Structural Investigation of Histidine Domain Protein Tyrosine Phosphatase and Its Interactions with Endosomal Sorting Complexes Required for Transport

Structural Investigation of Histidine Domain Protein Tyrosine Phosphatase and its Interactions with Endosomal Sorting Complexes Required for Transport

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

2017

Graham J Heaven

School of Chemistry

List of contents

List of tables and figures ...... 5 Abbreviations ...... 7 Abstract ...... 10 Declaration ...... 12 Copyright statement ...... 13 Acknowledgements ...... 14 Chapter 1: Introduction ...... 15 1.1 Membrane budding reactions in cell biology ...... 15 1.2 Endosomal trafficking of receptor proteins by ESCRTs...... 17 1.3 Other roles of ESCRTs ...... 19 1.4 Molecular architecture of the ESCRT machinery ...... 21 1.4.1 ESCRT-0 ...... 23 1.4.2 ESCRT-I ...... 24 1.4.3 ESCRT-II ...... 26 1.4.4 ESCRT-III ...... 27 1.4.5 Vps4 ...... 30 1.5 The Bro1 family ESCRT-adaptor proteins ...... 31 1.6 Yeast Bro1 proteins ...... 32 1.7 ALIX ...... 33 1.7.1 Regulatory mechanisms of ALIX ...... 34 1.8 HD-PTP ...... 38 1.8.1 Structural biology of HD-PTP ...... 38 1.8.2 HD-PTP phosphatase domain ...... 39 1.9 Reasons to study HD-PTP ...... 40 1.9.1 HD-PTP in cancer ...... 41 1.9.2 Viral budding ...... 42 1.9.3 Ubap1 in neurodegeneration ...... 42 1.9.4 HD-PTP in signalling ...... 43 1.11 Aims of the PhD ...... 43 References for Chapter 1 ...... 45 Chapter 2: Methodology ...... 50 2.1 EPR spectroscopy ...... 50

1 2.1.1 Continuous-wave EPR ...... 51 2.1.2 Pulsed EPR ...... 54 2.1.3 Distance measurements by double electron-electron resonance spectroscopy .. 54 2.1.4 Simulating interspin distance distributions ...... 58 2.2 Paramagnetic relaxation enhancement ...... 58 2.3 Labelling confirmation by intact protein mass spectrometry ...... 59 2.4 Monitoring structural changes using circular dichroism ...... 60 2.5 X-ray crystallography of proteins ...... 61 2.5.1 Crystallisation of proteins ...... 64 2.5.2 Tackling the phase problem in protein crystallography ...... 64 2.5.3 Molecular replacement ...... 65 2.6 Hydrodynamic protein measurements...... 66 2.6.1 Analytical ultracentrifugation ...... 67 2.6.2 Multi-angle light scattering ...... 69 2.7 Small angle X-ray scattering ...... 72 2.7.1 Flexibility analysis ...... 74 2.7.2 Radius of gyration ...... 74 2.7.3 Real space atom pair distribution ...... 75 2.7.4 Molecular modelling with SAXS data ...... 75 2.8 Summary of methods ...... 76 References for Chapter 2 ...... 78 Chapter 3: Journal format thesis outline ...... 81 Chapter 4: Paper 1 ...... 85 Summary ...... 86 Introduction ...... 86 Results ...... 88

HD-PTPCC Adopts an Open and Extended Conformation ...... 88

HD-PTPCC Shows Limited Local Conformational Flexibility but No Large Structural Rearrangements ...... 92 Identification of the Minimal UBAP1 Binding Region to HD-PTP ...... 97

Mapping the Molecular Interface between HD-PTPCC and UBAP1C ...... 100 Structural Basis for the Specific Functional Interaction of UBAP1 with HD-PTP .. 106 Functional Validation of HD-PTP-UBAP1 Interactions...... 111 Discussion ...... 112 Experimental Procedures ...... 114

2 Acknowledgments ...... 118 References for Chapter 4 ...... 119 Chapter 5: Paper 2 ...... 122 Abstract ...... 123 Introduction ...... 123 Results ...... 126 Variation in crystal structure conformations of Alix and HD-PTP ...... 126

Probing ligand-induced conformational changes in HD-PTPCC using DEER ...... 128

Crystallography of HD-PTPCC with extended Ubap1 peptides...... 133 Discussion ...... 136 Methods ...... 137 Acknowledgements ...... 139 References for Chapter 5 ...... 140 Chapter 6: Paper 3 ...... 142 Abstract ...... 143 Introduction ...... 143 Results ...... 147

The extended architecture of HD-PTPCC, HD-PTPBro1-CC and HD-PTPBro1-CC-PRR rules out a self-inhibited conformation...... 147 Hydrodynamic analyses confirm HD-PTP is a monomer with an extended conformation ...... 150 The HD-PTP open conformation is functionally active ...... 152 Discussion ...... 155 Materials and Methods ...... 158 Acknowledgements ...... 163 References for Chapter 6 ...... 165 Supplementary Information ...... 168 Chapter 7: Paper 4 ...... 169 Abstract ...... 170 Introduction ...... 170 Results ...... 173 HD-PTP and STAM2 coelute by size exclusion as a stable complex...... 173 MALS and AUC support 1:1 stoichiometry of HD-PTP:STAM2 binding...... 176 SAXS data show rigid HD-PTP binds flexible STAM2 to form rigid complex...... 177 Solution structure of STAM2 SH3-GAT ...... 179

3 Solution structure of HD-PTP:STAM2 complex...... 181 Discussion ...... 184 Methods ...... 186 Acknowledgements ...... 188 References for Chapter 7 ...... 189 Chapter 8: Paper 5 ...... 191 Abstract ...... 192 Introduction ...... 192 Results ...... 196 BASL labelling of the three cysteine CC domain ...... 196 BASL labelling of the 6 cysteine Bro1 domain ...... 199 Protein-protein interactions permitted by selective spin labelling ...... 202 DEER distance measurement between BASL label pair possible in single domain, full length protein and in heterodimer complex ...... 203 Discussion ...... 206 Materials and methods ...... 207 CW EPR ...... 208 Mass spectrometry ...... 208 Acknowledgements ...... 210 References for Chapter 8 ...... 210 Chapter 9: Conclusions ...... 214 9.1 Summary of results ...... 214

9.1.1 HD-PTPCC domain adopts elongated structure unlike Bro1 and ALIX V domains...... 214

9.1.2 HD-PTPCC domain conformation not modulated by Ubap1 or monoubiquitin binding...... 215 9.1.3 HD-PTP Bro CC and PRR domains extend into a long binding platform for ESCRT binding...... 215 9.1.4 STAM2 SH3 and Core domains are connected by a flexible linker which stretch out across the extended CC domain to bind HD-PTP Bro1 and PRR simultaneously...... 216 9.1.5 Use of a bromoacrylaldehyde spin label allows probing of DEER distances in multidomain HD-PTP...... 217 9.2 Final discussion ...... 218 References for Chapter 9 ...... 222 Final word count: 50809

4 List of tables and figures

Figure 1.1. Illustration of various membrane scission reactions in the cell ...... 16 Figure 1.2. ESCRT function at the endosome to facilitate ILV formation ...... 19 Figure 1.3. Illustration of some of the diverse roles of ESCRTs in the cell ...... 21 Figure 1.4. Overall assembly of ESCRT complexes in ILV formation ...... 22 Figure 1.5. Human ESCRT-0 structure ...... 23 Figure 1.6. Core of the ESCRT-I complex ...... 25 Figure 1.7. ESCRT-II complex ...... 26 Figure 1.8. ESCRT-III ...... 29 Figure 1.9. Domain structures of Bro1 family proteins ...... 31 Figure 1.10. ALIX and HD-PTP Bro1 ESCRT adaptor proteins both allow the ESCRT pathway to bypass ESCRT-II...... 32 Figure 1.11. Mechanism of ALIX-ESCRT regulation ...... 37 Figure 1.12. Domain organisation of HD-PTP with known interactions to different ESCRT partners...... 39

Figure 2.1. MTSL labelling ...... 53 Figure 2.2. DEER spectroscopy ...... 55 Figure 2.3. Example CD spectra ...... 61 Figure 2.4. The Bragg equation...... 62 Figure 2.5. Flowchart for protein structure solution by X-ray crystallography...... 63 Figure 2.6. AUC ...... 69 Figure 2.7. Multi-angle light scattering ...... 70 Figure 2.8. Dynamic light scattering...... 72 Figure 2.9. SAXS data analysis workflow...... 73 Figure 2.10. Summary of methods ...... 78

Figure 4.1. Crystallographic Structure of HD-PTPCC ...... 90 Figure 4.2. DEER Spectroscopy of HD-PTPCC ...... 94 Figure 4.3. Minimal Binding Region in UBAP1 Responsible for Interaction with HD- PTP ...... 99 Figure 4.4. Crystal Structure of the HD-PTPCC-UBAP1C Complex and Analysis ...... 103 Figure 4.5. NMR Analysis of the HD-PTPCC-UBAP1C Binding Interface...... 104 Figure 4.6. Comparison of the Binding Sites in the HD-PTPCC-UBAP1C, and Alix-Gag Peptide Complexes ...... 108 Figure 4.7. Biochemical and Functional Validation of the UBAP1C Binding Interface with HD-PTP ...... 110

Supplemental Figure 4.1. (Related to Figure 4.2) DEER spectroscopy of HD-PTPCC and Circular Dichrosim analysis of MTSL-labelled HD-PTPCC and mutants used in the EPR studies ...... 95

5 Supplemental Figure 4.2. (Related to Figure 4.2) Torsion-angle molecular dynamics simulations using DEER distance constraints ...... 96 Supplemental Figure 4.3. (Related to Figure 4.3) Co-immunoprecipitation of UBAP1 with HD-PTP ...... 100 Supplemental Figure 4.4. (Related to Figure 4.5) NMR and PRE studies of UBAP1 peptide in the presence of HD-PTP ...... 105 Supplemental Figure 4.5. (Related to Figure 6) Sequence alignment of the coiled-coil domains of human HD-PTP and Alix ...... 109

Table 4.1. Crystallographic Data Collection and Refinement Statistics ...... 91

Figure 5.1 Bro1 family proteins ...... 125 Figure 5.2 Conformational variation in crystal structures ...... 128 Figure 5.3 Crystal packing against HD-PTPCC helix 2 ...... 130 Figure 5.4 HDPTPCC DEER with binding partners ...... 132 Figure 5.5 Crystallography of HD-PTPCC with Ubap1 peptides ...... 135

Figure 6.1. Domain organisation of HD-PTP and Alix ...... 145 Figure 6.2. Analysis of HD-PTP by Small Angle X-ray Scattering ...... 148 Figure 6.3. Hydrodynamic analyses of HD-PTP ...... 151 Figure 6.4. The molecular shapes of HD-PTP and Alix differ...... 153 Figure 6.5. Analysis of HD-PTP binding to ESCRT subunits...... 154 Figure 6.6. HD-PTP is a scaffold for ESCRT binding ...... 157

Supplemental Figure 6.1. Analysis of SAXS data ...... 168

Table 6.1. Hydrodynamic and dimensional experimental data for HD-PTP ...... 164

Figure 7.1 Domain organisation of HD-PTP and STAM2-Hrs ...... 172 Figure 7.2 Mutational analysis of HD-PTP:STAM2 binding ...... 175 Figure 7.3 MALS and AUC of HD-PTP, STAM2 and HD-PTP:STAM2 complex ..... 176 Figure 7.4. SAXS flexibility analysis ...... 178 Figure 7.5 Solution structure of STAM2...... 180 Figure 7.6. HD-PTP:STAM2 complex SAXS ...... 182

Table 7.1. Hydrodynamic parameters for STAM2 model ...... 180 Table 7.2. Hydrodynamic parameters for HD-PTP:STAM2 models...... 184

6 Abbreviations Biological abbreviations: AAA ATPase ATPase associated with diverse cellular activities ALIX Apoptosis linked gene-2 interacting protein X ATP Adenosine triphosphate Bro1 Bypass for C kinase 1 resistance to osmotic shock BroX BRO1 domain-containing protein BROX CC Coiled-coil CHMP Charged multivesicular body protein DUIM double-sided ubiquitin-interacting motif EAP ESCRT-II associated protein EGFR Epidermal growth factor receptor Endofin Endosome associated FYVE domain protein ESCRT Endosomal sorting complex required for transport FYVE Fab 1, YOTB, Vac 1, EEA1 gag Group-specific antigen GAT GGA (Golgi-localised, γ-ear-containing, Arf-binding protein) and TOM (target of myb) domain HD-PTP Histidine domain protein tyrosine phosphatase HIV Human immunodefficiency virus HRS Hepatocyte growth factor regulated tyrosine kinase substrate ILV Intralumenal vesicle MIM MIT-interacting motif MIT microtubule interacting and transport domain MVB Multivesicular body PDGFR Platelet-derived growth factor receptor PEST Rich in proline, glutamic acid, serine and threonine PRR Proline rich region PSAP Proline-serine-alanine-proline motif PTAP Proline-threonine-alanine-proline motif PTP Protein tyrosine phosphatase Rfu1 Regulator for free ubiquitin chains 1 SARA Smad anchor for receptor activation Src Proto-oncogene c-Sarcoma SH3 Src homology 3 Smad Suppressor of mothers against decapentaplegic SOUBA Solenoid of overlapping ubiquitin-associated domains STAM Signal transducing adaptor molecule TSG101 Tumour suppressor gene 101 UBA Ubiquitin-associated domains Ubap1 Ubiquitin-associated protein 1 UE2 Ubiquitin E2 variant VHS Vps-27, Hrs and STAM VPS Vacuolar protein sorting YPXnL Tyrosine-proline-(n number of any amino acid)-leucine motif

Methodological abbreviations: AUC Analytical ultracentrifugation BASL Bromoacrylaldehyde spin label CD Circular dichroism CNS Crystallography and NMR System (MD software) CW EPR Continuous-wave EPR DAMMIF (SAXS ab initio calculation software) DEER Double electron-electron resonance DLS Dynamic light scattering DQC double quantum coherence EPR Electron paramagnetic resonance spectroscopy FOXS Fast X-ray scattering (SAXS curve prediction software) LIC Ligation independent clonin MALS Multi-angle light scattering Mass spec Mass spectrometry/ spectrum/ spectra MD Molecular dynamics MMM Multiscale modelling of macromolecules (DEER prediction software) MTSL Methane thiosulfonate spin label NMR Nuclear magnetic resonance spectroscopy RIDME relaxation-induced dipolar modulation enhancement rmsd Root mean squared deviation (of distance between atoms) SAD Single anomolous diffraction SAXS Small angle X-ray scattering SDS-PAGE Sodium dodecylsulfate polyacrylamide gel electrophoresis SEC Size exclusion chromatography SOMO SOlution MOdeller (hydrodynamic properties prediction software) SPR Surface plasmon resonance svAUC sedimenation velocity Analytical ultracentrifugation TAMDS Torsional angle molecular dynamics simulations

8 Symbols: Å Angstrom = 10-10 m B(t) DEER background factor B0 Magnetic field strength of EPR magnet D Diffusion coefficient Dmax Maximum distance in SAXS atom pair distance distribution f Frictional coefficient F(t) DEER form factor f/f0 Frictional ratio g g-value h Planck’s constant (6.6261 x 10-34 J T-1) I Nuclear spin quantum number I(q) SAXS signal intensity at scattering angle, q Mr Molecular weight 34 -1 N0 Avagadro's number (6.022 x 10 mol ) P(r) SAXS atom pair distance distribution q SAXS scattering angle Rg Radius of gyration

Rh Hydrodynamic radius S Svedberg = 10-13 s S Electron spin quantum number S Svedberg (units for sedimentation coefficent, 1 S = 1 x 10-13sec)

s20,w Sedimentaion coefficient (corrected for solvent density and viscosity, and solute concentration) α Regularisation coefficient (in DEER analysis and SAXS modelling) -24 -1 βe Bohr Magneton (9.2740 x 10 J T ) Δ DEER modulation depth ω Angular velocity of AUC rotor ωee Electron-electron dipolar coupling

9 Abstract

Biogenesis of the multivesicular body (MVB) organelle is an important process for regulation of signalling in the cell. Signal receptors embedded within the outer MVB membrane can be sorted into intralumenal vesicles which bud away from the cytosol to within the MVB preventing further signalling. Sorting of receptors, invagination of the membrane and release of vesicles into the MVB lumen are mediated by the endosomal sorting complexes required for transport (ESCRT) along with a range of accessory proteins including histidine domain protein tyrosine phosphatase (HD-PTP).

HD-PTP is a multidomain protein which makes several interactions with ESCRT partners, including ESCRT-0, ESCRT-I and ESCRT-III. This thesis focusses specifically on the interaction between HD-PTP CC domain and Ubap1 (ESCRT-I), and the two interactions of HD-PTP Bro and PRR domains with STAM2 (ESCRT-0) SH3 and Core domains.

To address the structure of HD-PTP, multiple techniques were used: X-ray crystallography, which gives high resolution structural information; small angle X-ray scattering (SAXS), which gives low resolution data for large non-crystallisable units in their solution state; and double electron-electron resonance (DEER) spectroscopy, which gives high resolution nanometre-range distance constraints between cysteines labelled with methanethiosulfonate spin label (MTSL).

It was shown by X-ray crystallography that HD-PTP has an elongated CC domain, in stark contrast to its homologues ALIX and Bro1 which both have V-shaped CC domains.

The CC domain showed limited flexibility both by SAXS and DEER. Further investigation showed that there was no significant conformational change upon binding its ESCRT-I partner Ubap1.

10 The multidomain structure of HD-PTP Bro1-CC-PRR was described by SAXS, showing that these domains form an extended arrangement in solution. In addition, SAXS was also used to analyse the structure of these domains in complex with STAM2 (ESCRT-0), which showed that STAM2 is simultaneously tethered by the Bro1 domain and PRR.

The Bro-CC-PRR portion of HD-PTP, has 9 cysteines, so with the aim of measuring local structural information in the CC domain alone, alternative spin labelling methods were investigated. Use of a bromoacrylaldehyde spin label (BASL), instead of MTSL, allowed more selective labelling of surface exposed cysteines, and avoided labelling most of the cysteines in the Bro1 domain. This novel method allowed the shape of the CC domain to be monitored during STAM2 binding and showed that there is no induced conformational change.

11 Declaration

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.

12 Copyright statement

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 he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes.

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

The ownership of certain Copyright, patents, designs, trade marks 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.

Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property University

IP Policy (see http://documents.manchester.ac.uk/display.aspx?DocID=24420), in any relevant Thesis restriction declarations deposited in the University Library, The

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

13 Acknowledgements

Firstly, I would like to thank my supervisors: Alistair Fielding, for your unyielding positivity and support; Lydia Tabernero, for encouraging me to reach my full potential; and Phil Woodman, for sharing your knowledge and friendly advice throughout my PhD.

I feel grateful to have had the opportunity to work in two different labs during my PhD, and I am indebted to the countless people who have helped me along the way.

Thank you to the Fielding group and to the Manchester EPR group members, past and present; and to Floriana Tuna for organising engaging EPR group meetings.

I want to express my thanks to the Tabernero group past and present. I would especially like to thank Deepankar Gahloth for training me in protein expression and purification.

To James Adams, Eyong Egbe, Paulina Fernández and Petronela Buiga, thanks for the fun times and I wish you the best of luck for the future. Also thank you to the Michael

Smith C wingers for being such fun neighbours.

Thank you to the Woodman group past and present, especially to Lydia Wunderley for always coming to my help and responding to my questions with such lively enthusiasm.

I owe special thanks to: Jordi Bella for your invaluable advice and training in MD;

Colin Levy for your time and expertise with X-ray crystallography; Daniel Sells and

Adam Brookfield for your help with EPR; Tom Jowitt, Marj Howard and Diana Ruiz

Nivia for your help with MALS and AUC; Mike Lockhart for running SAXS; and David

Knight, Emma-Jane Keevill, and Reynard Spiess for running mass spectrometry.

I am very grateful to my parents, family and friends who have supported and encouraged me throughout my studies. And, lastly, thank you Mike for your love and support.

14 Chapter 1: Introduction

1.1 Membrane budding reactions in cell biology

The cell is the fundamental unit of life, with all living organisms consisting of one or more cells. Universal to all cells is the lipid bilayer which is needed to protect genetic material, facilitate cell division and establish proton gradients in photosynthesis and respiration. In eukaryotes, evolution of intricate membrane organelles allows compartmentalisation of different cellular reactions and signalling processes. An essential mechanism for maintenance of plamsa membranes and communication within the cell is the budding of membranes into vesicles, transporting lipids, proteins and other membrane components from one compartment to another.

Membrane budding reactions are active processes involving specialised macromolecular machinery, with various protein complexes having evolved for different types of membrane transport (Figure 1.1). These complexes include the clathrin and COPI/II coat complexes which produce transport vesicles throughout the cell in endocytosis (import of material) and exocytosis (export of material);1,2 BAR (Bin1/amphiphysin/Rvs167) proteins which transport material between the Golgi complex to the endosome, with additional roles in phagocytosis and cytokinesis;3 and ESCRT (endosomal sorting complexes required for transport) complexes which act at the endosome and have a range of other roles including microvesicle formation and viral budding.4–6

Figure 1.1. Illustration of various membrane scission reactions in the cell. Reactions are annotated with the relevant membrane scission proteins involved: Coats (clathrin and COPI/II coat complexes), BAR (Bin1/amphiphysin/Rvs167) and ESCRT (endosomal sorting complexes required for transport). Figure adapted from Field et. al., 2011.7

All membrane budding machinery operate from the cytosolic membrane face. However, for certain events it is necessary for membranes to either be pulled into the cytosol

(“normal” topology), whilst for other transport events they are pushed outwards from the cytosol (“reverse” topology).

The most thoroughly studied complexes for lipid membrane budding are clathrin coats, which transport small lipid vesicles from the cell surface membrane (plasma membrane) into the cytosol. Clathrin coats are also found at several other intracellular transport steps, including transport of vesicles between the Golgi complex and endosome. These pathways operate using normal topology; clathrin subunits attach via adaptins to activated receptors on the cytosolic face of the plasma membrane and distort the membrane inwards forming a clathrin coated pit. Membrane scission by the dynamin GTPase releases the

16 vesicle from the plasma membrane (Figure 1.2A).2 This topology of cytosolic proteins pulling vesicles inward is also appropriate for a wide range of membrane budding events, budding of vesicles that transport cargo from the endoplasmic reticulum to the Golgi complex, or vesicles that bud from the trans-Golgi network.

A prominent example of reverse topology membrane budding occurs on endosomal membranes during the formation of intraluminal vesicles (ILVs); vesicles which bud away from the cytosol to within the lumen of endosomes, taking their transmembrane protein cargo with them.8 This process, which generates a multivesicular body (MVB) is an important regulatory process for deciding the fate of transmembrane proteins and is carried out by the ESCRTs. The ESCRTs assemble on the cytosolic surface of the endosome and push vesicles outward into the endosome lumen (Figure 1.2A).

1.2 Endosomal trafficking of receptor proteins by ESCRTs

The role of receptor proteins is to receive and transmit messages from outside to inside the cell. For example, mitogenic receptors such as EGFR (epidermal growth factor receptor) and PDGFR (platelet derived growth factor receptor) bind to extracellular growth factors and trigger cellular responses such as growth or division. These signals are amplified in the cell through signalling cascades, initiated by phosphorylation of the receptor’s cytosolic domain, and so must be “downregulated” to control their strength and duration. These signals can be downregulated by various mechanisms including dephosphorylation of their cytosolic domain. Another important mechanism by which this is achieved is through ubiquitination of the surface receptors, which causes their internalisation into transport vesicles by clathrin2 or other complexes.9 Normal topology budding proceeds with retention of membrane configuration, meaning that the cytosolic signalling domain remains in contact with the cytosol and its signalling partners. Once these vesicles fuse with the endosome, depending on the receptors’ ubiquitination

17 status,10 they may be recycled via recycling endosomal tubules back to the plasma membrane, or ESCRTs can be recruited to block this recycling and instead sort the receptors into ILVs. In this case, the signalling domains are removed from contact with the cytosol, which interrupts downstream signalling cascades. Hence, ESCRT-mediated

ILV formation is an efficient method of signal downregulation.

Mature endosomes containing ILVs are referred to as multivesicular bodies (MVBs) and their transmembrane cargo will eventually be transferred by fusion of the MVB with the lysosome, an organelle where proteins are hydrolysed by proteases. (Figure 1.2B)11 There is also some evidence, however, that ILVs can back-fuse with the endosomal membrane, a process also mediated by ESCRTs, which may protect some sorted cargo from lysosomal degradation.12

Figure 1.2. ESCRT function at the endosome to facilitate ILV formation. (A) Inward budding vs outward budding. 1) clathrin-mediated endocytosis at the cell membrane is an example of inward budding: a clathrin/adaptin coat assembles on the inside face of the cell membrane and pulls a budding vesicle into the cytosol. 2) ESCRT-mediated intralumenal vesicle formation is an outward budding process: ESCRTs 0-III assemble on the outside of the endosome and push a vesicle away from the cytosol into the endosomal lumen. (B) Maturation of early endosome into a multivesicular body then a degradative lysosome.

1.3 Other roles of ESCRTs

Although the ESCRTs were named for their endosomal function, it is now clear that they have a much wider range of roles throughout the cell (Figure 1.3).6,13,14 ESCRTs are

19 known to be involved in cytokinesis, the final scission step in cell division.15 ESCRTs are also involved in releasing two types of vesicles to the extracellular space: exosomes, which are released when an MVB attaches to the plasma membrane; and microvesicles which bud directly from the plasma membrane.16 Some enveloped viruses can also hijack these ESCRT complexes to facilitate their viral budding at the plasma membrane.17,18

Some of the other roles of ESCRTs in cells include plasma membrane repair,19 nuclear envelope reformation20 and dendrite scission in nerve cells.21 From a broad perspective, all these processes involve similar “reverse” topology events: the membrane are all deformed and bud away from the cytoplasm. However, to be able to participate in such widespread cell functions, the ESCRTs requires a degree of adaptability. Firstly, this is enabled by the modular nature of ESCRT recruitment which are recruited selectively and sequentially to the target membrane. Furthermore, each complex composition is variable with alternative isoforms possible for many of the protein subunits. Finally, additional adaptor proteins are present in some processes which allow ESCRT pathways to be bypassed.

Figure 1.3. Illustration of some of the diverse roles of ESCRTs in the cell.

1.4 Molecular architecture of the ESCRT machinery

The proteins which make up the ESCRT machinery were first identified in yeast by studying the effects of genetic mutations on protein transport to the vacuole (the yeast and plant equivalent of the lysosome). Certain “Class E” vps (vacuolar protein sorting) mutants were identified which, although appearing to give normal vacuoles, also led to the growth of prominent prevacuolar “Class E” compartments.22 The contents of these compartments were revealed, by immunostaining, to include the transmembrane protein

V-ATPase; whereas ALP (alkaline phosphatase), which is another transmembrane protein, was still being correctly transported to the vacuole. Further studies showed that the class E vps genes are responsible for ILV formation, so when they are mutated the

21 endosomal sorting pathway of transmembrane proteins like V-ATPase to the vacuole is blocked. However, not all vacuole transport occurs via the endosome. For example, ALP is transported directly from the Golgi complex to the lysosome. The class E vps proteins have now been shown to interact with one another as a series of stable cytosolic complexes referred to as ESCRT-0, ESCRT-I, ESCRT-II and ESCRT-III (endosomal sorting complex required for transport-0, I, II and III).

The four ESCRT complexes generate interactions with membranes, ubiquitinated substrates and each other to promote ILV formation. ESCRT-0, I and II form quaternary complexes in the cytosol and are each recruited to the endosome en bloc, whereas

ESCRT-III exists as monomers in the cytosol until polymerising into filaments on the endosomal membrane. Substantial evidence has shown that the ESCRTs are recruited sequentially, in numerical order, from ESCRT-0 to ESCRT-III, culminating in ESCRT-

III subunits polymerising to form filaments, which curve around the neck of the budding vesicle to generate membrane curvature (Figure 1.4). Afterwards, the VPS4 ATPase is recruited. This uses energy from ATP hydrolysis to break apart the ESCRT-III into soluble monomers and catalyses membrane fission.

Figure 1.4. Overall assembly of ESCRT complexes in ILV formation.

22 1.4.1 ESCRT-0

The ESCRT-0 complex is a 1:1 heterodimer of the two subunits HRS (hepatocyte growth factor regulated tyrosine kinase substrate) and STAM1 or STAM2 (signal transducer adaptor molecular 1 or 2) interacting via their coiled-coil GAT (GGAs and

TOM) domains (Figure 1.5).23,24 Both of the subunits, HRS and STAM1/2 contain ubiquitin binding domains to recruit ESCRT-0 to ubiquitinated cargo, and the complex also anchors to membranes through binding of HRS FYVE (Fab 1, YOTB, Vac 1,

EEA1) zinc finger domain to the endosome-specific phospholipid Ptdlns3P

(phosphatidylinositol-3-phosphate).25 After associating with the target membrane and binding ubiquitinated cargo ESCRT-0 can recruit the ESCRT-I complex through interactions between HRS and the TSG101 (tumour suppressor gene 101) subunit of

ESCRT-I.26

Figure 1.5. Human ESCRT-0 structure. Hrs forms a heterodimer with either STAM1 or STAM2. A) crystal structure of core regions of Hrs (green) and STAM1 (yellow) [PDB: 3F1I]24 B) Domain structure of Hrs, STAM1 and STAM2. Interaction between Hrs core and STAM1 or STAM2 core are shown with dashes. Domain boundaries from uniport database. Diagram was constructed using IBS web server (http://ibs.biocuckoo.org/). 23 1.4.2 ESCRT-I

The ESCRT-I complex is a heterotetrameric complex of: TSG101, VPS28 (A or B spliceoforms), VPS37 (A, B, C or D) and MVB12 (A, B or UBAP1) (Figure 1.6).27–31

Yeast has single versions of VPS37 and MVB12 and the X-ray crystal structure of the yeast ESCRT-I reveals an elongated coiled-coil stalk of TSG101, VPS37 and MVB12 which meet at a globular head group along with VPS28.32 Since ESCRT-I only associates with membranes by weak electrostatic interactions involving TSG101, its recruitment to the endosome is dependent on the ESCRT-0 adaptor complex.32 Yeast ESCRT-I is able to recruit ESCRT-II via interactions between the globular head group domain Vps28p with ESCRT-II Vps36p.33 This interaction is dependent on N2F insertions in the GLUE domains which are not present in the human homologs and no human ESCRT-I:ESCRT-

II interaction motif has been identified.

In higher eukaryotes, all ESCRT-I subunits except TSG101 have several possible isoforms/ gene variants meaning there is wide structural and functional variation available to ESCRT-I complexes. However certain subunits are required to be present to allow others to also be associated. For example, the HD-PTP-interacting subunit Ubap1 can replace MVB12A/B in the presence of VPS37A but not VPS37B/C.34 Ubap1 forms interactions with ubiquitin through its triple UBA (ubiquitin-associated domains),

SOUBA (solenoid of overlapping ubiquitin-associated) domain,34 which allows this

Ubap1-containing ESCRT-I complex to be involved in the sorting of ubiquitinated endosomal cargoes.35

Figure 1.6. Core of the ESCRT-I complex. A) The crystal structure of yeast ESCRT-I shows a heterotetrameric stalk and head structure. The stalk is made up of long helical regions of Vps23p/TSG101 (purple), Vps37p/VPS37 (cyan), Mvb12p/MVB12 (magenta) and at the head these interact with the final subunit Vps28p/VPS28 (blue).32 [PDB: 2P22] B) Domain structure of human TSG101, VPS28, MVB12(A,B), Ubap1 and VPS37(A,B,C,D). Domain boundaries from uniport database. Diagram was constructed using IBS web server (http://ibs.biocuckoo.org/).

25 1.4.3 ESCRT-II

Mammalian ESCRT-II is a heterotetrameric complex of EAP30 (ESCRT-II associated protein 30), EAP45 (ESCRT-II associated protein 45), and two copies of EAP20

(ESCRT-II associated protein 20) (Figure 1.7). They associate in a capital Y shape, where the N termini of EAP30 and EAP45 which, together, form the body are each tethered to a EAP20 subunit which form the two Y arms.36–38 EAP45 in the core of the complex binds phophoinositides as well as ubiquitin.39 The EAP20 arms have been shown to bind the

ESCRT-III component CHMP6 and scaffold the formation of ESCRT-III filaments.40

Figure 1.7. ESCRT-II complex. Top: Crystal structure of yeast ESCRT-II shows a tetramer of Vps36p/EAP45 (yellow), Vps22p/EAP30 (orange), and 2 copies of Vps25p/EAP20 (red).38 [PDB: 2ZME] Bottom: domain structure of human EAP45, EAP30 and EAP20. Domain boundaries from uniport database. Diagram was constructed using IBS web server (http://ibs.biocuckoo.org/).

26 1.4.4 ESCRT-III

ESCRT-III filaments assemble at the neck of budding intralumenal vesicles and distort the membrane. The ESCRT-III proteins comprise: CHMP1 (A, B), CHMP2 (A,B),

CHMP3, CHMP4 (A, B, C), CHMP5, CHMP6, CHMP7 and IST1 (Figure 1.8E). Unlike the other ESCRTs (0, I and II), ESCRT-III does not form a cytosolic complex. Instead

ESCRT-III subunits exist as closed-state monomers, where α5 packs against α1 and α2

(Figure 1.8B,D).

Dissociation of α5 results in a conformational change of ESCRT-III proteins from closed to open forms is linked to the activation for polymerisation into filaments. (Figure

1.8A,C).41–44 This activation can be induced when recruited by ESCRT-II or Bro1. Once in their open forms, CHMP proteins expose their C terminal MIM (MIT (microtubule interacting and transport domain) interacting motifs) motifs which can bind MIT motifs of the VPS4 complex (VPS4 aaaATPase and its activator Vta1/LIP5).45 Microscopy of fluorescently labelled ESCRT-III proteins during MVB formation showed that the proteins are recruited in the order: CHMP6, CHMP4, CHMP3 then CHMP2.46 The final recruitment of VPS4 complex by CHMP2 allows recycling of the ESCRT-III for multiple rounds of membrane scission.47

The cryo-electron microscopy structure of CHMP1B:IST1 by McCullough et. al. demonstrates how two ESCRT-III can polymerise into a spiral structure (Figure 1.8F).

However, this structure challenges the conventional ESCRT-III polymerisation mechanism in two ways: firstly it demonstrates an example of ESCRT filaments which contain both open (CHMP1b, Figure 1.8C) and closed conformation subunits (IST1,

Figure 1.8D), and secondly it is example of “normal topology” membrane deformation i.e. the filaments assemble on the outside of a tubule rather than the inside.48

27 There are multiple theories about how the filaments assemble at the ILV neck and drive changes to membrane curvature. One theory is that the filaments may form spiral shaped dome structures, which initiating from ESCRT-II and grow either towards or away from the cytosol within the neck.49 A current prominent theory is the buckling theory, which is where concentric rings of filaments grow within the neck, from narrow to wide, out towards the cytosol; the dome then flattens towards the cytosol which pinches the neck together (Figure 1.8G).50–52

Figure 1.8. ESCRT-III. A) Snf7 (CHMP4), open conformation. [PDB B) CHMP3, closed conformation. C) CHMP1B, open conformation. D) IST1, closed conformation. E) Domain structure based on annotations by Schöneberg and colleagues and secondary structure prediction using PSIPred (http://bioinf.cs.ucl.ac.uk/psipred/).52 Diagram was constructed using IBS web server (http://ibs.biocuckoo.org/). M1 = MIM1 M2 = MIM2. F) EM structure of CHMP1:IST1 ESCRT-III filaments. G) Buckling model of ESCRT- III filament scission: concentric circles form, growing from small to large towards the cytosol, buckling the vesicle neck; then the filaments flatten, causing the neck to pinch.

29 1.4.5 Vps4

Vps4 AAA-ATPase (vacuolar protein sorting 4 ATPase associated with diverse cellular activities) uses ATP hydrolysis to disassemble the ESCRT-III filaments and mediate membrane remodelling.43,53 The full AAA-ATPase is formed of two subunits Vps4 and

Vta1/LIP5 which are inactive until their recruitment to ESCRT-III via interactions of

Vps4 MIT domains with ESCRT-III MIM1 and MIM2 domains. It has been shown that in its active form, the Vps4-Vta1/LIP5 complex assembles into an oligomeric structure.54

30 1.5 The Bro1 family ESCRT-adaptor proteins

The Bro1 (Bypass for C kinase 1 resistance to osmotic shock) domain containing proteins act as accessory proteins to the ESCRTs. Yeast have two family members: Bro1 (the origin of the family name) and Rim20. Mammals have several Bro1 proteins but the most highly studied are ALIX (Apoptosis linked gene-2 interacting protein X), HD-PTP

(Histidine domain protein tyrosine phosphatase) and BROX (BRO1 domain-containing protein BROX) (Figure 1.9). The Bro1 proteins have a similar domain architecture with most containing Bro1 and neighbouring coiled-coil (CC) domains (sometimes termed V domains), and some containing proline rich regions (PRR). HD-PTP additionally has a protein tyrosine phosphatase (PTP) domain and a PEST region (rich in proline, glutamic acid, serine and threonine).

Figure 1.9. Domain structures of Bro1 family proteins. ALIX, HD-PTP and BroX from Homo sapiens (hs) and Bro1 and Rim20 from Saccharomyces cerevisiae (sc). Domain boundaries are from uniprot database (http://www.uniprot.org/) or from crystal structures where available. CC domains have “V” annotated only where X-ray crystallography has verified the V shape. Diagram was constructed using IBS web server (http://ibs.biocuckoo.org/).

31 Some of the common features of all Bro1 domains are the interaction of Bro1 domain with CHMP4 (Snf7 in yeast)55,56 and the interaction of their CC domains (when present)

57 with YP(X)nL (where n = 1 to 3) motif-containing substrates. Additional interactions and the specific partners of their CC domains confer a range of different functions.

Both human Bro1 proteins, ALIX and HD-PTP, bind to additional ESCRTs throughout the ESCRT pathway, with ALIX binding ESCRT-I17 and HD-PTP binding both ESCRT-

I34,58 and ESCRT-0.59 However, neither of these human Bro1 proteins bind to ESCRT-II, which means that ESCRT pathways involving ALIX and HD-PTP can bypass ESCRT-II

(Figure 1.10). This is another way that ESCRTs achieve the adaptability required to be able to carry out such widespread cellular functions.

Figure 1.10. ALIX and HD-PTP Bro1 ESCRT adaptor proteins both allow the ESCRT pathway to bypass ESCRT-II.

1.6 Yeast Bro1 proteins

There are two yeast Bro1 proteins, Bro1 (Bypass for C kinase 1 resistance to osmotic shock) and Rim20. The yeast Bro1 is involved in ILV formation at the MVB60,61 whereas

Rim20 is involved in the Rim101 alkaline response pathway.62

Yeast Bro1 is involved in ubiquitinated cargo sorting at the endosome. Its coiled-coil domain which binds regulator for free ubiquitin chains 1 (Rfu1) as well as binding

32 ubiquitin directly.63,64 The Bro1 domain binds Snf7 (yeast CHMP4 orthologue)55 allowing Bro1 to recruit ESCRT-III filaments. Bro1 dependent activation of ESCRT-III has been shown to act in parallel with the ESCRT-I/ESCRT-II activation of ESCRT-III.61

The Bro1 domain of Rim20 also binds snf7 but its coiled-coil domain binds to

Rim101.62,65 Rim101 in a transcriptional activator in yeast which is enters the nucleus after proteolysis in response to increased extracellular pH.66 Rim20’s connection with this pathway as well as its links with the ESCRT machinery suggest a role in pH responsive

MVB formation.65

1.7 ALIX

Apoptosis linked gene-2-interacting protein X (ALIX), as the name implies, was originally identified as an interacting protein partner of ALG-2,67,68 which is a protein responsible for controlled apoptosis-cell death in response to certain signals. However,

ALIX was later shown to be involved in several different ESCRT-dependent processes.

Like yeast Bro1, the ALIX Bro1 domain interacts with CHMP4, acting as a recruiter of

ESCRT-III filaments56 (Figure 1.10). ALIX also forms an interaction with ESCRT-I through its PRR, with a PSAP motif responsible for recognition by the UEV (Ubiquitin

E2 variant) domain within TSG101.17,18,69

Most attention has been given to ALIX-mediated exosome formation, which is a process whereby MVBs fuse with the plasma membrane and their content ILVs are secreted outside the cell as extracellular vesicles, or exosomes.70 The formation of ILVs for eventual exosome secretion is promoted by the interaction of ALIX with syntenin through

16 its CC/V domain. Syntenin contains two YP(X)nL motifs (LYPSLEDL and

LYPRLYPELSQYMGL) which interact with the G(X)2FY(X)2L site of ALIX V.

33 ALIX has also been shown to be involved in viral budding. Several viral gag domains contain YP(X)nL or similar motifs which allow them to bind the syntenin binding site of

ALIX V. These gag domains also contain PTAP motifs to bind TSG101, so together

ALIX and TSG101 are used to hijack the ESCRT exosome secretion system to facilitate viral budding.17,71–74

In addition to exosome formation/viral budding, ALIX is also involved in the final membrane scission stage of cell division (cytokinesis)15 and is also involved in receptor downregulation, demonstrated through its trafficking of EGFR.75

1.7.1 Regulatory mechanisms of ALIX

Structures of the ALIX Bro1 and V domains, showing their interaction with CHMP4 and viral gag peptides, and the structure of the dual domain ALIX Bro1-V have been solved by crystallography.56,71,72 Several publications have investigated the structure of ALIX further to determine how ALIX regulates access of CHMP4 to its Bro1 domain for

ESCRT-III filament assembly.

