University of Southern California Dissertations and Theses

Protein-protein interaction and protein-lipid interaction of membrane proteins

(USC Thesis Other)

Protein-protein interaction and protein-lipid interaction of membrane proteins

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content PROTEIN-PROTEIN INTERACTION AND PROTEIN-LIPID INTERACTION OF MEMBRANE PROTEINS by Meixin Tao A Dissertation Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY MEDICAL BIOPHYSICS May 2020 Copyright 2020 Meixin Tao ii Acknowledgements I would like to take this opportunity to express my sincere thanks to my mentor, Dr. Ralf Langen, for taking me as his graduate student, getting me enrolled into the on-going exciting work of the Langen Lab. None of the studies presented in this thesis would have been achieved without his patient guidance, support and encouragement over the years. I am also grateful to my other qualifying exam and thesis committee members: Dr. Ansgar Siemer, Dr. Ian Haworth, Dr. Tobias Ulmer, Dr. Derek Sieburth and Dr. Robert Chow for their suggestions on the projects. Additionally, I would like to thank my lab members: Drs. Mario Isas, Nitin Pandey, Jobin Varkey, Jose Bravo, Anoop Rawat as well as previous lab members: Mark Ambroso, Alan Okada, Natalie Kegulian and Kazuki Teranishi for their technical training and friendship. I particularly like the productive and amiable atmosphere of our lab. Together we have confronted frustrations of the failed experiments and cheered over the successful ones. For our recently joined new lab members: Elissa Fultz, Joshua Lugo and Bryan Chau, I wish them all the best of luck and would like to pass them the courage for upcoming challenges. Specifically, for the huntingtin project, I would like to thank our collaborator from UC Santa Barbara: Dr. Songi Han and her student Ryan Barnes, for their knowledge and assistance with the ODNP studies. For the annexin A7 project, sincere thanks go to Dr. Jeannie Chen and Yun Yao for suggestions and technical assistance with the cell work and Dr. Seth Ruffins for training and troubleshooting of the microscopy imaging. Finally, my special thanks go to my parents, my grandparents and my best friends. I feel so lucky and grateful to have you all by my side. iii Table of Contents Acknowledgements ii List of Tables vi List of Figures vii Abstract ix Preface xi Chapter 1: Introduction 1 1.1 Lipid Membranes and Membrane Proteins 1 1.2 EPR and Its Application in Membrane Protein Structural Studies 6 1.2.1 Spin mobility 10 1.2.2 Membrane accessibility 13 1.3 Amyloid Protein and Its Membrane Interaction 16 1.3.1 Introduction to amyloid proteins 16 1.3.2 Membrane interaction and membrane-mediated aggregation of amyloid proteins 20 1.4 Protein Phase Separation 23 1.4.1. Introduction to protein liquid-liquid phase separation (LLPS) 23 1.4.2. Protein phase separation and amyloid fibril formation 27 Chapter 2: Annexin B12 Trimer Formation is Governed by a Network of Protein-Protein and Protein-Lipid Interactions 29 Abstract 29 2.1 Introduction 29 2.2 Results 34 2.2.1 Some protein-protein interactions, but not the salt bridges, have a strong effect on ANXB12 trimer formation. 38 2.2.2 Some Ca 2+ -ligands in AB membrane binding sites significantly contribute to ANXB12 trimer formation on membranes. 40 2.2.3 ANXA2-mimicking Lys mutations at the tip of AB binding loops cause a modest reduction of trimer formation. 41 2.2.4 The AB’ ligand E105 is important for trimer formation 42 2.3 Discussion 42 2.4 Materials and Methods 47 2.4.1 Mutagenesis, Expression, Purification and Spin Labeling of ANXB12 mutants 47 2.4.2 Preparation of Lipid Vesicles 48 2.4.3 CW-EPR Spectroscopy Analysis 49 2.4.4 Fluorescence Microscopy 49 Chapter 3: Structure of Membrane-Bound Huntingtin Exon 1 Reveals Membrane Interaction and Aggregation Mechanisms 54 Abstract 54 iv 3.1 Introduction 54 3.2 Results: 57 3.2.1 Httex1(Q25) membrane interaction is mediated by its N-terminal region 57 3.2.2 Accessibility measurements reveal formation of a short amphipathic α-helix in N17 with shallow membrane penetration. 59 3.2.3 ODNP measurements indicate solvent-exposed, C-terminal region that does not interact with the membrane. 63 3.2.4 Httex1(Q25) membrane binding depends on vesicle size and membrane curvature 64 3.2.5 Electrostatic interactions contribute to Httex1 membrane interaction 66 3.2.6 Phosphomimetic mutations in N17 potently decrease the membrane binding affinity of Httex1(Q25) and protect Httex1(Q46) from membrane-mediated aggregation. 68 3.3 Discussion 69 3.4 Materials and Methods 81 3.4.1 Preparation of R1-labeled Httex1 Derivatives 81 3.4.2 Preparation of Small and Large Unilamellar Vesicles 81 3.4.3 Transmission Electron Microscopy 82 3.4.4 EPR Measurements 82 3.4.5 ODNP Measurements 84 3.4.6 Circular Dichroism 85 Chapter 4: The N-terminus of Membrane Protein Annexin A7 is a Functional Domain Mediating Liquid-Liquid Phase Separation and Vesicle Clustering 86 Abstract 86 4.1 Introduction 86 4.2 Results 89 4.2.1 Endogenous ANXA7 is recruited into granular structures upon arsenite-induced cellular stress 90 4.2.2 Overexpression of ANXA7 variants containing the N-terminal sequence leads to the formation of cellular aggregates. 90 4.2.3 Recombinant ANXA7 undergoes pH- and salt-sensitive liquid-liquid phase separation in vitro, mediated by its low-complexity N-terminus 91 4.2.4 FRAP analysis indicates a high percentage of mobile component within the ANXA7 droplets. 93 4.2.5 The C-terminal core domain is sufficient for Ca 2+ -mediated membrane interaction of ANXA7 but vesicle clustering requires the additional contribution from the N-terminal domain. 94 4.3 Discussion 96 4.4 Materials and Methods 98 4.4.1 Plasmid DNA Preparation 98 4.4.2 Cell Culture and Immunostaining. 99 4.4.3 Recombinant Protein Expression and Purification 99 4.4.4 Preparation of Lipid Vesicles 100 4.4.5 Confocal Fluorescence Microscopy Imaging 100 4.4.6 Time Course Measurement of Solution Turbidity 101 4.4.7 FRAP Measurement 101 Concluding Remarks and Future Directions 112 v References 114 Appendices 132 Appendix A: Potential Role of Annexin A7 in Huntingtin Cellular Aggregation 132 Appendix B: The Establishment of ANXA7-/- Cell Line by CRISPR-Cas9 137 Appendix C: Analysis of Huntingtin-targeting Peptides 141 Section 1: Analyzing the potential of anti-Htt peptides as diagnostic probes 141 Section 2: Analyzing the potential of anti-Htt dipeptides to inhibit Htt aggregation 142 vi List of Tables Table 2-1. Summary of ANXB12 mutants in the study ................................................................ 46 Table B-1. Summary of gRNA sequence targeting ANXA7 ...................................................... 137 Table B-2. Summary of genome sequencing primers for ANXA7-/- verification ..................... 138 vii List of Figures Figure 1-1. Diversity in the distribution and structure of lipid molecules ...................................... 2 Figure 1-2. Highlights of some structural and dynamic properties of lipid bilayer ........................ 4 Figure 1-3. Brief introduction of EPR ............................................................................................ 7 Figure 1-4. Site-directed spin labeling with MTSL ........................................................................ 9 Figure 1-5. EPR spin mobility and its applications ...................................................................... 11 Figure 1-6. EPR power saturation and its applications ................................................................. 14 Figure 1-7. Amyloid proteins and their fibrillar structures ........................................................... 17 Figure 1-8. Nucleation growth mechanism of amyloid formation ............................................... 19 Figure 1-9. Membrane interaction and membrane-mediated aggregation of amyloid proteins ....................................................................................................................... 21 Figure 1-10. Liquid-liquid phase separation and its cellular roles ................................................ 24 Figure 1-11. LLPS-mediated fibril formation ............................................................................... 28 Figure 2-1. ANXB12 monomer and trimer structures showing Ca 2+ - and lipid-binding sites and inter subunit salt bridges ............................................................................. 33 Figure 2-2. EPR and fluorescence-based detection of ANXB12 membrane binding and trimer formation ......................................................................................................... 35 Figure 2-3. EPR spectra of ANXB12 derivatives with mutations designed to disrupt protein-protein or protein-membrane interactions ..................................................... 37 Figure 2-4. ANXB12 trimer formation estimated from EPR amplitudes ..................................... 39 Figure 2-5. Convex surface of the ANXB12 trimer showing the location of disrupted protein-lipid contacts ................................................................................................. 43 Figure 2-S1. Sequence alignment of non-trimer forming ANXA2 and trimer-forming ANXB12 .................................................................................................................. 50 Figure 2-S2. EPR spectra from Figure 3 shown at same amplitude ............................................. 51 Figure 2-S3. Protein-protein contact surface disrupted by IF-A mutations .................................. 52 Figure 2-S4. Verification of labeling efficiency for the top four mutants most significantly disrupted in trimer formation. ............................................................. 53 Figure 3-1. CW-EPR spectra of singly R1-labeled Httex1(Q25) derivatives ............................... 58 Figure 3-2. EPR accessibility measurements and CD indicate the formation of helical structure upon membrane binding .............................................................................. 60 Figure 3-3. Httex1(Q25) ODNP measurement ............................................................................. 62 Figure 3-4. Membrane curvature, ionic strength, and charge modulate Httex1(Q25) membrane binding ...................................................................................................... 65 Figure 3-5. Phosphomimetic (S13D/S16D) mutations decrease membrane-binding affinity of Httex1(Q25) and protect Httex1(Q46) from membrane-mediated aggregation ................................................................................................................. 67 Figure 3-6. Schematic illustration of Httex1 membrane interaction ............................................. 71 Figure 3-S1. Httex1 domain organization, R1 labeling sites, and analysis of vesicle as well as protein stability over time ............................................................................ 75 Figure 3-S2. CW-EPR spectra of singly R1-labeled Httex1(Q25) stably bound to SUVs ........... 76 viii Figure 3-S3. EPR-based accessibility measurements ................................................................... 77 Figure 3-S4. k𝜎(P) fitting and site-specific kσ of Httex1(Q25) ..................................................... 78 Figure 3-S5. EPR spectra for membrane curvature, salt and lipid composition dependence of membrane binding ............................................................................................... 79 Figure 3-S6. EPR and CD spectral analyses of the S13D/S16D mutations on Httex1(Q25) membrane binding and EPR time course of Httex1(Q46) aggregation ................... 80 Figure 4-1. Endogenous ANXA7 is recruited into granular structures upon arsenite- induced cellular stress .............................................................................................. 103 Figure 4-2. Overexpression of ANXA7 variants containing the N-terminal sequence leads to the formation of cellular aggregates .................................................................... 104 Figure 4-3. Recombinant ANXA7 undergoes pH-sensitive liquid-liquid phase separation in vitro, mediated by its low-complexity N-terminus .............................................. 105 Figure 4-4. Recombinant ANXA7 undergoes salt-sensitive liquid-liquid phase separation in vitro, mediated by its low-complexity N-terminus .............................................. 106 Figure 4-5. FRAP analysis indicates a high percentage of mobile component within the ANXA7 droplets. ..................................................................................................... 107 Figure 4-6. Membrane interaction and vesicle clustering of ANXA7 ........................................ 108 Figure 4-S1. Disorder analysis of ANXA7 sequences ................................................................ 109 Figure 4-S2. Composition bias of ANXA7 N-terminal sequences ............................................. 110 Figure 4-S3. Single channel confocal fluorescence images for Figure 4-6A ............................. 111 Appendix Figure A-1. Full-length ANXA7 can form worm-like aggregates surrounding Httex1(Q39) puncta in live HEK293T cells ........................................... 134 Appendix Figure A-2. Full-length ANXA7eGFP forms granular structures surrounding Httex1 puncta in fixed HEK293T cells .................................................. 135 Appendix Figure A-3. ANXA7 accelerates Httex1 puncta formation in cells ........................... 136 Appendix Figure B-1. Workflow chart for CRISPR-Cas9 mediated cell genome editing ......... 139 Appendix Figure B-2. Western blot result for selective ANXA7-/- candidate clones ............... 140 Appendix Figure C-1. Clone 1 peptide immunostaining of HEK293T cells 48 hours post Httex1(Q39)mRFP or Httex1(Q72)mRFP transfection .......................... 143 Appendix Figure C-2. Clone 8 peptide immunostaining of HEK293T cells 48 hours post Httex1(Q39)mRFP or Httex1(Q72)mRFP transfection .......................... 144 Appendix Figure C-3. Clone 1 or clone 8 peptide immunostaining of HEK293T cells 48 hours post transfection with PBS, Httex1(Q39) or Httex1(Q72) ........... 145 Appendix Figure C-4. HEK293T cells 24 hour post Httex1(Q72)mRFP transfection ............... 146 Appendix Figure C-5. HEK293T cells 24 hour post Httex1(Q72)mRFP/dipeptide cotransfection .......................................................................................... 147 ix Abstract Membrane proteins participate in multiple cellular processes, fulfilling diverse functional roles. The focus of my thesis is to understand: 1) The interaction between the membrane proteins (protein-protein interaction); 2) The interaction between the membrane protein and the lipid membrane (protein-lipid interaction); 3) The interplay between the above two types of interaction. To achieve these goals, structural, biophysical and cellular studies were designed and performed on three model membrane proteins, with emphasis on different perspectives. Protein oligomerization is a critical step in many cellular processes including the transduction of signaling cascade, the assembly of functional macromolecules and the induction of membrane curvatures. Using site-directed spin labeling and EPR, we were able to investigate the driving force underlying this phenomenon by studying model membrane protein: annexin B12 (ANXB12). ANXB12 can form trimers upon binding to phospholipid-containing lipid membranes in the presence of Ca 2+ . Through systematic mutagenesis, we found that trimer formation requires the joint contributions from two different types of interaction: protein-protein interaction (other than the salt bridges) and specific Ca 2+ -mediated protein- membrane interaction at the trimer contact interface. Trimer disruption can be achieved by disrupting either of the two interactions. The study highlighted the fact that protein-lipid interaction, a force being largely neglected previously, could play a role as important as protein-protein interaction in driving on-membrane protein oligomerization. Many amyloid proteins are also membrane proteins and undergo a process known as membrane- mediated amyloid aggregation, which is linked to accelerated formation of toxic species. We next extended our study to amyloid proteins to see whether the same types of interaction persist as driving force and if so, would like to find ways to attenuate the process. Huntingtin exon 1 (Httex1) is the pathology-relevant protein fragment of Huntington’s disease. Previous study suggested that the presence of lipid membranes could accelerate the aggregation kinetics of this amyloid protein fragment. Using EPR, ODNP, CD and TEM, we managed to characterize its membrane-bound structure. The N-terminus of the protein (residues x 3-13) forms an amphipathic short α-helix with shallow membrane penetration and is sensitive to membrane curvature, lipid charges and ionic strength. The C-terminus of the protein (residue 23-92) is solvent exposed with no membrane-induced conformational change. The N-terminus and the C-terminus are linked by a transition region (14-22) that is no longer α-helical but undergoes membrane-mediated structural changes. On top of that, we found that phosphomimetic mutations could successfully attenuate both membrane interaction and membrane-mediated aggregation of the protein. Protein phase separation is a phenomenon recently being acknowledged to play a role in the amyloid formation as well as the biogenesis of membraneless organelles. We would like to explore how this special form of protein-protein interaction affects the behavior of membrane proteins. Annexin A7 (ANXA7) is another subfamily of annexin but unlike ANXB12, it has a long low complexity N-terminus. The role of its signature N-terminal domain was explored through cell studies combined with biochemical studies. We found that the N-terminal domain can mediate the recruitment of endogenous ANXA7 into stress granules as well as the self-association of protein when overexpressed. Interestingly, the domain can also promote protein phase separation in vitro and is indispensable for the normal function of the full-length protein in clustering lipid vesicles. The study revealed how the interplay and cooperation between a protein- protein interaction motif and a protein-lipid interaction motif could be related to a protein’s cellular function. Our work shows that membrane-protein interaction can induce conformational changes of the membrane protein (Httex1). It can drive the assembly of functional oligomer formation (ANXB12) as well as the membrane-mediated aggregation of amyloid protein (Httex1). Specific motif in the non-membrane binding region could mediate the protein-protein interaction between membrane proteins, leading to protein-protein phase separation (ANXA7). Collectively, these studies help elucidate the molecular mechanism of protein-protein interaction and protein-lipid interaction of membrane proteins. xi Preface Several research articles have been or will be published based on the work presented in this thesis. The work presented in Chapter 2 is reformatted from a research article published in Scientific Reports: Tao, M., Isas, J.M. & Langen, R., 2020. Annexin B12 Trimer Formation is Governed by a Network of Protein-Protein and Protein-Lipid Interactions. Sci Rep 10, 5301. The work presented in Chapter 3 is reformatted from a research article published in Structure: Tao, M., Pandey, N.K., Barnes, R., Han, S. & Langen, R., 2019. Structure of Membrane-Bound Huntingtin Exon 1 Reveals Membrane Interaction and Aggregation Mechanisms. Structure. 27(10): 1570-1580 e1574. The work presented in Chapter 4 is included in a manuscript currently under preparation: The N-terminus of Membrane Protein Annexin A7 is a Functional Domain Mediating Liquid-Liquid Phase Separation and Vesicle Clustering 1 Chapter 1: Introduction 1.1 Lipid Membranes and Membrane Proteins Lipid membranes are one of the most crucial components in biological systems. Structurally, plasma membranes define the boundaries that separate the intracellular environment from the extracellular environment, managing the selective permeabilization of substances into and out of the cells. Lipid membranes inside the cells subdivide the intracellular space further into smaller compartments (i.e. cell organelles), enabling biochemical events to proceed in a spatially and temporally organized manner. Additionally, lipid membranes provide the platform for important cellular processes (e.g. signal transduction and enzymatic reaction). The lipid composition of biological membranes varies drastically depending on the species, the tissue type, the cell type and the type of organelle (Figure 1-1A) (Harayama and Riezman, 2018; Leventis and Grinstein, 2010). The biophysical and biochemical properties of a certain membrane type is affected by the overall lipid constitution. An amphipathic lipid molecule is composed of two main parts: a hydrophilic head group and a hydrophobic tail group (Figure 1-1B). Both parts are of great diversity, giving rise to the many different categories of lipid molecules found in nature (Fahy et al., 2011). As shown in Figure 1-1C, the tail group could be straight/kinked, single/double acyl-chained and vary in length. The head group, on the other hand, could be as simple as a carboxylic group or as complex as to contain multi- ring sterol group. In an aqueous environment, lipids spontaneously assemble into membranes driven by energy minimization. And thanks to the great structural diversity of lipids, different lipid compositions can give rise to membranes of flat, negative or positive innate curvatures (Figure 1-1D) (Harayama and Riezman, 2018). The net electric charge of a lipid molecule is mainly determined by its hydrophilic head group (Poyry and Vattulainen, 2016). Neutrally charged lipid with zwitterionic head group is the most abundant category of lipids represented by phosphatidylcholine and phosphatidylethanolamine. Commonly identified negatively charged lipids include phosphatidylinositol, phosphatidylserine, phosphatidylglycerol and 2 Figure 1-1. Diversity in the distribution and structure of lipid molecules (A) Phospholipid content and cholesterol/PS molar ratio of different organelles. a Each phospholipid species is expressed as a percentage of total phospholipids. Abbreviations: PC, phosphatidylcholine; PE, phosphatidylethanolamine; PI, phosphatidylinositol; PS, phosphatidylserine; Chol, cholesterol. Table adapted from Leventis et al, 2010. (B) Space-filling model showing the basic parts of a lipid molecule illustrated with phosphatidylcholine. Figure adapted from Molecular Biology of the Cell. 4th edition. (C) The structural diversity of lipid molecules. Figure adapted from Fahy et al, 2011. (D) Spontaneous curvature generation by various lipid species. Figure adapted from Harayama et al, 2018. 3 cardiolipin. Generally speaking, positively charged lipids are rare component in naturally-derived membranes but they have wide applications in biological experiments, for example, artificial cationic liposomes are used commonly in cell transfection, mediating fusion with negatively-charged biomembranes. Although charged lipids constitute only a minor proportion of total lipids, they fulfill critical cellular functions. Cells can actively maintain an asymmetrical distribution of charged lipids in their inner and outer leaflet, which is important to membrane potential (Ma et al., 2017) and cellular processes including apoptosis and platelet coagulation (Ikeda et al., 2006). Furthermore, binding of cations and anions (in solution) to these charged lipids could largely neutralize the surface charge of the membrane, tuning the spontaneous curvature generation (Ali Doosti et al., 2017; Macdonald and Seelig, 1988). Since the first fluid mosaic model was proposed for plasma membranes (Singer and Nicolson, 1972), our knowledge on the lipid membranes have been significantly broadened thanks to the advancement of new techniques and the numerous scientific studies performed in the last few decades. Listed here are some of the interesting findings related to our studies (see Figure 1-2): (1) Lipid bilayers have different phases, namely solid-gel phase, liquid-ordered phase and liquid-disordered phase, with decreasing degree of order and increasing membrane fluidity (van Meer et al., 2008). (2) The organization of lipid molecules on the membrane is heterogeneous. Lipids can phase separate and coexist in more than one phase (Bernardino de la Serna et al., 2004). (3) In addition to rotational and lateral diffusion, lipid molecules can also undergo flip-flop across the membrane bilayers (Allhusen and Conboy, 2017). Unidirectional flip-flop events could be facilitated by translocases (Leventis and Grinstein, 2010). (4) The curvature and shape of lipid membranes in the cells are generated and maintained by specialized membrane curvature-inducing proteins (McMahon and Boucrot, 2015; McMahon and Gallop, 2005). 4 Figure 1-2. Highlights of some structural and dynamic properties of lipid bilayer (A) Three phase states of lipid bilayer. Figure adapted from van Meer et al., 2008. (B) Phase separation of lipid molecules on giant unilamellar vesicles made from pulmonary surfactant membranes with coexisting fluid order phase (red) and fluid disordered phase (green). Figure adapted from Bernardino de la Serna et al., 2004. (C) Translocation of phospholipids from the exoplasmic side of the membrane to the cytoplasmic side (flipping) and vice versa (flopping). Schematic illustration adapted from Allhusen and Conboy, 2017. (D) Membrane curvature is ubiquitously present in the cells. Schematic illustration adapted from McMahon and Boucrot, 2015. 5 Membrane proteins are proteins that reside on lipid membranes, well-situated to regulate the exchange of information and substances between cell/environment, cell/cell or organelle/organelle. Expectedly, membrane proteins have various cellular functions and can act as ion channels (Unwin, 1989), receptors (Cuatrecasas, 1974), enzymes (Coleman, 1973) or membrane curvature sensors (Antonny, 2011). Corresponding to their many functions, they are also diverse in structures and can be peripheral, transmembrane or lipid-anchored. Noticeably, due to their critical roles in various cellular processes, membrane protein is the largest class of drug targets (Arinaminpathy et al., 2009; Overington et al., 2006). Some of the well-known examples of druggable membrane proteins are: SERCA pump in cardiac disease (Tadini-Buoninsegni et al., 2018), HER-2 receptor in breast cancer (Ross et al., 2009) and NMDAR in neurological disorder (Paoletti et al., 2013). Therefore, it is of great therapeutic significance to characterize the structure of membrane proteins and to understand how they interact with the lipid membranes. However, because of their tight association with lipid membranes, membrane proteins are more challenging to study compared to most other proteins. Many membrane proteins contain hydrophobic portions to mediate their membrane interaction/insertion and sometimes, the correct folding of membrane protein itself depends on membrane interaction. For expressing in the recombinant systems, such as in E. Coli, a high cytoplasmic concentration will be reached. Protein-protein interaction between the hydrophobic domains is favored under high protein concentrations, leading to the accumulation of a high level of unstable misfolded proteins, promoting the formation of protein aggregates and lowing the yield for desired protein products. Apart from expression, purification of the membrane proteins could be problematic as well due to similar reasons. Upon lysis, membrane proteins tend to be unstable and aggregation-prone. The addition of detergents during purification could introduce additional artifacts. Rounds of extra optimizations are usually required in order to obtain a reasonable amount of membrane proteins to work with. Another major obstacle comes from structural determination, especially in characterizing the structure of membrane proteins in their native membrane-bound state, which is of great therapeutic interest. Commonly employed experimental approaches for studying membrane proteins include x-ray diffraction, cryo-EM as well as spectral tools like solution/solid-state NMR (Liang and Tamm, 2016) and EPR. These approaches each 6 have its advantages and disadvantages. In the next section, I will explain in more detail how EPR could be a useful tool in the study of membrane protein. The technique will be used intensively later in my studies. It is also worth noticing that due to the various difficulties in studying the membrane proteins experimentally, researchers often times seek alternatively to computational tools. 1.2 EPR and Its Application in Membrane Protein Structural Studies Electron Paramagnetic Resonance (EPR) is a spectroscopy technique that can detect the signals of unpaired electrons. The basic structure of a typical EPR spectrometer (Berliner, 2016) can be found in Figure 1-3A. It is composed of an external magnetic field, a microwave source as well as other electronic accessories for digital signal process and detection. EPR is an application of the Zeeman Effect (discovered in 1896 by Pieter Zeeman). An electron has two intrinsic spin states, corresponding to two different angular momentums (Ms = ± 1/2). In the absence of an external magnetic field, the two states are degenerated to be equal in energy and their distribution is isotropic. According to the Zeeman Effect (Figure 1-3B), when an external magnetic field (B 0) is applied on the electrons, the two spin states will split into two energy levels: align with B 0 (low energy state β e, Ms = -1/2) or against B 0 (high energy state α e, Ms = +1/2) (Brudvig, 1995). The energy gap between the two states increases linearly with the strength of the external magnetic field B 0 (𝛥𝛦 =𝑔𝜇 " 𝐵 $ ) (Bolton, 2007). The distribution between the two states is no longer isotropic but instead follows Boltzmann Distribution: 𝑛 % =𝑛 $ 𝑒 &( !" #$ ) . As we increase the strength of the external magnetic field, ΔΕ keeps increasing till its value matches the energy input from the microwave radiation (provided at a constant frequency ν). This is the very moment when an electron in low energy state can absorb the radiation energy and transition to the high energy state (Planck’s Law ΔE = hν). By measuring the amount of microwave radiation entered and left the sample chamber, the EPR spectrometer can detect the amount of microwave radiation absorbed by the sample and generate an EPR spectrum, which is the first derivative of the radiation absorption spectrum. By analyzing the many parameters of the EPR spectrum, we can extract useful information on the energy states of the unpaired electrons in the system. 7 Figure 1-3. Brief introduction of EPR (A) Schematic of an EPR spectrometer. Figure adapted from Berliner et al, 2016. (B) Under the application of an external magnetic field, the two spin states of electron will split into two energy states, a phenomenon known as the Zeeman Effect. Figure adapted and modified from Brudvig et al, 1995. αe: electron at high energy state βe: electron at low energy state n0: number of electrons at high energy state n1: number of electrons at low energy state k: Boltzmann constant T: Environment temperature Ms: Spin angular momentum (Quantum spin number) h: Planck’s constant ν: Frequency of microwave radiation g: g-factor μB: Bohr magneton of electron B0: Strength of external magnetic field 8 Many transition metal ions (Pilbrow, 1988) and organic free radicals (Spasojevic, 2011) contain unpaired electrons and are naturally paramagnetic. They are ideal and traditional samples used for the EPR studies. But most biological macromolecules of therapeutic interests do not contain any unpaired electrons. In order to “visualize” them in the EPR spectrometer, it is necessary to artificially introduce a spin label containing unpaired electrons. This approach somewhat resembles microscopy analysis, where we label biomolecules with dyes so as to enable observation under a light/fluorescence microscope. For protein molecules, the introduction of a spin label is usually achieved through a method known as site-directed spin labeling (SDSL) (Altenbach et al., 1989; Altenbach et al., 1990). Figure 1-4 gives an example of how to conjugate a protein molecule with a commonly used nitroxide spin label: MTSL (methanethiosulfonate spin label), which was originally synthesized in 1982 (Berliner et al., 1982). The first step of SDSL is to check whether there are any native cysteines in the amino acid sequence of the target protein. Because the MTSL conjugation is thiol-specific, the presence of any native cysteines would cause off-target labeling. Therefore, the first task is to replace any existing native cysteines with non-cysteine residues (usually an alanine or a serine). Next, the target amino acid of interest is replaced with cysteine (site-directed mutagenesis). The thiol group of the cysteine can react with the methanethiosulfonate (MTS) functional group of MTSL and form a disulfide bond. The protein thus labeled is usually referred to as the R1-labeled derivative. When a protein underwent conformational changes, bound to proteins/membranes or exposed to a different solvent environment, the conjugated R1 label can sense these changes and give rise to a new EPR spectrum. By analyzing various EPR parameters established by previous studies, we can obtain structural information on the target residue. When necessary, different residues can be systematically mapped by labeling one residue at a time with MTSL. Here we will focus on two of the most widely-used EPR parameters relevant to our studies: spin mobility and membrane accessibility, though other applications such as spin-spin distance measurement can also be achieved. And of course, we should always keep in mind that the application of EPR is not limited to in vitro studies. Nitroxide imaging probes such as methoxycarbamyl-proxyl (MCP) can be injected into live animals, where in vivo EPR imager can be used to track real-time physiological processes as the reduction half-life of these probes are affected by the 9 Figure 1-4. Site-directed spin labeling with MTSL Native cysteines, if present, are replaced in the first step. A target amino acid is then picked and converted into cysteine by site-directed mutagenesis. Next, the MTSL spin labels are allowed to react with the protein. The nucleophilic thiol group of the introduced cysteine can specifically react with the electrophilic sulfur in the methanethiosulfonate group of the MTSL spin label (red arrow). The nitroxide spin functional group is thus conjugated with the protein through a disulfide bond (highlighted in red). The spin-labeled protein thus derived is commonly referred to as the R1-labeled derivative. 10 onset of diseases or therapeutic treatments (Cheng et al., 2019; Emoto et al., 2019; Epel et al., 2014). Figure 1-5A shows the EPR spectrum of free MTSL spin labels, which has three sharp peaks. From the chemical structure of MTSL shown in Figure 1-4, we know that it contains only one unpaired electron, so why three peaks? This is due to a phenomenon known as hyperfine splitting resulted from electron- nucleus interactions. Basically, the unpaired electron of MTSL is under the influence of the nuclei spin coming from its neighboring 14 N (M I = +1, 0 and -1). Thus, each of the two electron spin states (MS = ±1/2) will further split into three energy states, giving rise to a total of six energy states and three different transition energies ΔE (Figure 1-5B). The three transition energies correspond to three absorption events as the strength of the external magnetic field increases. The absorption peaks are later converted into its first derivative (due to signal-to-noise considerations) and hence the three sharp peaks in the final EPR spectrum (Figure 1-5A). Because the add-on ΔΕ contributing from the neighboring 14 N follows the order of +1 > 0 > -1 and the EPR monitors the absorption of electrons, electrons under the influence of +1 14 N would reach the resonant moment first, followed by those under the influence of 0 and -1 14 N. In other words, under the influence of the nucleus magnetic field, the electrons with ν 1 frequency (Figure 1-5B) possess the highest starting energy, thus require the least energy input from microwaves to reach the threshold requirement of resonance energy. 1.2.1 Spin mobility Spin mobility is an EPR spectrum parameter used to describe the dynamics of the spin label. A decrease in spin mobility is reflected in the lineshape of the EPR spectrum as line broadenings (Figure 1- 5C). The slower the motion of the spin label, the longer the rotational correlation time (τ) and the broader the EPR spectrum. It can be quantified by the inverse second moment <H -2 > or the inverse of the central line width ΔΗ 0 -1 . The line width (ΔΗ) is defined as the distance between the highest point and the lowest point for each peak (see Figure 1-5A, red). When the two EPR spectra measured at the same parameter are normalized to the same double-integrated signal intensity, i.e. for them to have the same total area under the absorption spectrum (I), spectrum with decreased spin mobility will have lower amplitude because I~AΔH ) (Klug and Feix, 2008). Therefore, in practice, the central line amplitude can alternatively be used 11 Figure 1-5. EPR spin mobility and its applications (A) The nucleus of the nearby nitrogen atom can interact with the unpaired electron and cause hyperfine splitting into six states, corresponding to three possible transition energies. (B) Hyperfine splitting leads to three possible electron transition frequencies ν 1, ν 2, ν 3, corresponding to decreasing ΔE. Figures adapted and modified from B. G. Hegde, 2015. (C) EPR line broadening shows a decrease in spin mobility and an increase in rotational correlation time. Figure adapted from Fanucci et al, Encyclopedia of Biophysics. (D) Study on T4 lysozyme suggests that R1 side chains in a given structural class are clustered together in mobility, suggesting a correlation between spin mobility parameters and protein secondary structures. Figure adapted from McHaourab et al., 1996. 12 to compare the relative spin mobility between spectra. The spin mobility reported from the EPR spectra is a superposition outcome from three different sources (Sahu et al., 2013). The first contribution comes from the tumbling and rotation of the protein/peptide on which the spin label is attached to. This is especially a concern when the protein or peptide has a small molecular weight (< 15kDa) and thus a relatively fast tumbling rate (Klug and Feix, 2008). To mute this contribution from the others, viscous solutes such as sucrose and glycerol can be added to slow down the tumbling (McHaourab et al., 1997). The second and third contribution comes from the fluctuation of the α-carbon backbone and the mobility of the label relative to the peptide backbone. These two contributions provide the basis for EPR to be used routinely as a tool to determine the protein backbone dynamics and the existence of tertiary contacts. The foundations of the theory are derived from a model protein: T4 lysozyme, for which crystal structure was available and systematic studies were conducted to characterize the relationship between protein structure and side-chain mobility of spin labels (McHaourab et al., 1996). As shown in Figure 1-5D, there is a clear correlation between the two. Residues located in the loop regions are fully exposed to the solvent and have the highest degree of freedom and backbone flexibility, giving rise to the highest spin mobility. For residues located in the solvent-accessible surface of the helix, even though their side chains have high degree of freedom, they are restricted by ordered fluctuation of the backbone within the helix and are second in spin mobility. Residues having tertiary contacts with neighboring side chains or backbone atoms are further restricted in their side chain rotation, but not as restricted as the fully buried residues. It is worth noticing that tertiary contacts are not necessarily limited to proteins, binding to ligand compounds or lipid membranes could also result in restricted side chain mobility of the attached spin labels, leading to broadened EPR spectra. For example, in Chapter 3, we observed a shift in spin mobility of a membrane protein upon its binding to lipid membranes. On the other hand, spin-spin coupling could also lead to decrease in spin mobility. The two spins could be contributed from the same protein at different sites or from two different protein molecules. In Chapter 2, we studied the multimerization behavior of a membrane protein by a method established based on restricted spin mobility due to spin-spin coupling. 