High-Value Oxy-Pharmaceuticals from P450 BM3 ‘Gatekeeper’ Mutations

Home , Adrenalone, Cilnidipine, Cinepazide, Cytochrome P450 BM3, Iminodiacetic acid, Piceid

High-value oxy-pharmaceuticals from P450 BM3 ‘gatekeeper’ mutations

A thesis submitted to The University of Manchester for the degree of Doctor of Philosophy in the Faculty of Science and Engineering

2018

Laura N. Jeffreys School of Chemistry

BLANK PAGE

Table of Contents

Figures……. ………………………………………………………………………………..8

Tables……………………………………………………………………………………...10

Supplementary Figures………………………………………………………………….. 11

List of Abbreviations…………………………………………………………………….. 13

Abstract…………………………………………………………………………………... 15

Acknowledgements ………………………………………………………………………16

Declaration……………………………………………………………………………….. 17

Preface to the Journal Format Thesis………………………………………………….. 19

Author contributions……………………………………………………………………. 21

Chapter 1: General Introduction……………………………………………………... 23

1.1. An Overview of Cytochromes P450……………………………………... 23 1.1.1. The Evolution and Nomenclature of Cytochromes P450…………….23

1.1.2. The History of Cytochrome P450 Research ………………………….28

1.1.3. The P450 Catalytic Cycle …………………………………………….32

1.1.4. The Structure of P450 Enzymes ……………………………………...38

1.1.5. Unusual P450 Proteins ……………………………………………….43

1.2. The Natural Fusion Protein P450 BM3 (CYP102A1) ……………………47 1.2.1. The Structure of P450 BM3 ………………………………………….48

1.2.2. Electron Transfer Within P450 BM3 ………………………………...54

1.2.3. P450 BM3 Mutagenesis and the Gatekeeper Mutants ……………….58

1.3. Real-World Applications of P450 Enzymes ……………………………...61 1.3.1. Using P450 BM3 in the Pharmaceutical Industry ……………………63

1.3.2. Using other P450 Enzymes in Industry ………………………………65

1.4. Project Aims ……………………………………………………………...67 1.5. References ………………………………………………………………..68

Chapter 2: Characterization of the Structure and Interactions of the Active form of

P450 BM3 using Hybrid Mass Spectrometry Approaches……………………………. 82

2.1. Abstract …………………………………………………………………...82 2.2. Introduction……………………………………………………………….82 2.3. Materials and Methods …………………………………………………...86 2.3.1. P450 BM3 expression and purification ………………………………86

2.3.2. Native MS and IM-MS ……………………………………………….88

2.3.3. Collision-induced unfolding (CIU) …………………………………..89

2.3.4. Calculating CCS values from X-ray crystallographic structures …….89

2.3.5. Hydrogen-deuterium exchange mass spectrometry (HDX-MS) ……..89

2.4. Results…………………………………………………………………….91 2.4.1. Native mass spectrometry ……………………………………………91

2.4.2. Collision-induced unfolding and ion-mobility mass spectrometry …..91

2.4.3. Hydrogen-deuterium exchange mass spectrometry ………………….93

2.5. Discussion ……………………………………………………………….100 2.6. Conclusions ……………………………………………………………..105 2.7. References……………………………………………………………….107 2.8. Supplementary Information ……………………………………………..110

Chapter 3: The Promiscuous Nature of P450 BM3 and its Ability To Bind

Pharmaceutical Compounds Using a Novel Library Screen………………………… 115

3.1. Abstract ………………………………………………………………….115 3.2. Introduction……………………………………………………………...115 3.3. Materials and Methods ………………………………………………….119 3.3.1. Protein expression and purification …………………………………119

3.3.2. Screening of the FDA-approved compound library ………………...120

3.3.3. Binding affinity determination of pharmaceutical compounds ……..122

3.4. Results …………………………………………………………………..123 3.4.1. FDA-approved compound library screening ……………………….123

3.4.2. Determination of binding affinities for WT and DM BM3 variants ..125

3.5. Discussion ……………………………………………………………….128 3.6. Conclusions ……………………………………………………………..131 3.7. References……………………………………………………………….132 3.8. Supporting Information …………………………………………………135

Chapter 4: Novel Insights into P450 BM3 Interactions with FDA-approved

Antifungal Azole Drugs…………………………………………………………………157

4.1. Abstract ………………………………………………………………….157 4.2. Introduction……………………………………………………………...157 4.3. Materials and Methods ………………………………………………….161 4.3.1. Expression and purification of WT and A82F/F87V (DM) BM3 heme

domain proteins ……………………………………………………………………..161

4.3.2. UV-Visible spectroscopic assays of azole drug binding to WT and DM

BM3 heme domains …………………………………………………………………162

4.3.3. EPR spectroscopy studies of WT and DM BM3 heme domains bound

to azole drugs..………………………………………………………………………163

4.3.4. X-ray crystallography of the DM BM3 heme domain with azole

compounds………………………………………………………………………….. 163

4.4. Results…………………………………………………………………...164 4.4.1. Ligand binding assays of azole drugs to the WT and A82F/F87V (DM)

BM3 heme domains …………………………………………………………………164

4.4.2. Electron paramagnetic resonance (EPR) studies of WT and

A82F/F87V (DM) BM3 heme domains bound to azole drugs ……………………...168

4.4.3. X-ray crystallography of azole-bound DM heme domain

complexes……………………………………………………………………………170

4.5. Discussion ……………………………………………………………….173 4.6. Conclusions……………………………………………………………...177 4.7. References……………………………………………………………….179 4.8. Supporting Information …………………………………………………184

Chapter 5: Screening Antidiabetic Binding to P450 BM3 and the Production of

Human Metabolites…………………………………………………………………….. 187

5.1. Abstract ………………………………………………………………….187 5.2. Introduction……………………………………………………………...187 5.3. Materials and Methods ………………………………………………….190 5.3.1. P450 BM3 protein expression and purification …………………….190

5.3.2. Antidiabetic binding affinity determination... ………………………191

5.3.3. Electron paramagnetic resonance (EPR) spectroscopy…………….. 192

5.3.4. Structural determination by X-ray crystallography …………………192

5.3.5. Product determination by HPLC, LCMS, and LCMS/MS………….193

5.3.6. Product determination by NMR …………………………………….195

5.4. Results …………………………………………………………………..195 5.4.1. Binding affinity determination for antidiabetic compounds ………..195

5.4.2. EPR spectroscopy of the troglitazone-bound DM heme domain …...197

5.4.3. The X-ray crystal structure of the troglitazone-bound DM heme

domain……………………………………………………………………………….197

5.4.4. Production of human metabolites …………………………………...199

5.5. Discussion ……………………………………………………………….200 5.6. Conclusions……………………………………………………………...204 5.7. References……………………………………………………………….205 5.8. Supporting Information …………………………………………………208

Chapter 6: Binding of Fibrates to P450 BM3 Reveals Novel Changes to the P450

BM3 Landscape………………………………………………………………………… 215

6.1. Abstract ………………………………………………………………….215 6.2. Introduction……………………………………………………………...215 6.3. Materials and Methods ………………………………………………….219 6.3.1. Protein expression and purification of P450 BM3 ………………….219

6.3.2. Binding affinity determination ……………………………………...221

6.3.3. EPR analysis of ligand binding to WT and DM BM3 ……………...221 6

6.3.4. Native MS studies …………………………………………………..222

6.3.5. Collision-induced unfolding (CIU) analyzes ……………………….222

6.3.6. HDX-MS characterization ………………………………………….223

6.3.7. Product determination by HPLC, LCMS and LCMS/MS ………….224

6.3.8. Product determination by NMR …………………………………….225

6.4. Results …………………………………………………………………..226 6.4.1. Fibrate drug binding affinity determination ………………………...226

6.4.2. EPR spectroscopy of WT and DM BM3 heme domain in complex with

fibrate drugs…………………………………………………………………………228

6.4.3. Production of human metabolites from fibrate drugs…………...... 229

6.4.4. Native MS and CIU studies …………………………………………229

6.4.5. HDX-MS analysis of substrate binding to WT and DM BM3 heme

domains……...………………………………………………………………………231

6.5. Discussion………………………………………………………………. 233 6.6. Conclusions ……………………………………………………………..239 6.7. References……………………………………………………………….240 6.8. Supporting Information …………………………………………………244

Chapter 7: Conclusions and Future Work………………………………………….. 248

7.1. Summary ………………………………………………………………...248 7.2. Conclusions ……………………………………………………………..252 7.3. Future Work ……………………………………………………………..253 7.4. References……………………………………………………………….255

Word Count: 73,921

Figures

Figure 1: An overview of xenobiotic metabolism ...... 25

Figure 2: Different P450 redox class systems ...... 27

Figure 3: Binding of carbon monoxide to P450 BM3 elicits a Soret shift to 450 nm ... 30

Figure 4: Electron rearrangement in the d-orbitals of the heme iron atom upon

substrate binding ...... 33

Figure 5: An overview of the cytochrome P450 catalytic cycle ...... 37

Figure 6: The structures of the three most common heme prosthetic groups found in

hemoproteins ...... 38

Figure 7: The core of a P450 protein and the P450 fold...... 41

Figure 8: The primary and secondary structure of P450 BM3 ...... 50

Figure 9: The structures of the P450 BM3 heme and CPR domains ...... 53

Figure 10: The proposed CPR conformational changes that occur during electron

transfer to a cytochrome P450 ...... 57

Figure 11: A comparison of the flexibility of the P450 BM3 active site in the WT and

DM proteins ...... 62

Figure 12: Collision-induced unfolding of the P450 BM3 domains and the full-length

P450 BM3 enzyme ...... 92

Figure 13: HDX-MS of the surface of the isolated BM3 domains in comparison to the

full-length dimeric protein ...... 94

Figure 14: Binding of ligands to intact P450 BM3 elicits structural rearrangements

across the heme domain as visualized by HDX MS ...... 97

Figure 15: Binding of ligands to intact P450 BM3 elicits structural rearrangements

across the CPR domain as visualized by HDX MS ...... 99

Figure 16: A model of the full-length dimeric P450 BM3 enzymes using data collected

from hybrid MS techniques ...... 106

Figure 17: FDA-approved compound library hits presented according to their extent

of high-spin percentage induced ...... 124

Figure 18: A wide variety of structures were found to bind to the DM BM3 variant

...... 125

Figure 19: Binding of tioconazole to WT and DM BM3 heme domains ...... 166

Figure 20: Structures of azole antifungal drugs used in binding studies with the P450

BM3 WT and DM heme domains ...... 167

Figure 21: X-band EPR spectra for the WT and DM BM3 heme domain complexes

with azole drugs ...... 169

Figure 22: Crystal structures of the BM3 heme domain A82F/F87V mutants in

complex with different azole drugs ...... 171

Figure 23: Metformin-induces a type I (substrate) shift with the DM BM3 heme

domain ...... 196

Figure 24: The EPR spectrum of the troglitazone-bound DM BM3 heme domain ... 198

Figure 25: Troglitazone binding in the active site of the DM BM3 heme domain ..... 198

Figure 26: Production of stereo- and regioselectively oxidized products from P450

BM3 ...... 200

Figure 27: The structures of a variety of fibrate compounds ...... 226

Figure 28: Spectral binding titrations for the modified fatty acid NPG and the fibrate

drug bezafibrate with the BM3 DM heme domain ...... 228

Figure 29: CIU demonstrates differences in the unfolding patterns of DM ligand-

bound proteins ...... 230

Figure 30: Comparing the binding of NPG in WT and DM heme domain proteins by

HDX-MS ...... 232

Figure 31: Comparing the binding of bezafibrate and NPG ligands to the DM BM3

heme domain variant by HDX-MS ...... 234

Tables

Table 1: Binding affinity (Kd) values for WT and DM BM3 heme domains for

compounds displaying type II shifts ...... 126

Table 2: Binding affinity values for WT and DM BM3 heme domains with ligands

displaying type I Soret spectral shifts ...... 127

Table 3: Azole drug binding constants and associated spectral shifts for WT and DM

BM3 heme domains ...... 165

Table 4: Binding affinities for BM3 DM and WT heme domains with antidiabetic

compounds ...... 195

Table 5: Binding constants for fibrate class molecules demonstrate higher affinity for

the DM BM3 heme domain than for the WT BM3 heme domain ...... 227

Supplementary Figures

Figure S1: Native MS of intact P450 BM3 and its component domains in ligand-free

and ligand-bound states…………………………………………………………... 110

Figure S2: Experimentally and theoretically derived collision cross-section (CCS)

values for P450 BM3……………………………………………………………… 111

Figure S3: Relative deuterium uptake plots for ligand-free and ligand-bound full-

length P450 BM3 dimeric protein ……………………………………………….112

Figure S4: The coverage maps for the ligand-free P450 BM3 heme and CPR domains

………………………………………………………………………………………113

Figure S5: The coverage maps for the ligand-bound, full-length P450 BM3 dimeric

protein ……………………………………………………………………………..114

Figure S6: Percentage spin-state shift values obtained under near-saturating ligand

concentrations for the DM BM3 heme domain ………………………………15639

Figure S7: Table of crystallographic data for the azole bound DM complexes …….184

Figure S8: X-band EPR data sets for the WT BM3 heme domain bound to azole

compounds …………………………………………………………………………185

Figure S9: X-band EPR data sets for the DM BM3 heme domain bound to azole

compounds …………………………………………………………………………186

Figure S10: Table of EPR g-values for the WT and DM heme domain proteins bound

to a variety of antidiabetic compounds …………………………………………..208

Figure S11: Table of crystallographic data for the troglitazone bound DM variant

………………………………………………………………………………………209

Figure S12: LCMS/MS of rosiglitazone suggests the production of two distinct

metabolites …………………………………………………………………………210

Figure S13: LCMS/MS of metformin shows a potential oxidative deamination

reaction …………………………………………………………………………….211

Figure S14: 1H-NMR shows the production of a single pioglitazone metabolite

corresponding to an oxidation reaction by the DM BM3 variant………………212

Figure S15: 1H-NMR shows the production of a single troglitazone metabolite

corresponding to an oxidation reaction by the DM BM3 variant………………212

Figure S16: 1H-NMR shows the production of five metabolites corresponding to

hydroxylation reactions by the DM BM3 variant with repaglinide ……………213

Figure S17: 1H-NMR shows the production of a single metabolite and two eliminated

smaller molecules corresponding to an oxidation reaction by the DM BM3

variant with darglitazone …………………………………………………………214

Figure S18: X-band EPR data sets for the DM BM3 heme domain in complex with

fibrate compounds …………………………………………………………………244

Figure S19: 1H-NMR of fenofibrate shows the production of three metabolites …...245

Figure S20: 1H-NMR of gemfibrozil shows the production of two human metabolites

………………………………………………………………………………………246

Figure S21: Collision cross-section (CCS) values of the DM variant bound a variety of

ligands ………………………………………………………………………………247

List of Abbreviations

µL Microlitre µM Micromolar Å Angstrom ADR Adrenodoxin reductase Adx Adrenodoxin AI Auto-induction medium with terrific broth base AKTA Fast protein liquid chromatography (FPLC) ATP Adenosine triphosphate B. megaterium Bacillus megaterium BM3 The third P450 isolated from Bacillus megaterium BMR BM3 reductase CCS Collision cross-section CIU Collision-induced unfolding COSY Homonuclear correlation spectroscopy Cpd 0 Compound 0 Cpd I Compound I Cpd II Compound II CPR Cytochrome P450 reductase CYP Cytochrome P450 CW Continuous wave DEAE Diethylaminoethyl DM P450 BM3 double mutant (A82F/F87V) DMSO Dimethyl sulfoxide DNA Deoxyribonucleic acid DSC Differential scanning calorimetry E.coli Escherichia coli EDTA Ethylenediaminetetraacetic acid EPR Electron Paramagnetic Resonance ESI Electrospray ionisation FAD Flavin adenine dinucleotide FADH Reduced, semiquinone form of FAD FADH2 Reduced, hydroquinone form of FAD FDA The US Food and Drug Administration FDR Ferredoxin reductase Fdx Ferredoxin FMN Flavin mononucleotide FMNH Reduced, semiquinone form of FMN FMNH2 Reduced, hydroquinone form of FMN G Gauss g Gram GHz GigaHertz HA Hydroxyapatite HDX Hydrogen-deuterium exchange HDX-MS Hydrogen-deuterium exchange mass spectrometry HPLC High-performance liquid chromatography HS High-spin Hz Hertz IM Ion-mobility IMS Ion-mobility spectrometry IMS-MS Ion-mobility mass spectrometry 13 kcat Enzyme turnover rate constant kcat/Km Catalytic efficiency specificity constant Kd Dissociation constant KDa KiloDalton Km Michaelis constant KPi Potassium phosphate LCMS Liquid chromatography-mass spectrometry LCMS/MS Liquid chromatography-tandem mass spectrometry fragmentation LS Low-spin M Molar MgCl2 Magnesium chloride mL Millilitre mM Millimolar mmol Millimole MS Mass spectrometry MWCO Molecular Weight Cut-Off NaCl Sodium chloride NAD+ Nicotinamide adenine dinucleotide (oxidized form) NADH Reduced form of NAD NADP+ Nicotinamide adenine dinucleotide phosphate (oxidized form) NADPH Reduced form of NADP nESI Nano-electrospray ionisation Ni-IDA Nickel iminodiacetic acid nM Nanomolar NMR Nuclear magnetic resonance NOS Nitric oxide synthase NPG N-palmitoyl glycine OH Hydroxyl group PDB Protein data bank PEG Polyethylene glycol pKa Acid dissociation constant RMSD Root mean square deviation RMSF Root mean square fluctuation ROS Reactive oxygen species RPM Revolutions per minute RT Retention time Rz Reinheitszahl value SDS-PAGE Sodium dodecyl sulfate-polyacrylamide gel electrophoresis TB Terrific broth medium TOCSY Total correlation spectroscopy WT Wild-type UV-Vis Ultraviolet-Visible

Abstract

Title: High-value oxy-pharmaceuticals from P450 BM3 ‘gatekeeper’ mutations Author: Laura N Jeffreys Institution: The University of Manchester Faculty: Science and Engineering Degree: Doctor of Philosophy Submitted: 01/10/18

P450 BM3 is a natural fusion protein containing a heme catalytic domain fused to a CPR- like domain that binds two flavin cofactors (FAD and FMN). This allows this enzyme to have one of the highest catalytic rates of any P450. Such characteristics make P450 BM3 an attractive enzyme for mutagenesis to produce non-natural products. Unfortunately, no full-length structure revealing the natural domain interactions has been elucidated using X- ray crystallography or other means.

HDX-MS is a relatively new technique to explore the surface of a protein and any solvent accessible channels, such as active sites and allosteric binding sites. HDX-MS and other MS techniques were utilized to probe the surface of WT P450 BM3 to gain new insights into the structure of the enzyme including potential dimeric interfaces. The observation of many shielded areas has given insights into conformational changes during ligand binding and on the dimeric interface of the protein, aiding in piecing together the full-length structure of this enzyme.

WT BM3 binds specifically to fatty acids and typically produces hydroxylated products. Previously, two mutations were discovered that greatly affect the substrate and product profile of the double mutant (DM, A82F/F87V) enzyme. The substrate/ligand-binding profile of the DM enzyme was investigated using an FDA-approved library. From the library, 59% of the drugs caused significant shifts in the UV-Vis spectrum. These compounds have a variety of targets, masses and structures. Around 80 compounds were chosen due to their structural and binding properties for further analysis, including binding affinity determination, EPR, HPLC, LC-MS, LC-MS/MS, NMR and X-ray crystallography. From these analyzes, the binding of 18 inhibitors was identified, including a novel binding mode involving a substituted pyrimidine ring. Many substrates were also identified, which were found to be metabolized to known human metabolites during turnover.

Acknowledgements

My thanks go to my supervisor, Prof. Andrew Munro for his unwavering support and guidance throughout my Ph.D. I would also like to thank my industrial CASE co- supervisor Dr. Michael Voice at Cypex Ltd. and the BBSRC for their funding of a CASE PhD studentship [Grant number BB/L017253/1]. I would also like to thank Prof. Jon Waltho and Prof. David Leys for their help with data interpretation throughout the project, especially in the crucial final months. Also, many thanks go to Dr. Colin Levy, Dr. Katherine Hollywood and Dr. Matthew Cliff for all their help, expertise and running of experiments throughout the four years.

I have received much guidance and support from my lab group, especially by our lab technician Marina Golovanova who taught me the basics at the beginning of the Ph.D. and solved all the little problems throughout, without which I wouldn’t have been able to complete this thesis. Also, to our postdoctoral researchers Kirsty, Hazel, Jim, Kang-Lan, Harsh, Sarah and Richard who have all been there to talk through ideas and problems. Thanks also go to all the Ph.D. students within our group from the first year to final year; Shalini, George, Manca, Jude, Mark, Charlie, Dayana, Dom, Alessia, Emily, James and Irwin.

I’d also like to thank members of the MIB crew who have helped me outside the lab; Mike, Alex, Robin, Charlotte, Morgane, Jean-Marc and our adoptive members Jake, Anna, Jinesh, Tom (Stevie) and Emily. Without all of you the time in Manchester wouldn’t have been nearly as entertaining.

Finally, and perhaps most importantly, I would like to thank my fiancé and family who have supported me during the Ph.D. and with everything leading up to this point.

Declaration

Title: High-value oxy-pharmaceuticals from P450 BM3 ‘gatekeeper’ mutations Author: Laura N. Jeffreys

I declare that no portion of the work referred to in the thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning;

Signed:

Date: 01/10/2018

The following four notes on copyright and the ownership of intellectual property rights must be included as written below:

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

ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance Presentation of Theses Policy You are required to submit your thesis electronically Page 11 of 25 with licensing agreements which the University has from time to time. This page must form part of any such copies made.

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

iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=24420), in any relevant Thesis restriction declarations deposited in the University Library,

v. The University Library’s regulations (see http://www.library.manchester.ac.uk/about/regulations/) and in The University’s policy on Presentation of Theses.

Preface to the Journal Format Thesis

This thesis in presented in the University of Manchester journal style of PhD thesis. This allows the incorporation of data already published and/or sections that are in a format suitable for submission for publication in a peer-reviewed journal. The structure of each chapter follows that of the journal in which it is prepared for publications, meaning a separate methods chapter is not included as they are in the individual results chapters, and heading labels for each chapter may differ. The layout of the journal format thesis is as follows; an abstract, introduction, three results chapters (papers), and a summary, conclusions and future work chapter. Each chapter contains self-contained references. The general formatting of these papers has been kept consistent throughout this thesis, along with the page numbering. The journal format thesis was introduced in the University of Manchester to help students develop their skills in writing papers for peer-reviewed submission and to be overall more relevant to scientific research. Successful scientific publications are often collaborative and therefore as part of the journal format the contributions of each co-author are stated below.

Papers included as results chapters:

Chapter 2 Laura N. Jeffreys, Kamila J. Pacholarz, Linus O. Johannissen, Hazel M. Girvan, Perdita E. Barran, and Andrew W. Munro (2018). Characterization of the Structure and Interactions of the Active Form of P450 BM3 using Hybrid Mass Spectrometry Approaches. This paper is currently unpublished but it is hoped that it will be published in Structure.

Chapter 3 Laura N. Jeffreys, Hazel M. Girvan, Richard J. Milnes, Marina Golovanova, Kirsty J. McLean, Michael W. Voice, and Andrew W. Munro (2018). The promiscuous nature of P450 BM3 and its ability to bind pharmaceutical compounds using a novel library screen. This paper is currently unpublished but it is hoped that it will be published in Nature Communications.

Chapter 4 Laura N. Jeffreys, Harshwardhan Poddar, Marina Golovanova, Colin W. Levy, Hazel M. Girvan, Kirsty J. McLean, Michael W. Voice, David Leys, and Andrew W. Munro (2018). Novel insights into P450 BM3 interactions with FDA-approved antifungal azole drugs. This paper has been accepted by Scientific Reports and is awaiting publication.

Chapter 5 Laura N. Jeffreys, Harshwardhan Poddar, Marina Golovanova, Katherine Hollywood, Matthew J. Cliff, Colin W. Levy, Hazel M. Girvan, Kirsty J. McLean, Michael W. Voice, Jon P. Waltho, David Leys, and Andrew W. Munro (2018). Screening antidiabetic binding to P450 BM3 and the production of human metabolites. This paper is currently unpublished but it is hoped that it will be published in The Journal of the American Chemical Society.

Chapter 6 Laura N. Jeffreys, Kamila J. Pacholarz, Marina Golovanova, Katherine Hollywood, Matthew J. Cliff, Hazel M. Girvan, Kirsty J. McLean, Perdita E. Barran, Michael W. Voice, Jon P. Waltho, and Andrew W. Munro (2018). Binding of fibrates to P450 BM3 reveals novel changes to the P450 BM3 landscape. This paper is currently unpublished but it is hoped that it will be published in The Journal of Biological Chemistry.

Author contributions

As Ph.D. supervisor, Prof. Andrew W. Munro contributed to the manuscript preparation for all papers.

Chapter 2 Dr. Kamila Pacholarz ran all HDX-MS, native MS, and CIU experiments. Dr. Linus Johannissen calculated all CSS values from X-ray crystallography structures. Dr. Hazel Girvan expressed and purified all truncated reductase domains. Prof. Perdita Barran supervised all HDX-MS, native MS, and CIU experiments.

Chapter 3 Dr. Hazel Girvan collected all EPR data and assisted with binding titrations. Mr. Richard J. Milnes assisted with generating Python scripts for percentage HS determination. Mrs. Marina Golovanova assisted with AKTA protein purification for crystallography. Dr. Kirsty McLean aided with binding titrations and issues involving DMSO. Dr. Michael Voice contributed to direction and supervision of the project.

Chapter 4 Dr. Harshwardhan Poddar assisted with all X-ray crystallography data analysis for the azole bound crystals. Mrs. Marina Golovanova assisted with AKTA protein purification for crystallography. Dr. Colin Levy collected X-ray diffraction data from P450 BM3 crystals at the Diamond synchrotron. Dr. Hazel Girvan collected all EPR data. Dr. Kirsty McLean aided with binding titrations. Dr. Michael Voice contributed to direction and supervision of the project. Prof. David Leys contributed to the solving of ligand-bound P450 BM3 heme domain crystal structures and prepared the figures for publication.

Chapter 5 Dr. Harshwardhan Poddar assisted with all X-ray crystallography data analysis for the antidiabetic bound crystals. Mrs. Marina Golovanova assisted with AKTA protein purification for crystallography. 21

Dr. Katherine Hollywood assisted with LC-MS and LC-MS/MS experiments. Dr. Matthew Cliff assisted with NMR spectroscopy, product analysis and in the preparation of figures. Dr. Colin Levy collected X-ray diffraction data from P450 BM3 crystals at the Diamond synchrotron. Dr. Hazel Girvan collected all EPR data. Dr. Kirsty McLean aided with binding titrations. Dr. Michael Voice contributed to direction and supervision of the project. Prof. Jon Waltho assisted with all product determination by NMR. Prof. David Leys contributed to the solving of ligand-bound P450 BM3 heme domain crystal structures and prepared the figures for publication.

Chapter 6 Dr. Kamila Pacholarz ran all HDX-MS, native MS, and CIU experiments. Mrs. Marina Golovanova assisted with AKTA protein purification for crystallography. Dr. Katherine Hollywood assisted with LC-MS and LC-MS/MS experiments. Dr. Matthew Cliff assisted with NMR spectroscopy, product analysis and in the preparation of figures. Dr. Hazel Girvan collected all EPR data. Dr. Kirsty McLean aided with binding titrations. Prof. Perdita Barran supervised all HDX-MS, native MS, and CIU experiments. Dr. Michael Voice contributed to direction and supervision of the project. Prof. Jon Waltho assisted with all product determination by NMR.

Chapter 1: General Introduction

1.1. An Overview of Cytochromes P450

The Cytochrome P450 enzyme superfamily is one of the largest in nature. These enzymes are involved in xenobiotic metabolism in eukaryotes, prokaryotes and archaebacteria (Danielson, 2002). There is great catalytic variety between different P450 classes, but there are also substantial similarities, including conserved amino acid residue motifs, the common P450 fold and the cysteine thiolate-coordinated heme prosthetic group

(Graham and Peterson, 1999). The heme group facilitates the catalytic activity of each

P450 through oxygen activation, allowing for a variety of reactions, most commonly hydroxylation at a carbon atom as shown in the chemical equation below, where RH is a substrate and ROH indicates an oxidized (hydroxylated) product (Munro et al., 1996a).

Within this introduction, Cytochrome P450 evolution, structure, applications, and the catalytic cycle will be discussed, in particular for the prokaryotic protein P450 BM3.

− + 푅퐻 + 푂2 + 2푒 + 2퐻 → 푅푂퐻 + 퐻2푂

1.1.1. The Evolution and Nomenclature of Cytochromes P450

Cytochromes P450 are thought to have evolved as oxygen levels increased in the atmosphere over 3.5 billion years ago, before eubacteria and eukaryotes split in their evolutionary paths (Loomis, 1988, Whitehouse et al., 2012). This is evident by the presence of these enzymes in almost all eukaryotes and prokaryotes. It is estimated that by the year 2020 one million P450 enzymes will have been identified and classified (Nelson,

2018).

P450 enzymes are crucial for phase I metabolism in organisms in which oxidative and peroxidative reactions are needed, as shown in Figure 1 (Nelson, 2003). This is followed by phases II and III in which the site of oxygenation is targeted for further modification, such as by glucuronidation or through a variety of conjugation reactions catalyzed by other cellular enzymes (Nebert and Gonzalez, 1987). This allows molecules to be targeted throughout the cell for excretion, as in the case of environmental toxins.

Many P450 enzymes have evolved to bind specific molecules, leading to the formation of

P450 families with similar amino acid sequences. However, some gene families exhibit overlapping substrate specificities due to divergent evolution. For example, as animals were exposed to plant metabolites, enzymes from P450 families I-IV evolved and diverged to oxidize these molecules (Nebert and Gonzalez, 1987). Now P450 enzymes exist that can metabolize substrates from fatty acids (Miura and Fulco, 1975) to sterols (Mast et al.,

2017) to pharmaceutical drugs (Guengerich, 2003) and so on. One common feature for these enzymes is the axial cysteine residue that interacts with the heme prosthetic group, contributing to the P450 catalytic activity and suggesting an ancient common ancestor in the P450 superfamily (Nebert and Gonzalez, 1985).

A potential candidate for a P450 common ancestor is CYP51 which is involved in sterol metabolism (specifically demethylation) in animals, plants, and fungi. This P450 enzyme is currently the only clan of P450 enzymes found in plants and animals, due to the different environmental pressures placed upon the species (Nelson, 2018). Discovering ancestral proteins is of importance to some researchers, due to the theory that these ancestral proteins will have greater stability and substrate promiscuity as they evolved in much harsher conditions before evolving substrate specificity (Gumulya and Gillam,

2017).

Figure 1: An overview of xenobiotic metabolism Many enzymes are important in xenobiotic metabolism and are able to modify the relevant compound for excretion. Compounds enter the cell through transporters (blue sphere) and are modified by Phase I proteins such as cytochromes P450. If further modification is required Phase II and III proteins are utilized (Jakoby and Ziegler, 1990). This figure was produced using ChemBioDraw (PerkinElmer, Waltham, MA).

Due to the sheer number of P450 enzymes, a classification system was derived in which all genes bear the identifier “CYP”, followed by a number referring to the particular

P450 family, a letter referring to the subfamily and another number referring to the isoform, e.g. CYP102A1 which is the gene encoding P450 BM3. The P450s are assigned by sequence similarity, with members of the same family sharing at least 40% amino acid sequence identity, and members of the same subfamily sharing at least 55% sequence identity (Bak et al., 2011). The exception to these rules occurs in the naming of plant P450 enzymes, where gene duplication and shuffling leads to an alternative nomenclature

(Nelson and Werck-Reichhart, 2011). Certain plant species have many P450 enzymes, such as Arabidopsis thaliana which has 244 cytochrome P450 genes within its genome

(Bak et al., 2011). Of interest, all of the plant P450 proteins currently identified are membrane-bound whilst most bacterial P450 enzymes are soluble, highlighting the 25 differences between the Kingdoms (Bak et al., 2011). Families can be grouped in clans due to proximity in their phylogenetic trees, suggesting shared structures and/or roles. Some clans only contain one P450 family, such as CYP51 and CYP74 (Nelson, 2018). All naming of cytochrome P450 genes to date has been undertaken by Prof. David Nelson to ensure consistency.

P450 proteins can also be divided into classes depending on the electron donation systems utilized, shown in Figure 2. Class I and II P450 enzymes require redox partner enzymes for electron donation. Class I proteins utilize a flavin adenine dinucleotide

(FAD)-binding ferredoxin reductase (FdR) and an iron-sulfur cluster-binding ferredoxin

(Fdx) partners (in prokaryotes), with electron transfer from the reduced cofactors nicotinamide adenine dinucleotide phosphate (NADPH) or nicotinamide adenine dinucleotide (NADH). In eukaryotes, the NADPH-dependent adrenodoxin reductase (AdR) and ferredoxin adrenodoxin (binding a 2Fe-2S cluster) reduce mitochondrial P450 enzymes. The Adx/AdR systems are membrane-bound while the prokaryotic Fdx/FdR systems are soluble enzymes that may utilize NADPH or NADH for electron delivery to the P450. One of the most widely researched P450 enzymes belongs to this class: P450cam

(CYP101A1) (Katagiri et al., 1968). Class II P450 proteins utilize cytochrome P450 reductases (CPRs) which are diflavin reductases with a FAD-binding domain and a flavin mononucleotide (FMN)-binding domain that use NADPH as their electron source. All enzymes in class II systems are membrane-bound, such as the most pharmaceutically important human P450 CYP3A4 (Paine et al., 2005). Class III P450s are soluble, catalytically self-sufficient fusion proteins containing a CPR-like diflavin domain linked to a P450 catalytic domain through a short linker region (Munro et al., 1999). These proteins use NADPH as a source of electrons, allowing two electrons to be shuttled from a CPR module to the heme domain for catalysis (discussed more in section 1.2.2.).

Figure 2: Different P450 redox class systems Four different cytochromes P450 classes are shown with FAD-binding domains/proteins in green, FMN-binding domains/proteins in blue, proteins/domains containing FeS clusters in yellow and the cytochromes P450 enzymes or their heme catalytic domain in red. The cofactors used to donate electrons are also shown. Class I P450s can be soluble or membrane-bound with electrons shuttling between a flavin adenine dinucleotide (FAD)- containing reductase, a ferredoxin (Fdx)/adrenodoxin (Adx) containing an iron-sulfur cluster (FeS) and a cytochrome P450. Class II P450s contains a membrane-bound diflavin reductase with a membrane-bound cytochrome P450, usually within the endoplasmic reticulum. These systems shuttle electrons from the FAD-binding domain of the reductase to the FMN-binding domain, and then to the P450 enzyme. Class III and IV P450s are catalytically self-sufficient, soluble proteins with multiple domains joined by linker regions. Class III contains FAD- and FMN-binding domains. Class IV contains an FMN- and 2Fe-2S binding phthalate dioxygenase-like domain. This figure was produced using ChemBioDraw.

The first example of the class III P450s was the soluble bacterial protein P450

BM3, which has the highest catalytic activity of any P450 monooxygenase due to its domain configuration and high rate of inter-cofactor electron transfer (Noble et al., 1999). 27

Class IV is the newest class discovered to date. This class describes soluble proteins, such as P450RhF from Rhodococcus sp. NCIMB 9784, in which the P450 catalytic domain is fused to a phthalate dioxygenase reductase-like partner containing an FMN-binding domain and a [2Fe-2S] ferredoxin-like component (Roberts et al., 2002). These proteins utilize NADH or NADPH as an electron donor (Warman et al., 2012). A number of other different types of P450 redox systems have also been recognized in recent years.

1.1.2. The History of Cytochrome P450 Research

P450 proteins were first reported in the literature 60 years ago when two scientists independently reported the presence of a pigmented peak at 450 nm when carbon monoxide was bubbled through reduced microsomal liver fractions (Garfinkel, 1958,

Klingenberg, 1958), such as the shift observed in Figure 3. These experiments were to investigate the properties of other heme-containing proteins, such as cytochrome b5.

Garfinkel also noted the presence of an unknown electron acceptor in other cytochrome b5 experiments (Estabrook, 2003). The phrase cytochrome P450 was coined by Omura and

Sato in two papers characterizing these proteins (Omura and Sato, 1964a, Omura and Sato,

1964b). Considerable research was being conducted on liver enzymes at the time, and after review it was apparent that the oxidases observed in many experiments (including the

Garfinkel 1958 and Klingenberg 1958 papers) were cytochromes P450 and these enzymes were discovered to be vital in mammals for the hydroxylation of steroids (Estabrook et al.,

1963) and drug oxidation (Cooper et al., 1965). There are over 45,000 papers with the term

‘P450’ in the title published to date.

The ability of the P450s to form a 450 nm peak led to the discovery that these proteins contained a heme b prosthetic group (Omura and Sato, 1964a, Omura and Sato,

1964b). The formation of the 450 nm peak also led to the finding that the heme prosthetic

28 group is ligated axially by a cysteine thiolate, which has an electron-donating nature critical for P450-mediated catalysis (Figure 5) (Dawson, 1988). In contrast, a 420 nm peak can be produced under the same conditions if the heme prosthetic group is instead ligated to a cysteine thiol, which could result from improper P450 folding, or be due to the protonation of the native (thiolate) form of the cysteine (Perera et al., 2003, Ogura et al.,

2004, Driscoll et al., 2011). Originally, studies on P450s were conducted on homogenized mammalian livers. Such samples contain many P450 proteins, although this was unknown at the time. For example, humans have 57 P450 enzymes from 15 families, and though not all of these are localized within the liver, most exist within this organ as it is the main site of xenobiotic metabolism mediated by P450s (Nebert et al., 2013).

UV-Vis spectroscopy has been a crucial technique for P450 research, and provides useful information on the heme environment and e.g. on ligand (substrate and inhibitor) binding. A number of features are observed in a typical P450 spectrum; the alpha and beta bands (~565 and ~540 nm respectively in the oxidised, substrate-free form), the gamma band also known as the Soret peak (at ~418 nm in the oxidised, substrate-free form) and the delta band (~360-365 nm in the oxidised, substrate-free form). In addition, there is a peak (at ~280 nm) that originates mainly from the absorbance of the aromatic side chains of tryptophan, tyrosine and phenylalanine residues within the protein. The

418/280 nm ratio can be used to give a “purity” or Reinheitszahl (Rz) value for P450s and other heme-containing proteins (Theorell and Maehly, 1950). These peaks are displayed for ligand-free P450 BM3 in Figure 3. As P450 BM3 is a fusion protein, signals from the oxidized flavins within the CPR domain are also visible on either side of the Soret peak as shoulders at ~390 nm and ~480 nm.

The substrate-free Soret peak absorbance maximum can vary significantly

Figure 3: Binding of carbon monoxide to P450 BM3 elicits a Soret shift to 450 nm By bubbling carbon monoxide into a cuvette containing reduced, full-length WT BM3, a shift from the ligand-free 418 nm peak (black line) to the 450 nm ferrous-CO complex (red line) can be observed. The ligand-free, oxidized BM3 (black line) exhibits the (Soret)and peaks as labelled. The band is obscured in the ferrous-CO state by the absorbance (from 400 nm downwards) from the sodium dithionite reductant, and the dithionite also fully reduces the reductase FAD and FMN cofactors and bleaches their absorbance signal in the UV-visible spectrum. This figure was produced using Origin Lab 9.1 (OriginLab, Massachusetts, USA) in conjunction with ChemBioDraw.

between P450 enzymes. For example, the peroxygenase P450 enzyme OleTJE with fatty acid hydroxylase/decarboxylase activity has a Soret maximum at 423 nm for the ligand- free protein (Belcher et al., 2014). In the ligand-free form of most P450 proteins, a water molecule is axially ligated to the distal face of the heme iron, with a cysteine thiolate as the proximal ligand. Upon substrate binding, the water molecule is typically replaced by the substrate, causing a wavelength shift for the Soret peak (typically from ~418 nm to ~394 nm) with the delta band at ~360-365 nm. The substrate-induced Soret blue shift is termed a 30 type I spectral shift. The binding of inhibitors elicits different types of shift in the P450 spectrum, termed type II shifts. Here the Soret peak red shifts from ~418 nm to longer wavelengths, e.g. to ~420-427 nm depending on the inhibitor (Locuson et al., 2007), but also to much longer wavelengths for certain ligands (e.g. to ~435 nm for nitric oxide). The binding of a substrate also typically causes an electron reconfiguration of the iron atom from a low-spin to a high-spin state, which is usually critical for the catalytic activity of the protein through increasing the heme iron potential. This is discussed further in section

1.1.3.

Various other techniques have been used to analyze P450 function, structure, molecular interactions, ligand binding and other properties. Through these approaches, many useful P450 enzymes have been discovered which produce compounds of interest e.g. CYP82Y2 in the production of opiates in poppies (Winzer et al., 2015). Other P450s can provide a good template for mutagenesis in order to design systems of interest e.g. using P450 BM3 mutants to catalyze a key reaction in the production of the anti-malarial drug artemisinin (Dietrich et al., 2009), or as targets for treatment of disease e.g. CYP46A1 as a potential target for Alzheimer's (Djelti et al., 2015).

Some P450 enzymes have been studied intensively due to their ease of expression and purification. The widely studied enzyme P450 BM3 is discussed in more detail in section 1.2. Another example is P450cam (CYP101A1), which is a soluble, type I enzyme from the CYP101 family isolated from Pseudomonas putida. This P450 enzyme was the first to be crystallized in both ligand-free and ligand-bound states, and is still often a first choice protein for mechanistic, structural and other P450 research (Poulos et al., 1985,

Poulos et al., 1987). For example, P450cam was used to determine the importance of the cysteine residue in P450 catalysis in several experiments, including resonance Raman

(Champion et al., 1982), with observations of the effects of mutagenesis on the formation 31 of the inactive 420 nm complex in the CO-bound form (Yoshioka et al., 2001). Binding of various inhibitors and of the substrate camphor have been extensively analyzed, alongside the conformational changes that such ligands elicit (Shiro et al., 1989, Diprimo et al.,

1993). The structure of this protein has also been successfully elucidated in complex with its putidaredoxin electron donating partner (Hiruma et al., 2013, Tripathi et al., 2013).

Considerable research into the P450 catalytic cycle has been conducted using P450cam, which is discussed further in section 1.1.3. (Schlichting et al., 2000, Makris et al., 2002,

Denisov et al., 2001b). Another enzyme extensively studied for the understanding of the catalytic cycle is the acidothermophile P450 CYP119 isolated from Sulfolobus solfataricus. This enzyme has been used in the crucial characterization of the reactive species Compound 0 (FeIII-OOH), Compound I (FeIV=O) and Compound II (FeIV-OH), which is discussed further in section 1.1.3. (Denisov et al., 2001a, Rittle and Green, 2010).

Of note, the human P450 enzyme CYP3A4 is often used for drug metabolism research as ~50% of all drugs ingested are metabolized by this protein (Ince et al., 2013).

However, this protein is difficult to express and purify in comparison to soluble bacterial proteins such as P450cam and P450 BM3, due to its membrane-bound nature and the need for other membrane-bound redox partner enzymes.

1.1.3. The P450 Catalytic Cycle

As mentioned previously, the heme prosthetic group acts as the catalytic centre of a P450 protein. As shown in Figure 5(i) the heme iron exists in a hexacoordinated state in the ligand-free conformation. In this form the heme iron atom has four equatorial interactions from pyrrole nitrogen atoms in the tetrapyrrole ring. It also makes one bond to a proximally-ligated cysteine thiolate and one to a distally-ligated water molecule in the resting state (Luthra et al., 2011). In some mammalian P450 enzyme structures this water

32 molecule is not observed in X-ray crystallography studies, due to weak interactions, poor resolution or mechanisms not yet understood (Guengerich and Johnson, 1997). The cysteine thiolate ligand remains in place, contributing to electron donation throughout the catalytic cycle (Cramer et al., 1978, Hahn et al., 1982).

Figure 4: Electron rearrangement in the d-orbitals of the heme iron atom upon substrate binding Displacement of the distal water molecule by substrate binding causes no change to the oxidation state of the iron atom, but does result in an electron rearrangement of the d- orbitals due to a decrease in crystal field splitting energy (E). This causes the change from low spin (LS) to high spin (HS) as all electrons in the d-orbitals become unpaired. This image was drawn using ChemBioDraw.

At this stage, the iron atom exists in the ferric state and so has the electron configuration 1s2 2s2 2p6 3s2 3p6 3d5. In this octahedral complex there is a difference in energy between the d-orbitals as the t2g orbitals are in the correct symmetry for pi () interactions but not sigma () interactions, causing these orbitals to become lower in energy than the eg orbitals, which have the correct symmetry for sigma () interactions but not pi () interactions (Figure 4). In the ligand-free (‘resting’ state) the energy difference between the d-orbitals is quite large (E1) causing the electrons to pair in the t2g orbitals

(Figure 4) (Luthra et al., 2011, Gibson and Skett, 2001). Upon substrate binding, the distally-ligated water molecule is displaced by a hydrophobic substrate (RH), allowing the 33 heme iron atom to become pentacoordinated (Figure 5(ii)). This causes the energy difference between the d-orbitals to decrease (E2) and the electrons to fill the d-orbitals in an unpaired configuration, as it is more energetically favourable, resulting in 3 electrons in the t2g orbitals and 2 electrons in the eg orbitals (Figure 4) (Gibson and Skett, 2001). The spin state of the iron changes from low spin (LS) = ½ to high spin (HS) = 5/2, resulting in changes to the UV-Vis spectrum. The nomenclature for the ligand-free protein is referred to as low spin at ~418 nm, whilst a substrate-bound form is referred to as high spin with absorbance at ~394 nm, corresponding to changes in the spin state.

The binding of the substrate begins the catalytic cycle through displacing the distal water and generating a pentacoordinate form with an increased (more positive) potential of the ferric heme iron which facilitates the donation of electrons from partner enzymes using NAD(P)H reductant, although this process can also occur without substrate binding in various cases (Guengerich and Johnson, 1997). In comparison, inhibitors ligate to the iron atom directly on the distal side, often through a nitrogen atom and resulting in a hexacoordinated low spin species. This blocks substrate access and prevents the redox potential from increasing, therefore inhibiting heme iron reduction and subsequent molecular oxygen binding. The degree of shift exhibited on inhibitor binding within the

UV-Vis spectrum (typically ~420-427 nm, although more substantial inhibitor-induced shifts are also observed) is thought to be due to the nature of the electronic structure of the relevant coordinating group (Locuson et al., 2007).

The first electron donation from a redox partner causes the formation of the ferrous state (Figure 5(iii)). This reduction is needed for oxygen binding in a similar manner to myoglobin and hemoglobin (Berg et al., 2007). The binding of molecular oxygen (O2) causes the formation of the ferric-superoxy intermediate which has a characteristic ‘bent’ shape with respect to the oxygen atom orientation, and which is in 34 equilibrium with a ferrous-oxy form (not shown) (Figure 5(iv)) (Macdonald et al., 1999).

The protein is then reduced again by a redox partner, forming the ferric-peroxy intermediate (Figure 5(v)). Protonation forms the ferric-hydroperoxy intermediate from the ferric-peroxy state (Figure 5(vi)), also known as Compound 0 (Cpd 0). This intermediate was captured by cryogenic radiolysis using P450cam (Denisov et al., 2001b) and horseradish peroxidase (HRP) (Denisov et al., 2002). The Sligar lab used X-ray crystallography in attempts to trap Cpd 0. However, only partial occupancy was observed for the heme prosthetic group intermediate (Schlichting et al., 2000). Cpd 0 is observed in many heme-containing proteins and other attempts at trapping Cpd 0 using X-ray crystallography have been successful (Kuhnel et al., 2007). This intermediate has been characterized for P450cam using other techniques, such as resonance Raman spectroscopy

(Mak et al., 2007).

A second protonation event causes the scission of the oxygen-oxygen bond in Cpd

0, generating a water molecule (Figure 5(vii)). The intermediate produced during this step is a reactive ferryl-oxo porphyrin radical cation species, also known as Compound I (Cpd

I). Cpd I is described as a ‘chameleon species’ as the key features of its geometry are shaped by the polarity of the active site, the residues interacting with the heme prosthetic group and its cysteine thiolate ligand (Ogliaro et al., 2000). This species is capable of interacting with the substrate by abstraction of a hydrogen atom from a C-H bond, forming the substrate radical and a ferryl-hydroxo intermediate, also known as Compound II (Cpd

II) (Figure 5(viii)). The substrate radical and the ferryl-hydroxo species then recombine by

‘radical rebound’ to form the hydroxylated product (Figure 5 (viii-ix)) (Groves and

McClusky, 1978). Cpd I and II were first characterized using CYP119. Cpd I was confirmed by observing spectroscopic and kinetic changes using Mössbauer and EPR spectroscopy (Rittle and Green, 2010). The X-ray absorption spectroscopy (XAS) technique was also used to observe the bond distances between the iron and oxygen atoms 35 to confirm the existence of Cpd II (Newcomb et al., 2008). Work by Makris’ group on the peroxygenase P450 OleTJE (using mainly transient kinetic approaches) also resulted in the characterization of both the Cpd I and Cpd II species, which were reported to catalyze fatty acid hydroxylation and decarboxylation reactions, respectively (Grant et al., 2016). Finally, the oxidized (hydroxylated) product is released and the cycle is ‘reset’ as a water molecule binds to the heme iron once again (Figure 5(i)).

Other catalytic pathways are shown in Figure 5. The oxidase shunt involves the reduction of Cpd I and leads to the release of a water molecule and to the reformation of the pentacoordinated, substrate-bound species (FeIII-RH). Similarly, the FeIII-RH species is reformed in the autoxidation shunt as superoxide is released from the ferric-superoxy species. In both of these shunts, electrons are consumed yet no oxidized product is formed.

This is also known as uncoupling (Luthra et al., 2011). In contrast, the peroxide shunt is a pathway that can successfully catalyze oxidation, or alternative reactions such as decarboxylation, by utilizing alternative oxygen sources such as peroxide ions and peracids. This shunt can occur in both directions between ferric, substrate-bound protein

(FeIII-RH) (intermediate ii) and Cpd 0 (vi), through reaction of the FeIII-RH species with peroxide (H2O2) in the forward direction (forming Cpd 0) and through loss of H2O2 in the reverse direction. For example, decarboxylation reactions catalyzed by the peroxygenase

P450 OleTJE utilize the peroxide shunt Figure 5(iii-v) (Grant et al., 2016, Hsieh et al.,

2017, Munro et al., 2018). This reaction does not require electron donating partners as all steps requiring reduction are ‘bypassed’ (Matsunaga et al., 1996).

Figure 5: An overview of the cytochrome P450 catalytic cycle The tetrapyrrole ring is shown with the heme iron in red and the axial cysteine (thiolate) in yellow. When a substrate (RH, pink) displaces the distal water molecule (blue), the catalytic cycle begins, leading to the scission of molecular oxygen (green), the formation of various ferrous, ferric and ferryl intermediates (ii-ix), and finally to the dissociation of the oxidized (hydroxylated) substrate. A more detailed explanation of the catalytic cycle mechanism can be found above. Pathways for the breakdown of iron-oxo intermediate species are shown in grey as (i) the autoxidation shunt (ferric superoxy to ferric with release of superoxide to regenerate the ferric, substrate-bound species, FeIII-RH), (ii) the reversible peroxide shunt (ferric to ferric hydroperoxy and vice versa using the peroxide shunt mechanism or through the collapse of the ferric-hydroperoxy species to regenerate FeIII-RH), and (iii) the oxidase shunt (in which water is displaced from compound I to regenerate FeIII-RH). This image was adapted from (Luthra et al., 2011) and was drawn using ChemBioDraw.

1.1.4. The Structure of P450 Enzymes

P450 enzymes contain a heme prosthetic group buried deep within the protein and accessible by substrate-access channels. This heme molecule is a heme b prosthetic group

(Figure 6) that interacts with the axial cysteine residue through the iron atom and to other residues through its propionate groups (Omura and Sato, 1964a, Poulos and Howard,

1987).

Figure 6: The structures of the three most common heme prosthetic groups found in hemoproteins Heme prosthetic groups are found in many proteins. The iron centre allows for catalytic activity mediated by electron transfer between ferric and ferrous states (discussed in section 1.1.3) as well as through e.g. reactions with oxygen. Within the P450 enzyme class heme b is utilized. The tetrapyrrole rings are similar in hemes b and c, with esterification of the two vinyls to two cysteine residues in heme c to provide covalent heme attachment to the protein. In heme a, the addition of an isoprenoid chain is seen, with the oxidation of a methyl group to a formyl group to produce a hydroxyethylfarnesyl group in heme a compared to heme b. The differences in the heme structures allow for different interactions with the active site of the protein. This image was drawn using ChemBioDraw.

The amino acid sequences of P450 enzymes can vary greatly between different

P450 families, yet these proteins exhibit remarkable structural similarity for such a large superfamily. There are three widely conserved sequences in most P450 enzymes (Werck-

Reichhart and Feyereisen, 2000). Within certain organisms the number of conserved sequences increases. For example, there are 5 conserved sequences in insect P450 enzymes

(Yu et al., 2015). The first widely conserved sequence is the WXXXR motif in the C-helix.

This motif is important for the interaction with the heme prosthetic group as it coordinates one of the heme propionate groups through the Arg residue (El-Garj et al., 2016). The Trp residue of this motif varies between P450 enzymes. For example, in prokaryotes the Trp residue is often replaced by a His residue. In addition, Gln residues have also been observed at this position (Nelson, 1998). Interestingly, some enzymes have substitutions of the important Arg residue. For example, CYP307 has the motif WXXXQ (Feyereisen,

2012). The second widely conserved sequence is EXXR found in the K-helix. This motif is also thought to interact with the heme prosthetic group as well as forming crucial salt bridges, hence its highly conserved nature (Hasemann et al., 1995, Rupasinghe et al.,

2006). The CYP157C family does not contain the EXXR motif. Instead, these proteins contain a QXXW sequence. Mutating these residues to a variety of residues, including the

EXXR sequence, resulted in a P420 form of the protein in which the cysteine thiolate is converted to a thiol (Rupasinghe et al., 2006). The final motif is the FXXGXXXCXG sequence containing the crucial cysteine thiolate residue that ligates to the P450 heme iron.

As more P450 enzymes have been discovered, many exceptions to this motif have been discovered. These include proteins that do not contain the cysteine residue, such as

CYP408 (discussed further in section 1.1.5). Of note, a highly conserved amino acid in the

I-helix is also a hallmark of P450 catalysis. This is usually a Thr residue that is adjacent to an acidic residue (e.g. Asp251 and Thr252 in cytochrome P450cam, CYP101A1). This motif is considered to catalyze the transfer of protons to iron-oxo species in the P450 catalytic cycle. However, many exceptions exist, such as the erythromycin producing P450eryF which catalyzes the hydroxylation of 6-deoxyerythronolide B (6-DEB) in a pathway to the production of erythromycin. P450eryF contains an Asn residue at this position, which forces the substrate to adopt a particular conformation within the active site and leads to regio- and stereoselective oxidation of 6-DEB to form erythronolide B (Slessor et al., 2011).

In the P450 tertiary structure, similarities are also apparent in the triangular prism shape exhibited by many P450 enzymes - a feature called the “P450 fold”. The P450 fold is a strongly conserved core structural arrangement among P450 proteins. A study compared five P450 structures and found this P450 core superimposed within a 2-

Angstroms root mean square deviation (RMSD) for each structure, confirming the high core structural similarity necessary for P450 catalysis. One of the P450 enzymes overlaid in this study (P450 BM3) is shown in Figure 7, highlighting important regions of the protein and the P450 protein fold (Graham and Peterson, 1999). The P450 core consists of the alpha helices D, E, I and L within a four-helix bundle. Other components include several beta sheets which contribute to a hydrophobic substrate-access channel. In certain thermostable P450 enzymes, the hairpin loop of the beta sheets  and segments of the

G helix are absent, suggesting aspects of this P450 core are not necessary for function in various P450s from thermophiles (Harris et al., 2018).

The structures of P450 enzymes have been analyzed using a variety of techniques.

In early studies, UV-Vis spectroscopy was utilized to identify P450 heme absorbance bands and to gain information on the heme environment (Omura and Sato, 1964a). The most accurate descriptions of P450 structure come from X-ray crystallography. Between

1985 and 2014, 449 P450 enzyme structures were deposited in the Protein Data Bank

(PDB). Many of these come from the same P450 enzymes with different ligands bound

(Poulos and Johnson, 2015, Poulos et al., 1985). Due to the inherent flexibility of P450 proteins, many have not been successfully crystallized without the use of cross-linkers, truncations to their sequences or through the introduction of mutations. However, many

P450 enzymes do not have X-ray crystal structures. An example of this is the exceptionally large multidomain protein P450 BM3, which is a natural fusion protein formed of a 55 kDa heme (P450) domain and a 65 kDa reductase domain (containing FAD and FMN

Figure 7: The core of a P450 protein and the P450 fold Using the P450 BM3 structure 4KEY (Butler et al., 2013) several key components of the P450 fold are exhibited. The I-helix (yellow) and the F/G-helices (red) are highly flexible components of the P450 fold. Other areas of the P450 fold, such as the B’-helix (blue) and the C-terminus (magenta), surround the heme prosthetic group (green). A selection of alpha helices have also been labelled. This figure was made using PyMOL (The PyMOL Molecular Graphics System, Version 1.7.4.5 Schrödinger, LLC).

cofactors), which also dimerizes (Girvan et al., 2011, Neeli et al., 2005). The structure of this protein has not been solved for the intact P450 BM3, or for the reductase domaindespite research being conducted on this enzyme for over 40 years (discussed further in section 1.2.). However, the crystal structure of the individual heme (P450) domain has been determined in various ligand-free and ligand-bound forms, and the crystal structures of the component FAD/NADPH-binding (ferredoxin reductase-like) and FMN-

41 binding (flavodoxin-like) forms have also been solved (Sevrioukova et al., 1999, Joyce et al., 2012).

Other spectroscopic techniques used for P450 characterization include nuclear magnetic resonance (NMR), electron paramagnetic resonance (EPR), circular dichroism

(CD), resonance Raman and (more recently) cryo-electron microscopy and hydrogen- deuterium exchange mass spectrometry (HDX-MS). Although NMR studies have been very successful for some P450s, one of the key limitations of this technique is the ability to label the protein effectively, especially for large proteins such as P450 enzymes (Mouro et al., 1999). Another limitation is that the inherent paramagnetism of the cytochromes P450 contribute additional isotropic shifts and cause faster relaxation times, which interferes with the resonance of backbone amides within 10 Å of the paramagnetic centre (Cheng and

Markley, 1995). New advances in this area have allowed real-time solution-state NMR, allowing interactions between partner enzymes to be observed (Rennella et al., 2012,

Estrada et al., 2013). Due to the paramagnetic nature of the ferric heme iron, cytochrome

P450s can be analyzed by EPR to give information on the heme prosthetic group environment. Historically, EPR has been used to great success for P450 proteins. In particular, EPR has the ability to determine changes in the P450 heme environment due to ligand interactions within the active site (Girvan et al., 2007). Proteins with iron centres require very low temperatures (e.g. ~10 K for P450s) for the effective use of this technique due to the fast relaxation time of iron which affects ligand binding. The HS (substrate- bound) species is favoured at higher temperatures for EPR and many other techniques, such as UV-Vis spectrophotometry. However, at higher temperatures the signals observed in the EPR spectrum broaden and the signal-to-noise ratio increases (Luthra et al., 2011,

Lipscomb, 1980). EPR spectroscopy also has a number of related techniques, such as

DEER (double electron-electron resonance), which can give additional information such as determining the distance between radical species in a protein (Liou et al., 2014). Circular 42 dichroism has been used frequently in the study of P450s to provide information on the secondary and tertiary structure of P450s and other proteins (Cramer et al., 1978,

Andersson et al., 1997). Resonance Raman spectroscopy has also been used widely to characterize P450 heme iron spin-state, redox state, interactions of ligands with heme iron and to characterize transient intermediates formed during the catalytic cycle (Denisov et al., 2008); as well as providing structural insights into the P450 active site (Hudecek et al.,

2000, Anzenbacher and Hudecek, 2001). Cryo-electron microscopy has also been used to determine structures of proteins that have proved resistant to crystallogenesis. These include P450 BM3. However, the structural resolution achieved using cryo-EM methods is not comparable to that from X-ray crystallography, and thus an accurate structural analysis has not yet been achieved for P450 BM3 using cryo-EM (Zhang et al., 2018). HDX-MS and other MS methods can give information on the protein in real-time, as is also the case for solution NMR. In HDX-MS, the protein is diluted into a deuterated buffer for varying times, quenched with ice-cold acid and then passed through a protease column to break the protein into many different lengths of amino acid sequences. These fragments are then passed through a series of mass spectrophotometers to separate the fragments based on size, charge, and shape. This information can be used to determine the surface properties of a protein and how they change upon ligand binding or through interactions with partner enzyme binding sites or other domains within the same protein (Anderson et al., 2016).

1.1.5. Unusual P450 Proteins

There are many P450 enzymes that stand out due to their structure, redox partner systems and ligand/substrate profile. One such protein is the decarboxylating enzyme

OleTJE from the CYP152 family, which has a ligand-free Soret peak absorption at 423 nm

(Belcher et al., 2014). As mentioned in section 1.1.4, this protein makes use of the

43 peroxide shunt to for catalytic activity (Matthews et al., 2017). OleTJE is currently being researched due to its ability to make short chain alkenes for the production of biofuels.

Another interesting P450 enzyme is the protein CYP170A1 from Streptomyces coelicolor A3(2) which has a ferrous-CO heme peak at 440 nm or 447 nm, depending on the laboratory reporting the information (Lamb et al., 2002, Zhao et al., 2008). CYP170A1 catalyzes the two-step oxidation of epi-isozizaene through hydroxylation to produce either

4(S)- or 4(R)-albaflavenol intermediates in the first step, followed by a second oxidative reaction in which both of these intermediates are converted to albaflavenol. The epi- isozizaene substrate is produced by a sesquiterpene synthase which catalyzes cyclization of farnesyl diphosphate to the novel tricyclic hydrocarbon, epi-isozizaene. The sco5222 gene encodes the synthase and is co-transcribed with the CYP170A1 gene. Both the (i) sesquiterpene synthase and (ii) CYP170A1 have additional binding sites for (i) the production of isomers of farnesene from farnesyl diphosphate, and (ii) for a novel terpene synthase active site that “moonlights” in the P450 structure to enable albaflavenone antibiotic synthesis in the soil bacterium Streptomyces coelicolor (Zhao et al., 2009).