There are three main structural arrangements which have been discovered to regulate

ALIX-ESCRT activity: dimerization of ALIX via the V domain; opening and closing of the V domain arms; and autoinhibition by intramolecular interaction of the PRR with the

Bro1 domain.

ALIX has been shown to dimerise, evidenced both by multi-angle light scattering

(MALS) of recombinantly bacterially expressed protein76 and by pull down experiments using epitope-tagged ALIX expressed in human cells.75 The first study of ALIX by small angle X-ray scattering showed that its V-shaped structure was maintained in solution as expected both with the V domain alone and in the presence of the N terminal Bro1 domain.76 SAXS analysis of the ALIX Bro1-V dimer was inconsistent with the crystal

34 structure, and supported a model where the V hinge opened into an extended conformation. Hydrogen/deuterium exchange measured by mass spec identified a dimerization interface involving residues 638-645, which forms part of the ALIX V hinge.

The secondary structure of the ALIX V domain forms 6 helical regions, and folds back on itself twice resulting in the formation of two trihelical “arms” with three “passes” through its looped hinge region.71,72 The hinge is stabilised by interactions between residues in the two arms and passes. First, E650 (arm 2, pass 3) forms salt bridges with

K640 (hinge, pass 3) and N400 (arm 1, pass 1) (Figure 1.11Bi). Second, R649 (arm2, pass 3) hydrogen bonds to ordered waters which in turn hydrogen bond to T412 (hinge, pass 1) and N531 (arm 2, pass 2) (Figure 1.11Bii).

Attempts to mutate the hinge dimerization interface successfully resulted in exclusively monomeric ALIX V. However, these mutations gave the additional effect of causing an opening of the V domain into its extended conformation, presumably by interrupting the

K640-N400 salt bridge.76 Further work by another group also showed using SAXS that the V can be opened into an extended conformation by a single point mutation of the hydrogen bonding R649 residue to aspartic acid (R649E).77 Together these results showed that there is a functional link between the dimerization and hinge mechanisms.

Various lines of evidence have shown that the presence of the ALIX PRR leads to autoinhibition of V domain binding to YP(X)nL proteins and of Bro1 domain binding to

CHMP4 and Src.78–80 Structural studies using small-angle X-ray scattering (SAXS) have further showed how the PRR folds back against the V and Bro1 domains into a compact structure.77 Release of the PRR domain from the Bro1 can be triggered by a number of activation processes, depending on the scenario: the PRR can be phosphorylated which releases it from the Bro1 domain, during exosome formation; binding to CEP55 releases

35 the PRR during cell division and binding of the PRR to ALG-2 releases it during ILV formation at the MVB.81

Combined, these mechanisms have been proposed to involve first, the release of the PRR from the Bro1 domain, followed by opening of the V domain and the formation of an functional dimeric ALIX (Figure 1.11C).77 However, some aspects of this mechanism as still unclear. For example, the crystal structure of ALIX Bro1-V in its closed form does not appear to be compatible with PRR being able to reach the CHMP4 binding site of the

Bro1 domain, indicating that there may be some interdomain flexibility between the Bro1 and V domains. Alternatively, the PRR-Bro1 interaction could occur across the ALIX dimer meaning that this interaction is not relieved until later in the mechanism. Since

ALIX is involved in a wide range of functions, it may be possible that its monomer-dimer and open-closed behaviour may be situation dependent with no universal mechanism to describe all cases.

Figure 1.11. Mechanism of ALIX-ESCRT regulation. (A) Crystal structure of ALIX Bro1-CHMP4A complex (PDB:3C3R)56 superimposed on ALIX BroV-HIVgag structure (PDB:2R02)73, with PRR - TSG101 interaction draw as cartoon. (B) ALIX V hinge interactions. (i) N400 and K640 both form salt bridges to E650. (ii) R649 forms hydrogen bonds with ordered waters also hydrogen bonding to T412 and N531. (C) Cartoon of autoinhibited ALIX with PRR interacting with Bro1. ALIX dimers have been proposed to be formed where the V domains interact as extended conformations and PRR-Bro1 interactions are lost, allowing recruitment of CHMP4A/B/C.

37 1.8 HD-PTP

Histidine domain protein tyrosine phosphatase (HD-PTP) is a Bro1 protein found in human cells and conserved in Drosophila (where it is named myopic), with ESCRT- interacting roles largely limited to endosomal sorting. Studies have identified a number of different transmembrane proteins which require HD-PTP for endosomal trafficking, including EGFR,82 PDGFR,83 src84 and α5β1integrin.85

In terms of domain structure, HD-PTP also has the familial Bro1 and CC domains. The

HD-PTP-PRR is much longer than those of Bro1 or ALIX and is rich in histidine

(sometimes referred to as the “his domain”), and the PRR has also been suggested to form a potential zinc finger, although no experimental study has verified this claim.86 Distal to the PRR is a protein tyrosine phosphatase (PTP) domain which is not present in any other

Bro1 members. Both the PRR/ his domain and PTP domain of HD-PTP have, however, been shown to be dispensable for HD-PTP activity in ESCRT trafficking of EGFR.82

1.8.1 Structural biology of HD-PTP

The Bro1 domain of HD-PTP, like ALIX yeast Bro1, interacts with the ESCRT-III component CHMP4 (Figure 1.12). HD-PTP, like ALIX, also binds the ESCRT-I component TSG101 through a 2x PSAP motif in its PRR. However, the V domain

(FYADL) motif, instead of binding substrates as is the case for ALIX, binds Ubap1, the non-core component of ESCRT-I which is required for MVB sorting.34,58 HD-PTP, therefore, makes two interactions with ESCRT-I, and specifically recruits the Ubap1- specific ESCRT-I.

Unlike Bro1 and ALIX, HD-PTP also interacts with ESCRT-0.59 This occurs at two sites: the Bro1 domain binds the STAM2-core (GAT) domain, and the PRR binds the STAM2-

SH3 domain. There are interactions which compete with HD-PTP-STAM2 binding.

38 Firstly, the site on HD-PTP-Bro1 that binds STAM2-core overlaps with its the Bro1-

CHMP4 binding site. Secondly, the site on HD-PTP-PRR that binds STAM2-SH3 overlaps with the PTAP motif which binds TSG101. And lastly, the STAM2-SH3 also binds to the ubiquitin specific peptidase 8 (USP8/UBPY). This leads to a mechanism where ESCRT-0/I/III proteins and the ubiquitinase USP8/UBPY interact with HD-PTP by competitive displacement of one another.

Figure 1.12. Domain organisation of HD-PTP with known interactions to different ESCRT partners.

1.8.2 HD-PTP phosphatase domain

The most striking disparity between HD-PTP and all other Bro1 family proteins is the additional C-terminal protein tyrosine phosphatase domain. As mentioned above this domain has been shown to be dispensable for HD-PTP mediated vesicle budding into the endosome.82 However, loss of the PTP domain or mutation of the catalytic cysteine residue to serine diminishes the tumour suppressing activity of HD-PTP in cell experiments where cancer was stimulated by ras activation.87

Phosphatases are enzymes which catalyse the hydrolysis of phosphate post-translational modifications, antagonising the action of kinases which are responsible for introducing these modifications. Phosphate groups, which are large and negative, modulate the activity of enzymes, allowing phosphatases and kinases to act as versatile regulators of a wide range of signalling pathways.88 HD-PTP is classed as a non-receptor PTP,

39 sometimes being referred to as PTPN23, but there has been some debate about the catalytic activity of HD-PTP. There have been some conflicting publications on the in vitro catalytic activity of HD-PTP, with some reporting activity84,89 and others arguing against it.90

A rigorous study by Gingras and coworkers argued the case that HD-PTP is inactive and that previous reports of weak activity were within experimental error of their assays.90

All phosphatases contain several consensus motifs, including the crucial motif 9

(PXXVHCSAGXGRTG) at the active site, which includes the catalytic cysteine residue.

They noticed that although the HD-PTP sequence at this site (PIIVHCSSGVGRTG) does contain the catalytic cysteine, it differs from the motif by a serine at residue 1394 instead of an alanine; a substitution which is conserved in HD-PTP across several species. They demonstrated using mammalian expressed HD-PTP and employing the highly sensitive

8-difluoro-4-methylumbiliferyl phosphate assay substrate that there was no catalytic activity in wild type HD-PTP. However, by mutating S1349A they could reintroduce activity, highlighting the importance of this substitution for modifying HD-PTP’s catalytic activity.

The function of the HD-PTP phosphatase domain remains unknown. One possibility is that the PTP may act as a non-catalytic tether, allowing it to bind (but not hydrolyse) phosphorylated binding partners and promoting the recruitment of the ESCRT complexes.

1.9 Reasons to study HD-PTP

With interactions spanning across early and late ESCRTs, HD-PTP has links with a range of areas in health and disease.

40 1.9.1 HD-PTP in cancer

As proteins with roles in downregulation of signalling receptors it seems likely that

ESCRTs and associated proteins would behave as tumour suppressors. Indeed, several

ESCRT members have specifically been identified as tumour suppressors, including

TSG10191 and VPS37A (aka HCRP1).92

HD-PTP from its discovery was suspected to be a tumour suppressor86 due the positioning of its gene on chromosome 3 in an area mutated in cancer.93 Homozygous deletions (loss of both genes) of HD-PTP resulted in embryonic lethality in mice.94 Further studies in mice exploring hemizygous deletions (i.e. only a single copy of the gene is present), resulting in halved expression of HD-PTP, showed an increased susceptibility to tumourigenesis.95 Furthermore, it has been shown that these hemizygous deletions of HD-

PTP are common amongst human cancer patients. Although removal of the PTP domain increased tumorigenesis in cells,87 the more apparent link between HD-PTP and cancer arises from the ESCRT binding and endosomal trafficking roles performed by its N- terminal domains.

HD-PTP has been shown to be involved in the downregulation of a range of transmembrane receptors including EGFR,82 PDGFR-β,83 α5β1 integrin85 and Src.84,96

Depletion of HD-PTP decreases downregulation of these receptors and leads to boosted signalling for cell proliferation, cell migration and invasion.83–85,97

To summarise there is genetic and functional evidence for tumour suppressor activity of

HD-PTP from studies in cells, mice and humans. Understanding the mechanisms of HD-

PTP-mediated ESCRT-assembly will be crucial to understanding cancer in ESCRT biology and is a key step in research towards ESCRT targeted cancer therapy.

41 1.9.2 Viral budding

ALIX but not HD-PTP has been shown to have a role in the budding of retroviruses including HIV, SIV and EIAV, through their targeting to ALIX V domain.17,71–74 This corresponds with the role of ALIX and not HD-PTP in the apparently equivalent process of exosome formation at the plasma membrane.16 Presumably retroviruses have evolved to selectively mimic the ALIX-syntenin interaction, but to avoid the similar recognition in motif in HD-PTP through which they would find themselves doomed to the lysosome.

Appreciating the structural differences between HD-PTP and ALIX, and the key to selectivity between HD-PTP-Ubap1 and ALIX-syntenin/gag is a crucial step towards understanding the mechanisms of viral infection and potential therapeutics.

1.9.3 Ubap1 in neurodegeneration

Many neurodegenerative diseases have links to membrane trafficking proteins,98 and both

CHMP2B (ESCRT-III)99,100 and Ubap1 (ESCRT-I)101 have been implicated in frontal temporal lobar degeneration (FTLD). FTLD is a disease causing behavioural changes in middle aged patients, characterised by decline in social conduct, disinhibition and emotional blunting as well as a range of other symptoms including dietary changes and repetitive behaviour.102 Some FTLD patients also go on to develop motor neurone disease. Analysis of brain tissue of patients with a pathological subtype FTLD called

FTLD-U shows the presence of ubiquitinated aggregates, which contain TAR binding protein (TDP-43), and this was also shown be colocalised with Ubap1.101

Nearly half of FTLD patients have a family history of neurodegenerative disease, and genetic studies of patients of FTLD highlighted a number of mutations present in Ubap1 that were associated with FTLD.101 One of these, P256L is close to the 268FPK271L interaction motif which binds 678FYAD682L motif of HD-PTP CC domain.34 Therefore, it

42 will be important to also investigate whether there are any other additional interactions in this mutated area with HD-PTP, and whether this mutation causes any structural rearrangement.

1.9.4 HD-PTP in signalling

Despite being responsible for receptor degradation, there is evidence that the early

ESCRT complexes may also be involved in signalling upregulation. For example, Hrs

(ESCRT-0) has been shown to be involved in both positive and negative regulation of signalling,103,104 and studies have shown that the so-called tumour suppressor TSG101 is actually upregulated rather than depleted in multiple types of cancer.105–108 This work shows that there is a role for signalling in the early ESCRT machinery.

HD-PTP, unlike ALIX, interacts with ESCRT-0, via its Bro1 domain and PRR,59 and one way which HD-PTP:ESCRT-0 may be involved in signalling is through the interactions that the HD-PTP Bro1 domain also makes with SARA and endofin. These are both endosomal membrane anchored proteins involved in Smad signalling pathways.109,110

Describing the nature of HD-PTPs interaction with STAM2 (ESCRT-0) will therefore not only be important for understanding the endosomal trafficking mechanisms but also be needed to fully understand signalling by SARA and endofin.

1.11 Aims of the PhD

The overall aims of the PhD were to understand the structure and mechanism of

HD-PTP’s interaction with ESCRTs:

• Firstly, to investigate the previously uncharacterised structure of the HD-PTP CC

domain along with its interactions with Ubap1 (ESCRT-I).

43 • To study, in detail, the conformations of the HD-PTP CC domain for comparison

with ALIX V

• To investigate the overall conformation of HD-PTPBro1-CC-PRR, which represents

the minimum unit for HD-PTP/ESCRT-mediated ILV formation

• To explore the dual mode interaction of HD-PTPBro1-CC-PRR with STAM2SH3-Core

and to determine whether any changes any conformation of HD-PTP are induced

by binding.

To address these aims, a range of structural methods were used including X-ray crystallography, SAXS, MALS and AUC, which have all been previously used to study

ALIX.

Given the important role of conformational change in the regulation of ALIX, a large focus of the PhD was to investigate how similar mechanisms may work in HD-PTP. So, for detailed analysis of the conformations of HD-PTP CC domain, double electron- electron resonance (DEER) spectroscopy was employed to derive distance constraints for structural modelling. This label-probe method has previously been applied to some of the other proteins in this field including ESCRT-I111 and ESCRT-III112 but not to Bro1 proteins.

DEER is much simpler to interpret when there are only pairs of cysteines. The HD-PTP

CC domain contains only three cysteines, which made it ideal for study. However, the

Bro1-CC-PRR region contains nine cysteines, making it much more challenging to approach using DEER. Therefore, an additional aim was to investigate spin labelling approaches for studying the multi-domain structure of HD-PTP.

44 References for Chapter 1

1 J. S. Bonifacino and J. Lippincott-Schwartz, Nat. Rev. Mol. Cell Biol., 2003, 4, 409– 414. 2 H. T. McMahon and E. Boucrot, Nat. Rev. Mol. Cell Biol., 2011, 12, 517–533. 3 G. Ren, P. Vajjhala, J. S. Lee, B. Winsor and A. L. Munn, Microbiol. Mol. Biol. Rev., 2006, 70, 37–120. 4 W. M. Henne, N. J. Buchkovich and S. D. Emr, Dev. Cell, 2011, 21, 77–91. 5 J. McCullough, L. A. Colf and W. I. Sundquist, Annu. Rev. Biochem., 2013, 82, 663– 692. 6 J. H. Hurley, EMBO J., 2015, 34, 2398–2407. 7 M. C. Field, A. Sali and M. P. Rout, J. Cell Biol., 2011, 193, 963–972. 8 J. Huotari and A. Helenius, EMBO J., 2011, 30, 3481–3500. 9 L. Johannes, R. G. Parton, P. Bassereau and S. Mayor, Nat. Rev. Mol. Cell Biol., 2015, 16, 311–321. 10 M. J. Clague, H. Liu and S. Urbé, Dev. Cell, 2012, 23, 457–467. 11 B. A. Davies, J. R. E. Lee, A. J. Oestreich and D. J. Katzmann, Chem. Rev., 2009, 109, 1575–1586. 12 C. Bissig and J. Gruenberg, Trends Cell Biol., 2014, 24, 19–25. 13 C. Campsteijn, M. Vietri and H. Stenmark, Curr. Opin. Cell Biol., 2016, 41, 1–8. 14 Y. Olmos and J. Carlton, Curr. Opin. Cell Biol., 2016, 38, 1–11. 15 J. G. Carlton and J. Martin-Serrano, Science, 2007, 316, 1908–1912. 16 M. F. Baietti, Z. Zhang, E. Mortier, A. Melchior, G. Degeest, A. Geeraerts, Y. Ivarsson, F. Depoortere, C. Coomans, E. Vermeiren, P. Zimmermann and G. David, Nat. Cell Biol., 2012, 14, 677–685. 17 B. Strack, A. Calistri, S. Craig, E. Popova and H. G. Göttlinger, Cell, 2003, 114, 689–699. 18 J. Martin-Serrano, A. Yaravoy, D. Perez-Caballero and P. D. Bieniasz, Proc. Natl. Acad. Sci., 2003, 100, 12414–12419. 19 A. J. Jimenez, P. Maiuri, J. Lafaurie-Janvore, S. Divoux, M. Piel and F. Perez, Science, 2014, 343, 1247136. 20 Y. Olmos, L. Hodgson, J. Mantell, P. Verkade and J. G. Carlton, Nature, 2015, 522, 236–239. 21 N. Loncle, M. Agromayor, J. Martin-Serrano and D. W. Williams, Sci. Rep., , DOI:10.1038/srep08461. 22 C. K. Raymond, I. Howald-Stevenson, C. A. Vater and T. H. Stevens, Mol. Biol. Cell, 1992, 3, 1389–1402. 23 G. Prag, H. Watson, Y. C. Kim, B. M. Beach, R. Ghirlando, G. Hummer, J. S. Bonifacino and J. H. Hurley, Dev. Cell, 2007, 12, 973–986. 24 X. Ren, D. P. Kloer, Y. C. Kim, R. Ghirlando, L. F. Saidi, G. Hummer and J. H. Hurley, Structure, 2009, 17, 406–416. 25 C. Raiborg, B. Bremnes, A. Mehlum, D. J. Gillooly, A. D’Arrigo, E. Stang and H. Stenmark, J. Cell Sci., 2001, 114, 2255–2263. 26 M. J. Clague and S. Urbé, Trends Cell Biol., 2003, 13, 603–606. 27N. Bishop and P. Woodman, J. Biol. Chem., 2001, 276, 11735–11742. 28 K. G. Bache, T. Slagsvold, A. Cabezas, K. R. Rosendal, C. Raiborg and H. Stenmark, Mol. Biol. Cell, 2004, 15, 4337–4346.

45 29 M. D. Stuchell, J. E. Garrus, B. Müller, K. M. Stray, S. Ghaffarian, R. McKinnon, H.-G. Kräusslich, S. G. Morham and W. I. Sundquist, J. Biol. Chem., 2004, 279, 36059–36071. 30 S. W. Eastman, J. Martin-Serrano, W. Chung, T. Zang and P. D. Bieniasz, J. Biol. Chem., 2005, 280, 628–636. 31 E. Morita, V. Sandrin, S. L. Alam, D. M. Eckert, S. P. Gygi and W. I. Sundquist, Cell Host Microbe, 2007, 2, 41–53. 32 M. S. Kostelansky, C. Schluter, Y. Y. C. Tam, S. Lee, R. Ghirlando, B. Beach, E. Conibear and J. H. Hurley, Cell, 2007, 129, 485–498. 33 H. Teo, D. J. Gill, J. Sun, O. Perisic, D. B. Veprintsev, Y. Vallis, S. D. Emr and R. L. Williams, Cell, 2006, 125, 99–111. 34 F. Stefani, L. Zhang, S. Taylor, J. Donovan, S. Rollinson, A. Doyotte, K. Brownhill, J. Bennion, S. Pickering-Brown and P. Woodman, Curr. Biol., 2011, 21, 1245–1250. 35 M. Agromayor, N. Soler, A. Caballe, T. Kueck, S. M. Freund, M. D. Allen, M. Bycroft, O. Perisic, Y. Ye, B. McDonald, H. Scheel, K. Hofmann, S. J. D. Neil, J. Martin-Serrano and R. L. Williams, Structure, 2012, 20, 414–428. 36 H. Teo, O. Perisic, B. González and R. L. Williams, Dev. Cell, 2004, 7, 559–569. 37 A. Hierro, J. Sun, A. S. Rusnak, J. Kim, G. Prag, S. D. Emr and J. H. Hurley, Nature, 2004, 431, 221–225. 38 Y. J. Im and J. H. Hurley, Dev. Cell, 2008, 14, 902–913. 39 T. Slagsvold, R. Aasland, S. Hirano, K. G. Bache, C. Raiborg, D. Trambaiolo, S. Wakatsuki and H. Stenmark, J. Biol. Chem., 2005, 280, 19600–19606. 40 D. Teis, S. Saksena, B. L. Judson and S. D. Emr, EMBO J., 2010, 29, 871–883. 41 Y. Lin, L. A. Kimpler, T. V. Naismith, J. M. Lauer and P. I. Hanson, J. Biol. Chem., 2005, 280, 12799–12809. 42 S. Shim, L. A. Kimpler and P. I. Hanson, Traffic, 2007, 8, 1068–1079. 43 S. Lata, M. Roessle, J. Solomons, M. Jamin, H. G. Gottlinger, D. I. Svergun and W. Weissenhorn, J. Mol. Biol., 2008, 378, 818–827. 44 M. Bajorek, H. L. Schubert, J. McCullough, C. Langelier, D. M. Eckert, W.-M. B. Stubblefield, N. T. Uter, D. G. Myszka, C. P. Hill and W. I. Sundquist, Nat. Struct. Mol. Biol., 2009, 16, 754–762. 45 J. J. Skalicky, J. Arii, D. M. Wenzel, W.-M. B. Stubblefield, A. Katsuyama, N. T. Uter, M. Bajorek, D. G. Myszka and W. I. Sundquist, J. Biol. Chem., 2012, 287, 43910– 43926. 46 D. Teis, S. Saksena and S. D. Emr, Dev. Cell, 2008, 15, 578–589. 47 T. Wollert, C. Wunder, J. Lippincott-Schwartz and J. H. Hurley, Nature, 2009, 458, 172–177. 48 J. McCullough, A. K. Clippinger, N. Talledge, M. L. Skowyra, M. G. Saunders, T. V. Naismith, L. A. Colf, P. Afonine, C. Arthur, W. I. Sundquist, P. I. Hanson and A. Frost, Science, 2015, 350, 1548–1551. 49 G. Fabrikant, S. Lata, J. D. Riches, J. A. G. Briggs, W. Weissenhorn and M. M. Kozlov, PLOS Comput. Biol., 2009, 5, e1000575. 50 Q.-T. Shen, A. L. Schuh, Y. Zheng, K. Quinney, L. Wang, M. Hanna, J. C. Mitchell, M. S. Otegui, P. Ahlquist, Q. Cui and A. Audhya, J Cell Biol, 2014, 206, 763–777. 51 N. Chiaruttini, L. Redondo-Morata, A. Colom, F. Humbert, M. Lenz, S. Scheuring and A. Roux, Cell, 2015, 163, 866–879. 52 J. Schöneberg, I.-H. Lee, J. H. Iwasa and J. H. Hurley, Nat. Rev. Mol. Cell Biol., 2017, 18, 5–17.

46 53 C. P. Hill and M. Babst, Biochim. Biophys. Acta-Mol. Cell Res., 2012, 1823, 172– 181. 54 Z. Yu, M. D. Gonciarz, W. I. Sundquist, C. P. Hill and G. J. Jensen, J. Mol. Biol., 2008, 377, 364–377. 55 J. Kim, S. Sitaraman, A. Hierro, B. M. Beach, G. Odorizzi and J. H. Hurley, Dev. Cell, 2005, 8, 937–947. 56 J. McCullough, R. D. Fisher, F. G. Whitby, W. I. Sundquist and C. P. Hill, Proc. Natl. Acad. Sci., 2008, 105, 7687–7691. 57 Y. Kimura, M. Tanigawa, J. Kawawaki, K. Takagi, T. Mizushima, T. Maeda and K. Tanaka, Eukaryot. Cell, 2015, 14, 976–982. 58 L. Wunderley, K. Brownhill, F. Stefani, L. Tabernero and P. Woodman, J. Cell Sci., 2013, jcs.140673. 59 N. Ali, L. Zhang, S. Taylor, A. Mironov, S. Urbé and P. Woodman, Curr. Biol., 2013, 23, 453–461. 60 G. Odorizzi, D. J. Katzmann, M. Babst, A. Audhya and S. D. Emr, J. Cell Sci., 2003, 116, 1893–1903. 61 S. Tang, N. J. Buchkovich, W. M. Henne, S. Banjade, Y. J. Kim and S. D. Emr, eLife, 2016, 5, e15507. 62 J. H. Boysen and A. P. Mitchell, Mol. Biol. Cell, 2006, 17, 1344–1353. 63 N. Pashkova, L. Gakhar, S. Winistorfer and R. Piper, Mol. Biol. Cell. 64 N. Pashkova, L. Gakhar, S. C. Winistorfer, A. B. Sunshine, M. Rich, M. J. Dunham, L. Yu and R. C. Piper, Dev. Cell, 2013, 25, 520–533. 65 W. Xu, F. J. Smith, R. Subaran and A. P. Mitchell, Mol. Biol. Cell, 2004, 15, 5528– 5537. 66 W. Li and A. P. Mitchell, Genetics, 1997, 145, 63–73. 67 P. Vito, L. Pellegrini, C. Guiet and L. D’Adamio, J. Biol. Chem., 1999, 274, 1533– 1540. 68 M. Missotten, A. Nichols, K. Rieger and R. Sadoul, Cell Death Differ., 1999, 6, 124–129. 69 U. K. von Schwedler, M. Stuchell, B. Müller, D. M. Ward, H.-Y. Chung, E. Morita, H. E. Wang, T. Davis, G.-P. He, D. M. Cimbora, A. Scott, H.-G. Kräusslich, J. Kaplan, S. G. Morham and W. I. Sundquist, Cell, 2003, 114, 701–713. 70 A. Bobrie, M. Colombo, G. Raposo and C. Théry, Traffic, 2011, 12, 1659–1668. 71 R. D. Fisher, H.-Y. Chung, Q. Zhai, H. Robinson, W. I. Sundquist and C. P. Hill, Cell, 2007, 128, 841–852. 72 S. Lee, A. Joshi, K. Nagashima, E. O. Freed and J. H. Hurley, Nat. Struct. Mol. Biol., 2007, 14, 194–199. 73 Q. Zhai, R. D. Fisher, H.-Y. Chung, D. G. Myszka, W. I. Sundquist and C. P. Hill, Nat. Struct. Mol. Biol., 2008, 15, 43–49. 74 Q. Zhai, M. B. Landesman, H. Robinson, W. I. Sundquist and C. P. Hill, J. Virol., 2011, 85, 632–637. 75 S. Sun, X. Zhou, W. Zhang, G. E. Gallick and J. Kuang, Biochem. J., 2015, 466, 475–487. 76 R. Pires, B. Hartlieb, L. Signor, G. Schoehn, S. Lata, M. Roessle, C. Moriscot, S. Popov, A. Hinz, M. Jamin, V. Boyer, R. Sadoul, E. Forest, D. I. Svergun, H. G. Göttlinger and W. Weissenhorn, Structure, 2009, 17, 843–856. 77 Q. Zhai, M. B. Landesman, H.-Y. Chung, A. Dierkers, C. M. Jeffries, J. Trewhella, C. P. Hill and W. I. Sundquist, J. Virol., 2011, 85, 9222–9226.

47 78 X. Zhou, S. Pan, L. Sun, J. Corvera, S.-H. Lin and J. Kuang, Biochem. J., 2008, 414, 215–220. 79 X. Zhou, S. Pan, L. Sun, J. Corvera, Y.-C. Lee, S.-H. Lin and J. Kuang, Biochem. J., 2009, 418, 277–284. 80 X. Zhou, J. Si, J. Corvera, G. E. Gallick and J. Kuang, Biochem. J., 2010, 432, 525– 538. 81 S. Sun, L. Sun, X. Zhou, C. Wu, R. Wang, S.-H. Lin and J. Kuang, Dev. Cell, 2016, 36, 331–343. 82 A. Doyotte, A. Mironov, E. McKenzie and P. Woodman, Proc. Natl. Acad. Sci., 2008, 105, 6308–6313. 83 H. Ma, P. Wardega, D. Mazaud, A. Klosowska-Wardega, A. Jurek, U. Engström, J. Lennartsson and C.-H. Heldin, Cell. Signal., 2015, 27, 2209–2219. 84 M. Mariotti, S. Castiglioni, J. M. Garcia-Manteiga, L. Beguinot and J. A. M. Maier, Int. J. Biochem. Cell Biol., 2009, 41, 687–693. 85 D. Kharitidi, P. M. Apaja, S. Manteghi, K. Suzuki, E. Malitskaya, A. Roldan, M.-C. Gingras, J. Takagi, G. L. Lukacs and A. Pause, Cell Rep., 2015, 13, 599–609. 86 S. Toyooka, M. Ouchida, Y. Jitsumori, K. Tsukuda, A. Sakai, A. Nakamura, N. Shimizu and K. Shimizu, Biochem. Biophys. Res. Commun., 2000, 278, 671–678. 87 L. Cao, L. Zhang, P. Ruiz-Lozano, Q. Yang, K. R. Chien, R. M. Graham and M. Zhou, J. Biol. Chem., 1998, 273, 21077–21083. 88 N. K. Tonks, FEBS J., 2013, 280, 346–378. 89 A. J. Barr, E. Ugochukwu, W. H. Lee, O. N. F. King, P. Filippakopoulos, I. Alfano, P. Savitsky, N. A. Burgess-Brown, S. Müller and S. Knapp, Cell, 2009, 136, 352–363. 90 M.-C. Gingras, Y. L. Zhang, D. Kharitidi, A. J. Barr, S. Knapp, M. L. Tremblay and A. Pause, PLoS ONE, 2009, 4, e5105. 91 L. Li and S. N. Cohen, Cell, 1996, 85, 319–329. 92 Z. Xu, L. Liang, H. Wang, T. Li and M. Zhao, Biochem. Biophys. Res. Commun., 2003, 311, 1057–1066. 93 Y. Cheng, N. E. Poulos, M. L. Lung, G. Hampton, B. Ou, M. I. Lerman and E. J. Stanbridge, Proc. Natl. Acad. Sci., 1998, 95, 3042–3047. 94 M.-C. Gingras, D. Kharitidi, V. Chenard, N. Uetani, M. Bouchard, M. L. Tremblay and A. Pause, Int. J. Dev. Biol., 2009, 53, 1069–1074. 95 S. Manteghi, M.-C. Gingras, D. Kharitidi, L. Galarneau, M. Marques, M. Yan, R. Cencic, F. Robert, M. Paquet, M. Witcher, J. Pelletier and A. Pause, Cell Rep., 2016, 15, 1893–1900. 96 G. Lin, V. Aranda, S. K. Muthuswamy and N. K. Tonks, Genes Dev., 2011, 25, 1412–1425. 97 D.-Y. Chen, M.-Y. Li, S.-Y. Wu, Y.-L. Lin, S.-P. Tsai, P.-L. Lai, Y.-T. Lin, J.-C. Kuo, T.-C. Meng and G.-C. Chen, J. Cell Sci., 2012, 125, 4841–4852. 98 D. Wang, C.-C. Chan, S. Cherry and P. R. Hiesinger, Cell. Mol. Life Sci., 2013, 70, 2919–2934. 99 G. Skibinski, N. J. Parkinson, J. M. Brown, L. Chakrabarti, S. L. Lloyd, H. Hummerich, J. E. Nielsen, J. R. Hodges, M. G. Spillantini, T. Thusgaard, S. Brandner, A. Brun, M. N. Rossor, A. Gade, P. Johannsen, S. A. Sørensen, S. Gydesen, E. M. Fisher and J. Collinge, Nat. Genet., 2005, 37, 806–808. 100 J.-A. Lee, A. Beigneux, S. T. Ahmad, S. G. Young and F.-B. Gao, Curr. Biol., 2007, 17, 1561–1567. 101 S. Rollinson, P. Rizzu, S. Sikkink, M. Baker, N. Halliwell, J. Snowden, B. J. Traynor, D. Ruano, N. Cairns, J. D. Rohrer, S. Mead, J. Collinge, M. Rossor, E. Akay, 48 R. Guerreiro, R. Rademakers, K. E. Morrison, P. Pastor, E. Alonso, P. Martinez-Lage, N. Graff-Radford, D. Neary, P. Heutink, D. M. A. Mann, J. Van Swieten and S. M. Pickering-Brown, Neurobiol. Aging, 2009, 30, 656–665. 102 J. S. Snowden, D. Neary and D. M. A. Mann, Br. J. Psychiatry, 2002, 180, 140– 143. 103 H. Chanut-Delalande, A. C. Jung, M. M. Baer, L. Lin, F. Payre and M. Affolter, PLOS ONE, 2010, 5, e10245. 104 D. R. Scoles, V. D. Nguyen, Y. Qin, C.-X. Sun, H. Morrison, D. H. Gutmann and S.-M. Pulst, Hum. Mol. Genet., 2002, 11, 3179–3189. 105 R.-T. Liu, C.-C. Huang, H.-L. You, F.-F. Chou, C.-C. A. Hu, F.-P. Chao, C.-M. Chen and J.-T. Cheng, Oncogene, 2002, 21, 4830–4837. 106 T. W. Young, D. G. Rosen, F. C. Mei, N. Li, J. Liu, X.-F. Wang and X. Cheng, Clin. Cancer Res. Off. J. Am. Assoc. Cancer Res., 2007, 13, 3848–3854. 107 K. B. Oh, M. J. Stanton, W. W. West, G. L. Todd and K.-U. Wagner, Oncogene, 2007, 26, 5950–5959. 108 X. R. Ma, U. H. Edmund Sim, B. Pauline, L. Patricia and J. Rahman, Trop. Biomed., 2008, 25, 46–57. 109 L. Walker, PhD Thesis, University of Manchester, 2014. 110 W. Shi, C. Chang, S. Nie, S. Xie, M. Wan and X. Cao, J. Cell Sci., 2007, 120, 1216–1224. 111 E. Boura, B. Różycki, D. Z. Herrick, H. S. Chung, J. Vecer, W. A. Eaton, D. S. Cafiso, G. Hummer and J. H. Hurley, Proc. Natl. Acad. Sci., 2011, 108, 9437–9442. 112 S. Tang, W. M. Henne, P. P. Borbat, N. J. Buchkovich, J. H. Freed, Y. Mao, J. C. Fromme and S. D. Emr, eLife, 2015, 4, e12548.

49 Chapter 2: Methodology

The five results chapters of this thesis are in journal format so contain a wide range of methods, some of which were performed by various co-authors. This section gives background information on the methods that were performed by GH, as highlighted in author-contributions in Chapter 3.

2.1 EPR spectroscopy

Electron paramagnetic resonance (EPR) spectroscopy is a technique for measuring the magnetic properties of unpaired electrons. This can be applied to a range of biological systems including free radicals, metalloenzymes, redox active organic cofactors and, increasingly, proteins which have been artificially modified by spin labelling.1

For studying the structure of proteins, unpaired electrons (or spins) can be artificially introduced by reacting exposed cysteine residues with thiol reactive nitroxide “spin labels” such as MTSL (methanethiosulfonate spin label) (Figure 2.1A). The unpaired electron in this label is situated in the nitrogen oxygen bond and is protected from reduction by the two adjacent germinal dimethyl groups.2

MTSL labels can be used to give a range of information on protein structure. The continuous-wave (CW) spectrum can be used to monitor local flexibility by singly labelling different sites across a protein sequence;3 MTSL labels are also sometimes used in conjunction with NMR (nuclear magnetic resonance) spectroscopy where the paramagnetic centre quenches nearby NMR signals measured using paramagnetic relaxation enhancement experiments;4 but one of the most widely used experiments is double electron-electron resonance (DEER) spectroscopy for the determination of intramolecular distance measurements between spin labelled cysteine pairs.5–7

50 2.1.1 Continuous-wave EPR

The origin of a nitroxide EPR signal comes from its unpaired electron (electron spin, S =

1/2) situated in the nitrogen-oxygen bond.8,9 In a continuous-wave (CW) EPR experiment, the sample is situated in a magnetic field and microwave frequency radiation is applied. Electron spins have an associated magnetic dipole moment which acts in the opposite direction to the spin. In an external magnetic field, magnetic dipole moments will align themselves so that they are either parallel (mS = -1/2) or antiparallel (mS = +1/2) to the external magnetic field. The energy, E, of each of these states is given by:

E = mSgβeH (Equation 2.1) where g is the g-value for the electron (which is variable but ranges between 2.003 and

-24 -1 2.006 for nitroxide radicals), βe is the electronic Bohr Magneton (9.2740x10 J T ) and

H is magnetic field of the EPR magnet. The energy will be lower when mS = -1/2, so a slight excess of electrons will be in this state. In CW EPR, a fixed microwave frequency radiation is applied and the magnetic field is swept; when the field strength separates the energy levels to the same energy as the applied microwaves, the radiation is absorbed, and resonance can be detected, i.e. the following must be satisfied:

hv = gβeH (Equation 2.2) where h is Planck’s constant (6.6261x10-34 J T-1) and v is the microwave frequency, which is fixed, whilst the magnetic field, H, is varied.

Nitroxide signals are affected by hyperfine splitting, A, due to electron spin interacting with nuclear spins of the 14N nitrogen atom (with nuclear quantum number, I = 1). This means that in addition to the mS = -1/2 or +1/2 states, the angular momentum of the nuclear spin mI = -1, 0 or +1, causes splitting to give a total of six states (Figure 2.1B).

51 EPR transitions are only allowed that satisfy the selection rules ΔmS=±1 and ΔmI=0. For an isotropic nitroxide this results in three resonance positions.

Due to the field modulation with phase sensitive detection method used to measure CW

EPR, the spectra appear as first derivatives of the corresponding absorbtion spectra: i.e. each resonance has as a positive peak (positive absorbance slope), a crossing through the baseline (maximum of absorbance peak), then a negative trough (negative absorbance slope).

The three line nitroxide EPR spectrum is also affected by the rate of tumbling. A nitroxide molecule is not symmetrical so has different g and A values in three cartesian directions gx, gy, gz and Ax, Ay, and Az. When a molecule is tumbling in solution fast enough these values average together, giving the isotropic values giso and Aiso (Figure 2.1B). However, as molecular tumbling slows down (increased rotational correlation time, τc), the spectrum becomes more anisotropic. These effects are particularly apparent in spin

-10 labelled proteins, as free spin labels have τc ≈ 10 s giving a mostly isotropic spectrum,

-8 whereas labels attached to large proteins have τc in range of 10 s resulting in broadening of the spectrum (Figure 2.1C).

Figure 2.1. MTSL labelling. (A) MTSL disulfide bond formation with a cysteine thiol sidechain. (B) Six energy levels of a S = 1/2 electron coupled to a nitrogen (I = 1) nucleus in a magnetic field, B0. Transitions allowed by the two selection rules ΔmS = ±1 and ΔmI = 0, resulting in three “line” field swept EPR spectrum. Isotropic EPR spectrum of MTSL nitroxide. (C) Anisotropic line broadening of nitroxide spectrum with increasing 10 rotational correlation time, τc, simulated using EasySpin.

53 2.1.2 Pulsed EPR

Pulsed EPR experiments use a range of microwave frequencies which switch on for set periods of time defined by pulse sequences. At equilibrium, the net magnetisation of a sample will be parallel to the magnetic field i.e. positive direction on the Z-axis. When a resonant frequency radiation is applied, spins respond in bulk by tipping into the X-Y plane. A π/2 pulse results in a z magnetisation of 0 and a π pulse completely inverts the population so that the magnetisation is transferred to the negative position on the Z-axis.

The π pulse also has the effect of rephasing spins: spin packets reverse their direction and rephase, resulting in a “spin echo”. The intensity of these echos can be monitored by integration in pulsed EPR experiments.

2.1.3 Distance measurements by double electron-electron resonance spectroscopy

Double electron-electron resonance (DEER) spectroscopy is a pulsed EPR method for measuring distances through detection of dipolar coupling interactions.5–7 It is often used in combination with site directed spin labelling: mutagenesis to produce pairs of cysteines, then, spin labelling of the recombinant protein.

The technique takes advantage of the distance dependent dipolar coupling interaction between unpaired electrons or “spins”. The splitting of energy levels due to dipolar coupling is often smaller than the EPR spectral line widths, so manifests as line broadening. Even so, this broadening is only detectable for relatively strong interactions at short separations (1.0 - 2.5 nm).9 For longer distances (1.7 - 10 nm)6,11–13 this interaction can be determined using pulsed techniques. Several methods for this have been developed which include DQC (double quantum coherence)14 and RIDME

(relaxation-induced dipolar modulation enhancement).15,16 However, for nitroxide- nitroxide distances on a standard pulse EPR instrument, DEER is the most widely applicable. Variants of DEER have been developed including three-,5 four-,6 and five- pulse17 DEER, but the four-pulse method is currently the most common. 54 Four pulse DEER uses two different frequencies across the nitroxide spectrum: the pump frequency and the observe frequency (Figure 2.2).6 The pump frequency pulse is used to flip the spins of the partner labels, and the effect of this on the observe label’s spins are monitored by “observation” pulses.