13 1.2.2 Membrane accessibility Membrane accessibility is another EPR parameter that evaluates how deep a spin label is buried inside the lipid membranes. Through analyzing the accessibility of a spin label to the membrane relative to its aqueous environment, we can estimate the location of the residue attached to the label. A collection of information on different residues will help us decipher the topology of membrane proteins and better understand membrane-protein interaction. The measurement of this parameter is based on a phenomenon known as power saturation. As discussed briefly before, the generation of EPR signal requires the excitation of electrons from the low energy spin state to the high energy spin state at resonant microwave frequency. At thermal equilibrium, the two spin states follow Boltzmann distribution, with a slight excess population of electrons in low energy spin state. At the very beginning, the excitation process is the rate-limiting step due to insufficient supply of microwave power. Therefore, as we gradually increase the microwave power from low to high, more electrons in the low energy state can absorb energy from the microwave and get excited to the high energy state (Figure 1-6A). The EPR signal increases and is proportional to the square root of the power. Simultaneously, mediated by the spin-lattice relaxation, the excited electrons are returning back to the low energy state. This replenishes the pool of electrons in low energy state, which can be again excited. But because the total number of electrons is limited, eventually, excitation and relaxation will reach a dynamic equilibrium at which no further increase in the resonance absorption events can be obtained through increasing the power. At this point, the EPR signal plateaus (i.e. reaches saturation). Collisions to fast-relaxing paramagnetic species (colliders) in the solution can increase the rate of spin-lattice relaxation. Therefore, power saturation can be delayed by adding colliders. That is, the saturation will now be reached at a higher level of microwave power. O 2 and NiEDDA are two commonly- used colliders. As shown in Figure 1-6B (left panel), a shift in the saturation curve to the right will be observed in the presence of O 2 or NiEDDA, corresponding to a larger P 1/2. The degree of shifting is proportional to the concentration of the colliders. Noticeably, O 2 is nonpolar and preferentially partitions into the lipid membranes while NiEDDA is polar and preferentially partitions into the aqueous 14 Figure 1-6. EPR power saturation and its applications (A) At thermal equilibrium under the application of an external magnetic field, the distribution of the electrons in two energy states follows Boltzmann distribution. The low energy species can absorb microwave energy and get excited to high energy states till saturation is reached. Relaxation down to low energy state is required for absorption to occur again. (B) The presence of colliders can increase the relaxation rate and shift the saturation curve to the right, giving rise to higher half maximum P values (P 1/2). O 2 and NiEDDA are two common colliders, with preferential partition into nonpolar and polar environment, respectively. (C) The periodicity between O 2 and NiEDDA Pi values can help identify conformations of membrane proteins. Figure adapted from Hubbell et al, 1998. 15 environment. An inversely correlated gradient of the two colliders can be achieved at the membrane-water interface (Figure 1-6B, right panel). We can then perform the power saturation experiment on spin-labeled membrane proteins under conditions where we have only O 2 or NiEDDA present as colliders during an individual measurement. Their respective P 1/2 values can be recorded and converted into accessibility parameters: ΠO 2 and ΠNiEDDA. By analyzing the periodicities of the two values for a number of selected singly R1-labeled derivatives, it is possible to characterize the different conformation of membrane proteins (Hubbell et al., 1998). Figure 1- 6C illustrated the method by selected residues within various helices. As shown, a helix fully exposed to water will have a constantly high ΠNiEDDA and low ΠO 2. A transmembrane helix, on the other hand, will have high ΠO 2/low ΠNiEDDA in the helix center and low ΠO 2/ high ΠNiEDDA at the termini of the helix. For a helix running in parallel along the membrane surface and peripherally bound to a single leaflet, an out-of-phase periodicity is expected. For practical convenience, ΠO 2 is usually measured in the presence of air. And ΠNiEDDA is measured under a constant flow of N 2 and later deducted by values obtained under N 2 flow but in the absence of NiEDDA. To estimate the immersion depth of a membrane protein, ΠO 2 and ΠNiEDDA can be further integrated into an immersion depth parameter (Φ), which is defined as Φ = In (ΠΟ 2/ ΠNiEDDA). It is a measure of the relative membrane accessibility versus water accessibility. The higher the Φ values, the deeper the spin label is embedded in the lipid membrane and the less accessible it is to the aqueous environment. Using commercially available lipids spin-labeled at known depth on the aliphatic chains, we can perform the same power saturation experiment using label-free proteins and plot a standard curve which gives the correlation between the Φ value and the distance away from the lipid headgroup. This way, Φ values derived from different R1-labeled protein derivatives can be converted directly into membrane immersion depth distances (Altenbach et al., 1994). In Chapter 3, we employed the membrane accessibility parameter to characterize the immersion depth of a peripherally bound membrane protein. 16 1.3 Amyloid Protein and Its Membrane Interaction 1.3.1 Introduction to amyloid proteins Amyloid proteins were originally discovered from pathological studies on a number of diseases including Alzheimer’s disease, Parkinson’s disease and diabetes. Within those patient tissue samples, starch-like proteinaceous deposits were observed in the form of tangles or plaques (Iadanza et al., 2018; Rocken and Sletten, 2003). Although these highly insoluble deposits are heterogenous in nature and contain polysaccharides, lipids, calcium, cell debris and so on, they are more than just amorphous aggregates. Amyloid proteins were detected in and isolated from these deposits: proteins able to self-assemble into insoluble ordered species with fibrillar morphologies. Figure 1-7A presents four different types of fibrils examined under the electron microscope (Gharibyan et al., 2020; Guerrero-Ferreira et al., 2018; Isas et al., 2017; Luhrs et al., 2005). As shown, there are morphological diversities between different fibril types. Traditionally, amyloid proteins are challenging subjects to study for structural biologists due to their intrinsic properties: thermodynamically unstable and highly prone to aggregation. They also undergo dynamic structural changes in response to different preparation conditions (e.g. pH, salt and temperature) as well as the presence of other proteins. Fibrils, the end product of amyloid protein aggregation, initially attracted the most attention. This is not only because they were typically observed in vivo and reproduced in vitro but also because they were deemed as the culprit for disease progression. Using a number of advanced biophysical tools including X-ray diffraction, solid-state NMR, EPR and Cryo-EM, more and more fibril structures have been successfully depicted on the molecular level. It is widely-recognized that amyloid macromolecular assemblies share a generic conformation known as the cross-β-sheet (Geddes et al., 1968). Basically, β-strands running perpendicular to the axis of the fibril are first packed through backbone hydrogen bonds to form pleated β-sheet. The β-sheets are further packed laterally with neighboring β-sheets mediated by side chain interactions (Sabate and Ventura, 2013). As illustrated in Figure 1-7B, X-ray diffraction experiments on fibrillar crystals can identify this type of supersecondary structure: two major signature diffraction patterns can be observed respectively at ~4.7 Å and ~10 Å, corresponding to the inter β-strand spacing and inter β-sheet spacing (Gallardo et al., 2019). 17 Figure 1-7. Amyloid proteins and their fibrillar structures (A) Electron microscopy images of fibrils formed from (a) amyloid β (1-42), figure adapted from Luhrs et al, 2005. (b) islet amyloid polypeptide precursor (IAPP), figure adapted from Gharibyan et al, 2020. (c) α-synuclein (αS), figure adapted from Guerrero-Ferreira et al, 2018. (d) huntingtin exon 1 protein fragment with 46 glutamines (Httex1Q46), figure adapted from Isas et al, 2017. (B) The X-ray diffraction pattern for amyloid shows major reflections at ∼4.7 Å (hydrogen bonding distances between β-strands) and ∼10 Å (side-chain packing between β-sheets) indicating cross-β structure where β-strands align perpendicular to the fibril axis. Graphical models are shown for a few types of tau and αS fibrils. Figure adapted from Chuang et al, 2018. (C) Varieties of cross-β structures in amyloid fibrils. (a) in-register parallel cross-β structure. (b) antiparallel cross-β structure. (c) double-layered in-register parallel structure. (d) double-layered antiparallel cross-β structure. (e) double-layered antiparallel β-hairpin structure. Figure adapted from Tycko et al, 2015. 18 Certain dyes such as Congo Red and Thioflavin T have high binding affinity for cross-β-sheet conformation and are used widely as amyloid-specific probes (Biancalana and Koide, 2010; Stopa et al., 2003). However, we need to be cautious when making definitive conclusions based solely on these dyes since potential limitations and non-specific staining have also been reported (Cooper, 1969; Khurana et al., 2001). Apart from this common architecture of cross-β-sheet, increasing evidence is pointing towards a polymorphism in the fibril packing (Fandrich et al., 2018; Gallardo et al., 2019; Tycko, 2015). For example, the orientation of the fibril alignment could be parallel or antiparallel, staggered or non-staggered and the cross-β interactions could be intra- or intermolecular, etc. Figure 1-7B and 1-7C present some examples on the different varieties of existing cross-β structures. Astonishingly, depending on the preparation procedure and the different protein fragments used, different fibril structures can even arise from the same amyloid protein. So how are these highly ordered fibrillar macromolecules built-up? Figure 1-8A provides a schematic illustration for fibril assembly through the nucleation growth mechanism (Iadanza et al., 2018). According to this proposed mechanism, the basic building blocks are monomeric amyloid proteins, which are either intrinsically or unfolded to be partially/fully disordered. They first form dynamic heterogenous oligomers, which can then seed the growth and elongation of fibrils. Fibril formation is commonly-accepted to be entropy-driven (Kardos et al., 2004). Because the process is energetically favorable, fibril formation is not readily reversible although depolymerization of fibrils can happen (Yagi et al., 2013). Interestingly, patterned arrangement of selected types of amino acids, for example: aromatic and charged residues, are often times found in amyloid proteins (Gazit, 2002; Marshall et al., 2011). These amino acids could enhance inter β-sheet interactions through side chain aromatic stacking or electrostatic interactions, therefore facilitating the assembly and strengthening the stability of amyloid fibrils. Increasing numbers of new studies have shifted their focus to deciphering the structure of amyloid proteins in non-fibrillar state (e.g. monomeric or oligomeric states). These states are comparatively unstable and transient and therefore more challenging to capture, but they could serve as better therapeutic targets. 19 Figure 1-8. Nucleation growth mechanism of amyloid formation Schematic illustration for multi-step amyloid formation through the nucleation growth mechanism. Native proteins are in dynamic equilibrium with their less-structured, partially folded and/or unfolded states. One (or possibly several) of these states can be assembled into oligomeric species, higher-order oligomers and ultimately a fibril nucleus. This process occurs in the lag time (nucleation phase) of assembly. Fibrils grow by recruiting unfolded monomers into the fibril nucleus. They also fragment, yielding more fibril ends that are capable of elongation. This process occurs as an exponential growth of fibril until nearly all free monomer is depleted. Fibrils can also associate further with each other, with other proteins and factors to form the amyloid plaques and intracellular inclusions. Figure adapted from Iadanza et al, 2018. 20 Existing hypothesis of pathogenic mechanism argued that the soluble monomeric protein fulfills normal function in the cell and a loss-of-function upon misfolding/aggregation could lead to toxicity (Chung et al., 2018; Winklhofer et al., 2008). Thus, therapeutics that act through stabilizing monomers retain the protein’s normal function and effectively inhibit toxicity. To achieve that goal, a well- characterized monomeric structure is required. Additionally, many studies have shown that the oligomeric species could also be toxic (Cremades et al., 2012; Klein et al., 2004; Walsh et al., 2002). Therefore, high- resolution oligomeric structures could serve as potential drug targets as well and allow arresting of the disease progression in its early stage before fibrils or amyloid deposits are abundantly formed. 1.3.2 Membrane interaction and membrane-mediated aggregation of amyloid proteins Many of the well-known amyloid proteins are also membrane proteins. For example, Aβ is a transmembrane protein while as IAPP, αS, and tau are all peripheral membrane proteins. As illustrated in Figure 1-9A, pathogenic proteins have multiple interacting partners on the lipid membrane including specific lipid molecules, other membrane proteins and extracellular matrix components. Here, we focus on protein-membrane interactions through direct contacts with the lipid molecules, which has important physiological as well as pathological consequences (Shrivastava et al., 2017). From the positive perspective, membrane interaction may contribute to the normal physiological role of the nonpathogenic/functional variant of amyloid protein. Lipid membranes provide a 2D platform so that the protein can network with other membrane proteins efficiently. Furthermore, membrane interaction can trigger conformational changes in a number of amyloid proteins (Davidson et al., 1998; Jayasinghe and Langen, 2005; Terzi et al., 1997). As stated earlier in Section 1.1, lipid membrane is composed of a high diversity of lipid molecules, including charged lipids. Many amyloid proteins, on the other hand, contain intrinsically disordered domains that lack defined structure or exist as an equilibrium of interconvertible conformations of similar energy levels in the aqueous solution. Lipid molecules can induce and stabilize a defined conformation that can be distinguished from its original conformation in solution through electrostatic or hydrophobic interaction with a number of particular residues within these intrinsically disordered regions. Post translational modifications of these residues made it possible to 21 Figure 1-9. Membrane interaction and membrane-mediated aggregation of amyloid proteins (A) Binding partners of pathogenic proteins on the membrane. Pathogenic proteins may bind to various biomolecules on the plasma membrane including lipids (1), cholesterol (2), receptors and channels (3), and ECM components, including glycoproteins (4) or carbohydrate chains (5). Figure adapted from Shrivastava et al, 2017. (B) Schematic of membrane-mediated aggregation of amyloid proteins. Native IAPP or α-synuclein molecules bind to the membrane and adopt helical conformation. The reduced dimensionality on the membrane surface increases local protein concentration which facilitates intermolecular interaction through their respective amyloidogenic region, promoting the formation of oligomers and fibrils. Red cylinder: helical region; Double-headed arrow: intermolecular interaction in the amyloidogenic region. The C-terminal of α-synuclein is not depicted. Figure adapted from Rawat et al, 2018. 22 regulate membrane interaction of those proteins. Additionally, membrane interaction can change the orientation of protein-protein interaction. Part of the protein could be buried/partially buried in the membrane, thus less accessible to other interacting partners in solution or on membranes. These membrane- induced differences could have switch-like roles in regulating the normal function of the nonpathogenic form of the amyloid protein. From the negative perspective, however, membrane interaction could be contributing to the intracellular/intercellular propagation and spreading of various toxic amyloid species. Additionally, membrane interaction of amyloid proteins in their monomeric, oligomeric or fibrillar state have been reported to disrupt the integrity of lipid membranes (Bode et al., 2019; Brender et al., 2008; Milanesi et al., 2012), potentially leading to cytotoxicity by disrupting the plasma or organelle membranes. And on top of these, membrane interaction of amyloid proteins bearing pathogenic mutations/variations could increase the complexity of the disease by providing an alternative toxicity pathway. This is achieved through a phenomenon known as membrane-mediated aggregation (Figure 1-9B). Studies have revealed that membrane interaction can facilitate the aggregation and promote the formation of oligomers/fibrils for a number of amyloid proteins including αS, Aβ and tau (Chi et al., 2008; Jones et al., 2012; Zhu et al., 2003). Protein aggregation on membrane-water interface is generally favored over in solution. Even though the biophysical mechanism for membrane-mediated aggregation is still under debate, there are a few widely-accepted explanations. First of all, membrane interaction can increase the local concentration of proteins, which will enhance the incidence of protein-protein interaction and oligomerization. Additionally, membrane interaction decreases the translational diffusion of proteins, bringing down the dimensionality of protein-protein interaction from 3D to 2D. This reduction in the degree of freedom further increases the incidence of protein-protein collision (Jayasinghe and Langen, 2007; Rawat et al., 2018). Compared to the aqueous environment, lipid membrane has a low dielectric constant, favoring hydrophobic interactions between proteins and promoting the self-association of membrane-bound monomers into oligomers, which ultimately nucleate and elongate into amyloid fibrils. To better understand the process of membrane-mediated aggregation of amyloid proteins, it would 23 be helpful to depict the protein structure in its membrane-bound state. Membranes are ubiquitously present in a biological system, thus the membrane-bound state of an amyloid protein is physiologically-relevant and has great potential for the development of future therapeutic treatment. In Chapter 3, we will depict the membrane-bound structure of an amyloid protein, huntingtin exon 1 fragment and investigate how membrane interaction mediates the aggregation of its pathogenic variant. 1.4 Protein Phase Separation 1.4.1. Introduction to protein liquid-liquid phase separation (LLPS) Liquid-liquid phase separation is a phenomenon intensively-studied in soft matter physics (Khabibullaev, 2013). An everyday example of LLPS is the oil-in-water emulsion system. Phase separation of lipids can be visualized under fluorescent microscope as shown earlier in this chapter (Figure 1-2B). The phenomenon is not limited to lipids but apply to all multi-component solution systems, which are metastable when supersaturated and will spontaneously demix into thermodynamically more stable co- existing phases (Myerson, 1993). The process is illustrated in Figure 1-10A with a two-component system containing solute A and solvent B. In the phase diagram, any points on the borderline of one-phase and two-phase region is known to be marginally stable. Under that state, even a small shift in the environmental factors (in this case the temperature and the concentration of solute A) could trigger a phase separation event. A new phase in the form of liquid droplets significantly enriched in solute A would separate from the continuous phase B, in which residual low amount of solute A remains. The new phase here is generally referred to as the solute-rich phase and the continuous phase is referred to as the solute-poor phase (Style, 2018). In the biological system, LLPS is also ubiquitously present and fulfill various functional rules including mitosis (Tiwary and Zheng, 2019), embryonic development (Tatavosian et al., 2019) and protein quality control (Frottin et al., 2019). Recently, the ability for proteins to undergo LLPS has attracted much attention and revolutionized our traditional understanding of how the cellular activities are orchestrated spatially and temporally (Boeynaems et al., 2018). The most well-recognized biological application of phase separation is the assembly of subcellular structures such as the spindle apparatus and the centrosomes 24 Figure 1-10. Liquid-liquid phase separation and its cellular roles (A) Schematic diagram of simple liquid-liquid phase separation in a two-component system with an upper-critical solution temperature. Point (i) on the phase diagram indicates a marginally stable mixture where the majority component B is saturated with a dilute component A. Upon cooling to (ii), the system spontaneously separates into two phases, a continuous phase (iii) with a low concentration of A and a droplet phase rich in A (iii’). Figure adapted from Style et al, 2018. LLPS can regulate the biogenesis of membraneless organelles. (B) Multiphase nucleoli after actin disruption in X. laevis. Figure adapted from Feric et al, 2016. (C) Stress granule assembly (G3BP1- eGFP) during NaAsO2 stress. Figure reproduced from the video in Wheeler et al, 2016. LLPS can function in the nucleus as a protein quality control. (D) Model of nucleolar protein quality control for proteins entering the nucleolus from the nucleoplasm. Figure adapted from Frottin et al, 2019. 25 (Tiwary and Zheng, 2019) as well as the formation of many membraneless organelles (Mitrea and Kriwacki, 2016). These organelles include cytoplasmic bodies [e.g. stress granules (Wheeler et al., 2016) and P- bodies (Luo et al., 2018)] and nuclear bodies (Zhu and Brangwynne, 2015) (e.g. PML bodies, Cajal bodies and the nucleoli). Representative images showing multiphase nucleoli (Feric et al., 2016) and assembled stress granules (Wheeler et al., 2016) can be found in Figure 1-10B and 1-10C. The fluid diffusion within the protein droplets is significantly retarded and slowed down due to the crowding effect and the variety of enclosed molecular components are rather restricted. As a result, membraneless organelles provide the cell with an alternative option for subcompartmentation, where molecular collisions and biochemical reactions can take place without membrane borders. They are ideal sites for temporary storage of misfolded protein, where proper refolding could be mediated by chaperone proteins also recruited into the droplets (Figure 1-10D) (Frottin et al., 2019). But different from their membrane-encompassed counterparts, these “droplet organelles” are more dynamic, undergoing more frequent fusion and fission and exchanging substances more readily with their surroundings. Recall that for membrane-enclosed organelles, the selective cross-membrane transportation usually requires membrane proteins and is energetically costly. In contrast, phase separation event, once triggered, can immediately withdraw large amounts of substances out of the local environment. Therefore, it is highly efficient as a mechanism in depleting unwanted substances and maintaining relative local homeostasis. This mechanism has application in the control of protein expression noise (Klosin et al., 2020) as well as the formation of stress granules (Wheeler et al., 2016) (Figure 1-10C). The formation of protein droplets is highly sensitive to micro-environmental factors such as the protein concentration, pH, ionic strength, post-translational modification and temperature. Thus, LLPS can function as a bio-switch, allowing cells to act promptly in response to minor changes in the surroundings. Additionally, there is growing evidence that the occurrence of LLPS is sequence-dependent: Intrinsically disordered regions with low sequence complexities have a high tendency to undergo LLPS (Martin and Mittag, 2018; Statt et al., 2020). In general, there are a few major forces that are known to play crucial roles in the formation of 26 protein droplets including the hydrophobic interactions, hydrogen bonding and π-π stacking. Studies employing in vitro systems suggested that the formation of disordered droplets are primarily driven by hydrophobic interactions while as hydrogen bonding and π-π stacking are forces mediating the maturation and evolution of these droplets into fibrillar structures (Yuan et al., 2019a). One category of low-complexity sequence with high tendency to phase separate is enriched in both positively charged residues (Arg, Lys) and aromatic residues (Trp, Tyr, Phe). The two types of residues can form cation-π interactions, promoting the self-association of proteins into droplets. Interestingly, compared to Phe and Tyr, Trp has been proposed to mediate stronger cation-pi interactions due to its double-ring structure. And compared to Lys, Arg can drive stronger interaction with Tyr and Phe because its guanidine group is a Y-shaped planar cross-conjugated pi system. The phase separation behavior of this category can be regulated by Arg/Lys acetylation. FUS and hnRNPA2 are two representative proteins that belong to this category (Qamar et al., 2018; Ryan et al., 2018). Another category of low-complexity sequence with high droplet-forming tendency, however, lacks charged residues. Droplet formation is therefore likely to be driven by π-π stacking between aromatic residues or by hydrophobic interactions. Other categories may also exist. Our current understanding and knowledge in this new field is relatively limited and further exploration is required to unveil the mystery. As a step forward to figure out the direct correlation between the amino acid sequence and the phase separation, several databases have been established for proteins undergoing LLPS including LLPSDB (Li et al., 2020), PhaSePro (Meszaros et al., 2020) and PhaSepDB (You et al., 2020). In Chapter 4, we explored a low-complexity protein sequence that lacks charged residues and analyzed its capability to undergo LLPS. So far, we mainly focused on the functional role of protein droplet formation. However, there’s a high cost that comes with their increased switching sensitivity and dynamic properties. These droplets do not have lipid molecules as the outer most layer. According to the Oswald step rule, the outer most layer of protein molecules are thermodynamically least stable and have a tendency to crystalize out first. In fact, the aberrant protein droplet formation has been implicated in many pathological conditions and diseases, which will be discussed with more details in the next section. 27 1.4.2. Protein phase separation and amyloid fibril formation Though seems manifestly, it is tempting to overlook the fact that amyloid fibril formation is also a form of protein phase separation (liquid-to-solid). During the process of amyloid fibril formation, we start with a system of liquid phase in which amyloid protein is fully dissolved in aqueous solution. The protein then gradually transitions into insoluble aggregates that separate from the liquid surroundings, resulting in a two-phase system (solid phase demixed from the liquid phase). Abnormal LLPS of amyloid proteins was found to be an early step preceding their irreversible aggregation into fibrils as has been observed for a number of amyloid proteins including FUS and tau (Ambadipudi et al., 2017; Patel et al., 2015) as shown in Figure 1-11A and 1-11B. Liquid droplets of amyloid protein could be an intermediate species in the pathway of amyloid fibril formation. The droplets are highly dynamic and mobile when freshly formed. But liquid diffusion within the droplet is still slowed down compared to the aqueous surroundings. Computational simulation studies suggest that slowing down of diffusion in crowded solution could be correlated with the formation of transient clusters (Nawrocki et al., 2017). Indeed, these protein droplets of amyloid protein can later mature into less mobile gel phase/glass phase, which can be further converted into solid state as amyloid fibrils. In a traditional fibril formation process of liquid-to-solid transition, the energy barrier to form nuclei directly from a homologous solution is high. However, LLPS provides an alternative pathway with decreased energy barrier by introducing a metastable liquid phase as illustrated by LLPS-mediated nucleation-elongation nanofibril formation mechanism shown in Figure 1-11C (Yuan et al., 2019a). Prior studies focusing on inhibiting amyloid fibril formation are largely limited in targeting monomer to oligomer or oligomer to fibril transitions. However, these therapeutic measures will not be as effective if amyloid protein can get around the solution nucleation step by taking the alternative pathway of droplet formation. Therefore, a deeper understanding of the relationship between LLPS and amyloid fibril formation could provide new strategic rationales in the disease therapeutic in arresting the fibril maturation. 28 Figure 1-11. LLPS-mediated fibril formation Aberrant LLPS has been implicated to underlie the formation of irreversible amyloid aggregates. (A) Liquid-like droplets of FUS gradually convert into solid-like structures. Figure adapted from Patel et al, 2015. (B) Temperature-sensitive tau droplets formation. Figure adapted from Ambadipudi et al, 2017. (C) LLPS-mediated nucleation-elongation mechanism of fibril formation. Figure adapted from Yuan et al, 2019. 29 Chapter 2: Annexin B12 Trimer Formation is Governed by a Network of Protein-Protein and Protein-Lipid Interactions Abstract Membrane protein oligomerization mediates a wide range of biological events including signal transduction, viral infection and membrane curvature induction. However, the relative contributions of protein-protein and protein-membrane interactions to protein oligomerization remain poorly understood. Here, we used the Ca 2+ -dependent membrane-binding protein ANXB12 as a model system to determine the relative contributions of protein-protein and protein-membrane interactions toward trimer formation. Using an EPR-based detection method, we find that some protein-protein interactions are essential for trimer formation. Surprisingly, these interactions are largely hydrophobic, and they do not include the previously identified salt bridges, which are less important. Interfering with membrane interaction by mutating selected Ca 2+ -ligands or by introducing Lys residues in the membrane-binding loops had variable, strongly position- dependent effects on trimer formation. The strongest effect was observed for the E226Q/E105Q mutant, which almost fully abolished trimer formation without preventing membrane interaction. These results indicate that lipids engage in specific, trimer-stabilizing interactions that go beyond simply providing a concentration-enhancing surface. The finding that protein-membrane interactions are just as important as protein-protein interactions in ANXB12 trimer formation raises the possibility that the formation of specific lipid contacts could be a more widely used driving force for membrane-mediated oligomerization of proteins in general. 2.1 Introduction A plethora of biological events, including signal transduction (Diaz et al., 2016; Janosi et al., 2012; Pawson and Scott, 1997; Simons and Toomre, 2000), viral infection (Dick et al., 2013; Panchal et al., 2003) and membrane curvature induction (McMahon and Boucrot, 2015; Zimmerberg and Kozlov, 2006) rely on the ability of peripheral membrane proteins to oligomerize on membranes. In addition to physiologically relevant oligomerization of peripheral membrane proteins, it has also been suggested that membranes can 30 promote the pathological oligomerization of amyloidogenic proteins (Bucciantini and Cecchi, 2010; Gorbenko and Kinnunen, 2006; Rawat et al., 2018). Oligomerization requires specific protein-protein contacts to be made in a process generally known to be facilitated by shape complementarity and hydrophobic as well as electrostatic interactions at the protein contact interface (Chanphai, 2015; Li et al., 2013b; Sheinerman et al., 2000). Membranes could further facilitate the formation of such protein-protein contacts in a variety of ways. First of all, membranes provide a common interaction surface that increases the local protein concentration and that reduces the dimensionality of protein diffusion from 3D to 2D. Both mechanisms have been frequently invoked in order to explain the membrane-mediated aggregation of amyloidogenic proteins (Burke et al., 2013b; Jayasinghe and Langen, 2005, 2007; Rawat et al., 2018; Shrivastava et al., 2017). Beyond such a simple surface effect, membranes could also act by more specific interactions, where membranes become a more integral part of the complexes or where modulation of membrane thickness promotes oligomerization. Support has been obtained for transmembrane proteins (Kahraman et al., 2016b; Stangl and Schneider, 2015) and computational studies suggest that similar mechanisms could also apply to peripheral membrane proteins (Kahraman et al., 2018; Morozova et al., 2011). However, little experimental support has been obtained thus far in favor of the notion that thickness deformation or other lipid-binding related factors can specifically promote the oligomerization of peripheral membrane proteins. Here we sought to develop an experimental model system that makes it possible to dissect the effects of protein-protein and protein-lipid interaction on the oligomerization of peripheral membrane proteins. Toward this end, we used trimer-forming annexin B12 (ANXB12) as a model system. ANXB12 belongs to the annexin superfamily of proteins, which are characterized by reversible, Ca 2+ -dependent binding to negatively charged membranes that typically contain phosphatidylserine (Lizarbe et al., 2013; Moss and Morgan, 2004; Swairjo et al., 1995). Annexins have a number of membrane-related functions, including membrane trafficking (Creutz, 1992; Zaks and Creutz, 1990), membrane repair (Boye and Nylandsted, 2016; Draeger et al., 2011; Lauritzen et al., 2015), calcium signaling (Gerke et al., 2005; Hawkins et al., 2000) and ion channel formation (Berendes et al., 1993; Cohen et al., 1995; Isas et al., 2000). 31 The annexin superfamily of proteins is highly conserved and annexins are typically monomeric in solution, yet some annexins form trimers upon Ca 2+ -mediated membrane binding, while others do not. Examples of trimer-forming annexins are annexin A5 (ANXA5) and ANXB12. In vitro experiments found that trimer formation is relatively rapid (sub-second) and trimers can subsequently further assemble into 2D networks on a minute to hour time scale (Langen et al., 1998). In the case of ANXA5, it has been suggested that these membrane-mediated assemblies are related to annexin’s role in anticoagulation (Andree et al., 1992) as well as the repair patch formation in cell membrane wound healing (Bouter et al., 2011). Interestingly, human ANXA5 and hydra ANXB12 can also form heterotrimers (Patel et al., 2005), suggesting that different trimer-forming annexins share a common mechanism of trimer assembly that has remained conserved throughout evolution. In contrast, annexin A1 (ANXA1) and annexin A2 (ANXA2) are non-trimer forming. Trimer and non-trimer forming annexins also vary in terms of their membrane binding ability. Specifically, the trimer-forming annexins bind membranes with much higher calcium stoichiometry (~11 Ca 2+ -ions in the case of ANXB12 as opposed to ~2 in the case of ANXA2 (Patel et al., 2005)). Trimer-forming annexins also strongly inhibit inter-leaflet flip-flop of lipid molecules, while the non-trimer forming annexins are less effective. Furthermore, non-trimer forming annexins can bind to liquid and gel phase membranes, yet trimer-forming annexins only bind to liquid phase membranes (Patel et al., 2005). These findings show that trimer and non-trimer forming annexins have distinctively different modes of membrane binding, but it remains unclear whether these differences are directly related to whether the various annexins form trimers or not. ANXB12 has four repeats, each composed of a four helical bundle (helices A, B, D, E) and a fifth helix (helix C), which runs perpendicular to the bundle (Figure 2-1A). Membrane interaction is mediated by the loop regions between helices A and B (AB loop) and helices D and E (DE loop). These loop regions, which are located on the convex surface of ANXB12 (dashed line in Figure 2-1A), harbor up to three Ca 2+ - binding sites (labeled AB, AB’ and DE in Figure 2-1A) in each of the four repeats, totaling twelve potential binding sites per protein. The Ca 2+ -binding sites promote membrane binding using a calcium bridging mechanism by which the lipids and the protein jointly coordinate Ca 2+ -ions at the protein-membrane 32 interface. The primary Ca 2+ and membrane binding sites are the AB loops, which make backbone carbonyl contacts to the Ca 2+ -ion (Figure 2-1). In addition, a key interaction comes from the side chain carboxylate from a Glu or Asp residue in the DE loop within the same repeat (Supplemental Fig. S1). Selective disruption of a subset of AB binding sites has been shown to affect lattice formation, but not membrane binding (Kenis et al., 2004). However, it is not yet known whether the mutations also affect trimer formation. Trimer formation is very rapid (in milliseconds) and occurs independently of lattice formation, which requires minutes to hours (Langen et al., 1998). Loss of trimer formation under these conditions would only be expected, if protein-protein interactions alone are insufficient to induce trimer formation. According to crystal structures of ANXB12 and ANXA5 trimers, a small number of amino acids are found at the contact surface between subunits. The most recognized contacts in the literature (Mo et al., 2003) have been the salt bridges between individual subunits (shown for ANXB12 in Figure 2-1B). Disruption of these salt bridges in ANXA5 has also been shown to reduce lattice formation (Bouter et al., 2011). It is therefore possible that the salt bridges are the key factors in governing trimer formation. However, the direct effect of the salt bridges on membrane-mediated trimer formation has not yet been determined and it is not clear whether other protein-protein contacts exist to have significant contributions to trimer formation. One of the central goals of the present study was therefore to obtain quantitative experimental data for the effects of protein-protein and protein-lipid interaction on ANXB12 trimer formation. Surprisingly, we found that the impact of salt bridges on trimer formation, though detectable, was relatively small. In contrast, mutations of other residues at subunit interface, had a much more pronounced effect on trimer formation. These residues engaged in a combination of hydrophobic and electrostatic interactions. While protein-protein interactions were important, they were not sufficient for inducing trimer formation as disruption of some Ca 2+ - and lipid-binding sites also strongly inhibited trimer formation on membranes. Collectively, our data suggest that trimer formation of ANXB12 is simultaneously driven by a network of protein-protein and protein-membrane interactions. 33 Figure 2-1. ANXB12 monomer and trimer structures showing Ca 2+ - and lipid-binding sites and inter subunit salt bridges (A) The monomeric subunit of ANXB12 contains four repeats shown in different colors: repeat 1, blue; repeat 2, yellow; repeat 3, green; and repeat 4, red. The membrane-facing surface of ANXB12 is curved (dashed line, top panel). Ca 2+ -ions are shown as orange spheres. The twelve Ca 2+ - and lipid-binding sites belong to three categories: AB, AB’ and DE (bottom panel). The side chain of AB’ Ca 2+ - ligand E105 in repeat 2 is red. Each repeat is composed of five helices (A to E) as illustrated with the zoomed-in depiction of repeat 3 (right). (B) The top panel shows the crystal structure of the ANXB12 trimer with subunits highlighted in green, light blue and grey. Residue 132, which is used for the EPR-based detection of trimer formation, is located near the 3-fold symmetry axis and is shown in yellow. The location of positively and negatively charged salt bridge residues are indicated by blue + and red - symbols. The bottom panel zooms in on the salt-bridge interactions which are shown using a side view for one of the trimer interfaces. Residue R23 (which could potentially pair with D188 upon loss of K27) is colored in light blue. All illustrations were created using the ANXB12 crystal structure (PDB code: 1aei). 34 2.2 Results To determine how protein-protein and protein-lipid interactions affect Ca 2+ - and membrane- dependent trimer formation, we specifically disrupted these interactions in a total of 22 mutants. A list of all mutants can be found in Table 1 and details of mutated residues are summarized in Figure 2-S1. All mutants were purified using reversible, Ca 2+ -dependent membrane binding, indicating that they were functional with respect to membrane binding. The extent of trimer formation was quantified using a previously established, EPR-based method. This method detected trimer formation using spin-spin interaction that occurs as three spin labels at residue 132 (132R1) come into close proximity (Langen et al., 1998) (Figure 2-1B, yellow). Residue 132 is located in a loop region, where it is highly exposed and accessible, further ensuring high labeling efficiencies. In fact, essentially quantitative labeling has previously been demonstrated for this residue (Langen et al., 1998). In solution, ANXB12 has been shown to be monomeric (Fischer et al., 2007; Langen et al., 1998), and the EPR spectrum of 132R1 gave rise to three sharp lines, characteristic of dynamic spin label motion without significant spin-spin interaction (Figure 2-2A). No significant EPR spectral changes were observed when large unilamellar vesicles (LUV) were added to 132R1 in the absence of additional Ca 2+ (Figure 2-2B), consistent with prior reports that no membrane binding occurs under these conditions (Fischer et al., 2007; Langen et al., 1998). However, membrane binding was induced by the addition of Ca 2+ . According to fluorescence microscopy the vesicles remained largely intact under these conditions (Figure 2-2F) without being remodeled into smaller vesicles, as sometimes observed with other proteins (Varkey and Langen, 2017). As previously described, these conditions led to trimer formation and resulted in a distinctively different EPR spectrum with significant amplitude drop and broad lines indicative of strong spin-spin interaction (Figure 2-2C). The effect of spin- spin interaction was alleviated by mixing labeled protein with 90% unlabeled ANXB12 (B12 Cysless), which gave rise to sharp spectral lines of significantly increased signal amplitude (Figure 2-2D). A similar EPR spectrum with sharp lines and high intensity was also obtained for fully R1-labeled, membrane-bound ANXA2 spin-labeled at residue 152 (152R1), a position that is equivalent to 132R1 in ANXB12 (Figure 2-2E). Since ANXA2 does not undergo membrane-mediated trimerization, the 152R1 spectrum does not 35 Figure 2-2. EPR and fluorescence-based detection of ANXB12 membrane binding and trimer formation All black EPR spectra in panels (A) – (E) are shown normalized to the same number of spins. (A) EPR spectrum of ANXB12 132R1 was recorded in solution (20 mM HEPES, 100 mM NaCl, pH 7.4). (B) In the absence of added Ca 2+ , the addition of 1000 nm LUV (POPS/POPC molar ratio 2 to 1) at a protein-to-lipid molar ratio of 1:450 does not cause membrane binding and results in spectrum analogous to that in (A). (C) Membrane binding was induced using the same condition as in (B), but in the presence of 1 mM CaCl 2. The spectrum is of much lower amplitude and significantly broadened. To better visualize the line shapes, the red spectrum shows the same spectrum at 20X magnification. (D) The EPR spectrum of membrane-bound 132R1 as in (C) but diluted with 90% unlabeled protein (ANXB12 Cysless). This spin dilution leads to much sharper lines and an increase in EPR central line amplitude relative to the spectrum in (C). (E) The EPR spectrum of the 152R1 derivative of ANXA2 in membrane-bound form (same condition as in (C)). 