CYP55A1 (P450nor) is an enzyme from the fungus Fusarium oxysporum that does not require redox partners (fused or separate) for catalysis. Instead, this soluble protein is reduced by NADH which binds in the active site of the P450. Electrons are passed to two molecules of nitrogen monoxide (nitric oxide or NO). The NADH reduces a

NO molecule bound to ferric heme iron to generate an intermediate with a Soret absorbance at 444 nm and thought to be a ferric-hydroxylamine radical complex. The intermediate then reacts with the second molecule of NO to produce N2O (nitrous oxide) in the final stage of the respiratory process in the fungus (Nakahara et al., 1993). The P450nor enzyme has an NADH binding site that is located beneath a beta-helix and allows for this catalytic activity (Oshima et al., 2004). 44

CYP154A1 is another example of a P450 enzyme that does not require reductase proteins or other interacting domains for catalysis. Remarkably, this P450 enzyme is reported to catalyze a cyclisation reaction without either oxidation or reduction occurring

(Cheng et al., 2010). The enzyme catalyzes the coupling of the C5 carbonyl to the C11-

C12 double bond of a dipentaenone substrate to form a product with an oxetane ring. Such non-redox, molecular rearrangement reactions are quite rare in the P450s. This protein exhibits a 180-degree rotation of the heme prosthetic group (relative to the heme binding mode in most other P450s) that is believed to be essential for this reaction (Podust et al.,

2004).

The Mycobacterium tuberculosis P450 enzyme CYP121A1 also displays mixed conformations of the heme prosthetic group, as revealed in X-ray crystal structures of the enzyme. The heme conformations are observed in a ratio of ~70%/30% with the former being the “typical” conformation and the latter the heme rotated by 180° in the active site

(Leys et al., 2003). This phenomenon also occurs in other heme b-containing proteins, such as the small cytochrome b5 proteins (which are redox partners for P450s) in which the heme prosthetic group can alter its orientation mode in response to external factors such as temperature (Mortuza and Whitford, 1997). In P450 enzymes the heme prosthetic group is deeply buried and so the orientation of the heme group is defined during folding/heme insertion and cannot be changed during the lifetime of the protein (Leys et al., 2003).

The P450 XplA has a novel structure in which the P450 heme catalytic domain is fused to a flavodoxin partner. This protein was first isolated from the bacterium

Rhodococcus rhodochrous and is able to break down explosive molecules, most notably the carcinogenic explosive RDX (1,3,5-Trinitro-1,3,5-triazinane) in the soil (Sabbadin et al., 2009). Work has been done to express XplA alongside its flavodoxin reductase (XplB) 45 in plants in order to produce organisms that can use RDX as their sole source of nitrogen, allowing for a marker of explosive presence and a method to remove explosives from the soil (Jackson et al., 2007). Interestingly, it was discovered that the t2g orbitals of XplA are closer in energy to the porphyrin orbitals than are those in other P450 enzymes, such as

P450 BM3, by using near-infrared magnetic circular dichroism (NIR MCD) (Bui et al.,

2012).

Members of the CYP408 family lack the highly conserved proximal cysteine residue and the conserved threonine in the I-helix, yet have high sequence similarity and so are classed as P450 enzymes (Lamb and Waterman, 2013). These enzymes have so far been found in five insect species from two orders that diverged over 350 million years ago

(Sezutsu et al., 2013). Currently, no structural or catalytic data are available for these proteins and so it is not known if they contain a heme prosthetic group or function as a cytochrome P450. Studies have been undertaken to determine the importance of the cysteine residue, including mutagenesis studies which resulted in inactive forms of the protein (Yoshioka et al., 2001). In other studies, mutating the cysteine residue to a histidine resulted in non-natural reactions such as carbene- and nitrene-transfer reactions (McIntosh et al., 2015).

The P450 enzyme TxtE (CYP1048A1) isolated from Streptomyces scabiei has a very unusual catalytic cycle. The cycle progresses as normal until the ferric superoxy intermediate is produced. At this point, nitric oxide is donated from a NOS isoform, causing the formation of a peroxynitrite intermediate which catalyzes the nitration of the L- tryptophan substrate to form the product (Barry et al., 2012).

Certain P450 enzymes can be covalently modified after synthesis by phosphorylation, glycosylation, ubiquitination and nitration to become ‘unusual’ members 46 of the P450 superfamily. For example, phosphorylation of the hepatic drug-metabolizing

CYP3A4 leads to increased ubiquitination of the protein, resulting in its aggregation and protein degradation (Wang et al., 2009). For the human 25-hydroxycholesterol 7-alpha- hydroxylase CYP7B1, the catalytic properties of the enzyme change for different locations of the P450 in the human body due to different protein interactions, for example with

CYP17A1 (Katagiri et al., 1995) and CYP7B1 (Stiles et al., 2009).

1.2. The Natural Fusion Protein P450 BM3 (CYP102A1)

P450 BM3 was the third P450 enzyme isolated from the soil bacterium Bacillus megaterium. This is a gram-positive bacterium capable of growing across a wide temperature range (3-45 °C). B. megaterium is currently used in industry for the production of antibiotic precursors, fungicides, vitamin B12 and other compounds (Vary et al., 2007).

This widely studied P450 enzyme was characterized in the 1970s by Armand Fulco’s lab in a series of papers (Matson et al., 1977, Miura and Fulco, 1974, Hare and Fulco, 1975).

This protein was later discovered to be a fatty acid monooxygenase capable of hydroxylation at the -1, -2 and -3 positions of a range of different fatty acids of chain lengths from C12 to C20 (Miura and Fulco, 1975, Ho and Fulco, 1976). The enzyme has very high activity towards the polyunsaturated fatty acid arachidonic acid, which contains four carbon-carbon double bonds. Interestingly, unsaturated fatty acids are toxic to B. megaterium. It is postulated that the P450 BM3 enzyme may have evolved to detoxify such fatty acids produced by plants (English et al., 1994, Makita et al., 1996). Another hypothesis is that branched-chain saturated fatty acids are the natural substrates for P450

BM3, because these compounds have been shown to induce P450 BM3 expression

(English et al., 1997). As mentioned previously, P450 BM3 is a self-sufficient, natural fusion protein with the highest reported catalytic rate for any cytochrome P450 monooxygenase (Noble et al., 1999). Many researchers have used wild-type and mutant 47 forms of P450 BM3 for biotechnological and other applications due to its ease of expression and purification, and its high turnover number.

1.2.1. The Structure of P450 BM3

The full-length P450 BM3 protein is comprised of a ~55 kDa heme-binding catalytic domain and a ~65 kDa diflavin reductase domain forming a ~119 kDa monomer

(Narhi and Fulco, 1986, Narhi and Fulco, 1987). The diflavin reductase domain, also known as the CPR (cytochrome P450 reductase) or BMR (BM3 reductase) domain, can be further divided into a FAD/NADPH-binding domain of 44 kDa and a 21 kDa FMN- binding domain (Miles et al., 1992, Govindaraj and Poulos, 1997, Joyce et al., 2012). The

1048 amino acid sequence and the secondary structure of the full-length protein are shown in Figure 8 and the tertiary structures of the P450 and CPR domains from X-ray crystallography are shown in Figure 9.

The P450 catalytic domain (Figure 9) contains the heme prosthetic group, and displays amino acid sequence and structural similarity to other P450 enzymes, such as

P450cam (Graham and Peterson, 1999). The active site of this enzyme is flexible and stable, contributing to its attractive nature for mutagenesis (Anzenbacher and Hudecek, 2001).

The WXXRR motif is found in the C-helix. P450 BM3 contains a Trp residue, similar to eukaryotic P450 enzymes, and potentially due to its interactions with the reductase domain.

This differs from other bacterial P450s, which often have a His residue in this location, which is thought to be the ancestral residue (Ravichandran et al., 1993, Lewis, 1995).

Similar reductase contacts are thought to occur in the K-helix in close proximity to the C- helix. P450 BM3 also contains a motif located between the K- and L-helices containing the sequence: ZxxPxxZxPxxZ, where Z represents an aromatic amino acid and x denotes any amino acid. The final four amino acids are the PERF motif, which is often only observed in

48 eukaryotic P450 enzymes (Lewis et al., 1998). This motif may also contribute to reductase interactions. The EXXR motif is present in the K-helix and is thought to be essential for heme incorporation and correct protein folding (Lewis et al., 1998, Rupasinghe et al.,

2006). As the heme domain contains the active site, most P450 BM3 research is focused on this region, including substrate/inhibitor binding titrations, X-ray crystallography for structural analysis and EPR to investigate ligand binding modes. However, experiments involving e.g. reaction kinetics and product formation cannot be accomplished with this domain only. Interestingly, removal of the linker region through proteolysis or by using truncated forms (i.e. the individual P450 and CPR domains) results in dramatically

MGSSHHHHHH SSGLVPRGSH

(1 20)

Heme domain

A helix

MTIKEMPQPK TFGELKNLPL LNTDKPVQAL MKIADELGEI FKFEAPGRVT 1 50

B Helix B’ Helix C Helix

RYLSSQRLIK EACDESRFDK NLSQALKFVR DFAGDGLFTS WTHEKNWKKA 51 100 100

C’ D Helix E Helix

HNILLPSFSQ QAMKGYHAMM VDIAVQLVQK WERLNADEHI EVPEDMTRLT 101 150

F Helix

LDTIGLCGFN YRFNSFYRDQ PHPFITSMVR ALDEAMNKLQ RANPDDPAYD 151 200

G Helix H Helix

ENKRQFQEDI KVMNDLVDKI IADRKASGEQ SDDLLTHMLN GKDPETGEPL 201 250

I Helix J Helix

DDENIRYQII TFLIAGHETT SGLLSFALYF LVKNPHVLQK AAEEAARVLV 251 300

J’ Helix K Helix

DPVPSYKQVK QLKYVGMVLN EALRLWPTAP AFSLYAKEDT VLGGEYPLEK 301 350

GDELMVLIPQ LHRDKTIWGD DVEEFRPERF ENPSAIPQHA FKPFGNGQRA 351 400

L Helix

CIGQQFALHE ATLVLGMMLK HFDFEDHTNY ELDIKETLTL KPEGFVVKAK 49

401 450 FMN Domain

SKKIPLGGIP SPSTEQSAKK VRKKAENAHN TPLLVLYGSN MGTAEGTARD 451 500

LADIAMSKGF APQVATLDSH AGNLPREGAV LIVTASYNGH PPDNAKQFVD 501 550

WLDQASADEV KGVRYSVFGC GDKNWATTYQ KVPAFIDETL AAKGAENIAD 551 600

RGEADASDDF EGTYEEWREH MWSDVAAYFN LDIENSEDNK STLSLQFVDS 601 650 FAD Domain

AADMPLAKMH GAFSTNVVAS KELQQPGSAR STRHLEIELP KEASYQEGDH 651 700

LGVIPRNYEG IVNRVTARFG LDASQQIRLE AEEEKLAHLP LAKTVSVEEL 701 750

LQYVELQDPV TRTQLRAMAA KTVCPPHKVE LEALLEKQAY KEQVLAKRLT 751 800

MLELLEKYPA CEMKFSEFIA LLPSIRPRYY SISSSPRVDE KQASITVSVV 801 851

SGEAWSGYGE YKGIASNYLA ELQEGDTITC FISTPQSEFT LPKDPETPLI 851 900

MVGPGTGVAP FRGFVQARKQ LKEQGQSLGE AHLYFGCRSP HEDYLYQEEL 901 950

ENAQSEGIIT LHTAFSRMPN QPKTYVQHVM EQDGKKLIEL LDQGAHFYIC 951 1000

GDGSQMAPAV EATLMKSYAD VHQVSEADAR LWLQQLEEKG RYAKDVWAG 1001

Figure 8: The primary and secondary structure of P450 BM3 In the amino acid sequence, secondary structures are represented by different coloured highlights; red corresponding to alpha helices, green to beta sheets and yellow to turns. The secondary structure is also represented using shapes: rectangle, arrow and oval shapes correspond to alpha helices, beta sheets and turns respectively. A shape with faded colour represents a continuation on the next line or from the previous line, depending on the direction of shading. The His-tag is represented in light blue at the start of the sequence. The beginning of each domain is denoted with a blue arrow. This figure was made using UniProt deposit P14779 to identify the helices.

50 decreased catalytic rates, highlighting the importance of the integrity of the intact P450

BM3 enzyme for efficient catalysis (Narhi and Fulco, 1987, Boddupalli et al., 1990, Munro et al., 1994).

The BM3 CPR domain has not yet been crystallised in its intact state and so is typically modelled by the alignment of its individual FMN-binding domain crystal structure (from PDB 1BVY) and the FAD/NADPH-binding domain crystal structure (e.g. from PDB 4DQK) against mammalian CPR structures (such as 1AMO, the rat CPR structure). The individual domains of the P450 BM3 CPR domain are shown in Figure 9.

The diflavin reductase domain of P450 BM3 has high sequence similarity to mammalian

CPR proteins and exhibits a strong preference for NADPH over NADH, unlike most other prokaryotic P450 systems (Sevrioukova and Peterson, 1995, Lewis et al., 1998). The electron transfer process from the BM3 CPR domain to the heme domain has been hypothesized to occur in a similar manner to that of nitric oxide synthase (NOS); a mammalian fusion protein comprised of a CPR-like reductase with an arginine oxidizing heme domain that is structurally dissimilar to the cytochromes P450 (Siddhanta et al.,

1998, Stuehr et al., 2004, Girvan et al., 2011). A study found the BM3 reductase domain could reduce various electron acceptor molecules and was also able to donate electrons to human CYP1A2 and CYP2E1 to promote catalysis (Park et al., 2012). The majority of the members of the CPR family have a lysine residue two amino acids downstream from a conserved aspartic acid necessary for NADPH binding and dissociation. P450 BM3 is an exception to this as the sequence contains a serine residue in this position and is the only member of the diflavin reductase family that has this substitution (Xia et al., 2018). In other studies, mutating the human CPR Arg636 residue to alanine or serine (or using an

A635G/R636S double mutant) led to a small increase in CPR activity towards cytochrome c reduction (Mothersole et al., 2016).

The BM3 CPR domain is highly dynamic and flexible, and is less stable than the heme domain (Munro et al., 1996b). To date, there is no crystal structure of the BM3 CPR module. However, a BM3 P450-FMN domain fusion construct was crystallized. However, the structure obtained showed a non-stoichiometric domain arrangement, with two heme domains for each FMN-binding domain in the asymmetric unit, and with the linker region cleaved between each monomer (Sevrioukova et al., 1999). Despite the cleavage of the

FMN domain from the heme domain, the authors suggested that electron transfer could occur between FMN and heme. However, computational studies indicated that the extended FMN-to-heme pathway proposed would result in each electron transfer event taking ~50 years (Munro et al., 2002). Only two FAD/NADP(H)-binding domain structures have been determined by X-ray crystallography; these being the ligand-free

(4DQK) and NADP+-bound (4DQL) forms. However, this was only successful following the mutation of two surface cysteine residues to prevent inter-domain dimerization (Joyce et al., 2012). The BM3 CPR domain is known to form dimers, whilst the heme domain does not, suggesting that the CPR domain is responsible for the functional dimerization in the full-length P450 BM3 (see section 1.2.2.) (Neeli et al., 2005, Joyce et al., 2012).

The full-length structure of the highly dynamic P450 BM3 protein has proved difficult to determine. No wild-type, mutated or crosslinked full-length structure has yet been produced. The structure of the dynamic linker region has also not been elucidated.

The linker region is of great interest with respect to forming fusion proteins that use alternative P450 enzymes as the heme domain (see section 1.2.3). Many techniques have been used to determine structural features of the P450 BM3 protein, including small-angle

X-ray scattering (SAXS), NMR, EPR and, more recently, electron microscopy (discussed further in section 4.1).

Figure 9: The structures of the P450 BM3 heme and CPR domains The alpha helices are shown in cyan, the beta sheets in magenta, random coils in light pink and cofactors/prosthetic groups in yellow. The heme domain is the catalytic domain and is presented as the 1BU7 structure (Sevrioukova et al., 1999). The CPR domain contains the FMN-binding domain and the FAD-binding domain, shown by the structures 1BVY (Sevrioukova et al., 1999) and 4DQK structure (Joyce et al., 2012) respectively. These structures were produced using PyMOL.

1.2.2. Electron Transfer Within P450 BM3

NADPH binds to the BM3 CPR FAD-binding domain, donating a hydride ion, and leading to a two-electron reduced FAD cofactor (FADH2), known as the hydroquinone form (HQ). One electron is shuttled to the FMN cofactor, resulting in the formation of a neutral FADH semiquinone (SQ) and an anionic (FMNH-) semiquinone (Sevrioukova and

Peterson, 1995, Sevrioukova et al., 1996a). These species are detectable in both the isolated, single electron reduced FAD and FMN domains, and in the full-length CPR using

UV-visible spectroscopy, as the FADH SQ produces a blue colour with an absorbance peak at ~585 nm and the FMNH SQ produces a red colour with a peak at ~388 nm

(Sevrioukova et al., 1996a, Sevrioukova et al., 1996b). However, the latter SQ is more difficult to produce in a stable manner by reductive titrations with the FMN domain and has been observed more successfully using EPR with rapid freezing following reduction

(Boddupalli et al., 1992, Munro et al., 1996a). The FMN semiquinone (FMNH-) form is regenerated by electron transfer from the FAD semiquinone (FADH), and is again oxidized by the shuttling of a single electron to the heme domain (Sevrioukova and Peterson, 1995,

Sevrioukova et al., 1996a). The FMNH- SQ has a more negative redox potential than the

FMNH2 HQ form and so is the preferred form for electron transfer from the FMN to the heme prosthetic group (Daff et al., 1997). Excess amounts of NADPH (Sevrioukova and

Peterson, 1995, Murataliev et al., 1997) or lack of substrate (Matson et al., 1977,

Ruettinger and Fulco, 1981) cause the FMNH2 hydroquinone (HQ) form to build up. The

FMN HQ is an ineffective reductant of the heme iron and so affects catalytic rates and product formation. Due to this, turnover experiments using P450 BM3 must be carefully planned in terms of stoichiometry with electron regeneration systems.

In summary, the BM3 reductase operates a 0-2-1-0 electron transfer cycle in which the digits indicate the number of electrons donated to the CPR domain. The resting form of the flavins are in the oxidized state (corresponding to 0). NADPH donates two electrons as a hydride ion to the FAD, and one of these is transferred to the FMN to produce a diradical (FAD neutral SQ/FMN anionic SQ) species. The FMN SQ provides the first electron to the substrate-bound ferric heme iron, and the FMN SQ is restored by electron transfer from the FAD SQ. A second electron delivery from the FMN SQ produces the ferric-peroxy species, leaving the BM3 reductase in an oxidized state

(McLean et al., 2015b). In comparison, other diflavin reductase proteins, such as mammalian CPRs, exhibit a different electron transfer mechanism. In such systems, electrons are transferred to the heme iron by the FMNH2 HQ form, regenerating the

FMNH- anionic semiquinone form. Therefore, eukaryotic CPRs operate a 1-3-2-1 electron transfer cycle. The resting form has one electron on the FMN, and is reduced to a 3- electron state by hydride transfer from NADPH, resulting in the formation of an FMN HQ and a FAD SQ. The FMN HQ passes one electron to the P450 heme iron (forming a ferrous species) and the HQ is then regenerated by single electron transfer from the FAD

SQ. A second electron transfer from the FMN HQ to the P450 heme (forming a ferric peroxy species) then restores the eukaryotic CPR 1-electron reduced resting state. The difference in mechanisms is due to differences in reduction potentials between the SQ/HQ forms of the eukaryotic CPR enzymes and BM3 (Daff et al., 1997). This is attributed to the removal of a glycine residue in the BM3 sequence in the FMN binding loop (Sevrioukova et al., 1999). Removing this residue from NOS changes the electron transfer mechanism observed within this protein, resulting in the donating of electrons from the FAD HQ form to the SQ form of FMN, the same mechanism as observed in BM3 (Li et al., 2008).

Interestingly, the electrons shuttled from the reductase domain are transferred to the heme domain of the other monomer in the dimeric structure of P450 BM3 (Girvan et 55 al., 2011). This is also observed in the NOS enzymes, which also use diflavin reductases as redox partners (Siddhanta et al., 1998). In the case of P450 BM3, preventing dimer formation dramatically decreases the enzyme’s catalytic rate, but does not affect inter- flavin electron transfer (Neeli et al., 2005). As the full-length structure of the BM3 CPR domain has not been determined by X-ray crystallography or other means (to suitable resolution), the conformational changes occurring during the BM3 electron transfer cycle have yet to be fully characterized. However, CPR enzymes have also been extensively studied in mammals and many observations have been made on their conformational changes during the catalytic cycle, and these may be similar to those occurring in the BM3 reductase domain. For CPR in the NADPH-bound reduced form, the FAD and FMN cofactors are close together as the reductase domain is in a closed conformation. In this orientation, the cofactors interact through their xylene rings at the 7- and 8-methyl positions. The oxidized CPR protein has its FMN-binding domain rotated away from the

FAD-binding domain on a hinge in an open conformation for ~50% of the time, suggesting that more than one type of movement occurs within this state. This has been observed using SAXS (Ellis et al., 2009), small-angle neutron scattering (SANS) (Freeman et al.,

2017) and X-ray crystallography (Xia et al., 2018, Hamdane et al., 2009). A full electron transfer cycle has been proposed using X-ray crystallographic data for various mutants, as well as kinetic analysis, as discussed previously. A model is shown in Figure 10.

The FMN cofactor and the heme prosthetic group cannot make direct interactions as the heme prosthetic group is buried deep within the P450 heme domain. Various residues considered likely to be involved in transferring electrons from the FMN cofactor to the P450 heme have been investigated by mutagenesis. However, an “exact” pathway is still unknown and different residues may be involved in different P450s. Modelling has been attempted in order to identify important regions and residues involved in electron transfer between the CPR and P450. Unfortunately, for P450 BM3 the modelling relies on 56 the non-natural heme-FMN structure (1BVY) for such calculations. In one study, the residues identified as important were found in the K- and L-helices, in addition to residues in random coil segments of the heme domain and M490 of the FMN-binding domain

(Verma et al., 2013). Modelling conformational changes with substrates demonstrated the importance of the C-helix for CPR domain interactions. The I-helix in these models serves as the bridge between the heme domain and the C-helix, allowing for electron transfer from the mobile FMN-binding domain, which then docks onto the proximal face of the

Figure 10: The proposed CPR conformational changes that occur during electron transfer to a cytochrome P450 Upon NADPH binding, conformational changes occur internally in the FAD-binding domain. As electrons are shuttled from NADPH (dark green) to the FAD cofactor (brown) and on to the FMN cofactor (blue), NADP+ (light green) is ejected from the FAD-binding domain. The FMN-binding domain rotates away from the FAD-binding domain, allowing for its interactions with the cytochrome P450 (red). After electron donation to the P450, the FMN-binding domain rotates back to interact with the FAD-binding domain for a further electron transfer reaction. This is followed by another rotation back to allow for the second electron donation to the cytochrome P450. This figure is simplified and adapted from Xia et al. 2018, made using ChemBioDraw.

57 heme domain and transfers an electron to the heme iron of the heme prosthetic group in a process that may also involve amino acid side chains close the heme (Dubey and Shaik,

2018). Using the heme-FMN structure (1BVY), a number of potential redox partner interaction sites (RPISs) were determined across the proximal face of the protein in the B-,

C-, C’-, I-, J’-, K- and L-helices, as well as in certain random coil segments (Gricman et al., 2016). Without a more natural heme-FMN domain structure, full-length reductase structure and/or full-length P450 BM3 structure, more detailed analysis of electron transfer and conformational changes during the catalytic cycle may be unattainable.

1.2.3. P450 BM3 Mutagenesis and the Gatekeeper Mutants

Mutations of P450 BM3 typically focus on the heme domain in order to alter substrate profiles and product formation, or on the reductase domain to alter reaction rates and to facilitate the formation of the functional dimer. A number of CPR domain mutations have been discussed, such as those required to determine the structure of the FAD-binding domain (section 1.2.1.) and mutations that affect electron transfer from the FMN-binding domain to the heme domain in the functional dimer (section 1.2.2.). Many researchers have fused other cytochromes P450 to the CPR domain of P450 BM3 to form catalytically self- sufficient chimeras with increased reaction rates. However, no BM3 CPR domain fusion has been reported with a catalytic rate as high as WT P450 BM3. One such example is the fusion of the Mycobacterium tuberculosis CYP130 to the BM3 CPR for improved catalysis with the human drug dextromethorphan (Ortega Ugalde et al., 2018). Many P450 chimeras have been produced with insoluble, membrane-bound P450 enzymes fused to the BM3 reductase domain. These chimeras are often soluble and catalytically self-sufficient. For example, BM3 CPR chimeras produced with CYP2C9, CYP2C19 and CYP3A4 were found to produce their natural products, as observed for diclofenac, omeprazole and erythromycin. Interestingly, in this study the chimeras were observed to form as dimers

(Dodhia et al., 2006). Many P450 chimeras have been produced using the natural linker region found in P450 BM3, compared to other chimeras in which the length of linker needed is unknown for correct orientations of the component domains in solution.

Investigating alternative linker regions displayed how their different lengths and compositions affect the stability, NADPH oxidation rate, heme iron reduction potential and

Vmax for testosterone for CYP3A4-BM3 reductase chimeras (Degregorio et al., 2017).

P450 BM3 chimeras with the heme domain are reported less frequently in the literature compared to BM3 reductase domain chimeras. However, researchers have fused the P450 BM3 enzyme to a phosphite dehydrogenase enzyme. This allowed the turnover of rosiglitazone and omeprazole in a reaction driven by phosphite, with the covalently bound phosphite dehydrogenase recycling NADPH efficiently to improve NADPH-driven catalysis of BM3 (Beyer et al., 2017). The catalytic properties of the BM3 heme domain

(relative to the intact WT BM3 enzyme) were also shown to be enhanced by immobilizing the heme domain on an electrode surface (Fantuzzi et al., 2006).

As the WT BM3 has a large, flexible active site within the heme domain, many researchers have introduced mutations in order to change its substrate profile. In many cases, several mutations can be required for efficient stereo- and regioselective reactions.

For example, four mutations were required for the successful conversion of n-butane to 2- butanol (Yang et al., 2017). For the production of 16-hydroxy- or 16-hydroxy- testosterone from testosterone, six mutations were required (Acevedo-Rocha et al., 2018).

The introduction of a large number of mutations, as stated above, into P450 BM3 or other proteins can result in increased instability, depending on the specific nature and positions of these mutations. The destabilizing nature of multiple mutations can be counteracted somewhat by mutations in ‘stabilizing regions’ of the P450, such as in the Cys400 loop, the lid domain and sheet 1 (Geronimo et al., 2016). The lid domain is comprised of the 59

B’-, F- and G-helices, and forms part of the substrate access channel (Joyce et al., 2004).

This region is highly labile and is one of the first sections of the protein to unfold as temperatures increase (Geronimo et al., 2016). Further examples of BM3 mutagenesis are discussed in terms of their use to industry in sections 1.3.1 and 1.3.3.

Many BM3 mutagenesis studies have been conducted with only one or two mutations, but these can also cause great changes to the structure and properties of the enzyme. The most commonly mutated residue is phenylalanine-87 (Whitehouse et al.,

2012). By mutating Phe87 to residues with smaller side chains, a bulky aromatic group is removed from the active site, creating additional space in the immediate environment of the heme. The F87V mutation was found to change the BM3 reaction products obtained for arachidonic acid from ~20% epoxidation at the C14/C15 double bond and ~80% hydroxylation at C18 to 100% epoxidation (GrahamLorence et al., 1997). In comparison, the F87Y mutation caused the arachidonic acid reaction to become completely uncoupled as similar concentrations of NADPH were consumed as for the WT BM3, but no products were formed (GrahamLorence et al., 1997). Other common positions mutated for desired effects on P450 BM3 catalysis are Arg47, Ala74, Leu75, Ala82, Leu181, Ala184, Leu188,

Thr260, Ile262, Ala264, Ala328, Phe393 and Leu437 (Whitehouse et al., 2012).

Within our laboratory, we have produced a double mutant (DM) variant containing the point mutations A82F and F87V, which greatly expand the substrate specificity profile of the enzyme. As mentioned previously, the F87V mutation removes steric bulk from the immediate environment of the active site, allowing larger ligands to bind closer to the heme prosthetic group. Ala82 is located in a binding pocket near the active site. The A82F mutation adds bulk to this residue and causes secondary structural interactions that induce substantial conformational changes in the protein, mimicking the conformation typically observed upon substrate binding to the P450, and leading to the 60 displacement of the F- and G- helices at the mouth of the active site to facilitate substrate binding. The A82F mutation thus causes the P450 to be in a ‘catalytically primed’ state, as the conformational changes normally seen upon substrate binding are induced even in the absence of substrate (Huang et al., 2007). These two mutations were described as gatekeeper mutations as they allowed the binding of several non-natural, very large molecules in the active site. A comparison between substrate-bound active sites is shown in Figure 11 with the binding of a modified fatty acid to WT BM3 (A) and of a Food and

Drug Administration (FDA) approved drug to the DM BM3 (B). Butler et al., 2013 demonstrated how the A82F mutation had the largest effect of the two mutations on the substrate binding profile and substrate binding affinities, with the double mutation

(A82F/F87V) also having slightly greater activity than the single A82F mutant. From these studies, human metabolites of proton pump inhibitor drug substrates (including omeprazole) were also characterized (Butler et al., 2013, Butler et al., 2014).

1.3. Real-World Applications of P450 Enzymes

As P450 enzymes can catalyze a wide range of reactions allowing for the formation of many valuable oxidized and other compounds, many industries could make use of these enzymes, such as the pharmaceutical industry (Guengerich, 2003, Gao et al.,

2013). Metabolism resulting in toxic metabolites can cause reduced drug half-lives, side effects and other problems resulting in failed drug trials, loss of money and time. The FDA states that any metabolite produced at 10% of the parent drug compound must be identified and characterized. For this reason, many P450 screens are commercially available for metabolite screening. P450 enzymes are also being used to produce compounds useful to many industries, due to their ability to catalyze many reactions that are otherwise difficult to perform synthetically without expensive and often toxic shielding groups. Many of these

61 enzymes have been adapted for such reactions using mutagenesis in order to change their substrate profile and thermostability.

Figure 11: A comparison of the flexibility of the P450 BM3 active site in the WT and DM proteins By mutation of two key residues (A82F and F87V), the substrate-free conformation of the protein is destabilized (A82F) and steric bulk is removed from the active site (F87V). Comparing the WT P450 BM3 structure with N-palmitoyl glycine bound (4KPA) (Panel A) and the DM with omeprazole bound (4KEY) (Panel B), it is clear that several much larger ligands are able to bind within the active site of the DM protein, while the WT enzyme favours mid- to long-chain fatty acids as substrates (Butler et al., 2013, Catalano et al., 2013). This figure was made using PyMol with alpha helices shown in cyan, beta sheets in magenta, random coils in light pink, the gatekeeper residues in red and the heme prosthetic group/ligand shown in yellow.

1.3.1. Using P450 BM3 in the Pharmaceutical Industry

As mentioned previously, there have been many studies done to introduce mutations into P450 BM3 for the production of human metabolites such as omeprazole

(Butler et al., 2013, Butler et al., 2014) and testosterone (Acevedo-Rocha et al., 2018).

Mutants have also been successfully produced using site-directed and random mutagenesis to enable the hydroxylation of 17--estradiol (Cha et al., 2014). Using P450 BM3 to produce human metabolites is an attractive alternative to synthetic chemistry approaches for metabolite production, particularly when compared with the low catalytic activity and instability exhibited by human P450 enzymes, as well as the difficulties associated with their expression (Whitehouse et al., 2012).

WT BM3 is able to bind short chain alkyl compounds such as hexyne and octyne

(Waltham et al., 2011). The R47E mutation allows P450 BM3 to bind much larger molecules such as alkyl trimethylammonium compounds, and to catalyze their oxidation at the -1, -2 and -3 positions (Oliver et al., 1997). Such quaternary ammonium salts have many uses, such as disinfectants, detergents, cosmetics and eye drops. These compounds have been shown to be toxic to mammals and to aquatic life due to cytotoxic effects (Pang and Willis, 1997). Hexadecyltrimethylammonium bromide (CTAB) has been explored as a treatment for cancer, highlighting the importance of such compounds and their interaction with human proteins (Ito et al., 2009). Further research is being conducted to understand the cytotoxic nature of these molecules, using P450 enzymes such as P450 BM3.

Styrene is a precursor to polystyrene, which is widely used for packaging across the world. If ingested, styrene is metabolized by human P450 enzymes and other proteins by oxidation or epoxidation and this can lead to toxic and mutagenic effects in mice and humans (NTP, 2014). Several P450 BM3 variants can metabolize styrene to (R)-styrene

63 oxide. The double mutant containing the gatekeeper mutation A82F with T438F can catalyze this epoxide reaction (Huang et al., 2011). However, a single F87G mutation can generate much higher levels of the product (Eiben et al., 2006).

P450 BM3 mutants can be used in the production of pharmaceutical compounds by converting cheaper precursors to high-value drug compounds. For example, the P450

BM3-catalyzed conversion of terpenes can be used to produce potential anti-tumour drugs such as perillyl alcohol from plants (Seifert et al., 2011), or antibacterial and antifungal compounds such as -cembrenediol from tobacco (Le-Huu et al., 2015). In another example, the anti-malarial drug artemisinin was successfully hydroxylated in a process using P450 BM3 mutants to produce high-value compounds, such as artemether (Zhang et al., 2012). Another example is the use of a quadruple P450 BM3 mutant with a His-ligated

BM3 heme iron for the catalysis of the enantioselective synthesis of the core precursor for the anti-depressant levomilnacipran, involving a novel epoxidation reaction (Wang et al.,

2014).

Many P450 BM3 mutants have been generated by the Wong lab in Oxford. In one study a library of single, double, triple and quadruple mutants was produced from the group’s previous directed evolution studies. These mutants were found to be capable of producing metabolites from diclofenac, naproxen, testosterone, chlorzoxazone, amitriptyline and lidocaine (Ren et al., 2015). In another study, P450 BM3 and P450cam were mutated to develop fragrant molecules from sesquiterpenes, such as the high-value grapefruit fragrance nootkatone (Sowden et al., 2005).

Another example of mutants of P450 BM3 being able to produce high-value fragrance molecules is the production of (E)-astringen (Le et al., 2017). This compound is found in pine tree extracts (Teguo et al., 1996) and wine (Vitrac et al., 2005) and has been 64 shown to be a powerful antioxidant molecule (Merillon et al., 1997) with potential uses as a cancer chemopreventive (Waffo-Teguo et al., 2001). In a similar reaction, resveratrol can be converted into anticancer metabolites such as piceatannol (Kim et al., 2009).

1.3.2. Using other P450 Enzymes in Industry

Several useful P450 enzymes have been mentioned in section 1.1.5, such as the explosive-degrading enzyme XplA (Jackson et al., 2007), the alkene-producing protein

OleTJE which can be used for biofuel manufacture (Belcher et al., 2014) and CYP121, which is currently being used for fragment screening to produce new tuberculosis treatments (Kavanagh et al., 2015, Kavanagh et al., 2016).

In some cases, reasons underlying drug toxicity can be traced back to a toxic metabolite produced by a P450 enzyme. For example, troglitazone is an antidiabetic compound that caused the death of over 60 patients during the two years that it was on the market. Using a range of human P450 enzymes, a quinone-type metabolite was found to be the cause of mortality through scavenging radicals in a similar manner to the paracetamol quinone-type metabolite responsible for overdoses (Yamazaki et al., 1999, Tolman, 2000).

Many drugs can be produced by cytochrome P450 reactions. For example, the enzyme P450eryF from the CYP107 family can produce the antibiotics erythromycin and tetracenomycin (Shafiee and Hutchinson, 1987, Shafiee and Hutchinson, 1988). The widely used compound hydrocortisone can be produced by 11β-hydroxylation of cortexolone using P450 enzymes within immobilized Curvularia lunata mycelium

(Sonomoto et al., 1983). Similarly, cortisone can be produced by P450-mediated hydroxylation of progesterone by Rhizopus arrhizus fermentation (Peterson et al., 1952).

Another example of P450 catalysis is seen in the production of the anticancer agent taxol

65 which has an inefficient synthetic pathway with multiple catalytic steps requiring many enzymes (Jennewein and Croteau, 2001). The P450 CYP725A4 catalyzes the first step of this pathway through an oxidation reaction that converts taxadiene to taxadiene-5-ol

(Rouck et al., 2017, Hefner et al., 1996).

Many one-pot reactions have been developed which use P450 enzymes that must overcome problems such as pH and temperature in order to achieve optimum product formation of intermediates while ensuring the stability of the different enzymes present within the mixture. A particularly elegant system was designed within our lab for the production of the cholesterol-lowering drug pravastatin from the cheaper precursor compactin (McLean et al., 2015a). In this work, the relevant CYP105AS1 enzyme was mutated to allow for specific and optimized product conversion using an industrial

Penicillium chrysogenum strain for the large-scale fermentation of the compactin. A

CYP105AS1 mutant with five mutations that led to the stereoselective formation of the desired pravastatin product was developed and named P450Prava. The production of pravastatin by P450 enzymes was first developed in the Streptomyces carbophilus

P450sca-2 system. However, this system uses a 2-step production/fermentation process and does not produce as high quantities of product as the P450Prava system (Hosobuchi et al.,

1993). Other statins can also be produced by reactions with P450 enzymes. These include lovastatin and simvastatin produced by CYP109E1, which can also catalyze the reaction to produce pravastatin from compactin (Putkaradze et al., 2017).

P450 enzymes can also be therapeutic targets for the treatment of disease. For example, azole inhibitors were developed to inhibit CYP51 (sterol demethylase) enzymes in fungal cells. CYP51 is responsible for lanosterol demethylation, producing other sterols essential in the cell, including sterols essential for the integrity of the cell wall (Lepesheva and Waterman, 2007). Older azole drugs can also inhibit human P450 enzymes and so 66 cannot be taken orally. Over time, new azoles have been produced which are less inhibitory to human CYP51 and to other P450 enzymes, leading to fewer unwanted side effects (Drouhet and Dupont, 1980). Another example is the CYP46A1 enzyme, which is localized in the brain in humans and is responsible for the elimination of cholesterol (in the form of 24-hydroxycholesterol) across the blood-brain barrier. The activity of this enzyme can be increased by the binding of the human immunodeficiency virus (HIV) drug efavirenz to an allosteric binding site. Due to this research, efavirenz is now being trialled as a treatment for Alzheimer’s (Anderson et al., 2016).

Finally, the intermediates of the opiate morphine can be produced using a fusion protein of the plant P450 CYP82Y2 to an aldo-keto reductase (AKR). The P450 domain catalyzes the transformation of (S)-reticuline to 1,2-dehydroreticuline. This product is then converted to (R)-reticuline by the AKR domain. If used in a one-pot system, then another

P450 (CYP719B1) can also be utilized to catalyze the formation of the alkaloid salutaridine (Farrow et al., 2015, Galanie et al., 2015, Gesell et al., 2009).

1.4. Project Aims

At first, the project aim was to produce human metabolites of pharmaceutical drugs from P450 BM3, mainly by using the DM (A82F/F87V) variant. To accomplish this aim, a large FDA-approved drug compound library was screened for molecules capable of binding to DM BM3, as described in Chapter 3. Certain drug classes were chosen for further investigation, including product analysis. Two drug targets were investigated in- depth; the antidiabetic compounds which include several drug classes (Chapter 5) and the fibrate class for the treatment of hyperlipidaemia (Chapter 6). The binding modes of these ligands were analyzed from both classes, using EPR, drug binding titrations, X-ray crystallography and HDX-MS. For both of these classes, the known human metabolites 67 were produced using HPLC, LC-MS, LC-MS/MS and NMR using the BM3 DM enzyme.

This highlights the ability of the P450 BM3 gatekeeper mutants to mimic human P450 enzyme activities. In addition, the anti-fungal azole class was investigated due to their interesting inhibitory characteristics (Chapter 4). This included structural studies using X- ray crystallography, EPR and inhibitor binding titrations leading to Kd determinations.

During the duration of the project a second aim was investigated; to provide further information on the full-length P450 BM3 protein and to analyze its dimeric interface. To accomplish this aim, a collaboration was undertaken enabling the use of the hydrogen- deuterium exchange mass spectrometry (HDX-MS) technique and other MS techniques such as native MS and collision-induced unfolding (CIU). These experiments provided novel insights into the dimeric interface of the reductase domain and other interactions made in the full-length BM3 protein. In addition, protein dynamics induced on ligand binding were also observed. These results are discussed in Chapter 2.

1.5. References

L. 2012. Cytochrome P450-catalyzed L-tryptophan nitration in thaxtomin phytotoxin biosynthesis. Nat Chem Biol, 8, 814-6. BELCHER, J., MCLEAN, K. J., MATTHEWS, S., WOODWARD, L. S., FISHER, K., RIGBY, S. E. J., NELSON, D. R., POTTS, D., BAYNHAM, M. T., PARKER, D. A., LEYS, D. & MUNRO, A. W. 2014. Structure and Biochemical Properties of the Alkene Producing Cytochrome P450 OleTJE (CYP152L1) from the Jeotgalicoccus sp 8456 Bacterium. J Biol Chem, 289, 6535-6550. BERG, J. M., TYMOCZKO, J. L. & STRYER, L. 2007. The biosynthesis of membrane lipids and steroids. In: BERG, J. M., TYMOCZKO, J. L. & STRYER, L. (eds.) Biochemistry. Sixth ed. New York: W. H. Freeman and company. BEYER, N., KULIG, J. K., BARTSCH, A., HAYES, M. A., JANSSEN, D. B. & FRAAIJE, M. W. 2017. P450BM3 fused to phosphite dehydrogenase allows phosphite-driven selective oxidations. Appl Microbiol Biotechnol, 101, 2319-2331. BODDUPALLI, S. S., ESTABROOK, R. W. & PETERSON, J. A. 1990. Fatty-Acid Monooxygenation by Cytochrome-P-450bm-3. J Biol Chem, 265, 4233-4239. BODDUPALLI, S. S., PRAMANIK, B. C., SLAUGHTER, C. A., ESTABROOK, R. W. & PETERSON, J. A. 1992. Fatty acid monooxygenation by P450BM-3: product identification and proposed mechanisms for the sequential hydroxylation reactions. Arch Biochem Biophys, 292, 20-8. BUI, S. H., MCLEAN, K. J., CHEESMAN, M. R., BRADLEY, J. M., RIGBY, S. E., LEVY, C. W., LEYS, D. & MUNRO, A. W. 2012. Unusual spectroscopic and ligand binding properties of the cytochrome P450-flavodoxin fusion enzyme XplA. J Biol Chem, 287, 19699-714. BUTLER, C. F., PEET, C., MASON, A. E., VOICE, M. W., LEYS, D. & MUNRO, A. W. 2013. Key mutations alter the cytochrome P450 BM3 conformational landscape and remove inherent substrate bias. J Biol Chem, 288, 25387-99. BUTLER, C. F., PEET, C., MCLEAN, K. J., BAYNHAM, M. T., BLANKLEY, R. T., FISHER, K., RIGBY, S. E., LEYS, D., VOICE, M. W. & MUNRO, A. W. 2014. Human P450-like oxidation of diverse proton pump inhibitor drugs by 'gatekeeper' mutants of flavocytochrome P450 BM3. Biochem J, 460, 247-59. CATALANO, J., SADRE-BAZZAZ, K., AMODEO, G. A., TONG, L. & MCDERMOTT, A. 2013. Structural evidence: a single charged residue affects substrate binding in cytochrome P450 BM-3. Biochemistry, 52, 6807-15. CHA, G. S., RYU, S. H., AHN, T. & YUN, C. H. 2014. Regioselective hydroxylation of 17beta-estradiol by mutants of CYP102A1 from Bacillus megaterium. Biotechnol Lett, 36, 2501-6. CHAMPION, P. M., STALLARD, B. R., WAGNER, G. C. & GUNSALUS, I. C. 1982. Resonance Raman Detection of an Fe-S Bond in Cytochrome-P450cam. J Am Chem Soc, 104, 5469-5472. CHENG, H. & MARKLEY, J. L. 1995. NMR spectroscopic studies of paramagnetic proteins: iron-sulfur proteins. Annu Rev Biophys Biomol Struct, 24, 209-37. CHENG, Q. A., LAMB, D. C., KELLY, S. L., LEI, L. & GUENGERICH, F. P. 2010. Cyclization of a Cellular Dipentaenone by Streptomyces coelicolor Cytochrome P450 154A1 without Oxidation/Reduction. J Am Chem Soc, 132, 15173-15175. COOPER, D. Y., LEVIN, S., NARASIMHULU, S. & ROSENTHAL, O. 1965. Photochemical Action Spectrum of the Terminal Oxidase of Mixed Function Oxidase Systems. Science, 147, 400-2. CRAMER, S. P., DAWSON, J. H., HODGSON, K. O. & HAGER, L. P. 1978. Studies on Ferric Forms of Cytochrome-P-450 and Chloroperoxidase by Extended X-Ray Absorption Fine-Structure - Characterization of Fe-N and Fe-S Distances. J Am Chem Soc, 100, 7282-7290.

DAFF, S. N., CHAPMAN, S. K., TURNER, K. L., HOLT, R. A., GOVINDARAJ, S., POULOS, T. L. & MUNRO, A. W. 1997. Redox control of the catalytic cycle of flavocytochrome P-450 BM3. Biochemistry, 36, 13816-23. DANIELSON, P. B. 2002. The cytochrome P450 superfamily: biochemistry, evolution and drug metabolism in humans. Curr Drug Metab, 3, 561-97. DAWSON, J. H. 1988. Probing structure-function relations in heme-containing oxygenases and peroxidases. Science, 240, 433-9. DEGREGORIO, D., D'AVINO, S., CASTRIGNANO, S., DI NARDO, G., SADEGHI, S. J., CATUCCI, G. & GILARDI, G. 2017. Human Cytochrome P450 3A4 as a Biocatalyst: Effects of the Engineered Linker in Modulation of Coupling Efficiency in 3A4-BMR Chimeras. Front Pharmacol, 8, 121. DENISOV, I. G., HUNG, S. C., WEISS, K. E., MCLEAN, M. A., SHIRO, Y., PARK, S. Y., CHAMPION, P. M. & SLIGAR, S. G. 2001a. Characterization of the oxygenated intermediate of the thermophilic cytochrome P450 CYP119. J Inorg Biochem, 87, 215-26. DENISOV, I. G., MAK, P. J., MAKRIS, T. M., SLIGAR, S. G. & KINCAID, J. R. 2008. Resonance Raman Characterization of the Peroxo and Hydroperoxo Intermediates in Cytochrome P450. J Phys Chem A, 112, 13172-13179. DENISOV, I. G., MAKRIS, T. M. & SLIGAR, S. G. 2001b. Cryotrapped reaction intermediates of cytochrome p450 studied by radiolytic reduction with phosphorus- 32. J Biol Chem, 276, 11648-52. DENISOV, I. G., MAKRIS, T. M. & SLIGAR, S. G. 2002. Cryoradiolysis for the study of P450 reaction intermediates. Methods Enzymol, 357, 103-15. DIETRICH, J. A., YOSHIKUNI, Y., FISHER, K. J., WOOLARD, F. X., OCKEY, D., MCPHEE, D. J., RENNINGER, N. S., CHANG, M. C. Y., BAKER, D. & KEASLING, J. D. 2009. A Novel Semi-biosynthetic Route for Artemisinin Production Using Engineered Substrate-Promiscuous P450(BM3). ACS Chem Biol, 4, 261-267. DIPRIMO, C., HOA, G. H. B., DEPREZ, E., DOUZOU, P. & SLIGAR, S. G. 1993. Conformational Dynamics of Cytochrome-P-450(Cam) as Monitored by Photoacoustic Calorimetry. Biochemistry, 32, 3671-3676. DJELTI, F., BRAUDEAU, J., HUDRY, E., DHENAIN, M., VARIN, J., BIECHE, I., MARQUER, C., CHALI, F., AYCIRIEX, S., AUZEIL, N., ALVES, S., LANGUI, D., POTIER, M. C., LAPREVOTE, O., VIDAUD, M., DUYCKAERTS, C., MILES, R., AUBOURG, P. & CARTIER, N. 2015. CYP46A1 inhibition, brain cholesterol accumulation and neurodegeneration pave the way for Alzheimer's disease. Brain, 138, 2383-2398. DODHIA, V. R., FANTUZZI, A. & GILARDI, G. 2006. Engineering human cytochrome P450 enzymes into catalytically self-sufficient chimeras using molecular Lego. J Biol Inorg Chem, 11, 903-16. DRISCOLL, M. D., MCLEAN, K. J., CHEESMAN, M. R., JOWITT, T. A., HOWARD, M., CARROLL, P., PARISH, T. & MUNRO, A. W. 2011. Expression and characterization of Mycobacterium tuberculosis CYP144: common themes and lessons learned in the M. tuberculosis P450 enzyme family. Biochim Biophys Acta, 1814, 76-87. DROUHET, E. & DUPONT, B. 1980. Chronic mucocutaneous candidosis and other superficial and systemic mycoses successfully treated with ketoconazole. Rev Infect Dis, 2, 606-19. DUBEY, K. D. & SHAIK, S. 2018. Choreography of the Reductase and P450BM3 Domains Toward Electron Transfer Is Instigated by the Substrate. J Am Chem Soc, 140, 683-690.

EIBEN, S., KAYSSER, L., MAURER, S., KUHNEL, K., URLACHER, V. B. & SCHMID, R. D. 2006. Preparative use of isolated CYP102 monooxygenases -- a critical appraisal. J Biotechnol, 124, 662-9. EL-GARJ, F. M. A., WAJIDI, M. F. F. & AVICOR, S. W. 2016. Identification and analysis of a processed cytochrome P450 pseudogene of the disease vector Aedes aegypti. Asian Pac J Trop Med, 9, 973-978. ELLIS, J., GUTIERREZ, A., BARSUKOV, I. L., HUANG, W. C., GROSSMANN, J. G. & ROBERTS, G. C. 2009. Domain motion in cytochrome P450 reductase: conformational equilibria revealed by NMR and small-angle X-ray scattering. J Biol Chem, 284, 36628-37. ENGLISH, N., HUGHES, V. & WOLF, C. R. 1994. Common pathways of cytochrome P450 gene regulation by peroxisome proliferators and barbiturates in Bacillus megaterium ATCC14581. J Biol Chem, 269, 26836-41. ENGLISH, N., PALMER, C. N., ALWORTH, W. L., KANG, L., HUGHES, V. & WOLF, C. R. 1997. Fatty acid signals in Bacillus megaterium are attenuated by cytochrome P-450-mediated hydroxylation. Biochem J, 327 ( Pt 2), 363-8. ESTABROOK, R. W. 2003. A passion for P450s (rememberances of the early history of research on cytochrome P450). Drug Metab Dispos, 31, 1461-73. ESTABROOK, R. W., COOPER, D. Y. & ROSENTHAL, O. 1963. The Light Reversible Carbon Monoxide Inhibition of the Steroid C21-Hydroxylase System of the Adrenal Cortex. Biochem Z, 338, 741-55. ESTRADA, D. F., LAURENCE, J. S. & SCOTT, E. E. 2013. Substrate-modulated Cytochrome P450 17A1 and Cytochrome b7(5) Interactions Revealed by NMR. J Biol Chem, 288, 17008-17018. FANTUZZI, A., MEHARENNA, Y. T., BRISCOE, P. B., SASSONE, C., BORGIA, B. & GILARDI, G. 2006. Improving catalytic properties of P450 BM3 haem domain electrodes by molecular Lego. Chem Commun (Camb), 1289-91. FARROW, S. C., HAGEL, J. M., BEAUDOIN, G. A., BURNS, D. C. & FACCHINI, P. J. 2015. Stereochemical inversion of (S)-reticuline by a cytochrome P450 fusion in opium poppy. Nat Chem Biol, 11, 728-32. FEYEREISEN, R. 2012. Insect CYP Genes and P450 Enzymes (Chapter 8). In: GILBERT, L. I. (ed.) Insect Molecular Biology and Biochemistry. INRA Sophia Antipolis, France: Academic Press. FREEMAN, S. L., MARTEL, A., RAVEN, E. L. & ROBERTS, G. C. K. 2017. Orchestrated Domain Movement in Catalysis by Cytochrome P450 Reductase. Sci Rep, 7, 9741. GALANIE, S., THODEY, K., TRENCHARD, I. J., FILSINGER INTERRANTE, M. & SMOLKE, C. D. 2015. Complete biosynthesis of opioids in yeast. Science, 349, 1095-100. GAO, H., JACOBS, A., WHITE, R. E., BOOTH, B. P. & OBACH, R. S. 2013. Meeting report: metabolites in safety testing (MIST) symposium-safety assessment of human metabolites: what's REALLY necessary to ascertain exposure coverage in safety tests? AAPS J, 15, 970-3. GARFINKEL, D. 1958. Studies on pig liver microsomes. I. Enzymic and pigment composition of different microsomal fractions. Arch Biochem Biophys, 77, 493- 509. GERONIMO, I., DENNING, C. A., ROGERS, W. E., OTHMAN, T., HUXFORD, T., HEIDARY, D. K., GLAZER, E. C. & PAYNE, C. M. 2016. Effect of Mutation and Substrate Binding on the Stability of Cytochrome P450BM3 Variants. Biochemistry, 55, 3594-606. GESELL, A., ROLF, M., ZIEGLER, J., DIAZ CHAVEZ, M. L., HUANG, F. C. & KUTCHAN, T. M. 2009. CYP719B1 is salutaridine synthase, the C-C phenol-

coupling enzyme of morphine biosynthesis in opium poppy. J Biol Chem, 284, 24432-42. GIBSON, G. G. & SKETT, P. 2001. Enzymology and molecular mechanisms (Chaper 2). In: GIBSON, G. G. & SKETT, P. (eds.) Introduction to drug metabolism. Third ed. Bath, England: Nelson Thornes Ltd. GIRVAN, H. M., DUNFORD, A. J., NEELI, R., EKANEM, I. S., WALTHAM, T. N., JOYCE, M. G., LEYS, D., CURTIS, R. A., WILLIAMS, P., FISHER, K., VOICE, M. W. & MUNRO, A. W. 2011. Flavocytochrome P450 BM3 mutant W1046A is a NADH-dependent fatty acid hydroxylase: Implications for the mechanism of electron transfer in the P450 BM3 dimer. Arch Biochem Biophys, 507, 75-85. GIRVAN, H. M., SEWARD, H. E., TOOGOOD, H. S., CHEESMAN, M. R., LEYS, D. & MUNRO, A. W. 2007. Structural and spectroscopic characterization of P450 BM3 mutants with unprecedented P450 heme iron ligand sets. New heme ligation states influence conformational equilibria in P450 BM3. J Biol Chem, 282, 564-72. GOVINDARAJ, S. & POULOS, T. L. 1997. The domain architecture of cytochrome P450BM-3. J Biol Chem, 272, 7915-7921. GRAHAM, S. E. & PETERSON, J. A. 1999. How similar are P450s and what can their differences teach us? Arch Biochem Biophys, 369, 24-9. GRAHAMLORENCE, S., TRUAN, G., PETERSON, J. A., FALCK, J. R., WEI, S. Z., HELVIG, C. & CAPDEVILA, J. H. 1997. An active site substitution, F87V, converts cytochrome P450 BM-3 into a regio- and stereoselective (14S,15R)- arachidonic acid epoxygenase. J Biol Chem, 272, 1127-1135. GRANT, J. L., MITCHELL, M. E. & MAKRIS, T. M. 2016. Catalytic strategy for carbon- carbon bond scission by the cytochrome P450 OleT. Proc Natl Acad Sci U S A, 113, 10049-54. GRICMAN, L., WEISSENBORN, M. J., HOFFMANN, S. M., BORLINGHAUS, N., HAUER, B. & PLEISS, J. 2016. Redox Partner Interaction Sites in Cytochrome P450 Monooxygenases: In Silico Analysis and Experimental Validation. Chemistry Select, 1, 1243-1251. GROVES, J. T. & MCCLUSKY, G. A. 1978. Aliphatic hydroxylation by highly purified liver microsomal cytochrome P-450. Evidence for a carbon radical intermediate. Biochem Biophys Res Commun, 81, 154-60. GUENGERICH, F. P. 2003. Cytochromes P450, drugs, and diseases. Mol Interv, 3, 194- 204. GUENGERICH, F. P. & JOHNSON, W. W. 1997. Kinetics of ferric cytochrome P450 reduction by NADPH-cytochrome P450 reductase: rapid reduction in the absence of substrate and variations among cytochrome P450 systems. Biochemistry, 36, 14741-50. GUMULYA, Y. & GILLAM, E. M. 2017. Exploring the past and the future of protein evolution with ancestral sequence reconstruction: the 'retro' approach to protein engineering. Biochem J, 474, 1-19. HAHN, J. E., HODGSON, K. O., ANDERSSON, L. A. & DAWSON, J. H. 1982. Endogenous Cysteine Ligation in Ferric and Ferrous Cytochrome-P-450 Direct Evidence from X-Ray Absorption-Spectroscopy. J Biol Chem, 257, 934-941. HAMDANE, D., XIA, C., IM, S. C., ZHANG, H., KIM, J. J. & WASKELL, L. 2009. Structure and function of an NADPH-cytochrome P450 oxidoreductase in an open conformation capable of reducing cytochrome P450. J Biol Chem, 284, 11374-84. HARE, R. S. & FULCO, A. J. 1975. Carbon monoxide and hydroxymercuribenzoate sensitivity of a fatty acid (omega-2) hydroxylase from Bacillus megaterium. Biochem Biophys Res Commun, 65, 665-72. HARRIS, K. L., THOMSON, R. E. S., STROHMAIER, S. J., GUMULYA, Y. & GILLAM, E. M. J. 2018. Determinants of thermostability in the cytochrome P450 fold. Biochim Biophys Acta - Proteins and Proteomics, 1866, 97-115. 72

HASEMANN, C. A., KURUMBAIL, R. G., BODDUPALLI, S. S., PETERSON, J. A. & DEISENHOFER, J. 1995. Structure and Function of Cytochromes-P450 - a Comparative-Analysis of 3 Crystal-Structures. Structure, 3, 41-62. HEFNER, J., RUBENSTEIN, S. M., KETCHUM, R. E., GIBSON, D. M., WILLIAMS, R. M. & CROTEAU, R. 1996. Cytochrome P450-catalyzed hydroxylation of taxa- 4(5),11(12)-diene to taxa-4(20),11(12)-dien-5alpha-ol: the first oxygenation step in taxol biosynthesis. Chem Biol, 3, 479-89. HIRUMA, Y., HASS, M. A., KIKUI, Y., LIU, W. M., OLMEZ, B., SKINNER, S. P., BLOK, A., KLOOSTERMAN, A., KOTEISHI, H., LOHR, F., SCHWALBE, H., NOJIRI, M. & UBBINK, M. 2013. The structure of the cytochrome P450cam- putidaredoxin complex determined by paramagnetic NMR spectroscopy and crystallography. J Mol Biol, 425, 4353-65. HO, P. P. & FULCO, A. J. 1976. Involvement of a single hydroxylase species in the hydroxylation of palmitate at the omega-1, omega-2 and omega-3 positions by a preparation from Bacillus megaterium. Biochim Biophys Acta, 431, 249-56. HOSOBUCHI, M., KUROSAWA, K. & YOSHIKAWA, H. 1993. Application of Computer to Monitoring and Control of Fermentation Process - Microbial Conversion of Ml-236b Na to Pravastatin. Biotechnol Bioeng, 42, 815-820. HSIEH, C. H., HUANG, X., AMAYA, J. A., RUTLAND, C. D., KEYS, C. L., GROVES, J. T., AUSTIN, R. N. & MAKRIS, T. M. 2017. The Enigmatic P450 Decarboxylase OleT Is Capable of, but Evolved To Frustrate, Oxygen Rebound Chemistry. Biochemistry, 56, 3347-3357. HUANG, W. C., CULLIS, P. M., RAVEN, E. L. & ROBERTS, G. C. 2011. Control of the stereo-selectivity of styrene epoxidation by cytochrome P450 BM3 using structure- based mutagenesis. Metallomics, 3, 410-6. HUANG, W. C., WESTLAKE, A. C., MARECHAL, J. D., JOYCE, M. G., MOODY, P. C. & ROBERTS, G. C. 2007. Filling a hole in cytochrome P450 BM3 improves substrate binding and catalytic efficiency. J Mol Biol, 373, 633-51. HUDECEK, J., ANZENBACHEROVA, E., ANZENBACHER, P., MUNRO, A. W. & HILDEBRANDT, P. 2000. Structural similarities and differences of the heme pockets of various P450 isoforms as revealed by resonance Raman spectroscopy. Arch Biochem Biophys, 383, 70-78. INCE, I., KNIBBE, C. A., DANHOF, M. & DE WILDT, S. N. 2013. Developmental changes in the expression and function of cytochrome P450 3A isoforms: evidence from in vitro and in vivo investigations. Clin Pharmacokinet, 52, 333-45. ITO, E., YIP, K. W., KATZ, D., FONSECA, S. B., HEDLEY, D. W., CHOW, S., XU, G. W., WOOD, T. E., BASTIANUTTO, C., SCHIMMER, A. D., KELLEY, S. O. & LIU, F. F. 2009. Potential use of cetrimonium bromide as an apoptosis-promoting anticancer agent for head and neck cancer. Mol Pharmacol, 76, 969-83. JACKSON, R. G., RYLOTT, E. L., FOURNIER, D., HAWARI, J. & BRUCE, N. C. 2007. Exploring the biochemical properties and remediation applications of the unusual explosive-degrading P450 system XpIA/B. Proc Natl Acad Sci U S A, 104, 16822- 16827. JAKOBY, W. B. & ZIEGLER, D. M. 1990. The enzymes of detoxication. J Biol Chem, 265, 20715-8. JENNEWEIN, S. & CROTEAU, R. 2001. Taxol: biosynthesis, molecular genetics, and biotechnological applications. Appl Microbiol Biotechnol, 57, 13-19. JOYCE, M. G., EKANEM, I. S., ROITEL, O., DUNFORD, A. J., NEELI, R., GIRVAN, H. M., BAKER, G. J., CURTIS, R. A., MUNRO, A. W. & LEYS, D. 2012. The crystal structure of the FAD/NADPH-binding domain of flavocytochrome P450 BM3. FEBS J, 279, 1694-1706.