Figure 2.2. DEER spectroscopy. (A) 4-pulse DEER sequence. (B) Echo detected field sweep spectrum of a MTSL labelled protein, with pump and observe field positions indicated. It is also possible to set up DEER experiments with these positions swapped over to optimise modulation depth at the cost of signal to noise. (C) Raw DEER trace with calculated exponential background shown as a dotted red line. (D) Fourier transform of background corrected DEER trace in the frequency domain. (E) Distance distribution calculated from background corrected DEER trace.

55 The observation pulse sequence consists of a π/2 flip pulse, followed by a π rephasing pulse, then a final π rephrasing pulse (Figure 2.2A). At the pump frequency, a single π pulse is applied, but the delay of this pulse is incremented over repetitions of the pulse sequence within a window length of t = 0 to t = tmax. The echo intensity of the observer frequency refocussed echo is recorded, which modulates with respect to the pump time delay, t. The pump and observe frequencies are chosen to correspond to the most intense positions on the field sweep spectrum (Figure 2.2B), with either the observe pulse at the most intense point to maximise signal-to-noise or the pump pulse at the field maximum to optimise modulation depth.

The raw DEER trace (stimulated echo height vs t), V(t) (Figure 2.2C), can be expressed as a product of a background factor, B(t) and form factor, F(t):18

V(t) = B(t) F(t) (Equation 2.3)

The background factor, B(t), represents the intermolecular dipolar interactions whilst the form factor, F(t) represents the intramolecular dipolar interactions, which are usually of interest in DEER experiments. The background factor for soluble proteins can be modelled as a three-dimensional exponential decay and is most easily modelled when the signal is recorded until after tdecay when the signal is dominated by the background factor:

V (t>tdecay) = (1 – Δ) B(t) (Equation 2.4)

B(t) = exp(-kt) (Equation 2.5)

Background can be modelled, in this case, by optimising two parameters: Δ the modulation depth and k which is related to the concentration of spins. Once background is removed, analysis of the form factor allows the dipolar coupling between spin label pairs to be determined. Fourier transform into the frequency domain gives a spectrum

56 characteristic of dipolar interactions known as the Pake pattern, allowing direct readout of the dipolar coupling frequency peak (Figure 2.2D). However, often more detailed information is extracted from DEER data by conversion to a probable inter-spin distance distribution P(r) (Figure 2.2E).18

Transformation of F(t) to P(r) is described as “moderately ill-posed”, which means that the solutions are ambiguous: many P(r) distributions could describe the same F(t).

However, back calculation of F(t) from P(r) is relatively straightforward and unambiguous so the P(r) function can be optimised using a protocol known as Tikhonov regularisation, where the target function Gα(P) is minimized:

2 푑2 G (P) = ‖KP(r) − F(t)‖2 + α ‖ 푃(푟)‖ (Equation 2.6) 훼 푑푟2 where the first term is the mean squared deviation between calculated (KP) and experimental (F) form factors (smaller = better fit) and the second term is the square norm of the second derivative of the distance distribution (small = smoother peaks). Weighting of the smoothness term by α prevents overfitting of noise and avoids unrealistically narrow peaks in the distance distribution.

Measurement of longer distances requires a longer window length to be able to observe the modulating DEER signal. The tmax for the pump pulse is limited by the length of the

τ2 separation in the observer pulse sequence, which decreases the amplitude of the observed echo, increasing noise. In practice, a compromise must be made between signal- to-noise and window length. Longer τ2 is possible when the phase memory time, Tm, of the system is increased, which is easiest to achieve by adding glycerol (preferably deuterated) to the sample and using deuterated water. Glycerol increases viscosity, slowing tumbling, and substitution for deuterons removes the spin relaxation pathway through hyperfine interaction with proton nuclei. Full deuteration of the protein sample 57 can be achieved by feeding E.coli with deuterated media, and using this method, remarkably long window lengths have been reported allowing measurement of very long distances (10 nm).13

2.1.4 Simulating interspin distance distributions

DEER measures interspin distances, which is regarded to be the distance between the midpoints of the nitroxide N-O bond. An MTSL side chain is roughly 0.7 nm long so, unless crystal structures of spin labelled proteins are available, simulations are necessary to predict interspin distributions for comparison with data.

A range of software has been developed for in silico spin labelling including MMM,19

MTSSLWizard20 and RosettaEPR.21 These three programmes use rotamer libraries based on known MTSL-labelled protein crystal structures and further intense simulations to predict the full flexibility of the MTSL side chain, which can then be used to probe the local environment of the input model and select possible non-clashing conformations.

Alternatively, molecular dynamics software such as CNS22 can be used which can be useful if carrying out other structural simulations.

2.2 Paramagnetic relaxation enhancement

In NMR, proteins containing spin labels can be used to provide structural information using paramagnetic relaxation enhancement (PRE) experiments. This takes advantage of the shortened relaxation times of NMR active nuclei near the unpaired electrons of spin labels, due to distance dependent dipole-dipole interactions between the nuclear spins and unpaired electrons.23

A PRE experiment with a nitroxide spin labelled protein is carried out by making NMR measurements of the paramagnetic sample; afterwards the nitroxide group is reduced to a hydroxylamine group, by adding a reducing agent such as ascorbic acid, to allow

58 measurements of the diamagnetic sample.4 By measuring the difference in relaxation times between reduced and unreduced spin labelled samples, PRE experiments can be used quantitatively to produce distance constraints to supplement 3D structural determination by NMR.23 Alternatively, with the knowledge that PRE effects are strongest for distances under 40 Å,23 PRE experiments can be used to qualitatively assign nuclei which are close to the spin label by observing which NMR signals are missing in the paramagnetic spectra.24

The qualitative approach to PRE experiments was used in this work, by measuring 1H

NMR and 1H-13C HSQC (heteronuclear single quantum coherence) spectra of nitroxide spin labelled samples. This complemented crystallography experiments, by ruling out alternative peptide orientations which could also be modelled into the electron density map.

2.3 Labelling confirmation by intact protein mass spectrometry

Electrospray ionisation (ESI) mass spectrometry is commonly used to confirm labelling of proteins by quantifying the mass of intact proteins before and after labelling. The electrospray charges the protein sample by the addition of N number of bound protons.25

Since mass spectra record the mass over charge ratio, the resulting spectrum includes a range of peaks with m/z by:

m/z = (mass of protein + N (1.0073)) / N (Equation 2.7)

Enough protonation states are present that, at reasonable resolution, adjacent clusters of peaks will correspond to sequential consecutive protonation numbers: N, N-1, N-2 etc.

(from low to high m/z). The spectra can be deconvoluted to the neutral mass by solving simultaneous equations of adjacent peaks to determine the values of N. Once N is known, it is trivial to rearrange the m/z expression to determine mass. Software for spectral

59 deconvolution can be used to assign N for each peak, determine mass, and construct an average distribution of neutral deconvoluted masses. The open source software mMass was used in this work for its ability to deconvolute spectra measured from multiple instruments.26

2.4 Monitoring structural changes using circular dichroism

Proteins are polymers of amino acid monomers which, apart from glycine, are all chiral.

This means that proteins exhibit circular dichroism (CD): a difference in the absorption

27 level of clockwise (ER) and anticlockwise (EL) circularly polarised light. The resultant summed vectors of ER and EL trace the shape ellipse. Experimentally ΔE is measured over a range of UV wavelengths:

ΔE = EL – ER (Equation 2.8) but this is usually converted to ellipticity, in radians, which is related by:

θ = 3298ΔE (Equation 2.9)

Ordered arrangement of protein backbone into alpha helical and beta sheet secondary structures give rise to characteristic CD spectra (Figure 2.3A). CD can be used to monitor protein unfolding by the loss of these characteristic features for secondary structure, for example figure 2.3B shows the effect of heat induced denaturation on the CD spectrum of Na/K transporter ATPase.28 In this work, it has been used to monitor the effects of mutagenesis and of spin labelling. An important limitation of using CD in this way, is that it is unable to detect induced changes in tertiary structure.

Figure 2.3. Example CD spectra. (A) CD spectra of an alpha helix rich protein (sensory rhodopsin 2, 81% alpha helix, 1% beta sheet) and a beta sheet rich protein (outer membrane protein G, 68% beta sheet, 1% alpha helix).29 (B) CD spectra monitoring the heat induced denaturation of Na/K transporter ATPase (34% alpha helix, 11% beta sheet) with spectra shown at 20⁰C and 80⁰C.28 CD data was obtained from the Protein Circular Dichroism Data Bank (PCDDB, http://pcddb.cryst.bbk.ac.uk).

2.5 X-ray crystallography of proteins

X-ray crystallography was the first method to allow atomic resolution three-dimensional structure determination of protein structures. At the time of writing, 90 % of current structure depositions in the protein data bank (http://www.rcsb.org/pdb/home/home.do) are from X-ray crystallography data and, despite the rise in electron microscopy, the proportion remains high with 93 % X-ray structures in the past 5 years.

Protein crystals are formed of repeating arrays proteins related by translational symmetry.

They can be formed by slow precipitation from a protein solution, usually in the presence of a precipitant such as the polymer polyethylene glycol (PEG).30

The regular spacing of atoms in the crystal causes diffraction of X-rays, which is described as being analogous to reflections of X-rays from sets of parallel Bragg planes

(Figure 2.4).31 When parallel incident X-ray approach the Bragg planes, they can penetrate the planes to differing levels. Once the deeper penetrating X-ray emerge their extra path lengths may have given them wavelengths which are out of phase to one

61 another, causing destructive interference. The Bragg condition of constructive interference is satisfied when the extra pathlength is equal to an integer multiple of the

X-ray wavelength. This only occurs at certain incident angles, which are described by the

Bragg equation:

2dsinθ=nλ (Equation 2.10)

Where d is the lattice plane spacing, n is an integer and λ is the X-ray wavelength.

Constructive interference at only certain angles gives rise to a series of spots (or reflections) known as the diffraction pattern. Each reflection in the pattern depends on the distribution of electron density in the crystal. These patterns are recorded over a range of crystal orientations and the position and intensity of these reflections form the raw data required for producing an electron density map, in which to fit a protein structural model.

Figure 2.4. The Bragg equation. The X-rays R1 and R2 will interfere constructively only at angles where the extra distance travelled by R2, when penetrating the lattice (highlighted in red), is equal to an integer multiple of the X-ray wavelength. This is described by the Bragg equation: 2dsinθ=nλ. Diagram adapted from Rhodes, 2006.31

A complete transformation from X-ray diffraction raw data to an electron density map requires knowledge of the 3D arrangement of reflections, their intensities and their phases. Unfortunately, detection of X-rays in a diffraction experiment only measures their location and intensity. This means that efforts must be made to obtain phases for these

62 reflections before an electron density map can be determined. There are various methods to approach this “phase problem” which are discussed below.

Once the electron density map has been constructed, a protein model can be placed within it. Prior knowledge of the protein sequence and likely bond angles can help to build an appropriate protein model. The process of building the protein into the map can help to improve the estimated phases, improving the quality of the electron density map; this is known as refinement. The broad workflow for the protein structure by X-ray crystallography is given in figure 2.5.

Figure 2.5. Flowchart for protein structure solution by X-ray crystallography.

63 2.5.1 Crystallisation of proteins

Protein crystallisation is a phenomenon which occurs at a border between precipitation and dissolution. It is commonly induced by a vapour diffusion approach. First water will evaporate from the protein solution, increasing protein and precipitant concentrations and encouraging nucleation. In favourable conditions, crystallisation occurs on nucleation sites and, because of this consumption of protein, soluble protein concentration in the surrounding solution falls, avoiding precipitation.30

Purification to homogeneity on an SDS-PAGE gel and a single symmetrical peak on gel filtration is the best starting point for crystallisation trials. High throughput trials are most often used to find the best conditions for crystallisation, where protein concentration, temperature, buffers, precipitant conditions, and protein: precipitant ratio can be varied.

With a limited amount of protein, it has been shown that the best approach is to prioritise screening a wide range of precipitant conditions.32

2.5.2 Tackling the phase problem in protein crystallography

X-ray crystallography gives incomplete data for direct reconstruction of an electron density map. The spots seen on an X-ray detector plate can only be used to determine X- ray intensity. But the phase of the X-rays which contributed to these spots is unknown so various approaches have been developed to get around this problem. In small molecule crystallography, the problem is simpler so usually these structures can be solved by direct methods. But in protein crystallography the problem is much more complex with some methods specific to the study of large macromolecules like proteins.

There are three main approaches to obtaining phases for protein X-ray crystallography:33 isomorphous replacement of heavy atoms, which involves reproducing crystals with and without heavy atom derivatives; anomalous scattering of heavy atoms which may be

64 intrinsic to the protein or introduced through unnatural amino acids such as selenomethionine; or molecular replacement, where a non-derivatised crystal data set can be analysed using phase estimates from previously solved structures.

2.5.3 Molecular replacement

Molecular replacement relies upon having a previously solved protein crystal structure, referred to as a search model, which is similar in shape to the current protein of interest.

Of course, it will not be certain whether the search model is similar in shape until the new structure is solved, but this method is usually effective when sequence identity is above

30 %. Consequently, usually a homologue from the same family is chosen, or the model could be the same protein if, for example, the apo structure is known and the new experiment is a complex of this protein.

A known 3D protein structure can be used to calculate structure factors, so long as a unit cell, space group and orientation of the molecule is chosen. Unit cell dimensions can be determined from the measured data, and the Patterson map (plotting a map of the data with all phases set to zero) allows the space group to be determined, or narrowed down, from symmetry relationships and systematic absences. The greater problem lies with determining the correct orientation of the search model: the range of possible space is huge and for a correct solution to be found the rmsd between search model and solution needs to be within 2 Å.34

Calculated structure factors will consist of intensities and phases. But the data only consist of intensities, so these can be used to judge the quality of fit to the data of the search model. Regardless, the number of possible orientations of the search model is immense so a number of approaches have been taken to tackle this computationally intensive procedure. The programme Phaser, which has been used in this work, separates searches

65 into rotations then translations, and uses maximum likelihood scoring function to assess the quality of fit.35,36

Once a suitable fit is found, phases can be copied over from the calculated set to give a complete initial set of experimental structure factors. These initial phases are improved by model building refinement: mutating or extending the known amino acid sequence of the target structure, manual fitting to the electron density map, and can be further guided by known geometrical constraints of protein backbone and side chain bond parameters.

Much of this processing is automated by a model building programme such as

PhenixAutobuild37 and the refine programmes Refmac38 or PhenixRefine.39 Manual editing of the three dimensional protein structure with electron density shown in three dimensional is possible using the programme Coot.40

2.6 Hydrodynamic protein measurements

An effective way to study the size and shape of proteins is to measure their hydrodynamic properties. Frictional drag affects how proteins travel through solution and depends on a property known as the hydrodynamic radius, Rh, which represents the radius of a sphere which migrates at the same rate as the sample protein. These are related by Stokes law:

f = 6πηRh (Equation 2.11)

-1 -1 where f is the coefficient of friction (g s ), η is buffer viscosity (g cm s), and Rh is hydrodynamic radius in cm.

Rh can be determined from sedimentation behaviour under a centrifugal force, which is measured in analytical ultracentrifugation. Alternatively, Rh can be determined via the diffusion coefficient which is measured using dynamic light scattering.

66 Rh can be used in protein structural modelling but the additional modelling parameter f/f0 is also often used. This represents the ratio between the observed frictional ratio and the minimum frictional coefficient, f0, which would be expected if the protein was completely compact. This is useful because it gives a measure of how asymmetric a protein is behaving in solution. f0 is determined, first, by approximation of the minimum hydrodynamic radius R0:

3 3Mrv̅ R0 = √ (Equation 2.12) 4πN0

-1 where Mr is molecular weight (in g mol , which is calculated from protein sequence assuming a known oligomerisation state), v̅ is the proteins partial specific volume (ml g-1, which can be estimated based on sequence), and N0 is Avogadro’s number (6.022 x

34 -1 10 mol ). Estimating R0 allows calculation of f0 through Stokes law (Equation 2.11), then f derived from the data divided by f0 gives the frictional ratio f/f0.

2.6.1 Analytical ultracentrifugation

A sedimentation velocity analytical ultracentrifugation (AUC) experiment measures the rate of sedimentation of a protein solution whilst spinning in a high-speed centrifuge. The distribution of protein concentration across a cell is measured by scanning UV absorbance across the length of the cell through the experiment (Figure 2.6A). A second compartment containing buffer (with slightly higher volume) is used for a background subtraction.

-2 -1 Flux of material due to sedimentation, Js (in g cm s ) given by:

2 Js = csω r (Equation 2.13) where c is protein concentration (g cm-3), s is the sedimentation coefficient (in units of seconds, but more commonly in Svedbergs, S = 10-13 s), ω is angular velocity of rotor 67 (rpm x π/30, in units of s-1) and r is the distance of the concentration boundary from the centre of the rotor (cm).

Measuring the position of the concentration boundary at time t and plotting ln r against

ω2t gives a straight-line graph where sedimentation coefficient can be determined from the gradient. Sedimentation coefficient can then be used to determine the frictional coefficient using equation 2.14, which allows calculation of Rh from stokes law (equation

2.11).

M(1 − v̅ρ) f = (Equation 2.14) N0s

Alternatively, a more rigorous approach can be used to produce a continuous distribution of sedimentation coefficients by analysis using the Lamm equation (Figure 2.6B).41 This approach is used by AUC analysis programs such as Sedfit.

Figure 2.6. AUC. (A) Example of background subtracted sedimentation velocity data. Only few time points are displayed to demonstrate the movement of the protein concentration boundary across the cell. (B) The same data, interpreted by Lamm equation fitting to give a continuous sedimentation coefficient distribution. Data analysed using Sedfit.41

2.6.2 Multi-angle light scattering

A multi-angle light scattering (MALS) instrument performs a series of measurements to determine properties of proteins purified through a size exclusion column (Figure

2.7A).42,43 UV absorbance is used to detect eluting proteins, dynamic light scattering

(DLS) is used to determine hydrodynamic radius (Rh) and refractometry is used to determine molecular weight (Mr). Separation by size exclusion allows measurements to be obtained for each elution peak which can be useful for identifying oligomers of a purified protein sample (Figure 2.7B,C).

Figure 2.7. Multi-angle light scattering (MALS). (A) Summary of experiment: samples are eluted from a size exclusion column before passing through three different detectors. (B) Example MALS experiment for a 40 kDa protein. The protein elutes as two peaks indicating a monomer and dimer. (C) Summary of data obtained from the example MALS experiment.

Like AUC, hydrodynamic radius is determined from DLS by measuring how the protein moves through a solution. But in DLS, instead of determining Rh through measurement of the sedimentation coefficient (s), Rh is determined through measurement of the diffusion coefficient (D). Light scattering signal intensity in the microsecond timescale oscillates slowly for large slow diffusing proteins (Figure 2.8Ai) and faster for smaller quickly diffusing proteins (Figure 2.8Aii). An autocorrelation function is applied to

70 determine the rate of signal dephasing, where successive pairs of intensities at time intervals are multiplied together and averaged for each time interval length:

g(t) = (Equation 2.15)

At the smallest time intervals, this mean product tends towards the mean of the average intensity squared (Equation 2.16). At the longest time intervals, this tends towards the square of the mean intensity (Equation 2.17).

lim g(t) = t→0 t (Equation 2.16)

2 limt→∞g(t) = (Equation 2.17)

2 2 will always be smaller than so plotting g(t) against t gives a decaying curve, where the more rapidly oscillating signals, corresponding to smaller proteins, decay faster

(Figure 2.8B).

Diffusion coefficient, D, is determined by:

< Γ > D = (Equation 2.18) q2 where Γ is the decay time (= 1/τ), and q is the scattering angle. Finally, D can be converted to Rh using the Stokes-Einstein equation:

kT R = (Equation 2.19) h 6πηD where k is Boltzmann constant and T is temperature (which is kept constant during the

AUC experiment).

Figure 2.8. Dynamic light scattering. (A) Dynamic light scattering intensity fluctuations for (i) large protein and (ii) small protein. (B) Autocorrelation function decays more rapidly for quickly oscillating signals.

Measurement of molecular weight from MALS is important for the determination of oligomerisation or the stoichiometry of complexes. This information is not available using standard intact mass spectrometry as quaternary interactions are lost during the conditions of the experiment, although some specialised equipment has been developed more recently to determine intact mass of complexes.44 Molecular weight is determined using refractometry.

2.7 Small angle X-ray scattering

Small angle X-ray scattering (SAXS) experiments allow measurement of the size and shape of proteins in solution.45 X-rays are shone through a protein solution, and electron density in the protein causes X-ray scattering, which is recorded by an X-ray detector.

A SAXS experiment consists of two measurements, a protein sample measurement and a background measurement for subtraction, both of which must have good signal-to-noise

72 ratio for analysis. The data consist of a curve of X-ray intensity, I(q), vs scattering angle, q. The scattering occurs in two dimensions across the plate, but is averaged into one radial measurement.

Analysis of SAXS data can yield a large amount of information on the size, shape and flexibility of proteins. In this work, SAXS data was analysed as summarised in figure 2.9: first, the data was inspected for evidence of protein flexibility; then parameters for structural modelling were obtained by Guinier analysis and real space analysis; finally, structural models were generated using ab initio methods and rigid body modelling whenever suitable 3D models were available.

Figure 2.9. SAXS data analysis workflow.

73 2.7.1 Flexibility analysis

The shape of a SAXS curve can give information about the flexibility of the samples. A useful way of interpreting this is to plot the data raised to different powers.46 According to the Porod-Debye law, when plotted as q4 x I(q) vs q4, SAXS data for a rigid protein will form an asymptotic plateau; plotted as q3 x I(q) vs q3, data for partially flexible proteins will form a plateau; and plotted as q2 x I(q) vs q2, data for completely flexible proteins will form a plateau.

2.7.2 Radius of gyration

SAXS can be used to determine the radius of gyration, Rg, of a protein. Rg is the weight average radius from all atoms to the protein’s centre of mass. This is different to the radius of hydration, Rh, but it will be in the same order of magnitude. Rg can be used along with

Rh as an additional modelling parameter.

Rg can be determined from the low q portion of a SAXS curve (up to q.Rg<1.3), which is known as the Guinier region. In this region, scattered X-ray intensity is given by:45

2 2 −q Rg I(q) = I(0)exp ( ) (Equation 2.20) 3 where I(0) is the extrapolated intensity at q=0. This can be rearranged to:

2 Rg ln I(q) = ln I(0) − q2 (Equation 2.21) 3

2 Therefore, plotting SAXS Guinier region as lnI(q) vs q allows Rg to be determined from the slope.

74 2.7.3 Real space atom pair distribution

SAXS data can be transformed into a real space atom pair distance distribution by Fourier transform of the X-ray scattering intensity.47 The shape of the P(r) distribution gives an indication of asymmetry and gives a maximum distance (Dmax) at the point where the P(r) curve intersects with the x-axis. P(r) fitting relies upon some human judgement, Dmax must be chosen manually, and a balance must be sought between a smoothness of the distribution and the quality of fit. This represents the same issue encountered with DEER distance distribution fitting (Section 2.1.2), and for this reason, Tikhonov regularisation procedure is also used to minimise the function, Φ:

Φ = Χ2 + αP(p) (Equation 2.22) where X2 is a function of the fitting of the calculated data to the measured data, and P(p) is a function of the smoothness of the distance distribution. α, is the weighting factor for the second function to avoid overfitting of noise and to create a biologically feasible distribution. The P(r) can also be verified by calculation of real space Rg and I(0) for comparison with those determined by Guinier fitting.

2.7.4 Molecular modelling with SAXS data

There are two main approaches to model building with SAXS data: ab initio modelling, using only SAXS data; and rigid-body modelling, using structures generated from other techniques such as X-ray crystallography or NMR (nuclear magnetic resonance) spectroscopy.

Ab initio modelling by programmes the programmes DAMMIN48 and DAMMIF49 use a simulated annealing procedure to produce a bead model to fit the SAXS data. The simulations begin with a large sphere of beads, which is then altered to minimise the

75 deviation between calculated SAXS curve and measured data. Again, like P(r) fitting, a regularisation procedure is used to avoid over-fitting, by minimising the function:48

f(x) = X2 + αP(X) (Equation 2.23)

Again, Χ2 is the quality of fit between the simulated and measured SAXS curve. In this case, P(X) is a measure of “looseness” of the bead model, making the role of this weighting factor α to avoid ab initio models which are unrealistically compact.

Rigid body modelling again relies upon methods to simulate the SAXS scattering curve from molecular models. A useful resource for this modelling is the FoXS webserver.50,51

In this work, this server has been used in combination with generating many hundreds of possible models using the molecular dynamics programme CNS.22 Generating rigid body models assumes that the input domains are invariant in structure, but is necessary to be able to interpret molecular details of the proteins. It also allows calculation of hydrodynamic properties such as s20,w, f/f0 and Rh using a programme such as SOMO, which allows models to be generated, which are altogether consistent with SAXS, MALS and AUC.

2.8 Summary of methods

The methods described in this chapter can be used together to provide complementary information needed to fully understand the structure and mechanism of HD-PTP (Figure

2.10).

X-ray crystallography gives the most highly detailed structural description of protein structure and would always be desirable to obtain. However, it lacks detail in disordered protein regions and is often only applicable to individual domains as large multidomain proteins can suffer from poor crystallisability and more difficult phase solutions.

76 SAXS only gives low resolution protein structures but can handle large multidomain proteins. Samples for SAXS are in solution which, as well as requiring simple sample preparation compared to crystallography, also means that multiple protein conformations can be present which allows information on flexibility to be interpreted.

DEER spectroscopy can, selectively, give high resolution measurements of individual distances and, since this method involves measurements of a frozen sample, DEER is also virtually unaffected by the size of the protein. However, to prepare samples for DEER, mutagenesis needs to be carried out followed by separate expression and purification of mutants for site directed spin labelling. Furthermore, larger proteins tend to have more cysteines which means numerous rounds of mutagenesis are required before cysteine pairs can be introduced for labelling.

AUC and MALS both give hydrodynamic radius; in AUC, this is measured by protein sedimentation under a centrifugal field and in MALS, this is measured by protein diffusion which is detected by dynamic light scattering. Both methods separate out mixtures allowing a distribution of species (monomers, dimers, oligomers, aggregates) to be detected. In AUC, the high concentration induced by sedimentation to the sample cell wall allows higher oligomers to be detected which may be lost in a MALS experiment due to sample dilution in the size exclusion column. Molecular weight can by estimated from AUC but can be measured directly in MALS via the differential refractive index.

Figure 2.10. Summary of methods and their advantages (+) and disadvantages (-). Parameters available from each technique are given: Rg, radius of gyration; Dmax, maximum dimension length; Rh, hydrodynamic radius; s20,w, sedimentation coefficient; f/f0, frictional ratio; Mr, molecular weight.

References for Chapter 2

1 W. L. Hubbell, C. J. López, C. Altenbach and Z. Yang, Curr. Opin. Struct. Biol., 2013, 23, 725–733. 2 W. R. Couet, R. C. Brasch, G. Sosnovsky and T. N. Tozer, Magn. Reson. Imaging, 1985, 3, 83–88. 3 C. J. López, S. Oga and W. L. Hubbell, Biochemistry (Mosc.), 2012, 51, 6568–6583. 4 B. Liang, J. H. Bushweller and L. K. Tamm, J. Am. Chem. Soc., 2006, 128, 4389– 4397. 5 R. G. Larsen and D. J. Singel, J. Chem. Phys., 1993, 98, 5134–5146. 6 M. Pannier, S. Veit, A. Godt, G. Jeschke and H. . Spiess, J. Magn. Reson., 2000, 142, 331–340. 7 G. Jeschke, Annu. Rev. Phys. Chem., 2012, 63, 419–446. 78 8 S. S. Eaton and G. R. Eaton, in Analytical Instrumentation Handbook, Third Edition, CRC Press, 2004, pp. 349–397. 9 J. A. Cooke and L. J. Brown, in Protein Folding, Misfolding, and Disease, eds. A. F. Hill, K. J. Barnham, S. P. Bottomley and R. Cappai, Humana Press, 2011, pp. 73–96. 10 S. Stoll and A. Schweiger, J. Magn. Reson., 2006, 178, 42–55. 11 A. Milov, K. Salikhov and M. Shchirov, Fiz. Tverd. Tela, 1981, 23, 975–982. 12 A. D. Milov, A. B. Ponomarev and Y. D. Tsvetkov, Chem. Phys. Lett., 1984, 110, 67–72. 13 R. Ward, A. Bowman, E. Sozudogru, H. El-Mkami, T. Owen-Hughes and D. G. Norman, J. Magn. Reson. San Diego Calif 1997, 2010, 207, 164–167. 14 P. P. Borbat and J. H. Freed, Chem. Phys. Lett., 1999, 313, 145–154. 15 L. V. Kulik, S. A. Dzuba, I. A. Grigoryev and Y. D. Tsvetkov, Chem. Phys. Lett., 2001, 343, 315–324. 16 S. Milikisyants, F. Scarpelli, M. G. Finiguerra, M. Ubbink and M. Huber, J. Magn. Reson. San Diego Calif 1997, 2009, 201, 48–56. 17 P. P. Borbat, E. R. Georgieva and J. H. Freed, J. Phys. Chem. Lett., 2013, 4, 170– 175. 18 G. Jeschke, V. Chechik, P. Ionita, A. Godt, H. Zimmermann, J. Banham, C. R. Timmel, D. Hilger and H. Jung, Appl. Magn. Reson., 2006, 30, 473–498. 19 Y. Polyhach, E. Bordignon and G. Jeschke, Phys. Chem. Chem. Phys., 2011, 13, 2356–2366. 20 G. Hagelueken, R. Ward, J. H. Naismith and O. Schiemann, Appl. Magn. Reson., 2012, 42, 377–391. 21 N. S. Alexander, R. A. Stein, H. A. Koteiche, K. W. Kaufmann, H. S. Mchaourab and J. Meiler, PLOS ONE, 2013, 8, e72851. 22 A. T. Brünger, P. D. Adams, G. M. Clore, W. L. DeLano, P. Gros, R. W. Grosse- Kunstleve, J. S. Jiang, J. Kuszewski, M. Nilges, N. S. Pannu, R. J. Read, L. M. Rice, T. Simonson and G. L. Warren, Acta Crystallogr. D Biol. Crystallogr., 1998, 54, 905–921. 23 G. M. Clore and J. Iwahara, Chem. Rev., 2009, 109, 4108–4139. 24 H. Tamaki, A. Egawa, K. Kido, T. Kameda, M. Kamiya, T. Kikukawa, T. Aizawa, T. Fujiwara and M. Demura, J. Biomol. NMR, 2016, 64, 87–101. 25 R. Bakhtiar and F. L. S. Tse, Mutagenesis, 2000, 15, 415–430. 26 M. Strohalm, D. Kavan, P. Novák, M. Volný and V. Havlíček, Anal. Chem., 2010, 82, 4648–4651. 27 N. J. Greenfield, Nat. Protoc., 2006, 1, 2876–2890. 28 A. J. Miles, B. A. Wallace and M. Esmann, Biochim. Biophys. Acta BBA - Biomembr., 2011, 1808, 2573–2580. 29 L. Whitmore and B. A. Wallace, Biopolymers, 2008, 89, 392–400. 30 G. Rhodes, in Crystallography Made Crystal Clear (Third Edition), ed. G. Rhodes, Academic Press, Burlington, 2006, pp. 7–30. 31 G. Rhodes, in Crystallography Made Crystal Clear (Third Edition), ed. G. Rhodes, Academic Press, Burlington, 2006, pp. 49–89. 32 J. T. Ng, C. Dekker, P. Reardon and F. von Delft, Acta Crystallogr. Sect. Struct. Biol., 2016, 72, 224–235. 33 G. Rhodes, in Crystallography Made Crystal Clear (Third Edition), ed. G. Rhodes, Academic Press, Burlington, 2006, pp. 109–143. 34 C. Abergel, Acta Crystallogr. D Biol. Crystallogr., 2013, 69, 2167–2173. 35 R. J. Read, Acta Crystallogr. D Biol. Crystallogr., 2001, 57, 1373–1382.

79 36 A. J. McCoy, R. W. Grosse-Kunstleve, L. C. Storoni and R. J. Read, Acta Crystallogr. D Biol. Crystallogr., 2005, 61, 458–464. 37 T. C. Terwilliger, R. W. Grosse-Kunstleve, P. V. Afonine, N. W. Moriarty, P. H. Zwart, L.-W. Hung, R. J. Read and P. D. Adams, Acta Crystallogr. D Biol. Crystallogr., 2008, 64, 61–69. 38 G. N. Murshudov, A. A. Vagin and E. J. Dodson, Acta Crystallogr. D Biol. Crystallogr., 1997, 53, 240–255. 39 P. V. Afonine, R. W. Grosse-Kunstleve, N. Echols, J. J. Headd, N. W. Moriarty, M. Mustyakimov, T. C. Terwilliger, A. Urzhumtsev, P. H. Zwart and P. D. Adams, Acta Crystallogr. D Biol. Crystallogr., 2012, 68, 352–367. 40 P. Emsley, B. Lohkamp, W. G. Scott and K. Cowtan, Acta Crystallogr. D Biol. Crystallogr., 2010, 66, 486–501. 41 P. Schuck, Biophys. J., 2000, 78, 1606–1619. 42 P. J. Wyatt, Anal. Chim. Acta, 1993, 272, 1–40. 43 N. C. Santos and M. A. Castanho, Biophys. J., 1996, 71, 1641–1650. 44 A. Doerr, Nat. Methods, 2013, 10, 38–38. 45 A. G. Kikhney and D. I. Svergun, FEBS Lett., 2015, 589, 2570–2577. 46 R. P. Rambo and J. A. Tainer, Biopolymers, 2011, 95, 559–571. 47 H. D. T. Mertens and D. I. Svergun, J. Struct. Biol., 2010, 172, 128–141. 48 D. I. Svergun, Biophys. J., 1999, 76, 2879–2886. 49 D. Franke and D. I. Svergun, J. Appl. Crystallogr., 2009, 42, 342–346. 50 D. Schneidman-Duhovny, M. Hammel, J. A. Tainer and A. Sali, Biophys. J., 2013, 105, 962–974. 51 D. Schneidman-Duhovny, M. Hammel, J. A. Tainer and A. Sali, Nucleic Acids Res., 2016, 44, W424–W429.

80 Chapter 3: Journal format thesis outline

The work in this PhD research programme, investigating HD-PTP, has been part of a multi-disciplinary collaboration involving several people and its impact is best illustrated as a part of the full publications in which it features. The results section is split into five chapters written as separate journal articles: paper 1 has already been published, paper 3 has been submitted, and papers 2, 4 and 5 are draft manuscripts.

The following is a list of papers, author contributions and a summary of how the work of the thesis contributed to each paper.

Paper 1: Structural Basis for Selective Interaction between the ESCRT Regulator

HD-PTP and UBAP1

Deepankar Gahloth, Colin Levy, Graham Heaven, Flavia Stefani, Lydia Wunderley,

Paul Mould, Matthew J. Cliff, Jordi Bella, Alistair J. Fielding, Philip Woodman, and

Lydia Tabernero, 2016, Structure, 24, 2115-2126.

Published in: Structure.

Author Contributions: CL collected X-ray diffraction data, solved, and refined the crystal structures. DG cloned HD-PTP constructs, prepared protein samples, and crystallized the proteins. GH prepared spin-labeled proteins and collected DEER data;

GH and AJF analyzed DEER data. GH collected CD data and analyzed it with JB. MJC collected and analyzed NMR data. PM and DG collected SPR data and PM analyzed it.

GH and JB performed torsion-angle molecular dynamics simulations. FS performed the yeast two-hybrid and co-immunoprecipitation experiments. LW performed the mitochondrial localization and RNAi rescue experiments. AJF designed and supervised the EPR spectroscopy. PW supervised and analyzed all cell biology data. LT designed

81 and supervised the project and analyzed the structural and biophysical data. LT and PW wrote the paper with input from other authors.

Summary: This publication reports the first X-ray structure of HD-PTP CC domain. Spin labelling and DEER spectroscopy was used to verify the open conformation observed in the X-ray structure. DEER also identified a discrepancy in distance measurements to the second helix, helix 2. Torsional angle molecular dynamics was used to model conformations of helix 2, which are compatible with DEER.

Paper 2: Investigating the conformational flexibility of HD-PTP CC domain

Graham Heaven, Colin Levy, Jordi Bella, Philip Woodman, Lydia Tabernero and

Alistair J. Fielding.

Target journal: Biophysical Journal.

Author contributions: GH prepared protein samples. GH and AJF performed EPR experiments and GH analyzed the data. GH and JD carried out MD for spin label modeling. CB collected and processed SAXS data and analyses were done by GH. LT,

PW and AJF designed and supervised the project. GH and AJF wrote the paper with input from other authors.

Summary: This publication explores further the structure of HD-PTP CC domain. Using

DEER, the conformation of the CC domain was investigated in the presence of binding partners. The results showed that the CC domain conformation is not affected by binding of Ubap1 or mono-ubiquitin. Furthermore, a frontotemporal lobar degeneration linked mutation of Ubap1 was investigated and these results showed that the mutant peptide was not prevented from binding HD-PTP CC, neither did this mutation affect the conformation as measured by DEER.

82 Paper 3: The open architecture of HD-PTP phosphatase provides new insights into the mechanism of regulation of ESCRT function

Deepankar Gahloth, Graham Heaven, Thomas Jowitt, Paul Mould, Jordi Bella, Clair

Baldock, Philip Woodman and Lydia Tabernero.

Submitted to: Scientific Reports.

Author contributions: DG cloned HD-PTP constructs, prepared protein samples and analysed SAXS data. GH, DG and TJ collected and analyzed MALS and AUC data. CB collected and processed SAXS data and analyses were done together with DG and JB. JB performed MD simulations for SAXS analyses and generated molecular models. LT designed, supervised the project and analysed the structural and biophysical data. LT and

PW wrote the paper with input from other authors.

Summary: This publication uses a combined approach of SAXS, MALS and AUC to examine the multidomain structure of HD-PTP in solution. MALS and AUC were used to measure hydrodynamic radii and frictional ratios which were used, in concert with parameters from SAXS, to generate solution models of HD-PTP. The results showed that

HD-PTP forms an open binding platform for ESCRT partners.

Paper 4: Structural characterization of HD-PTP:STAM2 ESCRT-0 complex

Graham Heaven, Michael Lockhart, Clair Baldock, Thomas Jowitt, Hilda D. Ruiz

Nivia, Jordi Bella, Colin Levy, Alistair J. Fielding, Philip Woodman and Lydia

Tabernero.

Target journal: EMBO Reports.

Author contributions: GH cloned STAM2 construct, mutated HD-PTP constructs and prepared protein samples. HDRN and GH collected and analyzed MALS data. TJ and GH

83 collected and analyzed AUC data. ML and CB collected and processed SAXS data and analyses were done together with GH and LT. GH and JB performed MD simulations for

SAXS analyses and generated molecular models. CL collected X-ray diffraction data. LT,

PW and AJF designed and supervised the project. GH and LT wrote the paper with input from other authors.

Summary: This publication investigates the solution structure of a HD-PTP:STAM2 complex by SAXS, MALS and AUC. The results supported a model where STAM2 binds

HD-PTP at both the Bro and PRR domains.

Paper 5: Surface selective spin labelling for DEER distance measurements in a cysteine rich protein

Graham Heaven, Michael Hollas, Philip Woodman, Lydia Tabernero and Alistair J.

Fielding.

Target journal: JACS.

Author contributions: GH prepared spin labelled protein samples. MH synthesized the spin label. GH and AJF performed EPR experiments and GH analyzed the data. GH analyzed mass spectra. AJF, MH and GH designed the project. AJF, LT and PW supervised the project. GH and AJF wrote the paper with input from other authors.

Summary: This publication investigates strategies for measuring DEER with multiple domains of HD-PTP with its increased number of cysteines. By using a bromoacrylaldehyde spin label (BASL), only the most exposed cysteines were labelled, which were situated in the CC domain, which allowed DEER to be carried out with the

CC domain alone, in the presence of the Bro and PRR domains and in the context of the

HD-PTP:STAM2 complex.

84 Chapter 4: Paper 1

Structural Basis for Selective Interaction between the ESCRT Regulator HD-PTP and UBAP1

Deepankar Gahloth,1,3 Colin Levy,1,3 Graham Heaven,2 Flavia Stefani,1 Lydia Wunderley,1 Paul Mould,1 Matthew J. Cliff,1 Jordi Bella,1 Alistair J. Fielding,2 Philip Woodman,1 and Lydia Tabernero1,4, 2016, Structure, 24, 2115-2126.

1School of Biological Sciences, Faculty of Biology, Medicine and Health 2School of Chemistry and Photon Science Institute University of Manchester, Manchester M13 9PT, UK 3Co-first author 4Lead Contact

Published in: Structure

Copyright information:

This article is available under the terms of the Creative Commons Attribution License (CC BY). You may copy and distribute the article, create extracts, abstracts and new works from the article, alter and revise the article, text or data mine the article and otherwise reuse the article commercially (including reuse and/or resale of the article) without permission from Elsevier. You must give appropriate credit to the original work, together with a link to the formal publication through the relevant DOI and a link to the Creative Commons user license above. You must indicate if any changes are made but not in any way that suggests the licensor endorses you or your use of the work. Permission is not required for this type of reuse.