152R1 in ANXA2 is homologous to 132R1 in ANXB12. ANXA2 is non-trimer forming and the EPR spectrum of its 152R1 derivative does not exhibit spin-spin interaction. (F) Fluorescent images showing ANXB12 (green) bound to vesicles (red). ANXB12 was labeled using Alexa Fluor 488 and mixed with 0.2% rhodamine DOPE-labeled, sucrose- filled vesicles (POPS/POPC 2 to 1, red) in buffer (20 mM HEPES, 100 mM NaCl, 1 mM CaCl 2, pH 7.4). 36 exhibit spin-spin interaction. Collectively, these data confirm prior reports that spin-spin interaction can be used to detect ANXB12 trimer formation and that loss of trimer formation causes loss of spin-spin interaction (Langen et al., 1998; Patel et al., 2005). Using the 132R1-based readout, we next determined how trimer formation was affected by mutations at the protein-protein or protein-membrane interface. The membrane-bound state of each annexin mutant was obtained by mixing protein with vesicles (1:450 molar ratio) in the presence of 1 mM Ca 2+ . These conditions were chosen to ensure membrane binding in all cases. The membrane-bound ANXB12 mutants were then concentrated by centrifugation. All mutants pelleted nearly quantitatively with negligible EPR signals remaining in the supernatant. The ability to bind to membranes in all cases was consistent with the fact that all proteins could be purified using reversible, Ca 2+ -dependent membrane binding. As shown in Figure 2-3A to 3D, the amplitude and shape of the EPR spectra of all derivatives varied widely. The differences in amplitude were not due to different concentrations, as all spectra were normalized to the same number of spins. Rather spectral subtractions showed that all spectra were composed of varying ratios of two very different spectral components, illustrated with AB23 mutant in Figure 2-3E. One of these spectral components corresponds to the broad, spin-spin interaction containing 132R1 spectrum (Figure 2-2C), which is indicative of trimer formation. The other spectral component had narrow lines and was of high amplitude. This spectrum was different from that of ANXB12 132R1 in solution (Figure 2-2A) and therefore could not be attributed to unbound protein. Rather, this spectrum was similar to those in Figure 2-2D and 2E and indicated the formation of a membrane-bound state that lacked spin-spin interaction and that no longer formed the trimer. Due to the drastically different amplitudes of the spectral components, it was possible to estimate percent trimer formation from the respective amplitudes, as larger amplitudes indicated less trimer formation. In the following sections, we therefore quantified the spectral amplitudes in triplicates and converted them into percent trimer formation (see Materials and Methods). 37 Figure 2-3. EPR spectra of ANXB12 derivatives with mutations designed to disrupt protein- protein or protein-membrane interactions All mutants contain a spin label at position 132 to monitor trimer formation. Additional mutations are made to disrupt protein-protein (A) or protein-lipid interactions (B-D). The acronyms for all mutants are described in Table 2-1. The EPR spectra of ANXB12 mutants with disrupted Ca 2+ - ligands in their AB Ca 2+ -binding sites are given in (B). Lysine substitutions at AB Ca 2+ -binding sites either alone or in combination with Ca 2+ -ligand disruptions are presented in (C). EPR spectra of mutants in which selected AB and AB’ Ca 2+ -ligands are simultaneously mutated are shown in (D). All spectra have varying amounts of two spectral components arising from a trimer and a non-trimer state. Deconvolution of the respective spectral components is shown for the AB23 mutant in (E). The trimer component is of very low amplitude and to better visualize its line shape, this spectrum is also plotted at 20X magnification (red). All spectra shown in black were normalized to the same number of spins throughout the figure. This representation illustrates the different amplitudes but makes some of the line shapes more difficult to see. The spectra are therefore replotted at uniform amplitude in Figure 2-S2. 38 2.2.1 Some protein-protein interactions, but not the salt bridges, have a strong effect on ANXB12 trimer formation. To test the effect of protein-protein interactions, we first mutagenized three pairs of salt bridges (R16-E163, E20-R191, K27 (or R23)-D188), which are the primary interaction sites between the monomers (Figure 2-1B). It is worth noting that K27 as well as R23 could potentially salt bridge with D188, as has been suggested in case of the homologous residues in ANXA5 (Mo et al., 2003). Both residues were therefore mutated. Two different types of salt bridge-disrupting mutants were made: SB-A and SB-E. For the SB-A mutant, the positively charged salt bridge residues (Figure 2-S1, blue) were replaced by alanine, thereby leaving the negatively charged residues without counterion. For the SB-E mutant, the positively charged residues were replaced with negatively charged glutamic acids. This mutant was designed to be more severe as it replaced the three salt bridges with negatively charged residues that would be expected to repel each other. When compared to the 132R1 reference spectrum (Figure 2-3A, magnified in Figure 2- S2), increased amounts of sharp lines could be observed for SB-A and for SB-E this effect was more pronounced. Nonetheless, all spectra were still of relatively low amplitude and contained clear evidence of abundant trimer formation. In fact, based on the EPR spectra we estimated that trimer formation was ~96% and 92% for SB-A and SB-E, respectively (Figure 2-4A). This result was particularly surprising in the case of the SB-E mutant, where charge repulsion might have been expected to result in stronger effects. These data suggested that the salt bridges alone were not the primary drivers of trimer formation, rather additional and possibly stronger driving forces had to be present. To test the effects of other protein-protein interactions, we mutagenized additional residues located at the subunit interface (F59, Q148 and R149, Figure 2-S1, orange) into alanines, resulting in the interface alanine mutant IF-A. As illustrated in Figure 2-S3, this mutant is disrupted in numerous contacts, most of which were hydrophobic in nature, although they also disrupted electrostatic interactions. In contrast to the salt bridge mutants, this mutant had a much more pronounced effect, resulting in only 35% trimer formation. When combined with the salt bridge alanine mutant (SBIF-A), trimer formation was almost completely abolished (3% trimer). These results indicated that specific protein-protein interactions play an important 39 Figure 2-4. ANXB12 trimer formation estimated from EPR amplitudes The amplitudes of the EPR spectra for the various mutants are shown as fold increase relative to the amplitude of 132R1. All amplitudes are those of spin normalized EPR spectra obtained in triplicates. The second y-axis shows the percentage of trimer formation, which is inversely proportional to the relative amplitude. The mutants are grouped into disruption of (A) protein- protein interactions, (B) AB Ca 2+ -binding ligands, (C) lysine introduction into the AB loops and (D) simultaneous disruption of ligands in the AB and AB’ Ca 2+ -binding sites. The mutant names are explained in Table 1. The bars represent average values of independent repeats (n = 3) with standard deviation. Two-tailed unpaired t-tests were performed for selected mutant pairs and P- values are shown (red). Data analysis was performed in GraphPad Prism 8.0. 40 role in ANXB12 trimer formation, but that a combination of mutations was required to observe strong effects. 2.2.2 Some Ca 2+ -ligands in AB membrane binding sites significantly contribute to ANXB12 trimer formation on membranes. We next examined the contribution of protein-lipid interaction to ANXB12 trimer formation. As described above, AB Ca 2+ -binding sites play a major role in ANXB12 membrane interaction (Figure 2- 1A). To test the influence of these binding sites on trimer formation, we mutagenized the conserved Glu/Asp residues (see Figure 2-S1 and Figure 2-5, green) within the DE loops to their respective amidated forms (i.e. E into Q or D into N). Although the mutated residues reside in the DE loops, they coordinate Ca 2+ in the AB loops. We therefore referred to these mutants as AB1, AB2, AB3 and AB4, where the number indicated the repeat in which the mutations were made. As shown with the EPR spectra in Figure 2-3B and the analysis in Figure 2-4B, AB1, AB2 and AB4 had only a very modest effect, while AB3 caused more significant disruption of trimer formation (72% trimer). Surprisingly, the disruption of trimer formation caused by a single ligand (AB3) was more pronounced than for either of the salt bridges mutants. To further investigate how combinations of Ca 2+ -binding site mutants affected trimer formation, we generated six additional proteins that had mutations in two AB binding sites in different repeats: AB12, AB13, AB14, AB23, AB24 and AB34 (again numbers indicate repeat in which mutations were made). The impact of each double mutant on trimer formation is shown with the EPR spectra in Figure 2-3B and the analysis in Figure 2-4B. All double mutants containing a mutation in repeat 3 (AB23, AB13, and AB34) showed a significant drop in trimer formation. All other double mutants had only minor effects. In fact, even the triple mutant, where all ligands in repeats 1, 2 and 4 were mutated (AB124) had little effect on trimer formation (Figure 2-4B). While ligand mutation in repeat 3 was clearly the dominant event, differences were observed depending on where the second mutation was placed. That is, the AB23 mutant had a very low degree of trimer formation (36% trimer), much lower than that of the individual AB2 (97% trimer) and AB3 (72% trimer) mutations alone. Thus, there was strong synergy when these two ligands were mutated simultaneously. The second highest disruption was seen for the AB13 (76% trimer) mutation, 41 which was similar to that of AB3 alone. The lowest disruption of trimer formation for the AB3-containing mutants was observed for AB34 (83% trimer). 2.2.3 ANXA2-mimicking Lys mutations at the tip of AB binding loops cause a modest reduction of trimer formation. Sequence alignment between trimer-forming ANXB12 and non-trimer forming ANXA2 showed that ANXA2 has lysine substitutions for the membrane-facing Ile/Leu residues within the AB loop (Figure 2-S1 and Figure 2-5, purple). According to a computational simulation study, positively charged lysine residues on the membrane-facing convex surface can facilitate ANXA2-membrane interaction via electrostatic interactions with negatively charged lipids (Hakobyan et al., 2017). Whether these interactions also affect trimer formation, however, is unknown. To test whether ANXA2-mimicking lysine substitutions could destabilize trimer formation of ANXB12, we generated another five mutants: K1234 (simultaneously substituting all four Ile/Leu residues into lysines) and AB1K1, AB2K2, AB3K3, AB4K4 (mutating the Glu/Asp residue and the Ile/Leu residue within each repeat). Again, the numbers refer to the repeat in which the mutations were introduced. The introduction of the four lysines alone reduced trimer formation to 97% trimer. This effect was relatively small and comparable to that of the SB-A salt bridge mutant. Thus, Lys mutations alone are not major disruptors of trimer formation and the corresponding Lys residues are likely not responsible by themselves for the lack of trimer formation in ANXA2. Nonetheless, the introduction of lysines also further potentiated the effects of the ligand mutations. The AB3 mutation alone already had significant effects on trimer formation (72% trimer) and the AB3K3 mutant still further enhanced this effect (64% trimer). Moreover, while the single ligand mutations in the other repeats had no significant trimer disrupting effects, the additional lysine mutations caused robust trimer disruption for AB1K1, AB2K2 and AB4K4 (Figure 2-3C and Figure 2-4C). Among these three mutants, the strongest effect was seen for AB2K2 and AB1K1 (72% and 84% trimer respectively), while a much more modest effect was observed for AB4K4 (97% trimer). Interestingly, these data give the same order of importance for the repeats as the 42 ligand double mutants mentioned above with the importance decreasing in the following manner: repeat 3>> repeat 2> repeat 1>> repeat 4. 2.2.4 The AB’ ligand E105 is important for trimer formation The Ca 2+ - and membrane-binding mutations thus far only involved the AB Ca 2+ -binding sites. However, other binding sites could also contribute to trimer formation. To test this notion, we focused on E105 (Fig. 1a, red), a ligand for the AB’ Ca 2+ -binding site in repeat 2 (Figure 2-S1 and Figure 2-5, cyan). It has previously been suggested that AB’ Ca 2+ -binding sites contribute to lipid binding by specifically coordinating the negatively charged serine moiety of the phosphatidyl serine head group (Swairjo et al., 1995). Prior studies revealed that mutating E105 can cause marked alterations in membrane properties (Sokolov et al., 2000). Therefore, we generated the E105Q in combination with the AB2 or the AB3 mutants described above. The resulting mutants are referred to as AB2AB’2 and AB3AB’2, respectively. As shown in Figure 2-3D and Figure 2-4D, both of these mutants caused significantly more reduction in trimer formation than individual AB mutants alone with the AB3AB’2 almost completely abolishing trimer formation (2% trimer). Thus, AB’ sites like E105 also contributed to trimer stability. Our prior studies showed that the labeling at position 132 is essentially quantitative (Langen et al., 1998). To further ensure that the line shape changes seen here were not a consequence of underlabeling in some cases, we evaluated the labeling efficiencies of mutants which exhibited the strongest line shape changes, including AB3AB’2. As illustrated in Figure 2-S4, all mutants were as fully labeled as the K132C reference mutant. Thus, lack of labeling was not the reason for the line shape changes. Moreover, even if underlabeling of about 10% were present, its spectral effects would be limited, as loss of one spin label in the trimer (statistically most likely scenario (Langen et al., 1998)) would still leave two spin labels in close proximity and this would also cause strong amplitude reductions (Figure 2-S4D). 2.3 Discussion The present study investigated the effects of protein-protein and protein-membrane interactions on ANXB12 trimer formation. While selective mutations of only three or four residues at the protein-protein 43 Figure 2-5. Convex surface of the ANXB12 trimer showing the location of disrupted protein-lipid contacts Each monomer of the ANXB12 trimer is illustrated with a different color. Mutagenized Glu/Asp ligands in AB Ca 2+ -binding sites are colored in green. The E105 ligand of the repeat 2 AB’ Ca 2+ -binding site is colored in cyan. Locations where lysine residues were introduced are purple. The illustration was created from the ANXB12 crystal structure (PDB code: 1aei). Most of the protein-membrane contacts that are critical to trimer formation of ANXB12 (repeat 1, 2 and 3, circled with dashed lines) are located in close proximity to the subunit interface within the trimer. 44 interaction surface had only small effects, a combined mutant in which seven contact residues were substituted (SBIF-A mutant) nearly completely abolished trimer formation. This indicated that protein- protein interactions were essential and that multiple individual contacts were needed to bring about trimer formation. Perhaps more surprisingly, we found that similar trimer disruption could be achieved by mutating only two residues (AB3AB2’) on the membrane binding surface. Protein-membrane interactions are therefore just as important as protein-protein interactions in terms of promoting trimer formation and both interactions are required for trimer formation. By selectively targeting only a subset of up to 12 Ca 2+ - and lipid-binding sites, it was possible to obtain mutants that retained their functional ability to bind to membranes. In other words, all of these mutants still allowed the proteins to be concentrated on the membrane and to experience two-dimensional diffusion. The role of membranes in ANXB12 trimer formation must, therefore, go beyond simply providing a two-dimensional surface. Rather, lipids are likely to engage in some specific interactions that promote trimer formation. While protein-protein and protein-membrane interactions were clearly essential, not all of the interactions were equally important. For example, we found that the salt bridge mutants did not have a very strong effect on trimer formation (4-8%). This was rather surprising as the salt bridges have typically been considered important for trimer formation. A potential reason for why the salt bridge mutations had a relatively small effect was that the energetic gains from charge interactions in the trimer could have been offset by the unfavorable loss of solvation of the same residues in the monomer. In fact, it has previously been suggested that salt bridges may not always represent strong driving forces for protein-protein interactions (Lumb and Kim, 1995). In contrast, the IF-A mutant (F59A/Q148A/R149A) yielded a much bigger effect (65% disruption). This mutation disrupted a variety of interactions, many of which are hydrophobic (Figure 2-S3). However, even this mutant alone only partially disrupted trimer formation and near complete inhibition of trimer formation required the combined SBIF-A mutant, in which seven residues are mutated. Although the ANXB12 repeats are highly homologous, disrupting the AB Ca 2+ - and lipid-binding sites had very different effects, depending on which repeat was targeted. For example, mutations in repeat 45 4 had little or no effect on trimer formation, while mutations in repeats 3, 2, and 1 had much more pronounced effects. Interestingly, all of these repeats were located at the subunit interface in the trimer, while repeat 4 was not (Figure 2-5). This pattern is consistent with a model, in which specific protein oligomerization above the membrane (trimer formation) is coupled to attractive forces between lipids near the trimer interface. The nature of these interactions is still unclear, but it could be related to two characteristic features of trimer-forming annexins, like ANXB12. One of these features is that trimer- forming annexins strongly reduce lipid and protein movement (Cezanne et al., 1999; Gilmanshin et al., 1994; Junker and Creutz, 1993; Megli et al., 1998). This immobilization is caused by the formation of a specific lipid network that is accommodated by a complementary spacing between the tightly packed lipids and the membrane binding sites of the assembled trimer (Patel et al., 2005). Based on prior studies it is likely that formation of this network is strongly affected by E105 mutation (Sokolov et al., 2000). Moreover, non-trimer forming annexins do not result in lipid immobilization and they do not appear to form a specific lipid network. Another characteristic feature of ANXB12 is that the membrane-binding surface is highly curved and located on the convex side of the protein. Our prior studies (Isas et al., 2002; Isas et al., 2004) revealed a continuous contact between the convex surface and the lipid molecules. By following the convex shape of ANXB12 (see dashed line in Figure 2-1A), the membrane is likely to experience thickness deformations. In the trimer, such thickness deformations would be complementary as the most stretched lipid regions are located at the contact surface between monomers in the trimer (Figure 2-1). According to recent computational work, the local clustering of stretched or compressed lipid regions is energetically favorable (Kahraman et al., 2016a, b). Thus, it is possible that thickness deformations might contribute to trimer formation of ANXA5 and ANXB12, where binding is largely driven by the numerous Ca 2+ - and lipid-binding sites on the convex surface. This effect might be less pronounced for ANXA1 and ANXA2, which have much fewer such binding sites in lieu of potentially more flexible Lys-containing membrane- binding loops. Future studies will have to show the extent to which membrane thickness deformations contribute to trimer formation in annexins. 46 Table 2-1. Summary of ANXB12 mutants in the study In addition to the mutations listed, all ANXB12 derivatives listed in the table are also R1 spin-labeled at residue 132 to allow EPR-based detection of trimer formation. The 22 mutants are divided into two main categories where protein-protein or protein-lipid contacts were disrupted. The right column specifies the acronym of each mutant as it is referred to in the paper. 47 The finding that protein-lipid contacts can actively control oligomerization of ANXB12 suggests the possibility that similar mechanisms could also apply to other proteins, perhaps even allowing the hetero- oligomerization of different proteins. It has been well established that cellular membranes can be organizing centers that bring a subset of proteins into the same local membrane domain where they promote oligomerization by enhancement of local concentration. By making specific lipids parts of the oligomeric complex it may thus be possible to regulate protein-protein interactions through control of lipid binding interactions and perhaps lipid compositions. 2.4 Materials and Methods 2.4.1 Mutagenesis, Expression, Purification and Spin Labeling of ANXB12 mutants ANXB12 and ANXA2 were expressed using pSE420 plasmids. The ANXB12 cysless mutant, K132C mutant and ANXA2 K152C mutant were generated as described previously (Langen et al., 1998; Patel et al., 2005). All other mutant plasmids were prepared with Q5 Site-Directed Mutagenesis Kit (NEB) using ANXB12 K132C mutant as the template and were sequence confirmed. The mutant plasmids were transformed into DH5a cells (Zymo5α, #T3007) and cells were spread on LB plates (100 μg/ml ampicillin) at 37 °C overnight. A single colony was picked and inoculated into 5 ml TB media with ampicillin (100 μg/ml) for overnight incubation (225 rpm, 37 °C). Cultures were then pelleted and the pellet was resuspended in 5 ml fresh TB media with ampicillin. Of these 500 μl were used to inoculate 50 ml fresh media. This starting culture was incubated for another 3 hours before expanding into two flasks of 500 ml fresh TB media with ampicillin. Protein expression was induced by 1 mM IPTG when O.D. 600nm reached 1.2, followed by overnight incubation (180 rpm, 18 °C). Bacteria were harvested the next morning by centrifugation (3,500 x g, 4 °C, 20 minutes). Protein purification was based on the reversible Ca 2+ - dependent binding to phosphatidylserine-containing vesicles. Giant unilamellar vesicles were made with the Reeves-Dowben method (Reeves and Dowben, 1969). Bacterial pellets were incubated in lysis buffer (20 mM HEPES, 100 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 0.15 mg/ml egg lysozyme, pH 7.4) for about 30 minutes, then sonicated in the presence of 0.3 μM PMSF, followed by centrifugation 48 (26,000 x g, 4 °C, 40 minutes). The supernatant was transferred and mixed with Reeves-Dowben vesicles in the presence of 5 mM CaCl 2 followed by centrifugation (26,000 x g, 4°C, 30 minutes). The pellet was washed by resuspending in washing buffer (20 mM HEPES, 100 mM NaCl, 1 mM CaCl 2, pH 7.4) followed by centrifugation (18,000 x g, 4 °C, 20 minutes). This process was repeated three times. The pellet was then resuspended in 10 ml elution buffer (20 mM HEPES, 100 mM NaCl, 10 mM EGTA) and again subjected to centrifugation (26,000 x g, 4 °C, 30 minutes). The supernatant was again subjected to centrifugation. The final supernatant was concentrated to 0.5 ml using a 10 kDa Centricon Filter (Millipore #UFC901096). Next, the protein was purified with a Superdex 200 10/300 GL size exclusion column (GE Healthcare) using buffer (20 mM HEPES, 100 mM NaCl, 1 mM EDTA, 1 mM DTT, pH 7.4). The protein eluted in the second peak at ~16 ml elution volume. For MTSL labeling, DTT was first removed using a PD10 column (GE Healthcare). MTSL spin label ((1-Oxyl-2,2,5,5-tetramethyl-∆3-pyrroline-3-methyl) Methanethiosulfonate) (Toronto Research Chemicals, #O875000) was added at > 5:1 (label/protein) molar ratio. According to the labeling kinetics for ANXB12 residue 132, the reaction is complete after 30 minutes incubation at 4 °C. To ensure complete labeling and high label efficiency, the labeling reaction was allowed to proceed overnight. The labeled protein was then passed through a HiTrap Q XL anionic exchange column (GE Healthcare) to wash away any free labels and for final purification. The end product was verified on SDS-PAGE gels as a single band at around 35 kDa, consistent with the molecular weight of annexin monomers. In order to match the osmolarity of lipid vesicles used in the EPR experiments later, buffer exchange (20 mM HEPES, 100 mM NaCl, pH 7.4) was performed for the purified proteins using PD10 columns. Protein concentration was determined by UV absorbance at 280 nm using an extinction coefficient of 14,900 M -1 cm -1 (ExPASy, ProtParam Tool). To label the K132C reference mutant diamagnetically, (1- Acetoxy-2,2,5,5-tetramethyl-∆3-pyrroline-3-methyl) Methanethiosulfonate (Toronto Research Chemicals, #A167900) was used in replacement of the paramagnetic MTSL. 2.4.2 Preparation of Lipid Vesicles 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC) and 1-palmitoyl-2-oleoyl-sn-glycero- 3-[phospho-L-serine] (POPS) were purchased as chloroform solutions from Avanti Polar Lipids, Inc (#850457C 49 and #840034C). Lipids were first mixed (POPS/POPC molar ratio 2 to 1) and then dried under a gentle nitrogen flow to form a thin layer of lipid film and kept in a desiccator overnight at room temperature. The film was later resuspended in buffer (20 mM HEPES, 100 mM NaCl, pH 7.4) to a final concentration of 15 mg/ml. After 10 cycles of freeze-and-thaw, lipid vesicles were extruded through a 1.0 μm Nucleopore Track-Etch Membrane (Whatman, #800319) back and forth for 21 times using glass syringes. 2.4.3 CW-EPR Spectroscopy Analysis For EPR measurements, annexin protein (ANXB12 or ANXA2) were mixed with 600 μg of the extruded vesicles (POPS/POPC molar ratio 2 to 1) in 20 mM HEPES, 100 mM NaCl, pH 7.4 with 1 mM CaCl 2 (protein-to-lipid molar ratio ~1:450) in a total volume of 1000 ul. The lipid vesicles as well as the membrane-bound annexins were harvested by centrifugation (21,000 x g, room temperature, 20 minutes). After carefully removing the supernatant, the pellet was collected into a borosilicate glass capillary to record EPR spectra using a Bruker EMX spectrophotometer fitted with ER4119HS resonator (scan width 150 G, modulation amplitude 1.5 G). All spectra were normalized to the same number of spins using double integration. The central line amplitudes from the normalized spectra were converted to fold increase over the 100% labeled ANXB12-K132C mutant and plotted. Estimates of trimer formation were then obtained from the spectral amplitudes of the respective normalized spectra assuming a linear combination of the individual spectral components of Figure 2-3E. 2.4.4 Fluorescence Microscopy The ANXB12 K132C mutant was labeled with Alexa Fluor 488 C5 Malemide (Invitrogen, #A10254) at 4 °C overnight. Residual free dye was removed by PD10 column (GE Healthcare #17-0851- 01). Sucrose-filled (240 mM) Reeves Dowben vesicles composed of POPS/POPC (2 to 1 molar ratio) and labeled with 0.2% rhodamine-DOPE (Avanti Polar Lipids, #810150) were prepared and washed with buffer (20 mM HEPES, 100 mM NaCl, pH 7.4) for 10 minutes at 21,000 x g. Protein and lipid vesicles were mixed at an estimated 1 to 450 protein-to-lipid ratio in the same buffer containing 1 mM CaCl 2 at room temperature for 20 minutes. The images were acquired using a Zeiss LSM 780 inverted confocal microscope with a 63x oil immersion NA 1.4 objective lens and then processed using Fiji/ImageJ software. 50 Figure 2-S1. Sequence alignment of non-trimer forming ANXA2 and trimer-forming ANXB12 The alignment was performed using Clustal Omega. Regions adopting α helical secondary structures are indicated by colored ribbons (color coded the same as in Fig. 1a). Each helix is denoted with its respective repeat number (from 1 to 4) followed by helix letter (from A to E). Residues mutated in the present study are highlighted. Positively charged salt bridge residues of ANXB12 are colored in blue. IF-A mutant residues (F59, Q148 and R149) are orange. Mutated Glu/Asp ligands (E70, E142, E226 and D301) of ANXB12 AB Ca 2+ -binding sites are green. The E105 ligand of repeat 2 AB’ Ca 2+ -binding site is cyan. Residues at the tip of the AB loops (I29, L101, I185 and L260) are purple. All of these residues are hydrophobic in ANXB12, but three of the equivalent residues in ANXA2 are lysines. 51 Figure 2-S2. EPR spectra from Figure 3 shown at same amplitude The EPR spectra in Figure 2-3 were shown normalized to the same number of spins, which emphasizes the amplitude differences caused by varying amounts of trimer formation. To better illustrate the line shapes of the individual spectra, all spectra are shown here at the same central line amplitude. 52 Figure 2-S3. Protein-protein contact surface disrupted by IF-A mutations (A) The trimer contact surface is composed of repeat 1 from one subunit and repeats 2/3 from the neighboring subunit. IF-A mutant residues are highlighted in orange. F59 is located in repeat 1 while as Q148/R149 are located in repeat 2. (B) Stick representations showing backbones and sidechains of amino acids in contact with IF-A residues according to the crystal structure (PDB code: 1aei). 53 Figure 2-S4. Verification of labeling efficiency for the top four mutants most significantly disrupted in trimer formation. (A) Solution EPR spectra for reference mutant K132C and the top four mutants most significantly disrupted in trimer formation (IF-A, SBIF-A, AB23 and AB3AB’2) were measured in buffer (20 mM HEPES, 100 mM NaCl, pH 7.4). Their nearly perfect spectral overlap is consistent with an identical local environment. (B) To evaluate labeling efficiencies for these four mutants, we estimated the relative protein and spin label concentrations. The former is proportional to 280 nm UV absorbance while the latter is proportional to the double integration of the EPR signal. The two parameters were measured in triplicates and plotted with y-axis showing EPR double integration and x-axis showing UV absorbance at 280 nm. Data fitting shows highly linear relationship (R 2 = 0.99). No significant underlabeling relative to the reference mutant K132C was detected, as in all cases the labeling was directly linearly related with protein concentration. Data analysis was performed in GraphPad Prism 8.0. (C) The ratios of the two parameters shown in (B) were used to estimate relative label efficiency. Triplicate measurement values of reference mutant K132C were averaged and the ratio was set to be 100% label efficiency. This was done based on the essentially complete spin coupling observed for the spectrum of spin-labeled K132C recorded in its membrane-bound state. Relative labeling efficiencies for the four mutants is then derived from the same ratio in triplicate repeats as shown in the table. (D) EPR spectra of membrane-bound ANXB12 K132C mutant (black line) were measured as described in Figure 2-2C. When replacing 10% MTSL-labeled protein with 10% diamagnetically- labeled protein (DMSL K132C), a small but detectable change in line shape was observed (red line). When we overlay the two spectra with a composite spectrum from 90% membrane-bound trimer and 10% membrane-bound monomer (blue line), the latter has much higher central line amplitude. Thus, underlabeling affects the spectra in a way that is non-linear with respect to apparent monomer concentration. In fact, 10% underlabeling has the same spectral effect as ~2% monomer formation. All spectra were normalized to the same number of spins. 54 Chapter 3: Structure of Membrane-Bound Huntingtin Exon 1 Reveals Membrane Interaction and Aggregation Mechanisms Abstract Huntington’s Disease is caused by a polyQ expansion in the first exon of huntingtin (Httex1). Membrane interaction of huntingtin is of physiological and pathological relevance. Using EPR and ODNP, we find that the N-terminal residues 3-13 of wild-type Httex1(Q25) form a membrane-bound, amphipathic α-helix. This helix is positioned in the interfacial region, where it is sensitive to membrane curvature and electrostatic interactions with headgroup charges. Residues 14-22, which contain the first five residues of the polyQ region, are in a transition region that remains in the interfacial region without taking up a stable, α-helical structure. The remaining C-terminal portion is solvent-exposed. The phosphomimetic S13D/S16D mutations, which are known to protect from toxicity, inhibit membrane binding and attenuate membrane- mediated aggregation of mutant Httex1(Q46) due to electrostatic repulsion. Targeting the N-terminal membrane anchor using post-translational modifications or specific binders could be a potential means to reduce aggregation and toxicity in vivo. 3.1 Introduction Huntington’s Disease (HD) is a fatal neurodegenerative disorder (Gomez-Tortosa et al., 1998; Kirkwood et al., 2001; Paulsen, 2011; Smith et al., 2000) with no existing cure. The disease is caused by the expression of mutant huntingtin protein with long, expanded polyglutamine stretches (>36Q) that forms amyloid-like aggregates, which can be found in HD patient tissue samples as well as cell and animal models (Iuchi et al., 2003; Scherzinger et al., 1997). A number of membrane-related functions have been reported for huntingtin, including vesicular trafficking (Gauthier et al., 2004; Gunawardena et al., 2003; Lee et al., 2004; Pal et al., 2006; Velier et al., 1998) and autophagy (Gelman et al., 2015; Martin et al., 2015; Ochaba et al., 2014). In addition to these physiological roles, membrane interaction is likely also of relevance in disease. In vitro, membrane interaction can potently accelerate aggregation and fibril formation of mutant huntingtin, in a misfolding process that can cause membrane damage and the formation of potentially toxic 55 misfolded species (Pandey et al., 2018). In vivo, mutant huntingtin or its naturally occurring N-terminal fragments are present in membrane fractions (Kim et al., 2001; Suopanki et al., 2006; Velier et al., 1998). Moreover, mutant huntingtin can alter the fluidity of cell membranes (Sameni et al., 2018) and disrupt the morphology of membranous organelles (Liu et al., 2015; Squitieri et al., 2006). Interestingly, perturbations in lipid homeostasis and biosynthesis have also been reported in HD (Aditi et al., 2016; Block et al., 2010; Kreilaus et al., 2016; Valenza et al., 2005). These dysregulations and perturbations could contribute to neuronal toxicity and loss (Albin et al., 1992; Davies et al., 1997; Reiner et al., 1988; Richfield et al., 1995). A frequently studied and biologically relevant fragment of the huntingtin protein is its exon 1 (Httex1). Full-length huntingtin is a large protein with 67 exons, but among these exons, Httex1 is of importance for a number of reasons. Httex1 contains the disease-causing polyQ region (Figure 3-S1A) and forms amyloid-like fibrils in a Q-length dependent fashion. It is naturally generated in vivo, as Httex1 or fragments of similar size arise from aberrant splicing or proteolysis (Gafni et al., 2004; Graham et al., 2006; Neueder et al., 2017; Sathasivam et al., 2013; Wellington et al., 1998; Wellington et al., 2000). Importantly, multiple cell and animal models revealed that mutant Httex1 is sufficient to induce toxicity and mimic disease pathology (Chan et al., 2014; Duennwald and Lindquist, 2008; Mangiarini et al., 1996; Yang et al., 2010). Httex1 contains an important membrane-binding region at its N-terminus (N17 domain) that is thought to be a membrane anchor for interaction with a wide range of intracellular membranes, including those of ER, Golgi and mitochondria (Atwal et al., 2007; Rockabrand et al., 2007). The N-terminus also contains a nuclear export signal and is the target of multiple post-translational modifications (Atwal and Truant, 2008; Atwal et al., 2007; DiGiovanni et al., 2016; Lee et al., 2013; Zheng et al., 2013). Of note are the phosphorylation sites at residues S13 and S16, as phosphomimetic (S13D/S16D) mutations have been protective against mutant huntingtin toxicity in both cell and animal models (Di Pardo et al., 2012; Gu et al., 2009; Mishra et al., 2012). Bordering the polyQ domain at the C-terminus is a 50 amino acid long proline rich domain (PRD) that mediates the interactions of Httex1 with its protein binding partners containing SH3, WW or EVH1 domains (Gao et al., 2006; Kay et al., 2000; Zarrinpar et al., 2003). 56 In fibrils of Httex1, each of the three domains takes up very different structures (Bugg et al., 2012; Isas et al., 2017; Sivanandam et al., 2011). The polyQ region forms the β-sheet core of the fibrils. While the N17 also becomes ordered and tightly packed, it contains mainly α-helical structure. The C-terminal PRD takes up a mixture of polyproline II helical and random coil structure, and according to the bottle brush model of Httex1 fibrils (Bugg et al., 2012; Isas et al., 2017), these regions form bristles that face outward away from the fibril core. The formation of Httex1 fibrils can be promoted by the presence of membranes (Pandey et al., 2018). Unfortunately, no detailed structural information for membrane-bound Httex1 is currently available. Most of the existing structural information has been obtained from studies of smaller, Httex1-mimicking peptides with various membrane-mimetic molecules (Ceccon et al., 2018a; Ceccon et al., 2018b; Levy et al., 2018; Michalek et al., 2013a; Michalek et al., 2013b). These studies concluded that the N17 can adopt α-helical structure in its membrane-bound state, but there is little consensus on the length of the helix and the residues which partake in helix formation (Ceccon et al., 2018b; Levy et al., 2018; Michalek et al., 2013b). The structural differences in prior studies could have been caused by the different peptide lengths or the different membrane mimetic reagents that were used in those studies. The precise peptide length could become important if regions outside the N17 modulate membrane interaction. In fact, it has been suggested that the C-terminal PRD might also influence membrane interaction (Burke et al., 2013a). The use of different lipid or membrane-mimetic conditions could have further contributed to the structural difference in prior studies. Such effects have already been reported for other amyloidogenic proteins, such as α-synuclein (Bortolus et al., 2008; Drescher et al., 2008; Georgieva et al., 2008; Jao et al., 2008; Ulmer et al., 2005; Varkey et al., 2013). The main goals of the present study were, therefore, to (1) determine how the entire Httex1 protein binds to intact phospholipid bilayers (rather than membrane mimetics), (2) measure the depth of Httex1 insertion into membranes, and (3) determine mechanistically which forces promote membrane binding, membrane-mediated aggregation and fibril formation. Toward this end, we used a combination of biophysical techniques, including transmission electron microscopy (TEM), circular dichroism (CD), 57 continuous-wave electron paramagnetic resonance (CW-EPR) and Overhauser dynamic nuclear polarization (ODNP). 