JOYCE, M. G., GIRVAN, H. M., MUNRO, A. W. & LEYS, D. 2004. A single mutation in cytochrome P450 BM3 induces the conformational rearrangement seen upon substrate binding in the wild-type enzyme. J Biol Chem, 279, 23287-93. KATAGIRI, M., GANGULI, B. N. & GUNSALUS, I. C. 1968. A soluble cytochrome P- 450 functional in methylene hydroxylation. J Biol Chem, 243, 3543-6. KATAGIRI, M., KAGAWA, N. & WATERMAN, M. R. 1995. The role of cytochrome b5 in the biosynthesis of androgens by human P450c17. Arch Biochem Biophys, 317, 343-7. KAVANAGH, M., COYNE, A., JAMES, G., MCLEAN, K., HUDSON, S., DE CARVALHO, L. P., MUNRO, A. & ABELL, C. 2015. Fragment-based drug discovery of potent and selective CYP121 inhibitors for tuberculosis. Abstr Pap Am Chem Soc, 250. KAVANAGH, M. E., COYNE, A. G., MCLEAN, K. J., JAMES, G. G., LEVY, C. W., MARINO, L. B., DE CARVALHO, L. P. S., CHAN, D. S. H., HUDSON, S. A., SURADE, S., LEYS, D., MUNRO, A. W. & ABELL, C. 2016. Fragment-Based Approaches to the Development of Mycobacterium tuberculosis CYP121 Inhibitors. J Med Chem, 59, 3272-3302. KIM, D. H., AHN, T., JUNG, H. C., PAN, J. G. & YUN, C. H. 2009. Generation of the Human Metabolite Piceatannol from the Anticancer-Preventive Agent Resveratrol by Bacterial Cytochrome P450 BM3. Drug Metab Dispos, 37, 932-936. KLINGENBERG, M. 1958. Pigments of rat liver microsomes. Arch Biochem Biophys, 75, 376-86. KUHNEL, K., DERAT, E., TERNER, J., SHAIK, S. & SCHLICHTING, I. 2007. Structure and quantum chemical characterization of chloroperoxidase compound 0, a common reaction intermediate of diverse heme enzymes. Proc Natl Acad Sci U S A, 104, 99-104. LAMB, D. C., FOWLER, K., KIESER, T., MANNING, N., PODUST, L. M., WATERMAN, M. R., KELLY, D. E. & KELLY, S. L. 2002. Sterol 14 alpha- demethylase activity in Streptomyces coelicolor A3(2) is associated with an unusual member of the CYP51 gene family. Biochem J, 364, 555-562. LAMB, D. C. & WATERMAN, M. R. 2013. Unusual properties of the cytochrome P450 superfamily. Philos Trans R Soc Lond B Biol Sci, 368. LE-HUU, P., HEIDT, T., CLAASEN, B., LASCHAT, S. & URLACHER, V. B. 2015. Chemo-, Regio-, and Stereoselective Oxidation of the Monocyclic Diterpenoid beta-Cembrenediol by P450 BM3. ACS Catal, 5, 1772-1780. LE, T. K., JANG, H. H., NGUYEN, H. T. H., DOAN, T. T. M., LEE, G. Y., PARK, K. D., AHN, T., JOUNG, Y. H., KANG, H. S. & YUN, C. H. 2017. Highly regioselective hydroxylation of polydatin, a resveratrol glucoside, for one-step synthesis of astringin, a piceatannol glucoside, by P450 BM3. Enzyme Microb Technol, 97, 34- 42. LEPESHEVA, G. I. & WATERMAN, M. R. 2007. Sterol 14 alpha-demethylase cytochrome P450 (CYP51), a P450 in all biological kingdoms. Biochimica Et Biophysica Acta-General Subjects, 1770, 467-477. LEWIS, D. F., WATSON, E. & LAKE, B. G. 1998. Evolution of the cytochrome P450 superfamily: sequence alignments and pharmacogenetics. Mutat Res, 410, 245-70. LEWIS, D. F. V. 1995. 3-Dimensional Models of Human and Other Mammalian Microsomal P450s Constructed from an Alignment with P450102 (P450(Bm3)). Xenobiotica, 25, 333-366. LEYS, D., MOWAT, C. G., MCLEAN, K. J., RICHMOND, A., CHAPMAN, S. K., WALKINSHAW, M. D. & MUNRO, A. W. 2003. Atomic structure of Mycobacterium tuberculosis CYP121 to 1.06 A reveals novel features of cytochrome P450. J Biol Chem, 278, 5141-7.

LI, H., DAS, A., SIBHATU, H., JAMAL, J., SLIGAR, S. G. & POULOS, T. L. 2008. Exploring the electron transfer properties of neuronal nitric-oxide synthase by reversal of the FMN redox potential. J Biol Chem, 283, 34762-72. LIOU, S. H., MYERS, W. K., BRITT, R. D. & GOODIN, D. B. 2014. Structural Studies of the Docking Complex between Cytochrome P450 (CYP101) and Putidaredoxin (PDX) by Double Electron-Electron Resonance (DEER). Biophys J, 106, 49a-49a. LIPSCOMB, J. D. 1980. Electron paramagnetic resonance detectable states of cytochrome P-450cam. Biochemistry, 19, 3590-9. LOCUSON, C. W., HUTZLER, J. M. & TRACY, T. S. 2007. Visible spectra of type II cytochrome P450-drug complexes: evidence that "incomplete" heme coordination is common. Drug Metab Dispos, 35, 614-22. LOOMIS, W. F. 1988. Four billion years: An essay on the evolution of genes and organisms, Sunderland, Massachsetts, Sinauer Associates, Inc. LUTHRA, A., DENISOV, I. G. & SLIGAR, S. G. 2011. Spectroscopic features of cytochrome P450 reaction intermediates. Arch Biochem Biophys, 507, 26-35. MACDONALD, I. D. G., SLIGAR, S. G., CHRISTIAN, J. F., UNNO, M. & CHAMPION, P. M. 1999. Identification of the Fe-O-O bending mode in oxycytochrome P450cam by resonance Raman spectroscopy. J Am Chem Soc, 121, 376-380. MAK, P. J., DENISOV, I. G., VICTORIA, D., MAKRIS, T. M., DENG, T., SLIGAR, S. G. & KINCAID, J. R. 2007. Resonance Raman detection of the hydroperoxo intermediate in the cytochrome P450 enzymatic cycle. J Am Chem Soc, 129, 6382- 3. MAKITA, K., FALCK, J. R. & CAPDEVILA, J. H. 1996. Cytochrome P450, the arachidonic acid cascade, and hypertension: new vistas for an old enzyme system. FASEB J, 10, 1456-63. MAKRIS, T. M., DAVYDOV, R., DENISOV, I. G., HOFFMAN, B. M. & SLIGAR, S. G. 2002. Mechanistic enzymology of oxygen activation by the cytochromes P450. Drug Metab Rev, 34, 691-708. MAST, N., ANDERSON, K. W., LIN, J. B., LI, Y., TURKO, I. V., TATSUOKA, C., BJORKHEM, I. & PIKULEVA, I. A. 2017. Cytochrome P450 27A1 Deficiency and Regional Differences in Brain Sterol Metabolism Cause Preferential Cholestanol Accumulation in the Cerebellum. J Biol Chem, 292, 4913-4924. MATSON, R. S., HARE, R. S. & FULCO, A. J. 1977. Characteristics of a cytochrome P- 450-dependent fatty acid omega-2 hydroxylase from Bacillus megaterium. Biochim Biophys Acta, 487, 487-94. MATSUNAGA, I., YAMADA, M., KUSUNOSE, E., NISHIUCHI, Y., YANO, I. & ICHIHARA, K. 1996. Direct involvement of hydrogen peroxide in bacterial alpha- hydroxylation of fatty acid. FEBS Lett, 386, 252-4. MATTHEWS, S., BELCHER, J. D., TEE, K. L., GIRVAN, H. M., MCLEAN, K. J., RIGBY, S. E. J., LEVY, C. W., LEYS, D., PARKER, D. A., BLANKLEY, R. T. & MUNRO, A. W. 2017. Catalytic Determinants of Alkene Production by the Cytochrome P450 Peroxygenase OleT(JE). J Biol Chem, 292, 5128-5143. MCINTOSH, J. A., HEEL, T., BULLER, A. R., CHIO, L. & ARNOLD, F. H. 2015. Structural Adaptability Facilitates Histidine Heme Ligation in a Cytochrome P450. J Am Chem Soc, 137, 13861-5. MCLEAN, K. J., HANS, M., MEIJRINK, B., VAN SCHEPPINGEN, W. B., VOLLEBREGT, A., TEE, K. L., VAN DER LAAN, J. M., LEYS, D., MUNRO, A. W. & VAN DER BERG, M. A. 2015a. Single-step fermentative production of the cholesterol-lowering drug pravastatin via reprogramming of Penicillium chrysogenum. Proc Natl Acad Sci U S A, 112, 2847-2852. MCLEAN, K. J., LEYS, D. & MUNRO, A. W. 2015b. Microbial Cytochromes P450 (Chapter 8). In: ORTIZ DE MONTELLANO, P. R. (ed.) Cytochrome P450: Structure, Mechanism and Biochemistry. 4 ed. Switzerland: Springer International. 75

MERILLON, J. M., FAUCONNEAU, B., TEGUO, P. W., BARRIER, L., VERCAUTEREN, J. & HUGUET, F. 1997. Antioxidant activity of the stilbene astringin, newly extracted from Vitis vinifera cell cultures. Clin Chem, 43, 1092- 1093. MILES, J. S., MUNRO, A. W., ROSPENDOWSKI, B. N., SMITH, W. E., MCKNIGHT, J. & THOMSON, A. J. 1992. Domains of the catalytically self-sufficient cytochrome P-450 BM-3. Genetic construction, overexpression, purification and spectroscopic characterization. Biochem J, 288 ( Pt 2), 503-9. MIURA, Y. & FULCO, A. J. 1974. (Omega -2) hydroxylation of fatty acids by a soluble system from Bacillus megaterium. J Biol Chem, 249, 1880-8. MIURA, Y. & FULCO, A. J. 1975. Omega-1, Omega-2 and Omega-3 hydroxylation of long-chain fatty acids, amides and alcohols by a soluble enzyme system from Bacillus megaterium. Biochim Biophys Acta, 388, 305-17. MORTUZA, G. B. & WHITFORD, D. 1997. Mutagenesis of residues 27 and 78 modulates heme orientation in cytochrome b5. FEBS Lett, 412, 610-4. MOTHERSOLE, R. G., MEINTS, C. E., LOUDER, A. & WOLTHERS, K. R. 2016. Role of active site loop in coenzyme binding and flavin reduction in cytochrome P450 reductase. Arch Biochem Biophys, 606, 111-9. MOURO, C., BONDON, A., JUNG, C., HOA, G. H. B., DE CERTAINES, J. D., SPENCER, R. G. S. & SIMONNEAUX, G. 1999. Proton nuclear magnetic resonance study of the binary complex of cytochrome P450cam and putidaredoxin: interaction and electron transfer rate analysis. FEBS Lett, 455, 302-306. MUNRO, A. W., DAFF, S., COGGINS, J. R., LINDSAY, J. G. & CHAPMAN, S. K. 1996a. Probing electron transfer in flavocytochrome P-450 BM3 and its component domains. Eur J Biochem, 239, 403-9. MUNRO, A. W., LEYS, D. G., MCLEAN, K. J., MARSHALL, K. R., OST, T. W., DAFF, S., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MOSER, C. C., PAGE, C. C. & DUTTON, P. L. 2002. P450 BM3: the very model of a modern flavocytochrome. Trends Biochem Sci, 27, 250-7. MUNRO, A. W., LINDSAY, J. G., COGGINS, J. R., KELLY, S. M. & PRICE, N. C. 1994. Structural and enzymological analysis of the interaction of isolated domains of cytochrome P-450 BM3. FEBS Lett, 343, 70-4. MUNRO, A. W., LINDSAY, J. G., COGGINS, J. R., KELLY, S. M. & PRICE, N. C. 1996b. Analysis of the structural stability of the multidomain enzyme flavocytochrome P-450 BM3. Biochim Biophys Acta, 1296, 127-137. MUNRO, A. W., MCLEAN, K. J., GRANT, J. L. & MAKRIS, T. M. 2018. Structure and function of the cytochrome P450 peroxygenase enzymes. Biochem Soc Trans, 46, 183-196. MUNRO, A. W., NOBLE, M. A., MILES, C. S., DAFF, S. N., GREEN, A. J., QUARONI, L., RIVERS, S., OST, T. W., REID, G. A. & CHAPMAN, S. K. 1999. Flavocytochrome P-450 BM3: a paradigm for the analysis of electron transfer and its control in the P-450s. Biochem Soc Trans, 27, 190-6. MURATALIEV, M. B., KLEIN, M., FULCO, A. & FEYEREISEN, R. 1997. Functional interactions in cytochrome P450BM3: flavin semiquinone intermediates, role of NADP(H), and mechanism of electron transfer by the flavoprotein domain. Biochemistry, 36, 8401-12. NAKAHARA, K., TANIMOTO, T., HATANO, K., USUDA, K. & SHOUN, H. 1993. Cytochrome-P-450-55a1 (P-450dnir) Acts as Nitric-Oxide Reductase Employing NADH as the Direct Electron-Donor. J Biol Chem, 268, 8350-8355. NARHI, L. O. & FULCO, A. J. 1986. Characterization of a catalytically self-sufficient 119,000-dalton cytochrome P-450 monooxygenase induced by barbiturates in Bacillus megaterium. J Biol Chem, 261, 7160-9.

NARHI, L. O. & FULCO, A. J. 1987. Identification and characterization of two functional domains in cytochrome P-450BM-3, a catalytically self-sufficient monooxygenase induced by barbiturates in Bacillus megaterium. J Biol Chem, 262, 6683-90. NEBERT, D. W. & GONZALEZ, F. J. 1985. Cytochrome-P450 Gene-Expression and Regulation. Trends in Pharmacological Sciences, 6, 160-164. NEBERT, D. W. & GONZALEZ, F. J. 1987. P450 genes: structure, evolution, and regulation. Annu Rev Biochem, 56, 945-93. NEBERT, D. W., WIKVALL, K. & MILLER, W. L. 2013. Human cytochromes P450 in health and disease. Philos Trans R Soc Lond B Biol Sci, 368, 20120431. NEELI, R., GIRVAN, H. M., LAWRENCE, A., WARREN, M. J., LEYS, D., SCRUTTON, N. S. & MUNRO, A. W. 2005. The dimeric form of flavocytochrome P450BM3 is catalytically functional as a fatty acid hydroxylase. FEBS Lett, 579, 5582-5588. NELSON, D. & WERCK-REICHHART, D. 2011. A P450-centric view of plant evolution. Plant J, 66, 194-211. NELSON, D. R. 1998. Cytochrome P450 Nomenclature (Chapter 2). In: Cytochrome P450 protocols. Totowa, New Jersey: Humana Press Inc. NELSON, D. R. 2003. Comparison of P450s from human and fugu: 420 million years of vertebrate P450 evolution. Arch Biochem Biophys, 409, 18-24. NELSON, D. R. 2018. Cytochrome P450 diversity in the tree of life. Biochim Biophys Acta, 1866, 141-154. NEWCOMB, M., HALGRIMSON, J. A., HORNER, J. H., WASINGER, E. C., CHEN, L. X. & SLIGAR, S. G. 2008. X-ray absorption spectroscopic characterization of a cytochrome P450 compound II derivative. Proc Natl Acad Sci U S A, 105, 8179-84. NOBLE, M. A., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MACKAY, A. C., REID, G. A., HANZLIK, R. P. & MUNRO, A. W. 1999. Roles of key active-site residues in flavocytochrome P450 BM3. Biochem J, 339 ( Pt 2), 371-9. NTP 2014. NTP. Review of the Styrene Assessment in the National Toxicology Program 12th Report on Carcinogens: Workshop Summary. Washington (DC). OGLIARO, F., COHEN, S., FILATOV, M., HARRIS, N. & SHAIK, S. 2000. The High- Valent Compound of Cytochrome P450: The Nature of the Fe-S Bond and the Role of the Thiolate Ligand as an Internal Electron Donor. Angew Chem Int Ed Engl, 39, 3851-3855. OGURA, H., NISHIDA, C. R., HOCH, U. R., PERERA, R., DAWSON, J. H. & ORTIZ DE MONTELLANO, P. R. 2004. EpoK, a cytochrome P450 involved in biosynthesis of the anticancer agents epothilones A and B. Substrate-mediated rescue of a P450 enzyme. Biochemistry, 43, 14712-21. OLIVER, C. F., MODI, S., PRIMROSE, W. U., LIAN, L. Y. & ROBERTS, G. C. K. 1997. Engineering the substrate specificity of Bacillus megaterium cytochrome P- 450 BM3: hydroxylation of alkyl trimethylammonium compounds. Biochem J, 327, 537-544. OMURA, T. & SATO, R. 1964a. The Carbon Monoxide-Binding Pigment of Liver Microsomes. I. Evidence for Its Hemoprotein Nature. J Biol Chem, 239, 2370-8. OMURA, T. & SATO, R. 1964b. Carbon Monoxide-Binding Pigment of Liver Microsomes. II. Solubilization Purification + Properties. J Biol Chem, 239, 2379- &. ORTEGA UGALDE, S., LUIRINK, R. A., GEERKE, D. P., VERMEULEN, N. P. E., BITTER, W. & COMMANDEUR, J. N. M. 2018. Engineering a self-sufficient Mycobacterium tuberculosis CYP130 by gene fusion with the reductase-domain of CYP102A1 from Bacillus megaterium. J Inorg Biochem, 180, 47-53. OSHIMA, R., FUSHINOBU, S., SU, F., ZHANG, L., TAKAYA, N. & SHOUN, H. 2004. Structural evidence for direct hydride transfer from NADH to cytochrome P450nor. J Mol Biol, 342, 207-217. 77

PAINE, M. J. I., SCRUTTON, N. S., MUNRO, A. W., GUTIERREZ, A., ROBERTS, G. C. K. & WOLF, C. R. 2005. Electron transfer partners of cytochrome P450. In: ORTIZ DE MONTELLANO, P. R. (ed.) Cytochrome P450: Structure, Mechanism and Biochemistry. 3 ed. Springer. PANG, S. & WILLIS, L. 1997. Final report on the safety assessment of Cetrimonium Chloride, Cetrimonium Bromide, and Steartrimonium Chloride. Int J Toxic, 16, 195-220. PARK, S. H., KANG, J. Y., KIM, D. H., AHN, T. & YUN, C. H. 2012. The Flavin- Containing Reductase Domain of Cytochrome P450 BM3 Acts as a Surrogate for Mammalian NADPH-P450 Reductase. Biomol Ther, 20, 562-568. PERERA, R., SONO, M., SIGMAN, J. A., PFISTER, T. D., LU, Y. & DAWSON, J. H. 2003. Neutral thiol as a proximal ligand to ferrous heme iron: Implications for heme proteins that lose cysteine thiolate ligation on reduction. Proc Natl Acad Sci U S A, 100, 3641-3646. PETERSON, D. H., MURRAY, H. C., EPPSTEIN, S. H., REINEKE, L. M., WEINTRAUB, A., MEISTER, P. D. & LEIGH, H. M. 1952. Microbiological Transformations of Steroids .1. Introduction of Oxygen at Carbon-11 of Progesterone. J Am Soc Chem, 74, 5933-5936. PODUST, L. M., BACH, H., KIM, Y., LAMB, D. C., ARASE, M., SHERMAN, D. H., KELLY, S. L. & WATERMAN, M. R. 2004. Comparison of the 1.85 A structure of CYP154A1 from Streptomyces coelicolor A3(2) with the closely related CYP154C1 and CYPs from antibiotic biosynthetic pathways. Protein Sci, 13, 255- 268. POULOS, T. L., FINZEL, B. C., GUNSALUS, I. C., WAGNER, G. C. & KRAUT, J. 1985. The 2.6-A crystal structure of Pseudomonas putida cytochrome P-450. J Biol Chem, 260, 16122-30. POULOS, T. L., FINZEL, B. C. & HOWARD, A. J. 1987. High-resolution crystal structure of cytochrome P450cam. J Mol Biol, 195, 687-700. POULOS, T. L. & HOWARD, A. J. 1987. Crystal structures of metyrapone- and phenylimidazole-inhibited complexes of cytochrome P-450cam. Biochemistry, 26, 8165-74. POULOS, T. L. & JOHNSON, E. F. 2015. Structures of Cytochrome P450 Enzymes In: ORTIZ DE MONTELLANO, P. R. (ed.) Cytochrome P450: Structure, Mechanism and Biochemistry. 4 ed. Switzerland: Springer International Springer. PUTKARADZE, N., LITZENBURGER, M., ABDULMUGHNI, A., MILHIM, M., BRILL, E., HANNEMANN, F. & BERNHARDT, R. 2017. CYP109E1 is a novel versatile statin and terpene oxidase from Bacillus megaterium. Appl Microbiol Biotechnol, 101, 8379-8393. RAVICHANDRAN, K. G., BODDUPALLI, S. S., HASEMANN, C. A., PETERSON, J. A. & DEISENHOFER, J. 1993. Crystal-Structure of Hemoprotein Domain of P450bm-3, a Prototype for Microsomal P450s. Science, 261, 731-736. REN, X., YORKE, J. A., TAYLOR, E., ZHANG, T., ZHOU, W. & WONG, L. L. 2015. Drug Oxidation by Cytochrome P450BM3 : Metabolite Synthesis and Discovering New P450 Reaction Types. Chemistry, 21, 15039-47. RENNELLA, E., CUTUIL, T., SCHANDA, P., AYALA, I., FORGE, V. & BRUTSCHER, B. 2012. Real-Time NMR Characterization of Structure and Dynamics in a Transiently Populated Protein Folding Intermediate. J Am Soc Chem, 134, 8066- 8069. RITTLE, J. & GREEN, M. T. 2010. Cytochrome P450 compound I: capture, characterization, and C-H bond activation kinetics. Science, 330, 933-7. ROBERTS, G. A., GROGAN, G., GRETER, A., FLITSCH, S. L. & TURNER, N. J. 2002. Identification of a new class of cytochrome P450 from a Rhodococcus sp. J Bacteriol, 184, 3898-908. 78

ROUCK, J. E., BIGGS, B. W., KAMBALYAL, A., ARNOLD, W. R., DE MEY, M., AJIKUMAR, P. K. & DAS, A. 2017. Heterologous expression and characterization of plant Taxadiene-5alpha-Hydroxylase (CYP725A4) in Escherichia coli. Protein Expr Purif, 132, 60-67. RUETTINGER, R. T. & FULCO, A. J. 1981. Epoxidation of unsaturated fatty acids by a soluble cytochrome P-450-dependent system from Bacillus megaterium. J Biol Chem, 256, 5728-34. RUPASINGHE, S., SCHULER, M. A., KAGAWA, N., YUAN, H., LEI, L., ZHAO, B., KELLY, S. L., WATERMAN, M. R. & LAMB, D. C. 2006. The cytochrome P450 gene family CYP157 does not contain EXXR in the K-helix reducing the absolute conserved P450 residues to a single cysteine. FEBS Lett, 580, 6338-6342. SABBADIN, F., JACKSON, R., HAIDER, K., TAMPI, G., TURKENBURG, J. P., HART, S., BRUCE, N. C. & GROGAN, G. 2009. The 1.5-angstrom Structure of XplA-heme, an Unusual Cytochrome P450 Heme Domain That Catalyzes Reductive Biotransformation of Royal Demolition Explosive. J Biol Chem, 284, 28467-28475. SCHLICHTING, I., BERENDZEN, J., CHU, K., STOCK, A. M., MAVES, S. A., BENSON, D. E., SWEET, R. M., RINGE, D., PETSKO, G. A. & SLIGAR, S. G. 2000. The catalytic pathway of cytochrome p450cam at atomic resolution. Science, 287, 1615-22. SEIFERT, A., ANTONOVICI, M., HAUER, B. & PLEISS, J. 2011. An efficient route to selective bio-oxidation catalysts: an iterative approach comprising modeling, diversification, and screening, based on CYP102A1. Chembiochem, 12, 1346-51. SEVRIOUKOVA, I., SHAFFER, C., BALLOU, D. P. & PETERSON, J. A. 1996a. Equilibrium and transient state spectrophotometric studies of the mechanism of reduction of the flavoprotein domain of P450BM-3. Biochemistry, 35, 7058-68. SEVRIOUKOVA, I., TRUAN, G. & PETERSON, J. A. 1996b. The flavoprotein domain of P450BM-3: expression, purification, and properties of the flavin adenine dinucleotide- and flavin mononucleotide-binding subdomains. Biochemistry, 35, 7528-35. SEVRIOUKOVA, I. F., LI, H., ZHANG, H., PETERSON, J. A. & POULOS, T. L. 1999. Structure of a cytochrome P450-redox partner electron-transfer complex. Proc Natl Acad Sci U S A, 96, 1863-8. SEVRIOUKOVA, I. F. & PETERSON, J. A. 1995. NADPH-P-450 reductase: structural and functional comparisons of the eukaryotic and prokaryotic isoforms. Biochimie, 77, 562-72. SEZUTSU, H., LE GOFF, G. & FEYEREISEN, R. 2013. Origins of P450 diversity. Philos Trans R Soc Lond B Biol Sci, 368. SHAFIEE, A. & HUTCHINSON, C. R. 1987. Macrolide Antibiotic Biosynthesis - Isolation and Properties of 2 Forms of 6-Deoxyerythronolide-B Hydroxylase from Saccharopolyspora-Erythraea (Streptomyces, Erythreus). Biochemistry, 26, 6204- 6210. SHAFIEE, A. & HUTCHINSON, C. R. 1988. Purification and Reconstitution of the Electron-Transport Components for 6-Deoxyerythronolide B-Hydroxylase, a Cytochrome-P-450 Enzyme of Macrolide Antibiotic (Erythromycin) Biosynthesis. J Bacteriol, 170, 1548-1553. SHIRO, Y., IIZUKA, T., MAKINO, R., ISHIMURA, Y. & MORISHIMA, I. 1989. N-15 Nmr-Study on Cyanide (C-15n-) Complex of Cytochrome-P-450cam - Effects of D- Camphor and Putidaredoxin on the Iron Ligand Structure. J Am Soc Chem, 111, 7707-7711. SIDDHANTA, U., PRESTA, A., FAN, B. C., WOLAN, D., ROUSSEAU, D. L. & STUEHR, D. J. 1998. Domain swapping in inducible nitric-oxide synthase -

Electron transfer occurs between flavin and heme groups located on adjacent subunits in the dimer. J Biol Chem, 273, 18950-18958. SLESSOR, K. E., FARLOW, A. J., CAVAIGNAC, S. M., STOK, J. E. & DE VOSS, J. J. 2011. Oxygen activation by P450(cin): Protein and substrate mutagenesis. Arch Biochem Biophys, 507, 154-62. SONOMOTO, K., HOQ, M. M., TANAKA, A. & SUKUI, S. 1983. 11-Beta- Hydroxylation of Cortexolone (Reichstein Compound-S) to Hydrocortisone by Curvularia-Lunata Entrapped in Photo-Cross-Linked Resin Gels. Appl Environ Microbiol, 45, 436-443. SOWDEN, R. J., YASMIN, S., REES, N. H., BELL, S. G. & WONG, L. L. 2005. Biotransformation of the sesquiterpene (+)-valencene by cytochrome P450cam and P450BM-3. Org Biomol Chem, 3, 57-64. STILES, A. R., MCDONALD, J. G., BAUMAN, D. R. & RUSSELL, D. W. 2009. CYP7B1: One Cytochrome P450, Two Human Genetic Diseases, and Multiple Physiological Functions. J Biol Chem, 284, 28485-28489. STUEHR, D. J., SANTOLINI, J., WANG, Z. Q., WEI, C. C. & ADAK, S. 2004. Update on mechanism and catalytic regulation in the NO synthases. J Biol Chem, 279, 36167-36170. TEGUO, P. W., DECENDIT, A., KRISA, S., DEFFIEUX, C., VERCAUTEREN, J. & MERILLON, J. M. 1996. The accumulation of stilbene glycosides in Vitis vinifera cell suspension cultures. J Nat Prod, 59, 1189-1191. THEORELL, H. & MAEHLY, A. C. 1950. Untersuchungen an Kunstlichen Peroxydasen. Acta Chemica Scandinavica, 4, 422-434. TOLMAN, K. G. 2000. Thiazolidinedione hepatotoxicity: a class effect? Int J Clin Pract Suppl, 29-34. TRIPATHI, S., LI, H. & POULOS, T. L. 2013. Structural basis for effector control and redox partner recognition in cytochrome P450. Science, 340, 1227-30. VARY, P. S., BIEDENDIECK, R., FUERCH, T., MEINHARDT, F., ROHDE, M., DECKWER, W. D. & JAHN, D. 2007. Bacillus megaterium--from simple soil bacterium to industrial protein production host. Appl Microbiol Biotechnol, 76, 957-67. VERMA, R., SCHWANEBERG, U. & ROCCATANO, D. 2013. Conformational Dynamics of the FMN-Binding Reductase Domain of Monooxygenase P450BM-3. J Chem Theory Comput, 9, 96-105. VITRAC, X., BORNET, A., VANDERLINDE, R., VALLS, J., RICHARD, T., DELAUNAY, J. C., MERILLON, J. M. & TEISSEDRE, P. L. 2005. Determination of stilbenes (delta-viniferin, trans-astringin, trans-piceid, cis- and trans-resveratrol, epsilon-viniferin) in Brazilian wines. J Agric Food Chem, 53, 5664-5669. WAFFO-TEGUO, P., LEE, D., CUENDET, M., MERILLON, J. M., PEZZUTO, J. M. & KINGHORN, A. D. 2001. Two new stilbene dimer glucosides from grape (Vitis vinifera) cell cultures. J Nat Prod, 64, 136-138. WALTHAM, T. N., GIRVAN, H. M., BUTLER, C. F., RIGBY, S. R., DUNFORD, A. J., HOLT, R. A. & MUNRO, A. W. 2011. Analysis of the oxidation of short chain alkynes by flavocytochrome P450 BM3. Metallomics, 3, 369-78. WANG, Y. Q., LIAO, M. X., HOE, N., ACHARYA, P., DENG, C. H., KRUTCHINSKY, A. N. & CORREIA, M. A. 2009. A Role for Protein Phosphorylation in Cytochrome P450 3A4 Ubiquitin-dependent Proteasomal Degradation. J Biol Chem, 284, 5671-5684. WANG, Z. J., RENATA, H., PECK, N. E., FARWELL, C. C., COELHO, P. S. & ARNOLD, F. H. 2014. Improved cyclopropanation activity of histidine-ligated cytochrome P450 enables the enantioselective formal synthesis of levomilnacipran. Angew Chem Int Ed Engl, 53, 6810-3.

WARMAN, A. J., ROBINSON, J. W., LUCIAKOVA, D., LAWRENCE, A. D., MARSHALL, K. R., WARREN, M. J., CHEESMAN, M. R., RIGBY, S. E., MUNRO, A. W. & MCLEAN, K. J. 2012. Characterization of Cupriavidus metallidurans CYP116B1--a thiocarbamate herbicide oxygenating P450-phthalate dioxygenase reductase fusion protein. FEBS J, 279, 1675-93. WERCK-REICHHART, D. & FEYEREISEN, R. 2000. Cytochromes P450: a success story. Genome Biol, 1, REVIEWS3003. WHITEHOUSE, C. J. C., BELL, S. G. & WONG, L. L. 2012. P450(BM3) (CYP102A1): connecting the dots. Chem Soc Rev, 41, 1218-1260. WINZER, T., KERN, M., KING, A. J., LARSON, T. R., TEODOR, R. I., DONNINGER, S. L., LI, Y., DOWLE, A. A., CARTWRIGHT, J., BATES, R., ASHFORD, D., THOMAS, J., WALKER, C., BOWSER, T. A. & GRAHAM, I. A. 2015. Morphinan biosynthesis in opium poppy requires a P450-oxidoreductase fusion protein. Science, 349, 309-312. XIA, C. W., RWERE, F., IM, S., SHEN, A. L., WASKELL, L. & KIM, J. J. P. 2018. Structural and Kinetic Studies of Asp632 Mutants and Fully Reduced NADPH- Cytochrome P450 Oxidoreductase Define the Role of Asp632 Loop Dynamics in the Control of NADPH Binding and Hydride Transfer. Biochemistry, 57, 945-962. YAMAZAKI, H., SHIBATA, A., SUZUKI, M., NAKAJIMA, M., SHIMADA, N., GUENGERICH, F. P. & YOKOI, T. 1999. Oxidation of troglitazone to a quinone- type metabolite catalyzed by cytochrome P-450 2C8 and P-450 3A4 in human liver microsomes. Drug Metab Dispos, 27, 1260-6. YANG, C. L., LIN, C. H., LUO, W. I., LEE, T. L., RAMU, R., NG, K. Y., TSAI, Y. F., WEI, G. T. & YU, S. S. 2017. Mechanistic Study of the Stereoselective Hydroxylation of [2-(2) H1 ,3-(2) H1 ]Butanes Catalyzed by Cytochrome P450 BM3 Variants. Chemistry, 23, 2571-2582. YOSHIOKA, S., TAKAHASHI, S., HORI, H., ISHIMORI, K. & MORISHIMA, I. 2001. Proximal cysteine residue is essential for the enzymatic activities of cytochrome P450cam. Eur J Biochem, 268, 252-9. YU, L., TANG, W., HE, W., MA, X., VASSEUR, L., BAXTER, S. W., YANG, G., HUANG, S., SONG, F. & YOU, M. 2015. Characterization and expression of the cytochrome P450 gene family in diamondback moth, Plutella xylostella (L.). Sci Rep, 5, 8952. ZHANG, H., YOKOM, A. L., CHENG, S., SU, M., HOLLENBERG, P. F., SOUTHWORTH, D. R. & OSAWA, Y. 2018. The full-length cytochrome P450 enzyme CYP102A1 dimerizes at its reductase domains and has flexible heme domains for efficient catalysis. J Biol Chem, 293, 7727-7736. ZHANG, K., SHAFER, B. M., DEMARS, M. D., 2ND, STERN, H. A. & FASAN, R. 2012. Controlled oxidation of remote sp3 C-H bonds in artemisinin via P450 catalysts with fine-tuned regio- and stereoselectivity. J Am Chem Soc, 134, 18695- 704. ZHAO, B., LEI, L., VASSYLYEV, D. G., LIN, X., CANE, D. E., KELLY, S. L., YUAN, H., LAMB, D. C. & WATERMAN, M. R. 2009. Crystal Structure of Albaflavenone Monooxygenase Containing a Moonlighting Terpene Synthase Active Site. J Biol Chem, 284, 36711-36719. ZHAO, B., LIN, X., LEI, L., LAMB, D. C., KELLY, S. L., WATERMAN, M. R. & CANE, D. E. 2008. Biosynthesis of the sesquiterpene antibiotic albaflavenone in Streptomyces coelicolor A3(2). J Biol Chem, 283, 8183-8189.

Chapter 2: Characterization of the Structure and Interactions of the Active form of P450 BM3 using Hybrid Mass Spectrometry Approaches

2.1. Abstract

Flavocytochrome P450 BM3 is a natural fusion enzyme comprised of two major domains: a cytochrome P450 (heme) catalytic domain and a NADPH-cytochrome P450 reductase (CPR) domain containing FAD and FMN cofactors. A structure for the intact enzyme has yet to be determined, although it has been established that the enzyme is active as a dimer. In this study we gain insights into the interface of the P450 and CPR domains, and that of the full-length dimeric WT P450 BM3 using hybrid mass spectrometry techniques, namely native ion mobility mass spectrometry (IM-MS), collision induced unfolding (CIU) and hydrogen-deuterium exchange mass spectrometry (HDX-MS). These methods determine the shape, stoichiometry and how each of the domains interact in the

240 kDa dimeric complex. High coverage (88-99%) was obtained for the intact P450 BM3 enzyme and its component domains. Previously identified protein interaction sites were confirmed, in addition to novel interaction sites that correspond to P450-CPR interactions at the dimeric interface.

2.2. Introduction

Flavocytochrome P450 BM3 (BM3) is a natural fusion enzyme in which a fatty acid-binding cytochrome P450 (heme) domain (~55 kDa) is fused to a NADPH- cytochrome P450 reductase (~65 kDa) domain through a flexible inter-domain linker region (Munro et al., 2002, Govindaraj and Poulos, 1996). P450 BM3 was first isolated from the soil bacterium Bacillus megaterium and was found to hydroxylate a range of

82 different saturated fatty acids with ~12-18 carbon chain lengths at the -1 to -3 positions

(Miura and Fulco, 1975). The CPR domain binds NADPH and passes electrons from this cofactor through the CPR’s FAD and FMN cofactors, and then from the FMN cofactor to the heme iron in the P450 domain of the enzyme (Daff et al., 1997). The consecutive transfer of two single electrons to the P450 heme iron enables the formation of first a ferric-superoxo species, and then a ferric-peroxo species which undergoes two rapid protonation steps followed by a dehydration reaction to form the highly reactive compound

I (FeIV-oxo porphyrin radical cation) species which catalyzes oxygen insertion into the substrate (Rittle and Green, 2010). The P450 BM3 structural arrangement allows for efficient electron transport from NADPH through the FAD, FMN and heme cofactors and results in P450 BM3 having the highest reported catalytic activity for a P450 monooxygenase enzyme (285 s-1 with arachidonic acid substrate) (Noble et al., 1999). In efforts to enhance activities of P450 enzymes, various groups have fused the reductase domain of BM3 to other P450s in order to produce catalytically self-sufficient flavocytochromes with improved catalytic rates (Dodhia et al., 2006). Other homologues of

P450 BM3 have also been characterized (Gustafsson et al., 2004).

Although several crystal structures have been determined for wild-type and mutant forms of the BM3 heme domain, no structures have been solved for the intact CPR domain, or for the full-length (monomeric or dimeric) P450 BM3 protein. However, structures are available for the FAD/NADPH-binding (ferredoxin reductase-like, FAD domain) domain of the BM3 CPR (Joyce et al., 2012), and the structure of the FMN- binding (flavodoxin-like, FMN domain) domain has also been solved as part of a P450

BM3 structure from which the terminal FAD/NADPH-binding domain was removed. In this truncated heme-FMN domain structure, the FMN domain was cleaved from its heme domain, but its structure could still be resolved as part of a structure which contained two heme domains and one FMN domain in each asymmetric unit (Sevrioukova et al., 1999). 83

Intact P450 BM3 is a functional dimer, with the dimeric interface present in the CPR domain (Neeli et al., 2005b). In the FAD/NADPH-binding domain structure, two surface cysteine residues were mutated to prevent adventitious disulfide bridge formation and dimer formation of the FAD/NADPH-binding domain protein. This strategy enabled the successful crystallization of the monomeric form of this domain.

Aside from X-ray crystallographic approaches, several other methods have been used to characterize the structural and spectroscopic properties of P450 BM3 and its component domains. These include techniquessuch as circular dichroism (for secondary structural analysis), Mössbauer spectroscopy (to probe ferryl species in BM3 and

P450cam), resonance Raman (for analysis of heme structure and heme iron coordination) and EPR (for characterization of heme radical species) (Munro et al., 1996, Behan et al.,

2006, Liu et al., 2018, Bui et al., 2012). However, studies of the intact P450 BM3 and its component heme and CPR domains have not been conducive to structural analysis by

NMR techniques, due to the large sizes of these proteins and the paramagnetic nature of the ferric iron in the heme prosthetic group.

Recently, a full-length structure of a P450 BM3 dimer was modelled using data collected from negative stain and 2D electron microscopy (EM) studies. In the EM structural model, the two CPR domains interact with one another in an extended conformation, with the more flexible N-terminal heme domains predicted to interact with the CPR FMN domains at the proximal face of the heme. In this model, the CPR domains are tightly associated while the heme domains can occupy multiple conformations.

However, the exact positions of the heme domains could not be accurately determined and so specific residues or regions of interaction with the CPR FAD-/FMN-binding domains could not be defined (Zhang et al., 2018). These data are in contrast to a solution state small angle X-ray scattering study, in which the human CPR enzyme was shown to exist in 84 two conformations that could be described as “open” and “closed” forms with respect to the packing of the FMN domain against the FAD domain, and in which the FMN domain was described as a ‘ball on a string’ (Ellis et al., 2009). This model is supported by other studies on the P450 BM3 enzyme which suggest that the FMN-binding domain shuttles electrons to the P450 heme iron by a FMN domain movement towards the heme domain, following it accepting electronsfrom the FAD domain. In studies by Neeli et al. and Girvan et al., electron transfer was shown to occur within a P450 BM3 dimer and between the

FMN domain of one of the monomers and the heme domain of the other (Neeli et al.,

2005a, Girvan et al., 2011a). This model is consistent with the properties of the related flavocytochrome nitric oxide synthase (NOS) (Siddhanta et al., 1998). In contrast,

Kitazume et al. suggested a model in which the FAD domain of one monomer reduces the

FMN domain of the other monomer within the CPR(Kitazume et al., 2007). Thus, there are still several questions to be answered regarding the catalytic mechanism and molecular dynamics of the biotechnologically important P450 BM3 enzyme.

In previous studies, we have used a variety of MS techniques including

Hydrogen-Deuterium Exchange Mass Spectrometry (HDX-MS), native MS, and CIU to analyze structural properties and dynamics within proteins that have been difficult to capture by alternative methods, such as X-ray crystallography (Pacholarz et al., 2017). The

HDX-MS technique involves diluting the target protein in deuterated buffer for different lengths of time in order to allow for the exchange between backbone (amide) hydrogens and deuterium from the solvent. The protein is then digested on a protease column and subsequently passed through a series of MS instruments to separate the protein fragments based on their size, mass, charge and shape. The Pikuleva lab published the first HDX-MS data for a P450 enzyme and used these data to identify an allosteric binding site on

CYP46A1 (Anderson et al., 2016). In this paper, we have used HDX-MS to map the

85 protein surface of P450 BM3 in order to gain insights into the dimeric interface(s) of the enzyme.

2.3. Materials and Methods

2.3.1. P450 BM3 expression and purification

WT P450 BM3 and its heme domain were expressed using pET14b (heme domain) and pET15b (intact P450 BM3 protein) vectors, as described in our previous studies (Butler et al., 2013). The plasmids were transformed into the E. coli BL21 (DE3) strain for gene expression. The BM3 heme domain and the intact P450 BM3 enzyme were expressed in TB medium (Formedium, Hunstanton UK) with cell growth for 24 hours at

37°C. Cells were harvested by centrifugation (6000 g, 20 min, 4 °C). The cell pellets were resuspended in ice-cold buffer A (50 mM potassium phosphate [KPi] containing 350 mM

KCl and 10% glycerol, pH 8) containing protease inhibitors (1 tablet per 100 mL, EDTA- free cOmpleteTM tablets, Roche Applied Science, Burgess Hill, UK), and DNase (10 µg mL-1, Merck, Nottingham UK). Cells were lysed by sonication on ice using a Bandelin

Sonopuls instrument at 37% amplitude with 12 pulses for 40 s, and with 60 s breaks between pulses. The cell extract was clarified by centrifugation (4600 g, 60 min, 4°C).

30% w/v ammonium sulfate was added to the clarified extract and contaminant proteins were removed by incubation for 1 hr at 4°C using gentle agitation. Precipitated material was removed using centrifugation (4600 g, 15 minutes, 4°C).

For the intact P450 BM3 and its heme domain, the proteins were purified using affinity chromatography with a nickel-IDA column, followed by a further chromatography step using hydroxyapatite. The clarified cell extracts in both cases were incubated with nickel-IDA resin overnight at 4°C in buffer A. The intact BM3 and heme domain-bound

86 resins were then applied to columns and a stepwise gradient of 10 mM (300 mL), 20 mM

(200 mL) and 200 mM (60 mL) imidazole was applied to elute the proteins. The eluted proteins were dialyzed into 25 mM KPi, pH 6.5 (buffer B) and applied to separate 28 mL columns containing CHT hydroxyapatite type 1 resin (Bio-Rad Laboratories, Watford

UK). A linear gradient of 25-300 mM KPi, pH 6.5 (600 mL) was applied in both cases to fractionate intact BM3 and its heme domain.

The CPR domain was cloned into pET11a and expressed in BL21-Gold (DE3) cells. The BM3 CPR domain was expressed in 2xYT medium with cell growth at 37°C.

CPR gene expression was induced when the OD600 reached 0.8 by addition of 500 µM

IPTG (Melford Laboratories Ltd., Ipswich UK). Thereafter, the growth temperature was lowered to 30°C and cell culture was continued for 24 hours. Cells were harvested and resuspended in 50 mM Tris containing 1 mM EDTA, pH 7.2 (buffer C). Subsequently, the cells were ruptured by sonication and the CPR partially purified using 30% w/v ammonium sulfate, as described above for the intact BM3 and heme domain proteins. The

CPR protein required dialysis before further purification to remove the ammonium sulfate.

The dialysis was completed using buffer C. The CPR was purified using a 150 mL DEAE

Sepharose™ fast flow resin (GE Healthcare Life Sciences, Little Chalfont UK) with a linear gradient of 0-300 mM KCl in buffer C (1500 mL). The protein was dialyzed into 50 mM KPi, pH 7.2 (buffer D) for further purification on a Q-Sepharose™ Fast Flow column.

The CPR protein was then eluted using a gradient of 0-300 mM KCl in buffer D. Finally, the protein was purified using a hydroxyapatite column as described above for the intact

P450 BM3 and heme domain proteins.

Before analysis of the purified proteins, the samples were desalted using a gel filtration column (HiLoad™ GF S200 16/600 Superdex™ 200 pg, GE Healthcare Life

Sciences). 50 mM ammonium acetate, pH 7 (buffer E) was the buffer chosen due to its volatility for use in native MS experiments.

2.3.2. Native MS and IM-MS

On the day of analysis, the protein was further desalted by exchange into 100 mM ammonium acetate (buffer E) using micro Bio-Spin Chromatography columns (Bio-Rad,

Micro Bio-Spin 6 Columns), following the manufacturer’s instructions. The buffer was exchanged three times for the intact P450 BM3 enzyme and once for the heme and CPR domains to ensure adequate protein compatibility for native MS. Native MS and IM-MS data were acquired on a Synapt G2S HDMS instrument (Waters, Manchester UK).

NanoESI (nESI) capillaries were prepared in-house from thin-walled borosilicate capillaries (inner diameter 0.9 mm, outer diameter 1.2 mm, World Precision Instruments,

Stevenage UK) using a Flaming/Brown P-97 micropipette puller (Sutter Instrument

Company, Novato, CA, USA). A positive voltage was applied to the solution through a platinum wire (Goodfellow Cambridge Ltd., Huntington, UK) inserted into the capillary.

Gentle source conditions were applied to preserve the native-like structure: capillary voltage 1.2 – 1.5 kV, sampling cone 40 V, source temperature 40°C. Trap collision energy was 4 V and transfer collision energy was set to 0 V. IM-MS experiments were acquired using a modified Waters Synapt G2 with a 25.05 cm RF-confining linear drift cell filled with ~2 Torr helium, 298 K. A multi-field approach (20 V drift voltage increments) was applied using WREnS and in-house software (ORIGAMI) (Migas et al., 2018). Capillary voltage 1.3 – 1.6 kV, cone 60 V and source temperature 40°C. The arrival time distributions (ATDs) were converted to collision cross-section (CCS). Measurements were performed in helium as well as in nitrogen as the carrier gas. External calibration of the spectra was achieved using solutions of cesium iodide (2 mg/mL in 50:50 water:isopropanol). Data were acquired and processed with MassLynx software (Waters).

2.3.3. Collision-induced unfolding (CIU)

Experiments were performed on a Waters Synapt G2S instrument using nESI and trap-activated ion mobility; capillary voltage 1.2 – 1.5 kV, cone 40 V and source temperature 40°C. The helium cell and the IMS gas flows were 180 and 90 mL/min, respectively; the IMS wave velocity was 400 m/s, and the IMS wave height was 35 V.

Nitrogen was the carrier gas. The most intense charge state for each protein species was mass selected using the quadrupole prior to the trap region. Activation was induced by elevating the trap collision energy. ORIGAMI (Migas et al., 2018) was used to automatically acquire data for collision energies from 4 – 200 V in 2 V increments, as well as for data processing.

2.3.4. Calculating CCS values from X-ray crystallographic structures

Structures were first energy minimized in the gas phase (with an infinite non- bonded cut-off) using ambertools 14 (Case et al., 2005) and the Amber14 force field

(Maier et al., 2015). CCS values were then calculated using the trajectory method (Mesleh et al., 1996) as implemented in the IMoS suite (Larriba-Andaluz et al., 2015, Larriba and

Hogan, 2013b, Larriba and Hogan, 2013a). The CCS values were calculated for the crystal structure 1BU7 for the BM3 heme domain. For the CPR domain, a model was created using 1BVY (the BM3 FMN domain) and 4DQK (the BM3 FAD domain) aligned to

1AMO (the rat CPR structure), as no intact BM3 CPR structure has been solved to date.

2.3.5. Hydrogen-deuterium exchange mass spectrometry (HDX-MS)

The same protein stocks for native MS were also used for HDX-MS. The protein was incubated with appropriate ligands for a few hours on ice; specifically 0.5 mM N- 89 palmitoylglycine (NPG) and/or 1 mM nicotinamide adenine dinucleotide phosphate

(NADP+), which bind in the P450 active site and to the FAD domain, close to the FAD cofactor, respectively. NPG and NADP+ were purchased from the Cayman Chemical

Company (Ann Arbor, Michigan, USA) and Bio Basic Canada Inc. (Markham, Canada).

Samples were frozen at -80°C until required.

The HDX-MS set-up was comprised of a Waters nano-Acquity UPLC system with ESI MS detection coupled to a LEAP Technologies dual-armed robot for sample preparation, incubation and inlet injection. A Waters Synapt G2S mass spectrometer was operated in positive ion/resolution mode, with data acquired over the m/z range 290-2500.

30 μM protein solutions were diluted 20-fold into 10 mM KPi in either H2O or D2O, pH/pD 7, and the mixture incubated at 20°C for 0 minutes (in H2O), or for 1, 10, 30 or 120 minutes (in D2O), before the quench step. HDX quenching was achieved by mixing the reaction solution 1:1 with cooled 100 mM KPi (pH 2.5, 0°C). ~37.5 pmol were then injected into the HDX module (0°C), and washed over a pepsin column (Waters Enzymate

-1 BEH Pepsin, 2.1 x 30 mm) with 0.1% HCOOH in H2O, pH 2.5, at 200 μL min . Resulting peptides were trapped on a VanGuard C18 trap column. Peptide separation was achieved on a C18 column (Waters Acquity UPLC BEH C18 1.7 µm, 1.0 x 10 mm) at 40 µL/min flow over 16 mins with the following gradient: 0 min: 5% B, 7 min: 35% B, 8 min: 85% B,

11 min: 5% B, 12 min: 95% B, 13 min: 5% B, 14 min: 95% B, 15 min: 5% B (mobile phases: A, water + 0.1 % formic acid; and B, acetonitrile + 0.1 % formic acid). The mass spectrometer was operated in ToF only mode. LeuEnk peptide was used as Lock Spray.

Data were acquired using Waters MassLynx software v4.1, with the LEAP robot controlled by HDx Director 1.0.3.9. Data processing and analysis were carried out with Waters

ProteinLynx Global Server 3.0.1 and Waters DynamX 3.0 software, respectively. Python scripts of the deuterium uptake were generated using the data collected at 1 hour. Python scripts were mapped on to structures in PyMOL (The PyMOL Molecular Graphics System, 90

Version 1.7.4.5 Schrödinger, LLC). High coverage was generated for each state; full- length BM3 vs heme domain (91.5%), full-length BM3 vs CPR domain (89.8%), full- length BM3 ligand-free vs full-length BM3 ligand-bound (99.1%).

2.4. Results

2.4.1. Native mass spectrometry

Native MS was undertaken before CIU to determine the quality of the protein, the stoichiometry of ligand binding and to identify which charge state was the most prevalent for CIU and collision cross-section (CCS) determination. This work confirmed that the full-length protein exists predominantly as a dimer in solution (Figure S1A). The monomeric heme domain exhibited the same three charged species in its ligand-free and

NPG-bound states, with the 14+ state being the most prominent (Figure S1B/C). For the

CPR (diflavin reductase) domain, multiple species were observed, corresponding to one, two, three and four molecules of NADP+ bound to the reductase domain (Figure S1D/E).

2.4.2. Collision-induced unfolding and ion-mobility mass spectrometry

Collision-induced unfolding (CIU) is a technique where increasing voltages are used to observe protein unfolding, similar to methods in which increasing temperature is used to unfold the protein, such as differential scanning calorimetry (DSC) (Butler et al.,

2013). The unfolding intermediates are also visualized using this method. CIU was undertaken with ligand-free domains or full-length dimeric protein (Figure 12A/B/D) and ligand-bound domains (Figure 12C/E/F). The ligand-free full-length dimeric protein showed 6 discrete unfolding events (Figure 12A). The heme domain showed 3 discrete unfolding intermediates (Figure 12B), whilst the reductase (CPR) domain undergoes unfolding across a narrow range of overlapping, unfolding intermediates (Figure 12D). As

Figure 12: Collision-induced unfolding of the P450 BM3 domains and the full-length P450 BM3 enzyme CIU was undertaken on P450 BM3 using the most prevalent charge states found in the native mass spectrometry shown in Figure S1. Panel A: The ligand-free, intact P450 BM3 using charge state [33+] shows 6 unfolding events. Panel B: The ligand-free BM3 heme domain using charge state [14+] shows 3 unfolding events. Panel C: The NPG-bound BM3 heme domain using charge state [14+] shows an increase in stability of the protein, as slightly more energy is required before unfolding is initiated. Panel D: The BM3 reductase (CPR) domain using charge state [16+] shows a range of overlapping intermediates. Panels E/F: The BM3 CPR domain bound to NADP+ using charge state [16+] shows no change to the unfolding pattern that was observed with the ligand-free CPR domain. The stoichiometry used was 1:1 (Panel E) and 1:2 (Panel F) protein to NADP+.

92 stated previously, the CPR module contains two domains: the FAD-binding domain and the FMN-binding domain, yet the CPR domain exhibits a pattern more suggestive of a single domain structure. In contrast, the heme domain behaves more like a multi-domain structure during CIU experiments, due to the presence of 3 discrete unfolding events. The full-length dimeric protein and its domains remain stable in the unfolded states up to 200 V

TW as there is no loss of signal. The collision cross-section ( CCSHe) values derived from

IM-MS were 34.6 ± 0.2, 37.5 ± 1.1 and 99.7 ± 1.2 /nm2 for the heme domain, dimeric CPR domain, and full-length dimeric protein respectively. As mentioned previously, a CPR model structure was produced using the FMN-binding domain structure (1BVY) and the

FAD-binding domain (4DQK) aligned to NADP+-bound rat CPR (1AMO). This model was used throughout the work described here and will be referred to as the reductase model from this point forward. These structures gave CCS values of 39.1 nm2 for the heme domain (13% difference to the experimental data) and 46.1 nm2 for the reductase domain model (23% difference to the experimental data).

The binding of NPG to the heme domain slightly increased the unfolding energy barrier of the heme domain (Figure 12C). The undefined edges are due to the ligand dissociating at high voltages. The presence of NADP+ did not appear to have any significant effect on the unfolding energy barrier of the CPR domain (Figures 12E/F).

2.4.3. Hydrogen-deuterium exchange mass spectrometry

HDX-MS is a technique that enables visualization of the solvent accessibility of amino acid residues to determine protein structure, ligand binding and protein dynamics.

Here we have performed experiments comparing two states to each other, for example by contrasting a full-length protein with a truncated version, or by comparing a ligand-free protein with a ligand-bound protein in order to determine which

Figure 13: HDX-MS of the surface of the isolated BM3 domains in comparison to the full-length dimeric protein Comparisons of the isolated heme and CPR domains to the full-length dimeric BM3 shows areas of increased shielding (cyan and dark blue) and solvent accessibility due to deshielding (orange and red). Residues with no coverage are displayed in grey. The crystal structures used are 1BU7 (heme domain) and 1BVY/4DQK (FMN domain and FAD/NADPH-binding domain, respectively) aligned to the rat CPR 1AMO structure as a CPR model. Panels A/B: Comparing the truncated heme domain to the heme domain of the full-length dimeric protein shows some areas of change. The area with the greatest shielding (dark blue) when the reductase domain is present is from residues 22-30 on the distal face of the protein, seen clearly in panel A. The area of greatest solvent accessibility centres around glutamic acid 424 (red). Panel D/E: The CPR domain exhibits a highly dynamic structure, as shown by large changes in deuterium uptake. Many areas within the intact P450 BM3 protein are deshielded, leading to increasing deuterium exchange (orange-red). In particular, a portion of the hinge domain of the FAD-binding domain 94

(residues 721-729 in yellow and residues 732-748 in orange) and the corresponding shielded areas (residues 709-720, 766-767 and 773-781 in dark blue) suggest significant protein movement. Panel C/F: Butterfly plots show the difference in relative deuterium uptake for each peptide fragment for the heme domain and CPR domain, respectively. The upper states in both plots correspond to the full-length protein and the lower states to the isolated heme (Panel C) and CPR (Panel F) domains. Multiple time points for deuterium uptake are observed for each peptide; 1 minute (orange), 10 minutes (red), 30 minutes (grey), 60 minutes (light blue), 180 minutes (dark blue) and 480 minutes (black). Vertical lines enclose regions of greatest change, as exhibited in Panels A/B/D/E, with blue corresponding to shielding and red corresponding to deshielding for the full-length dimeric protein compared to its domains.

areas become more shielded (less deuterium uptake) or more solvent accessible/deshielded

(more deuterium uptake) on conformational rearrangement. Figure 13 compares the heme and CPR domains of P450 BM3 to the dimeric full-length BM3, showing areas of increased shielding (green to blue) and solvent accessibility due to deshielding (yellow to red). Panels A and B of Figure 13 show the differences between the heme domain of the intact P450 BM3 and the isolated heme domain mapped onto 1BU7 and rotated by ~180 degrees. A large area of Figure 13B in the heme domain from the intact P450 BM3 shows a slight increase in shielding (cyan), including residues 279-315 (encompassing the I-, J- and J’-helices, and the β5 sheet). Other areas that see changes include the F-helix (residues

178-185) and beta sheets (residues 432-445 in the β3/4 sheets) in cyan. The area with the greatest shielding (dark blue) when the CPR domain is present is from residues 22-30 in the heme domain A-helix, as seen clearly in Figure 13A. These areas are potential sites for forming the interface between the CPR and heme domains. Interestingly, there are certain areas (shown in Figure 13A) that have increased solvent accessibility when the CPR domain is present in the protein. In particular, glutamic acid 424 (red) and the residues surrounding it (yellow). Figure 13C shows the butterfly difference plot for the HDX-MS data, highlighting areas of greatest change for the heme domain in blue (indicating increased shielding) and red (indicating increased deshielding). Similarly, the greatest changes observed for the CPR domain, by comparing the truncated domain to the full- length protein, are highlighted in Figure 13F. 95

Figure 13D/E shows that the CPR domain exhibits greater changes in the uptake of deuterium than for intact P450 BM3 or the heme domain, as observed for the blue and red regions of the protein. These data have been mapped onto the CPR model structure

(1BVY/4DQK aligned to 1AMO). Many areas within the intact P450 BM3 protein are deshielded, leading to increased deuterium exchange (orange-to-red). In particular, a portion of the hinge domain (linking the FAD and FMN domains, including residues 721-

729 in yellow and residues 732-748 in orange), a region adjacent to the NADPH-binding site of the FAD-binding domain (residues 967-974 in orange and residues 980-990 in yellow) and a portion of the FMN domain (residues 593-598 in red). These data show how flexible the CPR domain is and how much it moves in the presence of the heme domain within the intact P450 BM3 protein. Examination of the shielded areas for the interaction sites of the two domains reveals a number of interesting sites, indicated in dark blue in

Figure 13D/E. In the FAD-binding domain, three areas are identified in the hinge domain: residues 709-720, 766-767 and 773-781. In the FMN domain, the residues most affected are 559-566 (cyan), Thr592 (dark blue) and 599-601 (dark blue). The combination of shielded and deshielded areas in the CPR hinge domain suggests a dramatic protein movement, allowing one side to become more solvent accessible and the other more shielded.

Figures 14 and 15 compare the effects of ligand binding to the intact P450 BM3 protein. As no full-length structure exists presently, the data have been mapped onto the ligand-bound heme domain crystal structure 1JPZ and the CPR model used previously

(1BVY/4DQK aligned to 1AMO). The ligands used for these experiments were the substrate NPG (which binds within the active site) and nicotinamide adenine dinucleotide

Figure 14: Binding of ligands to intact P450 BM3 elicits structural rearrangements across the heme domain as visualized by HDX MS Comparisons of the isolated P450 BM3 heme domain with the intact P450 BM3 enzyme in substrate-free or ligand-bound forms reveal areas of increased shielding (cyan, dark blue and purple) and solvent accessibility due to deshielding (orange and red). The crystal structure used is that of the NPG-bound heme domain (1JPZ). Panels A/B: Comparisons of the ligand-free heme domain with that of the NADP+-bound protein reveal very little change in the heme domain structure. Panels C/D: Comparisons of the ligand-free heme domain with the NPG-bound protein show that the majority of the protein has undergone deshielding, as revealed by the widespread orange colouring. Panels E/F: Comparisons of the ligand-free heme domain to the NADP+/NPG-bound protein show slight differences across the heme domain compared to the NPG-bound form (panels C/D).

phosphate (NADP+, which binds adjacent to the FAD cofactor of the CPR domain). The

NADP+-binding site is immediately adjacent to the FAD isoalloxazine ring, but the oxidized NADP+ cannot reduce the FAD and instead acts as a structural mimic of NADPH.

In Figures 14A/B, the ligand-free protein is compared to the NADP+-bound protein, revealing very little structural change in the heme domain. There is a slight increase in solvent accessibility visible around the N-terminus of the heme domain. In particular, three residues of the A-helix, near the N-terminus, exhibit deshielding in all ligand bound states.

The greatest change to the heme domain was exhibited on NPG binding or NADP+/NPG binding to the intact P450 BM3 in Figures 14C/D and 14E/F, respectively. Both states exhibit deshielding around residue 138 and a portion of the J-helix (residues 294-305).

However, these areas are more extensive for the NPG-bound protein. No significant shielding is observed for the NPG-bound protein. However, deshielding is observed across the protein upon NPG-binding in the B-, B’-, C-, C’-, D-, E-, F-, I-, J-, J’- and K-helices. In addition, a beta sheet at the C-terminus of the heme domain exhibits significant deshielding. However, the greatest deshielding is observed in the E-helix (residues 149-

155 in red). In contrast, the NADP+/NPG-bound protein exhibits fewer deshielded areas, many identical to the NPG-bound protein. This differs for the large deshielded area containing much of the L-helix, observed on NADP+/NPG-binding. Interestingly, shielding is observed for the NADP+/NPG-bound protein near the N-terminus (residues 15-20 in purple) and the helices and random coil segments between the F- and G-helices (residues

187-197 in cyan).

