were separated by SDS-PAGE and transferred to a membrane for Western blotting.

Under normal cellular conditions HRAS is exclusively farnesylated and RHOA is predominantly geranylgeranylated. In agreement with the labelling pattern achieved by in-gel fluorescence, the four farnesyl probes all labelled HRAS under conditions of statin pre-treatment. In the absence of statin pre-treatment, HRAS was very efficiently detected by YnF, weakly detected by AzC15

58 and no signal was observed with PeC15 or AzF. The geranylgeranyl probe AzC20 was not incorporated into HRAS. Statin pre-treatment appeared to decrease total levels of cellular HRAS. This effect of statins on HRAS has previously been observed in small-cancer lung cell line H-69 and lung carcinoma cell line A549254.

Statin pre-treatment resulted in incorporation of farnesyl probes AzC15, PeC15 and YnF into RHOA, suggesting that depletion of endogenous isoprenoids by statins does in fact induce non- physiological prenylation patterns. Interestingly, statins had a deleterious effect on RHOA labelling by AzC20, mirroring the results obtained by in-gel fluorescence scanning. A similar effect has previously been reported in Jurkat cells204. Overall RHOA protein levels were, however, increased in response to statin treatment, which has previously been observed in several cell types including colorectal cancer cell line HCT11635, 255, erythroleukemia cell line K562252 and leukaemia monocytic cell line THP-1256. The disparity between increased levels of total RHOA but decreased levels of labelled RHOA remains unclear. However, downregulation of GGTase-1 in response to statins has been noted in mice bronchial smooth muscle cells257. I did not explore the effect of statins on GGTase-1 levels in EA.hy926 cells; however, a decreased level of this protein would account for the reduction in labelling.

2.6. Evaluation of Shorter YnC14 Probe

YnF, which exhibited excellent ability to label farnesylated proteins, is one carbon longer in chain length than farnesol. As such, we aimed to investigate whether shortening the structure by one carbon would yield an even better probe, and to this extent YnC14 was designed. YnC14 was accessed via the same synthetic route as YnF (further detailed in section 2.7), but coupled to an acetylene moiety rather than a propyne group in the final stages of the synthesis (Scheme 2). Initial attempts to generate an acetylide by treatment of acetylene-TMS with NaH and displacing the iodine of 5 failed. Instead, reacting iodide 5 with lithium acetylide ethylenediamine complex proved a successful method to generate YnC14.

Scheme 2. Synthesis of farnesyl probe YnC14. a) Ethynyltrimethylsilane, n-BuLi, THF, 0°C – RT; b) Lithium acetylide ethylenediamine complex, Et2O, DMSO, 0°C; c) Pyridinium p- toluenesulfonate, EtOH, 60°C, 46% (2 steps).

The labelling efficiency of YnC14 versus YnF was compared by in-gel fluorescence and HRAS immunoblotting. As evident in Figure 21 the YnC14 probe did not result in improved labelling; rather, it performed much poorer than YnF. Although YnC14 is the same chain length as farnesol, the replacement of the terminal isoprene unit by an alkyl chain leads to a more flexible structure. The increased flexibility probably results in an altered binding mode in the farnesyl transferase enzyme, resulting in non-optimal placement of the pyrophosphate group in relation to the catalytic zinc.

Figure 21. Comparison of YnC14 versus YnF labelling efficiency. EA.hy926 cells were cultured in medium supplemented with YnC14 or YnF (20 μM) or vehicle (DMSO) for 24 hours. The cell lysate was subjected to CuAAC with AzTB, separated by SDS-PAGE and visualised by in-gel fluorescence. The protein loading was assessed by Coomassie Blue staining. Labelled proteins were immobilised on NeutrAvidin beads. The purified proteins were subjected to Western blot analysis to detect HRAS. Note: Lanes have been cropped in the Western blots to omit irrelevant data, resulting in background signal variation. Full-size blots are available in Appendix B. 2.7. Development of Geranylgeranyl Probe YnGG

With the novel farnesyl probe YnF in hand, we next focused on the development of a probe for geranylgeranylated proteins. Previous studies and our initial screening indicated that AzC20 was a potential probe suited to study geranylgeranylated proteins201, 204. However, as previously discussed, the AzC20 probe suffered from several issues such as high background labelling and chemical instability. Furthermore, the use of AzC20 in parallel with YnF would require the use of two different capture reagents and hence more control samples to process. Importantly, any background labelling would not be directly comparable, which would particularly hinder the analysis of proteomics data.

Based on the success of the YnF probe, the analogous probe for geranylgeranylation, termed YnGG, was designed. An initial synthetic strategy for this probe was developed based on Dr Remigiusz Serwa’s synthesis of YnF (unpublished data) and previously reported synthetic methodology of similar compounds199, 258, 259 as outlined in Scheme 3. Epoxidation of the terminal double bond of geranylgeraniol followed by oxidative cleavage using sodium periodate yielded aldehyde 10. After THP-protection of the hydroxyl group, aldehyde 11 was reduced to alcohol 12 via facile treatment with sodium borohydride. Although this synthetic strategy was successful to this point, not enough material was obtained to complete the proposed synthesis of YnGG. Due to the prohibitively high cost of the geranylgeraniol precursor (>£400/g) this synthetic strategy was abandoned. Rather than reiterating the synthesis on a larger scale it was deemed sensible to re-design the synthesis using a shorter and cheaper isoprenoid starting material.

Scheme 3. Attempted synthetic route to YnGG. a) N-bromosuccinimide, H2O, THF, 0°C, 34% ; b) K2CO3, MeOH, 87%; c) NaIO4, H5IO6, H2O, THF, 0°C; d) 3,4-dihydro-2H-pyran, pyridinium p-toluenesulfonate, DCM, 53% (2 steps); e) NaBH4, THF, 0°C, 53%; f) PPh3, Imidazole, I2, 0°C, 21%. The second synthetic strategy attempted is outlined in Scheme 4. The same procedure was also applied to synthesise YnF on a larger scale. Alcohols 15a and 15b were generated in moderate yields through oxidation of the allylic methyl group of THP-protected isoprenoids 14a and 14b, a reaction already trialled in the synthesis of propargyl ether and azido probes (see Scheme 1). Initial attempts to convert the alcohols into bromides failed despite using established protocols, probably due to decomposition of the allylic bromide during work-up and purification. The bromides 16a and 16b were finally accessed when the reaction was conducted according to a modified procedure by Dauben et al.260, using lower reaction temperatures and, importantly, never heating the compound above room temperature on the rotary evaporator. The bromides were successfully coupled to the copper enolate of ethyl acetate to generate esters 17a and 17b261. Reduction to generate the alcohols proceeded smoothly upon standard treatment with lithium aluminium hydride. From this point the synthetic strategy converged with the previously proposed route (Scheme 3). Alcohols 18a and 18b were converted to iodides via an Apel

61 reaction. Subsequent displacement of the iodine by 1-(trimethylsilyl)-1-propyne under basic conditions and two simple de-protection steps yielded the final compounds YnF and YnGG.

Scheme 4. Optimised synthesis of YnF and YnGG. a) 3,4-dihydro-2H-pyran, pyridinium p-toluenesulfonate, DCM, 87% (14a), 94% (14b); b) SeO2, t-BuOOH, DCM, 0°C, 32% (15a), 22% (15b); c) N-bromosuccinimide, dimethyl sulfide, DCM, -35°C – 0°C, 40% (16a), 73% (16b); d) Diisopropylamine, n-BuLi, EtOAc, CuI, THF, -78°C, 50% (17a), 67% (17b); e) LiAlH4, THF, 0°C, 82% (18a), 69% (18b); f) PPh3, Imidazole, I2, 0°C, 56% (19a), 84% (19b); g) 1-(trimethylsilyl)propyne, n-BuLi, THF, -25°C, 83% (20a), 80% (20b); h) Pyridinium p-toluenesulfonate, EtOH, 60°C, 85% (21a), 82% (21b); i) TBAF, THF, 59% (YnF), 33% (YnGG). The synthetic route developed, although reliable, could benefit from several improvements. The overall yields were moderate, mainly due to the inefficiency of the allylic oxidation by selenium dioxide. Selenium dioxide is also highly toxic and thus an undesirable reagent to work with. Furthermore, a modular approach to access both the YnF and YnGG probes from one precursor would cut down synthesis time. However, in the interest of time no further improvements to the synthesis were attempted.

2.8. Labelling by YnF and YnGG is Time and Concentration Dependent

Labelling efficiency of the novel geranylgeranyl probe YnGG was compared to the labelling of AzC20 and YnF in EA.hy926 cells (Figure 22). YnGG and AzC20 labelling was near identical, with one band in the 12 kDa region appearing more strongly with the YnGG probe. Furthermore, YnGG showed excellent incorporation into RHOA, but did not appear to label HRAS, as assessed by immunoblotting. This highly encouraging result suggests that YnGG is an excellent alternative to AzC20, with the added benefit of lower background labelling and improved chemical stability. The orthogonal labelling pattern achieved by in-gel fluorescence with YnF versus YnGG, coupled to selective labelling of HRAS and RHOA respectively, indicate that these two probes can be utilised to differentiate between farnesylated and geranylgeranylated proteins. 62

Figure 22. Labelling by YnF and YnGG is concentration dependent. EA.hy926 cells were cultured in medium supplemented with prenyl probes or vehicle (DMSO) for 24 hours. The cell lysate was subjected to CuAAC with AzTB, separated by SDS-PAGE and visualised by in-gel fluorescence. Labelled proteins were immobilised on NeutrAvidin beads. The purified proteins were subjected to Western blot analysis to detect HRAS or RHOA. The upper bands apparent in the input blots likely correspond to labelled protein which migrates more slowly due to the added mass of the capture reagent, as observed by others using the AzTB capture reagent186, 194. To pin-point the optimal conditions for labelling, concentration and time courses were performed (Figure 22 and Figure 23). For YnF optimal labelling was achieved with 10 µM probe. Interestingly, a further increase in YnF probe concentration resulted in diminished labelling. The same affect was also observed in other cell types (sections 5.2.1 and 5.4.1). There are several plausible reasons for this effect. It is possible that a high concentration of isoprenoid alcohols disrupts the isoprenoid rescue pathway that converts the probes into pyrophosphates, leading to a decreased concentration of the pyrophosphate substrates. Alternatively, high levels of isoprenoids could result in decreased cellular levels of proteins that are substrates for prenylation. Exogenous isoprenoids can reverse the effect of statin-induced upregulation of small GTPases, indicating that isoprenoid levels (or prenylation) may serve to regulate protein

63 levels251. However, in the same study the authors mention that no effect on protein levels is seen in response to incubation with isoprenoids alone.

Labelling with YnGG did not exhibit the same dramatic dose-dependent decrease in labelling observed with YnF in EA.hy926 cells. In other cell types this effect was, however, observed (see chapter 5). A 10 µM concentration of YnGG appeared to be optimal, as higher concentrations did not lead to any great improvement in labelling. Labelling with the new prenyl probes was also time-dependent, as observed by in-gel fluorescence in Figure 23. For YnF, labelling efficiency plateaued after 24 hours and further incubation did not appear to significantly improve incorporation. 24 hours of incubation also resulted in robust labelling by YnGG. Further evaluation of optimal labelling conditions was performed by proteomics and is discussed in Chapter 3.

Figure 23. Labelling by YnF and YnGG is time dependent. EA.hy926 cells were cultured in medium supplemented with prenyl probe (10 μM) or vehicle (DMSO) for the indicated time period. (A) The cell lysate was subjected to CuAAC with AzTB, separated by SDS-PAGE and visualised by in-gel fluorescence. The protein loading was assessed by Coomassie Blue staining. 2.9. Labelling by YnF and YnGG is Sensitive to Natural Substrate Competition

To further validate the novel prenyl probes a competition experiment using the natural isoprenoid substrates was conducted. The rationale was that if YnF and YnGG were incorporated onto prenylated proteins then the labelling should be sensitive to competition with the natural substrates. EA.hy926 cells were cultured in presence of the prenyl probes and increasing concentrations of farnesol or geranylgeraniol. As apparent in Figure 24, addition of the natural isoprenoids lead to a dose-dependent reduction in labelling by both probes. YnGG labelling was exquisitely sensitive to competition by GGOH as well as FOH with loss of labelling achieved at a 2:1 and 1:1 isoprenoid:probe ratio respectively. YnF was less sensitive to isoprenoid competition than YnGG.

Figure 24. Labelling by YnF and YnGG is outcompeted by natural isoprenoid substrate. EA.hy926 cells were cultured in medium supplemented with prenyl probe (5 μM) and increasing concentrations of farnesol (FOH) or geranylgeraniol (GGOH) for 8 hours. The cell lysate was subjected to CuAAC with AzTB, separated by SDS-PAGE and visualised by in-gel fluorescence. The protein loading was assessed by Coomassie Blue staining. The interpretation of the competition experiments is complex, and highlights the some of the challenges encountered when studying prenylation. A simple interpretation of the results may suggest that both probes are substrates for all prenyl transferase enzymes as labelling is sensitive to both FOH and GGOH. However, such an interpretation is highly simplistic. Firstly, FOH and GGOH can bind to all the prenyl transferase enzymes, but FOH and GGOH can only act as an efficient substrate for FTase and GGTase-1/2 respectively. This means that FOH can effectively act as a competitive inhibitor for GGTase-1/2 and GGOH as an inhibitor for FTase. Therefore, it is hard difficult to distinguish whether loss of labelling should be attributed to prenyl transferase inhibition rather than substrate competition. Furthermore, as the probes and natural isoprenoids need to enter the isoprenoid rescue pathway for conversion to pyrophosphates, the observed effect could in fact be due to competition for the rescue pathway. In mammals, only the kinases responsible to conversion of farnesol to farnesyl mono- and di-phosphates have been characterised211, 212. In archaebacterium Sulfolobus acidocaldarius, geranylgeraniol kinase and geranylgeranyl monophosphate kinase are able to phosphorylate farnesol as well262, raising the question whether the mammalian farnesol rescue pathway processes both isoprenoids.

The analysis is further complicated by the fact that farnesol may exhibit toxic effects on cells. Farnesol has been shown to induce apoptosis and cell cycle arrest in cancer cells, mediated

65 through several mechanisms including downregulation of HMG-CoA reductase and ER stress263. However, the concentrations at which these effects were seen far exceed the concentrations of farnesol used in the competition experiment. Hence, the loss of labelling observed upon the addition of farnesol is unlikely to be due to general cell toxicity effects.

2.10. Conclusions

This chapter describes the evaluation of isoprenoid probes to label prenylated proteins. A library of novel as well as previously reported alkyne and azide-tagged analogues of isoprenoid alcohols was synthesised and tested in cell labelling experiments. Our initial evaluation by in- gel fluorescence suggested that the azido probes AzC15 and AzC20 labelled a greater number of proteins than the corresponding propargyl ether probes PeC15 and PeC20. However, issues of background labelling and chemical instability urged me to abandon the work with the azido probes. Instead, an expanded set of farnesyl probes was explored. Comparison of labelling efficiency by in-gel fluorescence and immunoblotting identified YnF as a novel probe for farnesylation which, in contrast to previously reported probes, showed robust incorporation without requiring depletion of the endogenous pool of isoprenoids through statin treatment. The encouraging results achieved with YnF prompted development of the structurally analogous probe for geranylgeranylation, YnGG. A successful synthetic strategy was devised, which was also applicable to the synthesis of YnF. Labelling experiments with YnGG indicated that this probe was comparable to AzC20, with the added benefit of lower background. Importantly, YnGG showed no stability issues.

The YnF and YnGG probes provide a new toolkit for studying prenylation in live cells. Although previous in vitro studies have provided important information on the identity of prenylated proteins and the prenyl transferase enzymes, studies in intact cells are more physiologically relevant, for example by providing spatial control of the prenyl transferase enzymes and protein substrates. Metabolic labelling using radioisotope-labelled mevalonate or isoprenoids has widely been used to identify prenylated proteins. However, apart from issues of poor signal and long exposure times, this method can only be used to study one protein at a time. The use of our prenyl probes allows enrichment of the prenylated proteome, an option not available through other experimental methods such as radiolabelling. Hence our method workflow provides the opportunity to study the whole prenylated proteome within one experiment and perform unbiased de novo identification of prenylated proteins. Subsequent chapters describe the use of the novel probes in proteomics studies to identify novel prenylated proteins, explore prenylation dynamics in response to prenyl transferase inhibitors, and attempts to study prenylation in primary cells and in vivo.

Chapter 3: Development and Optimisation of Proteomics Platform

3.1. Introduction

The rapid development of proteomics technologies in recent years has enabled the study thousands of proteins within one single experiment264. The availability of increasingly sensitive and affordable mass spectrometers means that proteomics analysis is now readily available for routine application. Although post-translational modifications can be studied directly by mass spectrometry by identification of the PTM-bearing peptide, this strategy has limited applicability for prenylation and generally only results in a handful of identifications265. The large, greasy isoprenoid modification hampers peptide ionisation leading to poor signal. Furthermore, peptides must be of sufficient length to be confidently assigned to a specific protein. Before proteomic analysis, proteins are subjected to proteolytic cleavage, typically by trypsin which cleaves at the carboxyl-terminal side of lysine or arginine residues. Many prenylated proteins contain a trypsin cleavage site very close to the C-terminus, resulting in the release of a very short C-terminal peptide which cannot be assigned to a unique protein sequence.

The labelling strategy developed in Chapter 2 lends itself particularly to incorporation into a proteomics workflow. Traditionally, the identification and quantification of prenylated proteins is hampered by the lack of inherent handles for isolation of the prenylome. Our prenyl probes overcome these issues by introducing a chemical tag in the form of an alkyne onto prenylated proteins, which facilitates down-stream derivatisation with any label of choice. By employing affinity handles, our strategy enables facile purification of the prenylome thus allowing unbiased de novo identification of prenylated proteins.

This chapter describes the development and optimisation of a chemical proteomics platform incorporating the novel prenyl probes YnF and YnGG. We coupled in-cell labelling experiments to quantitative proteomic methodologies such as stable isotope labelling with amino acids in cell culture (SILAC) to enable relative quantification of prenylated proteins. In addition, we describe the development of novel capture reagents to enable the identification of prenylated peptides.

3.2. Optimising Sample Processing: Does Reduction/Alkylation Affect Labelling?

During preparation of proteomics samples, proteins are usually subjected to a reduction/alkylation protocol to break disulfide bridges and subsequently cap the free sulfhydryl groups before digestion by trypsin. Dithiothreitol (DTT) is commonly used for the reduction step and iodoacetamide (IAA) for the alkylation. In contrast to palmitoylation, which is a reversible and labile cysteine modification, prenylation is considered irreversible under physiological

67 conditions. However, prenyl groups can be cleaved using chemical means such as treatment of proteins with methyl iodide under acidic conditions266.

As cleavage of the prenyl would lead to loss of enriched proteins we wanted to ensure that our prenyl probe modification was stable under the reduction/alkylation conditions used during proteomics sample preparation. We assessed the integrity of the probe modification by monitoring the release of fluorescently labelled proteins from beads subjected to different steps of the reduction/alkylation workflow. Briefly, protein lysate from YnF-fed EA.hy926 cells was subjected to CuAAC with capture reagent AzTB followed by pulldown on NeutrAvidin beads. The beads were subsequently subjected to various steps of the proteomics preparation workflow as outlined in Figure 25. In addition to treatment with IAA, we also chose to treat the beads with N-ethylmaleimide (NEM) which is another cysteine-reactive alkylating agent. The proteins were released from the beads, separated by SDS-PAGE and imaged by in-gel fluorescence. Based on the in-gel fluorescence signal intensity it appeared that no step of the reduction/alkylation protocol led to loss of labelling when compared to proteins released from untreated beads (lanes 4-7 versus lane 3). Furthermore, no fluorescent proteins were detected in the bead supernatant after reduction/alkylation (lanes 8-9). Thus, we concluded that the prenyl probe modification was stable under reduction/alkylation conditions adopted for preparation of proteomics samples.

Figure 25. The prenyl probe modification is unaffected by reduction and alkylation steps of the proteomics preparation workflow. YnF-tagged EA.hy926 lysate was subjected to CuAAC with AzTB (lane 1) and labelled proteins immobilised on NeutrAvidin resin (lane 2 [pulldown supernatant] and lane 3 [beads]). Beads were subjected to various steps of the reduction/alkylation protocol of the proteomics preparation workflow. Reduction (lane 5), and reduction followed by alkylation with iodoacetamide (lane 6) or N-ethylmaleimide (lane 7) did not result in loss of protein from the beads, which was further validated by the absence of fluorescent proteins in the bead supernatant (lanes 8 and 9).

3.3. Initial Proteomics Evaluation of YnF, AzC15 and AzC20

Following the development of the isoprenoid probes described in Chapter 2 an initial evaluation of the utility of our probes in identifying prenylated proteins by proteomics was undertaken. This preliminary experiment was conducted using farnesyl probes YnF and AzC15 and geranylgeranyl probe AzC20. Briefly, EA.hy926 cells were incubated with YnF, AzC15, AzC20 or DMSO vehicle. After cell lysis the samples were subjected to CuAAC with capture reagent AzTB or YnTB. Labelled proteins were immobilised on NeutrAvidin beads, digested by trypsin and analysed by LC-MS/MS as per the workflow outlined in Figure 15 (Chapter 2). The results of the processed data are summarised in Figure 26 and Table 2.

Figure 26. Number of proteins identified in initial proteomics experiment by novel alkyne farnesyl probe YnF and previously reported azido probes AzC15 and AzC20. EA.hy926 cells were grown in medium supplemented with YnF, AzC15, AzC20 or DMSO. Post-lysis the protein samples were subjected to click reaction with AzTB (YnF and DMSO samples) or YnTB (AzC15, AzC20 and DMSO samples), immobilised on NeutrAvidin resin and subjected on to on-bead tryptic digest. Peptide analysis by LCMS/MS and protein identification was performed by Dr Andrew Bottrill at the University of Leicester Proteomics Facility. A minimum of 3 unique peptides were required for protein identification. n=1 for all samples. (A) Comparison of proteins identified in YnF, AzC15 and DMSO control samples. The DMSO grouping includes proteins identified in samples treated with AzTB and/or YnTB. B) Number of proteins identified in initial proteomics experiment by AzC20. DMSO includes proteins identified in samples treated with YnTB. C) Comparison of prenylation motif proteins identified in initial proteomics experiment by novel alkyne farnesyl probe YnF, previously reported azido probes AzC15 and AzC20. 1 protein detected in DMSO control samples was discarded from this comparison. D) Comparison of background proteins identified using capture reagents AzTB or YnTB in DMSO control samples.

A total of 81 proteins were identified by YnF, 132 proteins by AzC15 and 171 proteins by AzC20 (Figure 26A & B). After discarding proteins that were also identified in the DMSO control samples (treated with AzTB or YnTB), YnF was found to label 33 proteins, AzC15 labelled 53 proteins and AzC20 labelled 78 proteins. The data were further filtered to identify proteins bearing a potential prenylation motif (proteins with a C-terminal CXXX motif where X denotes any amino acid and members of the RAB family). The normal consensus motif for prenylation by FTase or GGTase-1 is a CAAX motif (where A is any aliphatic amino acid) but we chose to apply the more unbiased CXXX motif in our filtering to avoid losing information about potential novel substrate sequences. Within this subset, YnF identified 28 proteins, AzC15 identified 25 proteins and AzC20 identified 21 proteins when disregarding any proteins that were also present in the DMSO controls (Figure 26C). Thus, although AzC15 identified more proteins than YnF in total, YnF was able to identify more proteins bearing a prenylation motif, suggesting that the AzC15 probe suffers from greater non-specific labelling. A summary of the CXXX and RAB proteins identified can be found in Table 2.

A comparison of the CXXX and RAB proteins identified by the three probes (Figure 26C) suggested that YnF was more selective than AzC15. However, due to the lack of replicate samples in the experiment the data cannot be considered conclusive. 6 proteins were uniquely identified in the sample treated with YnF including well characterised farnesylated targets such as GTP-binding protein Rheb (RHEB), peroxisomal biogenesis factor 19 (PEX19) and small GTPase KRAS (KRAS). YnF also identified zinc finger CCCH-type antiviral protein 1 (ZC3HAV1), a protein which was recently described as a novel farnesylated target in a chemical proteomics study conducted by the Hang lab using propargyl ether probe PeC15207. Additionally, two novel proteins bearing a CXXX motif were uniquely identified by YnF: DDB1- and CUL4-associated factor 8 (DCAF8) and EH domain-binding protein 1-like protein 1 (EHBP1L1). In contrast, all proteins identified by AzC15 were also detected by YnF and/or AzC20. YnF and AzC15 both identified the putative novel prenylated proteins leucine-rich repeat flightless-interacting protein 1 (LRRFIP1) and protein DPCD. Further discussion and validation of these putative targets follows in Chapter 4.

AzC20 uniquely identified 9 CXXX and RAB proteins and showed greater overlap with AzC15 (10 proteins) than with YnF (7 proteins). All the proteins identified by AzC20 are known or predicted to be geranylgeranylated, except endothelin-converting enzyme-1 (ECE1), which contains the non-canonical prenylation motif CEVW. The prenylation state of this protein has previously been studied, but it was concluded that ECE1 is in fact not prenylated. Instead, the C-terminal cysteine is thought to be involved in disulfide bridge formation267. Thus it was unclear at this point whether the identification of ECE1 resulted from non-specific labelling.

Table 2. Summary of CXXX and RAB proteins identified in initial proteomics experiment with probes YnF, AzC15 and AzC20. A minimum of 3 unique peptides were required for identification. The prenylated cysteine(s) (verified or predicted) are highlighted in red in the C-terminal sequence. Proteins identified in previous chemical proteomics studies by the Hang207, Zhao195, Tamanoi204, Alexandrov200 and Distefano209, 268 labs are indicated by “Y”. Annotation of prenylation in Uniprot (accessed 27/07/2016) is indicated by F (farnesylated) or GG (geranylgeranylated). Annotations marked by (*) are inferred based on sequence similarity. Predictions of FTase, GGTase-1 and GGTase-2 substrate specificity was performed with the Prenylation Prediction Suite84 (PrePS, http://mendel.imp.ac.at/sat/PrePS/index.html). †KRAS exists in two isoforms with different C-termini which could not be differentiated in this study. LC-MS/MS and data analysis was performed by Dr Andrew Bottrill at the University of Leicester Proteomics Facility.

# Unique peptides Previous ID? PrePS prediction

Uniprot C-terminal Uniprot Gene name accession MW sequence annotation GGTase- GGTase-

YnF AzC15 AzC20 FTase number (kDa) 1 2 Hang Zhao Tamanoi Alexandrov Distefano 1 Distefano 2 BROX Q5VW32 7 6 0 46 CYIS Y F + -- --- CDC42 P60953 6 5 6 21 CVLL Y Y Y GG ++ +++ - DCAF8 Q5TAQ9 3 0 0 67 CMPS + ------DNAJA1 P31689 14 12 0 45 CQTS Y F + ------DNAJA2 O60884 14 11 0 46 CAHQ Y Y F ------DPCD Q9BVM2 3 4 0 23 CKTQ + ------ECE1 P42892 0 0 7 87 CEVW ------EHBP1L1 Q8N3D4 7 0 0 162 CVLS ++ -- --- GNG12 Q9UBI6 0 0 3 8 CIIL GG ++ +++ --- GNG5 P63218 3 3 6 7 CSFL GG* ------INF2 Q27J81 14 11 0 136 CVIQ +++ -- --- KRAS† P01116 3 0 0 22 CIIM/CVIM Y Y F ++/+++ +++/+++ -/- LMNA P02545 13 9 0 74 CSIM Y F ++ +++ --- LMNB1 P20700 8 6 0 66 CAIM Y Y Y Y F ++ +++ --- LMNB2 Q03252 7 3 0 68 CYVM Y Y F* + + --- LRRFIP1 Q32MZ4 6 8 0 89 CTMS + ------NAP1L1 P55209 8 11 0 45 CKQQ Y F + ------NRAS P01111 4 5 0 21 CVVM Y F ++ ++ -

Table 2 cont.

# Unique peptides Previous ID? PrePS prediction

Uniprot C-terminal Uniprot Gene name accession MW sequence annotation GGTase- GGTase-

YnF AzC15 AzC20 FTase number (kDa) 1 2 Hang Zhao Tamanoi Alexandrov Distefano 1 Distefano 2 PEX19 P40855 3 0 0 33 CLIM Y F ++ +++ --- PTP4A1 Q93096 5 3 0 20 CCIQ Y F +++ -- --- RAB10 P61026 0 0 4 23 SKCC Y Y Y GG* ------+++ RAB11B Q15907 7 6 11 24 CCQNL Y Y GG -- -- +++ RAB14 P61106 0 0 6 24 GCGC Y Y Y GG ------+++ RAB1A P62820 5 4 10 23 GGCC Y GG ------+++ RAB21 Q9UL25 0 0 3 24 CCSSG Y GG* - --- +++ RAB2A P61019 0 3 9 24 GGCC Y Y Y Y Y GG ------+++ RAB5C P51148 0 4 8 23 CCSN Y Y Y Y GG* -- --- +++ RAB6A P20340 0 0 5 24 GCSC Y Y Y GG* ------++ RAB7A P51149 0 3 8 23 SCSC Y Y Y Y Y GG* ------+++ RAB8A P61006 0 0 4 24 CVLL Y GG ++ +++ +++ RAC1 P63000 4 0 6 21 CLLL Y GG + +++ - RALB P11234 0 0 3 23 CCLL GG ++ ++ - RAP1B P61224 3 0 4 21 CQLL GG + ++ - RAP2C Q9Y3L5 3 4 0 21 CVVQ Y Y Y GG* ++ -- - RHEB Q15382 5 0 0 20 CSVM Y F ++ ++ - RHOA P61586 8 10 11 22 CLVL Y GG* ++ +++ - RHOG P84095 0 3 7 21 CILL Y GG* + ++ - RRAS P10301 0 0 3 23 CVLL GG* ++ +++ - RRAS2 P62070 4 7 0 23 CVIF Y F ++ ++ - UCHL1 P09936 0 3 3 25 CKAA F ------YKT6 O15498 8 8 0 22 CAIM Y F ++ +++ --- ZC3HAV1 Q7Z2W4 3 0 0 101 CVIS Y ++ -- ---

Analysis of the two DMSO control samples indicated that capture reagent YnTB resulted in much greater background labelling than AzTB (Figure 26D), mirroring the results observed by in-gel fluorescence labelling (Chapter 2). Whereas 93 proteins were detected in YnTB sample only 30 proteins were identified in AzTB control sample. The differences in background labelling further strengthened our resolve to move away from the use of azido probes, and focus solely on the use of alkyne probes.

3.4. Initial Comparison of YnF and YnGG

As discussed in Chapter 2, the promising early results achieved by in-gel fluorescence and immunoblotting experiments with YnF led us to develop the analogous geranylgeranyl probe YnGG. With the two probes in hand, we proceeded to apply these in our proteomics workflow to gain an initial comparison of their putative targets.

The samples were prepared on a large scale (3 mg/sample) as it had been observed in chemical proteomics studies of myristoylation that increasing the scale lead to a large increase in protein identifications269. However, in our case an increase in background labelling was observed with peptides from several prenylated proteins such as LMNA and RHOA identified non-specifically in the DMSO-treated control sample. As this preliminary experiment was only prepared in a single replicate, any in-depth analysis of this data set was limited. Figure 27 provides a summary of the number of proteins hits for each probe.

Figure 27. Number of proteins identified in initial proteomics experiment by novel alkyne probes YnF and YnGG. EA.hy926 cells were grown in medium supplemented with YnF, YnGG or DMSO. Post-lysis the protein samples were subjected to CuAAC with AzTB, immobilised on NeutrAvidin resin and subjected on to on-bead tryptic digest followed by LC-MS/MS analysis on a Q Exactive™ instrument (Thermo Scientific™). A minimum of 3 unique peptides were required for identification. n=1. (A) Comparison of number of proteins identified in YnF, YnGG and DMSO control samples. B) Comparison of the number of CXXX-motif (left) and non-CXXX motif (right) Rab proteins identified by YnF and YnGG.

When disregarding identifications in the control sample a total of 61 versus 66 prenylation motif proteins (CXXX and RAB proteins) were found by YnF and YnGG respectively (Figure 27A). This represents a large increase in protein hits with YnF compared to the results set presented in the previous section. This difference can most probably be accounted for by the use of different mass spectrometers for the LC-MS/MS analysis. Samples described in the previous section were analysed at the University of Leicester Proteomics facility using an LTQ-Orbitrap™ Velos mass spectrometer (Thermo Scientific™). However, during the course of this project a Q Exactive™ (Thermo Scientific™) mass spectrometer became available in-house and this machine was used for the analysis of the samples in the current section. The Q Exactive™ instrument provides greatly improved resolution and mass accuracy, enabling more proteins identifications264.

Interestingly, several of the putative novel farnesylated proteins identified in the preliminary comparison of YnF and AzC15 discussed in the previous section were again identified by YnF in this experiment including EHBP1L1, DCAF8, LRRFIP1 and DPCD (Table 3). In addition, serine/threonine protein kinase ULK3, centrosomal protein of 85 kDa (CEP85), EH-domain binding protein 1 (EHBP1), E3 SUMO-protein ligase RanBP2 (RANBP2) and S-phase kinase- associated protein 1 (SKP1) were also detected exclusively by YnF, with no peptides detected in the DMSO control. For further validation, the peptides identified for each of the proteins were inspected. In the case of SKP1 the unmodified C-terminal peptide was detected, suggesting that this protein was a non-specific hit. Affinity enrichment of the probe-labelled proteins relies on tethering of the C-terminal cysteine to the NeutrAvidin beads via the probe-capture reagent conjugate. The C-terminal peptide would remain on the bead after tryptic digest, and hence would not be detected LC-MS/MS analysis. No C-terminal peptide was detected for the other proteins listed above.

The YnGG probe uniquely detected two potential novel prenylated proteins: ubiquitin-like protein 3 (UBL3) and 2'-5'-oligoadenylate synthase 1 (OAS1) (Table 3). UBL3 is prenylated in Arabidopsis thaliana270 but its prenylation has not been investigated in mammalian systems. ECE1, which was robustly detected by AzC20, was detected by 2 peptides in both the YnGG and DMSO sample. Thus, it is likely that the labelling observed with AzC20-labelling was in fact background.

As previously mentioned, we found that probably due to the large scale preparation of the samples high background labelling was apparent, for example several peptides for LMNA were detected in the DMSO sample. As such, in the original analysis of protein hits (Figure 27) the threshold for protein identification was set at a minimum of 3 unique peptides. Thus, proteins that were detected with 3 or more unique peptides by YnF and/or YnGG but 1 or 2 unique

74 peptides in the DMSO sample were classified as probe hits. A number of well characterised prenylated proteins fit within these selection criteria, including RAS-related C3 botulinum toxin substrate 1 (RAC1) and BRO1 domain-containing protein BROX (BROX). In addition, a number of CXXX-motif proteins with no previous evidence of prenylation in the literature were identified within this group including nucleosome assembly protein 1-like 4 (NAP1L4), activator of 90 kDa heat shock protein ATPase homolog 1 (AHSA1), histone-arginine methyltransferase CARM1, NADH-cytochrome b5 reductase 3 (CYB5R3), 60S ribosomal protein L12 (RPL12), E3 ubiquitin/ISG15 ligase TRIM25 (TRIM25) and sulfotransferase 1A4/1A3/1A1 (SULT1A4/3/1). Inspection of the identified peptides allowed us to discard AHSA1, CARM1, RPL12 and SULT1A4/3/1 as probable background hits due to the presence of a C-terminal peptide or peptides that originated from protein isoforms not bearing a prenylation motif. A summary of the putative novel prenylated proteins identified in this experiment are presented in Table 3 below.

Uniprot # Unique peptides C- PrePS prediction MW Gene name accession terminal GGTase- GGTase- DMSO YnF YnGG (kDa) FTase number sequence 1 2 ULK3 Q6PHR2 0 9 0 53 CTLQ ++ ------EHBP1 Q8NDI1 0 6 0 132 CVLQ ++ -- --- CEP85 Q6P2H3 0 7 0 80 CVTQ ++ ------RANBP2 P49792 0 5 0 358 CGQI ------UBL3* O95164 0 0 3 13 CVIL ++ +++ --- OAS1 P00973 0 1 3 35 CTIL ++ ++ --- NAP1L4† Q99733-2 1 11 1 43 CKQQ + ------CYB5R3 P00387 2 6 5 34 CFVF - - --- TRIM25 Q14258 2 10 4 71 CSPK ------Table 3. Putative novel prenylated proteins in initial proteomics study with probes YnF and YnGG. Data were filtered to retain CXXX-motif proteins that were identified by a minimum of 3 unique peptides in the YnF and/or YnGG samples and a maximum of 2 peptides in the DMSO control sample. n=1. *Prenylated in Arabidopsis270. †Only isoform 2 has CXXX-motif. A major part of the YnGG data set consisted of RAB proteins. RAB prenylation has been intensely studied by the Alexandrov group200. They have developed an in vitro prenylation assay where lysates from statin-treated cells are re-prenylated by addition of exogenous RABGGTase-2 and a biotin-tagged isoprenoid pyrophosphate (B-GPP). YnGG detected 37 RAB proteins by a minimum of 3 unique peptides in this initial study, which compares very favourably to the 42 RAB proteins identified by Nguyen et al. using their optimised methodology. A comparison of the two RAB data sets is summarised in Table 4 below.

YnGG Overlap Alexandrov RAB12 RAB1A RAB4B RAB8B RAB21 RAB34 RAB3A RAB37 RAB23 RAB1B RAB5A RAB9A RAB22A RAB35 RAB3C RAB39B RAB24 RAB2A RAB5B RAB10 RAB27A RAB39A RAB6A RAB41 RAB29 RAB2B RAB5C RAB11B RAB27B RAB43 RAB6C RAB32 RAB3B RAB6B RAB13 RAB30 RAB9B RAB3D RAB7A RAB14 RAB31 RAB11A RAB4A RAB8A RAB18 RAB33B RAB15 Table 4. Comparison of RAB proteins identified by YnGG in this study and a comprehensive study by the Alexandrov group utilising B-GPP200. 3.5. Optimising Labelling Conditions for Proteomics Analysis

Two conflicting challenges present themselves when employing chemical proteomics strategies to detect prenylated proteins: allowing sufficient incubation time and probe concentration for adequate incorporation onto prenylated proteins versus limiting potential probe metabolism and off-target effects. As prenylation is an irreversible post-translational modification, labelling is dependent on cellular protein turn-over. Proteins with short turn-over times will be labelled within shorter time-frames whereas proteins with slow turn-over will require prolonged probe exposure for sufficient probe incorporation. However, prolonged incubation with the probes may lead to off-target effects if the probes are metabolised by the cell. It is highly likely that the farnesyl probe YnF will undergo elongation by geranylgeranyl pyrophosphate synthase (GGPPS) in the cell, resulting in conversion into YnGG as outlined in Figure 28 below. Indeed, recent studies have shown that alkyne fatty acids can be used to trace lipid metabolism271 and, more significantly, that exogenously administered anilinogeraniol (AGOH) (a farnesol analogue) is converted into the longer chain anilinofarnesyl pyrophosphate in cells213 (Figure 28B).

Figure 28. Cellular elongation of FPP and farnesyl analogues. (A) Cells synthesise GGPP by elongation of FPP. (B) Exogenously added AGOH, a farnesol analogue, is phosphorylated via the rescue pathway and elongated by GGPPS in cells213. (C) Potential elongation of phosphorylated YnF (YnFPP) by GGPPS would produce YnGGPP. We considered the possibility of using a GGPPS inhibitor to investigate potential elongation of the probe. However, this would have result in the depletion of GGPP. As discussed in chapter 2, we previously encountered altered prenylation patterns in cells treated with statins and thus we envisaged that depletion of GGPP would probably also upset the normal prenylation dynamics. This could potentially be overcome by supplementing the cell with exogenous GGOH; however, as shown in section 2.9 (Chapter 2), YnF labelling is outcompeted upon addition of GGOH and thus this option would not provide a useful strategy.

To pinpoint a probe concentration and labelling time allowing adequate detection of prenylated proteins by proteomics a concentration and time course was devised. To achieve relative quantification of protein labelling for the different conditions a spike-in SILAC methodology233 was adopted in the proteomics workflow. A schematic of the experimental method and data analysis for the probe concentration study is outlined in Figure 29.

Standard medium MS1 quantification + probe Spike-in Spike-in Spike-in

2 µM probe Mass Mass Mass

Intensity Intensity Intensity shift shift shift CuAAC Enrichment m/z m/z m/z 5 µM probe On-bead digest Light Light Light LC-MS/MS analysis Ratio = Ratio = Ratio = 1 Heavy 2 Heavy 3 Heavy

10 µM probe Normalised ratios:

Heavy (R10K8) Light / Heavy Light Ratio([2 µM] / [10 µM]) = = medium + probe Light / Heavy Light Light / Heavy Light Ratio([5 µM] / [10 µM]) = = Light/Heavy Light Spike-in standard Light / Heavy Light Ratio([10 µM] / [10 µM]) = = = 1 10 µM probe Light Light / Heavy Figure 29. Schematic of spike-in SILAC methodology employed for relative quantification of probe incorporation. Experimental samples were grown in standard medium supplemented with increasing concentrations of prenyl probe YnF or YnGG. A spike-in standard cultured in R10K8- labelled SILAC medium was also labelled with prenyl probe. After cell lysis the experimental samples were mixed with spike-in standard in a 1:1 ratio. The lysates were processed by CuAAC to append capture reagent AzTB, immobilised on NeutrAvidin beads and subjected to tryptic digest followed by LC-MS/MS analysis. For relative comparison of probe incorporation the SILAC ratios generated from the MS1 spectra were normalized with respect to the sample fed the highest probe concentrations (10 µM). Briefly, EA.hy926 cells were cultured for 24 hours in standard culture medium containing increasing concentrations of prenyl probe. In parallel, a probe-labelled spike-in standard was generated from EA.hy926 cells cultured in heavy SILAC-labelled culture medium. The R10K8 SILAC medium contained arginine and lysine isotopically labelled with 13C and 15N, leading to a mass shift of 10 or 8 Da for each arginine- or lysine-containing peptide respectively originating from the spike-in. Each experimental sample was combined with spike-in standard post-lysis. After sample processing by CuAAC, enrichment and protein digest the samples were analysed by LC-MS/MS. While the peak intensity for peptides originating from the spike-in remained constant across the samples, the peaks originating from the experimental samples differed in intensity depending on the amount of probe labelling achieved. Thus the spike-in provided an internal standard allowing relative quantification of the probe incorporation. An analogous methodology was used in the time course study by varying the probe incubation time while keeping the probe concentration constant at 10 µM.