3.2 Results: 3.2.1 Httex1(Q25) membrane interaction is mediated by its N-terminal region In order to investigate the membrane-binding induced conformational changes by EPR, we generated tag-free, singly R1-labeled Httex1(Q25) derivatives and recorded their EPR spectra in the absence and presence of small unilamellar vesicles (SUVs) containing 25% POPS and 75% POPC. Under these conditions, SUVs remained visible by transmission electron microscopy over a period of at least seven hours regardless of whether Httex1 was absent or present (Figure 3-S1B and S1C). As illustrated with three sites for each domain (N17, polyQ and PRD), all spectra of R1-labeled derivatives of monomeric Httex1(Q25) gave rise to CW-EPR spectra with relatively sharp line shapes (Figure 3-1, black spectra), consistent with a significant degree of structural dynamics and fast rotational tumbling in the absence of membranes (Bravo-Arredondo et al., 2018; Newcombe et al., 2018). In contrast, the Httex1(Q25) derivatives with spin labels introduced within the N17 domain gave rise to line broadening and pronounced reduction in amplitude in the presence of lipid vesicles (Figure 3-1, red spectra), indicating that the N17 domain becomes less dynamic upon membrane interaction. This ordering also affected an N-terminal site in the polyQ region (21R1), while the more C-terminal sites (35R1, 42R1, 55R1, 80R1, 90R1) did not reveal any significant spectral changes upon membrane interaction. To further verify whether membrane- induced structural changes were localized to the N-terminal region of Httex1(Q25), we generated 20 additional R1-labeled derivatives (Figure 3-S1A) and recorded their EPR spectra in the presence of membranes (Figure 3-S2). The mobility information contained in all spectra was summarized by the inverse central line width (ΔH 0 -1 ), a commonly employed mobility parameter (Figure 3-2A, red circles). In the presence of membranes, the lowest ΔH 0 -1 values were obtained for the first 22 residues while the values in the C-terminal region were much larger, indicating much higher mobility in the latter regions. The result indicated that most membrane-induced ordering occurred in the first ~ 22 amino acids, thus including a 58 Figure 3-1. CW-EPR spectra of singly R1-labeled Httex1(Q25) derivatives The X-band EPR spectra of Httex1(Q25) derivatives with spin labels in N17 (4R1, 9R1, 17R1), polyQ (21R1, 35R1, 42R1), or PRD (55R1, 80R1, 90R1) in the absence of lipid vesicles (black) and in the presence of lipid vesicles (red). Scan width was 100 G and modulation amplitude was 1.5 G. All spectra were normalized to have the same central line amplitude for protein in the absence of lipid vesicles. EPR central line amplitude reduction and line broadening upon membrane binding are visible for 4R1, 9R1, 17R1, and 21R1, but not for the more C-terminal sites. See also Figures 3-S1 and 3-S2. 59 region that encompasses the N17 and the first five N-terminal residues in the polyQ region. From residue 23 onwards, mobility increased quickly, further supporting the notion that the effect of membrane interaction was less pronounced for more C-terminal residues. While the EPR spectra for the first 22 amino acids indicated structural ordering, they also revealed a lack of tertiary or quaternary packing interactions (Figure 3-S2). Such interactions would have led to more strongly immobilized spectral components and larger line widths, neither of which were observed here. Collectively, the line shapes suggested that no significant oligomerization or aggregation of the proteins occurred on the lipid membrane surface for Httex1(Q25). Under the present conditions, we also found no evidence for aggregation of the Httex1(Q25) over a period of seven hours by TEM (Figure 3-S1C). Moreover, we monitored the EPR spectrum of Httex1(Q25) labeled at position 35 (35R1) (Figure 3-S1D) during the same time period in the presence of membranes. Residue 35 was chosen because it is located within the center of the polyQ domain, a region that adopts a β-sheet structure in fibrils (Isas et al., 2017). The EPR spectral lines remained consistently sharp, indicating that the site resides in a highly dynamic, freely moving region, further supporting the lack of aggregation under the present conditions. The finding that this non-pathogenic variant (i.e. short Q-length containing Httex1(Q25)) does not aggregate under these conditions is important, as we previously found that membranes can potently accelerate fibril formation of Httex1 with longer, pathogenic Q-lengths (Pandey et al., 2018) (also see below). 3.2.2 Accessibility measurements reveal formation of a short amphipathic α-helix in N17 with shallow membrane penetration. To characterize the structure and membrane immersion depth of Httex1(Q25) in more details, we measured the O 2 and NiEDDA accessibilities (PO 2 and PNiEDDA) of the R1-labeled derivatives. This method is based on the preferential partitioning of O 2 into the hydrophobic membrane environment, whereas NiEDDA is predominantly located in the hydrophilic aqueous environment (Altenbach et al., 1994). It is well established that the ratio of the two accessibilities is a measure of membrane immersion depth (Altenbach et al., 1994). The natural log of this ratio represents the immersion depth parameter Φ (Φ = 60 Figure 3-2. EPR accessibility measurements and CD indicate the formation of helical structure upon membrane binding (A) Depth parameter F (blue) and inverse central line width DH01 (red) are plotted for the various labeling positions. Error bars show ±standard deviation of independent repeats (n = 3). (B) Helical wheel representation of Httex1(Q25) amphipathic helix (residues 3–13). Residues with F values >0.8 (more membrane exposed) are colored gray and fall onto one surface of the helical wheel. (C) Normalized EPR central line amplitudes of 15 mM Httex1(Q25) 21R1 in the presence of increasing concentrations of 25% POPS and 75% POPC SUV lipids (0, 0.15, 0.75, 1.5, 3, and 10 mM). All spectra were normalized, and the spectral amplitude of the sample in the absence of lipid vesicles was set to 1. The gradually stabilized curve indicates that membrane binding saturates between 1 and 2 mM of lipid. Error bars show ±standard deviation of independent repeats (n = 3). (D) CD spectra of Httex1(Q25) 21R1 in the absence and presence of 3 mM 25% POPS and 75% POPC SUV lipid. The spectra are represented as mean residual ellipticities [θ] MRE222nm plotted as function of wavelength (200-260 nm). (E) The difference spectrum between the membrane-bound spectrum and the solution spectrum shown in (D) had minima near ~208 nm and 222 nm, suggesting increased a helicity upon membrane interactions. See also Figures 3-S2 and 3-S3. 61 In(PO 2/PNiEDDA)). This measure is directly proportional to the immersion depth with larger Φ values corresponding to deeper membrane insertion. As shown in Figure 3-2A (blue circles), the Φ values are generally elevated in the first 22 residues, indicating that this region, which also experienced the most pronounced membrane-mediated structuring, is in contact with the membrane. While the Φ values were generally high in the first 22 amino acids, some fluctuations could still be detected. These were particularly pronounced for residues 3 to 13 for which the Φ values underwent a periodic oscillation where residues 3, 4, 6, 7, 8, 10 and 11 had the highest values (more membrane exposed), while residues 5, 9, 12, and 13 are local minima (more solvent exposed). As illustrated with the helical wheel representation in Figure 3-2B, the two groups of residues fall onto opposing surfaces of an amphipathic helix. In agreement with the notion of an asymmetrically solvated, amphipathic helix, we also find that the O 2 and NiEDDA accessibilities for this region are out-of-phase, with a periodicity of an α- helix (~3.6 amino acids/turn) (Figure 3-S3). Out-of-phase periodicity is typically observed for asymmetrically solvated helices as residues facing toward the acyl chain region experience the highest O 2 and the lowest NiEDDA accessibilities, while the opposite is the case for residues on the opposing, solvent- facing surface (Hubbell et al., 1998). To further verify that membrane interaction induces α-helical structure in Httex1, we performed circular dichroism (CD) experiments. Since the high UV light absorbance at the highest lipid concentrations used in the EPR experiments (up to 20 mM) complicated the CD experiments, we first used EPR to ensure that Httex1 was still membrane-bound at lower vesicle concentrations. Using the EPR amplitude change of Httex1(Q25) 21R1, we found that vesicle binding saturates at lipid concentrations between 1 and 2 mM (Figure 3-2C). Using CD, we were able to record spectra with up to 3 mM lipid and obtained reasonable signal-to-noise for a wavelength range between 200 to 260 nm. Indeed, the addition of lipid vesicles induced a shift in the CD spectra (Figure 3-2D). The difference spectrum obtained for the protein in the absence and presence of 3 mM lipid had the typical shape expected for α- helix formation (Figure 3-2E). These data further support the notion that membrane interaction increases α-helical structure in Httex1(Q25). 62 Figure 3-3. Httex1(Q25) ODNP measurement (A) k σ retardation factors of Httex1(Q25) versus residue number. (B) Distance to phosphate level versus residue number, where 0 is at, - is below, and + is above phosphate level of the lipid head group. k σ retardation factors of Annexin B12 are taken from a previous publication (Cheng et al., 2013) as the reference to calibrate the distance of Httex1(Q25) (error bars show ±standard deviation, n = 3). See also Figure 3-S4. 63 In order to determine the degree of membrane immersion for the membrane-bound regions of Httex1(Q25), we further analyzed the EPR data by converting the Φ values of membrane-exposed residues into immersion depth using calibration standards (see Materials and Methods). Based on this calibration, we find that the side chains of the membrane-exposed residues in the helix are located just below the level of the phosphate group (3R1, 4R1, 7R1 and 11R1 at -1, -1, -1 and -5 Å, respectively). Here, the negative values indicate a location in the hydrophobic portion of the bilayer relative to the phosphate level which is defined as zero. Considering that the membrane-exposed side chains are the most deeply inserted part of an amphipathic helix, we can estimate that the center of the helix is in the interfacial region, above the phosphate level of the lipid headgroups. 3.2.3 ODNP measurements indicate solvent-exposed, C-terminal region that does not interact with the membrane. Due to the shallow insertion of Httex1 into the membrane, O 2 and NiEDDA accessibility measurements could not give reliable membrane proximity estimates for much of the Httex1(Q25) protein. In order to extend the range for membrane proximity measurements into the aqueous solution, we performed ODNP measurements that are more sensitive to distances above the phosphate level (Fisette et al., 2016). ODNP measures water accessibility and dynamics, which is quantified using the electron-proton spin cross relaxivity, k σ, between the unpaired electron spin of the R1 label and the proton spin of water. We then compute the k σ retardation factor as determined by the ratio of k σ of bulk water over k σ of surface hydration water, where a value of 1 corresponds to bulk water diffusivity and increasing k σ retardation values to slower water diffusivity. Two key factors have been established that allow us to deduce from the k σ retardation factor information about the distance of the R1 label from the lipid membrane phosphate level. On the one hand, the k σ retardation factor has been shown to have an approximately linear relationship with the retardation in surface water diffusivity in the retardation factor range up to 6, and still monotonically increase with water diffusivity retardation beyond this range (Barnes et al., 2017). On the other hand, it has been established previously that the retardation in surface water diffusivity inversely scales with the distance of the R1 label to the phosphate level on lipid membrane surfaces, with water 64 diffusivity monotonically approaching bulk water diffusivity values at greater distance from the membrane surface (Cheng et al., 2013; Fisette et al., 2016). Consequently, greater k σ retardation corresponds to distances closer to the lipid membrane. The k σ retardation factors were measured for six singly R1-labeled Httex1(Q25) derivatives (11R1, 22R1, 30R1, 42R1, 60R1 and 90R1) (Figure 3-3A) and the ODNP data were summarized in Figure 3-S4. The data follow a qualitatively similar trend as the EPR-based mobility and accessibility studies. The strongest water diffusion retardation was seen for residue in the N-terminal region (11R1) and the values decreased toward the C-terminal region, with residues 20-30 being in a transition region (22R1) and the smallest retardation values observed from residue 30 onward. When converted into distances using an empirical calibration (Fisette et al., 2016), we find that the 11R1 position penetrates ~1 Å below the phosphate level of the lipid headgroup (Figure 3-3B). This result is consistent with the shallow membrane penetration obtained from EPR power saturation. In addition, residue 22 was estimated to be ~5 Å above the phosphate level. The further C-terminal residues (30, 42, 60 and 90) were at distances more than 10 Å above the phosphate level, i.e. farther enough from the level of the phosphate to approach the dynamics of the bulk water environment. Collectively, the ODNP and the EPR data suggest a simple model of membrane-bound Httex1 in which residues 3 to 13 form an α-helical structure, the ordered secondary structure is then gradually lost in the transition region from residue 14 to 22, which are located still within the interfacial membrane contact region, while residues 23 to 92 are in the bulk aqueous environment (Figure 3-6A). 3.2.4 Httex1(Q25) membrane binding depends on vesicle size and membrane curvature The relatively shallow insertion of Httex1(Q25) into the membrane could have implications for the membrane curvature dependence of membrane binding. That is, by being located up in the headgroup region, one might expect Httex1 to have stronger binding to membranes with positive curvature as such membrane can more easily accommodate additional protein mass in their headgroups (Figure 3-6B). To test this model, we incubated Httex1(Q25) 21R1 with vesicles of increasing diameter (i.e. decreasing positive curvature). 65 Figure 3-4. Membrane curvature, ionic strength, and charge modulate Httex1(Q25) membrane binding Percent membrane binding of Httex1(Q25) 21R1 was obtained from the EPR spectra under varying conditions. (A) Effect of decreasing membrane curvature by using SUVs, 200-nm large unilamellar vesicles (LUVs), and 1,000-nm LUVs in 20 mM HEPES, 100 mM NaCl (pH 7.4) buffer. (B) Increasing concentrations of NaCl (0, 100, 500, 1,000 mM in 20 mM HEPES [pH 7.4]) and SUVs with 25% POPS and 75% POPC are used. (C) SUVs with increasing molar percentages of negatively charged POPS lipids (0%, 5%, 10%, 25%, 50%) are used in 20 mM HEPES, 100 mM NaCl (pH 7.4) buffer. One-way ANOVA analyses with Dunnett’s multiple comparisons test were performed relative to the experimental group using 25% POPS and 75% POPC SUVs in 20 mM HEPES, 100 mM NaCl (pH 7.4). Statistically significant differences are shown as **** adjusted p < 0.0001 and ** 0.001 < adjusted p < 0.01; n.s., not significant. Error bars show ±standard deviation of independent repeats (n = 3). All data analyses were performed in GraphPad Prism 7.0. For EPR spectra, see Figure 3-S5. 66 For all experiments, we used 1.5 mM lipid, the concentration at which binding was close to saturation in the prior experiments (Figure 3-2C). Again, we used the EPR amplitude changes as a measure of membrane binding. Using this readout, we found that binding to the highly curved SUVs was much more potent than to larger vesicles with 200 nm or 1000 nm diameters (Figure 3-4A, for EPR spectra see Figure 3-S5A). As described in the Materials and Methods section, it was possible to convert the spectral and amplitude changes into percent binding. This yielded 91, 32 and 23% binding for SUVs, 200 nm and 1000 nm vesicles, respectively. Thus, vesicle size and membrane curvature strongly influence membrane binding affinity. This is consistent with prior observations made on Httex1-mimicking peptides interacting with lipids tethered to different solid supports (Chaibva et al., 2014). 3.2.5 Electrostatic interactions contribute to Httex1 membrane interaction According to the just outlined structural model (Figure 3-6A), the N17 comes into close contact with the membrane. This region of Httex1(Q25) has a net positive charge (3 Lys residues and 2 Glu residues), making it possible that charge interactions with negatively charged membranes could promote binding. In order to test for electrostatic components that could promote membrane interactions, we investigated the ionic strength and lipid charge dependence of Httex1 membrane interaction. To test the effect of ionic strength, we incubated the R1-labeled protein Httex1(Q25) 21R1 with SUVs containing 25% POPS and 75% POPC at different ionic strengths (i.e. NaCl concentrations), and monitored the resulting changes in EPR amplitude. Increasing ionic strength attenuated the membrane-dependent drop in signal amplitude in a relatively small, but clearly detectable manner (Figure 3-4B), indicating less membrane binding under those conditions (for EPR spectra see Figure 3-S5B). This suggests that electrostatic interactions modulate the binding event. Next, we evaluated how membrane binding depends on the molar fraction of the negatively charged POPS lipids. Increasing amounts of POPS strongly enhanced membrane binding, but it should be noted that clearly detectable (40%) binding was also observed for vesicles containing the net neutral 100% POPC (Figure 3-4C, for EPR spectra see Figure 3-S5C). Collectively, these experiments show that charge interactions significantly contribute to the membrane interaction of 67 Figure 3-5. Phosphomimetic (S13D/S16D) mutations decrease membrane-binding affinity of Httex1(Q25) and protect Httex1(Q46) from membrane-mediated aggregation (A) The central line amplitudes of the EPR spectra for the 21R1 derivative at 15 mM are shown as function of lipid concentration. Data from proteins containing S13/S16 are shown in black, while those from S13D/S16D are shown in red. All spectral amplitudes were normalized relative to the spectrum in the absence of lipid, which was set to 1. Error bars show ±standard deviation of independent repeats (n = 3). For EPR spectra, see Figures S6A and S6B. (B) Changes in CD mean residual ellipticity Δ[θ] MRE222nm were plotted against lipid concentration, colored as in (A). Error bars show ±standard deviation of independent repeats (n = 3). For CD spectra, see Figures S6C and S6D. (C) For monitoring aggregation kinetics of Httex1(Q46), EPR spectra were recorded for 15 mM Httex1(Q46) 35R1 with or without S13D/S16D mutations in the presence or absence of 375 mM 25% POPS and 75% POPC SUV lipids. The central line amplitudes were normalized by setting the t0 amplitude to 1, and the resulting values from 1 h to 2 h post mixing were plotted for comparison (S13/S16 in black and S13D/S16D in red). Unlike the Httex1(Q46) 35R1 group, no statistically significant membrane-mediated aggregation was observed for Httex1(Q46) 35R1 with S13D/S16D mutations (t tests, **p < 0.01, n = 3). Error bars show standard deviation of independent repeats (n = 3). All data analyses were performed in GraphPad Prism 7.0. For a more complete time course, see Figure S6E. 68 Httex1, but that there are additional energy contributions that do not require a net negative membrane surface charge. 3.2.6 Phosphomimetic mutations in N17 potently decrease the membrane binding affinity of Httex1(Q25) and protect Httex1(Q46) from membrane-mediated aggregation. Having found that charge interactions contribute to the membrane interaction of Httex1(Q25), we next wanted to know how the introduction of negative charges in the N17 modulates membrane interactions. A well-studied alteration of N17 charge is phosphorylation at positions S13 and S16. This modification has frequently been mimicked by S13D/S16D mutations, which introduce two additional negative charges into the N17. Phosphomimetic (S13D/S16D) mutations have previously been shown to reduce huntingtin toxicity in cell and animal models of HD (Di Pardo et al., 2012; Gu et al., 2009; Mishra et al., 2012). To test whether the additional negative charges at positions 13 and 16 affected membrane interaction, we performed lipid titration experiments where binding was monitored using EPR and CD as described above. In these experiments, the 21R1 derivatives of Httex1(Q25) or Httex1(Q25) S13D/S16D were incubated with SUVs at varying lipid concentrations (up to 20 mM). As shown in Figure 3-5A (for EPR spectra see Figure 3-S6A and S6B), the phosphomimetic mutations reduced membrane binding affinity, as indicated by the ~ 15 times higher amounts of lipid required for saturation of membrane binding. The reduced membrane binding propensity of the S13D/S16D mutant was further verified by CD, where lipid addition produced much smaller changes in the mean residual ellipticity Δ[θ] MRE222nm (Figure 3-5B, for CD spectra see Figure 3-S6C and 3-S6D). Thus, the EPR and CD data support the notion that the S13D/S16D mutations reduced membrane binding. For the non-pathogenic Httex1(Q25) discussed thus far, we find no evidence for oligomerization or aggregation. However, we recently reported that interaction with negatively charged membranes promotes aggregation and fibril formation in case of the pathogenic, long Q-length containing Httex1(Q46) (Pandey et al., 2018). Inasmuch the S13D/S16D mutations attenuate membrane interaction of Httex1(Q25), we sought to test how these mutations affect membrane-mediated aggregation. In order to monitor misfolding of Httex1(Q46) in the presence of membranes, we again used the previously developed EPR- 69 based readout, which monitors the β-sheet formation of the polyQ region by tracking the spectral amplitude decrease of 35R1. A faster reduction in the signal amplitude corresponds to faster rates of aggregation. First, we verified that the vesicles used in the present study (SUVs) could potently accelerate the aggregation of Httex1 for its long Q-length variant (Q46). In the presence of SUVs, we found that the EPR amplitudes decayed more rapidly than in the absence of SUVs (Figure 3-5C, for a more complete time course see Figure 3-S6E). In contrast, the aggregation of 35R1 Httex1(Q46) with the S13D/S16D mutations was less sensitive to the addition of liposomes. After 1 or 2 hours, the EPR amplitudes were very similar regardless of whether lipids were added or not. It should also be noted that the S13D/S16D mutations slowed down the aggregation of Httex1(Q46) in solution, which is in agreement with previous studies (Gu et al., 2009; Mishra et al., 2012). The minor effect of membranes on promoting the aggregation kinetics of Httex1(Q46) with S13D/S16D mutations compared to Httex1(Q46) was consistent with the aforementioned attenuation of membrane interaction caused by the S13D/S16D mutation. Together these data indicate that the S13D/S16D mutations are generally inhibitory for solution or membrane-mediated aggregation pathways. 3.3 Discussion Here we investigated the structure of Httex1 stably-bound to intact phospholipid membranes using a combination of biophysical techniques. We found that negative membrane charge and strong membrane curvature (SUV) led to strong binding. Inasmuch as Httex1(Q25) was stably bound without aggregating into misfolded species, it was possible to systematically investigate its structural features, including the local secondary structure and the membrane proximity of the protein. As schematically illustrated in Figure 3-6A, we found that the N-terminal portion of the N17 forms a short amphipathic α-helix extending approximately from residue 3 to 13. This α-helix is largely located in the headgroup region. This region was C-terminally flanked by a transition region that extends from residue 14 to 22. While this region did not exhibit the characteristic helical structure of the preceding region, it nonetheless became less dynamic (presumably partially structured) and remained in the proximity of the headgroup region of the lipid membrane. For the more C-terminal residues, the EPR spectra of R1-labeled Httex1(Q25) in solution and 70 bound to membranes became more and more superimposable, suggesting that these regions did not experience a significant structural change upon membrane interaction. Altogether we found that the changes in structure and dynamics of Httex1 were predominantly within the first 22 residues. This region encompasses not only the N17, but also the first five Gln residues. This perhaps somewhat unexpected involvement of residues in the polyQ region shows that shorter peptides, which only contain the N17, may not be ideal models for studying Httex1 membrane interaction. Our data on Httex1(Q25) also indicated that negative membrane charge and positive membrane curvature (outside surface of highly curved SUVs) significantly promote Httex1 membrane interaction (Figure 3-4). This binding behavior can be rationalized by the structure and the increased headgroup spacing in case of positive curvature (Figure 3-6B). Highly curved membranes, such as those in an SUV have a much higher outer than inner surface area. This means, a headgroup for a given lipid in the outer leaflet has to cover a much larger surface area than the acyl chain tails. This reduced packing density in the outer leaflet increases progressively with increasing distance away from the membrane interior. As a consequence, positively curved membranes are well known to have packing defects in the headgroup region, which can cause water penetration into the acyl chain region. By binding higher up in the headgroup region, the Httex1(Q25) amphipathic helix is well-situated to fill the voids in the headgroup region, thereby reducing packing defects. This idea is furthermore consistent with computational studies showing that a headgroup location of an amphipathic helix favors positive rather than negative membrane curvature (Campelo et al., 2008). The importance of negatively charged lipids in the membrane interaction of Httex1 is consistent with the predominantly positively charged characteristics of the N17. The N17 is overall positively charged, containing three lysines (K6, K9, and K15) and two glutamic acids (E5 and E12). The location of the charged residues also likely impacts the structure of the membrane-bound state. K6 and K9 as well as E5 and E12 are located within the amphipathic helix. As is commonly observed for membrane-bound amphipathic helices, all these residues are located on the hydrophilic surface. In contrast, K15 is just 71 Figure 3-6. Schematic illustration of Httex1 membrane interaction (A) The interaction of Httex1(Q25) with curved membranes is shown. PolyQ and PRD domains are colored brown and green, respectively. N17 is blue with some residues highlighted. Selected solvent- exposed residues (E5 and K9) are highlighted in cyan, while selected membrane-exposed residues (L4, L7, and L11) are in red, phosphorylation residues (S13 and S16) are in yellow, and Q22 is in magenta. The a-helical structure for residues 3–13 is shown by ribbon representation. The C terminus of N17 and the N terminus of polyQ domain comprise the transition region (14–22) and the rest of the protein (residues 23 onward) is completely solvent exposed. (B) Illustration of how the N-terminal Httex1 helix (red circle) preferentially interacts with curved membranes. Vesicles with positive membrane curvature have greater lipid-packing defects in the head- group region, allowing the helix to more readily bind to membranes. (C) Membrane-mediated aggregation of long Q-length Httex1 illustrated with Httex1(Q46). Monomeric Httex1(Q46) collisionally encounter each other on the membrane, and the polyQ regions from different molecules come into contact and initiate intermolecular b-sheet formation, ultimately leading to the formation of fibrils and aggregates (color code of domains is as in A). 72 flanking the helical region. Interestingly, K15 would be located on the hydrophobic surface of the helix if the helix were to be further extended. This would be a very uncommon position for a Lys residue and could explain why the helical region does not extend all the way to K15. Although K15 is outside the helix, it is nonetheless in the interfacial region, where it can interact with the negatively charged lipid headgroup and promote membrane interaction. The notion that positive charges on the N17 promote membrane interaction is further supported by the observation that acetylation of the N17 Lys residues reduces membrane-binding affinity (Chaibva et al., 2016). Conversely, N17 phosphorylation at residues S13 and S16 would introduce two negatively charged moieties into the interfacial region. While this location avoids exposure to the non- polar region of the membrane, bringing additional negative charges into the net negatively charged interfacial region would nonetheless be expected to be unfavorable. Here we mimicked phosphorylation by introducing S13D/S16D mutations, which are well characterized, and which have been shown to block toxicity in cell and animal models (Di Pardo et al., 2012; Gu et al., 2009; Mishra et al., 2012). As expected from their locations, the mutations potently inhibited membrane binding by an order of magnitude. This effect is likely even more significant in case of actual phosphorylation. While phosphomimetic mutations as well as phosphorylation introduce negative charge, there are some chemical differences. Depending on the pH, a phosphorylated side-chain can have up to two charges, whereas an Asp side chain only has one. The larger number of negative charges introduced with phosphorylation could therefore reduce membrane affinity even more. One way of accommodating an acidic side chain upon membrane interaction is to protonate it. While one of the negative charges on a phosphate can be protonated at mildly acidic pH, the second charge has a pKa that is much lower than that of Asp. Thus, it would be more difficult to protonate the second charge of a phosphate as compared to that of an Asp. While we did not directly compare our results to the phosphorylated Httex1, a recent study found that phosphorylation of S13 and S16 was indeed more potent at inhibiting membrane interaction than the phosphomimetic mutants for N-terminal peptides (Deguire et al., 2018). Despite the significant importance of electrostatic interactions in membrane binding of Httex1, other factors are likely to contribute as well. We conclude this from the fact that interaction with 73 uncharged membranes is still possible (albeit strongly reduced). This residual binding is likely driven by hydrophobic residues such as L4, L7, and F11, which are facing toward the membrane. A common feature in all the EPR spectra of membrane-bound Httex1(Q25) was the lack of strongly immobilized components (Figure 3-S2), which would have been expected in case of tertiary or quaternary contacts. This means that Httex1(Q25) is largely monomeric under the present conditions, and that it does not contain any globular structures. This behavior was reminiscent of what we observed in the early spectra of membrane-bound Httex1(Q46) just prior to aggregation and fibril formation, as these spectra also lacked strongly immobilized components immediately after the addition of lipids (Pandey et al., 2018). This means that membrane binding did not immediately induce an oligomeric state, but that oligomerization was a subsequent step. Oligomerization was found to be polyQ-length dependent, as shorter Q-lengths, like the Httex1(Q25) studied here, did not exhibit any evidence of oligomerization or aggregation. How then could the largely monomeric, membrane-bound state facilitate aggregation of Httex1(Q46)? Compared to the inefficient, three-dimensional diffusional encounter of monomeric proteins in solution, Httex1 molecules can more readily find each other when diffusing in two-dimensions on the membrane surface. This reduced dimensionality is thought to be a major driver of membrane-mediated aggregation of various amyloidogenic proteins (Jayasinghe and Langen, 2007; Rawat et al., 2018). For Httex1, the N-terminal helix is the primary membrane anchor that promotes two-dimensional diffusion. The contacts between multiple Httex1 molecules could then initiate intermolecular β-sheet formation in the polyQ region, ultimately leading to fibril formation (Figure 3-6C). It is also possible that membrane binding causes a conformational change in the monomer structure, especially in the transition region. In addition, formation of β-sheet structure in the polyQ region during aggregation could be further promoted by the relatively low dielectric constant of the interfacial region, which can favor the formation of secondary structure. It is also interesting to note that the membrane-bound α-helix does not have to be converted into β-sheet structure upon fibril formation. In fact, the helical region in the membrane-bound state (residues 3 to 13) closely corresponds to the α- helical region found in fibrils (residues 4 to 11) (Sivanandam et al., 2011). 74 Our prior study identified parallel pathways (in solution and on membranes), by which Httex1 with expanded Q-lengths can aggregate into fibrils. It is therefore possible that an effective therapeutic strategy might have to take all of these pathways into account. The phosphomimetic mutations have previously been shown to be highly protective against huntingtin toxicity (Di Pardo et al., 2012; Gu et al., 2009; Mishra et al., 2012) and they can block aggregation in solution as well as on membranes. Considering the importance of the N17 for aggregation in both cases, targeting the N17 by binders or post-translational modifications might therefore be a powerful strategy for blocking multiple aggregation pathways in the cell. 75 Figure 3-S1. Httex1 domain organization, R1 labeling sites, and analysis of vesicle as well as protein stability over time (A) Httex1 are composed of three domains: N17, polyQ and PRD. The table shows the domain organization for Httex1 using the residue numbering for Httex1(Q25). The sequence for the individual domains is given in the second column and the sites at which R1-labeled side chains were introduced, one amino acid at-a-time are given in red. The residue number of R1-labeled site is provided in the third column. EM images show no significant morphological changes for SUVs (B) before or (C) after seven hours of incubation with wild type Httex1(Q25) (protein-to-lipid ratio 1:1300). Also, no visible protein aggregates are found in (C). (D) EPR spectra overlay shows that Httex1(Q25) 35R1 spectra are the same at 0 minute (black) or after seven hours of incubation with SUVs (red) indicating that no fibril formation or aggregation occurred in this time frame. 76 Figure 3-S2. CW-EPR spectra of singly R1-labeled Httex1(Q25) stably bound to SUVs All vesicles contained 25% POPS and 75% POPC. Spectra are grouped by the domain, in which the respective spin labels were introduced (N17 domain (A), polyQ domain (B) and PRD domain (C)). Some spectra were scaled as indicated. 77 Figure 3-S3. EPR-based accessibility measurements (A) Oxygen and NiEDDA accessibilities (ΠO 2 (red) and ΠNiEDDA (blue)) are plotted as function of labeling position. Error bars show ±standard deviation of independent repeats (n = 3). (B) Zoom-in of (A) highlights the out-of-phase oscillation between ΠO 2 and ΠNiEDDA. To better distinguish between the two different accessibility values, they are plotted on different scales using the same color code as in (A). 78 Figure 3-S4. k𝜎(P) fitting and site-specific kσ of Httex1(Q25) (A) The strip value of each row is the sequence of independent repeats. The strip value of each column is the site at which Httex1(Q25) was labeled. Black points show the recorded k𝜎(P) plotted against varying power P (Watt). Data of each experiment was fitted using equation k σ(P) = k σP/(P+constant). Red lines show the fitted results. (B) ODNP data table. Errors are standard deviations from independent repeats (n = 3). 79 Figure 3-S5. EPR spectra for membrane curvature, salt and lipid composition dependence of membrane binding Representative X-band EPR spectra showing the effects of (A) decreasing membrane curvatures, (B) increasing NaCl concentrations, and (C) increasing POPS percentages. The respective conditions are indicated in the figure. All spectra were normalized by double integration to the same number of spins. 80 Figure 3-S6. EPR and CD spectral analyses of the S13D/S16D mutations on Httex1(Q25) membrane binding and EPR time course of Httex1(Q46) aggregation EPR spectra were recorded for 15 μΜ Httex1(Q25) 21R1 (A) S13/S16 and (B) S13D/S16D in the presence of increasing concentrations of 25% POPS and 75% POPC SUV lipids. All spectra were normalized by double integration to the same number of spins. The corresponding CD spectra for (C) S13/S16 and (D) S13D/S16D are given using the mean residual ellipticities [θ] MRE222nm. (E) For monitoring aggregation kinetics of Httex1(Q46), EPR spectra were recorded for 15 μΜ Httex1(Q46) 35R1 with (red) or without (black) S13D/S16D mutations in the presence (solid circles) or absence (empty circles) of 375 μM 25% POPS and 75% POPC SUV lipids every 15 minutes over a time course of two hours. Error bars show ±standard deviation of independent repeats (n = 3). The central line amplitudes were normalized by setting the t 0 amplitude to 1 and plotted over time. Note, the error bars for S13D/S16D in the presence of SUV are smaller than the symbols used. 81 3.4 Materials and Methods 3.4.1 Preparation of R1-labeled Httex1 Derivatives The native amino acid sequence of human Httex1 fragment contains no cysteine. Cysteine mutations as well as phosphomimetic (S13D/S16D) mutations were introduced either by Q5® Site-Directed Mutagenesis Kit (E0554S, NEB) or NEBuilder ® HiFi DNA Assembly Cloning Kit (E5520S, NEB) based on templates previously synthesized by GenScript. All plasmids were verified by DNA sequencing. Monomeric Httex1 fragments were expressed from pET-28b(+) plasmid as a Trx-Httex1 fusion protein, which had a thioredoxin (Trx) fused to the N-terminus of Httex1. This Trx-Httex1 fusion protein was purified, spin-labeled and cleaved to separate Httex1 from Trx using a previously described protocol (Pandey et al., 2018). The only modification was that the present Trx-Httex1 contained a 6xHis-tag at the N-terminus of the Trx tag rather than the C-terminus of Httex1. This was done to ensure that the 6xHis-tag could later be cleaved off together with the thioredoxin fusion partner (Pandey et al., 2018). As in the prior study, the resulting Httex1 was lyophilized after the final purification step, dissolved using 0.5% TFA (v/v) in methanol and dried completely under gentle nitrogen flow. The resulting protein film was solubilized either directly with buffer or by SUVs prepared in the same buffer. 3.4.2 Preparation of Small and Large Unilamellar Vesicles 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC) and 1-palmitoyl-2-oleoyl-sn- glycero-3-[phospho-L-serine] (POPS) were purchased as chloroform solution from Avanti Polar Lipids, Inc (#850457C, #840034C). SUVs and LUVs of varying POPS to POPC ratios were prepared in the respective buffers used for studying Httex1 interaction. For all vesicle preparations, lipids were first mixed and dried under a gentle nitrogen flow to form a thin layer of lipid film and placed at room temperature in a desiccator overnight. For preparation of SUVs, lipid films were resuspended in buffer and sonicated until clear. LUVs (25% POPS and 75% POPC) were made by freeze-thawing the solubilized lipid suspension for 10 cycles, followed by 21 times of back-and-forth extrusions through either 200 nm or 1000 nm Nucleopore Track-Etch Membrane (Whatman ® #800281, #800319) using glass syringes. For testing the effect of negatively charged lipids on Httex1 membrane binding, SUVs were prepared containing 0, 5, 10, 82 25 and 50% POPS and the remaining lipids were POPC. For all other experiments, 25% POPS and 75% POPC were used. The buffer was 20 mM Hepes, 100 mM NaCl, pH 7.4 except in the following cases: For NiEDDA accessibility measurement, SUVs were prepared with an additional 3 mM NiEDDA. For testing the effect of ionic strength on Httex1 membrane binding, SUVs were prepared in 20 mM Hepes, pH 7.4 containing increasing concentrations of NaCl (0, 100, 500, 1000 mM). SUVs for circular dichroism were prepared in 20 mM sodium phosphate, pH 7.0. No additional salt was added to avoid high absorbance in the UV. For consistency, the same SUVs were used for EPR lipid titration and EPR kinetics experiments. 3.4.3 Transmission Electron Microscopy Electron microscopy was used to verify SUV size and shape and to detect potential protein aggregates. 10 mM SUVs were pre-spun for 10 minutes at 15,000 rpm. The vesicle suspension (supernatant) was used to solubilize Httex1(Q25) to a concentration of 7.5 μΜ. SUVs alone as well as the SUV/Httex1(Q25) mixture were then incubated at room temperature. After seven hours, the samples were transferred to 150 mesh carbon coated copper EM grids (#FCF-150-Cu, Electron Microscopy Sciences). The protein-containing sample was incubated for 10 minutes while the SUV only sample was incubated for 1 hour to compensate for the poor absorption in the absence of proteins. Both samples were subsequently stained with 2% uranyl acetate for 1 minute. The grids were then examined under JEOL JEM-1400 transmission electron microscope operated at 100 kV and detected using a Gatan ORIUS TM CCD camera. 