In Figure 15, the CPR domain exhibits greater surface changes than does the heme domain on ligand binding. Comparing the ligand-free CPR domain to the NADP+- bound form reveals very little change in solvent accessibility across the CPR domain,

Figure 15: Binding of ligands to intact P450 BM3 elicits structural rearrangements across the CPR domain as visualized by HDX MS Comparisons of the CPR domain with the intact P450 BM3 protein in the substrate-free or ligand-bound forms reveal areas of increased shielding (cyan, dark blue and purple) and solvent accessibility due to deshielding (orange and red). The crystal structure used is the rat CPR model produced by aligning the BM3 FMN domain from the 1BVY structure and the BM3 FAD-binding domain from the 4DQK structure to the rat CPR structure (1AMO). Panels A/B: Comparisons of the ligand-free CPR to the NADP+-bound protein revealed little difference in solvent accessibility, except for one area of deshielding within the FAD domain. Panels C/D: Comparisons of the ligand-free CPR with the NPG-bound form of the intact P450 BM3 protein show that much of the CPR domain has undergone a change to orange and red, indicating greater solvent accessibility that is likely due to the CPR domain adopting a less compact conformation. The greatest decrease in deuterium uptake is seen on the hinge domain of the FAD-binding domain. In comparison, the greatest increase in deuterium uptake for the FAD-binding domain is located on the opposite face

99 of the hinge domain (purple). Panels E/F: Binding of NPG elicits the greatest change across the CPR domain, whether NPG only or NADP+/NPG were bound, as observed in panels C/D/E/F. There is a greater shift from grey to orange indicating a larger conformational reorganization. The areas with the greatest shielding observed are present in the NPG- and NADP+/NPG-bound protein.

shown in grey (Figure 15A/B). There is also a slight increase in shielding in the FAD domain for residues 1007-1011 (cyan).Surprisingly, binding of NPG had a greater effect on the conformational changes exhibited in the CPR domain compared to those induced by the binding of the CPR domain ligand NADP+, as seen in Figures 15C/D (NPG-bound) and 15E/F (NADP+/NPG-bound). The figures illustrate that the CPR domain has undergone many structural changes, including extensive deshielding of the FMN-binding domain, which is further amplified in the NADP+/NPG-bound structure. In particular, as evidenced by orange segments observed in the FMN domain (residues 559-567) and in the

FAD domain (residues 711-720 and 822-829). In contrast, shielding is observed for the

FAD-binding domain in two areas; residues 967-974 (in cyan) and residues 787-792. The latter area is located on the opposite face to the greatest deshielded area of the protein

(residues 713-718). For the NADP+/NPG-bound CPR structure, there is an additional increase in solvent accessibility in two areas: residues 1029-1034 of the FAD domain (red) and residues 559-564 of the FMN domain.

2.5. Discussion

Comparison of the experimental CCS values to the CCS values calculated from the crystal structures showed slight differences for the heme and CPR domains. Due to the lack of a full-length crystal structure of intact P450 BM3, no CCS values could be predicted in this case. The heme domain CCS value discrepancy is within the 14% range described as an acceptable error in the literature (Jurneczko and Barran, 2011). However, the CCS value for the CPR domain is outside this error range, suggesting that our model may not be a highly accurate representation of the CPR in solution. Our experimental data 100 suggest that the P450 BM3 CPR domain adopts a more compact conformation than rat

CPR when in its ligand-free form.

From thermal denaturation (differential scanning calorimetry, DSC) experiments, the intact P450 BM3 was found to have three distinct unfolding events, with the heme domain having two unfolding events and the CPR domain a single unfolding event (Munro et al., 1996, Butler et al., 2013). Our CIU data suggest additional unfolding events for the intact P450 BM3 and the heme domain. For intact P450 BM3, three unfolding events are very clearly defined with prolonged drift time over a large range of collision energies.

There are also three much shorter, overlapping unfolding events which might make these events difficult to identify using techniques such as DSC. A similar overlapping series of unfolding events is observed in the heme domain CIU experiments. Other experiments also confirmed the stabilization of the heme domain upon substrate binding, with the tight- binding NPG increasing the major heme domain melting temperature by ~4.7 °C (Butler et al., 2013).

Comparisons between the intact P450 BM3 protein and its heme domain reveal many areas of shielding, including those on the proximal surface of the protein where the redox partner (CPR) domain interacts to catalyze electron transfer to the P450 heme iron.

Gricman et al. and others have studied P450-redox partner interactions in P450 BM3 using the BM3 heme-FMN domain crystal structure (1BVY). However, this structure is formed with a non-stoichiometric ratio of two heme domains to one FMN domain, with the single

FMN domain cleaved from its heme domain and positioned distant from the heme domains. Computational analysis indicated that each electron transfer from FMN-to-heme in the conformation observed in the crystal structure would take ~50 years to complete

(Gricman et al., 2016, Munro et al., 2002, Sevrioukova et al., 1999). Modelling of the interactions between the BM3 heme domain and the FMN domain identifies larger regions 101 of interest from the 1BVY structure. In particular, molecular modelling reveals a number of residues and regions that are highly mobile, resulting in three collective movements that lead to a more compact conformation of the heme-FMN complex that would be required for a high catalytic rate of FMN-to-heme iron electron transfer to be achieved (Verma et al., 2014).

Comparing the identified residues with the HDX-MS approach shows that they align to the highly dynamic areas in the heme domain, as observed by comparing the heme domain to the intact P450 BM3, shaded red and cyan in Figure 13 (Verma et al., 2014).

Darimont et al. mutated these residues (found within the K- and L-helices, as well as random coil segments, of the heme domain and M490 of the FMN-binding domain) and found that electron transfer coupling efficiency and enzyme activity could be greatly altered by engineering these residues, but only when the redox reaction was driven by

NADPH (the physiological cofactor), and not when an electrode system was used

(Darimont et al., 2018). Our data reveal areas that are not explained by these collective movements, but which exhibit some of the greatest changes in deuterium uptake. This suggests that other factors may contribute, such as shielding and deshielding by structural changes in the dimeric interface. These areas include residues 22-30, which have the greatest shielding observed across the heme domain surface, and a large amino acid stretch from 294-305 (within the I- and J-helices). Through modelling the surface charge using

PyMOL, the segment containing amino acids 22-30 was shown to have modest positive charge, while the larger segment from amino acids 294-305 exhibits a small negative charge. The shielded regions are adjacent to each other and close to other deshielded areas, suggesting a movement in this area when comparing intact P450 BM3 with the heme domain protein, which may be attributed to the conformational changes during dimerization.

102

For the intact P450 BM3, the heme domain did not exhibit structural changes as dramatic as those observed for the CPR domain on ligand binding. Comparisons of the data described here to those for the human cholesterol 24-hydroxylase CYP46A1(the only other published example of P450 HDX-MS data) revealed similar changes in deuterium uptake (Anderson et al., 2016). Areas of shielding were formed near the N-terminus on

NADP+/NPG binding, in the same region as observed when comparing the full-length dimeric protein to the heme domain. In contrast, the other area of shielding (in the J-helix) showed a slight deshielding on NPG binding. Also of interest is the consistent deshielding of Thr149 in the E-helix observed on NADP+ binding. This residue is not well represented in the literature and no mutagenesis studies have targeted it to date.

Unfortunately, there are no crystal structures for P450 BM3 in which the catalytically important heme-FMN linker region is intact, and so we were unable to map our HDX data to this critically important, but likely highly dynamic, region. We were able to map the majority of the CPR domain data using the CPR model (1BVY/4DQK aligned to 1AMO), revealing a number of shielded areas throughout this domain. However, our focus was drawn towards the FMN domain, as this protein module is known to be responsible for shuttling NADPH-derived electrons between the BM3 CPR and heme domains (Girvan et al., 2011b, Kitazume et al., 2007, Neeli et al., 2005a). The highly dynamic regions of the FMN domain determined by modelling 1BVY were found to be

Lβ2/3 and Lα2 (Verma et al., 2014). Comparisons between the CPR domain and the intact

P450 BM3 revealed that these highly dynamic beta sheets exhibited profound shielding, whilst the alpha helix had slightly increased deshielding. In particular, there is a substantially shielded area adjacent to a highly solvent accessible area of the FMN domain from residues 592-602. This corresponds to a highly negatively charged area of the protein.

103

Within the CPR domain, lysine residues 778 and 791, previously described as key residues involved in forming the dimeric interface from the EM structure, sit either side of the most shielded area in the hinge region of the FAD domain when comparing intact P450

BM3 to the CPR domain (Zhang et al., 2018). When comparing NPG-bound to ligand-free data, the area of greatest shielding within the hinge domain becomes slightly shifted in location and is focused around lysine 791 (residues 787-792). Interestingly, the other side of this hinge domain has greatly increased solvent accessibility, indicating substantial protein movement in this area. The cysteine residues that affect the crystallization of the

FAD-binding domain display reduced deuterium uptake corresponding to a highly shielded area when comparing intact P450 BM3 to the CPR domain (Figure 13D/E). A C773A mutation was originally introduced to prevent dimerization of the isolated FAD-binding domain and to facilitate its crystallization, and is located in the CPR hinge domain region

(Joyce et al., 2012). Coverage for this residue was missing for the ligand-bound forms.

However, where coverage was available near this residue, shielding was observed. The areas of greatest deuterium uptake observed in the CPR domain, particularly from the intact P450 BM3 to CPR domain comparison, are also those regions that were suggested for interactions in the small angle X-ray scattering (SAXS) models for the human CPR

(Ellis et al., 2009). The CPR domain is able to dimerize in solution, suggesting that the shielding we observe in this hinge domain is also consistent with other data in the literature, and is associated with the site of the CPR domain dimerization. Consolidating our domain data with our ligand-bound data suggests that the CPR dimerization interface undergoes a slight rotation on ligand-binding, causing the shielding of residues 787-792 and deshielding of residues 711-720.

NADP+ binding had a small effect on the structure of the BM3 CPR domain, with various shielding and deshielding events observed across the protein surface. The largest changes observed occur in the hinge domain. NPG binding elicits substantial changes 104 across the CPR domain in intact P450 BM3, which are further amplified in the

NADP+/NPG-bound structure. This suggests that binding of a substrate (NPG) to the heme domain induces major conformational changes that are transmitted through to the CPR domain, and that the binding of NADP+ produces other conformational changes that are conducive for catalysis. The areas of greatest deshielding are seen in (i) the hinge domain of the CPR, (ii) within the beta-sheets of the CPR FMN domain, and (iii) in a peripheral alpha helix within the FAD domain (L2). Studies have suggested that NADP+ binding to the CPR domain induces a closed conformation in which the FAD and FMN cofactors are a short distance apart (ca 4 Å) (Xia et al., 2018). On reduction by NADPH, an open conformation is adopted as the FAD/NADPH- and FMN-binding domains move apart to enable FMN domain/heme domain interactions (Xia et al., 2018). An important residue for the dissociation of NADP+ is Ser634. Unfortunately, as no full-length structure has been determined for the BM3 CPR domain, the Ser634 residue is not represented in the CPR model and so the HDX data for Ser634 and its associated loop region cannot be mapped accurately (Xia et al., 2018).

2.6. Conclusions

In this paper we have utilized hybrid mass spectrometry techniques to probe the protein surface of the intact P450 BM3 enzyme and its component heme-binding (P450) and reductase (CPR) domains, in addition to analysis of the protein-protein interactions made by the intact P450 BM3 enzyme and its CPR domain in ligand-free and ligand-bound

(NPG- and/or NADP+-bound) forms. The results provide key data for P450 BM3, a biotechnologically important enzyme. Very high coverage percentages were achieved.

Important information was obtained on various exposed and buried regions of the intact

P450 BM3, as well as new insights into conformational rearrangements that occur on

105

Figure 16: A model of the full-length dimeric P450 BM3 enzymes using data collected from hybrid MS techniques Using the HDX-MS data collected from NADP+/NPG-bound protein compared to the ligand-free enzyme, a number of areas with decreased deuterium uptake are observed. By assuming the areas of greatest shielding are due to the interaction of the protein within the dimeric interface, a model can be produced. Linker region models (shown with curved black lines) were placed at the relevant terminus of each domain. Interactions between the domains are shown with dotted lines. Data were mapped onto the crystal structure 1JPZ for the heme domain, and onto 4DQK and 1BVY aligned to 1AMO for the CPR domain. Colour coding for each domain is displayed in Figure 14 and 15 for the heme domain and CPR domain, respectively.

binding the NPG substrate and NADP+ cofactor to the heme domain and CPR domain, respectively. HDX-MS was also used to probe the CPR domain dimerization site, which is

106 crucial for the formation of the functional P450 BM3 dimer. In addition, other interaction sites in P450 BM3 have been located through determination of reduced deuterium uptake

(shielding) in areas that may correspond to interaction sites for the construction of heme domain-CPR domain interface(s) and for other dimeric interface(s). These include, in particular, residues 15-20 and 187-197 of the heme domain, and residues 967-974 of the

CPR domain. Collectively, these data provide new insights into the molecular dynamics and protein-protein interactions underpinning the catalytic activity of the widely studied and catalytically versatile P450 BM3 enzyme. A model of the full-length dimer and its interactions are proposed in Figure 16, with the linker region placed at the relevant of each domain using the NADP+/NPG-bound data. The greatest deshielding occurs adjacent to the linker region connecting the heme- and FMN-binding domains. With the data acquired, the heme domains interact with the CPR domains of the opposite monomer. No conclusions can be drawn for the path of electron transfer due to the highly flexible nature of the protein.

2.7. References

ANDERSON, K. W., MAST, N., HUDGENS, J. W., LIN, J. B., TURKO, I. V. & PIKULEVA, I. A. 2016. Mapping of the Allosteric Site in Cholesterol Hydroxylase CYP46A1 for Efavirenz, a Drug That Stimulates Enzyme Activity. J Biol Chem, 291, 11876-86. BEHAN, R. K., HOFFART, L. M., STONE, K. L., KREBS, C. & GREEN, M. T. 2006. Evidence for basic ferryls in cytochromes P450. J Am Chem Soc, 128, 11471-4. BUI, S. H., MCLEAN, K. J., CHEESMAN, M. R., BRADLEY, J. M., RIGBY, S. E., LEVY, C. W., LEYS, D. & MUNRO, A. W. 2012. Unusual spectroscopic and ligand binding properties of the cytochrome P450-flavodoxin fusion enzyme XplA. J Biol Chem, 287, 19699-714. BUTLER, C. F., PEET, C., MASON, A. E., VOICE, M. W., LEYS, D. & MUNRO, A. W. 2013. Key mutations alter the cytochrome P450 BM3 conformational landscape and remove inherent substrate bias. J Biol Chem, 288, 25387-99. CASE, D. A., CHEATHAM, T. E., 3RD, DARDEN, T., GOHLKE, H., LUO, R., MERZ, K. M., JR., ONUFRIEV, A., SIMMERLING, C., WANG, B. & WOODS, R. J. 2005. The Amber biomolecular simulation programs. J Comput Chem, 26, 1668- 88. DAFF, S. N., CHAPMAN, S. K., TURNER, K. L., HOLT, R. A., GOVINDARAJ, S., POULOS, T. L. & MUNRO, A. W. 1997. Redox control of the catalytic cycle of flavocytochrome P-450 BM3. Biochemistry, 36, 13816-23.

107

DARIMONT, D., WEISSENBORN, M. J., NEBEL, B. A. & HAUER, B. 2018. Modulating proposed electron transfer pathways in P450BM3 led to improved activity and coupling efficiency. Bioelectrochemistry, 119, 119-123. DODHIA, V. R., FANTUZZI, A. & GILARDI, G. 2006. Engineering human cytochrome P450 enzymes into catalytically self-sufficient chimeras using molecular Lego. J Biol Inorg Chem, 11, 903-16. ELLIS, J., GUTIERREZ, A., BARSUKOV, I. L., HUANG, W. C., GROSSMANN, J. G. & ROBERTS, G. C. 2009. Domain motion in cytochrome P450 reductase: conformational equilibria revealed by NMR and small-angle X-ray scattering. J Biol Chem, 284, 36628-37. GIRVAN, H. M., DUNFORD, A. J., NEELI, R., EKANEM, I. S., WALTHAM, T. N., JOYCE, M. G., LEYS, D., CURTIS, R. A., WILLIAMS, P., FISHER, K., VOICE, M. W. & MUNRO, A. W. 2011. Flavocytochrome P450 BM3 mutant W1046A is a NADH-dependent fatty acid hydroxylase: Implications for the mechanism of electron transfer in the P450 BM3 dimer. Arch Biochem Biophys, 507, 75-85. GOVINDARAJ, S. & POULOS, T. L. 1996. Probing the structure of the linker connecting the reductase and heme domains of cytochrome P450BM-3 using site-directed mutagenesis. Protein Sci, 5, 1389-93. GRICMAN, L., WEISSENBORN, M. J., HOFFMANN, S. M., BORLINGHAUS, N., HAUER, B. & PLEISS, J. 2016. Redox Partner Interaction Sites in Cytochrome P450 Monooxygenases: In Silico Analysis and Experimental Validation. Chemistryselect, 1, 1243-1251. GUSTAFSSON, M. C., ROITEL, O., MARSHALL, K. R., NOBLE, M. A., CHAPMAN, S. K., PESSEGUEIRO, A., FULCO, A. J., CHEESMAN, M. R., VON WACHENFELDT, C. & MUNRO, A. W. 2004. Expression, purification, and characterization of Bacillus subtilis cytochromes P450 CYP102A2 and CYP102A3: flavocytochrome homologues of P450 BM3 from Bacillus megaterium. Biochemistry, 43, 5474-87. JOYCE, M. G., EKANEM, I. S., ROITEL, O., DUNFORD, A. J., NEELI, R., GIRVAN, H. M., BAKER, G. J., CURTIS, R. A., MUNRO, A. W. & LEYS, D. 2012. The crystal structure of the FAD/NADPH-binding domain of flavocytochrome P450 BM3. FEBS J, 279, 1694-1706. JURNECZKO, E. & BARRAN, P. E. 2011. How useful is ion mobility mass spectrometry for structural biology? The relationship between protein crystal structures and their collision cross sections in the gas phase. Analyst, 136, 20-8. KITAZUME, T., HAINES, D. C., ESTABROOK, R. W., CHEN, B. & PETERSON, J. A. 2007. Obligatory intermolecular electron-transfer from FAD to FMN in dimeric P450BM-3. Biochemistry, 46, 11892-901. LARRIBA-ANDALUZ, C., FERNANDEZ-GARCIA, J., EWING, M. A., HOGAN, C. J., JR. & CLEMMER, D. E. 2015. Gas molecule scattering & ion mobility measurements for organic macro-ions in He versus N2 environments. Phys Chem Chem Phys, 17, 15019-29. LARRIBA, C. & HOGAN, C. J. 2013a. Free molecular collision cross section calculation methods for nanoparticles and complex ions with energy accommodation. J Comput Phys, 251, 344-363. LARRIBA, C. & HOGAN, C. J., JR. 2013b. Ion mobilities in diatomic gases: measurement versus prediction with non-specular scattering models. J Phys Chem A, 117, 3887-901. LIU, Y. L., MCLEAN, K. J., MUNRO, A. W. & KINCAID, J. R. 2018. Resonance Raman studies of Bacillus megaterium cytochrome P450 BM3 and biotechnologically important mutants. J Raman Spectrosc, 49, 287-297. MAIER, J. A., MARTINEZ, C., KASAVAJHALA, K., WICKSTROM, L., HAUSER, K. E. & SIMMERLING, C. 2015. ff14SB: Improving the Accuracy of Protein Side 108

Chain and Backbone Parameters from ff99SB. J Chem Theory Comput, 11, 3696- 713. MESLEH, M. F., HUNTER, J. M., SHVARTSBURG, A. A., SCHATZ, G. C. & JARROLD, M. F. 1996. Structural information from ion mobility measurements: Effects of the long-range potential. J Phys Chem, 100, 16082-16086. MIGAS, L. G., FRANCE, A. P., BELLINA, B. & BARRANN, P. E. 2018. ORIGAMI: A software suite for activated ion mobility mass spectrometry (aIM-MS) applied to multimeric protein assemblies. Int J of Mass Spectrom, 427, 20-28. MIURA, Y. & FULCO, A. J. 1975. Omega-1, Omega-2 and Omega-3 hydroxylation of long-chain fatty acids, amides and alcohols by a soluble enzyme system from Bacillus megaterium. Biochim Biophys Acta, 388, 305-17. MUNRO, A. W., LEYS, D. G., MCLEAN, K. J., MARSHALL, K. R., OST, T. W., DAFF, S., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MOSER, C. C., PAGE, C. C. & DUTTON, P. L. 2002. P450 BM3: the very model of a modern flavocytochrome. Trends Biochem Sci, 27, 250-7. MUNRO, A. W., LINDSAY, J. G., COGGINS, J. R., KELLY, S. M. & PRICE, N. C. 1996. Analysis of the structural stability of the multidomain enzyme flavocytochrome P-450 BM3. Biochim Biophys Acta, 1296, 127-37. NEELI, R., GIRVAN, H. M., LAWRENCE, A., WARREN, M. J., LEYS, D., SCRUTTON, N. S. & MUNRO, A. W. 2005a. The dimeric form of flavocytochrome P450BM3 is catalytically functional as a fatty acid hydroxylase. FEBS Lett, 579, 5582-5588. NOBLE, M. A., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MACKAY, A. C., REID, G. A., HANZLIK, R. P. & MUNRO, A. W. 1999. Roles of key active-site residues in flavocytochrome P450 BM3. Biochem J, 339 ( Pt 2), 371-9. PACHOLARZ, K. J., BURNLEY, R. J., JOWITT, T. A., ORDSMITH, V., PISCO, J. P., PORRINI, M., LARROUY-MAUMUS, G., GARLISH, R. A., TAYLOR, R. J., DE CARVALHO, L. P. S. & BARRAN, P. E. 2017. Hybrid Mass Spectrometry Approaches to Determine How L-Histidine Feedback Regulates the Enzyzme MtATP-Phosphoribosyltransferase. Structure, 25, 730-738 e4. RITTLE, J. & GREEN, M. T. 2010. Cytochrome P450 compound I: capture, characterization, and C-H bond activation kinetics. Science, 330, 933-7. SEVRIOUKOVA, I. F., LI, H., ZHANG, H., PETERSON, J. A. & POULOS, T. L. 1999. Structure of a cytochrome P450-redox partner electron-transfer complex. Proc Natl Acad Sci U S A, 96, 1863-8. SIDDHANTA, U., PRESTA, A., FAN, B. C., WOLAN, D., ROUSSEAU, D. L. & STUEHR, D. J. 1998. Domain swapping in inducible nitric-oxide synthase - Electron transfer occurs between flavin and heme groups located on adjacent subunits in the dimer. J Biol Chem, 273, 18950-18958. VERMA, R., SCHWANEBERG, U. & ROCCATANO, D. 2014. Insight into the Redox Partner Interaction Mechanism in Cytochrome P450BM-3 Using Molecular Dynamics Simulations. Biopolymers, 101, 197-209. XIA, C., RWERE, F., IM, S., SHEN, A. L., WASKELL, L. & KIM, J. P. 2018. Structural and Kinetic Studies of Asp632 Mutants and Fully Reduced NADPH-Cytochrome P450 Oxidoreductase Define the Role of Asp632 Loop Dynamics in the Control of NADPH Binding and Hydride Transfer. Biochemistry, 57, 945-962. ZHANG, H., YOKOM, A. L., CHENG, S., SU, M., HOLLENBERG, P. F., SOUTHWORTH, D. R. & OSAWA, Y. 2018. The full-length cytochrome P450 enzyme CYP102A1 dimerizes at its reductase domains and has flexible heme domains for efficient catalysis. J Biol Chem, 293, 7727-7736.

109

2.8. Supplementary Information

Figure S1: Native MS of intact P450 BM3 and its component domains in ligand-free and ligand-bound states Panel A: The native MS spectrum of intact P450 BM3 shows multiple charge states as well as the presence of dimeric (major component) and monomeric (minor component) protein. Panel B: The native MS spectrum of the heme domain protein shows only three prevalent charge states. Panel C: The MS spectrum of the heme domain does not change significantly upon association withof the tight binding BM3 substrate NPG. Panel D: The MS spectrum of the CPR domain shows a number of different charge states. Panel E: The MS spectrum of the CPR domain changes substantially on the binding of NADP+ to the CPR. Multiple species are observed corresponding to 1, 2, 3 and 4 molecules of NADP+ binding to the CPR domain.

110

Experimental Theoretical Expected mass Measured mass

Protein CCS (nm2) STDEV ImoS (nm2) Da Da

Full-length protein 99.66 1.21 - 239889.20 244915.91 ± 76.08

Heme domain 34.56 0.17 39.1 56147.08 56637.51 ± 0.71

Heme domain + NPG 34.71 0.24 - 56406.58 56952.95 ± 1.56

CPR domain 37.53 1.08 46.1 64145.98 64536.33 ± 30.81

CPR domain + 1 NADP+ 37.05 0.38 - 64889.38 65229.05 ± 11.67

CPR domain + 2 NADP+ 37.39 0.15 - 65632.78 66054.80 ± 2.79

CPR domain + 3 NADP+ 38.03 0.48 - 66376.18 66809.45 ± 14.95

CPR domain + 4 NADP+ 38.31 0.43 - 67119.58 67563.77 ± 20.17

Figure S2: Experimentally and theoretically derived collision cross-section (CCS) values for P450 BM3 CCS values were determined for the full-length dimeric P450 BM3 protein and its component heme and reductase (CPR) domains using IM-MS with helium as the carrier gas. In addition, CCS values were determined for ligand-bound domains. Theoretical CCS values were calculated using crystallographic structures (1BU7, 4DQK and 1BVY) using the ImoS suite with helium as the carrier gas. These calculations gave CCS values of 39.1 nm2 for the heme domain (13% difference to the experimental data) and 46.1 nm2 for the reductase domain model (23% difference to the experimental data).

111

Figure S3: Relative deuterium uptake plots for ligand-free and ligand-bound full- length P450 BM3 dimeric protein Using a LEAP Technologies dual-armed robot, accurate deuterium incubations could be accomplished for HDX-MS analysis. Observing the deuterium uptake from 1 minute to 180 minutes for each peptide fragment allows determination of the solvent accessibility of the protein. Four states are compared above; ligand-free (red), NADP+-bound (blue), NPG- bound (green) and NADP+/NPG-bound (purple). Plots were generated using Waters, DynamX software.

112

Figure S4: The coverage maps for the ligand-free P450 BM3 heme and CPR domains Panel A: The heme domain HDX-MS data show 91.5% coverage with 3.18 redundancy. Panel B: The CPR domain data show 89.8% coverage with 3.75 redundancy.

113

Figure S5: The coverage maps for the ligand-bound, full-length P450 BM3 dimeric protein All four states of the protein (ligand-free, NADP+-bound, NPG-bound and NADP+/NPG- bound) show 99.1% coverage and 8.51 redundancy.

114

Chapter 3: The Promiscuous Nature of P450 BM3 and its Ability To Bind Pharmaceutical Compounds Using a Novel Library Screen

3.1. Abstract

P450 BM3 (BM3) is a high activity cytochrome P450-P450 reductase fusion protein best known for its ability to oxidize medium- to long-chain fatty acids efficiently.

By introducing two mutations (F87A and A82F) into the BM3 P450 domain, structural reorganization occurs leading to the enzyme becoming more promiscuous and catalyzing the oxidation of novel compounds, including drug molecules. To explore the breadth of substrate specificity for this BM3 variant, we screened for binding of novel ligands using an FDA-approved compound library and identified potential substrates through their ability to induce a P450 Soret heme iron shift towards the high spin state. 59% of the compounds tested elicited significant Soret shifts indicative of drug binding. Our studies demonstrate the plasticity of the P450 fold and the ability of this P450 BM3 variant to adapt to the binding of structurally diverse compounds, highlighting P450 BM3’s utility as a model system for studying pharmaceutical compounds.

3.2. Introduction

The cytochromes P450 constitute one of the largest superfamilies of proteins, and are found across all kingdoms (Nelson, 2018). The P450s are monooxygenase enzymes that catalyze an extensive range of reactions on diverse types of substrates. These include fatty acids, steroids, polyketides, terpenes and pharmaceuticals (Guengerich and Munro,

2013, Bernhardt, 2006, Munro et al., 2007). In humans and other eukaryotic organisms, the

P450s play crucial roles in the phase I metabolism of drugs and other xenobiotics, which

115 leads to their excretion either following direct oxidation of these molecules, or after the further modification of oxidized molecules catalyzed by phase II metabolic enzymes such as glutathione-S-transferases (Board and Menon, 2013). In humans and other mammals, the majority of these reactions occur in the liver, where the P450s and their redox partner

(NADPH-cytochrome P450 reductase or CPR) are both membrane-associated enzymes (a type I P450 redox system). However, in prokaryotes and archaea the P450s and their redox partners are soluble, cytoplasmic enzymes that lack the N-terminal membrane-spanning helix that is typical of eukaryotic P450s and their CPR partners. The majority of prokaryotic P450s obtain electrons from a ferredoxin redox partner, which in turn receives electrons from a NAD(P)H-dependent, FAD-binding ferredoxin reductase (a type II P450 redox system) (McLean et al., 2015). A third class of the P450s is best characterized by the soluble P450 BM3 (CYP102A1) enzyme from Bacillus megaterium and its orthologues in other microbes. BM3 is a natural fusion of a fatty acid hydroxylase to a CPR enzyme, neither modules of which are membrane-bound (Munro et al., 2002). It has the highest catalytic rate reported for a P450 enzyme (ca 17,000 min-1) and has proven to be a versatile and biotechnologically useful enzyme that can be engineered to bind and oxidise molecules that are structurally diverse from its “natural” fatty acid substrates (e.g. alkylbenzenes, flavonoids, alkanes and substituted naphthalenes (Whitehouse et al., 2012, Chu et al.,

2016, Chen et al., 2012, Misawa et al., 2011).

P450 BM3 binds a heme b prosthetic group which produces a red to red/brown colour depending up the redox state of the heme iron and whether a substrate or inhibitor is bound to the P450 (heme) domain (Munro et al., 2002). In its resting, substrate-free ferric state the heme iron is hexacoordinate and is axially ligated by a proximal cysteine thiolate and a loosely associated water molecule in the distal pocket adjacent to the active site. In this state an absorbance maximum is seen at ~418 nm for the Soret band of the P450 heme in its low-spin form (S = 1/2) (Miles et al., 1992). On binding of fatty acid substrates to the 116 wild-type (WT) P450 BM3 enzyme, the distal water is displaced, the heme becomes pentacoordinate and electrons in the 3d orbitals redistribute to give a high-spin state with a

Soret maximum at ~390 nm (S = 5/2) (Denisov et al., 2005, Luthra et al., 2011). The extent of the heme spectral shift induced on fatty acid substrate binding varies according to the chain length of the substrate, with longer chain fatty acids such as palmitic acid (C16:0) and arachidonic acid (C20:4) inducing much greater conversion to the high-spin state than do shorter chain substrates (e.g. lauric acid) (Miles et al., 1992, Noble et al., 1999). The extent of the heme Soret spectral shift can be estimated accurately with knowledge of the extinction coefficients for the low-spin and high-spin forms of P450 BM3. This enables comparative analysis of the efficiency of the binding of different substrates (Tran et al.,

2012). The binding affinity for potential substrates can easily be determined by UV-vis titrations with each compound and by data fitting to determine binding constant (Kd) values in each case (Jeffreys et al., 2018). In contrast, many inhibitors of P450 enzymes displace the distal water ligand and ligate to the heme iron through a heteroatom, frequently a nitrogen atom as in the case of azole drugs (e.g. fluconazole) that inhibit the fungal lanosterol 14-demethylase (Mast et al., 2010) (McLean et al., 2008). Azole drug binding induces a red shift in the P450 Soret spectrum from ~418 nm to ~420-427 nm, depending on the specific drug bound (Locuson et al., 2007).

P450 BM3 has been the subject of numerous protein engineering studies aimed at altering its substrate selectivity with a view towards biotechnological applications

(Whitehouse et al., 2012). Examples include the development of BM3 heme domain-based dopamine biosensors (Shapiro et al., 2010), the production of propane-hydroxylating BM3 variants for propanol production (Fasan et al., 2007), and the efficient hydroxylation of the gastric proton pump inhibitor drug omeprazole (forming the same metabolite as that produced by the human CYP2C19 enzyme) by a A82F/F87V BM3 double mutant variant

(Butler et al., 2013). In the latter case, preceding studies revealed that the A82F mutation 117 caused a structural rearrangement near the core of the P450 active site, and resulted in a conformational change similar to that induced by substrate (e.g. palmitoleic acid or N- palmitoyl glycine) binding to the WT form of the P450 BM3, and resulting in a

“catalytically primed” form of the enzyme (Huang et al., 2007). The F87V mutation exchanges a bulky phenyl group from the heme environment, creating additional space in the active site for novel substrate binding (Huang et al., 2007). Our subsequent studies on the A82F/F87V mutant demonstrated its ability to oxidatively modify other proton pump inhibitor drugs, such as esomeprazole, rabeprazole, lansoprazole and pantoprazole, forming known human metabolites (Butler et al., 2014). We refer to this promiscuous

A82F/F87V BM3 variant as a “gatekeeper” mutant in light of its ability to alter heme domain structure and facilitate the binding of several non-native substrates. In the case of human P450 enzymes, the characterization of their metabolites is now important for their approval by regulatory bodies including the US food and drug administration (FDA).

Many pharmaceutical compounds are metabolized by human P450 enzymes (50-70%) and so the production and analysis of their metabolites is now a crucial step in the assessment of their safety profiles prior to these drugs entering the market (Guengerich, 2003, Ince et al., 2013).

In this manuscript, we describe studies to characterize the scope of substrates recognized by the BM3 A82F/F87V double mutant (DM), using the heme (P450) domain of this enzyme and by UV-visible spectroscopic analysis of the optical perturbations induced on interaction of the DM heme domain with the chemical compounds from an

FDA-approved compound library. Around 60% of these compounds were found to induce significant Soret peak shifts, most of which were high-spin (blue) Soret spectral shifts indicative of substrate binding. Various compounds identified in this way included compounds from several drug classes, and these and other hit compounds were titrated with the WT and DM heme domain variants in order to determine binding affinities (Kd 118 values) for these molecules. Many new compounds were discovered to bind to the P450

BM3 DM enzyme, indicating its potential to produce human drug metabolites and the capacity of the P450 BM3 DM enzyme as a human P450 enzyme model that can produce human drug metabolites for applications in drug safety testing.

3.3. Materials and Methods

3.3.1. Protein expression and purification

WT and DM P450 BM3 heme domain proteins were used for the compound binding and titration experiments discussed in this paper. The ~55 kDa WT and DM heme domains were cloned into the pET14b vector (Butler et al., 2013). The plasmids were transformed into the E. coli BL21 (DE3) strain and expressed in terrific broth medium

(Formedium, Hunstanton, UK) for the WT BM3 heme domain, and in autoinduction

Terrific Broth medium (Formedium) for the BM3 DM heme domain. Cells were grown for

24 hours at 37°C and at 170 rpm in an orbital shaker. 100 µM δ-aminolevulinic acid was added once the OD600 reached 0.6 in order to promote heme incorporation in the P450.

Cells were harvested by centrifugation at 4°C (6000 g, 20 min).

The cell pellets were resuspended in ice-cold phosphate buffer A (50 mM potassium phosphate [KPi] containing 350 mM KCl and 10% glycerol, pH 8) containing protease inhibitors (1 tablet per 100 mL EDTA-free cOmpleteTM tablets, Roche Applied

Science, Burgess Hill, UK), and DNase (Merck, Nottingham UK) (10 µg mL-1). Cells were lysed by sonication on ice using a Bandelin Sonopuls instrument at 37% amplitude, with

12 pulses for 40 s, and with 60 s breaks between pulses. The cell extracts were clarified by centrifugation (4600 g for 60 minutes at 4°C). Contaminant proteins were removed by incubation with 30% w/v ammonium sulfate for 1 hour at 4°C with gentle stirring.

119

Precipitated material was removed using centrifugation (4600 g, 15 minutes, 4°C). The proteins were purified using two-column purification steps, using a nickel-IDA column followed by an anion exchange (hydroxyapatite) chromatography step. The clarified cell extracts were incubated with nickel-IDA resin overnight at 4°C in 50 mM KPi containing

350 mM KCl and 10% glycerol, pH 8. The DM and WT BM3 heme domains resins were then transferred to columns, and a stepwise gradient of 10 mM (300 mL), 20 mM (200 mL) and 200 mM (60 mL) imidazole was applied to elute the proteins. The eluted proteins were then dialyzed into 25 mM KPi, pH 6.5 (buffer B) and applied to 28 mL columns containing

CHT hydroxyapatite type 1 resin (Bio-Rad Laboratories, Watford UK). The proteins were then eluted using a linear gradient of 25-300 mM KPi, pH 6.5 (600 mL). Potential ligands bound to the WT and DM heme domains were removed from the proteins using the hydrophobic resin Lipidex-1000 (Perkin Elmer, Beaconsfield UK) and by washing the column using 100 mM KPi, pH 7. Purified proteins were stored at -80°C. When required for library screening, purified heme domain batches were defrosted and stored on ice.

Proteins were then dialyzed into 25 mM KPi (pH 6.5) and purified using a hydroxyapatite column, as described above. This resulted in substantial amounts of highly purified proteins (e.g. ~4 g of the DM heme domain), with approximately half of this sample used for library screening to detect compounds inducing heme spectral shifts.

3.3.2. Screening of the FDA-approved compound library

An FDA-approved compound screen was acquired and contained 978 FDA- approved compounds in 10 mM stocks (in 100% DMSO) from Selleck Chemicals (L1300)

(Houston, Texas, USA). The screen contained a wide variety of FDA-approved compounds targeting many different drug target proteins, and with molecular weights ranging from

100-800 g/mol. By adding 150 M of each compound to 3 M of the WT or DM heme domain proteins in 0.5 mL 100 mM KPi, pH 7 binding was observed. P450 Soret

120 absorbance shifts were monitored on a UV-Vis spectrophotometer (Agilent Cary 50). The high stoichiometry of ligand to protein was used to ensure saturation with the ligand occurred and that the extent of the ligand-induced spectral change was easily determined for both the WT and DM BM3 heme domains. These experiments were conducted using

1.5 mL semi-micro disposable cuvettes in a total volume of 500 µl. In some cases the ligand was coloured or had poor solubility, leading to increased absorbance across the spectrum in the latter case. To address this issue, the baseline used was the buffer plus the ligand, rather the buffer only (which was used for the remaining ligands). In this way, evidence for type I or type II spectral shifts could be obtained in most cases. For type II shifts elicited by inhibitors, the final wavelength for the Soret peak shift was recorded. For type I shifts elicited by substrate binding, the percentage high-spin shift was calculated using a modified high-spin calculation equation (Tran et al., 2012) using extinction coefficients derived for the WT and DM BM3 heme domains, and by using N-palmitoyl glycine (NPG) as a model for substrate binding in view of its ability to induce a near- complete conversion to the high-spin state. For all the results described here, the percentage spin shift induced by DMSO solvent alone (3.5%) was subtracted from the original percentage. A Python script was produced to extract the necessary LS or HS absorbance maxima from CSV files and input them into the percentage high spin calculation (Equation 1). The script was produced to find the closest wavelength to

393/394 nm and 417/418 nm in all cases. In addition, the script produced graphical images for manual determination of turbidity, protein aggregation or other causes of error.

퐿푆 퐿푆 휀 퐴393 − 휀 퐴417 % 퐻푆 = 417 393 퐿푆 퐻푆 퐿푆 퐻푆 퐴 (휀 − 휀 ) − 퐴 (휀 − 휀 ) 393 417 417 417 393 393

퐿푆 퐿푆 퐻푆 퐻푆 = 95.00 mM cm-1 = 53.98 mM cm-1 = 56.05 mM cm-1 = 84.29 mM cm-1 417 393 417 393

Equation 1: Percentage high spin equation for DM BM3

121

The application of the percentage high-spin equation enables rapid and accurate comparisons to be made between the extents of high-spin conversion that occur in the various substrate-bound forms of WT and DM BM3 heme domains. The method relies on calculating extinction coefficients for the ligand-bound enzyme. The modified fatty acid N- palmitoyl glycine (NPG) was used as a standard due to its ability to shift to the P450 BM3 heme Soret peak to close to 100% high-spin (Butler et al., 2013). This allows the percentage high-spin values to be calculated as a percentage relative to that for NPG at

100%. Due to volume constraints, this method was chosen for library result comparisons.

This method can also be used to highlight differences in extents of binding when similar binding affinities are observed, for example between mutant variants (Tran et al., 2012).

The method used successfully identified several novel ligands for the DM and WT P450

BM3 heme domains, enabling subsequent titrations studies to determine binding constants for these compounds.

3.3.3. Binding affinity determination of pharmaceutical compounds

Around 80 compounds were chosen for Kd determinations on the basis of their structures or drug class, and the percentage high-spin or low-spin shift elicited by substrate/inhibitor binding. The binding affinities for these compounds were determined using a UV-Vis spectrophotometer at 30°C. Titrations were undertaken using ~3 µM of the

WT and DM heme domains in 100 mM KPi (pH 7) in a quartz cuvette by titrating small amounts of the ligand into the P450 samples until no further spectral changes were observed. Binding affinity for each ligand was determined by data fitting using the

Michaelis-Menten (Michaelis and Menten, 1913), Morrison (Morrison, 1969) or Hill (Hill,

1910) equations, according to the nature of the dependence of induced absorbance change on ligand concentration and the relative binding affinity of the ligand.

122

3.4. Results

3.4.1. FDA-approved compound library screening

A compound library was acquired containing 978 FDA-approved compounds in

10 mM stocks (100% DMSO) from Selleck Chemicals (L1300) (Houston, Texas, USA).

The library contained a wide variety of compounds acting on different drug target enzymes, with molecular weights ranging from 100-800 g/mol. Additions of 150 M of each ligand were made to 3 M BM3 DM heme domain protein (ensuring high ligand-to- protein stoichiometry to ensure protein saturation) and heme spectral shifts induced were recorded on a UV-Vis spectrophotometer (Agilent Cary 50 instrument). Dimethyl sulfoxide (DMSO) solvent was used for compound solubilization and analysis of control samples indicated that the DMSO in isolation induced a BM3 DM heme domain spin-state shift towards high-spin of ~3.5%. As a result, this proportion was subtracted from the calculated spin-state shifts observed for the numerous compounds that elicited more substantial high-spin shifts.

Many compounds emanating from a variety of different drug classes (e.g. steroids, antibacterials, anti-inflammatories, azoles and statins) were found to induce shifts in the

BM3 DM Soret peak. The extents of the heme spectral shifts induced are shown as percentages in Figure S6 and graphically in Figure 17. Only spectral shifts greater than

10% were classified as significant in order to remove the chance of false positives. This resulted in 59.3% of the compounds tested producing a significant shift of the Soret peak.

A small selection of these compound structures is shown in Figure 18. The library was not balanced in terms of mass, target or logP and so no trends could be deduced. In addition, no trend could be deduced by structure, as the compounds from the library were grouped by target. 318 compounds induced small high-spin (type I) heme spectral shifts between

10-19%, compared to 30 compounds with a high-spin Soret spectral shift of over 50%. 61

123

Figure 17: FDA-approved compound library hits presented according to their extent of high-spin percentage induced The graph shows ligands that produce high-spin shifts in the DM BM3 heme domain, presented according to the percentage high-spin conversion that these ligands induce and (in colour code) the compound classes to which the ligands belong. The percentage spin- state change is shown with reference to the binding of N-palmitoyl glycine to the DM heme domain, where a near 100% conversion to the high-spin state is observed. 59% of the 978 compounds in the library were shown to induce either type I (substrate-like) or type II (inhibitor-like) Soret absorbance shifts. Compound library hits from a variety of drug and other compound classes were identified.

inhibitors were identified through their producing a red-shifted (type II) Soret spectrum, with Soret peak shifts observed from 419 nm through to 430 nm dependent on the specific inhibitor compound. A number of other compounds resulted in protein aggregation and/or heme depletion, as observed by decreased Soret peak intensity at 418 nm (with little effect on the 280 nm peak) suggesting protein instability on the addition of these compounds.

124

Figure 18: A wide variety of structures were found to bind to the DM BM3 variant Examples of compounds found to elicit large type I and type II shifts on binding to the DM BM3 heme domain. Inhibitors are shown with blue labels and substrates with red labels.

3.4.2. Determination of binding affinities for WT and DM BM3 variants

In view of the substantial numbers of compounds that elicited substrate-like type I spectral shifts, several compounds were selected from the library for further analysis based on the extent of high-spin shift induced, their structure and their drug class. For greater expansion of the substrate profile, compounds related to those identified as hits within the same drug class were also purchased and tested for binding to the BM3 DM and the BM3

125

WT heme domains, such as tegafur and galeterone. Binding titrations were performed as described in the Methods section, and binding constants (Kd values) were determined by plotting ligand-induced heme spectral shift against ligand concentration, and by fitting

WT BM3 DM BM3 FDA-approved compound MW (Da) Kd (M) Shift Kd (M) Shift

Apatinib† 493.6 NB - 1.6 ± 0.1 419 nm

Abiraterone acetate 391.6 NB - NB -

Azathioprine† 277.3 NB - 7.6 ± 0.1 430 nm

Cyproterone acetate 416.9 NB - NB -

Fenticonazole nitrate 518.4 1.89 ± 0.05 419 nm 0.52 ± 0.03 424 nm

Galeterone 388.6 NB - NB -

Orteronel 307.4 NB - 3.0 ± 0.2 423 nm

Sulconazole nitrate 460.8 8.93 ± 0.54 421 nm 0.52 ± 0.01 424 nm

Table 1: Binding affinity (Kd) values for WT and DM BM3 heme domains for compounds displaying type II shifts Several drugs were shown to bind to DM BM3 heme domains by eliciting type II (inhibitor) spectral shifts, but not to the WT heme domain. The table shows the relevant Kd values for each compound and the final Soret wavelength shift observed (from the starting wavelength of 418 nm in each case) at apparent saturation. NB denotes instances where no binding is apparent in UV-visible spectroscopic titrations. † denotes the presence of this compound in the initial library screen. The Kd values were determined as described in the Methods section, using either a hyperbolic function (Michaelis-Menten), the Hill equation or the Morrison equation for tight-binding ligands, as appropriate. Further inhibitors are described in Chapter 4.

WT BM3 DM BM3 FDA-approved compound MW (Da) Kd (M) Kd (M)

Artemether† 298.4 NB 15.8 ± 0.9

Artemisinin† 282.3 NB 51.0 ± 1.6

Betamethasone valerate† 476.6 NB 0.17 ± 0.03

Carmofur† 257.3 75.6 ± 4.1 3.88 ± 0.28

Cetrimonium bromide 320.0 Inhibitor 2.7 ± 0.2

126

Cetylpyridinium chloride 340.0 Inhibitor 10.4 ± 1.3

Chenodeoxycholic acid† 392.6 39.4 ± 3.0 1.3 ± 0.3

Dominphen bromide 414.5 NB 16.8 ± 3.6

Drospirenone† 366.5 NB 29.6 ± 0.1

Levonorgestrel† 312.5 NB 11.3 ± 0.1

Lithocholic acid† 376.6 NB 0.86 ± 0.09

Norethindrone† 298.4 NB 12.5 ± 0.2

Nystatin† 926.1 NB 27.3 ± 1.9

Retinol 286.5 NB 1.3 ± 0.19

Tegafur 200.2 14.0 ± 0.7 58.0 ± 2.6

Tretinoin (Retinoic acid)† 300.4 NB 2.1 ± 0.1

Trilostane† 329.4 NB 15.4 ± 0.7

Table 2: Binding affinity values for WT and DM BM3 heme domains with ligands displaying type I Soret spectral shifts Several drugs were shown to bind to the WT and DM BM3 heme domains by eliciting a type I (substrate-like) Soret spectral shift. The table shows examples of such molecules and the relevant Kd values determined for these compounds by optical titration. NB denotes instances (for the WT BM3 heme domain) where no evidence of ligand binding was apparent from UV-visible spectroscopic titrations. † denotes the presence of this compound in the initial library screen. The Kd values were determined as described in the Methods section, using either a hyperbolic function (Michaelis-Menten), the Hill function or the Morrison equation for tight-binding ligands, as appropriate.

these data using standard equations. Several substrates and inhibitors were shown to have low Kd values, indicating their tight binding to the BM3 DM. In comparison, far fewer compounds were able to bind to the BM3 WT heme domain (as evidenced from spectral shifts) and in incidences where binding was observed the heme spectral shift induced was much less than for the DM variant and Kd values were much higher (Table 1 and Table 2).

The only exception was for the anticancer agent tegafur, where the Kd for the WT heme domain is lower than that for the DM heme domain (14 M compared to 58 M for the

WT and DM heme domains, respectively). For the quaternary ammonium compounds tested, a type I shift was observed in each case for their binding to the BM3 DM, whereas

127 no spectral perturbation was observed on the addition of these compounds to the BM3 WT heme domain.

3.5. Discussion

Many papers have been published in which P450 enzymes and mutants thereof are screened for substrate binding and product formation. For example, studies reporting the targeted screening of P450 BM3 mutants for flucloxacillin hydroxylation (Luirink et al.,

2018) or for the production of human metabolites from eight commonly prescribed drugs

(Ren et al., 2015). However, few papers exist in which one type of P450 protein is screened for ligand profile identification. Examples of this type of study include fragment screening and inhibitor compound development for the Mycobacterium tuberculosis P450 enzyme CYP121, and (based on the similarities between fragment binding profiles of the

M. tuberculosis CYP121 and CYP144 enzymes) the development and application of inhibitors for the CYP144 P450 (Kavanagh et al., 2016, Kavanagh et al., 2017). Therefore, the screening of such a large library of compounds to probe for ligands specific for P450

BM3 and its DM variant, as done in this paper, appears to be novel in the literature.

Some of the most extensive high-spin shift results observed from our compound screening were for steroid or steroid-like compounds. Steroids are known to bind to many human P450 enzymes, including CYP17A1 where its activities include 17-hydroxylation of pregnenolone and progesterone (Petrunak et al., 2014). The compound library contained the steroid betamethasone in three different forms: betamethasone, betamethasone dipropionate and betamethasone valerate. While all of these forms of the steroid induce a substrate-like high-spin shift, the extent of the shift observed varied greatly for each form

(8.94%, 29.50%, and 69.09% high-spin respectively), likely due to the enhanced solubility of the steroid salts in comparison to betamethasone itself. Similar phenomena were 128 observed for different preparations of other compounds from the screen (such as omeprazole, ozagrel and more). The CYP17A1 inhibitor orteronel was also found to bind the DM variant, but as an inhibitor rather than a substrate, inducing a spectral shift from

418 to 422 nm and suggesting that the DM variant could act as a human CYP17A1 model.

However, other inhibitors of CYP17A1, such as the steroidal anti-androgens abiraterone acetate, cyproterone acetate and galeterone, did not induce Soret spectral shifts for either the BM3 DM or WT heme domains.

The BM3 DM variant was also found to bind to vitamin A (retinol) and to one of its metabolites (retinoic acid) in a substrate-like manner. Many human P450s were shown to bind (retinol) and to convert this compound to retinal, including CYP1B1 (Zhang et al.,

2000) and CYP2D6 (Chen et al., 2000). In the case of CYP1B1, the enzyme catalyzes the further oxidation of retinal to retinoic acid (Zhang et al., 2000). Other human P450s also bind these compounds as inhibitors, including CYP2C8 and CYP2C9 (Yamazaki and

Shimada, 1999).

The anti-malarial drugs artemether and artemisinin, also used as anticancer agents, were also found to induce high-spin shifts in the BM3 DM (with Kd values of 15.8 M and

51.0 M, respectively), but not for the WT BM3 heme domain. Human metabolites of these antimalarial drugs are produced primarily by CYP2B6, or by CYP3A4 in individuals with low CYP2B6 expression levels (Svensson and Ashton, 1999). In previous studies, human metabolites of artemisinin were produced using P450 BM3 variants containing mutations at Phe87 and Ala82, albeit with different amino acid substitutions at these positions and together with other active site mutations (Zhang et al., 2012). Many anticancer drugs, such as apatinib (inhibitor), lomustine (substrate) and carmofur (substrate), were also found to be ligands for the DM BM3 from library screening. The chemotherapeutic drugs carmofur and tegafur are interesting examples of prodrugs that are 129 converted to their active form by a P450-catalyzed hydroxylation reaction, causing a bond breakage that releases the anticancer agent 5-fluorouracil. It is believed that human

CYP2A6 catalyzes the cleavage of tegafur to form 5-fluorouracil (Kajita et al., 2003). Both of these compounds were shown to bind to WT and DM BM3 heme domains. Carmofur and tegafur exhibited greater extents of high-spin shift for the DM compared to the WT heme domain. However, while carmofur bound more tightly to the DM than the WT BM3 heme domain (Kd values of 3.88 M and 75.6 M, respectively), tegafur was found instead to bind more avidly to the WT than to the DM heme domain (Kd values of 14.0 M and

58.0 M, respectively). This is an unusual result for these BM3 gatekeeper mutant studies, as the DM BM3 typically has greater affinity for natural and non-natural ligands than does the WT enzyme (Butler et al., 2013, Butler et al., 2014).

A number of quaternary ammonium compounds were present in the compound library (Figure S6). These compounds are used to sterilize surgical equipment and are used in many commercial products, such as toothpaste, eye drops etc. These compounds bind with greater affinity to the DM BM3 than to the WT BM3 heme domain (Table 2).

Interestingly, the shift elicited for the DM heme domain was found to be type I for each compound. However, for the WT BM3 heme domain, a type II inhibitor-like shift was observed for the binding of cetylpyridinium chloride (CPC) and cetrimonium bromide

(CTAB), whilst the domiphen bromide ligand did not induce either a type I or II spectral shift. Binding affinities for these compounds could not be calculated accurately for the WT heme domain, as the critical micelle concentration (CMC) was reached before saturation of the heme spectral response could be achieved, with heme loss evident by decreases in

Soret peak magnitude towards the end of these titrations. For CPC and CTAB, their respective pyridinium and quaternary ammonium groups are located close to one at the one end of the molecule and likely interact with the heme iron to induce a type II binding

130 spectrum. For domiphen bromide, the quaternary ammonium group is located more centrally in the fatty acid chain, and this likely affects its binding mode and prevents the quaternary ammonium group interacting with the WT BM3 to induce a type II spectral shift.

61 different inhibitors were identified from our UV-visible spectroscopic analysis of ligand binding to the DM heme domain. Many of these were from the azole drug class, including fluconazole, clotrimazole and butoconazole. These are well-known inhibitors of cytochrome P450 enzymes that inactivate their target enzymes by coordinating directly to the P450 heme through a nitrogen atom from an imidazole or triazole ring. In our recent studies, a novel mode of azole binding to the DM BM3 heme domain was observed in a voriconazole-bound complex. In this case, the heme iron-coordinating nitrogen atom comes from a fluorinated pyrimidine ring instead of the voriconazole triazole group.

Interesting conformational changes for the DM BM3, similar to those induced by substrate binding, were observed upon fluconazole binding using X-ray crystallography for structural elucidation. Structural analysis for the interactions of selected FDA-approved azoles are discussed in further detail in our recent publication (Chapter 4).

3.6. Conclusions

In conclusion, we present a study of the interactions and binding affinity of a conformationally flexible and promiscuous P450 BM3 double mutant (A82F/F87V) with a diverse range of chemical compounds from an FDA-approved 978 compound library. This screening process involved the identification of likely substrates through their inducing a type I (blue) shift of the heme Soret spectrum that is characteristic of substrate-like behaviour. Consistent with these data, we have demonstrated that library compounds from the fibrate and glitazone classes are bona fide substrates for the BM3 “gatekeeper” mutant, 131 in addition to our previous studies that demonstrated this mutant was catalytically active with the compound library drug omeprazole (and others from the proton pump inhibitor class) and other drugs from this class (Butler et al., 2013, Butler et al., 2014). Protein engineering studies using P450 BM3 have resulted in the development of biocatalysts that are highly specific for particular oxidative reactions (Acevedo-Rocha et al., 2018, Glieder et al., 2002). However, there is clearly a wider scope for the application of the promiscuous

DM BM3 enzyme in areas such as the synthesis of e.g. pharmaceutical metabolites for drug safety testing, and oxidatively modified terpenes and hydroxylated steroids for biotechnological and biomedical applications.

3.7. References

ACEVEDO-ROCHA, C. G., GAMBLE, C. G., LONSDALE, R., LI, A. T., NETT, N., HOEBENREICH, S., LINGNAU, J. B., WIRTZ, C., FARES, C., HINRICHS, H., DEEGE, A., MULHOLLAND, A. J., NOV, Y., LEYS, D., MCLEAN, K. J., MUNRO, A. W. & REETZ, M. T. 2018. P450-Catalyzed Regio- and Diastereoselective Steroid Hydroxylation: Efficient Directed Evolution Enabled by Mutability Landscaping. ACS Catal, 8, 3395-3410. BERNHARDT, R. 2006. Cytochromes P450 as versatile biocatalysts. J Biotechnol, 124, 128-45. BOARD, P. G. & MENON, D. 2013. Glutathione transferases, regulators of cellular metabolism and physiology. Biochim Biophys Acta, 1830, 3267-88. BUTLER, C. F., PEET, C., MASON, A. E., VOICE, M. W., LEYS, D. & MUNRO, A. W. 2013. Key mutations alter the cytochrome P450 BM3 conformational landscape and remove inherent substrate bias. J Biol Chem, 288, 25387-99. BUTLER, C. F., PEET, C., MCLEAN, K. J., BAYNHAM, M. T., BLANKLEY, R. T., FISHER, K., RIGBY, S. E., LEYS, D., VOICE, M. W. & MUNRO, A. W. 2014. Human P450-like oxidation of diverse proton pump inhibitor drugs by 'gatekeeper' mutants of flavocytochrome P450 BM3. Biochem J, 460, 247-59. CHEN, H., HOWALD, W. N. & JUCHAU, M. R. 2000. Biosynthesis of all-trans-retinoic acid from all-trans-retinol: catalysis of all-trans-retinol oxidation by human P-450 cytochromes. Drug Metab Dispos, 28, 315-22. CHEN, M. M., SNOW, C. D., VIZCARRA, C. L., MAYO, S. L. & ARNOLD, F. H. 2012. Comparison of random mutagenesis and semi-rational designed libraries for improved cytochrome P450 BM3-catalyzed hydroxylation of small alkanes. Protein Eng Des Sel, 25, 171-8. CHU, L. L., PANDEY, R. P., JUNG, N., JUNG, H. J., KIM, E. H. & SOHNG, J. K. 2016. Hydroxylation of diverse flavonoids by CYP450 BM3 variants: biosynthesis of eriodictyol from naringenin in whole cells and its biological activities. Microb Cell Fact, 15, 135.

132

DENISOV, I. G., MAKRIS, T. M., SLIGAR, S. G. & SCHLICHTING, I. 2005. Structure and chemistry of cytochrome P450. Chem Rev, 105, 2253-77. FASAN, R., CHEN, M. M., CROOK, N. C. & ARNOLD, F. H. 2007. Engineered alkane- hydroxylating cytochrome P450(BM3) exhibiting nativelike catalytic properties. Angew Chem Int Ed Engl, 46, 8414-8. GLIEDER, A., FARINAS, E. T. & ARNOLD, F. H. 2002. Laboratory evolution of a soluble, self-sufficient, highly active alkane hydroxylase. Nat Biotechnol, 20, 1135- 9. GUENGERICH, F. P. 2003. Cytochromes P450, drugs, and diseases. Mol Interv, 3, 194- 204. GUENGERICH, F. P. & MUNRO, A. W. 2013. Unusual cytochrome p450 enzymes and reactions. J Biol Chem, 288, 17065-73. HILL, A. V. 1910. The possible effects of the aggregation of the molecules of haemoglobin on its dissociation curves. J Physiol, 40, 4-7. HUANG, W. C., WESTLAKE, A. C., MARECHAL, J. D., JOYCE, M. G., MOODY, P. C. & ROBERTS, G. C. 2007. Filling a hole in cytochrome P450 BM3 improves substrate binding and catalytic efficiency. J Mol Biol, 373, 633-51. INCE, I., KNIBBE, C. A., DANHOF, M. & DE WILDT, S. N. 2013. Developmental changes in the expression and function of cytochrome P450 3A isoforms: evidence from in vitro and in vivo investigations. Clin Pharmacokinet, 52, 333-45. JEFFREYS, L. N., GIRVAN, H. M., MCLEAN, K. J. & MUNRO, A. W. 2018. Characterization of Cytochrome P450 Enzymes and Their Applications in Synthetic Biology. In: Methods in Enzymology. Elsevier. KAJITA, J., FUSE, E., KUWABARA, T. & KOBAYASHI, H. 2003. The contribution of cytochrome P450 to the metabolism of tegafur in human liver. Drug Metab Pharmacokinet, 18, 303-9. KAVANAGH, M. E., CHENGE, J., ZOUFIR, A., MCLEAN, K. J., COYNE, A. G., BENDER, A., MUNRO, A. W. & ABELL, C. 2017. Fragment Profiling Approach to Inhibitors of the Orphan M. tuberculosis P450 CYP144A1. Biochemistry, 56, 1559-1572. KAVANAGH, M. E., COYNE, A. G., MCLEAN, K. J., JAMES, G. G., LEVY, C. W., MARINO, L. B., DE CARVALHO, L. P. S., CHAN, D. S. H., HUDSON, S. A., SURADE, S., LEYS, D., MUNRO, A. W. & ABELL, C. 2016. Fragment-Based Approaches to the Development of Mycobacterium tuberculosis CYP121 Inhibitors. J Med Chem, 59, 3272-3302. LOCUSON, C. W., HUTZLER, J. M. & TRACY, T. S. 2007. Visible spectra of type II cytochrome P450-drug complexes: evidence that "incomplete" heme coordination is common. Drug Metab Dispos, 35, 614-22. LUIRINK, R. A., DEKKER, S. J., CAPOFERRI, L., JANSSEN, L. F. H., KUIPER, C. L., ARI, M. E., VERMEULEN, N. P. E., VOS, J. C., COMMANDEUR, J. N. M. & GEERKE, D. P. 2018. A combined computational and experimental study on selective flucloxacillin hydroxylation by cytochrome P450 BM3 variants. J Inorg Biochem, 184, 115-122. LUTHRA, A., DENISOV, I. G. & SLIGAR, S. G. 2011. Spectroscopic features of cytochrome P450 reaction intermediates. Arch Biochem Biophys, 507, 26-35. MAST, N., CHARVET, C., PIKULEVA, I. A. & STOUT, C. D. 2010. Structural basis of drug binding to CYP46A1, an enzyme that controls cholesterol turnover in the brain. J Biol Chem, 285, 31783-95. MCLEAN, K. J., CARROLL, P., LEWIS, D. G., DUNFORD, A. J., SEWARD, H. E., NEELI, R., CHEESMAN, M. R., MARSOLLIER, L., DOUGLAS, P., SMITH, W. E., ROSENKRANDS, I., COLE, S. T., LEYS, D., PARISH, T. & MUNRO, A. W. 2008. Characterization of Active Site Structure in CYP121 a cytochrome P450

133

essential for viability of Mycobacterium tuberculosis H37V. J Biol Chem, 283, 33406-33416. MCLEAN, K. J., LUCIAKOVA, D., BELCHER, J., TEE, K. L. & MUNRO, A. W. 2015. Biological diversity of cytochrome P450 redox partner systems. Adv Exp Med Biol, 851, 299-317. MICHAELIS, L. & MENTEN, M. L. 1913. The kenetics of the inversion effect. Biochem Z, 49, 333-369. MILES, J. S., MUNRO, A. W., ROSPENDOWSKI, B. N., SMITH, W. E., MCKNIGHT, J. & THOMSON, A. J. 1992. Domains of the catalytically self-sufficient cytochrome P-450 BM-3. Genetic construction, overexpression, purification and spectroscopic characterization. Biochem J, 288 ( Pt 2), 503-9. MISAWA, N., NODATE, M., OTOMATSU, T., SHIMIZU, K., KAIDO, C., KIKUTA, M., IDENO, A., IKENAGA, H., OGAWA, J., SHIMIZU, S. & SHINDO, K. 2011. Bioconversion of substituted naphthalenes and beta-eudesmol with the cytochrome P450 BM3 variant F87V. Appl Microbiol Biotechnol, 90, 147-57. MORRISON, J. F. 1969. Kinetics of the reversible inhibition of enzyme-catalysed reactions by tight-binding inhibitors. Biochim Biophys Acta, 185, 269-86. MUNRO, A. W., GIRVAN, H. M. & MCLEAN, K. J. 2007. Variations on a (t)heme-- novel mechanisms, redox partners and catalytic functions in the cytochrome P450 superfamily. Nat Prod Rep, 24, 585-609. MUNRO, A. W., LEYS, D. G., MCLEAN, K. J., MARSHALL, K. R., OST, T. W., DAFF, S., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MOSER, C. C., PAGE, C. C. & DUTTON, P. L. 2002. P450 BM3: the very model of a modern flavocytochrome. Trends Biochem Sci, 27, 250-7. NELSON, D. R. 2018. Cytochrome P450 diversity in the tree of life. Biochim Biophys Acta, 1866, 141-154. NOBLE, M. A., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MACKAY, A. C., REID, G. A., HANZLIK, R. P. & MUNRO, A. W. 1999. Roles of key active-site residues in flavocytochrome P450 BM3. Biochem J, 339 ( Pt 2), 371-9. PETRUNAK, E. M., DEVORE, N. M., PORUBSKY, P. R. & SCOTT, E. E. 2014. Structures of human steroidogenic cytochrome P450 17A1 with substrates. J Biol Chem, 289, 32952-64. REN, X., YORKE, J. A., TAYLOR, E., ZHANG, T., ZHOU, W. & WONG, L. L. 2015. Drug Oxidation by Cytochrome P450BM3 : Metabolite Synthesis and Discovering New P450 Reaction Types. Chemistry, 21, 15039-47. SHAPIRO, M. G., WESTMEYER, G. G., ROMERO, P. A., SZABLOWSKI, J. O., KUSTER, B., SHAH, A., OTEY, C. R., LANGER, R., ARNOLD, F. H. & JASANOFF, A. 2010. Directed evolution of a magnetic resonance imaging contrast agent for noninvasive imaging of dopamine. Nat Biotechnol, 28, 264-70. SVENSSON, U. S. & ASHTON, M. 1999. Identification of the human cytochrome P450 enzymes involved in the in vitro metabolism of artemisinin. Br J Clin Pharmacol, 48, 528-35. TRAN, N. H., HUYNH, N., CHAVEZ, G., NGUYEN, A., DWARAKNATH, S., NGUYEN, T. A., NGUYEN, M. & CHERUZEL, L. 2012. A series of hybrid P450 BM3 enzymes with different catalytic activity in the light-initiated hydroxylation of lauric acid. J Inorg Biochem, 115, 50-6. WHITEHOUSE, C. J. C., BELL, S. G. & WONG, L. L. 2012. P450(BM3) (CYP102A1): connecting the dots. Chem Soc Rev, 41, 1218-1260. YAMAZAKI, H. & SHIMADA, T. 1999. Effects of arachidonic acid, prostaglandins, retinol, retinoic acid and cholecalciferol on xenobiotic oxidations catalysed by human cytochrome P450 enzymes. Xenobiotica, 29, 231-41. ZHANG, K., SHAFER, B. M., DEMARS, M. D., 2ND, STERN, H. A. & FASAN, R. 2012. Controlled oxidation of remote sp3 C-H bonds in artemisinin via P450 134

catalysts with fine-tuned regio- and stereoselectivity. J Am Chem Soc, 134, 18695- 704. ZHANG, Q. Y., DUNBAR, D. & KAMINSKY, L. 2000. Human cytochrome P-450 metabolism of retinals to retinoic acids. Drug Metab Dispos, 28, 292-7.

3.8. Supporting Information

Molecular % HS - Drug Name Indicator weight DMSO 1-Hexadecanol 242.44 N/A 0.00% Acyclovir (Aciclovir) 225.2 Infection 0.00% Allopurinol (Zyloprim) 136.11 Neurological Disease 0.00% Amiloride hydrochloride dihydrate 302.12 Cardiovascular Disease 0.00% Avanafil 483.95 Cardiovascular Disease 0.00% Benzbromarone 424.08 N/A 0.00% Betaxolol hydrochloride (Betoptic) 343.89 Cardiovascular Disease 0.00% Bicalutamide (Casodex) 430.37 Endocrinology 0.00% Cabazitaxel (Jevtana) 835.93 Neurological Disease 0.00% Cephalexin (Cefalexin) 347.39 Infection 0.00% Chlorpheniramine Maleate 390.86 Neurological Disease 0.00% Chlorprothixene 315.86 Neurological Disease 0.00% Cinchophen 249.26 Immunology 0.00% Clemastine Fumarate 459.96 Immunology 0.00% Clobetasol propionate 466.97 Neurological Disease 0.00% Crizotinib (PF-02341066) 450.34 Cancer 0.00% Daunorubicin HCl (Daunomycin HCl) 563.98 Cancer 0.00% Diminazene Aceturate 515.52 Vermifuge 0.00% Dipyridamole (Persantine) 504.63 Cardiovascular Disease 0.00% Dopamine hydrochloride (Inotropin) 189.64 Infection 0.00% Doripenem Hydrate 438.52 Infection 0.00% Ethionamide 166.24 Infection 0.00% Fenbendazole 299.35 N/A 0.00% Flavoxate HCl 427.92 Neurological Disease 0.00% Floxuridine 246.19 Cancer 0.00% Gatifloxacin 375.39 N/A 0.00% Gemcitabine (Gemzar) 263.2 Metabolic Disease 0.00% Levobetaxolol HCl 343.89 Cardiovascular Disease 0.00% Levofloxacin (Levaquin) 361.37 Infection 0.00% Lonidamine 321.16 Cardiovascular Disease 0.00% Lovastatin (Mevacor) 404.54 Respiratory Disease 0.00% MDV3100 (Enzalutamide) 464.44 Cancer 0.00% Mequinol 124.14 Infection 0.00% Methimazole (Tapazole, Northyx) 114.17 Endocrinology 0.00% Metolazone (Zaroxolyn) 365.83 Cardiovascular Disease 0.00% 135

Nadifloxacin 360.38 Neurological Disease 0.00% Nifuroxazide 275.22 Infection 0.00% Nitrendipine 360.3699 Neurological Disease 0.00% Oxeladin Citrate 527.6 N/A 0.00% Pralatrexate (Folotyn) 477.47 Metabolic Disease 0.00% Ranitidine (Zantac) 350.86 Metabolic Disease 0.00% Rifabutin (Mycobutin) 847 Infection 0.00% Rimantadine (Flumadine) 179.3 Infection 0.00% Ronidazole 200.15 Neurological Disease 0.00% Sildenafil citrate 666.7 Cardiovascular Disease 0.00% Sparfloxacin 392.4 Infection 0.00% Sulfasalazine (Azulfidine) 398.39 Inflammation 0.00% Sulfathiazole 255.32 Infection 0.00% Sulindac (Clinoril) 356.41 Cancer 0.00% Tigecycline 585.65 Infection 0.00% tinidazole 247.27 Infection 0.00% Vincristine sulfate 923.04 Cancer 0.00% Zolmitriptan (Zomig) 287.36 Neurological Disease 0.00% Tiotropium Bromide hydrate 490.43 Infection 0.09% Trichlormethiazide (Achletin) 380.66 Cardiovascular Disease 0.25% Difluprednate 508.55 Endocrinology 0.27% Olmesartan medoxomil (Benicar) 558.59 Cardiovascular Disease 0.27% Cyclosporine (Neoral) 1202.61 Immunology 0.27% Dofetilide (Tikosyn) 441.56 Cardiovascular Disease 0.29% Dextrose (D-glucose) 180.16 Infection 0.40% Acipimox 154.12 Cardiovascular Disease 0.46% Ketoprofen (Actron) 254.28 Inflammation 0.54% Moxalactam Disodium 564.44 Infection 0.58% Acetanilide (Antifebrin) 135.16 Neurological Disease 0.60% Medroxyprogesterone acetate 386.52 Infection 0.74% Phentolamine mesilate 377.46 Cardiovascular Disease 0.75% Indomethacin (Indocid, Indocin) 357.79 Inflammation 0.77% Nepafenac 254.28 Inflammation 0.77% TAME 342.41 Cancer 0.88% Vecuronium Bromide 637.73 Neurological Disease 0.97% Trifluridine (Viroptic) 296.2 Infection 1.00% Nithiamide 187.18 N/A 1.09% Prednisolone (Hydroretrocortine) 360.44 Infection 1.20% Doxylamine Succinate 388.46 Neurological Disease 1.28% Ranolazine dihydrochloride 500.46 Cardiovascular Disease 1.33% Cytidine 243.22 Cardiovascular Disease 1.37% Adenine hydrochloride 171.59 Cancer 1.39% Amfenac Sodium (monohydrate) 295.27 Inflammation 1.43% Genistein 270.24 Cancer 1.47% Rapamycin (Sirolimus) 914.18 Immunology 1.47% Dutasteride 528.53 Endocrinology 1.58% Etoposide (VP-16) 588.56 Cancer 1.71% 136

Probucol 516.84 Cardiovascular Disease 1.87% Rivastigmine tartrate (Exelon) 400.42 Cardiovascular Disease 1.87% Sitafloxacin Hydrate 436.84 N/A 1.91% Fesoterodine fumarate (Toviaz) 527.65 Immunology 1.95% Sotalol (Betapace) 308.82 Neurological Disease 1.97% Tizanidine HCl 290.17 Neurological Disease 2.00% Moclobemide 268.74 Neurological Disease 2.03% Indapamide (Lozol) 365.83 Cardiovascular Disease 2.07% Sulfanilamide 172.2 Infection 2.16% Ethacridine lactate monohydrate 361.39 Infection 2.16% Tylosin tartrate 1066.19 Neurological Disease 2.20% Trimebutine 387.47 Neurological Disease 2.21% Roxatidine acetate HCl 384.9 Digestive system disease 2.26% Enoxacin (Penetrex) 320.32 Infection 2.27% Paroxetine HCl 365.83 Infection 2.33% Vidarabine (Vira-A) 267.24 Infection 2.34% Enalaprilat dihydrate 348.4 Cardiovascular Disease 2.45% Asenapine 401.84 Neurological Disease 2.52% Raltegravir (MK-0518) 444.42 Immunology 2.61% Propafenone (Rytmonorm) 377.9 Cardiovascular Disease 2.63% Niacin (Nicotinic acid) 123.11 Metabolic Disease 2.65% Sulfaguanidine 214.24 Infection 2.71% Metoprolol tartrate 684.81 Cardiovascular Disease 2.72% Cisatracurium besylate (Nimbex) 1243.48 Neurological Disease 2.80% Mesalamine (Lialda) 153.14 Inflammation 2.89% Oxytetracycline dihydrate 496.46 Infection 2.95% Idarubicin HCl 533.95 Cancer 2.99% Zaltoprofen 298.36 Inflammation 3.01% Atracurium besylate 1243.48 Neurological Disease 3.01% Almotriptan malate (Axert) 469.55 Cardiovascular Disease 3.04% Rifapentine (Priftin) 877.03 Infection 3.06% Solifenacin succinate 480.55 Cardiovascular Disease 3.15% Anagrelide HCl 292.55 Endocrinology 3.15% Diphenhydramine HCl (Benadryl) 291.82 Immunology 3.16% Rifaximin (Xifaxan) 785.88 Infection 3.20% Chlorpromazine (Sonazine) 355.33 Neurological Disease 3.27% Epirubicin HCl 579.98 N/A 3.32% Isoniazid (Tubizid) 137.14 Infection 3.39% Telbivudine (Sebivo, Tyzeka) 242.23 Infection 3.54% Prilocaine 220.31 Neurological Disease 3.56% Meropenem 383.46 Infection 3.59% Linezolid (Zyvox) 337.35 Infection 3.60% Erythromycin (E-Mycin) 733.93 Infection 3.77% Ranolazine (Ranexa) 427.54 Cardiovascular Disease 3.79% Scopine 155.19 Metabolic Disease 3.87% Flurbiprofen (Ansaid) 244.26 Inflammation 3.89% Megestrol Acetate 384.51 Infection 3.93% 137

Isradipine (Dynacirc) 371.39 Neurological Disease 4.01% Bortezomib (Velcade) 384.24 Cancer 4.04% Dapoxetine hydrochloride (Priligy) 341.87 Neurological Disease 4.07% Adefovir Dipivoxil (Preveon, Hepsera) 501.47 Infection 4.10% Memantine HCl (Namenda) 215.76 Neurological Disease 4.10% Apixaban 459.5 Cardiovascular Disease 4.18% Lincomycin hydrochloride (Lincocin) 443 Cancer 4.25% Ornidazole 219.63 Endocrinology 4.25% Vitamin C (Ascorbic acid) 176.12 Respiratory Disease 4.26% Sumatriptan succinate 413.49 Neurological Disease 4.26% Clorsulon 380.66 Cancer 4.33% Flutamide (Eulexin) 276.21 Cancer 4.38% Triamcinolone (Aristocort) 394.43 Inflammation 4.41% Mycophenolate mofetil (CellCept) 433.49 Immunology 4.45% Adenosine (Adenocard) 267.24 Cardiovascular Disease 4.57% Iloperidone (Fanapt) 426.48 Neurological Disease 4.58% Pyridostigmine Bromide (Mestinon) 261.12 Cardiovascular Disease 4.66% Geniposide 388.37 N/A 4.67% Trimipramine Maleate 410.51 N/A 4.79% Mirtazapine (Remeron, Avanza) 265.35 Immunology 4.88% Prednisone (Adasone) 358.43 Immunology 4.93% Silodosin (Rapaflo) 495.53 Cardiovascular Disease 4.93% Argatroban 508.63 Cardiovascular Disease 5.00% OSI-420 (Desmethyl Erlotinib) 415.87 Cancer 5.05% Cyproheptadine HCl (Periactin) 323.86 Neurological Disease 5.08% Imidapril (Tanatril) HCl 441.91 Cardiovascular Disease 5.09% Fluoxetine HCl 345.79 Neurological Disease 5.17% Axitinib 386.47 Cancer 5.23% Candesartan (Atacand) 440.45 Cardiovascular Disease 5.24% Ivacaftor (VX-770) 392.49 Respiratory disease 5.25% Fenoprofen calcium hydrate 558.63 Immunology 5.30% Vinblastine sulfate 909.05 Neurological Disease 5.30% Ethambutol HCl 277.23 Neurological Disease 5.49% Diltiazem HCl (Tiazac) 450.98 Cardiovascular Disease 5.61% Diclofenac Potassium 334.24 Infection 5.64% Altretamine (Hexalen) 210.28 Cancer 5.65% Amantadine hydrochloride (Symmetrel) 187.7 Cardiovascular Disease 5.68% Tenofovir 287.21 N/A 5.68% Resveratrol 228.24 Infection 5.70% Lidocaine (Alphacaine) 234.34 Neurological Disease 5.71% Niflumic acid 282.22 Infection 5.84% Melatonin 232.28 Endocrinology 5.85% Pergolide mesylate 410.59 Neurological Disease 5.87% Arecoline 236.11 Endocrinology 5.88% Tebipenem pivoxil (L-084) 497.63 Cardiovascular Disease 5.90% Xylose 150.13 Metabolic Disease 5.96% Rivaroxaban (Xarelto) 435.88 Metabolic Disease 5.99% 138

Primidone (Mysoline) 218.25 Neurological Disease 6.01% Thiamphenicol (Thiophenicol) 356.22 Cardiovascular Disease 6.05% Risperidone (Risperdal) 410.48 Neurological Disease 6.06% Adiphenine HCl 347.88 Cardiovascular Disease 6.07% 10-Deacetylbaccatin-III 544.59 N/A 6.23% Epinephrine bitartrate (Adrenalinium) 333.29 Cancer 6.30% Flunarizine 2HCl 477.42 Cancer 6.40% Pitavastatin calcium (Livalo) 880.98 Cardiovascular Disease 6.44% Uridine 244.2 Vermifuge 6.48% Tolterodine tartrate (Detrol LA) 475.57 Neurological Disease 6.48% Nitrofurazone (Nitrofural) 198.14 Infection 6.50% Mevastatin 390.51 Cardiovascular Disease 6.55% Broxyquinoline 302.95 Vermifuge 6.58% Ketotifen fumarate (Zaditor) 425.5 Neurological Disease 6.59% Etodolac (Lodine) 287.35 Inflammation 6.62% Pefloxacin mesylate 429.46 Infection 6.70% Nifedipine (Adalat) 346.33 Cardiovascular Disease 6.73% Methscopolamine (Pamine) 398.29 Neurological Disease 6.74% Stavudine 224.21 Infection 6.81% Chlormezanone (Trancopal) 273.74 Respiratory Disease 6.94% Formoterol hemifumarate 402.4 Neurological Disease 6.94% Temocapril HCl 513.07 Cancer 6.97% Prochlorperazine Dimaleate 606.09 N/A 7.00% Rofecoxib (Vioxx) 314.36 Digestive system disease 7.08% Ritodrine hydrochloride (Yutopar) 323.81 Infection 7.08% Moroxydine 207.66 Cancer 7.10% Sulphadimethoxine 310.33 Infection 7.15% Teniposide (Vumon) 656.65 Cancer 7.27% Clindamycin hydrochloride (Dalacin) 461.44 Neurological Disease 7.35% Reboxetine mesylate 409.5 Neurological Disease 7.43% Pyrazinamide (Pyrazinoic acid amide) 123.11 Infection 7.44% Quetiapine fumarate (Seroquel) 883.09 Neurological Disease 7.44% Famotidine (Pepcid) 337.45 Cardiovascular Disease 7.45% Tadalafil (Cialis) 389.4 Cardiovascular Disease 7.54% Methyclothiazide 360.24 Cardiovascular Disease 7.59% Vinorelbine Tartrate 1079.11 N/A 7.68% Enalapril maleate (Vasotec) 492.52 Cardiovascular Disease 7.70% Propranolol HCl 295.8 Cardiovascular Disease 7.76% Oxytetracycline (Terramycin) 460.43 Infection 7.83% Chlorothiazide 295.72 Cardiovascular Disease 7.98% Flumequine 261.25 Metabolic Disease 8.09% Nalmefene HCl 375.89 Neurological Disease 8.11% Lafutidine 431.55 Infection 8.12% Pancuronium (Pavulon) 732.67 Cardiovascular Disease 8.13% Bleomycin sulfate 1512.62 Cancer 8.16% Maraviroc 513.67 Inflammation 8.26% Pazopanib HCl 473.98 Cancer 8.28% 139

Emtricitabine (Emtriva) 247.25 Infection 8.29% Diclofenac 318.13 Neurological Disease 8.33% Nialamide 298.34 Neurological Disease 8.34% Phenacetin 179.22 Infection 8.38% Imipramine HCl 316.87 N/A 8.38% Brucine 510.56 N/A 8.45% Busulfan (Myleran, Busulfex) 246.3 Cardiovascular Disease 8.49% Furaltadone HCl 360.75 Infection 8.50% Flucytosine (Ancobon) 129.09 Infection 8.55% Nabumetone 228.29 Inflammation 8.58% Cilostazol 369.46 Cardiovascular Disease 8.65% Chloramphenicol (Chloromycetin) 323.13 Infection 8.66% Tropisetron 320.81 Neurological Disease 8.67% Divalproex sodium 310.41 Neurological Disease 8.73% Diclofenac Diethylamine 369.29 Neurological Disease 8.74% Mecarbinate 233.26 Metabolic Disease 8.74% valganciclovir hydrochloride 390.82 Endocrinology 8.74% Fluorometholone Acetate 418.5 Inflammation 8.78% Oxcarbazepine 252.27 Neurological Disease 8.81% Azelastine hydrochloride (Astelin) 418.36 Neurological Disease 8.84% Methacycline hydrochloride 478.88 Cancer 8.86% (Physiomycine) Betamethasone (Celestone) 392.46 Inflammation 8.94% Phenindione (Rectadione) 222.24 Cardiovascular Disease 8.96% Methazolamide 236.27 Neurological Disease 9.01% Maprotiline hydrochloride 313.86 Neurological Disease 9.01% Articaine HCl 320.84 Neurological Disease 9.06% Betaxolol (Betoptic) 307.43 Neurological Disease 9.08% Suplatast tosylate 499.64 Cardiovascular Disease 9.08% Genipin 226.23 N/A 9.08% Procaine (Novocaine) HCl 272.77 Neurological Disease 9.19% Fenoprofen calcium 522.6 Inflammation 9.20% Gefitinib (Iressa) 446.9 Cancer 9.27% Isoxicam 335.34 N/A 9.29% Nalidixic acid (NegGram) 232.24 Infection 9.30% Meglumine 195.21 N/A 9.33% Mianserin hydrochloride 300.83 Neurological Disease 9.35% Ambrisentan 378.42 Neurological Disease 9.51% Praziquantel (Biltricide) 312.41 Vermifuge 9.52% Chlorpropamide 276.74 Infection 9.63% Telmisartan (Micardis) 514.62 Cardiovascular Disease 9.65% Pramoxine HCl 329.86 Neurological Disease 9.70% Darunavir Ethanolate (Prezista) 593.73 Infection 9.71% Dimethyl Fumarate 144.13 Inflammation 9.73% Mometasone furoate 521.43 Inflammation 9.74% Azithromycin (Zithromax) 748.98 Cancer 9.81% Pranoprofen 255.27 Inflammation 9.91%

140

Ivermectin 875.09 Vermifuge 10.00% Amoxicillin sodium (Amox) 387.39 Infection 10.08% Proadifen HCl 390.01 N/A 10.09% Sulfisoxazole 267.3 Infection 10.11% Clorprenaline HCL 250.16 Cardiovascular Disease 10.13% Doxapram HCl 432.98 Neurological Disease 10.19% Doxorubicin (Adriamycin) 579.98 Cancer 10.22% Bendamustine HCL 394.72 Cancer 10.26% Zalcitabine 211.22 Infection 10.27% Menadione 172.18 Endocrinology 10.28% Desonide 416.51 Inflammation 10.29% Nafamostat mesylate 539.58 Cardiovascular Disease 10.33% Venlafaxine 313.86 Neurological Disease 10.36% Noscapine HCl 449.88 N/A 10.39% Cilnidipine 492.52 Cardiovascular Disease 10.41% Dyclonine HCl 325.87 Inflammation 10.44% Guanabenz acetate 291.13 Endocrinology 10.46% Abitrexate (Methotrexate) 454.44 Cancer 10.50% Evista (Raloxifene Hydrochloride) 510.04 Endocrinology 10.50% Pidotimod 244.27 Immunology 10.66% Anisotropine Methylbromide 362.35 Neurological Disease 10.66% Bepotastine Besilate 547.06 Cancer 10.70% Acetylcholine chloride 181.66 Neurological Disease 10.73% Potassium iodide 166 Endocrinology 10.81% Trospium chloride (Sanctura) 427.96 Cardiovascular Disease 10.83% Ramipril (Altace) 416.51 Cardiovascular Disease 10.88% Acarbose 645.6 Metabolic Disease 10.90% Ibuprofen (Advil) 206.28 Inflammation 10.94% Aztreonam (Azactam, Cayston) 435.43 Infection 10.94% Dyphylline (Dilor) 254.24 Respiratory Disease 10.94% Spironolactone 416.57 Infection 10.99% Methylprednisolone 374.47 Immunology 11.00% sulfacetamide sodium 236.22 Cardiovascular Disease 11.08% Pentamidine 413.34 Infection 11.10% Phenylephrine HCl 203.67 Endocrinology 11.10% Clomipramine HCl 351.31 N/A 11.13% Sulfamerazine 264.3 Infection 11.14% Amitriptyline HCl 313.86 Infection 11.22% Valnemulin HCl 601.28 Infection 11.22% Ondansetron hydrochloride (Zofran) 329.82 Infection 11.22% Hydrochlorothiazide 297.74 Cardiovascular Disease 11.23% Thiabendazole 201.25 Vermifuge 11.25% Terazosin HCl (Hytrin) 459.92 Infection 11.31% Dasatinib (BMS-354825) 488.01 Cancer 11.33% Naproxen (Aleve) 252.24 Inflammation 11.37% Sasapyrine 258.23 Inflammation 11.39% Scopolamine hydrobromide 384.26 Respiratory Disease 11.40% 141

Cefoperazone (Cefobid) 645.67 Infection 11.41% Pyrimethamine 248.71 Immunology 11.45% Carbamazepine (Carbatrol) 236.27 Neurological Disease 11.46% Hydrocortisone (Cortisol) 362.46 Infection 11.46% Oxaliplatin (Eloxatin) 397.29 Cancer 11.48% Pizotifen malate 429.53 Inflammation 11.49% Nicotinamide (Niacinamide) 122.12 Neurological Disease 11.51% Dibucaine HCL 379.92 Endocrinology 11.51% Phenformin hydrochloride 241.72 Metabolic Disease 11.51% Hydroxyzine 2HCl 447.83 Neurological Disease 11.51% Triflupromazine HCl 388.88 N/A 11.51% Disopyramide Phosphate 437.47 Cardiovascular Disease 11.52% tetrahydrozoline hydrochloride 236.74 Inflammation 11.61% Cepharanthine 606.71 Metabolic Disease 11.61% Azacyclonol 267.37 Neurological Disease 11.62% Rocuronium bromide 609.68 Neurological Disease 11.69% Bisoprolol fumarate 441.52 N/A 11.69% Edaravone (MCI-186) 174.2 Cardiovascular Disease 11.80% Deferasirox (Exjade) 373.36 Endocrinology 11.84% Amfebutamone (Bupropion) 276.2 Infection 11.87% Cyclophosphamide monohydrate 279.1 Cancer 11.89% Noradrenaline bitartrate monohydrate 337.28 Metabolic Disease 11.89% (Levophed) Tiopronin (Thiola) 163.19 Cardiovascular Disease 11.89% Naltrexone HCl 377.86 Neurological Disease 11.91% Ethamsylate 263.31 Cardiovascular Disease 11.93% Flunixin meglumin 491.46 Immunology 12.01% Roxithromycin (Roxl-150) 837.0673 Metabolic Disease 12.03% Trometamol 121.14 N/A 12.03% Nelarabine (Arranon) 297.27 Cancer 12.04% Lenalidomide 259.26 Cardiovascular Disease 12.06% Nefiracetam (Translon) 246.3 Neurological Disease 12.10% Vardenafil (Vivanza) 579.11 Infection 12.12% Fluocinolone acetonide (Flucort-N) 452.4999 Infection 12.13% Atazanavir sulfate 802.93 Cancer 12.13% Urapidil HCl 423.94 Respiratory Disease 12.17% Sertraline HCl 342.69 Inflammation 12.17% Albendazole Oxide (Ricobendazole) 281.33 Infection 12.22% Didanosine (Videx) 236.23 Infection 12.22% Cladribine 285.69 Cancer 12.26% Clofarabine 303.68 Cancer 12.30% Thioridazine HCl 407.04 N/A 12.31% Carvedilol 406.47 Cardiovascular Disease 12.36% Brinzolamide 383.51 Neurological Disease 12.36% Valsartan (Diovan) 435.52 Cardiovascular Disease 12.39% Brompheniramine 435.31 Infection 12.41% Methocarbamol (Robaxin) 241.24 Neurological Disease 12.46%

142

Lornoxicam (Xefo) 371.82 Inflammation 12.50% Suprofen (Profenal) 260.31 Inflammation 12.51% Lamivudine (Epivir) 229.26 Infection 12.61% Dexamethasone 392.46 Inflammation 12.63% Azatadine dimaleate 522.55 Infection 12.63% Atomoxetine HCl 291.82 Neurological Disease 12.64% Alibendol 251.28 Neurological Disease 12.68% Clarithromycin (Biaxin, Klacid) 747.95 Neurological Disease 12.69% Glipizide (Glucotrol) 445.54 Endocrinology 12.69% Cimetidine (Tagamet) 252.34 Inflammation 12.75% Rebamipide 370.79 Infection 12.86% Clozapine (Clozaril) 326.82 Cardiovascular Disease 12.87% Celecoxib 381.37 Inflammation 12.98% Trifluoperazine 2HCl 480.42 Neurological Disease 13.00% Antazoline HCl 301.81 Neurological Disease 13.00% Amisulpride 369.48 Neurological Disease 13.01% Pyrilamine Maleate 401.46 N/A 13.05% Erythromycin Ethylsuccinate 862.05 Infection 13.19% Ivabradine HCl (Procoralan) 505.05 Neurological Disease 13.25% Zanamivir 332.31 N/A 13.26% Antipyrine 188.23 Infection 13.27% XL-184 (Cabozantinib) 501.51 Cancer 13.28% Aminophylline (Truphylline) 420.43 Respiratory Disease 13.30% Aceclidine HCl 205.68 N/A 13.42% Bimatoprost 415.57 Cardiovascular Disease 13.46% Mexiletine HCl 215.72 Cardiovascular Disease 13.49% Balofloxacin 389.42 Metabolic Disease 13.50% Dropropizine 236.31 Respiratory Disease 13.51% Diclazuril 407.64 Infection 13.53% Ciclopirox ethanolamine 268.35 Infection 13.53% Pramipexole (Mirapex) 211.33 Neurological Disease 13.55% Orlistat (Alli, Xenical) 495.73 Metabolic Disease 13.56% Atropine sulfate monohydrate 694.83 Respiratory Disease 13.58% Nilotinib (AMN-107) 529.52 Cancer 13.59% Glyburide (Diabeta) 494 Endocrinology 13.60% Desloratadine 310.82 Cardiovascular Disease 13.61% Benzydamine Hydrochloride 345.87 Inflammation 13.62% 2-Thiouracil 128.15 Endocrinology 13.63% Hydroxyurea (Cytodrox) 76.05 Cancer 13.65% Tetracycline HCl 480.9 Infection 13.68% Thalidomide 258.23 Immunology 13.71% Voglibose 267.28 Metabolic Disease 13.74% Hyoscyamine (Daturine) 289.37 Neurological Disease 13.75% Pramiracetam 269.38 Endocrinology 13.89% Tacrine HCl 234.72 N/A 13.96% Candesartan cilexetil (Atacand) 610.66 Cardiovascular Disease 14.06% Topiramate 339.36 Neurological Disease 14.08% 143

Proparacaine HCl 330.85 Neurological Disease 14.11% 5-Aminolevulinic acid hydrochloride 167.59 Neurological Disease 14.13% Sulfameter (Bayrena) 280.3 Infection 14.14% Terbinafine (Lamisil, Terbinex) 291.43 Infection 14.14% Vorinostat (SAHA) 264.3 Cancer 14.19% Erdosteine 249.31 Respiratory Disease 14.19% Azaguanine-8 152.11 Cancer 14.21% Zileuton 236.29 Respiratory Disease 14.32% Aminosalicylate sodium 211.15 Neurological Disease 14.34% Zidovudine (Retrovir) 267.24 Cardiovascular Disease 14.37% Reserpine 608.68 Cardiovascular Disease 14.41% Fludarabine Phosphate (Fludara) 365.21 Cancer 14.48% Benztropine mesylate 403.53 Infection 14.50% D-Mannitol (Osmitrol) 182.17 Cardiovascular Disease 14.54% Monobenzone (Benoquin) 200.23 Metabolic Disease 14.55% Spiramycin 843.058 Infection 14.56% Furosemide (Lasix) 330.74 Cardiovascular Disease 14.57% Mosapride citrate 614.02 Neurological Disease 14.58% Flumazenil 303.29 Neurological Disease 14.61% Trimethoprim 290.32 Infection 14.62% Mupirocin 500.62 N/A 14.63% Amprolium HCl 315.24 Metabolic Disease 14.65% Alfuzosin hydrochloride (Uroxatral) 425.91 Cardiovascular Disease 14.65% Acemetacin (Emflex) 415.82 Infection 14.66% Dirithromycin 835.07 Infection 14.70% Agomelatine 243.3 N/A 14.74% Gabexate mesylate 417.48 Cardiovascular Disease 14.75% Mirabegron (YM178) 396.51 Cancer 14.79% Sodium 4-aminohippurate Hydrate 216.17 N/A 14.80% Everolimus (RAD001) 958.22 Cancer 14.80% Guaifenesin (Guaiphenesin) 198.22 Respiratory Disease 14.81% Phthalylsulfacetamide 362.36 N/A 14.84% Dicyclomine HCl 345.95 N/A 14.87% Ouabain 728.77 Neurological Disease 14.88% Orphenadrine citrate (Norflex) 461.5 Infection 14.90% Salicylanilide 213.23 Infection 14.97% Buflomedil HCl 343.85 Neurological Disease 15.00% Nevirapine (Viramune) 266.3 Infection 15.06% Uracil 112.09 N/A 15.06% Novobiocin sodium (Albamycin) 634.61 Cardiovascular Disease 15.08% Vitamin B12 1355.37 Metabolic Disease 15.09% Cetirizine Dihydrochloride 461.81 Inflammation 15.14% Diperodon HCl 433.93 N/A 15.26% Aprepitant (MK-0869) 534.43 Neurological Disease 15.32% Esmolol HCl 331.83 Cardiovascular Disease 15.33% Moguisteine 339.41 Respiratory Disease 15.34% Methylthiouracil 142.18 Infection 15.34% 144

Cysteamine HCl 113.61 Metabolic Disease 15.35% Diphemanil methylsulfate 389.51 Neurological Disease 15.39% Miglitol (Glyset) 207.22 Neurological Disease 15.39% Nebivolol (Bystolic) 441.9 Cardiovascular Disease 15.39% Levetiracetam 170.21 Neurological Disease 15.42% Ponatinib (AP24534) 532.56 Cancer 15.50% Ropinirole HCl 296.84 Neurological Disease 15.53% Mepenzolate Bromide 205.68 N/A 15.56% Nateglinide (Starlix) 317.42 Immunology 15.62% Haloperidol (Haldol) 375.86 Neurological Disease 15.62% Propylthiouracil 170.23 Endocrinology 15.75% Loxapine Succinate 445.9 Neurological Disease 15.81% Aniracetam 219.24 Neurological Disease 15.86% Simvastatin (Zocor) 418.57 Cardiovascular Disease 15.90% Bosentan 551.61 N/A 15.93% Triamcinolone Acetonide 434.5 Inflammation 15.99% Aminocaproic acid (Amicar) 131.17 Cardiovascular Disease 16.03% Procyclidine HCl 323.9 N/A 16.03% Riluzole (Rilutek) 234.2 Neurological Disease 16.06% Ceftiofur HCl 560.02 N/A 16.08% Isoprenaline hydrochloride 247.72 Infection 16.09% Ruxolitinib (INCB018424) 306.37 Cancer 16.17% Sodium salicylate 161.11 Infection 16.24% Finasteride 372.54 Endocrinology 16.27% Ibutilide fumarate 885.23 Cardiovascular Disease 16.30% Flumethasone 410.45 Endocrinology 16.32% Rufinamide (Banzel) 238.19 Neurological Disease 16.36% Tolbutamide 270.35 Cancer 16.36% Olanzapine (Zyprexa) 312.4398 Infection 16.38% Disulfiram (Antabuse) 296.54 Neurological Disease 16.41% Amiloride hydrochloride (Midamor) 266.09 Metabolic Disease 16.42% Rasagiline mesylate 267.34 Cardiovascular Disease 16.43% Metaproterenol Sulfate 520.59 Respiratory Disease 16.43% Captopril (Capoten) 217.29 Metabolic Disease 16.46% Clindamycin 424.98 Infection 16.48% Cisplatin 300.05 N/A 16.48% Meticrane 275.34 Cardiovascular Disease 16.53% Pomalidomide 273.24 Cancer 16.54% Telaprevir (VX-950) 679.85 Infection 16.66% Deflazacort (Calcort) 441.52 Endocrinology 16.67% Methenamine (Mandelamine) 140.19 Inflammation 16.67% Famprofazone 377.52 Inflammation 16.70% Procodazole 190.2 N/A 16.72% Carfilzomib (PR-171) 719.91 Cardiovascular Disease 16.72% Spectinomycin hydrochloride 405.27 Cardiovascular Disease 16.76% Bindarit 324.37 Cancer 16.78% Zonisamide 212.23 Neurological Disease 16.87% 145

Piromidic Acid 288.32 N/A 16.87% Acetarsone 275.09 N/A 16.96% Estriol 288.39 Neurological Disease 17.02% Rosuvastatin calcium (Crestor) 500.57 Infection 17.03% Pramipexole dihydrochloride 302.26 Neurological Disease 17.03% monohydrate Sulfamethoxazole 253.28 Infection 17.04% Dienogest 311.42 Endocrinology 17.13% Valproic acid sodium salt (Sodium 166.19 Cardiovascular Disease 17.13% valproate) Homatropine Bromide 356.25 Infection 17.14% Temsirolimus (Torisel) 1030.29 Cancer 17.15% Irsogladine 256.09 Neurological Disease 17.20% Tenofovir Disoproxil Fumarate 635.51 N/A 17.29% Mycophenolic (Mycophenolate) 320.34 Infection 17.30% Docetaxel (Taxotere) 807.88 Cancer 17.35% Pridinol Methanesulfonate 391.52 N/A 17.38% Oxybuprocaine HCl 344.88 Neurological Disease 17.40% Lamotrigine 256.09 Cancer 17.44% Ftorafur 200.17 Cancer 17.51% Phenylbutazone (Butazolidin, Butatron) 308.37 Cancer 17.63% Chloroxine 214.05 Infection 17.68% Deoxycorticosterone acetate 372.5 Endocrinology 17.71% Benzthiazide 431.94 Cardiovascular Disease 17.71% Gimeracil 145.54 Neurological Disease 17.73% Bufexamac 223.27 Metabolic Disease 17.75% Naloxone HCl 363.84 Cancer 17.77% Bumetanide 364.42 Cardiovascular Disease 17.78% Ginkgolide A 408.4 Cardiovascular Disease 17.79% Aspartame 294.3 Metabolic Disease 17.80% Nizatidine 331.46 Metabolic Disease 17.82% Bupivacaine hydrochloride (Marcain) 324.89 Neurological Disease 17.87% Dexrazoxane Hydrochloride 304.73 Cardiovascular Disease 17.90% Mifepristone (Mifeprex) 429.59 Metabolic Disease 17.92% Loperamide hydrochloride 513.5 Infection 17.92% Ampiroxicam 447.46 Cardiovascular Disease 17.92% Cyromazine 166.18 Vermifuge 18.00% Cinepazide maleate 533.57 Inflammation 18.04% Tolperisone HCl 281.82 Neurological Disease 18.09% Amprenavir (Agenerase) 505.63 Infection 18.13% Metaraminol Bitartrate 317.29 N/A 18.16% Tioxolone 168.17 Endocrinology 18.19% Diphenylpyraline HCl 317.85 Neurological Disease 18.23% Serotonin HCl 212.68 Neurological Disease 18.27% (+,-)-Octopamine HCl 189.64 Immunology 18.30% Dydrogesterone 312.45 Endocrinology 18.32% Arbidol HCl 513.88 Cardiovascular Disease 18.37% Sulbactam sodium (Unasyn) 255.22 Infection 18.39% 146

Coumarin 146.14 N/A 18.44% Rotigotine 315.47 N/A 18.54% Rizatriptan Benzoate 391.47 N/A 18.74% Ractopamine HCl 337.84 N/A 18.75% Carbidopa 226.23 Neurological Disease 18.80% Vildagliptin (LAF-237) 303.4 Metabolic Disease 18.86% Aminothiazole 100.14 Infection 18.90% Fluvastatin sodium (Lescol) 433.45 Cardiovascular Disease 18.93% Bisacodyl 361.39 Cardiovascular Disease 19.03% Tripelennamine HCl 291.82 Neurological Disease 19.06% Carbachol 182.65 N/A 19.09% Fexofenadine HCl 538.12 Neurological Disease 19.09% Saxagliptin (BMS-477118,Onglyza) 315.41 Infection 19.13% Estrone 270.37 Endocrinology 19.21% Ribavirin (Copegus) 244.2086 Metabolic Disease 19.22% Linagliptin (BI-1356) 472.54 Cancer 19.22% Milrinone (Primacor) 211.22 Cardiovascular Disease 19.25% Alverine Citrate 473.56 Digestive system disease 19.31% Sodium ascorbate 201.13 Endocrinology 19.43% Licofelone 379.88 N/A 19.49% Decitabine 228.21 Cardiovascular Disease 19.51% Eprosartan Mesylate 520.62 Cardiovascular Disease 19.57% Chlorzoxazone 169.57 Metabolic Disease 19.60% FK-506 (Tacrolimus) 804.02 N/A 19.71% Pimecrolimus 810.45 N/A 19.71% Daidzein 254.24 Cardiovascular Disease 19.73% L-Adrenaline (Epinephrine) 183.2 Cardiovascular Disease 19.74% PMSF (Phenylmethylsulfonyl Fluoride) 174.19 Inflammation 19.77% Chlorocresol 142.58 N/A 19.80% Rifampin (Rifadin, Rimactane) 822.94 Infection 19.86% Isovaleramide 101.15 Neurological Disease 19.87% Sulfadiazine 250.28 Infection 19.87% Cyclamic acid 179.24 Inflammation 19.89% Cefdinir (Omnicef) 395.41 Infection 19.89% Mesna (Uromitexan, Mesnex) 164.18 Cancer 19.91% Sulbactam 233.24 Infection 19.93% Meptazinol HCl 269.81 N/A 19.95% Dexamethasone acetate 434.5 Inflammation 19.99% Aspirin (Acetylsalicylic acid) 180.16 Cancer 20.06% Exemestane 296.4 Endocrinology 20.06% Pasiniazid 290.27 N/A 20.13% Ethynodiol diacetate 384.51 Endocrinology 20.17% Entecavir hydrate 295.29 Infection 20.17% Decamethonium bromide 418.29 Neurological Disease 20.30% Elvitegravir (GS-9137) 447.88 Immunology 20.30% Naphazoline hydrochloride (Naphcon) 246.74 Neurological Disease 20.50% Azlocillin sodium salt 484.48 Neurological Disease 20.56% 147

Bemegride 155.19 N/A 20.64% Homatropine Methylbromide 370.28 N/A 20.69% Gallamine triethiodide (Flaxedil) 891.53 Inflammation 20.72% Sulfamethazine 278.33 Endocrinology 20.89% Piperacillin Sodium 539.54 Infection 20.95% DL-Carnitine hydrochloride 197.66 Cardiovascular Disease 21.04% Xylazine HCl 256.79 Cardiovascular Disease 21.10% Darifenacin HBr 507.46 Infection 21.10% Nelfinavir Mesylate 663.89 N/A 21.12% Acebutolol HCl 372.89 Neurological Disease 21.18% Sucralose 397.63 N/A 21.18% Felbamate 238.24 Neurological Disease 21.25% Ziprasidone hydrochloride 449.4 Neurological Disease 21.28% Doxifluridine 246.19 Immunology 21.34% Olopatadine hydrochloride (Opatanol) 373.87 Neurological Disease 21.39% Bethanechol chloride 196.68 Neurological Disease 21.45% Oxymetazoline hydrochloride 296.84 Immunology 21.45% Naratriptan HCl 371.93 Neurological Disease 21.51% Penicillin G Sodium 356.37 Infection 21.55% Conivaptan HCl (Vaprisol) 535.04 Cardiovascular Disease 21.58% Betamipron 193.2 Infection 21.63% Dicloxacillin Sodium 510.32 Infection 21.65% Retapamulin 517.76 Neurological Disease 21.72% Moxonidine 241.68 Digestive system disease 21.76% Aliskiren hemifumarate 609.83 Cardiovascular Disease 21.79% Budesonide 430.53 Endocrinology 21.80% Nicotine Ditartrate 462.46 N/A 21.82% Adrucil (Fluorouracil) 130.08 Cancer 21.86% Carbenicillin disodium 422.36 Infection 21.89% Carprofen 273.71 Inflammation 22.00% Albendazole (Albenza) 265.33 Vermifuge 22.05% Benzocaine 165.19 Respiratory Disease 22.12% Liothyronine Sodium 672.96 Endocrinology 22.17% Ciclopirox (Penlac) 207.27 Neurological Disease 22.20% Isosorbide 146.14 N/A 22.25% Atorvastatin calcium (Lipitor) 1155.34 Cardiovascular Disease 22.30% Aclidinium Bromide 564.55 Neurological Disease 22.36% Acetylcysteine 163.19 Respiratory Disease 22.41% Quinine hydrochloride dihydrate 396.91 Cardiovascular Disease 22.46% Moexipril HCl 535.03 Digestive system disease 22.52% Glimepiride 490.62 Metabolic Disease 22.53% Gemfibrozil (Lopid) 250.33 Cardiovascular Disease 22.53% Sorbitol (Glucitol) 182.17 Digestive system disease 22.60% Febuxostat (Uloric) 316.37 Inflammation 22.61% Acadesine 258.23 Cardiovascular Disease 22.73% Quinapril HCl (Accupril) 474.98 Inflammation 22.80% Fosfomycin Tromethamine 259.19 N/A 22.85% 148

Amidopyrine 231.29 Neurological Disease 22.94% Sodium Picosulfate 481.41 Metabolic Disease 22.98% Fosaprepitant dimeglumine 1004.83 Cardiovascular Disease 23.02% Otilonium Bromide 563.57 Cardiovascular Disease 23.05% Benserazide 293.7 Neurological Disease 23.14% Ampicillin sodium 371.39 Infection 23.19% Ampicillin Trihydrate 403.45 Infection 23.19% Dehydroepiandrosterone (DHEA) 288.43 Endocrinology 23.20% Prednisolone acetate (Omnipred) 402.48 Immunology 23.23% Azithromycin Dihydrate 785.02 Infection 23.29% Bazedoxifene HCl 507.06 Metabolic Disease 23.30% Oxaprozin 293.32 Inflammation 23.35% Betahistine 2HCl 209.12 N/A 23.49% Pravastatin sodium 446.51 Metabolic Disease 23.49% Lomerizine HCl 541.46 Cardiovascular Disease 23.49% Choline Chloride 139.62 N/A 23.53% Sodium nitrite 69 Neurological Disease 23.68% Azacitidine (Vidaza) 244.2 Cancer 23.69% Ropivacaine HCl 310.86 Infection 23.75% Camylofin Chlorhydrate 393.39 N/A 23.78% Mezlocillin Sodium 561.56 Infection 23.84% Benazepril hydrochloride 460.95 Cardiovascular Disease 23.90% Famciclovir (Famvir) 321.3386 Cancer 23.92% Biotin (Vitamin B7) 244.31 Infection 23.97% Beclomethasone dipropionate 521.04 Inflammation 24.01% Florfenicol 358.21 Infection 24.04% Allylthiourea 116.18 Metabolic Disease 24.04% Teriflunomide 270.21 Immunology 24.05% Catharanthine 336.43 N/A 24.08% Tetracaine hydrochloride (Pontocaine) 300.82 Endocrinology 24.10% Mercaptopurine 152.18 Cancer 24.14% Escitalopram oxalate 414.43 Infection 24.17% Fenspiride HCl 296.79 Inflammation 24.17% Sulfamethizole (Proklar) 270.33 Infection 24.24% nafcillin sodium monohydrate 454.47 Endocrinology 24.26% Bromhexine HCl 412.59 Cardiovascular Disease 24.30% Clofibric acid 214.65 Metabolic Disease 24.40% Ganciclovir 255.23 Infection 24.46% Ramelteon (TAK-375) 259.34 Neurological Disease 24.49% Capecitabine (Xeloda) 359.35 Cancer 24.54% Nifenazone 308.33 N/A 24.55% Mepivacaine HCl 282.81 Metabolic Disease 24.55% Methoxsalen (Oxsoralen) 216.19 Inflammation 24.56% Streptozotocin (Zanosar) 265.22 Cancer 24.58% Penfluridol 523.97 Neurological Disease 24.63% Azaperone 327.4 Neurological Disease 24.64% Cortisone acetate (Cortone) 402.48 Cancer 24.66% 149

Methazolastone 194.15 Cancer 24.83% Etravirine (TMC125) 435.28 Neurological Disease 24.88% Mepiroxol 125.13 N/A 24.89% Triflusal 248.16 Infection 24.89% Halobetasol Propionate 484.96 Inflammation 24.89% BIBR-1048 (Dabigatran) 627.73 Infection 24.94% Vandetanib (Zactima) 475.35 Cancer 24.96% Pentoxifylline 278.31 N/A 25.11% Tolvaptan (OPC-41061) 448.94 Metabolic Disease 25.14% Ceftazidime Pentahydrate 636.65 Infection 25.15% Azilsartan (TAK-536) 456.45 Neurological Disease 25.16% Dibenzothiophene 184.26 N/A 25.35% Bacitracin 1408.67 Infection 25.46% Calcium Gluceptate 490.42 N/A 25.46% Oxfendazole 315.35 Vermifuge 25.50% Penciclovir 253.26 Infection 25.58% Meloxicam (Mobic) 351.4 Inflammation 25.68% Isoetharine Mesylate 335.42 Cardiovascular Disease 25.68% Meclofenamate Sodium 318.13 N/A 25.94% Pazopanib 437.52 Cardiovascular Disease 25.97% Levosulpiride (Levogastrol) 341.43 Neurological Disease 26.04% Ursodiol (Actigal Urso) 392.57 Metabolic Disease 26.08% Fidaxomicin 1058.04 Infection 26.15% Guanidine HCl 95.53 Vermifuge 26.23% Altrenogest 310.43 Neurological Disease 26.37% Losartan potassium 462.01 Cardiovascular Disease 26.63% Carbimazole 186.23 Infection 26.66% Droperidol 379.43 Neurological Disease 26.69% Ifosfamide 261.09 Cancer 26.78% Tilmicosin 869.13 Infection 26.82% Pheniramine Maleate 356.42 Neurological Disease 26.83% Doxofylline 266.25 Metabolic Disease 26.84% Lansoprazole 369.36 Infection 26.91% Geniposidic acid 374.34 N/A 26.97% Tofacitinib citrate (CP-690550 citrate) 504.49 N/A 27.08% Ethoxzolamide 258.32 Neurological Disease 27.11% Itraconazole (Sporanox) 705.6503 Cancer 27.15% Cloxacillin sodium (Cloxacap) 475.88 Cardiovascular Disease 27.18% Azelnidipine 582.65 Neurological Disease 27.35% Caspofungin acetate 1213.42 Infection 27.50% Xylometazoline HCl 280.84 Infection 27.51% Beta Carotene 536.87 N/A 27.57% Mitoxantrone Hydrochloride 517.4 Cardiovascular Disease 27.63% Leflunomide 270.21 Inflammation 27.71% Paeoniflorin 480.46 N/A 27.76% Troxipide 294.35 Digestive system disease 27.91% Gliclazide (Diamicron) 323.41 Neurological Disease 27.98% 150

Ammonium Glycyrrhizinate 839.96 N/A 27.99% Fludarabine (Fludara) 285.23 Cancer 28.09% Amoxicillin (Amoxycillin) 365.4 Neurological Disease 28.41% Loratadine 382.88 Inflammation 28.50% Closantel Sodium 685.06 Vermifuge 28.58% Pyrithione zinc 317.7 Infection 28.72% Tianeptine sodium 458.93 Neurological Disease 28.89% Sodium Nitroprusside 261.92 Cardiovascular Disease 28.89% Betapar (Meprednisone) 372.455 Inflammation 28.97% Oxybutynin (Ditropan) 357.49 Neurological Disease 29.29% Betamethasone Dipropionate 504.59 N/A 29.50% Lopinavir (ABT-378) 628.8 Infection 29.50% Mitotane (Lysodren) 320.04 Cancer 29.50% Cefditoren pivoxil 620.72 Infection 29.62% Ulipristal 475.62 Infection 29.73% Tolfenamic acid 261.7 Inflammation 29.86% Azilsartan Medoxomil (TAK-491) 568.53 Cardiovascular Disease 30.61% Idebenone 338.44 Inflammation 30.94% L-Thyroxine 776.87 Neurological Disease 31.11% Ketorolac (Toradol) 376.4 Neurological Disease 31.26% Alfacalcidol 400.64 Endocrinology 31.26% Terfenadine 471.67 N/A 31.29% Estradiol valerate 356.5 Endocrinology 31.38% Chromocarb 190.15 Cardiovascular Disease 31.42% 2-Methoxyestradiol 302.41 Cancer 31.67% Racecadotril (Acetorphan) 385.48 Infection 32.01% Metronidazole (Flagyl) 171.15 Infection 32.12% toltrazuril 425.38 Infection 32.14% Loteprednol etabonate 466.95 N/A 32.15% Atovaquone (Atavaquone) 366.84 Neurological Disease 32.37% Deoxyarbutin 194.23 Cardiovascular Disease 32.39% Diethylstilbestrol (Stilbestrol) 268.35 Cancer 32.45% Oxybutynin chloride 393.95 Neurological Disease 32.49% Estradiol Cypionate 396.56 N/A 32.49% Bextra (valdecoxib) 314.36 Neurological Disease 32.53% Clofoctol 365.34 N/A 32.63% Dexlansoprazole 369.36 Cardiovascular Disease 32.67% Phenoxybenzamine HCl 340.3 Endocrinology 32.99% Clopidogrel (Plavix) 419.9 Cardiovascular Disease 33.02% Vitamin D2 396.65 Endocrinology 33.30% Vismodegib (GDC-0449) 421.3 Cancer 33.31% Benzoic acid 122.12 N/A 33.68% Rolipram 275.34 Inflammation 33.74% Probenecid (Benemid) 285.36 Metabolic Disease 34.04% Phenytoin sodium (Dilantin) 274.25 Metabolic Disease 34.09% Clindamycin palmitate HCl 699.85 Infection 34.14% Pyridoxine hydrochloride 205.64 Endocrinology 34.15% 151

Cleviprex (Clevidipine) 456.32 Cardiovascular Disease 34.20% Prasugrel (Effient) 373.44 Cardiovascular Disease 34.25% Butenafine HCl 353.93 Neurological Disease 34.43% Calcitriol (Rocaltrol) 416.64 Endocrinology 34.50% Fulvestrant (Faslodex) 606.77 Cancer 34.50% Carbenoxolone Sodium 614.72 N/A 34.85% Camptothecin 348.35 Cancer 35.06% Oxethazaine 467.64 Neurological Disease 35.44% Zafirlukast (Accolate) 575.68 Inflammation 35.68% Erlotinib HCl 429.9 Cancer 35.71% Difloxacin HCl 259.19 N/A 36.17% PCI-32765 (Ibrutinib) 440.5 Neurological Disease 36.29% Ethinyl Estradiol 296.4 Endocrinology 37.01% Benzethonium chloride 448.08 Neurological Disease 37.63% Progesterone (Prometrium) 314.46 Endocrinology 37.74% Estradiol 272.38 Endocrinology 38.05% Doxercalciferol (Hectorol) 412.65 Endocrinology 39.50% Indacaterol Maleate 508.56 Infection 39.50% Pimozide 461.55 N/A 39.88% Artemisinin 282.33 Infection 40.97% Vitamin D3 (Cholecalciferol) 384.64 Cardiovascular Disease 41.73% Naftopidil (Flivas) 392.49 Endocrinology 42.19% S-(+)-Rolipram 275.34 Cardiovascular Disease 42.72% AMG-073 HCl (Cinacalcet hydrochloride) 393.87 Endocrinology 42.77% Trilostane 329.43 Endocrinology 43.07% Paclitaxel (Taxol) 853.91 Cancer 45.50% Terbinafine hydrochloride (Lamisil) 327.89 Infection 45.83% Drospirenone 366.49 Endocrinology 45.84% Repaglinide 452.59 Endocrinology 46.56% Manidipine (Manyper) 610.7 Metabolic Disease 46.78% Gestodene 310.43 Endocrinology 47.10% Lithocholic acid 376.57 Neurological Disease 47.19% Phenytoin (Lepitoin) 252.27 Endocrinology 47.92% Pimobendan (Vetmedin) 334.37 Cardiovascular Disease 48.26% Pioglitazone (Actos) 356.44 Cancer 49.05% Vemurafenib (PLX4032) 489.92 Cancer 49.85% Idoxuridine 354.1 Infection 50.85% Lacidipine (Lacipil, Motens) 455.54 Cardiovascular Disease 51.86% Pioglitazone hydrochloride (Actos) 392.9 Metabolic Disease 52.55% Levonorgestrel (Levonelle) 312.45 Endocrinology 53.45% Rosiglitazone HCl 393.89 Cardiovascular Disease 53.56% DAPT (GSI-IX) 432.46 Cancer 54.30% Rosiglitazone maleate 473.5 Infection 54.76% Isotretinoin 300.44 Metabolic Disease 55.43% Rosiglitazone (Avandia) 357.43 Cancer 55.71% Nystatin (Mycostatin) 926.09 Infection 56.43% Chlorquinaldol 228.07 Infection 56.54% 152

Lomustine (CeeNU) 233.7 Cancer 56.61% Bexarotene 348 Cardiovascular Disease 57.29% Doxazosin mesylate 547.58 Cardiovascular Disease 60.88% Artemether (SM-224) 298.37 Cancer 61.17% Chenodeoxycholic acid 392.57 Infection 61.59% Esomeprazole sodium (Nexium) 367.4 Cancer 62.64% Carmofur 257.26 Cancer 68.01% Betamethasone valerate (Betnovate) 476.58 Inflammation 69.09% Domperidone (Motilium) 425.91 Neurological Disease 70.89% Pranlukast 481.5 Immunology 71.26% Adapalene 412.52 Inflammation 75.05% Esomeprazole magnesium (Nexium) 713.12 Digestive system disease 75.50% Amorolfine Hydrochloride 353.97 Infection 75.79% norethindrone 298.42 Neurological Disease 76.25% Rimonabant (SR141716) 463.79 Inflammation 76.40% Cyclandelate 276.37 Neurological Disease 80.12% Omeprazole (Prilosec) 345.42 Metabolic Disease 80.50% Cephalomannine 831.9 Cancer 80.78% Bezafibrate 361.82 Metabolic Disease 87.03% Aminoglutethimide (Cytadren) 232.28 Endocrinology Inhibitor Amlodipine (Norvasc) 408.88 Cardiovascular Disease Inhibitor Amlodipine besylate (Norvasc) 567.05 Cardiovascular Disease Inhibitor Amphotericin B 924.08 N/A Inhibitor Anastrozole 293.37 Endocrinology Inhibitor Apatinib (YN968D1) 493.58 Cancer Inhibitor Azathioprine (Azasan, Imuran) 277.26 Immunology Inhibitor Bifonazole 310.39 Infection Inhibitor Butoconazole nitrate 474.79 Infection Inhibitor Carbazochrome sodium sulfonate 322.27 Cancer Inhibitor Cetrimonium Bromide 364.45 Infection Inhibitor Cetylpyridinium Chloride 339.99 Infection Inhibitor Climbazole 292.76 Infection Inhibitor Clotrimazole (Canesten) 344.84 Infection Inhibitor Cobicistat (GS-9350) 776.02 Cancer Inhibitor Curcumin 368.38 N/A Inhibitor Detomidine HCl 222.71 Cardiovascular Disease Inhibitor Dexmedetomidine 200.28 Neurological Disease Inhibitor Dexmedetomidine HCl (Precedex) 236.74 Neurological Disease Inhibitor Domiphen Bromide 414.46 Infection Inhibitor Duloxetine HCl (Cymbalta) 333.88 Neurological Disease Inhibitor Econazole nitrate (Spectazole) 444.7 Neurological Disease Inhibitor Etomidate 244.29 Neurological Disease Inhibitor Fenticonazole nitrate 518.41 Neurological Disease Inhibitor Fluconazole 306.27 Infection Inhibitor Fluvoxamine maleate 434.41 Neurological Disease Inhibitor Imatinib (Gleevec) 493.6 Neurological Disease Inhibitor Imatinib Mesylate 589.71 Cancer Inhibitor 153

Isoconazole nitrate (Travogen) 479.14 Infection Inhibitor Ketoconazole 531.43 Infection Inhibitor Letrozole 285.3 Endocrinology Inhibitor Masitinib (AB1010) 498.64 Respiratory Disease Inhibitor Medetomidine HCl 236.74 Infection Inhibitor Miconazole (Monistat) 416.13 Neurological Disease Inhibitor Miconazole nitrate 479.14 Infection Inhibitor Milnacipran HCl 282.81 Endocrinology Inhibitor Nicardipine HCl 515.99 Neurological Disease Inhibitor Nicorandil (Ikorel) 211.17 Cardiovascular Disease Inhibitor Nilvadipine (ARC029) 385.37 Cancer Inhibitor Nimesulide 308.31 Cardiovascular Disease Inhibitor Nimodipine (Nimotop) 418.44 Cardiovascular Disease Inhibitor Ospemifene 378.89 N/A Inhibitor Ozagrel 228.25 Cardiovascular Disease Inhibitor Ozagrel HCl 264.71 Cardiovascular Disease Inhibitor Pilocarpine HCl 244.72 Neurological Disease Inhibitor Piroxicam (Feldene) 331.35 Inflammation Inhibitor Posaconazole 700.78 Infection Inhibitor Primaquine Diphosphate 455.34 N/A Inhibitor Protionamide (Prothionamide) 180.27 Infection Inhibitor Ritonavir 720.94 Infection Inhibitor Roflumilast (Daxas) 403.21 Neurological Disease Inhibitor Sertaconazole nitrate 500.78 Infection Inhibitor Sulconazole Nitrate 460.76 Infection Inhibitor Tazarotene (Avage) 351.46 Inflammation Inhibitor Tioconazole 387.71 Infection Inhibitor Tiratricol 621.93 Endocrinology Inhibitor Tolcapone 273.24 Metabolic Disease Inhibitor Tretinoin (Aberela) 300.4 Cancer Inhibitor Tropicamide 284.35 Neurological Disease Inhibitor Valaciclovir HCl 360.8 Infection Inhibitor Voriconazole 349.31 Infection Inhibitor Zoxazolamine 168.58 Neurological Disease N/A Adrenalone HCl 217.65 Cardiovascular Disease NSB Afatinib (BIBW2992) 485.94 Cancer NSB Avobenzone (Parsol 1789) 310.39 N/A NSB Bephenium Hydroxynaphthoate 443.53 N/A NSB Bergapten 216.19 Cancer NSB Bosutinib (SKI-606) 530.45 Cancer NSB Clonidine hydrochloride (Catapres) 266.5 Infection NSB Dacarbazine (DTIC-Dome) 182.18 Cancer NSB Glafenine HCl 409.26 N/A NSB Irinotecan 586.68 Cancer NSB Malotilate 288.38 Metabolic Disease NSB Marbofloxacin 362.36 N/A NSB Mefenamic acid 241.29 Cardiovascular Disease NSB 154

Mesoridazine Besylate 544.75 Neurological Disease NSB Moxifloxacin hydrochloride 437.89 Infection NSB Natamycin (Pimaricin) 665.73 Infection NSB Regorafenib (BAY 73-4506) 482.82 Cancer NSB Secnidazole (Flagentyl) 185.18 Infection NSB Sorafenib (Nexavar) 637.03 Cancer NSB Tolmetin Sodium 315.3 N/A NSB Tranilast (SB 252218) 327.33 Respiratory Disease NSB 9-Aminoacridine 194.23 N/A PD Acitretin 326.43 Infection PD Alprostadil (Caverject) 354.48 Endocrinology PD Amiodarone HCl 681.77 Cardiovascular Disease PD Aripiprazole (Abilify) 448.39 Neurological Disease PD Benidipine hydrochloride 542.02 Cardiovascular Disease PD Carbadox 262.22 Infection PD Clofazimine 473.4 Infection PD Clomifene citrate (Serophene) 598.08 Cancer PD Closantel 663.07 Vermifuge PD Crystal violet 407.98 Infection PD Dabrafenib (GSK2118436) 519.56 Infection PD Dronedarone HCl (Multaq) 593.22 Neurological Disease PD Eltrombopag (SB-497115-GR) 564.63 Cancer PD Entacapone 305.29 Neurological Disease PD Epalrestat 319.4 Inflammation PD Ezetimibe (Zetia) 409.4 Cardiovascular Disease PD Felodipine (Plendil) 384.25 Cardiovascular Disease PD Fenofibrate (Tricor, Trilipix) 360.83 Cardiovascular Disease PD Fluocinonide (Vanos) 494.52 Endocrinology PD Fluticasone propionate (Flonase, 500.57 Inflammation PD Veramyst) Gliquidone 527.63 Metabolic Disease PD Ipratropium bromide 412.37 Respiratory Disease PD Irinotecan HCl Trihydrate (Campto) 677.18 Neurological Disease PD Lapatinib 581.06 Neurological Disease PD Lapatinib Ditosylate (Tykerb) 925.46 Cancer PD Levosimendan 280.28 Metabolic Disease PD Mestranol 310.43 Endocrinology PD Montelukast Sodium 608.17 Respiratory Disease PD Nisoldipine (Sular) 388.41 Cardiovascular Disease PD Nitazoxanide (Alinia, Annita) 307.28 Vermifuge PD olsalazine sodium 346.2 Inflammation PD Phenazopyridine HCl 249.7 Cardiovascular Disease PD Phenothrin 350.45 N/A PD Pregnenolone 316.48 Neurological Disease PD Sunitinib Malate (Sutent) 532.56 Cancer PD Tamoxifen Citrate (Nolvadex) 563.64 Endocrinology PD Tenoxicam (Mobiflex) 337.3783 Infection PD

155

Ticagrelor 522.57 Cardiovascular Disease PD Tolnaftate 307.41 Cancer PD Topotecan HCl 457.91 Cancer PD Toremifene Citrate (Fareston, Acapodene) 598.08 Endocrinology PD triamterene 253.26 Inflammation PD Triclabendazole 359.66 Vermifuge PD Verteporfin (Visudyne) 718.79 Endocrinology PD

Figure S6: Percentage spin-state shift values obtained under near-saturating ligand concentrations for the DM BM3 heme domain A 978 compound library was screened for binding with the DM BM3 variant. For each substrate the percentage high-spin shift was calculated using a Python script and the percentage high spin equation, as described in the Methods section. The relevant compound/drug class is organised by the percentage high spin observed. Type II (inhibitor- like) Soret red shifts are labelled as “inhibitors”. N/A indicates a drug missing from the library received. NSB indicates that “no steady baseline” could be achieved due to solubility issues with the relevant ligands. PD indicates “protein depletion” as observed by heme dissociation from the DM heme domain, accompanied by protein aggregation/turbidity in many cases. The compound names (and where applicable the trade names shown in brackets), the compound mass and indicator (a rough idea of drug usage) were provided by the supplier of the FDA-approved library; Selleck Chemicals.