We were interested in the conditions that would maximise the number of proteins quantified. A summary of the number of quantifications of RAB and CXXX-motif proteins achieved using the two probes under the different labelling conditions is presented in Table 5. As each experimental condition was only performed once, conclusions from the data are limited. Quantifications for

YnF ranged from 64 to 73 proteins, and 76 to 78 proteins for YnGG. Thus, it appears that robust labelling can be achieved with low probe concentrations and short incubation times.

Further quality parameters pertaining to the identification and quantification of CXXX and RAB proteins achieved for the different conditions including the number of ratio counts, peptide counts and sequence coverage are summarised in Table 6. Higher ratio and peptide counts leads to more accurate protein quantification, and peptide counts and sequence coverage are important factors in protein identification. Increasing probe concentration and incubation time appeared to increase the number of ratio counts achieved for both probes. YnF-labelling was more dependent on probe concentration and exposure time with respect to number of peptide identifications and sequence coverage, whereas these parameters were very similar across the different samples for the YnGG-labelled data sets.

Conclusions about ideal labelling conditions based on the data in Table 5 and Table 6 were hard to draw. Firstly, as the experiment was only performed once due to limited availability of LC-MS/MS time, the results should be approached with some caution. Secondly, all samples were mixed with a spike-in standard incubated with 10 µM probe for 24 hours, representing the maximum probe concentration and labelling time employed in the experimental samples. Thus, the intensity of the isotopically labelled peptides was the same across all conditions, which may have resulted in a boost of protein identifications and quantifications in samples exposed to low probe concentration or short incubation times. Perhaps greater differences would have been apparent in a duplex-SILAC experiment where the light and heavy lysates were prepared under the same conditions. Such experimental set-up would not allow relative quantification of probe labelling across the different conditions, but would provide a more accurate picture of how many protein identifications and quantifications would be achievable for each condition.

For optimised proteomics studies it was decided that a probe concentration of 10 µM and an incubation time of 8 hours should be employed. Although a 24 hours incubation time resulted in slightly improved incorporation, 8 hours was sufficient to achieve robust labelling.

Table 5. Number of RAB and CXXX proteins quantified in concentration gradient and time course studies with YnF and YnGG. Results were filtered to retain proteins that were identified by a minimum of 3 razor & unique peptides in at least one sample within each study. n=1.

Concentration Gradient Time Course YnF YnGG YnF YnGG

2 M 5 M 10 M 2 M 5 M 10 M 4h 8h 24h 4h 8h 24h µ µ µ µ µ µ Number of protein quantifications 64 74 73 76 77 78 71 67 72 78 77 78

Table 6. Summary of data quality parameters achieved for RAB and CXXX proteins in (A) concentration gradient and (B) time course studies with YnF and YnGG. The mean and median number of ratio counts, razor & unique peptide counts and % sequence coverage within each data set. Results were filtered to retain proteins that were identified by a minimum of 3 razor & unique peptides in at least one sample within each study. n=1.

A YnF Concentration Gradient YnGG Concentration Gradient Razor & unique % Sequence Razor & unique % Sequence Ratio counts Ratio counts peptides coverage peptides coverage [Probe] 2 µM 5 µM 10 µM 2 µM 5 µM 10 µM 2 µM 5 µM 10 µM 2 µM 5 µM 10 µM 2 µM 5 µM 10 µM 2 µM 5 µM 10 µM Mean 6.3 9.0 9.9 6.1 8.3 9.2 25.4 33.0 37.0 9.2 11.9 12.7 8.2 8.9 9.3 47.5 50.7 52.0 Median 5 7 8 5 7 7 25.1 35.2 37.5 7 9.5 10.5 7 8 8 52.9 55.2 54.2

B YnF Time Course YnGG Time Course Razor & unique % Sequence Razor & unique % Sequence Ratio counts Ratio counts peptides coverage peptides coverage Time 4h 8h 24h 4h 8h 24h 4h 8h 24h 4h 8h 24h 4h 8h 24h 4h 8h 24h Mean 8.3 8.1 10.0 7.6 7.5 9.3 29.6 29.4 36.9 10.7 11.8 12.5 8.6 8.8 9.2 48.1 49.4 51.4 Median 6 6 8 6 6 7 30.0 28.2 36.9 9 9 10 8 8 8 53.3 52.1 53.6

To provide a relative quantification of labelling, all samples were normalised with respect to the highest probe concentration or incubation time point. The concentration gradient results are shown in Figure 30 and the time course results in Figure 31. Encouragingly, RAB proteins (red lines) and the majority of proteins bearing a CXXX-motif on the C-terminus (blue lines) showed a dose-dependent increase in labelling. In contrast, the majority of proteins without a prenylation motif remained at a constant level suggesting that they are a result of background labelling.

Figure 30. The effect of probe concentration on labelling of proteins by YnF and YnGG. Data were filtered to retain proteins that were quantified across all concentration points, with a minimum of 3 unique and razor peptides in at least one sample. n=1. Red = RAB proteins; Blue = CXXX proteins; Grey = other proteins. In the YnF concentration study, 4 CXXX-motif proteins appeared unaffected by probe concentration, suggesting that they are not probe targets. Interestingly, this subset included ubiquitin carboxyl-terminal hydrolase isozyme L1 (UCHL1), which has previously been reported to be modified by farnesylation272. UCHL1 was also detected by YnGG and labelling was similarly unaffected by probe concentration. This protein and questions arising regarding its prenylation status is discussed in further detail in Chapter 4.

Some of the novel putative farnesylated proteins identified in the initial studies with YnF were also identified in these experiments. LRRFIP1, ULK3, DCAF8, NAP1L4, DPCD, CEP85 and EHBP1L1 showed the same dose-dependent increase in labelling as known farnesylated proteins, suggesting that they are in fact modified by our YnF probe. OAS1, a potentially novel geranylgeranylated protein, showed concentration-dependent increase in labelling with the YnGG probe. Also detected by the YnGG probe was the protein EF-hand calcium-binding domain-containing protein 4B (EFCAB4B/CRACR2A). This protein has recently been found to exist in a long isoform in endothelial cells, with a C-terminal extension which makes the long isoform a member of the RAB GTPase family273. The C-terminal SCCG sequence is prenylated274. EFCAB4B was only detected by YnGG, and inspection of the identified peptides did indeed confirm that several originated from the sequence unique to the long isoform of the protein.

Figure 31. The effect of incubation time on labelling of proteins by YnF and YnGG. Data were filtered to retain proteins that were quantified across all concentration points, with a minimum of 3 unique and razor peptides in at least one sample. n=1. Red = RAB proteins; Blue = CXXX proteins; Grey = other proteins. The results from the time course study (Figure 31) were more difficult to interpret than the concentration study as the rate of probe incorporation is additionally dependent on the rate of

82 protein turn-over. For example, RHOB labelling with YnGG was found to increase in a concentration-dependent manner, but was not time-dependent. This is consistent with the short half-life of RHOB (2 hours)275 facilitating rapid incorporation of the prenyl probe onto this protein. In contrast, labelling of proteins with longer half-life such as HRAS (~24 hours)276, responded in a time dependent manner within the time-frame of the experiment.

While the majority of proteins with a prenylation motif showed no change or a time-dependent increase in labelling with time, a number of proteins with no prenylation motif showed a decrease in level over time. Annexin A2 (ANXA2), which was detected in both YnF- and YnGG-fed samples showed a particularly big decrease (top grey line in both plots in Figure 31). ANXA2, which has a C-terminus sequence ending CGGDD was identified in a chemical proteomics study with AzC15 by the Zhao lab195 where the authors suggested that this protein may have an atypical prenylation motif. In contrast, a study employing AGOH and 2D immunoblotting drew the conclusion that ANXA2 is not prenylated277. Our data suggest that this protein is not prenylated, as it was reliably detected in the DMSO control samples in initial proteomics experiments and the peptide list included the C-terminal sequence.

The reason for the decrease in ANXA2 is unclear, but may be the result of altered protein expression or degradation in response to increased isoprenoid levels. ANXA2 is present in multiple compartments of the cell, including the plasma membrane and cytoplasm where it is involved in a multitude of functions including regulating membrane dynamics and intracellular vesicles278. Annexins are phospholipid-binding and it has been suggested that phospholipid binding recruits ANXA2 to the membrane279, where it incorporates into detergent-insoluble rafts280. It is unknown whether ANXA2 binds to isoprenoids, but if this is the case the decrease in ANXA2 could reflect an accumulation into detergent-insoluble domains mediated by YnFPP or YnGGPP binding. ANXA2 also binds and inhibits proprotein convertase subtilisin/kexin- type 9 (PCSK-9), a protein which regulates the degradation of the low-density lipoprotein receptor (LDLR)281. The LDLR is a key regulator of cholesterol homeostasis within the cell by mediating uptake of LDL from the plasma, which releases cholesterol within the cell282. Cholesterol is also biosynthesised by cells through the mevalonate pathway. As isoprenoids are intermediates of the cholesterol biosynthesis pathway, increased isoprenoid levels may signal to the cell to downregulate its cholesterol levels. Thus, a decrease in ANXA2 may act as an acute protective mechanism by releasing PCSK-9 and thus inducing LDLR degradation and preventing cellular uptake of LDL. These potential mechanisms of ANXA2 regulation in the presence of YnF and YnGG are highly speculative and further analysis is required to understand the dramatic decrease in ANXA2.

A number of non-prenylation motif proteins showed a time-dependent increase in YnGG samples similar to the trends observed for RAB and CXXX-proteins. These included cell- surface antigens CD9, CD44, CD63, CD81 and CD151, podocalyxin (PODXL), DnaJ homolog subfamily C member 5 (DNAJC5), transferrin receptor protein 1 (TFRC), endothelial protein C receptor (PROCR) and Niemann-Pick C1 protein (NPC1). None of the proteins were detected in the YnF-labelled samples. All of the proteins are localised at membranes, suggesting that they may have been purified by association with prenylated partners. For example, CD9, CD63, CD81 and CD151 are found in RAB-containing exosomes released from cells283 and PODXL is thought to co-localise with RAB11A and RAB8A. NPC1 is a lipid-binding protein which is localised to late endosomes and lysosome membranes and acts as a vesicular shuttle for cholesterol and other lipids from late endosomes284. Thus its detection could be due to interaction of YnGG with its lipid-binding motif.

3.6. Quantitative Comparison of Probe Preference

A main goal of this project was to develop a labelling system that could distinguish between farnesylated and geranylgeranylated proteins. From the results obtained in the initial proteomics studies described within this chapter, it is clear that YnF and YnGG label a distinct but partially overlapping set of proteins. Although some proteins such as RHOB and RRAS2 can be either farnesylated or geranylgeranylated266, 285, most prenylated proteins are thought to be selectively modified by one isoprenoid only.

To better understand the selectivity of our probes and gain an understanding of the preferred prenylation status of the proteins identified in our proteomics studies we chose to quantify the relative incorporation of our probes by means of quantitative proteomics. EA.hy926 cells were isotopically labelled in heavy or light SILAC medium and subsequently fed either YnF or YnGG probe. Cell lysates were combined in a 1:1 ratio and processed for proteomics analysis. The relative levels of YnF versus YnGG labelling of CXXX and RAB proteins identified and quantified in the study are summarised in Figure 32. Additionally, the ratios of 29 non- prenylation motif proteins quantified were averaged and are presented as a comparison (denoted “No Motif”).

Figure 32. Relative comparison of YnF versus YnGG incorporation into CXXX and RAB proteins. EA.hy926 cells were cultured in heavy or light SILAC medium and labelled with either YnF or YnGG probe (25 µM) for 24 hours. The lysates were combined in a 1:1 ratio, subjected to CuAAC with capture reagent AzTB and enriched on NeutrAvidin beads. Following tryptic digest the proteins were analysed by LC-MS/MS on an LTQ-Orbitrap™ Velos mass spectrometer by Dr Andrew Bottrill at the University of Leicester proteomics facility. The data was processed in Maxquant. The data were filtered to retain proteins identified with a minimum of 2 razor & unique peptides in at least one replicate. *Average log2 ratio of 29 proteins with no prenylation motif. Red = RAB proteins. Blue = CXXX proteins. Error bars show standard deviation of 2 replicates. Of the set of prenylation motif proteins quantified only DNAJA2, a known substrate of FTase, was preferentially labelled by YnF. The remaining proteins are known substrates of GGTase- 1 or GGTase-2, and in agreement with this showed a higher level of YnGG incorporation. The non-prenylation motif proteins were equally labelled by both probes, consistent with equal levels of background labelling in all samples. Although none of the putative novel prenylated proteins discussed within this chapter were quantified in this experiment, the results obtained were encouraging as it highlighted the selectivity of our probes for either farnesylated or geranylgeranylated proteins. One caveat in the data set was the lack of proteins preferring YnF labelling. However, as relative quantification relies on the identification of a protein in both samples, proteins that exclusively incorporate only one of the probes cannot be quantified as no SILAC pair would appear in the mass spectrum. A schematic explaining this principle is outlined in Figure 33. below. Furthermore, the data were analysed on a LTQ-Orbitrap Velos

85 instrument which, as previously discussed, may not achieve sufficient sensitivity and mass resolution for identification of all SILAC pairs, particularly if one peak is of very low abundance.

Figure 33. Principle of relative quantification of probe incorporation from SILAC pairs. This schematic shows the different types of SILAC pairing that might be obtained in a duplex-SILAC experiment. Quantification is only possible when a SILAC pair is present (B-D), requiring some incorporation of both probes. In cases where proteins are exclusively modified by YnF or YnGG only (A and E), only one peak of the SILAC pair is detected rendering quantification impossible. It is clear that this experimental approach can provide valuable information about the isoprenoid preference of prenylated proteins in cases where proteins can be detected by both the YnF and YnGG probe. Although the results are likely to broadly represent the prenylation patterns observed in vivo, some care should be taken in interpreting the data. The detection of geranylgeranylated proteins with YnF and farnesylated proteins with YnGG could be due to a number of reasons. It may represent a true reflection of prenylation in intact cells and the high sensitivity of our method is enabling detection of very low levels of farnesylation or geranylgeranylation that have not previously been observed with less sensitive techniques such as radiolabelling. There is, however, the possibility that the addition of the exogenous isoprenoid analogues upsets the normal prenylation patterns in the cell. GGTase-1, for example, is able to use FPP as a substrate but only when its concentration reaches a critical level286. This possibility further points to the importance of using minimal concentrations of probe, as previously discussed. As highlighted in section 3.5, it cannot be ruled out that labelling originating from the YnF probe is a result of its elongation in the cell to YnGG.

3.7. Development of Novel Capture Reagents to Identify Probe-Modified Peptides

As previously discussed, enrichment and tryptic digest of the probe-labelled proteins leads to loss of the C-terminal modified peptide as it remains tethered to the beads through the probe/capture reagent modification. This peptide contains important information about the site of modification on the protein, and the nature of the modification itself. Identification of this peptide could answer several key questions:

 Are the probes acting on target, i.e. labelling cysteine residues within a prenylation motif? This would serve to validate the probes themselves in addition to providing a means to distinguish putative novel prenylated proteins from background labelling.

 What is the nature of the modification, i.e. is labelling originating from the administered probe or a metabolite thereof? This was particularly pertinent to answer with regards to YnF, which may undergo elongation by GGPS.

A key strategy to recover the modified peptide from the beads is to introduce a cleavable linker in the capture reagent and several such reagents already exist in the literature. Most of these reagents rely on an additional cleavage step by chemical or orthogonal enzymatic means in addition to protein digest by trypsin287, 288. Work in the Tate group has in recent years led to the development of capture reagents incorporating a trypsin cleavage site, allowing release of the modified peptide and tryptic digest of the protein in one step. These reagents have successfully been applied to profile myristoylated proteins in human cancer cells186 and in zebrafish289.

In parallel with on-going work in the group, 3 novel capture reagents bearing trypsin cleavage sites were designed and synthesised as part of this project. Figure 34 shows the structures of the reagents developed in this project compared to standard capture reagent AzTB and the cleavable reagent AzRTB developed by Dr Goska Broncel186. AzRB and AzKB incorporate an arginine and lysine sequence respectively for recognition by trypsin. AzK+RB, in addition to an arginine cleavage site, was designed such that a trimethyl-lysine residue would remain tethered to the modified peptide upon release. It was envisaged that the introduction of this charged residue could improve detection on the mass spectrometer by improving its ionisation. As the capture reagents were intended for use in proteomics applications, no fluorophore was incorporated to make the synthesis easier and more cost-effective.

Figure 34. Structure of trypsin-cleavable capture reagents. AzKB, AzRB and AzK+RB were designed and synthesised as part of this project. AzRTB was developed by Dr Goska Broncel186. The trypsin cleavage site is indicated by a red arrow. The capture reagents were synthesised using standard Fmoc-solid phase peptide synthesis (SPPS) on a biotin-conjugated resin as outlined in Scheme 5. After initial deprotection of the Fmoc-protected resin, the amino acid building blocks were sequentially coupled using standard HATU/DIPEA or HBTU/DIPEA chemistry. In the final step, azido-acetic acid was added to introduce the azido moiety. The reagents were cleaved from the resin using trifluoroacetic acid and purified by preparative RP-HPLC. The reagents were recovered in moderate yields of 26%, 37% and 26% for AzRB, AzKB and AzK+RB respectively.

Scheme 5. Synthesis of trypsin-cleavable capture reagents. The reagents were synthesised by standard Fmoc-solid phase peptide synthesis on a biotin-modified resin. The reagents were cleaved from the resin by treatment with TFA and purified by preparative RP-HPLC. In initial tests, the labelling efficiency of the capture reagents compared to standard reagent AzTB was assessed by Western blot using NeutrAvidin-HRP for detection of the biotin moiety. EA.hy926 cell lysate tagged with YnGG or vehicle control (DMSO) only was subjected to click chemistry with the novel capture reagents. Figure 35 shows the comparison of labelling efficiency of the reagents as assessed by NeutrAvidin Western blots. All the reagents appeared to label proteins in the 20-25kDa region, consistent with the molecular weight of geranylgeranylated RAB and Rho proteins. The labelling pattern is comparable to that achieved by the in-gel fluorescence results presented in Chapter 2 (section 2.8). AzRB and AzK+RB exhibit the strongest labelling efficiency, albeit also with higher background labelling. NeutrAvidin Western blot analysis is not as reliable or sensitive as in-gel fluorescence and additionally detects endogenously biotinylated proteins. Furthermore, detection is dependent on the quality of the protein transfer from the gel to the membrane. In this case the stronger labelling with AzRB and AzK+RB may be a reflection of improved transfer efficiency rather than superior conjugation of the capture reagent to the YnGG-tagged proteins during the click reaction. As the capture reagents were aimed to be used for proteomic applications we proceeded to validate their utility in chemical proteomics applications.

Figure 35. Test of novel capture reagents. EA.hy926 cells tagged with YnGG (or DMSO control) was lysed and subjected to CuAAC with standard capture reagent AzTB or novel, trypsin cleavable reagents AzRB, AzKB or AzK+RB. The protein samples were separated by SDS-page, transferred to a PVDF membrane and biotin-labelled protein detected by NeutrAvidin-HRP. A separate SDS-PAGE gel of the samples was stained with Coomassie Blue to assess control for total protein concentration across all samples. 3.8. Proof of Concept: Detection of Myristoylated Peptides in HeLa Cells with Novel Capture Reagents

To assess the ability of the novel capture reagents to detect modified peptides, an initial proteomics experiment was performed to identify myristoylated proteins. Myristoylation is a co- and post-translational attachment of the C14 fatty acid myristate to an N-terminal glycine residue of target proteins. This modification has been widely studied in the Tate group using an alkyne myristic acid analogue (YnMyr)290 and this modification can be robustly detected using cleavable capture reagents186, 289. As such, this modification was deemed a suitable model system for an initial proof-of-concept study.

YnMyr-labelled HeLa lysate was obtained from Dr Julia Morales-Sanfrutos, and processed for proteomic analysis using the novel capture reagents. Two replicate samples were prepared for each capture reagent. In addition, capture reagent AzTB was included as a non-cleavable control and AzRTB as a positive control as it had successfully enabled the identification of myristoylated peptides in previous studies186. A list of modified peptides containing a motif for myristoylation (N-terminal glycine) detected by the various reagents is given in Table 7 below. A ‘Y’ denotes the identification of a modified peptide, and a green colour indicates a protein

90 identification with a minimum of 3 razor and unique peptides in at least one sample. Additionally, the set of modified peptides was compared to those identified in previous studies by Thinon et al.186 and Broncel et al.289 using a trypsin-cleavable capture reagent.

AzRB and AzRTB identified 13 modified peptides each. AzK+RB and AzKB identified 8 and 1 peptides respectively. The superior performance of the arginine-containing reagents over those with a lysine cleavage site has been noted previously186, 289.

Modified ID? peptide?

Protein Name Modified Peptide Sequence AzK+RB AzKB AzRB AzRTB AzTB Thinon Broncel A-kinase anchor protein 12 Y Y Y GAGSSTEQR Y Y ADP-ribosylation factor 3 Y GNIFGNLLK Y Y ADP-ribosylation factor-like protein 1 Y GGFFSSIFSSLFGTR MICOS complex subunit MIC19 Y GGTTSTR Y Y MICOS complex subunit MIC25 Y GSTESSEGR Y Niban-like protein 1 Y Y Y GDVLSTHLDDAR Y Y Guanine nucleotide-binding protein G(i) subunit alpha-2 Y Y GCTVSAEDK Y Y Guanine nucleotide-binding protein G(k) subunit alpha Y Y Y GCTLSAEDK Y Y Golgi reassembly-stacking protein 2 Y Y GSSQSVEIPGGGTEGYHVLR Y Y Regulator complex protein LAMTOR1 Y Y GCCYSSENEDSDQDREER Y Y Protein phosphatase 1A Y Y GAFLDKPK Y Y Protein phosphatase 1B Y Y GAFLDKPK Y Y Protein phosphatase 1G Y Y Y GAYLSQPNTVK Y Y Calcineurin subunit B type 1 Y GNEASYPLEMCSHFDADEIKR Y Y cAMP-dependent protein kinase catalytic subunit alpha Y Y Y GNAAAAK Y Y 26S protease regulatory subunit 4 Y Y Y Y GQSQSGGHGPGGGK Y Y RING finger protein 141 Y GQQISDQTQLVINK Y Y

Total number of modified peptides 8 1 13 13 0 Table 7. Myristoylated peptides identified by LC-MS/MS analysis using novel capture reagents. A minimum of 3 “razor and unique” peptides were required in at least one sample for a positive identification. For modified peptide identification a Maxquant delta score >20 was required. Protein identifications are highlighted in green and modified peptide identifications are indicated by ‘Y’. n=2. Myristoylated peptides identified in previous work by Thinon et al and Broncel et al are indicated by ‘Y’. The number of modified peptides identified in this experiment was low compared to the 36 peptides discovered by Thinon et al. using reagent AzRTB. The comparatively modest number of modified peptides may be due to the small scale of the experiment. Only 2 replicates prepared from 100 µg protein per sample were analysed compared to 0.6-2 mg per sample utilised in the previous report. Furthermore, a double precipitation protocol after CuAAC

91 optimised for AzRTB was used in the preparation of all samples, which may have resulted in loss of labelled proteins and thus lower detection rates.

Based on these initial proteomics results we chose AzRB for use in our further work. AzRB was subsequently employed in a separate study by Dr Megan Wright to identify myristoylated proteins in the in the human pathogen Leishmania donovani187. In this study, 20 modified peptides were identified, including known myristoylated proteins such as LdARL1.

3.9. Optimising the Post-CuAAC Precipitation Protocol for AzRB

The general protocol employed for CuAAC of probe-tagged protein lysate involves a precipitation step after CuAAC to remove excess capture reagent. This decreases background in in-gel fluorescence imaging when fluorophore-containing capture reagents are used, and prevents excess capture reagent from blocking binding sites on the NeutrAvidin beads during affinity enrichment. As reagent AzRTB required a double precipitation step to sufficiently reduce background, this protocol was employed in the initial proteomic experiment with the novel capture reagents as discussed in the previous section. However, as we chose to use AzRB in our further work optimisation we needed to optimise the precipitation protocol for this reagent.

A number of standard methods of protein precipitation were trialled, including methanol/chloroform, methanol only and acetone precipitation as outlined in Figure 36. The first step of methanol/chloroform precipitation involves a centrifugation step where the protein pellet settles at the interface between an aqueous and organic layer. The standard protocol involves removal of the top layer only, leaving the protein pellet intact (lane 1). However, we chose to also investigate the effect of removing all solvent (lane 2). Methanol only (lane 3) and acetone (lane 4) precipitation involve an overnight incubation at -80°C and -20°C respectively. After the precipitation step the proteins were re-suspended and subjected to immunoblot analysis. Based on the results presented in Figure 36 it appeared that standard methanol/chloroform precipitation ensured the optimal recovery of labelled proteins, and this procedure was used in all subsequent experiments where AzRB was utilised.

Figure 36. Comparison of different precipitation methods on the recovery of labelled proteins after click reaction with AzRB. YnGG-tagged EA.hy926 lysate was subjected to bio-orthogonal ligation to AzRB and precipitated using methanol/chloroform with removal of top layer after first centrifugation step (1), methanol/chloroform with removal of both layers after first centrifugation step (2), methanol at -80°C overnight (3) and acetone at -20°C overnight (4). After resuspension, the protein samples were separated by SDS-page, transferred to a PVDF membrane and labelled protein detected by NeutrAvidin-HRP. The membrane was subsequently re-probed for α-tubulin to assess protein loading. 3.10. Applying AzRB for the Detection of Prenylated Peptides

Confident that our novel capture reagent was able to detect probe-modified peptides in an established model system we turned our attention to the detection of prenylated peptides. As discussed, the application of this methodology to identify prenylated peptides is limited by the C-terminal sequence of prenylated proteins. For example, KRAS has a lysine-rich C-terminal sequence (8 C-terminal residues: KIKKCIIM), meaning that if correctly processed by RCE1 and trypsin the modification would remain attached to a single cysteine amino acid. Even in cases where trypsin cleavage yields a peptide of sufficient length, C-terminal peptides are notoriously challenging to detect by mass spectrometry. A big advantage of using trypsin for protein digest is that it yields peptides with a lysine or arginine residue at the peptide C-terminus, leading to good ionisation and hence enhanced detection on the mass spectrometer. The peptides generated from protein C-termini are however lysine- and arginine-depleted, making detection more difficult291. Furthermore, many prenylated proteins undergo cleavage of the 3 C-terminal amino acids by RCE1 and subsequently carboxyl methylation of the cysteine by ICMT. These processing steps further complicate database searching and data analysis.

We reasoned that we would be more likely to identify peptides modified by YnF, as the longer and more hydrophobic YnGG would potentially lead to poorer ionisation characteristics. Thus, our initial efforts focused on this modification. However, preliminary analysis using the Maxquant software coupled to the Andromeda search engine did not result in the detection of any modified peptides.

Further work was performed by Dr Julia Morales-Sanfrutos using the PEAKS software suite292. PEAKS combines de novo sequencing and database searches to identify proteins from the mass spectrometry data. The de novo sequencing function makes it particularly amenable to detection of protein modifications and protein processing as it does not solely rely on matching peptides to a defined sequence database. By means of YnF and capture reagent AzRB, a number of probe-modified peptides were identified from labelling experiments in EA.hy926 cells. The identified peptides are listed in Table 8 below and a selection of spectra are displayed in Figure 37. Spectra for all modified peptides identified are available in Appendix C.

% Sequence Gene name Modified peptide ID CAAX motif coverage DNAJA1 GGVQCQTS CQTS 66 TQSPQNC CSIM 66 LMNA† TQSPQNC-Me LMNB2 GCYVM CYVM 48 NAP1L1 KDQNPAECK CKQQ 46 NAP1L4 (isoform 2) KEPSQPAECK CKQQ 29 PTP4A1 NNCCIQ CCIQ 52 RAP2C QDQCCTTC CVVQ 73 RHOA SGCLVL CLVL 81 ULK3 SSCTLQ CTLQ 31 Table 8. Modified peptides identified using prenyl probe YnF and capture reagent AzRB. The probe-modified cysteine is highlighted in red. †For LMNA both a non-methylated and carboxyl methylated peptide was found. CAAX motif refers to C-terminal prenylation motif. Results from Dr Julia Morales-Sanfrutos. Interestingly, a combination of Rce-1-processed and non-processed peptides were identified. LMNA and RAP2C were only identified in their RCE1-processed form, whereas the remaining peptides retained an intact C-terminus. In the case of NAP1L1 and NAP1L4, the lysine contained within the CAAX-motif most likely resulted in cleavage by trypsin during the sample preparation. The identification of the non-processed peptides raises a number of important questions. It is widely accepted that all prenylated CAAX-proteins undergo processing by RCE1 and ICMT7 although this has not been experimentally validated for all prenylated proteins. An important aspect to consider is whether the YnF modification interferes with post- prenylation processing. However, this is an unlikely scenario as LMNA appeared to be fully processed by both RCE1 and ICMT. Inspection of the identified sequence of the non- processed peptides is informative. Deletion of the –AAX residues would in most cases result in very short peptide that would not enable confident assignment to a unique protein sequence, which explains their absence from the analysis. It is therefore likely that the non-processed peptides originate from a low population of proteins at the transient stage between isoprenoid attachment and RCE1 cleavage stage of maturation.

Figure 37. Example spectra obtained from identification of modified peptides from LMNA, DNAJA1 and ULK3. The characteristic peaks resulting from the probe-capture reagent adduct are highlighted. Data was processed in PEAKs. Data from Dr Julia Morales-Sanfrutos. The identification of YnF-modified peptides sheds light on several important questions. Firstly, detection of the probe modification acts as direct proof that our probe is acting on target as a

95 farnesyl analogue. Secondly, the detection of ULK3 and NAP1L4 C-terminal peptides suggests that these are bona fida substrates and not the result of non-specific background labelling. Thirdly, the detection of YnF-labelled RHOA suggests that proteins that are traditionally considered as targets of geranylgeranylation only may in fact incorporate low levels of farnesyl that have escaped detection in previous studies using less sensitive methods such as radiolabelling. However, it is still unclear whether this farnesylation is relevant, or an effect of a non-physiologically high level of farnesyl analogue in the cell. Furthermore, it is not known whether YnF is acting as substrate of FTase of GGTase-1.

Although only a small number of prenylated peptides were identified compared to the data sets reported for e.g. myristoylation, these initial results are highly encouraging. As discussed, there are several inherent challenges associated with the detection of prenylated C-terminal peptides. We envisage that further studies, perhaps using a larger complement of proteases for the digestion step, will result in further identifications. Due to the inherent challenges of detecting C-terminal peptides from complex samples, adapting workflows to selectively enrich these peptides to enhance mass spectrometric signal should be explored.

3.11. Conclusions

This chapter outlines the use of our prenyl probes in combination with proteomic technologies to explore prenylation on a global scale. Initial comparisons of YnF, AzC15 and AzC20 highlighted the superior selectivity of YnF for detection of farnesylated proteins. Furthermore, this preliminary study strengthened our argument to abandon the use of azido probes due to higher levels of background labelling.

In addition to a large number of known targets, several novel putative farnesylated proteins, including LRRFIP1, ULK3, DCAF8, DPCD, EHBP1, EHBP1L1, NAP1L4 and CEP85 were identified by YnF. Our YnGG probe detected the putative geranylgeranylated proteins OAS1 and UBL3 as well as a large number of known targets of GGTase-1 and RABGGTase. Comparison of the YnF and YnGG datasets revealed distinct as well as over-lapping targets, highlighting the utility of a 2-probe system to maximise coverage of both farnesylated and geranylgeranylated proteins.

Our proteomics workflow was further developed to incorporate SILAC-labelling methods to enable quantification of the prenylome by mass spectrometry. A concentration and time- course study was devised to explore the optimal conditions detection of prenylated proteins with the YnF and YnGG probes. The majority of the identified CXXX-proteins and all RAB proteins showed a concentration-dependent increase in labelling with both probes. The rate of probe incorporation was more variable, probably due to the dependence on protein turn-

96 over times which differ across the population of prenylated proteome. To better understand the selectivity of our probes for farnesylated versus geranylgeranylated proteins we conducted a study to quantify the relative incorporation of each probe. Encouragingly, the YnF probe was preferred over YnGG by DNAJA2, a known substrate for farnesylation. In contrast, the remaining data set consisted of known geranylgeranylated proteins which all preferentially incorporated the YnGG probe.

We also describe the development of three novel capture reagents, AzRB, AzKB and AzK+RB, which incorporate a trypsin-cleavable linker to enable the release of probe-modified peptides. In a proof-of-concept study, all 3 reagents were able to detect myristoylated peptides from HeLa lysate fed with an alkyne-myristate analogue. AzRB gave the largest number of identifications and was chosen for use in further studies. Further work by Dr Julia Morales- Sanfrutos led to the discovery of several YnF-modified peptides. Importantly, the identification of the ULK3 and NAP1L4 C-terminal peptides indicates that these proteins are indeed novel farnesylated proteins.

In the following chapter we build on the proteomics methods described here-in to further validate our methods and the putative novel prenylated proteins identified in our initial studies. We also apply our quantitative workflow to explore the dynamics of protein prenylation in response to a variety of prenyl transferase inhibitors.

Chapter 4: De Novo Identification of Prenylated Proteins and Prenyl Transferase Inhibitor Dynamics

4.1. Introduction

A key strength of chemical proteomics methodologies for profiling PTMs is their utility in unbiased de novo identification of cellular targets and substrates. By combining chemical probes and mass spectrometric analysis, a proteome wide survey can be performed in one experiment. This stands in stark contrast to traditional biochemical methods such as radiolabelling which is limited to assessing one defined protein substrate at a time. Chemical proteomics is also a powerful method to profile the cellular targets of enzyme inhibitors. For example, a combination of alkyne-tagged myristic acid analogue and quantitative proteomics has been used to aid the development of N-myristoyl transferase inhibitors185.

In excess of 70 clinical trials have been conducted with inhibitors of the prenyl transferases37, most targeting FTase, yet no clinical application has been approved. To explain these past failures and direct further research, it is imperative that we gain a full understanding of all the cellular substrates of the prenyl transferases, which substrates are affected upon enzyme inhibition and in particular which substrates undergo alternative prenylation upon inhibitor treatment.

This chapter describes the application of our probes to validate a number of putative novel farnesylated substrates. Furthermore, we apply the probes in comprehensive experiments to decipher the cellular targets of a number of FTIs, as well as inhibitors or GGTase-1 and RABGGTase. We show-case how our probes can be used to determine an in-cell dose response to a number of widely used prenyl transferase inhibitors and their value in validating the on/off target effects of novel inhibitors.

4.2. Defining the Prenylome

Our initial proteomics profiling presented in Chapter 3 suggested that our prenyl probes were able to identify a number of novel prenylated substrates. YnF, in particular, labelled a number of putative CXXX-motif proteins. To validate whether these proteins were genuine prenylated substrates or the result of background labelling we designed a quantitative proteomics competition experiment (Figure 38). The rationale of the experimental approach was that the labelling of genuine substrates would be sensitive to competition by the natural isoprenoid substrates. Thus, we aimed to quantify loss of labelling in the presence of natural isoprenoids by mass spectrometry.

Figure 38. Schematic outlining the experimental approach to quantify isoprenoid competition by SILAC quantitative proteomics. R10K8-labelled (‘heavy’) EA.hy926 cells were cultured in medium supplemented with probe (YnF or YnGG) and the natural isoprenoid substrate (25 µM FOH or 10 µM GGOH) for 24 hours. Based on in-gel analysis (Figure 24, Chapter 2) and an initial proteomics experiment (data not shown) it was evident that YnGG was more sensitive to competition than YnF labelling, and hence GGOH was added at a lower concentration than FOH. R0K0-labelled (‘light’) control cells were grown in the presence of prenyl probe only. Following lysis, the heavy and light cells were combined in a 1:1 ratio for each probe. The lysates were subjected to CuAAC with capture reagent AzRB and labelled proteins enriched on NeutrAvidin beads. Following tryptic digest the eluted peptides were analysed by LC-MS/MS. The results of the data analysis are depicted in Figure 39.

We chose to focus on proteins where labelling exhibited a log2 fold change of <-2, equivalent to a >75% decrease in labelling in response to isoprenoid competition. Of the proteins that met this criterion, YnF-labelling of 23 proteins and YnGG-labelling of 40 proteins were significantly reduced by isoprenoid competition (p<0.05). All of these proteins harboured a motif for prenylation and all but one have previously been described as substrates of the prenyl transferases. Disappointingly, of the putative novel prenylated substrates presented in Chapter 3, only CEP85 was identified and quantified in our study. The reason for the relatively poor coverage of prenylation motif proteins identified in this study compared to our preliminary studies in Chapter 3 is unclear.

Figure 39. Quantitative analysis of YnF and YnGG labelling upon isoprenoid competition. R10K8- labelled EA.hy926 cells were cultured in medium containing (A) 10µM YnF + 25 µM FOH or (B) 10 µM YnGG + 10µM GGOH for 24 hours. R0K0-labelled cells were labelled with YnF or YnGG only (10 µM). 4 replicates of each condition were prepared. R10K8 and R0K0-labelled cells were mixed in a 1:1 ratio for each probe and subjected to CuAAC with capture reagent AzRB. Labelled proteins were immobilised on NeutrAvidin resin and subjected to tryptic digest. Peptide samples were analysed by LC-MS/MS on a Q Exactive™ instrument. Data was processed in Maxquant and statistical analysis performed in Perseus. A minimum of 3 “razor+unique” peptides in at least one replicate was required. Only proteins quantified in at least 3 replicates were retained. A one-sample t-test (p<0.05 or p<0.01) was performed to identify proteins with a log2 H/L ratio significantly different from 0. ULK3, which was identified in preliminary proteomics experiments (Chapter 3) as a potential novel farnesylated protein, was not quantified in the proteomics competition analysis. To address the potential farnesylation of this protein further immunoblot analysis was performed. Cell lysate labelled with YnF or YnGG with or without addition of natural isoprenoid substrate was subjected to CuAAC and pulldown on NeutrAvidin beads. Bound proteins were analysed by Western blot to detect ULK3. As evident in Figure 40 ULK3 is robustly detected by the YnF probe, and this labelling is sensitive to competition by FOH. ULK3 was absent from YnGG- labelled and DSMO control samples. This analysis, coupled to the identification of the probe- modified C-terminal peptide (Chapter 3) and the sensitivity of YnF-labelling to farnesyl transferase inhibition (see subsequent sections), leads us to define ULK3 as a novel farnesylated protein, in addition to CEP85.

100

Figure 40. Labelling of ULK3 is sensitive to competition by natural substrate. EA.hy926 cells were cultured in medium containing 10 µM YnF +/- 25 µM FOH, 10 µM YnGG +/- 10 µM GGOH or vehicle control (DMSO) for 24 hours. The cell lysate was subjected to CuAAC with AzTB and labelled proteins were immobilised on NeutrAvidin beads. The purified proteins were subjected to Western blot analysis to detect ULK3. The input blots shows presence of ULK3 in samples not subjected to affinity purification. 4.3. Profiling the Cellular Targets of Prenyl Transferase Inhibition

Inhibition of protein prenylation, and farnesylation in particular, has been widely studied as a potential therapy for cancer and numerous other pathologies37. Several FTIs (Tipifarnib, Lonafarnib, L-788.123 and BMS-214662) and one GGTase-1 inhibitor (GGTI-2148) have been assessed in clinical trials, either as a single agent or in combination therapy. However, thus far no clinical application has received approval. A major issue surrounding the use of PTIs is our incomplete knowledge of exactly which cellular targets are affected by inhibition, and patterns of alternative prenylation occurring as a consequence of selective inhibition of one prenyl transferase enzyme. In this section we sought to apply our quantitative proteomics workflow to assess the cellular targets of a variety of PTIs.

We chose to profile 3 FTIs (FTI-277, Manumycin A and Tipifarnib), one GGTase-1 inhibitor (GGTI-2133) and one RABGGTase inhibitor (Bon-15). FTI-277 and Manumycin A were chosen on the basis that they are commonly cited tools used to study FTase inhibition in cells, and Tipifarnib, which is the FTI that has entered the largest number of clinical studies, for its clinical relevance. GGTI-2133 is also widely cited in the literature as a probe for GGTase-1 inhibition. Bon-15 was recently reported as a first example of a selective RABGGTase inhibitor, and as such we wanted to further explore its cellular targets. The structure of the inhibitors tested in this project and all PTIs that have entered clinical trials are depicted in Figure 41 below.

101

Figure 41. Chemical structure of prenyl transferase inhibitors. A number of FTase inhibitors (Tipifarnib, Lonafarnib, L-788.123 and BMS-21466) and one GGTase-1 inhibitor (GGTI-2418) have undergone clinical trials. FTI-277, Manumycin A and GGTI-2133 are widely used as a tool to probe protein prenylation. Bon-15 was recently reported as a potent and selective inhibitor of RABGGTase. 4.3.1. In-gel Studies

In initial experiments we sought to address whether the labelling achieved by our probes YnF and YnGG was sensitive to inhibition by PTIs. As attachment of YnF and YnGG is mediated by the prenyl transferase enzymes, the addition of selective inhibitors should result in loss of labelling. Initial inhibition studies were assessed by in-gel fluorescence scanning of labelled proteins. Briefly, EA.hy926 cells were pre-incubated with inhibitor or vehicle control for 1-2 hours, followed by addition of YnF or YnGG. Cells were lysed and capture reagent AzTB was appended to probe-labelled targets by CuAAC. Proteins were then separated by SDS-PAGE and visualised by in-gel fluorescence.

102

FTase inhibitor FTI-277

Farnesyl transferase inhibitor FTI-277 is peptidomimetic methyl ester pro-drug of the potent

293 inhibitor FTI-276 . It inhibits HRAS farnesylation in cells with an IC50 of 200 nM, and is approximately 100-fold selective for FTase over GGTase-1294. FTI-277 does, however, cause changes in cellular levels of some prenylated proteins, for example by inducing the expression of RHOB in a large number of cancer cell lines295.

The complex dynamics of prenylation in response to FTIs is immediately evident from our in- gel fluorescence assessment of FTI-277 inhibition (Figure 42A). Although several bands labelled by YnF are lost upon treatment with FTI-277 (annotated by blue asterisk), indicating inhibition of farnesylation, other bands appear more strongly (red asterisk). Several explanations could account for this observation, such as transfer of this probe onto proteins that are normally geranylgeranylated. The accumulation of YnF by inhibition of FTase could potentially lead to an increased use of YnF by GGTase-1 or RABGGTase, both of which have some ability to use farnesyl pyrophosphate as a substrate74, 286. Alternatively, the accumulated pool of YnF is elongated by GGPPS, to yield YnGG in cells, and thus generating an alkyne- tagged substrate for the geranylgeranyl transferases. The enhanced bands could also correspond to prenylated proteins that are up-regulated in response to FTI-277.

Figure 42. The effect of farnesyl transferase inhibitor FTI-277 on labelling by YnF and YnGG in EA.hy926 cells. (A) EA.hy926 cells were pre-incubated with FTI-277 or vehicle control for 2 hrs, followed by YnF (10µM) labelling for 24 hrs. Post-lysis the samples were subjected to click chemistry with capture reagent AzTB and separated by SDS-PAGE. Gels were imaged by in-gel fluorescence (channel Cy3 for TAMRA and Cy5 for protein marker). Subsequently, gels were Coomassie Blue stained to assess protein loading. (B) Labelled proteins were immobilised on NeutrAvidin beads. Purified proteins were subjected to Western blot analysis to detect farnesylated protein HRAS.

103

YnGG labelling is also affected by FTI-277. Interestingly, YnGG labelling of proteins corresponding to the molecular weight of small GTPases (20-25 kDa) is enhanced by addition of FTI-277. This may suggest that proteins which are normally farnesylated are instead labelled by YnGG in the presence of an FTI; for example, KRAS is known to be geranylgeranylated in response to FTase inhibition94, 95. It is also plausible that the enhanced fluorescence results from increased expression of geranylgeranylated proteins in response to FTI-277. As the same effect is not apparent upon treatment with farnesyl transferase inhibitor Tipifarnib (Figure 43), it is probable that the enhanced fluorescence is caused by increased protein expression rather than a switch in prenylation status.

Immunoblot analysis indicated that YnF labelling of known farnesylated protein HRAS was decreased in the presence of FTI-277 (Figure 42B). This result suggests that the YnF probe is suitable to detect changes in prenylation in response to PTI treatment.

FTase Inhibitor Tipifarnib

Tipifarnib (Zarnestra) is a potent, orally active FTI developed by Johnson & Johnson Pharmaceuticals296. Despite a large number of clinical trials, both as a single agent and in combination therapy, it has failed to gain approval due to lack of efficacy37.

As assessed by in-gel fluorescence (Figure 43), Tipifarnib potently inhibits YnF labelling even at the lowest concentration tested (5 nM). It does not inhibit the labelling of YnGG, as expected from a highly selective FTase inhibitor. Interestingly, in contrast to results achieved with FTI-277, Tipifarnib does not appear to enhance the labelling of some protein bands by YnF and YnGG. This difference suggests that the key contributor to the enhanced labelling pattern observed upon FTI-277 treatment likely results from changes in protein expression levels, and that any changes in prenylation status play a negligible role to the increased labelling observed in FTI-277 treated cells.

Similarly to FTI-277, YnF labelling of HRAS is inhibited by Tipifarnib (Figure 43C). Interestingly, in the presence of Tipifarnib, HRAS appears to be labelled by YnGG, suggesting that this protein becomes geranylgeranylated in the presence of a FTI. In contrast to KRAS and NRAS, HRAS has been reported to be strictly farnesylated, and is thought not to change prenylation status upon FTI treatment95. However, considering the high sequence similarity between HRAS, KRAS and NRAS, it is highly possible that the antibody used recognises KRAS and NRAS to some extent in addition to HRAS. Thus, the apparent YnGG labelling of HRAS in response to Tipifarnib-treatment could in fact result from geranylgeranylation of KRAS or NRAS. Further confirmation, for example by analysis of unique peptides by proteomics, is required to draw any further conclusions. In the proteomics analysis detailed

104 later in this chapter (section 4.4) we did not find evidence of YnGG-labelling of HRAS in response to FTI treatment.

Figure 43. Labelling with farnesyl probe YnF can be inhibited by farnesyltransferase inhibitor Tipifarnib. (A & B) EA.hy926 cells were pre-incubated with increasing concentrations of Tipifarnib for 1 hr. Subsequently, prenyl probe (YnGG or YnF, 10 µM) was added and the cells cultured for a further 8 hrs. Post-lysis the samples were subjected to click chemistry with capture reagent AzTB and separated by SDS-PAGE. Gels were imaged by in-gel fluorescence (channel Cy3 for TAMRA and Cy5 for protein marker). Subsequently, gels were Coomassie Blue stained to assess protein loading. (C) Labelled proteins were immobilised on NeutrAvidin agarose resin. The purified proteins were subjected to Western blot analysis to assess for presence of farnesylated protein HRAS. The ‘input’ blot shows the presence of proteins in all samples before purification. FTase Inhibitor Manumycin A

Manumycin A is an antibiotic isolated from Streptomyces, which also shows potency as a

297 farnesyl transferase inhibitor with an IC50 of 5 µM in vitro . In previous reports, inhibition of RAS prenylation was decreased by 45% by 30 µM Manumycin A in Hep G2 cells, a RAS- activated human hepatoma cell line298. However, we found that concentrations above 10 µM resulted in cell death and this limited the concentration range which we could test. The in-gel analysis of Manumycin A suggests that concentrations up to 10 µM do not affect YnF or YnGG-labelling of prenylated proteins (Figure 44).

105

Figure 44. The effect of prenyl transferase inhibitor Manumycin A on labelling by YnF and YnGG in EA.hy926 cells. Cells were incubated with Manumycin A (2.5 or 10 µM) or vehicle control for 1 hr and subsequently labelled with (A) YnF or (B) YnGG (10 µM) for a further 8 hrs. Post-lysis the samples were subjected to click chemistry with capture reagent AzTB and separated by SDS-PAGE. Gels were imaged by in-gel fluorescence (channel Cy3 for TAMRA and Cy5 for protein marker). Subsequently, gels were Coomassie Blue stained to assess protein loading. GGTase-1 Inhibitor GGTI-2133

GGTI-2133 is a selective inhibitor of GGTase-1, with an IC50 of 38 nM in vitro and 10 µM in cells. Enzyme assay studies indicate that GGTI-2133 is approximately 140-selective for GGTase-1 over FTase299. Our in-gel analysis showed few effects on YnGG labelling upon addition of GGTI-2133 (Figure 45). However, the utility of in-gel analysis is limited due to the large number of RAB proteins that are also labelled by YnGG. Inhibition by GGTI-2133 is likely to affect predominantly geranylgeranylated RHO proteins, which migrate to the same molecular weight region as the RAB proteins (20-25 kDa). As RABs are prenylated by RABGGTase and thus would be unaffected by addition of inhibitor, they likely mask any changes in labelling due to GGTI-2133.

In both YnF- and YnGG-labelled cells, GGTI-2133 causes an increase in one protein band. As the same band appears to be affected in both samples, it is likely that this is a result of altered protein expression rather than a switch to farnesylation.

106

Figure 45. The effect of GGTase-1 inhibitor GGTI-2133 on labelling by YnF and YnGG in EA.hy926 cells. EA.hy926 cells were pre-incubated with GGTI-2133 or vehicle control for 2 hours, followed by YnF or YnGG (10µM) labelling for 24 hours. Post-lysis the samples were subjected to CuAAC with capture reagent AzTB and separated by SDS-PAGE. Gels were imaged by in-gel fluorescence (channel Cy3 for TAMRA and Cy5 for protein marker). Subsequently, gels were Coomassie Blue stained to assess protein loading. RABGGTase Inhibitor Bon-15

Bon-15, recently reported as a highly potent and selective inhibitor of RABGGTase, was obtained from Professor Herbert Waldmann (Max Planck Institute of Molecular Physiology, Dortmund)300. Bon-15 was developed by structure-guided design from dual FTase and RABGGTase inhibitor, BMS3301.

Our in-gel analysis depicted in Figure 46 suggested that Bon-15 had a more pronounced effect on YnF labelling than YnGG labelling. As discussed in the previous section, geranylgeranylation is mediated by both GGTase-1 and RABGGTase and thus inhibition of the latter would only affect a portion of YnGG-modified proteins. The YnGG-labelled protein bands unaffected by Bon-15 most probably represent proteins that are substrates of GGTase- 1.

107

Figure 46. The effect of RABGGTase inhibitor Bon-15 on labelling by YnF and YnGG in EA.hy926 cells. EA.hy926 cells were pre-incubated with Bon-15 or vehicle control for 2 hrs, followed by YnF or YnGG (10µM) labelling for 24 hrs. Post-lysis the samples were subjected to click chemistry with capture reagent AzTB and separated by SDS-PAGE. Gels were imaged by in-gel fluorescence (channel Cy3 for TAMRA and Cy5 for protein marker). Subsequently, gels were Coomassie Blue stained to assess protein loading. The loss of YnF labelling evident in response to Bon-15 is highly analogous to the pattern observed upon treatment with inhibitors of FTase, suggesting that Bon-15 may also be inhibiting this enzyme. The inhibitor validation carried out by Bon et al.300 showed that the inhibitor was selective for RABGGTase over FTase in vitro. However, although they assessed inhibition of RAB prenylation in cells, they did not assess the effects on FTase in a cellular setting. Bon-15 selectivity for RABGGTase over FTase is conferred by a carbamate-linked aryl group. Carbamates are commonly employed structural motifs in medicinal chemistry; however, they display varied rates of hydrolysis depending on substituents; aryl-OCO-NH- motifs are particularly prone to hydrolysis302. Thus, it is possible that metabolism by cellular enzymes results in carbamate hydrolysis, as outlined in Figure 47. Removal of the aryl carbamate group yields a compound (Bon-11) that in the same study was found to be a potent inhibitor of FTase.

108

Figure 47. Potential hydrolysis of Bon-15 carbamate. Cellular hydrolysis of selective RABGGTase inhibitor Bon-15 yields the highly potent FTI inhibitor Bon-11300. 4.3.2. Initial Proteomic Evaluation of Inhibitor Response

A key aim of this project was to enable the in-cell evaluation of response to prenyl transferase inhibition by unbiased proteomics analysis. As shown in the preceding sections detailing in- gel analysis, YnF and YnGG labelling is affected by various PTIs. Based on these encouraging results we sought to investigate whether the changes apparent in our in-gel analysis could be quantified on a protein basis through quantitative mass spectrometry.

A spike-in SILAC experiment was designed to assess whether we could quantify changes in prenylation in response to PTIs. One inhibitor of each of the three prenyl transferases, FTI- 277, GGTI-2133 and Bon-15, were chosen for this initial study. Briefly, EA.hy926 cells were pre-treated with inhibitor (10 µM) or vehicle control for 2 hours after which YnF or YnGG was added directly to the inhibitor-containing growth medium and the cells cultured for a further 24 hours. Cell lysates were mixed (1:1) with YnF- or YnGG-labelled R10K8 spike-in lysate and processed for proteomics analysis by CuAAC, enrichment and on-bead digest. After data processing, the SILAC ratios of the inhibitor-treated sample were normalised to the non- inhibitor-treated samples. The resultant log2 fold changes in response to inhibitor treatment are summarised in Table 9 (CXXX proteins) and Table 10 (RAB proteins).

In agreement with our in-gel analysis, the effect of specific inhibition of one prenyl transferase enzyme yields both increases and decreases in labelling of CXXX-motif and RAB proteins. YnF labelling of several known farnesylated substrates such as RHEB and nuclear lamins (LMNA, LMNB1, LMNB2) was sensitive to FTase inhibition by FTI-277. Labelling of EHBPL1, a putative novel farnesylated substrate, was also sensitive to inhibition by FTI-277, suggesting that this protein is a substrate for FTase. Several other putative substrates such as CEP85, DCAF8, EHBP1, LRRFIP1, ULK3 and DCAF8 were not quantified in samples treated with FTI- 277, possibly due to complete depletion of YnF labelling in response to inhibition. As they

109 were detected in samples treated with inhibitors of the other prenyl transferases (and inhibited by Bon-15), this suggests that they may also be bona fida FTase substrates.

Table 9. Initial quantification of prenylation of CXXX-proteins in response to prenyl transferase inhibitors. Log2(L/H ratios) were normalised to non-treated controls. Data was filtered to retain proteins quantified in at least one inhibitor sample in addition to non-treated controls. A minimum of 3 razor+unqiue peptides were required in non-inhibitor treated samples. n=1.

Colour code (log2 L/H ratio): x ≤ - 2 -2 < x ≤ -1 -1 < x ≤ -0.5 0.5 ≤ x < 1 x > 1

YnF YnGG Gene names FTI-277 Bon-15 GGTI-2133 FTI-277 Bon-15 GGTI-2133 NAP1L1 -3.939 -3.542 0.048 -0.319 -0.099 0.011 PEX19 -3.831 -2.075 -0.006 EHBP1L1 -3.787 -2.014 0.338 RHEB -3.692 -1.657 0.153 BROX -3.500 -2.335 0.344 DNAJA2 -3.493 -1.292 0.053 -0.663 0.777 0.195 DNAJA1 -3.152 -1.306 0.066 0.002 0.922 0.404 NAP1L4 -3.111 -3.353 0.113 YKT6 -2.940 -1.184 0.093 -3.085 -1.957 0.069 LMNB1 -2.834 -1.627 0.340 INF2 -2.505 -0.911 0.226 LMNB2 -2.238 -0.783 0.230 LMNA -1.821 -0.975 0.310 0.019 0.787 -0.083 PTP4A2 -1.647 -0.682 0.121 HRAS -1.561 -1.468 0.128 KRAS -0.975 -1.241 0.212 2.530 3.051 RRAS2 -0.843 -1.238 0.593 0.063 0.801 -2.785 RALB -0.705 -0.884 -1.180 -1.721 -0.373 -3.010 PTP4A1 -0.495 -0.018 0.322 RAP2C;RAP2A -0.223 -0.651 0.056 UCHL1 0.077 0.109 0.008 0.013 0.216 0.010 RHOG 0.711 -0.774 0.993 -0.228 -0.238 -0.329 CNP 0.714 -0.756 1.350 -0.303 -0.438 -1.054 RAC1;RAC3 0.742 -0.805 0.820 -0.133 -0.214 -0.353 RAP2B 0.750 -0.302 1.034 -0.237 -0.162 -0.644 RAP1B;RAP1A 0.797 -0.502 0.316 -0.205 -0.224 -0.823 CDC42 0.993 -0.674 1.459 -0.081 -0.264 0.076 RRAS 1.034 -0.338 0.744 -0.115 -0.290 -0.868 RHOA 1.077 -0.500 1.185 0.208 -0.170 -0.011 RAC2 1.271 -0.389 1.021 0.018 -0.104 -0.264

110

Table 9. cont.

Colour code (log2 L/H ratio): x ≤ - 2 -2 < x ≤ -1 -1 < x ≤ -0.5 0.5 ≤ x < 1 x > 1

YnF YnGG Gene names FTI-277 Bon-15 GGTI-2133 FTI-277 Bon-15 GGTI-2133 GNG5 1.685 -0.690 1.967 0.180 -0.234 0.060 RHOB 2.593 0.428 2.285 1.195 0.015 0.760 CEP85 N.D. -2.673 0.094 DCAF8 N.D. -3.406 0.101 DPCD N.D. -3.718 -0.029 EHBP1 N.D. -2.062 0.112 GBP1 N.D. N.D. 0.131 LRRFIP1 N.D. -2.875 0.049 RAP2A N.D. -0.534 0.255 TRIM25 N.D. -0.324 N.D. ULK3 N.D. -1.060 0.043 ZC3HAV1 N.D. N.D. 0.464 DNAJB2 -2.149 -0.122 -3.071 MIEN1 -1.747 -0.207 -2.521 RHOF -1.427 -0.529 -2.241 RALA -1.208 -0.195 -2.205 GNG12 0.041 0.109 -0.477 CYB5R3 0.055 -0.230 0.024 GNAI2;GNAI1; 0.940 0.765 0.828 GNAO1;GNAI3 FBXL20 N.D. 0.486 N.D.

Interestingly, both FTI-277 and GGTI-2133 resulted in increased YnF labelling of several substrates of GGTase-1, such as RHOA, RAC2 and RHOB in particular. As discussed, RHOB expression is known to be induced by FTI-277 and this result could thus be a reflection of increased protein levels. An analysis of total protein levels is required to draw further conclusions about this observation, as the increased labelling could also be due to incorporation of YnF by GGTase-1 (FTI-277-treated sample) or FTase-mediated YnF labelling (GGTI-2133-treated samples).

YnGG labelling of several known geranylgeranylated proteins such as RRAS2 and MIEN1 were sensitive to GGTI-2133 inhibition. Interestingly, a number of farnesylated substrates which were labelled by YnGG, such as YKT6, were also sensitive to FTI-277 inhibition. This phenomenon was also observed in further studies detailed later in this chapter. This could

111 mean that FTase is able to mediate attachment of YnGG to some of its substrates. Alternatively, it could indicate that these substrates are partly prenylated by GGTase-1, and FTI-277 is acting as an inhibitor of this transferase enzyme.

As previously suggested, Bon-15 appears to act as a dual FTase and RABGGTase inhibitor in cells. Although it showed potent inhibition of YnGG labelling of RAB proteins (Table 10), Bon-15 also inhibited YnF labelling of a large number of FTase substrates. This result highlights the discrepancy between inhibitor behaviour in vitro and in a more complex cell- based system and showcases the utility of our method in the development and validation of prenyl transferase inhibitors.

Interestingly, although YnGG labelling of the majority of RAB proteins was sensitive to inhibition by Bon-15, RAB8A and RAB8B were largely unaffected. Both proteins harbour a C-terminal CAAX motif and RAB8A has previously been shown to undergo FTase and GGTase-1-mediated prenylation in vitro85. RAB8A was detected in YnF-labelled samples, and this labelling was slightly inhibited by FTI-277 and GGTI-2133, suggesting that RAB8A prenylation may in part be mediated by the CAAX prenyl transferases in cells.

This initial analysis was only based on one replicate sample, and thus limited conclusions can be drawn from this data set. More comprehensive analysis of selected PTIs follows in the subsequent section.

Table 10. Initial quantification of prenylation of RAB proteins in response to prenyl transferase inhibitors. Log2(L/H ratios) were normalised to non-treated controls. Data was filtered to retain proteins quantified in at least one inhibitor sample in addition to non-treated controls. A minimum of 3 razor+ unique peptides were required in non-inhibitor treated samples. n=1. RAB proteins containing a CXXX motif are annotated as follows: (*) CCXXX or XCCXX, (**) XCXXX. (a) Note: In YnF-labelled samples RAB2A and RAB2B could not be differentiated.

Colour code (log2 L/H ratio): x ≤ - 2 -2 < x ≤ -1 -1 < x ≤ -0.5 0.5 ≤ x < 1

YnF YnGG Gene names FTI-277 Bon-15 GGTI-2133 FTI-277 Bon-15 GGTI-2133 RAB3B -0.687 N.D. -0.369 -0.496 -4.508 0.021 RAB12 -0.105 -4.436 0.182 RAB3D -0.179 -4.028 0.271 RAB5B* -0.307 N.D. -0.345 0.318 -3.982 0.180 RAB2B* (a) -0.633 -3.405 -0.394 -0.138 -3.942 0.107 RAB33B* 0.020 -3.906 0.104 RAB22A -0.073 -3.828 0.052 RAB4B 0.227 -3.780 -0.276 RAB1A -0.587 -2.219 -0.250 -0.223 -3.751 0.062

112

Table 10 cont.

Colour code (log2 L/H ratio): x ≤ - 2 -2 < x ≤ -1 -1 < x ≤ -0.5 0.5 ≤ x < 1

YnF YnGG Gene names FTI-277 Bon-15 GGTI-2133 FTI-277 Bon-15 GGTI-2133 RAB43 -0.415 -3.750 0.095 RAB35 -0.615 -1.132 -0.381 -0.218 -3.698 0.006 RAB30* -0.432 -3.628 0.104 RAB13** -0.234 -3.611 -0.012 RAB23** -0.120 -3.608 -0.007 RAB14 -0.671 -2.878 -0.459 -0.197 -3.599 0.070 RAB34 -0.167 -3.551 0.088 RAB29 0.112 -3.544 0.046 RAB32 0.042 -3.531 0.183 RAB4A 0.067 -3.274 0.034 RAB11A* -0.113 -2.914 0.104 RAB27A -0.094 -2.798 0.095 RAB2A (a) -0.633 -3.405 -0.394 -0.169 -2.681 0.028 RAB31 0.091 -2.588 -0.212 RAB9A 0.018 -2.455 0.034 RAB11B* -0.509 -4.684 -0.303 0.026 -2.073 0.115 RAB5A* -0.822 -1.714 -0.692 0.026 -2.049 0.097 RAB6B -0.142 -2.035 0.069 RAB5C* -0.234 -3.651 -0.468 -0.072 -1.980 0.003 RAB10 -0.654 -1.184 -0.335 -0.096 -1.889 0.168 RAB7A -0.330 -2.891 -0.432 0.150 -1.837 0.054 RAB6A -0.499 -1.461 -0.227 0.165 -1.668 0.153 RAB1B -0.718 N.D. -0.407 -0.311 -1.545 0.039 RAB18** -0.994 -0.488 -0.636 -0.272 -1.468 0.079 RAB21* -0.536 -2.710 -0.469 0.102 -0.860 0.133 RAB8B** -0.114 -0.442 0.122 RAB8A** -0.842 0.550 -0.609 -0.383 -0.244 0.021 RAB27B -0.470 N.D. -0.201 RAB39A -0.242 N.D. -0.079 RAB24* -0.136 N.D. 0.014

113

4.4. Evaluating an In-Cell Dose Response to Prenyl Transferase Inhibition

As our initial evaluation suggested that our quantitative proteomics workflow was capable of quantifying changes in prenylation in response to varied inhibitors of the prenyl transferases, we next embarked on more comprehensive profiling experiments. To this extent we employed a spike-in SILAC quantitative approach to establish an in-cell dose response to a number of FTIs (FTI-277, Manumycin A and Tipifarnib) and the GGTase-1 inhibitor GGTI-2133. Due to limited compound availability we were unable to carry out further work with Bon-15, and as our initial studies indicated that Bon-15 was not specific for RABGGTase we did not endeavour to re-synthesise this compound.

To establish an in-cell inhibitor dose response, EA.hy926 cells were pre-treated with increasing concentrations of inhibitor followed by YnF or YnGG labelling. Cell lysates were combined with probe-labelled R10K8 spike-in lysate as before to enable relative quantification of YnF and YnGG incorporation across all samples. Following sample processing and LC- MS/MS analysis, SILAC ratios were normalised to non-inhibitor-treated samples to access a relative inhibitor response.

A total of 100 proteins harbouring a prenylation motif were detected in the FTI-277 dataset, and 80 prenylation motif proteins in the Tipifarnib dataset. Statistical analysis (one-way ANOVA, Benjamini-Hochberg truncation, FDR = 0.05) was performed to identify proteins that exhibited a change in labelling upon inhibitor treatment. Subsequently the datasets were further filtered to retain proteins that showed a maximum relative inhibitor response <0.4, resulting in a list of 20 and 28 proteins for the FTI-277 and Tipifarnib sample sets respectively.

IC50 curves were fitted to the data to obtain an in-cell dose response. 2 proteins from the FTI- 277-treated sample set and 3 proteins from the Tipifarnib-treated sample set were discarded as they returned an ambiguous IC50 value. The dose response curves for FTI-277- and Tipifarnib-treated samples labelled with YnF are depicted in Figure 48 below.

In addition to known FTase substrates, the novel substrates CEP85, DCAF8, EHBPL1 and LRRFIP1 showed a dose-dependent inhibition in response to FTI-277 and Tipifarnib. Additionally, a dose-response for EHBP1 and ULK3 was obtained in Tipifarnib-treated samples. This robust inhibitor response adds further validity to the assignment of these proteins as novel farnesylated proteins.

114

Figure 48. In-cell dose-response of YnF-labelling by FTase inhibitors FTI-277 and Tipifarnib. EA.hy926 cells were pre-incubated with increasing concentrations of inhibitor or vehicle control for 1 hour, followed by YnF (10µM) labelling for 8 hrs. Post-lysis the samples were subjected to click chemistry with capture reagent AzRB, enriched on NeutrAvidin resin and digested by trypsin. Eluted peptides were analysed by LC-MS/MS on a Q Exactive™ instrument and data processed in Maxquant and Perseus. Proteins identified by at least 2 “razor & unique” peptides in at least 2 out of 3 of the non-inhibitor-treated samples were retained. L/H ratios were normalised to non-inhibitor-treated controls. Data were filtered to retain ANOVA significant hits (Benjamini-Hochberg truncation, FDR=0.05) with a relative response <0.4. Error bars show standard error of the mean (n=3). Tipifarnib-treated sample were prepared by Dr Julia Morales- Sanfrutos.

IC50 values of novel farnesylated proteins were derived from the dose-response curves and are summarised in Table 11 below.

115

Table 11. IC50 values of putative novel farnesylated proteins.

Gene names Tipifarnib (nM) FTI-277 (µM) CEP85 0.261 0.294 DCAF8 0.149 1.62 EHBP1L1 1.63 1.24 LRRFIP1 0.124 0.425 ULK3 6.54 Ambiguous EHBP1 1.55 Ambiguous

As a further validation step we assessed the labelling of two selected proteins, ULK3 and DCAF8, by Western blot. As apparent in Figure 49, YnF-labelling of both proteins showed a dose-dependent decrease upon increasing amount of inhibitor. Consistent with the absence of these proteins in YnGG-labelled proteomics samples, neither protein was detected in samples labelled with this probe, nor in DMSO control samples.

Figure 49. YnF-labelling of farnesylated proteins ULK3 and DCAF8 is sensitive to inhibition by Tipifarnib. EA.hy926 cells were pre-incubated with increasing concentrations of Tipifarnib (5, 10, 50, 100 nM) or vehicle control for 1 hr. Subsequently, prenyl probe (YnGG or YnF, 10 µM) was added and the cells cultured for a further 8 hrs. Post-lysis the samples were subjected to click chemistry with capture reagent AzTB and incubated with NeutrAvidin agarose resin to isolate labelled proteins. The purified proteins were subjected to Western blot analysis to assess the presence of ULK3 and DCAF8. The ‘input’ blot shows the presence of proteins in all samples before purification. We next assessed the effect of increasing concentrations of Manumycin A treatment on YnF labelling. Our statistical analysis (multi-sample ANOVA test, FDR=0.05) revealed no changes in the amount of protein labelled for any inhibitor concentration. A selection of known and novel farnesylated substrates are depicted in Figure 50, showing no significant change in prenylation upon increasing concentrations of Manumycin A. These proteomics results are consistent with the lack of effect observed by in-gel fluorescence. The in vitro IC50 of Manumycin A is 5 µM, and thus to achieve an in-cell response a much higher concentration

116 of Manumycin A may be required. However, we did not test concentrations above 10 µM as this resulted in cell toxicity.

Figure 50. Manumycin A treatment does not affect YnF labelling of selected known and novel FTase substrates. EA.hy926 cells were pre-incubated with increasing concentrations of Manumycin A or vehicle control for 1 hour, followed by YnF (10µM) labelling for 8 hrs. Post-lysis the samples were subjected to click chemistry with capture reagent AzRB, enriched on NeutrAvidin resin and digested by trypsin. Eluted peptides were analysed by LC-MS/MS on a Q Exactive™ instrument and data processed in Maxquant and Perseus. L/H ratios were normalised to non-inhibitor treated controls. A multi-sample ANOVA test (Benjamini-Hochberg truncation, FDR=0.05) did not detect any significant changes in labelling in response to Manumycin A. Error bars show standard error of the mean (n=3). A number of reports show that Manumycin A induces a number of toxic effects independent of any FTI activity, such as generation of reactive oxygen species303, 304. Inspection of the chemical structure of Manumycin A (Figure 41) reveals a number of Michael acceptor sites. Michael acceptors can react with cysteine thiols via conjugate addition to form covalent adducts. Indeed, Manumycin A has been shown to inhibit IκB kinase β activity in a cysteine- dependent manner by formation of covalent homodimers which the authors suggested were mediated by Manumycin A cross-linking305. The same study showed that Manumycin A covalently binds glutathione in vitro, highlighting its reactivity towards free cysteine thiols. A further report has shown that Manumycin A acts as an irreversible inhibitor of neutral sphingomyelinase306. Based on the lack of in-cell FTI activity observed in our study and the off-target effects demonstrated by others, we suggest that Manumycin A is an unsuitable tool for studying FTase inhibition, especially as highly potent and clinically relevant compounds such as Tipifarnib are readily available (albeit more expensive).

Next we explored the effects of prenyl transferase inhibitors on labelling by YnGG. Figure 51 shows the dose-response of proteins that showed a significant decrease (multi-sample ANOVA test, FDR=0.05) in YnGG labelling, with a maximum relative response of <0.5. 12

117 known substrates of GGTase-1 showed a dose-dependent decrease in labelling in response to GGTI-2133. Interestingly, and in line with our initial proteomics experiments, the labelling of a number of proteins was also inhibited by FTI-277, including both substrates of FTase (e.g. YKT6) and GGTase-1 (e.g. MIEN1). It is unclear from these results whether FTI-277 is affecting GGTase-1 or FTase activity.

Figure 51. In-cell dose-response of YnGG-labelling to inhibition by GGTI-2133 and FTI-277. EA.hy926 cells were pre-incubated with increasing concentrations of inhibitor or vehicle control for 1 hour, followed by YnGG (10µM) labelling for 8 hrs. Post-lysis the samples were subjected to click chemistry with capture reagent AzRB, enriched on NeutrAvidin resin and digested by trypsin. Eluted peptides were analysed by LC-MS/MS on a Q Exactive™ instrument and data processed in Maxquant and Perseus. Proteins identified by at least 2 “razor & unique” peptides in at least 2 out of 3 of the non-inhibitor-treated samples were retained. L/H ratios were normalised to non-inhibitor treated controls. Data was filtered to retain ANOVA significant hits (Benjamini-Hochberg truncation, FDR=0.05) with a relative response <0.5. Error bars show standard error of the mean (n=3). Finally, by means of YnGG labelling we were also able to show the utility of our method in deciphering the dynamics of prenylation in response to FTI inhibition. As discussed, the failure of FTI treatment to target RAS-driven cancers has been partly attributed to alternative geranylgeranylation of proteins that are normally farnesylated. As depicted in Figure 52, we were able to detect a significant increase in YnGG labelling of KRAS and RHOB in response to FTI treatment. This result is in agreement with previous studies showing increased levels of geranylgeranylation of these proteins upon FTase inhibition. No other FTase substrates identified behaved in this manner. However, as our analysis relied on comparison to a non- inhibitor-treated control, other proteins displaying this behaviour may not have been quantified due to lack of YnGG labelling in the absence of FTase inhibition. Further optimisation of our method, perhaps by using a spike-in dually labelled with YnF and YnGG to enable

118 quantification of all prenylated proteins in all experimental conditions, may result in identification of more proteins that switch prenylation status upon FTI treatment.

Figure 52. FTI-277 treatment increases YnGG labelling of farnesylated proteins KRAS and RHOB. EA.hy926 cells were pre-incubated with increasing concentrations of FTI-277 or vehicle control for 1 hour, followed by YnGG (10µM) labelling for 8 hrs. Post-lysis the samples were subjected to click chemistry with capture reagent AzRB, enriched on NeutrAvidin resin and digested by trypsin. Eluted peptides were analysed by LC-MS/MS on a Q Exactive™ instrument and data processed in Maxquant and Perseus. L/H ratios were normalised to non-inhibitor treated controls. Data was filtered to retain ANOVA significant CXXX-protein hits (Benjamini-Hochberg truncation, FDR=0.05) with a relative response >1.5. Error bars show standard error of the mean (n=3). 4.5. Cellular Roles of Putative Novel Farnesylated Substrates

DCAF8

DDB1- and CUL4-associated factor 8 (DCAF8/WDR42A) is a member of the WD40 repeat family of proteins. The WD40-repeat proteins bind the DDB1 component and act as recruiting factors for the DDB1-CUL4A-ROC1 E3 ubiquitin ligase machinery307, which is a key regulator of cell proliferation and survival and DNA repair308. DCAF8 is predominantly located to the nucleus, and its nuclear localisation is dependent on the small GTPase RAN309. Cell localisation studies shows that the C-terminally tagged DCAF8-EYFP fusion protein showed identical localisation as the N-terminally tagged HA-DCAF8 fusion309. As the EYFP tag fused onto the C-terminus of WDR42A will likely make prenylation impossible, it appears that farnesylation of DCAF8 does not impact its cellular localisation.

A recent study showed that DCAF8 promotes CDC25A ubiquitination and degradation310. CDC25A is a key regulator of cell cycle progression and cell apoptosis311, and is commonly over-expressed in cancer. The stability of CDC25A is tightly regulated throughout the cell cycle, partly by ubiquitination by E3 ligases and subsequent proteasomal degradation311. Overexpression of DCAF8 or DDB1 led to increased ubiquitination of CDC24A, and knock-

119 down of DCAF8 or DDB1 conversely decreased ubiquitination310. Thus, it appears that DCAF8 plays an important role in regulating the cellular levels of CDC25A throughout the cell cycle.

ULK3

ULK3 is a serine-threonine kinase which was recently shown to play a role in cytokinetic abscission312. When cells physically divide, an abscission checkpoint is in place to ensure that division does not take place until all chromosomes have physically segregated into the two daughter cells. The ESCRT-III complex is a key regulator of the physical separation of cells and of the abscission checkpoint. Caballe et al.312 showed that ULK3 interacts with several proteins in the ESCRT-III complex, in particular IST1. ULK3 phosphorylates IST1, and this phosphorylation event is required to sustain the checkpoint. ULK3-IST1 binding is mediated through a MIT domain, but does not require the C-terminal residues 450-472. Thus, it appears that farnesylation is not required for ULK3-IST1 binding.

ULK3 also plays a role as a positive regulator of the Hedgehog (HH) signalling pathway313, 314. The HH pathway plays a role in embryonic development, tissue homeostasis and tumourigenesis315. Modulators of HH signalling are under active investigation as therapeutics for a number of cancers316. The potential role of ULK3 farnesylation in HH signalling is unclear. The C-terminal domain of ULK3 is shown to have an inhibitory effect on HH signalling, whereas the kinase domain has an activating effect313, 314. If farnesylation is important to the role of the C-terminal domain, loss of farnesylation could result in enhanced activation of the HH pathway, which could promote tumourigenesis.

CEP85

The cellular function of CEP85 (CCDC21) has not been widely studied. Recent research showed that CEP85 acts as a binding partner for NIMA-related kinase 2A (NEK2A) and co- localises with NEK2A at centrosomes317. NEK2A plays a key role in promoting centrosome disjunction during mitosis by stimulating degradation of centrosomal linker proteins. Disjunction of the centrosomes is required for bipolar spindle formation. The bipolar spindle serves to separate the duplicated chromosomes into the two daughter cells318. CEP85 antagonises the effects of NEK2A, preventing premature centrosome disjunction. The role of CEP85 farnesylation is unclear, as mutational studies indicated that the C-terminal region of the protein was not required for centrosome localisation and NEK2A binding, nor for its ability to inhibit NEK2A activity317.

120

LRRFIP1

LRRFIP1 was also detected exclusively by YnF in our preliminary proteomics experiments, but was not quantified in the SILAC isoprenoid competition experiment. We were unable to obtain a working antibody for this protein, and as such were not able to perform immunoblot analysis to further validate its prenylation status. LRRFIP1 labelling was, however, sensitive to FTI inhibition in a dose-dependent manner, strongly indicating that this protein is farnesylated. The function of LRRFIP1 (GCF2/TRIP) has been under active investigation in the recent years and appears to play a role in several pathological processes, including cancer, cardiovascular disease and infection.

LRRFIP1 is expressed in most tissues of the human body319, 320, and highly expressed in certain cancer cell lines such as Burkitt’s lymphoma319. Genome-wide screening analysis estimated that LRRFIP1 mutations were present in 11% of human breast cancer tumours321. LRRFIP1 can bind both dsRNA322 and dsDNA323 and acts as a transcriptional repressor for the epidermal growth factor receptor (EGFR)324, platelet-derived growth factor A chain (PDGF- A)325, tumour necrosis factor alpha (TNF-α)323, 326 and a common polymorph of the glutamate transporter EAAT2 gene327.

LRRFIP1 appears to play a role in cancer invasion and metastasis, at least in part by controlling the activity of RHO GTPases. For example, LRRFIP1 interacts with Dishevelled (DVL)328, 329, a protein that regulates canonical Wnt/β-catenin and non-canonical Wnt signalling. Non-canonical Wnt signalling regulates the activity of small GTPases RHO, RAC and CDC42, which are key regulators of the actin cytoskeleton, cell migration, polarity, and adhesion330, 331 and play a key role in cancer initiation and progression332. LRRFIP1 specifically mediates Wnt-induced RHOA activation and cell migration328 and interacts with LARG, a RhoGEF protein which controls RHOA activation333. LRRFIP1 also interacts with the leucine rich repeat domain of the human flighless-1 homologue (FLII), an actin-binding protein that regulates the actin cytoskeleton and co-localises with small GTPases during cell migration320, 322, 334. Douchi et al.329 found that knock-down of LRRFIP1 disrupted the epithelial– mesenchymal transition (EMT) through the Wnt pathway. EMT leads to loss of cell-cell adhesion in epithelial cells, a hallmark process of cancer invasion and metastasis. LRRFIP1 was found to be highly expressed at the invasion front of pancreatic cancers and knock-down reduced the cancer cells’ invasive ability.

In addition to its role in cancer, LRRFIP1 may also be important in cardiovascular disease. Excessive cell proliferation is a key process in the development of atherosclerosis. MicroRNA- 132 silencing of LRRFIP1 was found to attenuate vascular smooth muscle cell proliferation and conversely, over-expression of LRRFIP1 led to increased cellular proliferation335. In a rat

121 model of arterial injury, microRNA-132 was found to reduce LRRFIP1 expression and reduce proliferation and migration of smooth muscle cells in the artery walls335.

Rupture of unstable atherosclerotic lesions in vessel walls results in thrombus (clot) formation. The thrombus may lead to occlusion of blood flow if it lodges in an artery, causing myocardial infarction (MI) or stroke. Platelets, a key component of blood, function to stop bleeding by aggregating into clots and are thus key to thrombus formation. A genome-wide expression study recently identified LRRFIP1 as a positive regulator of platelet function and a risk gene for MI336. Knockdown of LRRFIP1 in zebrafish significantly reduced thrombus formation following laser injury to the caudal artery. Further interactome analysis indicated that LRRFIP1 may exert effects on platelet function by interacting with regulators of the actin cytoskeleton336.

Finally, LRRFIP1 was also found to play a role in activating the innate immune response by inducing type I interferons (IFNs) in response to bacterial and viral infections337, 338. It was suggested that LRRFIP1 acts as a cytosolic sensor of dsRNA and dsDNA from pathogens. In response to infection LRRFIP1 interacts with β-catenin and promotes its activation and nuclear localisation, ultimately promoting expression of type I IFNs337.

4.6. Conclusions

This chapter describes the application of our chemical proteomics workflow to validate novel prenylated substrates and decipher the cellular effects of prenyl transferase inhibition. By means of isoprenoid competition analysis and selective FTase inhibition we propose that DCAF8, ULK3, CEP85, LRRFIP1, EHBPL1 and EHBP1 constitute a set of previously un- identified farnesylated proteins.

Our probes YnF and YnGG in combination with quantitative proteomics methods such as SILAC provide a comprehensive means to profile changes in protein prenylation in response to varied inhibitors of the different prenyl transferases. A strength of the method is its ability to investigate prenylation in an intact cell system, which is particularly important from a medicinal chemistry perspective to validate novel inhibitors and any potential off-target effects. For example, we show that the recently reported RABGGTase inhibitor Bon-15, which is highly selective in vitro, does in fact inhibit FTase in the cell. As discussed, this is probably a result of cellular metabolism leading revealing a metabolite which shows activity against FTase.

We further assessed the FTase inhibitor Manumycin A, and showed that concentrations up to 10 µM do not inhibit farnesylation. Coupled to off-target effects described by others and the toxic effects we observed at higher concentrations we advise against the use of this compound as a FTI in cell studies.

122

A key reason for the failure of FTIs in clinical trials is due to alternative geranylgeranylation of normally farnesylated substrates. Our quantitative analysis revealed that these dynamics can be described by our labelling system, as exemplified by increased incorporation of YnGG in KRAS and RHOB in response to FTI-277 treatment. Further optimisation of our methods may reveal further substrates that behave in this manner, as the current workflow is limited to quantifying proteins that are also detected in non-inhibitor-treated control samples. Thus, proteins that incorporate no YnGG in the absence of inhibitor treatment were not quantified.

We envisage that our methodology will be widely applicable to studying prenylation in both health and disease. The following chapter describes the use of our workflow to study prenylation in two disease systems: HGPS and hypoxia-induced PH.

123

Chapter 5: Applications in Disease Models and In Vivo

5.1. Introduction

Although prenylation has been most widely studied in the context of cancer, it plays a key role in numerous other diseases as outlined in Chapter 1. In this project we aimed to look closer at the role of prenylation in two pathologies: pulmonary hypertension (PH) and Hutchinson- Gilford progeria syndrome (HGPS).

PH is associated with remodelling of the pulmonary vasculature and affects both endothelial and smooth muscle cells140. Pathological features of PH include endothelial dysfunction, excessive proliferation of vascular smooth muscle cells and an imbalance in vasodilatory mediators. Increased resistance in the pulmonary vasculature leads to right ventricular hypertrophy, and eventually heart failure in affected patients. RHO GTPases are key mediators of vascular tone and remodelling and the role of RHOA/ROCK pathway in particular has been extensively studied in the context of PH. ROCK inhibitors are currently being assessed in clinical trials339.

Recent work in the Wojciak-Stothard group has focused on the role of RHOB in the development of hypoxia-induced PH143. Their studies showed that RHOB is upregulated in response to acute hypoxia in human pulmonary arterial smooth muscle cells (HPASMCs) and endothelial cells (HPAECs). Furthermore, RHOB stabilised hypoxia-inducible factor-1α (HIF- 1α), which is a key transcription factor regulating the cell’s adaptive response to low oxygen levels. They also showed that, in vivo, the development of chronic hypoxia-induced PH was attenuated in RHOB-/- mice. Recent in vivo work conducted by the group indicated that inhibition of FTase by Tipifarnib treatment may provide a novel means to target PH (Duluc et al., manuscript in preparation).

In contrast to RHOA which is a substrate for GGTase-1, RHOB can be either farnesylated or geranylgeranylated on its C-terminal CAAX motif266. Some studies suggest that the prenylation status of RHOB may affect its localisation and function in the cell, with farnesylated RHOB associating with the cell plasma membrane and geranylgeranylated RHOB localising to endosomes141. To better understand the role of RHOB and other prenylated proteins in the context of PH we sought to apply our prenyl probes to study prenylation in the pulmonary vasculature.

In this chapter we also describe the application our chemical proteomics methodology to probe prenylation in cells from HGPS patients. HGPS is a rare genetic disease that causes

124 accelerated ageing in affected patients, who typically do not live beyond their teenage years110. The unique phenotype makes HGPS an attractive model system to study ageing. The key cause of HGPS is thought to be nuclear accumulation of farnesylated progerin, a mutant version of LMNA. Low levels of progerin can also be detected in non-HGPS subjects and is thought to play a role in vascular pathology in the ageing population133. A recent clinical trial indicated that FTIs may ameliorate symptoms and improve life-span in HGPS patient129, 130 In light of the positive outcomes of the trial we sought to gain a better understanding of the role of prenylation in HGPS and to determine which proteins were affected by FTase inhibition.

Finally, this chapter also describes attempts to employ our probes to label prenylated proteins in in vivo. To this extent we performed labelling experiments in zebrafish. The ability to study prenylation in complex animal models would provide unique opportunities to access information about the role of prenylation in health and disease.

5.2. In-gel Labelling Studies in HPASMCs and HPAECs

Pulmonary arteries are complex structures, composed of several layers of cells and connective tissue, as shown in Figure 53. The pathophysiology of PH involves both dysfunction of the endothelial cells, which line the inner lumen of the artery, and smooth muscle cells, which compromise the middle layer of the artery. Hence, both cell types are of interest when studying mechanisms of the pulmonary vasculature.

Figure 53. The structural layers of human arteries. Arteries are composed of three layers: the tunica interna, tunica media and tunica adventitia. The tunica interna, the innermost layer, is composed of a layer of endothelial cells lining the lumen of the artery and a basement membrane. The middle layer, the tunica media, is mainly composed of smooth muscle cells surrounded by an elastic membrane. The tunica adventitia is the outermost layer of the artery and is composed of connective tissue. 5.2.1. Optimising Labelling of Prenylated Proteins in HPASMCs and HPAECs

As a first step we aimed to optimise the labelling conditions in HPAECs and HPASMCs by conducting time-course and concentration studies. The cells were cultured in medium supplemented with 10 µM YnF or YnGG for 8-48 hours (HPAECs) or 24-96 hours (HPSMCs).

125

Cell lysates were subjected to CuAAC with AzTB, separated by SDS-PAGE and labelled proteins visualised by in-gel fluorescence (Figure 54 and Figure 55). In both cell types robust labelling with both the YnF and YnGG probe was apparent after 24 hours, and maximal labelling achieved after 48 hours of incubation time. We chose not explore any further time- points beyond 48 hours in HPAECs as these cells require frequent media changes to sustain growth (every 2-3 days). In the HPASMC time-course the medium was replaced after 48 hours for the 72 and 96 hour time-point samples.

Figure 54. Optimisation of YnF and YnGG labelling conditions in HPAECs. (A) HPAECs were cultured in medium supplemented with prenyl probe (10 µM) or vehicle control for the indicated time periods. (B) HPAECs were cultured in medium supplemented with increasing concentrations of prenyl probe or vehicle control for 48 hours. Cell lysates were subjected by CuAAC with AzTB, separated by SDS-PAGE and visualised by in-gel fluorescence. Gels were stained by Coomassie Blue to assess protein loading. To further optimise the labelling protocol we conducted a probe concentration study. In both HPAECs and HPASMCs a probe concentration of 5 µM appeared optimal as judged by in-gel fluorescence (Figure 54B and Figure 55B). As previously observed in EA.hy926 cells (see chapter 2, section 2.8), labelling efficiency decreased at higher concentrations of prenyl probe.

126

Figure 55. Optimisation of YnF and YnGG labelling conditions in HPASMCs. (A) HPASMCs were cultured in medium supplemented with prenyl probe (10 µM) or vehicle control for the indicated time periods. Fresh probe containing medium was added to cells at 48 hours in the case of the 72 and 96 hour time-points. (B) HPASMCs were cultured in medium supplemented with increasing concentrations of prenyl probe or vehicle control for 48 hours. Cell lysates were subjected by CuAAC with AzTB, separated by SDS-PAGE and visualised by in-gel fluorescence. Gels were stained by Coomassie Blue to assess protein loading. 5.2.2. Attempts to Detect Farnesylated and Geranylgeranylated RHOB

RHOB is predominantly geranylgeranylated but can also be farnesylated266, and as such we aimed to use our two probes to detect differentially prenylated RHOB. In an initial experiment we employed the EA.hy926 cell line used in our previous labelling studies (Chapters 2-4). Following incubation with YnF or YnGG, EA.hy926 cells were lysed and subjected to immunoprecipitation to enrich RHOB. RHOB is expressed at low levels in unstimulated cells and as such we anticipated that an immunoprecipitation step would enhance our chances of detecting the protein. Previous attempt to visualise RHOB by biotin-mediated enrichment of all probe-labelled proteins followed by Western blot had not been successful (data not shown). The immunoprecipitate was subjected to on-bead CuAAC with capture reagent AzTB after which the bound proteins were eluted and separated by SDS-PAGE and the probe-labelled RHOB visualised by in-gel fluorescence. As evident in Figure 56A, YnGG enabled robust detection of RHOB. However, no signal was obtained with the YnF probe, which is probably a result of RHOB being preferentially geranylgeranylated.

127

In a second attempt to detect farnesylated RHOB, we switched to labelling in HPAECs as this was the cell type of interest in our further studies. RHOB levels are upregulated in response to acute hypoxia in HPAECs143 and hence we argued that hypoxic treatment might increase our chances of detecting RHOB. HPAECs were pre-incubated with YnF or YnGG for 24 hours followed by exposure to hypoxia (2% O2) for 2 hours. The pre-incubation period served to generate a pool of phosphorylated YnF and YnGG substrate in the cells prior to RHOB stimulation. Again, RHOB was robustly detected with YnGG, but not evident in YnF-labelled samples (Figure 56B).

Despite attempts to optimise the immunoprecipitation protocol (data not shown), we failed to detect RHOB with the YnF probe. As we had previous evidence that YnF-labelled RHOB could be detected by mass spectrometry, we decided to focus our efforts on this more sensitive method to dissect the prenylation of RHOB.

Figure 56. Attempts to detect YnF and YnGG-labelled RHOB. (A) EA.hy926 cells were cultured in medium supplemented with YnF, YnGG or vehicle (DMSO) for 24 hours. After cell lysis RHOB was immunoprecipitated and subjected to on-bead CuAAC with AzTB, followed by SDS-PAGE separation. Probe-labelled RHOB was visualised by in-gel fluorescence. Total RHOB was assessed by Western blot. (B) HPAECs were cultured in medium supplemented with YnF, YnGG or vehicle (DMSO) for 24 hours, followed to exposure to hypoxic (2% O2) or normoxic conditions for 2 hours. The lysates were processed as in (A). 5.3. Proteomics Studies in HPAECs and HPSMCs

5.3.1. Hypoxia-treated Cells

To decipher whether any changes in prenylated proteins occur in response to hypoxia, we performed a probe labelling experiment in hypoxia-treated HPAECs. Cells were incubated with

YnF or YnGG (5 µM) for 24 hours prior to hypoxic exposure (2% O2, 5% CO2) for 4 hours.

Normoxic controls (20% O2, 5% CO2) were prepared in an analogous manner. We chose the 4 hour time-point based on previous results showing that RHOB protein expression in response to acute hypoxia peaks at 2-4 hours143. For quantification purposes we prepared a probe-labelled SILAC spike-in standard from our EA.hy926 cell line. EA.hy926 cells are derived from human endothelial cells and as such we argued this would be suitable cell line to use for the spike-in standard. Following lysis, the HPAEC samples were mixed in a 2:1 ratio with spike-in lysate, and prepared for proteomics analysis using our standard protocol. The

128 samples were analysed by LC-MS/MS followed by data analysis in Maxquant to identify and quantify proteins present in the samples. The L/H ratios generated represent the relative amount of protein in the HPAEC samples versus the spike-in standard.

To assess changes in probe-labelled proteins in response to hypoxia, L/H ratios of each hypoxic sample were normalised to the L/H ratio of the corresponding normoxic control, generating a ratio representing the relative abundance in hypoxic versus normoxic samples. The normalised ratios were log2 converted and assessed in a one-sample t-test to identify any proteins that were differentially expressed in hypoxic conditions. The results from this analysis are shown in Figure 57 below.

Figure 57. The effect of hypoxia treatment on YnF and YnGG labelling in HPAECs. HPAECs were cultured in medium supplemented with 5 µM YnF (A) or 5 µM YnGG (B) for 24 hours, followed to exposure to hypoxic (2% O2, 5% CO2) or normoxic (20% O2, 5% CO2) conditions for 4 hours. 3 biological replicates per condition were prepared. Cell lysates were mixed 2:1 with a R10K8 SILAC spike-in standard prepared from YnF- or YnGG-labelled (10µM, 24 hours) EA.hy926 cells. Mixed lysates were subjected to CuAAC with capture reagent AzTB, enriched on NeutrAvidin resin and digested by trypsin. The eluted peptides were analysed by LC-MS/MS on a Q Exactive™ instrument. Data was processed in Maxquant and statistical analysis performed in Perseus. The L/H ratios of hypoxia-treated were normalised to the L/H ratios of normoxic controls. Normalised ratios were log2 converted and assessed in a one-sample t-test (p<0.05) to determine significant difference from 0. A minimum of 2 valid values per sample group were required. Closed circles: CXXX & RAB proteins, open circles: other proteins. RHOB is highlighted in blue.

In our initial analysis we focused on proteins that showed a significant (p<0.05) log2 fold change >1 or <-1. Only one protein fulfilled these criteria in the YnGG-labelled samples, and none in the YnF-labelled samples. The identified protein, COP9 signalosome complex subunit 1 (GPS1) does not harbour a prenylation motif. Its identification probably resulted from non- specific background labelling or co-purification by association with a prenylated protein.

129

Interestingly though, GPS1 is a regulator of hypoxia-inducible factor 1α (HIF-1α) and has previously been shown to be downregulated in response to hypoxia340.

Although we identified RHOB in both YnF- and YnGG-labelled samples we failed to detect any significant change of this protein in response to hypoxia. The previous study by Wojciak- Stothard et al. showed a doubling of RHOB protein levels after 4 hours of hypoxic exposure143. The reason for this discrepancy in results is unclear. The Wojciak-Stothard study analysed total cellular levels of RHOB, whereas our analysis was focused on the population of prenylated RHOB. It is possible that hypoxic upregulation of RHOB generates a pool of unprenylated protein which would not be quantified in our analysis. Another plausible explanation for the differences in results comes down to experimental design. For example, the lysis buffer employed in our study contained a lower detergent concentration. Poor release of the membrane-bound RHOB population may account for the lack of difference observed, as insoluble cellular material was discarded prior to sample processing.

One prenylation motif protein identified in the YnF-labelled samples, phospholipase D3 (PLD3), showed a slight but significant down-regulation in response to hypoxia. PLD3 harbours a CRLL motif, but the prenylation status of this protein has not been previously investigated. PLD3 was robustly detected in YnF-labelled samples (up to 10 unique peptides), but not in YnGG samples. This differential detection suggests that PLD3 was not merely identified as a result of non-specific background labelling and may be farnesylated or alternatively enriched through interaction with a farnesylated protein, although further validation is required to ascertain the prenylation status of PLD3. Other phospholipases such as cytosolic phospholipase A2-gamma (cPLA2γ) are, however, known to be farnesylated341, 342. The apparent downregulation of PLD3 observed in our analysis mirrors a previous study showing that PLD3 mRNA and protein levels are decreased in response to hypoxia343.

5.3.2. The Effect of Tipifarnib on YnF and YnGG Labelling in HPASMCs and HPAECs

Recent work in the Wojciak-Stothard group indicates that the farnesyl transferase inhibitor Tipifarnib may be a potential therapeutic for PH (Duluc et al., manuscript in preparation). Mice treated with Tipifarnib exhibited improved endothelial function and decreased vascular remodelling and right heart hypertrophy. To dissect which proteins may be involved in the beneficial effects of FTI treatment on the pulmonary vasculature we sought to profile changes in prenylation in Tipifarnib-treated HPSMCs and HPAECs.

In contrast to the spike-in SILAC methodology adopted for the study on hypoxia treatment, we instead chose to use dimethyl labelling for quantification purposes. The SILAC spike-in standard employed in the hypoxia study was generated in the endothelial cell line EA.hy926.

130

However, we argued that the EA.hy926 cells may not be representative of the proteome of smooth muscle cells. Dimethyl labelling circumvents the need for a spike-in standard, and in the interest of time we chose this approach rather than attempt to optimise the spike-in standard. The principle of dimethyl labelling is outlined in section 1.4.3 (Chapter 1).

Cells were pre-treated with Tipifarnib (100 nM) or vehicle control for 1 hour, followed by YnF or YnGG labelling for 24 hours. The protein lysates were subjected to CuAAC with capture reagent AzTB and enriched on NeutrAvidin beads. Following on-bead tryptic digest, the eluted peptides were subjected to on-stagetip dimethyl labelling, as per the protocol developed by Li and co-workers240. Non-inhibitor treated control samples were labelled as ‘light’ and Tipifarnib- treated samples as ‘medium’. Control and inhibitor-treated peptide samples were then combined in a 1:1 ratio and analysed by LC-MS/MS. The M/L ratios generated from data processing in Maxquant were log2 converted. A one-sample t-test (p<0.05) was performed to identify proteins that showed a significant change in labelling in response to Tipifarnib treatment. The results of the analysis are outlined in Figure 58.

In both cell types the labelling of a number of well-characterised farnesylated proteins, including KRAS and nuclear lamins, was sensitive to Tipifarnib treatment. YnF labelling of these proteins was significantly reduced when the cells were pre-treated with inhibitor (Figure 58A), consistent with inhibition of FTase and in line with the results presented in Chapter 4. The loss of RHEB farnesylation, in particular, may explain some of the beneficial effects of Tipifarnib treatment in hypoxia-induced PH. mTOR, the downstream target of RHEB (discussed in section 1.3.1, Chapter 1), is thought to play a key role in the pathogenesis of PH and is required for chronic hypoxia-induced pulmonary vascular cell proliferation and remodelling344. Previous studies have shown that farnesylation is required for RHEB to signal through mTOR345, 346.

Another protein of interest is RND3 (RHOE), which was recently shown to stabilise HIF-1α347. HIF-1α, which is also stabilised by RHOB143, is a mediator of hypoxia-induced remodelling in the pulmonary vasculature348. However, RHOE is also an inhibitor of ROCK349, suggesting that loss of RHOE function may have detrimental effects on the pulmonary vasculature by increasing RHOA/ROCK signalling.

131

Figure 58. The effect of Tipifarnib on prenylation in HPASMCs and HPAECs. HPASMCs (A&B) and HPAECs (C&D) were pre-incubated with Tipifarnib (100 nM) or vehicle control followed by labelling with 5 µM YnF (A&C) or 5 µM YnGG (B&D) for 24 hours. 3 biological replicates per condition were prepared. Lysates were subjected to CuAAC with capture reagent AzTB, enriched on NeutrAvidin resin and digested by trypsin. Peptides were subjected to on-stagetip dimethyl labelling. Tipifarnib-treated (labelled “M”) and non-inhibitor treated samples (labelled “L”) were subsequently combined in a 1:1 ratio for each biological replicate. The peptide samples were analysed by LC-MS/MS on a Q Exactive™ instrument. Data were processed in Maxquant and statistical analysis performed in Perseus. The log2(M/L) ratios were assessed in a one-sample t-test (p<0.05). A minimum of 2 valid values per sample group were required. Closed circles: CXXX & RAB proteins, open circles: other proteins. Proteins highlighted in blue satisfy the criteria of p<0.05 and log2 fold change >1 or <-1.

132

In HPASMCs, a number of proteins showed enhanced labelling by YnGG in the presence of Tipifarnib, indicating alternative prenylation upon FTase inhibition (Figure 58B). Both NRAS and KRAS showed a >3 log2 fold change in YnGG incorporation. This result highlights the dynamic and complex effects of prenyl transferase inhibitors in cellular systems, and showcases how our probes can be used to study this behaviour. YnGG labelling of RRAS2, OAS1 and interferon-induced GTP-binding protein MX1 (MX1) was also enhanced upon FTI treatment. RRAS2 is a dual substrate for both FTase and GGTase-1, and it is likely that loss of farnesylation of this protein is accompanied by a concomitant increase in geranylgeranylation, as previously observed for RHOB which is also a dual substrate for FTase and GGTase-1350.

Prenylation of MX1 has not previously been reported. MX1 exists in two isoform with differing C-terminal sequence. Isoform 2 has a CPAP C-terminal sequence, whereas isoform 1 does not carry a CXXX sequence. CPAP is not a canonical prenylation motif, and in vitro peptide substrate studies indicate that sequences terminating with a Pro residue are typically not substrates for FTase70. Further inspection of the 5 MX1 peptides identified in HPSMCs did not reveal any isoform-specific peptides. However, MX1 peptides specific to isoform 1, which does not carry a CXXX motif, were identified in HPAEC samples, suggesting that MX1 detection represents background labelling or co-purification with a prenylated interactor protein.

In HPSMCs, RHOB showed a small increase in YnGG incorporation upon Tipifarnib treatment, indicating that any loss of farnesylated RHOB is partly compensated by increased geranylgeranylation. Unfortunately RHOB was not detected in YnF-labelled HPSMCs, and with neither probe in HPAEC samples. It is clear that further optimisation of our labelling workflow is required to achieve robust and consistent identification of RHOB.

5.4. Labelling Studies in HGPS Patient Cells

In addition to studying prenylation in the pulmonary vasculature, we also sought to apply our methodology to study prenylation in the rare genetic disease HGPS. As discussed, HGPS is caused by the nuclear accumulation of farnesylated progerin, a mutant form of LMNA. Fibroblasts from HGPS patients and healthy donors were acquired from the Coriell Biorepository (catalog.coriell.org). Three patients with HGPS were selected based on harbouring the typical 2036C>T substitution in exon 11 of the LMNA gene112. Fibroblasts from healthy donors of a similar age to the HGPS patients were selected as controls. The selected cells lines and their characteristics are summarised in Table 12.

133

Table 12. Characteristics of fibroblasts obtained from Coriell Biorepository. (*) indicates the passage number at which proteomics samples (section 5.5) were prepared. Cell Code Phenotype Gender Age Passage* Mutation AG11513 HGPS Female 8 10 2036C>T AG01972 HGPS Female 14 20 2036C>T AG11498 HGPS Male 14 10 2036C>T GM08398 Healthy Male 8 14 - GM02037 Healthy Male 13 18 - GM01651 Healthy Female 13 20 -

5.4.1. Optimising Labelling in HGPS and Healthy Fibroblasts

Labelling efficiency of YnF and YnGG in our HGPS and control fibroblasts was assessed by in-gel fluorescence in a selection of the cell lines (Figure 59). Similarly to our studies in EA.hy926 cells (Chapter 2), maximum labelling was achieved with 10 µM of YnF or YnGG probe, with a decrease in labelling observed at higher probe concentrations. We chose a labelling time point of 24 hours as our previous studies in various cells types indicated that efficient labelling could be achieved within this time frame. Based on the in-gel analysis, no obvious differences in labelling pattern were observed in HGPS versus control cells, suggesting a similar pattern of prenylation in both cell types.

134

Figure 59. YnF labels proteins in fibroblasts from HGPS patients and healthy controls. Cells were cultured in medium containing increasing concentrations of YnF (A&B), YnGG (C) or DMSO vehicle for 24 hrs. Post-lysis the samples were subjected to CuAAC with AzTB and separated by SDS-PAGE. Gels were imaged by in-gel fluorescence (channel Cy3 for TAMRA and Cy5 for protein marker). Subsequently, gels were Coomassie Blue stained to assess protein loading. 5.4.2. Assessing the Effect of Tipifarnib on YnF and YnGG Labelling in HGPS Cells

Next we sought to investigate whether YnF and/or YnGG labelling was sensitive to FTI treatment. The recently conducted clinical trial in HGPS patients assessed the inhibitor Lonafarnib129, 130; however, in this initial study we decided to use Tipifarnib as it was available in our lab. Cells were pre-treated with increasing concentrations of Tipifarnib for 1 hour followed by YnF or YnGG labelling for a further 24 hours. The lysates were subjected to CuAAC with AzTB and the labelling assessed by in-gel fluorescence (Figure 60). Tipifarnib caused a dose-dependent decrease of some bands in the YnF-labelled samples but did not affect YnGG labelling. This effect is in line with our previous studies in EA.hy926 cells (Chapter

135

4), and consistent with the selective inhibition of FTase. As previously observed, several bands in the YnF-labelled samples were insensitive to Tipifarnib, suggesting that these represent proteins prenylated by GGTase-1 or RABGGTase.

Figure 60. The effect of farnesyl transferase inhibitor Tipifarnib on YnF and YnGG labelling in fibroblasts from HGPS patients and healthy controls. Cells were pre-incubated with increasing concentrations of Tipifarnib (0, 1, 10, 100 nM) for 1 hr, followed by YnF (A) or YnGG (A) (10µM) labelling for 24 hours. Post-lysis the samples were subjected to CuAAC with capture reagent AzTB and separated by SDS-PAGE. Gels were imaged by in-gel fluorescence (channel Cy3 for TAMRA and Cy5 for protein marker). Subsequently, gels were Coomassie Blue stained to assess protein loading. 5.5. Quantifying Prenylation in HGPS Patient Cells by Proteomics

After optimising labelling conditions in the HGPS and healthy donor fibroblasts we moved on to assess differential expression of farnesylated and geranylgeranylated proteins and the effect of FTI treatment in the two cell types. The key questions we aimed to answer were:

 What differences in protein farnesylation and/or geranylgeranylation exist in HGPS versus healthy cells?  Which prenylated proteins are affected by FTI (Tipifarnib) treatment in HGPS and healthy cells?  Does response to FTI treatment differ in HGPS versus healthy cells?

To this extent we prepared cells pre-treated with or without Tipifarnib (50 nM) for 1 hour followed by YnF or YnGG labelling for 24 hours. Two samples of each of the 6 cell lines (Table 12) were prepared, totalling 6 samples of HGPS and 6 samples of healthy cells for each condition. To enable relative quantification, a SILAC spike-in was produced from the healthy fibroblast population. Healthy fibroblasts were grown in R10K8 SILAC medium for 6 doubling prior to YnF or YnGG labelling. Cells from all three healthy cell lines were used for preparation of the spike-in. After lysis the three cell lines were mixed in a 1:1:1 ratio. We made no attempt

136 to generate R10K8-labelled HGPS cells. The HGPS cells showed poor growth under standard conditions, and hence we anticipated that they would not tolerate the prolonged culture required to ensure incorporation of the isotopically labelled amino acids, as also concluded by others351.

The lysates from the experimental samples were mixed with an aliquot of the spike-in and subjected to CuAAC with capture reagent AzRB. Labelled proteins were enriched on NeutrAvidin beads and digested by trypsin, followed by LC-MS/MS analysis of the eluted peptides. The results of the proteomics analysis are summarised in Figure 61, Figure 62 and Figure 63.

5.5.1. Comparison of Prenylation in HGPS versus Healthy Fibroblasts

The comparison of YnF and YnGG labelling in HGPS versus healthy cells is summarised in Figure 61 below. The main protein of interest, LMNA, showed a small but significant increase in YnF-labelling in HGPS versus healthy cells. This is consistent with the expected retention of the farnesyl modification on progerin, the mutant version of LMNA present in HGPS cells. Breakdown of the results by cell line (Figure 61C) indicated that the relative increase in LMNA differed markedly in the three HGPS cell lines. Whereas cell lines AG01972 and AG11498 showed a significant increase in LMNA (doubling in the case of AG11498), AG11513 displayed much lower levels of this protein. The reason for this difference is unclear, but could be related to the age of the HGPS donors (Table 12). Progerin accumulates in cells from healthy individuals with increasing age, and cultured HGPS cells display increased nuclear defects with increasing passage number352. The AG11513 donor was 8 years old compared to the donors of the other two cell lines, both of whom were 14 years. Thus, it is plausible that the older HGPS patients accumulate higher levels of progerin.

The relatively modest overall increase in LMNA in HGPS cells was unexpected. In normal cells farnesylated LMNA is virtually undetectable as it is rapidly processed to mature lamin A. Thus, we would have expected to detect a much larger increase in farnesylated LMNA in HGPS cells. The most likely reason for this discrepancy is the lysis conditions employed. To ensure compatibility with CuAAC, a modified RIPA buffer containing 1% Triton X-100 and 0.2% SDS was utilised. This detergent composition may not have achieved efficient nuclear lysis and thus any LMNA associated with the nucleus would not have been solubilised. Use of higher SDS concentrations in combination with mechanical lysis techniques may have circumvented this problem. However, significant optimisation of the sample preparation protocol would have been required as high detergent concentrations may inhibit the CuAAC reaction353. Attempts to detect progerin by Western blot in our samples were unsuccessful (data not shown), further supporting the hypothesis that lysis may not have been optimal.

137

Figure 61. Differential expression of prenylated proteins in HGPS versus healthy fibroblasts. Fibroblasts from HGPS patients (AG11523, AG01972 and AG11498) or healthy donors (GM02037, GM08398 and GM01654) were cultured in medium supplemented with 10 µM YnF (A) or 10 µM YnGG (B) for 24 hours. 2 replicates per cell line were prepared. Cell lysates were mixed 2:1 with a R10K8 SILAC spike-in standard prepared from YnF- or YnGG-labelled (10µM, 24 hours) healthy donor cells. Mixed lysates were subjected to CuAAC with capture reagent AzRB, enriched on NeutrAvidin resin and digested by trypsin. The eluted peptides were analysed by LC-MS/MS on a Q Exactive™ instrument. Data was processed in Maxquant and statistical analysis performed in Perseus. The log2(L/H) ratios of HGPS versus healthy cell lines were compared in a two- sample t-test (p<0.05). A minimum of 3 valid values per sample group were required. Closed circles: CXXX & RAB proteins, open circles: other proteins. (C) Log2(L/H) ratios of LMNA from the 3 HGPS cell lines labelled with YnF were normalised to the average LMNA log2(L/H) ratio of YnF-labelled healthy cells. A one-sample t-test was performed to assess whether ratios were significantly different from 0.*p<0.05, ***p<0.01, n=2. In the YnGG-labelled samples, a number of proteins showed differential levels of probe incorporation in HGPS versus healthy cells. We focused on proteins that showed a significant

(p<0.05) log2 fold change >1 or <-1. Three prenylation motif proteins, GNG2, UBL3 and RAB27B fulfilled these criteria. GNG and UBL3 showed a decrease in YnGG labelling in HGPS cells versus healthy cells whereas RAB27B showed an increase. As previously

138 mentioned in Chapter 3, the prenylation of UBL3 has not previously been experimentally validated in humans, although the homologous protein is known to be prenylated in Arabidopsis thaliana270. None of these proteins have been identified in previous quantitative proteomics studies351, 354, 355 exploring differential protein expression in HGPS cells and any functional role of these prenylated proteins in HGPS is unclear.

One non-prenylation motif protein, mitochondrial pyruvate carboxylase (PC) was increased in HGPS cells. PC is endogenously biotinylated in the cell and hence contains a covalently bound biotin molecule which would have mediated its enrichment on the NeutrAvidin beads. PC catalyses the ATP-dependent carboxylation of pyruvate to oxaloacetate, a key reaction in gluconeogenesis and lipogenesis356. A recent proteomics study in adipose tissue from Zmpste24-/- mice showed that PC was upregulated in the mutant strain357. Zmpste24 knock- out mice are used as a model for HGPS. Correct processing of normal LMNA requires the cleavage of the C-terminal portion of the protein including the farnesylated cysteine, which is mediated by Zmpste24. Zmpste24 knock-out hence results in accumulation of farnesylated LMNA, which mimics the pathology of HGPS.

5.5.2. Response to Tipifarnib

As a recent clinical trial showed positive outcomes for HGPS patients treated with a FTI, we wanted to gain a better picture of which farnesylated and geranylgeranylated proteins were affected by FTI treatment. Additionally, we wanted to understand whether HGPS and healthy cells responded in a similar manner to FTase inhibition. Figure 62 shows the effect of Tipifarnib treatment on YnF and YnGG-labelling in HGPS and healthy fibroblasts.

139

Figure 62. Effect of Tipifarnib treatment on YnF and YnGG labelling in HGPS and healthy fibroblasts. Fibroblasts from HGPS patients (AG11523, AG01972 and AG11498) or healthy donors (GM02037, GM08398 and GM01654) were pre-treated with Tipifarnib (50 nM) or vehicle (DMSO) for 1 hour followed by YnF (A&B) or YnGG (C&D) labelling (10 µM, 24 hours). 2 replicates per cell line were prepared. Cell lysates were mixed 2:1 with a R10K8 SILAC spike-in standard prepared from YnF- or YnGG-labelled (10µM, 24 hours) healthy donor cells. Lysates were subjected to CuAAC with capture reagent AzRB, enriched on NeutrAvidin resin and digested by trypsin. The eluted peptides were analysed by LC-MS/MS on a Q Exactive™ instrument. Data was processed in Maxquant and statistical analysis performed in Perseus. The log2(L/H) ratios of Tipifarnib-treated versus non-treated cells were compared in a two-sample t-test (p<0.05). A minimum of 3 valid values per sample group were required. Closed circles: CXXX & RAB proteins, open circles: other proteins. Proteins with p<0.05 and a log2 ratio >1 or <-1 are coloured blue.

140

We were particularly interested to observe whether LMNA could be alternatively geranylgeranylated in the presence of FTIs, analogous to the effects observed on KRAS and NRAS prenylation status. A previous study indicated that LMNA becomes geranylgeranylated upon FTI treatment in vitro358; however, this is not thought to occur to any extent which is physiologically relevant in vivo111. Although LMNA was robustly detected and quantified with the YnF probe (Figure 62A & B), in line with the preference for a farnesyl modification, we were unable to quantify any YnGG incorporation in LMNA (Figure 62C & D). A number of LMNA peptides were detected in YnGG-labelled HGPS samples treated with Tipifarnib, suggesting that some incorporation of YnGG may be occurring in the presence of FTIs. However, as no peptides were detected in the absence of FTI, a quantitative comparison was not possible. Optimisation of the quantitative protocol is required to access more information about the potential geranylgeranylation of LMNA. One approach might entail use of a spike-in standard which is dually labelled with both YnF and YnGG to aid quantification of all prenylated proteins.

In response to FTI, increased YnGG labelling was apparent for RRAS2 (Figure 62C & D), which mirrors results obtained in other cell types. Quantification of NRAS and KRAS dynamics was not possible, because as in the case of LMNA, these proteins were not detected by YnGG in non-inhibitor-treated samples. YnF labelling of numerous farnesylated proteins was inhibited by Tipifarnib; all of which were identified in our previous inhibitor studies (Chapter 4).

Comparison of the response to FTI in HGPS and healthy cells (Figure 63) shows a very similar story when compared to the comparison of non-inhibitor-treated cells (Figure 61). GNG2 and UBL3 showed a decreased level of labelling in HGPS cells, consistent with the decreased level of these proteins present in non-inhibitor-treated cells. Aldehyde dehydrogenase family 3 member B1 (ALDH3B1), which is known to be geranylgeranylated359, was also decreased in FTI-treated HGPS cells versus FTI-treated healthy cells. In our correlation between non- inhibitor-treated cells (Figure 61) ALDH3B1 was also downregulated, although the observed log2 fold change (-0.98) just fell outside our cut-off criteria. Aldehyde dehydrogenases are important for cellular conversion of reactive aldehydes to less toxic carboxylic acids, and such play a role in protecting cells from oxidative stress360.

141

Figure 63. Comparing the effect of Tipifarnib treatment on YnF and YnGG labelling in HGPS and healthy fibroblasts. Fibroblasts from HGPS patients (AG11523, AG01972 and AG11498) or healthy donors (GM02037, GM08398 and GM01654) were pre-treated with Tipifarnib (50 nM) for 1 hour followed by YnF (A&B) or YnGG (C&D) labelling (10 µM, 24 hours). 2 replicates per cell line were prepared. Cell lysates were mixed 2:1 with a R10K8 SILAC spike-in standard prepared from YnF- or YnGG-labelled (10µM, 24 hours) healthy donor cells. Lysates were subjected to CuAAC with capture reagent AzRB, enriched on NeutrAvidin resin and digested by trypsin. The eluted peptides were analysed by LC-MS/MS on a Q Exactive™ instrument. Data was processed in Maxquant and statistical analysis performed in Perseus. The log2(L/H) ratios of HGPS versus healthy cells were compared in a two-sample t-test (p<0.05). A minimum of 3 valid values per sample group were required. Closed circles: CXXX & RAB proteins, open circles: other proteins. 5.6. Proof-of-Concept: In vivo Labelling of Zebrafish

Zebrafish (Danio rerio) is an important animal model for vertebrate development and physiology and it’s high genetic similarity to humans makes it a valuable tool for studying disease361. In comparison to larger mammalian models such as rodents or sheep, zebrafish are cheap to breed and maintain. Furthermore, zebrafish develop from egg to fully mature adults within a few days of fertilisation, enabling rapid experimental turn-over. Zebrafish harbour the three prenyl transferases FTase, GGTase-1 and RABGGTase. However, in contrast to humans, the zebrafish genome only encodes for one REP protein362. Pharmacological and genetic studies in zebrafish indicate that prenylation is important for key developmental processes such as germ cell migration363 and heart tube formation364.

A fluorescent isoprenoid pyrophosphate analogue has previously been used to explore isoprenoid distribution in zebrafish203. The analogue was delivered via injection into the pre- cardiac sinus of anaesthetised embryos and was found to localise predominantly to the

142 digestive tract and circulatory system. Although this study shows that an isoprenoid analogue can be readily distributed around the zebrafish body, the potential of labelling prenylated proteins in this in vivo model has not been explored.

Recent studies in the Tate group have employed alkynyl probes to study protein cholesteroylation194 and myistoylation289 in zebrafish. Based on these results we sought to explore whether our prenyl probes would also be amenable to in vivo labelling. To assess whether our novel probes could be incorporated into prenylated proteins in zebrafish, a labelling study was conducted by Dr Paulina Ciepla using the YnF and YnGG prenyl probes.

Hatched zebrafish embryos (48 hours post-fertilisation (hpf)) were grown in water containing increasing concentrations of YnF or YnGG. The embryos were subsequently lysed and subjected to CuAAC with capture reagent AzTB. Figure 64 shows the in-gel fluorescence images of labelled proteins after separation by SDS-PAGE. Both probes exhibited a dose- dependent increase in labelling efficiency. In the case of YnF, concentrations above 10 µM resulted in severe toxicity and death.

Figure 64. YnF and YnGG labelling in zebrafish. Hatched zebrafish embryos (48 hpf) were grown in the presence of YnF or YnGG for 24 hours. After sacrifice the zebrafish lysate was subjected to CuAAC with AzTB, separated by SDS-PAGE and visualised by in-gel fluorescence. Gels were stained with Coomassie Blue to assess protein loading. Samples and gels prepared by Dr Paulina Ciepla. Further studies in zebrafish were beyond the scope of this project, however these preliminary results warrant further investigation. Identification of YnF and YnGG-labelled proteins by mass spectrometry will provide further validation of the utility of the probes for in vivo labelling studies.

143

5.7. Conclusions

This chapter describes the use of our novel tools YnF and YnGG to further our knowledge of prenylation in models of disease and in vivo. We show that YnF and YnGG can efficiently label primary cells from the pulmonary vasculature and HGPS patients, enabling identification prenylated proteins by mass spectrometry. By applying spike-in SILAC and dimethyl labelling we profiled changes in prenylation in response to hypoxia and FTI treatment.

Contradictory to previous results, we found no evidence of changes in protein prenylation in response to hypoxia in HPAECs, which may be due to the experimental protocol employed. Further work is required to optimise the dual labelling and hypoxic exposure conditions to ascertain the validity of our results. Tipifarnib treatment of HPAECs and HPASMCs inhibited the farnesylation of a number of FTase substrates. Concomitantly, we observed an increase in geranylgeranylation of proteins such as members of the RAS family.

Analysis of prenylation in HGPS and healthy cells indicated that 3 geranylgeranylated proteins showed a >1 or -<1 log2 fold change in HGPS cells. Whereas RAB27B was increased, GNG2 and UBL3 were decreased in HGPS. However, our inability to detect a robust increase in LMNA farnesylation indicates that our protocol may require further optimisation; in particular improvements in cell lysis conditions may be needed to fully release all prenylated proteins.

We also provide a first account of in vivo labelling of prenylated proteins. Zebrafish grown in increasing concentrations of YnF and YnGG incorporated to probes into distinct proteins, as assessed by in-gel fluorescence of fish tissue lysate. We envisage that application of our probes in vivo will provide new insights on the role of prenylation in health as well as in disease models.

5.7.1. Technical Challenges to Address

Although the initial data presented herein suggests that our tools may provide a useful means to answer questions related to the role of prenylation in PH and HGPS, methodological optimisation is required. Due to the time-frame of this project this was not possible at the current time.

Detection of RHOB

A key aim of this chapter was to explore the prenylation of RHOB in cells of the pulmonary vasculature. Although our studies in EA.hy926 cells enabled the detection of farnesylated RHOB, we were not successful in quantifying YnF-labelled RHOB in HPASMCs and HPAECs. RHOB is expressed at low basal levels, and is predominantly geranylgeranylated, making

144 detection challenging. Further work by Dr Julia Morales-Sanfrutos suggests that in EA.hy926 cells the relative incorporation ratio of YnF versus YnGG is 11:89 (manuscript in preparation), highlighting the preference of geranylgeranylation for this protein.

An option to improve detection would be to increase protein levels by over-expressing RHOB. Although we considered this option we decided that an over-expression system was not physiologically relevant. RHOB over-expression within a SILAC spike-in standard may however be a useful strategy to enhance detection and quantification. However, careful titration of the level of RHOB would be necessary. If the level of RHOB in the spike-in standard versus the experimental samples was several-fold greater, the ability to detect differences between experimental samples would be compromised. Another option could be to employ a spike-in standard dually labelled with both YnF and YnGG to expand the repertoire of prenylated proteins present in the spike-in standard. As previously discussed, this strategy would be particularly useful when exploring prenylation dynamics in response to inhibitors.

Scope to improve the labelling conditions also exists. Although our initial optimisation indicated that HPAECs and HPSMCs require 24-48 hours of incubation with the prenyl probes to gain maximal labelling, this result is related to prenylation on a whole proteome scale. To probe a specific protein, labelling should be further optimised on a protein basis, for example by assessing labelling efficiency of the protein of interest by Western blot. In our proteomics time- course studies in Chapter 3 we showed that probe incorporation into RHOB occurs within a few hours, consistent with the rapid turn-over of this protein. Our attempt to access changes in RHOB prenylation in response to hypoxia was preceded by a 24-hour incubation period with the prenyl probe prior to hypoxic exposure. The isoprenoid probes need to undergo phosphorylation by the isoprenoid rescue pathway before they can be used as substrates for the prenyl transferase enzymes. We had no means available to assess the rate at which uptake and phosphorylation of the probe occurs and the time-frame for our hypoxic exposure was relatively short (2-4 hours). Thus, the probe pre-incubation served to ensure that a pool of phosphorylated probe was already available within the cell from the start of hypoxic exposure to ensure incorporation of the probe on proteins such as RHOB that are rapidly induced in hypoxia. However, while we didn’t have information about the rate of probe phosphorylation, we also had no information about potential probe metabolism. Thus, it is plausible that after 24 hours there is very little probe still present in the cell due to cellular metabolism, and as such any proteins induced by hypoxia were not efficiently labelled. However, the fact that labelling intensity increases up to 48 hours as assessed by in-gel fluorescence suggests that probe pools should still be present in the cells at 24 hours.

145

Cell lysis

A further factor to consider is whether our lysis protocol requires optimisation. Our comparison of HGPS and healthy fibroblast cells failed to detect the greatly enhanced levels of farnesylated LMNA that are a hall-mark of HGPS pathology. This discrepancy was most likely a result of poor solubilisation of nuclear LMNA during cell lysis. As prenylated RHOB is predominantly membrane-bound, enhanced detection may be aided by sub-cellular fractionation. However, as a significant portion of prenylated proteins exist in cytosolic form bound to chaperones7, ideally analysis should be performed on whole cell lysates. The use of higher detergent concentrations during lysis should aid solubilisation of prenylated proteins. As the lysis buffer composition has implications for the efficiency of the CuAAC reaction353, significant optimisation work may be required to find the best lysis conditions.

146

Chapter 6: Conclusions and Future Work

This thesis presents a novel methodology to label, identify and quantify prenylated proteins. This chapter summarised the key findings of our work, and proposes further optimisation work.

6.1. Novel Probes for Metabolic Labelling of Prenylated Proteins

In Chapter 2 we present two novel probes YnF and YnGG aimed to enable metabolic labelling of prenylated proteins in cultured cells. An initial panel of novel and previously reported prenyl probes were assessed for their ability to label proteins as assessed by in-gel fluorescence and immunoblotting. YnF and YnGG was able to detect HRAS and RHOA, respectively, suggesting that they selectively label farnesylated versus geranylgeranylated proteins. Importantly, we found that our novel probes showed excellent incorporation without needing to deplete endogenous isoprenoids through statin treatment. As discussed, statins exhibit numerous effects on the cell including altered regulation of prenylated proteins, and affects prenylation status. A robust synthetic route to access both probes from cheap and readily available starting materials was developed.

6.1.1. Future Work

 During this project we were unable to investigate the enzyme kinetics of the YnF and YnGG probes, due to issues in obtaining active prenyl transferase enzymes. In future work the in vitro enzyme assays should be performed to compare the prenyl probes to the natural substrates. Importantly, the assay should be performed with a variety of known prenylated peptide substrates to assess whether the structural changes are impacting on the substrate selectivity of the enzymes.

 Structural studies, in particular X-ray crystallographic studies, would provide useful information about the binding of the probes. As the isoprenoid forms a substantial part of the peptide substrate binding site, gauging the interaction between the prenyl probes and a peptide substrates should inform about potential altered substrate selectivity that result from of the modifications installed on the probes.

 Scope exists to improve the synthesis of YnF and YnGG. The current approach relies on preparing each probe in parallel, however a modular approach might be more efficient. As the probe structure is analogous, only differing in the number of isoprenoid units, one might envisage a strategy where the alkyne-modified portion of the molecule is prepared separately such that it can be coupled onto any isoprenoid of choice in a late stage of the synthesis.

147

6.2. Development of a Proteomics Workflow to Identify and Quantify Prenylation

With our prenyl probes in hand we sought to apply these to enrich prenylated proteins for detection by mass spectrometry. Initial work (Chapter 3) indicated that our probes robustly enabled the detection of both known and putative novel substrates for all three prenyl transferases. In non-quantitative proteomics studies we identified in excess of 60 proteins harbouring prenylation motifs with each probe. These number of proteins contained in these data sets far exceed the number of proteins identified in previous studies employing similar methodologies207. While some substrate overlap between the YnF and YnGG were observed, both probes also identified unique sets of proteins, demonstrating the importance of a two probe system to detect both farnesylated and geranylgeranylated proteins. Our initial proteomics studies indicated that YnF was able to identify a number of putative novel farnesylated substrates, including DCAF8, ULK3, CEP85, LRRFIP1, EHBP1 and EHBP1L1.

Spike-in SILAC quantitative proteomics showed that the incorporation of our probes onto prenylation motif proteins was concentration- and time-dependent. We also employed SILAC to decipher the prenyl probe preference of known prenylated substrates, showing that YnGG was preferentially incorporated into geranylgeranylated proteins and YnF into farnesylated substrates. This result further highlights the selectivity of our probes.

In further validation work we conducted an isoprenoid competition study by quantitative proteomics (Chapter 4). Addition of the natural isoprenoid substrates farnesol and geranylgeraniol lead to a significant decrease in the labelling of prenylation motif proteins by YnF and YnGG respectively. All but one of the proteins which responded to isoprenoid competition have previously been described as prenylated. In contrast, proteins that did not harbour a prenylation motif were unaffected by isoprenoid competition. While all RAB proteins are geranylgeranylated, not all CXXX-motif proteins are prenylated. Thus, this methodology enabled us to decipher which proteins are true substrates and which are a result of non- specific background labelling.

In an attempt to further validate the on-target effects of our prenyl probes a number of novel capture reagents were developed. 3 reagents, AzRB, AzKB and AzK+RB, all containing a trypsin cleavage site, were synthesised by solid phase peptide synthesis. The trypsin cleavage site mediated the release of the probe-modified peptide after enrichment on NeutrAvidin beads, which was exemplified by the detection of myristoylated peptides modified by the myristic acid analogue YnMyr. In further work conducted by Dr Julia Morales-Sanfrutos a number of C-terminal peptides modified by YnF were detected. Of particular interest was the

148 detection of a probe-modified peptide from ULK3, supporting our claim that this protein is a bona fida farnesylated substrate.

6.2.1. Future Work

 Profiling of prenylated proteins in different cell types should be considered as a means to detect further substrates. Our work focused on identifying labelled proteins in endothelial cell line EA.hy926. Different cell types may express higher levels of putative targets that went undetected in our studies due to low abundance.

 Further optimisation of the cell lysis protocol, for example higher detergent concentrations, should be trialled. Prenylated proteins are typically associated with membranes and thus it is plausible that a different lysis protocol may result in better solubilisation of these proteins. Sub-cellular fractionation to enrich nuclear and membrane fractions may also provide additional benefits over whole cell lysate analysis.

 Work to optimise the detection of probe-modified peptides should be conducted to improve the identification rate. As discussed in Chapter 3, several inherent difficulties are encountered in the identification of prenylated peptides. Firstly, the modification exists on the C-terminal peptide, which due to their typically uncharged nature is more difficult to detect by mass spectrometry. Secondly, many prenylated proteins harbours Lys and Arg rich sequences near their C-terminus and thus trypsin digest yield peptides that are too short to confidently identify by proteomics. The use of different enzymes for the digest step and methods to enrich the C-terminal peptides may overcome some of these issues.

 Mutational studies should be performed to further validate the novel prenylated substrates described here-in. Specifically, mutation of the cysteine residue within the C-terminal CXXX should abrogate labelling of these substrates if the attachment of the prenyl probe is mediated by FTase.

6.3. Inhibition Studies

A key aim of this project was to develop a method that would enable the quantitative profiling of the effects of prenyl transferase inhibitors in a cell-based system. Our work outlined in Chapter 5 showcases how our prenyl probes in combination with in gel fluorescence analysis or SILAC quantification enables in-cell profiling of the targets of various prenyl transferase inhibitors. In our initial studies we assessed the proteins affected by inhibitors of the three prenyl transferases. Our analysis of the recently reported RABGGTase inhibitor Bon-15 indicated that this compound is in fact a dual inhibitor of both RABGGTase and FTase. This result provides an example of the how our method may be employed to survey the on/off-

149 target effects of novel inhibitors to aid in validation of novel compounds and to understand the cellular responses of novel compounds.

In more comprehensive studies we used spike-in SILAC quantification to determine a dose- response to a number of prenyl transferase inhibitors, including FTI-277, Tipifarnib, Manumycin A and GGTI-2133. We found that YnF-labelling of all of our putative novel farnesylated targets were sensitive to FTI treatment, and by fitting dose-response curves were able to determine an in-cell IC50 for Tipifarnib and FTI-277 inhibition. It has previously been shown that some proteins such as KRAS, which are normally farnesylated, can be geranylgeranylated by GGTase-1 upon inhibition of FTase. In line with this phenomenon we observed increased incorporation of YnGG in KRAS in response to FTI-277. The same effect could not be quantified for Tipifarnib in these experiments due to lack of detection of this protein in non-inhibitor-treated control samples.

We envisage that our methodology will provide important insight into the mechanisms of prenyl transferase inhibition, which may help to explain the lack of efficacy observed in clinical trials.

6.3.1. Future Work

 Further studies should be focused on optimising the detection and quantification rate of prenylated substrates. A small number of proteins were present in control samples but absent from inhibitor-treated samples, even at the lowest concentration tested, indicating that they may be exquisitely sensitive to inhibition. By testing a wider range of inhibitor concentration a dose-response for these proteins may be accessed.

 Further work should be conducted to explore the prenyl transferase dynamics associated with inhibitor treatment. Specifically, the use of a spike-in lysate dually labelled with both YnF and YnGG should enhance detection rates for proteins that show altered isoprenoid substrate incorporation when treated with inhibitor.

6.4. Prenylation in Disease Models and In Vivo

Prenylation has been most widely studies in the context of cancer due to the importance of prenylation for the correct function of the oncoprotein RAS. However, as discussed in the Introduction of this thesis, prenylation also plays a key role in numerous other diseases such as cardiovascular and neurodegenerative disease, viral and parasitic infections and HGPS. In Chapter 5 we describe the application of our prenyl probes to gain further understanding of the role of prenylation in two diseases: HGPS and hypoxia-induced PH.

150

A recent clinical trial provided evidence that FTI treatment may be of benefit for HGPS patients. HGPS patient harbour a mutant LMNA gene, which results in nuclear envelope defects due to accumulation of farnesylated LMNA at the nuclear rim. Inhibition of LMNA farnesylation is thought to abrogate the effects of the mutant LMNA protein. To gain further insight into the effect of FTI treatment in HGPS, we employed our probes to profile prenylation in cells derived from HGPS patients. Comparison of HGPS versus healthy cells did not show any robust differences in the level of prenylated proteins, including LMNA, in these two cell types. Treatment with Tipifarnib resulted in decreased level of YnF-labelling comparable to the results observed in our comprehensive studies in EA.hy926 cells. Importantly, the pattern of inhibition in HGPS cells mirrored that of healthy cells.

Recent work in the Wojciak-Stothard lab indicates that targeting farnesylation may provide a potential treatment for hypoxia-induced PH. To better understand the role of farnesylation in this disease and which proteins may contribute to the positive effects of FTI-treatment we employed our quantitative proteomics workflow to profile changes in prenylated proteins in response to hypoxia and Tipifarnib-treatment. In particular, we were interested in investigating the role of RHOB prenylation. Comparison of hypoxia and normoxia-treated HPAECs by quantitative proteomics failed to detect any changes in prenylated proteins. Although RHOB was detected by our YnGG probe, we were unable to detect YnF-labelled RHOB. Treatment of HPAECs and HPSMCs with Tipifarnib resulted in loss of YnF-labelling of a number of known farnesylated substrates.

Despite the challenges encountered within these preliminary studies, further optimisation of labelling protocols and sample processing should provide further insights. Encouragingly, these studies highlighted the ability of our probes to label prenylated proteins in a number of different cell types. Furthermore, we were also able to provide a first account of labelling in vivo.

6.4.1. Future Work

 As discussed, we failed to detect a large increase of farnesylated LMNA in HGPS cells compared to healthy cells. This could be attributed to inadequate lysis of the nuclear fraction and hence poor release of YnF-labelled LMNA, and hence work to optimise lysis conditions should be conducted.

 As discussed in detail in Chapter 5, further optimisation is required to achieve robust detection of YnF-labelled RHOB. This may be achieved by employing a spike-in lysate from cells over-expressing RHOB to enhance levels of YnF-labelled RHOB. Alternatively, targeted mass spectrometry could provide a means to specifically study RHOB.

151

Chapter 7: Experimental Methods

7.1. Biological methods

7.1.1. Reagents

Dulbecco’s Modified Eagle’s Medium (DMEM) was obtained from Sigma-Aldrich. Foetal bovine serum (FBS) and penicillin/streptomycin were obtained from Gibco®. R10K8 DMEM for SILAC cell culture was obtained from Dundee Cell products and dialysed serum from Sigma-Aldrich. Human pulmonary arterial endothelial cells (HPAECs), human pulmonary arterial smooth muscle cells (HPASMCs), Endothelial Cell growth medium, Endothelial Cell Growth medium 2 supplement and Smooth Muscle Cell Growth medium were obtained from PromoCell. Bovine fibronectin was obtained from Sigma-Aldrich. HGPS primary skin fibroblasts (AG01972, AG11498 and AG11513) and apparently healthy primary skin fibroblasts (GM01651, GM02037, GM08398) were obtained from Coriell Cell Repositories.

FTI-277, GGTI-2133 and Manumycin A were obtained from Sigma-Aldrich. Tipifarnib was obtained from Biorbyt Ltd. Bon-15 was obtained from Professor Herbert Waldmann (Max Planck Institute of Molecular Physiology, Munich). Farnesol (FOH) and geranylgeraniol (GGOH) were obtained from Sigma-Aldrich. Due to low purity GGOH was re-purified by column chromatography prior to use.

Sequence grade modified porcine trypsin was obtained from Promega.

Ultrapure water was obtained on a MilliQ water purification system (Millipore).

7.1.2. Cell culture

Immortalised human umbilical vein endothelial cells (cell line EA.hy926) were cultured in DMEM supplemented with 10% FBS, 100 units/mL penicillin and 100 µg/mL streptomycin. SILAC-labelled EA.hy926 cells were grown in R10K8 DMEM supplemented with 10% dialyzed serum and Penicillin/Streptomycin as above. Cells were maintained for a minimum of 6 passages prior to use in experiments.

Primary HPAECs were cultured on fibronectin-coated plates in Endothelial Cell Growth medium with Endothelial Cell Growth medium 2 supplement. Primary HPASMCs were cultured in Smooth Muscle Cell Growth medium 2 on un-coated plates. HPAECs and HPSMCs were used in experiment before passage 10.

HGPS fibroblasts (AG01972, AG11498 and AG11513) and healthy fibroblasts (GM01651, GM02037, GM08398) were cultured in DMEM supplemented with 15% FBS. SILAC-labelled

152 fibroblasts (GM01651, GM02037, GM08398) were grown in R10K8 medium supplemented with 15% dialysed serum for 6 passages prior to use in experiments.

All cells were maintained in a humidified incubator (37°C, 5% CO2). Cells were seeded on plates at least 24 hours before experiment.

7.1.3. General procedure for preparation of cell lysate tagged with YnF or YnGG

Near-confluent (ca. 90%) plates of cells were cultured in medium (as above) supplemented with YnF or YnGG (typically 10 µM, from 1000x stock in DMSO) or DMSO only for 8-24 hours. Cells were washed twice with cold PBS, aspirated and lysis buffer added (0.6 mL per 10 cm plate, 1% Triton X-100, 0.2% SDS in PBS supplemented with 1x Complete EDTA-free protease inhibitor cocktail (Roche)). Cells were scraped, transferred to a microcentifuge tube and lysed on ice for 30 min. The lysate was clarified by centrifugation (14,000 x g, 4°C, 30 min) and the supernatant recovered. The protein concentration was determined using the DC™ protein assay (Bio-Rad) following the manufacturer’s protocol.

7.1.4. Preparation of spike-in SILAC samples for concentration-gradient or time- course labelling studies in EA.hy926 cells

Spike-in standard cell lysate was prepared from EA.hy926 cells grown in R10K8 medium supplemented with YnF or YnGG (10 µM) for 24 hours. Concentration-gradient samples were cultured in normal DMEM containing increasing concentration of YnF or YnGG (2, 5 or 10 µM) for 24 hours. Time-course samples were cultured in normal DMEM containing 10 µM YnF or YnGG for 4, 8 or 24 hours. After cell lysis each inhibitor treated sample (300 µg) was mixed with spike-in lysate (300 µg). Protein concentration was normalised across all samples by addition of lysis buffer. The samples were processed for proteomics analysis by CuAAC (capture reagent AzTB), enrichment and tryptic digest as described below.

7.1.5. Preparation of isoprenoid competition SILAC samples in EA.hy926 cells

SILAC R0K0-labelled EA.hy926 cells were incubated with YnF or YnGG (10 µM) for 8 hours. R10K8 labelled cells were incubated with YnF (10 µM) + FOH (25 µM) or YnGG (10 µM) + GGOH (10 µM) for 8 hours. 4 replicates of each sample were prepared. Cells lysates were mixed in a 1:1 R0K0/R10K8 ratio (1 mg each) before click chemistry (capture reagent AzRB), enrichment and on-bead digest as described below.

7.1.6. Preparation of spike-in SILAC samples for inhibitor evaluation in EA.hy926 cells

Spike-in standard cell lysate was prepared from EA.hy926 cells grown in R10K8 medium supplemented with YnF or YnGG (10 µM) for 24 hours. Inhibitor treated EA.hy926 cells were

153 pre-incubated in normal medium containing inhibitor (FTI-277, GGTI-2133, Manumycin, Tipifarnib or DMSO control) for 1 hour. The medium was supplemented with YnF or YnGG (10 µM) and the cells incubated for a further 8 hours. Biological triplicates were prepared for each inhibitor concentration. After cell lysis each inhibitor treated sample (1 mg) was mixed with spike-in lysate (500 µg). Protein concentration was normalised across all samples by addition of lysis buffer. The samples were processed for proteomics analysis by CuAAC (capture reagent AzRB), enrichment and tryptic digest as described below.

7.1.7. Preparation of hypoxia-treated HPAEC spike-in SILAC samples

HPAECs were incubated with YnF or YnGG probe (5 µM) for 24 hours, followed by hypoxic

(2% O2, 5% CO2) or normoxic (20% O, 5% CO2) for 4 hours. 3 biological replicates for each condition were prepared. Spike-in standard was prepared from R10K8-labelled EA.hy926 cells grown in the presence of YnF or YnGG for 24 hours. HPAEC cell lysates (600 µg) were mixed with spike-in standard (300 µg) and processed for proteomics analysis by CuAAC (capture reagent AzTB), enrichment and tryptic digest as described below.

7.1.8. Preparation of Tipifarnib-treated HPAEC and HPASMC proteomics samples

HPAECs and HPSMCs were pre-incubated Tipifarnib (100 nM) or vehicle control (DMSO) for 1 hour, followed by addition of YnF or YnGG probe (5 µM) for 24 hours. 3 biological replicates for each condition were prepared. Cell lysates (400 µg) were processed for proteomics analysis by CuAAC (capture reagent AzTB), enrichment, tryptic digest and dimethyl labelling as described below.

7.1.9. Preparation of HGPS and healthy fibroblast spike-in SILAC samples

SILAC cell lysate was prepared from healthy primary skin fibroblasts (GM01651, GM02037, GM08398) grown in R10K8 medium supplemented with YnF or YnGG (10 µM) for 24 hours. The cell lysates from the three donors were mixed in a 1:1:1 protein ratio to produce a SILAC spike-in standard. Inhibitor treated cells (HGPS cells or healthy fibroblasts) were pre- incubated in normal medium containing Tipifarnib (50 nM) for 1 hour. Control cells were treated with an equivalent amount of DMSO vehicle only. After the pre-incubation the medium was supplemented with YnF or YnGG (10 µM) and the cells cultured for a further 24 hours. 2 replicates per cell line were prepared for each condition. Post lysis the samples (600 µg) were mixed with SILAC spike-in standard (300 µg). The samples were processed for proteomics analysis by CuAAC (capture reagent AzRB), enrichment and tryptic digest as described below.

154

7.1.10. Click chemistry (CuAAC) general protocol

Protein concentrations were normalised across all samples to 1-2 mg/mL by addition of lysis buffer. For each 100 µL of protein lysate click reagent mixture was prepared as follows: 1 µL capture reagent (10 mM in DMSO), 2 µL CuSO4 (50 mM in H2O) and 2 µL

Tris(2-carboxyethyl)phosphine hydrochloride (50 mM in H2O) were added sequentially to a microcentrifuge tube, mixed and left to stand for 2 min after which 1 µL Tris[(1-benzyl-1H- 1,2,3-triazol-4-yl)methyl]amine (10 mM in DMSO) was added. The click reagent mixture was added to the protein lysate and the lysate incubated on a shaker at room temperature for 1 hour. The click reaction was terminated by addition of 0.5M EDTA (5 µL). Methanol (400 µL), chloroform (100 µL) and ultrapure water (300 µL) were added sequentially, the mixture vortexed and centrifuged (14,000 x g, 5 min, RT). The top layer was removed, leaving the protein layer at the phase interface intact. Methanol (800 µL) was added, the sample vortexed and the protein pelleted as before. All solution was aspirated, and the protein pellet washed in methanol (800 µL) again, with vortexing and sonication to break the pellet. After a final centrifugation step and removal of the methanol the pellet was air-dried in an inverted tube for 5-10 min. Samples were re-suspended in 2% SDS/10 mM EDTA in PBS (10 µL). Once dissolved, the samples were diluted to 0.2% SDS by addition of PBS.

Capture reagent AzTB was used for in-gel fluorescence and WB applications. Capture reagent AzRB was used for proteomics applications. Reagent quantities were scaled accordingly for applications requiring larger quantities of protein lysate (>100 µL). For proteomics analysis the click reaction performed in 15 mL tubes and all centrifugation steps performed at room temperature (4000 x g, 30 min) to prevent precipitation of SDS.

7.1.11. SDS-PAGE and in-gel fluorescence imaging

Protein samples were combined with LDS loading buffer (4x stock) and incubated in a heat block at 95°C for 5 min. 10 µg of protein was separated on 12% Bis-Tris gels using MOPS running buffer at 100-160 V. Gels were fixed (40% methanol, 10% acetic acid, 50% ultrapure water) for 5 min and then washed in water for 5 min prior to imaging on a Ettan DIGE Imager (GE Healthcare).

7.1.12. Protein enrichment for Western blot analysis

Protein samples (170 µg) were subjected to click chemistry with AzTB capture reagent, precipitated and re-suspended as described, except the proteins were re-suspended to a final concentration of 2 mg/mL. An aliquot (20 µg) of protein was kept aside for blots of total protein input before enrichment. The samples were incubated with Dynabead® MyOne™ Streptavidin C1 resin (10 µL per 100 µg protein, Invitrogen) for 1 hour at room temperature. The beads

155 were washed with 0.2% SDS in PBS (3 x 0.5 mL) and eluted by boiling in 1x LDS sample loading buffer (95°C, 5 min). The proteins were separated by SDS-PAGE as described and transferred to a PVDF membrane (Immobilin-PSQ, Millipore) by wet-tank transfer in a Tris- Glycine transfer buffer (Novex®) supplemented with 20% methanol for 2hr at 100V. Membranes were blocked in 5% skimmed milk in Tris-buffered saline containing 0.01% Tween-20 (TBS-T). Subsequent incubation with primary antibody in blocking buffer was performed at room temperature for 1.5 hr using the following antibodies and dilutions: ULK3 (1:1000, ab124947, Abcam), DCAF8 (1:1000, ab54746, Abcam), HRAS (1:750, MAB3291, Millipore) or RHOA (1:250, SC-418, Santa Cruz Biotechnologies). After washing in TBST (3 x 15 min) the membranes were incubated with the appropriate HRP-conjugated secondary antibodies (α-mouse-HRP or α-rabbit-HRP, 1:5000, Advansta) in blocking buffer for 1 hr at room temperature. After washing as before, the membranes were incubated with HRP substrate (Luminata Crescendo, Millipore) and imaged on an ImageQuant LAS 4000 (GE Healthcare).

7.1.13. Immunoprecipitation and on-bead CuAAC for detection of RHOB

Probe-labelled protein lysate (300 µg for EA.hy926 cells, 200 µg for HPAECs) was pre-cleared with Protein-A sepharose resin (30 µL) for 1 hour at 4°C. The beads were removed and the protein lysate incubated with rabbit α-RHOB (SC-180, Santa Cruz) at 4°C overnight. Protein A sepharose resin (40 µL) was subsequently added followed by a further incubation period of 2.5 hours at 4°C. The supernatant was removed and the beads washed twice with cold lysis buffer (500 µL) and once with phosphate-buffered saline with 0.01% Tween-20 (PBS-T, 500 µL). The beads were resuspended in PBS-T (50 µL). Click mix with capture reagent AzTB was prepared as described in section 7.1.10 and 3 µL added to each bead sample. The beads were vortexed briefly, and incubated on ice for 1.5 hours, with a brief vortex performed every 15 minutes. The supernatant was removed and the beads washed once with PBS-T (500 µL). The beads were resuspended in 2x LDS sample loading buffer (50 µL) and incubated at 90°C for 10 minutes. The eluted protein samples (25 µL) were separated by SDS-PAGE on a 15% Bis-Tris gel and visualised by in-gel fluorescence. After imaging the proteins were transferred to a PVDF membrane and total RHOB assessed by Western blot as described in section 7.1.12 with rabbit α-RHOB (1:1000).

7.1.14. Protein enrichment and on-bead tryptic digest for proteomic analysis

Proteomics samples were prepared in a dust-free area using dedicated pipettes and pipette tips. Only low binding microcentrifuge tubes (Eppendorf® Protein LoBind) were used. All solutions were prepared fresh and filtered through a 0.22 µm syringe filter before use.

156

Protein lysates (preparation of protein lysates described above) were subjected to click chemistry with AzTB or AzRB capture reagent as described (7.1.10). After precipitation and re-suspension, the protein solution was centrifuged (4000 x g, 10 min, RT) to pellet any particulates. The clarified protein samples were incubated with NeutrAvidin® Agarose resin (30 µL per 1 mg protein, Thermo Scientific) for 2 hours at room temperature. The beads were pelleted (3,000 x g, 3 min) and the supernatant was removed. The beads were washed sequentially in 1% SDS in PBS (3 x 0.5 mL), 4M Urea in PBS (2 x 0.5 mL) and 50 mM ammonium bicarbonate (5 x 0.5 mL). For each wash step the beads were gently vortexed for 1 min followed by pelleting in a microcentrifuge (3,000 x g, 2-3 min).

After the final wash the beads were re-suspended in 50 mM ammonium bicarbonate (50 µL). DL-dithiothreitol (3 µL, 100 mM in 50 mM ammonium bicarbonate) was added and the beads incubated at 55°C for 30 minutes in a shaker. The beads were washed with 50 mM ammonium bicarbonate (2 x 0.5 mL) with vortexing and pelleting as before, leaving the beads covered in 50 µL solution after the second wash. Iodoacetamide (3 µL, 100 mM in 50 mM ammonium bicarbonate) was added and the beads incubated at room temperature for 30 minutes in the dark. The beads were washed as before. Sequence grade trypsin (5 µL, 0.2 µg/µL in 50 mM ammonium bicarbonate) was added and the beads incubated at 37°C overnight in a shaker. The beads were pelleted and the supernatant collected. The beads were washed with 0.1% formic acid in ultrapure water (80 µL) with gentle shaking for 10 minutes. The beads were pelleted and the supernatants pooled. The peptide solutions were purified on stage-tips or subjected to on-stagetip dimethyl labelling as described below before LC-MS/MS analysis.

7.1.15. Stage-tip purification of peptides

P200 pipette tips were fitted with 3 layers of SDB-XC extraction disks (Empore®) cut out using an in-house constructed tool. The pipettes were inserted into the hole of a microcentrifuge tube lid. The tips were activated by addition of methanol (150 µL) and the tips centrifuged (2,000 x g, 1-2 min). The wash was repeated with LC-MS/MS grade water (150 µL). The peptide solution was loaded into the tip and the tips centrifuged again. The water wash was repeated. The peptides were eluted into a clean microcentrifuge tube by addition of 79% acetonitrile in water (60 µL). The peptides were dried in a Savant SPD1010 SpeedVac® Concentrator (Thermo Scientific) and stored at -80°C until LC-MS/MS analysis. Peptides were reconstituted in 2% acetonitrile in water with 0.5% trifluoroacetic acid for LC-MS/MS analysis. Peptide samples were sonicated in a water bath for 20 minutes to aid solvation followed by centrifugation (13,000 xg, 10 min) to pellet any particulates before injection on the LC-MS/MS.

157

7.1.16. On-stagetip dimethyl labelling of peptides

On-stagetip dimethyl labelling was adapted from the procedure of Li et al.240

The following solutions were prepared freshly and kept on ice:

 PB 7.5: Mix of Na2HPO4 (2 mL, 50 mM) and 50 mM NaH2PO4 (7 mL, 50 mM)

 Light labelling buffer: PB 7.5 (90%, v:v), 4% (v:v) CH2O solution in water (5%, v:v), 0.6 M

NaBH3CN in water (5%, v:v)

 Medium labelling buffer: PB 7.5 (90%, v:v), 4% (v/v) CD2O solution in water (5%, v:v),

0.6 M NaBH3CN in water (5%, v:v)

The stagetips (constructed as in the previous section) were washed with methanol (150 µL) and water (150 µL), with centrifugation after each step to remove solvent (2,000 xg, 1 minute, RT). The tryptic peptides were added to the stagetip and loaded by centrifugation (2,000 xg, 2 minutes, RT). The peptides were washed with water (150 µL) followed by centrifugation (2,000 xg, 2 minutes, RT). Light or Medium labelling buffer (20 µL) was added and the samples centrifuged (100 xg, 5 minutes, RT). The centrifugation speed was adjusted to ensure that the peptides were exposed to labelling buffer for at least 5 min. The labelling step was repeated five times in total. The stagetips were centrifuged (2,000 xg, 2 minutes, RT) to remove any excess labelling buffer. The peptides were washed with water (150 µL) and centrifuged (2,000 xg, 1 minute, RT). The labelled peptides were eluted by addition of 79% (v:v) acetonitrile in water (60 µL) and centrifuging (2,000 xg, 2 minutes, RT). An equal volume of Light and Medium-labelled peptides were combined (1:1) and dried in a Savant SPD1010 SpeedVac® Concentrator (Thermo Scientific). The peptide pellet was stored at -80°C until LC-MS/MS analysis. Peptides were reconstituted in 2% acetonitrile in water with 0.5% trifluoroacetic acid as before for LC-MS/MS analysis.

7.2. Proteomics analysis

7.2.1. LC-MS/MS analysis

LC-MS/MS analysis was performed on an Easy nLC-1000 system coupled to a Q Exactive™ mass spectrometer via an easy-spray source (Thermo Fisher Scientific). Typically 2 µL peptide sample was loaded onto the column. Peptides were separated on an EASY-Spray™ Acclaim PepMap C18 column (50 cm × 75 μm inner diameter, Thermo Fisher Scientific) using a binary solvent system of 2% acetonitrile with 0.1% formic acid (Solvent A) and 80% acetonitrile with 0.1% formic acid (Solvent B) in an Easy nLC-1000 system (Thermo Fisher Scientific). 2 µL of peptide solution was loaded using Solvent A onto an Acclaim PepMap100

158

C18 trap column (2 cm x 75 μm inner diameter), followed by a linear gradient separation of 0- 100% Solvent B over 2 hours at a flow rate of 250 nL/min.

The Q Exactive™ was operated in data-dependent mode with survey scans acquired at a resolution of 75,000 at m/z 200 (transient time 256 ms). Up to 10 of the most abundant isotope patterns with charge +2 or higher from the survey scan were selected with an isolation window of 3.0 m/z and fragmented by HCD with normalized collision energies of 25. The maximum ion injection times for the survey scan and the MS/MS scans (acquired with a resolution of 17

6 500 at m/z 200) were 20 and 120 ms, respectively. The ion target value for MS was set to 10 and for MS/MS to 105, and the intensity threshold was set to 8.3 × 102.

7.2.2. Proteomics data processing in Maxquant

Processing of LC-MS/MS data was performed in MaxQuant version 1.5.4.1 or 1.5.5.1 using the built-in Andromeda search engine. Peptides were identified from the MS/MS spectra searched against the human reference proteome (Uniprot, accessed 16 July 2015). For non- quantitative analysis the multiplicity was set to 1. For duplex SILAC and spike-in SILAC experiments the multiplicity was set to 2 and “Arg10” and “Lys8” chosen as heavy labels. For di-methyl labelled samples the multiplicity was set to 2 and “DimethLys0” and “DimethNter0” were chosen as light labels and “DimethLys4” and “DimethNter4” were chosen as heavy labels. Cysteine carbamidomethylation was set as a fixed modification. Methionine oxidation and N-terminal acetylation were set as variable modifications. “Trypsin/P” was chosen as digestion mode enzyme. Minimum peptide length was set to 7 residues and maximum 2 missed cleavages were allowed. The “re-quantify” and “match between run” options were selected unless otherwise stated. “Unique and razor peptides” were chosen for protein quantification. PSM and protein FDR was set to 0.01. Other parameters were used as pre-set in the software. Processed data was analysed using Perseus version 1.5.4.1, Microsoft Office Excel 2010 and GraphPad Prism version 5.03.

7.2.3. Data analysis of initial proteomics comparison of YnF and YnGG probes

For Maxquant data processing, the “match between run” function was not enabled. Maxquant- processed data was loaded into Perseus. The data was filtered to remove proteins categorized as “Only ID by site”, “Reverse” and “Potential contaminant”. Data was filtered to retain proteins that were identified by a minimum of 3 unique peptides in the YnF and/or YnGG samples and a maximum of 2 peptides in the DMSO control sample. RAB proteins and proteins containing a CXXX C-terminal sequence were annotated.

159

7.2.4. Spike-in SILAC concentration-gradient or time-course labelling studies proteomics data analysis

For Maxquant data processing, the “match between run” function was not enabled. The “Ratio H/L normalized” values returned from Maxquant processing were loaded in Perseus. The data was filtered to remove proteins categorized as “Only ID by site”, “Reverse” and “Potential contaminant”. Data was filtered to retain proteins that were identified by a minimum of 3 unique + unique peptides in at least one time-point sample (for time-course study) or one probe concentration (for concentration study).

7.2.5. Identification of YnMyr-modified peptides

For Maxquant data processing, the “match between run” function was not enabled and a search for the following variable modifications was included to detect YnMyr peptides:

C25H42O5N6 (YnMyr-AzKB peptides), C25H42O5N8 (YnMyr-AzRB peptides), C34H60O6N10

(YnMyr-AzK+RB peptides) and C26H44O5N8 (YnMyr-AzRTB peptides). Modifications were set as specific for Gly residues on any peptide N-terminus.

Maxquant-processed data was loaded into Perseus. The data was filtered to remove proteins categorized as “Only ID by site”, “Reverse” and “Potential contaminant”. A minimum of 3 razor and unique peptides was required in at least one replicate sample (n=2) for a positive protein identification. For modified peptide identification a Maxquant score >20 was required.

7.2.6. Isoprenoid competition proteomics data analysis

7.2.7. Initial evaluation of prenyl transferase inhibitors (FTI-277, GGTI-2133 and Bon-15) data analysis

The “Ratio H/L normalized” values returned from Maxquant processing were loaded in Perseus. The data was filtered to remove proteins categorized as “Only ID by site”, “Reverse” and “Potential contaminant”. The ratios were inverted to obtain L/H ratios and log2 transformed. For each probe, the data-set was filtered to retain proteins that were identified with a minimum of 3 unique + razor peptides in the non-inhibitor treated control samples, and

160 further filtered to retain proteins quantified in at least 2 samples (non-inhibitor-treated control sample and at least one inhibitor-treated sample). All ratios were normalised with respect to the non-inhibitor treated samples.

7.2.8. Prenyl transferase inhibitor (FTI-277, Tipifarnib, Manumycin A and GGTI-2133) dose response data analysis

Data processed in Maxquant as described was analysed in Perseus. Each prenyl probe/inhibitor combination was processed separately. The data was filtered to remove proteins categorized as “Only ID by site”, “Reverse” and “Potential contaminant”. A minimum of 3 valid values per protein was required. “L/H ratios” were log2 transformed and a multi- sample ANOVA test was performed (Benjamini-Hochberg truncation, FDR 0.05).

To establish a relative response to inhibition the mean L/H ratio of each inhibitor concentration (n=3, n=2 for YnGG/GGTI 2.5 µM) was normalized by the mean L/H ratio of the non-inhibitor treated sample (n=3). The standard deviation of the normalized means was derived by the following relationship:

2 2 � �� = � × √(( �) + ( ) ) � � �

(where z is the normalized mean, x and y the original means, and σ the standard deviation).

Data was filtered to retain proteins with a prenylation motif (proteins bearing a C-terminal CXXX-motif and RAB proteins). ANOVA-significant proteins with a maximum relative response to inhibition of <0.4 were imported into Graphpad Prism (v. 5.03). For the purpose of IC50 calculations the non-inhibitor treated samples were approximated at an inhibitor concentration of 3 log units below the lowest inhibitor concentration tested (1 nM for FTI-277,

GGTI-2133 and Manumycin A samples, 1 pM for Tipifarnib samples). IC50 values were determined by non-linear regression using the “log(inhibitor) vs. response – Variable slope (four parameters)” function. Top values were constrained to equal 1 and bottom values to be greater than 0. Proteins that returned an ambiguous IC50 were excluded from further analysis.

7.2.9. Hypoxia-treated HPAEC data analysis

The “Ratio H/L normalized” values returned from Maxquant processing were loaded in Perseus. The data was filtered to remove proteins categorized as “Only ID by site”, “Reverse” and “Potential contaminant”. The ratios were inverted to obtain L/H ratios and L/H ratios of hypoxia-treated samples were normalised to the L/H ratios of normoxic controls. The

161 normalised L/H ratios were log2 transformed and a one-sample t-test was performed to identify ratios that were significantly different from 0 (p-value truncation, p<0.05).

7.2.10. Tipifarnib-treated HPAEC and HPSMC data analysis

7.2.11. HGPS versus healthy fibroblast data analysis

The “Ratio H/L normalized” values returned from Maxquant processing were loaded in Perseus. The data was filtered to remove proteins categorized as “Only ID by site”, “Reverse” and “Potential contaminant”. Data was filtered to retain proteins quantified in 3 of 6 replicates. The ratios were inverted to obtain L/H ratios and log2 transformed. The following comparisons were performed in a two-sample t-test (p-value truncation, p<0.05, min. 3 valid values per group): 1. HGPS versus healthy 2. Tipifarnib-treated HGPS versus HGPS (no inhibitor) 3. Tipifarnib-treated Healthy versus Healthy (no inhibitor) 4. Tipifarnib-treated HGPS versus Tipifarnib-treated Healthy To assess the level of YnF-labelling of LMNA in individual HGPS cell lines versus healthy cells, log2(L/H) ratio of LMNA from each of the 3 HGPS cell lines labelled with YnF (n=2 per cell line) were normalised to the average LMNA log2(L/H) ratio of all YnF-labelled healthy cells (n=6). A one-sample t-test was performed in Graphpad Prism to assess whether ratios were significantly different from 0 (p<0.05 and p<0.01).

7.3. Organic Synthesis

7.3.1. General

All reagents were purchased from commercial sources (WVR or Sigma Aldrich) and used without further purification. Oven-dried glassware was used for all anhydrous reactions and flasks flushed with inert gas (nitrogen or argon) prior to use. Reaction progress was monitored by thin layer chromatography (TLC). TLC was performed on TLC Silica gel 60 F254 aluminium sheets (MERCK) and spots were visualised by UV (where possible) or vanillin stain.

162

Flash column chromatography was performed on a Biotage Isolera™ One fitted with a dual wavelength detector, with fraction collection at 254 nm for UV active compounds. For non-UV active compounds all fractions were collected and presence of product confirmed by TLC.

Nuclear magnetic resonance spectra were recorded at room temperature on a 400 MHz Bruker instrument (400 MHz for 1H and 100 MHz for 13C). Chemical shifts (δ) are reported in

1 13 parts per million (ppm) relative to the CDCl3 solvent peak (7.26 for H and 77.16 for C). Coupling constants are reported in Hertz (Hz). High resolution mass spectrometry (HRMS) was performed by the Imperial College London Mass Spectrometry Service.

7.3.2. 3-Methyl-1-(tetrahydro-2H-pyran-2-yloxy)but-2-ene (1a)

To a solution of 3-methyl-2-buten-1-ol (3.8 g, 44 mmol) in DCM stirring at room temperature was added pyridinium p-toluenesulfonate (1.12 g, 4.4 mmol) and 3,4-dihydro-2H-pyran (16 mL, 175 mmol). The solution was stirred for 2 hours 15 min after which the solvent was removed in vacuo and the resulting residue taken up in Et2O. The organic layer was washed with saturated aqueous sodium bicarbonate (x3) and brine (x1), dried over Na2SO4, filtered and the solvent removed under reduced pressure. The crude residue was purified by flash column chromatography, eluting with 5% EtOAc in hexane. The product was isolated as a colourless oil (5.38 g, 32 mmol, 72%).

Rf = 0.45 (n-Hex – EtOAc, 9:1); δH/ppm (400 MHz, CDCl3) 5.44 – 5.32 (m, 1H), 4.69 – 4.61 (m, 1H), 4.25 (dd, J = 11.7, 6.6, 1H), 4.01 (dd, J = 11.7, 7.6, 1H), 3.95 – 3.85 (m, 1H), 3.60 –

3.47 (m, 1H), 1.92 – 1.48 (m, 6H), 1.78 (s, 3H), 1.71 (s, 3H); δC/ppm (100 MHz, CDCl3) 137.15, 120.84, 97.91, 63.69, 62.29, 30.72, 25.87, 25.51, 19.64, 17.99; m/z (Magnet CI+), 171 ([M +

+ + H] ); HRMS, found 171.1383 (C10H19O2, [M + H] , requires 171.1385).

7.3.3. (E)-2-Methyl-4-(tetrahydro-2H-pyran-2-yloxy)but-2-en-1-ol (2a)

To a solution of 1a (1.46 g, 8.6 mmol) in ethanol (15 mL) and pyridine (0.43 mL, 5.3 mmol) stirring at room temperature was added SeO2 (780 mg, 7.0 mmol) in portions over the duration of the reaction. The solution was stirred at room temperature overnight and then at 80°C for

24 hours. The reaction was cooled to room temperature, diluted with H2O and filtered to remove the selenium precipitate. The aqueous layer was extracted with EtOAc (x3) and the combined organic layers washed with brine (x3), dried over Na2SO4, filtered and evaporated

163 in vacuo. The crude residue was purified by column chromatography, eluting with a gradient of 5-50% EtOAc in n-hexane. Two products were collected, one corresponding to alcohol 2a, and the other to the corresponding aldehyde. The aldehyde (513 mg, 2.8 mmol) was dissolved in EtOH (25 mL) and treated with NaBH4 (104 mg, 2.8 mmol) at 0°C for 20 minutes. The reaction was diluted with EtOAc and the organic layer washed with H2O (x2) and brine (x2).

The organic layer was dried over Na2SO4, filtered and evaporated in vacuo. The crude product was combined with alcohol 2a obtained previously and again purified by column chromatography in a gradient of 5-50% EtOAc in n-hexane. The product was isolated as a pale yellow oil (487 mg, 2.6 mmol, 31%).

δH/ppm (400 MHz, CDCl3) 5.70 – 5.64 (m, 1H), 4.69 – 4.64 (m, 1H), 4.36 – 4.29 (m, 1H), 4.13 – 4.08 (m, 1H), 4.07 (s, 2H), 3.91 (m, 1H), 3.55 (m, 1H), 1.92 – 1.48 (m, 6H), 1.74 (s, 3H);

+ HRMS, found 204.1600 (C10H22NO3, [M + NH4] , requires 204.1600).

7.3.4. (E)-2-Methyl-1-(2-propynyloxy)-4-(tetrahydro-2H-pyran-2-yloxy)but-2-ene (3a)

A solution of 2a (38 mg, 0.20 mmol) and 18-crown-6 (422 mg, 1.60 mmol) in THF (1.5 mL) was added dropwise to a suspension of NaH (133 mg, 3.33 mmol, 60% dispersion in oil, pre- washed with THF) in THF (1.5 mL) stirring at 0°C under a nitrogen atmosphere. The resultant suspension was stirred at 0°C for 45 min, after which propargyl bromide (65 µL, 80% in toluene, 0.60 mmol) was added dropwise. The light brown suspension was heated to 50°C and stirred for 2.5 hours.

The reaction was cooled to room temperature, and quenched by addition of H2O. The suspension was diluted with DCM and the organic layer washed with H2O (x2) and brine (x2), dried over Na2SO4, filtered and evaporated in vacuo. The crude was taken up in a solution of EtOAc/hexane (1:1, v:v) and filtered through a silica plug. The solvent was evaporated. The product was carried through to the next step without further purification.

7.3.5. (E)-3-Methyl-4-(2-propynyloxy)but-2-en-1-ol (PeC5)

The crude product 3a was taken up in EtOH (1.5 mL). PPTS (6 mg, 0.024 mmol) was added and the solution was stirred at 60°C overnight. The solvent was removed in vacuo and the crude was purified by flash column chromatography, eluting in a gradient of 6-60% EtOAc in hexane. The product was isolated as a yellow oil (20.2 mg, 72% over two steps).

164

Rf = 0.17 (n-Hex – EtOAc, 7:3); δH/ppm (400 MHz, CDCl3) 5.72 – 5.65 (m, 1H), 4.21 (dd, J = 6.7, 0.5, 2H), 4.12 (d, J = 2.4, 2H), 3.96 (s, 2H), 2.43 (t, J = 2.4, 1H), 1.70 (s, 3H), 1.55 (br s,

1H); δC/ppm (100 MHz, CDCl3) 134.84, 127.04, 79.68, 74.89, 74.45, 59.06, 56.99, 14.07; m/z

+ + (Magnet CI+), 158 ([M + NH4] ), 141 ([M + H] ); HRMS, found 158.1182 (C8H16NO2,

+ [M + NH4] , requires 158.1181).

7.3.6. (E)-1-Azido-2-methyl-4-(tetrahydro-2H-pyran-2-yloxy)but-2-ene (4a)

To a solution of alcohol 2a (61 mg, 0.33 mmol) in toluene (2 mL) stirring at 0°C was added

(PhO)2PON3 (85 µL, 0.39 mmol) dropwise. The solution was stirred at 0°C under N2 for 45 min, after which diazabicyclo[5.4.0]undec-7-ene (72 µL, 0.49 mmol) was added and the reaction stirred at 0°C for 2 hours, then allowed to warm to room temperature and stirred for a further 20 hours. The reaction was quenched by addition of H2O, diluted with EtOAc and the organic layer washed with brine (2x). The brine layer was back-extracted with EtOAc, and the combined organic layers washed with 10% HCl, dried over Na2SO4, filtered and evaporated in vacuo. The crude was purified by flash column chromatography, eluting in a gradient of 2-20% EtOAc in hexane. The product was isolated as a colourless oil (26 mg, 0.12 mmol, 38%).

Rf = 0.33 (n-Hex – EtOAc, 9:1); δH/ppm (400 MHz, CDCl3) δ 5.69 – 5.57 (m, 1H), 4.69 – 4.58 (m, 1H), 4.34 – 4.23 (m, 1H), 4.14 – 4.04 (m, 1H), 3.92 – 3.82 (m, 1H), 3.71 (s, 2H), 3.58 – 3.46 (m, 1H), 1.75 (s, 3H), 1.91 – 1.48 (m, 6H).

7.3.7. 2-azido-3-methylbut-3-en-1-ol (major) and (2E)-4-azido-3-methylbut-2-en-1-ol (minor) (AzC5)

As for propargyl ether PeC5, using 4a (26 mg, 0.13 mmol) and PPTs (3.2 mg, 0.013 mmol) in EtOH (1 mL). The product was isolated as a pale yellow oil (8.0 mg, 0.063 mmol, 48%) containing two isomers with a relative abundance of 3:1 based on integration of NMR peaks.

δH/ppm (400 MHz, CDCl3) Major isomer: 5.11 – 5.02 (m, 2H), 4.08 (dd, J = 7.7, 4.8 Hz, 1H), 3.69 – 3.55 (m, 2H), 1.76 (s, 3H), Minor isomer: 5.68 (t, J = 6.0 Hz, 1H), 4.24 (d, J = 6.6 Hz, 2H), 3.72 (s, 2H), 1.75 (s, 3H).

165

7.3.8. (2E,6E,10E)-3,7,11,15-Tetramethyl-1-(tetrahydro-2H-pyran-2-yloxy)hexadeca- 2,6,10,14-tetraene (1b)

To a stirred solution of geranylgeraniol (116 mg, 0.40 mmol) in DCM (5 mL) under N2 was added pyridinium p-toluenesulfonate (10 mg, 0.04 mmol) and 3,4-dihydro-2H-pyran (150 µL, 1.6 mmol). The solution was stirred at room temperature for 26 hours after which the solvent was removed in vacuo and the resulting residue taken up in Et2O. The organic layer was washed with saturated sodium bicarbonate (x3) and brine (x1), dried over Na2SO4, filtered, and evaporated. The crude was purified by flash column chromatography eluting in a gradient of 3-12% EtOAc in hexane to yield the product as a colourless oil (125 mg, 0.33 mmol, 83%).

Rf = 0.61 (n-Hex – EtOAc, 7:3); δH/ppm (400 MHz, CDCl3) 5.39 (t, J = 6.9, 1H), 5.17 – 5.08 (m, 3H), 4.65 (dd, J = 4.1, 3.0, 1H), 4.27 (dd, J = 11.7, 6.3, 1H), 4.06 (dd, J = 11.9, 7.4, 1H), 3.96 – 3.88 (m, 1H), 3.59 – 3.49 (m, 1H), 2.19 – 2.04 (m, 8H), 2.04 – 1.95 (m, 4H), 1.92 – 1.73

(m, 2H), 1.71 (s, 6H), 1.63 (s, 9H), 1.61 – 1.53 (m, 4H); δC/ppm (100 MHz, CDCl3) 140.33, 135.29, 134.96, 131.29, 124.40, 124.22, 123.91, 120.55, 97.80, 63.66, 62.30, 39.73 (2C), 39.67, 30.73, 26.77, 26.65, 26.32, 25.72, 25.52, 19.64, 17.70, 16.45, 16.04, 16.02; m/z (ES+), 375 ([M + H]+)

7.3.9. (2E,6E,10E,14E)-2,6,10,14-Tetramethyl-16-(tetrahydro-2H-pyran-2- yloxy)hexadeca-2,6-10,14-tetraen-1-ol (2b)

To a solution of 1b (125 mg, 0.33 mmol) in DCM (1.5 mL) stirring under Argon at 0°C was

t added SeO2 (7.4 mg, 66 µmol) followed by dropwise addition of BuOOH (5 M in decane, 100 µL, 0.48 mmol). The reaction was stirred at 0°C for 130 min. The solvent was removed in vacuo and the resulting residue taken up in Et2O. The organic layer was washed with H2O (x2) and brine (x2), dried over Na2SO4, filtered and evaporated in vacuo. The product was purified by flash column chromatography and isolated as a colourless oil (17 mg, 0.045 mmol, 14%).

Rf = 0.39 (n-Hex – EtOAc, 7:3); δH/ppm (400 MHz, CDCl3) 5.48 – 5.33 (m, 2H), 5.13 (dd, J = 14.0, 7.0, 2H), 4.68 – 4.61 (m, 1H), 4.26 (dd, J = 11.7, 6.3, 1H), 4.09 – 4.04 (m, 1H), 4.02 (s, 2H), 3.96 – 3.85 (m, 1H), 3.59 – 3.48 (m, 1H), 2.20 – 1.96 (m, 12H), 1.71 (s, 3H), 1.70 (s, 3H),

+ 1.63 (s, 6H), 1.90 – 1.42 (m, 6H); HRMS, found 413.3040 (C25H42O3Na, [M + Na] , requires 413.3032).

166

7.3.10. (2E,6E,10E,14E)-2,6,10,14-Tetramethyl-1-(2-propynyloxy)-16-(tetrahydro-2H- pyran-2-yloxy)hexadeca-2,6,10,14-tetraene (3b)

As for 3a using 2b (10 mg, 0.026 mmol), NaH (60% in oil, 16.3 mg, 0.41 mmol), 18-crown-6 (55 mg, 0.21 mmol) and propargyl bromide (9 µL, 0.10 mmol). The crude product was carried through to the next step without purification or characterisation.

7.3.11. (2E,6E,10E,14E)-3,7,11,15-Tetramethyl-16-(2-propynyloxy)hexadeca-2,6,10,14- tetraen-1-ol (PeC20)

As for PeC5 using crude 3b, PPTS (1 mg, 4 µmol) in EtOH (1 mL). The product was isolated as a pale yellow oil (4.4 mg, 0.013 mmol, 49% over 2 steps).

δH/ppm (400 MHz, CDCl3) 5.49 – 5.41 (m, 2H), 5.14 (t, J = 6.7, 2H), 4.18 (d, J = 6.9, 2H), 4.10 (d, J = 2.4, 2H), 3.96 (s, J = 4.9, 2H), 2.44 (t, J = 2.3, 1H), 2.19 – 1.97 (m, 12H), 1.69 (d, J =

12.6, 3H), 1.68 (s, 3H), 1.63 (s, 6H); δC/ppm (100 MHz, CDCl3) 139.94, 138.27, 136.87, 131.14, 129.51, 124.55, 123.82, 123.32, 86.65, 75.88, 74.07, 59.42, 56.27, 39.67, 39.57,

39.22, 26.62, 26.38, 26.31, 16.30, 16.03, 15.97, 13.94; HRMS, found 367.2609 (C23H36O2Na, [M + Na]+, requires 367.2613).

7.3.12. (2E,6E,10E,14E)-1-Azido-2,6,10,14-tetramethyl-16-(tetrahydro-2H-pyran-2- yloxy)hexadeca-2,6,10,14-tetraene (4b)

To a solution of alcohol 2b (17.3 mg, 0.044 mmol) in toluene (1.5 mL) stirring at 0°C under Ar was added (PhO)2PON3 (35 µL, 0.12 mmol) in two portions over 2.5 hours, after which diazabicyclo[5.4.0]undec-7-ene (20 µL, 0.13 mmol) was added and the reaction stirred at 0°C for 2 hours, then allowed to warm to room temperature overnight. The reaction was quenched by addition of H2O, diluted with EtOAc and the organic layer washed with brine (x3), dried over

Na2SO4, and evaporated in vacuo. The crude residue was purified by flash column chromatography in a gradient of 0-30% EtOAc in hexane. The product was isolated as a pale yellow oil (10.7 mg, 0.026 mmol, 59%). The product was carried through to the deprotection in the next step without further characterisation.

167

7.3.13. (2E,6E,10E,14E)‐16-azido-3,7,11,15‐tetramethylhexadec‐2,6,10,14‐tetraen‐1‐ol (AzC20)

As for PeC5 using 4b (10 mg, 0.024 mmol) and PPTS (1.6 mg, 6 μmol) in EtOH (1 mL). The product was purified by column chromatography eluting in 15-50% EtOAc in n-hexane. The product was isolated as a pale yellow oil (7.2 mg, 0.022 mmol, 90%).

δH/ppm (400 MHz, CDCl3) 5.45 (t, J = 6.8 Hz, 1H), 5.14 (t, J = 6.8 Hz, 3H), 4.18 (d, J = 6.9 Hz, 2H), 3.67 (s, 2H), 2.28 – 1.95 (m, 12H), 1.71 (s, 6H), 1.63 (s, 6H).

7.3.14. (6E,10E)-6,10-Dimethyl-12-(tetrahydro-2H-pyran-2-yloxy)dodeca-6,10-dien-1- yne (7)

To a solution of 10 (25 mg, 0.27 mmol) in anhydrous Et2O (0.6 mL) and DMSO (0.6 mL) stirring under nitrogen at 0°C was added lithium acetylide ethylenediamine complex (20.6 mg, 53

µmol) in Et2O (0.6 mL). The solution was stirred at room temperature for 17 hours. The reaction was diluted with saturated ammonium chloride and extracted with Et2O (x3). The combined organic layers were washed with brine (x1), dried over Na2SO4 and evaporated in vacuo. The crude (13 mg) was used in the next step without further purification.

7.3.15. (2E,6E)-3,7-dimethyldodeca-2,6-dien-11-yn-1-ol (YnC14)

A solution of 11 (13 mg crude mass) and pyridinium p-toluenesulfonate (1.2 mg, 5 µmol) in EtOH (0.8 mL) was stirred at 60°C for 3 hours. The solvent was removed in vacuo and the crude residue purified by flash column chromatography, eluting with 25% EtOAc in n-hexane. The product was isolated as a pale yellow oil (5.0 mg, 24 µmol, 43% over 2 steps).

Rf = 0.52 (n-Hex – EtOAc, 1:1); δH/ppm (400 MHz, CDCl3) 5.47 – 5.40 (m, 1H), 5.19 – 5.12 (m, 1H), 4.18 (d, J = 6.7, 2H), 2.17 (td, J = 7.2, 2.7, 2H), 2.14 – 2.03 (m, 6H), 1.98 (t, J = 2.6,

1H), 1.70 (s, 3H), 1.69 – 1.62 (m, 2H), 1.62 (s, 3H); δC/ppm (100 MHz, CDCl3) 139.62, 134.24, 124.71, 123.50, 84.68, 68.29, 59.43, 39.48, 38.46, 26.62, 26.18, 17.74, 16.23, 15.82; HRMS,

+ found 189.1642 (C14H21, [M – H2O + H] , requires 189.1643).

168

7.3.16. (2E,6E,10E)-14-bromo-3,7,11,15-tetramethylhexadeca-2,6,10-triene-1,15-diol (8)

To a solution of geranylgeraniol (125 mg, 0.43 mmol) in THF (3 mL) and H2O (1.5 mL) stirring at 0°C was added N-bromosuccinimide (92 mg, 0.52 mmol). The solution was stirred at 0°C for 5 minutes, then at room temperature for 2.5 hours, after which the reaction was diluted with H2O and extracted with Et2O (x3). The combined organic layers were dried over Na2SO4, filtered and evaporated in vacuo. The crude was purified by flash column chromatography eluting in a gradient of 5-55% EtOAc in n-hexane. The product was isolated as a colourless oil (77 mg, 0.20 mmol, 46%).

Rf = 0.32 (n-Hex – EtOAc, 1:1); δH/ppm (400 MHz, CDCl3) 5.45 (td, J = 6.9, 1.1, 1H), 5.22 (t, J = 6.9, 1H), 5.14 (t, J = 6.3, 1H), 4.18 (d, J = 6.8, 2H), 4.01 (dd, J = 11.4, 1.8, 1H), 2.40 – 2.29 (m, 1H), 2.21 – 1.94 (m, 10H), 1.87 – 1.75 (m, 2H), 1.71 (s, 3H), 1.63 (s, 3H), 1.62 (s,

3H), 1.37 (s, 3H), 1.36 (s, 3H); δC/ppm (100 MHz, CDCl3) 139.81, 135.14, 133.02, 125.95, 123.97, 123.32, 72.48, 70.99, 59.43, 39.59, 39.55, 38.16, 32.11, 26.62, 26.56, 26.28, 25.78,

+ 16.32, 16.01, 15.86; HRMS, found 369.1784 (C20H34OBr, [M – H2O + H] , requires 369.1793).

7.3.17. (2E,6E,10E)-13-(3,3-dimethyloxiran-2-yl)-3,7,11-trimethyltrideca-2,6,10-trien-1- ol (9)

To a solution of 8 (77 mg, 0.20 mmol) in MeOH (3 mL) was added K2CO3 (56 mg, 0.40 mmol). The solution was stirred at room temperature for 1 hour 45 minutes. The reaction was diluted with Et2O and saturated ammonium chloride, and extracted with EtOAc (x3). The combined organic layers were dried over Na2SO4, filtered and evaporated in vacuo. The crude was purified by flash column chromatography, eluting in 20% EtOAc in n-Hex. The product was isolated as a colourless oil 39 mg, 0.13 mmol, 64%).

Rf = 0.54 (n-Hex – EtOAc, 1:1); δH/ppm (400 MHz, CDCl3) 5.44 (td, J = 6.9, 1.2, 1H), 5.18 (td, J = 6.9, 1.1, 1H), 5.13 (td, J = 6.7, 0.9, 1H), 4.18 (d, J = 6.9, 2H), 2.73 (t, J = 6.3, 1H), 2.22 –

1.96 (m, 12H), 1.71 (s, 3H), 1.64 (s, 3H), 1.63 (s, 3H), 1.33 (s, 3H), 1.29 (s, 3H); δC/ppm (100

MHz, CDCl3) 139.72, 135.23, 134.06, 124.83, 123.91, 123.39, 64.24, 59.41, 58.42, 39.61, 39.55, 36.31, 27.47, 26.57, 26.27, 24.93, 18.76 (2C), 16.30, 16.02; HRMS, found 329.2451

+ (C20H34O2Na, [M + Na] , requires 329.2457).

169

7.3.18. (4E,8E,12E)-14-Hydroxy-4,8,12-trimethyltetradeca-4,8,12-trienal (10)

To a solution of epoxide 9 (102 mg, 0.34 mmol) in THF (1.5 mL) stirring at 0°C was added

NaIO4 (145 mg, 0.67 mmol) in H2O (1.8 mL) followed by H5IO6 (38 mg, 0.17 mmol) in THF (2.5 mL). The reaction was stirred at 0°C for 1.5 hours. The reaction was quenched by addition of 1M HCl and extracted with EtOAc (x3). The combined organic layers were washed with sodium bicarbonate, dried over MgSO4, filtered, and evaporated in vacuo to yield a pale yellow oil which was used in the next step without further purification.

7.3.19. (4E,8E,12E)-4,8,12-Trimethyl-14-(tetrahydro-2H-pyran-2-yloxy)tetradeca-4,8,12- trienal (11)

To solution of 10 (91 mg crude weight) and pyridinium p-toluenesulfonate (10 mg, 0.40 mmol) in DCM (3 mL) stirring under nitrogen was added 3,4-dihydro-2H-pyran (150 µL, 1.64 mmol). The solution was stirred at room temperature overnight. The solvent was removed in vacuo and the residue taken up in Et2O, washed with sodium bicarbonate (x2) and brine (x2), dried over Na2SO4, filtered and evaporated in vacuo. The crude was purified by flash column chromatography eluting in a gradient of EtOAc in n-hexane. The product was isolated as a pale yellow oil (64 mg, 0.18 mmol, 53% over 2 steps).

Rf = 0.64 (n-Hex – EtOAc, 1:1); δH/ppm (400 MHz, CDCl3) 9.80 – 9.76 (m, 1H), 5.39 (t, J = 6.3, 1H), 5.20 – 5.08 (m, 2H), 4.68 – 4.63 (m, 1H), 4.27 (dd, J = 11.8, 6.4, 1H), 4.05 (dd, J = 11.8, 7.3, 1H), 3.96 – 3.88 (m, 1H), 3.58 – 3.50 (m, 1H), 2.54 (td, J = 7.3, 1.6, 2H), 2.34 (t, J = 7.4, 2H), 2.19 – 1.96 (m, 8H), 1.91 – 1.73 (m, 2H), 1.71 (s, 3H), 1.64 (s, 3H), 1.62 (s, 3H), 1.60 – 1.53 (m, 4H).

7.3.20. (4E,8E,12E)-4,8,12-trimethyl-14-(oxan-2-yloxy)tetradeca-4,8,12-trien-1-ol (12)

To a solution of aldehyde 11 (52 mg, 0.15 mmol) in MeOH (2.5 mL) stirring at 0°C was added

NaBH4 (8.5 mg, 0.22 mmol). The suspension was stirred at 0°C for 1 hour, after which the reaction was quenched by addition of H2O and the aqueous layer extracted with EtOAc (x3).

The combined organic layers were washed with brine (x1), dried over Na2SO4, filtered and

170 evaporated in vacuo. The crude was purified by flash column chromatography, eluting in a gradient of 2-20% EtOAc in n-hexane. The product was isolated as a colourless oil (33.5 mg, 0.10 mmol, 64%).

Rf = 0.46 (n-Hex – EtOAc, 1:1); δH/ppm (400 MHz, CDCl3) 5.39 – 5.31 (m, 1H), 5.18 – 5.06 (m, 2H), 4.65 – 4.59 (m, 1H), 4.28 – 4.19 (m, 1H), 4.03 (dd, J = 11.8, 7.4, 1H), 3.93 – 3.84 (m, 1H), 3.62 (t, J = 6.5, 2H), 3.55 – 3.47 (m, 1H), 2.15 – 1.94 (m, 10H), 1.89 – 1.69 (m, 2H), 1.68 (s, 3H), 1.67 – 1.62 (m, 2H), 1.61 (s, 3H), 1.59 (s, 3H), 1.58 – 1.47 (m, 4H).

7.3.21. (2E)-3,7-Dimethyl-1-(tetrahydro-2H-pyran-2-yloxy)octa-2,6-diene (14a)

To a solution of trans-geraniol (8.34 g, 54 mmol) and pyridinium p-toulenesulfonate (1.35 g, 5.4 mmol) in anhydrous dichloromethane (10 mL) under nitrogen was added 2,4-dihydropyran (19.7 mL, 216 mmol). The solution was stirred at room temperature overnight after which the solvent was removed in vacuo and the resulting residue taken up in diethyl ether. The organic layer was washed with saturated sodium bicarbonate (x2) and brine (x2), dried over Na2SO4, filtered, and evaporated. The crude was purified by flash column chromatography eluting with 5% ethyl acetate in n-hexane. The product was isolated as a colourless liquid (6.73 g, 28 mmol, 52%).

Rf = 0.42 (n-Hex – EtOAc, 9:1); δH/ppm (400 MHz, CDCl3) 5.40 – 5.29 (m, 1H), 5.12 – 5.02 (m, 1H), 4.64 – 4.57 (m, 1H), 4.22 (dd, J = 11.6, 6.2 Hz, 1H), 4.02 (dd, J = 11.9, 7.4 Hz, 1H), 3.88 (ddd, J = 11.2, 7.6, 3.5 Hz, 1H), 3.55 – 3.42 (m, 1H), 2.15 – 1.93 (m, 4H), 1.66 (s, 6H),

1.59 (s, 3H), 1.90-1.44 (m, 6H); δC/ppm (100 MHz, CDCl3) 140.35, 131.70, 124.14, 120.68, 97.85, 63.73, 62.38, 39.74, 30.83, 26.47, 25.79, 25.62, 19.74, 17.78, 16.50; HRMS, found

+ 256.2255 (C15H30NO2, [M + NH4] , requires 256.2277).

7.3.22. (2E,6E)-2,6-Dimethyl-8-(tetrahydro-2H-pyran-2-yloxy)octa-2,6-dien-1-ol (15a)

To a solution of 14a (8.60 g, 36 mmol) in anhydrous dichloromethane (40 mL) stirring under

t nitrogen at 0°C was added SeO2 (400 mg, 3.6 mmol) followed by BuOOH (5.5 M in decane, 7.25 mL, 40 mmol) dropwise. The solution was allowed to warm to room temperature and stirred for 24 hours. The reaction was diluted with dichloromethane, and the organic layer washed with water (x2) and brine (x2), dried over Na2SO4, filtered and evaporated in vacuo.

171

The crude was purified by flash column chromatography eluting in a gradient of 10-30% ethyl acetate in n-hexane. The isolated product was a pale yellow oil (3.27 g, 12.9 mmol, 36%).

Rf = 0.23 (n-Hex – EtOAc, 7:3); δH/ppm (400 MHz, CDCl3) 5.43 – 5.30 (m, 2H), 4.64 – 4.60 (m, 1H), 4.23 (dd, J = 11.9, 6.4 Hz, 1H), 4.01 (dd, J = 12.1, 7.6 Hz, 3H), 3.98 (s, 2H), 3.93 – 3.83 (m, 1H), 3.55 – 3.46 (m, 1H), 2.17 (td, J = 8.5, 7.7 Hz, 2H), 2.07 (t, J = 7.5 Hz, 2H), 1.90

– 1.45 (m, 6H), 1.67 (s, 3H), 1.66 (s, 3H); δC/ppm (100 MHz, CDCl3) 139.82, 135.27, 125.81, 121.15, 97.99, 69.10, 63.82, 62.42, 39.28, 30.81, 25.86, 25.61, 19.72, 16.47, 13.83; HRMS,

+ found 272.2230 (C15H30NO3, [M + NH4] , requires 272.2226).

7.3.23. (2E,6E)-1-Bromo-2,6-dimethyl-8-(tetrahydro-2H-pyran-2-yloxy)octa-2,6-diene (16a)

To a suspension of N-bromosuccinimide (2.48 g, 13.9 mmol) in anhydrous dichloromethane (20 mL) stirring under nitrogen at 0°C was added dimethylsulfide (1.23 mL, 16.7 mmol) dropwise. The bright orange suspension was stirred at 0°C for 15 minutes then cooled to - 40°C. 15a (2.36 g, 9.28 mmol) in DCM (1 mL) was added dropwise during which time the colour of the reaction faded to a pale yellow. The reaction was allowed to warm to room temperature overnight. The reaction was poured into brine, and the organic layer was washed with water (x2) and brine (x1), dried over Na2SO4, filtered and evaporated in vacuo without heating (IMPORTANT: heating this compound above room temperature will result in decomposition). The crude was purified by flash column chromatography eluting in a gradient of 1-20% ethyl acetate in n-hexane. The isolated product was a pale yellow oil (498 mg, 1.57 mmol, 16.9%).

Rf = 0.47 (n-Hex – EtOAc, 8:2); δH/ppm (400 MHz, CDCl3) 5.58 (t, J = 6.8 Hz, 1H), 5.36 (t, J = 6.8 Hz, 1H), 4.65 – 4.59 (m, 1H), 4.24 (dd, J = 11.9, 6.4 Hz, 1H), 4.02 (dd, J = 11.9, 7.3 Hz, 1H), 3.96 (s, 2H), 3.93 – 3.84 (m, 1H), 3.56 – 3.46 (m, 1H), 2.21 – 2.11 (m, 2H), 2.11 – 2.02

(m, 2H), 1.90 – 1.37 (m, 6H), 1.75 (s, 3H), 1.67 (s, 3H); δC/ppm (100 MHz, CDCl3) 139.45, 132.35, 130.95, 121.31, 98.02, 63.75, 62.45, 41.86, 38.79, 30.84, 26.64, 25.62, 19.76, 16.54,

+ 14.81; HRMS, found 334.1376 (C15H29NO2Br, [M + NH4] , requires 334.1382).

172

7.3.24. Ethyl(4E,8E)-4,8-dimethyl-10-(tetrahydro-2H-pyran-2-yloxy)deca-4,8-dienoate (17a)

Preparation of lithium di-isopropylamine (LDA): n-Butyllithium (2.5M in hexanes, 1.25 mL, 3.1 mmol) was added dropwise to a stirring solution of diisopropylamine (440 µL, 3.1 mmol) in anhydrous tetrahydrofuran (3 mL) at -78°C. The solution was stirred at 0°C for 15 minutes, and then cooled again to -78°C.

The LDA was transferred dropwise to a suspension of ethyl acetate (307 µL, 3.1 mmol) and CuI (1.2 g, 6.3 mmol) in dry THF (3 mL) stirring at -78°C. Transfer was done in 1 mL aliquots to prevent warming of the LDA. The resulting suspension was stirred at -78°C for 30 minutes after which a solution of bromide 16a (498 mg, 1.57 mmol) in THF (2 mL) was cooled to -78°C and added dropwise. The resulting brown suspension was stirred at -78°C for 10 min, quenched by addition of saturated ammonium chloride and extracted with Et2O (x3). The combined organic layers were washed with water, brine and saturated ammonium bicarbonate, dried over Na2SO4, filtered and evaporated in vacuo to yield a yellow oil. The crude product was purified by flash column chromatography eluting in a gradient of 3-10% ethyl acetate in n-hexane. The product was isolated as a colourless oil (474 mg, 1.46 mmol, 93%).

Rf = 0.39 (n-Hex – EtOAc, 8:2); δH/ppm (400 MHz, CDCl3) 5.39 – 5.30 (m, 1H), 5.17 – 5.10 (m, 1H), 4.62 (dd, J = 4.2, 3.0 Hz, 1H), 4.23 (dd, J = 11.9, 6.4 Hz, 1H), 4.11 (q, J = 7.1 Hz, 2H), 4.02 (dd, J = 11.8, 7.4 Hz, 1H), 3.89 (ddd, J = 11.2, 7.6, 3.5 Hz, 1H), 3.55 – 3.46 (m, 1H), 2.43 – 2.33 (m, 2H), 2.33 – 2.24 (m, 2H), 2.16 – 2.06 (m, 2H), 2.06 – 1.99 (m, 2H), 1.90 – 1.45

(m, 6H), 1.67 (s, 3H), 1.60 (s, 3H), 1.24 (t, J = 7.1 Hz, 3H); δC/ppm (100 MHz, CDCl3) 173.64, 140.18, 133.76, 124.85, 120.84, 97.97, 63.78, 62.44, 60.37, 39.59, 34.80, 33.39, 30.85, 26.35,

+ 25.63, 19.77, 16.55, 16.07, 14.40; HRMS, found 347.2193 (C19H32O4Na, [M + Na] , requires 347.2198).

7.3.25. (4E,8E)-4,8-Dimethyl-10-(tetrahydro-2H-pyran-2-yloxy)deca-4,8-dien-1-ol (18a)

To a solution of ester 17a (470 mg, 1.45 mmol) in dry THF (5 mL) stirring at 0°C under nitrogen was added LiAlH4 (83 mg mg, 2.17 mmol) and the reaction stirred for 30 minutes, then

173 quenched by addition of saturated ammonium chloride. The aqueous layer was extracted with

Et2O (x3) and the combined organic layers washed with saturated sodium bicarbonate (x1), brine (x2), dried over Na2SO4, filtered and evaporated in vacuo. The crude was purified by flash column chromatography eluting in a gradient of 1-30% acetone in n-hexane. The product was isolated as a pale yellow oil (349 mg, 1.23 mmol, 85%).

Rf = 0.58 (n-Hex – EtOAc, 1:1); δH/ppm (400 MHz, CDCl3 5.38 – 5.31 (m, 1H), 5.17 – 5.09 (m, 1H), 4.62 (dd, J = 4.3, 2.9 Hz, 1H), 4.28 – 4.20 (m, 1H), 4.02 (dd, J = 11.7, 7.3 Hz, 1H), 3.89 (ddd, J = 11.2, 7.5, 3.7 Hz, 1H), 3.61 (t, J = 6.5 Hz, 2H), 3.54 – 3.47 (m, 1H), 2.19 – 2.09 (m,

2H), 2.05 (t, J = 7.3 Hz, 4H), 1.89 – 1.46 (m, 8H), 1.66 (s, 3H), 1.60 (s, 3H); δC/ppm (100 MHz,

CDCl3) 139.91, 134.79, 124.54, 120.83, 97.99, 63.79, 62.47, 62.41, 39.51, 35.87, 30.71,

+ 30.42, 26.02, 25.48, 19.67, 16.31, 15.84; HRMS, found 300.2542 (C17H34NO3, [M + NH4] , requires 300.2539).

7.3.26. (4E,8E)-1-Iodo-4,8-dimethyl-10-(tetrahydro-2H-pyran-2-yloxy)deca-4,8-diene (19a)

To a solution of alcohol 18a (190 mg, 0.67 mmol), imidazole (64 mg, 0.94 mmol) and triphenylphosphine (265 mg, 1.01 mmol) in dry acetonitrile stirring at 0°C under nitrogen was added iodine (239 mg, 0.94 mmol). The solution was stirred at 0°C for 2 hours, and room temperature for 30 mins. The reaction was quenched by addition of saturated sodium thiosulfate and extracted with diethyl ether (3x). The combined organic layers were dried

(Na2SO4), filtered and evaporated to yield a yellow crude which was purified by flash column chromatography in a gradient of 1-10% ethyl acetate in n-hexane. The isolated product was a pale yellow oil (146 mg, 0.37 mmol, 56%).

Rf = 0.50 (n-hexane – ethyl acetate, 8:2); δH/ppm (400 MHz, CDCl3) 5.39 – 5.32 (m, 1H), 5.20 – 5.13 (m, 1H), 4.63 (dd, J = 4.1, 3.0 Hz, 1H), 4.27 – 4.20 (m, 1H), 4.02 (dd, J = 11.8, 7.4 Hz, 1H), 3.89 (ddd, J = 11.2, 7.7, 3.6 Hz, 1H), 3.55 – 3.46 (m, 1H), 3.14 (t, J = 6.9 Hz, 2H), 2.16 – 2.00 (m, 6H), 1.90 (t, J = 7.1 Hz, 2H), 1.87 – 1.45 (m, 6H), 1.68 (s, 3H), 1.58 (s, 3H); HRMS,

+ found 410.1533 (C17H33NO2I, [M +NH4] , requires 410.1556).

174

7.3.27. Trimethyl((7E,11E)‐7,11‐dimethyl‐13‐(tetrahydro‐2H‐pyran‐2‐yloxy)trideca‐ 7,11‐dien‐1‐yn‐1‐yl)silane (20a)

To a solution of 1-(trimethylsilyl)propyne (220 µL, 1.49 mmol) in dry THF (5 mL) stirring under nitorgen at -20°C was added n-BuLi (2.5 M, 0.6 mL, 1.49 mmol) dropwise. The solution was stirred at -20°C for 45 minutes during which a pale yellow colour developed. Iodide 19a (490 mg, 1.06 mmol, 1 eq.) in THF (1 mL) was added dropwise and the reaction stirred at -20°C for

1 hour. The reaction was diluted with H2O and extracted with diethyl ether (3x). The combined organic layers were washed with saturated sodium bicarbonate and brine, dried over Na2SO4, filtered and evaporated in vacuo. The crude was purified by flash column chromatography eluting in a gradient of 1-10% ethyl acetate in n-hexane. The isolated product was a colourless oil (116 mg, 0.31 mmol, 83%).

Rf = 0.50 (n-Hex – EtOAc, 9:1); δH/ppm (400 MHz, CDCl3) 5.40 – 5.29 (m, 1H), 5.10 (dt, J = 6.9, 3.4, 1H), 4.66 – 4.57 (m, 1H), 4.23 (dd, J = 11.9, 6.4, 1H), 4.02 (dd, J = 11.9, 7.4, 1H), 3.94 – 3.83 (m, 1H), 3.56 – 3.45 (m, 1H), 2.26 – 2.16 (m, 2H), 2.17 – 2.07 (m, 2H), 2.07 – 2.01 (m, 2H), 1.97 (t, J = 6.5, 2H), 1.90 – 1.38 (m, 10H), 1.67 (s, 3H), 1.58 (s, 3H), 0.14 (s, 9H);

δC/ppm (100 MHz, CDCl3) 140.35, 135.18, 124.25, 120.77, 107.77, 97.93, 84.50, 63.78, 62.42, 39.77, 39.15, 30.87, 28.18, 27.03, 26.41, 25.66, 19.85, 19.77, 16.56, 15.93, 0.32 (3C); HRMS,

+ found 377.2883 (C23H41O2Si, [M +H] , requires 377.2876).

7.3.28. (2E,6E)‐3,7‐Dimethyl-13-(trimethylsilyl)trideca‐2,6‐dien‐12‐yn‐1‐ol (21a)

A solution of 20a (116 mg, 0.31 mmol) and pyridinium p-toluenesulfonate (7.7 mg, 0.03 mmol) in ethanol (5 mL) was stirred at 60°C for 4 hours. The solvent was evaporated in vacuo and the crude purified by flash column chromatography eluting in a gradient of 5-30% acetone in n-hexane. The isolated product (77 mg, 0.26 mmol) was a colourless oil.

Rf = 0.25 (n-hexane – acetone, 8:2); δH/ppm (400 MHz, CDCl3) 5.45 – 5.37 (m, 1H), 5.14 – 5.06 (m, 1H), 4.15 (d, J = 6.9 Hz, 2H), 2.26 – 2.18 (m, 2H), 2.16 – 2.07 (m, 2H), 2.07 – 2.00 (m, 2H), 2.00 – 1.93 (m, 2H), 1.68 (s, 3H), 1.58 (s, 3H), 1.52 – 1.40 (m, 4H), 0.14 (s, 9H);

δC/ppm (100 MHz, CDCl3) 139.84, 135.27, 124.16, 123.56, 107.81, 84.53, 59.55, 39.66, 39.12,

175

28.14, 27.00, 26.36, 19.86, 16.40, 15.91, 0.31 (3C); HRMS, found 310.2579 (C18H36NOSi,

+ [M +NH4] , requires 310.2566).

7.3.29. (2E,6E)‐3,7‐Dimethyltrideca‐2,6‐dien‐12‐yn‐1‐ol (YnF)

To a solution of 21a (75 mg, 0.25 mmol) in dry THF (5 mL) stirring at room temperature under argon was added tetrabutyl-ammonium fluoride (1 M in THF, 0.5 mL, 0.5 mmol). The reaction was stirred for 15 minutes after which the solvents were removed in vacuo. The crude was purified by flash column chromatography eluting in a gradient of 3-30% acetone in n-hexane. The isolated product was a colourless oil (32 mg, 0.15 mmol, 59%).

Rf = 0.21 (n-hexane – acetone, 8:2); δH/ppm (400 MHz, CDCl3) 5.40 (tq, J = 6.9, 1.2 Hz, 1H), 5.14 – 5.04 (m, 1H), 4.14 (d, J = 6.9 Hz, 2H), 2.22 – 2.15 (m, 2H), 2.15 – 2.07 (m, 2H), 2.07 – 2.00 (m, 2H), 2.00 – 1.95 (m, 2H), 1.94 (t, J = 2.7 Hz, 1H), 1.67 (s, 3H), 1.58 (s, 3H), 1.52 –

1.44 (m, 4H), 1.42 (s, 1H); δC/ppm (100 MHz, CDCl3) 139.75, 135.17, 124.17, 123.56, 84.82, 68.28, 59.51, 39.63, 39.11, 28.05, 26.98, 26.31, 18.40, 16.36, 15.93; HRMS, found 238.2165

+ (C15H28NO, [M +NH4] , requires 238.2171).

7.3.30. (2E,6E)-3,7,11-Trimethyl-1-(tetrahydro-2H-pyran-2-yloxy)dodeca-2,6,10-triene (14b)

As for 14a except using trans-trans-farnesol (10 g, 45 mmol), pyridinium p-toluenesulfonate (1.13 g, 4.5 mmol) and 2,4-dihydropyran (16.4 mL, 180 mmol). The product was isolated as a colourless liquid (11.2 g, 36.4 mmol, 88%).

Rf = 0.25 (n-Hex – EtOAc, 95:5); δH/ppm (400 MHz, CDCl3) 5.39 – 5.30 (m, 1H), 5.14 – 5.02 (m, 2H), 4.65 – 4.57 (m, 1H), 4.27 – 4.17 (m, 1H), 4.01 (dd, J = 11.8, 7.4 Hz, 1H), 3.88 (ddd, J = 11.2, 7.7, 3.5 Hz, 1H), 3.54 – 3.43 (m, 1H), 2.16 – 1.90 (m, 8H), 1.90 – 1.40 (m, 6H), 1.66

(s, 6H), 1.58 (s, 6H); δC/ppm (100 MHz, CDCl3) 140.30, 135.30, 131.34, 124.45, 124.00, 120.74, 97.84, 63.72, 62.33, 39.80, 39.73, 30.82, 26.83, 26.39, 25.77, 25.62, 19.71, 17.76,

+ 16.50, 16.09; HRMS, found 324.2908 (C20H38NO2, [M + NH4] , requires 324.2903).

176

7.3.31. (2E,6E,10E)-2,6,10-Trimethyl-12-(tetrahydro-2H-pyran-2-yloxy)dodeca-2,6,10- trien-1-ol (15b)

As for 15a with minor modifications as detailed below, using 14b (12.6 g, 41 mmol), SeO2 (490 mg, 4.4 mmol) and tert-butyl hydroperoxide (8.22 mL, 45 mmol). After stirring overnight a further volume of tert-butyl hydroperoxide (4.0 mL, 22 mmol) was added and the reaction stirred for a further 48 hours. The crude was purified by flash column chromatography eluting in a gradient of 5-15% ethyl acetate in n-hexane. The product was isolated as a colourless oil (3.0 g, 9.3 mmol, 23%).

Rf = 0.18 (n-Hex – EtOAc, 8:2); δH/ppm (400 MHz, CDCl3) 5.44 – 5.30 (m, 2H), 5.15 – 5.06 (m, 1H), 4.65 – 4.58 (m, 1H), 4.27 – 4.18 (m, 1H), 4.07 – 4.00 (m, 1H), 3.98 (s, 2H), 3.88 (ddd, J = 11.2, 7.6, 3.4 Hz, 1H), 3.55 – 3.45 (m, 1H), 2.07 (m, 8H), 1.89 – 1.46 (m, 6H), 1.67 (s, 3H),

1.65 (s, 3H), 1.59 (s, 3H); δC/ppm (100 MHz, CDCl3) 140.23, 134.95, 134.88, 125.97, 124.35, 120.81, 97.84, 69.03, 63.74, 62.40, 39.69, 39.38, 30.81, 26.31, 26.21, 25.62, 19.72, 16.53,

+ 16.11, 13.82; HRMS, found 340.2861 (C20H38NO3, [M + NH4] , requires 340.2852).

7.3.32. (2E,6E,10E)-1-Bromo-2,6,10-trimethyl-12-(tetrahydro-2H-pyran-2- yloxy)dodeca-2,6,10-triene (16b)

As for 16a using alcohol 15b (1.41 g, 4.4 mmol), NBS (1.17 g, 6.6 mmol) and dimethyl sulfide (0.58 mL, 7.9 mmol). The product was isolated as a pale yellow oil (1.22 g, 3.2 mmol, 73%)

Rf = 0.47 (20% EtOAc in n-Hex); δH/ppm (400 MHz, CDCl3) 5.57 (t, J = 7.4 Hz, 1H), 5.40 – 5.32 (m, 1H), 5.15 – 5.07 (m, 1H), 4.66 – 4.59 (m, 1H), 4.24 (dd, J = 11.9, 6.4 Hz, 1H), 4.06 – 4.00 (m, 1H), 3.97 (s, 2H), 3.89 (ddd, J = 11.2, 7.6, 3.4 Hz, 1H), 3.55 – 3.46 (m, 1H), 2.17 –

1.94 (m, 8H), 1.91 – 1.46 (m, 6H), 1.75 (s, 3H), 1.68 (s, 3H), 1.59 (s, 3H); δC/ppm (100 MHz,

CDCl3) 140.28, 134.57, 132.06, 131.36, 124.67, 120.78, 97.97, 63.80, 62.45, 42.06, 39.69, 38.88, 30.86, 26.97, 26.38, 25.64, 19.78, 16.57, 16.11, 14.80; HRMS, found 402.2012

+ (C20H37NO2Br, [M + NH4] , requires 402.2008).

177

7.3.33. Ethyl (4E,8E,12E)-4,8,12-trimethyl-14-(tetrahydro-2H-pyran-2-yloxy)tetradeca- 4,8,12-trienoate (17b)

As for 17a using bromide 16b (1.10 g, 2.9 mmol), ethyl acetate (560 µL, 5.7 mmol), n-BuLi (2.28 mL, 5.7 mmol), diisopropylamine (800 µL, 5.7 mmol) and CuI (2.17 g, 11.4 mmol). The isolated product was a pale yellow oil (753 mg, 1.92 mmol, 67%)

Rf = 0.50 (n-Hex – EtOAc, 8:2); δH/ppm (400 MHz, CDCl3) 5.43 – 5.35 (m, 1H), 5.19 – 5.08 (m, 2H), 4.68 – 4.62 (m, 1H), 4.26 (dd, J = 11.8, 6.4 Hz, 1H), 4.14 (q, J = 7.1 Hz, 2H), 4.05 (dd, J = 11.8, 7.5 Hz, 1H), 3.92 (ddd, J = 11.2, 7.6, 3.3 Hz, 1H), 3.59 – 3.49 (m, 1H), 2.45 – 2.37 (m, 2H), 2.37 – 2.27 (m, 2H), 2.19 – 1.94 (m, 8H), 1.94 – 1.50 (m, 6H), 1.71 (s, 3H), 1.63

(s, 3H), 1.62 (s, 3H), 1.28 (t, J = 7.1 Hz, 3H); δC/ppm (100 MHz, CDCl3) 173.69, 140.42, 135.24, 133.45, 125.19, 124.16, 120.70, 97.95, 63.79, 62.44, 60.38, 39.77, 39.68, 34.82, 33.43, 30.86, 26.74, 26.43, 25.64, 19.77, 16.58, 16.16, 16.06, 14.41; HRMS, found 415.2820

+ (C24H40O4Na, [M + Na] , requires 415.2824).

7.3.34. (4E,8E,12E)-4,8,12-Trimethyl-14-(tetrahydro-2H-pyran-2-yloxy)tetradeca-4,8,12- trien-1-ol (18b)

As for 18a using ester 17b (750 mg, 1.9 mmol) and lithium aluminium hydride (108 mg, 2.9 mmol). The isolated product was a colourless oil (459 mg, 1.3 mmol, 69%).

Rf = 0.33 (n-Hex – acetone, 7:3); δH/ppm (400 MHz, CDCl3) 5.41 – 5.34 (m, 1H), 5.19 – 5.08 (m, 2H), 4.67 – 4.61 (m, 1H), 4.25 (dd, J = 11.6, 6.1, 1H), 4.09 – 4.01 (m, 1H), 3.95 – 3.87 (m, 1H), 3.64 (t, J = 6.5, 2H), 3.57 – 3.48 (m, 1H), 2.25 – 1.95 (m, 10H), 1.94 – 1.42 (m, 8H), 1.70

(s, 3H), 1.63 (s, 3H), 1.61 (s, 3H); δC/ppm (100 MHz, CDCl3) 140.43, 135.28, 134.84, 124.93, 124.26, 120.83, 98.00, 63.87, 63.00, 62.50, 39.81, 36.21, 30.97, 30.93, 26.68, 26.45, 25.72,

+ 19.87, 19.83, 16.62, 16.17, 16.08; HRMS, found 373.2731 (C22H38O3Na, [M + Na] , requires 373.2719).

178

7.3.35. (4E,8E,12E)-1-Iodo-4,8,12-trimethyl-14-(tetrahydro-2H-pyran-2-yloxy)tetradeca- 4,8,12-triene (19b)

As for 19a using alcohol 18b (450 mg, 1.3 mmol), imidazole (123 mg, 1.8 mmol), triphenylphosphine (506 mg, 1.9 mmol) and iodine (457 mg, 1.8 mmol). The isolated product was a pale yellow oil (497 mg, 1.1 mmol, 84%).

Rf = 0.60 (n-Hex – EtOAc, 8:2); δH/ppm (400 MHz, CDCl3) 5.40 – 5.32 (m, 1H), 5.20 – 5.13 (m, 1H), 5.13 – 5.06 (m, 1H), 4.62 (dd, J = 4.2, 3.0 Hz, 1H), 4.23 (dd, J = 11.9, 6.4 Hz, 1H), 4.03 (dd, J = 11.8, 7.4 Hz, 1H), 3.89 (ddd, J = 11.2, 7.6, 3.5 Hz, 1H), 3.55 – 3.45 (m, 1H), 3.14 (t, J = 7.0 Hz, 2H), 2.15 – 1.95 (m, 10H), 1.94 – 1.46 (m, 8H), 1.68 (s, 3H), 1.59 (s, 3H), 1.58

(s, 3H); δC/ppm (100 MHz, CDCl3) 140.36, 135.16, 132.91, 125.94, 124.25, 120.75, 97.93, 63.79, 62.43, 40.13, 39.77, 39.71, 31.71, 30.87, 26.68, 26.44, 25.65, 19.77, 16.58, 16.10,

+ 15.95, 6.79; HRMS, found 478.2196 (C22H41NO2I, [M + NH4] , requires 478.2182).

7.3.36. Trimethyl((7E,11E,15E)‐7,11,15‐trimethyl‐17‐(tetrahydro‐2H‐pyran‐2‐ yloxy)heptadeca‐7,11,15‐trien‐1‐yn‐1‐yl)silane (20b)

As for 20a using iodide 19b (490 mg, 1.1 mmol), 1-(trimethylsilyl)propyne (0.63 mL, 4.3 mmol), n-BuLi (1.70 mL, 4.3 mmol). The isolated product was a colourless oil (375 mg, 0.84 mmol, 80%).

Rf = 0.51 (n-Hex – EtOAc, 9:1); δH/ppm (400 MHz, CDCl3) 5.41 – 5.30 (m, 1H), 5.15 – 5.05 (m, 2H), 4.62 (dd, J = 4.3, 2.9 Hz, 1H), 4.28 – 4.19 (m, 1H), 4.03 (dd, J = 11.9, 7.4 Hz, 1H), 3.89 (ddd, J = 11.3, 7.6, 3.7 Hz, 1H), 3.55 – 3.45 (m, 1H), 2.25 – 2.18 (m, 2H), 2.15 – 1.93 (m,

10H), 1.90 – 1.40 (m, 10H), 1.68 (s, 3H), 1.59 (s, 3H), 1.58 (s, 3H), 0.14 (s, 9H); δC/ppm (100

MHz, CDCl3) 140.41, 135.38, 134.86, 124.59, 124.07, 120.73, 107.79, 97.93, 84.48, 63.79, 62.42, 39.85, 39.79, 39.16, 30.87, 28.18, 27.05, 26.77, 26.45, 25.66, 19.87, 19.77, 16.58,

+ 16.16, 15.91, 0.32 (3C); HRMS, found 462.3776 (C28H52NO2Si, [M + NH4] , requires 462.3767).

179

7.3.37. (2E,6E,10E)‐3,7,11‐Trimethyl-17-(trimethylsilyl)heptadeca‐2,6,10‐trien‐16‐yn‐1‐ ol (21b)

As for 21a using 20b (312 mg, 0.70 mmol), and PPTs (17 mg, 0.07 mmol). The crude was purified by flash column chromatography eluting in a gradient of 3-30% ethyl acetate in n- hexane. The isolated product was a colourless oil (207 mg, 0.57 mmol, 82%).

Rf = 0.34 (n-Hex – EtOAc, 8:2); δH/ppm (400 MHz, CDCl3) 5.46 – 5.37 (m, 1H), 5.15 – 5.04 (m, 2H), 4.15 (d, J = 7.0 Hz, 2H), 2.27 – 2.17 (m, 2H), 2.13 – 1.93 (m, 10H), 1.68 (s, 3H), 1.60

(s, 3H), 1.58 (s, 3H), 1.47 (m, 4H), 1.41 (br s, 1H), 0.14 (s, 9H); δC/ppm (100 MHz, CDCl3) 140.00, 135.50, 134.89, 124.55, 123.93, 123.45, 107.80, 84.50, 59.56, 39.83, 39.69, 39.16, 28.17, 27.04, 26.74, 26.44, 19.87, 16.44, 16.16, 15.91, 0.32 (3C); HRMS, found 378.3182

+ (C23H44NOSi, [M + NH4] , requires 378.3192).

7.3.38. (2E,6E,10E)-3,7,11-Trimethylheptadeca-2,6,10-trien-16-yn-1-ol (YnGG)

As for YnF using 21b (205 mg, 0.57 mmol) and TBAF (1.14 mL, 1.14 mmol). The crude purified by flash column chromatography eluting in a gradient of 3-20% acetone in n-hexane. The product was a colourless oil (58 mg, 0.19 mmol, 33%).

Rf = 0.29 (n-Hex – acetone, 8:2); δH/ppm (400 MHz, CDCl3) 5.46 – 5.36 (m, 1H), 5.15 – 5.04 (m, 2H), 4.15 (d, J = 6.9 Hz, 2H), 2.18 (td, J = 6.6, 2.6 Hz, 2H), 2.05 (m, 10H), 1.94 (t, J = 2.6

Hz, 1H), 1.68 (s, 3H), 1.60 (s, 3H), 1.58 (s, 3H), 1.52 – 1.45 (m, 4H), 1.33 (br s, 1H); δC/ppm

(100 MHz, CDCl3) 139.97, 135.46, 134.81, 124.59, 123.96, 123.46, 84.83, 68.27, 59.54, 39.81, 39.69, 39.17, 28.09, 27.03, 26.70, 26.44, 18.43, 16.43, 16.13, 15.94; HRMS, found

+ 306.2788 (C20H36NO, [M + NH4] , requires 306.2797).

7.3.39. Synthesis of capture reagents

The synthesis of AzRB is described in Wright et al.187

All capture reagents were synthesised by solid phase peptide synthesis on a Biotin-PEG NovaTag™ resin (Novabiochem®). Fmoc-protected amino acids and coupling reagents were sourced from AGTC Bioproducts. Azidoacetic acid was sourced from Sigma Aldrich. Ultrapure water was obtained from a MilliQ Millipore water purification system. LC-MS analysis and

180 purification was performed on a Waters HPLC system fitted with a Waters 515 HPLC pump, Waters 2767 autosampler, Waters XBridge C18 4.6 mm × 100 mm (analytical) or 19mm × 100 mm (preparative) column, Waters 3100 mass spectrometer and Waters 2998 photodiode array. NMR was performed on a Bruker AV-400 instrument. High resolution mass spectrometry was performed by the Imperial College Department of Chemistry Mass Spectrometry Service.

In general, the resin (100 was swelled in DMF for 30 min before Fmoc deprotection in 20% piperidine/DMF (3mL, 3 x 3 min). The resin was washed in DMF (3 x 3mL), DCM (3 x 3mL) and DMF (3 x 3mL). Each coupling step was performed as follows: the amino acid and coupling reagents were dissolved in DMF (2 mL) and activated for 5 min before addition to the resin. The coupling of Fmoc-ε-Ahx-OH was performed using HATU (4.9:5 mol ratio HATU:amino acid) and DIPEA (2:1 mol ratio DIPEA:amino acid) and for all other residues HBTU (4.9:5 mol ratio HBTU:amino acid) and DIPEA (2:1 mol ratio DIPEA:amino acid) were used. The resin was shaken for 1-3 hours, washed in DMF (3 x 3mL), DCM (3 x 3mL) and DMF (3 x 3mL). A fresh batch of activated amino acid was added and the resin shaken for a further 1-3 hours. After washing as before, Fmoc deprotection was performed as before, followed by washing. The procedure was then sequentially repeated for each residue, finishing with the coupling of the azidoacetic acid moiety, which required no Fmoc deprotection. After the final coupling the resin was washed in DMF (3 x 3mL), DCM (3 x 3mL), MeOH (3 x 3mL) and Et2O (3 x 3mL) and dried in a dessicator under vacuum overnight. The capture reagent was cleaved from the resin by incubation in a solution of 95% TFA, 2.5% TIS and 2.5% DTT (3 mL) for 3 hours. The cleavage mixture containing the cleaved product was transferred into a 15 mL falcon tube, and the resin washed twice with cleavage mixture (0.5 mL). The solution was evaporated to a volume of 1 mL under a stream of nitrogen, after which the product was precipitated by addition of ice-cold tert-butyl methyl ester (TBME) (10 mL). The product was pelleted by centrifugation. The pellet was washed twice with TBME (10 mL) and then dried overnight in a vacuum dessicator. The pellet was resuspended in ultrapure MilliQ water and purified by preparative RP-HPLC in a gradient of methanol and water both supplemented with 0.1% formic acid (0-10 min 5-98% MeOH, 10-12 min 98% MeOH). The fractions containing product were combined, the solvents removed on a speedvac (Genevac). The product was resuspended in MilliQ water (1 mL) and lyophilised to dryness.

181

AzRB (Yield: 24.3 mg, 26 µmol, 26%): Component Amount per coupling µmol Equivalents Biotin-PEG NovaTag™ resin 208 mg 100 1 Fmoc-ε-Ahx-OH 177 mg 500 5 Fmoc-Gly-OH 149 mg 500 5 Fmoc-Arg(Pbf)-OH 324 mg 500 5 Fmoc-Ala-OH 156 mg 500 5 Azidoacetic acid 37 µL 500 5

AzK+RB (Yield: 29.3 mg, 27 µmol, 26%): Component Amount per coupling µmol Equivalents Biotin-PEG NovaTag™ resin 212 mg 102 1 Fmoc-ε-Ahx-OH 180 mg 511 5 Fmoc-Gly-OH 152 mg 511 5 Fmoc-Arg(Pbf)-OH 332 mg 511 5 Fmoc-Ala-OH 159 mg 511 5 Fmoc-Lys(Me3Cl)-OH 114 mg 256 2.5 Azidoacetic acid 37 µL 500 5

AzKB (Yield: 33.4 mg, 37 µmol, 37%) Component Amount per coupling µmol Equivalents Biotin-PEG NovaTag™ resin 208 mg 100 1 Fmoc-ε-Ahx-OH 176 mg 500 5 Fmoc-Gly-OH 149 mg 500 5 Fmoc-Lys(Boc)-OH 234 mg 500 5 Fmoc-Ala-OH 156 mg 500 5 Azidoacetic acid 37 µL 500 5

182

Bibliography

1. International Human Genome Sequencing, C. Finishing the euchromatic sequence of the human genome. Nature 431, 931-45 (2004). 2. Walsh, C.T., Garneau-Tsodikova, S. & Gatto, G.J., Jr. Protein posttranslational modifications: the chemistry of proteome diversifications. Angew Chem Int Ed Engl 44, 7342-72 (2005). 3. Hentschel, A., Zahedi, R.P. & Ahrends, R. Protein lipid modifications--More than just a greasy ballast. Proteomics 16, 759-82 (2016). 4. Resh, M.D. Covalent lipid modifications of proteins. Curr Biol 23, R431-5 (2013). 5. Nile, A.H. & Hannoush, R.N. Fatty acylation of Wnt proteins. Nat Chem Biol 12, 60- 69 (2016). 6. Ciepla, P., Magee, A.I. & Tate, E.W. Cholesterylation: a tail of hedgehog. Biochem Soc Trans 43, 262-7 (2015). 7. Wang, M. & Casey, P.J. Protein prenylation: unique fats make their mark on biology. Nat Rev Mol Cell Biol (2016). 8. Casey, P.J., Solski, P.A., Der, C.J. & Buss, J.E. p21ras is modified by a farnesyl isoprenoid. Proc Natl Acad Sci U S A 86, 8323-7 (1989). 9. Marrari, Y., Crouthamel, M., Irannejad, R. & Wedegaertner, P.B. Assembly and trafficking of heterotrimeric G proteins. Biochemistry 46, 7665-77 (2007). 10. Farnsworth, C.C., Wolda, S.L., Gelb, M.H. & Glomset, J.A. Human lamin B contains a farnesylated cysteine residue. J Biol Chem 264, 20422-9 (1989). 11. Leung, K.F., Baron, R. & Seabra, M.C. Thematic review series: lipid posttranslational modifications. geranylgeranylation of Rab GTPases. J Lipid Res 47, 467-75 (2006). 12. Gutkowska, M. & Swiezewska, E. Structure, regulation and cellular functions of Rab geranylgeranyl transferase and its cellular partner Rab Escort Protein. Mol Membr Biol 29, 243-56 (2012). 13. Kamiya, Y., Sakurai, A., Tamura, S. & Takahashi, N. Structure of rhodotorucine A, a novel lipopeptide, inducing mating tube formation in Rhodosporidium toruloides. Biochem Biophys Res Commun 83, 1077-83 (1978). 14. Schmidt, R.A., Schneider, C.J. & Glomset, J.A. Evidence for post-translational incorporation of a product of mevalonic acid into Swiss 3T3 cell proteins. J Biol Chem 259, 10175-80 (1984). 15. Wolda, S.L. & Glomset, J.A. Evidence for modification of lamin B by a product of mevalonic acid. J Biol Chem 263, 5997-6000 (1988). 16. Hancock, J.F., Magee, A.I., Childs, J.E. & Marshall, C.J. All ras proteins are polyisoprenylated but only some are palmitoylated. Cell 57, 1167-77 (1989). 17. Farnsworth, C.C., Gelb, M.H. & Glomset, J.A. Identification of geranylgeranyl- modified proteins in HeLa cells. Science 247, 320-2 (1990). 18. Rilling, H.C., Breunger, E., Epstein, W.W. & Crain, P.F. Prenylated proteins: the structure of the isoprenoid group. Science 247, 318-20 (1990). 19. Boyartchuk, V.L., Ashby, M.N. & Rine, J. Modulation of Ras and a-factor function by carboxyl-terminal proteolysis. Science 275, 1796-800 (1997). 20. Ashby, M.N. CaaX converting enzymes. Curr Opin Lipidol 9, 99-102 (1998). 21. Dai, Q. et al. Mammalian prenylcysteine carboxyl methyltransferase is in the endoplasmic reticulum. J Biol Chem 273, 15030-4 (1998).

183

22. Smeland, T.E., Seabra, M.C., Goldstein, J.L. & Brown, M.S. Geranylgeranylated Rab proteins terminating in Cys-Ala-Cys, but not Cys-Cys, are carboxyl-methylated by bovine brain membranes in vitro. Proc Natl Acad Sci U S A 91, 10712-6 (1994). 23. Cushman, I., Cushman, S.M., Potter, P.M. & Casey, P.J. Control of RhoA methylation by carboxylesterase I. J Biol Chem 288, 19177-83 (2013). 24. Winter-Vann, A.M. & Casey, P.J. Post-prenylation-processing enzymes as new targets in oncogenesis. Nat Rev Cancer 5, 405-12 (2005). 25. Michaelson, D. et al. Postprenylation CAAX processing is required for proper localization of Ras but not Rho GTPases. Mol Biol Cell 16, 1606-16 (2005). 26. Mijimolle, N. et al. Protein farnesyltransferase in embryogenesis, adult homeostasis, and tumor development. Cancer Cell 7, 313-24 (2005). 27. Yang, S.H. et al. Severe hepatocellular disease in mice lacking one or both CaaX prenyltransferases. J Lipid Res 53, 77-86 (2012). 28. Sjogren, A.K. et al. GGTase-I deficiency reduces tumor formation and improves survival in mice with K-RAS-induced lung cancer. J Clin Invest 117, 1294-304 (2007). 29. Kim, E. et al. Disruption of the mouse Rce1 gene results in defective Ras processing and mislocalization of Ras within cells. J Biol Chem 274, 8383-90 (1999). 30. Bergo, M.O. et al. Isoprenylcysteine carboxyl methyltransferase deficiency in mice. J Biol Chem 276, 5841-5 (2001). 31. Roberts, P.J. et al. Rho Family GTPase modification and dependence on CAAX motif-signaled posttranslational modification. J Biol Chem 283, 25150-63 (2008). 32. Ahearn, I.M., Haigis, K., Bar-Sagi, D. & Philips, M.R. Regulating the regulator: post- translational modification of RAS. Nat Rev Mol Cell Biol 13, 39-51 (2012). 33. Kellogg, B.A. & Poulter, C.D. Chain elongation in the isoprenoid biosynthetic pathway. Curr Opin Chem Biol 1, 570-8 (1997). 34. McTaggart, S.J. Isoprenylated proteins. Cell Mol Life Sci 63, 255-67 (2006). 35. Matusewicz, L., Meissner, J., Toporkiewicz, M. & Sikorski, A.F. The effect of statins on cancer cells--review. Tumour Biol 36, 4889-904 (2015). 36. Yeganeh, B. et al. Targeting the mevalonate cascade as a new therapeutic approach in heart disease, cancer and pulmonary disease. Pharmacol Ther 143, 87-110 (2014). 37. Berndt, N., Hamilton, A.D. & Sebti, S.M. Targeting protein prenylation for cancer therapy. Nat Rev Cancer 11, 775-91 (2011). 38. Reiss, Y., Goldstein, J.L., Seabra, M.C., Casey, P.J. & Brown, M.S. Inhibition of purified p21ras farnesyl:protein transferase by Cys-AAX tetrapeptides. Cell 62, 81-8 (1990). 39. Casey, P.J., Thissen, J.A. & Moomaw, J.F. Enzymatic modification of proteins with a geranylgeranyl isoprenoid. Proc Natl Acad Sci U S A 88, 8631-5 (1991). 40. Seabra, M.C., Reiss, Y., Casey, P.J., Brown, M.S. & Goldstein, J.L. Protein farnesyltransferase and geranylgeranyltransferase share a common alpha subunit. Cell 65, 429-34 (1991). 41. Moomaw, J.F. & Casey, P.J. Mammalian protein geranylgeranyltransferase. Subunit composition and metal requirements. J Biol Chem 267, 17438-43 (1992). 42. Reiss, Y., Brown, M.S. & Goldstein, J.L. Divalent cation and prenyl pyrophosphate specificities of the protein farnesyltransferase from rat brain, a zinc metalloenzyme. J Biol Chem 267, 6403-8 (1992).

184

43. Chen, W.J., Moomaw, J.F., Overton, L., Kost, T.A. & Casey, P.J. High level expression of mammalian protein farnesyltransferase in a baculovirus system. The purified protein contains zinc. J Biol Chem 268, 9675-80 (1993). 44. Zhang, F.L. & Casey, P.J. Influence of metal ions on substrate binding and catalytic activity of mammalian protein geranylgeranyltransferase type-I. Biochem J 320 ( Pt 3), 925-32 (1996). 45. Park, H.W., Boduluri, S.R., Moomaw, J.F., Casey, P.J. & Beese, L.S. Crystal structure of protein farnesyltransferase at 2.25 angstrom resolution. Science 275, 1800-4 (1997). 46. Long, S.B., Casey, P.J. & Beese, L.S. Cocrystal structure of protein farnesyltransferase complexed with a farnesyl diphosphate substrate. Biochemistry 37, 9612-8 (1998). 47. Strickland, C.L. et al. Crystal structure of farnesyl protein transferase complexed with a CaaX peptide and farnesyl diphosphate analogue. Biochemistry 37, 16601-11 (1998). 48. Long, S.B., Casey, P.J. & Beese, L.S. The basis for K-Ras4B binding specificity to protein farnesyltransferase revealed by 2 A resolution ternary complex structures. Structure 8, 209-22 (2000). 49. Long, S.B., Hancock, P.J., Kral, A.M., Hellinga, H.W. & Beese, L.S. The crystal structure of human protein farnesyltransferase reveals the basis for inhibition by CaaX tetrapeptides and their mimetics. Proc Natl Acad Sci U S A 98, 12948-53 (2001). 50. Long, S.B., Casey, P.J. & Beese, L.S. Reaction path of protein farnesyltransferase at atomic resolution. Nature 419, 645-50 (2002). 51. Taylor, J.S., Reid, T.S., Terry, K.L., Casey, P.J. & Beese, L.S. Structure of mammalian protein geranylgeranyltransferase type-I. EMBO J 22, 5963-74 (2003). 52. Reid, T.S., Terry, K.L., Casey, P.J. & Beese, L.S. Crystallographic analysis of CaaX prenyltransferases complexed with substrates defines rules of protein substrate selectivity. J Mol Biol 343, 417-33 (2004). 53. Terry, K.L., Casey, P.J. & Beese, L.S. Conversion of protein farnesyltransferase to a geranylgeranyltransferase. Biochemistry 45, 9746-55 (2006). 54. Turek-Etienne, T.C., Strickland, C.L. & Distefano, M.D. Biochemical and structural studies with prenyl diphosphate analogues provide insights into isoprenoid recognition by protein farnesyl transferase. Biochemistry 42, 3716-24 (2003). 55. Huang, C.C., Casey, P.J. & Fierke, C.A. Evidence for a catalytic role of zinc in protein farnesyltransferase. Spectroscopy of Co2+-farnesyltransferase indicates metal coordination of the substrate thiolate. J Biol Chem 272, 20-3 (1997). 56. Wu, Z. et al. Farnesyl protein transferase: identification of K164 alpha and Y300 beta as catalytic residues by mutagenesis and kinetic studies. Biochemistry 38, 11239-49 (1999). 57. Hightower, K.E., De, S., Weinbaum, C., Spence, R.A. & Casey, P.J. Lysine(164)alpha of protein farnesyltransferase is important for both CaaX substrate binding and catalysis. Biochem J 360, 625-31 (2001). 58. Dolence, J.M., Rozema, D.B. & Poulter, C.D. Yeast protein farnesyltransferase. Site- directed mutagenesis of conserved residues in the beta-subunit. Biochemistry 36, 9246-52 (1997).

185

59. Hartman, H.L., Bowers, K.E. & Fierke, C.A. Lysine beta311 of protein geranylgeranyltransferase type I partially replaces magnesium. J Biol Chem 279, 30546-53 (2004). 60. Tschantz, W.R., Furfine, E.S. & Casey, P.J. Substrate binding is required for release of product from mammalian protein farnesyltransferase. J Biol Chem 272, 9989-93 (1997). 61. Moores, S.L. et al. Sequence dependence of protein isoprenylation. J Biol Chem 266, 14603-10 (1991). 62. Kinsella, B.T., Erdman, R.A. & Maltese, W.A. Posttranslational modification of Ha-ras p21 by farnesyl versus geranylgeranyl isoprenoids is determined by the COOH- terminal amino acid. Proc Natl Acad Sci U S A 88, 8934-8 (1991). 63. Casey, P.J. & Seabra, M.C. Protein prenyltransferases. J Biol Chem 271, 5289-92 (1996). 64. Hougland, J.L., Lamphear, C.L., Scott, S.A., Gibbs, R.A. & Fierke, C.A. Context- dependent substrate recognition by protein farnesyltransferase. Biochemistry 48, 1691-701 (2009). 65. Krzysiak, A.J., Aditya, A.V., Hougland, J.L., Fierke, C.A. & Gibbs, R.A. Synthesis and screening of a CaaL peptide library versus FTase reveals a surprising number of substrates. Bioorg Med Chem Lett 20, 767-70 (2010). 66. Zverina, E.A., Lamphear, C.L., Wright, E.N. & Fierke, C.A. Recent advances in protein prenyltransferases: substrate identification, regulation, and disease interventions. Curr Opin Chem Biol 16, 544-52 (2012). 67. Reiss, Y., Stradley, S.J., Gierasch, L.M., Brown, M.S. & Goldstein, J.L. Sequence requirement for peptide recognition by rat brain p21ras protein farnesyltransferase. Proc Natl Acad Sci U S A 88, 732-6 (1991). 68. Hicks, K.A., Hartman, H.L. & Fierke, C.A. Upstream polybasic region in peptides enhances dual specificity for prenylation by both farnesyltransferase and geranylgeranyltransferase type I. Biochemistry 44, 15325-33 (2005). 69. Hartman, H.L., Hicks, K.A. & Fierke, C.A. Peptide specificity of protein prenyltransferases is determined mainly by reactivity rather than binding affinity. Biochemistry 44, 15314-24 (2005). 70. Hougland, J.L. et al. Identification of novel peptide substrates for protein farnesyltransferase reveals two substrate classes with distinct sequence selectivities. J Mol Biol 395, 176-90 (2010). 71. Seabra, M.C., Goldstein, J.L., Sudhof, T.C. & Brown, M.S. Rab geranylgeranyl transferase. A multisubunit enzyme that prenylates GTP-binding proteins terminating in Cys-X-Cys or Cys-Cys. J Biol Chem 267, 14497-503 (1992). 72. Zhang, H., Seabra, M.C. & Deisenhofer, J. Crystal structure of Rab geranylgeranyltransferase at 2.0 A resolution. Structure 8, 241-51 (2000). 73. Rasteiro, R. & Pereira-Leal, J.B. Multiple domain insertions and losses in the evolution of the Rab prenylation complex. BMC Evol Biol 7, 140 (2007). 74. Guo, Z. et al. Structures of RabGGTase-substrate/product complexes provide insights into the evolution of protein prenylation. EMBO J 27, 2444-56 (2008). 75. Seabra, M.C., Brown, M.S., Slaughter, C.A., Sudhof, T.C. & Goldstein, J.L. Purification of component A of Rab geranylgeranyl transferase: possible identity with the choroideremia gene product. Cell 70, 1049-57 (1992). 76. Andres, D.A. et al. cDNA cloning of component A of Rab geranylgeranyl transferase and demonstration of its role as a Rab escort protein. Cell 73, 1091-9 (1993).

186

77. Rak, A. et al. Structure of the Rab7:REP-1 complex: insights into the mechanism of Rab prenylation and choroideremia disease. Cell 117, 749-60 (2004). 78. Baron, R.A. et al. Phosphonocarboxylates inhibit the second geranylgeranyl addition by Rab geranylgeranyl transferase. J Biol Chem 284, 6861-8 (2009). 79. Thomä, N.H., Niculae, A., Goody, R.S. & Alexandrov, K. Double Prenylation by RabGGTase Can Proceed without Dissociation of the Mono-prenylated Intermediate. Journal of Biological Chemistry 276, 48631-48636 (2001). 80. Thoma, N.H. et al. Allosteric regulation of substrate binding and product release in geranylgeranyltransferase type II. Biochemistry 40, 268-74 (2001). 81. Gomes, A.Q. et al. Membrane targeting of Rab GTPases is influenced by the prenylation motif. Mol Biol Cell 14, 1882-99 (2003). 82. Ali, B.R. & Seabra, M.C. Targeting of Rab GTPases to cellular membranes. Biochem Soc Trans 33, 652-6 (2005). 83. Thoma, N.H. et al. Phosphoisoprenoid binding specificity of geranylgeranyltransferase type II. Biochemistry 39, 12043-52 (2000). 84. Maurer-Stroh, S. et al. Towards complete sets of farnesylated and geranylgeranylated proteins. PLoS Comput Biol 3, e66 (2007). 85. Wilson, A.L., Erdman, R.A., Castellano, F. & Maltese, W.A. Prenylation of Rab8 GTPase by type I and type II geranylgeranyl transferases. Biochem J 333 ( Pt 3), 497-504 (1998). 86. Resh, M.D. Targeting protein lipidation in disease. Trends Mol Med 18, 206-14 (2012). 87. Bos, J.L. ras oncogenes in human cancer: a review. Cancer Res 49, 4682-9 (1989). 88. Bailey, P. et al. Genomic analyses identify molecular subtypes of pancreatic cancer. Nature 531, 47-52 (2016). 89. Bos, J.L. The ras gene family and human carcinogenesis. Mutat Res 195, 255-71 (1988). 90. Wilson, C.Y. & Tolias, P. Recent advances in cancer drug discovery targeting RAS. Drug Discov Today (2016). 91. Omerovic, J., Laude, A.J. & Prior, I.A. Ras proteins: paradigms for compartmentalised and isoform-specific signalling. Cell Mol Life Sci 64, 2575-89 (2007). 92. Holstein, S.A. & Hohl, R.J. Is there a future for prenyltransferase inhibitors in cancer therapy? Curr Opin Pharmacol 12, 704-9 (2012). 93. James, G.L., Goldstein, J.L. & Brown, M.S. Polylysine and CVIM sequences of K- RasB dictate specificity of prenylation and confer resistance to benzodiazepine peptidomimetic in vitro. J Biol Chem 270, 6221-6 (1995). 94. Rowell, C.A., Kowalczyk, J.J., Lewis, M.D. & Garcia, A.M. Direct demonstration of geranylgeranylation and farnesylation of Ki-Ras in vivo. J Biol Chem 272, 14093-7 (1997). 95. Whyte, D.B. et al. K- and N-Ras are geranylgeranylated in cells treated with farnesyl protein transferase inhibitors. J Biol Chem 272, 14459-64 (1997). 96. Heard, J.J., Fong, V., Bathaie, S.Z. & Tamanoi, F. Recent progress in the study of the Rheb family GTPases. Cell Signal 26, 1950-7 (2014). 97. Armijo, M.E. et al. Rheb signaling and tumorigenesis: mTORC1 and new horizons. Int J Cancer 138, 1815-23 (2016). 98. Mavrakis, K.J. et al. Tumorigenic activity and therapeutic inhibition of Rheb GTPase. Genes Dev 22, 2178-88 (2008).

187

99. Ashar, H.R. et al. Farnesyl transferase inhibitors block the farnesylation of CENP-E and CENP-F and alter the association of CENP-E with the microtubules. J Biol Chem 275, 30451-7 (2000). 100. Schafer-Hales, K. et al. Farnesyl transferase inhibitors impair chromosomal maintenance in cell lines and human tumors by compromising CENP-E and CENP-F function. Mol Cancer Ther 6, 1317-28 (2007). 101. Sebti, S.M. & Hamilton, A.D. Farnesyltransferase and geranylgeranyltransferase I inhibitors and cancer therapy: lessons from mechanism and bench-to-bedside translational studies. Oncogene 19, 6584-93 (2000). 102. Smithers, C.C. & Overduin, M. Structural Mechanisms and Drug Discovery Prospects of Rho GTPases. Cells 5 (2016). 103. Kusama, T. et al. Selective inhibition of cancer cell invasion by a geranylgeranyltransferase-I inhibitor. Clin Exp Metastasis 20, 561-7 (2003). 104. Gentry, L.R., Martin, T.D., Reiner, D.J. & Der, C.J. Ral small GTPase signaling and oncogenesis: More than just 15minutes of fame. Biochim Biophys Acta 1843, 2976- 2988 (2014). 105. Dasgupta, S., Cushman, I., Kpetemey, M., Casey, P.J. & Vishwanatha, J.K. Prenylated c17orf37 induces filopodia formation to promote cell migration and metastasis. J Biol Chem 286, 25935-46 (2011). 106. Evans, E.E. et al. C35 (C17orf37) is a novel tumor biomarker abundantly expressed in breast cancer. Mol Cancer Ther 5, 2919-30 (2006). 107. Kpetemey, M. et al. MIEN1, a novel interactor of Annexin A2, promotes tumor cell migration by enhancing AnxA2 cell surface expression. Mol Cancer 14, 156 (2015). 108. Dasgupta, S. et al. Novel gene C17orf37 in 17q12 amplicon promotes migration and invasion of prostate cancer cells. Oncogene 28, 2860-72 (2009). 109. Rajendiran, S. et al. MIEN1 promotes oral cancer progression and implicates poor overall survival. Cancer Biol Ther 16, 876-85 (2015). 110. Merideth, M.A. et al. Phenotype and course of Hutchinson-Gilford progeria syndrome. N Engl J Med 358, 592-604 (2008). 111. Young, S.G., Yang, S.H., Davies, B.S., Jung, H.J. & Fong, L.G. Targeting protein prenylation in progeria. Sci Transl Med 5, 171ps3 (2013). 112. Eriksson, M. et al. Recurrent de novo point mutations in lamin A cause Hutchinson- Gilford progeria syndrome. Nature 423, 293-8 (2003). 113. De Sandre-Giovannoli, A. et al. Lamin a truncation in Hutchinson-Gilford progeria. Science 300, 2055 (2003). 114. Sinensky, M. et al. The processing pathway of prelamin A. J Cell Sci 107 ( Pt 1), 61-7 (1994). 115. Corrigan, D.P. et al. Prelamin A endoproteolytic processing in vitro by recombinant Zmpste24. Biochem J 387, 129-38 (2005). 116. Prokocimer, M. et al. Nuclear lamins: key regulators of nuclear structure and activities. J Cell Mol Med 13, 1059-85 (2009). 117. Gruenbaum, Y. & Medalia, O. Lamins: the structure and protein complexes. Curr Opin Cell Biol 32, 7-12 (2015). 118. Toth, J.I. et al. Blocking protein farnesyltransferase improves nuclear shape in fibroblasts from humans with progeroid syndromes. Proc Natl Acad Sci U S A 102, 12873-8 (2005).

188

119. Lutz, R.J., Trujillo, M.A., Denham, K.S., Wenger, L. & Sinensky, M. Nucleoplasmic localization of prelamin A: implications for prenylation-dependent lamin A assembly into the nuclear lamina. Proc Natl Acad Sci U S A 89, 3000-4 (1992). 120. Yang, S.H. et al. Blocking protein farnesyltransferase improves nuclear blebbing in mouse fibroblasts with a targeted Hutchinson-Gilford progeria syndrome mutation. Proc Natl Acad Sci U S A 102, 10291-6 (2005). 121. Capell, B.C. et al. Inhibiting farnesylation of progerin prevents the characteristic nuclear blebbing of Hutchinson-Gilford progeria syndrome. Proc Natl Acad Sci U S A 102, 12879-84 (2005). 122. Mallampalli, M.P., Huyer, G., Bendale, P., Gelb, M.H. & Michaelis, S. Inhibiting farnesylation reverses the nuclear morphology defect in a HeLa cell model for Hutchinson-Gilford progeria syndrome. Proc Natl Acad Sci U S A 102, 14416-21 (2005). 123. Glynn, M.W. & Glover, T.W. Incomplete processing of mutant lamin A in Hutchinson- Gilford progeria leads to nuclear abnormalities, which are reversed by farnesyltransferase inhibition. Hum Mol Genet 14, 2959-69 (2005). 124. Young, S.G., Fong, L.G. & Michaelis, S. Prelamin A, Zmpste24, misshapen cell nuclei, and progeria--new evidence suggesting that protein farnesylation could be important for disease pathogenesis. J Lipid Res 46, 2531-58 (2005). 125. Yang, S.H. et al. A farnesyltransferase inhibitor improves disease phenotypes in mice with a Hutchinson-Gilford progeria syndrome mutation. J Clin Invest 116, 2115-21 (2006). 126. Fong, L.G. et al. A protein farnesyltransferase inhibitor ameliorates disease in a mouse model of progeria. Science 311, 1621-3 (2006). 127. Verstraeten, V.L., Ji, J.Y., Cummings, K.S., Lee, R.T. & Lammerding, J. Increased mechanosensitivity and nuclear stiffness in Hutchinson-Gilford progeria cells: effects of farnesyltransferase inhibitors. Aging Cell 7, 383-93 (2008). 128. Wang, Y., Ostlund, C. & Worman, H.J. Blocking protein farnesylation improves nuclear shape abnormalities in keratinocytes of mice expressing the prelamin A variant in Hutchinson-Gilford progeria syndrome. Nucleus 1, 432-9 (2010). 129. Gordon, L.B. et al. Clinical trial of a farnesyltransferase inhibitor in children with Hutchinson-Gilford progeria syndrome. Proc Natl Acad Sci U S A 109, 16666-71 (2012). 130. Gordon, L.B. et al. Impact of farnesylation inhibitors on survival in Hutchinson-Gilford progeria syndrome. Circulation 130, 27-34 (2014). 131. Yang, S.H., Andres, D.A., Spielmann, H.P., Young, S.G. & Fong, L.G. Progerin elicits disease phenotypes of progeria in mice whether or not it is farnesylated. J Clin Invest 118, 3291-300 (2008). 132. Yang, S.H. et al. Absence of progeria-like disease phenotypes in knock-in mice expressing a non-farnesylated version of progerin. Hum Mol Genet 20, 436-44 (2011). 133. Ragnauth, C.D. et al. Prelamin A acts to accelerate smooth muscle cell senescence and is a novel biomarker of human vascular aging. Circulation 121, 2200-10 (2010). 134. Cai, A., Zhou, Y. & Li, L. Rho-GTPase and Atherosclerosis: Pleiotropic Effects of Statins. J Am Heart Assoc 4 (2015). 135. Oesterle, A., Laufs, U. & Liao, J.K. Pleiotropic Effects of Statins on the Cardiovascular System. Circ Res 120, 229-243 (2017).

189

136. Wang, M. & Casey, P.J. Protein prenylation: unique fats make their mark on biology. Nat Rev Mol Cell Biol 17, 110-122 (2016). 137. Laufs, U. & Liao, J.K. Targeting Rho in cardiovascular disease. Circ Res 87, 526-8 (2000). 138. Rolfe, B.E., Worth, N.F., World, C.J., Campbell, J.H. & Campbell, G.R. Rho and vascular disease. Atherosclerosis 183, 1-16 (2005). 139. Loirand, G., Sauzeau, V. & Pacaud, P. Small G proteins in the cardiovascular system: physiological and pathological aspects. Physiol Rev 93, 1659-720 (2013). 140. Schermuly, R.T., Ghofrani, H.A., Wilkins, M.R. & Grimminger, F. Mechanisms of disease: pulmonary arterial hypertension. Nat Rev Cardiol 8, 443-55 (2011). 141. Storck, E.M. & Wojciak-Stothard, B. Rho GTPases in pulmonary vascular dysfunction. Vascul Pharmacol 58, 202-10 (2013). 142. Antoniu, S.A. Targeting RhoA/ROCK pathway in pulmonary arterial hypertension. Expert Opin Ther Targets 16, 355-63 (2012). 143. Wojciak-Stothard, B. et al. Role of RhoB in the regulation of pulmonary endothelial and smooth muscle cell responses to hypoxia. Circ Res 110, 1423-34 (2012). 144. Hu, J., Discher, D.J., Bishopric, N.H. & Webster, K.A. Hypoxia regulates expression of the endothelin-1 gene through a proximal hypoxia-inducible factor-1 binding site on the antisense strand. Biochem Biophys Res Commun 245, 894-9 (1998). 145. Roosing, S., Collin, R.W., den Hollander, A.I., Cremers, F.P. & Siemiatkowska, A.M. Prenylation defects in inherited retinal diseases. J Med Genet 51, 143-51 (2014). 146. Coussa, R.G. & Traboulsi, E.I. Choroideremia: a review of general findings and pathogenesis. Ophthalmic Genet 33, 57-65 (2012). 147. Seabra, M.C., Ho, Y.K. & Anant, J.S. Deficient geranylgeranylation of Ram/Rab27 in choroideremia. J Biol Chem 270, 24420-7 (1995). 148. Larijani, B., Hume, A.N., Tarafder, A.K. & Seabra, M.C. Multiple Factors Contribute to Inefficient Prenylation of Rab27a in Rab Prenylation Diseases. Journal of Biological Chemistry 278, 46798-46804 (2003). 149. Kohnke, M. et al. Rab GTPase prenylation hierarchy and its potential role in choroideremia disease. PLoS One 8, e81758 (2013). 150. Glenn, J.S., Watson, J.A., Havel, C.M. & White, J.M. Identification of a prenylation site in delta virus large antigen. Science 256, 1331-3 (1992). 151. Bordier, B.B. et al. A prenylation inhibitor prevents production of infectious hepatitis delta virus particles. J Virol 76, 10465-72 (2002). 152. Bordier, B.B. et al. In vivo antiviral efficacy of prenylation inhibitors against hepatitis delta virus. J Clin Invest 112, 407-14 (2003). 153. Koh, C. et al. Oral prenylation inhibition with lonafarnib in chronic hepatitis D infection: a proof-of-concept randomised, double-blind, placebo-controlled phase 2A trial. Lancet Infect Dis 15, 1167-74 (2015). 154. Ye, J. et al. Disruption of hepatitis C virus RNA replication through inhibition of host protein geranylgeranylation. Proc Natl Acad Sci U S A 100, 15865-70 (2003). 155. Wang, C. et al. Identification of FBL2 as a geranylgeranylated cellular protein required for hepatitis C virus RNA replication. Mol Cell 18, 425-34 (2005). 156. Imlay, L. & Odom, A.R. Isoprenoid metabolism in apicomplexan parasites. Curr Clin Microbiol Rep 1, 37-50 (2014). 157. Eastman, R.T., Buckner, F.S., Yokoyama, K., Gelb, M.H. & Van Voorhis, W.C. Thematic review series: lipid posttranslational modifications. Fighting parasitic disease by blocking protein farnesylation. J Lipid Res 47, 233-40 (2006).

190

158. Chakrabarti, D. et al. Protein farnesyltransferase and protein prenylation in Plasmodium falciparum. J Biol Chem 277, 42066-73 (2002). 159. Yokoyama, K., Lin, Y., Stuart, K.D. & Gelb, M.H. Prenylation of proteins in Trypanosoma brucei. Mol Biochem Parasitol 87, 61-9 (1997). 160. Buckner, F.S. et al. Cloning, heterologous expression, and substrate specificities of protein farnesyltransferases from Trypanosoma cruzi and Leishmania major. Mol Biochem Parasitol 122, 181-8 (2002). 161. Nallan, L. et al. Protein farnesyltransferase inhibitors exhibit potent antimalarial activity. J Med Chem 48, 3704-13 (2005). 162. Fletcher, S. et al. Potent, Plasmodium-Selective Farnesyltransferase Inhibitors That Arrest the Growth of Malaria Parasites: Structure−Activity Relationships of Ethylenediamine-Analogue Scaffolds and Homology Model Validation. Journal of Medicinal Chemistry 51, 5176-5197 (2008). 163. Eastman, R.T. et al. Resistance to a protein farnesyltransferase inhibitor in Plasmodium falciparum. J Biol Chem 280, 13554-9 (2005). 164. Hottman, D.A. & Li, L. Protein prenylation and synaptic plasticity: implications for Alzheimer's disease. Mol Neurobiol 50, 177-85 (2014). 165. Hooff, G.P., Wood, W.G., Muller, W.E. & Eckert, G.P. Isoprenoids, small GTPases and Alzheimer's disease. Biochim Biophys Acta 1801, 896-905 (2010). 166. Shankar, G.M. et al. Amyloid-beta protein dimers isolated directly from Alzheimer's brains impair synaptic plasticity and memory. Nat Med 14, 837-42 (2008). 167. Ostrowski, S.M., Wilkinson, B.L., Golde, T.E. & Landreth, G. Statins reduce amyloid- beta production through inhibition of protein isoprenylation. J Biol Chem 282, 26832- 44 (2007). 168. Zhou, Y. et al. Geranylgeranyl pyrophosphate stimulates γ-secretase to increase the generation of Aβ and APP-CTFγ. The FASEB Journal 22, 47-54 (2008). 169. Eckert, G.P. et al. Regulation of the brain isoprenoids farnesyl- and geranylgeranylpyrophosphate is altered in male Alzheimer patients. Neurobiol Dis 35, 251-7 (2009). 170. Cheng, S. et al. Farnesyltransferase haplodeficiency reduces neuropathology and rescues cognitive function in a mouse model of Alzheimer disease. J Biol Chem 288, 35952-60 (2013). 171. Storck, E.M., Serwa, R.A. & Tate, E.W. Chemical proteomics: a powerful tool for exploring protein lipidation. Biochem Soc Trans 41, 56-61 (2013). 172. Tate, E.W., Kalesh, K.A., Lanyon-Hogg, T., Storck, E.M. & Thinon, E. Global profiling of protein lipidation using chemical proteomic technologies. Curr Opin Chem Biol 24, 48-57 (2015). 173. Charron, G., Wilson, J. & Hang, H.C. Chemical tools for understanding protein lipidation in eukaryotes. Curr Opin Chem Biol 13, 382-91 (2009). 174. Hannoush, R.N. & Sun, J. The chemical toolbox for monitoring protein fatty acylation and prenylation. Nat Chem Biol 6, 498-506 (2010). 175. Hang, H.C. & Linder, M.E. Exploring protein lipidation with chemical biology. Chem Rev 111, 6341-58 (2011). 176. Grammel, M. & Hang, H.C. Chemical reporters for biological discovery. Nat Chem Biol 9, 475-84 (2013). 177. Thinon, E. & Hang, H.C. Chemical reporters for exploring protein acylation. Biochem Soc Trans 43, 253-61 (2015).

191

178. Best, M.D. Click chemistry and bioorthogonal reactions: unprecedented selectivity in the labeling of biological molecules. Biochemistry 48, 6571-84 (2009). 179. Carell, T. & Vrabel, M. Bioorthogonal Chemistry—Introduction and Overview. Topics in Current Chemistry 374, 1-21 (2016). 180. Jewett, J.C. & Bertozzi, C.R. Cu-free click cycloaddition reactions in chemical biology. Chemical Society Reviews 39, 1272-1279 (2010). 181. Sletten, E.M. & Bertozzi, C.R. From Mechanism to Mouse: A Tale of Two Bioorthogonal Reactions. Accounts of Chemical Research 44, 666-676 (2011). 182. van Berkel, S.S., van Eldijk, M.B. & van Hest, J.C.M. Staudinger Ligation as a Method for Bioconjugation. Angewandte Chemie-International Edition 50, 8806-8827 (2011). 183. de Araujo, A.D. et al. Diels-Alder ligation and surface immobilization of proteins. Angewandte Chemie-International Edition 45, 296-301 (2006). 184. Willems, L.I., Verdoes, M., Florea, B.I., van der Marel, G.A. & Overkleeft, H.S. Two- Step Labeling of Endogenous Enzymatic Activities by Diels-Alder Ligation. Chembiochem 11, 1769-1781 (2010). 185. Wright, M.H. et al. Validation of N-myristoyltransferase as an antimalarial drug target using an integrated chemical biology approach. Nat Chem 6, 112-21 (2014). 186. Thinon, E. et al. Global profiling of co- and post-translationally N-myristoylated proteomes in human cells. Nat Commun 5, 4919 (2014). 187. Wright, M.H. et al. Global analysis of protein N-myristoylation and exploration of N- myristoyltransferase as a drug target in the neglected human pathogen Leishmania donovani. Chem Biol 22, 342-54 (2015). 188. Wright, M.H., Paape, D., Price, H.P., Smith, D.F. & Tate, E.W. Global Profiling and Inhibition of Protein Lipidation in Vector and Host Stages of the Sleeping Sickness Parasite Trypanosoma brucei. ACS Infect Dis 2, 427-441 (2016). 189. Roth, A.F. et al. Global analysis of protein palmitoylation in yeast. Cell 125, 1003-13 (2006). 190. Yount, J.S. et al. Palmitoylome profiling reveals S-palmitoylation-dependent antiviral activity of IFITM3. Nat Chem Biol 6, 610-4 (2010). 191. Martin, B.R., Wang, C., Adibekian, A., Tully, S.E. & Cravatt, B.F. Global profiling of dynamic protein palmitoylation. Nat Methods 9, 84-9 (2012). 192. Serwa, R.A., Abaitua, F., Krause, E., Tate, E.W. & O'Hare, P. Systems Analysis of Protein Fatty Acylation in Herpes Simplex Virus-Infected Cells Using Chemical Proteomics. Chem Biol 22, 1008-17 (2015). 193. Heal, W.P. et al. Bioorthogonal chemical tagging of protein cholesterylation in living cells. Chem Commun (Camb) 47, 4081-3 (2011). 194. Ciepla, P. et al. New chemical probes targeting cholesterylation of Sonic Hedgehog in human cells and zebrafish. Chem Sci 5, 4249-4259 (2014). 195. Kho, Y. et al. A tagging-via-substrate technology for detection and proteomics of farnesylated proteins. Proc Natl Acad Sci U S A 101, 12479-84 (2004). 196. Rose, M.W. et al. Evaluation of geranylazide and farnesylazide diphosphate for incorporation of prenylazides into a CAAX box-containing peptide using protein farnesyltransferase. J Pept Res 65, 529-37 (2005). 197. Rose, M.W. et al. Enzymatic incorporation of orthogonally reactive prenylazide groups into peptides using geranylazide diphosphate via protein farnesyltransferase: implications for selective protein labeling. Biopolymers 80, 164-71 (2005).

192

198. Nguyen, U.T. et al. Exploiting the substrate tolerance of farnesyltransferase for site- selective protein derivatization. Chembiochem 8, 408-23 (2007). 199. Labadie, G.R., Viswanathan, R. & Poulter, C.D. Farnesyl diphosphate analogues with omega-bioorthogonal azide and alkyne functional groups for protein farnesyl transferase-catalyzed ligation reactions. J Org Chem 72, 9291-7 (2007). 200. Nguyen, U.T. et al. Analysis of the eukaryotic prenylome by isoprenoid affinity tagging. Nat Chem Biol 5, 227-35 (2009). 201. Berry, A.F. et al. Rapid multilabel detection of geranylgeranylated proteins by using bioorthogonal ligation chemistry. Chembiochem 11, 771-3 (2010). 202. Placzek, A.T. & Gibbs, R.A. New synthetic methodology for the construction of 7- substituted farnesyl diphosphate analogs. Org Lett 13, 3576-9 (2011). 203. Das, D. et al. Flexible and general synthesis of functionalized phosphoisoprenoids for the study of prenylation in vivo and in vitro. Chembiochem 13, 674-83 (2012). 204. Chan, L.N. et al. A novel approach to tag and identify geranylgeranylated proteins. Electrophoresis 30, 3598-606 (2009). 205. Ivanov, S.S., Charron, G., Hang, H.C. & Roy, C.R. Lipidation by the host prenyltransferase machinery facilitates membrane localization of Legionella pneumophila effector proteins. J Biol Chem 285, 34686-98 (2010). 206. Charron, G., Tsou, L.K., Maguire, W., Yount, J.S. & Hang, H.C. Alkynyl-farnesol reporters for detection of protein S-prenylation in cells. Mol Biosyst 7, 67-73 (2011). 207. Charron, G., Li, M.M., MacDonald, M.R. & Hang, H.C. Prenylome profiling reveals S- farnesylation is crucial for membrane targeting and antiviral activity of ZAP long- isoform. Proc Natl Acad Sci U S A 110, 11085-90 (2013). 208. Palsuledesai, C.C. et al. Metabolic Labeling with an Alkyne-modified Isoprenoid Analog Facilitates Imaging and Quantification of the Prenylome in Cells. ACS Chem Biol (2016). 209. Palsuledesai, C.C., Ochocki, J.D., Markowski, T.W. & Distefano, M.D. A combination of metabolic labeling and 2D-DIGE analysis in response to a farnesyltransferase inhibitor facilitates the discovery of new prenylated proteins. Mol Biosyst 10, 1094- 103 (2014). 210. Wang, Y.C. & Distefano, M.D. Synthetic isoprenoid analogues for the study of prenylated proteins: Fluorescent imaging and proteomic applications. Bioorg Chem 64, 59-65 (2015). 211. Westfall, D., Aboushadi, N., Shackelford, J.E. & Krisans, S.K. Metabolism of farnesol: phosphorylation of farnesol by rat liver microsomal and peroxisomal fractions. Biochem Biophys Res Commun 230, 562-8 (1997). 212. Bentinger, M., Grunler, J., Peterson, E., Swiezewska, E. & Dallner, G. Phosphorylation of farnesol in rat liver microsomes: properties of farnesol kinase and farnesyl phosphate kinase. Arch Biochem Biophys 353, 191-8 (1998). 213. Onono, F. et al. Efficient use of exogenous isoprenols for protein isoprenylation by MDA-MB-231 cells is regulated independently of the mevalonate pathway. J Biol Chem 288, 27444-55 (2013). 214. Crick, D.C., Andres, D.A. & Waechter, C.J. Novel Salvage Pathway Utilizing Farnesol and Geranylgeraniol for Protein Isoprenylation. Biochemical and Biophysical Research Communications 237, 483-487 (1997). 215. Hosokawa, A. et al. Evaluation of an Alkyne-containing Analogue of Farnesyl Diphosphate as a Dual Substrate for Protein-prenyltransferases. International Journal of Peptide Research and Therapeutics 13, 345-354 (2007).

193

216. Jennings, B.C. et al. Analogs of farnesyl diphosphate alter CaaX substrate specificity and reactions rates of protein farnesyltransferase. Bioorg Med Chem Lett 26, 1333-6 (2016). 217. Walther, T.C. & Mann, M. Mass spectrometry-based proteomics in cell biology. J Cell Biol 190, 491-500 (2010). 218. McLafferty, F.W. et al. Top-down MS, a powerful complement to the high capabilities of proteolysis proteomics. FEBS J 274, 6256-68 (2007). 219. Zhang, Y., Fonslow, B.R., Shan, B., Baek, M.C. & Yates, J.R., 3rd. Protein analysis by shotgun/bottom-up proteomics. Chem Rev 113, 2343-94 (2013). 220. Banerjee, S. & Mazumdar, S. Electrospray ionization mass spectrometry: a technique to access the information beyond the molecular weight of the analyte. Int J Anal Chem 2012, 282574 (2012). 221. Chernushevich, I.V., Loboda, A.V. & Thomson, B.A. An introduction to quadrupole- time-of-flight mass spectrometry. J Mass Spectrom 36, 849-65 (2001). 222. Hu, Q. et al. The Orbitrap: a new mass spectrometer. J Mass Spectrom 40, 430-43 (2005). 223. von Stechow, L., Francavilla, C. & Olsen, J.V. Recent findings and technological advances in phosphoproteomics for cells and tissues. Expert Rev Proteomics 12, 469-87 (2015). 224. Riley, N.M. & Coon, J.J. Phosphoproteomics in the Age of Rapid and Deep Proteome Profiling. Anal Chem 88, 74-94 (2016). 225. Hoffman, M.D. & Kast, J. Mass spectrometric characterization of lipid-modified peptides for the analysis of acylated proteins. J Mass Spectrom 41, 229-41 (2006). 226. Wotske, M., Wu, Y. & Wolters, D.A. Liquid chromatographic analysis and mass spectrometric identification of farnesylated peptides. Anal Chem 84, 6848-55 (2012). 227. Bhawal, R.P., Sadananda, S.C., Bugarin, A., Laposa, B. & Chowdhury, S.M. Mass spectrometry cleavable strategy for identification and differentiation of prenylated peptides. Anal Chem 87, 2178-86 (2015). 228. Neilson, K.A. et al. Less label, more free: approaches in label-free quantitative mass spectrometry. Proteomics 11, 535-53 (2011). 229. Bakalarski, C.E. & Kirkpatrick, D.S. A Biologist's Field Guide to Multiplexed Quantitative Proteomics. Mol Cell Proteomics 15, 1489-97 (2016). 230. Ong, S.E. & Mann, M. A practical recipe for stable isotope labeling by amino acids in cell culture (SILAC). Nat Protoc 1, 2650-60 (2006). 231. Zanivan, S., Krueger, M. & Mann, M. In vivo quantitative proteomics: the SILAC mouse. Methods Mol Biol 757, 435-50 (2012). 232. Geiger, T. et al. Initial quantitative proteomic map of 28 mouse tissues using the SILAC mouse. Mol Cell Proteomics 12, 1709-22 (2013). 233. Geiger, T. et al. Use of stable isotope labeling by amino acids in cell culture as a spike-in standard in quantitative proteomics. Nat Protoc 6, 147-57 (2011). 234. Geiger, T., Cox, J., Ostasiewicz, P., Wisniewski, J.R. & Mann, M. Super-SILAC mix for quantitative proteomics of human tumor tissue. Nat Methods 7, 383-5 (2010). 235. Shenoy, A. & Geiger, T. Super-SILAC: current trends and future perspectives. Expert Rev Proteomics 12, 13-9 (2015). 236. Pierce, A. et al. Eight-channel iTRAQ enables comparison of the activity of six leukemogenic tyrosine kinases. Mol Cell Proteomics 7, 853-63 (2008). 237. Ross, P.L. et al. Multiplexed protein quantitation in Saccharomyces cerevisiae using amine-reactive isobaric tagging reagents. Mol Cell Proteomics 3, 1154-69 (2004).

194

238. Dayon, L. et al. Relative quantification of proteins in human cerebrospinal fluids by MS/MS using 6-plex isobaric tags. Anal Chem 80, 2921-31 (2008). 239. Boersema, P.J., Raijmakers, R., Lemeer, S., Mohammed, S. & Heck, A.J. Multiplex peptide stable isotope dimethyl labeling for quantitative proteomics. Nat Protoc 4, 484-94 (2009). 240. Li, N. et al. Relative quantification of proteasome activity by activity-based protein profiling and LC-MS/MS. Nat Protoc 8, 1155-68 (2013). 241. Boersema, P.J., Aye, T.T., van Veen, T.A., Heck, A.J. & Mohammed, S. Triplex protein quantification based on stable isotope labeling by peptide dimethylation applied to cell and tissue lysates. Proteomics 8, 4624-32 (2008). 242. Choe, L. et al. 8-plex quantitation of changes in cerebrospinal fluid protein expression in subjects undergoing intravenous immunoglobulin treatment for Alzheimer's disease. Proteomics 7, 3651-60 (2007). 243. McAlister, G.C. et al. Increasing the multiplexing capacity of TMTs using reporter ion isotopologues with isobaric masses. Anal Chem 84, 7469-78 (2012). 244. Lau, H.T., Suh, H.W., Golkowski, M. & Ong, S.E. Comparing SILAC- and stable isotope dimethyl-labeling approaches for quantitative proteomics. J Proteome Res 13, 4164-74 (2014). 245. Wang, Y.C. & Distefano, M.D. Synthetic isoprenoid analogues for the study of prenylated proteins: Fluorescent imaging and proteomic applications. Bioorg Chem 64, 59-65 (2016). 246. Berndt, N. & Sebti, S.M. Measurement of protein farnesylation and geranylgeranylation in vitro, in cultured cells and in biopsies, and the effects of prenyl transferase inhibitors. Nat Protoc 6, 1775-91 (2011). 247. Umbreit, M.A. & Sharpless, K.B. Allylic Oxidation of Olefins by Catalytic and Stoichiometric Selenium Dioxide with Tert-Butyl Hydroperoxide. J Am Chem Soc 99, 5526-5528 (1977). 248. van Oost, B.A., Edgell, C.J., Hay, C.W. & MacGillivray, R.T. Isolation of a human von Willebrand factor cDNA from the hybrid endothelial cell line EA.hy926. Biochem Cell Biol 64, 699-705 (1986). 249. Charron, G. et al. Robust fluorescent detection of protein fatty-acylation with chemical reporters. J Am Chem Soc 131, 4967-75 (2009). 250. Gagneux, A., Winstein, S. & Young, W.G. REARRANGEMENT OF ALLYLIC AZIDES. Journal of the American Chemical Society 82, 5956-5957 (1960). 251. Holstein, S.A., Wohlford-Lenane, C.L. & Hohl, R.J. Isoprenoids influence expression of Ras and Ras-related proteins. Biochemistry 41, 13698-704 (2002). 252. Holstein, S.A., Wohlford-Lenane, C.L. & Hohl, R.J. Consequences of mevalonate depletion. Differential transcriptional, translational, and post-translational upregulation of Ras, Rap1a, RhoA, AND RhoB. J Biol Chem 277, 10678-82 (2002). 253. Holstein, S.A., Wohlford-Lenane, C.L., Wiemer, D.F. & Hohl, R.J. Isoprenoid pyrophosphate analogues regulate expression of Ras-related proteins. Biochemistry 42, 4384-91 (2003). 254. Khanzada, U.K. et al. Potent inhibition of small-cell lung cancer cell growth by simvastatin reveals selective functions of Ras isoforms in growth factor signalling. Oncogene 25, 877-87 (2006). 255. Zhu, Y., Casey, P.J., Kumar, A.P. & Pervaiz, S. Deciphering the signaling networks underlying simvastatin-induced apoptosis in human cancer cells: evidence for non- canonical activation of RhoA and Rac1 GTPases. Cell Death Dis 4, e568 (2013).

195

256. Cordle, A., Koenigsknecht-Talboo, J., Wilkinson, B., Limpert, A. & Landreth, G. Mechanisms of statin-mediated inhibition of small G-protein function. J Biol Chem 280, 34202-9 (2005). 257. Chiba, Y., Sato, S. & Misawa, M. Upregulation of geranylgeranyltransferase I in bronchial smooth muscle of mouse experimental asthma: its inhibition by lovastatin. J Smooth Muscle Res 46, 57-64 (2010). 258. Murai, M. et al. Characterization of the ubiquinone binding site in the alternative NADH-quinone oxidoreductase of Saccharomyces cerevisiae by photoaffinity labeling. Biochemistry 49, 2973-80 (2010). 259. Daines, A.M. & Abell, A.D. The preparation of side chain functionalized analogues of coenzyme Q for protein conjugation studies. Org Biomol Chem 2, 2371-5 (2004). 260. Dauben, W.G., Saugier, R.K. & Fleischhauer, I. Synthetic Studies Directed toward Cembranolides - Synthesis of the Basic Nucleus of Crassin Acetate. Journal of Organic Chemistry 50, 3767-3774 (1985). 261. Odejinmi, S.I. & Wiemer, D.F. Synthesis of arieianal, a prenylated benzoic acid from Piper arieianum. J Nat Prod 68, 1375-9 (2005). 262. Ohnuma, S., Watanabe, M. & Nishino, T. Identification and characterization of geranylgeraniol kinase and geranylgeranyl phosphate kinase from the Archaebacterium Sulfolobus acidocaldarius. J Biochem 119, 541-7 (1996). 263. Joo, J.H. & Jetten, A.M. Molecular mechanisms involved in farnesol-induced apoptosis. Cancer Lett 287, 123-35 (2010). 264. Michalski, A. et al. Mass spectrometry-based proteomics using Q Exactive, a high- performance benchtop quadrupole Orbitrap mass spectrometer. Mol Cell Proteomics 10, M111 011015 (2011). 265. Catherman, A.D. et al. Large-scale top-down proteomics of the human proteome: membrane proteins, mitochondria, and senescence. Mol Cell Proteomics 12, 3465- 73 (2013). 266. Adamson, P., Marshall, C.J., Hall, A. & Tilbrook, P.A. Post-translational modifications of p21rho proteins. J Biol Chem 267, 20033-8 (1992). 267. MacLeod, K.J., Fuller, R.S., Scholten, J.D. & Ahn, K. Conserved cysteine and tryptophan residues of the endothelin-converting enzyme-1 CXAW motif are critical for protein maturation and enzyme activity. J Biol Chem 276, 30608-14 (2001). 268. DeGraw, A.J. et al. Evaluation of alkyne-modified isoprenoids as chemical reporters of protein prenylation. Chem Biol Drug Des 76, 460-71 (2010). 269. Thinon, E. in Department of Chemistry (Imperial College London, 2013). 270. Downes, B.P., Saracco, S.A., Lee, S.S., Crowell, D.N. & Vierstra, R.D. MUBs, a family of ubiquitin-fold proteins that are plasma membrane-anchored by prenylation. J Biol Chem 281, 27145-57 (2006). 271. Thiele, C. et al. Tracing fatty acid metabolism by click chemistry. ACS Chem Biol 7, 2004-11 (2012). 272. Liu, Z. et al. Membrane-associated farnesylated UCH-L1 promotes alpha-synuclein neurotoxicity and is a therapeutic target for Parkinson's disease. Proc Natl Acad Sci U S A 106, 4635-40 (2009). 273. Wilson, L.A., McKeown, L., Tumova, S., Li, J. & Beech, D.J. Expression of a long variant of CRACR2A that belongs to the Rab GTPase protein family in endothelial cells. Biochem Biophys Res Commun 456, 398-402 (2015). 274. Srikanth, S. et al. A large Rab GTPase encoded by CRACR2A is a component of subsynaptic vesicles that transmit T cell activation signals. Sci Signal 9, ra31 (2016).

196

275. Lebowitz, P.F., Davide, J.P. & Prendergast, G.C. Evidence that farnesyltransferase inhibitors suppress Ras transformation by interfering with Rho activity. Mol Cell Biol 15, 6613-22 (1995). 276. Ulsh, L.S. & Shih, T.Y. Metabolic turnover of human c-rasH p21 protein of EJ bladder carcinoma and its normal cellular and viral homologs. Mol Cell Biol 4, 1647-52 (1984). 277. Onono, F.O. et al. A tagging-via-substrate approach to detect the farnesylated proteome using two-dimensional electrophoresis coupled with Western blotting. Mol Cell Proteomics 9, 742-51 (2010). 278. Hajjar, K.A. The Biology of Annexin A2: From Vascular Fibrinolysis to Innate Immunity. Trans Am Clin Climatol Assoc 126, 144-55 (2015). 279. Rescher, U. & Gerke, V. Annexins--unique membrane binding proteins with diverse functions. J Cell Sci 117, 2631-9 (2004). 280. Ayala-Sanmartin, J., Henry, J.P. & Pradel, L.A. Cholesterol regulates membrane binding and aggregation by annexin 2 at submicromolar Ca(2+) concentration. Biochim Biophys Acta 1510, 18-28 (2001). 281. Mayer, G., Poirier, S. & Seidah, N.G. Annexin A2 is a C-terminal PCSK9-binding protein that regulates endogenous low density lipoprotein receptor levels. J Biol Chem 283, 31791-801 (2008). 282. Goldstein, J.L. & Brown, M.S. The LDL receptor. Arterioscler Thromb Vasc Biol 29, 431-8 (2009). 283. Simons, M. & Raposo, G. Exosomes--vesicular carriers for intercellular communication. Curr Opin Cell Biol 21, 575-81 (2009). 284. Liscum, L. Niemann-Pick type C mutations cause lipid traffic jam. Traffic 1, 218-25 (2000). 285. Carboni, J.M. et al. Farnesyltransferase inhibitors are inhibitors of Ras but not R- Ras2/TC21, transformation. Oncogene 10, 1905-13 (1995). 286. Yokoyama, K., Zimmerman, K., Scholten, J. & Gelb, M.H. Differential prenyl pyrophosphate binding to mammalian protein geranylgeranyltransferase-I and protein farnesyltransferase and its consequence on the specificity of protein prenylation. J Biol Chem 272, 3944-52 (1997). 287. Yang, Y. & Verhelst, S.H. Cleavable trifunctional biotin reagents for protein labelling, capture and release. Chem Commun (Camb) 49, 5366-8 (2013). 288. Weerapana, E., Speers, A.E. & Cravatt, B.F. Tandem orthogonal proteolysis-activity- based protein profiling (TOP-ABPP)--a general method for mapping sites of probe modification in proteomes. Nat Protoc 2, 1414-25 (2007). 289. Broncel, M. et al. Multifunctional reagents for quantitative proteome-wide analysis of protein modification in human cells and dynamic profiling of protein lipidation during vertebrate development. Angew Chem Int Ed Engl 54, 5948-51 (2015). 290. Heal, W.P., Wright, M.H., Thinon, E. & Tate, E.W. Multifunctional protein labeling via enzymatic N-terminal tagging and elaboration by click chemistry. Nat Protoc 7, 105- 17 (2012). 291. Schilling, O., Barre, O., Huesgen, P.F. & Overall, C.M. Proteome-wide analysis of protein carboxy termini: C terminomics. Nat Methods 7, 508-11 (2010). 292. Zhang, J. et al. PEAKS DB: de novo sequencing assisted database search for sensitive and accurate peptide identification. Mol Cell Proteomics 11, M111 010587 (2012).

197

293. Lerner, E.C. et al. Ras CAAX peptidomimetic FTI-277 selectively blocks oncogenic Ras signaling by inducing cytoplasmic accumulation of inactive Ras-Raf complexes. J Biol Chem 270, 26802-6 (1995). 294. McGuire, T.F., Qian, Y., Vogt, A., Hamilton, A.D. & Sebti, S.M. Platelet-derived growth factor receptor tyrosine phosphorylation requires protein geranylgeranylation but not farnesylation. J Biol Chem 271, 27402-7 (1996). 295. Delarue, F.L. et al. Farnesyltransferase and geranylgeranyltransferase I inhibitors upregulate RhoB expression by HDAC1 dissociation, HAT association and histone acetylation of the RhoB promoter. Oncogene 26, 633-40 (2007). 296. End, D.W. et al. Characterization of the antitumor effects of the selective farnesyl protein transferase inhibitor R115777 in vivo and in vitro. Cancer Res 61, 131-7 (2001). 297. Hara, M. et al. Identification of Ras farnesyltransferase inhibitors by microbial screening. Proc Natl Acad Sci U S A 90, 2281-5 (1993). 298. Nagase, T. et al. Inhibition of cell growth of human hepatoma cell line (Hep G2) by a farnesyl protein transferase inhibitor: a preferential suppression of ras farnesylation. Int J Cancer 65, 620-6 (1996). 299. Vasudevan, A. et al. Potent, highly selective, and non-thiol inhibitors of protein geranylgeranyltransferase-I. J Med Chem 42, 1333-40 (1999). 300. Bon, R.S. et al. Structure-guided development of selective RabGGTase inhibitors. Angew Chem Int Ed Engl 50, 4957-61 (2011). 301. Lackner, M.R. et al. Chemical genetics identifies Rab geranylgeranyl transferase as an apoptotic target of farnesyl transferase inhibitors. Cancer Cell 7, 325-36 (2005). 302. Ghosh, A.K. & Brindisi, M. Organic carbamates in drug design and medicinal chemistry. J Med Chem 58, 2895-940 (2015). 303. Sears, K.T., Daino, H. & Carey, G.B. Reactive oxygen species-dependent destruction of MEK and Akt in Manumycin stimulated death of lymphoid tumor and myeloma cell lines. Int J Cancer 122, 1496-505 (2008). 304. Carey, G.B., Roy, S.K. & Daino, H. The natural tumorcide Manumycin-A targets protein phosphatase 1alpha and reduces hydrogen peroxide to induce lymphoma apoptosis. Exp Cell Res 332, 136-45 (2015). 305. Bernier, M. et al. Binding of manumycin A inhibits IkappaB kinase beta activity. J Biol Chem 281, 2551-61 (2006). 306. Arenz, C. et al. Manumycin A and its analogues are irreversible inhibitors of neutral sphingomyelinase. Chembiochem 2, 141-3 (2001). 307. Angers, S. et al. Molecular architecture and assembly of the DDB1-CUL4A ubiquitin ligase machinery. Nature 443, 590-3 (2006). 308. Lee, J. & Zhou, P. DCAFs, the missing link of the CUL4-DDB1 ubiquitin ligase. Mol Cell 26, 775-80 (2007). 309. Wu, F., Wang, S., Xing, J., Li, M. & Zheng, C. Characterization of nuclear import and export signals determining the subcellular localization of WD repeat-containing protein 42A (WDR42A). FEBS Lett 586, 1079-85 (2012). 310. Wu, Y. et al. A genome-scale CRISPR-Cas9 screening method for protein stability reveals novel regulators of Cdc25A. Cell Discov 2, 16014 (2016). 311. Shen, T. & Huang, S. The role of Cdc25A in the regulation of cell proliferation and apoptosis. Anticancer Agents Med Chem 12, 631-9 (2012). 312. Caballe, A. et al. ULK3 regulates cytokinetic abscission by phosphorylating ESCRT- III proteins. Elife 4, e06547 (2015).

198

313. Maloverjan, A. et al. Dual function of UNC-51-like kinase 3 (Ulk3) in the Sonic hedgehog signaling pathway. J Biol Chem 285, 30079-90 (2010). 314. Maloverjan, A., Piirsoo, M., Michelson, P., Kogerman, P. & Osterlund, T. Identification of a novel serine/threonine kinase ULK3 as a positive regulator of Hedgehog pathway. Exp Cell Res 316, 627-37 (2010). 315. Jiang, J. & Hui, C.C. Hedgehog signaling in development and cancer. Dev Cell 15, 801-12 (2008). 316. Ng, J.M. & Curran, T. The Hedgehog's tale: developing strategies for targeting cancer. Nat Rev Cancer 11, 493-501 (2011). 317. Chen, C. et al. Characterization of Cep85 - a new antagonist of Nek2A that is involved in the regulation of centrosome disjunction. J Cell Sci 128, 3290-303 (2015). 318. Nam, H.-J., Naylor, R.M. & van Deursen, J.M. Centrosome dynamics as a source of chromosomal instability. Trends in Cell Biology 25, 65-73 (2015). 319. Rikiyama, T. et al. GCF2: expression and molecular analysis of repression. Biochim Biophys Acta 1629, 15-25 (2003). 320. Fong, K.S. & de Couet, H.G. Novel proteins interacting with the leucine-rich repeat domain of human flightless-I identified by the yeast two-hybrid system. Genomics 58, 146-57 (1999). 321. Sjoblom, T. et al. The consensus coding sequences of human breast and colorectal cancers. Science 314, 268-74 (2006). 322. Wilson, S.A., Brown, E.C., Kingsman, A.J. & Kingsman, S.M. TRIP: a novel double stranded RNA binding protein which interacts with the leucine rich repeat of flightless I. Nucleic Acids Res 26, 3460-7 (1998). 323. Suriano, A.R. et al. GCF2/LRRFIP1 represses tumor necrosis factor alpha expression. Mol Cell Biol 25, 9073-81 (2005). 324. Reed, A.L. et al. Molecular cloning and characterization of a transcription regulator with homology to GC-binding factor. J Biol Chem 273, 21594-602 (1998). 325. Khachigian, L.M. et al. GC factor 2 represses platelet-derived growth factor A-chain gene transcription and is itself induced by arterial injury. Circ Res 84, 1258-67 (1999). 326. Shi, L. et al. Noncoding RNAs and LRRFIP1 regulate TNF expression. J Immunol 192, 3057-67 (2014). 327. Mallolas, J. et al. A polymorphism in the EAAT2 promoter is associated with higher glutamate concentrations and higher frequency of progressing stroke. J Exp Med 203, 711-7 (2006). 328. Ohtsuka, H. et al. GC-binding factor 2 interacts with dishevelled and regulates Wnt signaling pathways in human carcinoma cell lines. Int J Cancer 129, 1599-610 (2011). 329. Douchi, D. et al. Silencing of LRRFIP1 reverses the epithelial-mesenchymal transition via inhibition of the Wnt/beta-catenin signaling pathway. Cancer Lett 365, 132-40 (2015). 330. Etienne-Manneville, S. & Hall, A. Rho GTPases in cell biology. Nature 420, 629-35 (2002). 331. Jaffe, A.B. & Hall, A. Rho GTPases: biochemistry and biology. Annu Rev Cell Dev Biol 21, 247-69 (2005). 332. Vega, F.M. & Ridley, A.J. Rho GTPases in cancer cell biology. FEBS Lett 582, 2093- 101 (2008).

199

333. Ariake, K. et al. GCF2/LRRFIP1 promotes colorectal cancer metastasis and liver invasion through integrin-dependent RhoA activation. Cancer Lett 325, 99-107 (2012). 334. Davy, D.A., Campbell, H.D., Fountain, S., de Jong, D. & Crouch, M.F. The flightless I protein colocalizes with actin- and microtubule-based structures in motile Swiss 3T3 fibroblasts: evidence for the involvement of PI 3-kinase and Ras-related small GTPases. J Cell Sci 114, 549-62 (2001). 335. Choe, N. et al. The microRNA miR-132 targets Lrrfip1 to block vascular smooth muscle cell proliferation and neointimal hyperplasia. Atherosclerosis 229, 348-55 (2013). 336. Goodall, A.H. et al. Transcription profiling in human platelets reveals LRRFIP1 as a novel protein regulating platelet function. Blood 116, 4646-56 (2010). 337. Yang, P. et al. The cytosolic nucleic acid sensor LRRFIP1 mediates the production of type I interferon via a beta-catenin-dependent pathway. Nat Immunol 11, 487-94 (2010). 338. Bagashev, A. et al. Leucine-rich repeat (in Flightless I) interacting protein-1 regulates a rapid type I interferon response. J Interferon Cytokine Res 30, 843-52 (2010). 339. Montani, D. et al. Targeted therapies in pulmonary arterial hypertension. Pharmacol Ther 141, 172-91 (2014). 340. Bemis, L. et al. Distinct aerobic and hypoxic mechanisms of HIF-alpha regulation by CSN5. Genes Dev 18, 739-44 (2004). 341. Underwood, K.W. et al. A novel calcium-independent phospholipase A2, cPLA2- gamma, that is prenylated and contains homology to cPLA2. J Biol Chem 273, 21926-32 (1998). 342. Murakami, M., Masuda, S. & Kudo, I. Arachidonate release and prostaglandin production by group IVC phospholipase A2 (cytosolic phospholipase A2gamma). Biochem J 372, 695-702 (2003). 343. Valli, A. et al. Hypoxia induces a lipogenic cancer cell phenotype via HIF1alpha- dependent and -independent pathways. Oncotarget 6, 1920-41 (2015). 344. Goncharova, E.A. mTOR and vascular remodeling in lung diseases: current challenges and therapeutic prospects. FASEB J 27, 1796-807 (2013). 345. Basso, A.D. et al. The farnesyl transferase inhibitor (FTI) SCH66336 (lonafarnib) inhibits Rheb farnesylation and mTOR signaling. Role in FTI enhancement of taxane and tamoxifen anti-tumor activity. J Biol Chem 280, 31101-8 (2005). 346. Castro, A.F., Rebhun, J.F., Clark, G.J. & Quilliam, L.A. Rheb binds tuberous sclerosis complex 2 (TSC2) and promotes S6 kinase activation in a rapamycin- and farnesylation-dependent manner. J Biol Chem 278, 32493-6 (2003). 347. Yue, X. et al. Rnd3/RhoE Modulates Hypoxia-Inducible Factor 1alpha/Vascular Endothelial Growth Factor Signaling by Stabilizing Hypoxia-Inducible Factor 1alpha and Regulates Responsive Cardiac Angiogenesis. Hypertension 67, 597-605 (2016). 348. Shimoda, L.A. & Semenza, G.L. HIF and the lung: role of hypoxia-inducible factors in pulmonary development and disease. Am J Respir Crit Care Med 183, 152-6 (2011). 349. Riento, K., Villalonga, P., Garg, R. & Ridley, A. Function and regulation of RhoE. Biochem Soc Trans 33, 649-51 (2005). 350. Lebowitz, P.F., Casey, P.J., Prendergast, G.C. & Thissen, J.A. Farnesyltransferase inhibitors alter the prenylation and growth-stimulating function of RhoB. J Biol Chem 272, 15591-4 (1997).

200

351. Rivera-Torres, J. et al. Identification of mitochondrial dysfunction in Hutchinson- Gilford progeria syndrome through use of stable isotope labeling with amino acids in cell culture. J Proteomics 91, 466-77 (2013). 352. Scaffidi, P. & Misteli, T. Lamin A-dependent nuclear defects in human aging. Science 312, 1059-63 (2006). 353. Yang, Y., Yang, X. & Verhelst, S.H. Comparative analysis of click chemistry mediated activity-based protein profiling in cell lysates. Molecules 18, 12599-608 (2013). 354. Wang, L. et al. A proteomic study of Hutchinson-Gilford progeria syndrome: Application of 2D-chromotography in a premature aging disease. Biochem Biophys Res Commun 417, 1119-26 (2012). 355. Mateos, J. et al. iTRAQ-based analysis of progerin expression reveals mitochondrial dysfunction, reactive oxygen species accumulation and altered proteostasis. Stem Cell Res Ther 6, 119 (2015). 356. Jitrapakdee, S. & Wallace, J.C. Structure, function and regulation of pyruvate carboxylase. Biochem J 340 ( Pt 1), 1-16 (1999). 357. Peinado, J.R. et al. Proteomic profiling of adipose tissue from Zmpste24-/- mice, a model of lipodystrophy and premature aging, reveals major changes in mitochondrial function and vimentin processing. Mol Cell Proteomics 10, M111 008094 (2011). 358. Varela, I. et al. Combined treatment with statins and aminobisphosphonates extends longevity in a mouse model of human premature aging. Nat Med 14, 767-72 (2008). 359. Kitamura, T. et al. Substrate specificity, plasma membrane localization, and lipid modification of the aldehyde dehydrogenase ALDH3B1. Biochim Biophys Acta 1831, 1395-401 (2013). 360. Marchitti, S.A., Brocker, C., Orlicky, D.J. & Vasiliou, V. Molecular characterization, expression analysis, and role of ALDH3B1 in the cellular protection against oxidative stress. Free Radic Biol Med 49, 1432-43 (2010). 361. Santoriello, C. & Zon, L.I. Hooked! Modeling human disease in zebrafish. J Clin Invest 122, 2337-43 (2012). 362. Moosajee, M. et al. Single choroideremia gene in nonmammalian vertebrates explains early embryonic lethality of the zebrafish model of choroideremia. Invest Ophthalmol Vis Sci 50, 3009-16 (2009). 363. Thorpe, J.L., Doitsidou, M., Ho, S.Y., Raz, E. & Farber, S.A. Germ cell migration in zebrafish is dependent on HMGCoA reductase activity and prenylation. Dev Cell 6, 295-302 (2004). 364. D'Amico, L., Scott, I.C., Jungblut, B. & Stainier, D.Y. A mutation in zebrafish hmgcr1b reveals a role for isoprenoids in vertebrate heart-tube formation. Curr Biol 17, 252-9 (2007).

201

Appendix A: List of electronic files

File name Description File 1 - YnF, AzC15 All proteins identified in initial proteomics experiment and AzC20 comparison comparing YnF, AzC15 and AzC20. (Section 3.3) File 2 - YnF and YnGG All proteins identified in initial proteomics comparison comparison of YnF and YnGG. (Section 3.4) Spike-in SILAC quantitative proteomics comparison File 3 - YnF and YnGG of YnF and YnGG labelling in time-course and labelling optimisation concentration gradient studies. (Section 3.5) File 4 - Quantitative Quantitative comparison of YnF and YnGG comparison of probe incorporation by SILAC proteomics. (Section 3.6) preference SILAC quantitative proteomics analysis of YnF and File 5 - Initial inhibitor YnGG labelling in EA.hy926 cells in response to evaluation treatment with, FTI-277, GGTI-2133 or Bon-15. (Section 4.3.2) Spike-in SILAC quantitative proteomics analysis of File 6 - Inhibitor dose YnF and YnGG labelling in EA.hy926 cells in response in EA.hy926 response to treatment with FTI-277, GGTI-2133, cells Manumycin A and Tipifarnib. (Section 4.4) Spike-in SILAC quantitative proteomics comparison File 7 – Hypoxia-treated of YnF and YnGG labelling in hypoxia versus HPAECs normoxia in HPAECs. (Section 5.3.1) Dimethyl labelling quantitative proteomics File 8 - Tipifarnib- comparison of YnF and YnGG labelling in Tipifarnib- treated HPAECs and treated and non-treated HPAECs and HPSMCs. HPSMCs (Section 5.3.2) Spike-in SILAC quantitative comparison of YnF and File 9 - HGPS versus YnGG labelling in HGPS versus healthy cells with or healthy cells without Tipifarnib treatment. (Sections 5.5.1 & 5.5.2)

202

Appendix B: Unedited Western Blot Images

Figure B1. Unedited blots of total lysate (input) from Figure 20, Section 2.5

Figure B2. Unedited blots from Figure 21, Section 2.6

The column denoted “other” column refers to a condition not relevant for the experimental results presented and cropped from the blot presented in Figure 20.

203

Appendix C: Spectra of YnF-AzRB Modified Peptides

204

205

206

207

Appendix D: NMR Spectra of YnF

208

Appendix E: NMR Spectra of YnGG

209

Appendix F: Reprint Permission (Figure 9, Section 1.3.3)

210

211

212

213

214