3.4.4 EPR Measurements X-band continuous-wave EPR spectra for all R1-labeled Httex1(Q25) derivatives were recorded in a Bruker EMX spectrophotometer fitted with a Bruker ER4119HS resonator. For spectra of Httex1 in the absence of lipids, 15 μΜ protein was solubilized in buffer (20 mM Hepes, 100 mM NaCl, pH 7.4). For spectra of Httex1 in the presence of lipids, Httex1 film was solubilized directly in 10 mM SUVs to a final protein concentration of 7.5 μΜ. The vesicle-bound proteins were then harvested by ultracentrifugation at 60,000 rpm, 25 °C for 30 minutes (Beckman TM TLX Ultracentrifuge). Samples were then loaded into round borosilicate glass capillaries (#CV6084-B-100, VitroCom Inc.) and the respective X-band EPR spectra were recorded at room temperature using a scan width of 100 G and a modulation amplitude of 1.5 G. 83 EPR was also used to quantify Httex1 membrane binding based on the distinctively different spectra for the bound and unbound states of the 21R1 Httex1(Q25) derivative. For these experiments, Httex1 and membranes were not harvested by ultracentrifugation. Rather, the R1-labeled proteins were directly solubilized in the indicated vesicles using Httex1 concentrations of 15 μΜ. All spectra were normalized by double integration to the same number of spins. The assignment of percent binding was calibrated using the normalized spectra of the fully bound and fully unbound states as reference. This was possible because all spectra were the composite mixture of varying amounts of the same spectral components arising from bound and unbound protein. Binding percentage was then estimated based on the linear amplitude decay of the normalized EPR spectra that occurs as membrane binding immobilizes 21R1. Here the spectral amplitude of the free protein (in the absence of lipid) corresponded to 0% membrane binding while the spectral amplitude of the fully bound form corresponded to 100% binding. O 2 and NiEDDA accessibility measurements were performed using EPR power saturation. Samples were prepared using ultracentrifugation as described above to remove any potential unbound protein and loaded into a TPX capillary. EPR spectra at different incident powers were recorded using a Bruker EMX spectrophotometer fitted with a Bruker dielectric resonator (ER4123D) at room temperature (Altenbach et al., 1994). For O 2 accessibility measurements, the power dependence of the EPR spectra was first determined in the presence of O 2 from air. Then the same sample was exposed to a continuous flow of N 2 and the EPR experiments were repeated. For NiEDDA accessibility measurements, power saturation experiments were performed in the presence of 3 mM NiEDDA and N 2 flow. The respective power saturation data were converted into O 2 and NiEDDA accessibility parameters, and Φ values according to previously published methods (Altenbach et al., 1994). To convert the Φ values into immersion depth, calibration experiments were performed using SUVs containing 1% spin-labeled lipids (Avanti Polar Lipids): 1-palmitoyl-2-oleoyl-sn-glycero-3- phospho(tempo)choline (#810609C), 1-palmitoyl-2-stearoyl-(n-DOXYL)-sn-glycero-3-phosphocholine (n = 5, 7, 10; #810601C, #810602C, #810603C). The relationship between membrane immersion depth and 84 the depth parameter Φ was established previously (Altenbach et al., 1994) as d[Å] = aΦ + b. We found that under the current experiment conditions a = 9.6 and b = -9.7. EPR kinetics experiments were performed by solubilizing 15 μΜ 35R1 derivative of Httex1(Q46) with or without S13D/S16D mutations either directly with buffer or with 375 μM SUV lipids. The EPR spectra were then recorded at a time interval of 15 minutes for a period of 15 hours. The spectral amplitudes were normalized to that of the initial t = 0 spectrum, which was set to 1. 3.4.5 ODNP Measurements Six singly R1-labeled Httex1(Q25) derivatives (11R1, 22R1, 30R1, 42R1, 60R1 and 90R1) were selected for ODNP measurement. 7.5 μΜ protein was mixed with 10 mM SUVs and then concentrated using 100 kDa centrifugal filters (Amicon Ultra #UFC510096) followed by buffer washing (x 3 times) to ensure removal of potential unbound protein. Samples of 3.5 μl volume were filled into a 0.6 mm i.d and 0.84 mm o.d. quartz capillary to acquire ODNP data, as described previously (Franck et al., 2013; Kaminker et al., 2015). A “pass through” NMR probe that fits inside a 3 mm i.d., 6 mm o.d. quartz tube was used for the measurements. The quartz tube along with the NMR probe and sample was inserted into a microwave cavity (ER 4119HS-LC, Bruker Biospin), while ensuring that the position of the sample within the cavity was reproducible between measurements. The ODNP experiments were performed using a Bruker EMX CW-EPR and a Bruker Avance III NMR console. The samples were sealed in a capillary with Critoseal on top and beeswax on the bottom. The sample was irradiated with up to 6 W of microwaves at the EPR resonant frequency of the R1 label (Armstrong et al., 2008; Franck et al., 2013). The power to the microwave resonator was sampled by a 20 dB directional coupler (Narda 4015C-20) and further attenuated by a 10 dB attenuator (Narda 4778-10) and Coax microwave cable which was approximated to have a loss of 2 dB. The sampled power was measured with a Gigatronics 8541C power meter and 80401A power sensor. The magnetic field was set on resonance at the central electron hyperfine transition, here at 9.8 GHz. The R1 label concentration of each sample was determined from the double integral of its CW EPR spectrum. The ODNP-derived electron-proton spin cross relaxivity, k σ, was normalized to the sample concentration derived from spin counting per integration of the CW-EPR spectrum. The ODNP data 85 yielding the k σ values are shown in the table of Figure 3-S4B. All ODNP data were measured in triplicates, constituting the error bars in the ODNP data derived k σ values. The theory behind the physical basis of k σ values, as well as the data analysis to obtain k σ values and retardation factors from ODNP data is described in the literature (Barnes et al., 2017). 3.4.6 Circular Dichroism Circular dichroism (CD) measurements were performed at room temperature using a Jasco J-810 spectropolarimeter in a 0.1 mm quartz cuvette every 0.5 nm at a 50-nm/min scan rate. The scan wavelength range was from 200 nm to 260 nm. Experiments were done in triplicates and all spectra were baseline subtracted. Data were converted into mean residual ellipticity [θ] MRE222nm. 86 Chapter 4: The N-terminus of Membrane Protein Annexin A7 is a Functional Domain Mediating Liquid-Liquid Phase Separation and Vesicle Clustering Abstract ANXA7 is a stress-related membrane protein with implicated functions in autophagy, mitophagy and cancer. The C-terminal core domain of ANXA7 is a renowned motif mediating Ca 2+ -dependent membrane binding. The protein also contains a long low-complexity N-terminal domain that is predicted to be structurally disordered. Here, we show that endogenous ANXA7 in the cell assembles into granular structures under arsenite stress. Upon overexpression, the N-terminal domain mediates self-association of the protein into cellular aggregates. Further biochemical studies revealed that ANXA7 can undergo liquid- liquid phase separation in vitro, another event mediated by the N-terminal domain and is sensitive to pH and salt concentrations. Protein droplets derived from the N-terminus alone mature into fibril-like filamentous structures over time under low pH and low salt conditions. ANXA7 droplets contain highly mobile components as verified by FRAP analyses. Full-length ANXA7 can bind to and cluster PS- containing lipid vesicles in the presence of Ca 2+ . N-terminus truncation did not affect the protein’s ability to bind membranes but largely impaired its normal function in vesicle-clustering. The dual functions of the N-terminal domain in liquid-liquid phase separation and Ca 2+ -mediated membrane clustering of ANXA7 have great physiological significance. 4.1 Introduction Annexin A7 (ANXA7, also referred to as ANX VII or synexin) is a membrane protein that belongs to the annexin protein superfamily, notable for their involvement in calcium signaling and membrane trafficking (Gerke et al., 2005; Hawkins et al., 2000; Moss and Morgan, 2004). First isolated from human adrenal medulla (Creutz et al., 1978), ANXA7 was found to promote Ca 2+ -mediated aggregation of chromaffin granules. Since then, more and more important functional roles have emerged, including Ca 2+ homeostasis (Clemen et al., 2003; Clemen et al., 2001), plasma membrane repair (Sonder et al., 2019), autophagy (Huang et al., 2014; Wang et al., 2010), cancer (Torosyan et al., 2006; Torosyan et al., 2009; 87 Yuan et al., 2019b) and mitophagy (Meng et al., 2020). Noticeably, many of these roles are membrane- related. The basic structure of ANXA7 is composed of two domains: An N-terminal signature domain and a C-terminal core domain. The latter is comparatively well-studied because it is a renowned membrane binding motif that is conserved across the entire annexin protein superfamily. Enriched in α-helix, the C- terminal core domain has a defined structure containing four repeats and can undergo reversible binding to phospholipid-containing membranes in a Ca 2+ -dependent manner (Lizarbe et al., 2013). The N-terminal domain, on the other hand, is what differs ANXA7 from other members of the protein superfamily. It lacks a defined structure and its function is enigmatic. First of all, the N-terminal domain is much longer in ANXA7 than in most other members of the annexin superfamily, accounting for ~1/3 of the total protein length. In addition, it is deficient in charged residues, only two Asp and two Glu, three out of which are restricted within the isoform-specific sequence, leaving only one Glu in the smaller isoform. In fact, the N-terminal region is compositionally biased to contain only a limited number of amino acids. And even more interestingly, the sequence follows a particular pattern of arrangement in that aromatic residues (Tyr/Phe) are repetitively followed by prolines. Historically, theoretical models have been proposed for this type of sequence, named as pro-β helices (Matsushima et al., 1990). It was predicted that the characteristic sequence is encoding a protein structure consists of repeating segments of the polyproline helix interrupted by β turns. However, experimental evidence supporting these models are lacking and the molecular structure of the N-terminus remains undetermined. In general, however, low complexity sequence (i.e. high composition bias) is associated with a high inclination to form disordered structures. But ordered structures can also form in case of sequence repeats (Mier et al., 2019). And structural characterization of the flexible and disordered protein regions is often times challenging. Even though the N-terminal domain is of great structural ambiguity, its biological function still attracted much attention. It has been reported that the domain can establish reversible interactions with many other proteins (Matsushima et al., 2008) and a number of recognition sites have been successfully 88 identified. For example, the first 31 amino acids (containing the GYPP tandem repeat) are believed to mediate its Ca 2+ -dependent interaction with Sorin (Brownawell and Creutz, 1997; Verzili et al., 2000). SODD and KIAA0280 are also interacting with ANXA7 by targeting the N-terminus (Creutz, 2009). Apart from acting as a loading dock for binding partner recognition, the N-terminal domain may also modulate the membrane binding. The first 29 amino acids have been suggested to potentially influence membrane binding through interacting with the C-terminal end of the core domain (Naidu et al., 2005). And removal of a ten residue domain (Y11-A20) within the N-terminal domain has been suggested to diminish membrane binding (Chander et al., 2006). Furthermore, the N-terminus is implied in a TEM study to facilitate the self- association of ANXA7 into rod-like assemblies, a process that can be accelerated by Ca 2+ (Creutz et al., 1979). But in spite of all these attempts and efforts, the main functional role of the N-terminal domain remains elusive. Interestingly, ANXA7 was detected in a proteomics study on stress granule interactome (Markmiller et al., 2018). Stress granules (SGs) are liquid-like condensates formed in response to certain environmental stress (e.g. toxins, heat shock and nutrition deprivation) (Franzmann and Alberti, 2019). Formation of SGs promotes cell survival (Mahboubi and Stochaj, 2017) through several mechanisms. For instance, it can help cells better adapt to stress by regulating gene expression (Franzmann and Alberti, 2019; Namkoong et al., 2018) or inhibit apoptosis by reducing the level of ROS species (Takahashi et al., 2013). Underlying all these mechanisms is the ability of SGs to recruit or deplete selective cellular species, among which are nontranslating mRNAs and RNA binding proteins (Jain et al., 2016; Khong et al., 2017). Similar to many other membraneless organelles, SGs are generated spontaneously through a phenomenon known as liquid-liquid phase separation (LLPS) (Boeynaems et al., 2018; Gomes and Shorter, 2019; Wheeler et al., 2016). Emerging biological roles have been discovered for LLPS, including transcription regulation (Guthmann et al., 2019), protein quality control (Frottin et al., 2019) and the formation of skin barrier (Quiroz et al., 2020). Here, we would like to emphasize that LLPS is observed frequently in proteins with low-complexity sequences (Martin and Mittag, 2018; Statt et al., 2020). 89 Thus far, we learned that ANXA7 was both detected in the stress granule interactome and has a low-complexity N-terminal domain. Taken together the two facts, we hypothesized that the protein may undergo LLPS. Using cell models and biophysical tools, we find that the N-terminal domain promotes the self-association of ANXA7 in the cell and mediates the protein phase separation in vitro, a process sensitive to pH and salt concentrations. As mentioned earlier, ANXA7 is involved in a number of membrane-related functions and can promote Ca 2+ -dependent aggregation of lipid vesicles (e.g. chromaffin granules) (Creutz et al., 1978), the molecular basis of which was not fully understood. Here, we have successfully showed that while as the C-terminal core domain is sufficient for Ca 2+ -dependent membrane binding, the N-terminal domain plays a key role in bridging the membrane surfaces. Collectively, our study identified the N-terminal domain of ANXA7 as a functional motif in two different biological processes: liquid-liquid phase separation and vesicle clustering. This important finding provides new mechanistic insight into the many membrane-related, stress-related physiological functions of the protein. More importantly, ANXA7 may represent one class of membrane proteins with dual motifs linking together the process of phase separation and membrane interaction. 4.2 Results Phase separation (into liquid droplets) is a phenomenon frequently observed for proteins containing low-complexity domains, corresponding to their high propensity to be structurally disordered (Martin and Mittag, 2018; Statt et al., 2020). Therefore, before testing our hypothesis systematically, it would be helpful to first examine the ANXA7 protein sequence. In order to identify the presence and range of the disordered regions within ANXA7, we subjected the entire protein sequence to bioinformatic analysis using PrDOS online server (see Figure 4-S1). As shown, the N-terminal region was predicted to be highly disordered from residue M1 to T170 (highlighted in red, > 0.5 disorder probability). Only the last six amino acids (VTQVTQ) close to the transition region exhibited a minor tendency to become ordered. And the rest of the protein was predicted to be highly structured, consistent with prior reports that it is a conserved functional motif with defined structure. Further analysis of the N-terminal amino acid composition revealed 90 that the sequence consists of a very limited number of amino acids (e.g. Pro/Gly/Ser/Tyr) (see Figure 4- S2). In other words, this domain is compositionally biased and is of low-complexity. The results of the sequence analysis further strengthened our initial hypothesis that ANXA7 might be able to undergo phase separation. The N-terminal domain, prone to take on disordered conformations and being low-complex, could play a crucial role in driving the phase separation. We begin our systematic analysis by first verifying the recruitment of ANXA7 into stress granules. 4.2.1 Endogenous ANXA7 is recruited into granular structures upon arsenite-induced cellular stress In order to compare the cellular localization of ANXA7 under normal conditions and under conditions of cellular stress, HEK293T cells were treated with 0.5 mM NaAsO 2 for 30 minutes under its normal growing conditions (37 °C supplemented with 5% CO 2). Similar conditions were commonly used in the literature to trigger the formation of stress granules (Sfakianos et al., 2018; Wheeler et al., 2016). Endogenous ANXA7 within the cell was probed by anti-ANX VII antibody and imaged under confocal fluorescence microscope. Representative images are shown in Figure 4-1A and 4-1B. Under normal conditions, cells showed diffused expression pattern of ANXA7 in the entire cytoplasm (Figure 4-1A). In contrast, under arsenite stress, ANXA7 are found in granular structures within the cytoplasm (Figure 4- 1B). The formation of those granular structures was visible both in differential interference contrast (DIC) and in the green channel, with morphologies highly similar to the stress granules reported previously. We then further quantified the efficiency of ANXA7 recruitment into the granular structures induced by the arsenite treatment. For each group, numbers of total cells (counted by DAPI nucleus staining) and the cells with ANXA7 positive granular structures (counted by green fluorescence signals) were collected for five random frames each. And the percentage of cells containing ANXA7 positive granular structures were calculated. As plotted in Figure 4-1C, the difference is statistically significant. 4.2.2 Overexpression of ANXA7 variants containing the N-terminal sequence leads to the formation of cellular aggregates. In order to evaluate the tendency of ANXA7 to self-associate at high cellular concentrations. Three mammalian expression constructs were created, respectively encoding three different variants (see 91 Materials and Methods and the cartoon illustrations in Figure 4-2). For clarity, they are denoted as ANXA7 (encoding full-length ANXA7), ANXA7Nt (encoding the N-terminal domain residues M1-Q176 of ANXA7) and ANXA7ΔNt (encoding the C-terminal core domain of ANXA7 where S2-Q176 has been removed). All variants are eGFP-tagged at the C-terminal end of the protein fragment to allow fluorescence detection. Transfection of the constructs in HEK293T cells leads to drastically different overexpression patterns (see Figure 4-2). GFP positive cellular protein aggregates were observed for the ANXA7 and ANXA7Nt variant but not for the ANXA7ΔNt variant, where cytoplasmic diffused expression pattern was detected. Additionally, there are also morphological differences between the ANXA7 and ANXA7Nt variants. For ANXA7, the aggregates were overall larger in size and located mostly within the cytoplasm. In fact, the morphology is reminiscent of the rod-like structures previously reported in an electron microscopy study (Creutz et al., 1979). For ANXA7Nt, however, the aggregates were on average much smaller and nuclear localization were frequently observed with a tendency to exclude the DAPI positive chromatin, reminiscent of the nuclear inclusions. Despite these differences, it is quite obvious that ANXA7 does have a tendency to self-associate at high cellular concentration and the N-terminal domain is required for it to happen. 4.2.3 Recombinant ANXA7 undergoes pH- and salt-sensitive liquid-liquid phase separation in vitro, mediated by its low-complexity N-terminus To test whether ANXA7 could undergo LLPS in vitro, three bacteria expression constructs were created for the three ANXA7 variants: ANXA7, ANX7Nt and ANXA7ΔNt, also with C-terminal eGFP tag (see Materials and Methods). Recombinant proteins were then expressed and purified. Because low- complexity sequences are generally known to have a high tendency to self-associate, 2 M Urea was present throughout the purification process to avoid protein aggregations. The ability of the three ANXA7 variants to undergo phase-separation was first evaluated at different pH. Specifically, we employed fluorescence microscopy to monitor the evolvement of various protein species. Prior to imaging, protein stock in 2 M Urea was diluted into urea-free buffers containing 100 mM NaCl at different pH: 5.0, 6.0 and 7.0 (see Materials and Methods). Two samples were prepared 92 in identical ways so as to acquire images respectively at 5 minutes and 18 hours post dilution at room temperature. As shown in Figure 4-3A to 4-3C, liquid-like droplet species were observed for ANXA7 and ANXA7Nt at lower pH (5.0 and pH 6.0) but not at pH 7.0. For ANXA7, those liquid-like droplet species fused into larger patches and wet the glass surface after 18 hours of incubation. Smaller spherical species with irregular edges were also seen, potentially derived from the fission events of the larger patches. For ANXA7Nt, the liquid-like droplet species formed at pH 5.0 further matured into fibril-like filamentous structures, sprouting out from the edges of the droplets. At pH 6.0, however, only smaller spherical species with irregular edges were seen, but not extended filamentous structures. In contrast, for ANXA7ΔNt, formation of liquid-like droplet species was not observed under all tested conditions. Instead, smaller amorphous aggregates were observed after 5 minutes at lower pH conditions (pH 5.0 and pH 6.0) and larger amorphous aggregates were observed after 18 hours at all pH conditions. Similar amorphous aggregates were observed for the other two variants after 18 hours as well. Upon formation of protein droplets or aggregates, a protein solution will increase in turbidity. Therefore, it is possible to use sample turbidity (optical density at 600 nm) to quantify the tendency of ANXA7 variants to self-associate into larger species. Although an increase in turbidity does not always mean greater droplet formation, when combined with fluorescent imaging, it provides us with a quantitative reference of the process. The results were plotted and summarized in Figure 4-3D to 4-3F mirroring the same buffer conditions used in Figure 4-3A to 4- 3C. Next, we evaluated the effect of ionic strength (NaCl concentration) on the ANXA7 phase separation using the same method (Figure 4-4). Here, urea-free buffers at pH 6.0 containing increasing concentrations of NaCl (0 mM, 50 mM, 100 mM, 250 mM and 500 mM) were used. As shown, higher salt concentration (250 mM and 500 mM) weakened or completely abolished the formation of protein droplets for ANXA7 and ANXA7Nt. Similar fibril-like filamentous sprouting was also observed for ANXA7Nt at pH 6.0 in the absence of NaCl. And similar droplet fusion/wetting as well as smaller spherical species with irregular edges were observed for ANXA7 and ANXA7Nt after 18 hours at lower salt conditions. And again, for ANXA7ΔNt, formation of liquid-like droplet species was not observed under all tested conditions. 93 Collectively, these data support that ANXA7 can undergo liquid-liquid phase separation in vitro. Comparing between three ANXA7 variants, we can conclude that the N-terminal domain is required for phase separation since liquid-like droplet species did not form under all tested conditions when the N- terminal domain was missing (ANXA7ΔNt). Additionally, the results indicate that ANXA7 droplet formation was favored at lower pH and NaCl concentrations. 4.2.4 FRAP analysis indicates a high percentage of mobile component within the ANXA7 droplets. Having confirmed that the full-length ANXA7 and its N-terminal domain can both form liquid-like droplets, we want to characterize their fluid-like properties in more details. Fluorescence recovery after photobleaching (FRAP) was employed to evaluate the degree of dynamic diffusion within the protein droplets. Briefly, a small region of interest (ROI) was selected for photobleaching using the laser power, which was then allowed to spontaneously recover through diffusion. Average fluorescence intensity within the ROI was monitored over time to characterize the intrinsic dynamic property of the droplet. For the results to better reflect the protein-specific properties, five independent measurements were performed for each variant. Again, droplet formation was triggered by diluting the protein stock into urea-free buffer (50 mM MES, 100 mM NaCl, pH 6.0). It is necessary to point out that the formation of protein droplets is very sensitive to pipetting-related mechanical disturbance, special attention needs to be paid during sample preparation (see Materials and Methods). The results of the FRAP analysis are summarized in Figure 4- 5. According to the representative time series images (Figure 4-5A and 4-5E), protein droplets derived from both variants were able to undergo a gradual fluorescence recovery over time. This recovery process can be better visualized using their respective kymographs (Figure 4-5B and 4-5F). Individual recovery curves for five droplets were plotted in Figure 4-5C and 4-5G. Interestingly, full-length ANXA7 droplets seem to be more homogenous with regard to their recovery kinetics when compared to ANXA7Nt droplets. The average recovery curves were also plotted and fitted using a two-phase association model (Figure 4-5D and 4-5H). As shown, both types of droplet are composed of a small fraction of immobile component, which is slightly lower in full-length ANXA7 droplets (6.75%) than in ANXA7Nt droplets 94 (12.95%). Nevertheless, we can still conclude that the main component for both droplets are highly mobile, corresponding to highly dynamic protein droplets with liquid-like properties. During fitting, we noticed that the curves fit better with two-phase than one-phase association model. This signals that the recovery of the mobile fraction is a composite of two processes: a fast one and a slow one, though the specific rate constants vary between the two protein variants. To summarize briefly, compared to the full-length ANXA7 droplets, the ANXA7Nt droplets are more heterogenous and include higher percentage of immobile components. Recall earlier that ANXA7Nt droplets, but not the full-length ANXA7 droplets, were able to mature into fibril-like filamentous species (Figure 4-3B and 4-4B). These results collectively are supporting a higher propensity for the N-terminal domain to mature into immobile species, which is consistent with its low-complexity sequence. 4.2.5 The C-terminal core domain is sufficient for Ca 2+ -mediated membrane interaction of ANXA7 but vesicle clustering requires the additional contribution from the N-terminal domain. As mentioned earlier, ANXA7 has a number of membrane-related functions, therefore, it would be interesting to investigate the role of the N-terminal domain in membrane interaction. Many well-studied members of the annexin superfamily (e.g. ANXA2, A5 and B12) can readily tolerate high Ca 2+ concentration (1 mM). However, the protein stability of ANXA7 is very sensitive to the presence of Ca 2+ . In fact, according to a previous calcium titration experiment using ANXA7 isolated from bovine adrenal glands, Ca 2+ can promote and accelerate the formation of protein aggregates with a half maximal aggregation-activation Ca 2+ concentration of 200 μΜ (Creutz et al., 1979). This was verified in our preliminary study, where recombinant ANXA7 exhibited significant aggregation in the Ca 2+ range of ~100 to 300 μΜ. Therefore, in order to minimize ANXA7 aggregation during our tests of Ca 2+ -mediated membrane binding, we seek to even lower Ca 2+ concentration (20 μΜ). Additionally, we also used a low protein concentration (5 μΜ) to decrease the incidence of protein collision. To enhance membrane interaction at such a low Ca 2+ concentration, a high protein-to-lipid ratio was chosen (molar ratio 1:2000). For microscopy analysis, the three eGFP-tagged ANXA7 variants were mixed with Rhodamine-DOPE 95 labeled MLVs (POPS/POPC molar ratio 2 to 1) in buffer (50 mM HEPES, 100 mM NaCl, 1 mM EDTA, pH 7.0). The results of the Ca 2+ -mediated membrane binding tests are summarized in Figure 4-6. As shown, in the presence of 1 mM EDTA and in the absence of Ca 2+ , none of the three ANXA7 variants exhibited any significant membrane binding (Figure 4-6A, panel a, c and e, for single channel images, see Figure 4-S3). With the addition of 1.02 mM CaCl 2 to neutralize the EDTA and to introduce 20 μΜ net Ca 2+ in the system (assuming perfect EDTA/Ca 2+ chelation at 1:1 molar ratio), we observed strong membrane binding for ANXA7 and ANXA7ΔNt but not for ANXA7Nt (Figure 4-6A, panel b, d and f). The results indicate that the C-terminal core domain remains active upon N-terminal truncation and is sufficient to mediate ANXA7 membrane binding. Interestingly, in addition to membrane interaction, Ca 2+ addition also triggered potent clustering of the lipid vesicles (Figure 4-6A, panel b) for full-length ANXA7. Clustered vesicles are huge in size and are less mobile under the fluorescence microscope compared to the free ones. This phenomenon was unique to full-length ANXA7, suggesting that the N-terminal domain plays a key role in the vesicle clustering. In earlier sections, we have proved that the N-terminus alone has a high tendency to self-associate into aggregates (in the cell) and form protein droplets (in vitro). Here, the self-association of ANXA7 can similarly bridge the membrane surfaces between lipid vesicles, promoting their clustering. Interestingly, the event can be efficiently reversed by the addition of another 1 mM EDTA to chelate the Ca 2+ (Figure 4- 6A, panel g). The results of the membrane interaction tests were similarly mirrored in the turbidity measurements performed at the same protein-to-lipid ratio (plotted in Figure 4-6A, panel h). A schematic illustration depicting the Ca 2+ -induced vesicle clustering and the reverse by EDTA chelation of full-length ANXA7 is presented in Figure 4-6B. Noticing that in earlier results, ANXA7 (10 μΜ) did not undergo phase-separation at pH 7.0 in the presence of 100 mM NaCl and in the absence of lipid membranes (Figure 4-3A), suggesting that a weak tendency to self-associate at these conditions. However, in the presence of lipid membranes, the protein potently self-associated to promote vesicle clustering at even lower protein concentration (5 μΜ). In preliminary tests, we have shown that no significant protein aggregation was observed in the presence of 96 20 μΜ Ca 2+ . Therefore, this tendency of self-association is likely due to the increase in local ANXA7 protein concentration upon membrane binding. 4.3 Discussion The N-terminal domain has been implied in previous studies to mediate Ca 2+ -dependent reversible self-association of ANXA7 (Creutz, 2009; Creutz et al., 1979). In the current study, we showed that the N- terminal domain itself can undergo Ca 2+ -independent self-association into highly mobile protein droplets. Similar behavior was also observed for full-length ANXA7 but not for ANXA7ΔNt variant, suggesting the pivotal role of the low-complexity N-terminal domain in mediating the protein liquid-liquid phase separation. The process is highly sensitive to environmental factors such as the pH and the salt concentration. Furthermore, the N-terminal domain was suggested previously to modulate membrane binding of ANXA7 as removal of specific regions within the N-terminus reduced membrane binding affinity (Chander et al., 2006; Naidu et al., 2005)). Here, we successfully showed that the N-terminal domain is not essential for membrane interaction, since truncation of the entire N-terminal region (residue 2-176) did not lead to significant defects in membrane binding. The seemly inconsistency between our study and the previous studies might be because we performed the experiment at a really high lipid-to-protein ratio, at which membrane interaction is strongly favored. On the other hand, however, we find that the N-terminal domain is playing a crucial role in vesicle clustering. The ability of the N-terminal domain to bring together neighboring membrane surfaces could be attributed to its self-association tendency at increased local protein concentration upon membrane binding. The above two functional roles of the N-terminal domain are highly relevant to ANXA7’s biological roles. In the cell, where physiological level of calcium is present, Ca 2+ -dependent and Ca 2+ - independent self-association can both take place under diverse conditions. The recruitment of endogenous ANXA7 into stress granule-like structures (Figure 4-1) can be mediated by the Ca 2+ -independent mechanism. By shifting cellular environment, cellular stress can switch on the formation of ANXA7 droplets, which are favored at specific pH and salt conditions. However, stress-induced disturbance of the 97 Ca 2+ signaling can also trigger self-association of the protein through the Ca 2+ -dependent mechanism, thus further complicating the scenario. On the other hand, the morphology of full-length ANXA7 aggregates observed in Figure 4-2 highly resembles those reported earlier in the Ca 2+ -dependent self-association (Creutz et al., 1979). In fact, the formation of protein aggregates rather than droplets upon ectopic overexpression is a common artifact observed for proteins undergoing LLPS (McSwiggen et al., 2019). This is mostly due to the fact that LLPS is highly sensitive to protein concentration. At abnormally high protein concentration way beyond its physiological level, LLPS is disturbed, leading towards the formation of amorphous protein aggregates. Unlike many proteins, which rely on the cation-aromatic interactions between positively charged residues (Arg/Lys) and aromatic residues (Phe/Trp/Tyr), the N-terminal domain of ANXA7 is completely depleted of positively-charged amino acids. Therefore, its phase separation must be driven by forces other than cation-aromatic interactions, potentially hydrophobic interactions or hydrogen bonds. Additionally, unlike most other proteins undergoing LLPS, no post translational modification site has been identified so far within the N-terminal domain, although serines and tyrosines are abundantly present and could act as potential phosphorylation sites. Annexin A7 is not the only annexin whose N-terminus has membrane-related functions. For example, the N-terminus of ANXA1 can form helical structure upon membrane interaction and act as a membrane anchor for Ca 2+ -independent membrane interaction (Hu et al., 2008). ANXA2 can promote membrane clustering by disulfide bond formation of Cys9 at the N-terminus (Grill et al., 2018). ANXA13 can undergo N-terminal refolding into a continuous helix, regulating the membrane binding (McCulloch et al., 2019). It is amazing that so many diverse mechanisms exist to modulate how the N-terminus of different annexins fulfill their membrane-related functions. As mentioned earlier, ANXA7 can interact with multiple binding partners (Creutz, 2009; Li et al., 2013a; Shibata et al., 2008; Verzili et al., 2000). LLPS introduces a new mechanism for these intracellular interactions. Instead of interacting individually in solution, proteins containing similar low-complexity domains may simultaneously get recruited into droplets for a network of interactions. Additionally, with 98 the assistance of the C-terminal core domain of ANXA7, the established protein droplet network could be transported to the surface of various cellular membranes in response to a Ca 2+ signaling. ANXA7 may even act as a vehicle for the transportation and spreading of biological species (e.g. RNAs, toxic amyloid proteins) between different organelles and cells. Similar behavior has been observed previously in ANXA11 and RNA granules (Liao et al., 2019). Membrane proteins with intrinsically disordered domains can be potent sensors for membrane curvature, the rationale being that curved surfaces allow greater conformational entropy of the disordered amino acid chain (Zeno et al., 2018). In fact, we have also observed suspicious preferential binding to smaller vesicles during fluorescence imaging. However, quantification is difficult due to two main reasons. First, vesicles with the highest membrane curvatures are the smallest in size, which is below the detection resolution limit of common microscopes. Second, upon membrane binding of the full-length protein, significant vesicle clustering was observed (Figure 4-6A, panel b). Tightly clustered vesicles are often morphologically deformed and their curvature cannot be defined by diameter. On top of this, it is challenging to assign protein signals at the borderline between two or more neighboring vesicles. To explore the potential curvature-sensitive feature of the N-terminus, future studies need to employ more advanced techniques to get around the intrinsic limitations of the current methods. 4.4 Materials and Methods 4.4.1 Plasmid DNA Preparation ANXA7 cDNA (transcript variant 1) was purchased from OriGene Technologies, Inc (#SC126802). For cellular expression, full-length ANXA7 cDNA with a C-terminal eGFP (A206K variant) tag was cloned into pcDNA4.0 backbone. For ANXA7Nt plasmid, residues 177-488 were truncated. For ANXA7ΔΝt plasmid, residues 2-176 were truncated. For expressing recombinant ANXA7, the same eGFP-tagged ANXA7 cDNA or its modified variants were cloned into pET28b(+) backbone with an N-terminal 6xHis tag. 99 4.4.2 Cell Culture and Immunostaining. HEK293T cells obtained from ATCC were maintained in high glucose DMEM media supplemented with 10% FBS and 100 μg/ml penicillin-streptomycin and grown in a humidified incubator at 37 °C supplemented with 5% CO 2. Cells were first seeded on poly-D-lysine coated glass cover slips overnight and then transfections were performed using lipofectamine â LTX with PLUS TM reagent (#15338030, Invitrogen TM ) according to the manufacturer’s instruction. For immunocytochemistry analysis, seeded cells were fixed with 4% PFA for 10 minutes (followed by PBS wash) and then permeabilized with 0.1% Triton-X for 15 minutes (followed by PBS wash). The cells were blocked with 1% BSA for 1 hour at room temperature and then stained with anti-ANX VII primary antibody (#A4475, Sigma) by incubating at 4 °C for overnight (1:200 dilution). Next day, the cells were washed with PBS (x 3 times) and then stained with donkey anti-mouse Alexa Fluor â 488 secondary antibody (#R37114, Invitrogen TM ) for 1 hour (1:500 dilution) at room temperature. For both antibody staining, 0.1% BSA was present to reduce non-specific binding. The cells were again washed with PBS (x 3 times). DAPI (1 μg/ml) staining was performed at room temperature for 10 minutes and the extra dyes were rinsed off with PBS. The coverslips with the cells were mounted onto glass slides using VECTASHIELD â Antifade Mounting Medium (#H-1000, Vector Laboratories). 4.4.3 Recombinant Protein Expression and Purification For expressing recombinant ANXA7, DNA plasmids were transformed into BL21(DE3) E.Coli and cells were spread on LB plates (30 μg/ml kanamycin) at 37 °C overnight. A single colony was picked and inoculated into 25 ml LB media with kanamycin and grown at 37 °C for 3-4 hours. The culture was then expanded to 1 L (37 °C, 225 rpm). Protein expression was induced with 1 mM IPTG at ~ 0.6 OD 600nm and the culture was incubated at 18 °C, 170 rpm for two days overnight. Bacteria pellet was harvested by centrifugation at 4 °C, 4000 rpm for 20 minutes. The pellet was incubated for 30 minutes on ice in lysis buffer (50 mM Tris, 100 mM NaCl, 2 M Urea, 1 mM DTT, 0.1% Triton-X, pH 8.0). 0.3 μM PMSF was added to the lysed cells followed by tip sonication. The sonicated mixture was pelleted down at 4 °C, 18,000 100 rpm for 20 minutes and the supernatant was loaded onto Ni-column (His60 Ni Superflow Resin, Takara Cat#635662) and incubated at 4 °C for 30 minutes with gentle shaking to allow binding. The Ni-column was then washed with wash buffer (50 mM Tris, 100 mM NaCl, 2 M Urea, 20 mM Imidazole, 1 mM DTT, pH 8.0) and eluted with elution buffer (50 mM Tris, 100 mM NaCl, 2 M Urea, 200 mM Imidazole, pH 8.0). Immediately after elution, 5 mM DTT and 10 mM EDTA were added to the sample to prevent aggregation and to chelate Ni 2+ ions potentially been carried over from the Ni-column elution. The protein solution was then passed through a 0.2 μm filter to remove any potential aggregates before being concentrated to ~500 ul using centrifugal filters (Millipore #UFC901096). The concentrated protein was then loaded onto a PD10 column (GE Healthcare) to remove the imidazole and for buffer exchange (50 mM HEPES, 100 mM NaCl, 2 M Urea, 1 mM EDTA, pH 8.0). 4.4.4 Preparation of Lipid Vesicles 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC) and 1-palmitoyl-2-oleoyl-sn-glycero- 3-[phospho-L-serine] (POPS) were purchased as chloroform solutions from Avanti Polar Lipids, Inc (#850457C and #840034C). Lipids were first mixed (POPS/POPC molar ratio 2 to 1) and labeled with 0.2% rhodamine- DOPE (Avanti Polar Lipids, #810150). The lipid mixture was then dried under a gentle nitrogen flow to form a thin layer of lipid film and kept in a desiccator overnight at room temperature. The film was later resuspended in buffer (50 mM HEPES, 100 mM NaCl, 1 mM EDTA, pH 7.0) to a final concentration of 10 mM. 4.4.5 Confocal Fluorescence Microscopy Imaging For examining ANXA7 cell localization, glass slides were examined under a 63x oil immersion objective lens on a Zeiss LSM 780 inverted confocal microscope. For examining ANXA7 droplet formation in vitro, 1 mM protein stock in buffer (50 mM HEPES, 100 mM NaCl, 2 M Urea, 5 mM DTT, 1 mM EDTA, pH 7.0) was diluted into a final concentration of 10 μΜ using respective urea-free buffer. We noticed that for ANXA7, direct dilution from 1 mM protein stock introduces artifact by triggering immediate aggregation and therefore protein stock was pre-diluted to 500 μM. For analyzing the effect of pH, buffer containing 100 mM NaCl was used with either 50 mM NaAc, pH 5.0; 50 mM MES, pH 6.0 or 50 mM HEPES, pH 7.0. For analyzing the effect of ionic strength, buffers (50 mM MES, pH 6.0) containing increasing NaCl concentration: 0 mM, 50 101 mM, 100 mM, 250 mM and 500 mM were used. All samples were prepared in 35mm glass-bottomed dishes (MatTek, P35G-1.5-14-C). The images were acquired using 40x oil immersion objective lens on a Leica SP8X inverted confocal microscope. For membrane interaction experiments, protein and lipid vesicles (POPS/POPC molar ratio 2 to 1, 0.2% rhodamine-DOPE labeled) were mixed at 1 to 2000 final protein-to-lipid ratio in buffer (50 mM HEPES, 100 mM NaCl, pH 7.0) in a total volume of 100 ul system. Final protein concentration was 5 μΜ and final lipid concentration was 10 mΜ. To induce Ca 2+ -mediated membrane interaction, 1 ul of CaCl 2 (102 mM) was spiked into the lipid vesicles to reach a final Ca 2+ concentration of 1.02 mM and mixed thoroughly before adding the protein. This was so designed as to avoid protein aggregation induced by high local Ca 2+ concentration. And after addition of the protein, the mixture was once again mixed thoroughly before imaged on a Zeiss LSM 780 inverted confocal microscope with a 40x oil immersion objective lens. Unbinding of ANXA7 from the membrane was triggered by spiking in 1 mM EDTA and thorough mixing. All images captured were processed using Fiji/ImageJ software. 