156

Chapter 4: Novel Insights into P450 BM3 Interactions with FDA-approved Antifungal Azole Drugs

4.1. Abstract

Flavocytochrome P450 BM3 is a natural fusion protein constructed of cytochrome

P450 and NADPH-cytochrome P450 reductase domains. P450 BM3 binds and oxidizes several mid- to long-chain fatty acids, typically hydroxylating these lipids at the -1, -2 and -3 positions. However, protein engineering has led to variants of this enzyme that are able to bind and oxidize diverse compounds, including steroids, terpenes and various human drugs. The wild-type P450 BM3 enzyme binds inefficiently to many azole antifungal drugs. However, we show that the BM3 A82F/F87V double mutant (DM) variant binds substantially tighter to numerous azole drugs than does the wild-type BM3, and that their binding occurs with more extensive heme spectral shifts indicative of complete binding of several azoles to the BM3 DM heme iron. We report here the first crystal structures of P450 BM3 bound to azole antifungal drugs – with the BM3 DM heme domain bound to the imidazole drugs clotrimazole and tioconazole, and to the triazole drugs fluconazole and voriconazole. This is the first report of any protein structure bound to the azole drug tioconazole, as well as the first example of voriconazole heme iron ligation through a pyrimidine nitrogen from its 5-fluoropyrimidine ring.

4.2. Introduction

The cytochromes P450 (P450s or CYPs) are a superfamily of heme b-binding enzymes that catalyze the oxidative modification of a huge number of organic substrates

(Munro et al., 2013). This is typically achieved through the formation of a highly reactive heme iron-oxo species known as compound I (a ferryl-oxo [FeIV=O] porphyrin radical

157 cation species) that inserts an oxygen atom into a substrate bound close to the heme in the

P450 active site (Rittle and Green, 2010, Grant et al., 2015). According to the nature of the substrate involved and its binding mode in the active site, various different catalytic outcomes can occur following the oxidation of the substrate, including hydroxylation, demethylation/dealkylation, epoxidation, deamination, decarboxylation and sulfoxidation

(Guengerich and Munro, 2013, Munro et al., 2007).

Human P450s play crucial roles including the metabolism and interconversion of steroids, the hydroxylation/epoxidation of saturated and unsaturated fatty acids (including prostaglandins), and the oxidation of pharmaceuticals and other xenobiotics to facilitate their metabolism and excretion (Yoshimoto and Auchus, 2015, Hoch et al., 2000, Abelo et al., 2000). The steroid oxidizing human P450s typically exhibit strict substrate selectivity and regioselectivity in their reactions, whereas many of the xenobiotic transforming P450s are known to be more promiscuous in substrate binding and can catalyze diverse types of oxidative reactions. Perhaps the best example of the latter class of human P450s is

CYP3A4, which is an extremely versatile enzyme that is able to metabolize a large number of drug molecules, including tamoxifen, erythromycin, codeine, steroids and a number of statin drugs (Guengerich, 2015). The conformational flexibility of CYP3A4 and its adaptability to accommodate and oxidize substrates of different sizes and chemical properties makes this P450 a prodigious catalyst in this enzyme superfamily.

In humans, the P450s are located in the endoplasmic reticulum of cells, or else in the mitochondrial inner membrane of adrenal glands (and other steroidogenic tissues), liver and kidneys (Guengerich, 2015). Human and other eukaryotic P450s are membrane- associated enzymes attached to membranes through an N-terminal transmembrane alpha- helical segment. These P450s present some challenges in protein purification in view of the needs for detergents to solubilize the enzymes, and due to limited yields of these P450s. In 158 contrast, P450s from bacteria and archaea are soluble enzymes that are devoid of membrane “anchor” regions (McLean, 2015). These enzymes are usually soluble and located in the cell cytoplasm, and they can typically be expressed in much higher amounts than their eukaryotic counterparts (Gustafsson et al., 2004, Driscoll et al., 2010). Among the bacterial P450s, one of the best-characterized enzymes is the Bacillus megaterium

CYP102A1 (P450 BM3), which Armand Fulco’s group identified as a fatty acid hydroxylase that could catalyze the hydroxylation of saturated fatty acid substrates, primarily at the -1, -2, and -3 positions (Miura and Fulco, 1975). P450 BM3 (BM3) is a natural fusion of a cytochrome P450 (N-terminal) to a FAD-, FMN- and NADP(H)- binding cytochrome P450 reductase (CPR). The BM3 CPR resembles the membrane- associated eukaryotic CPRs that transfer electrons to their cognate P450 enzymes, but is a soluble protein devoid of a membrane anchor region. BM3 has the highest catalytic rate for substrate oxidation yet reported for a P450 monooxygenase at ~285 s-1 with arachidonic acid as the substrate (Noble et al., 1999). The component P450 and CPR domains of BM3 were successfully expressed in isolation, although they no longer interacted efficiently for fatty acid hydroxylation (Munro et al., 1994, Narhi and Fulco, 1987). In addition, the

FAD/NADPH-binding (ferredoxin reductase-like) and FMN-binding (flavodoxin-like) modules were also produced in large amounts using E. coli expression systems (Daff et al.,

1997). Intact BM3 was shown to be a dimeric enzyme with NADPH-dependent electron transfer able to occur between the CPR domain of one monomer and the heme domain of the other in the BM3 dimer (Neeli et al., 2005).

Early studies on P450 BM3 demonstrated its high catalytic rate and selectivity towards medium- to long-chain fatty acid substrates. However, the catalytic proficiency of

BM3 and its convenience as a self-sufficient catalyst (requiring only NADPH and substrate for activity) led various researchers to use protein engineering strategies in order to alter its substrate specificity. There have been a number of successful studies in this area in recent 159 years, including the production of BM3 variants that can bind and hydroxylate propane to propanol, or that catalyze selective carbene transfer from diazoesters to olefins in intact E. coli cells (Peters et al., 2003, Coelho et al., 2013). Other researchers have developed mutants that can transform the sesquiterpene (+)-valencene into nootkatone and nootkatol products, with nootkatone being an important fragrance compound (Sowden et al., 2005).

More recent work in our group has used the double mutant (DM) form of the flavocytochrome P450 BM3 enzyme (A82F/F87V), in which the first mutation expands available substrate binding space in the active site, while the second mutation is more distant from the heme but causes a structural readjustment in the P450 that alters its conformational state. The DM variant appears much more flexible than wild-type (WT)

BM3, and can bind and oxidize drug molecules including omeprazole and related gastric proton pump inhibitors to produce human metabolites (e.g. 5-OH esomeprazole, rabeprazole desmethyl ether and lansoprazole sulfone) of these drugs (Butler et al., 2013,

Butler et al., 2014).

In view of the more promiscuous nature of the BM3 DM enzyme and its ability to bind a number of molecules that do not interact productively with WT BM3, we have explored the binding of a range of bulky azole antifungal drugs to the heme domain of the

BM3 DM enzyme. These azole compounds typically have modest binding affinities for

WT BM3, as evidenced by their inability to induce substantial heme spectral shifts that are indicative of either substrate-like or inhibitor-like P450 binding behaviour. The azoles were developed as inhibitors of the fungal 14-sterol demethylase (CYP51 family) enzymes, and characteristically enter the CYP51 active site and inhibit sterol demethylation by ligating to the P450 heme iron through a nitrogen atom from an imidazole or triazole group on the drug. In this paper we have quantified the binding of several azole drugs to both the WT and DM forms of the BM3 heme domain, demonstrating their much higher affinity for the DM protein. We also report the first 160 structures of the BM3 DM heme domain in complex with four different azole drugs, confirming the greater plasticity of this variant and its ability to structurally adapt to bind to several azole drug compounds.

4.3. Materials and Methods

4.3.1. Expression and purification of WT and A82F/F87V (DM) BM3 heme domain

proteins

The BM3 heme domain proteins (WT and the DM A82F/F87V variant) used in this study were expressed as described in our previous studies (Miles et al., 1992, Girvan et al., 2004). Delta-aminolevulinic acid (0.1 mM) was added to the E. coli BL21 (DE3) transformant culture to promote heme incorporation in the case of the DM heme domain when the OD600 of the culture reached 0.6. Cells were pelleted by centrifugation at 4 °C

(6000 g, 20 min). The cells were resuspended in a small volume of ice-cold buffer A (50 mM Tris plus 1 mM EDTA, pH 7.2). Protease inhibitors (two EDTA-free cOmpleteTM

-1 tablets, Roche Applied Science), MgCl2 and DNase (10 g mL ) were added to the mixture and the cell suspension was lysed by sonication on ice using a Bandelin Sonopuls sonicator at 40% power, with 12 pulses for 40 s and with 60 s between pulses). The supernatants containing the WT and DM heme domains were separated from the cell debris by centrifugation (46,000 g, 60 min, 4 °C). Ammonium sulfate was added to 30% saturation at 4 °C with slow stirring for ∼1 h to precipitate contaminant proteins. The sample was then centrifuged again (46,000 g, 15 min, 4 °C) to remove cell debris.

The WT and DM BM3 heme domain-containing suspensions were further purified on a 150 mL DEAE Sepharose™ fast flow resin (GE Healthcare Life Sciences) after buffer exchange by dialysis into buffer A. A linear gradient of 0-300 mM KCl in buffer A (1500

161 mL) was applied and the fractions containing the WT and DM heme domains (both red in color) were further purified by loading onto a 28 mL CHT hydroxyapatite type 1 resin

(Bio-Rad) column in 25 mM potassium phosphate (KPi) buffer (pH 6.5). A linear gradient of 25-300 mM KPi (600 mL) was applied to fractionate the heme proteins. For protein samples destined for X-ray crystallography, a further purification step was used. A Q-

Sepharose™ Fast Flow column was used with a gradient of 0-300 mM KCl in 50 mM KPi

(pH 7.2) (1000 mL), followed by further resolution on a HiLoad™ GF S200 16/600

Superdex™ 200 pg column (GE Healthcare Life Sciences) using a 25 mM KPi (pH 7) buffer containing 150 mM NaCl. Protein purity was determined by SDS-PAGE analysis and the Reinheitszahl (Rz) purity ratios (A418/A280) were determined by UV-Vis spectroscopy (Omura and Sato, 1964, Asahina and Omura, 1964). For X-ray crystallography, heme domain samples with Rz values over 1.5 were used. For other experiments (e.g. ligand binding), heme domain samples with Rz values of at least 1.3 were used. Lipids were removed from the proteins by passing the heme domains through a

20 mL Lipidex™-1000 column in 25 mM KPi (pH 7).

4.3.2. UV-Visible spectroscopic assays of azole drug binding to WT and DM BM3

heme domains

Experiments were performed using a Cary 50 UV-Visible spectrophotometer

(Agilent Technologies Ltd). WT and DM heme domains at concentrations of ~2-4 M were used, and binding titrations were done by the addition of small (typically 0.2-0.4 L) volumes of azole ligand stocks (1, 10 or 30 mM, to minimize amounts of solvent carrier used) to a 1 mL heme domain sample in a quartz cuvette. Hamilton syringes were used to deliver azole drugs samples until no further spectral shifts were observed. Azole drugs were obtained from Merck Chemicals (Nottingham, UK). Data were processed by the subtraction of the spectrum of each successive ligand-bound form (from a specific titration

162 set) from the spectrum of the ligand-free enzyme. This produced a set of difference spectra

(for each particular azole titration), from which wavelengths associated with the absorbance minimum (Amin) and maximum (Amax) were readily identified. For each titration point, an absorbance difference value was calculated as Amax minus Amin. These values were then plotted against the relevant azole ligand concentrations used. Data were fitted using either the Michaelis-Menten (hyperbolic) function, the Morrison equation (for tight binding substrates where the Kd value is ≤ 5x the P450 concentration) or the Hill equation for sigmoidal curves to determine Kd values (Denisov et al., 2007, Bui et al.,

2012, Morrison, 1969). Assays were done using 25 mM KPi (pH 7) at a constant temperature of 30 C.

4.3.3. EPR spectroscopy studies of WT and DM BM3 heme domains bound to azole

drugs

Samples of ligand bound protein were prepared by incubation overnight at 4 C with agitation at 10 rpm. Samples contained 200 M protein in 25 mM KPi (pH 7) with the addition of 5 mM azole drug in DMSO, or appropriate volumes of DMSO or buffer as controls. X-band EPR spectra were recorded on a Bruker ER-300D series electromagnet with a microwave source interfaced with a Bruker EMX control unit and fitted with an

ESR-9 liquid helium flow cryostat (Oxford Instruments), and a dual-mode microwave cavity from Bruker (ER-4116DM). Spectra were recorded at 10 K with a microwave power of 0.5 mW, a modulation frequency of 100 kHz and a modulation amplitude of 5 G.

4.3.4. X-ray crystallography of the DM BM3 heme domain with azole compounds

Samples of azole ligand-bound protein were prepared with 9 mg mL-1 protein and

5 mM ligand (dissolved in a 100% DMSO stock) in 25 mM KPi (pH 7), followed by an

163 overnight incubation at 4 C with mixing at 10 rpm. Any precipitate was removed by centrifugation at 14,000 rpm for ten minutes in a microfuge. Crystallography was performed using the sitting drop method at 4 C using Molecular Dimensions screening plates (PACT, SG1, MORP, JCSG+, BCS, Midas, LMB). 300 nL DM heme domain was added to plates using a Mosquito liquid handling robot (TTP LabTech Ltd). Where needed, seeding was implemented to improve crystal size and quality using Molecular Dimensions screening plates. In these cases, 250 nL of the DM heme domain (prepared as above) were mixed with 50 nL seed stock and 300 nL mother liquor. Data were collected at Diamond synchrotron beamlines and reduced and scaled using XDS (28). Structures were solved and refined by molecular replacement with a previously solved BM3 heme domain structure

(PDB 4KF2) using Phenix (Adams et al., 2010), PDB re-do (Joosten et al., 2014) and Coot

(Emsley et al., 2010).

4.4. Results

4.4.1. Ligand binding assays of azole drugs to the WT and A82F/F87V (DM) BM3

heme domains

In their resting, substrate-free forms, the WT and DM BM3 heme domains are in a predominantly low-spin ferric state with the heme iron axially coordinated by cysteine thiolate (proximal) and water (distal) ligands. The weakly bound distal water ligand is readily displaced by a substrate to produce a 5-coordinate high-spin (HS) ferric species that

(in the case of the intact BM3 enzyme) is readily reduced to the ferrous state by an electron delivered from the CPR domain and obtained from NADPH. This leads to dioxygen (O2) binding to heme iron and to the progression of the P450 catalytic cycle that leads to monooxygenation of the substrate (Denisov et al., 2005). UV-visible spectroscopic

164

WT BM3 DM BM3 MW Azole (Da) Kd (M) Shift Binding mode Kd (M) Shift Binding mode

Bifonazole 310.4 3.54 ± 0.05 423 nm Hill 0.13 ± 0.02 425 nm Morrison

Butoconazole nitrate 474.8 3.80 ± 0.12 422 nm Hill 0.47 ± 0.02 424 nm Hill

Climbazole 292.8 2.41 ± 0.10 424 nm Morrison 0.32 ± 0.01 424 nm Hill

Clotrimazole* 344.8 32.8 ± 1.9 421 nm Hill 0.42 ± 0.02 423 nm Hill

Econazole nitrate 444.7 19.3 ± 1.1 421 nm Michaelis-Menten 0.53 ± 0.04 423 nm Hill

Fenticonazole nitrate 518.4 1.89 ± 0.05 419 nm Hill 0.52 ± 0.03 424 nm Hill

Fluconazole* 306.3 NB - - 2.91 ± 0.20 421 nm Morrison

Isoconazole nitrate 479.1 12.0 ± 0.5 419 nm Michaelis-Menten 0.44 ± 0.02 422 nm Hill

Itraconazole 705.6 NB - - 0.05 ± 0.01 421 nm Morrison

Ketoconazole 531.4 4.41 ± 0.15 423 nm Hill 0.50 ± 0.02 423 nm Hill

Miconazole 416.1 6.94 ± 0.15 421 nm Hill 0.63 ± 0.03 423 nm Hill

Posaconazole 700.8 NB - - 2.69 ± 0.05 422 nm Hill

Ravuconazole 437.5 NB - - 0.34 ± 0.02 422 nm Morrison

Sertaconazole nitrate 500.8 2.44 ± 0.14 419 nm Hill 0.46 ± 0.01 423 nm Hill

Sulconazole nitrate 460.8 8.93 ± 0.54 421 nm Michaelis-Menten 0.52 ± 0.01 424 nm Hill

Tioconazole* 387.7 24.5 ± 1.1 422 nm Michaelis-Menten 0.39 ± 0.03 423 nm Hill

Voriconazole* 349.3 NB - - 1.90 ± 0.13 420 nm Morrison

Table 3: Azole drug binding constants and associated spectral shifts for WT and DM BM3 heme domains Several different azole drugs were shown to bind to the WT and DM BM3 heme domains. The table shows the relevant Kd values for each azole drug (NB indicates instances where no binding is apparent in UV-visible spectroscopic titrations) and the final Soret wavelength shift observed (from the starting wavelength of 418 nm in each case) at apparent saturation with each azole. The Kd values were determined as described in the Methods section, using either a hyperbolic function (Michaelis-Menten), the Hill function or the Morrison equation for tight-binding ligands, as appropriate. Green text indicates imidazole drugs and blue text indicates triazole drugs. The azoles marked with an asterisk are those for which crystal structures of azole-bound complexes of the BM3 DM heme domain were solved. No significant heme absorbance changes were observed for the WT BM3 heme domain on titration with the triazole drugs. 165 titrations can be used to quantify the binding of substrates to intact P450 BM3 and its heme domain based on low-spin (LS) to HS type I spectral shifts (from ~418 nm to ~390 nm).

The azole drugs described here all typically elicit a type II (red) Soret shift that is indicative of the displacement of the heme iron axial water ligand by nitrogen ligation to the heme iron atom. As shown in Table 3, heme spectral shifts of different magnitudes occur according to the particular azole drug used, with more substantial Soret shifts typically observed for the BM3 DM heme domain on binding to azole inhibitor ligands

(Figure 19). P450 UV-visible binding assays and other spectroscopic studies using azoles are often hindered by the low solubility of many of these drugs. However, we were able to determine binding affinities by UV-visible spectroscopic titration for almost all of the azole drugs used, with both the WT and DM heme domains. Figure 20 shows structures of the 17 azole drugs that were used in these binding studies. The azoles used range widely in

Figure 19: Binding of tioconazole to WT and DM BM3 heme domains The main panels show spectral titrations of the WT BM3 heme domain (Panel A) and of the DM BM3 heme domain (Panel B) with tioconazole. In both cases, tioconazole binding induces a Soret red shift accompanied by decreased Soret intensity. The arrows indicate the direction of absorption change in each case. Tioconazole induces a more substantial Soret shift (from 418 nm to 423 nm) in the DM BM3 heme domain, and binds ~50-fold tighter to the DM heme domain than to the WT heme domain. The insets show difference spectra generated by subtracting each successive ligand-bound spectrum from that of the ligand- free heme domain. Kd values were determined by subtracting the absorbance value at the trough from that at the peak in each case (using the same wavelength pair throughout the titration) and by data fitting as described in the Methods section.

166

Figure 20: Structures of azole antifungal drugs used in binding studies with the P450 BM3 WT and DM heme domains All the azoles shown are drugs that are currently (or were previously) marketed for the treatment of a variety of topical and/or systemic fungal infections. The binding of these compounds to the WT and A82F/F87V (DM) forms of the BM3 heme domain was analyzed using UV-visible spectroscopic titrations and by EPR analysis in a number of cases. Imidazoles are shown in green and triazoles are in blue.

size and structure, and contain either an imidazole or a triazole functional group (or two triazole groups in the case of fluconazole). The triazoles typically elicit less substantial red shifts and produce higher Kd values for the WT and DM heme domains in comparison to the imidazole drugs. Spectral binding data were fitted using hyperbolic (Michaelis-

Menten) or tight-binding (Morrison) equations, or by using the Hill function in cases where there was a clear sigmoidal dependence of heme absorbance change on azole drug concentration (Hill, 1910, Michaelis-Menten, 1913, Morrison, 1969).

167

The WT BM3 heme domain is able to bind many of the azole drugs. However, the

DM heme domain binds these azoles far more tightly, and much greater heme spectral shifts are induced on their binding to the DM heme domain. This is a result of the combined structural rearrangements that are induced by the A82F mutation (resulting in an enlarged active site) and by the F87A mutation (where removal of the Phe87 phenyl side chain creates further binding space close to the heme). Examples of azole drug binding titrations are shown in Figure 19 for the imidazole drug tioconazole with WT and DM heme domains. The binding affinities for the WT and DM heme domains are 18.4 ± 0.9

M and 0.39 ± 0.03 M, respectively. Table 3 shows the collated datasets for the binding of the 17 different azoles to WT and DM heme domains, including spectral shifts observed,

Kd values for azole binding, and the binding model used (Michaelis-Menten, Hill or

Morrison equations). Only in the cases of the five triazole drugs with the WT heme domain were negligible changes in the heme spectrum observed.

4.4.2. Electron paramagnetic resonance (EPR) studies of WT and A82F/F87V (DM)

BM3 heme domains bound to azole drugs

X-band EPR was used to analyze the binding of azole drugs to the WT and DM

BM3 heme domains in their ferric state. There was no significant HS component seen in any of the EPR spectra, consistent with the known properties of the BM3 WT and DM heme domains in their resting hexacoordinated states, and with the low temperature required for EPR spectroscopic analysis of ferric hemes (10 K in this case) (Belcher et al.,

2014). The azole drugs are strong ligands that either displace the water ligand to coordinate the heme iron directly through nitrogen atoms from imidazole or triazole rings, or in some cases may instead bind indirectly through a retained distal water ligand (Seward et al.,

2006). The WT, ligand-free heme domain shows a single EPR species with g-values of gz

168

Figure 21: X-band EPR spectra for the WT and DM BM3 heme domain complexes with azole drugs Selected spectra for the WT and DM BM3 heme domains are shown in (i) the azole drug- free state, (ii) in presence of DMSO (5% v/v), and (iii) in complex with the azole drugs bifonazole, fluconazole and sertaconazole (with 5% DMSO). The g-values are indicated in each spectrum, and the EPR data for these and other species are tabulated in Figure S8 and Figure S9. Spectra were recorded at 10 K with a microwave power of 0.5 mW, a modulation frequency of 100 kHz and a modulation amplitude of 5 G. DMSO was used to solubilize the azole drugs, but can itself also interact with heme iron in these enzymes.

= 2.40, gy = 2.25, gx = 1.92 (2.40/2.25/1.92), while the DM heme domain shows some heterogeneity with two sets of g-values at 2.44/2.25/1.89 and 2.41/2.25/1.91. The heterogeneity likely arises due to the greater conformational flexibility of the DM heme domain, for which the ligand-free X-ray crystal structure revealed a P450 conformation similar to that for a substrate-bound form of the WT heme domain (Butler et al., 2013).

Dimethyl sulfoxide (DMSO) was used as a solvent for the azole drugs and was found to influence the EPR spectra, likely through interactions made with the heme iron through its sulfur atom (Kuper et al., 2012). Figure 21 shows overlaid EPR spectra for the WT and

DM heme domains in the ligand-free state and when bound to various azole ligands, with 169 g- values shown. In general, the imidazole drugs induce more substantial changes to the

EPR spectrum than do the triazoles for the WT heme domain, although fluconazole and posaconazole both produce significant shifts in the g-values for the DM heme domain.

DMSO binding to the WT heme domain results in a triplet set of g-values, whereas its binding to the DM heme domain produces a homogeneous EPR spectrum. As observed in the spectral binding assays, no clear evidence was obtained from EPR that was indicative of triazole drug binding to the WT heme domain. The EPR spectral changes seen in the presence of the triazoles were essentially the same as those observed with DMSO alone.

4.4.3. X-ray crystallography of azole-bound DM heme domain complexes

Crystal structures were obtained for the DM BM3 heme domain bound to clotrimazole (PDB 6H1T), fluconazole (PDB 6H1S), tioconazole (PDB 6H1L) and voriconazole (PDB 6H1O). Voriconazole (Mast et al., 2010), clotrimazole (Montemiglio et al., 2016) and fluconazole (Podust et al., 2001) have previously been crystallized in complex with eukaryotic P450s. Structures of fluconazole complexes with M. tuberculosis

CYP121A1 and CYP51B1 enzymes have also been solved (Seward et al., 2006, Podust et al., 2001). However, the structure of a P450 bound to tioconazole has not been reported previously. In order to produce the DM heme domain complexes, the protein was co- crystallized with 5 mM ligand (added in a DMSO stock). Structures of DMSO-bound P450

BM3 heme domain crystals have been reported previously (Kuper et al., 2012). However, no DMSO was observed in any of the structures reported in this study. All crystals diffracted to a resolution of ~2.2-1.8 Å. A table of crystallographic data for the four DM heme domain azole-bound structures is shown as Figure S7.

Previous studies revealed the binding of an imidazole ligand to the BM3 DM heme iron with a bond distance of 1.80 Angstroms (PDB 4KF2) (Butler et al., 2013). Three

170

Figure 22: Crystal structures of the BM3 heme domain A82F/F87V mutants in complex with different azole drugs Panel A shows an overlay of various azole-complexed A82F/F87V (DM) BM3 heme domain structures, revealing that a similar overall conformation (“open”, in blue) is adopted by the majority of the monomers. In contrast, a single monomer in the fluconazole complex is in a “closed” stated (shown in green). Panels B-F depict the environment of the bound ligands (clotrimazole, tioconazole, voriconazole and fluconazole, respectively), with key residues shown in atom colored sticks (ligand with yellow carbons, protein with light blue carbons and heme with purple carbons). Panels G-K depict the corresponding omit electron density (contoured at 3 sigma) for each of the ligands as a blue mesh. With the exception of fluconazole, the conformation of the ligand is similar in the various monomers in the asymmetric unit, and in each instance monomer A is shown. In the case of the fluconazole ligand, both of the conformations observed are depicted in panels E (“open” conformation) and F (“closed” conformation), with the corresponding electron density shown in panels J and K. MPD: 2-methyl-2,4-pentanediol, EDO: 1,2-ethanediol.

of the four novel DM P450-azole drug crystal structures presented in this study show the direct coordination of imidazole or triazole nitrogen atoms to the DM heme iron, with slightly longer bond lengths than those in the imidazole-bound form (ranging from 2-2.2

Å). However, a novel ligand binding mode was observed for voriconazole (see below).

171

The voriconazole, clotrimazole and tioconazole structures contain multiple monomers in the asymmetric unit, and in each case the general protein conformation corresponds to the “open”, substrate-free conformation of the enzyme, with little significant difference in protein or ligand conformation between monomers. In contrast, the fluconazole complex contains two distinct monomer forms in terms of their conformation; one corresponding to the “open” form, and a second monomer where the

F/G helices have reoriented to resemble a “closed” P450 conformation. As a consequence, two distinct orientations are observed for the bound fluconazole ligand.

Previous studies revealed the binding of an imidazole ligand to the DM heme iron with a bond distance of 1.80 Å (PDB 4KF2) (Butler et al., 2013). Each of the novelstructures presented in this study also show the direct ligation of a hetero-aromatic nitrogen atom to the DM heme iron with slightly longer bond lengths than those in the imidazole-bound form (ranging from 2-2.2 Å). For all the azole-bound structures reported in this study, the hetero-aromatic nitrogen ligand is oriented so as to fit within the groove of the I-helix formed between Ala264 and Glu267/Thr268. While this places the distinct ligands in the same relative position, large differences occur in the nature and orientation of the various non-heme ligating substituents. As a result, despite a similar overall protein conformation, the positioning of the various hydrophobic amino acid side chains is dependent on the nature of the ligand. Furthermore, the conformation of the C-terminal loop protruding into the active site, containing Leu437, is also dependent on the particular ligand.

Figure 22 shows an overlay of the four crystal structures of the ligand-bound DM heme domain, together with images of the active sites of the DM heme domain in their complexes with clotrimazole, tioconazole, fluconazole and voriconazole. In the case of voriconazole, it was clear the 2R,3R-stereoisomer was bound, while for tioconazole the R- tioconazole was identified as the bound stereoisomer. With the exception of the heme- 172 ligand interaction, most contacts established with the protein are of a hydrophobic nature.

In the case of clotrimazole, one of the most hydrophobic ligands, two additional 2-methyl-

2,4-pentanediol (MPD) molecules derived from the mother liquor were observed to interact both with the ligand and the protein. As binding titrations and EPR studies were conducted without MPD or similar compounds, MPD is clearly not needed for azole binding.

Similarly, small mother liquor-derived compounds are present in the case of tioconazole (a molecule of 1,2-ethanediol) and voriconazole (a molecule of glycerol). One of the few direct polar interactions observed between the protein and one of the azole ligands is made between the triazole moiety of voriconazole and Glu267. Surprisingly, voriconazole coordinates the heme iron through a pyrimidine nitrogen from its 5-fluoropyrimidine ring.

All other voriconazole-bound P450 crystal structures on the PDB show coordination of the

P450 heme iron through a triazole nitrogen, and thus this a unique binding mode for voriconazole (Debnath et al., 2017, Sagatova et al., 2016, Hargrove et al., 2015, Mast et al.,

2010). No direct polar interactions are observed for fluconazole, the only other ligand with a free triazole group. For fluconazole, two different conformations are observed, linked to the distinct conformations of the DM heme domain protein itself. In the “open” structure, the electron density of the bound fluconazole is relatively poor, indicating multiple conformations for the non-heme bound triazole group. In contrast, the “closed” monomer contains a well-defined fluconazole ligand that does not display significant conformational heterogeneity.

4.5. Discussion

Azole drugs were developed in the 1960s and have been used widely as inhibitors of fungal 14-lanosterol demethylases (CYP51 family enzymes). The azole drug-mediated inhibition of sterol demethylation in fungi ultimately prevents the formation of ergosterol, which is a key regulator of membrane fluidity and permeability (Lepesheva and Waterman, 173

2007). The first azole drugs (clotrimazole, econazole and miconazole) developed in the

1960s had unacceptable side effects when administered orally (Fromtling, 1988). This stimulated the development of safer azoles such as ketoconazole, which became the industry leading azole until triazoles and advanced imidazole drugs were developed

(Drouhet and Dupont, 1980). For example, the antifungal drug clotrimazole is known to inhibit many human P450s, but newer generations of azole drugs, such as voriconazole, appear to inhibit fewer human P450s (Rendic, 2002). While azoles are best known for their antifungal activity in CYP51 enzymes, they have also been used widely as inhibitors of other eukaryotic and prokaryotic P450 enzymes. For example, the crystal structure of human CYP46A1 was solved in voriconazole- and clotrimazole-bound forms (Mast et al.,

2010), as well as in a posaconazole-bound state (Mast et al., 2013); while the structures of the Mycobacterium tuberculosis sterol demethylase CYP51B1 bound to fluconazole

(Podust et al., 2001) and the M. tuberculosis cyclodipeptide oxidase CYP121A1 bound to fluconazole have also been determined (Seward et al., 2006).

Previous studies of the WT P450 BM3 identified the binding of small imidazole- based compounds (e.g. 1-phenylimidazole and 4-phenylimidazole), and in other work we have reported the binding of omega-imidazolyl carboxylic acids (C10-C12) to the WT

P450 BM3 with Kd values in the range from ~0.2 M to 27 M (Noble et al., 1998). In this paper we present a systematic study of the binding of 17 different azole antifungal drugs to both (i) the WT heme domain and (ii) a double mutant (DM, A82F/F87V) form of the heme domain in which these mutations confer greater conformational flexibility on the

P450, in addition to generating more space in the active site cavity of the enzyme. Table 3 shows that the WT BM3 heme domain is able to bind to several different azole drugs that are in clinical use. In so doing, the azoles ligate to the ferric heme iron of the WT heme domain through nitrogen atoms. This occurs through imidazole nitrogens for clotrimazole/tioconazole, and through a triazole nitrogen for fluconazole. A novel DM 174 heme iron ligation occurs for voriconazole, through a pyrimidine nitrogen on the voriconazole 5-fluoropyrimidine ring. In general, the azole binding-induced Soret band

(red, type II) spectral shifts for the DM heme domain are more substantial than those for the WT heme domain (by ~2 nm on average), discounting those azoles that failed to induce any significant Soret spectral shifts in the WT heme domain. All five triazoles tested on the

WT heme domain (itraconazole, posaconazole, fluconazole, voriconazole and ravuconazole) failed to induce a Soret spectral shift of sufficient extent to allow Kd determination. The azole Kd values for the WT heme domain ranged from 1.89 M

(fenticonazole nitrate) to 32.8 M (clotrimazole). For the DM heme domain, all 17 azoles tested produced type II heme Soret spectral shifts, with Kd values ranging from 0.05 M

(itraconazole) to 4.53 M (fluconazole).

The crystal structures of the BM3 DM heme domain were solved in their complexes with the azoles clotrimazole, voriconazole, fluconazole and tioconazole (Figure

22). It is clear that the conformational flexibility of the A82F/F87V DM variant is an important property of the enzyme that facilitates the full binding of these ligands and enables the structural analysis of their crystallized complexes. In contrast, the WT heme domain is less flexible and typically displays less extensive spectral shifts on binding to these azoles, and no significant spectral shifts for itraconazole, posaconazole and ravuconazole. Several of the azoles used in this study showed apparent binding cooperativity (i.e. a sigmoidal dependence of spectral change on azole drug concentration) in their binding to WT and/or DM heme domains (e.g. with butoconazole nitrate and ketoconazole for both WT/DM heme domains). However, X-ray crystallographic analysis revealed single occupancy of the azole ligands for each of the structures solved. In EPR studies, many of the azole drugs used display heterogeneity (i.e. exhibited multiple sets of g-values). Reasons for such heterogeneity are rarely determined using EPR methods. For

175 example, Roberts et al., reported that no ligand-dependent spin-state equilibrium observed in EPR has been quantitatively understood for any ligands that display cooperativity

(Roberts et al., 2005). The binding of three azoles induced narrowing of the DM heme domain EPR spectra compared to the azole-free form (2.45/2.25/1.91): ketoconazole

(2.43/2.26/1.90), posaconazole (2.42/2.26/1.91) and fluconazole (2.42/2.25/1.91). In the case of fluconazole, this spectrum may correspond to the “closed” conformation observed in the X-ray crystal structure of the fluconazole-bound DM heme domain (6H1S).

In previous work, P450 BM3 heme domains were co-crystallized with 14%

DMSO (F87A) or 28% DMSO (WT and F87A) (Kuper et al., 2007). It was hypothesized that the DMSO sulfur atom competes with the distal water molecule. This is consistent with our EPR studies, which show that the WT heme domain X-band EPR spectra become more heterogeneous in the presence of DMSO. That is, DMSO-bound and water-ligated states can co-exist in the samples. However, our DM heme domain samples become more homogeneous in presence of DMSO, possibly due to DMSO displacing the distal water and interacting more effectively with the DM heme iron. DMSO effects are also seen in collision-induced unfolding (CIU), different forms of mass spectrometry and other methods (Chan et al., 2017). However, no DMSO molecules were observed in any of the four crystal structures solved.

The azole-bound DM heme domain structures solved in this study all contain a single molecule of their respective azoles, although chemicals from the relevant mother liquor (glycerol, 1,2-ethanediol and 2-methyl-2,4-pentanediol) are also observed in the active sites. However, the DM has a large active site volume and could potentially interact with a second molecule of the ligand. Indeed, recent structural studies with another P450

BM3 steroid-oxidizing mutant enzyme revealed the binding of two molecules of testosterone, one at the catalytic site, with the second binding closer to the active site entry 176 region (Acevedo-Rocha et al., 2018). In solution state ligand-binding titrations, several of the azole drugs produced sigmoidal binding curves indicative of cooperative binding

(Table 3). This could occur through e.g. co-binding of two ligands to the heme domain, or by interactions between heme domains. However, the WT BM3 heme domain is monomeric in solution. Many eukaryotic P450s exhibit cooperativity in ligand binding. For example, the major human drug-metabolizing protein CYP3A4. An allosteric binding site was identified on CYP3A4 which contributes to the cooperativity observed for both substrate and inhibitor binding (Roberts et al., 2005). WT BM3 was also reported to show chain length-dependent cooperativity with fatty acids (Rowlatt et al., 2011).

A number of residues were identified from our azole-bound structures which appear crucial to azole binding: Ala264, Glu267, Thr268, and Leu437. These residues are close to the heme prosthetic group and allow both electrostatic and hydrogen bonding interactions to occur with the ligand. Glu267 and Thr268 are important for oxygen binding and activation (Clark et al., 2006). Thr268 is also implicated in substrate recognition

(Yeom et al., 1995). The role of Leu437 is less clear. However, a hydrogen bond was observed between Leu437 and the FDA-approved drug omeprazole in the structure of the

A82F heme domain variant of BM3 (Butler et al., 2013). Ala264 also occupies a position close to the heme prosthetic group in BM3. This residue aligns with a conserved glutamate that can form an ester bond to a heme methyl group in mammalian family 4 P450 enzymes

(Girvan et al., 2004). CYP4 family enzymes, like P450 BM3, can bind a variety of fatty acid substrates (e.g. lauric acid and arachidonic acid), as well as steroids (e.g. testosterone) in humans (Rendic, 2002).

4.6. Conclusions

The azole drugs are used widely as antifungal medications. However, many pathogenic fungi are becoming resistant to current antifungal drugs, including several 177 drugs of the azole class (Perlin et al., 2017). Paths to combat this drug resistance may be to improve current azoles (and other antifungals) by increasing their bioavailability, reducing side effects and improving the spectrum of these drugs (Scorzoni et al., 2017). The P450

BM3 enzyme has long served as a model system for eukaryotic P450s, in view of its soluble nature and its natural interaction with a (fused) eukaryotic-like cytochrome P450 reductase enzyme (Munro et al., 2002). There has been no previous report of a P450 BM3 crystal structure in complex with an azole antifungal drug. Our current crystallographic study demonstrates how the structurally altered A82F/F87V variant of the P450 BM3 heme domain binds to four different azole drugs: two imidazoles (clotrimazole and tioconazole) and two triazoles (fluconazole and voriconazole). Key residues interacting with the azoles in these structures are Ala264, Glu267, Thr268, and Leu437. In previous crystallographic studies with the M. tuberculosis CYP121A1 P450, the major binding mode of fluconazole was observed to occur through its interaction with a retained distal water molecule on the heme iron (Seward et al., 2006). However, fluconazole coordinates directly to the DM heme iron in the crystal state, as is also the case for its interaction with the M. tuberculosis sterol demethylase enzyme CYP51B1 (Podust et al., 2001). The crystal structure of tioconazole-bound DM heme domain is the first shown to bind this azole drug. The binding mode of tioconazole to the DM heme domain is similar to those of clotrimazole and voriconazole. However, fluconazole has a different binding mode in the “closed” conformation with more extensive active site occupancy. It does not approach closely to the key conformation altering mutant residue (Phe82), allowing the phenyl side chain of this residue to rotate by ~90°.

In azole drug ligand-binding studies, numerous azoles of diverse size and structure were found to bind to both the WT and the DM heme domains. EPR was also used to explore azole drug binding, although the DMSO used to solubilize the azoles competed for heme iron binding in some instances. In all cases, the Kd values determined 178 by optical titrations for the DM were significantly lower than those for the WT heme domain, demonstrating that the F87V and A82F mutations are conducive to conformational reorganization in the DM heme domain that facilitates the improved binding of all of the azoles relative to their Kd values in the WT heme domain. In all cases, there is a larger azole-induced DM Soret spectral shift, which in turn is indicative of a greater extent of azole ligand binding to the DM heme iron. The greater level of azole drug occupancy achieved with the DM mutant heme domain clearly proved crucial for the successful crystallization and structural determination of the four different azole drug complexes of the DM heme domain. Other studies with the DM heme domain also enabled the determination of the first crystal structure of BM3 in complex with the human gastric proton pump inhibitor omeprazole (Butler et al., 2013). Very few structures are available to date for azole antifungal drugs in complex with human P450s. These are (i) the cholesterol hydroxylase CYP46A1 (bound to clotrimazole, voriconazole and posaconazole), which is located within the brain and eyes (Mast et al., 2010, Mast et al., 2013); and (ii) ketoconazole bound to the major human drug-metabolizing P450 CYP3A4 (Ekroos and

Sjogren, 2006) and to the human lanosterol 14-demethylase CYP51 (Strushkevich et al.,

2010). As a result, the BM3 DM heme domain (or variants thereof) could have important applications in analysis of the binding modes of other azole drugs. Novel insights gained would include understanding the types of molecular interactions that are favored by specific azole drug substituent groups, as well as establishing if indirect azole nitrogen binding to a retained distal water on the heme iron can also occur in BM3 mutants.

4.7. References

ABELO, A., ANDERSSON, T. B., ANTONSSON, M., NAUDOT, A. K., SKANBERG, I. & WEIDOLF, L. 2000. Stereoselective metabolism of omeprazole by human cytochrome P450 enzymes. Drug Metab Dispos, 28, 966-72. ACEVEDO-ROCHA, C. G., GAMBLE, C. G., LONSDALE, R., LI, A., NETT, N., HOEBENREICH, S., LINGNAU, J. B., WIRTZ, C., FARES, C., HINRICHS, H., 179

DEEGE, A., MULHOLLAND, A. J., NOV, Y., LEYS, D., MCLEAN, K. J., MUNRO, A. W. & REETZ, M. R. 2018. P450-Catalyzed Regio- and Diastereoselective Steroid Hydroxylation: Efficient Directed Evolution Enabled by Mutability Landscaping. ACS Catal, 8, 3395-3410. ADAMS, P. D., AFONINE, P. V., BUNKOCZI, G., CHEN, V. B., DAVIS, I. W., ECHOLS, N., HEADD, J. J., HUNG, L. W., KAPRAL, G. J., GROSSE- KUNSTLEVE, R. W., MCCOY, A. J., MORIARTY, N. W., OEFFNER, R., READ, R. J., RICHARDSON, D. C., RICHARDSON, J. S., TERWILLIGER, T. C. & ZWART, P. H. 2010. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr, 66, 213- 21. ASAHINA, K. & OMURA, K. 1964. Phenomenological Study of Paradoxical Phase and Reverse Paradoxical Phase of Sleep. Jpn J Physiol, 14, 365-72. BELCHER, J., MCLEAN, K. J., MATTHEWS, S., WOODWARD, L. S., FISHER, K., RIGBY, S. E., NELSON, D. R., POTTS, D., BAYNHAM, M. T., PARKER, D. A., LEYS, D. & MUNRO, A. W. 2014. Structure and biochemical properties of the alkene producing cytochrome P450 OleTJE (CYP152L1) from the Jeotgalicoccus sp. 8456 bacterium. J Biol Chem, 289, 6535-50. BUI, S. H., MCLEAN, K. J., CHEESMAN, M. R., BRADLEY, J. M., RIGBY, S. E., LEVY, C. W., LEYS, D. & MUNRO, A. W. 2012. Unusual spectroscopic and ligand binding properties of the cytochrome P450-flavodoxin fusion enzyme XplA. J Biol Chem, 287, 19699-714. BUTLER, C. F., PEET, C., MASON, A. E., VOICE, M. W., LEYS, D. & MUNRO, A. W. 2013. Key mutations alter the cytochrome P450 BM3 conformational landscape and remove inherent substrate bias. J Biol Chem, 288, 25387-99. BUTLER, C. F., PEET, C., MCLEAN, K. J., BAYNHAM, M. T., BLANKLEY, R. T., FISHER, K., RIGBY, S. E., LEYS, D., VOICE, M. W. & MUNRO, A. W. 2014. Human P450-like oxidation of diverse proton pump inhibitor drugs by 'gatekeeper' mutants of flavocytochrome P450 BM3. Biochem J, 460, 247-59. CHAN, D. S., KAVANAGH, M. E., MCLEAN, K. J., MUNRO, A. W., MATAK- VINKOVIC, D., COYNE, A. G. & ABELL, C. 2017. Effect of DMSO on Protein Structure and Interactions Assessed by Collision-Induced Dissociation and Unfolding. Anal Chem, 89, 9976-9983. CLARK, J. P., MILES, C. S., MOWAT, C. G., WALKINSHAW, M. D., REID, G. A., DAFF, S. N. & CHAPMAN, S. K. 2006. The role of Thr268 and Phe393 in cytochrome P450 BM3. J Inorg Biochem, 100, 1075-90. COELHO, P. S., BRUSTAD, E. M., KANNAN, A. & ARNOLD, F. H. 2013. Olefin cyclopropanation via carbene transfer catalyzed by engineered cytochrome P450 enzymes. Science, 339, 307-10. DAFF, S. N., CHAPMAN, S. K., TURNER, K. L., HOLT, R. A., GOVINDARAJ, S., POULOS, T. L. & MUNRO, A. W. 1997. Redox control of the catalytic cycle of flavocytochrome P-450 BM3. Biochemistry, 36, 13816-23. DEBNATH, A., CALVET, C. M., JENNINGS, G., ZHOU, W., AKSENOV, A., LUTH, M. R., ABAGYAN, R., NES, W. D., MCKERROW, J. H. & PODUST, L. M. 2017. CYP51 is an essential drug target for the treatment of primary amoebic meningoencephalitis (PAM). PLoS Negl Trop Dis, 11, e0006104. DENISOV, I. G., GRINKOVA, Y. V., MCLEAN, M. A. & SLIGAR, S. G. 2007. The one- electron autoxidation of human cytochrome P450 3A4. J Biol Chem, 282, 26865- 73. DENISOV, I. G., MAKRIS, T. M., SLIGAR, S. G. & SCHLICHTING, I. 2005. Structure and chemistry of cytochrome P450. Chem Rev, 105, 2253-77. DRISCOLL, M. D., MCLEAN, K. J., LEVY, C., MAST, N., PIKULEVA, I. A., LAFITE, P., RIGBY, S. E., LEYS, D. & MUNRO, A. W. 2010. Structural and biochemical 180

characterization of Mycobacterium tuberculosis CYP142: evidence for multiple cholesterol 27-hydroxylase activities in a human pathogen. J Biol Chem, 285, 38270-82. DROUHET, E. & DUPONT, B. 1980. Chronic mucocutaneous candidosis and other superficial and systemic mycoses successfully treated with ketoconazole. Rev Infect Dis, 2, 606-19. EKROOS, M. & SJOGREN, T. 2006. Structural basis for ligand promiscuity in cytochrome P450 3A4. Proc Natl Acad Sci U S A, 103, 13682-7. EMSLEY, P., LOHKAMP, B., SCOTT, W. G. & COWTAN, K. 2010. Features and development of Coot. Acta Crystallogr D Biol Crystallogr, 66, 486-501. FROMTLING, R. A. 1988. Overview of medically important antifungal azole derivatives. Clin Microbiol Rev, 1, 187-217. GIRVAN, H. M., MARSHALL, K. R., LAWSON, R. J., LEYS, D., JOYCE, M. G., CLARKSON, J., SMITH, W. E., CHEESMAN, M. R. & MUNRO, A. W. 2004. Flavocytochrome P450 BM3 mutant A264E undergoes substrate-dependent formation of a novel heme iron ligand set. J Biol Chem, 279, 23274-86. GRANT, J. L., HSIEH, C. H. & MAKRIS, T. M. 2015. Decarboxylation of fatty acids to terminal alkenes by cytochrome P450 compound I. J Am Chem Soc, 137, 4940-3. GUENGERICH, F. P. 2015. Human cytochrome P450 enzymes. In: ORTIZ DE MONTELLANO, P. R. (ed.) Cytochrome P450 Structure, Mechanism and Biochemistry. 4 ed. Switzerland: Springer International Publishing. GUENGERICH, F. P. & MUNRO, A. W. 2013. Unusual cytochrome p450 enzymes and reactions. J Biol Chem, 288, 17065-73. GUSTAFSSON, M. C., ROITEL, O., MARSHALL, K. R., NOBLE, M. A., CHAPMAN, S. K., PESSEGUEIRO, A., FULCO, A. J., CHEESMAN, M. R., VON WACHENFELDT, C. & MUNRO, A. W. 2004. Expression, purification, and characterization of Bacillus subtilis cytochromes P450 CYP102A2 and CYP102A3: flavocytochrome homologues of P450 BM3 from Bacillus megaterium. Biochemistry, 43, 5474-87. HARGROVE, T. Y., WAWRZAK, Z., LAMB, D. C., GUENGERICH, F. P. & LEPESHEVA, G. I. 2015. Structure-Functional Characterization of Cytochrome P450 Sterol 14alpha-Demethylase (CYP51B) from Aspergillus fumigatus and Molecular Basis for the Development of Antifungal Drugs. J Biol Chem, 290, 23916-34. HILL, A. V. 1910. The possible effects of the aggregation of the molecules of haemoglobin on its dissociation curves. J Physiol-London, 4, 4-7. HOCH, U., ZHANG, Z., KROETZ, D. L. & ORTIZ DE MONTELLANO, P. R. 2000. Structural determination of the substrate specificities and regioselectivities of the rat and human fatty acid omega-hydroxylases. Arch Biochem Biophys, 373, 63-71. JOOSTEN, R. P., LONG, F., MURSHUDOV, G. N. & PERRAKIS, A. 2014. The PDB_REDO server for macromolecular structure model optimization. IUCrJ, 1, 213-20. KUPER, J., TEE, K. L., WILMANNS, M., ROCCATANO, D., SCHWANEBERG, U. & WONG, T. S. 2012. The role of active-site Phe87 in modulating the organic cosolvent tolerance of cytochrome P450 BM3 monooxygenase. Acta Crystallogr Sect F Struct Biol Cryst Commun, 68, 1013-7. KUPER, J., WONG, T. S., ROCCATANO, D., WILMANNS, M. & SCHWANEBERG, U. 2007. Understanding a mechanism of organic cosolvent inactivation in heme monooxygenase P450 BM-3. J Am Chem Soc, 129, 5786-7. LEPESHEVA, G. I. & WATERMAN, M. R. 2007. Sterol 14alpha-demethylase cytochrome P450 (CYP51), a P450 in all biological kingdoms. Biochim Biophys Acta, 1770, 467-77.

181

MAST, N., CHARVET, C., PIKULEVA, I. A. & STOUT, C. D. 2010. Structural basis of drug binding to CYP46A1, an enzyme that controls cholesterol turnover in the brain. J Biol Chem, 285, 31783-95. MAST, N., ZHENG, W., STOUT, C. D. & PIKULEVA, I. A. 2013. Antifungal Azoles: Structural Insights into Undesired Tight Binding to Cholesterol-Metabolizing CYP46A1. Mol Pharmacol, 84, 86-94. MCLEAN, K. J. E. A. 2015. Microbial cytochromes P450. In: ORTIZ DE MONTELLANO, P. R. (ed.) Cytochrome P450 Structure, Mechanism and Biochemistry. Switzerland: Springer International Publishing. MICHAELIS, L. & MENTEN, M. L. 1913. The kenetics of the inversion effect. Biochem Z, 49, 333-369. MILES, J. S., MUNRO, A. W., ROSPENDOWSKI, B. N., SMITH, W. E., MCKNIGHT, J. & THOMSON, A. J. 1992. Domains of the catalytically self-sufficient cytochrome P-450 BM-3. Genetic construction, overexpression, purification and spectroscopic characterization. Biochem J, 288 ( Pt 2), 503-9. MIURA, Y. & FULCO, A. J. 1975. Omega-1, Omega-2 and Omega-3 hydroxylation of long-chain fatty acids, amides and alcohols by a soluble enzyme system from Bacillus megaterium. Biochim Biophys Acta, 388, 305-17. MONTEMIGLIO, L. C., PARISI, G., SCAGLIONE, A., SCIARA, G., SAVINO, C. & VALLONE, B. 2016. Functional analysis and crystallographic structure of clotrimazole bound OleP, a cytochrome P450 epoxidase from Streptomyces antibioticus involved in oleandomycin biosynthesis. Biochim Biophys Acta Gen Subj, 1860, 465-475. MORRISON, J. F. 1969. Kinetics of the reversible inhibition of enzyme-catalysed reactions by tight-binding inhibitors. Biochim Biophys Acta, 185, 269-86. MUNRO, A. W., GIRVAN, H. M., MASON, A. E., DUNFORD, A. J. & MCLEAN, K. J. 2013. What makes a P450 tick? Trends Biochem Sci, 38, 140-50. MUNRO, A. W., GIRVAN, H. M. & MCLEAN, K. J. 2007. Variations on a (t)heme-- novel mechanisms, redox partners and catalytic functions in the cytochrome P450 superfamily. Nat Prod Rep, 24, 585-609. MUNRO, A. W., LEYS, D. G., MCLEAN, K. J., MARSHALL, K. R., OST, T. W., DAFF, S., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MOSER, C. C., PAGE, C. C. & DUTTON, P. L. 2002. P450 BM3: the very model of a modern flavocytochrome. Trends Biochem Sci, 27, 250-7. MUNRO, A. W., LINDSAY, J. G., COGGINS, J. R., KELLY, S. M. & PRICE, N. C. 1994. Structural and enzymological analysis of the interaction of isolated domains of cytochrome P-450 BM3. FEBS Lett, 343, 70-4. NARHI, L. O. & FULCO, A. J. 1987. Identification and characterization of two functional domains in cytochrome P-450BM-3, a catalytically self-sufficient monooxygenase induced by barbiturates in Bacillus megaterium. J Biol Chem, 262, 6683-90. NEELI, R., GIRVAN, H. M., LAWRENCE, A., WARREN, M. J., LEYS, D., SCRUTTON, N. S. & MUNRO, A. W. 2005. The dimeric form of flavocytochrome P450 BM3 is catalytically functional as a fatty acid hydroxylase. FEBS Lett, 579, 5582-8. NOBLE, M. A., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MACKAY, A. C., REID, G. A., HANZLIK, R. P. & MUNRO, A. W. 1999. Roles of key active-site residues in flavocytochrome P450 BM3. Biochem J, 339 ( Pt 2), 371-9. NOBLE, M. A., QUARONI, L., CHUMANOV, G. D., TURNER, K. L., CHAPMAN, S. K., HANZLIK, R. P. & MUNRO, A. W. 1998. Imidazolyl carboxylic acids as mechanistic probes of flavocytochrome P-450 BM3. Biochemistry, 37, 15799-807. OMURA, T. & SATO, R. 1964. The Carbon Monoxide-Binding Pigment of Liver Microsomes. Ii. Solubilization, Purification, and Properties. J Biol Chem, 239, 2379-85. 182

PERLIN, D. S., RAUTEMAA-RICHARDSON, R. & ALASTRUEY-IZQUIERDO, A. 2017. The global problem of antifungal resistance: prevalence, mechanisms, and management. Lancet Infect Dis, 17, e383-e392. PETERS, M. W., MEINHOLD, P., GLIEDER, A. & ARNOLD, F. H. 2003. Regio- and enantioselective alkane hydroxylation with engineered cytochromes P450 BM-3. J Am Chem Soc, 125, 13442-50. PODUST, L. M., POULOS, T. L. & WATERMAN, M. R. 2001. Crystal structure of cytochrome P450 14alpha -sterol demethylase (CYP51) from Mycobacterium tuberculosis in complex with azole inhibitors. Proc Natl Acad Sci U S A, 98, 3068- 73. RENDIC, S. 2002. Summary of information on human CYP enzymes: human P450 metabolism data. Drug Metab Rev, 34, 83-448. RITTLE, J. & GREEN, M. T. 2010. Cytochrome P450 compound I: capture, characterization, and C-H bond activation kinetics. Science, 330, 933-7. ROBERTS, A. G., CAMPBELL, A. P. & ATKINS, W. M. 2005. The thermodynamic landscape of testosterone binding to cytochrome P450 3A4: ligand binding and spin state equilibria. Biochemistry, 44, 1353-66. ROWLATT, B., YORKE, J. A., STRONG, A. J., WHITEHOUSE, C. J., BELL, S. G. & WONG, L. L. 2011. Chain length-dependent cooperativity in fatty acid binding and oxidation by cytochrome P450BM3 (CYP102A1). Protein Cell, 2, 656-71. SAGATOVA, A. A., KENIYA, M. V., WILSON, R. K., SABHERWAL, M., TYNDALL, J. D. & MONK, B. C. 2016. Triazole resistance mediated by mutations of a conserved active site tyrosine in fungal lanosterol 14alpha-demethylase. Sci Rep, 6, 26213. SCORZONI, L., DE PAULA, E. S. A. C., MARCOS, C. M., ASSATO, P. A., DE MELO, W. C., DE OLIVEIRA, H. C., COSTA-ORLANDI, C. B., MENDES-GIANNINI, M. J. & FUSCO-ALMEIDA, A. M. 2017. Antifungal Therapy: New Advances in the Understanding and Treatment of Mycosis. Front Microbiol, 8, 36. SEWARD, H. E., ROUJEINIKOVA, A., MCLEAN, K. J., MUNRO, A. W. & LEYS, D. 2006. Crystal structure of the Mycobacterium tuberculosis P450 CYP121- fluconazole complex reveals new azole drug-P450 binding mode. J Biol Chem, 281, 39437-43. SOWDEN, R. J., YASMIN, S., REES, N. H., BELL, S. G. & WONG, L. L. 2005. Biotransformation of the sesquiterpene (+)-valencene by cytochrome P450cam and P450BM-3. Org Biomol Chem, 3, 57-64. STRUSHKEVICH, N., USANOV, S. A. & PARK, H. W. 2010. Structural basis of human CYP51 inhibition by antifungal azoles. J Mol Biol, 397, 1067-78. YEOM, H., SLIGAR, S. G., LI, H., POULOS, T. L. & FULCO, A. J. 1995. The role of Thr268 in oxygen activation of cytochrome P450BM-3. Biochemistry, 34, 14733- 40. YOSHIMOTO, F. K. & AUCHUS, R. J. 2015. The diverse chemistry of cytochrome P450 17A1 (P450c17, CYP17A1). J Steroid Biochem Mol Biol, 151, 52-65.

183

4.8. Supporting Information

Clotrimazole Fluconazole Tioconazole Voriconazole

PDB ID 6H1T 6H1S 6H1L 6H1O

Data Collection

Space group P21 P212121 P21 P21

Cell dimensions

a, b, c (Å) 79.0, 70.8, 209.6 60.9, 119.3, 146.5 59.0, 150.9, 61.0 59.8, 149.9, 61.0

, β, γ () 90.0, 95.3, 90.0 90.0, 90.0, 90.0 90.0, 95.7, 90.0 90.0, 96.8, 90.0

Resolution range (Å) 208-2.08 119.29-1.95 60.96-1.97 60.63-1.73

(2.12-2.08) (1.99-1.95) (2.00-1.97) (1.76-1.73)

Rmeas 0.080 (0.591) 0.113 (0.815) 0.047 (0.533) 0.088 (0.383)

CC1/2 1.0 (0.8) 1.0 (0.8) 1.0 (0.8) 1.0 (0.9)

I/I 11.8 (2.2) 10.7 (2.1) 15.2 (2.2) 8.2 (2.2)

Completeness (%) 91.7 (98.9) 100.0 (100.0) 98.6 (99.9) 90.6 (93.8)

Redundancy 3.4 (3.4) 6.6 (6.1) 3.4 (3.5) 3.0 (2.8)

Refinement

Resolution (Å) 2.08 1.95 1.97 1.73

Rwork / Rfree 0.170/0.213 0.162/0.200 0.189/0.215 0.159/0.199

R.m.s. deviations

Bond lengths (Å) 0.007 0.010 0.005 0.012

Bond angles () 0.963 1.007 0.754 1.265

Ramanchandran

Favoured (%) 97.1 96.4 95.4 96.5

Outlier (%) 0.3 0.6 0.8 0.3

Figure S7: Table of crystallographic data for the azole bound DM complexes All structures displayed good density with high resolution. All structures have been deposited on the PDB using the codes provided. 184

Rhombic LS signals (gz/gy/gx) BM3 Variant Major species Minor species

WT 2.41/2.25/1.92 None

WT + DMSO 2.41/2.25/1.91 2.39/2.25/1.93; 2.44/2.25/1.90

WT + Bifonazole 2.47/2.25/1.89 2.57/2.25/1.85

WT + Butoconazole nitrate 2.45/2.26/1.90 2.55/2.26/1.86

WT + Climbazole 2.47/2.26/1.91 2.50/2.26/1.88

WT + Clotrimazole* 2.41/2.25/1.91 2.39/2.25/1.93; 2.45/2.25/1.89

WT + Econazole nitrate 2.42/2.26/1.91 2.44/2.26/1.90; 2.52/2.26/1.87; 2.59/2.26/1.84

WT + Fenticonazole nitrate 2.45/2.25/1.91 2.39/2.25/1.94; 2.41/2.25/1.93; 2.53/2.25/2.53

WT + Fluconazole* 2.41/2.25/1.91 2.39/2.25/1.93; 2.45/2.25/1.87

WT + Isoconazole nitrate 2.41/2.25/1.91 2.39/2.25/1.92; 2.44/2.25/1.90; 2.52/2.25/1.87

WT + Itraconazole 2.41/2.25/1.91 2.39/2.25/1.92; 2.44/2.25/1.90

WT + Ketoconazole 2.42/2.25/1.91 2.44/2.25/1.90; 2.56/2.25/1.86

WT + Miconazole 2.44/2.25/1.91 2.42/2.25/1.92; 2.52/2.25/1.89

WT + Posaconazole 2.44/2.25/1.91 2.39/2.25/1.93; 2.41/2.25/1.90

WT + Ravuconazole 2.44/2.25/1.91 2.39/2.25/1.93; 2.41/2.25/1.90

WT + Sertaconazole nitrate 2.42/2.25/1.91 2.39/2.25/1.92; 2.44/2.25/1.90

WT + Sulconazole nitrate 2.45/2.25/1.91 2.39/2.25/1.92; 2.56/2.25/1.86

2.39/2.25/1.93;2.44/2.25/1.90; WT + Tioconazole* 2.41/2.25/1.91 2.53/2.25/1.87; 2.59/2.25/1.85 WT + Voriconazole* 2.41/2.25/1.91 2.39/2.25/1.93; 2.44/2.25/1.90

Figure S8: X-band EPR data sets for the WT BM3 heme domain bound to azole compounds Many compounds displayed differences to the DMSO spectrum. Successfully crystallized compounds are denoted with an asterisk.

185

Rhombic LS signals (gz/gy/gx) BM3 Variant Major species Minor species

DM 2.44/2.25/1.91 2.41/2.25/1.89

DM + DMSO 2.45/2.25/1.91 2.50/2.25/1.89

DM + Bifonazole 2.49/2.26/1.88 None

DM + Butoconazole nitrate 2.44/2.26/1.89 2.50/2.26/1.88

DM + Climbazole 2.46/2.26/1.89 None

DM + Clotrimazole* 2.46/2.26/1.90 None

DM + Econazole nitrate 2.45/2.26/1.89 2.49/2.26/1.88

DM + Fenticonazole nitrate 2.45/2.26/1.89 2.49/2.26/1.88

DM + Fluconazole* 2.42/2.25/1.91 2.49/2.25/1.89

DM + Isoconazole nitrate 2.46/2.25/1.89 2.44/2.26/1.90; 2.53/2.26/1.88

DM + Itraconazole 2.45/2.25/1.91 2.51/2.26/1.89

DM + Ketoconazole 2.43/2.26/1.90 2.53/2.261.87

DM + Miconazole 2.45/2.26/1.90 2.50/2.26/1.87

DM + Posaconazole 2.42/2.26/1.91 2.50/2.26/1.89

DM + Ravuconazole 2.46/2.26/1.89 2.52/2.26/1.87

DM + Sertaconazole nitrate 2.45/2.26/1.90 2.56/2.26/1.88

DM + Sulconazole nitrate 2.45/2.26/1.90 2.56/2.26/1.88

DM + Tioconazole* 2.45/2.26/1.91 2.51/2.26/1.89

DM + Voriconazole* 2.45/2.26/1.90 2.42/2.26/1.91; 2.51/2.26/1.89

Figure S9: X-band EPR data sets for the DM BM3 heme domain bound to azole compounds Many compounds displayed differences to the DMSO spectrum. Successfully crystallized compounds are denoted with an asterisk.

186

Chapter 5: Screening Antidiabetic Binding to P450 BM3 and the Production of Human Metabolites

5.1. Abstract

P450 BM3 is a natural fusion protein capable of binding a range of fatty acids.

The flexibility of the active site and high catalytic rate have led to this protein being extensively researched to modify substrate and product profiles. The introduction of the gatekeeper mutations (A82F/F87V) allows the binding of a range of molecules including pharmaceutical drugs. Previously, screening with a drug compound library showed the promiscuous nature of this protein and its ability to bind compounds such as antidiabetic compounds used to treat type II diabetes. The interaction of DM BM3 with antidiabetic compounds was investigated to observe ligand binding (using binding titrations, X-ray crystallography, and EPR) and metabolite production (using HPLC, LCMS, LCMS/MS, and NMR). Human metabolites were successfully produced for pioglitazone highlighting the use of P450 BM3 as a model for the main human metabolizing enzymes CYP2C8,

CYP2C9, and CYP3A4.

5.2. Introduction

P450 BM3 (BM3) is a soluble cytochrome P450monooxygenase enzyme isolated from Bacillus megaterium (Miura and Fulco, 1974, Hare and Fulco, 1975, Matson et al.,

1977). BM3 is a natural fusion enzyme containing a heme-binding catalytic domain (~55 kDa) linked to its cytochrome p450 reductase (CPR) domain partner (~65 kDa) (Narhi and

Fulco, 1986, Narhi and Fulco, 1987). This structural arrangement enables BM3 to have very high catalytic rates (>17,000 min-1 with arachidonic acid (Noble et al., 1999)). The

WT enzyme binds and hydroxylates fatty acids with chains lengths from ~C12-C18 (Miura

187 and Fulco, 1975, Ho and Fulco, 1976). However, BM3 has been the subject of numerous protein engineering studies, leading to its ability to bind non-native substrates and to the repurposing of variants of this efficient enzyme for catalytic functions including the production of the antimalarial drug artemether (Zhang et al., 2012), synthesis of the high- value grapefruit fragrance molecule nootkatone (Sowden et al., 2005), and the production of human metabolites for a range of pharmaceuticals, including the regioselective 16- hydroxylation of testosterone and other steroids (Acevedo-Rocha et al., 2018).

In recent studies, we have used a versatile P450 BM3 mutant with the introduction of two mutations that enable the binding of several new substrates, including large molecules such as pharmaceutical compounds (Butler et al., 2013). These mutations remove steric bulk from the active site (F87V) and cause conformational changes allowing the P450 to adopt the catalytically primed state, even when substrate-free (A82F). Human drug metabolites were successfully produced using these “gatekeeper” (A82F/F87V or double mutant, DM) mutants for omeprazole and other proton pump inhibitor drugs (Butler et al., 2013, Butler et al., 2014). In subsequent work, we have probed further the substrate specificity profile for the DM enzyme using a 978 FDA-approved compound library and have many other potential substrates. A measure of the degree of binding induced by these compounds was determined by determining the extent of the heme Soret band on ligand binding. On substrate association, the distal water ligated to the heme iron is displaced, causing the heme iron to become pentacoordinate and to adopt the high-spin (HS) state

(Guengerich and Johnson, 1997). This causes an electronic rearrangement in the ferric iron

3d-orbitals and a shift in the Soret peak from ~418 nm to ~394 nm (Gibson and Skett,

2001, Luthra et al., 2011). The Soret shift induced by each potential substrate was quantified to enable comparisons between compound hits and thus to identify novel substrates for the DM variant (Chapter 3). Through this screening process, a number of likely antidiabetic substrates for the DM BM3 heme domain were identified. 188

Type II diabetes is a disease in which patients can no longer control their blood sugar levels due to low levels of insulin or the development of resistance to insulin. In the year 2000 171 million people were living with this condition, but by the year 2030 this is expected to rise to 366 million (Wild et al., 2004). One reason for this increase is due to the growing number of overweight and obese individuals in developed countries who are more likely to develop insulin resistance (Zaninotto et al., 2009). Diabetic patients spend an average of 23 days per year in hospitals at great costs, thought to be due to poor glucose level control (Jonsson, 2002). The cost of pharmaceutical drugs in Europe contributes only

7% of the total cost of diabetic healthcare (Jonsson, 2002). These compounds can help prevent the burden on hospitals and lower costs associated with diabetes by controlling patient blood sugar through a variety of mechanisms.

In this paper, the antidiabetic compounds identified as substrate-type hits from our

FDA-approved library screen are investigated for their binding affinity with the DM BM3 heme domains, binding mode, and metabolite production. A human metabolite for pioglitazone was identified from this work, formed in a highly stereo- and regioselective manner from a reaction with the intact DM P450 BM3 enzyme. These data indicate that this high activity BM3 mutant can oxidize compounds from the glitazone drug class, as is also evidenced from subsequent studies reported herein that demonstrate both productive binding and catalytic production of human metabolites from other antidiabetic compounds from the glitazone and meglitinide drug classes. These data provide evidence that the DM

P450 BM3 variant can act as a model enzyme for human P450s and can produce human drug metabolites that are required for drug safety testing and other applications.

189

5.3. Materials and Methods

5.3.1. P450 BM3 protein expression and purification

The WT and DM heme domain proteins were expressed using pET14b (heme domain) and pET15b (full-length) plasmid constructs. Plasmids were transformed into

BL21 (DE3) and expressed in Terrific Broth medium (Formedium, Hunstanton, UK) for the WT protein, and in Auto Induction Terrific Broth (Formedium) for the DM variants.

Cells were grown for 24 hours at 37 °C at 170 rpm in an orbital shaker. The DM heme domain protein was supplemented with 100 µM δ-aminolevulinic acid when OD600 reached 0.6 in order to promote heme incorporation. Cells were harvested by centrifugation at 4 °C (6000 g, 20 min).

Cell pellets were resuspended in ice-cold buffer containing protease inhibitors (1 tablet per 100 mL, EDTA-free cOmpleteTM tablets, Roche Applied Science, Burgess Hill,

UK), and DNase (Merck, Nottingham UK) (10 µg mL-1). For the full-length (intact) BM3 enzyme, buffer A was used (50 mM potassium phosphate [KPi] containing 350 mM KCl and 10% v/v glycerol, pH 8). For the DM BM3 heme domain, buffer C was used (50mM

Tris 1 mM EDTA, pH 7.2) Cells were lysed by sonication on ice using a Bandelin

Sonopuls instrument at 37% amplitude with 12 pulses for 40 s, and with 60 s breaks between pulses. The cell extract was clarified by centrifugation (4600 g for 60 minutes at 4

°C). Contaminant proteins were removed using a 1-hour incubation with 30% w/v ammonium sulfate at 4 °C with gentle stirring using buffer A for intact protein and buffer

C for truncated heme domain. Precipitated proteins and other material were removed using centrifugation (4600 g for 15 minutes at 4 °C).

For the intact DM BM3 protein, purification was achieved in two steps using a nickel-IDA column, followed by an anion exchange (hydroxyapatite) column. The clarified

190 cell extracts were incubated with nickel-IDA resin overnight at 4 °C in buffer A. The protein-bound resin was applied to columns and eluted with a stepwise gradient of 10 mM

(300 mL), 20 mM (200 mL) and 200 mM (60 mL) imidazole. The eluted protein was dialyzed into 25 mM KPi, pH 6.5 (buffer B) and applied to a column containing 28 mL of

CHT hydroxyapatite type 1 resin (Bio-Rad Laboratories, Watford UK). Protein was eluted using a linear gradient of 25-300 mM KPi, pH 6.5 (600 mL).

For the DM BM3 heme domain, the protein was dialyzed into buffer B. A His-tag was not used for this construct in order to improve crystallogenesis and structural resolution. A four-column purification was utilized to achieve the high purity needed for

X-ray crystallography. Full details of this method are given in Chapter 4. In a final step, lipids were removed from the protein using the hydrophobic resin Lipidex-1000 using the relevant experimental buffer.

Heme protein concentration was quantified using the absorbance of the low-spin

P450 Soret peak (~418 nm) divided by the extinction coefficient (95 mM-1 cm-1 for the heme domain and 105 mM-1 cm-1 for the intact P450 BM3 protein).

5.3.2. Antidiabetic binding affinity determination

The binding affinities (Kd values) of all antidiabetic compounds were determined using the WT or DM BM3 heme domains. All antidiabetic compounds were sourced from

Sigma Aldrich (Haverhill, Cambridgeshire, UK). A Cary 60 UV-Vis spectrophotometer

(Agilent, Stockport, UK) with a Peltier at 30 °C was used for data collection. Darglitazone exhibited absorbance in the UV-Vis range and so titrations were undertaken on a Cary 300 dual beam spectrophotometer (Agilent) with a reference cell containing 100 mM KPi, (pH

7) and the same addition of ligand as made to the sample cell. Titrations were undertaken

191 with ~3 µM protein in a quartz cuvette (paired in the case of darglitazone) by microliter additions of concentrated drug ligands (in DMSO) until saturation was achieved. Kd values were calculated using either the Michaelis-Menten equation (for hyperbolic curves), the

Morrison equation for tight-binding ligands (where Kd values are typically ≤ 5x the P450 concentration), or the Hill equation for sigmoidal curves (Michaelis and Menten, 1913,

Morrison, 1969, Hill, 1910).

5.3.3. Electron paramagnetic resonance (EPR) spectroscopy

The binding of ligands to the WT or DM heme domain (in the ferric state) was analyzed using X-band EPR. Samples were prepared with 200 M protein in 25 mM KPi, pH 7 with the addition of 5 mM drug in DMSO, or appropriate volumes of DMSO or buffer as controls. Samples were incubated overnight at 4 C with stirring at 10 rpm.

Samples were then centrifuged (14,000 rpm for 10 minutes at 4 °C in a microfuge) before transfer to EPR tubes and freezing. For all data analyzed, a buffer control was subtracted.

X-band EPR spectra were recorded on a Bruker ER-300D series electromagnet with a microwave source interfaced with a Bruker EMX control unit and fitted with an ESR-9 liquid helium flow cryostat (Oxford Instruments), and a dual-mode microwave cavity from

Bruker (ER-4116DM). Spectra were recorded at 10 K with a microwave power of 0.5 mW, a modulation frequency of 100 kHz and a modulation amplitude of 5 G.

5.3.4. Structural determination by X-ray crystallography

Samples of ligand-bound DM heme domain were prepared with 9 mg mL-1 protein and 5 mM ligand (dissolved in a 100% DMSO stock) in 25 mM KPi, pH 7.

Samples were prepared as above. Liquid handling was achieved using a Mosquito liquid handling robot (TTP LabTech Ltd, Melbourn, Cambridgeshire, UK) for the addition of the

192 protein to Molecular Dimensions screening plates (PACT, SG1, MORP, JCSG+, BCS,

Midas, LMB) (Newmarket, Cambridgeshire, UK) and by mixing with the mother liquor provided within the plates. Crystallography was performed using the sitting drop method using 300 nL DM heme domain protein and 300 nL mother liquor. Seeding was required for improved crystal quality. Seeding was accomplished with 250 nL protein (prepared as above) mixed with 50 nL seed stock and 300 nL mother liquor. Plates were stored at 4 °C for crystallogenesis. X-ray diffraction data were collected at Diamond synchrotron beamlines. Data were reduced and scaled using XDS (Kabsch, 2010). Structures were solved and refined by molecular replacement with a previously solved BM3 DM heme domain structure (PDB 4O4P) using Phenix (Adams et al., 2010), PDB re-do (Joosten et al., 2014) and Coot (Emsley et al., 2010).

5.3.5. Product determination by HPLC, LCMS, and LCMS/MS

To ensure good turnover of compounds for initial HPLC, LCMS and LCMS/MS studies, a high stoichiometry of protein-to-ligand was used. Samples were produced by incubating 50 µM of the intact P450 BM3 DM enzyme with 100 µM ligand. A NADPH regeneration system containing 7.76 mM glucose-6-phosphate, 0.6 mM NADP+ and 0.75

U/mL glucose-6-phosphate dehydrogenase was used. The samples were made up to 500 μL in turnover buffer (50 mM KPi, 5 mM CaCl2, pH 7.4) in 9 mL glass vials. Turnover reactions were incubated at 37 °C, 170 rpm for 3 hours in an orbital shaker.

For HPLC and LCMS, reactions were quenched with 1:1 acetonitrile and clarified using 2 mL Impact® protein precipitation plates (Phenomenex, Macclesfield, UK). HPLC and LCMS analyzes were completed on an Agilent 1100 system with an Agilent 1100 diode array detector (DAD) and an Agilent 1100 LC/MSD ion trap). A linear gradient from

10 to 100% acetonitrile over 10 minutes was used, followed by a hold of 2 minutes at

193

100% acetonitrile before returning to 10% acetonitrile over 10 minutes. For each sample,

10 L was injected per run. The column used for HPLC and LCMS was a C18 Kinetex column (2.6 m x 100 mm, 2.1 mm internal diameter) with a constant flow rate of 1 mL min-1 at room temperature (Phenomenex).

The solvents used for HPLC, LCMS and LCMS/MS were Chromasolv® acetonitrile and Chromasolv® water (Thermo Fisher Scientific, Runcorn, UK) both with

0.1% formic acid (Sigma Aldrich).

LC-MS/MS was undertaken using a Q Exactive Plus instrument equipped with a heated electrospray ionization source and a U3000 UHPLC (Thermo Fisher Scientific).

Prior to analysis, reactions were purified using 1 mL, 33 mm polymetric reversed phase

Strata™-X columns (Phenomenex, Macclesfield, UK) and eluted into 1 mL Chromasolv® acetonitrile. The reactions were chromatographically separated using a gradient of water +

0.1 % formic acid (solvent A) and methanol + 0.1 % formic acid (solvent B) and a

Hypersil Gold column (100 x 2.1 mm, 3 m, Thermo Fisher Scientific). The gradient was programmed to hold at 90% A for 1.5 min, followed by a linear decrease to 1% A over 5 min, and then held for 1 min before returning to 90% A in 0.1 min, with a final hold time of 1.22 min. The flow rate was constant at 400 uL/min, the injection volume was 5 uL and the column was maintained at 40 °C. Mass spectrometry was collected in positive and negative ionization modes in separate acquisitions. Full scan mass spectrometry was conducted at a resolution of 70000 with an AGC target of 3e6 and a scan range of 90-1000 m/z. Data-dependent MSMS was conducted on the top 5 most abundant ions, with a resolution of 17500, AGC target of 5e4 and a collision energy of 30 or 50eV.

194

5.3.6. Product determination by NMR

NMR product determination required larger quantities of reaction products. To generate the required amounts, 3 µM of the DM BM3 enzyme was reacted with 45 µM of ligand in 60 mL flasks. The same NADPH regeneration system was used for the HPLC,

LCMS, and LCMS/MS studies. As above, turnover reactions were completed at 37 °C, 80 rpm for 3 hours in an orbital shaker. Reactions were purified using Strata-X columns

(Phenomenex) and eluted in 1 mL DMSO-d6. NMR spectra were recorded on a Bruker

AVIII 500MHz spectrophotometer with a 1H/13C TCI cryoprobe equipped with Z- gradients. All spectra were recorded at 298K. Spectra were processed and analyzed with

Topspin 3.2 (Bruker). 1H-1H NOESY spectra used excitation sculpting solvent suppression and gradients in T1. Where novel products were made, 1H-13C HSQCs were recorded using the gradient-selected sensitivity-enhanced method, and assigned on the basis of

HSQC-TOCSY spectra.

5.4. Results

5.4.1. Binding affinity determination for antidiabetic compounds

MW WT BM3 DM BM3 Antidiabetic compound (g/mol) Kd (M) Kd (M) Darglitazone* 420.5 17.40 ± 0.56 1.15 ± 0.07 Pioglitazone 356.4 NB 2.01 ± 0.09 Rosiglitazone 357.4 NB 0.85 ± 0.15 Troglitazone 441.5 NB 0.04 ± 0.01 Mitiglinide* 315.4 NB 1.17 ± 0.08 Nateglinide 317.4 NB 5.08 ± 0.30 Repaglinide 452.6 NB 12.53 ± 0.86 Metformin* 129.2 NB 0.43 ± 0.61

Table 4: Binding affinities for BM3 DM and WT heme domains with antidiabetic compounds The DM variant binds all the antidiabetics. Only darglitazone binds to the WT heme domain, with a much lower affinity than for the DM heme domain. Asterisks indicate antidiabetics that were not present in the FDA screen.

195

A number of antidiabetic compounds were present in the FDA-approved library

(described in Chapter 3). Many thiazolidinediones (glitazones) and meglitinides were found to bind to the DM BM3 heme domain (listed in Table 4). However, no first or second generation sulfonylurea compounds were found to elicit significant heme shifts

(tolbutamide, chlorpropamide, gliclazide, glipizide or glyburide). Similarly, no alphaglucosidase inhibitors were found to bind to the DM BM3 heme domain (miglitol, acarbose or voglibose). Only one biguanide compound was screened (phenformin), but induced no significant heme absorbance shift. To widen this class, metformin was acquired and screened separately, and was found to induce a substantial DM BM3 Soret shift.

Further compounds were also acquired to probe the binding of antidiabetic compounds to the DM heme domain, shown with asterisks in Table 4.

Figure 23: Metformin-induces a type I (substrate) shift with the DM BM3 heme domain The low spin substrate-free protein (blue) gives a Soret shift from 418 nm to a high spin substrate-bound protein (red) at 395 nm allowing for binding affinity determination.

196

To determine affinities for the various antidiabetic compounds that bound the DM heme domain, UV-Vis titrations were conducted (Table 4). Several compounds with diverse structures were shown to bind to the DM heme domain, but not to the WT heme domain. Darglitazone was the only compound that bound to WT BM3 heme domain, although its Kd value (17.4M) was much lower than that for the DM heme domain. All compounds bound avidly to the DM heme domain. Interestingly, metformin contains several nitrogen atoms, yet elicits a type I (substrate) shift rather than coordinating to the

DM heme iron (Figure 23).

5.4.2. EPR spectroscopy of the troglitazone-bound DM heme domain

EPR spectra were collected for the DM heme domain bound to each of the antidiabetic compounds (Figure S10). In most cases, the binding of these compounds induced only minor shifts to the g-values in the EPR spectra, and were similar to the spectra of the DMSO solvent control for the WT or DM ligand-bound proteins. However, the DM heme domain troglitazone-bound spectrum showed a substantial high-spin component (g = 7.94/3.64/1.70), suggesting that the tight-binding troglitazone substrate binds close to the heme iron, displacing the distal water and inducing high-spin heme formation (Figure 24).

5.4.3. The X-ray crystal structure of the troglitazone-bound DM heme domain

The DM BM3 heme domain in complex with troglitazone was successfully crystallized in the condition 0.2 M ammonium chloride and 20 % w/v PEG 3350. The troglitazone-bound structure (PDB 6HN8) adopts the closed conformation attributed to substrate binding. The molecule makes very close interactions with the residues Ala264,

Th268 and Tyr51 in the active site, and with the heme prosthetic group (at distances of 2.8,

197

Figure 24: The EPR spectrum of the troglitazone-bound DM BM3 heme domain The spectrum shows the development of a high-spin heme iron spectrum with g-values of 7.94/3.04/1.70 and other alterations in the low-spin spectrum induced by troglitazone binding to the P450 that are distinct from the low-spin native and DMSO-bound forms.

Figure 25: Troglitazone binding in the active site of the DM BM3 heme domain The substrate makes a number of close interactions with active site residues and the heme prosthetic group.

198

3.1, 3.1 and 3.0 Angstroms respectively), as shown in Figure 25. The troglitazone molecule shown is the 2R-5R enantiomer (PDB TDZ), the only troglitazone enantiomer currently found on the PDB. The molecule binds with its trimethylated/hydroxylated aromatic ring closest to the heme prosthetic group. All crystallographic data are presented in Figure S10.

5.4.4. Production of human metabolites

HPLC, LCMS, and LCMS/MS confirmed the production of metabolites for all antidiabetic compounds apart from nateglinide and mitiglinide. Metabolites could not be fully characterized by NMR for rosiglitazone and metformin and so only LCMS/MS results are discussed for these compounds (Figure S12 and Figure S13 respectively).

Analysis of fragments formed was done using Compound Discoverer (Thermo Scientific) and suggested that oxidative deamination occurred for metformin, while a variety of oxidation reactions for each of the glitazone compounds, and that multiple reactions also occurred with repaglinide. Due to crowded spectra, product determination proved difficult.

Separation of products using HPLC proved unsuccessful for isolating products for NMR.

However, NMR did confirm the production of human metabolites for pioglitazone, troglitazone and repaglinide (Figure 26). Pioglitazone is metabolized by the P450 BM3

DM enzyme to a single product, as observed by 1H-NMR (Figure S14). This product is a keto-type metabolite, a known metabolite of human P450 enzymes. Similarly, 1H-NMR also confirmed the production of a single troglitazone metabolite (Figure S15). This product is a quinone-like metabolite, very similar to the toxic metabolite produced in humans. The DM BM3 produced five different products during turnover experiments with repaglinide, two of which were confirmed to be human metabolites (Figure S16). In addition, products were modelled for darglitazone based on species observed in the NMR spectra. These are the first metabolites reported for this compound (Figure S17).

199

Figure 26: Production of stereo- and regioselectively oxidized products from P450 BM3 Antidiabetic compounds pioglitazone (top) and troglitazone (bottom) are converted to human metabolites by the DM variant. Atoms are colored by element.

5.5. Discussion

Members of the antidiabetic drug classes were shown to bind tightly to the

A82F/F87V (DM) heme domain and to induce substantial high-spin heme iron spectral shifts consistent with substrate-like behavior. Antidiabetics were also oxidized by the intact

P450 BM3 DM enzyme to produce metabolites of these drugs and to enable structural characterization these molecules. Crystallographic studies also defined the troglitazone binding mode in the BM3 DM heme domain. The glitazone class of antidiabetics bind to the transcription peroxisome proliferator-activated receptor gamma (PPAR) (Hauner,

2002). PPAR naturally binds fatty acids, similar to P450 BM3, and in response modulates the transcription of other proteins associated with fatty acid storage and glucose metabolism. Glitazones bind to this enzyme and cause the reduction of insulin resistance in diabetic patients (Hauner, 2002). This class of drug is the only one currently designed to

200 target insulin resistance, which is becoming increasingly necessary as the number of obese diabetic patients increase.

The antidiabetic drug troglitazone was the first glitazone marketed. However, within two years troglitazone caused a number of deaths, leading to its removal from the market. This compound was found to produce a toxic quinone-type metabolite in humans

(Yamazaki et al., 1999). Troglitazone has very tight affinities for the DM protein in the nanomolar range. While many antidiabetic drugs induced high-spin shifts in the DM heme domain, troglitazone was also the only substrate observed to produce a substantial high- spin component species in the EPR spectrum. P450 enzymes do not typically exhibit high- spin spectra at the low temperatures required for heme EPR (10 K), as the low-spin species is favored at low temperatures (Lipscomb, 1980, Luthra et al., 2011). From the X-ray crystal structure of the BM3 DM heme domain-troglitazone complex, the ligand was seen to make a number of short distance interactions with active site residues, in addition to its close approach to the heme prosthetic group. Comparing the troglitazone-bound DM heme domain structure to another DM structure (bound to the FDA-approved compound omeprazole), it is clear that the heme-troglitazone interaction distance is much shorter

(approximately 3 Å and 4 Å, respectively) (Butler et al., 2013). Both substrates interact with Tyr51, a residue at the mouth of the active site associated with binding of the carboxylate group of fatty acid substrates (Daff et al., 1997, Li and Poulos, 1997, Noble et al., 1999).

The enantiomer of troglitazone (PDB TDZ) found within this structure (PDB

6HN8) was also observed to bind to fatty acid binding protein 4 (PDB 2QM9) and

CYP2C8DH (PDB 2VN0). In the CYP2C8 structure, troglitazone is described as a competitive inhibitor, as the binding mode of the molecule is more than 7.5 Angstroms distant from the heme. However, other studies have found that CYP2C8 is one of the P450 201 enzymes necessary for troglitazone metabolism, particularly for the formation of the toxic quinone metabolite (Yamazaki et al., 1999). Interestingly, the carbon atom closest to the heme prosthetic group is the same for both structures (Schoch et al., 2008).

The low Kd value for troglitazone and its interactions observed within the crystal structure are consistent with this substrate binding very tightly to the active site. Despite this tight binding, enzyme turnover was successful and products were observed. Work was done to characterize the DM BM3 troglitazone metabolites. LCMS analysis reported a +16 peak, suggesting that an oxidation reaction had occurred. However, LCMS/MS analysis proved inconclusive and it is known that troglitazone undergoes photooxidation at room temperature (Fu et al., 1996). Steps were taken to overcome this issue for NMR experiments, including keeping samples away from light unless absolutely necessary.

Through proton and carbon NMR studies, the structure of the metabolite was determined.

The DM BM3 enzyme was shown to produce a compound very similar to the toxic human metabolite (Yamazaki et al., 1999). Specifically, the BM3 DM enzyme produces the protonated hydroquinone molecule rather than a quinone (Figure S15). Over time, the same sample produced signals in the 1H-NMR spectra suggesting oxidation to form the quinone metabolite without the presence of P450 BM3 or the electron regeneration system.

Pioglitazone is the best-selling glitazone drug and is widely prescribed in isolation or in tablet form containing other antidiabetic compounds, such as metformin. Pioglitazone is metabolized in a stereo- and regioselective manner to a single product by the DM BM3 enzyme – a surprising result for such a promiscuous protein (Figure S14). 90% of the parent compound was converted to this metabolite, with the remaining 10% containing unreacted parent drug, contaminants and, potentially, other metabolites. Pioglitazone is metabolized in vivo by CYP2C8, CYP2C9 and CYP3A4 to over ten known metabolites

(Eckland and Danhof, 2000, Shen et al., 2003). Only three of these metabolites are thought 202 to be active, one of which is the keto derivative shown in Figure 24 and Figure S14

(Baba, 2001). The keto derivative metabolite and a similar metabolite hydroxylated at the same position are the main metabolites found in human serum and comprise approximately

75-80% of the total active compounds observed in humans (Baba, 2001).

From LCMS experiments rosiglitazone and darglitazone were thought to undergo hydroxylation reactions. For rosiglitazone, the smallest hydroxylated fragment suggested the formation of a human metabolite, shown by an asterisk in Figure S12. In addition, rosiglitazone appears to undergo a demethylation reaction. LCMS/MS analysis suggested that this occurred on a nitrogen atom, corresponding to one of the known human metabolites (Cox et al., 2000). From NMR, at least two products are produced from darglitazone and at least five from rosiglitazone. Rosiglitazone produces a number of metabolites in humans including four different oxidized products (Cox et al., 2000).

Darglitazone failed stage I clinical trials due to reported organ toxicity and so no metabolites are in the public domain. Darglitazone is one of fourteen glitazone compounds that failed during clinical trials (Hong et al., 2018). In the 1H-NMR spectrum the signal for the methyl group, labelled atom 1 using IUPAC numbering, is lost. Due to this and several other observations in the HMBC and proton NMR experiments, an oxidation followed by an elimination reaction is modelled. This results in the formation of 4-(4-((2,4- dioxothiazolidin-5-yl)methyl)phenyl)-4-oxobutanoic acid containing the glitazone functional group (Figure S17). In addition, two small molecules are produced; benzamide and acetate. Signals for benzamide are present in the NMR spectra. However, the signal for acetate is missing, likely due to the loss of this sample during purification using Strata-X columns. The formation of the acutely toxic compounds benzamide and acetate may contribute to toxic side effects within the cell. However, during studies with human volunteers darglitazone did successfully increase insulin effectiveness in obese patients

(Chaiken et al., 1995). Interestingly, the toxic compounds troglitazone and darglitazone are 203 more potent drugs than the FDA-approved compounds pioglitazone and rosiglitazone

(Wright et al., 2014).

Nateglinide and mitiglinide were found to induce Soret high-spin spectral shifts, but did not undergo any significant catalytic turnover. The larger compound repaglinide generates five metabolites in humans, produced by CYP3A4 and CYP2C8, highlighting the promiscuous nature of the main pharmaceutical metabolizing P450 enzymes in humans

(Bidstrup et al., 2003). Our results also identified at least five metabolites using the DM

BM3 enzyme. Masses observed corresponded to a single hydroxylation occurring at multiple positions. Two hydroxylated products identified during these experiments were determined to be human metabolites. In addition, two novel hydroxylated compounds were observed. The fifth metabolite could not be assigned completely (Figure S16).

Metformin was also found to bind as a substrate to the DM P450 heme domain, similarly to CYP3A4 but with an absorbance trough at ~420 nm compared to the trough exhibited by CYP3A4 at ~415 nm (Guo et al., 2017). Metformin is a competitive inhibitor of CYP3A4 in humans and is a novel treatment method for breast cancer by inhibiting arachidonic acid (AA)-derived epoxyeicosatrienoic acids (EETs) (Guo et al., 2017, Guo et al., 2014). Results suggested that an oxidative deamination occurred, resulting in a mass loss of 16 (Figure S13). Further work is needed to fully identify this product.

5.6. Conclusions

In conclusion, a range of antidiabetic compounds were shown to bind to the DM heme domain variant. A crystal structure was obtained for the DM protein in complex with the drug troglitazone, and the DM BM3 enzyme was shown to produce a hydroquinone compound similar in structure to the toxic troglitazone metabolite which led to troglitazone’s removal from the market. A stereo- and regioselective reaction was also 204 observed for pioglitazone, producing one of the main human metabolites for this widely prescribed compound. As the cases of diabetes, in particular, type II diabetes, increase worldwide, maintaining a pipeline of existing antidiabetic drugs alongside the production of novel drugs will be of vital importance. Probing the reactions of these drugs in humans and modelling such reactions in vitro will allow greater understanding of the mechanisms of action of current and new antidiabetic compounds.

5.7. References

ACEVEDO-ROCHA, C. G., GAMBLE, C. G., LONSDALE, R., LI, A. T., NETT, N., HOEBENREICH, S., LINGNAU, J. B., WIRTZ, C., FARES, C., HINRICHS, H., DEEGE, A., MULHOLLAND, A. J., NOV, Y., LEYS, D., MCLEAN, K. J., MUNRO, A. W. & REETZ, M. T. 2018. P450-Catalyzed Regio- and Diastereoselective Steroid Hydroxylation: Efficient Directed Evolution Enabled by Mutability Landscaping. ACS Catal, 8, 3395-3410. ADAMS, P. D., AFONINE, P. V., BUNKOCZI, G., CHEN, V. B., DAVIS, I. W., ECHOLS, N., HEADD, J. J., HUNG, L. W., KAPRAL, G. J., GROSSE- KUNSTLEVE, R. W., MCCOY, A. J., MORIARTY, N. W., OEFFNER, R., READ, R. J., RICHARDSON, D. C., RICHARDSON, J. S., TERWILLIGER, T. C. & ZWART, P. H. 2010. PHENIX: a comprehensive Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr, 66, 213- 21. BABA, S. 2001. Pioglitazone: A review of Japanese clinical studies. Curr Med Res Opin, 17, 166-189. BIDSTRUP, T. B., BJORNSDOTTIR, I., SIDELMANN, U. G., THOMSEN, M. S. & HANSEN, K. T. 2003. CYP2C8 and CYP3A4 are the principal enzymes involved in the human in vitro biotransformation of the insulin secretagogue repaglinide. Br J Clin Pharmacol, 56, 305-14. BUTLER, C. F., PEET, C., MASON, A. E., VOICE, M. W., LEYS, D. & MUNRO, A. W. 2013. Key mutations alter the cytochrome P450 BM3 conformational landscape and remove inherent substrate bias. J Biol Chem, 288, 25387-99. BUTLER, C. F., PEET, C., MCLEAN, K. J., BAYNHAM, M. T., BLANKLEY, R. T., FISHER, K., RIGBY, S. E., LEYS, D., VOICE, M. W. & MUNRO, A. W. 2014. Human P450-like oxidation of diverse proton pump inhibitor drugs by 'gatekeeper' mutants of flavocytochrome P450 BM3. Biochem J, 460, 247-59. CHAIKEN, R. L., ECKERT-NORTON, M., PASMANTIER, R., BODEN, G., RYAN, I., GELFAND, R. A. & LEBOVITZ, H. E. 1995. Metabolic effects of darglitazone, an insulin sensitizer, in NIDDM subjects. Diabetologia, 38, 1307-12. COX, P. J., RYAN, D. A., HOLLIS, F. J., HARRIS, A. M., MILLER, A. K., VOUSDEN, M. & COWLEY, H. 2000. Absorption, disposition, and metabolism of rosiglitazone, a potent thiazolidinedione insulin sensitizer, in humans. Drug Metab Dispos, 28, 772-80. DAFF, S. N., CHAPMAN, S. K., TURNER, K. L., HOLT, R. A., GOVINDARAJ, S., POULOS, T. L. & MUNRO, A. W. 1997. Redox control of the catalytic cycle of flavocytochrome P-450 BM3. Biochemistry, 36, 13816-23.

205

ECKLAND, D. A. & DANHOF, M. 2000. Clinical pharmacokinetics of pioglitazone. Exp Clin Endocrinol Diabetes, 108, S234-S242. EMSLEY, P., LOHKAMP, B., SCOTT, W. G. & COWTAN, K. 2010. Features and development of Coot. Acta Crystallogr D Biol Crystallogr, 66, 486-501. FU, Y., SHEU, C., FUJITA, T. & FOOTE, C. S. 1996. Photooxidation of troglitazone, a new antidiabetic drug. Photochem Photobiol, 63, 615-20. GIBSON, G. G. & SKETT, P. 2001. Enzymology and molecular mechanisms. In: GIBSON, G. G. & SKETT, P. (eds.) Introduction to drug metabolism. Third ed. Bath, England: Nelson Thornes Ltd. GUENGERICH, F. P. & JOHNSON, W. W. 1997. Kinetics of ferric cytochrome P450 reduction by NADPH-cytochrome P450 reductase: rapid reduction in the absence of substrate and variations among cytochrome P450 systems. Biochemistry, 36, 14741-50. GUO, Z. J., CHAVEZ, K. J., ALVAREZ, J., ZHANG, X., NORRIS, B., MAHER, M., MORGAN, M., SCHUMACHER, R. J., CUELLAR, R., SEVRIOUKOVA, I. F., POULOS, T. L., DENISOV, I., SLIGAR, S. G., GUPTA, K., BLAIR, I. A., CAPDEVILA, J., KELEKAR, A., AMIN, E., GEORG, G. & POTTER, D. A. 2014. Breast cancer inhibition by a novel and potent biguanide, N1-hexyl-N5- benzyl-biguanide. Cancer Res, 74. GUO, Z. J., SEVRIOUKOVA, I. F., DENISOV, I. G., ZHANG, X., CHIU, T. L., THOMAS, D. G., HANSE, E. A., CUELLAR, R. A. D., GRINKOVA, Y. V., LANGENFELD, V. W., SWEDIEN, D. S., STAMSCHROR, J. D., ALVAREZ, J., LUNA, F., GALVAN, A., BAE, Y. K., WULFKUHLE, J. D., GALLAGHER, R. I., PETRICOIN, E. F., NORRIS, B., FLORY, C. M., SCHUMACHER, R. J., O'SULLIVAN, M. G., CAO, Q., CHU, H. T., LIPSCOMB, J. D., ATKINS, W. M., GUPTA, K., KELEKAR, A., BLAIR, I. A., CAPDEVILA, J. H., FALCK, J. R., SLIGAR, S. G., POULOS, T. L., GEORG, G. I., AMBROSE, E. & POTTER, D. A. 2017. Heme Binding Biguanides Target Cytochrome P450-Dependent Cancer Cell Mitochondria. Cell Chem Biol, 24, 1259-+. HARE, R. S. & FULCO, A. J. 1975. Carbon monoxide and hydroxymercuribenzoate sensitivity of a fatty acid (omega-2) hydroxylase from Bacillus megaterium. Biochem Biophys Res Commun, 65, 665-72. HAUNER, H. 2002. The mode of action of thiazolidinediones. Diabetes Metab Res Rev, 18 Suppl 2, S10-5. HILL, A. V. 1910. The possible effects of the aggregation of the molecules of haemoglobin on its dissociation curves. J Physiol-London, 4, 4-7. HO, P. P. & FULCO, A. J. 1976. Involvement of a single hydroxylase species in the hydroxylation of palmitate at the omega-1, omega-2 and omega-3 positions by a preparation from Bacillus megaterium. Biochim Biophys Acta, 431, 249-56. HONG, F., XU, P. & ZHAI, Y. 2018. The Opportunities and Challenges of Peroxisome Proliferator-Activated Receptors Ligands in Clinical Drug Discovery and Development. Int J Mol Sci, 19. JONSSON, B. 2002. Revealing the cost of Type II diabetes in Europe. Diabetologia, 45, S5-S12. JOOSTEN, R. P., LONG, F., MURSHUDOV, G. N. & PERRAKIS, A. 2014. The PDB_REDO server for macromolecular structure model optimization. IUCrJ, 1, 213-20. KABSCH, W. 2010. Xds. Acta Crystallogr D Biol Crystallogr, 66, 125-32. LI, H. & POULOS, T. L. 1997. The structure of the cytochrome P450BM-3 haem domain complexed with the fatty acid substrate, palmitoleic acid. Nat Struct Biol, 4, 140-6. LIPSCOMB, J. D. 1980. Electron paramagnetic resonance detectable states of cytochrome P-450cam. Biochemistry, 19, 3590-9.

206

LUTHRA, A., DENISOV, I. G. & SLIGAR, S. G. 2011. Spectroscopic features of cytochrome P450 reaction intermediates. Arch Biochem Biophys, 507, 26-35. MATSON, R. S., HARE, R. S. & FULCO, A. J. 1977. Characteristics of a cytochrome P- 450-dependent fatty acid omega-2 hydroxylase from Bacillus megaterium. Biochim Biophys Acta, 487, 487-94. MICHAELIS, L. & MENTEN, M. L. 1913. The kenetics of the inversion effect. Biochem Z, 49, 333-369. MIURA, Y. & FULCO, A. J. 1974. (Omega -2) hydroxylation of fatty acids by a soluble system from Bacillus megaterium. J Biol Chem, 249, 1880-8. MIURA, Y. & FULCO, A. J. 1975. Omega-1, Omega-2 and Omega-3 hydroxylation of long-chain fatty acids, amides and alcohols by a soluble enzyme system from Bacillus megaterium. Biochim Biophys Acta, 388, 305-17. MORRISON, J. F. 1969. Kinetics of the reversible inhibition of enzyme-catalysed reactions by tight-binding inhibitors. Biochim Biophys Acta, 185, 269-86. NARHI, L. O. & FULCO, A. J. 1986. Characterization of a catalytically self-sufficient 119,000-dalton cytochrome P-450 monooxygenase induced by barbiturates in Bacillus megaterium. J Biol Chem, 261, 7160-9. NARHI, L. O. & FULCO, A. J. 1987. Identification and characterization of two functional domains in cytochrome P-450BM-3, a catalytically self-sufficient monooxygenase induced by barbiturates in Bacillus megaterium. J Biol Chem, 262, 6683-90. NOBLE, M. A., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MACKAY, A. C., REID, G. A., HANZLIK, R. P. & MUNRO, A. W. 1999. Roles of key active-site residues in flavocytochrome P450 BM3. Biochem J, 339 ( Pt 2), 371-9. SCHOCH, G. A., YANO, J. K., SANSEN, S., DANSETTE, P. M., STOUT, C. D. & JOHNSON, E. F. 2008. Determinants of cytochrome P450 2C8 substrate binding: structures of complexes with montelukast, troglitazone, felodipine, and 9-cis- retinoic acid. J Biol Chem, 283, 17227-37. SHEN, Z., REED, J. R., CREIGHTON, M., LIU, D. Q., TANG, Y. S., HORA, D. F., FEENEY, W., SZEWCZYK, J., BAKHTIAR, R., FRANKLIN, R. B. & VINCENT, S. H. 2003. Identification of novel metabolites of pioglitazone in rat and dog. Xenobiotica, 33, 499-509. SOWDEN, R. J., YASMIN, S., REES, N. H., BELL, S. G. & WONG, L. L. 2005. Biotransformation of the sesquiterpene (+)-valencene by cytochrome P450cam and P450BM-3. Org Biomol Chem, 3, 57-64. WILD, S., ROGLIC, G., GREEN, A., SICREE, R. & KING, H. 2004. Global prevalence of diabetes: estimates for the year 2000 and projections for 2030. Diabetes Care, 27, 1047-53. WRIGHT, M. B., BORTOLINI, M., TADAYYON, M. & BOPST, M. 2014. Minireview: Challenges and opportunities in development of PPAR agonists. Mol Endocrinol, 28, 1756-68. YAMAZAKI, H., SHIBATA, A., SUZUKI, M., NAKAJIMA, M., SHIMADA, N., GUENGERICH, F. P. & YOKOI, T. 1999. Oxidation of troglitazone to a quinone- type metabolite catalyzed by cytochrome P-450 2C8 and P-450 3A4 in human liver microsomes. Drug Metab Dispos, 27, 1260-6. ZANINOTTO, P., HEAD, J., STAMATAKIS, E., WARDLE, H. & MINDELL, J. 2009. Trends in obesity among adults in England from 1993 to 2004 by age and social class and projections of prevalence to 2012. J Epidemiol Community Health, 63, 140-6. ZHANG, K., SHAFER, B. M., DEMARS, M. D., 2ND, STERN, H. A. & FASAN, R. 2012. Controlled oxidation of remote sp3 C-H bonds in artemisinin via P450 catalysts with fine-tuned regio- and stereoselectivity. J Am Chem Soc, 134, 18695- 704.

207

5.8. Supporting Information

Rhombic LS signals (gz/gy/gx) Compound Major species Minor species

WT Heme 2.41/2.25/1.91 None

WT Heme + DMSO 2.41/2.25/1.91 2.44/2.25/1.90

WT Heme + Troglitazone 2.44/2.25/1.90 2.41/2.25/1.91

DM Heme 2.43/2.25/1.91 2.40/2.25/1.92; 2.51/2.25/1.89

DM Heme + DMSO 2.45/2.25/1.91 2.51/2.25/1.89

DM Heme + Darglitazone 2.44/2.25/1.91 None

DM Heme + Pioglitazone 2.45/2.25/1.91 2.51/2.25/1.89

DM Heme + Rosiglitazone 2.45/2.25/1.91 2.51/2.25/1.89

DM Heme + Troglitazone LS 2.45/2.25/1.91; HS 7.94/3.64/1.70 LS 2.39/2.25/1.97; LS 2.51/2.25/1.89

DM Heme + Mitiglinide 2.45/2.25/1.91 2.51/2.25/1.89

DM Heme + Nateglinide 2.44/2.25/1.91 2.47/2.25/1.90

DM Heme + Repaglinide 2.44/2.25/1.91 2.47/2.25/1.90

DM Heme + Metformin 2.43/2.25/1.91 2.40/2.25/1.92; 2.51/2.25/1.89

Figure S10: Table of EPR g-values for the WT and DM heme domain proteins bound to a variety of antidiabetic compounds Only DM + troglitazone exhibited high spin species (HS). All other spectra displayed only low spin species (LS).

208

DM BM3 + Troglitazone

PDB ID 6HN8

Data collection

Space group P212121

Cell dimensions

a, b, c (Å) 60.9 120.4 146.0

() 90 90 90

Resolution range (Å) 29.5-2.00

Rmeas 0.149 (0.952)

I / I 10.8 (2.2)

Completeness (%) 99.4 (95.1)

Redundancy 6.6 (6.4)

Refinement

Resolution (Å) 29.5-2.00

No. reflections 73144

Rwork / Rfree 15.9/18.7 (20.8/23.3)

No. of residues

Protein 910

Water 903

Ligand/ion 2

B-Factors (Å2)

Protein 28.9

Ligand 33.2

R.m.s. deviations

Bond lengths (Å) 0.003

Bond angles () 0.86

Ramachandran

Favored (%) 96.9

Outlier (%) 0.1

Figure S11: Table of crystallographic data for the troglitazone bound DM variant The troglitazone-bound structure displays good density and resolution. The structure has been deposited on the PDB using the code provided.

209

Figure S12: LCMS/MS of rosiglitazone suggests the production of two distinct metabolites Overlapping peaks (RT – 3.52-4.06) corresponding to the parent compound and two products are present in the LCMS spectra. The smallest fragments observed by LCMS/MS are shown using pink bonds for the +16 and -14 products. If a human metabolite has been formed by the DM BM3 enzyme then the modification will have had to occur at positions marked with pink asterisks.

210

Figure S13: LCMS/MS of metformin shows a potential oxidative deamination reaction LCMS/MS of metformin oxidized by the DM P450 BM3 enzyme shows overlapping peaks (RT 0.6-0.95) corresponding to the parent compound and a potential product with a decrease of 16 amu. The smallest fragment observed by LCMS/MS is shown using pink bonds.

211

Figure S14: 1H-NMR shows the production of a single pioglitazone metabolite corresponding to an oxidation reaction by the DM BM3 variant The 1H spectrum for the pioglitazone starting material is shown with peaks labelled and integrated (red). The overlay shows pioglitazone with accepted IUPAC numbering generated from the InChI key, with peaks labelled accordingly (blue). Data were collected on a Bruker 500 MHz NMR in DMSO-d6.

Figure S15: 1H-NMR shows the production of a single troglitazone metabolite corresponding to an oxidation reaction by the DM BM3 variant The TOCSY experiment shows the relationship of protons in relation to their environment. The 1H spectrum for the racemic troglitazone starting material is shown with peaks labelled and integrated (red). The overlay shows pioglitazone with accepted IUPAC numbering generated from the InChI key, with peaks labelled accordingly (blue). Data were collected on a 500 MHz NMR in DMSO-d6.

212

Figure S16: 1H-NMR shows the production of five metabolites corresponding to hydroxylation reactions by the DM BM3 variant with repaglinide 1 The solvent used was DMSO-d6 for all ligands. The H spectrum for repaglinide with IUPAC numbering generated from the InChI key, with peaks labelled accordingly (red). Data were collected on a 500 MHz NMR in DMSO-d6.

213

Figure S17: 1H-NMR shows the production of a single metabolite and two eliminated smaller molecules corresponding to an oxidation reaction by the DM BM3 variant with darglitazone The 1H spectrum for the pioglitazone starting material is shown with peaks labelled and integrated (blue). The overlay shows pioglitazone with accepted IUPAC numbering generated from the InChI key, with peaks labelled accordingly (red). Further characterization of the metabolite was determined using the COSY spectra. Data were collected on a Bruker 500 MHz NMR in DMSO-d6.

214

Chapter 6: Binding of Fibrates to P450 BM3 Reveals Novel Changes to the P450 BM3 Landscape

6.1. Abstract

P450 BM3 is a natural cytochrome P450-cytochrome P450 reductase fusion protein that binds to fatty acid ranging in length from ~12 to 18 atoms in carbon chain length. The introduction of two active site mutations (A82F/F87V), termed gatekeeper mutations, expands the substrate selectivity profile of this BM3 variant. The mutations cause structural changes that allow diverse compounds, such as pharmaceuticals, to bind within the active site. Compounds from the fibrate class of pharmaceuticals were demonstrated to elicit substantial Soret high-spin heme iron shifts in ligand-binding UV-

Vis spectrophotometric titration assays, consistent with substrate-type behaviour. By probing the binding modes of these compounds through spectroscopic (binding titrations and EPR) and mass spectrometry-based techniques (native mass spectrometry [MS], collision-induced unfolding [CIU] and hydrogen-deuterium exchange mass spectrometry

[HDX-MS]), novel conformational changes were observed. In addition, human metabolites could be produced with the DM BM3 variant and the widely prescribed drug gemfibrozil.

6.2. Introduction

The cytochrome P450 superfamily contains monooxygenase enzymes vital for xenobiotic metabolism for organisms from all Kingdoms (Danielson, 2002). In the 1970s an unusual P450 enzyme was isolated from Bacillus megaterium (Miura and Fulco, 1974,

Hare and Fulco, 1975, Matson et al., 1977). This enzyme, termed P450 BM3 (CYP102A1), is a natural fusion enzyme containing a heme-binding catalytic domain (~55 kDa) linked to a diflavin reductase domain (~65 kDa) (Narhi and Fulco, 1986, Narhi and Fulco, 1987).

215

The diflavin reductase domain binds NADPH and shuttles electrons to the heme prosthetic group, allowing this protein to have the fastest reported catalytic rate of any P450 monooxygenase enzyme (Noble et al., 1999). P450 BM3 is able to bind to a range of fatty acid molecules (particularly in the C12-C18 range) (Miura and Fulco, 1975, Ho and Fulco,

1976). However, the physiological role of P450 BM3 remains uncertain and the natural substrate or substrates for the enzyme have been suggested to be either saturated (English et al., 1997) or unsaturated fatty acids (English et al., 1994, Makita et al., 1996). Due to its high catalytic rate and flexible active site, P450 BM3 has been the subject of many protein engineering studies with the aim of altering the BM3 substrate selectivity profile. There are several examples of BM3 mutants that allow this enzyme to bind diverse compounds, including large pharmaceutical molecules (Acevedo-Rocha et al., 2018, Zhang et al., 2012,

Ren et al., 2015). For example, the so-called “gatekeeper” mutations which create additional ligand binding space in the active site (F87V) (GrahamLorence et al., 1997) and cause conformational changes allowing the protein to favour a catalytically primed conformational state (A82F) (Huang et al., 2007). This P450 BM3 A82F/F87V double mutant (DM) enzyme was shown to produce human P450 metabolites from proton pump inhibitor drugs such as omeprazole (Butler et al., 2013, Butler et al., 2014).

P450 enzymes display several features in the UV-Vis spectrum associated with the heme prosthetic group. The major absorbance band is the Soret () peak (at ~418 nm for P450 BM3). In its hexacoordinated form (with four equatorial ligands from heme pyrrole nitrogens and axial ligands from a water molecule and a cysteine residue in its thiolate form) the heme iron atom resides in an inactive, low-spin (LS) state. Upon substrate binding, the distal axial water molecule is displaced and the heme iron spin-state equilibrium shifts towards the pentacoordinated high-spin (HS) form with a Soret peak shift to ~394 nm accompanied by an electronic rearrangement in the heme iron 3d-orbitals.

The extent of the HS shift induced by the binding of specific substrates can be used as an 216 indicator of the “goodness of fit” for the substrate and its ability to penetrate the active site close to the heme iron. A good example is seen in our previous studies with the fatty acid decarboxylating P450 OleTJE (CYP152L1). Here, the binding of the C20:0 substrate arachidic (eicosanoic) acid induces a near-complete conversion to the HS state, and a crystal structure of the OleTJE/arachidic acid complex shows that the large substrate occupies the vast majority of the OleTJE substrate binding cavity (Belcher et al., 2014). The dissociation constants (Kd values) for P450 substrates can usually be readily determined by plotting the extent of substrate-induced HS shift against the concentration of the substrate, and by fitting these data using a hyperbolic or other appropriate fitting function. In this way, both substrate binding affinity and extent of HS shift be compared (Tran et al., 2012).

Recently, we have screened a 978 drug compound library in order to identify molecules able to bind to the DM BM3 variant. By additions of each of the library compounds to samples of the DM BM3 heme domain, a substantial number of compounds were found to exhibit substrate-like behaviour through inducing HS heme absorbance shifts. The molecule found to induce the greatest percentage HS shift was bezafibrate, a widely prescribed hypolipidemic drug (Chapter 3).

The fibrates molecules were first developed in the 1950s. However, it was not until 1967 until the first fibrate (clofibrate) was approved for human use (Backes et al.,

2007). This compound was discontinued in 2002 as more effective fibrate molecules were produced which exhibited fewer side effects, including bezafibrate and fenofibrate (Oliver,

2012). The fibrate drugs are designed to bind to the transcription factor peroxisome proliferator-activated receptor alpha (PPAR). PPAR is known to regulate the expression of many proteins, including P450 enzymes of the CYP2C family in humans, by the binding of fatty acid molecules (Lalloyer and Staels, 2010). Downregulation of CYP2C11 expression has been demonstrated by treatment with fenofibrate (Vecera et al., 2011) and clofibric acid (the active form of clofibrate) (Lake et al., 1984, Fan et al., 2004). 217

Manipulation of PPAR was shown to decrease tumour sizes due to the reduction in levels of epoxyeicosatrienoic acids (EETs) (Pozzi et al., 2010, Pozzi et al., 2007). EETs are produced by the modification of arachidonic acid molecules by cytochrome P450 epoxygenases, many of which belong to the CYP2C family (Spector, 2009). The fibrate drugs are currently being researched as a potential method for combating multidrug resistance, specifically during chemotherapy (Cizkova et al., 2012). The production of the fibrates has also led to the development of other pharmaceutical drugs, such as the thiazolidinediones. These drugs, also referred to as glitazones, are antidiabetic molecules which also target peroxisome proliferator-activated receptors (PPARs) (Lalloyer and

Staels, 2010).

There are several drug classes used to treat hyperlipidaemia, including statins and drugs of the fibrate class. Depending on the specific drugs compared, fibrates appear to be more efficient at increasing high-density lipoprotein (HDL) levels, whilst statins appear to be more efficient at decreasing low-density lipoprotein (LDL) levels (Dembowski and

Davidson, 2008). Both of these effects are desirable for treatment of hyperlipidaemia, dyslipidaemia, and potentially for the treatment of cardiovascular diseases and diabetes.

Due to their different benefits, combination treatments have been proposed. However, these proved unsuccessful as many fibrates inhibit CYP2C8 and CYP2C9, which are the main metabolizing enzymes of the statin drug class and of many other pharmaceuticals

(Prueksaritanont et al., 2005, Zhu et al., 2009, Wen et al., 2001, Prueksaritanont et al.,

2002).

In this study, members of the fibrate drug class were chosen for further investigation due to their important pharmaceutical properties and in light of the extensive

HS shifts exhibited upon their binding to the DM BM3 heme domain. Bezafibrate was found to induce conformational changes using HDX-MS (hydrogen-deuterium exchange 218 mass spectrometry) that were quite different from the conformational changes observed using a tight binding fatty acid derivative, the modified fatty acid N-palmitoyl glycine

(NPG). These data suggest that bezafibrate interacts with the DM BM3 variant through an alternative binding mode to that used by fatty acids (including NPG). Other molecules in the fibrates class appear to use the more typical binding mode, such as the non-halogenated drug gemfibrozil. Known human metabolites were successfully produced from the reaction of the DM P450 BM3 variant with gemfibrozil.

6.3. Materials and Methods

6.3.1. Protein expression and purification of P450 BM3

For all binding titrations, EPR and DM HDX-MS data collection, the BM3 heme domain was used. The pET20b vector encoding the heme domain protein was transformed into the E. coli BL21 (DE3) strain and cells were grown in terrific broth medium

(Formedium, Hunstanton, UK) for the wild-type (WT) protein and in autoinduced terrific broth (Formedium) for the DM variant. Cells were grown for 24 hours at 37 °C and at 170 rpm in an orbital shaker. 100 µM δ-aminolevulinic acid was added when the cell culture reached an OD600 of 0.6 in order to promote heme incorporation into the enzyme. Cells were harvested by centrifugation at 4 °C (6000 g, 20 min). Protein was purified by resuspending the cell pellets in ice-cold Tris buffer (50 mM Tris-HCl, 1 mM EDTA, pH

7.2) containing protease inhibitors (1 tablet per 100 mL of EDTA-free cOmpleteTM tablets, Roche Applied Science, Burgess Hill, UK), and DNase (Merck, Nottingham UK)

(10 µg mL-1). Cells were lysed by sonication on ice using a Bandelin Sonopuls instrument at 37% amplitude with 12 pulses for 40 s, and with 60 s breaks between pulses. The cell extract was clarified by centrifugation (4600 g for 60 minutes at 4 °C). Contaminant proteins were removed by incubation with 30% w/v ammonium sulfate using a 1-hour

219 incubation at 4 °C with gentle agitation. Precipitated material was removed using centrifugation (4600 g for 15 minutes at 4 °C). A non-His tagged version of the protein was used in order to obtain high-resolution structures of the DM heme domain in complex with substrates identified from the compound screen in future studies. The non-tagged heme domain was purified in four successive chromatographic steps on DEAE Sepharose, hydroxyapatite, Q-Sepharose and HiLoad™ GF S200 16/600 Superdex™ 200 pg columns.

The method for the non-tagged DM HD purification is detailed in (Chapter 4).

For metabolite production full-length (intact) DM BM3 protein was used. Intact protein was expressed as above using the vector pET15b. Cells were harvested using 50 mM KPi containing 350 mM KCl and 10% v/v glycerol, pH 8. For cell lysis, this buffer also contained protease inhibitors (1 tablet per 100 mL, EDTA-free cOmpleteTM tablets), and DNase (10 µg mL-1). Cells were lysed and clarified by centrifugation and an incubation with ammonium sulfate as done as described as above. Intact DM BM3 protein was incubated with nickel-IDA resin overnight at 4 °C with 50 mM KPi containing 350 mM KCl and 10% glycerol, pH 8. The protein-bound resin was then applied to a column and a stepwise gradient of 10 mM (300 mL), 20 mM (200 mL) and 200 mM (60 mL) imidazole was applied to elute the protein. The eluted protein was dialyzed into 25 mM

KPi, pH 6.5 and applied to 28 mL columns containing CHT hydroxyapatite type 1 resin

(Bio-Rad Laboratories). Protein was eluted using a linear gradient of 25-300 mM KPi, pH

6.5 (600 mL). Before any experiments were undertaken, any hydrophobic ligands (e.g. fatty acids) were removed from the protein using the hydrophobic resin Lipidex-1000

(Perkin-Elmer, Beaconsfield, UK) using the relevant experimental buffer.

220

6.3.2. Binding affinity determination

All ligand compounds (fibrates and NPG) were sourced from Sigma Aldrich

(Haverhill, Cambridgeshire, UK). A Cary 60 UV-Vis spectrophotometer (Agilent,

Stockport, UK) with a Peltier at 30 °C was used for data collection. Titrations were undertaken with 2-4 µM of protein in 100 mM KPi, pH 7 in a quartz cuvette, with additions of small amounts of the ligand until saturation was achieved. The experimental buffer for all titration experiments was 100 mM KPi, pH 7 was. Binding affinities were calculated using the Michaelis-Menten equation for hyperbolic curves, the Morrison equation for tight binding ligands where the Kd value is ≤ 5x the P450 concentration, or the

Hill equation for sigmoidal curves (Michaelis and Menten, 1913, Morrison, 1969, Hill,

1910).

6.3.3. EPR analysis of ligand binding to WT and DM BM3

X-band EPR spectroscopy was used to analyze the binding of ligands to the WT and DM BM3 heme domains in their ferric state. Samples were prepared with 200 M protein in 25 mM Kpi, pH 7 with the addition of 5 mM drug in DMSO, or with appropriate volumes of DMSO or buffer as controls. Samples were incubated overnight at 4 C on a

Stuart roller/shaker device at 10 rpm (Cole-Parmer, Stone, UK). Samples were centrifuged in a microfuge (14,000 rpm for 10 minutes at 4 °C) before transferring to EPR tubes and freezing in liquid nitrogen. X-band EPR spectra were recorded on a Bruker ER-300D series electromagnet with a microwave source interfaced with a Bruker EMX control unit and fitted with an ESR-9 liquid helium flow cryostat (Oxford Instruments), and using a dual-mode microwave cavity from Bruker (ER-4116DM). Spectra were recorded at 10 K with a microwave power of 0.5 mW, a modulation frequency of 100 kHz and a modulation amplitude of 5 G.

221

6.3.4. Native MS studies

On the day of analysis, the buffer was exchanged into 100 mM ammonium acetate, pH 6.9 using micro Bio-Spin Chromatography columns (Bio-Rad, Micro Bio-Spin

6 Columns) following the instructions specified by the manufacturer. The buffer was exchanged twice to ensure no residual salts were present. Native MS and ion mobility mass spectrometry (IM-MS) data were acquired on a Synapt G2S HDMS instrument (Waters,

Manchester, UK). NanoESI capillaries were prepared in-house from thin-walled borosilicate capillaries (inner diameter 0.9 mm, outer diameter 1.2 mm, World Precision

Instruments, Stevenage, UK) using a Flaming/Brown P-97 micropipette puller (Sutter

Instrument Company, Novato, CA, USA). A positive voltage was applied to the solution via a platinum wire (Goodfellow Cambridge Ltd, Huntingdon, UK) inserted into the capillary. Gentle source conditions were applied to preserve the native-like structure: capillary voltage 1.2-1.5 kV, sampling cone 40 V, source temperature 40 °C. Trap collision energy was 4 V, transfer collision energy was set to 0 V. For the IM, the helium cell and the IMS gas flows were 180 and 90 mL/min, respectively; the IMS wave velocity was 400 m/s, and the IMS wave height was 35 V. Nitrogen was the carrier gas. External calibrations of the spectra were achieved using solutions of cesium iodide (2 mg/mL in 50:50 water: isopropanol). Data were acquired and processed with MassLynx software (Waters).

6.3.5. Collision-induced unfolding (CIU) analyzes

Experiments were performed on a Waters Synapt G2S (Waters) using nanoESI and trap-activated ion mobility; capillary voltage 1.2 – 1.5 kV, cone voltage 40 V and source temperature 40 °C. The helium cell and the IMS gas flows were 180 and 90 mL/min, respectively; the IMS wave velocity was 400 m/s, and the IMS wave height was

35 V. Nitrogen was the carrier gas. The most intense charge state for each protein species was mass selected using the quadrupole prior to the trap region. Activation was induced by 222 elevating the trap collision energy. ORIGAMI (Migas et al., 2018) was used to automatically acquire data for collision energies from 4 – 200 V in 2 V increments, as well as for data processing.

6.3.6. HDX-MS characterization

The HDX-MS setup was comprised of a Waters nano-Acquity UPLC system with

ESI MS detection coupled to a LEAP Technologies dual-armed robot for sample preparation, incubation and inlet injection. A Waters Synapt G2S mass spectrometer was operated in positive ion/resolution mode, with data acquired over the m/z range 290-2500.

30 μM protein solutions were diluted 20 fold into 10 mM KPi in either H2O or D2O, pH/pD 7, and the mixture incubated at 20 °C for 0 minutes (H2O), or 30 seconds, 1, 10, 30 or 60 minutes (D2O), before the quench step. HDX quenching was achieved by mixing the reaction solution 1:1 with cooled 100 mM KPi (pH 2.5, 0 °C). ~37.5 pmol were injected into the HDX module (0 °C), and washed over a pepsin column (Waters Enzymate BEH

-1 Pepsin 2.1 x 30 mm) with 0.1% HCOOH in H2O, pH 2.5, at 200 μL min . Resulting peptides were trapped on a VanGuard C18 trap column. Peptide separation was achieved on a C18 column (Waters Acquity UPLC BEH C18 1.7 µm, 1.0 x 10 mm) at 40 µL/min flow over 16 min with the following gradient: 0 min: 5% B, 7 min: 35% B, 8 min: 85% B,

11 min: 5% B, 12 min: 95% B, 13 min:5% B, 14 min: 95% B, 15 min: 5% B (mobile phases: A, water + 0.1 % formic acid; and B, acetonitrile + 0.1 % formic acid). The mass spectrometer was operated in ToF only mode. LeuEnk peptide was used as Lock Spray.

Data were acquired using Waters MassLynx software v4.1, with the LEAP robot controlled by HDx Director 1.0.3.9.

Data were analyzed using DynamX (DynamX HDX Data Analysis Software 3.0,

Waters). Python scripts of the deuterium uptake were generated using the data collected at

223

1 minute and 1 hour. Python scripts were mapped onto structures in PyMOL (The PyMOL

Molecular Graphics System, Version 1.7.4.5 Schrödinger, LLC). High coverage was achieved for these data (88%).

6.3.7. Product determination by HPLC, LCMS and LCMS/MS

To ensure efficient turnover of compounds for initial HPLC, LCMS and

LCMS/MS studies, a high stoichiometry of protein-to-ligand was utilized. No attempts were made to optimize turnover rates or product quantity. Samples were produced by incubating 50 µM DM P450 BM3 protein with 100 µM ligand. An electron regeneration system of 7.76 mM glucose-6-phosphate, 0.6 mM NADP+ and 0.75 U/mL glucose-6- phosphate dehydrogenase was used. The samples were made up to 500 L in turnover buffer (50 mM KPi, 5 mM CaCl2, pH 7.4) in 9 mL glass vials. Turnover reactions were completed at 37 °C, 170 rpm for 3 hours in an orbital shaker.

The solvents used for HPLC, LCMS and LCMS/MS were Chromasolv® acetonitrile and Chromasolv® water (Thermo Fisher Scientific, Runcorn, UK) both with

0.1% formic acid (Sigma Aldrich). For HPLC and LCMS reactions samples were quenched with 1:1 acetonitrile and clarified using 2 mL Impact® protein precipitation plates (Phenomenex, Macclesfield UK). HPLC and LCMS analyzes were completed on an

Agilent 1100 system with an Agilent 1100 diode array detector (DAD) and an Agilent

1100 LC/MSD ion trap. A linear gradient from 10 to 100% acetonitrile over 10 minutes was utilized, followed by a hold of 2 minutes at 100% acetonitrile before returning to 10% acetonitrile over 10 minutes. For each sample 10 L was injected per run. The column used for HPLC and LCMS was a C18 Kinetex column (100 x 2.1 mM, 2.6 M) with a constant flow rate of 1 mL min-1 kept at room temperature (Phenomenex, Macclesfield

UK). 224

LC-MS/MS was undertaken using a Q Exactive Plus instrument equipped with a heated electrospray ionization source and a U3000 UHPLC (Thermo Fisher Scientific).

Prior to analysis reactions were purified using 1 mL, 33 mm poly-metric reversed phase

Stata™-X columns (Phenomenex) and eluted into 1 mL Chromasolv® acetonitrile. The reactions were chromatographically separated using a gradient of water + 0.1 % formic acid (solvent A) and methanol + 0.1 % formic acid (solvent B), and a Hypersil Gold column (100 x 2.1 mm, 3 m, Thermo Fisher Scientific). The gradient was programmed to hold at 90% A for 1.5 min, followed by a linear decrease to 1% A over 5 min and then held for 1 min before returning to 90% A in 0.1 min and with a final hold time of 1.22 min. The flow rate was constant at 400 L/min, the injection volume was 5 L and the column was maintained at 40 °C. Mass spectrometry was collected in positive and negative ionization in separate acquisitions. Full scan mass spectrometry was conducted at a resolution of

70000 with an AGC target of 3e6 and a scan range of 90-1000 m/z. Data-dependent

MS/MS was conducted on the top 5 most abundant ions, with a resolution of 17500, AGC target of 5e4 and a collision energy of 30 or 50 eV.

6.3.8. Product determination by NMR

NMR product determination required larger quantities of reaction material. To accomplish this, 3 µM protein was reacted with 45 µM of ligand in 60 mL flasks. The same electron regeneration system was used as for the HPLC, LCMS, and LCMS/MS experiments.

Turnover reactions were completed at 37 °C, 80 rpm for 3 hours in an orbital shaker.

Reactions were purified using Stata-X columns (Phenomenex) and eluted in 1 mL dimethyl sulfoxide-d6. NMR spectra were recorded on a Burker AVIII 500 MHz spectrophotometer with a 1H/13C TCI cryoprobe equipped with Z-gradients. All spectra were recorded at 298

K. Spectra were processed and analyzed with Topspin 3.2 (Bruker). 1H-1H NOESY spectra

225 used excitation sculpting solvent suppression and gradients in T1. Where novel products were made, 1H-13C HSQCs were recorded using the gradient-selected sensitivity-enhanced method and assigned on the basis of HSQC-TOCSY spectra.

Figure 27: The structures of a variety of fibrate compounds The fibrate class of pharmaceutical drugs is widely prescribed for hyperlipidaemia. Some molecules are halogenated to improve drug-target binding affinity. Atoms are coloured by element.

6.4. Results

6.4.1. Fibrate drug binding affinity determination

The fibrate drugs bezafibrate, fenofibrate and gemfibrozil from an FDA-approved compound screen were found to bind to the DM heme domain through their ability to induce a high-spin shift in the P450 spectrum (Figure 27). To further investigate this class

226 of drugs, the binding affinities for these compounds (marked with an asterisk in Table 5) and others sourced separately from the same drug class (ciprofibrate, clinofibrate and clofibrate) were determined for both WT and DM P450 BM3 heme domain (Table 5).

Many compounds showed tight binding to the DM heme domain. In comparison, only bezafibrate and ciprofibrate were able to bind to the WT heme domain, albeit with much weaker affinities. Figure 28 compares the binding of bezafibrate to that for the modified fatty acid N-palmitoyl glycine (NPG). NPG is a very tight-binding substituted lipid which i nduces a near-complete Soret peak shift. However, the binding of bezafibrate also produces a substantial HS shift in the DM heme domain spectrum with ~87% conversion to the HS state. The Kd values determined for NPG and bezafibrate are 0.07 M and 1.04

M, respectively.

WT BM3 DM BM3 FDA-approved compound MW (Da) Kd (M) Kd (M)

Bezafibrate* 362.8 224.5 ± 4.5 1.04 ± 0.07

Ciprofibrate 289.1 325.2 ± 10.0 31.16 ± 1.68

Clinofibrate 468.6 NB 92.01 ± 6.14

Clofibrate 242.7 NB 45.65 ± 2.25

Fenofibrate* 360.8 NB 0.15 ± 0.01

Gemfibrozil* 250.3 NB 39.94 ± 2.08

Table 5: Binding constants for fibrate class molecules demonstrate higher affinity for the DM BM3 heme domain than for the WT BM3 heme domain Fibrate drugs were shown to bind to both the WT and DM BM3 heme domains. The table shows the relevant Kd values for each fibrate drug (NB indicates instances where no binding is apparent in UV-visible spectroscopic titrations for the WT BM3 heme domain). The Kd values for fibrates binding to the DM heme domain range from 0.15 M (fenofibrate) to 92.01 M (clinofibrate). Only two of the fibrate compounds (bezafibrate and ciprofibrate) were able to bind to the WT BM3 heme domain and induce a high-spin spectral shift. However, their binding affinities were weak (224.5 M and 325.2 M, respectively). The Kd values were determined as described in the Methods section, using either a hyperbolic function (Michaelis-Menten), the Hill function or the Morrison equation for tight-binding ligands, as appropriate. The fibrates marked with an asterisk were those identified as hits from a 978 compound library screen.

227

Figure 28: Spectral binding titrations for the modified fatty acid NPG and the fibrate drug bezafibrate with the BM3 DM heme domain The N-acyl amide BM3 substrate N-palmitoyl glycine (NPG) is a tight-binding P450 BM3 substrate that induces a substantial Soret peak shift. The hyperlipidaemia drug bezafibrate also induces a similar extent of Soret peak shift (~87% HS conversion) at apparent saturation. The Kd values calculated for NPG and bezafibrate are 0.07 ± 0.01 M and 1.04 ± 0.07 M respectively.

6.4.2. EPR spectroscopy of WT and DM BM3 heme domain in complex with

fibrate drugs

EPR was conducted on the BM3 DM heme domain in complex with fibrate drugs

(Figure S18). The addition of the carrier solvent DMSO to the DM heme domain causes a change to the major low-spin EPR trio from 2.43/2.25/1.91 to 2.45/2.25/1.91. Addition of each of the fibrate drugs does not induce any substantial change to the g-values for the various complexes. For the DMSO-free DM heme domain, two minor species are also observed at 2.40/2.25/1.92 and 2.51/2.25/1.89, with DMSO addition resulting in the loss of the latter species. The bezafibrate-bound heme domain exhibits on a single major species, while the complexes with ciprofibrate, clinofibrate, clofibrate and fenofibrate all show the same pair of rhombic low-spin species at 2.45/2.25/1.91 (major) and 2.52/2.25/1.89

(minor). The interactions of the DMSO solvent with the DM heme domain likely dominate the EPR spectral feature for these heme domain/fibrate complexes. However, the outlier is

228 the gemfibrozil complex, where modest changes occur for both the major and minor species (2.44/2.25/1.91 and 2.47/2.25/1.90, respectively).

6.4.3. Production of human metabolites from fibrate drugs

The fibrate drugs bezafibrate, ciprofibrate, clinofibrate and clofibrate failed to generate observable products in studies using LCMS, LCMS/MS and NMR. However, through

LCMS and LCMS/MS approaches, oxidized metabolites were observed for both fenofibrate and gemfibrozil. Further analysis by NMR fully identified these products.

Fenofibrate generated three products: (i) a hydroxylated product on the hydroxybenzoyl ring that leads to the loss of the chlorine atom, and (ii) the hydroxylation on one or the other of the terminal methyl groups at the opposite end of the molecule (Figure S19). In the case of gemfibrozil, two metabolites were identified from NMR studies. Both of these were hydroxylation reactions on one or the other methyl group on the dimethylated benzene ring (Figure S20). These two products are both known human metabolites of gemfibrozil.

6.4.4. Native MS and CIU studies

The native mass spectra of DM BM3 did not change on bezafibrate or NPG binding. Using the most prevalent charge peak from native MS, CIU and CCS (collision cross-section) values were calculated. There was no change in CCS values resulting from substrate binding or in the solvent control (Figure S21).

Analysis of the unfolding of the DM heme domain protein in the substrate-free and substrate-bound states using collision-induced unfolding (CIU) shows a number of

229

Figure 29: CIU demonstrates differences in the unfolding patterns of DM ligand- bound proteins By applying increasing collision voltage to substrate-free and substrate-bound DM heme domains, unfolding is observed. Substrate-free (Panel A), ethanol solvent control (Panel B) and NPG-bound DM heme domains (Panel C) show little differences in their unfolding pattern. For the bezafibrate sample (Panel D) and the 1:1 NPG/Bezafibrate sample (Panel E), stabilization of the unfolding intermediates is observed across the spectra.

230

differences in unfolding patterns. There is little difference between the substrate-free, ethanol solvent control and NPG-bound spectra (Figure 29A-C respectively). Bezafibrate binding appears to stabilize the unfolding intermediates so that all states are still present at

150 V, while other samples have fully unfolded at this voltage (Figure 29D). This mixture of intermediate states is also present to a lesser degree in the sample containing both NPG and bezafibrate in a 1:1 stoichiometry (Figure 29E). There is no change in the voltage needed to begin protein unfolding.

6.4.5. HDX-MS analysis of substrate binding to WT and DM BM3 heme domains

To investigate the binding mode of bezafibrate, HDX-MS was undertaken.

Previously, HDX-MS has been conducted on the WT enzyme in order to investigate the dimeric interface of the protein. During these experiments, the substrate-free and NPG- bound WT enzymes were analyzed by HDX-MS. These results are shown in more detail in

Chapter 2. The binding of NPG to the WT BM3 heme domain (PDB 1JPZ) causes substantial changes to the surface of the protein, as shown in Figure 30A-D. These changes become much more apparent after a 1-hour incubation with deuterium (Figure

30C/D) than after only a 1 minute of incubation with deuterium (Figure 30A-B).

Widespread changes in the solvent accessibility of residues are displayed as deshielding (red-orange) and shielding (cyan-dark blue) events. In comparison, NPG binding to the DM BM3 heme domain (PDB 4KEY) causes very little change, with practically no change visible after 1 minute of incubation with deuterium (Figure 30E-H).

231

Figure 30: Comparing the binding of NPG in WT and DM heme domain proteins by HDX-MS By observing the uptake of deuterium, conformational changes to the protein on substrate binding can be observed. The protein was incubated with deuterium for different lengths of time, allowing for the observation of structural changes. Orange-to-red indicates an increase in deuterium uptake due to the residues becoming more solvent accessible (deshielded) and cyan-to-dark blue indicates a decrease in deuterium uptake due to the 232 residues becoming shielded. Deuterium uptake values were calculated by subtracting the deuterium uptake of the substrate-free protein from that of the substrate-bound protein. These values were applied to the NPG-bound WT structure PDB 1JPZ or the omeprazole- bound DM structure (PDB 4KEY) accordingly. Panels A-D: Data collected in Chapter 2 and are presented here for comparison to the DM heme domain. Panels A/B show NPG- bound WT protein incubated with deuterium for 1 minute, and rotated by 180 degrees. Panels C/D show NPG-bound WT protein incubated with deuterium for 1 hour. Panels E- H: NPG-bound DM protein incubated with protein for 1 minute (E/F) and 1 hour (G/H), and rotated by 180 degrees. The mutations within the DM variant allow the protein to adopt a catalytically primed state, as evidenced by the lack of changes (grey) to the deuterium uptake for the DM heme domain on ligand binding.

The relative uptake of deuterium did not change drastically during N-palmitoyl glycine (NPG) binding. Only two areas show a marked change in deuterium uptake for the

DM NPG-bound protein. The first is a shielded area of the DM heme domain (cyan-dark blue) that is also exhibited in the WT protein containing residues ~13-20 at the N-terminus

(Figures 30E/F). The other is another shielded area (cyan) containing the F- and G- helices (residues 167-207). In comparison, bezafibrate causes widespread changes in deuterium uptake across the DM heme domain corresponding to deshielding, the most significant of which are shown in Figure 31A/B. Intriguingly, deshielding occurred across all -sheet regions. In particular, the A- and F- helices exhibit the most significant deshielding in small 10-15 residue regions (orange). When the DM heme domain was bound to a 1:1 mixture of NPG and bezafibrate, a mixture of the deuterium uptake changes observed for the binding of the individual ligands were observed, such as the shielding at the N-terminus for the NPG binding (though to a lesser extent than the NPG-bound protein) and the deshielding in the A- and F-helices for bezafibrate binding (Figure 30E/F and Figure 31A-D).

6.5. Discussion

The “gatekeeper” double mutant (DM, A82F/F87V) form of the P450 BM3 enzyme was shown to be a particularly promiscuous variant of this biotechnologically 233

Figure 31: Comparing the binding of bezafibrate and NPG ligands to the DM BM3 heme domain variant by HDX-MS By subtracting the deuterium uptake of the substrate-free protein from that of the substrate- bound protein differential values could be calculated. These values were applied to the omeprazole bound DM heme domain structure PDB 4KEY (Butler). Panels A/B: Bezafibrate binding causes large conformational changes across the protein. Most significantly to the A- and F- helices (orange). Panels C/D: Incubation of the DM heme domain with a mixture of NPG and bezafibrate results in an amalgamation of both of the deuterium uptake patterns for NPG and bezafibrate. Distinct shielding (cyan) and deshielding (orange) events for the binding of both ligands are observed.

important enzyme (Chapter 3). In this study, we identified three members of the fibrate class of hyperlipidaemic drugs (bezafibrate, fenofibrate and gemfibrozil) from an FDA- approved compound library as potential substrates for the DM enzyme through establishing that the binding of these compounds induced substantial high-spin P450 Soret absorbance shifts in the DM heme domain. Three other fibrate class compounds

(ciprofibrate, clinofibrate and clofibrate) were also selected as potential substrates, and

234 were also shown to bind to the DM protein and to induce high-spin shifts consistent with substrate-type behaviour. The binding constants (Kd values) for these fibrates with the DM heme domain ranged from 0.15 M (fenofibrate) through to 92M (clinofibrate).

However, bezafibrate (Kd = 1.04 M) induced the greatest extent of high-spin shift (ca

87%). There was a clear difference in binding affinity between the WT and DM proteins – with only bezafibrate and ciprofibrate showing any significant spectral binding to the WT enzyme with Kd values of 224 M and 325 M, respectively (Table 5).

Further investigations were done using EPR in order to investigate the binding of these fibrate drugs to the WT and DM heme domains. However, in most cases the effect of the carrier solvent (DMSO) for the fibrates masked any EPR changes specific to these drugs. However, an exception was gemfibrozil, which was the only fibrate drug to cause distinctive changes to both the major and minor sets of low-spin EPR signals a difference in the EPR spectrum (Figure S18). In addition, the reaction of the DM BM3 enzyme with gemfibrozil was the only one for which the fibrate was shown to produce human metabolites (Murai et al., 2004). The human enzymes responsible for these reactions have not yet been identified, as these products were identified from liver microsome samples

(Murai et al., 2004). Around 50-70% of pharmaceutical drugs are metabolized by cytochrome P450 enzymes, with CYP3A4, CYP2C8 and CYP2C9 being among the most prominent enzymes involved (Ince et al., 2013, Guengerich, 2003). However, gemfibrozil was shown to inhibit the main pharmaceutical metabolizing P450 enzymes CYP2C8 and

CYP2C9 (Wen et al., 2001, Tornio et al., 2017). Selected studies indicate that gemfibrozil does not inhibit CYP3A4, or does so at only a low level (Wen et al., 2001, Prueksaritanont et al., 2002). However, other studies suggest fibrates are metabolized by CYP3A4 (Miller and Spence, 1998). The two hydroxylated products of gemfibrozil formed in our study only contribute to ~30% of the NMR signal for the DM BM3 variant reaction. Gemfibrozil

235 was one of the first fibrates produced and does not contain a halogen atom, hence its deviation from the fibrate name (Creger et al., 1976, Rodney et al., 1976). In our study, fenofibrate was also successfully metabolized by the DM BM3 variant. However, the metabolites produced in this case were not consistent with known existing human metabolites (Elsom et al., 1976, Chapman, 1987). The DM BM3 enzyme produces metabolites in which (i) the terminal phenyl group becomes hydroxylated, and (ii) one or the other of the terminal methyl groups at the opposite end of the molecule is hydroxylated.

This suggests that two alternative binding modes can be occupied by fenofibrate in the active site. In the case of gemfibrozil, hydroxylations occur to one or the other of the two methyl groups on the phenyl substituent group. For gemfibrozil, this suggests that a single binding orientation likely exists in the DM BM3 active site, and that sufficient mobility of the substrate and/or protein can occur to enable the hydroxylations on the methyl groups on the opposite sides of the gemfibrozil phenyl group.

Older types of fibrate compounds, such as gemfibrozil, were found to inhibit many P450 enzymes, including the drug-metabolizing enzymes CYP2C8 and CYP2C9

(Wen et al., 2001, Tornio et al., 2017). As a result, great care must be taken when prescribing fibrate compounds as this could lead to negative side effects, especially when fibrates are given in combination with other compounds for which P450 metabolism is vital. Bezafibrate and fenofibrate are newer halogenated fibrate compounds with masses larger than their preceding fibrates. Fenofibrate is the best-selling fibrate in the US and

Canada (Jackevicius et al., 2011). Interestingly, bezafibrate and fenofibrate do not appear to cause CYP2C8 inhibition, as demonstrated through studies of the metabolism of the antidiabetic compound repaglinide (Kajosaari et al., 2004). However, there is some contradicting evidence to suggest that fenofibrate, in particular the fibric acid form (the main human metabolite), does mildly inhibit CYP2C8 and CYP2C9, as well as CYP2C19 and CYP2A6 (Zhu et al., 2009, Prueksaritanont et al., 2005). 236

Attempts to obtain fibrate-bound structures of the DM heme domain by X-ray crystallography were unsuccessful, including many trials with bezafibrate-, clofibrate- and fenofibrate-bound DM heme domains. A typical diffraction limit of ~7 Angstroms was typical in these cases. To further investigate the structural properties of the DM heme domain and to identify any underlying reasons that might explain the issues with protein crystallization, studies using native MS, CIU and HDX-MS were undertaken. During CIU experiments, bezafibrate binding appeared to stabilize the unfolding intermediates of the

DM heme domain when compared to the substrate-free, solvent control and NPG-bound spectra. Typically, motions within a protein cause a region to become more solvent accessible/deshielded (orange to red) and another region of the protein to become more shielded (cyan to dark blue) for deuterium uptake during HDX experiments (Figure 30 and

Figure 31). Surprisingly, bezafibrate binding induces slight deshielding across many areas of the protein, but little or no corresponding shielding. This may suggest that protein unfolding or an event causing the protein to become less tightly packed occurs with bezafibrate. However, this is not corroborated by the CCS values. Interestingly, a 1:1 mixture of NPG and bezafibrate resulted in a mixture of the two sets of conformational changes observed with NPG and bezafibrate in isolation. NPG has a Kd value approximately a 100-fold lower than that for bezafibrate, and so should out-compete bezafibrate for binding to the DM heme domain active site. This would cause substantial depletion of bezafibrate signals in HDX-MS and CIU experiments. A potential explanation is that bezafibrate binds to the outside or periphery of the protein, possibly occupying an allosteric binding site. This could explain the lack of products observed for bezafibrate and potentially for other fibrate compounds, such as ciprofibrate and clinofibrate. However, as stated previously, no shielding was observed in the bezafibrate-bound structure that might be indicative of a ligand binding to an allosteric site. In contrast, for example, shielding is observed for the human cholesterol 24-hydroxylase CYP46A1 on efavirenz binding 237

(Anderson et al., 2016). Additionally, the CCS values calculated showed similar protein dimensions for substrate-free, NPG-bound and bezafibrate-bound DM heme domains, suggesting no non-specific binding on the surface of the protein. Furthermore, bezafibrate induces substantial high-spin Soret peak shifts, suggesting that this drug interacts closely with the heme prosthetic group within the active site.

The A82F/F87V DM heme domain “gatekeeper” variant has not been crystallized in a ligand-free state. In absence of a ligand in the active site, it appears likely that the DM protein is conformationally flexible and does not readily crystallize. This is mainly due to the A82F mutation which destabilizes the substrate-free conformation and causes the protein to adopt a catalytically primed state. Crystal structures of the A82F mutant show how this protein adopts the substrate-bound conformation when no ligand is present in the active site (Huang et al., 2007, Butler et al., 2013). The BM3 F87V mutant in isolation was shown to convert arachidonic acid into (14S,15R)-epoxyeicosatrienoic acid (Graham-

Lorence et al., 1997) and F87 mutations appear frequently in various other P450 BM3 mutants, where they create additional space for substrate binding near to the heme. F87 mutations also appear regularly in other BM3 heme domain crystal structures and are used for e.g. in silico studies involving molecular dynamics (Capoferri et al., 2016, Geronimo et al., 2016).

In our previous studies, conformational changes in P450 BM3 were investigated using HDX with the WT P450 BM3 enzyme using the tight-binding NPG substrate

(Chapter 2). These data showed that the WT BM3 undergoes widespread conformational changes on NPG binding, as revealed by a number of shielding and deshielding events observed on the protein. In comparison, during investigations of the binding mode of bezafibrate to the A82F/F87V DM variant described herein, it was observed that the DM variant undergoes remarkably little change in deuterium uptake, in turn suggesting that 238 little conformational change occurs. This finding supports evidence that the DM heme domain remains predominantly in a catalytically primed state in which the P450 occupies a substrate-bound type conformation, even when in a substrate-free state (Huang et al.,

2007). Interestingly, the CCS values for the substrate-free and substrate-bound DM heme domain forms are slightly larger than those of the WT heme domain values (Chapter 2). In addition, the voltage required to initiate unfolding of the protein was essentially the same for substrate-bound WT and all DM heme domain samples.

6.6. Conclusions

In conclusion, the binding modes of various members of the fibrate class of drugs to the WT and DM forms of the BM3 heme domain were probed using a number of spectroscopic and MS-based techniques. Bezafibrate exhibited tight binding to the DM heme domain variant and exhibited a large (~87%) high-spin Soret peak shift. However, no obvious products were produced during turnover experiments. This compound and other fibrate drugs also failed to produce diffraction-quality crystals for structural analysis, despite some members of this group exhibiting high affinity for the DM heme domain.

Analysis of the conformational changes induced by bezafibrate upon ligand binding point to an alternative binding mode to that for the modified fatty acid N-palmitoyl glycine

(NPG). Distinctive sets of conformational changes were observed using during HDX-MS studies of the DM heme domain bound to an equimolar mixture of NPG and bezafibrate, and despite NPG having a much higher binding affinity than does bezafibrate. In addition, the binding of bezafibrate caused the stabilization of all unfolding intermediates observed during CIU compared to the ligand-free and the NPG-bound protein. These findings suggests that bezafibrate could occupy a different binding site to that for the NPG, possibly an allosteric binding site separated from the main active site region. The collision cross- section (CCS) calculated from ion-mobility mass spectrometry (IM-MS) revealed that was 239 no major change in the size of the bezafibrate-bound protein compared to the NPG-bound protein (or for the solvent control or the substrate-free protein form). While the binding of bezafibrate to an allosteric site may not appear consistent with the substantial Soret peak high-spin shift exhibited by this drug when delivered in isolation, an explanation could be that the NPG fully occupies the DM heme domain active site (consistent with its higher affinity), but that bezafibrate then occupies an allosteric site or a position closer to the mouth of the active site. A scenario of this type might explain how signals from both drugs can be observed despite a large difference in the Kd values for NPG and bezafibrate.

6.7. References

ACEVEDO-ROCHA, C. G., GAMBLE, C. G., LONSDALE, R., LI, A. T., NETT, N., HOEBENREICH, S., LINGNAU, J. B., WIRTZ, C., FARES, C., HINRICHS, H., DEEGE, A., MULHOLLAND, A. J., NOV, Y., LEYS, D., MCLEAN, K. J., MUNRO, A. W. & REETZ, M. T. 2018. P450-Catalyzed Regio- and Diastereoselective Steroid Hydroxylation: Efficient Directed Evolution Enabled by Mutability Landscaping. ACS Catal, 8, 3395-3410. ANDERSON, K. W., MAST, N., HUDGENS, J. W., LIN, J. B., TURKO, I. V. & PIKULEVA, I. A. 2016. Mapping of the Allosteric Site in Cholesterol Hydroxylase CYP46A1 for Efavirenz, a Drug That Stimulates Enzyme Activity. J Biol Chem, 291, 11876-86. BACKES, J. M., GIBSON, C. A., RUISINGER, J. F. & MORIARTY, P. M. 2007. Fibrates: what have we learned in the past 40 years? Pharmacotherapy, 27, 412-24. BELCHER, J., MCLEAN, K. J., MATTHEWS, S., WOODWARD, L. S., FISHER, K., RIGBY, S. E. J., NELSON, D. R., POTTS, D., BAYNHAM, M. T., PARKER, D. A., LEYS, D. & MUNRO, A. W. 2014. Structure and Biochemical Properties of the Alkene Producing Cytochrome P450 OleTJE (CYP152L1) from the Jeotgalicoccus sp 8456 Bacterium. J Biol Chem, 289, 6535-6550. BUTLER, C. F., PEET, C., MASON, A. E., VOICE, M. W., LEYS, D. & MUNRO, A. W. 2013. Key mutations alter the cytochrome P450 BM3 conformational landscape and remove inherent substrate bias. J Biol Chem, 288, 25387-99. BUTLER, C. F., PEET, C., MCLEAN, K. J., BAYNHAM, M. T., BLANKLEY, R. T., FISHER, K., RIGBY, S. E., LEYS, D., VOICE, M. W. & MUNRO, A. W. 2014. Human P450-like oxidation of diverse proton pump inhibitor drugs by 'gatekeeper' mutants of flavocytochrome P450 BM3. Biochem J, 460, 247-59. CAPOFERRI, L., LETH, R., TER HAAR, E., MOHANTY, A. K., GROOTENHUIS, P. D., VOTTERO, E., COMMANDEUR, J. N., VERMEULEN, N. P., JORGENSEN, F. S., OLSEN, L. & GEERKE, D. P. 2016. Insights into regioselective metabolism of mefenamic acid by cytochrome P450 BM3 mutants through crystallography, docking, molecular dynamics, and free energy calculations. Proteins, 84, 383-96. CHAPMAN, M. J. 1987. Pharmacology of fenofibrate. Am J Med, 83, 21-5.

240

CIZKOVA, K., KONIECZNA, A., ERDOSOVA, B., LICHNOVSKA, R. & EHRMANN, J. 2012. Peroxisome proliferator-activated receptors in regulation of cytochromes P450: new way to overcome multidrug resistance? J Biomed Biotechnol, 2012, 656428. CREGER, P. L., MOERSCH, G. W. & NEUKLIS, W. A. 1976. Structure/activity relationship of gemfibrozil (CI-719) and related compounds. Proc R Soc Med, 69 Suppl 2, 3-5. DANIELSON, P. B. 2002. The cytochrome P450 superfamily: biochemistry, evolution and drug metabolism in humans. Curr Drug Metab, 3, 561-97. DEMBOWSKI, E. & DAVIDSON, M. H. 2008. Statin-Fibrate Combination For The Treatment Of Dyslipidaemia. Eur Cardiol, 4, 10-11. ELSOM, L. F., HAWKINS, D. R. & CHASSEAUD, L. F. 1976. Identification of a major metabolite of the new hypolipidaemic agent, isopropyl 2-(4'(p-chlorobenzoyl) phenoxy)-2-methylpropionate (procetofene) in humans by gas chromatography- mass spectrometry. J Chromatogr, 123, 463-7. ENGLISH, N., HUGHES, V. & WOLF, C. R. 1994. Common pathways of cytochrome P450 gene regulation by peroxisome proliferators and barbiturates in Bacillus megaterium ATCC14581. J Biol Chem, 269, 26836-41. ENGLISH, N., PALMER, C. N., ALWORTH, W. L., KANG, L., HUGHES, V. & WOLF, C. R. 1997. Fatty acid signals in Bacillus megaterium are attenuated by cytochrome P-450-mediated hydroxylation. Biochem J, 327 ( Pt 2), 363-8. FAN, L. Q., BROWN-BORG, H., BROWN, S., WESTIN, S., MODE, A. & CORTON, J. C. 2004. PPARalpha activators down-regulate CYP2C7, a retinoic acid and testosterone hydroxylase. Toxicology, 203, 41-8. GERONIMO, I., DENNING, C. A., ROGERS, W. E., OTHMAN, T., HUXFORD, T., HEIDARY, D. K., GLAZER, E. C. & PAYNE, C. M. 2016. Effect of Mutation and Substrate Binding on the Stability of Cytochrome P450BM3 Variants. Biochemistry, 55, 3594-606. GRAHAM-LORENCE, S., TRUAN, G., PETERSON, J. A., FALCK, J. R., WEI, S., HELVIG, C. & CAPDEVILA, J. H. 1997. An active site substitution, F87V, converts cytochrome P450 BM-3 into a regio- and stereoselective (14S,15R)- arachidonic acid epoxygenase. J Biol Chem, 272, 1127-35. GRAHAMLORENCE, S., TRUAN, G., PETERSON, J. A., FALCK, J. R., WEI, S. Z., HELVIG, C. & CAPDEVILA, J. H. 1997. An active site substitution, F87V, converts cytochrome P450 BM-3 into a regio- and stereoselective (14S,15R)- arachidonic acid epoxygenase. J Biol Chem, 272, 1127-1135. GUENGERICH, F. P. 2003. Cytochromes P450, drugs, and diseases. Mol Interv, 3, 194- 204. HARE, R. S. & FULCO, A. J. 1975. Carbon monoxide and hydroxymercuribenzoate sensitivity of a fatty acid (omega-2) hydroxylase from Bacillus megaterium. Biochem Biophys Res Commun, 65, 665-72. HILL, A. V. 1910. The possible effects of the aggregation of the molecules of haemoglobin on its dissociation curves. J Physiol-London, 4, 4-7. HO, P. P. & FULCO, A. J. 1976. Involvement of a single hydroxylase species in the hydroxylation of palmitate at the omega-1, omega-2 and omega-3 positions by a preparation from Bacillus megaterium. Biochim Biophys Acta, 431, 249-56. HUANG, W. C., WESTLAKE, A. C., MARECHAL, J. D., JOYCE, M. G., MOODY, P. C. & ROBERTS, G. C. 2007. Filling a hole in cytochrome P450 BM3 improves substrate binding and catalytic efficiency. J Mol Biol, 373, 633-51. INCE, I., KNIBBE, C. A., DANHOF, M. & DE WILDT, S. N. 2013. Developmental changes in the expression and function of cytochrome P450 3A isoforms: evidence from in vitro and in vivo investigations. Clin Pharmacokinet, 52, 333-45.

241

JACKEVICIUS, C. A., TU, J. V., ROSS, J. S., KO, D. T., CARREON, D. & KRUMHOLZ, H. M. 2011. Use of fibrates in the United States and Canada. JAMA, 305, 1217-24. KAJOSAARI, L. I., BACKMAN, J. T., NEUVONEN, M., LAITILA, J. & NEUVONEN, P. J. 2004. Lack of effect of bezafibrate and fenofibrate on the pharmacokinetics and pharmacodynamics of repaglinide. Br J Clin Pharmacol, 58, 390-6. LAKE, B. G., TREDGER, J. M., GRAY, T. J., STUBBERFIELD, C. R., HODDER, K. D., GANGOLLI, S. D. & WILLIAMS, R. 1984. The effect of peroxisome proliferators on the metabolism and spectral interaction of endogenous substrates of cytochrome P-450 in rat hepatic microsomes. Life Sci, 35, 2621-6. LALLOYER, F. & STAELS, B. 2010. Fibrates, glitazones, and peroxisome proliferator- activated receptors. Arterioscler Thromb Vasc Biol, 30, 894-9. MAKITA, K., FALCK, J. R. & CAPDEVILA, J. H. 1996. Cytochrome P450, the arachidonic acid cascade, and hypertension: new vistas for an old enzyme system. FASEB J, 10, 1456-63. MATSON, R. S., HARE, R. S. & FULCO, A. J. 1977. Characteristics of a cytochrome P- 450-dependent fatty acid omega-2 hydroxylase from Bacillus megaterium. Biochim Biophys Acta, 487, 487-94. MICHAELIS, L. & MENTEN, M. L. 1913. The kenetics of the inversion effect. Biochem Z, 49, 333-369. MIGAS, L. G., FRANCE, A. P., BELLINA, B. & BARRANN, P. E. 2018. ORIGAMI: A software suite for activated ion mobility mass spectrometry (aIM-MS) applied to multimeric protein assemblies. Int J Mass Spectrom, 427, 20-28. MILLER, D. B. & SPENCE, J. D. 1998. Clinical pharmacokinetics of fibric acid derivatives (fibrates). Clin Pharmacokinet, 34, 155-62. MIURA, Y. & FULCO, A. J. 1974. (Omega -2) hydroxylation of fatty acids by a soluble system from Bacillus megaterium. J Biol Chem, 249, 1880-8. MIURA, Y. & FULCO, A. J. 1975. Omega-1, Omega-2 and Omega-3 hydroxylation of long-chain fatty acids, amides and alcohols by a soluble enzyme system from Bacillus megaterium. Biochim Biophys Acta, 388, 305-17. MORRISON, J. F. 1969. Kinetics of the reversible inhibition of enzyme-catalysed reactions by tight-binding inhibitors. Biochim Biophys Acta, 185, 269-86. MURAI, T., IWABUCHI, H. & IKEDA, T. 2004. Identification of gemfibrozil metabolites, produced as positional isomers in human liver microsomes, by on-line analyzes using liquid chromatography/mass spectrometry and liquid chromatography/nuclear magnetic resonance spectroscopy. J Mass Spectrom Soc Jpn, 52, 277-283. NARHI, L. O. & FULCO, A. J. 1986. Characterization of a catalytically self-sufficient 119,000-dalton cytochrome P-450 monooxygenase induced by barbiturates in Bacillus megaterium. J Biol Chem, 261, 7160-9. NARHI, L. O. & FULCO, A. J. 1987. Identification and characterization of two functional domains in cytochrome P-450BM-3, a catalytically self-sufficient monooxygenase induced by barbiturates in Bacillus megaterium. J Biol Chem, 262, 6683-90. NOBLE, M. A., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MACKAY, A. C., REID, G. A., HANZLIK, R. P. & MUNRO, A. W. 1999. Roles of key active-site residues in flavocytochrome P450 BM3. Biochem J, 339 ( Pt 2), 371-9. OLIVER, M. 2012. The clofibrate saga: a retrospective commentary. Br J Clin Pharmacol, 74, 907-10. POZZI, A., IBANEZ, M. R., GATICA, A. E., YANG, S., WEI, S., MEI, S., FALCK, J. R. & CAPDEVILA, J. H. 2007. Peroxisomal proliferator-activated receptor-alpha- dependent inhibition of endothelial cell proliferation and tumorigenesis. J Biol Chem, 282, 17685-95.

242

POZZI, A., POPESCU, V., YANG, S., MEI, S., SHI, M., PUOLITAIVAL, S. M., CAPRIOLI, R. M. & CAPDEVILA, J. H. 2010. The anti-tumorigenic properties of peroxisomal proliferator-activated receptor alpha are arachidonic acid epoxygenase-mediated. J Biol Chem, 285, 12840-50. PRUEKSARITANONT, T., RICHARDS, K. M., QIU, Y., STRONG-BASALYGA, K., MILLER, A., LI, C., EISENHANDLER, R. & CARLINI, E. J. 2005. Comparative effects of fibrates on drug metabolizing enzymes in human hepatocytes. Pharm Res, 22, 71-8. PRUEKSARITANONT, T., ZHAO, J. J., MA, B., ROADCAP, B. A., TANG, C., QIU, Y., LIU, L., LIN, J. H., PEARSON, P. G. & BAILLIE, T. A. 2002. Mechanistic studies on metabolic interactions between gemfibrozil and statins. J Pharmacol Exp Ther, 301, 1042-51. REN, X., YORKE, J. A., TAYLOR, E., ZHANG, T., ZHOU, W. & WONG, L. L. 2015. Drug Oxidation by Cytochrome P450BM3 : Metabolite Synthesis and Discovering New P450 Reaction Types. Chemistry, 21, 15039-47. RODNEY, G., UHLENDORF, P. & MAXWELL, R. E. 1976. The hypolipidaemic effect of gemfibrozil (CI-719) in laboratory animals. Proc R Soc Med, 69 Suppl 2, 6-10. SPECTOR, A. A. 2009. Arachidonic acid cytochrome P450 epoxygenase pathway. J Lipid Res, 50 Suppl, S52-6. TORNIO, A., NEUVONEN, P. J., NIEMI, M. & BACKMAN, J. T. 2017. Role of gemfibrozil as an inhibitor of CYP2C8 and membrane transporters. Expert Opin Drug Metab Toxicol, 13, 83-95. TRAN, N. H., HUYNH, N., CHAVEZ, G., NGUYEN, A., DWARAKNATH, S., NGUYEN, T. A., NGUYEN, M. & CHERUZEL, L. 2012. A series of hybrid P450 BM3 enzymes with different catalytic activity in the light-initiated hydroxylation of lauric acid. J Inorg Biochem, 115, 50-6. VECERA, R., ZACHAROVA, A., OROLIN, J., STROJIL, J., SKOTTOVA, N. & ANZENBACHER, P. 2011. Fenofibrate-induced decrease of expression of CYP2C11 and CYP2C6 in rat. Biopharm Drug Dispos, 32, 482-7. WEN, X., WANG, J. S., BACKMAN, J. T., KIVISTO, K. T. & NEUVONEN, P. J. 2001. Gemfibrozil is a potent inhibitor of human cytochrome P450 2C9. Drug Metab Dispos, 29, 1359-61. ZHANG, K., SHAFER, B. M., DEMARS, M. D., 2ND, STERN, H. A. & FASAN, R. 2012. Controlled oxidation of remote sp3 C-H bonds in artemisinin via P450 catalysts with fine-tuned regio- and stereoselectivity. J Am Chem Soc, 134, 18695- 704. ZHU, T., AWNI, W. M., HOSMANE, B., KELLY, M. T., SLEEP, D. J., STOLZENBACH, J. C., WAN, K., CHIRA, T. O. & PRADHAN, R. S. 2009. ABT-335, the choline salt of fenofibric acid, does not have a clinically significant pharmacokinetic interaction with rosuvastatin in humans. J Clin Pharmacol, 49, 63-71.

243

6.8. Supporting Information

Rhombic LS signals (gz/gy/gx) Compound Major species Minor species

DM Heme 2.43/2.25/1.91 2.40/2.25/1.92; 2.51/2.25/1.89

DM Heme + DMSO 2.45/2.25/1.91 2.51/2.25/1.89

DM Heme + Bezafibrate 2.45/2.25/1.91 None

DM Heme + Ciprofibrate 2.45/2.25/1.91 2.51/2.25/1.89

DM Heme + Clinofibrate 2.45/2.25/1.91 2.51/2.25/1.89

DM Heme + Clofibrate 2.45/2.25/1.91 2.51/2.25/1.89

DM Heme + Fenofibrate 2.45/2.25/1.91 2.51/2.25/1.89

DM Heme + Gemfibrozil 2.44/2.25/1.91 2.47/2.25/1.90

Figure S18: X-band EPR data sets for the DM BM3 heme domain in complex with fibrate compounds The binding of fibrates causes the protein signal to become more homogeneous. All fibrates show the same pattern in the EPR spectra which is indifferent to the solvent control. Gemfibrozil shows some changes in the EPR spectrum with slightly shifted g values compared to the solvent control.

244

Figure S19: 1H-NMR of fenofibrate shows the production of three metabolites By measuring the interaction of proton atoms three distinct products can be observed. The top panel shows a control reaction with fenofibrate labelled with IUPAC numbering. The bottom panel shows the three distinct metabolites with the modified atoms labelled in the COSY, based on the IUPAC numbering of the parent compound. Data were collected on a 500 MHz NMR in DMSO-d6.

245

Figure S20: 1H-NMR of gemfibrozil shows the production of two human metabolites By measuring the interaction of proton atoms two distinct hydroxylated products can be observed. The top panel shows a control reaction with gemfibrozil labelled with IUPAC numbering. The bottom panel shows the two distinct metabolites with the modified atoms labelled in the COSY, based on the IUPAC numbering of the parent compound. Data were collected on a 500 MHz NMR in DMSO-d6.

246

+ NPG and Charge state Substrate free + EtOH + NPG + Bezafibrate Bezafibrate

+15 39.38 39.38 39.38 39.38 39.38

+14 38.11 38.40 38.11 38.11 38.11

+13 37.11 37.88 37.38 37.38 37.38

Average 38.20 38.55 38.29 38.29 38.29

Standard 1.14 0.76 1.01 1.01 1.01 deviation

Figure S21: Collision cross-section (CCS) values of the DM variant bound a variety of ligands CCS values show no difference between substrate-free, solvent controls or ligand- bound samples.

247

Chapter 7: Conclusions and Future Work

7.1. Summary

P450 BM3 has been examined for over fifty years due to its high catalytic rate and interesting structure (Miura and Fulco 1974, Narhi and Fulco 1986, Narhi and Fulco 1987,

Noble et al., 1999). In addition, mutagenesis studies have allowed the observation of many different types of reaction catalyzed by P450 BM3 and the production of many useful compounds (e.g. terpenes, hydroxylated alkanes and modified steroids). Despite successful structural determination of its component domains, the structures of the intact P450 BM3 and of its reductase domain have not been solved, and the interdomain interactions that occur within the P450 BM3 dimer are still not fully elucidated. Recent work in our laboratory has focused on the production of human metabolites to give further evidence that the DM variant containing the gatekeeper mutations (A82F/F87V) can be used as a model for human P450 enzymes. This would be of great importance to industry as P450

BM3 has an easier expression and purification protocol, higher expression yield, increased stability and much higher catalytic rates than the human P450 enzymes. In addition, this protein is soluble and requires no partner enzymes as it is catalytically self-sufficient.

We have probed the WT enzyme for further information about the structure of the full-length (intact) enzyme, the interaction between the domains and the nature of the dimeric interface. Numerous ligand-free and ligand-bound structures of the P450 BM3 heme domain have been solved. Unfortunately, likely due to the inherent flexibility of the

CPR domain, there is currently no structure of the entire CPR domain. However, there are two FAD/NADPH-binding domain structures, both containing mutations to surface cysteine to prevent dimerization through disulfide bonds (Joyce et al., 2012). However, there is only one P450 BM3 FMN-binding domain structure, which was obtained from a crystal structure of the fused P450 and FMN domains of P450 BM3. In this structure, the 248

FMN and heme domains are cleaved apart and there is a non-stoichiometric ratio of domains in the crystal structure with two heme domains and one FMN-binding domain

(Sevrioukova et al., 1999). The FMN-binding domain is also located in a position in the crystal lattice that is too distant from both the heme domains for efficient electron transfer

(Munro et al., 2002). To investigate the P450 BM3 dimeric interface and its full-length structure, native MS, CIU and HDX-MS methods were utilized. By determining the collision cross-section values from experimental data and by using available crystallographic data, it was demonstrated that the CPR domain of P450 BM3 is more compact than is the rat CPR, a structure used frequently for CPR modelling. Through

HDX-MS studies, a number of important regions in the P450 BM3 structure were identified. In the heme domain, shielding was observed in the A-, F-, I-, J- and J’-helices, with different degrees of shielding, whilst the area surrounding Glu424 shows great deshielding when comparing the intact protein to the heme domain. Within the CPR domain substantial shielding and deshielding is observed in both the hinge region of the

FAD-binding domain and the FMN-binding domain. By probing the binding of ligands to the intact P450 BM3 protein, small deshielding and shielding events were seen to occur in the heme domain upon NPG binding. Shielding that is potentially important for catalytic activity is observed near the N-terminus. In the CPR domain, NADP+ binding caused surprisingly little change to the protein conformation. The greatest change was again observed in the hinge region of the FAD-binding domain. This area was shown to be very important for reductase domain dimerization by X-ray crystallographic studies and electron microscopy. The FMN-binding domain is highly dynamic and does not show areas of great shielding indicative of heme domain interactions.

HDX-MS was further used to probe a class of drugs (the fibrates) which exhibited the highest percentage high-spin shift on binding to the P450 BM3 DM heme domain in studies using an FDA-approved compound library screen. In these studies, the binding 249 mode of bezafibrate to the DM heme domain was probed using HDX-MS in view of difficulties encountered in producing good diffraction quality crystals of a DM heme domain-bezafibrate complex. The binding of bezafibrate caused a great deshielding event to occur across the protein, and particularly in the beta-sheets of the heme domain. This was surprising as a number of the alpha helices are considered to be the more dynamic areas of P450 enzymes. In addition, no corresponding shielding was observed in the beta sheets or other regions of the protein usually indicative of movement. In contrast, the CCS values did not suggest that the protein unfolded or became less compact on bezafibrate binding. Intriguingly, bezafibrate stabilized the unfolding intermediates observed during

CIU. For comparison, the binding of the modified fatty acid NPG to the DM heme domain was also probed, and showed a more typical pattern of shielding and deshielding upon ligand binding. However, these changes were somewhat muted when compared to the effects induced in the WT heme domain on NPG binding, as the gatekeeper mutations within the DM variant cause the protein to adopt a catalytically primed form, even when in a substrate-free state (Huang et al., 2007). The combined additions of NPG and bezafibrate to the DM heme domain showed a mixture of the conformational changes induced by the two compounds in isolation. This was surprising as NPG has a Kd value one hundred times lower than bezafibrate and so should out-compete bezafibrate and prevent bezafibrate- induced conformational changes. The ability of bezafibrate to induce conformational changes in the DM enzyme may involve its interactions on the protein surface, rather than in the active site. Bezafibrate, clofibrate, clinofibrate and ciprofibrate failed to produce products, whereas turnover was successful for fenofibrate and gemfibrozil. The DM variant produced two human metabolites from gemfibrozil.

As mentioned previously, almost 1,000 pharmaceutical compounds were screened for binding to the DM BM3 variant. 59% of compounds caused a significant shift to the heme UV-Vis spectrum. These included a variety of compounds with a range of masses, 250 logP values and targets. From these compounds, the fibrate, antidiabetic and azole classes of drugs were identified for further investigation, including structure elucidation by X-ray crystallography, along with EPR, UV-Vis spectrophotometry, HPLC, LCMS, LCMS/MS and NMR for further characterization of e.g. ligand binding affinity and product analysis.

Binding affinities for the compounds mentioned above revealed their tight binding to the

DM variant and, in the rare cases when these compounds were found to bind to the WT enzyme, the Kd values determined were considerably lower for the DM variant.

The antidiabetic drugs investigated were from six different compound types, with around half of these compounds able to bind to the DM variant. The only substrate to exhibit high-spin species within the EPR spectrum for the antidiabetic class of compounds, and indeed for any of the compounds investigated within this thesis, was the toxic human drug troglitazone. A ligand-bound crystal structure was solved for the DM variant in complex with troglitazone. This compound binds tightly in the active site and makes a number of interactions with very short distances to several key residues in the active site and approaches closely to the heme prosthetic group. Interestingly, the toxic human metabolite of troglitazone could be produced during turnover experiments. Known human metabolites for pioglitazone and repaglinide were also produced during turnover reactions with the DM variant.

Many compounds from the azole class of inhibitors were shown to induce changes to the g-values in the EPR spectrum for the WT and DM enzyme. This resulted from the coordination of the heme iron through imidazoles or triazole groups on the azoles, leading to a “widening” of the X-band EPR spectrum with gz and gx values moving to higher and lower g-values, respectively. Interestingly, fluconazole instead induced a “narrowing” of the g-values when in complex with the DM heme domain. Analysis of the fluconazole- bound structure determined using X-ray crystallography showed that one monomer of the 251 asymmetric unit exhibits a conformational change to the closed conformation, as is observed for the substrate-bound form of the protein. Three other structures were determined, allowing the analysis of the binding modes for imidazole and triazole compounds. Voriconazole was shown to ligate to the heme prosthetic group through a pyrimidine nitrogen from its 5-fluoropyrimidine ring, displaying a novel type of inhibitor ligation for P450 BM3 and potentially for other P450 enzymes.

7.2. Conclusions

To fulfill the aims of this thesis, a large number of pharmaceutical compounds were screened for their binding to the DM BM3 variant previously produced within our laboratory. The binding of these compounds was investigated through a variety of methods. In particular, novel binding modes were observed for the compounds voriconazole (exhibiting ligation to the heme prosthetic group through a novel interaction) and fluconazole (exhibiting an unusual conformational change in its crystal structure). In addition, novel conformational changes were observed for the binding of bezafibrate to

P450 BM3 using HDX-MS, which revealed that this compound causes deshielding of several regions across the entire protein. Different conformational changes were observed upon the binding of the modified fatty acid NPG. A number of high-value oxy- pharmaceuticals were obtained by producing human metabolites from the drugs pioglitazone, troglitazone, repaglinide and gemfibrozil. Work presented in this thesis also probed the structures and conformational changes of the full-length protein during domain interactions and ligand binding using HDX-MS methods. Through this work, the site at which the CPR domains of P450 BM3 interact to cause dimerization of the enzyme was further confirmed. In addition, a number of other sites were identified which may correspond to sites of interaction between the heme- and FMN-binding domains within the

P450 BM3 dimer. 252

7.3. Future Work

Many compounds from the library screen were not investigated due to time constraints, and their further analysis may lead to other interesting developments. For example, the prodrugs tegafur and carmofur which are metabolized by human P450 enzymes to the anticancer agent 5-fluorouracil. These compounds bind effectively to both the WT and DM BM3 heme domains. Turnover experiments with these P450 BM3 enzymes could replicate the human P450 catalytic reaction. Many other compounds in the screen could also generate valuable human metabolites and enable further insights into

P450 BM3 binding modes through crystallographic analysis.

Kinetic studies using compounds identified in this research programme were not investigated in detail. Such work could be done for selected compounds with the WT and

DM BM3 enzymes in order to identify BM3 WT/DM substrate complexes that result in high catalytic activity. Further work could also be done to optimize turnover conditions for these BM3 enzymes in order to produce different metabolites, or to increase the substrate turnover rate for improved production of characterized metabolites. In other work, soluble cytochrome b5 was engineered with a His-tag for housefly, rat and human isoforms of the protein. All three of these proteins were successfully expressed and purified. In addition, extinction coefficient values were determined for the proteins using the pyridine hemochromogen assay technique. Unfortunately, no further work was done with these proteins due to time constraints. These proteins were produced in the hope of the production of new metabolites from P450 BM3 due to the different redox potential between the cytochrome b5 enzymes and the CPR domain of P450 BM3. This method was shown by Estrada et al. to produce new metabolites from CYP17A1 (Estrada et al., 2014).

In addition, previous studies using the purified housefly b5 enzyme (with a different

253 construct) shown that this cytochrome b5 could successfully bind and donate electrons to the P450 BM3 heme domain (Munro et al, 1998, Noble et al., 2007).

The binding of bezafibrate and the conformational changes it induces are still somewhat of a mystery, as some techniques suggest an allosteric binding site, yet this compound can cause Soret peak shifts suggesting that binding may occur in the active site, and no changes in the cross-section of the protein were observed. Further structural studies may help determine the cause of these conformational changes, such as circular dichroism, resonance Raman spectroscopy or other spectroscopic methods. However, the greatest achievement would be to solve a structure of the intact P450 BM3 enzyme using X-ray crystallography. In addition, further HDX-MS analysis should be conducted as bezafibrate is also able to bind to the WT BM3, and further studies with this enzyme could identify conformational changes specific for the WT in comparison to the DM BM3 variant. As mentioned above, no attempt was made to determine reaction kinetics for any of the library compounds tested. In addition, transient kinetic studies (i.e. determination of koff/koff values using stopped-flow methods) for library compounds were also not investigated, and data from these experiments may shed further light on the lack of turnover from fibrate class substrates.

There was some debate during the project as to how to investigate further the shielded sites identified for the WT enzyme during HDX-MS experiments. One method suggested was the introduction of non-natural amino acids to allow for click chemistry to prevent dimerization using steric bulk rather than through mutations or cross-linking.

Although the facilities and expertise exist in the MIB building, this method was not used in view of other research priorities for the thesis. In addition, a single residue in the BM3 heme domain (Thr149) showed increased solvent accessibility on NADP+ binding consistently through many WT BM3 HDX-MS experiments. Interestingly, this residue also 254 exhibited strange deshielding in the DM BM3 variant, and in these instances also affected residues surrounding it in the E-helix (146-149). This residue has not been mutated before in P450 BM3 and so no obvious reason could be determined for these changes in deuterium uptake. Future work could involve mutating this residue to determine its role within the protein, which from HDX-MS may suggest a crucial function in the electron transport chain.

7.4. References

ESTRADA, D. F., LAURENCE, J. S. & SCOTT, E. E. 2013. Substrate-modulated Cytochrome P450 17A1 and Cytochrome b7(5) Interactions Revealed by NMR. J Biol Chem, 288, 17008-17018. HUANG, W. C., WESTLAKE, A. C., MARECHAL, J. D., JOYCE, M. G., MOODY, P. C. & ROBERTS, G. C. 2007. Filling a hole in cytochrome P450 BM3 improves substrate binding and catalytic efficiency. J Mol Biol, 373, 633-51. JOYCE, M. G., EKANEM, I. S., ROITEL, O., DUNFORD, A. J., NEELI, R., GIRVAN, H. M., BAKER, G. J., CURTIS, R. A., MUNRO, A. W. & LEYS, D. 2012. The crystal structure of the FAD/NADPH-binding domain of flavocytochrome P450 BM3. FEBS J, 279, 1694-1706. MIURA, Y. & FULCO, A. J. 1974. (Omega -2) hydroxylation of fatty acids by a soluble system from Bacillus megaterium. J Biol Chem, 249, 1880-8. MIURA, Y. & FULCO, A. J. 1975. Omega-1, Omega-2 and Omega-3 hydroxylation of long-chain fatty acids, amides and alcohols by a soluble enzyme system from Bacillus megaterium. Biochim Biophys Acta, 388, 305-17. MUNRO, A. W., LEYS, D. G., MCLEAN, K. J., MARSHALL, K. R., OST, T. W., DAFF, S., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MOSER, C. C., PAGE, C. C. & DUTTON, P. L. 2002. P450 BM3: the very model of a modern flavocytochrome. Trends Biochem Sci, 27, 250-7. MUNRO, A. W., NOBLE, M. A., MILES, C. S., DAFF, S. N., GREEN, A. J., QUARONI, L., RIVERS, S., OST, T. W., REID, G. A. & CHAPMAN, S. K. 1999. Flavocytochrome P-450 BM3: a paradigm for the analysis of electron transfer and its control in the P-450s. Biochem Soc Trans, 27, 190-6. NARHI, L. O. & FULCO, A. J. 1986. Characterization of a catalytically self-sufficient 119,000-dalton cytochrome P-450 monooxygenase induced by barbiturates in Bacillus megaterium. J Biol Chem, 261, 7160-9. NARHI, L. O. & FULCO, A. J. 1987. Identification and characterization of two functional domains in cytochrome P-450BM-3, a catalytically self-sufficient monooxygenase induced by barbiturates in Bacillus megaterium. J Biol Chem, 262, 6683-90. NOBLE, M. A., MILES, C. S., CHAPMAN, S. K., LYSEK, D. A., MACKAY, A. C., REID, G. A., HANZLIK, R. P. & MUNRO, A. W. 1999. Roles of key active-site residues in flavocytochrome P450 BM3. Biochem J, 339 ( Pt 2), 371-9. NOBLE, M. A., GIRVAN, H. M., SMITH, S. J., SMITH, W. E., MURATALIEV, M., GUZOV, V. M., FEYEREISEN, R. & MUNRO, A. W. 2007. Analysis of the

255

interactions of cytochrome b5 with flavocytochrome P450 BM3 and its domains. Drug Metab Rev, 39, 599-617. SEVRIOUKOVA, I. F., LI, H., ZHANG, H., PETERSON, J. A. & POULOS, T. L. 1999. Structure of a cytochrome P450-redox partner electron-transfer complex. Proc Natl Acad Sci U S A, 96, 1863-8.

256