DOI for publication: http://dx.doi.org/10.1016/j.str.2016.10.006

Creative commons license: https://creativecommons.org/licenses/by/4.0/

85 Summary

Endosomal sorting complexes required for transport (ESCRTs) are essential for ubiquitin-dependent degradation of mitogenic receptors, a process often compromised in cancer pathologies. Sorting of ubiquinated receptors via ESCRTs is controlled by the tumor suppressor phosphatase HD-PTP. The specific interaction between HD-PTP and the ESCRT-I subunit UBAP1 is critical for degradation of growth factor receptors and integrins. Here, we present the structural characterization by X-ray crystallography and double electron-electron resonance spectroscopy of the coiled-coil domain of HD-PTP and its complex with UBAP1. The coiled-coil domain adopts an unexpected open and rigid conformation that contrasts with the closed and flexible coiled-coil domain of the related ESCRT regulator Alix. The HD-PTP:UBAP1 structure identifies the molecular determinants of the interaction and provides a molecular basis for the specific functional cooperation between HD-PTP and UBAP1. Our findings provide insights into the molecular mechanisms of regulation of ESCRT pathways that could be relevant to anticancer therapies.

Introduction

Ubiquitination, endocytosis, and degradation of cell-surface receptors constitute a major mechanism of regulation of signal transduction by downregulating receptor availability for interaction with extracellular ligands. Receptor ubiquitination or degradation are often compromised in cancer pathologies, resulting in hyperactivation of signaling pathways promoting cell transformation and tumorigenesis. HD-PTP (His domain protein tyrosine phosphatase; PTPN23) is a non-receptor tumor suppressor phosphatase1,2 that regulates several ubiquitin-dependent endosomal trafficking processes such as downregulation of

EGFR and PDGFRβ signalling,3,4 recycling of Src,5,6 and degradation of α5β1 integrin.7

Consequently, loss of HD-PTP promotes cell proliferation, cell migration, and invasion.4–

86 8 It has been recently reported that HD-PTP haploinsufficiency predisposes mice to tumorigenesis, while hemizygous HD-PTP deletions are observed in many human cancers.9

HD-PTP drives the degradation of mitogenic receptors by coordinating their sorting into the multivesicular body (MVB) via specific recruitment of different endosomal sorting complexes required for transport (ESCRTs).10,11 ESCRTs (named ESCRT-0, -I, -II, and

-III) are multimeric protein complexes that drive membrane remodeling and scission in a number of important cellular events, including cytokinesis, autophagy, membrane repair, and virus budding.12 Pathway selectivity is defined by different subsets of ESCRTs and specialized adaptor Bro1 proteins: Bro1 in yeast, HD-PTP and Alix in animals.13–15

However, the mechanisms by which Bro1 proteins regulate the different ESCRT pathways remain poorly understood. For example, Alix is a general ESCRT-III regulator that supports multiple ESCRT-dependent processes.13,16–20 In contrast, HD-PTP function is largely restricted to the early endosome3,21 where it acts in close cooperation with

ESCRT-010 and UBAP1 (ESCRT-I) to downregulate multiple ubiquitinated cargoes.7,22

UBAP1 is an ubiquitin-binding ESCRT-I subunit that acts exclusively in MVB sorting11,23 and, like for HD-PTP2,24, haploinsufficiency of UBAP1 is linked to nasopharyngeal carcinoma.25 UBAP1 is also a risk factor for familial frontotemporal lobar degeneration.26

HD-PTP has a multidomain organization that allows coordinated binding to several

ESCRTs (Figure 4.1). Knowledge of the three-dimensional architecture of HD-PTP and its mode of binding to the different ESCRTs is critical for understanding how these functional interactions are regulated. To date, structural information on HD-PTP is only available for its N-terminal Bro1 domain,27,28 which binds ESCRT-0 and ESCRT-III.2,3,10

Structures for the coiled-coil (CC) domains of other Bro1 proteins, Alix and yeast Bro1,

87 have been reported29–31 and they show a V-shaped conformation, with two arms connected by a flexible hinge region (Figure 4.1). However, the CC domain of HD-PTP has very low homology (17%–19%) to those of Alix or Bro1, suggesting that significant structural disparity may exist. The CC domain of HD-PTP is solely responsible for the interaction with UBAP1, and residue F678 in this domain is essential for binding.11 This

32 Phe residue is part of a FYX2L motif (Figure 4.1) conserved in all Bro1 proteins. In

Alix, the FYX2L motif is located on the second arm of the CC domain and mediates

29,30,33 binding to viral Gag proteins. Paradoxically, the presence of the FYX2L motif in

Alix is not sufficient for binding to UBAP1,11 indicating that other structural determinants may be important to define selectivity.

To address this paradox and to provide new insight into the regulation of ESCRT function by HD-PTP, we have determined the crystal structures of the CC domain of HD-PTP

(HD-PTPCC) alone and in complex with UBAP1. The structures show an unexpected open conformation of the CC domain, strikingly different from the V domains of Alix and yeast

Bro1.29–31 We propose that this open conformation is critical to explain the selective binding of UBAP1 to HD-PTPCC but not to Alix. The HD-PTPCC binding interface with

UBAP1 was also analyzed by nuclear magnetic resonance (NMR), paramagnetic relaxation enhancement (PRE), and surface plasmon resonance (SPR), and validated through site-directed mutagenesis and functional cell-based assays, thus establishing the structural basis for the functional cooperation between HD-PTP and UBAP1.

Results

HD-PTPCC Adopts an Open and Extended Conformation

The crystallographic structure of HD-PTPCC (apo-HD-PTPCC), determined at 2.5 Å resolution (Table 4.1), shows an elongated architecture with seven α helices (H1–H7)

88 whose main feature is a central helix of 105 residues (H7) extending the whole length of the molecule (Figure 4.1B). The overall shape resembles that of an ice hockey stick, in which H1, H6, and the N-terminal region of H7 form the blade, and H2–H5 and the rest of H7 form the shaft (Figure 4.1B). The maximal dimension of HD-PTPCC is approximately 155 Å from end to end. Analysis of the structure using SOCKET34 identified two canonical coiled-coil motifs: one antiparallel two-stranded coil involving helices H6 and H7 in the blade and an extensive antiparallel, tightly packed three-stranded coil involving helices H3, H4, and H7 in the shaft (Figure 4.1C). The topology of HD-

PTPCC is such that the polypeptide chain crosses three times over the length of the protein

(Figure 4.1D).

The extended shape of HD-PTPCC differs markedly from the V domains of Alix (AlixV)

29–31 and yeast Bro1 (Bro1V) (Figures 1E and 1F). These V domains are built from two arms joined by three unstructured loops, forming a flexible hinge (Figures 4.1E and 4.1F).

By contrast, in HD-PTPCC, helix H7 forms a continuous backbone (Figures 4.1B and

4.1D) that confers rigidity to the molecule.

Figure 4.1. Crystallographic Structure of HD-PTPCC. (A) Diagram of HD-PTP domain structure indicating domain boundaries and sites of interaction for ESCRT partners. The position of the conserved FYX2L motif in the CC domain is shown, as well as the FPXL motif in UBAP1. (B) Cartoon diagram of the HD-PTPCC crystal structure. HD-PTPCC resembles an ice hockey stick, where the N-terminal region represents the blade and the C-terminal region is the shaft. The seven α helices are labelled H1 to H7, with H7 being the central and longest helix extending the whole length of the structure. (C) Coiled-coil motifs in the HD-PTPCC structure after analysis with SOCKET (Walshaw and Woolfson, 2001). HD-PTPCC contains two canonical coiled coils: one two-stranded coiled coil (2st- cc) in the blade and one three-stranded coiled coil (3st-cc) in the shaft. (D) Topology diagram of the structure of HD-PTPCC showing the arrangement of the seven α helices. (E) Superimposition of the HD-PTPCC (red) and AlixV (blue) structures. The AlixV crystal structure (PDB: 2OJQ) shows a V-shaped helical protein in a closed conformation, in contrast to the open and extended conformation of HD-PTPCC. The two arms in AlixV are labeled. (F) Superposition of the structures of AlixV (blue) and yeast Bro1V (orange, PDB: 4JIO). Both structures contain two arms joined by flexible loops.

90 Table 4.1. Crystallographic Data Collection and Refinement Statistics

apo-HD-PTPCC HD-PTPCC–UBAP1C

Data Collection

Space group P21 P21 21 21

Cell dimensions

a, b, c (Å) 53.5, 47.7, 172.7 48.9, 93.3, 102.2

α, β, γ (°) 90, 96, 90 90, 90, 90

Molecules per asymmetric unit 2 1

Resolution (Å) 2.54 (2.63–2.54)a 2.55 (2.64–2.55)

Rmerge 0.1 (0.8) 0.1 (0.8)

I/σI 19.3 (3.8) 13.6 (2.5)

Completeness (%) 98 (100) 100 (100)

Redundancy 13.5 (13.8) 6.3 (6.6)

Refinement

Resolution (Å) 2.54 2.55

No. of reflections 28,609 15,835

Rwork/Rfree 21.7/27.8 20.9/25.4

No. of atoms 5,156 2,725

Protein 4,991 2,639

Peptide NA 73

Water 80 31

B factors

Protein 49.1 73.6

Peptide NA 89.7

Water 45.6 56.5

Root-mean-square deviations

Bond lengths (Å) 0.009 0.003

Bond angles (°) 1.0 0.5

Each structure was determined from one crystal. aValues in parentheses are for the highest-resolution shell.

91 HD-PTPCC Shows Limited Local Conformational Flexibility but No Large

Structural Rearrangements

Conformational flexibility between the two arms of AlixV and Bro1V has been

31,35 reported. We investigated the potential flexibility of HD-PTPCC using double electron-electron resonance (DEER) spectroscopy (Figure 4.2). This technique allows the measurement of dipolar coupling interactions between methanethiosulfonate spin labels

(MTSL), covalently attached to cysteines (Figure 4.S1). This was used here to estimate average distances between the labels (Figure 4.2). Comparison of these distances with those obtained from labels modelled on the HD-PTPCC crystal structure provided information on the conformational dynamics of HD-PTPCC. We conducted DEER experiments on wild-type (WT) HD-PTPCC triply labelled at C425, C628, and C697, the three cysteines present in the CC domain, and on doubly labelled mutants where each cysteine in turn had been changed to serine (Figure 4.2).

Significantly, the C425S mutant, doubly labelled at C628-C697 (Figure 4.2Ai), gave no resolved dipolar coupling, indicating that the spin labels must be separated by more than

7 nm (the maximal mean distance observable in a 6 μs window) (Figure 4.S1). This is consistent with the extended conformation observed in the HD-PTPCC crystal structure, where the distance between labels on C628 and C697 is 9.9 nm (Figure 4.2C). Likewise, the interspin distance between C425 and C697 was estimated to be 6.3 nm (Figures 4.2Aii and

4.2Aiii), which matches very well with the 6.0 nm on the crystal structure (Figure 4.2C).

However, the interspin distance between C628 and C425 estimated at 4.2 nm (Figures

4.2Aii and 4.2Aiv) differs from the distance of 5.4 nm on the crystal structure

(Figure 4.2C), suggesting some local conformational flexibility around C425. This residue is located in helix H2, which is connected by two flanking loops to helices H1 and H3

(Figure 4.1). These loops are poorly defined in the electron density maps, indicating

92 backbone flexibility in that region. Molecular dynamics simulations around helix H2 also showed that this helix could easily adopt alternative orientations compatible with the

DEER results, without requiring large overall conformational changes (Figures 4.2D and

4.S2).

To confirm the above analyses, an additional Cys residue was engineered at position 521 in H4 and residues C425 and C697 were changed to Ser. In this case, the mean interspin distance between C628 and C521 (4.9 nm, Figure 4.2Av) was in good agreement with the distance on the crystal structure (5.4 nm, Figure 4.2E). The DEER analysis is therefore consistent with an overall rigid conformation for HD-PTPCC that can accommodate local flexibility without global rearrangements, ruling out a potential mechanism of regulation by large-scale flexibility of the CC domain.

Figure 4.2. DEER Spectroscopy of HD-PTPCC. (A) DEER distance distributions (black) and crystal structure-based predictions (red) are shown for all the experiments using triply labelled HD-PTPCC (ii), or doubly labelled (i, iii, iv, v). Coloured boxes from DEERAnalysis36 are shown: green is reliable mean, width, and shape; yellow is reliable mean and width; orange is reliable mean; and red indicates long-range distance. (B) Structure of MTSL-labelled cysteine. (C) HD-PTPCC structure with modelled MTSL labels at C425, C628, and C697. Mean experimental DEER (black) and predicted (red) distances are shown. (D) HD-PTPCC structure (red) and model obtained after TAMDS (blue) showing displacement of H2 (where C425 is located) and movement of flanking loops. The predicted distance (blue) between labels at C425 and C628 shows agreement with experimental DEER distances (black). (E) HD-PTPCC structure with modelled MTSL labels at C521 and C628. Mean experimental DEER (black) and predicted (red) distances are shown.

Supplemental Figure 4.1. (Related to Figure 4.2) DEER spectroscopy of HD-PTPCC and Circular Dichroism analysis of MTSL-labelled HD-PTPCC and mutants used in the EPR studies. (A) DEER traces of the triply MTSL-labelled HD-PTPCC and doubly MTSL- labelled mutants, as indicated inside each frame. Inset graphs show form factor fits after exponential background correction. Modulation is apparent in all triply and doubly labelled proteins except for the C425S mutant (top panel) where the distance between spins is larger than the detectable limits within this window size. (B) Circular dichroism of HD- PTPCC (unlabelled and triply labelled at C425, C628 and C697) and mutants C425S (labelled at C628 and C697), C628S (labelled at C425 and C697) and C697S (labelled at C425 and C628). Samples for circular dichroism were adjusted to 0.1 mg/mL and buffer exchanged into phosphate buffer (100 mM potassium phosphate, 100 mM potassium fluoride, pH 7.4). Spectra were recorded on a Jasco J-180 spectropolarimeter between 190 and 260 nm using cuvettes of 0.5 mm path length. A data pitch of 0.2 nm was used with a response time of 8 seconds per point. Circular dichroism mdeg units were converted to mean residue ellipticity. No significant changes are seen upon labelling or mutagenesis of HD- PTPCC, thus confirming that there are no alterations to the secondary structure. 95

Supplemental Figure 4.2. (Related to Figure 4.2) Torsion-angle molecular dynamics simulations using DEER distance constraints. Ribbon diagram of the HDPTPCC crystal structure (red) superimposed to 25 trajectory model structures at 1 ns intervals (blue) showing the flexibility of the loops that connect helix H2 with the rest of the CC domain. Several orientations of helix H2 resulting from the molecular dynamics simulations are compatible with the interspin distances estimated from the DEER experiments, without major global conformational changes of the CC domain structure.

96 Identification of the Minimal UBAP1 Binding Region to HD-PTP

Specific interaction with UBAP1 is central to the function of HD-PTP,11 and their cooperation is essential to regulate integrin signalling and cell migration.7 Therefore, we aimed to define this interaction at the molecular level. We previously identified the central region of UBAP1 (122–309) as binding to the CC domain of HD-PTP in a yeast two- hybrid (Y2H) screen.11 We now have confirmed the interaction by co- immunoprecipitation of in vitro translated full-length UBAP1 with bacterially expressed

HD-PTPCC (Figure 4.S3). In addition, we have also shown the interaction between

UBAP1 and HD-PTPCC in cells. For these experiments, a chimera of FKBP12 fused to

HD-PTPBro1-CC (FKBP-HD-PTPBro1-CC) was co-expressed with a chimera of FRB

(FKBP12-rapamycin binding) fused to a mitochondrial targeting sequence (mito-FRB).

Treatment of cells with rapamycin caused the efficient relocalization of HD- PTPBro1-CC to mitochondria. Under these conditions, UBAP1-GFP, but not GFP, also relocated to mitochondria (Figure 4.3A).

Further truncations of UBAP1 (122–309) identified residues 260–269 as the minimal region for effective binding to HD-PTP (Figure 4.3B). This region contains an FPXL motif that resembles the YPXnL motif conserved within viral Gag late domains and other substrate proteins that bind to the V domain of Bro1 proteins.30,32,33,37 The F268S mutation in this motif abolished binding to HD-PTP, while the P269A or L271A mutations had no obvious effect on binding in the Y2H assays (Figures 4.3A and 4.3B).

This suggests that UBAP1 F268 forms a critical interaction with HD-PTP.

To further characterize this interaction, immobilized HD-PTPCC, HD-PTPBro1-CC, and

HD-PTPBro1 were tested in biosensor binding experiments (SPR) using a UBAP1 peptide containing residues 261–280 (UBAP1C) as the analyte. Affinities of UBAP1C to HD-

PTPCC and HD-PTPBro1-CC were similar, with dissociation constants Kd of 66.3 μM and

97 31.9 μM, respectively (Figures 4.3C and 4.3D). No binding was observed to HD-PTPBro1

(Figure 4.3D), thus confirming that the main UBAP1-binding region is within the CC domain, and that the conformation of the CC domain is functionally competent, both on its own and in the presence of the Bro1 domain.

Figure 4.3. Minimal Binding Region in UBAP1 Responsible for Interaction with HD- PTP. (A) HeLa cells transfected with FKBP-HD-PTPBro1-CC-myc, Mito-FRB, and either GFP, UBAP1-GFP wild-type, or UBAP1-GFP F268S were treated with rapamycin and imaged by immunofluorescence for GFP, myc, and mitochondrial Hsp70. Scale bar, 10 μm. Upon treatment, UBAP1-GFP translocates to mitochondria colocalizing with FKBP-HD-PTPBro1-CC but not GFP or UBAP1-GFP F268S. (B) Yeast two-hybrid interactions between UBAP1 fragments and HD-PTPBro1-CC. +/− symbols indicate the degree of growth. (C) Biosensor binding isotherms for the different HD-PTP constructs to UBAP1C. Affinity to HD-PTPCC, and HD-PTPBro1-CC was similar, with dissociation constants Kd of 66.3 ± 0.35 μM and 31.9 ± 0.85 μM, respectively. Error bars represent the SEM, n = 3. (D) Biosensor sensograms for the immobilized HD-PTP constructs binding to UBAP1C peptide, showing that the CC domain is responsible for binding since HD-PTPBro1 fails to bind to UBAP1C.

Supplemental Figure 4.3. (Related to Figure 4.3) Co-immunoprecipitation of UBAP1 with HD-PTP. UBAP1-strep was translated in vitro and incubated with or without His6- HD-PTPBro1-CC. Samples were immunoprecipitated with anti-His antobodies. Top panel: phosphorimage. Bottom panel: Coomassie stained SDS-PAGE.

Mapping the Molecular Interface between HD-PTPCC and UBAP1C

In order to define the molecular interactions, we determined the crystal structure of HD-

PTPCC in complex with UBAP1C at 2.5 Å resolution (Table 4.1). UBAP1C adopts a rather extended conformation and binds to the shaft, near the core of the three-strand coiled coil, between H4 and H7 (Figure 4A and Figure 1C). Only residues 262–271 from UBAP1C were clearly identified in the electron density maps (See http://www.cell.com/structure/fulltext/S0969-2126(16)30314-8 for mp4 file download of

Supplemental Movie) and the side chains of I263, L266, F268, P269, and L271 are in direct contact with HD-PTPCC (Figure 4). Residues 272–280 were disordered and not visible, probably because of weaker affinity. Interactions between UBAP1C and HD-

100 PTPCC are mainly hydrophobic, with hydrogen bonds only present between K671 on HD-

PTPCC and the main chain of I263, K264, and L266 on UBAP1C (Figure 4.4B).

We confirmed the molecular interface between HD-PTPCC-UBAP1C using NMR by comparing the 1H 13C-heteronuclear single quantum coherence (HSQC) spectra of

UBAP1C in the absence and presence of HD-PTPCC. Significant peak broadening was only observed for residues 261–271 in the presence of HD-PTPCC (Figures 4.5A, 4.5B, and 4.S4), consistent with the interactions observed in the crystal structure, and with the minimal UBAP1 binding region identified by Y2H (Figure 4.3B).

The binding interface in HD-PTPCC is defined by three hydrophobic pockets (A–C in

Figure 4.4) and the conserved FYX2L motif plays a critical role in the interaction. The aryl ring of HD-PTPCC F678 forms a wall that divides pockets B and C and contributes to multiple hydrophobic interactions with UBAP1C (Figure 4.4). Pocket A is large and shallow and accommodates UBAP1C residues I263 and L266. This pocket is lined by

T511, L515, A518 (H4 helix) and A664, L668, K671 (H7 central helix). Pocket B is deep and narrow and accommodates UBAP1C F268 and P269. This pocket includes A508,

T511 (H4), L445 (H3), and G675, F678 and Y679 (H7), forming extensive contacts with the aryl ring of F268. Pocket C accommodates L271 and includes V504 (H4) and L682

(H7). This pocket is closed off by a high rim provided by the side chains of K500, Y501

(H4), and K685 (H7) (Figure 4.4B).

The tight fit of F268 in the B pocket explains the key role of this residue for binding to

HD-PTP, as shown by Y2H (Figure 4.3). Conversely, UBAP1C P269 and L271 show fewer contacts with HD-PTPCC: V504 and F678 near pocket B, and V504 and L682 in pocket C, respectively (Figure 4.4). The minor contribution of these two residues to the binding interface explains the weaker phenotype observed for the single mutations P269A and L271A (Figure 4.3), particularly in the presence of F268, which is the main anchor

101 into the binding site. Hydrogen bonds are only present between the side chain of HD-

PTPCC K671 in pocket A and the main chain of I263, K264, and L266 in UBAP1C

(Figure 4.4B).

Pocket A is partially occluded in the structure of apo-HD-PTPCC (Figure 4.4E) by the side chains of K671, D667, and E514. Binding of UBAP1C thus requires re-arrangement of these three side chains (Figure 4.4F). In the apo-HD-PTPCC structure, D667 and K671 form a salt bridge, and the carboxyl group of E514 is about 6.4–7 Å to D667 and K671

(Figure 4.4F). Upon UBAP1C binding, E514, D667, and K671 side chains are displaced breaking the K671-D667 salt bridge and opening the A pocket. K671-Nζ then forms hydrogen bonds with the C=O groups of I263, K264, and L266 in UBAP1C (Figure 4.4B).

102

Figure 4.4. Crystal Structure of the HD-PTPCC-UBAP1C Complex and Analysis of the Binding Interface. (A) Structure of HD-PTPCC (red cartoon) in complex with UBAP1C (space-filling) showing that the binding site is located in the middle of the shaft region. (B) Overview of UBAP1C (sticks) binding site in HD-PTP. The three helices in HD-PTP that form the binding site in the shaft are labeled (red) and represented as cartoons. Residue side chains that participate in interactions with UBAP1C are shown as sticks and labeled (black). Residue K671 forms hydrogen bond interactions (cyan dashed lines) with three carbonyl oxygens in the UBAP1C peptide (sticks). (C) Detail of the UBAP1C binding site (pocket B) showing the HD-PTPCC residues in the conserved FYXnL motif: F678, Y679, and L682 forming hydrophobic interactions with the UBAP1c F268, P269, and L271 in the FPXL motif. (D) Electrostatic surface of HD-PTPCC at the UBAP1C binding site showing three main pockets A–C (yellow circles) that accommodate the peptide. UBAP1C is shown as sticks. Residues in UBAP1C that interact with HD-PTPCC are labeled: I263, L266 bind to pocket A; F268 and P269 bind to pocket B; L271 binds to pocket C. (E) Electrostatic surface of the structure of apo-HD-PTPCC showing that pocket A is occluded by the side chains of K671 and D667, forming a salt bridge (yellow arrow). (F) Detail of the structure of apo-HD-PTPCC (white ribbon left) and HD- PTPCC in complex with UBAP1C (green ribbon center). In the complex, the side chains of K671 and D667 are shifted (middle panel) and the salt bridge that they form in the apo structure (left panel) is lost. The side chain of E514 is also displaced, making room to accommodate the peptide in pocket A (right panel).

103

Figure 4.5. NMR Analysis of the HD-PTPCC-UBAP1C Binding Interface. (A) Hα region 1 13 13 of H C-PUSH-HSQC of UBAP1C (natural abundance C) in PBS-D2O showing the 1 13 resonance assignment. (B) Hα region of H C-HSQC of UBAP1C (natural abundance 13 C) in the presence of sub-stoichiometric HD-PTPCC. The D2O content in these samples was approximately 80% and significant intensity arises from H2O, compromising interpretation of signals between 4.8 and 4.6 ppm. Residues close in sequence to UBAP1 F268 are significantly broadened by interaction with HD-PTPCC. Residues C-terminal of UBAP1 D272 are less affected by the presence of HD-PTPCC, and therefore are identified as not being involved in the binding site. (C) Per residue mean peak intensity ratios between UBAP1C samples containing paramagnetically (Iox) and diamagnetically (Ired) labeled HD-PTPCC (C425, C628), indicating the extent of paramagnetic relaxation enhancement. For some resonances, intensity ratios are low because of the line broadening induced by binding, and these are marked by red bars in the chart. Error bars are SDs estimated from the noise level in the HSQC spectra.

104

Supplemental Figure 4.4. (Related to Figure 4.5) NMR and PRE studies of UBAP1 peptide in the presence of HD-PTP. (A) Aromatic regions of 1H NMR spectra of UBAP1C, in the absence (upper panel) and presence (lower panel) of HD-PTPCC. Narrow line-width signals arising from the sidechain of Phe268 (Hδ# (7.21 ppm), Hε# (7.28 ppm) and Hζ(7.24 ppm)) are clearly visible in the top panel, but absent from the lower one. Present at approximately 25% intensity are signals (7.18, 7.29 and 7.30 ppm) arising from UBAP1 F268 in the cis proline isomer of the peptide, which are unaffected by the 1 13 presence of HD-PTPCC. (B-C) Effect of electron spin- label at C425: Hα region of H- C HSQC of UBAP1C in the presence of spin-labelled variant of HD-PTPCC (10:1 mixture UBAP1C:HD-PTPCC); (B) shows spectrum when spin-label is diamagnetic (reduced with excess ascorbate), and (C) shows spectrum when spin-label is paramagnetic.

105 The orientation of UBAP1C in the HD-PTPCC binding site was validated in PRE experiments with MTSL labels at C425 and C628. Dipolar interaction with the electron spin label causes an additional component of the transverse relaxation of the NMR signal, which has strong distance dependence with maximal effect below 40 Å.38 The paramagnetic enhancement in this case is dominated by the spin label attached to C425, which is closer to the UBAP1C binding site. Consistent with this, NMR resonances from the N-terminal end of the UBAP1C showed enhancement of their relaxation properties and therefore loss of peak intensity in the presence of the MTSL label (oxidized form), whereas those from the region beyond L271 (>40 Å from the label) did not (Figure 4.5C), thus confirming the orientation of UBAP1C at the binding site.

Structural Basis for the Specific Functional Interaction of UBAP1 with HD-PTP

All Bro1 proteins contain a conserved FYX2L motif essential for binding to their substrate

32 proteins, which contain a reciprocally conserved YPXnL motif. Several structures of

Alix in complex with retroviral Gag late-domain peptides29,30 (Figure 4.6) have shown that these conserved motifs interact with each other, thus confirming their functional importance. The FY pair in the FYX2L motif in Alix stacks against Y in the Gag peptide

29,30 YPXnL motif, thereby serving as the main anchoring point for their interaction.

Similar binding motifs are found in HD-PTP (FYADL) and UBAP1 (FPTL), yet the

11 binding of UBAP1 to HD-PTP is surprisingly highly selective. In our HD-PTPCC-

UBAP1C structure, the FY pair (F678, Y679) in HD-PTP forms stacking interactions with

F268 in the FPXL motif of UBAP1C (Figure 4.4). The critical importance of these residues for binding was confirmed by SPR, where both the HD-PTP F678D and UBAP1

F268S mutations abolished binding (Figure 4.7A).

Despite similarities in the molecular recognition motifs, important differences are apparent when comparing the structures of HD-PTPCC-UBAP1C and AlixBro1-V-Gag

106 peptide complexes. First, the binding site of UBAP1C is displaced with respect to that of the Gag peptides (Figure 4.6A): UBAP1C occupies pocket A in HD-PTPCC, which is poorly conserved in Alix and does not participate in Gag peptide binding (Figures 4.6 and

4.S5). This displacement is possible because the architecture of HD-PTPCC offers an open, extended interface suitable to accommodate UBAP1C. In contrast, in all the complexes of AlixBro1-V with Gag peptides, both arms of the V domain form an apex that makes pocket A inaccessible, thus preventing binding due to steric hindrance (Figure 6).

Second, in Alix, there is a hydrophobic binding groove that accommodates the Gag peptides and extends beyond pocket C. In HD-PTPCC, the side chains of K500, Y501, and

K685 form a high rim at the edge of this pocket (Figures 4.4 and 4.6), and the three- stranded coiled coil results in tight packing between H4 and H7, leaving no room for a groove. These structural features also explain why UBAP1C is displaced toward pocket

A with respect to the position of the Gag peptides on Alix (Figure 4.6).

We believe that the differences in the architecture and binding interface features observed between Alix and HD-PTP are critical in determining ligand-binding selectivity. We confirmed this by showing that in solution, AlixV does not bind to the UBAP1C peptide, and that a Gag peptide (SIV-GAG) does not bind to HD-PTPCC (Figure 4.7B), mirroring

11 the lack of binding of full-length UABP1 to AlixBro1-V. The conserved FYX2L motif is therefore necessary for binding but insufficient to determine specific selectivity between

Bro1 proteins and their biological partners. Instead, both the overall architecture and local structural determinants appear to be key in defining molecular recognition and binding specificity.

107

Figure 4.6. Comparison of the Binding Sites in the HD-PTPCC-UBAP1C, and Alix-Gag Peptide Complexes. (A) Structures of Alix (gray ribbon) in complex with HIV-Gag (magenta) and SIV-Gag (blue), superimposed on the structure of HD-PTPCC (gold ribbon) in complex with UBAP1C (green). The UBAP1C binding site is displaced with respect to the Gag peptides that bind along a distal hydrophobic groove not conserved in HD- PTP. Conversely, the closed conformation of the AlixV domain prevents full access to the region near the apex, making pocket A inaccessible. (B) Enlarged view into the binding site of UBAP1c (sticks) in HD-PTPCC (gold surface) and of the complex of SIV-Gag peptide (blue ribbon) in AlixV (gray surface). Only residues that interact are shown. Pocket A and the high rim in HD-PTP that closes off pocket C are labelled. In the complex of AlixBro1-V with SIV-Gag, the binding site extends toward a hydrophobic groove beyond pocket C, but no binding occurs in the equivalent to pocket A in HD-PTP. (C) Detailed view of pocket A in HD-PTP (left, gold ribbon) and Alix (right, gray ribbon) showing the side chains in sticks. The lack of sequence conservation in this pocket together with steric hindrance may explain the lack of interactions of the SIV-Gag peptide with Alix (right), whereas in HD-PTP, pocket A provides several hydrophobic interactions with residues in the UBAP1C peptide (labelled) in addition to hydrogen bonds with K671 in HD-PTP. (D) Detail view of pocket C showing the HD-PTP residues that form the high rim that closes off pocket C (left) and prevents binding of UBAP1C peptide beyond this point. In contrast, SIV-Gag forms interactions along a hydrophobic groove in Alix (right) that extends beyond pocket C. 108

Supplemental Figure 4.5. (Related to Figure 6) Sequence alignment of the coiled-coil domains of human HD-PTP and Alix. Alignment of the sequences of the coiled-coil domain for HD-PTP and Alix. Conservation is highlighted in yellow and identity in red. The conserved FYXnL motif is underlined in black. Alignment was done using ESPript 3 online server.39

109

Figure 4.7. Biochemical and Functional Validation of the UBAP1C Binding Interface with HD-PTP. (A) Biosensor sensograms of the binding of HD-PTPCC containing the F678D mutation to UBAP1C peptide (left) and of binding of HD-PTPBro1-CC and HD- PTPCC to UBAP1C peptide containing the F268S substitution (center and right). Both mutations abolished interaction. (B) Binding to the CC domain is selective despite conservation of the FYX2L motif in both Alix and HD-PTPCC: UBAP1C does not bind to Alix and the SIV-Gag peptide does not bind to HD-PTPCC. (C) Cells depleted of HD-PTP were transiently transfected with HA-tagged HD-PTPBro1-CC and stimulated with EGF for 3 hr before fixing and staining with anti-ubiquitin (Ub). Cells transfected with WT HD- PTPBro1-CC display even ubiquitin distribution throughout the cell, while untransfected cells or those transfected with the F678D mutant display very strong accumulations on cytoplasmic inclusions. These co-label with endosomal markers as previously reported3 (data not shown). Scale bar, 10 μm. (D) Scoring of rescue experiments. Cells transfected as indicated were scored for normal ubiquitin distribution. One hundred cells from three independent experiments were counted, and SDs between these experiments are shown. Two-way ANOVA analysis: F678D versus WT, p = 0.0025; K671A versus WT, p = 0.043; K671A versus F678D, p = 0.002 110 Functional Validation of HD-PTP-UBAP1 Interactions

The HD-PTPCC-UBAP1C interface was validated by RNAi rescue experiments using HD-

PTPBro1-CC and mutations at the binding site. In normal cells, EGFR that has been activated by EGF passes through the endosomal pathway and is degraded within lysosomes. In contrast, cells depleted of HD-PTP are characterized by the accumulation of ligand-activated EGFR in highly clustered early endosomes that label strongly for protein-ubiquitin conjugates.3 We have previously shown that reintroduction of HD-

PTPBro1-CC is sufficient to rescue these trafficking defects and represents the minimal functional region of HD-PTP.3 HeLa cells depleted of HD-PTP and pulsed for 3 hr with

EGF, showed intense ubiquitin labelling on cytoplasmic clusters (Figure 7C), which co- labelled with the endosomal marker EEA1 as previously reported3 (data not shown). As expected, transfection of these cells with WT HD-PTPBro1-CC restored a WT phenotype, in which ubiquitinated proteins were evenly distributed throughout the cell (Figure 7C).

In contrast, the HD-PTPBro1-CC F678D mutant was unable to rescue depletion of HD-PTP, with both transfected and untransfected cells displaying strong ubiquitin labelling on cytoplasmic inclusions (Figure 7C). The HD-PTPBro1-CC K671A mutant failed to completely rescue a WT phenotype, but showed milder defects than F678D (Figure 7D).

Although K671 forms hydrogen bond interactions with UBAP1C (Figures 4 and 6), its overall contribution to binding affinity is clearly less critical when compared with the contribution of the extensive hydrophobic interactions of F678. These data extend our previous findings,11 confirm the functional significance of the interface observed in the crystal structure, and, importantly, demonstrate that binding of HD-PTP to UBAP1 is essential for correct sorting of activated EGFR, since Alix does not function in this

ESCRT pathway.

111 Discussion

The close functional cooperation between HD-PTP and UBAP1 is physiologically crucial. Disruption of this cooperation by ablation or genetic defects leads to cancer, altered cell migration, and neurological pathologies.7,24–26 Our findings explain the structural basis for the interaction between HD-PTP and UBAP1 and reveal why UBAP1 binding is selective to HD-PTP despite conservation of the FYX2L binding motif across other Bro1 proteins. In addition, the remarkably different architecture of the HD-PTP CC domain compared with other Bro1 proteins provides further insights into the assembly of specialized ESCRTs at the endosome that drive downregulation of cell-surface receptors.

Our data show that HD-PTPCC adopts an open and extended architecture, where the CC domain maintains a rigid conformation by virtue of the long central α helix. This structure contrasts with the V-shaped CC domains of Alix and yeast Bro1,29–31 and is consistent with the low homology found between HD-PTPCC and these V domains.

We revealed the structural determinants for specific interaction between HD-PTP and

UBAP1 by X-ray crystallography, NMR, and mutagenesis. Our results confirm that the conserved HD-PTP residue F678 in the FYX2L motif is essential for binding to UBAP1.

Furthermore, key hydrophobic interactions between UBAP1 residues F268, P269, and

L271 (in the FPXL motif) and HD-PTP residues F678, Y679, and L682 (in the FYX2L motif) form the core of the binding interface. We demonstrated the critical importance of

F268 and F678 residues by mutagenesis, binding analysis, and functional cellular assays.

Our findings match the reported roles of Alix F676, yeast Bro1 F687, and Rim20 F623 in binding to Gag late domains, Rfu1, and Rim101, respectively.29,30,32 Paradoxically, conservation of the FYX2L motif in Bro1 proteins and that of the YPXnL motif in their biological targets does not result in promiscuous interactions.11,32 Our studies bring new

112 insight into understanding this paradox. Significant differences in the sequence, local structural features, and hydrophobicity of the UBAP1C binding site in HD-PTP, compared with that of the Gag peptide site in Alix, are critical to determine specificity. Furthermore, the open architecture of HD-PTPCC is essential to enable binding of UBAP1C to pocket

A, which is inaccessible in the V-shaped structure of Alix observed in the complexes with

Gag peptides. Thus, the conformational differences may provide additional molecular determinants for the binding selectivity that we observe.

AlixV has been reported to be flexible in solution and therefore able to adopt a more open

31,35 conformation. However, we found no evidence that Alix binds the UBAP1C peptide in our binding studies in solution, or of binding to the full-length UBAP1 by Y2H.11

We hypothesize that differences in the overall architecture of Bro1 proteins, and in the local features at the binding site, combine to elicit exquisite functional selectivity for

ESCRT pathway regulation. This is certainly the case for HD-PTP and Alix. Whether these principles apply to other Bro1 proteins and their targets will require further high- resolution analyses of their complexes.

We found that there is no evidence for large-scale flexibility of HD-PTPCC, as has been suggested for the V domains of Alix29,35 and yeast Bro1.31 In addition, the open conformation exhibited by HD-PTPCC most likely extends to the entire Bro1-CC region, since the UBAP1C peptide binds equally well to HD-PTPCC and HD-PTPBro1-CC. Indeed,

HD-PTPBro1-CC is functionally competent as demonstrated by its binding to UBAP1 in cells and its ability to rescue defects in EGFR sorting caused by RNAi-mediated depletion of endogenous HD-PTP. Altogether, these new findings point to fundamental differences in how HD-PTP and Alix are regulated and how they control ESCRT function to define pathway diversity at different sub-cellular locations.

113 Experimental Procedures

Cloning, Protein Expression, and Purification

Constructs for HD-PTPBro1 (1–361), HD-PTPCC (363–712), and AlixV (358–702) were subcloned into a pNIC28a-Bsa4 vector (gift from Opher Gileadi; Addgene no. 26103).

HD-PTPBro1-CC (1–714) was cloned into a pET28a vector with restriction sites Nde1 and

Xho1. Point mutants were generated by quick-change primers using Phusion DNA polymerase (New England Biolabs). Constructs were expressed in BL21(DE3)

Escherichia coli using 0.1 mM isopropyl β-D-1 thiogalactopyranoside induction overnight at 20°C. Cells were lysed in 20 mM HEPES (pH 7.0), 500 mM NaCl, 10 mM imidazole, 2 mM phenylmethylsulfonyl fluoride by sonication, and the supernatant was clarified by centrifugation at 12,400 × g for 1 hr. Proteins were purified by affinity chromatography using nickel-beads (QIAGEN) pre-equilibrated in binding buffer

(20 mM HEPES, 500 mM NaCl, 10 mM imidazole [pH 7.4]) followed by anion- exchange chromatography with a Mono Q 5/50 GL column (GE Healthcare) in 20 mM

HEPES (pH 7.4), 2 mM EDTA, 2 mM DTT, and super elongation complex (SEC) using a Superdex200 column (GE Healthcare) in the same buffer. Incorporation of L- selenomethionine was achieved by growing a culture in M9 minimal medium (Molecular

Dimensions) supplemented with essential amino acids (100 mg/L each) and selenomethionine (80 mg/L).

Crystallization and Structure Determination

For crystallization, the His6 tag of HD-PTPCC was removed by cleavage with tobacco etch virus protease followed by nickel-affinity chromatography and further purification as above. apo-HD-PTPCC (11 mg/mL) was crystallized in 0.1 M Bis-Tris (pH 6.0), 0.1–

0.2 M Na-formate, and 13%–15% PEG3350 at 21°C. Crystals of the HD-PTPCC-UBAP1C complex were obtained by mixing HD-PTPCC (1 mg/mL) with 1 mM UBAP1 peptide and

114 concentrated to 11 mg/mL from a reservoir solution of 0.2 M KSCN, 20% PEG3350. All crystals were cryo-protected in perfluoropolyether cryo oil (Hampton Research) prior to freezing in liquid nitrogen. Data were collected at I02 and I03 beamlines in Diamond

40 Light Source (UK) and processed with XDS. The structure of apo-HD-PTPCC was determined by selenium single-wavelength anomalous diffraction (Se-SAD) using

PHENIX AutoSol, and the initial model built using PHENIX AutoBuild41 and COOT42 and refined using PHENIX Refine. The structure of HD-PTPCC-UBAP1C was determined by molecular replacement using the apo structure as the search model using PHENIX

Phaser and model building and refinement in COOT and PHENIX Refine.

EPR-DEER Measurements

Proteins were purified as describe above. A 15-fold molar excess of MTSL (Toronto

Research Chemicals) was mixed with the protein sample and incubated at 4°C overnight followed by SEC on a Superdex200 column (GE Healthcare) equilibrated in 20 mM

HEPES, 250 mM NaCl, 5 mM EDTA (pH 7.4). Labelling was confirmed by mass spectrometry. All mutants and the WT were checked by circular dichroism to confirm the secondary structure (Figure S1). DEER samples were prepared by buffer exchange into deuterated buffer (20 mM HEPES, 250 mM NaCl, in D2O [pD 7.4]) with 30% (v/v) glycerol-d8 to a final protein concentration of 60 μM and flash frozen in 4 mm quartz tubes (Wilmad). The four-pulse DEER experiments were carried out on a pulsed

ELEXSYS E580 (9 GHz) spectrometer (Bruker), cooled to 50 K with a continuous-flow helium CF935 cryostat and an isothermal calorimetry 502 temperature control system

(Oxford Instruments) and analyzed with DEERAnalysis 2013.2.36 Distance distribution predictions were calculated using MMM 2013.2 (Multiscale Modeling of

Macromolecular systems).43

Torsion-Angle Molecular Dynamics

115 CNS (Crystallography & NMR System, version 1.3)44 was used to run torsion-angle molecular dynamics simulations (TAMDS) of HD-PTPCC models with MTSL spin- labelled side chains at selected positions. Torsional angle flexibility was limited to the

MTSL side chains and the loop regions between H1 and H2 (residues 42–58) and between

H2 and H3 (residues 71–73), with the remaining residues treated as rigid. DEER distance constraints were introduced between the nitrogen atoms of the MTSL groups: 42.4 ± 10 Å for C628-C425 and 61.7 ± 10 Å for C425-C697.

NMR and PRE Measurements

UBAP1C peptide was dissolved at a concentration of 1 mg/mL (∼0.5 mM) in PBS in D2O

[pD 7.5]. 1H13C gradient-selected HSQC spectra were recorded at natural abundance 13C

(1%) at 800 MHz, using a Bruker AVANCE III spectrometer equipped with a TCI (1H-

13C-15N/2H) cryoprobe with z gradients. Pure-shift (PUSH) HSQC spectra45 were recorded for the uncomplexed peptide. Assignment of the H-C correlations was by

HSQC-total correlation spectroscopy and nuclear Overhauser effect spectroscopy spectra, and confirmed by the behaviour in the protein complexes. Complexation with

HD-PTPCC was detected with the addition of 60 μM of the protein. PRE measurements were recorded by taking MTSL-labelled HD-PTPCC C697S mutant under the same conditions, recording 1H13C HSQCs, and then reducing the MTSL label (rendering it diamagnetic) with 10-fold excess of sodium ascorbate (added from 1 M stock).

Biosensor Binding Studies

Binding studies were performed at 25°C on a multiplex system ProteOn XPR36 (Bio-

Rad Laboratories) in 10 mM HEPES (pH 7.4), 150 mM NaCl, 0.05% Tween 20 as running buffer, with 50–100 μg/mL of protein immobilized on an HTE chip (Bio-Rad

Laboratories). Peptide solutions (50 μL) were injected at 100 μL/min. Data were analyzed with ProteOn Manager software (Bio-Rad Laboratories), using the equilibrium binding 116 model: Response = [A] × Rmax/([A] + KD) where [A] is the analyte concentration and Rmax is the maximum response.

Yeast Two-Hybrid Analysis

HD-PTPBro1-CC, HD-PTPBro1-CC F678D, and AlixBro1-V cloned into pGBKT7 were used as described previously.11 UBAP1 (122–308) was cloned into pGADT7. Further deletion and missense mutations in UBAP1 as indicated were generated by standard PCR-based mutagenesis. Interactions were tested using the Matchmaker Gold system (Clontech) as described previously.11 Each triplicate experiment was repeated at least three times.

UBAP1 In Vitro Translation and Binding to HD-PTP

UBAP1-strep encoded on a pTriex5 vector11 was amplified using Pwo Polymerase

(Roche). UBAP1 RNA was synthesized from the PCR product using T7 RNA polymerase

(Promega). Protein was translated in nuclease-treated rabbit reticulocyte lysate (Promega) containing 35S-methionine (PerkinElmer), and 100 units of RNasin (Promega) for 1 hr at

30°C, followed by 10 min in the presence of 1 mM puromycin. Then 20 μL of translated protein was incubated with 5 μg His6- HD-PTPBro1-CC in 250 μL of immunoprecipitation

(IP) buffer (20 mM HEPES [pH 7.4], 100 mM NaCl, 1 mM MgCl2, 1% [w/v] Triton X-

100) for 2 hr at 4°C, then overnight with 3 μL of anti-His antibody (Clone HIS-1, Sigma).

Samples were incubated with 20 μL of protein A-Sepharose beads (Invitrogen) for 2 hr at 4°C, then washed three times in IP buffer.

Mitochondrial Targeting Experiments

Mitochondrial targeting experiments were performed as previously described.46 HeLa cells were transiently transfected with HD-PTPBro1-CC containing an N-terminal FKBP sequence and a C-terminal myc tag (cloned into pcDNA5), Mito-FRB (a gift from Martin

Lowe, Manchester, UK), and either GFP, WT UBAP1-GFP, or F268S UBAP1-GFP (in 117 pEGFP). Mitochondrial relocation of FKBP-HD-PTPBro1-CC-myc was induced by addition of 1 μM rapamycin (Sigma) for 3 hr. Cells were then prepared for immunofluorescence microscopy as above. Mitochondrial relocation of GFP-tagged constructs was scored in three independent experiments (100 cells counted per experiment).

Cell Culture, Transfection, and siRNA Rescue Experiments

HeLa cells were grown in DMEM with 1% NEAA, 10% fetal calf serum (HyClone;

Perbio) and 1% Pen-Strep Fugene 6 (Roche) was used for DNA transfections. Interferin

(QBiogene) was used for siRNA, using an HD-PTP nucleotide as previously published,3 or a non-targeting siRNA (Dharmacon) as a control. Efficient HD-PTP knockdown was confirmed by western blotting, as previously reported.3 For siRNA rescue experiments, cells were knocked down for 24 hr, then transfected with WT or specified mutants of HD-

PTPBro1-CC (note that the siRNA oligo targets a C-terminal sequence within HD-PTP) for

48 hr. Rescues were assessed by visual quantification of phenotypes. Cells containing clustered and strongly labelled foci of FK2 staining (previously identified as endosomal)3 were considered knocked down. Rescued cells displayed a WT, diffuse cytoplasmic distribution of FK2 staining. Scoring was performed for three independent experiments, with at least 100 transfected cells examined each time. Data were subjected to statistical analysis (two-way ANOVA) using Prism5 software (GraphPad). For graphical representation of rescue data, mean scores ± SD are provided for the three determinations.

Acknowledgments

This work was support by the Medical Research Council ( MR/K011049/1 to L.T. and

P.W.). G.H. is funded by a BBSRC Doctoral Training Studentship. A.J.F. thanks Bruker for sponsorship. All EPR experiments were carried out at the EPSRC National EPR

118 Research Facility & Service. We thank Diamond Light Source for access to beamlines

I02 and I03 (MX8997) and their staff for assistance, the staff of the Biomolecular

Analysis Facility of the Faculty of Biology Medicine and Health, and the staff at the

Advanced Photon Science facilities, University of Manchester. for assistance. We thank

Peristera Roboti for help with mitochondrial targeting experiments and Elaine Small for generating FKBP-HD-PTPBro1-CC-myc. We thank Melanie Vollmar, Simon Tanley and

Efrain Ceh Pavia for advice.

Accession Numbers

The atomic coordinates have been deposited in the PDB (PDB: 5LM1, 5LM2) and resonance assignments have been deposited in the Biological Magnetic Resonance Bank

(BMRB: 26914).

References for Chapter 4

1 K. Kok, S. L. Naylor and C. H. C. M. Buys, Adv. Cancer Res., 1997, 71, 27–92. 2 S. Toyooka, M. Ouchida, Y. Jitsumori, K. Tsukuda, A. Sakai, A. Nakamura, N. Shimizu and K. Shimizu, Biochem. Biophys. Res. Commun., 2000, 278, 671–678. 3 A. Doyotte, A. Mironov, E. McKenzie and P. Woodman, Proc. Natl. Acad. Sci., 2008, 105, 6308–6313. 4 H. Ma, P. Wardega, D. Mazaud, A. Klosowska-Wardega, A. Jurek, U. Engström, J. Lennartsson and C.-H. Heldin, Cell. Signal., 2015, 27, 2209–2219. 5 D.-Y. Chen, M.-Y. Li, S.-Y. Wu, Y.-L. Lin, S.-P. Tsai, P.-L. Lai, Y.-T. Lin, J.-C. Kuo, T.-C. Meng and G.-C. Chen, J. Cell Sci., 2012, 125, 4841–4852. 6 G. Lin, V. Aranda, S. K. Muthuswamy and N. K. Tonks, Genes Dev., 2011, 25, 1412–1425. 7 D. Kharitidi, P. M. Apaja, S. Manteghi, K. Suzuki, E. Malitskaya, A. Roldan, M.-C. Gingras, J. Takagi, G. L. Lukacs and A. Pause, Cell Rep., 2015, 13, 599–609. 8 M. Mariotti, S. Castiglioni, J. M. Garcia-Manteiga, L. Beguinot and J. A. M. Maier, Int. J. Biochem. Cell Biol., 2009, 41, 687–693. 9 S. Manteghi, M.-C. Gingras, D. Kharitidi, L. Galarneau, M. Marques, M. Yan, R. Cencic, F. Robert, M. Paquet, M. Witcher, J. Pelletier and A. Pause, Cell Rep., 2016, 15, 1893–1900. 10 N. Ali, L. Zhang, S. Taylor, A. Mironov, S. Urbé and P. Woodman, Curr. Biol., 2013, 23, 453–461. 11 F. Stefani, L. Zhang, S. Taylor, J. Donovan, S. Rollinson, A. Doyotte, K. Brownhill, J. Bennion, S. Pickering-Brown and P. Woodman, Curr. Biol., 2011, 21, 1245–1250. 12 J. H. Hurley, EMBO J., 2015, 34, 2398–2407. 119 13 C. Bissig and J. Gruenberg, Trends Cell Biol., 2014, 24, 19–25. 14 F. Ichioka, E. Takaya, H. Suzuki, S. Kajigaya, V. L. Buchman, H. Shibata and M. Maki, Arch. Biochem. Biophys., 2007, 457, 142–149. 15 J. Kim, S. Sitaraman, A. Hierro, B. M. Beach, G. Odorizzi and J. H. Hurley, Dev. Cell, 2005, 8, 937–947. 16 J. G. Carlton and J. Martin-Serrano, Science, 2007, 316, 1908–1912. 17 E. Morita, V. Sandrin, H.-Y. Chung, S. G. Morham, S. P. Gygi, C. K. Rodesch and W. I. Sundquist, EMBO J., 2007, 26, 4215–4227. 18 M. H. H. Schmidt, D. Hoeller, J. Yu, F. B. Furnari, W. K. Cavenee, I. Dikic and O. Bögler, Mol. Cell. Biol., 2004, 24, 8981–8993. 19 S. Sun, X. Zhou, W. Zhang, G. E. Gallick and J. Kuang, Biochem. J., 2015, 466, 475–487. 20 U. K. von Schwedler, M. Stuchell, B. Müller, D. M. Ward, H.-Y. Chung, E. Morita, H. E. Wang, T. Davis, G.-P. He, D. M. Cimbora, A. Scott, H.-G. Kräusslich, J. Kaplan, S. G. Morham and W. I. Sundquist, Cell, 2003, 114, 701–713. 21 M. D. J. Parkinson, S. C. Piper, N. A. Bright, J. L. Evans, J. M. Boname, K. Bowers, P. J. Lehner and J. P. Luzio, Biochem. J., 2015, 471, 79–88. 22 M. Agromayor, N. Soler, A. Caballe, T. Kueck, S. M. Freund, M. D. Allen, M. Bycroft, O. Perisic, Y. Ye, B. McDonald, H. Scheel, K. Hofmann, S. J. D. Neil, J. Martin-Serrano and R. L. Williams, Structure, 2012, 20, 414–428. 23 L. Wunderley, K. Brownhill, F. Stefani, L. Tabernero and P. Woodman, J. Cell Sci., 2013, jcs.140673. 24 Y. Cheng, N. E. Poulos, M. L. Lung, G. Hampton, B. Ou, M. I. Lerman and E. J. Stanbridge, Proc. Natl. Acad. Sci., 1998, 95, 3042–3047. 25 J. Qian, J. Yang, X. Zhang, B. Zhang, J. Wang, M. Zhou, K. Tang, W. Li, Z. Zeng, X. Zhao, S. Shen and G. Li, J. Cancer Res. Clin. Oncol., 2001, 127, 613–618. 26 S. Rollinson, P. Rizzu, S. Sikkink, M. Baker, N. Halliwell, J. Snowden, B. J. Traynor, D. Ruano, N. Cairns, J. D. Rohrer, S. Mead, J. Collinge, M. Rossor, E. Akay, R. Guerreiro, R. Rademakers, K. E. Morrison, P. Pastor, E. Alonso, P. Martinez-Lage, N. Graff-Radford, D. Neary, P. Heutink, D. M. A. Mann, J. Van Swieten and S. M. Pickering-Brown, Neurobiol. Aging, 2009, 30, 656–665. 27 J. Lee, K.-J. Oh, D. Lee, B. Y. Kim, J. S. Choi, B. Ku and S. J. Kim, PLoS ONE, 2016, 11, e0149113. 28 P. Sette, R. Mu, V. Dussupt, J. Jiang, G. Snyder, P. Smith, T. S. Xiao and F. Bouamr, Structure, 2011, 19, 1485–1495. 29 R. D. Fisher, H.-Y. Chung, Q. Zhai, H. Robinson, W. I. Sundquist and C. P. Hill, Cell, 2007, 128, 841–852. 30 S. Lee, A. Joshi, K. Nagashima, E. O. Freed and J. H. Hurley, Nat. Struct. Mol. Biol., 2007, 14, 194–199. 31 N. Pashkova, L. Gakhar, S. C. Winistorfer, A. B. Sunshine, M. Rich, M. J. Dunham, L. Yu and R. C. Piper, Dev. Cell, 2013, 25, 520–533. 32 Y. Kimura, M. Tanigawa, J. Kawawaki, K. Takagi, T. Mizushima, T. Maeda and K. Tanaka, Eukaryot. Cell, 2015, 14, 976–982. 33 Q. Zhai, R. D. Fisher, H.-Y. Chung, D. G. Myszka, W. I. Sundquist and C. P. Hill, Nat. Struct. Mol. Biol., 2008, 15, 43–49. 34 J. Walshaw and D. N. Woolfson, J. Mol. Biol., 2001, 307, 1427–1450. 120 35 R. Pires, B. Hartlieb, L. Signor, G. Schoehn, S. Lata, M. Roessle, C. Moriscot, S. Popov, A. Hinz, M. Jamin, V. Boyer, R. Sadoul, E. Forest, D. I. Svergun, H. G. Göttlinger and W. Weissenhorn, Structure, 2009, 17, 843–856. 36 G. Jeschke, V. Chechik, P. Ionita, A. Godt, H. Zimmermann, J. Banham, C. R. Timmel, D. Hilger and H. Jung, Appl. Magn. Reson., 2006, 30, 473–498. 37 Q. Zhai, M. B. Landesman, H. Robinson, W. I. Sundquist and C. P. Hill, J. Virol., 2011, 85, 632–637. 38 G. M. Clore and J. Iwahara, Chem. Rev., 2009, 109, 4108–4139. 39 X. Robert and P. Gouet, Nucleic Acids Res., 2014, 42, W320-324. 40 W. Kabsch, Acta Crystallogr. D Biol. Crystallogr., 2010, 66, 125–132. 41 P. D. Adams, P. V. Afonine, G. Bunkóczi, V. B. Chen, I. W. Davis, N. Echols, J. J. Headd, L.-W. Hung, G. J. Kapral, R. W. Grosse-Kunstleve, A. J. McCoy, N. W. Moriarty, R. Oeffner, R. J. Read, D. C. Richardson, J. S. Richardson, T. C. Terwilliger and P. H. Zwart, Acta Crystallogr. D Biol. Crystallogr., 2010, 66, 213–221. 42 P. Emsley, B. Lohkamp, W. G. Scott and K. Cowtan, Acta Crystallogr. D Biol. Crystallogr., 2010, 66, 486–501. 43 Y. Polyhach, E. Bordignon and G. Jeschke, Phys. Chem. Chem. Phys., 2011, 13, 2356–2366. 44 A. T. Brunger, Nat. Protoc., 2007, 2, 2728–2733. 45 L. Paudel, R. W. Adams, P. Király, J. A. Aguilar, M. Foroozandeh, M. J. Cliff, M. Nilsson, P. Sándor, J. P. Waltho and G. A. Morris, Angew. Chem. Int. Ed., 2013, 52, 11616–11619. 46 K. Sato, P. Roboti, A. A. Mironov and M. Lowe, Mol. Biol. Cell, 2015, 26, 537–553.

121 Chapter 5: Paper 2

Investigating the conformational flexibility of HD-PTP CC domain

Graham Heaven1, Colin Levy1, Philip Woodman2, Lydia Tabernero2 and Alistair J. Fielding1,*

1 School of Chemistry, Faculty of Science and Engineering 2 School of Biological Sciences, Faculty of Biology Medicine and Health University of Manchester, Manchester, M13 9PL, UK. * correspondence: [email protected]

Target journal: Biophysical journal

122 Abstract

Conformational flexibility is important for the regulation of the membrane remodelling

ESCRTs (endosomal sorting complexes required for transport). Both ESCRT-III and

Bro1 family protein Alix use mechanisms of autoinhibition where regions fold back to inhibit activity until activation by recruiting binding partners. Alix also has a V shaped coiled-coil “V domain” which can undergo rearrangement into an extended crescent shaped domain. We recently reported the X-ray crystal structure of Alix homologue HD-

PTP (his domain-protein tyrosine phosphatase) coiled-coil domain, showing that it forms an open conformation both in its crystal structure and in solution shown by DEER spectroscopy. Here we follow up this work, using DEER spectroscopy to probe the conformations of HD-PTP CC domain in the presence of binding partners. Our results show that although there is limited flexibility in this domain, binding of Ubap1 or monoubiquitin cause no changes to conformation.

Introduction

Endosomal Sorting Complexes Required for Transport (ESCRTs) are protein complexes which remodel membranes with diverse functions in the cell, ranging from sorting of ubiquitinated receptors into intralumenal vesicles at the multivesicular body,1,2 cell membrane repair,3 nuclear envelope pore formation,4,5 midbody abscission in cell division6 and exosome formation.7 There are four major ESCRT complexes: ESCRT-0,

ESCRT-I, ESCRT-II and ESCRT-III which act together with a range of accessory proteins including VPS4 ATPase, deubiquitinases and Bro1family ESCRT-adaptor proteins to carry out their specified function.

The Bro1 family has members across organisms from yeast to humans. They are multidomain proteins with various ESCRT interaction motifs (Figure 5.1). Common to all Bro1 proteins is the N-terminal Bro1 domain which interact with ESCRT-III

123 CHMP48–10 in humans or Sn7 in yeast.11,12 Bro1, Rim20, Alix and HD-PTP also have coiled-coil domains which are referred to as V domains in Bro1, Rim20 and Alix, and a

CC domain in HD-PTP. The coiled-coil domains are followed by proline rich regions

(PRR) which act as the C-terminal domains for all except HD-PTP which also has an additional protein tyrosine phosphatase domain (PTP) and PEST (proline glutamate serine threonine rich) region.

The coiled-coil (CC/ V) domains all share a long region of sequence similarity which act

13 to recognize corresponding YPXnL motifs in their specific protein partners: HD-PTPCC domain binds Ubiquitin-associated protein 1 (Ubap1) which is a component of the MVB-

14,15 specific ESCRT-I and a risk factor in frontal temporal lobar degeneration. AlixV domain binds syntenin7 for recruitment to exosome formation and also binds several viral gag domains which allows the ESCRT system to be hijacked for viral budding.16,17 The

18 yeast Bro1V domain binds Regulator for free ubiquitin chains (Rfu1) and yeast Rim20V domain binds Rim101.12 There is also evidence that HD-PTP, Alix and yeast Bro1 coiled- coil domains bind to ubiquitin, which link them to ubiquitinated cargo.19,20

The structures of the V domains of Alix (Figure 5.2A) and Bro1 have both been solved by X-ray crystallography, showing V-shaped structures with two trihelical bundles connected by three loops at the vertex of the V.20,21 The CC domain of HD-PTP was recently solved by us (Figure 5.2D), which showed a much more extended non-V conformation; it still has two trihelical bundles but these are connected by two looped regions and one continuous helix which braces the arms into an open conformation.15

124

Figure 5.1. Bro1 family proteins in Homo sapiens (hs) and Saccharomyces cerevisiae (sc). Domain boundaries are from uniprot database (http://www.uniprot.org/) or from crystal structures where available. Interacting partners are shown above each domain in italics.

Apart from their YPXnL motif-recognition regions, there is no other significant stretch of similarity between all four CC domains. Crucially, there is no significant similarity between the hinge regions of Alix V and the corresponding regions in Bro1 V domain or

HD-PTP CC domain, which means it is difficult to predict whether other homologues like

Rim20 will contain a closed V-like or open CC-like structure.

Regulation of Alix activity has been shown to involve autoinhibition, dimerization and opening of the V domain. Experiments showed that the presence of the PRR causes blocked binding to the V domain and to the Bro1 domain,22–24 and that this can be alleviated by a number of processes including PRR binding to ALG-2 or CEP55 and PRR

125 phosphorylation.25 Evidence for Alix dimerization from both human cell expressed26 and

E.coli expressed27 Alix. When the monomers and dimers purified from E.coli were analysed by SAXS, it was shown that the monomer has the same shape as the crystal structure however the dimeric structure was consistent with crescent shaped models where the central V domain opened up across the hinge.

For HD-PTPCC, on the other hand, there is no evidence for dimerization nor is there any evidence for conformational switching between CC and V shapes. However, our previous investigation using distance measurements through spin labelling and double electron- electron resonance (DEER) spectroscopy, showed some deviation from the crystal structure, suggesting localised conformational change around its second helix.15

Here, we look in more detail at the structure and flexibility of AlixV and HD-PTPCC domains. We investigate further the helix movement reported in our previous study of

HD-PTP and, by using DEER and X-ray crystallography, monitor the conformation of

HD-PTPCC in the presence of its binding partners Ubap1 and ubiquitin.

Results

Variation in crystal structure conformations of Alix and HD-PTP

Currently, there is a total of ten crystal structures of Alix V domain in the protein data bank (including cases where there are multiple molecules in their asymmetric units).

Three of these are isolated Alix V domains (apo);16,21 one is a multidomain BroV (apo) structure;16 four are also multidomain BroV structures but co-crystallised with YPX(n)L ligands;17,28 and two of these are of Alix V in a right-angle conformation, obtained when co-crystallisation with monoubiquitin was attempted.20 It has previously been acknowledged that the V domain structures exhibit a degree of flexibility primarily about the V domain, meaning that each arm can be aligned separately but the other will fan out at a different angle (Figure 5.2A).

126 We compared the root mean squared deviations of all the Alix V domains structures

(Figure 5.2B). We noticed that although there is up to 1.61 Å rmsd between the apo Alix

V crystal structures, when directly comparing the BroV apo and gag structures there is only 0.67 Å rmsd and between the four different BroV-gag structures there is only a maximum of 0.55 Å rmsd. This suggests that the crystallisation of the isolated V domain may exhert forces on the structure to deviate from its native conformation, but when in the presence of its N-terminal Bro1 domain the V structure is stabilized into a more rigid

V structure (Figure 5.2C). The remaining “open conformation” Alix structures display significant deviation (>6 Å) to all other structures and the mechanistic origin of this conformational change is not currently known.

For the HD-PTP CC domain there are two apo structures and one ligand bound structure.

Despite not being a V shape domain with any obvious hinge region, the apo HD-PTP CC structures, like Alix V, have a large rmsd; small deviations throughout the structure means that when the N terminal ends of the CC structures are aligned there is a significant swing in the orientation of the C-terminus (Figure 5.2D). There is also a large difference between the apo and Ubap1-bound structures, which is not observed in Alix, however comparison with Alix is limited because the only crystal structures of Alix with ligands are in the context of BroV.

127

Figure 5.2. Conformational variation in crystal structures of Alix V and HD-PTP CC. (A) Alix V domain apo (green) (PDB:2OJQ21,2OEXa,b16); and open V upon crystallisation with monoubiquitin (yellow) (PDB:4JJYa,b20). (B) tables for maximum root mean squared deviation of Alix V and HD-PTP CC crystal structures calculated using Pymol. (C) Alix BroV apo (blue) (PDB:2OEV16); and BroV(red) in complex with gag peptides (orange) (PDB:2R02,2R0317,2XS1,2XS829). d) HD-PTP CC apo (blue) (PDB:5LM2a,b15); and CC (red) in complex with Ubap1 peptide (orange) 15 (PDB:5LM1 ).

Probing ligand-induced conformational changes in HD-PTPCC using DEER

In our previous work, we used DEER spectroscopy to study inter-spin distances in HD-

PTPCC. This method measures the dipolar coupling interaction between spin label pairs, and distances are derived using the inverse cube relationship between dipolar interaction frequency and distance. Labelling of the three native cysteines, (HD-PTPCC WT) gave two peaks in the DEER distance distribution; these were deconvoluted by mutagenesis to

15 correspond to C628-C425 at 4.2 nm and C425-C697 at 6.3 nm. The third C628-C697 spin label pair did not give a detectable dipolar coupling response which, for the window length of

128 the DEER experiment, indicated a distance of at least 7 nm, consistent with the extended shape determined by crystallography.

Simulation of the expected distance distributions by in silico labelling of the crystal structure gave a large discrepancy to DEER derived distance (4.2 nm predicted vs 5.4 nm measured). We accounted for this by simulating small movements in the helix 2, which in the structure is flanked by two loops. We wanted to investigate this discrepancy further, and analysed in more detail the three HD-PTPCC structures.

One possible cause for discrepancy between solution and crystal structures is crystal packing. In the three crystal structures of HD-PTPCC, there is a common packing interface between CC and its symmetry related molecules (Figure 5.3). This interface involves a surface on helix 2, which could explain why this helix 2 is pushed down in the crystal structures whereas it appears to move in a wide arc based on DEER distance measurements.

For this reason, DEER is a suitable method to measure further changes in this helix conformation which would not be observed by crystallography. Since the wild type distribution has two well separated peaks we chose to use this system as a conformational probe upon substrate binding.

129

Figure 5.3. Crystal packing against HD-PTPCC helix 2. (A) Crystal packing in CC apo crystal structure. (B) Crystal packing in CC-Ubap1 crystal structure.

Our previous work investigated a Ubap1 peptides with boundaries 261-280 and crystallography showed a clear binding motif involving residues 262-271. We noticed that the end of N terminal end of Ubap1 was ~ 2 nm from the flexible helix 2 so we chose to investigate peptides which included the binding motif and additional residues towards the Ubap1 N-terminal end: residues 250-272. This region of Ubap1 is of interest because it contains a proline residue which has been shown to be mutated to leucine (P256L) in

14 patients with frontotemporal lobar degeneration. We added Ubap1250-272 WT and P256L peptides to triple MTSL-labelled HD-PTPCC and measured DEER.

130 DEER spectra and resulting distance distributions were very similar for the apo and

Ubap1-bound, giving no evidence for any conformational change (Figure 5.4). Neither did the P256L mutated Ubap1 give any significant change.

Since Alix had undergone such a large scale rearrangement, by crystallography, in the presence of monoubiquitin we also tested this HD-PTPCC DEER sample. The measured spectrum was noisier than the others, but it was still possible to make out the bimodal distance distribution with no significant change from the apo or Ubap1-bound samples.

The peak at the 8 nm limit of the distance distribution, seen in all measurements, is indicative of a larger distance beyond the upper measurable limit; which is most likely a small contribution from the third long distance between C628 and C697.

131

Figure 5.4 HD-PTPCC DEER with binding partners. (A) Apo, (B) Ubap1 wt, (C) Ubap1 (P256L) and (D) monoubiquitin. Left) Raw DEER traces and inset graphs showing form factor fits after exponential background correction. Right) Distance distributions calculated using DEERAnalysis.30 Shaded regions give an estimate of error for each distance, based on signal-to-noise in the DEER traces and uncertainty in the background correction. 132 Crystallography of HD-PTPCC with extended Ubap1 peptides

We also investigated the Ubap1250-272 WT and P256L peptides using X-ray crystallography. Both peptides when combined with purified HD-PTPCC gave crystals in a commercial screen PACT screen. X-ray diffraction data were recorded and we were able to phase the data using molecular replacement, with the previous HD-PTP crystal structure as a search model (Figure 5.5A).

The original crystal structure showed binding of Ubap1 residues N262-L271.15 In these new structures no significant extra density was visible with N terminally extended peptides, and it was only possible to map the exact same length of Ubap1 as found in our previous work (Figure 5.5B).

Between the three crystal structures of CC-Ubap1 there is 0.437 Å rmsd (Figure 5.5C) which might suggest that Ubap1 bound CC has restricted flexibility whereas the apo CC crystal structures had 2.09 Å rmsd (Figure 5.2B). However, since all three crystals were in the same space group and have similar unit cell dimensions it is most likely that this is the most stable crystal form, and is simply a repeat experiment of the same crystallisation.

The additional N-terminal Ubap1 extensions presumably form disordered extensions without impacting the crystallisation, in neither the WT or P256L variant.

133 Table 5.1. Crystallographic Data Collection and Refinement Statistics

HD-PTPCC-Ubap250-272wt HD-PTPCC–UBAP1250-272P256L

Data Collection

Space group P21 21 21 P21 21 21

Cell dimensions

a, b, c (Å) 48.9, 93.4, 102.5 49.2, 91.1, 102.9

α, β, γ (°) 90, 90, 90 90, 90, 90

Molecules per asymmetric unit 1 1

Resolution (Å) 2.58 (2.59–2.58)a 2.64 (2.64-2.65)

Rmerge 0.08 (0.96) 0.04 (1.02)

I/σI 15.0 (2.1) 17.1 (1.6)

Completeness (%) 100 (98) 100 (97)

Redundancy 6.5 (6.7) 4.9 (5.2)

Refinement

Resolution (Å) 69.11 - 2.58 68.29 - 2.64

No. of reflections 15315 / 765 14155 / 687

Rwork/Rfree 0.230/ 0.295 0.229/ 0.289

No. of atoms

Protein 5340 5340

Peptide 181 171

B factors

Protein 66.8 88.7

Peptide 90.9 121.7

Root-mean-square deviations

Bond lengths (Å) 0.0121 0.0122

Bond angles (°) 1.493 1.53

Each structure was determined from one crystal. aValues in parentheses are for the highest-resolution shell.

134

Figure 5.5. Crystallography of HD-PTPCC with Ubap1 peptides. A) All three HD- PTPCCUbap1 structures overlaid. (B) Close up of Ubap1 binding region, contoured to 0.8 sigma. (C) rmsd between all three structures.

135 Discussion

The ESCRT-adaptor Bro1 protein Alix displays significant conformational flexibility in its coiled-coil V domain, as indicated by the large rmsd between aligned crystal

16,21 structures. Additionally, AlixV domain has been shown to undergo largescale structural changes in solution which are linked to its activation for the recruitment of

25,27,29 ESCRT-III. Similarly, crystal structures of HD-PTPCC also display a large rmsd.

Although it was not clear whether this would also correspond to potential large-scale changes observed with AlixV. Our previous study using DEER showed that there was some localized conformational change in the helix 2 of HD-PTP but did not support a V- like closing mechanism.

We have now investigated the effects of HD-PTP’s binding partners Ubap1 and monoubiquitin. The results show that the conformational change observed by DEER spectroscopy is not modulated by binding of Ubap1, neither is it affected by FTLD-linked mutation in FTLD. Furthermore, addition of monoubiquitin, which generated an 90⁰ Alix

V structure (Figure 5.2A), caused no significant changes in HD-PTPCC DEER spectra ruling out any similar conformational rearrangements.

Our inspection of the AlixBro1-V apo versus viral gag peptide-bound structures showed that there was no induced conformational change. Although the crystal structures of

HD-PTPCC apo and Ubap1-bound structures do show a fairly large rmsd, it is no more variation than in the HD-PTPCC apo structures. Observation that the AlixV is stabilized in conformation in the presence of the Bro1 shows that for better comparison, HD-PTP would have to be investigated in the presence of its N-terminal Bro1 domain.

It is becoming apparent that HD-PTPCC is structurally distinct to Alix. Although solution structures of Alix, show an open V structure which bears some resemblance to HD-PTPCC the inverse relationship with HD-PTPCC displaying V-like character is not supported. The

136 opening of Alix into its extended conformation allows it to become activated for the recruitment of binding partners to its V and Bro1 domains.22,25,29 If HD-PTP is constitutively open in structure, alternative mechanisms for regulation will be required.

These will likely rely upon the order of interactions of competing binding partners as suggested previously, and also the distinction from Alix that HD-PTP can bind to

ESCRT-0 component STAM2.31

Methods

Expression and purification and spin labelling for DEER spectroscopy

15 HD-PTPCC in pNIC-28 Bsa4 was used from our previous work. The HD-PTPCC plasmid was transformed into BL21(DE3) cells and expressed by induction with 0.1 mM IPTG with shaking at 20 ⁰C for 18 hrs.

HDPTPCC cell pellets (~10 g after expression by autoinduction) were resuspended in 20 mM HEPES pH 7.4, 500 mM NaCl, 2 mM DTT, 0.1 % Triton x-100, 0.5 mg/mL lysozyme, and a Complete protease inhibitor tablet. For lysis this was passed through a

French Press at 1000 PSI three times. After centrifugation at 13,000 rpm lysates were purified by nickel affinity chromatography. Elution fractions were combined, diluted with

20 mM HEPES pH 7.4, 5 mM EDTA, 2 mM DTT and loaded onto a MonoQ 5/50 GL

(GE Healthcare) column eluting with a gradient of 0-500 mM NaCl in the same buffer but with no DTT. A 10-fold molar excess of MTSL (methanethiosulfonate spin label)

(Toronto Research Chemicals) was then added from a 100 mM stock solution in methanol, diluting if necessary to ensure that the overall volume of methanol in the sample did not exceed 1 %. This mixture was incubated at 4 °C overnight. A further 5- fold excess of MTSL was added and then the spin labelled protein was loaded onto a

Superdex200 column equilibrated in 20 mM HEPES, 250 mM NaCl, 5 mM EDTA, pH

7.4. Labelling was previously confirmed by mass spectrometry.

137 Double electron-electron resonance (DEER) sample preparation. After size-exclusion the buffer was exchanged into deuterated buffer (20 mM HEPES, 250 mM NaCl, pD 7.4 = pH 7.0 using a standard pH probe). DEER samples were prepared with 30% (v/v) glycerol-d8 to a final protein concentration of 60 μM. 120 μL samples were frozen inside

4 mm quartz tubes (Wilmad) by flash freezing with liquid nitrogen, and stored in a liquid nitrogen dewar until measurement.

Purification for crystallization

Protein was purified by affinity chromatography as described above and the histidine tag was cleaved by addition of TEV protease (1 mg/mL, 500 μL). After overnight cleavage this was diluted with 20 mM HEPES pH 7.4 and loaded back onto the HisTrap. The low imidazole/ low salt flowthrough of the HisTrap was then loaded onto a MonoQ 5/50 GL

(GE Healthcare) column eluting with a gradient of 200-500 mM NaCl (in 20 mM HEPES pH 7.4, 2 mM EDTA, 2 mM DTT). This was concentrated to 50 μL and loaded onto a

Superdex200 column equilibrated in 20 mM HEPES pH 7.4, 250 mM NaCl, 2 mM

EDTA, 2 mM DTT.

After size-exclusion protein was mixed with 2 mM Ubap1 peptides (250-272 and 250-

272 with P256L mutation, synthesised by Generon) before concentrating to an absorbance of ~ 10 at 280 nm (286 μM). Using a mosquito®, sitting drop plates were set up with 30 μL reservoir solution and 200 nL protein plus 200 nL reservoir drops. For seeded trials, crystal drops were suspended with 20 μL of mother liquor vortexed with an additional 180 μL reservoir solution with a seed bead. Seeded plates were set up with 30

μL reservoir solution and 180 nL protein plus 20 nL seed stock plus 200 nL reservoir drops.

138 DEER spectroscopy

DEER experiments were carried out on a pulsed ELEXSYS E580 (9 GHz, “X-band”) spectrometer (Bruker), cooled to 50 K with a continuous-flow helium CF935 cryostat

(Oxford Instruments) and a ITC 502 temperature control system (Oxford Instruments).

The four-pulse DEER sequence π/2νobs−τ1−π νobs −t−πνpump−(τ1+τ2−t)−πνobs−τ2−echo was applied, with π/2νobs pulse length of 16 ns, πνobs pulse length of 32 ns and πνpump pulse length of 32 ns. Pump pulses were applied at the maximum of the field sweep spectrum with the observe pulses 65 MHz lower. τ1 was varied by incrementing the first πνobs pulse position over eight steps of 56 ns for averaging of the deuterium nuclear modulation. Phase-cycling was applied. DEERAnalysis2013.2 was used to subtract the exponential background decay due to intermolecular interactions and to calculate the interspin distance distribution by Tikhonov regularization. Distance distribution simulations were calculated by in silico spin labelling of CC crystal structures or models using the R1 side chain rotamer library incorporated into program MMM 2013.2

(Multiscale Modelling of Macromolecular systems).

Acknowledgements

This work was support by Medical Research Council (MR/K011049/1 to L.T. and P.W).

G.H. is funded by a BBSRC DTP Studentship. A.J.F. thanks Bruker for sponsorship. All

EPR experiments were carried out at the EPSRC National EPR Research Facility &

Service. We thank the staff of the Biomolecular Analysis Facility of the Faculty of

Biology Medicine and Health, and the staff at the Advanced Photon Science facilities,

University of Manchester. for assistance.

139 References for Chapter 5

1 P. I. Hanson and A. Cashikar, Annu. Rev. Cell Dev. Biol., 2012, 28, 337–362. 2 J. McCullough, L. A. Colf and W. I. Sundquist, Annu. Rev. Biochem., 2013, 82, 663– 692. 3 A. J. Jimenez, P. Maiuri, J. Lafaurie-Janvore, S. Divoux, M. Piel and F. Perez, Science, 2014, 343, 1247136. 4 C.-P. Lee, P.-T. Liu, H.-N. Kung, M.-T. Su, H.-H. Chua, Y.-H. Chang, C.-W. Chang, C.-H. Tsai, F.-T. Liu and M.-R. Chen, PLOS Pathog., 2012, 8, e1002904. 5 B. M. Webster, P. Colombi, J. Jäger and C. P. Lusk, Cell, 2014, 159, 388–401. 6 J. G. Carlton and J. Martin-Serrano, Science, 2007, 316, 1908–1912. 7 M. F. Baietti, Z. Zhang, E. Mortier, A. Melchior, G. Degeest, A. Geeraerts, Y. Ivarsson, F. Depoortere, C. Coomans, E. Vermeiren, P. Zimmermann and G. David, Nat. Cell Biol., 2012, 14, 677–685. 8 J. McCullough, R. D. Fisher, F. G. Whitby, W. I. Sundquist and C. P. Hill, Proc. Natl. Acad. Sci., 2008, 105, 7687–7691. 9 F. Ichioka, E. Takaya, H. Suzuki, S. Kajigaya, V. L. Buchman, H. Shibata and M. Maki, Arch. Biochem. Biophys., 2007, 457, 142–149. 10 F. Ichioka, R. Kobayashi, K. Katoh, H. Shibata and M. Maki, FEBS J., 2008, 275, 682–692. 11 J. Kim, S. Sitaraman, A. Hierro, B. M. Beach, G. Odorizzi and J. H. Hurley, Dev. Cell, 2005, 8, 937–947. 12 J. H. Boysen and A. P. Mitchell, Mol. Biol. Cell, 2006, 17, 1344–1353. 13 Y. Kimura, M. Tanigawa, J. Kawawaki, K. Takagi, T. Mizushima, T. Maeda and K. Tanaka, Eukaryot. Cell, 2015, 14, 976–982. 14 S. Rollinson, P. Rizzu, S. Sikkink, M. Baker, N. Halliwell, J. Snowden, B. J. Traynor, D. Ruano, N. Cairns, J. D. Rohrer, S. Mead, J. Collinge, M. Rossor, E. Akay, R. Guerreiro, R. Rademakers, K. E. Morrison, P. Pastor, E. Alonso, P. Martinez-Lage, N. Graff-Radford, D. Neary, P. Heutink, D. M. A. Mann, J. Van Swieten and S. M. Pickering-Brown, Neurobiol. Aging, 2009, 30, 656–665. 15 D. Gahloth, C. Levy, G. Heaven, F. Stefani, L. Wunderley, P. Mould, M. J. Cliff, J. Bella, A. J. Fielding, P. Woodman and L. Tabernero, Structure, 2016, 24, 2115–2126. 16 R. D. Fisher, H.-Y. Chung, Q. Zhai, H. Robinson, W. I. Sundquist and C. P. Hill, Cell, 2007, 128, 841–852. 17 Q. Zhai, R. D. Fisher, H.-Y. Chung, D. G. Myszka, W. I. Sundquist and C. P. Hill, Nat. Struct. Mol. Biol., 2008, 15, 43–49. 18 Y. Kimura, J. Kawawaki, Y. Kakiyama, A. Shimoda and K. Tanaka, J. Biol. Chem., 2014, 289, 21760–21769. 19 N. Pashkova, L. Gakhar, S. Winistorfer and R. Piper, Mol. Biol. Cell. 20 N. Pashkova, L. Gakhar, S. C. Winistorfer, A. B. Sunshine, M. Rich, M. J. Dunham, L. Yu and R. C. Piper, Dev. Cell, 2013, 25, 520–533. 21 S. Lee, A. Joshi, K. Nagashima, E. O. Freed and J. H. Hurley, Nat. Struct. Mol. Biol., 2007, 14, 194–199. 22 X. Zhou, S. Pan, L. Sun, J. Corvera, S.-H. Lin and J. Kuang, Biochem. J., 2008, 414, 215–220.

140 23 X. Zhou, S. Pan, L. Sun, J. Corvera, Y.-C. Lee, S.-H. Lin and J. Kuang, Biochem. J., 2009, 418, 277–284. 24 X. Zhou, J. Si, J. Corvera, G. E. Gallick and J. Kuang, Biochem. J., 2010, 432, 525– 538. 25 S. Sun, L. Sun, X. Zhou, C. Wu, R. Wang, S.-H. Lin and J. Kuang, Dev. Cell, 2016, 36, 331–343. 26 S. Sun, X. Zhou, W. Zhang, G. E. Gallick and J. Kuang, Biochem. J., 2015, 466, 475–487. 27 R. Pires, B. Hartlieb, L. Signor, G. Schoehn, S. Lata, M. Roessle, C. Moriscot, S. Popov, A. Hinz, M. Jamin, V. Boyer, R. Sadoul, E. Forest, D. I. Svergun, H. G. Göttlinger and W. Weissenhorn, Structure, 2009, 17, 843–856. 28 Q. Zhai, M. B. Landesman, H. Robinson, W. I. Sundquist and C. P. Hill, J. Virol., 2011, 85, 632–637. 29 Q. Zhai, M. B. Landesman, H.-Y. Chung, A. Dierkers, C. M. Jeffries, J. Trewhella, C. P. Hill and W. I. Sundquist, J. Virol., 2011, 85, 9222–9226. 30 G. Jeschke, V. Chechik, P. Ionita, A. Godt, H. Zimmermann, J. Banham, C. R. Timmel, D. Hilger and H. Jung, Appl. Magn. Reson., 2006, 30, 473–498. 31 N. Ali, L. Zhang, S. Taylor, A. Mironov, S. Urbé and P. Woodman, Curr. Biol., 2013, 23, 453–461.

141 Chapter 6: Paper 3

The open architecture of HD-PTP phosphatase provides new insights into the mechanism of regulation of ESCRT function Deepankar Gahloth1, Graham Heaven2, Thomas A. Jowitt1,3, A. Paul Mould1,3, Jordi Bella1, Clair Baldock1, Philip Woodman1 and Lydia Tabernero1,*

1School of Biological Sciences, Faculty of Biology Medicine and Health, University of Manchester, Manchester Academic Health Science Centre 2School of Chemistry and Photon Science Institute, University of Manchester 3Biomolecular Analysis Core Facility, Faculty of Biology Medicine and Health, University of Manchester, Manchester Academic Health Science Centre

*Corresponding author. School of Biological Sciences, Faculty of Biology Medicine and Health, University of Manchester, Manchester Academic Health Science Centre, Manchester, M13 9PT, UK. e-mail: [email protected]

Submitted to: Scientific Reports

142 Abstract

HD-PTP is a tumour suppressor phosphatase that controls endocytosis, down-regulation of mitogenic receptors and cell migration. Central to its role is the specific recruitment of critical endosomal sorting complexes required for transport (ESCRTs). However, the molecular mechanisms that enable HD-PTP to regulate ESCRT function are unknown.

We have characterised the molecular architecture of the entire ESCRT binding region of

HD-PTP using small angle X-ray scattering and hydrodynamic analyses. We show that

HD-PTP adopts an open and extended conformation, optimal for concomitant interactions with multiple ESCRTs, which contrasts with the compact conformation of the related

ESCRT regulator Alix. We demonstrate that HD-PTP open conformation is functionally competent for binding cellular protein partners. Our analyses rationalise the functional cooperation of HD-PTP with ESCRT-0, ESCRT-I and ESCRT-III and provide a model for regulation of ESCRT function by displacement of ESCRT subunits, which is crucial in determining the fate of ubiquitinated cargo.

Introduction

His Domain Protein Tyrosine Phosphatase (HD-PTP) is essential for the lysosomal degradation of multiple membrane receptors, including activated EGFR and PDGFR-β,1,2

MHC Class I3 and α5β1 integrin.4 This process requires sorting of the ubiquitinated receptors to intralumenal vesicles (ILVs) within the multivesicular body (MVB), and their subsequent delivery to the lysosome.5 Central to HD-PTP function during MVB sorting is the binding to specific endosomal sorting complexes required for transport

(ESCRTs).1,6,7 However, little is known at the molecular level about how HD-PTP regulates ESCRT function.

ESCRTs are multimeric protein complexes (numbered 0, I, II, III) that control many membrane remodelling and scission events critical to cell physiology.8,9 These events

143 include ILV formation at the MVB [5], midbody abscission during cytokinesis,10,11 exosome formation,12 autophagy,13 plasma membrane wound repair,14 nuclear envelope remodelling15,16 and neuron pruning.17,18 Viruses also exploit the ESCRT machinery to facilitate virion budding.19

Different subsets of ESCRTs and specialised adaptor proteins define pathway selectivity but they all culminate in the assembly of membrane sculpting ESCRT-III polymers at the point of membrane scission.8,9,20,21 The mechanisms that regulate the different ESCRT pathways are still poorly understood and deciphering how ESCRT-III assembly is ultimately coordinated at each location remains an open question. HD-PTP and the related

ESCRT adaptor protein, Alix, are central to such functional specialisation. Alix directly regulates ESCRT-III recruitment and assembly at a range of sites.10,11,21–23 In contrast,

HD-PTP function is largely restricted to MVB biogenesis.1,3

HD-PTP has a multidomain organisation24 (Fig. 6.1), with an N-terminal Bro1 domain that binds ESCRT-0/STAM2 and ESCRT-III/CHMP4,7,25–28 followed by a coiled-coil domain (CC) that interacts with the ESCRT-I subunit UBAP1,6,29 a proline-rich region

(PRR) with binding sites for the ESCRT-I subunit TSG10127 as well as a further site for

STAM2, a protein tyrosine phosphatase (PTP) domain, and a C-terminal PEST domain.

Alix shares a similar Bro1-CC-PRR organisation, including binding sites for CHMP4 and

TSG101, but lacks the PTP and PEST domains (Fig. 6.1).

144

Figure 6.1. Domain organisation of HD-PTP and Alix. Residue numbers indicate domain boundaries and arrows show locations of binding for the different interacting partners. Domain names: Bro1; CC, coiled-coil; PRR, proline rich region; PTP, protein tyrosine phosphatase; PEST, rich in proline (P), glutamate (E), serine (S) and threonine (T). The proximal region of PRR in HD-PTP is indicated as a brown box (residues 714-738).

Alix function is highly regulated and it exists in the cytoplasm as an inactive form that is prevented from binding to ESCRT-III and viral Gag protein substrates.30,31 Self-inhibition is favoured by the compact architecture of the Bro1-CC region, where the CC domain adopts a closed V shape32 that brings the proximal region of the PRR to interact with the

Bro1 domain.33 This arrangement blocks access of CHMP4 and Gag proteins to their binding sites on the Alix surface. Alix activation occurs by several mechanisms including phosphorylation,34 binding of the membrane adaptors CEP5535 and ALG-234 to the PRR, and subsequent conformational rearrangements and dimerization.36,37

The role of HD-PTP as a tumour suppressor is incompatible with a self-inhibited inactive form as the one described for Alix. Thus, regulation of HD-PTP binding to ESCRTs must involve alternative mechanisms. We have previously reported that STAM2 competes with CHMP4 for binding to the Bro1 domain of HD-PTP, but does not bind to Alix.7

STAM2 also binds to the HD-PTP PRR at a site that overlaps with the TSG1017 (Fig.

6.1). In addition, UBAP1 binds to the CC domain of HD-PTP in a selective manner.6,29

Our structural analyses of HD-PTP in complex with different ESCRT subunits and endosomal effectors have shown that both amino acid sequence and molecular

145 determinants are key to define the specificity and selectivity observed for these interactions.25,29 Altogether these data highlight that coupling between the Bro1, CC and proximal PRR regions of HD-PTP may control the exchange of ESCRT partners, which is essential for HD-PTP function. The importance of the Bro1-CC region of HD-PTP is underscored by findings that it represents the minimal functional region of HD-PTP.1

Knowledge of the molecular architecture of HD-PTP is critical for understanding the mechanism of regulation of its interactions with ESCRTs. Whilst high-resolution structures of the Bro1 domain25,38,39 and the CC domain29 have been reported, the structural coupling of these two domains remains unexplored. Here, we present a structural and biophysical analysis of the entire ESCRT-binding region of HD-PTP

(encompassing the Bro1 and CC domains and the proximal region of PRR) by small angle

X-ray scattering (SAXS), analytical ultracentrifugation (AUC), size-exclusion chromatography (SEC), and multi-angle light-scattering (MALS). The structures of

HDPTP fragments encompassing the CC domain (HD-PTPCC), the Bro1 and CC domains

(HDPTPBro1-CC), and the Bro1 and CC domains plus the proximal region of PRR (HD-

PTPBro1-CC-PRR) all show an open and extended conformation where the Bro1 and CC domains are spread out horizontally, thus providing a large scaffolding architecture.

Hydrodynamic parameters obtained from AUC and SEC-MALS confirmed the extended conformation and the monomeric nature of all three proteins. The open architecture of

HD-PTP offers an ideal platform for binding of multiple ESCRTs to the Bro1, CC and

PRR domains, and suggests a mechanism of regulation by which displacement of ESCRT subunits controls access to ESCRT-III.

146 Results

The extended architecture of HD-PTPCC, HD-PTPBro1-CC and HD-PTPBro1-CC-PRR rules out a self-inhibited conformation.

Crystal structures of AlixBro1-V have shown a compact conformation in which the Bro1 and V domains are packed against each other. In addition, the two arms of the V domain

(V1 and V2) form a bent, rather than extended molecular shape.32 Further studies indicated that the PRR region binds to the Bro1 domain, locking Alix in a functionally inactive conformation.33,34 To assess if HD-PTP undergoes a similar regulation, we investigated the molecular shape and dimensions of HD-PTP in solution by collecting small-angle X-ray scattering (SAXS) measurements on HD-PTPCC, HD-PTPBro1-CC, and

HDPTPBro1-CC-PRR (Fig. 6.2).

Estimates of the radius of gyration (Rg) for each protein were obtained from the Guinier

40 region using PRIMUS, and the maximum dimension (Dmax) was obtained from indirect

Fourier transformation of the SAXS profiles using GNOM41 (Table 6.1, Suppl. Fig.

S6.1). For HD-PTPCC the Dmax of 15.3 nm is similar to that measured in its crystal

29 structure. For HD-PTPBro1-CC and HD-PTPBro1-CC-PRR the Dmax obtained were 19.3 nm and 20.4 nm respectively (Table 1). The probable atom-pair distribution functions p(r) for the three proteins (Suppl. Fig. S6.1B) present asymmetrical curves consistent with elongated molecules. Particle shapes were restored ab initio from the SAXS profiles using

DAMMIN and GASBOR,42,43 and they were all consistent with an open and extended molecular architecture (Fig. 6.2). The particle shape of HD-PTPBro1-CC-PRR shows a visible extension at the C-terminal end of the CC domain, corresponding to the additional PRR residues (Fig. 6.2C).

147

Figure 6.2. Analysis of HD-PTP by Small Angle X-ray Scattering. (A-C) Plots of X-ray scattering intensity log(I(q)) as a function of the scattering vector q (Å-1) for HD-PTPCC (A), HD-PTPBro1-CC (B) and HD-PTPBro1-CC-PRR (C). SAXS data are shown as grey open circles and the normalised fit from the best individual molecular model is shown as a red line (calculated with FoXS,44 goodness of fit indicated with its Χ2 value). Ab initio reconstructions calculated with DAMAVER45 are shown as transparent envelopes with the best individual models superimposed manually in Chimera.46 (D) Variation of the goodness of fit (Χ) with the HD-PTPCC SAXS data as a function of the distance between the centre of mass of subdomains CC1 and CC2 (dCC1-CC2). Fits from crystal structure coordinates of HD-PTPCC (blue) and AlixV (grey, closed forms; green, open forms) are individually shown (selected PDB ID codes indicated). Fits from HD-PTPCC models with varying CC1-CC2 distances are shown as red triangles. Filled triangles correspond to models with calculated sedimentation coefficients (SOMO47) consistent with the experimentally observed value (Table 1). Best Χ2 values and agreement with

148 hydrodynamic data are observed for the extended forms. (E) The angle between the CC1 and CC2 subdomains has a small degree of variability. The individual model with the best χ (red ribbon) is superimposed at CC2 with the coordinates from the HD-PTPCC crystal structures (blue, 5LM2 chain B; cyan, 5LM1; purple, 5LM2 chain A). (F) Analysis 48 by EOM of HDPTPBro1-CC interdomain flexibility. Black line: distribution of maximum distances for the pool of HD-PTPBro1-CC conformers selected on the basis of their hydrodynamic parameters (see Text). Red line: distribution of maximum distances of the conformers selected for the optimal ensembles. Representative members of the two main groups (blue and green ribbons) are shown superimposed manually (Chimera) to the ab initio reconstruction of HD-PTPBro1-CC. (G) The EOM analysis suggests a degree of flexibility between the Bro1 and CC domains of HD-PTPBro1-CC. Representative conformers from the two groups selected by EOM (blue, green) are shown superimposed at the CC domain.

Molecular models for each construct of HD-PTP were produced to help in the interpretation of the SAXS data (see Methods). The HD-PTPCC crystal structure coordinates29 show relatively poor fits to the SAXS profile (Fig. 6.2D). Coordinates of

AlixV domains obtained from Alix crystal structures (open and closed forms) give even poorer fits (Fig. 6.2d). We therefore generated a library of HD-PTPCC conformers where the angle between the “blade” (CC1) and “shaft” (CC2) subdomains29 was changed by variation of the distance d(CC1-CC2) between their centre of masses (see Methods for details). In general, closed conformations fit the SAXS data much worse than open conformations (Fig. 6.2D). Extended models with d(CC1-CC2) of 85-90 Å give the best fit to the SAXS profile (Fig. 6.2D). The best individual model (Χ2 = 0.97, Fig. 6.2A) differs slightly from the crystal structure coordinates (with d(CC1-CC2) = 82-88 Å) in the relative orientation between CC1 and CC2 (Fig. 6.2E) and in the conformation of the flexible loop connecting helices H1 and H2.29

A library of HD-PTPBro1-CC conformers was generated using torsion angle molecular dynamics (TAMD) to model the flexibility between the Bro1 and CC domains (see

Methods). Hydrodynamic parameters for each HD-PTPBro1-CC conformer were calculated using SOMO,47 and the best models were selected on the basis of their hydrodynamic

149 parameters (calculated versus experimentally determined) and their fit to the experimental SAXS profile of HD-PTPBro1-CC. Model selection and SAXS analysis proceeded as described in our previous work.49 The best individual model (Χ2 = 0.94) shows an extended linear arrangement of the Bro1 and CC domains (Fig. 6.2B). Analysis of HD-PTPBro1-CC conformational variability with the ensemble optimization method

(EOM)48 suggested some flexibility in the relative orientation between the Bro1 and CC domains but no evidence of extensive shape changes (Fig. 6.2F,G). The Porod-Debye plot for HDPTPBro1-CC (Suppl. Fig. S6.1) shows a clear plateau at high values of the scattering vector, indicative of limited flexibility.50,51

The pool of HD-PTPBro1-CC models fit reasonably well the experimental SAXS profile of

2 the longer HD-PTPBro1-CC-PRR, and the best individual models (Χ = 1.1) were the same as for the HD-PTPBro1-CC case. This indicates that HD-PTPBro1-CC and HD-PTPBro1-CC-PRR have very similar conformations in solution and that addition of the proximal PRR region

(up to residue 738, Fig. 6.1) does not affect the global architecture of the Bro1-CC domains or their relative orientation. Models including the additional PRR residues (see

Methods) showed a slightly better agreement with the experimental data (Χ2 = 1.0, Fig.

6.2C).

Hydrodynamic analyses confirm HD-PTP is a monomer with an extended conformation

Hydrodynamic parameters were determined using SEC-MALS and AUC (Fig. 6.3, Table

6.1). All three proteins behaved as monomers according to their elution profile (Fig.

6.3A), with average molecular weights of 38 kDa for HD-PTPCC, 86 kDa for HD-PTPBro1-

36,37 CC and 88 kDa for HD-PTPBro1-CC-PRR. Hence, in contrast to Alix, HD-PTP appears to lack an intrinsic ability to dimerise in solution. Furthermore, sedimentation velocity experiments showed single species for each protein with sedimentation coefficients of

150 2.47S for HD-PTPCC, 3.86S for HD-PTPBro1-CC and 3.63S for HD-PTPBro1-CC-PRR (Fig.

6.3B, Table 6.1).

Figure 6.3. Hydrodynamic analyses of HD-PTP. (A) SEC-MALS profile of HDPTPCC, HD-PTPBro1-CC and HD-PTPBro1-CC-PRR. The chromatogram shows differential refractive index (normalised) and molecular weight versus elution volume. The elution profile shows single peaks indicating monodisperse samples. The average molecular weights of 38 kDa for HD-PTPCC, 86 kDa for HD-PTPBro1-CC and 88 for HD-PTPBro1-CC-PRR indicate monomeric particles for each protein. Injected protein sample concentrations for MALS were 1-3 mg/mL. (B) AUC derived sedimentation coefficient distributions for HD- PTPCC, HD-PTPBro1-CC and HD-PTPBro1-CC-PRR show single peaks for each construct, labelled with the corresponding buffer-corrected s20,w values. Other hydrodynamic parameters are shown in Table 6.1. Protein sample concentrations for AUC were 0.2 mg/ml. 151

The hydrodynamic radius and frictional ratio of HD-PTPCC (Rh = 4.07 nm, f/f0 = 1.76) confirmed an anisotropic and elongated shape consistent with that observed in its crystal

29 structure. Hydrodynamically, HD-PTPCC is clearly different from the V-shaped

52 structure of AlixV (Rh = 3.22 nm, f/f0 = 1.44). Likewise, both HD-PTPBro1-CC and HD-

PTPBro1-CC-PRR adopted elongated conformations (Rh = 5.04 nm, f/f0 = 1.74 and Rh = 5.52

32 nm, f/f0 = 1.88, respectively) that contrast with the compact structure of AlixBro1-V (Rh =

4.32 nm and f/f0 = 1.52) (Fig. 6.4, Table 6.1). Together these data demonstrate that, unlike

Alix, HD-PTP exists in solution as an open and extended monomer, with no evidence for a closed, self-inhibited conformation, even in the presence of a functionally important region of the PRR.

The HD-PTP open conformation is functionally active

We have previously reported that HD-PTPBro1-CC is the minimal functional region, able to rescue endosomal trafficking defects of ubiquitinated cell-surface receptors,1 and also sufficient for UBAP1 binding in a cellular setting.29 Here we show that in contrast to

AlixBro1-V-PRR, HD-PTPBro1-CC-PRR is also fully functional for binding to ESCRT partners.

We used biosensor binding experiments (surface plasmon resonance, SPR) to demonstrate that both HD-PTPBro1 and HD-PTPBro1-CC-PRR are able to bind full length

CHMP4B (FL-CHMP4B) with similar affinity (Fig. 6.5A). HD-PTPBro1-CC-PRR also binds peptides from the C-terminus of CHMP4B (residues 205-224) and the central region of

UBAP1 (residues 261-280) (Fig. 6.5B) with similar affinity to that observed for the individual Bro1 or CC domains respectively.29 These results indicate that each binding site is fully accessible in HD-PTPBro1-CC-PRR and that the architecture observed in solution is compatible with ESCRT binding and remains functionally active in the presence of the

PRR proximal region.

152

Figure 6.4. The molecular shapes of HD-PTP and Alix differ. (A) Superimposition of the best individual model of HD-PTPBro1-CC (red) with the crystal structure of AlixBro1- V (purple) (PDB ID 2OEV) showing the contrast between the open and extended structure of HD-PTP versus the more compact conformation of Alix. (B) Ribbon diagram of HD-PTPBro1-CC with superimposed models of the ESCRT subunits STAM2 (cyan), CHMP4B (green) and UBAP1 (green) bound to their respective binding sites.25,29,39 The arrow shows the approximate distance between binding sites.

153

Figure 6.5. Analysis of HD-PTP binding to ESCRT subunits. (A) Equilibrium biosensor binding analysis for immobilised full length CHMP4B to HD-PTPBro1 (Kd 4.88 ± 0.3 μM) and HD-PTPBro1-CC-PRR (Kd 1.1± 0.1 μM), Biosensor sensograms shown in the insets. (B) Equilibrium biosensor binding analysis for immobilised HD-PTPBro1-CC-PRR to CHMP4B peptide (Kd 30.9 ± 0.2 μM) and to UBAP1 peptide (Kd 57.5 ± 3.1 μM). Biosensor sensograms shown in the insets.

154 Discussion

Our findings provide significant new insights into how HD-PTP controls the ESCRT pathway at the endosome, behaving differently from the canonical Bro1 domain containing protein Alix. First, they show that HD-PTPBro1-CC offers a far more extended

ESCRT binding platform than AlixBro1-V (Fig. 6.4). Indeed, the open architecture of HD-

PTP is compatible with binding to multiple ESCRT partners, allowing it to act as a hub for the traffic of endocytic cargo along the ESCRT pathway. Second, our findings show that the mechanisms of self-inhibition elucidated for Alix do not apply to HD-PTP.

Together these results invoke a new mechanism of ESCRT regulation, by which HD-PTP could act as central orchestrator to determine whether ubiquitinated cargoes are degraded, recycled, or held at the endosome, in keeping with its function as a tumour suppressor.

Our data show that HD-PTPBro1-CC adopts an open and extended architecture that is also maintained in HD-PTPBro1-CC-PRR, a construct including the proximal region of PRR. The

CC domain maintains the extended conformation observed in its crystallographic structure,29 and the Bro1 domain provides a largely linear extension to the molecule, with limited flexibility between the two domains (Fig. 6.2). This extended conformation is functionally competent, as both HD-PTPBro1 and HD-PTPBro1-CC-PRR are able to bind full length CHMP4B and HD-PTPBro1-CC-PRR also binds to the central region of UBAP1, as demonstrated by biosensor binding experiments (Fig. 6.5 and in our previous work29).

We have previously demonstrated the ability of HD-PTPBro1-CC to bind UBAP1 in cells and to rescue defects in EGFR sorting.1,29

Such an extended architecture is consistent with the proposed role for HD-PTP as a scaffold platform for ESCRT coordination.1,6,7 In particular, the long distance between the CHMP4B binding site in the Bro1 domain and the UBAP1 site in the CC domain

(~150 Å, Fig. 6.4) indicates that HD-PTP would be able to interact simultaneously with

155 ESCRT-I and ESCRT-III. The ability to link ESCRT-I with ESCRT-III may be especially relevant to the reported essential requirement of HD-PTP, but not Alix, in promoting cargo trafficking in the absence of ESCRT-II.3,18

Alix is subject to auto-inhibition. It adopts a closed conformation that prevents effector binding to the Bro1 and V domains and undergoes large-scale structural rearrangements and dimerisation upon activation.32,33,37 In contrast, we find no evidence for dimerisation, large-scale rearrangements, or the presence of closed self-inhibited forms in solution of

HD-PTP. Furthermore, the presence of the proximal region of PRR does not alter the open architecture of HD-PTP and, crucially, does not preclude binding of HD-PTP to critical ESCRT subunits including CHMP4B and UBAP1.

Since HD-PTPBro1-CC lacks the conformational variability of Alix, alternative mechanisms of regulating access to ESCRT-III must come into play. An attractive hypothesis is that interactions of both ESCRT-I and ESCRT–III with HD-PTP are controlled by ESCRT-0, consistent with the role of ESCRT-0 in deciding whether ubiquitinated cargo is recycled or sorted into the MVB pathway.53 In support of this idea is the evidence that the ESCRT-

0 subunit STAM2 binds to the Bro1 domain of HD-PTP at the conserved CHMP4B binding region,7,39 whilst also binding to the proximal region of PRR at a site which overlaps with the binding site for TSG101.7

156

Figure 6.6. HD-PTP is a scaffold for ESCRT binding. Our model of HD-PTP regulation of ESCRT function involves coordinated binding and displacement of different ESCRTs during endosomal trafficking of ubiquitinated cargo. First, upon internalisation of activated EGFR, ubiquitinated cargo traffics to ESCRT-0. In Stage 1, the ESCRT-0 subunit STAM2 associates with HD-PTP by binding both at the Bro1 domain with its GAT domain (pink oval), and at the PRR (KPPPR) with its SH3 domain (pink ¾-circle). In Stage 2, STAM2/ESCRT-0 is then released from HD-PTPPRR by the ESCRT-I core subunit TSG101, allowing access of UBAP1 to the CC region and stable binding of ESCRT-I to HD-PTP. UBAP1 binds to the conserved region FYX2L (yellow) in the CC domain and TSG101 binds to the “PTAP” motif (green) in the PRR. In Stage 3, binding of ESCRT-III (C4: CHMP4 subunit, purple) to the Bro1 domain of HD-PTP further displaces ESCRT-0, driving the cargo into the MVB pathway. Additional factors, including the presence of deubiquitinating enzymes and ESCRT-III polymerisation, would also drive MVB sorting.

Therefore, binding of STAM2 to HD-PTP would prevent the access of ESCRT-I subunits,

UBAP1 (to the CC) and TSG101 (to the PRR), whilst also blocking CHMP4 from binding to the Bro1 domain (Fig. 6.6; Stage 1). Release of ESCRT-0 in favour of ESCRT-I and

ESCRT-III would then be necessary to drive MVB sorting, and could occur by two, possibly simultaneous, competition events. First, the core ESCRT-I subunit TSG101 would bind the PTAP motif in HD-PTP and hence displace the STAM2 SH3 domain from binding the overlapping KPPPRP sequence, with binding of UBAP1 to the HD-PTP CC domain further supporting the engagement of ESCRT-I (Fig. 6.6, Stage 2). The deubiquitinating enzyme UBPY may assist this competition reaction by binding the

STAM2 SH3 domain.7 Second, binding of CHMP4 would release the GAT domain of

157 STAM2 from the HD-PTP Bro1 domain (Fig. 6.6, stage 3). Hence, ESCRT-I and ESCRT-

III recruitment to HD-PTP may represent a potentially irreversible decision for sorting ubiquitinated cargo into the MVB pathway for degradation. Progression through this pathway would also depend on environmental factors, including the balance between ubiquitin ligases and deubiquitinating enzymes54,55 and the polymerisation of CHMP4

(ESCRT-III).

Materials and Methods

Cloning, protein expression and purification

Four HD-PTP constructs were used in this work: HD-PTPBro1-CC (1-714) and HD-PTPBro1-

CC-PRR (1-738), cloned into a pET28a vector with restriction sites Nde1 and Xho1;

HDPTPBro1 (1-361), cloned into pNIC-28a-Bsa4 (Gift from Opher Gileadi (Addgene

#26103); and HD-PTPCC (362-704), cloned into pGEX-4T1 vector using EcoR I and Xho

I restriction sites. The GST-tagged FL-CHMP4B construct has been described previously.1 All constructs were confirmed by DNA sequencing.

All HD-PTP constructs were transformed in BL-21(DE3) E. coli cells and protein expression was induced with 0.1 mM IPTG overnight at 20°C. The His6-tagged proteins

(HD-PTPBro1, HD-PTPBro1-CC and HD-PTPBro1-CC-PRR) were purified by metal-affinity column chromatography using Nickel-beads (Qiagen) in 20 mM HEPES pH 7.4, 500 mM

NaCl, 10 mM imidazole and eluted with 250 mM imidazole, followed by anion-exchange chromatography using a MonoQ 5/50 GL column (GE Healthcare) equilibrated in 20 mM

HEPES pH 7.4, 2 mM EDTA, 2 mM DTT and eluted with a gradient of NaCl, and finally by SEC using a Superdex200 column (GE Healthcare) equilibrated with 20 mM HEPES pH 7.4, 0.3 M NaCl, 2 mM EDTA and 2 mM DTT.

158 GST-tagged HD-PTPCC protein was purified on a 5 ml GSTrap FF column (GE

Healthcare) in 20 mM HEPES pH 7.4, 250 mM NaCl, 2 mM DTT and eluted with 20 mM HEPES pH 7.4, 250 mM NaCl, 20 mM reduced glutathione. Fractions containing

GST-tagged HDPTPCC were pooled, and concentrated and buffer-exchanged in 20 mM

HEPES pH 7.4, 250 mM NaCl buffer to remove glutathione. The GST-tag was cleaved overnight by thrombin digestion at 4°C. This digested mixture was passed over a GSTrap

FF column, and digested HD-PTPCC was collected in the flowthrough. Digested HD-

PTPCC was further purified by anion-exchange chromatography using a MonoQ 5/50 GL column and SEC using a Superdex200 column.

FL-CHMP4B was expressed in C41 E. coli cells by IPTG induction as above, and purified by lysing bacterial cells in 20 mM HEPES pH 7.4, 250 mM NaCl, 2 mM DTT, followed by sonication and centrifugation at 12,400 ×g for 1 h. Cleared supernatant was loaded on a GSTrap FF (5 ml) column equilibrated with 20 mM HEPES pH 7.4, 250 mM NaCl, 2 mM DTT, and eluted with 20 mM reduced glutathione in the same buffer.

Small-angle X-Ray Scattering Analysis

The following concentrations and buffers were used for each sample: HD-PTPCC, 2.8 mg/ml in 20 mM HEPES pH 7.4, 300 mM NaCl, 2 mM EDTA and 2 mM DTT; HD-

PTPBro1-CC, 12 mg/ml in 50 mM Tris-Cl pH 8.0, 100 mM NaCl, 2 mM EDTA and 2 mM

DTT; HDPTPBro1-CC-PRR, 7.5 mg/ml in 50 mM Tris-Cl pH 8.0, 250 mM NaCl, 2 mM

EDTA and 2 mM DTT. SAXS data were collected at beamlines X33 Hamburg DESY

(HD-PTPBro1-CC) and BM29 ESRF (HD-PTPCC and HD-PTPBro1-CC-PRR). Data processing was performed with the ATSAS suite.56 The forward scattering, I(0) and the radius of

40 gyration Rg were estimated with PRIMUS using the Guinier approximation (Suppl. Fig.

S6.1). GNOM41 was used to compute the pairwise intra-particle distance distribution function p(r) and the maximum distance Dmax (Suppl. Fig. S6.1). Particle shapes were

159 restored ab-initio using DAMMIN and GASBOR.43 Twenty simulations were performed and the outputs were averaged and filtered using DAMAVER45 to produce the final envelopes (Fig. 6.2) with a normal spatial discrepancy value of 0.61-0.67 for the

DAMMIN models and 1.4-1.8 for the GASBOR models.

Molecular models and conformational sampling

Libraries of molecular models were generated for each construct as follows. The initial

29 model for HD-PTPCC was built from the coordinates of its crystal structure (PDB ID

5LM2, chain B). Disordered residues from the loop connecting the H1 and H2 helices

57 were rebuilt with standard geometry in CNS. HD-PTPCC conformational variability was explored by varying the distance between the centers of masses of two separate subdomains, CC1 encompassing the “blade” and CC2 encompassing the “shaft”

(according to the nomenclature used in our previous publication29). A library of conformers was thus generated using torsion angle molecular dynamics (TAMD) as implemented in CNS.57,58 For each conformer, the fit to the SAXS data was calculated using FoXS,44 and its hydrodynamic parameters calculated with SOMO47 (Fig. 6.2d).

The initial model for HD-PTPBro1-CC was built from the coordinates of the crystal structure

38 of HD-PTPBro1 (PDB ID 3RAU) and those of the individual HD-PTPCC model with the best fit to the SAXS data (Fig. 6.2A). Residues connecting the Bro1 and CC domains were rebuilt with standard geometry in CNS. A library of conformers was generated using

TAMD to model the interdomain flexibility of HD-PTPBro1-CC. Different lengths for the flexible linker between the Bro1 and CC domains were tried to ensure effective conformational sampling between the two domains. Best results were obtained with a linker spanning residues 360 to 370. Hydrodynamic and dimensional parameters for all the HD-PTPBro1-CC models were calculated with SOMO and used to select suitable conformers compatible with the experimentally determined parameters. The selected pool

160 of models was contrasted to the experimental SAXS profile using FoXS,44 as previously described.49 The conformational variability between the Bro1 and CC domains was further explored using the ensemble optimization method (EOM)48 on the pool of conformers generated by TAMD and selected on the basis of their hydrodynamic parameters.

Models for HD-PTPBro1-CC-PRR were generated by adding the proximal region of PRR (up to residue 738), to the coordinates of the best HD-PTPBro1-CC individual model (Fig. 6.2B).

The additional residues were modelled with standard geometry in CNS and their flexible conformation was sampled with TAMD.

Multi-angle light scattering (MALS) and analytical ultracentrifugation (AUC)

For MALS analyses, samples (1-3 mg/mL) were injected onto a Superdex-200 10/300

GL column (GE Healthcare) equilibrated with 20 mM Tris-Cl pH 8.0, 100 mM NaCl, 1 mM TCEP (HDPTPCC and HD-PTPBro1-CC) or 20 mM HEPES pH 7.4, 250 mM NaCl, 2 mM EDTA) (HD-PTPBro1-CC-PRR) and the eluted proteins passed through a Wyatt Helios

18-angle laser photometer with Wyatt EOS QELS detector. Concentrations were measured using a Wyatt rEX differential refractive index detector. Light scattering intensities were measured at different angles relative to the incident beam and data analysis was performed with ASTRA 6 software (Wyatt Technology Corp., CA, USA).

Protein fractions from MALS (adjusted to 0.2 mg/mL) were then used in sedimentation velocity experiments using either Optima XL-I (HD-PTPCC) or Optima XL-A (HD-

PTPBro1-CC and HD-PTPBro1-CC-PRR) ultracentrifuges (Beckman Instruments) at 50,000 rpm

(18,200 ×g) at 20°C and scanning every 60 or 90 seconds respectively, using a wavelength of 280 nm for a total of 200 scans. The sedimentation boundaries were analysed using the program Sedfit v8.7.59 Solvent corrected sedimentation coefficients

161 (S20,w), hydrodynamic radii (Rh) and frictional ratios (f/fo) were calculated with

Sednterp.60

Biosensor Binding studies

Biosensor protein binding studies were performed using the multiplex system ProteOn

XPR36 surface plasmon resonance instrument (Bio-Rad Laboratories) in 10 mM HEPES pH 7.4, 150 mM NaCl, 0.05% Tween-20 as running buffer. His6-tagged proteins were immobilised on a HTE chip (Bio-Rad Laboratories) at a concentration of 50-100 μg/ml.

This gave an immobilization level of proteins typically of 5000-8000 response units (RU).

A GLC sensor chip (Bio-Rad Laboratories) was derivatised with anti-GST-antibody (GE

Healthcare) using amine coupling, and then used to immobilise FL-CHMP4B at a surface density of 3000-4000 response units (RU).

All experiments were performed at 25°C. Purified protein HD-PTPBro1, HD-PTPBro1-CC-

PRR and synthetic peptides CHMP4B (205KKKEEEDDDMKELENWAGSM224) and

UBAP1 (261SNIKSLSFPKLDSDDSNQKT280) (Generon Ltd, UK) were used as analytes in equilibrium binding measurements. Analyte stocks were prepared just prior to the binding experiments and injected (50-100 μl at 100 μl/min) in the horizontal orientation, using serially diluted analyte concentrations chosen to give a suitable spread of responses below and above half-maximal binding. All the binding sensograms were collected, processed and analysed using the integrated ProteOn Manager software (Bio-

Rad Laboratories), using the equilibrium binding model: Response=[A] * Rmax / ([A] +

KD) where [A] is the analyte concentration and Rmax is the maximum response.

Data Availability

The datasets generated during this study are available from the corresponding author upon reasonable request.

162 Acknowledgements

This work was support by Medical Research Council (MR/K011049/1 to L.T. and P.W).

G.H. is funded by a Biotechnology and Biological Sciences Research Council Doctoral

Training Programme Studentship. We thank the staff of the beamline BM29 at ESRF

(MX1483/1580) and X33 at DESY, Hamburg for assistance with SAXS data collection.

163 Table 6.1. Hydrodynamic and dimensional experimental data for HD-PTP. Experimental parameters were determined from MALS, AUC and SAXS 29 52 32 measurements. Hydrodynamic parameters for the crystal structures of HD-PTPCC, AlixV and AlixBro1-V and the best SAXS models of HD-PTPCC, 47 HD-PTPBro1-CC and HD-PTPBro1-CC-PRR were calculated with SOMO. Rh hydrodynamic radius; s20,w sedimentation coefficient in water at 20°C; f/f0 frictional coefficient; Rg radius of gyration; Dmax, maximal linear dimension of the particles.

Hydrodynamic Parameters MALS AUC SAXS

Experimental Rh (nm) Mr (Da) Rh (nm) f/f0 s20,W Rg (nm) Dmax (nm)

HD-PTPCC 3.93 ± 0.12 38,200 4.07 1.76 2.47 ± 0.14 4.53 15.3

HD-PTPBro1-CC 4.97 ± 0.12 85,900 5.04 1.74 3.86 ± 0.11 5.55 19.3

HD-PTPBro1-CC-PRR 5.47 ± 0.05 87,900 5.52 1.88 3.63 ± 0.21 5.83 20.4

Calculated (SOMO) Coordinates Mr (Da) Rh (nm) f/f0 s20,W Rg (nm) Dmax (nm)

HD-PTPCC 5LM2 39,158 3.62 1.61 2.60 4.47 16.3

HD-PTPCC model 39,760 3.69 1.64 2.59 4.43 16.1

HD-PTPBro1-CC model 80,306 4.74 1.66 3.97 5.82 22.0

HD-PTPBro1-CC-PRR model 83,695 5.20 1.79 3.78 6.32 25.0

AlixV 4JJY 38,611 3.61 1.61 2.52 3.88 13.2

AlixV 2OJQ 38,697 3.22 1.44 2.83 2.91 11.2

AlixBro1-V 2OEV 77,979 4.32 1.52 4.16 4.49 16.5

References for Chapter 6

1 A. Doyotte, A. Mironov, E. McKenzie and P. Woodman, Proc. Natl. Acad. Sci., 2008, 105, 6308–6313. 2 H. Ma, P. Wardega, D. Mazaud, A. Klosowska-Wardega, A. Jurek, U. Engström, J. Lennartsson and C.-H. Heldin, Cell. Signal., 2015, 27, 2209–2219. 3 M. D. J. Parkinson, S. C. Piper, N. A. Bright, J. L. Evans, J. M. Boname, K. Bowers, P. J. Lehner and J. P. Luzio, Biochem. J., 2015, 471, 79–88. 4 D. Kharitidi, P. M. Apaja, S. Manteghi, K. Suzuki, E. Malitskaya, A. Roldan, M.-C. Gingras, J. Takagi, G. L. Lukacs and A. Pause, Cell Rep., 2015, 13, 599–609. 5 P. I. Hanson and A. Cashikar, Annu. Rev. Cell Dev. Biol., 2012, 28, 337–362. 6 F. Stefani, L. Zhang, S. Taylor, J. Donovan, S. Rollinson, A. Doyotte, K. Brownhill, J. Bennion, S. Pickering-Brown and P. Woodman, Curr. Biol., 2011, 21, 1245–1250. 7 N. Ali, L. Zhang, S. Taylor, A. Mironov, S. Urbé and P. Woodman, Curr. Biol., 2013, 23, 453–461. 8 J. Schöneberg, I.-H. Lee, J. H. Iwasa and J. H. Hurley, Nat. Rev. Mol. Cell Biol., 2017, 18, 5–17. 9 L. Christ, C. Raiborg, E. M. Wenzel, C. Campsteijn and H. Stenmark, Trends Biochem. Sci., 2017, 42, 42–56. 10 J. G. Carlton and J. Martin-Serrano, Science, 2007, 316, 1908–1912. 11 E. Morita, V. Sandrin, H.-Y. Chung, S. G. Morham, S. P. Gygi, C. K. Rodesch and W. I. Sundquist, EMBO J., 2007, 26, 4215–4227. 12 J. F. Nabhan, R. Hu, R. S. Oh, S. N. Cohen and Q. Lu, Proc. Natl. Acad. Sci. U. S. A., 2012, 109, 4146–4151. 13 R. Sahu, S. Kaushik, C. C. Clement, E. S. Cannizzo, B. Scharf, A. Follenzi, I. Potolicchio, E. Nieves, A. M. Cuervo and L. Santambrogio, Dev. Cell, 2011, 20, 131– 139. 14 A. J. Jimenez, P. Maiuri, J. Lafaurie-Janvore, S. Divoux, M. Piel and F. Perez, Science, 2014, 343, 1247136. 15 Y. Olmos, L. Hodgson, J. Mantell, P. Verkade and J. G. Carlton, Nature, 2015, 522, 236–239. 16 M. Vietri, K. O. Schink, C. Campsteijn, C. S. Wegner, S. W. Schultz, L. Christ, S. B. Thoresen, A. Brech, C. Raiborg and H. Stenmark, Nature, 2015, 522, 231–235. 17 N. Issman-Zecharya and O. Schuldiner, Dev. Cell, 2014, 31, 461–473. 18 N. Loncle, M. Agromayor, J. Martin-Serrano and D. W. Williams, Sci. Rep., , DOI:10.1038/srep08461. 19 J. E. Garrus, U. K. von Schwedler, O. W. Pornillos, S. G. Morham, K. H. Zavitz, H. E. Wang, D. A. Wettstein, K. M. Stray, M. Côté, R. L. Rich, D. G. Myszka and W. I. Sundquist, Cell, 2001, 107, 55–65. 20 Y. Olmos and J. Carlton, Curr. Opin. Cell Biol., 2016, 38, 1–11. 21 C. Bissig and J. Gruenberg, Trends Cell Biol., 2014, 24, 19–25. 22 U. K. von Schwedler, M. Stuchell, B. Müller, D. M. Ward, H.-Y. Chung, E. Morita, H. E. Wang, T. Davis, G.-P. He, D. M. Cimbora, A. Scott, H.-G. Kräusslich, J. Kaplan, S. G. Morham and W. I. Sundquist, Cell, 2003, 114, 701–713. 23 S. Sun, X. Zhou, W. Zhang, G. E. Gallick and J. Kuang, Biochem. J., 2015, 466, 475–487. 24 S. Toyooka, M. Ouchida, Y. Jitsumori, K. Tsukuda, A. Sakai, A. Nakamura, N. Shimizu and K. Shimizu, Biochem. Biophys. Res. Commun., 2000, 278, 671–678.

165 25 D. Gahloth, C. Levy, L. Walker, L. Wunderley, A. P. Mould, S. Taylor, P. Woodman and L. Tabernero, Struct. Lond. Engl. 1993, 2017, 25, 1011–1024.e4. 26 J. Kim, S. Sitaraman, A. Hierro, B. M. Beach, G. Odorizzi and J. H. Hurley, Dev. Cell, 2005, 8, 937–947. 27 F. Ichioka, E. Takaya, H. Suzuki, S. Kajigaya, V. L. Buchman, H. Shibata and M. Maki, Arch. Biochem. Biophys., 2007, 457, 142–149. 28 Y. Kimura, M. Tanigawa, J. Kawawaki, K. Takagi, T. Mizushima, T. Maeda and K. Tanaka, Eukaryot. Cell, 2015, 14, 976–982. 29 D. Gahloth, C. Levy, G. Heaven, F. Stefani, L. Wunderley, P. Mould, M. J. Cliff, J. Bella, A. J. Fielding, P. Woodman and L. Tabernero, Structure, 2016, 24, 2115–2126. 30 X. Zhou, S. Pan, L. Sun, J. Corvera, Y.-C. Lee, S.-H. Lin and J. Kuang, Biochem. J., 2009, 418, 277–284. 31 X. Zhou, J. Si, J. Corvera, G. E. Gallick and J. Kuang, Biochem. J., 2010, 432, 525– 538. 32 R. D. Fisher, H.-Y. Chung, Q. Zhai, H. Robinson, W. I. Sundquist and C. P. Hill, Cell, 2007, 128, 841–852. 33 Q. Zhai, M. B. Landesman, H.-Y. Chung, A. Dierkers, C. M. Jeffries, J. Trewhella, C. P. Hill and W. I. Sundquist, J. Virol., 2011, 85, 9222–9226. 34 S. Sun, L. Sun, X. Zhou, C. Wu, R. Wang, S.-H. Lin and J. Kuang, Dev. Cell, 2016, 36, 331–343. 35 J. G. Carlton, M. Agromayor and J. Martin-Serrano, Proc. Natl. Acad. Sci., 2008, 105, 10541–10546. 36 U. M. Munshi, J. Kim, K. Nagashima, J. H. Hurley and E. O. Freed, J. Biol. Chem., 2007, 282, 3847–3855. 37 R. Pires, B. Hartlieb, L. Signor, G. Schoehn, S. Lata, M. Roessle, C. Moriscot, S. Popov, A. Hinz, M. Jamin, V. Boyer, R. Sadoul, E. Forest, D. I. Svergun, H. G. Göttlinger and W. Weissenhorn, Structure, 2009, 17, 843–856. 38 P. Sette, R. Mu, V. Dussupt, J. Jiang, G. Snyder, P. Smith, T. S. Xiao and F. Bouamr, Structure, 2011, 19, 1485–1495. 39 J. Lee, K.-J. Oh, D. Lee, B. Y. Kim, J. S. Choi, B. Ku and S. J. Kim, PLoS ONE, 2016, 11, e0149113. 40 P. V. Konarev, V. V. Volkov, A. V. Sokolova, M. H. J. Koch and D. I. Svergun, J. Appl. Crystallogr., 2003, 36, 1277–1282. 41 D. I. Svergun, J. Appl. Crystallogr., 1992, 25, 495–503. 42 D. I. Svergun, Biophys. J., 1999, 76, 2879–2886. 43 D. I. Svergun, M. V. Petoukhov and M. H. Koch, Biophys. J., 2001, 80, 2946–2953. 44 D. Schneidman-Duhovny, M. Hammel and A. Sali, Nucleic Acids Res., 2010, 38, W540-544. 45 V. V. Volkov and D. I. Svergun, J. Appl. Crystallogr., 2003, 36, 860–864. 46 E. F. Pettersen, T. D. Goddard, C. C. Huang, G. S. Couch, D. M. Greenblatt, E. C. Meng and T. E. Ferrin, J. Comput. Chem., 2004, 25, 1605–1612. 47 N. Rai, M. Nöllmann, B. Spotorno, G. Tassara, O. Byron and M. Rocco, Struct. Lond. Engl. 1993, 2005, 13, 723–734. 48 P. Bernadó, E. Mylonas, M. V. Petoukhov, M. Blackledge and D. I. Svergun, J. Am. Chem. Soc., 2007, 129, 5656–5664. 49 H. Tariq, J. Bella, T. A. Jowitt, D. F. Holmes, M. Rouhi, Z. Nie, C. Baldock, D. Garrod and L. Tabernero, Proc. Natl. Acad. Sci. U. S. A., 2015, 112, 5395–5400. 50 M. Pelikan, G. L. Hura and M. Hammel, Gen. Physiol. Biophys., 2009, 28, 174–189.

166 51 R. P. Rambo and J. A. Tainer, Biopolymers, 2011, 95, 559–571. 52 S. Lee, A. Joshi, K. Nagashima, E. O. Freed and J. H. Hurley, Nat. Struct. Mol. Biol., 2007, 14, 194–199. 53 C. Raiborg, K. G. Bache, A. Mehlum and H. Stenmark, Biochem. Soc. Trans., 2001, 29, 472–475. 54 D. Popovic, D. Vucic and I. Dikic, Nat. Med., 2014, 20, 1242–1253. 55 R. C. Piper, I. Dikic and G. L. Lukacs, Cold Spring Harb. Perspect. Biol., , DOI:10.1101/cshperspect.a016808. 56 M. V. Petoukhov and D. I. Svergun, Biophys. J., 2005, 89, 1237–1250. 57 A. T. Brünger, P. D. Adams, G. M. Clore, W. L. DeLano, P. Gros, R. W. Grosse- Kunstleve, J. S. Jiang, J. Kuszewski, M. Nilges, N. S. Pannu, R. J. Read, L. M. Rice, T. Simonson and G. L. Warren, Acta Crystallogr. D Biol. Crystallogr., 1998, 54, 905–921. 58 A. T. Brunger, Nat. Protoc., 2007, 2, 2728–2733. 59 P. Schuck, Biophys. J., 2000, 78, 1606–1619. 60 T. M. Laue, B. D. Shah, T. M. Ridgeway and S. L. Pelletier, in Analytical ultracentrifugation in biochemistry and polymer science, Royal Society of Chemistry, 1992, pp. 90–125.

167 Supplementary Information

Supplemental Figure 6.1. Analysis of SAXS data for HD‐PTPCC (top row), HD-PTPBro1‐ CC (middle row) and HD-PTPBro1‐CC-PRR (bottom row). The Guinier fitting plots (A) show the logarithm of the X‐ray intensity ln [I(q)] as a function of the square of the modulus of 2 ‐2 the scattering vector q (in Å ) for each construct. The radii of gyration (Rg) measured 2 from the slopes of the linear regions at low values of q are: 4.53 nm for HD-PTPCC (top), 5.55 nm for HD-PTPBro1‐CC (middle) and 5.83 nm for HD-PTPBro1-CC-PRR (bottom). Experimental data are shown as blue empty circles and the lines of best fit are shown in red. Residual plots show random distributions of differences between experimental values and those calculated from the lines of best fit. P(r) distribution plots (center column) show the probability of distance between scatterers P(r) against the distance r (in nm) for each construct. The maximum inter-scatterer distances Dmax are, respectively, 15.3 nm for HD-PTPCC (top), 19.3 nm for HD-PTPBro1-CC (B), and 20.4 nm for HD- PTPBro1-CC-PRR (bottom). Each plot has been normalised so that the area under the curve equals 1. Porod-Debye plots for each construct (C) show q4 × I(q) against q4 where I(q) is the X-ray intensity and q is the modulus of the scattering vector in Å-1. Data reaching a plateau at high q4 values indicate compact particles whereas a pronounced positive slope indicates particles with large flexibility. The three plots are consistent with compact particles in solution with limited flexibility.51

168 Chapter 7: Paper 4

Structural characterization of the HD-PTP:STAM2 ESCRT-0 complex

Graham Heaven1, Michael Lockhart2, Clair Baldock2, Thomas Jowitt2, Hilda D. Ruiz Nivia2, Jordi Bella2, Alistair J. Fielding1, Philip Woodman2 and Lydia Tabernero2.

Target Journal: EMBO Journal

169 Abstract

His Domain-Protein Tyrosine Phosphatase (HD-PTP) is a regulator of the Endosomal

Sorting Complexes Required for Transport (ESCRT) pathway. Through association with

ESCRT-0, I and III, HD-PTP is responsible for the downregulation of mitogenic receptors, by sorting them into intralumenal vesicles within the multivesicular body

(MVB) and allowing them to access the lysosomal degradation pathway. STAM2, a component of ESCRT-0, is known to bind HD-PTP at both the Bro1 domain and proline rich region (PRR) but until now only the STAM2-Bro1 interaction has been structurally characterised. Here, we confirm that HD-PTP and STAM2 bind in a 1:1 complex and use

MALS, AUC and SAXS to build a solution structure of the full HD-PTP:STAM2 complex. Our results reveal how HD-PTP forms both interactions with STAM2 simultaneously and tethers the flexible STAM2 domains into a rigid complex. We also show HD-PTP and STAM2 can form a stable complex when either binding site at the

Bro1 or PRR is lost. This supports a mechanism where HD-PTP and STAM2 remain attached whilst each binding site can be displaced by additional binding partners.

Introduction

HD-PTP (His Domain-Protein Tyrosine Phosphatase) is a member of the Bro1 domain family of proteins which regulate ESCRTs (endosomal sorting complexes required for transport). HD-PTP has a specific ESCRT-dependent function in supporting the degradative endosomal trafficking of a range of transmembrane receptors including

EGFR,1 PDGFR-β,2 α5β1 integrin.3 It also promotes the recycling of a range of receptors including Src.4,5 HD-PTP is hemizygously deleted in various human cancers,6 and this has been shown to increase tumorous growth in mice.6 Furthermore, HD-PTP depletion has been widely shown to promote tumorous growth and migration in cell experiments.2–

4,7

170 HD-PTP differs from other members of the Bro1 domain family, including human ALIX and yeast Bro1, in that HD-PTP can bind a component of the ESCRT-0 complex,

STAM2.8 ESCRT-0 is regarded as the furthest upstream complex in the ESCRT pathway, responsible for the recognition of ubiquitinated cargo, and the membrane recruitment of downstream ESCRT-I.9–11 Hence, the binding between HD-PTP and STAM2 is likely the first HD-PTP-ESCRT contact that occurs during HDPTP-dependent trafficking.

Understanding the structural arrangement of this complex is crucially important towards fully explaining the mechanism of HD-PTP in this process.

Yeast two hybrid studies on the binding of HD-PTP with STAM2 (ESCRT-0) identified two binding interactions (Figure 7.1): HD-PTPBro1 binds the STAM2GAT (or Core domain), whilst HD-PTPPRR binds the STAM2SH3.8 Regulation of the HD-PTP-dependent

ESCRT pathway is achieved by competition at both interaction sites of this complex. The

Bro:GAT interaction is competed by the interaction of HD-PTPBro1 with CHMP4

(ESCRT-III)8 and the PRR:SH3 interaction is competed for by TSG101. Additionally, the PRR of UBPY also competes for the STAM2SH3 binding site.

171

Figure 7.1. Domain organisation of HD-PTP and STAM2-Hrs ESCRT-0 complex. Interacting domains are indicated: * HD-PTPBro1:STAM2GAT; † HD-PTPPRR:STAM2SH3; and § HrsGAT:STAM2GAT. (VHS = Vps-27, Hrs and STAM; FYVE = Fab 1, YOTB, Vac 1, EEA1; DUIM = double-sided ubiquitin-interacting motif, SH3 = Src homology 3; Bro1 = Bypass for C kinase 1 resistance to osmotic shock; CC = coiled-coil; PRR = proline rich region; PEST = rich in proline, glutamic acid, serine and threonine). Diagram was constructed using IBS web server (http://ibs.biocuckoo.org/).

The structure of the HD-PTPBro1 and STAM2GAT interaction has been solved by X-ray crystallography12 but the structure of the HD-PTPPRR STAM2SH3 interface has not been investigated. In this work the full HDPTPBro-CC-PRR:STAM2SH3-GAT complex, with both sets of molecular interactions, has been structurally characterised. Radius measurements from multi-angle laser light scattering, sedimentation velocity analytical ultracentrifugation and small angle x-ray scattering show that HDPTP and STAM2 are both elongated molecules in solution, and interact with each other to form an elongated complex with 1:1 stoichiometry.

Because there is a relatively short linker between the SH3 and GAT domains of STAM2 we wondered whether it would be able to reach both binding sites at once. Our SAXS data indicates that the STAM2 monomer is flexible, whilst the HD-PTP monomer is rigid.

The HDPTP:STAM complex is rigid with HD-PTP binding at both ends of the flexible

STAM2 causing it to stretch out into an extended conformation.

172 Results

HD-PTP and STAM2 coelute by size exclusion as a stable complex through either of their two binding sites

Two interaction sites between HD-PTP and STAM2 have previously been identified by yeast-two-hybrid and confirmed biochemically: HD-PTPBro1 with STAM2GAT, and

HD-PTPPRR with STAM2SH3.8 The Bro1 interaction with GAT peptides has further been characterised by X-ray crystallography and a moderate affinity binding constant of 6 μM determined by isothermal titration calorimetry.12

The PRR-SH3 interaction has not been characterised further, however information from interactions in homologous domains is known. HD-PTPPRR binds the SH3 domain from

Mona/Gads with an exceptionally strong KD of 0.17 μM.13 STAM2SH3 domain, on the other hand, binds the PMVNRENKP motif of UBPY with a relatively weak KD of 27

μM.14 This suggests that the moderate Bro:GAT binding may or may not dominate the

HD-PTP:STAM2 interaction, whereas the secondary PRR:SH3 interaction could range anywhere from ~0.1 – 30 μM.

To assess the contribution of the two HD-PTP:STAM2 binding interactions to the formation of a HD-PTP STAM2 complex, we performed size exclusion chromatography using mixtures of STAM2 (wild type) with various constructs of HD-PTP (Figure 7.2).

When wild type HD-PTP was mixed with STAM2 it eluted 1.3 mL earlier than the HD-

PTP monomer, and coelution was confirmed by SDS-PAGE (Figure 7.2A). HD-PTPL202D mutant with interrupted GAT binding still coeluted with STAM2, 0.7 mL earlier than the

HD-PTPL202D monomer (Figure 2B). HDPTP without the proline rich region, HD-

PTPΔPRR, also coeluted with STAM2, 0.8 mL earlier than HD-PTPΔPRR monomer (Figure

2C). However, when the proline rich region was missing and the Bro1 domain was mutated (HD-PTPΔPRR,L202D) there was no coelution with STAM2. Instead peaks were 173 observed corresponding to HD-PTPΔPRR,L202D monomer and STAM2 monomer (Figure

2D). These results suggest that a complex can form via interactions between Bro-GAT or

PRR-SH3, as interruption of either still produces a complex. Interrupting both interaction sites abolishes complex formation.

Judging from the magnitude of the peak shifts and the intensity of SDS-PAGE gel bands, it appears that the complex formed in the presence of both binding sites is more stable than the two mutants. However it is not completely clear from these data whether the wild type HD-PTP:STAM2 complex is composed of 1 molecule of HDPTP and 1 molecule of

STAM binding via the Bro-GAT or the SH3-PRR interactions; or whether binding occurs at both binding sites at once. Furthermore, other stoichiometries are possible, such as a

1:2 complex of HD-PTP with a STAM2 molecule bound at each end. To explore structural assembly of the complex, we used the eluted HD-PTP:STAM2 complex and measured MALS, AUC and SAXS.

174

Figure 7.2. Mutational analysis of HD-PTP:STAM2 binding. Superdex200 profiles and SDS-PAGE gels STAM2 mixed with HD-PTP variants: (A) wild type HD-PTP, wt (B) Bro binding site substitution, L202D; (C) PRR deletion, ΔPRR; (D) both ΔPRR and L202D . V0 = void volume, determined by an injection of blue dextran.

175 MALS and AUC support 1:1 stoichiometry of HD-PTP:STAM2 binding.

MALS and AUC were performed on HD-PTP, STAM2 and the HD-PTP:STAM2 complex. Both methods show that HD-PTP and STAM2 behave as monomers in solution, and further show that the complex has 1:1 complex stoichiometry (observed mass 100 kDa, expected mass 109 kDa). This measured mass is inconsistent with a higher order oligomer of HD-PTP and STAM2, for example a 1:2 complex (expected MW 131 kDa) or a 2:1 complex (expected MW 195 kDa).

The hydrodynamic parameters Rh, s20,w and f/f0 were used to aid modelling of STAM2 and the HD-PTP:STAM2 complex later in this paper (Table 7.1 and 7.2). The Rh of HD-

PTP:STAM2 complex measured by AUC was 0.68 nm larger than by MALS, however, our modelling later in this paper shows that the Rh measured by MALS agrees better with the SAXS fitted models.

Figure 7.3. MALS and AUC of HD-PTP (solid black lines), STAM2 (grey lines) and the HD-PTP:STAM2 complex (dashed black lines). Injected protein sample concentrations for MALS were 1-3 mg/mL and protein sample concentrations for AUC were 0.2 mg/ml.

176 SAXS data show that rigid HD-PTP binds flexible STAM2 to form a rigid complex.

To investigate the solution structure of HD-PTP, STAM2 and the HD-PTP:STAM2 complex, the three samples were all measured by size-exclusion chromatography-coupled small angle X-ray scattering (SAXS) (Figure 7.4A).

SAXS measures X-ray intensity I(q) as a function of scattering angle (q), and the total intensity is given by q x I(q). We first investigated flexibility in the samples by curve shape analysis after transformation of the data to different power orders, as described by

Rambo and Tainer.15 Following the Porod-Debye law, at low values of q, data are expected to curve towards a clear plateau in a q4 x I(q) vs q4 plot for rigid proteins, whereas a plateau in a q3 x I(q) vs q3 plot indicates partial flexibility and a plateau in a q2 x I(q) vs q2 plot is indicative of completely flexible proteins. When plotted in these three forms, the data for our samples clearly showed that HD-PTP and the HD-

PTP:STAM2 complex form a plateau in the q4 x I(q) vs q4 (Figure 7.4B), whereas

STAM2 instead showed a plateau in the q3 x I(q) vs q3 plot (Figure 7.4C). The q2 x I(q) vs q2 plot (Figure 7.4D) shows no plateau for any of the three samples, ruling out full disorder.

This analysis indicated first that STAM2 behaves as a partially flexible protein, but also that upon complexing HD-PTP this flexibility is lost. The secondary structure of STAM2 between the SH3 and GAT regions is predicted to have two alpha helices (Figure 7.4E), so these data indicate that either these helices do not form as predicted or that their flanking loop regions confer significant flexibility between the two ordered domains.

Reduction of flexibility upon binding HD-PTP implies that STAM2 is being tethered at both ends by the two binding interactions.

177

Figure 7.4. SAXS flexibility analysis. (A) Total intensity (q x I(q) vs q) plot of HD-PTP, STAM2 and HD-PTP:STAM2. (B) q4 x I(q) vs q4 plot (aka Porod-Debye plot). (C) q3 x I(q) vs q3 plot (aka Siblys plot). (D) q2 x I(q) vs q2 plot (aka Kratky-Debye plot). (E) Secondary structure prediction for STAM2 calculated using PSIPred.16

178 Solution structure of STAM2 SH3-GAT

We used the SAXS data to construct solution models of monomeric STAM2, first by transformation into real space atom pair distribution and ab initio reconstruction and afterwards by molecular modelling based on known 3D structures for SH3 and GAT domains.

The atom pair distance distribution was fitted to the STAM2 SAXS data (Figure 7.5A).

The distribution shape shows that STAM2 is fairly elongated in solution with a Dmax of

141 nm (Figure 7.5B). Using these data, DAMMIF17 was used to construct 21 ab initio models. Despite the flexibility indicated by the Porod-Debye analysis, there was little variation in the models produced by this approach (average NSD = 0.660). Dismissing only one outlier which was over two standard deviations above the average NSD, the remaining 20 models were averaged using DAMAVER,18 revealing an S-shaped ab initio structure (Figure 7.5C).

The structure of the STAM2 SH3 has been solved by NMR spectroscopy (PDB:1X2Q), and the GAT domain of human STAM1 (73% identity, 87% similarity to STAM2) has been solved by X-ray crystallography (PDB:3F1I).19 These two structures were used for structural modelling; the two domains were kept rigid while the linking region was varied by torsional angle molecular dynamics. Pools of models were produced with various linker lengths until a good fit to the SAXS data was achieved. The best model had a simulated SAXS curve (using FoXS server20,21) which fitted the recorded data with a Χ2 of 0.96. Agreement with the hydrodynamic properties from MALS and AUC was also satisfied by this best model, and it adopted a similar shape to the ab initio models (Figure

7.5D, Table 7.1).

179

Figure 7.5. Solution structure of STAM2. (A) Real space fitting and molecular model simulated spectrum versus measured SAXS data. (B) Atom pair distance distribution for STAM2. (C) Average ab initio model for STAM2. (D) Best fitting molecular model of STAM2 overlaid with the ab initio model.

Table 7.1. Hydrodynamic parameters for STAM2 model vs experimental data.

(SAXS) (MALS) (AUC) Experimental Rg Dmax Mr Rh Rh f/f0 s20,w (nm) (nm) (Da) (nm) (nm) (S)

STAM2 4.53 141 25,100 2.78 2.77 1.48 1.94

Model simulations χ2 Rg Dmax Mr Rh f/f0 s20,w (nm) (nm) (Da) (nm) (S) STAM2 molecular model 0.96 3.57 130 22,577 3.17 1.69 1.7

180 Solution structure of HD-PTP:STAM2 complex.

We went on to investigate the structure of HD-PTP in complex with STAM2 by SAXS.

The solution structure of three N-terminal domains of HD-PTP alone (Bro, CC and proximal PRR) has already been described previously by SAXS, MALS and AUC

(Gahloth 2017, Chapter 5, Paper 2 of this thesis). The Bro and CC domains are arranged in an elongated structure with a maximum dimension (Dmax) of 19.3 nm, and addition of the promixal PRR extends this length to 20.4 nm with no effect on the orientation of the

Bro and CC domains.

Comparison of the atom pair distributions of our data for HD-PTP:STAM2 with HD-PTP shows that the complex is slightly longer and thicker than HD-PTP alone (Figure 7.5A).

This also corresponds to an increase in Dmax and Rg upon binding STAM2 (from Dmax 218 to 228 nm and Rg 5.85 to 6.65 nm).

Ab initio models were constructed using the atom pair distributions for HD-PTP and HD-

PTP:STAM2. Overlaying the structures clearly shows how the complex is thicker along the full length of the HD-PTP, which indicates that STAM2 is interacting alongside HD-

PTP rather than projecting off one of the two binding sites.

To further rationalize the shape of the HD-PTP:STAM2 complex, we constructed molecular models using known structural information about the two binding sites. The

STAM2 GAT HD-PTP Bro crystal structure was used to generate distance constraints to model binding at the Bro1 site and HD-PRR-Mona SH3 crystal structure, by homology modelling, was used to generate distance constraints to model binding at the PRR.

181

Figure 7.6. HD-PTP:STAM2 complex SAXS. (A) Atom pair distance distribution of HD- PTP and HD-PTP:STAM2. (B) Ab initio models of HD-PTP and HD-PTP:STAM2 calculated using Damaver.18 (C) Molecular modelling of the HD-PTP:STAM2 with overlaid ab initio models: (i) STAM2 with restraints at Bro binding site, (ii) STAM2 with restraints at PRR binding site, (iii) STAM2 with restraints at both binding sites, (iv) HD- PTP with a STAM2 at both binding sites. (D) SAXS data with real space fitting curve (used to contruct P(r) distribution in panel A) and dual binding site simulation (from molecular model in panel Ciii). (E) Comparison of STAM2 monomeric SAXS structure and STAM2 structure in complex with HD-PTP.

182 Simply superimposing the STAM2 monomeric structure onto each binding site gave very poor fits to the data (Χ2 = 8.41 when bound to Bro1 and Χ2 = 4.28 when bound to the

PRR). Since STAM2 showed evidence for flexibility from the Porod-Debye analysis, but

HD-PTP did not, we explored further models with rigid a HD-PTP and flexible STAM2.

We performed restrained torsional angle molecular dynamics (TAMD) to create models of STAM2 with interaction restrains to HD-PTP at either the Bro1 or the PRR, using

FoXS server to simulate SAXS curves for comparison to the measured data. We found that regardless of whether the STAM2 was bound via the Bro1 or PRR, the simulations converged onto similar models where STAM2 extended across the length of HD-PTP (Χ2

= 1.56 Bro bound and Χ2 = 1.32 PRR bound). We then performed simulations with restraints for both binding sites, biasing the models to dual domain binding. These dual binders gave an improved fit the SAXS data with a best fit Χ2 of 1.12.

For the avoidance of doubt over the stoichiometry of the complex, we also created a pool of models with STAM2 bound at both ends of HD-PTP. As expected, these models gave poor fit to the data (best Χ2 = 2.19). Furthermore, modelling of hydrodynamic parameters was carried out using SOMO (Solution MOdeller)22 to confirm the HD-PTP:STAM2 complex structure (Table 7.2). Comparison against Rh, s20,w and f/f0 derived from MALS and AUC, favored the 1:1 stoichiometry, dual site binding model. This also showed how simulations binding only the Bro and SH3, selected for fitting against SAXS data, also converge onto models with similar hydrodynamic properties.

The bound STAM2 structure in these complex models is significantly longer than the structure we described for the STAM2 monomer data (Figure 7.5E). When we extracted

STAM2 molecule from the complex structure and fitted this to the STAM2 monomer data, it gave an extremely poor fit (Χ2 = 31.8). This suggests that HD-PTP binding

183 imposes structural constraints on STAM2 and is selecting conformers that are unfavorable in isolated STAM2.

Table 7.2. Hydrodynamic parameters for HD-PTP:STAM2 models vs experimental data.

(SAXS) (MALS) (AUC)

Experimental Rg Dmax Mr Rh Rh f/f0 s20,w (nm) (nm) (Da) (nm) (nm) (S)

HD-PTP:STAM2 6.65 22.8 100,000 5.77 6.45 2.04 3.94

Models χ2 Rg Dmax Mr Rh f/f0 s20,w (nm) (nm) (Da) (nm) (S) Bro and PRR restraints 1.12 6.878 24.89 105,765 5.87 1.87 4.26 Bro restraints, rigid STAM2 8.41 6.24 24.82 105,765 5.81 1.85 4.3 PRR restraints, rigid STAM2 4.28 7.67 25.91 105,765 5.8 1.85 4.31 Bro restraints, flexible STAM2 1.56 6.81 24.93 105,765 5.92 1.98 4.22 PRR restraints, flexible STAM2 1.32 7.1 25.93 105,765 5.66 1.8 4.42 HD-PTP with two STAM2 2.19 9.32 35.47 128,324 7.59 2.27 4.01

Discussion

HD-PTP acts as a multidomain binding platform for several ESCRT proteins. Many of the HD-PTP ESCRT binding interactions occur at overlapping binding sites which suggests a mechanism of competitive binding. It is difficult to narrow down the exact sequence of interactions between HD-PTP with each ESCRT. ESCRT-0 is canonically seen as the first complex to arrive at endosome for ILV formation. Even though it is known that HD-PTP binds the STAM2 subunit of ESCRT-0 at two different sites, it was unclear until now whether these interactions could occur at the same time.

Our results show that HD-PTP can indeed bind STAM2 at both the SH3 and GAT domains simultaneously. The STAM2 SH3-GAT linker region exhibits flexibility which is lost upon binding to HD-PTP. The indication that STAM2 is tightly tethered also suggests a reason for HD-PTP to maintain its long rigid platform. In contrast, HD-PTP’s homologue, ALIX, which does not bind STAM2, undergoes large scale conformational changes.23,24

184 We have previously suggested that HD-PTP manages controlled and sequential recruitment of ESCRT partners through competitive binding at overlapping sites of interaction.8 The homologue ALIX, is regulated by large scale conformational changes in its CC domain (or “V domain”) and autoinhibition of its Bro1 domain binding site by its own PRR. Our previous results have shown, firstly that HD-PTP does not form a V shaped CC domain or show any sign of induced conformational change; secondly that the

Bro1-CC-PRR structure is highly extended, which would make PRR interaction with the

Bro1 highly unlikely. These results now confirm that neither does the conformation of

HD-PTP alter when binding to STAM2. Together, these results highlights the importance of the competitive binding mechanism for HD-PTP regulation.

The close association of STAM2 along the length of HD-PTP suggests that, in addition to inhibited binding of other partners at the Bro1 domain and PRR sites, binding of

ESCRT-I component Ubap1 to the CC domain may also be inhibited.

Evidence for cooperativity has been found in the binding of STAM2 to diubiquitin.25–27

These interactions occur through its two ubiquitin binding domains VHS

(Vps27/Hrs/STAM) and UIM (ubiquitin interacting motif) and, similarly to the HD-PTP binding domain pair, these ubiquitin binding domains are also connected by a flexible linker. We showed that loss of interaction at either binding site still allowed stable coelution of a complex by size exclusion, however it may be possible that losing the divalent tethered binding interaction would significantly reduce the local binding affinity of the second site. A detailed investigation of binding affinities will be needed to understand the contribution of cooperativity of these interactions.

185

Methods

Cloning, Expression and Purification of HD-PTPBro-CC-PRR and STAM2SH3-GAT

STAM2SH3-GAT (204-377) was sub-cloned into a pNIC28a-Bsa4 vector (gift from Opher

Gileadi; Addgene no. 26103) by ligation independent cloning. HD-PTPBro1-CC-PRR in pNIC28-Bsa4 and HD-PTPBro1-CC in pET-28a vector were used from our previous work

(Chapter 6: Paper 3). Plasmids were transformed into BL21(DE3) cells and expressed by induction with 0.1 mM IPTG with shaking at 20 ⁰C for 18 hrs.

HD-PTPBro1-CC-PRR and STAM2SH3-GAT were purified separately by nickel affinity chromatography and size exclusion chromatography. Cell pellets were resuspended in 20 mM HEPES pH 7.4, 500 mM NaCl, 10 mM imidazole, 2 mM DTT, 0.1 % Triton x-100,

0.5 mg/mL lysozyme, and a Complete protease inhibitor tablet. After centrifugation at

13,000 rpm soluble supernatants were loaded onto 1 mL HisTrap columns at 0.5 mL/ min, washed with 20 mM HEPES pH 7.4, 500 mM NaCl, 10 mM imidazole, and eluted with a gradient up to 250 mM imidazole. 5 mM DTT and 5 mM EDTA were added to elution fractions. The cleanest fractions were concentrated to 250 μL using 10,000 MWCO PES membrane Vivaspin 2 columns, then loaded onto a Superdex 200 10/300 GL column (GE healthcare) through a 500 μL loop at 0.5 mL/min.

Coelution of HDPTPBro1-CC-PRR:STAM2SH3-GAT complex

An equimolar ratio of STAM2 and HDPTP was mixed for 30 mins with rocking and rolling in a 15 mL tube to allow efficient mixing. The mixture was then concentrated to

250 μL using 10,000 MWCO PES membrane Vivaspin 2 columns. Samples were loaded onto a Superdex 200 10/300 GL column (GE healthcare) column through a 500 μL loop at 0.5 mL/min, eluting in 20 mM HEPES, 250 mM NaCl, 2 mM EDTA, pH 7.4. Coelution

186 of the complex was confirmed by SDS-PAGE, using 10% NUPAGE gels and NOVEX

Sharp Prestained Protein Ladder.

Multiangle laser light scattering

Samples (1-3 mg/mL) were injected onto a Superdex 200 10/300 GL column (GE

Healthcare) in 20 mM HEPES, 250 mM NaCl, 2 mM EDTA, pH 7.4. MALS was measured using a Wyatt Helios 18-angle laser photometer with Wyatt EOS QELS detector. Concentrations were measured using a Wyatt rEX differential refractive index detector. Data were analysed and estimated molecular weights and hydrodynamic radii were calculated using ASTRA 6 software (Wyatt Technology Corp., CA, USA).

Sedimentation velocity – analytical ultracentrifugation

Samples (0.2 mg/mL) were measured using a XL-A ultracentrifuge (Beckman

Instruments) at 50,000 rpm (HDPTP and complex) or 45,000 rpm (STAM2) at 20 ˚C and detecting at a wavelength of 280 nm for a total of 200 scans. Boundary analysis of the raw data, yielding sedimentation coefficients was performed using Sedfit v8.7.22 s20,w, hydrodynamic radii (Rh) and frictional ratios (f/f0) were calculated using Sednterp.28

Partial specific volume, molecular weight and estimated hydration were calculated from the amino acid sequence using Sednterp. Solvent viscosity and density were calculated for the specific buffer components using Sednterp.

Small angle X-ray scattering

The individual proteins and complex were purified as described above. Samples were concentrated to 2.2 mg/mL, 5.2 mg/mL and 2.8 mg/mL for HDPTP, STAM2 and

HDPTP:STAM2 complex respectively. Samples were transported to the ESRF (European

Synchotron Radiation Facility, Grenoble, France) and size exclusion coupled-SAXS was carried out using a Superdex 200 Increase 3.2/300 on beamline 29 (MX1783). Guinier

187 fitting, D(max), P(r) distribution calculation and flexibility analysis was carried out using

SCÅTTER (http://www.bioisis.net/tutorial/9). Ab initio models were generated using

DAMMIF17 and averaged using DAMAVER.18

Torsional angle dynamics simulations

STAM2 models were built as follows: Homology modelling (using Swissmodel)29 was used to create a STAM2 GAT model from the STAM1 GAT crystal structure

(PDB:3F1I).19 The STAM2 NMR structure (1X2Q) was connected to the STAM2 GAT with a loop region of the correct protein sequence using CNS (Crystallography and NMR

System).30,31 The HD-PTPBro1-CC-PRR model was used from our previous publication

(Chapter 5, Paper 2). The interaction between the two proteins was modelled using CNS, with distance constraints representing the two binding interfaces: The HD-PTPBro-

STAM2GAT crystal structure (PDB:5crv)12 was used to generate constraints the Bro1 domain and the HD-PTPPRR-MonaSH3 crystal structure (PDB:2W10)13, by homology modelling, was used to generate distance constraints at the PRR.

Torsional angle simulations were used to carry out molecular modelling of the STAM2 monomer and of the HD-PTP-STAM2 complex. Models were selected based on their fit to the SAXS data using FOXS server20,21 and their fit to hydrodynamic parameters calculated using SOMO (SOlution MOdeller).22

Acknowledgements

This work was support by Medical Research Council (MR/K011049/1 to L.T. and P.W).

G.H. is funded by a BBSRC DTP Studentship. A.J.F. thanks Bruker for sponsorship. We thank the ESRF for access to beamline 29 (MX1783) and their staff for assistance. We thank the staff of the Biomolecular Analysis Facility of the Faculty of Biology Medicine and Health.

188 References for Chapter 7

1 A. Doyotte, A. Mironov, E. McKenzie and P. Woodman, Proc. Natl. Acad. Sci., 2008, 105, 6308–6313. 2 H. Ma, P. Wardega, D. Mazaud, A. Klosowska-Wardega, A. Jurek, U. Engström, J. Lennartsson and C.-H. Heldin, Cell. Signal., 2015, 27, 2209–2219. 3 D. Kharitidi, P. M. Apaja, S. Manteghi, K. Suzuki, E. Malitskaya, A. Roldan, M.-C. Gingras, J. Takagi, G. L. Lukacs and A. Pause, Cell Rep., 2015, 13, 599–609. 4 M. Mariotti, S. Castiglioni, J. M. Garcia-Manteiga, L. Beguinot and J. A. M. Maier, Int. J. Biochem. Cell Biol., 2009, 41, 687–693. 5 G. Lin, V. Aranda, S. K. Muthuswamy and N. K. Tonks, Genes Dev., 2011, 25, 1412–1425. 6 S. Manteghi, M.-C. Gingras, D. Kharitidi, L. Galarneau, M. Marques, M. Yan, R. Cencic, F. Robert, M. Paquet, M. Witcher, J. Pelletier and A. Pause, Cell Rep., 2016, 15, 1893–1900. 7 D.-Y. Chen, M.-Y. Li, S.-Y. Wu, Y.-L. Lin, S.-P. Tsai, P.-L. Lai, Y.-T. Lin, J.-C. Kuo, T.-C. Meng and G.-C. Chen, J. Cell Sci., 2012, 125, 4841–4852. 8 N. Ali, L. Zhang, S. Taylor, A. Mironov, S. Urbé and P. Woodman, Curr. Biol., 2013, 23, 453–461. 9 C. Raiborg, B. Bremnes, A. Mehlum, D. J. Gillooly, A. D’Arrigo, E. Stang and H. Stenmark, J. Cell Sci., 2001, 114, 2255–2263. 10 K. G. Bache, A. Brech, A. Mehlum and H. Stenmark, J. Cell Biol., 2003, 162, 435– 442. 11 E. Mizuno, K. Kawahata, M. Kato, N. Kitamura and M. Komada, Mol. Biol. Cell, 2003, 14, 3675–3689. 12 J. Lee, K.-J. Oh, D. Lee, B. Y. Kim, J. S. Choi, B. Ku and S. J. Kim, PLoS ONE, 2016, 11, e0149113. 13 M. Harkiolaki, T. Tsirka, M. Lewitzky, P. C. Simister, D. Joshi, L. E. Bird, E. Y. Jones, N. O’Reilly and S. M. Feller, Structure, 2009, 17, 809–822. 14 T. Kaneko, T. Kumasaka, T. Ganbe, T. Sato, K. Miyazawa, N. Kitamura and N. Tanaka, J. Biol. Chem., 2003, 278, 48162–48168. 15 R. P. Rambo and J. A. Tainer, Biopolymers, 2011, 95, 559–571. 16 L. J. McGuffin, K. Bryson and D. T. Jones, Bioinforma. Oxf. Engl., 2000, 16, 404– 405. 17 D. Franke and D. I. Svergun, J. Appl. Crystallogr., 2009, 42, 342–346. 18 V. V. Volkov and D. I. Svergun, J. Appl. Crystallogr., 2003, 36, 860–864. 19 X. Ren, D. P. Kloer, Y. C. Kim, R. Ghirlando, L. F. Saidi, G. Hummer and J. H. Hurley, Structure, 2009, 17, 406–416. 20 D. Schneidman-Duhovny, M. Hammel, J. A. Tainer and A. Sali, Biophys. J., 2013, 105, 962–974. 21 D. Schneidman-Duhovny, M. Hammel, J. A. Tainer and A. Sali, Nucleic Acids Res., 2016, 44, W424–W429. 22 E. Brookes, B. Demeler, C. Rosano and M. Rocco, Eur. Biophys. J. EBJ, 2010, 39, 423–435. 23 R. Pires, B. Hartlieb, L. Signor, G. Schoehn, S. Lata, M. Roessle, C. Moriscot, S. Popov, A. Hinz, M. Jamin, V. Boyer, R. Sadoul, E. Forest, D. I. Svergun, H. G. Göttlinger and W. Weissenhorn, Structure, 2009, 17, 843–856.

189 24 Q. Zhai, M. B. Landesman, H.-Y. Chung, A. Dierkers, C. M. Jeffries, J. Trewhella, C. P. Hill and W. I. Sundquist, J. Virol., 2011, 85, 9222–9226. 25 X. Ren and J. H. Hurley, EMBO J., 2010, 29, 1045–1054. 26 A. Lange, C. Castañeda, D. Hoeller, J.-M. Lancelin, D. Fushman and O. Walker, J. Biol. Chem., 2012, 287, 18687–18699. 27 M. Hologne, F.-X. Cantrelle, G. Riviere, F. Guillière, X. Trivelli and O. Walker, J. Mol. Biol., 2016, 428, 4544–4558. 28 T. M. Laue, B. D. Shah, T. M. Ridgeway and S. L. Pelletier, in Analytical ultracentrifugation in biochemistry and polymer science, Royal Society of Chemistry, 1992, pp. 90–125. 29 K. Arnold, L. Bordoli, J. Kopp and T. Schwede, Bioinformatics, 2006, 22, 195–201. 30 A. T. Brünger, P. D. Adams, G. M. Clore, W. L. DeLano, P. Gros, R. W. Grosse- Kunstleve, J. S. Jiang, J. Kuszewski, M. Nilges, N. S. Pannu, R. J. Read, L. M. Rice, T. Simonson and G. L. Warren, Acta Crystallogr. D Biol. Crystallogr., 1998, 54, 905–921. 31 A. T. Brunger, Nat. Protoc., 2007, 2, 2728–2733.

190 Chapter 8: Paper 5

Selective spin labelling of surface cysteines using a a bromoacrylaldehyde spin label

Graham Heaven1, Michael A. Hollas1, Philip Woodman2, Lydia Tabernero2 and Alistair J. Fielding1,*

Target journal: JACS

191 Abstract

Structural investigations of proteins and their complexes are now frequently complemented by distance constraints between spin labelled cysteines generated using double electron-electron resonance (DEER) spectroscopy. The most widely used spin label is MTSL (methane thiosulfonate spin label). In this paper we introduce the use of

BASL (bromoacrylaldehyde spin label) as a cysteine spin label. We show that using

BASL has a significant advantage over MTSL due to its increased selectivity for surface cysteines. Applied to the multidomain protein, his domain-protein tyrosine phosphatase

(HD-PTP), we show that BASL can be easily added in excess with selective labelling whereas MTSL causes protein precipitation and furthermore, using DEER, we were able to measure a single cysteine pair distance in a 3 cysteine domain, 9 cysteine protein and

10 cysteine protein complex.

Introduction

Modern studies in structural biology are facing increasingly complex systems and a range of techniques is required that can study not only small crystallisable subunits, but large multidomain proteins and their complexes. Double electron-electron resonance (DEER) spectroscopy has proven to be a useful addition to the structural biologist’s toolkit. It is used to extract distance constraints, in the range of 17-100 Å, between unpaired electrons in spin labelled cysteines, by measuring their electron-electron dipolar coupling interaction.1–3 This is a powerful method for investigating conformational changes in the presence of binding partners.

One of the key advantages of using DEER in biological studies is that it is selective for the radical signal regardless of its environment and therefore protein size, under appropriate conditions: i) The radical probe is in high enough concentration to measure. ii) The probe is not rapidly reduced by its environment. iii) There is no significant

192 contribution from other radical signals. This contrasts with nuclear magnetic resonance

(NMR) studies where the protein size poses a much greater challenge due to spectral crowding and line broadening.4,5 However, with increased protein size comes an increased number of cysteine residues, which requires more rounds of mutagenesis potentially making spin labelling approaches prohibitive in the case of large protein complexes.

The most widely used spin label is MTSL (methanethiosulfonate spin label) (Figure

8.1A). Site directed spin labelling (SDSL) studies are usually carried out on proteins whose 3D structures are known by crystallography, and labelling sites are chosen which are exposed to the surface. These surface cysteines are preferred because the resulting modified side chains are less likely to perturb structure; and MTSL can sample more easily predictable rotamers, allowing more accurate simulations of DEER distance distributions. Alternatively, it has also been shown that useful information can be extracted from multi-spin labelled proteins. Methods for extraction of distances from

DEER experiments, can cause multi-spin interactions to appear as additional “ghost” peaks in distance distributions, however experimental and analytical methods have been developed to identify and reduce these effects.6,7 These approaches has been successfully demonstrated using model compounds and homo-oligomeric protein complexes.6,8

We have previously used DEER to investigate the structure of the multidomain phosphatase HD-PTP (his domain protein tyrosine phosphatase).9 HD-PTP makes several interactions with endosomal sorting complexes required for transport (ESCRT) to mediate intralumenal vesicle formation at the endosome.10–13 The three native cysteines of HD-PTPCC domain were labelled by MTSL and individual distance measurements were determined by single cysteine to serine substitutions, and this was used to probe structure upon CC binding to ESCRT-I component Ubap1.

193 The interaction of HD-PTP with ESCRT-0 component STAM2 involves two interactions: one at the N terminal Bro1 domain, and another at the PRR which means that for analysis of structure in context of the double binding of HD-PTP to STAM2 requires expression

13,14 of a HD-PTPBro1-CC-PRR. One of the questions which this raised was whether binding of STAM2 and the Bro1 and PRR domain simultaneously would cause a conformational rearrangement of the CC domain to bring these binding sites closer together. This could be easily answered if the CC domain could again be labelled with MTSL and DEER measured with and without STAM2. However, this three-domain portion of HD-PTP contains 9 cysteines, so proves to be a more significant challenge for SDSL-DEER.

There are many alternative labels to MTSL which have been investigated for their various properties including radical stability, linker length and flexibility.15,16 One cysteine reactive compound that has been previously dismissed as a useful protein spin label is 3-

Bromo-4-formyl-2,2,5,5-tetramethyl-1H-pyrrol-1-yloxyl radical (Figure 8.1B). This compound has the same 5-membered pyrroline ring nitroxide “proxyl” moiety as MTSL but instead of a methane thiosulfonate (MTS) group has a bromine at position 4 and a formaldehyde group at position 3 of the ring. Separated by a double bond these two groups can be described as a bromoacrylaldehyde (BA) group so this compound is referred to as bromoacrylaldehyde spin label (BASL).

BASL was recently used in our group for spin labelling of gold nanoparticles.17 The bifunctional BA group allowed reaction with a thiol by nucleophilic substitution of the bromine and reaction with an amine by reductive amination of the aldehyde. Use of the thiol reactivity of BASL to label cysteine residues has already been demonstrated but potential reaction with amine side chains of lysine residue caused this to be described as an inappropriate cysteine label.18

194

Figure 8.1. Comparison of cysteine spin labels (A) MTSL and (B) BASL.

In this work, we show that BASL can be used as a selective cysteine spin label for DEER.

We find that, in comparison to MTSL, BASL is significantly more selective for surface cysteines. Using the multidomain phosphatase HD-PTP, we demonstrate several advantages of using the surface selective BASL, instead of MTSL. Firstly, BASL labelling here allows us to avoid multiple rounds of mutagenesis by labelling only 2 out of 9 possible cysteines, in contrast with MTSL which gives a heterogeneous mixture with up to 7 out of 9 labels. Secondly, when BASL is added in large excess there is no risk to protein stability whereas we find that MTSL causes precipitation by over-labelling.

Finally, we show that the native BASL labelled HD-PTP is able to interact with its binding partner STAM2 thus allowing conformational probing by DEER, whereas the multi-MTSL labelled HD-PTP is prevented from complex formation.

195

Results

BASL labelling of the three cysteine CC domain

HD-PTP CC domain contains three cysteines in the wild type sequence: C425, C628 and

C697. Our previous work showed that, using MTSL, all three cysteines were fully labelled by mass spectrometry (Supplemental Figure 8.1) which allowed us to use DEER to measure inter-spin distances with the three single cysteine mutations.9

HD-PTPCC wt was labelled with BASL, under the same conditions as MTSL, which resulted in addition of only two labels by mass spectrometry (Figure 8.2Ai). To show that

BASL was labelling cysteines rather than lysines, and to find out which cysteine was left unlabeled, we also carried out the same procedure on the three cysteine to serine mutants.

C697S (Figure 8.2Aii) gave a major mass peak at +2 BASL labels, like the wild type, indicating that Cys697 was not being labelled. C628S and C425S (Figure 8.2Aiii and

8.2Aiv) both had major peaks of +1 BASL, one less than the wild type, indicating that both these cysteines were labelled in the wild type sample.

All the mass spectra contained an extra small peak at around m + 387, highlighted in

Figure 8.2 with asterisks. The mass difference relative to the unlabeled proteins did not correspond to multiples of 166 for BASL modification but presumably represented a small molecular adduct with the protein which we were unable to identify. These extra peaks are also clear in the raw data for all ESI mass spectra (Supplemental Figure 8.2).

Continuous-wave (CW) EPR was carried out on the labeled wild type and mutant CC samples (Figure 8.2B). The spectra show the expected three peaks for nitroxide radicals with a significant broadening due to the slow tumbling of the proteins. These spectra also corroborate that BASL is a cysteine labeler, with differences in the spectra upon

196 mutagenesis, for example the loss of the feature at around 333 mT is present in all spectra except C425S (indicated with an arrow in the figure).

To rationalize the labelling preference of MTSL and BASL we performed cysteine site scans of the CC structure using MMM.19,20 This program superimposes MTSL rotamers at each cysteine and uses internal and interaction energies to predict the most stable rotamers. A partition function (PF) is assigned for each labelling site, which can be used as a measure of accessibility. Ideally a similar approach would be applied to the BASL label if crystal structures were known of the label and full rotamer simulations could be calculated. However, here our use of partition function for the MTSL side chain has given a useful prediction of accessibility for both labels. The average PF over the three available

CC crystal structures for C628 was 1.39, corresponding to an easily labelled site. C425 and C697 had a PF of 0.34 and 0.28, respectively, indicating similar accessibility for both sites. Our previous work showed that MTSL labelling combined with DEER was consistent with models of the protein where the central helix is flexible (Figure 8.2C).

Repeating the MMM site scan using one of these molecular models gave C425 a PF of

1.23, close to the PF for C628. Together these analyses support a steric hindrance argument for why BASL is able to label C628 and C425 but not C697.

197

Figure 8.2. CC domain labelling. (A) Deconvoluted mass spec of BASL-labelled HD- PTP CC wt (i) and mutants (ii, iii and iv). (B) CW EPR of BASL-labelled HD-PTP CC wt and mutants. (C) Cartoon structure of CC domain crystal structure (red) and helix 2 hinge model (grey) with cysteines shown as spheres and sulfur atoms in yellow. (D) Partition functions, as a measure of labelling site accessibility, calculated for all three CC cysteines using the X-ray crystal structure, and a second calculation for C425 using the molecular dynamics model.19,20

198 BASL labelling of the 6 cysteine Bro1 domain

HD-PTP Bro1 domain has 6 cysteines which are much more buried than in the CC domain. Our initial attempts to label this with MTSL resulted in protein precipitation; in as little as one hour, all protein had fallen from the solution preventing any sample from being obtained. We next tried to use moderate conditions, with MTSL at a stoichiometric molar ratio (1 label: 1 cysteine). This gave a heterogenous mixture of labelled protein with mainly 4 labels but also 2,3 and 5-labelled protein (Figure 8.3B), and after excess label removal through size-exclusion chromatography, no further precipitation was observed. The explanation for this seems to be that MTSL has the potential to label every cysteine in the Bro1 domain but this is avoided when the label is removed earlier; the protein can handle having 1-5 labels but once this reaches 6 labels the protein denatures and precipitates. The decrease in signal intensity of CW EPR spectra with longer MTSL incubation times also demonstrates this precipitation from solution (Figure 8.3D).

In contrast, when BASL was added in the 30x excess this caused no precipitation even after overnight labelling. Mass spectrometry after BASL labelling showed mainly unlabeled protein, with a small contribution of singly labelled protein Figure 8.3A. The presence of a small amount of spin label was confirmed by CW EPR (Figure 8.3C).

We also analyzed the Bro1 labelling sites with MMM.19,20 Only one cysteine is completely obstructed (C116, PF=0.00000), which is consistent with the MTSL mass spectrum showing up to 5 labels, leaving 1 cysteine unlabeled. C218 has the highest accessibility (PF=1.09) with the remaining four cysteines with PFs ranging from 0.40-

0.60. This C218 is probably the cysteine which is labelled to a low extent by BASL, with the others all too buried to be labelled.

199

Figure 8.3. Bro1 domain labelling. (A) Deconvoluted mass spectrum of BASL-labelled HD-PTP Bro1. (B) Deconvoluted mass spectrum of MTSL-labelled HD-PTP Bro1 after 1 hr incubation. (C) cwEPR of BASL-labelled HD-PTP Bro1. (D) cwEPR of MTSL- labelled HD-PTP Bro1 with various label incubation times. (E) Cartoon structure of Bro1 domain crystal structure (blue) with cysteines shown as spheres and sulfur atoms in yellow.21 (F) Partition functions, as a measure of labelling site accessibility, calculated for all six Bro1 cysteines using the X-ray crystal structure.19,20

200 The Bro1 labelling and MMM data together with the CC data indicate that MTSL (in low excess, 1 hour labelling) requires a minimum PF ~ 0.2 whereas BASL requires PF ~ 0.1

(Figure 8.4).

The results here also disagree with the previous assertion that BASL would label lysines as well as cysteines. HD-PTP has multiple lysine residues with 29 in the Bro1 domain and 23 in the CC domain but there is no evidence here of lysine labelling. The reported reaction conditions were in acetonitrile solvent in the presence of a base, 1,8- diazabicyclo[5.4.0]undec-7-ene at room temperature,18 whereas the reaction here was performed in pH 8 aqueous buffer at 4 ⁰C which appears to avoid any amine reaction.

Figure 8.4. Summary of calculated partition functions for MTSL labelling of cysteines in HD-PTP CC and Bro1. Ranges are shown for the minimum partition functions required for cysteine labelling by BASL (dashes) vs MTSL (dots).

201 Protein-protein interactions permitted by selective spin labelling

DEER spectroscopy is a powerful technique for measuring changes in conformations when in presence of ligand binding. Since MTSL labels all three CC domain cysteines and up to 5 of 6 Bro1 domain cysteines whereas BASL only labelled HD-PTP with 2 labels and leaves most Bro1 unlabelled, we predicted that the MTSL labelling would be more likely to inhibit HD-PTP from interacting with its binding partners.

To compare the binding of HD-PTP with the two different labels to the binding partner

STAM2, we performed size exclusion chromatography on mixes of unlabelled and MTSL and BASL labelled HD-PTP with STAM2 (Figure 8.5). Without any labelling, HD-PTP mixed with STAM2 gave an earlier peak, by size exclusion, than either monomeric proteins, consistent with an increase in size and was supported by SDS-PAGE clearly showing both proteins in this peak (Figure 8.5A). HD-PTP labelled with MTSL, on the other hand, was prevented from forming a complex with STAM2, with almost perfect overlap of the eluted mixture with the original monomeric HD-PTP and STAM2 peaks

(Figure 8.5C). However, when BASL-labelled HD-PTP was mixed with STAM2, they were able to coelute as a complex, similar to the unlabelled protein (Figure 8.5B). The

BASL complex elution peak is not shifted as drastically as the unlabelled protein, however, and this may indicate that the complex is only forming interactions at the Bro1-

GAT site and not the additional PRR-SH3 site which is prevented by the CC domain labels, whereas MTSL-HD-PTP PRR-SH3 is blocked by the CC labels and Bro1-GAT is blocked by Bro1 labels.

It is also noteworthy that BASL labelled HD-PTP in complex with STAM2 had been through two size exclusion columns since labelling, yet retained a strong CW EPR signal, which is good evidence that the BASL labelling reaction is irreversible.

202 DEER distance measurement between BASL label pair possible in single domain, full length protein and in heterodimer complex

DEER spectra for BASL-labelled WT and C697S CC domain are almost identical (Figure

8.6), which corroborates the evidence from mass spectrometry and CW EPR that C697 is not labelled by BASL, and, therefore, the distance distribution was between labels attached to C628 and C425 only. The distance distribution for both showed a broad and asymmetric peak with a maximum at 4.6 nm. As anticipated for a shorter label, this interspin distance was longer than the MTSL inter-spin distance of these same cysteine residues (4.2 nm).9

DEER was measured for BASL-labelled HD-PTP with and without STAM2, which both gave a broad distribution over a similar range to BASL-labelled WT and C697S CC domain (Figure 8.6). However, their distribution appeared to be more bimodal with a second maximum peak at around 3 nm. Compared to the CC DEER this could indicate that there are conformational changed in the CC domain when in the presence of the Bro1 and PRR domains or upon binding STAM2. However, these two DEER traces are visibly noisier than the apo-CC trace so it is difficult to draw strong conclusion. However, between the HD-PTP and HD-PTP:STAM2, of a similar noise level and DEER time window length, there is a marked similarity, which provides evidence that STAM2 binding to HD-PTP does not influence the conformation of the CC domain.

203

Figure 8.5. Complex formation of labelled HD-PTP. Size exclusion profiles and SDS- PAGE gels for STAM2 mixed with (A) unlabelled HD-PTP, (B) BASL-labelled HD-PTP and (C) MTSL-labelled HD-PTP. These profiles show that unlabelled HDPTP and BASL-labelled HD-PTP coelute with STAM2 as a complex, whereas MTSL-labelled HD-PTP is prohibited from binding. (D) CW EPR spectra of free BASL, BASL-labelled CC wt, BASL-labelled HD-PTP and BASL-labelled HD-PTP in complex with STAM2.

204

Figure 8.6. DEER spectroscopy of BASL labelled HD-PTP CC domain, C697S mutant and wild type, HDPTPBro1-CC-PRR and HDPTP:STAM2 complex.

205 Discussion

We have used SDSL and DEER to study the structure of HD-PTP9 (and Chapter 6:Paper

3). However, we have found ourselves somewhat limited in the scope of this methodology by the number of cysteines on this protein, and it is comparatively much easier to use techniques such as SAXS, MALS, AUC or crystallography to study structure. Using

BASL as an alternative to MTSL goes some way to tackling this limitation of SDSL-

DEER.

The study of proteins with multiple MTSL labels is possible, and is increasingly easier to handle with advances in the theory and practice of multi-spin systems.6–8 However, we demonstrated in this paper how MTSL labelling of buried cysteines lead to protein precipitation, and how MTSL interfered with protein function i.e. blocking protein- protein interaction of HD-PTP and STAM2.

BASL is selective for surface cysteines and the unfavorable buried positions are avoided without need for mutagenesis. Of course, mutagenesis of all 6 Bro1 cysteines and CC

C697 could be used to obtain the same doubly labelled HD-PTP sample reported here, but mutating up to 7 cysteines would increase the likelihood of affecting HD-PTP STAM2 binding interactions and could also affect expression or solubility of the resulting recombinant HD-PTP. By using BASL labelling as an alternative to the intensive mutagenesis approach needed for an MTSL study, we demonstrate how a useful EPR conformational probe can be gained with relatively minimal effort.

We had anticipated that STAM2 binding to HD-PTP at both its Bro1 and PRR domains may cause a contraction of the CC domain in between. However, the results here are able to show that there is no significant change in distances extracted through DEER measurements in the uncomplexed and complexed BASL-labelled HD-PTP. This is a

206 useful piece of evidence towards our understanding of HD-PTP structure which would not have been possible using MTSL, without multiple rounds of mutagenesis.

We have used native cysteines in this case, but this could be applied to any new protein study by first labelling with excess of BASL to identify any labelled cysteines to be mutated away, then introducing new surface sites for selective labelling. One way to quickly identify the labelled cysteines, not used here, is tryptic digestion coupled mass spectrometry.

Further investigation of BASL will be needed to better predict cysteine reactivity and to gain structural information needed to predict rotamers for DEER distance distributions.

Since MTSL seems to have moderate selectivity against buried cysteines whereas BASL is more selective, the investigation of this property of other spin labels with different reactive groups could also be investigated.

Materials and methods

Purification of HD-PTP

Cell pellets (~1.5 g after expression by IPTG induction in LB) were resuspended in 20 mM HEPES pH 7.4, 500 mM NaCl, 2 mM DTT, 0.1 % Triton x-100, 0.5 mg/mL lysozyme, and a Complete protease inhibitor tablet. For lysis, cells were sonicated 8 x 30 seconds on, 30 seconds off at 30 % amplitude. After centrifugation at 13,000 rpm lysates were purified by nickel affinity chromatography.

MTSL labelling

Nickel column elution fractions combined and 5 mM EDTA and 10 mM DTT were added before concentrating to 250 μL and loading onto a Superdex200 column equilibrated in

20 mM HEPES pH 7.4, 250 mM NaCl, 2 mM EDTA. This first size exclusion removed

207 DTT from the sample, allowing the cysteines to be labelled. Labelling was carried out in the presence of various amounts of the spin label MTSL (1-15x molar excess) for 1 hr at

4 ⁰C with gentle rocking and rolling. After this the sample was concentrated again for size exclusion to remove excess spin label.

BASL labelling

Protein fractions from the nickel column were combined as before but instead the

Superdex200 was equilibrated in 20 mM Tris pH 8.5, 250 mM NaCl, 2 mM EDTA.

Proteins were labelled with 15x excess of BASL overnight 4 ⁰C with gentle rocking and rolling. Size exclusion was used to remove excess spin label.

DEER sample preparation

After size-exclusion the buffer was exchanged into deuterated buffer (20 mM HEPES,

250 mM NaCl, pD 7.4 = pH 7.0 using a standard pH probe). DEER samples were prepared with 30% (v/v) glycerol-d8 to a final protein concentration of 60 μM. 120 μL samples were frozen inside 4 mm quartz tubes (Wilmad) by flash freezing with liquid nitrogen, and stored in a liquid nitrogen dewar until measurement.

CW EPR

CW EPR samples were prepared in capillary tubes inserted inside 4 mm quartz tubes

(Wilmad). CW EPR spectra were recorded on a Bruker EMX micro (9 GHz, “X-band”), accumulating 100 scans with a sweep width of 150 G, power attenuation of 20 mW, and a modulation amplitude of 1G.

Mass spectrometry

HD-PTPCC-MTSL mass spectra were recorded on an Agilent 6520 Q-TOF with an

Agilent 1200 LC system. BASL labelled samples were recorded on an Agilent 1290

Infinity II with an Agilent 6560 Ion Mobility Q-TOF-LC/MS or an Agilent 1200 series

208 with an Agilent 6510 Q-TOF LC/MS. Charges were assigned to peaks in the m/z spectra to allow deconvolution into a neutral mass spectrum using the open source program mMass.22

DEER spectroscopy

DEER experiments were carried out on a pulsed ELEXSYS E580 (9 GHz, “X-band”) spectrometer (Bruker), cooled to 50 K with a continuous-flow helium CF935 cryostat

(Oxford Instruments) and a ITC 502 temperature control system (Oxford Instruments). 4- pulse DEER sequence (π/2νobs−τ1−πνobs−t−πνpump−(τ1+τ2−t)−πνobs−τ2−echo) was applied,

1 with π/2νobs pulse length of 16 ns, πνobs pulse length of 32 ns. Pump pulses were applied at the maximum of the field sweep spectrum with the observe pulses 65 MHz lower. τ1 was varied by incrementing the first πνobs pulse position over eight steps of 56 ns for averaging of the deuterium nuclear modulation. Phase-cycling was applied. Matlab based program DEERAnalysis was used to correct for exponential background decay due to intermolecular interactions and to calculate the inter-spin distance distribution.23

Labelling site simulations

Labelling simulations to predict accessibility were performed using the Matlab based program MMM (multiscale modelling of macromolecules).19,20 In silico MTSL labelling was carried out using crystal structures of HD-PTP CC domain and HD-PTP Bro1 domain.9,21

BASL synthesis

BASL was prepared according to previous reports.17,18 Briefly, 4-oxo-2,2,5,5- tetramethyl-1H-pyrrol-1-yloxyl radical (4-oxo-TEMPO) was brominated and converted via Favorskii rearrangement to 3-bromo-4-carboxy-2,2,5,5-tetramethyl-1H-pyrrol-1- yloxyl radical. Thionyl chloride was used to convert this to its acyl chloride before

209 reduction with sodium borohydride to give the alcohol 3-bromo-4-hydroxymethyl-

2,2,5,5-tetramethyl-1H-pyrrol-1-yloxyl. This was oxidised with pyridinium dichromate to give the aldehyde 3-bromo-4-formyl-2,2,5,5-tetramethyl-1H-pyrrol-1-yloxyl radical

(BASL).

Acknowledgements

This work was support by Medical Research Council (MR/K011049/1 to LT and PW).

GH is funded by a BBSRC DTP Studentship and MAH is funded by an EPSRC DTA studentship. AJF thanks Bruker for sponsorship. All EPR experiments were carried out at the EPSRC National EPR Research Facility & Service. We thank Reynard Spiess for performing mass spectrometry experiments at the Manchester Institute of Biotechnology.

We thank the staff of the Biomolecular Analysis Facility of the Faculty of Biology

Medicine and Health, and the staff at the Advanced Photon Science facilities, University of Manchester. for assistance.

References for Chapter 8

1 M. Pannier, S. Veit, A. Godt, G. Jeschke and H. . Spiess, J. Magn. Reson., 2000, 142, 331–340. 2 G. Jeschke, Annu. Rev. Phys. Chem., 2012, 63, 419–446. 3 R. Ward, A. Bowman, E. Sozudogru, H. El-Mkami, T. Owen-Hughes and D. G. Norman, J. Magn. Reson. San Diego Calif 1997, 2010, 207, 164–167. 4 A. G. Tzakos, C. R. R. Grace, P. J. Lukavsky and R. Riek, Annu. Rev. Biophys. Biomol. Struct., 2006, 35, 319–342. 5 M. P. Foster, C. A. McElroy and C. D. Amero, Biochemistry (Mosc.), 2007, 46, 331– 340. 6 G. Jeschke, M. Sajid, M. Schulte and A. Godt, Phys. Chem. Chem. Phys., 2009, 11, 6580–6591. 7 A. Giannoulis, R. Ward, E. Branigan, J. H. Naismith and B. E. Bode, Mol. Phys., 2013, 111, 2845–2854. 8 S. Valera, K. Ackermann, C. Pliotas, H. Huang, J. H. Naismith and B. E. Bode, Chem. – Eur. J., 2016, 22, 4700–4703. 9 D. Gahloth, C. Levy, G. Heaven, F. Stefani, L. Wunderley, P. Mould, M. J. Cliff, J. Bella, A. J. Fielding, P. Woodman and L. Tabernero, Structure, 2016, 24, 2115–2126. 10 F. Ichioka, E. Takaya, H. Suzuki, S. Kajigaya, V. L. Buchman, H. Shibata and M. Maki, Arch. Biochem. Biophys., 2007, 457, 142–149. 11 A. Doyotte, A. Mironov, E. McKenzie and P. Woodman, Proc. Natl. Acad. Sci., 2008, 105, 6308–6313.

210 12 F. Stefani, L. Zhang, S. Taylor, J. Donovan, S. Rollinson, A. Doyotte, K. Brownhill, J. Bennion, S. Pickering-Brown and P. Woodman, Curr. Biol., 2011, 21, 1245–1250. 13 N. Ali, L. Zhang, S. Taylor, A. Mironov, S. Urbé and P. Woodman, Curr. Biol., 2013, 23, 453–461. 14 J. Lee, K.-J. Oh, D. Lee, B. Y. Kim, J. S. Choi, B. Ku and S. J. Kim, PLoS ONE, 2016, 11, e0149113. 15 W. L. Hubbell, C. J. López, C. Altenbach and Z. Yang, Curr. Opin. Struct. Biol., 2013, 23, 725–733. 16 A. J. Fielding, M. G. Concilio, G. Heaven and M. A. Hollas, Molecules, 2014, 19, 16998–17025. 17 M. A. Hollas, S. J. Webb, S. L. Flitsch and A. J. Fielding, Angew. Chem. Int. Ed., 2017, 56, 9449–9453. 18 T. Kálai, M. Balog, J. Jekö and K. Hideg, Synthesis, 1998, 1998, 1476–1482. 19 Y. Polyhach and G. Jeschke, J. Spectrosc., 2010, 24, 651–659. 20 Y. Polyhach, E. Bordignon and G. Jeschke, Phys. Chem. Chem. Phys., 2011, 13, 2356–2366. 21 P. Sette, R. Mu, V. Dussupt, J. Jiang, G. Snyder, P. Smith, T. S. Xiao and F. Bouamr, Structure, 2011, 19, 1485–1495. 22 M. Strohalm, D. Kavan, P. Novák, M. Volný and V. Havlíček, Anal. Chem., 2010, 82, 4648–4651. 23 G. Jeschke, V. Chechik, P. Ionita, A. Godt, H. Zimmermann, J. Banham, C. R. Timmel, D. Hilger and H. Jung, Appl. Magn. Reson., 2006, 30, 473–498.

211

Supplemental Figure 8.1. Mass spec of HD-PTP CC wt labelling with MTSL.

212

Supplemental Figure 8.2. Raw ESI mass spec of CC BASL-labelled samples.

213 Chapter 9: Conclusions

9.1 Summary of results

Combined use of structural methods has given an insight into the arrangement of the multidomain ESCRT adaptor protein, HD-PTP. The previously uncharacterised CC domain and its interaction with Ubap1 has been described, along with the arrangement of the three N-terminal ESCRT-interacting domains of HD-PTP. The nature of the binding interaction of HD-PTP with STAM2 (ESCRT-0) has been determined, and furthermore this interaction has been investigated by use of a novel spin label probe.

9.1.1 HD-PTPCC domain adopts elongated structure unlike Bro1 and ALIX V domains.

Paper 1 investigated the structure of HD-PTPCC domain and its interaction with Ubap1 by X-ray crystallography. Unexpectedly, it revealed a CC domain structure which was significantly more extended than the V-shaped structures of HD-PTP homologues Bro1 and ALIX.

The contribution of double electron-electron resonance (DEER) spectroscopy to this work allowed additional verification of this extended structure. Significantly, labelling of the most distal cysteines, C628 and C697 resulted in a DEER trace with negligible dipolar coupling, indicative of a large inter-spin distance. Interestingly, the comparison of DEER derived distances and crystal structure based simulations indicated that there was a deviation of the position of the second alpha helix in solution.

214 9.1.2 HD-PTPCC domain conformation not modulated by Ubap1 or monoubiquitin binding.

Paper 2 followed up the discrepancy between DEER measurements and crystallography of HD-PTPCC. It seemed likely that, given the proximity of CC helix 2 to the Ubap1 binding site, the orientation of CC helix 2 may be affected by Ubap1 interactions. The influence of frontal temporal lobar degeneration-linked mutation of Ubap1, P256L was also investigated. It seemed plausible that this mutation, only six residues upstream of the previously mapped CC binding site motif, may also have some influence of the CC structural conformation.

However, when Ubap1 was added to the CC DEER samples, no significant changes were observed in the DEER distance distributions, showing that the helix 2 position was unaffected, and neither did the P256L mutation have any affect. Further investigation by

X-ray crystallography clearly demonstrated that this region of Ubap1 formed no stable interaction with CC, and interactions were limited to the previously published motif.

9.1.3 HD-PTP Bro CC and PRR domains extend into a long binding platform for

ESCRT binding.

Paper 3 investigated the overall domain structure of the three N-terminal domains of

HD-PTP (HD-PTPBro1-CC-PRR), which contains all of its ESCRT-interaction motifs.

Solution models were simulated based on the combined information from multi-angle laser light scattering (MALS), analytical ultracentrifugation (AUC) and small-angle X- ray scattering (SAXS) experiments. The results showed that the Bro1 and CC domains are arranged in an extended conformation spanning 19.3 nm long, with addition of the

PRR domain taking this to 20.4 nm.

215 One of the questions raised was how HD-PTP is regulated. Its homologue ALIX prevents

Bro1 binding by autoinhibition by self-interaction of the PRR with the Bro1 domain across the folded V domain but in the case of HD-PTP this would not be possible as the

PRR is too far away. This indicated the importance of competitive binding of HD-PTP binding partners, to avoid indiscriminate binding of all ESCRTs, simultaneously.

Another unanswered question was how the relatively small protein STAM2, which binds both HD-PTP Bro1 and PRR, can bind such a long-elongated protein and whether

STAM2 might cause HD-PTP to contract.

9.1.4 STAM2 SH3 and Core domains are connected by a flexible linker which stretch out across the extended CC domain to bind HD-PTP Bro1 and PRR simultaneously.

In paper 4 the structure of ESCRT-0 component STAM2 and its interaction with HD-PTP were investigated, again using MALS, AUC and SAXS. STAM2 SH3 and core domains are separated by 54 amino acid residues. STAM2 studied alone, showed that it formed a relatively compact structure with a maximum dimension of 14.1 nm. SAXS analysis of the HD-PTP STAM2 complex showed an elongated structure which is slightly longer and thicker than HD-PTP alone, suggesting that STAM2 stretches out across the full length of HD-PTP and does not induce any contraction of the HD-PTP across the CC domain.

The bound structure of STAM2 is significantly longer than its monomeric structure, indicating that its flexible region is tethered by dual domain binding to HD-PTP. This dual site binding also indicates a possible cooperative binding mechanism between

STAM2 and HD-PTP, which has already been described with STAM2 binding to diubiquitin.

216 9.1.5 Use of a bromoacrylaldehyde spin label allows probing of DEER distances in multidomain HD-PTP.

The portion of HD-PTP which interacts with STAM2 spans three domains. Investigating this region using spin labelling and DEER spectroscopy was more challenging as this contained 9 cysteines, which would require multiple rounds of mutagenesis to obtain a doubly labelled sample, or involve the challenging interpretation of multispin DEER.

Paper 5 investigated a novel approach to this problem by comparing bromoacrylaldehyde spin label (BASL) with the commonly used methanethiosulfonate spin label (MTSL). The results showed that BASL had significantly increased selectively for surface cysteines over MTSL, which gave BASL certain advantages over MTSL. Firstly, BASL could label two out of three cysteines in the CC domain, without the need for mutagenesis. Secondly,

BASL incubation with the Bro1 domain gave only a small amount of singly labelled Bro1, in contrast to MTSL which gave a heterogeneous mixture of multiple labels and also suffered from the risk of protein precipitation in any attempts to increase the labelling efficiency. Finally, BASL labelling of the full Bro1-CC-PRR region of HD-PTP allowed measurement of the CC domain inter-cysteine distance with and without STAM2, whereas HD-PTP labelled with MTSL, was prohibited from STAM2 binding due to the large number of labels on the Bro1 domain.

The DEER results ruled out any contraction of the CC domain into a V-like domain upon

HD-PTP binding to STAM2. This label could also be used to tackle similarly challenging multidomain proteins in the future.

217 9.2 Final discussion

The combination of methods to characterise the structure of HD-PTP in this work have, together, been able to improve our understanding of the mechanism HD-PTP involvement in the ESCRT-mediated endolysosomal pathway.

One of the most prominent findings was the nature of the structurally distinct coiled-coil domain of HD-PTP. Its elongated structure differed dramatically with the V-shaped domains of both the yeast homologue, Bro1,1 and the human homologue, ALIX.2

Furthermore, there is significant evidence that the function of ALIX is dependent on dimerization via the V domain,3,4 and that this dimerization accompanies large-scale conformational changes in the V domain and the detachment of the autoinhibitory PRR from the Bro1 domain;1,5,6 but for HD-PTP, there was no evidence for dimerization and neither was there any evidence for such a large-scale conformational change (Figure 9.1).

Figure 9.1. Comparison of ALIX and HD-PTP. ALIX activation involves dimerization and a large scale conformational change in the V domain; whereas data for HD-PTP does not support dimerization or large scale conformational changes in the CC domain.

The extended CC structure, significantly, appeared to rule out an autoinhibitory function of the PRR, as observed in ALIX. This was further supported after structural studies of 218 HD-PTPBro1-CC-PRR showed an extended multidomain structure with the PRR far away from the Bro1 domain. This appeared to show that regulation of HD-PTP binding to

ESCRTs would have to rely strongly on competitive binding to its multiple partners

(Figure 9.2), rather than an ALIX-like mechanism of inactive-to-active conformational switching.

Figure 9.2. Competitive displacement mechanism for HD-PTP interaction with ESCRTs.

This work, along with previous studies,7–9 suggest a mechanism for HD-PTP interaction with ESCRTs where: first, HD-PTP binds to cargo-bound ESCRT-0 subunit STAM2 by its PRR which binds STAM2 SH3 domain and its Bro1 domain which binds STAM2 core domain (Figure 9.2, Stage 1); next, HD-PTP is displaced from STAM2 SH3 by binding to both ESCRT-I subunit TSG101 at its PRR and ESCRT-I subunit Ubap1 at its CC domain (Figure 9.2, Stage 2); finally, HD-PTP disengages from STAM2 completely after losing its Bro1 interaction with STAM2 core in favour of binding ESCRT-III subunit

CHMP4 (Figure 9.2, Stage 3).

The further structural study of HD-PTP in complex with STAM2 supported the mechanism that STAM2 can bind across the full length of HD-PTP which would likely block binding of Ubap1 at the CC domain until the SH3-PRR interaction is alleviated.

219 Furthermore, mutants of Bro1 to block STAM2 core binding and HD-PTP lacking the

PRR to block STAM2 SH3 binding both resulted in complexes which were stable in solution, supporting the presence of these intermediates.

Although HD-PTP-mediated ESCRT pathway can bypasses ESCRT-II,10 The overall shape of HD-PTPBro-CC-PRR is long and slightly curved, which bears greater similarity to

ESCRT-I than to ESCRT-II (Figure 9.3). In the case of ESCRT-I, this crescent shape is thought to induce membrane curvature in preparation for membrane budding.11 The solution structure of ALIX also forms a similar crescent shape5 and ALIX Bro1 domain has previously been shown to interact with membranes.5,12 However, it has also been reported that for the membrane binding of ALIX does not appear to induce membrane curvature.5 Instead, the curvature may be needed to sense curved membrane surfaces for bridging its proteins binding partners. Given the common crescent shape between

ESCRT-I, ALIX and HD-PTP, it seems likely that HD-PTP may also bind membranes.

220

Figure 9.3. Comparison of HD-PTP with ESCRT-I and ESCRT-II. Structures of HD- PTP(PDB:5LM2), ESCRT-I (PDB:2P22)13 and ESCRT-II (PDB:2ZME)14 are shown along with a structural alignment of HD-PTP and ESCRT-I, performed using Supcomb.15

Interpretation of the evidence for conformational dynamics within ALIX V domain is complicated. Initial crystallography experiments showed V shaped structures,2,16 then solution structures by SAXS showed ALIX forming much more elongated crescent shaped dimers,5 and later crystallography also gave evidence that the V domain can form an intermediate open right angle conformation.1 The flexibility has been interpreted as being required for ALIX autoregulation, along with the autoinhibition of the Bro1 domain by the PRR and monomer to dimer exchange.5,6 Given what we now know about the contrasting lack of flexibility in HD-PTP, another possibility is that the flexibility of

221 ALIX contributes to its ability to interact with membrane in such a wide range of ESCRT events in the cell.

ALIX operates at ESCRT mediated scission at various membrane surfaces: ILV formation on the endosome;3 cytokinesis at the midbody;17 and exosome formation18 and viral budding19 at the plasma membrane. The different shapes and curvatures of these membrane surfaces may benefit from having an ESCRT adaptor, such as ALIX, which is variable in shape and can bridge a variety of membrane curvatures. HD-PTP, on the other hand, has a limited role within endosomal trafficking, which is consistent with its more rigid structure which may only match a limited range of membrane curvatures.

One approach to investigate this idea would be to use DEER spectroscopy on spin labelled

ALIX V domain to monitor the arm spacing in the presence of a range of different lipid vesicles sizes. Changes in the conformations of the V domain with different vesicles would support the idea that ALIX could vary in shape to sense a wide range of curvatures.

These experiments would ideally need to be carried out using full length ALIX, which contains 11 cysteines. To avoid having to mutate every cysteine, any of which could have functional or structural importance, this further work could benefit from the use of the surface selective BASL spin label, as demonstrated with HD-PTP.

References for Chapter 9

1 N. Pashkova, L. Gakhar, S. C. Winistorfer, A. B. Sunshine, M. Rich, M. J. Dunham, L. Yu and R. C. Piper, Dev. Cell, 2013, 25, 520–533. 2 S. Lee, A. Joshi, K. Nagashima, E. O. Freed and J. H. Hurley, Nat. Struct. Mol. Biol., 2007, 14, 194–199. 3 S. Sun, X. Zhou, W. Zhang, G. E. Gallick and J. Kuang, Biochem. J., 2015, 466, 475– 487. 4 S. Sun, L. Sun, X. Zhou, C. Wu, R. Wang, S.-H. Lin and J. Kuang, Dev. Cell, 2016, 36, 331–343.

222 5 R. Pires, B. Hartlieb, L. Signor, G. Schoehn, S. Lata, M. Roessle, C. Moriscot, S. Popov, A. Hinz, M. Jamin, V. Boyer, R. Sadoul, E. Forest, D. I. Svergun, H. G. Göttlinger and W. Weissenhorn, Structure, 2009, 17, 843–856. 6 Q. Zhai, M. B. Landesman, H.-Y. Chung, A. Dierkers, C. M. Jeffries, J. Trewhella, C. P. Hill and W. I. Sundquist, J. Virol., 2011, 85, 9222–9226. 7 A. Doyotte, A. Mironov, E. McKenzie and P. Woodman, Proc. Natl. Acad. Sci., 2008, 105, 6308–6313. 8 F. Stefani, L. Zhang, S. Taylor, J. Donovan, S. Rollinson, A. Doyotte, K. Brownhill, J. Bennion, S. Pickering-Brown and P. Woodman, Curr. Biol., 2011, 21, 1245–1250. 9 N. Ali, L. Zhang, S. Taylor, A. Mironov, S. Urbé and P. Woodman, Curr. Biol., 2013, 23, 453–461. 10 M. D. J. Parkinson, S. C. Piper, N. A. Bright, J. L. Evans, J. M. Boname, K. Bowers, P. J. Lehner and J. P. Luzio, Biochem. J., 2015, 471, 79–88. 11 E. Boura, B. Różycki, H. S. Chung, D. Z. Herrick, B. Canagarajah, D. S. Cafiso, W. A. Eaton, G. Hummer and J. H. Hurley, Struct. Lond. Engl. 1993, 2012, 20, 874–886. 12 I. Kufareva, M. Lenoir, F. Dancea, P. Sridhar, E. Raush, C. Bissig, J. Gruenberg, R. Abagyan and M. Overduin, Biochem. Cell Biol. Biochim. Biol. Cell., 2014, 92, 555– 563. 13 M. S. Kostelansky, C. Schluter, Y. Y. C. Tam, S. Lee, R. Ghirlando, B. Beach, E. Conibear and J. H. Hurley, Cell, 2007, 129, 485–498. 14 Y. J. Im and J. H. Hurley, Dev. Cell, 2008, 14, 902–913. 15 M. B. Kozin and D. I. Svergun, J. Appl. Crystallogr., 2001, 34, 33–41. 16 R. D. Fisher, H.-Y. Chung, Q. Zhai, H. Robinson, W. I. Sundquist and C. P. Hill, Cell, 2007, 128, 841–852. 17 J. G. Carlton and J. Martin-Serrano, Science, 2007, 316, 1908–1912. 18 M. F. Baietti, Z. Zhang, E. Mortier, A. Melchior, G. Degeest, A. Geeraerts, Y. Ivarsson, F. Depoortere, C. Coomans, E. Vermeiren, P. Zimmermann and G. David, Nat. Cell Biol., 2012, 14, 677–685. 19 B. Strack, A. Calistri, S. Craig, E. Popova and H. G. Göttlinger, Cell, 2003, 114, 689–699.

223