4.4.6 Time Course Measurement of Solution Turbidity For examining ANXA7 droplet formation, the protein was diluted the same way as in fluorescence imaging and its optical density (O.D.) at 600 nm was measured over a time course of ~30 minutes at room temperature. Respective buffers were allowed to stabilize for 60s before protein was spiked into the system. After 20s of mixing, the sample was put back for further measurements. For examining ANXA7-mediated vesicle clustering, O.D. at 600 nm was measured by mixing MLVs (POPS/POPC 2 to 1 molar ratio but not rhodamine-DOPE labeled) with the same protein variants at the same protein-to-lipid molar ratio (1 to 2000) but at 10x lower final concentrations to avoid signal saturation. For reversing protein binding in the ANXA7 group in the presence of Ca 2+ , 1 mM EDTA was spiked in at 20 minutes. 4.4.7 FRAP Measurement The formation of protein droplets is very sensitive to pipetting-related mechanical disturbance. Without any pipette mixing, local droplet concentration will be very high, favoring quick fusion events 102 between neighboring mini-droplets. The net outcome is the formation of relatively large droplets. Although larger droplets are in general better candidates for FRAP bleaching and analysis, performing a FRAP measurement without any disturbance is difficult with concurring frequent fusion events. On the other hand, fierce mixing should also be avoided because it will trigger the formation of tiny droplets, which are very difficult to target during bleaching due to their sizes. Here, for each independent measurement, droplets were freshly prepared. Prior to bleaching, the protein stock was carefully spiked in at the bottom of the glass dish (remain as a relatively stable separate oil drop) followed by three gentle pipetting. FRAP analysis was performed on the LSM780 inverted confocal microscope. Freshly prepared droplets were imaged under a 63x oil lens. For each variant, protein was diluted to 10 μΜ final concentration using buffer (50 mM MES, 100 mM NaCl, pH 6.0). Immediately after dilution, a region of interest (ROI) at the center of the droplet was bleached for 8 iterations at maximum laser intensity and stopped until the average fluorescence intensity within the region dropped to 25% of its original intensity. The fluorescence intensity was then allowed to recover over a time course ~15 minutes. Images were captured for every 5 seconds. All images captured were processed using Fiji/ImageJ software. 103 Figure 4-1. Endogenous ANXA7 is recruited into granular structures upon arsenite-induced cellular stress Representative confocal microscopy images are shown for HEK293T cells immunostained with anti- ANXA7 antibody. The cells are either (B) treated or not treated (A) with 0.5 mM NaAsO2 for 30 minutes before fixation prior to the immunostaining. DAPI staining shows the location of cellular nucleus. As shown, the endogenous ANXA7 appeared as granular structures in the NaAsO2-treated group when compared to the non-treated group. From left to right, first column: differential interference contrast (DIC) channel; second column: green channel showing the signal of endogenous ANXA7; third column: blue channel showing the signal of DAPI. (C) Quantification of NaAsO2- induced formation of ANXA7-containing granular structures. The percentage of total cell population containing the granular structures were counted respectively for the NaAsO2 treated and non-treated groups. Three independent experiments were performed and five frames in the size of 350 x 350 μm were picked randomly for z-stack imaging in each experiment. Projections of the z-stacked images were used to count the number of total cells and the number of cells containing granular structures using cell-counter plugin of Fiji/Image J software. The ratio of the two numbers when then computed and the average values were plotted with error bars showing standard deviation. Two-tailed unpaired t-test was performed between the two groups, p < 0.0001, suggesting statistically significant difference between them. Scale bar = 20 μm. 104 Figure 4-2. Overexpression of ANXA7 variants containing the N-terminal sequence leads to the formation of cellular aggregates HEK293T cells were transfected with one of the three ANXA7 variant constructs: (A) full-length ANXA7, (B) N-terminus of ANXA7 (residues 1-176) or (C) N-terminus truncated fragment of ANXA7 (residues 2-176 removed). All three ANXA7 variants were eGFP-tagged at the C-terminal end of the protein sequence. 48 hours post transfection, cells were fixed. DAPI staining shows the location of the cellular nucleus. From left to right, first column: green channel showing ANXA7 signals; second column: blue channel showing DAPI signals; third column: merged overlay between green and blue channels, regions of interest highlighted with dashed white squares; Fourth column: 4X magnified view of the regions highlighted in the third column. As shown cellular aggregates are formed in (A) and (B) but not in (C), suggesting that the low-complexity N-terminus region is critical in promoting the self-association of ANXA7. Scale bar = 20 μm. 105 Figure 4-3. Recombinant ANXA7 undergoes pH-sensitive liquid-liquid phase separation in vitro, mediated by its low-complexity N-terminus (A) Full-length ANXA7, (B) N-terminus of ANXA7 (residues 1-176) and (C) N-terminus truncated fragment of ANXA7 (residues 2-176 removed) were expressed recombinantly with eGFP tagged at the c-terminus. Protein purified and stored in buffer containing 2M Urea was diluted to a final protein concentration of 10 μM using urea-free buffer at varying pH containing 100mM NaCl (50 mM NaAc, pH 5.0; 50 mM MES, pH 6.0; 50 mM HEPES, pH 7.0) to induce the liquid-liquid phase separation of ANXA7 variants. Samples were imaged 5 minutes or 18 hours post dilution. Representative confocal images are shown in (A) to (C). At lower pH (5.0 or 6.0), GFP positive droplets were observed for both full-length and n-terminus of ANXA7. After 18 hours of incubation at room temperature, fibril-like filamentous structures were sprouting out of the droplets formed from the N-terminus of ANXA7 at pH 5.0. No droplets were observed for the N-terminus truncated ANXA7 fragment under the same condition, although small aggregates were seen after 18 hours of incubation. Scale bars = 10 μm. (D) to (F): The proteins were diluted the same way as in (A) to (C) and their optical density (O.D.) at 600 nm were monitored over a time course of ~30 minutes. Respective buffers were allowed to stabilize for 60s before protein samples were spiked-in. The protein was allowed to mix completely for 20s before put back for the reading. UV absorbance values are plotted overtime as mean ±standard deviation (n = 3). Results for pH 5.0, 6.0 and 7.0 are colored respectively in red, blue and green. O.D. 600 nm was recorded using a 0.2 cm light path to avoid saturation of the signal. 106 Figure 4-4. Recombinant ANXA7 undergoes salt-sensitive liquid-liquid phase separation in vitro, mediated by its low-complexity N-terminus (A) to (C): Full-length, N-terminus or N-terminus truncated ANXA7 fragment was diluted to a final protein concentration of 10 μM using 50 mM MES buffer, pH 6.0 containing varying concentrations of NaCl: 0 mM, 50 mM, 100 mM, 250 mM and 500 mM. Samples were imaged 5 minutes post dilution. Scale bars = 10 μm. (D) to (F): The protein was diluted the same way as in (A) to (C) and its optical density (O.D.) at 600 nm was measured over a time course of ~30 minutes. Respective buffers were allowed to stabilize for 60s before protein was spiked into the system. The protein was allowed 20s of mixing before put back for measurements. Data are represented as mean ±standard deviation, n = 3. O.D. 600 nm was recorded using a 0.2 cm light path. 107 Figure 4-5. FRAP analysis indicates a high percentage of mobile component within the ANXA7 droplets. FRAP analyses were performed for protein droplets formed from full-length ANXA7 (A) to (D) and N- terminus of ANXA7 (E) to (H). For each variant, protein was diluted to 10 μΜ final concentration using buffer (50 mM HEPES, 100 mM NaCl, pH 6.0). Immediately after dilution, a region of interest (ROI) at the center of the droplet was bleached for 8 iterations at maximum laser intensity and stopped until the average fluorescence intensity within the region dropped to 25% of its original intensity. The fluorescence intensity was then allowed to recover over time. Images were captured for every 5 seconds. (A) and (E): Representative confocal microscopy images showing the droplets at the beginning and end of the bleaching. The recovery status at 1, 2, 4, 8 and 12 minutes after the bleaching end was also shown. Scale bar = 5 μm. (B) and (F): Representative kymographs are shown of individual FRAP experiments performed in (A) and (E) to illustrate the photobleaching dynamics. (C) and (G): individual recovery curves were shown for five droplets. The intensity was normalized using a non-bleached ROI within the same droplet. (D) and (H): the average of the five repeats were plotted with mean ±standard deviation. The data was fit using nonlinear two-phase association model in GraphPad Prism 8.0. 108 Figure 4-6. Membrane interaction and vesicle clustering of ANXA7 (A) MLVs composed of POPS to POPC (molar ratio) 2 to 1 (0.2% rhodamine-DOPE labeled) were mixed with one of the three variants of ANXA7 in buffer (50 mM HEPES, 100 mM NaCl, 1 mM EDTA, pH 7.0) at final lipid concentration of 10 mM and protein concentration of 5 μM in the absence (panel a, c, e) or presence (panel b, d, f) of 1.02 mM Ca 2+ . Representative confocal microscopy images are shown. Merged image overlaid between the green and the red channels. Scale bars = 10 μm. The vesicle clustering observed for ANXA7 can be reversed by adding 1 mM EDTA (panel g). Small protein aggregates were also observed. (h): O.D. at 600 nm was measured by mixing the same MLVs (not labeled) with the same protein variants at the same protein-to-lipid molar ratio (1 to 2000) but at 10x lower final concentrations to avoid signal saturation. As shown, significant increase in turbidity (O.D. at 600 nm) was observed for full-length ANXA7 in the presence of Ca 2+ (red) compared to all other groups. Spiking in an additional 1 mM EDTA at 20 minutes can reset the absorbance back to the basal level (green). (B) Schematic illustration of ANXA7-mediated vesicle clustering as regulated by the addition of Ca 2+ and EDTA. For images showing individual green or red channels, see Figure 4-S3. 109 Figure 4-S1. Disorder analysis of ANXA7 sequences Residues predicted to be disordered are shown in red and those ordered are shown in black. False positive (FP) rate was set as 5%. As shown the N-terminal region (residues 1-170) was predicted to be highly intrinsically disordered. Prediction was performed using PrDOS. 110 Figure 4-S2. Composition bias of ANXA7 N-terminal sequences Residues accounting for > 5% of amino acid composition within the N-terminal region (residues 1-176) are listed. Their respective percentages in the total protein composition are also shown. 111 Figure 4-S3. Single channel confocal fluorescence images for Figure 4-6A For each panel, left column: green channel for ANXA7 signals; right column: red channel for lipid vesicles; MLVs composed of POPS to POPC (molar ratio) 2 to 1 (0.2% rhodamine-DOPE labeled) were mixed with one of the three variants of ANXA7 in buffer (50 mM HEPES, 100 mM NaCl, 1 mM EDTA, pH 7.0) at final lipid concentration of 10 mM and protein concentration of 5 μM in the absence (panel A) or presence of 1.02 mM Ca 2+ (panel B). The vesicle clustering observed for ANXA7 can be reversed by adding 1 mM EDTA (panel C). Small protein aggregates were also observed. Scale bars = 10 μm. 112 Concluding Remarks and Future Directions Membrane proteins play diverse and critical roles in the biological system and their importance can never be overstated. They are involved in two major types of contact: the protein-protein interaction and the protein-lipid interaction. Importantly, there are also interesting interplays between the two contacts and therefore special attention should be paid not to judge one of them while completely neglecting the other. Failure to do so would result in segmental conclusions. Meanwhile, for different types of membrane proteins, the respective role of the two interactions could vary and so can the interplay. This increases our challenges during the investigation but at the same time adds to the intricacy and delicacy of nature and life. Thus far, we have investigated the protein-protein and protein-membrane interactions using three model membrane proteins, with emphasis on different perspectives. For annexin B12, the clustering of the protein is mediated jointly by a network of protein-protein interaction and protein-lipid interaction. Near complete abolishment of trimer formation can be achieved by disrupting either of the two interactions. For Httex1, protein-membrane interaction promotes the protein-protein interaction, leading to the clustering of protein (amyloid fibril formation). For annexin A7, its N-terminal domain is mediating the protein-protein interaction (protein phase separation) while as its C-terminal domain is mediating the protein-lipid interaction (Ca 2+ -mediated membrane binding). And the cooperation between the two domains is required for the normal function of the protein in clustering lipid vesicles. In all three cases, we can identify interplay between the two types of interaction, regardless of whether it is one interaction enhancing the other (Httex1) or the two interactions in collaboration (ANXB12, ANXA7). Common to all these diverse and complicated process, thermodynamic energy played critical roles. Protein-lipid interaction contributed significantly to the membrane-mediated trimerization of annexin B12, the energy minimization of stretched lipids at the edges of protein upon trimer formation could be potentially driving the process. Amyloid fibril formation of huntingtin exon 1 is entropy-driven as monomers are continuously recruited into fibrils, which is thermodynamically more stable. Membrane interaction decreases the energy requirement for fibril formation by inducing/stabilizing a defined 113 membrane-bound structure, which can then self-associate at a reduced dimensionality. On the other hand, liquid-liquid phase separation decreases the energy barrier for fibril formation by providing an intermediate step on the nucleation-elongation pathway to forming protein droplets. The extent of mobility decreases in the order of: solution > droplets > oligomers/fibrils. The decreased mobility within the droplets favors the nucleation into even more immobile species. The low-complexity N-terminus of ANXA7 is highly prone to self-association because it contains so many aromatic residues and hydrophobic residues. Self- association between hydrophobic species away from the aqueous environment is therefore thermodynamically favorable as it minimizes the system energy. Overall, protein-lipid interaction, protein-protein interaction and the interplay between the two interactions are regulating the normal function of cellular functions. Aberrant interactions will lead to pathological conditions. For therapeutic purposes, we should take good advantage of our current understanding of these interactions. For example, design small molecules or modify protein structures to block/enhance certain type of interactions. It is a challenging task and requires the consistent work of future researchers and the continuous accumulation of a comprehensive pool of knowledge. 114 References Aditi, K., Shakarad, M.N., and Agrawal, N. (2016). Altered lipid metabolism in Drosophila model of Huntington's disease. Sci Rep 6, 31411. Albin, R.L., Reiner, A., Anderson, K.D., Dure, L.S.t., Handelin, B., Balfour, R., Whetsell, W.O., Jr., Penney, J.B., and Young, A.B. (1992). Preferential loss of striato-external pallidal projection neurons in presymptomatic Huntington's disease. Ann Neurol 31, 425-430. Ali Doosti, B., Pezeshkian, W., Bruhn, D.S., Ipsen, J.H., Khandelia, H., Jeffries, G.D.M., and Lobovkina, T. (2017). Membrane Tubulation in Lipid Vesicles Triggered by the Local Application of Calcium Ions. Langmuir 33, 11010-11017. Allhusen, J.S., and Conboy, J.C. (2017). The Ins and Outs of Lipid Flip-Flop. Acc Chem Res 50, 58-65. Altenbach, C., Flitsch, S.L., Khorana, H.G., and Hubbell, W.L. (1989). Structural studies on transmembrane proteins. 2. Spin labeling of bacteriorhodopsin mutants at unique cysteines. Biochemistry 28, 7806-7812. Altenbach, C., Greenhalgh, D.A., Khorana, H.G., and Hubbell, W.L. (1994). A collision gradient method to determine the immersion depth of nitroxides in lipid bilayers: application to spin-labeled mutants of bacteriorhodopsin. Proc Natl Acad Sci U S A 91, 1667-1671. Altenbach, C., Marti, T., Khorana, H.G., and Hubbell, W.L. (1990). Transmembrane protein structure: spin labeling of bacteriorhodopsin mutants. Science 248, 1088-1092. Ambadipudi, S., Biernat, J., Riedel, D., Mandelkow, E., and Zweckstetter, M. (2017). Liquid-liquid phase separation of the microtubule-binding repeats of the Alzheimer-related protein Tau. Nat Commun 8, 275. Andree, H.A., Stuart, M.C., Hermens, W.T., Reutelingsperger, C.P., Hemker, H.C., Frederik, P.M., and Willems, G.M. (1992). Clustering of lipid-bound annexin V may explain its anticoagulant effect. J Biol Chem 267, 17907-17912. Antonny, B. (2011). Mechanisms of membrane curvature sensing. Annu Rev Biochem 80, 101-123. Arinaminpathy, Y., Khurana, E., Engelman, D.M., and Gerstein, M.B. (2009). Computational analysis of membrane proteins: the largest class of drug targets. Drug Discov Today 14, 1130-1135. Armstrong, B.D., Lingwood, M.D., McCarney, E.R., Brown, E.R., Blumler, P., and Han, S. (2008). Portable X-band system for solution state dynamic nuclear polarization. J Magn Reson 191, 273-281. Atwal, R.S., and Truant, R. (2008). A stress sensitive ER membrane-association domain in Huntingtin protein defines a potential role for Huntingtin in the regulation of autophagy. Autophagy 4, 91-93. Atwal, R.S., Xia, J., Pinchev, D., Taylor, J., Epand, R.M., and Truant, R. (2007). Huntingtin has a membrane association signal that can modulate huntingtin aggregation, nuclear entry and toxicity. Hum Mol Genet 16, 2600-2615. Barnes, R., Sun, S., Fichou, Y., Dahlquist, F.W., Heyden, M., and Han, S. (2017). Spatially Heterogeneous Surface Water Diffusivity around Structured Protein Surfaces at Equilibrium. J Am Chem Soc 139, 17890-17901. 115 Berendes, R., Voges, D., Demange, P., Huber, R., and Burger, A. (1993). Structure-function analysis of the ion channel selectivity filter in human annexin V. Science 262, 427-430. Berliner, L.J. (2016). The evolution of biomedical EPR (ESR). Biomedical Spectroscopy and Imaging 5, 5-26. Berliner, L.J., Grunwald, J., Hankovszky, H.O., and Hideg, K. (1982). A novel reversible thiol-specific spin label: papain active site labeling and inhibition. Anal Biochem 119, 450-455. Bernardino de la Serna, J., Perez-Gil, J., Simonsen, A.C., and Bagatolli, L.A. (2004). Cholesterol rules: direct observation of the coexistence of two fluid phases in native pulmonary surfactant membranes at physiological temperatures. J Biol Chem 279, 40715-40722. Biancalana, M., and Koide, S. (2010). Molecular mechanism of Thioflavin-T binding to amyloid fibrils. Biochim Biophys Acta 1804, 1405-1412. Block, R.C., Dorsey, E.R., Beck, C.A., Brenna, J.T., and Shoulson, I. (2010). Altered cholesterol and fatty acid metabolism in Huntington disease. J Clin Lipidol 4, 17-23. Bode, D.C., Freeley, M., Nield, J., Palma, M., and Viles, J.H. (2019). Amyloid-beta oligomers have a profound detergent-like effect on lipid membrane bilayers, imaged by atomic force and electron microscopy. J Biol Chem 294, 7566-7572. Boeynaems, S., Alberti, S., Fawzi, N.L., Mittag, T., Polymenidou, M., Rousseau, F., Schymkowitz, J., Shorter, J., Wolozin, B., Van Den Bosch, L., et al. (2018). Protein Phase Separation: A New Phase in Cell Biology. Trends Cell Biol 28, 420-435. Bolton, J.A.W.a.J.R. (2007). Electron paramagnetic resonance: Elementary Theory and Practical Applications (Wiley-Interscience). Bortolus, M., Tombolato, F., Tessari, I., Bisaglia, M., Mammi, S., Bubacco, L., Ferrarini, A., and Maniero, A.L. (2008). Broken helix in vesicle and micelle-bound alpha-synuclein: insights from site- directed spin labeling-EPR experiments and MD simulations. J Am Chem Soc 130, 6690-6691. Bouter, A., Gounou, C., Berat, R., Tan, S., Gallois, B., Granier, T., d'Estaintot, B.L., Poschl, E., Brachvogel, B., and Brisson, A.R. (2011). Annexin-A5 assembled into two-dimensional arrays promotes cell membrane repair. Nat Commun 2, 270. Boye, T.L., and Nylandsted, J. (2016). Annexins in plasma membrane repair. Biol Chem 397, 961-969. Bravo-Arredondo, J.M., Kegulian, N.C., Schmidt, T., Pandey, N.K., Situ, A.J., Ulmer, T.S., and Langen, R. (2018). The Folding Equilibrium of Huntingtin Exon-1 Monomer Depends on its Polyglutamine Tract. J Biol Chem. Brender, J.R., Lee, E.L., Cavitt, M.A., Gafni, A., Steel, D.G., and Ramamoorthy, A. (2008). Amyloid fiber formation and membrane disruption are separate processes localized in two distinct regions of IAPP, the type-2-diabetes-related peptide. J Am Chem Soc 130, 6424-6429. Brownawell, A.M., and Creutz, C.E. (1997). Calcium-dependent binding of sorcin to the N-terminal domain of synexin (annexin VII). J Biol Chem 272, 22182-22190. 116 Brudvig, G.W. (1995). Electron paramagnetic resonance spectroscopy. Methods Enzymol 246, 536-554. Bucciantini, M., and Cecchi, C. (2010). Biological membranes as protein aggregation matrices and targets of amyloid toxicity. Methods Mol Biol 648, 231-243. Bugg, C.W., Isas, J.M., Fischer, T., Patterson, P.H., and Langen, R. (2012). Structural features and domain organization of huntingtin fibrils. J Biol Chem 287, 31739-31746. Burke, K.A., Kauffman, K.J., Umbaugh, C.S., Frey, S.L., and Legleiter, J. (2013a). The interaction of polyglutamine peptides with lipid membranes is regulated by flanking sequences associated with huntingtin. J Biol Chem 288, 14993-15005. Burke, K.A., Yates, E.A., and Legleiter, J. (2013b). Biophysical insights into how surfaces, including lipid membranes, modulate protein aggregation related to neurodegeneration. Front Neurol 4, 17. Campelo, F., McMahon, H.T., and Kozlov, M.M. (2008). The hydrophobic insertion mechanism of membrane curvature generation by proteins. Biophys J 95, 2325-2339. Ceccon, A., Clore, G.M., and Tugarinov, V. (2018a). Decorrelating Kinetic and Relaxation Parameters in Exchange Saturation Transfer NMR: A Case Study of N-Terminal Huntingtin Peptides Binding to Unilamellar Lipid Vesicles. J Phys Chem B. Ceccon, A., Schmidt, T., Tugarinov, V., Kotler, S.A., Schwieters, C.D., and Clore, G.M. (2018b). Interaction of Huntingtin Exon-1 peptides with lipid-based micellar nanoparticles probed by solution NMR and Q-band pulsed EPR. J Am Chem Soc. Cezanne, L., Lopez, A., Loste, F., Parnaud, G., Saurel, O., Demange, P., and Tocanne, J.F. (1999). Organization and dynamics of the proteolipid complexes formed by annexin V and lipids in planar supported lipid bilayers. Biochemistry 38, 2779-2786. Chaibva, M., Burke, K.A., and Legleiter, J. (2014). Curvature enhances binding and aggregation of huntingtin at lipid membranes. Biochemistry 53, 2355-2365. Chaibva, M., Jawahery, S., Pilkington, A.W.t., Arndt, J.R., Sarver, O., Valentine, S., Matysiak, S., and Legleiter, J. (2016). Acetylation within the First 17 Residues of Huntingtin Exon 1 Alters Aggregation and Lipid Binding. Biophys J 111, 349-362. Chan, A.W., Xu, Y., Jiang, J., Rahim, T., Zhao, D., Kocerha, J., Chi, T., Moran, S., Engelhardt, H., Larkin, K., et al. (2014). A two years longitudinal study of a transgenic Huntington disease monkey. BMC Neurosci 15, 36. Chander, A., Naidu, D.G., and Chen, X.L. (2006). A ten-residue domain (Y11-A20) in the NH2-terminus modulates membrane association of annexin A7. Biochim Biophys Acta 1761, 775-784. Chanphai, P., L. Bekale, and H. A. Tajmir-Riahi. (2015). Effect of hydrophobicity on protein–protein interactions. European Polymer Journal 67, 224-231. Cheng, C.Y., Varkey, J., Ambroso, M.R., Langen, R., and Han, S. (2013). Hydration dynamics as an intrinsic ruler for refining protein structure at lipid membrane interfaces. Proc Natl Acad Sci U S A 110, 16838-16843. 117 Cheng, G., Pan, J., Podsiadly, R., Zielonka, J., Garces, A.M., Dias Duarte Machado, L.G., Bennett, B., McAllister, D., Dwinell, M.B., You, M., et al. (2019). Increased formation of reactive oxygen species during tumor growth: Ex vivo low-temperature EPR and in vivo bioluminescence analyses. Free Radic Biol Med 147, 167-174. Chi, E.Y., Ege, C., Winans, A., Majewski, J., Wu, G., Kjaer, K., and Lee, K.Y. (2008). Lipid membrane templates the ordering and induces the fibrillogenesis of Alzheimer's disease amyloid-beta peptide. Proteins 72, 1-24. Chung, C.G., Lee, H., and Lee, S.B. (2018). Mechanisms of protein toxicity in neurodegenerative diseases. Cell Mol Life Sci 75, 3159-3180. Clemen, C.S., Herr, C., Hovelmeyer, N., and Noegel, A.A. (2003). The lack of annexin A7 affects functions of primary astrocytes. Exp Cell Res 291, 406-414. Clemen, C.S., Herr, C., Lie, A.A., Noegel, A.A., and Schroder, R. (2001). Annexin VII: an astroglial protein exhibiting a Ca2+-dependent subcellular distribution. Neuroreport 12, 1139-1144. Cohen, B.E., Lee, G., Arispe, N., and Pollard, H.B. (1995). Cyclic 3'-5'-adenosine monophosphate binds to annexin I and regulates calcium-dependent membrane aggregation and ion channel activity. FEBS Lett 377, 444-450. Coleman, R. (1973). Membrane-bound enzymes and membrane ultrastructure. Biochim Biophys Acta 300, 1-30. Cooper, J.H. (1969). A n evaluation of current methods for the diagnostic histochemistry of amyloid. J Clin Pathol 22, 410-413. Cremades, N., Cohen, S.I., Deas, E., Abramov, A.Y., Chen, A.Y., Orte, A., Sandal, M., Clarke, R.W., Dunne, P., Aprile, F.A., et al. (2012). Direct observation of the interconversion of normal and toxic forms of alpha-synuclein. Cell 149, 1048-1059. Creutz, C.E. (1992). The annexins and exocytosis. Science 258, 924-931. Creutz, C.E. (2009). Novel protein ligands of the annexin A7 N-terminal region suggest pro-beta helices engage one another with high specificity. Gen Physiol Biophys 28 Spec No Focus, F7-F13. Creutz, C.E., Pazoles, C.J., and Pollard, H.B. (1978). Identification and purification of an adrenal medullary protein (synexin) that causes calcium-dependent aggregation of isolated chromaffin granules. J Biol Chem 253, 2858-2866. Creutz, C.E., Pazoles, C.J., and Pollard, H.B. (1979). Self-association of synexin in the presence of calcium. Correlation with synexin-induced membrane fusion and examination of the structure of synexin aggregates. J Biol Chem 254, 553-558. Cuatrecasas, P. (1974). Membrane receptors. Annu Rev Biochem 43, 169-214. Davidson, W.S., Jonas, A., Clayton, D.F., and George, J.M. (1998). Stabilization of alpha-synuclein secondary structure upon binding to synthetic membranes. J Biol Chem 273, 9443-9449. 118 Davies, S.W., Turmaine, M., Cozens, B.A., DiFiglia, M., Sharp, A.H., Ross, C.A., Scherzinger, E., Wanker, E.E., Mangiarini, L., and Bates, G.P. (1997). Formation of neuronal intranuclear inclusions underlies the neurological dysfunction in mice transgenic for the HD mutation. Cell 90, 537-548. Deguire, S.M., Ruggeri, F.S., Fares, M.B., Chiki, A., Cendrowska, U., Dietler, G., and Lashuel, H.A. (2018). N-terminal Huntingtin (Htt) phosphorylation is a molecular switch regulating Htt aggregation, helical conformation, internalization, and nuclear targeting. J Biol Chem. Di Pardo, A., Maglione, V., Alpaugh, M., Horkey, M., Atwal, R.S., Sassone, J., Ciammola, A., Steffan, J.S., Fouad, K., Truant, R., et al. (2012). Ganglioside GM1 induces phosphorylation of mutant huntingtin and restores normal motor behavior in Huntington disease mice. Proc Natl Acad Sci U S A 109, 3528- 3533. Diaz, M., Sanchez-Barrena, M.J., Gonzalez-Rubio, J.M., Rodriguez, L., Fernandez, D., Antoni, R., Yunta, C., Belda-Palazon, B., Gonzalez-Guzman, M., Peirats-Llobet, M., et al. (2016). Calcium-dependent oligomerization of CAR proteins at cell membrane modulates ABA signaling. Proc Natl Acad Sci U S A 113, E396-405. Dick, R.A., Kamynina, E., and Vogt, V.M. (2013). Effect of multimerization on membrane association of Rous sarcoma virus and HIV-1 matrix domain proteins. J Virol 87, 13598-13608. DiGiovanni, L.F., Mocle, A.J., Xia, J., and Truant, R. (2016). Huntingtin N17 domain is a reactive oxygen species sensor regulating huntingtin phosphorylation and localization. Hum Mol Genet 25, 3937- 3945. Draeger, A., Monastyrskaya, K., and Babiychuk, E.B. (2011). Plasma membrane repair and cellular damage control: the annexin survival kit. Biochem Pharmacol 81, 703-712. Drescher, M., Veldhuis, G., van Rooijen, B.D., Milikisyants, S., Subramaniam, V., and Huber, M. (2008). Antiparallel arrangement of the helices of vesicle-bound alpha-synuclein. J Am Chem Soc 130, 7796- 7797. Duennwald, M.L., and Lindquist, S. (2008). Impaired ERAD and ER stress are early and specific events in polyglutamine toxicity. Genes Dev 22, 3308-3319. Emoto, M.C., Sato-Akaba, H., Matsuoka, Y., Yamada, K.I., and Fujii, H.G. (2019). Non-invasive mapping of glutathione levels in mouse brains by in vivo electron paramagnetic resonance (EPR) imaging: Applied to a kindling mouse model. Neurosci Lett 690, 6-10. Epel, B., Redler, G., and Halpern, H.J. (2014). How in vivo EPR measures and images oxygen. Adv Exp Med Biol 812, 113-119. Fahy, E., Cotter, D., Sud, M., and Subramaniam, S. (2011). Lipid classification, structures and tools. Biochim Biophys Acta 1811, 637-647. Fandrich, M., Nystrom, S., Nilsson, K.P.R., Bockmann, A., LeVine, H., 3rd, and Hammarstrom, P. (2018). Amyloid fibril polymorphism: a challenge for molecular imaging and therapy. J Intern Med 283, 218-237. 119 Feric, M., Vaidya, N., Harmon, T.S., Mitrea, D.M., Zhu, L., Richardson, T.M., Kriwacki, R.W., Pappu, R.V., and Brangwynne, C.P. (2016). Coexisting Liquid Phases Underlie Nucleolar Subcompartments. Cell 165, 1686-1697. Fischer, T., Lu, L., Haigler, H.T., and Langen, R. (2007). Annexin B12 is a sensor of membrane curvature and undergoes major curvature-dependent structural changes. J Biol Chem 282, 9996-10004. Fisette, O., Paslack, C., Barnes, R., Isas, J.M., Langen, R., Heyden, M., Han, S., and Schafer, L.V. (2016). Hydration Dynamics of a Peripheral Membrane Protein. J Am Chem Soc 138, 11526-11535. Franck, J.M., Pavlova, A., Scott, J.A., and Han, S. (2013). Quantitative cw Overhauser effect dynamic nuclear polarization for the analysis of local water dynamics. Prog Nucl Magn Reson Spectrosc 74, 33-56. Franzmann, T.M., and Alberti, S. (2019). Protein Phase Separation as a Stress Survival Strategy. Cold Spring Harb Perspect Biol 11. Frottin, F., Schueder, F., Tiwary, S., Gupta, R., Korner, R., Schlichthaerle, T., Cox, J., Jungmann, R., Hartl, F.U., and Hipp, M.S. (2019). The nucleolus functions as a phase-separated protein quality control compartment. Science 365, 342-347. Gafni, J., Hermel, E., Young, J.E., Wellington, C.L., Hayden, M.R., and Ellerby, L.M. (2004). Inhibition of calpain cleavage of huntingtin reduces toxicity: accumulation of calpain/caspase fragments in the nucleus. J Biol Chem 279, 20211-20220. Gallardo, R., Ranson, N.A., and Radford, S.E. (2019). Amyloid structures: much more than just a cross- beta fold. Curr Opin Struct Biol 60, 7-16. Gao, Y.G., Yan, X.Z., Song, A.X., Chang, Y.G., Gao, X.C., Jiang, N., Zhang, Q., and Hu, H.Y. (2006). Structural insights into the specific binding of huntingtin proline-rich region with the SH3 and WW domains. Structure 14, 1755-1765. Gauthier, L.R., Charrin, B.C., Borrell-Pages, M., Dompierre, J.P., Rangone, H., Cordelieres, F.P., De Mey, J., MacDonald, M.E., Lessmann, V., Humbert, S., et al. (2004). Huntingtin controls neurotrophic support and survival of neurons by enhancing BDNF vesicular transport along microtubules. Cell 118, 127-138. Gazit, E. (2002). A possible role for pi-stacking in the self-assembly of amyloid fibrils. FASEB J 16, 77- 83. Geddes, A.J., Parker, K.D., Atkins, E.D., and Beighton, E. (1968). "Cross-beta" conformation in proteins. J Mol Biol 32, 343-358. Gelman, A., Rawet-Slobodkin, M., and Elazar, Z. (2015). Huntingtin facilitates selective autophagy. Nat Cell Biol 17, 214-215. Georgieva, E.R., Ramlall, T.F., Borbat, P.P., Freed, J.H., and Eliezer, D. (2008). Membrane-bound alpha- synuclein forms an extended helix: long-distance pulsed ESR measurements using vesicles, bicelles, and rodlike micelles. J Am Chem Soc 130, 12856-12857. Gerke, V., Creutz, C.E., and Moss, S.E. (2005). Annexins: linking Ca2+ signalling to membrane dynamics. Nat Rev Mol Cell Biol 6, 449-461. 120 Gharibyan, A.L., Islam, T., Pettersson, N., Golchin, S.A., Lundgren, J., Johansson, G., Genot, M., Schultz, N., Wennstrom, M., and Olofsson, A. (2020). Apolipoprotein E Interferes with IAPP Aggregation and Protects Pericytes from IAPP-Induced Toxicity. Biomolecules 10. Gilmanshin, R., Creutz, C.E., and Tamm, L.K. (1994). Annexin IV reduces the rate of lateral lipid diffusion and changes the fluid phase structure of the lipid bilayer when it binds to negatively charged membranes in the presence of calcium. Biochemistry 33, 8225-8232. Gomes, E., and Shorter, J. (2019). The molecular language of membraneless organelles. J Biol Chem 294, 7115-7127. Gomez-Tortosa, E., del Barrio, A., Garcia Ruiz, P.J., Pernaute, R.S., Benitez, J., Barroso, A., Jimenez, F.J., and Garcia Yebenes, J. (1998). Severity of cognitive impairment in juvenile and late-onset Huntington disease. Arch Neurol 55, 835-843. Gorbenko, G.P., and Kinnunen, P.K. (2006). The role of lipid-protein interactions in amyloid-type protein fibril formation. Chem Phys Lipids 141, 72-82. Graham, R.K., Deng, Y., Slow, E.J., Haigh, B., Bissada, N., Lu, G., Pearson, J., Shehadeh, J., Bertram, L., Murphy, Z., et al. (2006). Cleavage at the caspase-6 site is required for neuronal dysfunction and degeneration due to mutant huntingtin. Cell 125, 1179-1191. Grill, D., Matos, A.L.L., de Vries, W.C., Kudruk, S., Heflik, M., Dorner, W., Mootz, H.D., Jan Ravoo, B., Galla, H.J., and Gerke, V. (2018). Bridging of membrane surfaces by annexin A2. Sci Rep 8, 14662. Gu, X., Greiner, E.R., Mishra, R., Kodali, R., Osmand, A., Finkbeiner, S., Steffan, J.S., Thompson, L.M., Wetzel, R., and Yang, X.W. (2009). Serines 13 and 16 are critical determinants of full-length human mutant huntingtin induced disease pathogenesis in HD mice. Neuron 64, 828-840. Guerrero-Ferreira, R., Taylor, N.M., Mona, D., Ringler, P., Lauer, M.E., Riek, R., Britschgi, M., and Stahlberg, H. (2018). Cryo-EM structure of alpha-synuclein fibrils. Elife 7. Gunawardena, S., Her, L.S., Brusch, R.G., Laymon, R.A., Niesman, I.R., Gordesky-Gold, B., Sintasath, L., Bonini, N.M., and Goldstein, L.S. (2003). Disruption of axonal transport by loss of huntingtin or expression of pathogenic polyQ proteins in Drosophila. Neuron 40, 25-40. Guthmann, M., Burton, A., and Torres-Padilla, M.E. (2019). Expression and phase separation potential of heterochromatin proteins during early mouse development. EMBO Rep 20, e47952. Hakobyan, D., Gerke, V., and Heuer, A. (2017). Modeling of annexin A2-Membrane interactions by molecular dynamics simulations. PLoS One 12, e0185440. Harayama, T., and Riezman, H. (2018). Understanding the diversity of membrane lipid composition. Nat Rev Mol Cell Biol 19, 281-296. Hawkins, T.E., Merrifield, C.J., and Moss, S.E. (2000). Calcium signaling and annexins. Cell Biochem Biophys 33, 275-296. Hu, N.J., Bradshaw, J., Lauter, H., Buckingham, J., Solito, E., and Hofmann, A. (2008). Membrane- induced folding and structure of membrane-bound annexin A1 N-terminal peptides: implications for annexin-induced membrane aggregation. Biophys J 94, 1773-1781. 121 Huang, S., Liu, N., Li, H., Zhao, J., Su, L., Zhang, Y., Zhang, S., Zhao, B., and Miao, J. (2014). TIA1 interacts with annexin A7 in regulating vascular endothelial cell autophagy. Int J Biochem Cell Biol 57, 115-122. Hubbell, W.L., Gross, A., Langen, R., and Lietzow, M.A. (1998). Recent advances in site-directed spin labeling of proteins. Curr Opin Struct Biol 8, 649-656. Iadanza, M.G., Jackson, M.P., Hewitt, E.W., Ranson, N.A., and Radford, S.E. (2018). A new era for understanding amyloid structures and disease. Nat Rev Mol Cell Biol 19, 755-773. Ikeda, M., Kihara, A., and Igarashi, Y. (2006). Lipid asymmetry of the eukaryotic plasma membrane: functions and related enzymes. Biol Pharm Bull 29, 1542-1546. Isas, J.M., Cartailler, J.P., Sokolov, Y., Patel, D.R., Langen, R., Luecke, H., Hall, J.E., and Haigler, H.T. (2000). Annexins V and XII insert into bilayers at mildly acidic pH and form ion channels. Biochemistry 39, 3015-3022. Isas, J.M., Langen, A., Isas, M.C., Pandey, N.K., and Siemer, A.B. (2017). Formation and Structure of Wild Type Huntingtin Exon-1 Fibrils. Biochemistry 56, 3579-3586. Isas, J.M., Langen, R., Haigler, H.T., and Hubbell, W.L. (2002). Structure and dynamics of a helical hairpin and loop region in annexin 12: a site-directed spin labeling study. Biochemistry 41, 1464-1473. Isas, J.M., Langen, R., Hubbell, W.L., and Haigler, H.T. (2004). Structure and dynamics of a helical hairpin that mediates calcium-dependent membrane binding of annexin B12. J Biol Chem 279, 32492- 32498. Iuchi, S., Hoffner, G., Verbeke, P., Djian, P., and Green, H. (2003). Oligomeric and polymeric aggregates formed by proteins containing expanded polyglutamine. Proc Natl Acad Sci U S A 100, 2409-2414. Jain, S., Wheeler, J.R., Walters, R.W., Agrawal, A., Barsic, A., and Parker, R. (2016). ATPase-Modulated Stress Granules Contain a Diverse Proteome and Substructure. Cell 164, 487-498. Janosi, L., Li, Z., Hancock, J.F., and Gorfe, A.A. (2012). Organization, dynamics, and segregation of Ras nanoclusters in membrane domains. Proc Natl Acad Sci U S A 109, 8097-8102. Jao, C.C., Hegde, B.G., Chen, J., Haworth, I.S., and Langen, R. (2008). Structure of membrane-bound alpha-synuclein from site-directed spin labeling and computational refinement. Proc Natl Acad Sci U S A 105, 19666-19671. Jayasinghe, S.A., and Langen, R. (2005). Lipid membranes modulate the structure of islet amyloid polypeptide. Biochemistry 44, 12113-12119. Jayasinghe, S.A., and Langen, R. (2007). Membrane interaction of islet amyloid polypeptide. Biochim Biophys Acta 1768, 2002-2009. Jones, E.M., Dubey, M., Camp, P.J., Vernon, B.C., Biernat, J., Mandelkow, E., Majewski, J., and Chi, E.Y. (2012). Interaction of tau protein with model lipid membranes induces tau structural compaction and membrane disruption. Biochemistry 51, 2539-2550. 122 Junker, M., and Creutz, C.E. (1993). Endonexin (annexin IV)-mediated lateral segregation of phosphatidylglycerol in phosphatidylglycerol/phosphatidylcholine membranes. Biochemistry 32, 9968- 9974. Kahraman, O., Koch, P.D., Klug, W.S., and Haselwandter, C.A. (2016a). Architecture and Function of Mechanosensitive Membrane Protein Lattices. Sci Rep 6, 19214. Kahraman, O., Koch, P.D., Klug, W.S., and Haselwandter, C.A. (2016b). Bilayer-thickness-mediated interactions between integral membrane proteins. Phys Rev E 93, 042410. Kahraman, O., Langen, R., and Haselwandter, C.A. (2018). Directed Supramolecular Organization of N- BAR Proteins through Regulation of H0 Membrane Immersion Depth. Sci Rep 8, 16383. Kaminker, I., Barnes, R., and Han, S. (2015). Overhauser Dynamic Nuclear Polarization Studies on Local Water Dynamics. Methods Enzymol 564, 457-483. Kardos, J., Yamamoto, K., Hasegawa, K., Naiki, H., and Goto, Y. (2004). Direct measurement of the thermodynamic parameters of amyloid formation by isothermal titration calorimetry. J Biol Chem 279, 55308-55314. Kay, B.K., Williamson, M.P., and Sudol, M. (2000). The importance of being proline: the interaction of proline-rich motifs in signaling proteins with their cognate domains. FASEB J 14, 231-241. Kenis, H., van Genderen, H., Bennaghmouch, A., Rinia, H.A., Frederik, P., Narula, J., Hofstra, L., and Reutelingsperger, C.P. (2004). Cell surface-expressed phosphatidylserine and annexin A5 open a novel portal of cell entry. J Biol Chem 279, 52623-52629. Khabibullaev, P.K., and Abdulla Saidov. (2013). Phase separation in soft matter physics: micellar solutions, microemulsions, critical phenomena, Vol 138. Khong, A., Matheny, T., Jain, S., Mitchell, S.F., Wheeler, J.R., and Parker, R. (2017). The Stress Granule Transcriptome Reveals Principles of mRNA Accumulation in Stress Granules. Mol Cell 68, 808-820 e805. Khurana, R., Uversky, V.N., Nielsen, L., and Fink, A.L. (2001). Is Congo red an amyloid-specific dye? J Biol Chem 276, 22715-22721. Kim, Y.J., Yi, Y., Sapp, E., Wang, Y., Cuiffo, B., Kegel, K.B., Qin, Z.H., Aronin, N., and DiFiglia, M. (2001). Caspase 3-cleaved N-terminal fragments of wild-type and mutant huntingtin are present in normal and Huntington's disease brains, associate with membranes, and undergo calpain-dependent proteolysis. Proc Natl Acad Sci U S A 98, 12784-12789. Kirkwood, S.C., Su, J.L., Conneally, P., and Foroud, T. (2001). Progression of symptoms in the early and middle stages of Huntington disease. Arch Neurol 58, 273-278. Klein, W.L., Stine, W.B., Jr., and Teplow, D.B. (2004). Small assemblies of unmodified amyloid beta- protein are the proximate neurotoxin in Alzheimer's disease. Neurobiol Aging 25, 569-580. Klosin, A., Oltsch, F., Harmon, T., Honigmann, A., Julicher, F., Hyman, A.A., and Zechner, C. (2020). Phase separation provides a mechanism to reduce noise in cells. Science 367, 464-468. 123 Klug, C.S., and Feix, J.B. (2008). Methods and applications of site-directed spin labeling EPR spectroscopy. Methods Cell Biol 84, 617-658. Kreilaus, F., Spiro, A.S., McLean, C.A., Garner, B., and Jenner, A.M. (2016). Evidence for altered cholesterol metabolism in Huntington's disease post mortem brain tissue. Neuropathol Appl Neurobiol 42, 535-546. Langen, R., Isas, J.M., Luecke, H., Haigler, H.T., and Hubbell, W.L. (1998). Membrane-mediated assembly of annexins studied by site-directed spin labeling. J Biol Chem 273, 22453-22457. Lauritzen, S.P., Boye, T.L., and Nylandsted, J. (2015). Annexins are instrumental for efficient plasma membrane repair in cancer cells. Semin Cell Dev Biol 45, 32-38. Lee, C.Y., Cantle, J.P., and Yang, X.W. (2013). Genetic manipulations of mutant huntingtin in mice: new insights into Huntington's disease pathogenesis. FEBS J 280, 4382-4394. Lee, W.C., Yoshihara, M., and Littleton, J.T. (2004). Cytoplasmic aggregates trap polyglutamine- containing proteins and block axonal transport in a Drosophila model of Huntington's disease. Proc Natl Acad Sci U S A 101, 3224-3229. Leventis, P.A., and Grinstein, S. (2010). The distribution and function of phosphatidylserine in cellular membranes. Annu Rev Biophys 39, 407-427. Levy, G.R., Shen, K., Gavrilov, Y., Smith, P.E.S., Levy, Y., Chan, R., Frydman, J., and Frydman, L. (2018). Huntingtin's N-Terminus Rearrangements in the Presence of Membranes: A Joint Spectroscopic and Computational Perspective. ACS Chem Neurosci. Li, H., Huang, S., Wang, S., Wang, L., Qi, L., Zhang, Y., Zhang, S., Zhao, B., and Miao, J. (2013a). Relationship between annexin A7 and integrin beta4 in autophagy. Int J Biochem Cell Biol 45, 2605- 2611. Li, Q., Peng, X., Li, Y., Tang, W., Zhu, J., Huang, J., Qi, Y., and Zhang, Z. (2020). LLPSDB: a database of proteins undergoing liquid-liquid phase separation in vitro. Nucleic Acids Res 48, D320-D327. Li, Y., Zhang, X., and Cao, D. (2013b). The role of shape complementarity in the protein-protein interactions. Sci Rep 3, 3271. Liang, B., and Tamm, L.K. (2016). NMR as a tool to investigate the structure, dynamics and function of membrane proteins. Nat Struct Mol Biol 23, 468-474. Liao, Y.C., Fernandopulle, M.S., Wang, G., Choi, H., Hao, L., Drerup, C.M., Patel, R., Qamar, S., Nixon- Abell, J., Shen, Y., et al. (2019). RNA Granules Hitchhike on Lysosomes for Long-Distance Transport, Using Annexin A11 as a Molecular Tether. Cell 179, 147-164 e120. Liu, K.Y., Shyu, Y.C., Barbaro, B.A., Lin, Y.T., Chern, Y., Thompson, L.M., James Shen, C.K., and Marsh, J.L. (2015). Disruption of the nuclear membrane by perinuclear inclusions of mutant huntingtin causes cell-cycle re-entry and striatal cell death in mouse and cell models of Huntington's disease. Hum Mol Genet 24, 1602-1616. Lizarbe, M.A., Barrasa, J.I., Olmo, N., Gavilanes, F., and Turnay, J. (2013). Annexin-phospholipid interactions. Functional implications. Int J Mol Sci 14, 2652-2683. 124 Luhrs, T., Ritter, C., Adrian, M., Riek-Loher, D., Bohrmann, B., Dobeli, H., Schubert, D., and Riek, R. (2005). 3D structure of Alzheimer's amyloid-beta(1-42) fibrils. Proc Natl Acad Sci U S A 102, 17342- 17347. Lumb, K.J., and Kim, P.S. (1995). Measurement of interhelical electrostatic interactions in the GCN4 leucine zipper. Science 268, 436-439. Luo, Y., Na, Z., and Slavoff, S.A. (2018). P-Bodies: Composition, Properties, and Functions. Biochemistry 57, 2424-2431. Ma, Y., Poole, K., Goyette, J., and Gaus, K. (2017). Introducing Membrane Charge and Membrane Potential to T Cell Signaling. Front Immunol 8, 1513. Macdonald, P.M., and Seelig, J. (1988). Anion binding to neutral and positively charged lipid membranes. Biochemistry 27, 6769-6775. Mahboubi, H., and Stochaj, U. (2017). Cytoplasmic stress granules: Dynamic modulators of cell signaling and disease. Biochim Biophys Acta Mol Basis Dis 1863, 884-895. Mangiarini, L., Sathasivam, K., Seller, M., Cozens, B., Harper, A., Hetherington, C., Lawton, M., Trottier, Y., Lehrach, H., Davies, S.W., et al. (1996). Exon 1 of the HD gene with an expanded CAG repeat is sufficient to cause a progressive neurological phenotype in transgenic mice. Cell 87, 493-506. Markmiller, S., Soltanieh, S., Server, K.L., Mak, R., Jin, W., Fang, M.Y., Luo, E.C., Krach, F., Yang, D., Sen, A., et al. (2018). Context-Dependent and Disease-Specific Diversity in Protein Interactions within Stress Granules. Cell 172, 590-604 e513. Marshall, K.E., Morris, K.L., Charlton, D., O'Reilly, N., Lewis, L., Walden, H., and Serpell, L.C. (2011). Hydrophobic, aromatic, and electrostatic interactions play a central role in amyloid fibril formation and stability. Biochemistry 50, 2061-2071. Martin, D.D., Ladha, S., Ehrnhoefer, D.E., and Hayden, M.R. (2015). Autophagy in Huntington disease and huntingtin in autophagy. Trends Neurosci 38, 26-35. Martin, E.W., and Mittag, T. (2018). Relationship of Sequence and Phase Separation in Protein Low- Complexity Regions. Biochemistry 57, 2478-2487. Matsushima, N., Creutz, C.E., and Kretsinger, R.H. (1990). Polyproline, beta-turn helices. Novel secondary structures proposed for the tandem repeats within rhodopsin, synaptophysin, synexin, gliadin, RNA polymerase II, hordein, and gluten. Proteins 7, 125-155. Matsushima, N., Yoshida, H., Kumaki, Y., Kamiya, M., Tanaka, T., Izumi, Y., and Kretsinger, R.H. (2008). Flexible structures and ligand interactions of tandem repeats consisting of proline, glycine, asparagine, serine, and/or threonine rich oligopeptides in proteins. Curr Protein Pept Sci 9, 591-610. McCulloch, K.M., Yamakawa, I., Shifrin, D.A., Jr., McConnell, R.E., Foegeding, N.J., Singh, P.K., Mao, S., Tyska, M.J., and Iverson, T.M. (2019). An alternative N-terminal fold of the intestine-specific annexin A13a induces dimerization and regulates membrane-binding. J Biol Chem 294, 3454-3463. McHaourab, H.S., Lietzow, M.A., Hideg, K., and Hubbell, W.L. (1996). Motion of spin-labeled side chains in T4 lysozyme. Correlation with protein structure and dynamics. Biochemistry 35, 7692-7704. 125 McHaourab, H.S., Oh, K.J., Fang, C.J., and Hubbell, W.L. (1997). Conformation of T4 lysozyme in solution. Hinge-bending motion and the substrate-induced conformational transition studied by site- directed spin labeling. Biochemistry 36, 307-316. McMahon, H.T., and Boucrot, E. (2015). Membrane curvature at a glance. J Cell Sci 128, 1065-1070. McMahon, H.T., and Gallop, J.L. (2005). Membrane curvature and mechanisms of dynamic cell membrane remodelling. Nature 438, 590-596. McSwiggen, D.T., Mir, M., Darzacq, X., and Tjian, R. (2019). Evaluating phase separation in live cells: diagnosis, caveats, and functional consequences. Genes Dev 33, 1619-1634. Megli, F.M., Selvaggi, M., Liemann, S., Quagliariello, E., and Huber, R. (1998). The calcium-dependent binding of annexin V to phospholipid vesicles influences the bilayer inner fluidity gradient. Biochemistry 37, 10540-10546. Meng, K., Lu, S., Yan, X., Sun, Y., Gao, J., Wang, Y., Yin, X., Sun, Z., and He, Q.Y. (2020). Quantitative Mitochondrial Proteomics Reveals ANXA7 as a Crucial Factor in Mitophagy. J Proteome Res. Meszaros, B., Erdos, G., Szabo, B., Schad, E., Tantos, A., Abukhairan, R., Horvath, T., Murvai, N., Kovacs, O.P., Kovacs, M., et al. (2020). PhaSePro: the database of proteins driving liquid-liquid phase separation. Nucleic Acids Res 48, D360-D367. Michalek, M., Salnikov, E.S., and Bechinger, B. (2013a). Structure and topology of the huntingtin 1-17 membrane anchor by a combined solution and solid-state NMR approach. Biophys J 105, 699-710. Michalek, M., Salnikov, E.S., Werten, S., and Bechinger, B. (2013b). Membrane interactions of the amphipathic amino terminus of huntingtin. Biochemistry 52, 847-858. Mier, P., Paladin, L., Tamana, S., Petrosian, S., Hajdu-Soltesz, B., Urbanek, A., Gruca, A., Plewczynski, D., Grynberg, M., Bernado, P., et al. (2019). Disentangling the complexity of low complexity proteins. Brief Bioinform. Milanesi, L., Sheynis, T., Xue, W.F., Orlova, E.V., Hellewell, A.L., Jelinek, R., Hewitt, E.W., Radford, S.E., and Saibil, H.R. (2012). Direct three-dimensional visualization of membrane disruption by amyloid fibrils. Proc Natl Acad Sci U S A 109, 20455-20460. Mishra, R., Hoop, C.L., Kodali, R., Sahoo, B., van der Wel, P.C., and Wetzel, R. (2012). Serine phosphorylation suppresses huntingtin amyloid accumulation by altering protein aggregation properties. J Mol Biol 424, 1-14. Mitrea, D.M., and Kriwacki, R.W. (2016). Phase separation in biology; functional organization of a higher order. Cell Commun Signal 14, 1. Mo, Y., Campos, B., Mealy, T.R., Commodore, L., Head, J.F., Dedman, J.R., and Seaton, B.A. (2003). Interfacial basic cluster in annexin V couples phospholipid binding and trimer formation on membrane surfaces. J Biol Chem 278, 2437-2443. Morozova, D., Guigas, G., and Weiss, M. (2011). Dynamic structure formation of peripheral membrane proteins. PLoS Comput Biol 7, e1002067. 126 Moss, S.E., and Morgan, R.O. (2004). The annexins. Genome Biol 5, 219. Myerson, A.S., and A. F. Izmailov. (1993). The structure of supersaturated solutions. In Fundamentals, pp. 249-306. Naidu, D.G., Raha, A., Chen, X.L., Spitzer, A.R., and Chander, A. (2005). Partial truncation of the NH2- terminus affects physical characteristics and membrane binding, aggregation, and fusion properties of annexin A7. Biochim Biophys Acta 1734, 152-168. Namkoong, S., Ho, A., Woo, Y.M., Kwak, H., and Lee, J.H. (2018). Systematic Characterization of Stress-Induced RNA Granulation. Mol Cell 70, 175-187 e178. Nawrocki, G., Wang, P.H., Yu, I., Sugita, Y., and Feig, M. (2017). Slow-Down in Diffusion in Crowded Protein Solutions Correlates with Transient Cluster Formation. J Phys Chem B 121, 11072-11084. Neueder, A., Landles, C., Ghosh, R., Howland, D., Myers, R.H., Faull, R.L.M., Tabrizi, S.J., and Bates, G.P. (2017). The pathogenic exon 1 HTT protein is produced by incomplete splicing in Huntington's disease patients. Sci Rep 7, 1307. Newcombe, E.A., Ruff, K.M., Sethi, A., Ormsby, A.R., Ramdzan, Y.M., Fox, A., Purcell, A.W., Gooley, P.R., Pappu, R.V., and Hatters, D.M. (2018). Tadpole-like Conformations of Huntingtin Exon 1 Are Characterized by Conformational Heterogeneity that Persists regardless of Polyglutamine Length. J Mol Biol. Ochaba, J., Lukacsovich, T., Csikos, G., Zheng, S., Margulis, J., Salazar, L., Mao, K., Lau, A.L., Yeung, S.Y., Humbert, S., et al. (2014). Potential function for the Huntingtin protein as a scaffold for selective autophagy. Proc Natl Acad Sci U S A 111, 16889-16894. Overington, J.P., Al-Lazikani, B., and Hopkins, A.L. (2006). How many drug targets are there? Nat Rev Drug Discov 5, 993-996. Pal, A., Severin, F., Lommer, B., Shevchenko, A., and Zerial, M. (2006). Huntingtin-HAP40 complex is a novel Rab5 effector that regulates early endosome motility and is up-regulated in Huntington's disease. J Cell Biol 172, 605-618. Panchal, R.G., Ruthel, G., Kenny, T.A., Kallstrom, G.H., Lane, D., Badie, S.S., Li, L., Bavari, S., and Aman, M.J. (2003). In vivo oligomerization and raft localization of Ebola virus protein VP40 during vesicular budding. Proc Natl Acad Sci U S A 100, 15936-15941. Pandey, N.K., Isas, J.M., Rawat, A., Lee, R.V., Langen, J., Pandey, P., and Langen, R. (2018). The 17- residue-long N terminus in huntingtin controls stepwise aggregation in solution and on membranes via different mechanisms. J Biol Chem 293, 2597-2605. Paoletti, P., Bellone, C., and Zhou, Q. (2013). NMDA receptor subunit diversity: impact on receptor properties, synaptic plasticity and disease. Nat Rev Neurosci 14, 383-400. Patel, A., Lee, H.O., Jawerth, L., Maharana, S., Jahnel, M., Hein, M.Y., Stoynov, S., Mahamid, J., Saha, S., Franzmann, T.M., et al. (2015). A Liquid-to-Solid Phase Transition of the ALS Protein FUS Accelerated by Disease Mutation. Cell 162, 1066-1077. 127 Patel, D.R., Isas, J.M., Ladokhin, A.S., Jao, C.C., Kim, Y.E., Kirsch, T., Langen, R., and Haigler, H.T. (2005). The conserved core domains of annexins A1, A2, A5, and B12 can be divided into two groups with different Ca2+-dependent membrane-binding properties. Biochemistry 44, 2833-2844. Paulsen, J.S. (2011). Cognitive impairment in Huntington disease: diagnosis and treatment. Curr Neurol Neurosci Rep 11, 474-483. Pawson, T., and Scott, J.D. (1997). Signaling through scaffold, anchoring, and adaptor proteins. Science 278, 2075-2080. Pilbrow, J.R. (1988). New insights and trends in transition metal ion electron paramagnetic resonance. Bull Magn Reson 10, pp.32-64. Poyry, S., and Vattulainen, I. (2016). Role of charged lipids in membrane structures - Insight given by simulations. Biochim Biophys Acta 1858, 2322-2333. Qamar, S., Wang, G., Randle, S.J., Ruggeri, F.S., Varela, J.A., Lin, J.Q., Phillips, E.C., Miyashita, A., Williams, D., Strohl, F., et al. (2018). FUS Phase Separation Is Modulated by a Molecular Chaperone and Methylation of Arginine Cation-pi Interactions. Cell 173, 720-734 e715. Quiroz, F.G., Fiore, V.F., Levorse, J., Polak, L., Wong, E., Pasolli, H.A., and Fuchs, E. (2020). Liquid- liquid phase separation drives skin barrier formation. Science 367. Rawat, A., Langen, R., and Varkey, J. (2018). Membranes as modulators of amyloid protein misfolding and target of toxicity. Biochim Biophys Acta Biomembr. Reeves, J.P., and Dowben, R.M. (1969). Formation and properties of thin-walled phospholipid vesicles. J Cell Physiol 73, 49-60. Reiner, A., Albin, R.L., Anderson, K.D., D'Amato, C.J., Penney, J.B., and Young, A.B. (1988). Differential loss of striatal projection neurons in Huntington disease. Proc Natl Acad Sci U S A 85, 5733- 5737. Richfield, E.K., Maguire-Zeiss, K.A., Vonkeman, H.E., and Voorn, P. (1995). Preferential loss of preproenkephalin versus preprotachykinin neurons from the striatum of Huntington's disease patients. Ann Neurol 38, 852-861. Rockabrand, E., Slepko, N., Pantalone, A., Nukala, V.N., Kazantsev, A., Marsh, J.L., Sullivan, P.G., Steffan, J.S., Sensi, S.L., and Thompson, L.M. (2007). The first 17 amino acids of Huntingtin modulate its sub-cellular localization, aggregation and effects on calcium homeostasis. Hum Mol Genet 16, 61-77. Rocken, C., and Sletten, K. (2003). Amyloid in surgical pathology. Virchows Arch 443, 3-16. Ross, J.S., Slodkowska, E.A., Symmans, W.F., Pusztai, L., Ravdin, P.M., and Hortobagyi, G.N. (2009). The HER-2 receptor and breast cancer: ten years of targeted anti-HER-2 therapy and personalized medicine. Oncologist 14, 320-368. Ryan, V.H., Dignon, G.L., Zerze, G.H., Chabata, C.V., Silva, R., Conicella, A.E., Amaya, J., Burke, K.A., Mittal, J., and Fawzi, N.L. (2018). Mechanistic View of hnRNPA2 Low-Complexity Domain Structure, Interactions, and Phase Separation Altered by Mutation and Arginine Methylation. Mol Cell 69, 465-479 e467. 128 Sabate, R., and Ventura, S. (2013). Cross-beta-sheet supersecondary structure in amyloid folds: techniques for detection and characterization. Methods Mol Biol 932, 237-257. Sahu, I.D., McCarrick, R.M., and Lorigan, G.A. (2013). Use of electron paramagnetic resonance to solve biochemical problems. Biochemistry 52, 5967-5984. Sameni, S., Malacrida, L., Tan, Z., and Digman, M.A. (2018). Alteration in Fluidity of Cell Plasma Membrane in Huntington Disease Revealed by Spectral Phasor Analysis. Sci Rep 8, 734. Sathasivam, K., Neueder, A., Gipson, T.A., Landles, C., Benjamin, A.C., Bondulich, M.K., Smith, D.L., Faull, R.L., Roos, R.A., Howland, D., et al. (2013). Aberrant splicing of HTT generates the pathogenic exon 1 protein in Huntington disease. Proc Natl Acad Sci U S A 110, 2366-2370. Scherzinger, E., Lurz, R., Turmaine, M., Mangiarini, L., Hollenbach, B., Hasenbank, R., Bates, G.P., Davies, S.W., Lehrach, H., and Wanker, E.E. (1997). Huntingtin-encoded polyglutamine expansions form amyloid-like protein aggregates in vitro and in vivo. Cell 90, 549-558. Sfakianos, A.P., Mellor, L.E., Pang, Y.F., Kritsiligkou, P., Needs, H., Abou-Hamdan, H., Desaubry, L., Poulin, G.B., Ashe, M.P., and Whitmarsh, A.J. (2018). The mTOR-S6 kinase pathway promotes stress granule assembly. Cell Death Differ 25, 1766-1780. Sheinerman, F.B., Norel, R., and Honig, B. (2000). Electrostatic aspects of protein-protein interactions. Curr Opin Struct Biol 10, 153-159. Shibata, H., Suzuki, H., Kakiuchi, T., Inuzuka, T., Yoshida, H., Mizuno, T., and Maki, M. (2008). Identification of Alix-type and Non-Alix-type ALG-2-binding sites in human phospholipid scramblase 3: differential binding to an alternatively spliced isoform and amino acid-substituted mutants. J Biol Chem 283, 9623-9632. Shrivastava, A.N., Aperia, A., Melki, R., and Triller, A. (2017). Physico-Pathologic Mechanisms Involved in Neurodegeneration: Misfolded Protein-Plasma Membrane Interactions. Neuron 95, 33-50. Simons, K., and Toomre, D. (2000). Lipid rafts and signal transduction. Nat Rev Mol Cell Biol 1, 31-39. Singer, S.J., and Nicolson, G.L. (1972). The fluid mosaic model of the structure of cell membranes. Science 175, 720-731. Sivanandam, V.N., Jayaraman, M., Hoop, C.L., Kodali, R., Wetzel, R., and van der Wel, P.C. (2011). The aggregation-enhancing huntingtin N-terminus is helical in amyloid fibrils. J Am Chem Soc 133, 4558- 4566. Smith, M.A., Brandt, J., and Shadmehr, R. (2000). Motor disorder in Huntington's disease begins as a dysfunction in error feedback control. Nature 403, 544-549. Sokolov, Y., Mailliard, W.S., Tranngo, N., Isas, M., Luecke, H., Haigler, H.T., and Hall, J.E. (2000). Annexins V and XII alter the properties of planar lipid bilayers seen by conductance probes. J Gen Physiol 115, 571-582. Sonder, S.L., Boye, T.L., Tolle, R., Dengjel, J., Maeda, K., Jaattela, M., Simonsen, A.C., Jaiswal, J.K., and Nylandsted, J. (2019). Annexin A7 is required for ESCRT III-mediated plasma membrane repair. Sci Rep 9, 6726. 129 Spasojevic, I. (2011). Free radicals and antioxidants at a glance using EPR spectroscopy. Crit Rev Clin Lab Sci 48, 114-142. Squitieri, F., Cannella, M., Sgarbi, G., Maglione, V., Falleni, A., Lenzi, P., Baracca, A., Cislaghi, G., Saft, C., Ragona, G., et al. (2006). Severe ultrastructural mitochondrial changes in lymphoblasts homozygous for Huntington disease mutation. Mech Ageing Dev 127, 217-220. Stangl, M., and Schneider, D. (2015). Functional competition within a membrane: Lipid recognition vs. transmembrane helix oligomerization. Biochim Biophys Acta 1848, 1886-1896. Statt, A., Casademunt, H., Brangwynne, C.P., and Panagiotopoulos, A.Z. (2020). Model for disordered proteins with strongly sequence-dependent liquid phase behavior. J Chem Phys 152, 075101. Stopa, B., Piekarska, B., Konieczny, L., Rybarska, J., Spolnik, P., Zemanek, G., Roterman, I., and Krol, M. (2003). The structure and protein binding of amyloid-specific dye reagents. Acta Biochim Pol 50, 1213-1227. Style, R.W., Sai, T., Fanelli, N., Ijavi, M., Smith-Mannschott, K., Xu, Q., Wilen, L.A. and Dufresne, E.R. (2018). Liquid-liquid phase separation in an elastic network. Physical Review X 8, 011028. Suopanki, J., Gotz, C., Lutsch, G., Schiller, J., Harjes, P., Herrmann, A., and Wanker, E.E. (2006). Interaction of huntingtin fragments with brain membranes--clues to early dysfunction in Huntington's disease. J Neurochem 96, 870-884. Swairjo, M.A., Concha, N.O., Kaetzel, M.A., Dedman, J.R., and Seaton, B.A. (1995). Ca(2+)-bridging mechanism and phospholipid head group recognition in the membrane-binding protein annexin V. Nat Struct Biol 2, 968-974. Tadini-Buoninsegni, F., Smeazzetto, S., Gualdani, R., and Moncelli, M.R. (2018). Drug Interactions With the Ca(2+)-ATPase From Sarco(Endo)Plasmic Reticulum (SERCA). Front Mol Biosci 5, 36. Takahashi, M., Higuchi, M., Matsuki, H., Yoshita, M., Ohsawa, T., Oie, M., and Fujii, M. (2013). Stress granules inhibit apoptosis by reducing reactive oxygen species production. Mol Cell Biol 33, 815-829. Tatavosian, R., Kent, S., Brown, K., Yao, T., Duc, H.N., Huynh, T.N., Zhen, C.Y., Ma, B., Wang, H., and Ren, X. (2019). Nuclear condensates of the Polycomb protein chromobox 2 (CBX2) assemble through phase separation. J Biol Chem 294, 1451-1463. Terzi, E., Holzemann, G., and Seelig, J. (1997). Interaction of Alzheimer beta-amyloid peptide(1-40) with lipid membranes. Biochemistry 36, 14845-14852. Tiwary, A.K., and Zheng, Y. (2019). Protein phase separation in mitosis. Curr Opin Cell Biol 60, 92-98. Torosyan, Y., Dobi, A., Naga, S., Mezhevaya, K., Glasman, M., Norris, C., Jiang, G., Mueller, G., Pollard, H., and Srivastava, M. (2006). Distinct effects of annexin A7 and p53 on arachidonate lipoxygenation in prostate cancer cells involve 5-lipoxygenase transcription. Cancer Res 66, 9609-9616. Torosyan, Y., Simakova, O., Naga, S., Mezhevaya, K., Leighton, X., Diaz, J., Huang, W., Pollard, H., and Srivastava, M. (2009). Annexin-A7 protects normal prostate cells and induces distinct patterns of RB- associated cytotoxicity in androgen-sensitive and -resistant prostate cancer cells. Int J Cancer 125, 2528- 2539. 130 Tycko, R. (2015). Amyloid polymorphism: structural basis and neurobiological relevance. Neuron 86, 632-645. Ulmer, T.S., Bax, A., Cole, N.B., and Nussbaum, R.L. (2005). Structure and dynamics of micelle-bound human alpha-synuclein. J Biol Chem 280, 9595-9603. Unwin, N. (1989). The structure of ion channels in membranes of excitable cells. Neuron 3, 665-676. Valenza, M., Rigamonti, D., Goffredo, D., Zuccato, C., Fenu, S., Jamot, L., Strand, A., Tarditi, A., Woodman, B., Racchi, M., et al. (2005). Dysfunction of the cholesterol biosynthetic pathway in Huntington's disease. J Neurosci 25, 9932-9939. van Meer, G., Voelker, D.R., and Feigenson, G.W. (2008). Membrane lipids: where they are and how they behave. Nat Rev Mol Cell Biol 9, 112-124. Varkey, J., and Langen, R. (2017). Membrane remodeling by amyloidogenic and non-amyloidogenic proteins studied by EPR. J Magn Reson 280, 127-139. Varkey, J., Mizuno, N., Hegde, B.G., Cheng, N., Steven, A.C., and Langen, R. (2013). alpha-Synuclein oligomers with broken helical conformation form lipoprotein nanoparticles. J Biol Chem 288, 17620- 17630. Velier, J., Kim, M., Schwarz, C., Kim, T.W., Sapp, E., Chase, K., Aronin, N., and DiFiglia, M. (1998). Wild-type and mutant huntingtins function in vesicle trafficking in the secretory and endocytic pathways. Exp Neurol 152, 34-40. Verzili, D., Zamparelli, C., Mattei, B., Noegel, A.A., and Chiancone, E. (2000). The sorcin-annexin VII calcium-dependent interaction requires the sorcin N-terminal domain. FEBS Lett 471, 197-200. Walsh, D.M., Klyubin, I., Fadeeva, J.V., Cullen, W.K., Anwyl, R., Wolfe, M.S., Rowan, M.J., and Selkoe, D.J. (2002). Naturally secreted oligomers of amyloid beta protein potently inhibit hippocampal long-term potentiation in vivo. Nature 416, 535-539. Wang, L., Dong, Z., Huang, B., Zhao, B., Wang, H., Zhao, J., Kung, H., Zhang, S., and Miao, J. (2010). Distinct patterns of autophagy evoked by two benzoxazine derivatives in vascular endothelial cells. Autophagy 6, 1115-1124. Wellington, C.L., Ellerby, L.M., Hackam, A.S., Margolis, R.L., Trifiro, M.A., Singaraja, R., McCutcheon, K., Salvesen, G.S., Propp, S.S., Bromm, M., et al. (1998). Caspase cleavage of gene products associated with triplet expansion disorders generates truncated fragments containing the polyglutamine tract. J Biol Chem 273, 9158-9167. Wellington, C.L., Singaraja, R., Ellerby, L., Savill, J., Roy, S., Leavitt, B., Cattaneo, E., Hackam, A., Sharp, A., Thornberry, N., et al. (2000). Inhibiting caspase cleavage of huntingtin reduces toxicity and aggregate formation in neuronal and nonneuronal cells. J Biol Chem 275, 19831-19838. Wheeler, J.R., Matheny, T., Jain, S., Abrisch, R., and Parker, R. (2016). Distinct stages in stress granule assembly and disassembly. Elife 5. Winklhofer, K.F., Tatzelt, J., and Haass, C. (2008). The two faces of protein misfolding: gain- and loss- of-function in neurodegenerative diseases. EMBO J 27, 336-349. 131 Yagi, H., Hasegawa, K., Yoshimura, Y., and Goto, Y. (2013). Acceleration of the depolymerization of amyloid beta fibrils by ultrasonication. Biochim Biophys Acta 1834, 2480-2485. Yang, D., Wang, C.E., Zhao, B., Li, W., Ouyang, Z., Liu, Z., Yang, H., Fan, P., O'Neill, A., Gu, W., et al. (2010). Expression of Huntington's disease protein results in apoptotic neurons in the brains of cloned transgenic pigs. Hum Mol Genet 19, 3983-3994. You, K., Huang, Q., Yu, C., Shen, B., Sevilla, C., Shi, M., Hermjakob, H., Chen, Y., and Li, T. (2020). PhaSepDB: a database of liquid-liquid phase separation related proteins. Nucleic Acids Res 48, D354- D359. Yuan, C., Levin, A., Chen, W., Xing, R., Zou, Q., Herling, T.W., Challa, P.K., Knowles, T.P.J., and Yan, X. (2019a). Nucleation and Growth of Amino Acid and Peptide Supramolecular Polymers through Liquid-Liquid Phase Separation. Angew Chem Int Ed Engl 58, 18116-18123. Yuan, H.F., Li, Y., Ye, W.H., Liu, Y., Zhang, Z.D., Tan, B.B., Fan, L.Q., Zhao, Q., Wang, D., Jia, N., et al. (2019b). Downregulation of annexin A7 decreases proliferation, migration, and invasion of gastric cancer cells by reducing matrix metalloproteinase 1 and 9 expression. Am J Transl Res 11, 2754-2764. Zaks, W.J., and Creutz, C.E. (1990). Evaluation of the annexins as potential mediators of membrane fusion in exocytosis. J Bioenerg Biomembr 22, 97-120. Zarrinpar, A., Bhattacharyya, R.P., and Lim, W.A. (2003). The structure and function of proline recognition domains. Sci STKE 2003, RE8. Zeno, W.F., Baul, U., Snead, W.T., DeGroot, A.C.M., Wang, L., Lafer, E.M., Thirumalai, D., and Stachowiak, J.C. (2018). Synergy between intrinsically disordered domains and structured proteins amplifies membrane curvature sensing. Nat Commun 9, 4152. Zheng, Z., Li, A., Holmes, B.B., Marasa, J.C., and Diamond, M.I. (2013). An N-terminal nuclear export signal regulates trafficking and aggregation of Huntingtin (Htt) protein exon 1. J Biol Chem 288, 6063- 6071. Zhu, L., and Brangwynne, C.P. (2015). Nuclear bodies: the emerging biophysics of nucleoplasmic phases. Curr Opin Cell Biol 34, 23-30. Zhu, M., Li, J., and Fink, A.L. (2003). The association of alpha-synuclein with membranes affects bilayer structure, stability, and fibril formation. J Biol Chem 278, 40186-40197. Zimmerberg, J., and Kozlov, M.M. (2006). How proteins produce cellular membrane curvature. Nat Rev Mol Cell Biol 7, 9-19. 132 Appendices Appendix A: Potential Role of Annexin A7 in Huntingtin Cellular Aggregation In Chapter 3 and 4, we investigated two proteins: Httex1 and ANXA7. They are both membrane proteins that can undergo liquid-liquid phase separation (Peskett et al., 2018). Similar to many other amyloid proteins (Manoharan et al., 2016), the expression of Httex1 with pathogenic Q length leads to the formation of cellular puncta and induces cellular stress (Shacham et al., 2019). Recall that ANXA7 gets recruited into stress granules (see Chapter 4), we therefore wondered whether ANXA7 might be playing a role in the cellular aggregation of Httex1. To test the hypothesis, C-terminally mRFP tagged huntingtin exon 1 protein fragment containing 16, 39 or 72 polyglutamine was cloned into mammalian expression vector (pcDNA4.0), hereafter referred to as Httex1(Q16)mRFP, Httex1(Q39)mRFP and Httex1(Q72)mRFP. When Httex1(Q39)mRFP was cotransfected with the three eGFP-tagged ANXA7 variants (see Chapter 4 for details) into HEK293T cells, we observed the formation of worm-like aggregates surrounding the Httex1 puncta for full-length ANXA7 (Appendix Figure A-1) but not for the other two variants. To test the Q-length dependence of this phenomenon, full-length ANXA7 was then cotransfected with all three Httex1 variants. As shown in Appendix Figure A-2, the protein form granular structures surrounding the Httex1 puncta. Our preliminary time-lapse data (not shown) suggested that full-length ANXA7 could accelerate the phase separation of pathogenic Httex1 when cotransfected with the latter. For more conclusive evaluations, we performed quantitative analyses in live HEK293T cells. As shown in Appendix Figure A- 3, the percent cells containing puncta were counted for three time points: 12h, 24h and 48h post transfection (three random frames per group). Two different pathogenic Q-length: Q39, Q72 were analyzed. The results support that at each time point examined, Httex1(Q72)mRFP forms cellular puncta faster than Httex1(Q39)mRFP. Additionally, cotransfection of ANXA7 could further accelerate the puncta formation of Httex1 for both Q-length. 133 References Manoharan, S., Guillemin, G.J., Abiramasundari, R.S., Essa, M.M., Akbar, M., and Akbar, M.D. (2016). The Role of Reactive Oxygen Species in the Pathogenesis of Alzheimer's Disease, Parkinson's Disease, and Huntington's Disease: A Mini Review. Oxid Med Cell Longev 2016, 8590578. Peskett, T.R., Rau, F., O'Driscoll, J., Patani, R., Lowe, A.R., and Saibil, H.R. (2018). A Liquid to Solid Phase Transition Underlying Pathological Huntingtin Exon1 Aggregation. Mol Cell 70, 588-601 e586. Shacham, T., Sharma, N., and Lederkremer, G.Z. (2019). Protein Misfolding and ER Stress in Huntington's Disease. Front Mol Biosci 6, 20. 134 Appendix Figure A-1. Full-length ANXA7 can form worm-like aggregates surrounding Httex1(Q39) puncta in live HEK293T cells Httex1(Q39)mRFP was cotransfected with the three eGFP-tagged variants of ANXA7 for 48 hours. From left to right: green channel showing the signal of ANXA7; red channel showing the signal of Httex1(Q39); blue channel showing the DRAQ5 signal for nucleus; merged image of the three channels. (A) the full-length ANXA7. (B) the N-terminus of ANXA7 (residue 1-176). (C) the N-terminal truncated ANXA7 (residue 2-176 deleted). White arrows are pointing to the formation of ANXA7 aggregates surrounding Httex1 puncta and the lack of aggregates in the two truncated variants of ANXA7 lacking either the N-terminal or the C-terminal domain. Confocal microscopy images were captured for live cells and processed using Fiji/Image J software. Scale bar = 10 μm. 135 Appendix Figure A-2. Full-length ANXA7eGFP forms granular structures surrounding Httex1 puncta in fixed HEK293T cells Httex1mRFP containing 16, 39 or 72 polyglutamine was transfected for 48 hours either alone (A) or together with full-length ANXA7 (B) in HEK293T cells. From left to right: Q16, Q39 and Q72. Granular ANXA7 structures (green signals) were observed surrounding Httex1 puncta (red signals). Cells were fixed with 4% PFA and then imaged under confocal fluorescence microscope and processed using Fiji/Image J software. Scale bar = 10 μm. 136 Appendix Figure A-3. ANXA7 accelerates Httex1 puncta formation in cells Httex1(Q39)mRFP or Httex1(Q72)mRFP was transfected alone or together with full-length ANXA7eGFP in HEK293T cells. 12, 24 and 48-hour post transfection, cells were imaged under confocal microscope. The nuclei were labeled with DRAQ5. Three random frames were taken and representative images are shown in (A). Percent cells containing cellular puncta was counted and summarized as table and plotted in (B) and (C). 137 Appendix B: The Establishment of ANXA7-/- Cell Line by CRISPR-Cas9 ANXA7 is a stress-related protein with wide physiological implications in cancer, apoptosis and mitophagy. It would be interesting to further explore the correlation between the unique characteristics of ANXA7 and its cellular function. Therefore, we went further to establish an ANXA7 knockout HEK293T cell line using the CRISPR-Cas9 system (Shalem et al., 2014). The workflow chart for the entire process is given as Appendix Figure B-1. Human Annexin VII gene contains 14 exons and spans approximately 34 kb of genomic DNA. As a human embryonic kidney cell line, only the shorter isoform of ANXA7 is expressed in HEK293T cells. And because exon 6 encodes for the isoform-specific sequence, we intentionally avoided it as the target. The selection of guide RNA sequence was performed using online webtool CHOPCHOP (Labun et al., 2016) (https://chopchop.cbu.uib.no). Four 20-mer gRNA sequences with high specificity and low off-target probability were selected, each targeting a specific exon and named by their score rank (Table B-1). Table B-1. Summary of gRNA sequence targeting ANXA7 The four selected gRNA sequences were inserted into LentiCRISPR construct in conjugation with gRNA scaffold. All four constructs were simultaneously transfected into HEK293T cells for higher knockout efficiency. 48 hours post transfection, the clones were selected using puromycin for two days. The surviving cells were allowed to recover and proliferate. Single cells were sorted by FACS into 96 wells. Those single cells were allowed to grow into colonies and expanded to a larger cell population. Initial screening of ANXA7-/- clones was performed with western blot (Appendix Figure B-2). For promising clones, genomic DNA was extracted. Two target genome sequence respectively covering the exon 3+4 and gRNA#2 CGCAACCCGTAACTCACCTC Exon 3 gRNA#6 CCGTAACTCACCTCCGGGAT Exon 3 gRNA#13 TTATGGAGGTGGTCCAGCAC Exon 4 gRNA#23 TAGCCTGTTGGGGGATAGCC Exon 2 138 exon 2 of ANXA7 were PCR-amplified and sent for sequence-verification (primer sequences are listed in Table B-2). Exon 2-Forward Primer CTGGTGCTAAGCACCTGC Exon 2-Reverse Primer CCAGGTGTGGTTCTGGAG Exon 3/4-Forward Primer GGTGGTTTTACTATTGGTACAG Exon 3/4-Reverse Primer GGTCAGGAGTCATCTTTTCC Table B-2. Summary of genome sequencing primers for ANXA7-/- verification Though due to sequence quality and other complicating factors, some of the sequencing results were rather ambiguous, we managed to obtain conclusive results for two clones: Clone #12 was verified to contain partial truncation in exon 4 and missense frameshift between exon 4 and 5. Clone #26 was verified to contain partial truncation in exon 3 and 4 as well as missense frameshift between exon 3 and 4. The two clones also lack protein band as verified by western blot (Appendix Figure B-1). Therefore, the knock was complete both on the DNA level and the protein expression level. Thus far, we have successfully established the HEK293T ANXA7-/- cell line. I would like to thank Dr. Jeannie Chen for suggestions on CRISPR-Cas9 and Yun Yao for her technical assistance in the colony screening and western blot screening. References Labun, K., Montague, T.G., Gagnon, J.A., Thyme, S.B., and Valen, E. (2016). CHOPCHOP v2: a web tool for the next generation of CRISPR genome engineering. Nucleic Acids Res 44, W272-276. Shalem, O., Sanjana, N.E., Hartenian, E., Shi, X., Scott, D.A., Mikkelson, T., Heckl, D., Ebert, B.L., Root, D.E., Doench, J.G., et al. (2014). Genome-scale CRISPR-Cas9 knockout screening in human cells. Science 343, 84-87. 139 Appendix Figure B-2. Workflow chart for CRISPR-Cas9 mediated cell genome editing gRNA/Cas9 expression vectors were transfected into the HEK293T cell line to allow gRNA-mediated Cas9 genome editing. Puromycin was used as the selection marker to kill the non-transfected cells. The survived cells were left to allow further proliferation, followed by single cell sorting using FACS into 96-well plates. The survived cell colonies were expanded step-wise. The cells were then subjected to screening using western blot. Genomic DNAs from positive clones were extracted and the target region was PCR-amplified for sequencing verification. 140 Appendix Figure B-3. Western blot result for selective ANXA7-/- candidate clones The left most lane labeled with HEK is the total cell lysate from wild type HEK293T as a positive control. The rest of the lanes are labeled by their clone number. The membrane was first probed by anti-ANX VII primary antibody (#A4475, Sigma), followed by anti-mouse HRP secondary antibody (Santa Cruz), which was then detected by ECL reagent (GE Healthcare). 141 Appendix C: Analysis of Huntingtin-targeting Peptides Section 1: Analyzing the potential of anti-Htt peptides as diagnostic probes The anti-Htt peptide sequences were screened and optimized by our collaborator: Dr. Richard Roberts’ lab. The work has not been published and the sequence information remains confidential. We will therefore refer to them simply as anti-Htt Clone 1 and anti-Htt Clone 8 peptides in this section. Both peptides were synthesized commercially and labeled with Alexa 488 C5 Maleimide (Invitrogen, #A10254). Unlabeled free dyes were removed by reverse phase chromatography (prepared by Dr. J. Mario Isas). In order to analyze the ability of these peptides to identify Httex1 of pathogenic Q length, the same mammalian expression constructs encoding mRFP-tagged huntingtin exon 1 fragment containing 39 or 72 glutamines (discussed in Appendix A) were transfected into HEK293T cells at ~ 50% confluency using lipofectamine â LTX with PLUS TM reagent (#15338030, Invitrogen TM ) according to the manufacturer’s instruction. 48 hours post transfection, the cells were fixed with 4% PFA for 10 minutes and permeabilized with 0.1% triton-x for 15 minutes. After PBS rinsing, permeabilized cells were blocked with 1% BSA for 1 hour and rinsed again for PBS (3 times). Cells were then incubated with 1 μΜ Alexa488-labeled Clone 1 or Clone 8 peptides in the presence of 0.1% BSA for overnight at 4 °C. Unbound peptides were washed off with three repetitive rinsing in PBS. The cell nucleus was stained with DAPI. The prepared cell slides were imaged under LSM780 Zeiss confocal microscope using 63x oil lens. As shown in Appendix Figure C-1 and C-2, both peptides exhibit specific binding affinity to cellular puncta in the Httex1(Q72)mRFP group. In contrast, for the Httex1(Q39)mRFP group, the clone 8 peptide exhibited much higher binding affinity than the clone 1 peptide. This difference could be due to the fact that the two peptides were designed to target different protein species. To further verify that the specific binding of the two peptides are targeting the Httex1 fragment sequence and not due to interaction with the mRFP, we repeated the experiment with Httex1(Q39) and Httex1(Q72) constructs lacking fluorescent tags, the results are shown in Appendix Figure C-3. As shown, the peptides could still identify some droplet-like cellular puncta, which is not seen in the PBS control group. 142 We also noticed that the species identified by the peptides look morphologically different depending on the presence of the mRFP tag. Without the tag, the identified species were more spherical, reminiscent of early-stage protein droplets. With the tag, the identified species, though still in ball-shape, had rather loosened and irregular edges, reminiscent of the more mature gel-like or solid-like species. This could suggest that the mRFP fused protein behaves differently in aggregation kinetics or that the presence of mRFP is introducing some artifact in peptide-protein interactions. Further experiments are required to draw definitive conclusions. Section 2: Analyzing the potential of anti-Htt dipeptides to inhibit Htt aggregation In order to evaluate whether the peptides have the potential to inhibit the cellular aggregation of Httex1, eGFP-tagged Clone 1 and Clone 8 sequence was cloned into mammalian expression vector pcDNA4.0. We will hereafter refer to it as eGFP-dipeptide construct. C-terminally mRFP-tagged Httex1(Q72) was either transfected alone or cotransfected with eGFP-dipeptide construct at 1:1 molar DNA ratio into HEK293T cells at ~50% confluency. 24 hours post transfection, cells were fixed by 4% PFA and the nucleus was stained with DAPI. The cell slides were imaged on a Leica SP8 confocal microscope under 20x air lens. Due to the extremely high fluorescence intensity of Httex1(Q72)mRFP puncta, in order to quantitatively compare between the two experiment groups, each sample was imaged with increasing laser power at 0.1, 1 and 3 while keeping all other parameters constant. The results are shown in Appendix Figure C-4 and C-5. As can be concluded most directly from the images taken under the 0.1 laser power, 24 hours post transfection, RFP positive droplet-like puncta have already formed in many cells (Appendix Figure C-4A, red channel). In contrast, when the dipeptides are cotransfected, barely any puncta were visible (Appendix Figure C-5A, red channel). At higher laser power, we can see that Httex1(Q72)mRFP was also expressed successfully, but present mostly in diffused form within the cytoplasm (RFP-positive signal in cellular shape rather than droplet-like puncta shape). Thus, we can conclude that the cotransfection of anti-Htt dipeptide can efficiently inhibit the Httex1(Q72)mRFP puncta formation 24 hour post transfection. 143 Appendix Figure C-1. Clone 1 peptide immunostaining of HEK293T cells 48 hours post Httex1(Q39)mRFP or Httex1(Q72)mRFP transfection HEK293T cells were transfected with Httex1(Q39)mRFP (Panel A) or Httex1(Q72)mRFP (Panel B). 48 hours post transfection, the cells were fixed with 4% PFA and permeabilized with 0.1% triton-x and blocked with 1% BSA for 1 hour. After blocking, cells were incubated with 1 μΜ Alexa488-labeled clone 1 peptides in the presence of 0.1% BSA for overnight at 4 °C. Unbound peptides were washed off with three repetitive rinsing in PBS. 144 Appendix Figure C-2. Clone 8 peptide immunostaining of HEK293T cells 48 hours post Httex1(Q39)mRFP or Httex1(Q72)mRFP transfection HEK293T cells were transfected with Httex1(Q39)mRFP (Panel A) or Httex1(Q72)mRFP (Panel B). 48 hours post transfection, the cells were fixed with 4% PFA and permeabilized with 0.1% triton-x and blocked with 1% BSA for 1 hour. After blocking, cells were incubated with 1 μΜ Alexa488-labeled clone 8 peptides in the presence of 0.1% BSA for overnight at 4 °C. Unbound peptides were washed off with three repetitive rinsing in PBS. 145 Appendix Figure C-3. Clone 1 or clone 8 peptide immunostaining of HEK293T cells 48 hours post transfection with PBS, Httex1(Q39) or Httex1(Q72) HEK293T cells were transfected with PBS, Httex1(Q39) or Httex1(Q72) free of fluorescent tag. 48 hours post transfection, the cells were fixed with 4% PFA and permeabilized with 0.1% triton-x and blocked with 1% BSA for 1 hour. After blocking, cells were incubated with 1 μΜ Alexa488-labeled clone 1 (Panel A) or clone 8 peptides (Panel B) in the presence of 0.1% BSA for overnight at 4 °C. Unbound peptides were washed off with three repetitive rinsing in PBS. As shown, the peptides also specifically identified some droplet-like cellular puncta not observed in the PBS control group, suggesting that the anti-Htt peptides indeed specifically identify the protein sequence, not the fluorescent tag. 146 Appendix Figure C-4. HEK293T cells 24 hour post Httex1(Q72)mRFP transfection From top row to bottom row: the same frame was imaged at increasing laser power respectively at 0.1, 1 and 3 (Panel A, B and C). From left to right: first column, DAPI channel; second column, GFP channel; third column, RFP channel showing Httex1(Q72); fourth column, merged image of the three channels. Scale bar = 50 μm. 147 Appendix Figure C-5. HEK293T cells 24 hour post Httex1(Q72)mRFP/dipeptide cotransfection From top row to bottom row: the same frame was imaged at increasing laser power respectively at 0.1, 1 and 3 (Panel A, B and C). From left to right: first column, DAPI channel; second column, GFP channel showing dipeptides; third column, RFP channel showing Httex1(Q72); fourth column, merged image of the three channels. Scale bar = 50 μm.

Abstract (if available)

Abstract Membrane proteins participate in multiple cellular processes, fulfilling diverse functional roles. The focus of my thesis is to understand: 1) The interaction between the membrane proteins (protein-protein interaction)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Structure and kinetics of the Orb2 functional amyloid

PDF

The structure and function of membrane curving proteins on different membrane shapes and their regulation by post-translational modifications

PDF

Before they were amyloid: understanding the toxicity of disease-associated monomers and oligomers prior to their aggregation

PDF

Interaction of monoclonal antibodies MW1 and PHP1 with huntingtin exon1 protein

PDF

Membrane curvature sensors and inducers studied by site-directed spin labeling

PDF

Structural studies on functional amyloids and the mechanism of aggregation and disaggregation of Huntingtin Exon 1

PDF

The folding, misfolding and refolding of membrane proteins and the design of a phosphatidylserine-specific membrane sensor

PDF

Overexpression and interaction of Orb2 in S2 insect cells

PDF

Molecular basis of CD33 receptor function, and effects of a destabilizing transmembrane motif on αXβ2 integrin

PDF

Facilitating unambiguous NMR assignment by solid–state NMR using segmental isotope labeling through split-inteins

PDF

Structural studies of the IAPP membrane-mediated aggregation pathway

PDF

Developing a method for measuring kinetic rates of RNA/protein interaction using switchSENSE® technology on the DRX2

PDF

The physics of emergent membrane phenomena

PDF

Expression and purification of extracellular domain of human CD33 in Escherichia coli and Pichia pastoris

PDF

Regulating functional amyloid formation: the promiscuous behavior of the Orb2A N-terminal amphipathic region

PDF

Controlling membrane protein folding using photoresponsive surfactant

PDF

Probing the effects of transmembrane domains on the continuum mechanics of lipid bilayers

PDF

Oligomer formation of functional amyloid protein - Orb2A

PDF

Enhancing and inhibiting diabetic amyloid misfolding

PDF

Role of CD33 structure and function in Alzheimer’s disease

Asset Metadata

Creator Tao, Meixin (author)

Core Title Protein-protein interaction and protein-lipid interaction of membrane proteins

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Medical Biophysics

Publication Date 04/21/2021

Defense Date 03/16/2020

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag annexin A7,annexin B12,huntingtin,membrane protein oligomerization,membrane-mediated amyloid aggregation,OAI-PMH Harvest,protein phase separation,protein-lipid interaction,protein-protein interaction

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Siemer, Ansgar (committee chair), Haworth, Ian (committee member), Langen, Ralf (committee member)

Creator Email meixinta@usc.edu,meixintao2019@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-286472

Unique identifier UC11675368

Identifier etd-TaoMeixin-8285.pdf (filename),usctheses-c89-286472 (legacy record id)

Legacy Identifier etd-TaoMeixin-8285.pdf

Dmrecord 286472

Document Type Dissertation

Rights Tao, Meixin

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA