By Yulong Sun a Thesis Submitted in Conform

MOLECULAR MECHANISMS OF AMYOTROPHIC LATERAL SCLEROSIS AND FRONTOTEMPORAL DEMENTIA: NEW INSIGHTS INTO THE FORMATION OF TDP-43 PROTEIN ASSEMBLIES

Yulong Sun

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Department of Medical Biophysics University of Toronto

Molecular Mechanisms of Amyotrophic Lateral Sclerosis and Frontotemporal Dementia: New Insights into the Formation of TDP-43 Protein Assemblies

Yulong Sun

Doctor of Philosophy

Department of Medical Biophysics University of Toronto

2018 Abstract

Advances in modern medicine in the past century have dramatically improved the average life expectancy in the western world. Unfortunately, the molecular mechanisms that maintain the integrity of proteins in the body appear to be unable to keep pace. This has led to a growing prevalence of late-onset diseases involving abnormal accumulation of proteins, especially in the last century. The increase in occurrence of neurodegenerative diseases such as amyotrophic lateral sclerosis (ALS), Alzheimer’s disease (AD), Parkinson’s disease (PD), and transmissible spongiform encephalopathies such as prion disease, has become a great burden to the healthcare system. All of these diseases are currently incurable and fatal, but they share the common hallmark of misfolding and aggregation of proteins within the effected neurons. The discovery and characterization of such proteins have often led to the identification of potential targets for treatment and drug design. In the case of ALS, progressive death of upper and lower motor neurons leads to full-body paralysis, and patient death from respiratory failure. The cause of ALS is currently unknown, but remarkably, regardless of the type of ALS (familial or sporadic), the RNA binding protein, TDP-43, is found in

97% of cases as neuronal inclusions, suggesting a mechanistic role in disease pathogenesis.

ii In this thesis, several techniques are used to enable detailed biophysical characterization the TDP-43 aggregation process in solution and in model membranless organelles. Equilibrium turbidity measurements of the protein under aggregating conditions and the inhibitory effects of native-state stabilizing oligonucleotides on aggregation are presented. The modulatory effects of physiological concentrations of electrolytes on TDP-43 aggregation and their implications are also discussed. A novel technique called spatially targeted optical microproteomics (STOMP) is presented as a method to interrogate the proteomic contents of small cellular features in mammalian tissue in hope of identifying common proteins in neuronal inclusions and stress granules. Although the STOMP technique still requires refinement, the biophysical studies on TDP-43 presented here begin to unravel the complex and largely unknown etiology of what is currently a devastating and incurable disease.

iii

ACKNOWLEDGMENTS

I am indebted to a number of individuals who supported me over the course of my education in my graduate studies. Foremost, I am grateful to my family – particularly my parents - Dr. Ping Sun and Jane Luo – for providing me the opportunity to receive a western education. Although their efforts of pushing me to pursue a medical degree was not fruitful, I hope my sister Ruthie does so of her own choosing.

I thank also my Ph.D. supervisor, Dr. Avi Chakrabartty, for his expert guidance while allowing me to think independently and design my own projects. I thank the members of supervisory committee, Dr. Paul Fraser, Dr. Thomas Kislinger, and previously, Dr. Brad Wouters, for their thoughtful suggestions and criticisms, which have helped to shape the work presented here.

I owe thanks to the past and present members of the Chakrabartty Lab who provided words of encouragement and laid the foundations for the work presented here. Dr. P. Eli Arslan, Dr. Kevin C. Hadley, Dr. Philbert Ip, Dr. Aaron Kerman, Dr. Rishi Rakhit, Dr. Priya R. Sharda, Dr. Vanessa Morris, and Natalie Galant have my thanks, as does our collaborators from the labs of Dr. Chris M. Yip and Dr. Andrew Emili from the University of Toronto, and Dr. Sultan Darvesh from Dalhousie University. I also thank the newest members of the Chakrabartty lab and the research students – Alison, Joe, Ryan, Meghan and Jethro - for infusing some young blood back into the group.

I wish to thank my fiancée, Dr. Linda Chau, for her 4 years of patience as I completed my degree.

TABLE OF CONTENTS

Acknowledgments ...... iv Table of Contents ...... v List of Figures and Tables ...... ix

CHAPTER I Introduction: TDP-43, a Central Protein in the Amyotrophic Lateral Sclerosis/Frontotemporal Dementia (ALS/FTD) Disease Spectrum ...... 1

Introduction ...... 3 Rising prevalence of neurodegenerative diseases ...... 3 ALS/FTD is a spectrum disorder ...... 4 Physiological Functions and Pathobiology of TDP-43 ...... 6 Physiological functions of TDP-43...... 6 Pathological functions of TDP-43 ...... 7 Insights into TDP-43 Aggregation from Structural Studies ...... 9 Domain structure of TDP-43 ...... 9 The ubiquitin-like fold of the NTD ...... 11 The N-terminus is involved in aggregation and splicing ...... 11 Tandem RRMs contain canonical folds but are uniquely arranged ...... 14 Role of RRM domains in aggregation ...... 15 The C-terminal region has a dynamic structure ...... 15 The C-terminal Domain: Prions, Droplets and Aggregation ...... 17 The C-terminal domain resembles yeast prions ...... 17 Human proteins containing PrLD form membraneless organelles ...... 18 TDP-43 undergoes phase transitions through its PrLD ...... 19 LLPS as drivers for aggregation ...... 21 Spreading and propagation of ALS/FTD ...... 22 Chapter Remarks ...... 24 Concluding remarks on current literature ...... 24 Acknowledgements ...... 25 Thesis Rationale ...... 26 References ...... 29

CHAPTER II Binding of TDP-43 to the 3’UTR of its Cognate mRNA Enhances its Solubility ...... 45

Introduction ...... 47 Results ...... 49 Recombinant vYFP-TDP-43 is natively dimeric ...... 49 TG12 inhibits TDP-43 aggregation at sub-stoichiometric concentrations by maintaining dimer configuration ...... 51 Naturally occurring nucleotide targets reduce the level of TDP-43 aggregation ...... 54 Aggregation inhibition is achieved through RRM1 binding ...... 55 Effect of oligonucleotides on pre-formed vYFP-TDP-43 aggregates ...... 56 Morphology of TDP-43 aggregates ...... 57 Discussion ...... 58 Insights into TDP-43 misfolding and inhibition mechanisms ...... 58 v

Implications of TDP-43 aggregation inhibition by naturally occurring targets ...... 60 Materials and Methods ...... 63 Protein expression and purification ...... 63 Dynamic light scattering measurements ...... 63 Size exclusion chromatography ...... 64 Urea denaturation ...... 64 Circular dichroism spectroscopy ...... 64 In vitro aggregation of vYFP-TDP-43 ...... 64 Right angle light scattering ...... 64 Fluorescence microscopy ...... 65 Atomic force microscopy ...... 65 Chapter Remarks ...... 66 Acknowledgements ...... 66 Supplemental material ...... 66 References ...... 67

CHAPTER III Physiological Electrolytes as Regulators of TDP-43 Aggregation and Droplet-Phase Behavior ...... 72

Introduction ...... 74 Results ...... 76 TDP-43 aggregation is modulated by electrolyte concentration ...... 76 TDP-43 forms non-fibrillar aggregates ...... 81 TDP-43 aggregation is reversible ...... 83 Kinetics properties of yTDP-43 aggregation ...... 84 Insertion of yTDP-43 into Ddx4N1 droplets ...... 86 TDP-43 behavior in droplets is affected by environmental conditions ...... 91 Discussion ...... 95 TDP-43 aggregation is modulated by physiological electrolytes ...... 95 Kirkwood’s theory of electrolyte-protein interactions applied to TDP-43 ...... 96 Effect of electrolytes on TDP-43 aggregate morphology ...... 96 Kinetics of electrolyte-induced aggregation correlates with morphological changes ...... 97 Insights into TDP-43 behavior in a droplet structure...... 98 Phase separating properties of TDP-43 CTD as a driver for aggregation ...... 99 Role of electrolytes in physiological conditions ...... 100 Materials and Methods ...... 103 Protein expression and purification ...... 103 Dynamic light scattering measurements ...... 104 In vitro aggregation of vYFP-TDP-43 ...... 104 Curve fitting of aggregation kinetics ...... 104 Transmission electron microscopy ...... 105 Immunofluorescence staining of formalin-fixed human brain tissue ...... 105 In vitro Thioflavin T fluorescence ...... 106 Circular Dichroism spectroscopy ...... 106 Fluorescence microscopy ...... 106 Quantitation of yTDP-43 insertion into Ddx4 droplets ...... 107 Fluorescence recovery after photobleaching (FRAP) ...... 107 Chapter Remarks ...... 109 vi

Acknowledgements ...... 109 Supplemental material ...... 109 References ...... 112

CHAPTER IV Determining Composition of Micro-Scale Protein Deposits in Neurodegenerative Disease by Spatially Targeted Optical Microproteomics (STOMP)...... 119

Introduction ...... 121 Results and discussion ...... 124 The STOMP technique ...... 124 The resolution of STOMP ...... 126 STOMP analysis of amyloid plaques in a transgenic mouse model of AD ...... 127 Identification of photo-tagged amyloid plaque proteins by mass spectrometry ...... 128 Validation of the STOMP results in TgCRND8 mice with immunofluorescence and immunohistochemistry ...... 131 STOMP analysis of senile plaques from post-mortem AD brain ...... 133 Comparison with previously published data ...... 136 Conclusions ...... 136 Materials and Methods ...... 138 Photo-tag synthesis ...... 138 Murine tissue sectioning and preparation ...... 138 Microscopy and photoactivation ...... 139 Solubilization and affinity purification ...... 139 Mass spectrometry ...... 140 Gel electrophoresis and silver staining ...... 141 Photo-tagging volume measurement ...... 141 References ...... 143

CHAPTER V Cost-Effective Elimination of Lipofuscin Fluorescence from Formalin Fixed Brain Tissue by White Phosphor Light Emitting Diode Array ...... 147

Introduction ...... 149 Results and Discussion ...... 151 Photobleaching significantly reduces autofluorescence ...... 151 Immunostaining of tau-positive inclusions ...... 153 Materials and Methods ...... 157 Photobleaching apparatus ...... 157 Sample preparation and immunofluorescence ...... 157 Fluorescence microscopy and image quantitation ...... 158 Chapter Remarks ...... 159 Acknowledgements ...... 159 Statement of ethics ...... 159 References ...... 160

CHAPTER VI Probing TDP-43 Disease Mechanisms Using STOMP Technology: Challenges and Future Directions ...... 161 vii

Introduction ...... 163 Results and Discussion ...... 166 Application of STOMP to SGs in cultured cells ...... 166 Application of STOMP to FTLD-U inclusions with TDP-43 in archived brain tissue...... 170 The feasibility of STOMP in on TDP-43 positive cellular structures ...... 174 Materials and Methods ...... 175 Induction and visualization of stress granules ...... 175 Immunofluorescence staining of FTLD-U cases ...... 175 STOMP analysis and purification ...... 176 Concluding Remarks and Future Directions ...... 177 Impact and significance ...... 177 Misfolded TDP-43 as a biomarker and therapeutic target for ALS ...... 178 References ...... 180

viii

LIST OF FIGURES AND TABLES

I. Introduction: TDP-43, a Central Protein in the Amyotrophic Lateral Sclerosis/Frontotemporal Dementia (ALS/FTD) Disease Spectrum Figure 1.1: Domain arrangement and secondary structures of TDP-43 ...... 10 Figure 1.2: Combined, chimeric molecular structure of TDP-43 bound to a single strand of AUG12 RNA ...... 13 Figure 1.3: Graphical representation of TDP-43 aggregation model ...... 25

II. Binding of TDP-43 to the 3’UTR of its Cognate mRNA Enhances its Solubility Figure 2.1: Characterization of vYFP-TDP-43 using various biochemical techniques ...... 50 Figure 2.2: Size distribution by mass of vYFP-TDP-43 upon aggregation determined by dynamic light scattering and sample turbidity measured by right angle light scattering ...... 52 Figure 2.3: Inhibition of vYFP-TDP-43 aggregation using TG12 monitored by right angle light scattering ...... 53 Figure 2.4: Inhibition of wild-type (wt) and F147L/F149L mutant vYFP-TDP- 43 aggregation using various natural oligonucleotide binding targets ...... 55 Figure 2.5: Fluorescence microscopy of vYFP-TDP-43 aggregates ...... 56 Figure 2.6: Tapping mode atomic force microscopy images of non-fibrillar vYFP-TDP-43 aggregates ...... 57 Figure S2.1: Schematic representations of vYFP-TDP-43 constructs and proposed aggregation mechanism ...... 66

III. Physiological Electrolytes as Regulators of TDP-43 Aggregation and Droplet-Phase Behavior Figure 3.1: The effect of NaCl on purified Venus YFP-tagged, full-length human TDP-43 (yTDP-43) aggregation ...... 77 Figure 3.2: Aggregation of TDP-43 is induced by various electrolytes ...... 79 Figure 3.3: Non-amyloid nature of yTDP-43 aggregation ...... 82 Figure 3.4: Reversibility of TDP-43 aggregation ...... 83 Figure 3.5: Effect of protein concentration and NaCl concentration on yTDP-43 aggregation kinetics ...... 85 Table 3.1: Compositional similarity between Ddx4N1 (1-236) and known stress granule proteins with intrinsically disordered regions ...... 87 Figure 3.6: Insertion of yTDP-43 into pre-formed Ddx4N1 droplets is mediated by the C-terminal domain of TDP-43 ...... 89 Figure 3.7: Alteration to yTDP-43 droplet morphology by droplet persistence and environmental electrolyte concentrations ...... 93 Figure S3.1: Negative stain EM of TDP-43 aggregate induced by NaCl and NH4OAc ...... 109 Figure S3.2: Complex yTDP-43 structures induced by freeze-thaw treatment ...... 110 Figure S3.3: Aggregates of yTDP-43 do not recover after photobleaching ...... 110 Figure S3.4: Aggregates of yTDP-43 within Ddx4N1 droplets do not recover after photobleaching...... 111 Figure S3.5: TDP-43 aggregation is dependent on the CTD ...... 111

IV. Determining Composition of Micro-Scale Protein Deposits in Neurodegenerative Disease by Spatially Targeted Optical Microproteomics (STOMP) Figure 4.1: Overview of STOMP technology ...... 125 Figure 4.2: The smallest photo-tagging volume is less than 0.5 μm3 ...... 126 Figure S4.1: Reproducibility of STOMP technique across technical and biological replicates ...... 129 Table 4.1: Proteins statistically significantly enriched in the amyloid plaques of TgCRND8 mouse brain identified and retrieved by STOMP ...... 130 Figure 4.3: Immunofluorescent confirmation of synaptic proteins in amyloid plaques of TgCRND8 mice ...... 132 Figure 4.4: Common synaptic or disease-associated proteins in plaques of TgCRND8 mice not detected by STOMP are also absent by immunofluorescence ...... 132 Table 4.2: Proteins statistically significantly enriched in senile plaques from a patient with AD identified and retrieved by STOMP ...... 134 Figure 4.5: Immunofluorescent confirmation of results of STOMP analysis of senile plaques in a case of AD ...... 135 Figure 4.6: Microphotographs of Synaptophysin (A), VAMP2 (B) and SNAP25 (C) immunohistochemistry on the brain of human Alzheimer's disease cases ...... 136

V. Cost-Effective Elimination of Lipofuscin Fluorescence from Formalin Fixed Brain Tissue by White Phosphor Light Emitting Diode Array Figure 5.1: Time-dependent photobleaching of formalin-fixed brain tissue using a white phosphor LED array ...... 152 Figure 5.2: Quantification of LED-induced signal intensity reduction of lipofuscin fluorescence in two fields of view ...... 153 Figure 5.3: Immunofluorescence imaging of phospho-tau stained FTLD-T formalin-fixed brain tissue ...... 154 Figure 5.4: Immunofluorescence imaging of phospho-tau and Nissl stained FTLD-T formalin fixed brain tissue using photobleaching and TrueBlack™ treatments ...... 155

VI. Probing TDP-43 Disease Mechanisms Using STOMP Technology: Challenges and Future Directions Figure 6.1: Modified STOMP work-flow for stress granule analysis ...... 166 Figure 6.2: Preliminary photobleaching of SGs in microscope slides containing fixed HeLa cells under arsenite-induced stress ...... 167 Figure 6.3: STOMP analysis of SGs in HeLa cell culture ...... 168 Figure 6.4: Verification of photo-tag performance...... 168 Figure 6.5: Application of STOMP to FTLD-U inclusions ...... 171 Figure 6.6: Silver stain and western blot of brain homogenate and Ni-NTA- eluted fractions of a FTLD-U tissue section analyzed by STOMP ...... 172 Figure 6.7: Enrichment of His6-tagged material from UV-irradiated tissue (UV) compared to untreated (dark) control ...... 173

CHAPTER I INTRODUCTION: TDP-43, A CENTRAL PROTEIN IN THE AMYOTROPHIC LATERAL SCLEROSIS/FRONTOTEMPORAL DEMENTIA (ALS/FTD) DISEASE SPECTRUM

This chapter first appeared in Biochemistry as a review article: Y. Sun and A. Chakrabartty (2017). Phase to Phase with TDP-43. Biochemistry 56 (6): 809–823. It was written by Y.S with input from A.C.

1 2

Chapter abstract

TDP-43 is a dimeric nuclear protein that plays a central role in RNA metabolism. In recent years, this protein has become a focal point of research in the amyotrophic lateral sclerosis and frontotemporal dementia (ALS/FTD) disease spectrum, as pathognomonic inclusions within affected neurons contain post-translationally modified TDP-43. A key question in TDP-43 research involves determining the mechanisms and triggers that cause TDP-43 to form pathological aggregates. This chapter gives a brief overview of the physiological and pathological roles of TDP-43 and focuses on the structural features of its protein domains and how they may contribute to normal protein function and to disease. A special emphasis is placed on the C-terminal prion-like region thought to be implicated in pathology, as it is where nearly all ALS/FTD-associated mutations reside. Recent structural studies on this domain revealed its crucial role in the formation of phase-separated liquid droplets through a partially populated α-helix. This new discovery provides further support for the theory that liquid droplets such as stress granules may be precursors to pathological aggregates, linking environmental effects such as stress to the potential etiology of the disease. The transition of TDP-43 among soluble, droplet, and aggregate phases and the implications of these transitions for pathological aggregation are summarized and discussed.

Abbreviations used in this chapter

Aβ, amyloid-beta; AD, Alzheimer's disease; ALS, amyotrophic lateral sclerosis; bvFTLD, behavioral variant FTLD; CDK6, cyclin-dependent kinase 6; CFTR, cystic fibrosis transmembrane regulator; CTD, C-terminal domain; fALS, familial ALS; FTD, frontotemporal dementia; FTLD, frontotemporal lobar degeneration; FUS/TLS, Fused in Sarcoma/Translocated in Sarcoma; HIV-1, human immunodeficiency virus type 1; hnRNP, heterogeneous nuclear ribonucleoprotein; LLPS, liquid-liquid phase separation; lncRNA, long non-coding RNA; LTR, long terminal repeat; ncRNA, non-coding RNA; NES, nuclear export signal; NLS, nuclear localization signal; NMD, nonsense mediated decay; NMR, nuclear magnetic resonance; NTD, N-terminal domain; PLAAC, prion-like amino acid composition; PNFA, primary non-fluent aphasia; pRb, retinoblastoma protein; PrLD, prion-like domain; PrP, prion protein; RNP, ribonucleoprotein; RRM, RNA recognition motif; sALS, sporadic ALS; SAXS, small angle X-ray scattering; SD, semantic dementia; SG, stress granule; TAR, transactive response; TDP-43, TAR element DNA binding protein of 43 kDa; TTR, transthyretin.

Introduction Rising prevalence of neurodegenerative diseases

Recent advances in modern medicine have led to dramatically longer lifespans in the human population. Ironically, through this increase in life expectancy, the imperfections in our evolutionary mechanisms that maintain the robustness of proteins appear to have been exposed. Molecular evolution has produced tremendously sturdy proteins that have little to no turnover in an organism’s lifetime such as proteins of the lens. However, modifications can accumulate in even these proteins, and even these proteins can aggregate in individuals of advanced age, leading to the disruption of optical properties of the lens and senile cataracts (Swamy and Abraham 1987). As suggested by Christopher Dobson, the rapid increase in human life expectancy may have outpaced molecular evolution, giving rise to numerous diseases that are caused by protein misfolding and aggregation (Dobson 2002; Dobson 2003). Many of these diseases, such as senile cardiac amyloidosis (caused by the deposition of the protein transthyretin in heart tissue) and certain forms of cancer (in which the key regulatory protein p53 is known to aggregate) have only emerged in the past century (Galant et al. 2016; Ishimaru et al. 2009; Soragni et al. 2016). Neurons seem particularly vulnerable to late-onset protein misfolding diseases, possibly because of the lack of neuronal cellular turnover and age- dependent deficits in protein quality control. Indeed, the abnormal accumulation of protein in affected neurons has emerged as a common hallmark of neurodegenerative diseases.

Identification and characterization of major protein components of these aggregates have often lead to transformative breakthroughs in uncovering the mechanisms of disease pathogenesis. Classic examples include the discovery of the prion protein (PrP) and formulation of the protein-only hypothesis of prion disease and the identification and study of the amyloid-β (Aβ) peptide in formulating the amyloid cascade model of Alzheimer’s disease (AD; Bolton, McKinley, and Prusiner 1982; Glenner and Wong 1984). This chapter will focus on the molecular mechanisms of misfolding and aggregation of transactive response (TAR) element DNA binding protein of 43 kDa (TDP-43), a key protein found in neuronal inclusions of patients with amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD).

ALS/FTD is a spectrum disorder

ALS, also known as Lou Gehrig’s disease, is a devastating neurological disorder characterized by the progressive loss of lower motor neurons in the anterior horn of the spinal cord and upper motor neurons in the motor cortex, leading to initial paralysis of extremities followed by fatal paralysis of the diaphragm (Brooks et al. 2000; Tandan and Bradley 1985). ALS has a prevalence of 3.9 cases per 100000 in the United States (Mehta et al. 2016). The disease is incurable and progresses rapidly, resulting in an average life expectancy of 3−5 years after disease onset, usually occurring in mid-adult life (Tandan and Bradley 1985; Majoor-Krakauer, Willems, and Hofman 2003). Ninety percent of ALS cases are sporadic (sALS) with unknown etiology, while ten percent of cases have family history (fALS), although the two are clinically indistinguishable. Distinct from ALS, FTLD, presented clinically as frontotemporal dementia (FTD), belongs to a broad range of disorders leading to progressively cognitive, behavioral, and/or language deficits (Mackenzie and Neumann 2016). It is the second most common form of dementia in people younger than 65 years of age after AD (Neary et al. 1998). Clinically, FTD can present as behavioral variant FTD (bvFTD) with predominantly behavioral changes, primary non-fluent aphasia (PNFA), affecting speech, or semantic dementia (SD), affecting comprehension (Marc Cruts et al. 2013). Approximately 30−50% of FTD cases show family history, with the majority of cases caused by mutations in three major genes: microtubule protein tau (MAPT), progranulin (GRN), and chromosome 9 open reading frame 72 (C9ORF72; Hutton et al. 1998; M Cruts et al. 2006; Baker et al. 2006; DeJesus-Hernandez et al. 2011; Renton et al. 2011; Lashley et al. 2015).

To date, a large number of genes have been identified to be causative for ALS and FTD (reviewed by Weishaupt, Hyman, and Dikic 2016). Some of them are related to clinically pure ALS or FTD, but a large portion of genes are found in both diseases, suggesting a common disease mechanism. Additionally, despite distinction in disease presentation, co-occurrences of ALS and FTD have been widely reported. Approximately 15% of FTD patients develop ALS, and ∼50% of ALS patients show some signs of cognitive impairment, meeting diagnostic criteria of FTD in ∼5% of cases, indicating significant clinical overlap (Marc Cruts et al. 2013; Mackenzie 2007). A major breakthrough linking the two diseases occurred in 2006, when TDP-43 was found to be ubiquitinated, hyperphosphorylated, and fragmented in neuronal inclusions of patients with sporadic and familial forms of ALS and FTD (Arai et al. 2006; Neumann et al. 2006). Because of their clinical, genetic, and pathological overlap, it is now believed that the two diseases belong to a spectrum of disorders termed TDP-43

5 proteinopathies and that the disease phenotype arises from differences in the primary sites of neurodegeneration: motor neurons in ALS and cortical neurons in FTD. Despite the numerous genes involved in the disease spectrum, the fact that aggregates of TDP-43 have now been found in 97% of cases of ALS and 45% of cases of FTD suggests that it is directly linked to the disease mechanism (Ling, Polymenidou, and Cleveland 2013). The study of TDP-43’s folding and aggregation is therefore invaluable for determining the cause of ALS/FTD. The identification of TDP-43 as a major component of ALS/FTD pathology catapulted the investigation of the protein’s structure, function (native and pathological), and biophysical characteristics. Since the discovery of its involvement in ALS/FTD 12 years ago, more than 1700 publications on TDP-43 have been produced. The biophysical features of this multidomain dimeric protein and how the behavior of these domains may contribute to disease pathogenesis are discussed herein.

Physiological Functions and Pathobiology of TDP-43 Physiological functions of TDP-43

TDP-43 was first discovered as a ubiquitously expressed cellular factor that binds to the TAR element, an element in the long terminal repeat (LTR) region of human immunodeficiency virus type 1 (HIV- 1) that is critical for the control of gene expression in the virus (Ou et al. 1995). The 43 kDa protein was named for its function as TAR DNA binding protein of 43 kDa upon its initial discovery. The role of the protein in human disease was first studied in the early 2000, where it was found to regulate splicing of the gene coding the cystic fibrosis transmembrane regulator (CFTR) protein at exon 9 by binding to a region near its 3’-splice site consisting of repeats of UG nucleotides (Buratti et al. 2001). TDP-43 binds preferentially to TG-rich and UG-rich sequences (vide infra) and appears to be involved in a number of roles in splicing and transcription (Acharya et al. 2006; Buratti et al. 2004; Mercado et al. 2005; Disset et al. 2005; Buratti et al. 2001; Buratti and Baralle 2001). To date, TDP-43 has been found to participate in a large number of nuclear and cytoplasmic functions as it is shuttled between the two cellular milieus (Ayala et al. 2008). In brief, TDP-43 is known to be involved in pre-mRNA processing and splicing, microRNA processing and regulation, control of long noncoding RNA (lncRNA) and noncoding RNA (ncRNA) expression, mRNA transport, mRNA stability through recruitment into stress granules (SGs), and mRNA translation (reviewed in Ratti and Buratti 2016). TDP-43 is also involved in various aspects of cell proliferation and apoptosis. It regulates the phosphorylation of retinoblastoma protein (pRb), a tumour suppressor dysfunctional in several major cancers, through the repression of cyclin-dependent kinase 6 (CDK6) expression (Ayala, Misteli, and Baralle 2008). Mutant forms of TDP-43 are also more likely to cause neural apoptosis in chick embryos (Sreedharan et al. 2008). Disruption to cell cycle and apoptotic proteins by TDP-43 mutations may implicate the protein in neuronal cancers. Because of its many functions, TDP-43 levels are tightly regulated through a negative feedback loop. TDP-43 binds to the 3’UTR of its own mRNA, leading to nonsense mediated decay (NMD)-independent mRNA degradation and a decrease in the level of TDP-43 production (Ayala et al. 2011; Budini and Buratti 2011; Bembich et al. 2014). Recent findings also suggest that RNA or DNA binding modulates TDP-43 solubility (Pesiridis et al. 2011; Y.-C. Huang et al. 2013; Sun et al. 2014). In cells, TDP-43 localizes to the nucleus in both diffuse and speckled distributions (I.-F. Wang et al. 2012). During the stress response to heat shock or sodium arsenite, TDP-43 coalesces into SGs in the cytoplasm and modulates SG assembly and dynamics (Udan-Johns et al. 2014; Parker et al. 2012; Dewey et al. 2011; Aulas, Stabile, and Vande Velde 2012; McDonald et al. 2011). Alterations to these SG processes have been suggested to play a key role in

TDP-43 aggregation and pathology (Wolozin 2012; Li et al. 2013; Bentmann, Haass, and Dormann 2013). The involvement of SGs also links environmental effects to protein aggregation, which may provide an explanation for the mostly sporadic nature of ALS (vide infra). Given these crucial RNA- related functions of TDP-43, it is not surprising that homozygous knockouts of TDP-43 are embryonically lethal in mouse models while heterozygous mice are not as affected, possibly because of the ability for TDP-43 to tightly control its expression levels through negative feedback (Wu et al. 2010; Sephton et al. 2010). The gene encoding TDP-43 protein (TARDBP) is also highly conserved in humans, mice, Drosophila melanogaster, and Caenorhabditis elegans, with very low rates of divergence among the four eukaryotes, suggesting that TDP-43 likely has crucial roles as a gene regulator (H.-Y. Wang et al. 2004). TDP-43’s many functions have led to the suggestion that disruptions to TDP-43 expression level and function are at least partially responsible for neurotoxicity.

Pathological functions of TDP-43

Aggregation of wild-type TDP-43 primarily in the cytoplasm of neurons is a prominent feature in ALS/FTD, and research on the mechanistic relationship between aggregation and disease is ongoing. In pathognomonic, cytoplasmic aggregates, TDP-43 is aberrantly ubiquitinated, phosphorylated, acetylated, sumoylated, and cleaved into C-terminal fragments (Arai et al. 2006; Neumann et al. 2006; Seyfried et al. 2010; Cohen et al. 2015). The nature of how these post translational modifications relate to disease pathology is still under investigation. Unlike AD and prion disease, the aggregates in ALS/FTD neurons are amorphous and non-amyloid, and TDP-43 aggregates created in vitro and in vivo often share this property (Capitini et al. 2014; Kerman et al. 2010; Sun et al. 2014). Cytoplasmic aggregation is usually accompanied by depletion of native TDP-43 from the nucleus as well as sequestration of other RNA-binding proteins into these aggregates (Dammer et al. 2012; Collins et al. 2012; Giordana et al. 2010; Dormann and Haass 2011). Whether depletion of TDP-43 and other RNA binding proteins can cause RNA disruption and the extent of this disruption on neurotoxicity remain unclear. Current evidence suggests that the cytoplasmic aggregates themselves are toxic to cells and cause cell death through a toxic gain of function, although alternative theories of TDP-43 aggregates as cytoprotective structures do exist in Drosophila models (Johnson et al. 2009; Capitini et al. 2014; Igaz et al. 2009; T. Zhang et al. 2011; Nonaka et al. 2013; Langellotti et al. 2016). The current consensus is that the disease likely arises from a combination of loss of TDP-43 native function and gain of toxic function from aggregation (Lee, Lee, and Trojanowski 2012). A number of factors can contribute to the aggregation of the protein, including cytoplasmic accumulation, changes in TDP-43 expression levels, aberrant cleavage and fragmentation, loss of native state binding partners, or the production of

8 truncated isoforms through alternative splicing (Budini and Buratti 2011; Nishimoto et al. 2010; Xiao et al. 2015; Sun et al. 2014; Barmada et al. 2010). Environmental factors such as stress have long been suspected to be a contributing factor in ALS/FTD pathogenesis, and the body of evidence of this suspicion has recently been growing. One mechanism mammalian cells use to cope with environmental stress is to transiently repress the translation of mRNAs for proteins not essential to survival by organizing these arrested mRNAs and their RNA binding proteins into small (≤ 5 μm) non-membrane-bound cytoplasmic domains called stress granules (SGs). The assembly of SGs can be induced by oxidative, genotoxic, osmotic and thermal stress (Anderson and Kedersha 2009; Henao- Mejia and He 2009). SG assembly and disassembly are dynamic processes mediated by a number of proteins, including TDP-43 (Gilks et al. 2004; Aulas, Stabile, and Vande Velde 2012; Kedersha et al. 2002). Knockdown of TDP-43 reduces the levels of expression G3BP and increases TIA-1 levels, two proteins known to affect SG assembly, causing SGs to form more slowly, take more time to reach the average size of normal SGs, and take less time to disassemble (McDonald et al. 2011). The disruption to SG regulation and persistence is predicted to cause cytoplasmic inclusions similar to those observed in ALS/FTD neurons (Dewey et al. 2011). How TDP-43 mutations impact the in vivo response to stress in motor neurons is disease relevant and remains to be explored. It has been suggested that a predisposing event that enriches a population of cytoplasmic, aggregation-prone TDP-43 (through mutation or other events) followed by chronic environmental stress and persistence of SGs can cause normally reversible, functional TDP-43 aggregation in SGs to form pathological aggregates as seen in disease. However, it is still unclear whether SGs are direct precursors to TDP-43 aggregates or whether they are formed independently and recruited to SGs afterward (Bentmann, Haass, and Dormann 2013). The assembly of these membraneless organelles, also known as ribonucleoprotein (RNP) granules, occurs through the physical process of liquid-liquid phase separation (LLPS; March, King, and Shorter 2016). Current research into the formation of these RNP granules has made the SG precursor hypothesis increasingly popular, as it serves as a long-missing link between environmental effects and ALS/FTD pathology. Recent structural studies of TDP-43 have shed light on the potentially detailed molecular mechanisms of how TDP-43 diverges from folding into reversible RNP granule assemblies versus pathological aggregates.

Insights into TDP-43 Aggregation from Structural Studies Domain structure of TDP-43

TDP-43 is 414 amino acid residues in length and is comprised of an N-terminal domain (NTD; residues 1-102) that includes a predicted nuclear localization signal (NLS, residues 82-98), two RNA recognition motifs (RRMs) composed of residues 106-177 (RRM1), and residues 192-259 (RRM2), including a nuclear export signal (NES) from residues 239-250 and a C-terminal domain (CTD, residues 274-414; Figure 1.1A; Ayala et al. 2008; la Cour et al. 2004). In vitro biophysical characterization and crosslinking studies in cell culture and mouse brains all suggest that TDP-43 is intrinsically dimeric and that dimer formation may be mediated by a number of regions across the entirety of TDP-43, including the NTD, RRM2, and/or the CTD (Y. J. Zhang et al. 2013; P.-H. Kuo et al. 2009; Sun et al. 2014; I.-F. Wang et al. 2012; Chang et al. 2012). The CTD is particularly relevant to disease, as it is where nearly all ALS/FTD-associated mutations are found (Kovacs et al. 2009; Sreedharan et al. 2008; Kabashi et al. 2008; Winton et al. 2008; Guerreiro et al. 2008; Kirby et al. 2010; Corrado et al. 2009; Borroni et al. 2009; Van Deerlin et al. 2008; Pamphlett et al. 2009; Luquin et al. 2009; Del Bo et al. 2009; Williams et al. 2009; Benajiba et al. 2009; Gitcho et al. 2008; Cairns et al. 2010; Bäumer, Parkinson, and Talbot 2009; Rutherford et al. 2008; Tamaoka et al. 2010; Ju et al. 2016; Yokoseki et al. 2008; Kühnlein et al. 2008; Daoud et al. 2009; Kamada et al. 2009; Chiò et al. 2011; Orrù et al. 2012; H.-H. Chiang et al. 2012; Xiong et al. 2010; Nozaki et al. 2010; Huey et al. 2012; Iida et al. 2012; Zou et al. 2012; Lemmens et al. 2009; Fujita et al. 2011; Ticozzi et al. 2011; Tsai et al. 2011; Origone et al. 2010; R. Huang et al. 2012; Conforti et al. 2011; Millecamps et al. 2010; Van Blitterswijk et al. 2012; Moreno et al. 2015; Borroni et al. 2010). This is a flexible region containing only a transient α-helical structure, and contains Q/N-rich residues implicated in aggregation (Figure 1.1D). The structural study of full-length TDP-43 has been difficult because of its high aggregation propensity and difficulty of purification, as well as its flexible CTD (Johnson et al. 2009). No crystal or nuclear magnetic resonance (NMR) structures have been produced for the protein in its entirety, but structures of the individual domains of TDP-43 have been determined.

Figure 1.1: Domain arrangement and secondary structures of TDP- 43. (A) Schematic representation of the TDP-43 domains. Domain boundaries are numbered according to full length TDP-43. TDP-43 consists of a N-terminal domain (brown) that contains the nuclear localization signal (grey), tandem RNA recognition motifs (dark yellow) containing the predicted nuclear export signal (grey), and the C-terminal domain (blue). Nearly all known ALS/FTD associated mutation occur in the CTD (green box). (B) Secondary structure of the NTD. Residues 1-77 contain a ubiquitin-like fold consisting of six β- strands (orange) and one α-helix (green). The nuclear localization sequence is underlined. (C) Secondary structure of the tandem RRM domains of TDP-43. Both RRMs share similar secondary structure consisting of five β-strands (dark yellow) and helices α1 (light blue) and α2 (dark blue). (D) Secondary structure of C-terminal domain. Only sequences containing observed or predicted secondary structures are shown. These include the helix-turn-helix motif (dark blue) and predicted β-strands (underlined).

The ubiquitin-like fold of the NTD

The N-terminus of TDP-43 contains a ubiquitin-like fold from residues 1-77 and a nonstructured region from residues 78 to 102, but depending on the experimental conditions, NMR spectroscopy of residues 1-77 from two groups has reported conflicting results for the stability of this domain. While Qin et al. reports that it is at an equilibrium with an unstructured state, Mompeán et al. report a single stable structure at high resolution (Qin et al. 2014; Mompeán et al. 2016). Under the latter condition, residues 1-77 appear to adopt a well-folded structure that consists of six β-strands and an α-helix arranged in a β1β2α1β3β4β5β6 topology (Figure 1.1B and Figure 1.2A). The low-resolution structure of Qin et al. reports a similar conformation except β4 and β5 were not observed and appeared as a single β strand. Strands β1β2β3 and β6 form one β sheet that is similar to ubiquitin, while a second smaller sheet composed of β4 and β5 appears to be a novel feature unique to the TDP-43 NTD, resembling the structure of the C-terminal Dix domain of scaffolding protein axin 1 (Mompeán et al. 2016). The remaining residues (78-102) in the N-terminus are rich in positively charged amino acids. This region appears to bind non-specifically to DNA only at pH 4.0, likely because of charge-charge interactions (Qin et al. 2014; Mompeán et al. 2016; I.-F. Wang et al. 2012). The position of the NTD relative to the rest of TDP-43 is unclear, but small angle X-ray scattering (SAXS) data suggest that it may dock close to the tandem RRMs (Figure 1.2; Y. T. Wang et al. 2013).

The N-terminus is involved in aggregation and splicing

The NTD of TDP-43 appears to be essential for TDP-43 normal function but is also required for pathological aggregation. Cytosolic localization of ectopically expressed TDP-43 without a nuclear localization signal caused the formation of TDP-43 inclusions in HEK293T cells, but expression of the same construct without the first 10 N-terminal residues or constructs containing mutations of residues 6-9 (RVTE) to glycine residues showed diffuse cytoplasmic localization, suggesting that the N-terminus is required for aggregation (Y. J. Zhang et al. 2013). Additionally, the same mutations resulted in loss of TDP-43 splicing activity when the mutants were expressed in conjunction with knockdown of endogenous TDP-43 in cell culture (Y. J. Zhang et al. 2013). This suggests that loss or mutation of these first 10 residues may adversely affect the ability of TDP-43 to form its native dimeric structure and its ability to recruit proteins needed in the splicing machinery. This is expected because residues 6-8 also form the first β-strand in the N-terminal fold and residues 6-9 are involved in stabilizing the first β-sheet of the N-terminal domain (Figure 1.1B). Mutations of these residues would disrupt the structure and dimerization of NTD, which agrees with predictions made by computer

simulations (Y. J. Zhang et al. 2013). In a cellular aggregation model of TDP-43 in which exogenous TDP-43 protein containing 12 additional tandem repeats of its aggregation-prone QN-rich sequence at its C-terminus was expressed, aggregates resembling pathological inclusions formed and sequestered endogenous TDP-43, causing loss of TDP-43 exon skipping function (Budini et al. 2012). However, when the same construct without the N-terminal 75 residues was expressed, aggregates formed but without sequestration of native TDP-43 or loss of splicing function (Budini et al. 2015). Furthermore, the N-terminal TDP-43 fragment containing residues 1-105 also appears to oligomerize into larger species in a concentration dependent manner, and the constructs containing the NTD and RRM domains show improved DNA binding activity compared to that of the tandem RRMs alone (Chang et al. 2012).

Taken together, these pieces of evidence suggest that the NTD of TDP-43, and specifically the β- sheet structural motif, contributes to both native TDP-43 function and aggregation. This region of the protein is required for the initial dimerization of TDP-43 and recruitment of other RNA binding proteins, an event required for RNA splicing and perhaps the formation of RNP granules such as SGs. The N-terminal region effectively increases the local concentration of TDP-43 and other RNA binding proteins, which enhances RNA binding and splicing functions. On the other hand, this very mechanism of congregating proteins in the proximity may also serve as a prerequisite for aggregation of the protein or recruitment of native proteins into established aggregates (Budini et al. 2015). The only ALS/FTD-associated mutation in this region is A90V located at the predicted nuclear localization signal at residues 83-98 (Winton et al. 2008; H.-H. Chiang et al. 2012). The location in the nuclear localization signal suggests a possible mechanism of pathology through disruption to nuclear localization leading to cytosolic accumulation.

Figure 1.2: Combined, chimeric molecular structure of TDP-43 bound to a single strand of AUG12 RNA. Separate PDB files were joined together for a conceptual representation of the TDP- 43 molecule. Dotted lines represent gaps in the sequence containing unknown structure. The speculative position of the N-terminal domain relative to the tandem RRM domains is based on small angle X-ray scattering data (Y. T. Wang et al. 2013). NLS and NES sequences are shown in magenta boxes. The color scheme of the secondary structures is consistent with that of Figure 1.1. (A) The N- terminal ubiquitin-like fold consists of residues 1-77, derived from the NMR structure of PDB entry 2N4P (Mompeán et al. 2016). β-strands 1-3, and 6 form β-sheet 1 (dark orange), while strands β4 and β5 form β-sheet 2 (light orange). The α1 helix is coloured in green. (B) Structure of tandem TDP-43 RRM domains bound to AUG12 (pink line drawing) generated from the NMR structure of PDB entry 4BS2 (Lukavsky et al. 2013). Colour scheme of the secondary structures in each domain is consistent, with a β-sheet consisting of strands 1-5 (yellow) sandwiched between helices α1 (cyan) and α2 (light blue). The prime (’) notation denotes the matching secondary structures in RRM2. Key loop regions loop 1 (pink), loop 3 (orange) and the loop joining the two RRMs (green) are indicated. These loop regions combined with the extensive β-sheet surface of both RRMs creates the binding surface for RNA target AUG12 (5′-GUGUGAAUGAAU-3′). (C) Secondary structure of TDP-43 amyloidogenic region derived from NMR structure of PDB entry 2N3X (L.-L. Jiang et al. 2016). This region contains extensive loops of unstructured regions except for residues 320-343 which consists of a helix-turn- helix motif (dark blue).

Tandem RRMs contain canonical folds but are uniquely arranged

TDP-43 contains two RRMs in tandem, separated by a 15-amino acid linker. Initial co-crystal structures of individual RRM1 and RRM2 domains of TDP-43 bound to single-stranded DNA demonstrated that the structures of these RRMs and the molecular interactions involved in oligonucleotide binding are congruent with typical RRM domains (P. H. Kuo et al. 2014; P.-H. Kuo et al. 2009; C.-H. Chiang et al. 2016). Structurally, they consist of a β-sheet sandwiched between two α-helices arranged in a β1α1β2β3α2β4β5 topology, where the β4 strand can also be referred to as β- hairpin (Lukavsky et al. 2013). Two segments of six and eight amino acids rich in aromatic residues on the strand β1 and β3 form the typical interacting surface on the β-sheet for nucleotide binding through direct stacking interactions with the ribonucleic bases, while a few amino acids on the loop regions between β1 and α1 (loop1) as well as between β2 and β3 (loop3) provide hydrogen bonding interactions (Figure 1.1B and 1.2B; P.-H. Kuo et al. 2009; Maris, Dominguez, and Allain 2005). Both RRMs share this canonical structure, except that RRM1 possesses a loop3 region longer than that of RRM2, which is thought to contribute to RRM1’s higher affinity for targets due to the more numerous amino acid-DNA interactions generated from this longer loop region (Figure 1.2B; P. H. Kuo et al. 2014).

The two RRMs are individually capable of binding to relatively short poly-UG RNA sequences. RRM1 binds six UG repeats (Kd = 65.2 nM) while RRM2 binds to three UG repeats (Kd = 379 nM); however, both RRM domains are required for high-affinity, synergistic binding to sequences with more than six

UG repeats (Kd = 14.2 nM; Buratti et al. 2004; P.-H. Kuo et al. 2009). Indeed, the RNA binding targets of TDP-43 are quite numerous and not all possess short UG repeats. The 3’UTR sequence through which TDP-43 modulates autoregulation contains 34 nucleotides and some targets of TDP-43 can extend up to 100 nucleotides in length (Tollervey et al. 2011; Xiao et al. 2011). Recent NMR studies have produced a structure of tandem RRM domains interacting with a single RNA strand with the sequence 5′-GUGUGAAUGAAU-3′, termed AUG12, revealing the role of both RRMs in binding to this target. The tandem domains reside side by side upon RNA binding and use both of their hydrophobic β-sheet regions and their loops to generate an extended groove to accommodate the RNA molecule (Figure 1.2B). However, unlike usual tandem RRM domains in which RNA binds to the grove in a 3’-to-5’ direction from RRM1 to RRM2, the TDP-43 tandem RRMs are arranged in reverse. Subsequently, the linker between the two RRMs that canonically spans only two β-strands now spans across a larger area, across four β-strands, which allows this linker region to participate in more extensive interactions with RNA targets. This also allows for interactions between the RRMs

themselves and potentially with other regions of the protein such as the N-terminus (Figure 1.2B; Lukavsky et al. 2013). The structural study also reveals that a degenerate consensus sequence of 5′- GNGUGNNUGN-3′ is recognized by the tandem RRMs, unlike other RRMs that require a continuous stretch of six to nine nucleotides for high-affinity binding (Maris, Dominguez, and Allain 2005). It is possible that this extended inter-RRM linker region provides the basis for TDP-43’s ability to participate in a large number of RNA processing functions because of its ability to recognize a large number of specific sequences and potentially interact with other RNA binding molecules (Lukavsky et al. 2013).

Role of RRM domains in aggregation

The role of RRM domains are most often associated with the native RNA processing functions of the protein. In vitro studies have shown, however, that binding of oligonucleotide targets such as 12 tandem repeats of poly-TG DNA (TG12) to TDP-43 through its RRMs prevents TDP-43 aggregation, suggesting some direct or indirect involvement of RRMs in aggregation (Sun et al. 2014; Y.-C. Huang et al. 2013). Only two ALS-associated mutations at the RRM domains have been identified. The recently identified P112H mutation resides on the loop1 region of RRM1 between β1 and α1, which may affect RNA binding interactions (Figure 1.1C). However, because of the novelty of the study, the effect of this mutation on TDP-43 structure or function has not been well-characterized (Moreno et al. 2015). The other mutant caused by the D169G mutation, located at a short loop region between β4 and β5 of RRM1, shares the same overall structure as the wild-type protein and actually has slightly higher binding affinity for oligonucleotide targets, suggesting that it is unlikely to disrupt normal binding functions. Furthermore, this mutant increased the thermal stability of RRM1 due to increased number of hydrophobic interactions from the D to G mutation. Interestingly, this mutant is more susceptible to caspase 3 cleavage between D208 and V209, which effectively separates α1 from β2 in RRM2 and potentially exposes one side of the β-sheet within the RRM to aberrant protein-protein or protein-DNA interactions, while causing cytoplasmic mislocalization though loss of the nuclear localization signal (Figure 1.2B). The effect of this cleavage reaction may contribute to aggregation or disruption to native protein function (C.-H. Chiang et al. 2016).

The C-terminal region has a dynamic structure

Perhaps the most rigorously studied region of the protein is the CTD (residues 274-414). Because nearly all ALS-associated mutations are found in this region, it has been implicated as an important contributor to pathogenesis. 35 and 25 kDa C-terminal fragments of TDP-43 can be generated by

caspase 3 cleavage or alternative splicing and are found in pathological inclusions. The exact mechanism of how these fragments are generated in disease is still a subject of debate (Y.-J. Zhang et al. 2009; Y.-J. Zhang et al. 2007; Nishimoto et al. 2010; Xiao et al. 2015). This low-complexity region is glycine-rich and resembles sequences of yeast prions (Gitler and Shorter 2011). The CTD appears to adopt transient and dynamic secondary structures ranging from α-helix to β-strand conformations. The structural study of the CTD is difficult because of its flexible nature, but current studies have focused primarily on a segment of the CTD approximately between residues 318 to 369. This region is considered to be the amyloidogenic core of the protein, as it contains the QN-rich segment at residues 331-369 that is capable of forming amyloid-like β-sheet structures implicated in aggregation, although TDP-43 aggregates formed in vitro do not stain positively with amyloid-specific dyes such as Congo Red and Thioflavin S (Johnson et al. 2009; L. L. Jiang et al. 2013; Liu et al. 2013; Mompeán et al. 2015; Capitini et al. 2014). Expression of 12 tandem Q/N-repeats (12×QN) within this region is sufficient to induce the formation of phosphorylated and ubiquitinated inclusions in a cell culture model (Budini et al. 2012). Residues 321-366 have also been implicated in protein-protein interactions and specifically binding to heterogeneous nuclear ribonucleoprotein (hnRNP) A2/B1 and several other members of the hnRNP family (D’Ambrogio et al. 2009; Buratti et al. 2005). Structurally, the amyloidogenic core can be largely divided into two segments, one approximately between residues 320-343, which is capable of forming a transient helix-turn-helix structure and the remaining residues 341-366, which are predicted to form two antiparallel β-sheet structures in molecular dynamics simulations (L. L. Jiang et al. 2013; Conicella et al. 2016; Mompeán et al. 2014; Mompeán et al. 2015). Interestingly, the α-helical region can also undergo α-to-β secondary structure transitions as measured by CD spectroscopy (L. L. Jiang et al. 2013). Recent studies have implied that the key α-helix formed cooperatively from residues 321-330 is required for in the formation of liquid droplets containing TDP-43 through LLPS, which may be a key mechanism of how TDP-43 performs its native and pathological functions (Conicella et al. 2016; Schmidt and Rohatgi 2016).

The C-terminal Domain: Prions, Droplets and Aggregation The C-terminal domain resembles yeast prions

The most studied region of TDP-43 is its CTD because of its direct involvement in aggregation and pathology. This region can be considered a prion-like domain (PrLD) because of the similarity of its sequence to yeast prions. Recent findings suggest that this PrLD is not entirely disordered but can fold dynamically into α-helices or β-strands, which may govern both native protein function in reaction to environmental stress and pathological aggregation. The PrLD is not a unique property of TDP-43 and may trace its evolutionary history to early eukaryotes. Although identified as pathological, infectious agents in prion disease in humans, prions in yeast play a major role in yeast metabolism and may confer selective advantages (Cox 1965; Bolton, McKinley, and Prusiner 1982). Classically, yeast protein Sup35 can fold into its prion state Ψ+ via its N-terminal domains into the typical amyloid structures composed of cross β-sheets (Cox 1965; Tuite, Staniforth, and Cox 2015). Sup35 is a translation termination factor in yeast, but upon folding into its prion conformation, it loses this function and leads to read-through of nonsense mutations. In the laboratory, it was demonstrated that in yeast strains harboring a premature stop codon in their ADE1 gene, cells without the capacity to form Sup35 prion state (Ψ- strain) become auxotrophic for adenine, whereas Ψ+ cells can grow on media lacking adenine. The Ψ+ state can be propagated through template-directed misfolding during mitosis, and amyloid aggregates in the diploid cells are segregated in the four spores during meiosis, leading to non-Mendelian propagation of the Ψ+ state to yeast progeny and retention of this selective advantage in future generations. Another yeast prion Mot3 is a transcription factor that regulates mating, carbon metabolism and stress response under its native state, but the prion state [MOT3+] allows for facultative multicellular growth phenotypes (Holmes et al. 2013). These examples show that under certain environmental conditions, the ability to form prions can confer a survival advantage. However, these phenotypes are not without disadvantages, because prion formation triggers the increased activity of HSP40/70 in yeast and Ψ+ strains have reduced growth rates compared to their Ψ- counterparts, implying that the presence of prions induces a certain degree of cellular stress (Reed B. Wickner et al. 2011). Yeast prions appear to be bet-hedging mechanisms that allow adaptive response to environmental stress; albeit with the risk of gaining pathology (March, King, and Shorter 2016; Holmes et al. 2013; Halfmann and Lindquist 2010; R. B. Wickner et al. 2016).

Human proteins containing PrLD form membraneless organelles

In addition to TDP-43, PrLD are also found in other human proteins; 70% of human proteins predicted to contain PrLDs by the PLAAC (Prion-Like Amino Acid Composition) search algorithm have molecular functions related to RNA/DNA binding, transcription factor activity or mRNA processing (March, King, and Shorter 2016). Several of these proteins are implicated in human neurodegenerative disease, such as ataxin 1 and ataxin 2 in spinocerebellar ataxias, and more significantly, hnRNPA1, hnRNPA2/B1, TDP-43 and the RNA binding protein FUS (FUS/TLS; Fused in Sarcoma/Translocated in Sarcoma), which are proteins whose mutations are known to cause ALS/FTD (Tsai et al. 2011; Millecamps et al. 2010; Kwiatkowski et al. 2009; Vance et al. 2009; Kim et al. 2013). These proteins all share the common feature of having a disordered PrLD and RRMs to mediate RNA binding. In the case of TDP-43, regions within the PrLD are often considered the amyloidogenic core of the protein, responsible for TDP-43 aggregation (Johnson et al. 2009; L. L. Jiang et al. 2013; Liu et al. 2013). Despite the association of these PrLD with disease, there would be no selective pressure to retain these domains if they only confer detrimental effects of protein misfolding in the form of neurodegenerative diseases, yet regions of the PrLD such as the α-helical segment of TDP-43 PrLD is well-conserved in vertebrates (Conicella et al. 2016; Chong and Forman- Kay 2016). Thus, it is likely that in humans, the PrLD of proteins have functions that may confer selective advantages with risk of pathology, similar to yeast prions.

One possible functional advantage of PrLDs is their role in the formation of membraneless organelles or RNP granules. The lack of a lipid-rich barrier to enclose its constituents is advantageous because it allows environmental changes to rapidly alter the internal equilibrium of the organelle (Mitrea and Kriwacki 2016). The physical properties of RNP granules were initially studied in germline P granules in C. elegans embryos, where P granules showed classic liquid droplet properties such as a spherical morphology, fusion, dissolution, and concentration-dependent condensation, strongly implicating LLPS as their mechanism of formation (Brangwynne et al. 2009). Many other RNP granules have since been reported such as processing bodies (P-bodies) that are sites of mRNA decay and SGs that are assemblies of translationally stalled ribosomal subunits and its associated mRNAs that form during cellular stress (Buchan 2014; Buchan, Nissan, and Parker 2010). Proteins such as hnRNPA1 and TDP- 43 are known to be recruited to SGs, and TDP-43 also modulates SG formation and dynamics (Molliex et al. 2015; Liu-Yesucevitz et al. 2010). Membraneless organelles such as SGs allow for transient and reversible aggregation of unneeded transcripts and allows for cell survival under stress conditions (Buchan 2014). Recently, RNA binding proteins containing PrLD, such as FUS, hnRNPA1

and TDP-43, have been reported to undergo LLPS through their PrLD, which is thought to be the underlying mechanism of the formation of RNP granules or bodies (Chong and Forman-Kay 2016; Elbaum-Garfinkle and Brangwynne 2015; Molliex et al. 2015; Burke et al. 2015; Patel et al. 2015; Schmidt and Rohatgi 2016; Conicella et al. 2016; Murakami et al. 2015). The structural changes that occur within these proteins during LLPS, however, are not well understood and are still being actively studied. While a NMR study of FUS droplets suggests that the PrLD remains entirely disordered, a study using mass spectrometry-based footprinting of the PrLD in hnRNPA2 droplets asserts that it adopts a structure rich in cross-β sheets (Burke et al. 2015; Xiang et al. 2015). Whether liquid droplets formed in vitro are reflective of the phase architecture of membraneless organelles remain uncertain.

TDP-43 undergoes phase transitions through its PrLD

The component of TDP-43 responsible for phase separation into liquid droplets emerged very recently. Unlike FUS, which appears to have no apparent structure at its PrLD in the droplet state, NMR studies on a TDP-43 PrLD construct consisting of residues 267-414 revealed that it can assemble into liquid droplets through a cooperatively folded, partially populated α-helix, initiated by the addition of 150 mM NaCl or addition of yeast RNA extract (Conicella et al. 2016). Although NMR spectroscopy of the entire CTD indicates that the region is almost entirely disordered, this α-helix formed by residues 321-330 appears to be populated 50% of the time and interacts with helices from other TDP-43 CTD molecules during liquid droplet formation (Conicella et al. 2016). This apparent RRM-independent RNA interaction may be mediated through a RRG motif on the C-terminus (Phan et al. 2011). ALS/FTD-associated mutations A321G, Q331K and M337V disrupted phase separation and encouraged conversion to aggregates (Conicella et al. 2016). Although residues Q331 and M337 do not reside within the transiently formed α-helix, structural studies indicate that these residues belong to the helix-loop-helix region formed by resides 319-341, a region perfectly conserved among vertebrates and rich in aliphatic residues (L.-L. Jiang et al. 2016). Additionally, in a cell culture system in which the full length TDP-43 construct was modified by replacing its RRMs with a GFP reporter, the expressed construct formed nuclear droplets with “bubbles” containing nuclear milieu (Schmidt and Rohatgi 2016). In agreement with the NMR studies, ALS-causing mutation M337V altered the dynamics of these droplets, whereas mutations N345K and A382T outside the α-helical region did not have an effect that was as significant (Schmidt and Rohatgi 2016). This suggests that the α-helical segment of the CTD and its intermolecular interactions are critical for the formation of liquid droplets. It is unclear how the N-terminus biophysically affects TDP-43 droplet formation, as LLPS through the C-terminal critical residues has occurred with or without an N-terminal component (Conicella et

20 al. 2016; Schmidt and Rohatgi 2016). It is possible that the environment within a test tube allows for much higher than physiological concentrations of TDP-43, thus the N-terminus that would normally be required for oligomerization and consolidation of TDP-43 would no longer be necessary.

In addition to liquid droplets, FUS, TDP-43 and hnRNPA1/A2 can also form amyloid-like folds rich in β-structure (Kato et al. 2012; Kim et al. 2013; Mompeán et al. 2014; Mompeán et al. 2015). TDP- 43 is predicted to form β-rich structures through its QN-rich region consisting of residues 341-366, while hnRNPA1 is intrinsically prone to forming irreversible amyloid fibrils (Kim et al. 2013; Mompeán et al. 2014; Mompeán et al. 2015). Segments of FUS/TLS, hnRNPA2, and TDP-43 PrLD are also capable of phase separating into a gel phase known as “hydrogels” (Saini and Chauhan 2011; Burke et al. 2015; Murakami et al. 2015; Xiang et al. 2015; Kato et al. 2012). Unlike the liquid droplet phase, the structural basis of hydrogels appears to be distinctly amyloid-like (Kato et al. 2012; Xiang et al. 2015). Electron microscopy and X-ray diffraction of FUS and hnRNPA2 hydrogels reveal that they are composed of cross-β, amyloid-like folds, but unlike typical, irreversible amyloid, these hydrogels are readily solubilized by SDS or mild heating (Kato et al. 2012). The relationship between hydrogels and liquid droplets is not well-defined, but recent studies suggest that unlike the present model of SGs in which they behave as purely liquid compartments, SGs may contain “cores” of denser material held together by strong interactions within the PrLD (Jain et al. 2016). These core particles can be purified by conventional centrifugation. They are surrounded by a liquid shell in SGs that allows for the free exchange of materials with other liquid compartments, while exchange of material between the core and the shell is an ATP-dependent process that involves ribonucleoprotein remodeling mechanisms (Jain et al. 2016). It remains uncertain whether these cores are formed first, followed by recruitment of the outer shell, or whether the SG is formed, followed by condensation of the liquid phase into more stable core structure. We speculate that these core structures may be held together by β-rich, amyloid-like interactions, similar to those found in hydrogels, that PrLDs are capable of forming. It is however uncertain whether TDP-43 forms β-rich core structures in SGs as they were not found in the centrifugation-based purification procedure, but it is becoming likely that like other proteins with PrLDs, TDP-43 may be capable of forming both liquid droplets mediated by α-helical interactions and amyloid-like hydrogel structures through β-rich folds. While the liquid states appear to be a part of physiological SG function, the conversion of these structures into aggregates may be associated with pathology.

LLPS as drivers for aggregation

The formation and maintenance of liquid droplets appear to be a delicate process and a number of factors can cause their transition into aggregates. In fact, liquid droplets tend to become less dynamic and less reversible over time and appear to have an intrinsic propensity to aggregate (Li et al. 2013; Bentmann, Haass, and Dormann 2013). For instance, constructs containing the PrLD of several RNA binding proteins including FUS and hnRNPA1, all produce liquid droplets in vitro under low-salt conditions. The PrLD of different proteins can be recruited into the same droplet, forming a heterogeneous mixture, similar to the environment within SGs. However, over time spans of 24 h, these droplets lose their liquid-like properties and change from dynamic spheres into more stable, irregular or filamentous structures (Lin et al. 2015). ThT-positive fibrils eventually form on the surface of full length hnRNPA1 liquid droplets after 24 h and time-dependent “maturation” of SGs that lead to stable and β-rich SG core structures or amyloid-like fibrillization is also observed (Molliex et al. 2015; Jain et al. 2016). In the case of TDP-43, droplets formed by the PrLD construct remain stable only on the time scale of hours before conversion to aggregates, suggesting that these droplets are transient structures that inevitably enter an irreversible state over time unless otherwise maintained (Conicella et al. 2016). This property of TDP-43 liquid droplets may be attributed to the transient nature of the α-helix segment of the PrLD, which readily undergoes α-β transitions, and ALS/FTD mutations that disrupt LLPS encouraged this conversion process (L. L. Jiang et al. 2013; Conicella et al. 2016).

In SGs, the β-rich core structures may be a means of liquid droplet maintenance to sequester proteins entering the aggregating phase and convert them to the droplet state through active, ATP-dependent remodeling mechanisms. SGs normally persist for only hours, and it is possible that prolonged or repeated environmental stresses can overwhelm the remodeling mechanisms and cause irreversible aggregates to form (Buchan, Yoon, and Parker 2011; Bentmann, Haass, and Dormann 2013). The conversion of cellular liquid droplets such as SGs into aggregates is a plausible pathway through which TDP-43 can form pathological inclusions. It follows that any factors that enhances TDP-43’s propensity to enter the aggregate state such as disruption of the α-helix interactions, an increase in the propensity to form β-folds, or alterations to RNA content of SGs would accelerate the process of droplet-aggregate conversion. Notably, a significant number of ALS/FTD-associated mutations in FUS and hnRNPA1 as well as nearly all disease-associated mutations in TDP-43 occur in their PrLD, and these mutations cause defects in droplet formation leading to decreased reversibility and higher propensity to convert into fibrils (Lagier-Tourenne, Polymenidou, and Cleveland 2010; Kim et al.

2013; Burke et al. 2015; Molliex et al. 2015; Murakami et al. 2015). In FUS, ALS/FTD-associated mutations reduces the speed at which FUS traverses through the liquid droplet and causes formation of starburst-shaped aggregates after in vitro aging (Patel et al. 2015). In TDP-43, ALS/FTD mutations that occur at the α-helix segment cause disruptions to the intermolecular interactions of the PrLD transient α-helices and altered droplet formation and reversibility (Conicella et al. 2016). Although the effect of mutations outside the α-helix segment on the properties of liquid droplets remains to be tested, we speculate that these mutations may affect LLPS formation through altered protein-protein interactions. Under physiological conditions, membraneless organelles are not purely protein droplets but contain a collection of different proteins and their associated RNA targets, and the biophysical properties of these droplets such as viscosity and droplet fusion rates, can be altered depending on the identity of the RNA to which the proteins are bound (H. Zhang et al. 2015). Thus, mutations on the PrLD that can potentially cause aberrant recruitment of RNA or RNA-protein complexes can contribute to aggregate formation in this manner. The specific occurrence of ALS/FTD mutations on the PrLD suggests that the disease pathology is intimately linked to the formation and aggregation within SGs.

Spreading and propagation of ALS/FTD

An unanswered question of ALS/FTD is the mechanism of cell-to-cell spread of the disease. It is tempting to attribute the spreading agent to the aggregation-prone, β-rich fold of the TDP-43 PrLD, as prions spread by template-directed misfolding in prion disease. In vitro aggregates of recombinant TDP-43 can seed the aggregation of endogenous TDP-43 when it is transduced into HEK293T cells, and multiple TDP-43 CTD fragments containing residues 287-322 are capable of forming amyloid that possesses the same seeding properties as classic amyloid fibrils (Furukawa et al. 2011; Chen et al. 2010; Liu et al. 2013). Moreover, transduction of insoluble fractions from ALS/FTD patient brain lysates into SH-SY5Y neuronal cells can lead to the formation of phosphorylated and ubiquitinated aggregates (Nonaka et al. 2013). For cell-to-cell transmission to occur, the aggregation-prone form of TDP-43 must be expelled from the affected neuron, through mechanisms such as cell death, and cross the cell membrane into the extracellular milieu and be taken up by a subsequent neuron. One proposed mechanism of membrane entry is through the microvesicle/exosome pathway through axonal terminals, while others have proposed that aggregates can rupture unstructured macropinosomes through “membrane ruffling” to gain cell entry (Feiler et al. 2015; Zeineddine et al. 2015). Recent findings suggest that segments containing the α-helical region of the PrLD (311-343) can interact with DMPC/DHPC bicelles, suggesting the possibility of membrane disruption through membrane-helix

23 interactions (Liu et al. 2013; Lim et al. 2016). The formation of α-helices may also play a more direct role in cell-to-cell spreading, because α-helices are also known to form large supramolecular structures through helix-helix interactions between varying numbers of helices such as three-helix micelles that are used as nanocarrier polymers that can cross a multitude of biological barriers for the delivery of its contents (Ang et al. 2016). Further studies to identify the mechanism and agent of cell-to-cell spread can open new avenues of research for the treatment of ALS/FTD by potentially blocking the entry of these agents from the affected neuron to neighboring healthy neurons or sequester these agents using conformation-specific antibodies before they can gain cellular entry.

Chapter Remarks Concluding remarks on current literature

Recent structural findings in the TDP-43 CTD have produced unprecedented insights into the molecular mechanism of TDP-43 function and aggregation. In particular, the recruitment of TDP-43 into liquid droplets through its PrLD has provided a strong line of evidence of the intimate link between SG formation and pathological aggregation. The molecular mechanisms of pathogenesis are starting to come into focus (summarized in Figure 1.3). The PrLD containing the amyloidogenic core plays a central role in TDP-43 phase changes and there is a clear correlation between secondary structure of the PrLD and the propensity to form functional versus pathological states. The conversion between these conformations may be a key step in pathogenesis. TDP-43 PrLD appears to wobble at a cusp of the protein folding energy landscape, where any number of factors can tip the balance and trigger the conversion from its native function to its pathological one. This structural conversion could be template-directed in nature, and its occurrence in soluble cytosolic TDP-43 and subsequent formation of pathological aggregates cannot be ruled out. However, as a more exquisite alternative, the conversion may occur within liquid droplets such as SGs, where the increased protein density can accelerate the propagation of β-rich folds. SGs have an intrinsic propensity to aggregate, and mutations that render TDP-43 more prone to adopt a β-fold at the PrLD, which alter concentration through cytoplasmic mislocalization or can alter protein-protein interactions and in turn alter RNA composition, can all affect SG assembly and dynamics. Additionally, environmental factors such as persistent or chronic stress due to exposure to extreme temperatures, toxins or physical harm, can also accelerate this intrinsic property of SGs. The clearance of these β-rich structures may eventually become stagnant because of the inevitable age-dependent decline of the protein quality control mechanisms, leading to accumulation of aggregates and neurotoxicity.

Figure 1.3: Graphical representation of TDP-43 aggregation model. TDP-43 is represented by a round blue circle (N- terminus) and two orange squares (tandem RRMs) bound to RNA (red half-ladders), attached to a C-terminal helix (blue lines) in its native state as a dimer. In the stress granule, TDP-43 C-terminal helix interact to form liquid droplets containing other proteins (green circles) and RNA (green half-ladders). The structural transition of the prion-like domain of TDP-43 to β-rich folds is represented with pink antiparallel arrows. The fibril-like arrangement of the antiparallel arrows reflects amyloid-like folds associated with stable stress granule cores or pathological aggregates in the cytoplasm. Post-translational modifications to these aggregates are represented by yellow circles labeled P for phosphorylation, or purple circles labeled U to represent ubiquitination. The pink-blue gradient on the left represents the conversion of α-helix to β-rich folds in the prion-like domain, reflecting the transition from native to pathological TDP-43 states. The schematic presents both stress granule- dependent and -independent TDP-43 aggregation pathways.

Acknowledgements

This study was supported in whole or in part by the Canadian Consortium of Neurodegeneration and Aging (CCNA), the Canadian Institute of Health Research (CIHR), the ALS Society of Canada (ALS Canada) and the Alzheimer Society of Canada (ASC). The authors would like to thank Kevin C. Hadley for helpful discussions and critically reading the document.

Thesis Rationale Current studies indicate that aggregates of TDP-43 are found in 97% of ALS cases and 50% of FTD cases, suggesting that TDP-43 likely plays a key role in the ALS/FTD disease mechanism. Recent findings of TDP-43’s recruitment into phase-separated organelles such as stress granules (SGs), and the ability of the protein’s CTD to form both phase-separated protein droplets and disease-like aggregates in vitro suggests the interplay of environmental stress and ALS pathology. It has been suggested that ALS pathogenesis may arise from failure of TDP-43 to disassemble in SGs due to chronic or persistent stress, leading to its conversion into pathological inclusions. Thus, the detailed molecular characterization of TDP-43 aggregation and phase separation and its relevance to the formation of stress granules is immensely important in understanding the mechanisms of ALS/FTD pathology. The link between liquid-liquid phase separation and protein aggregation has become a focus of ALS research only recently. The potential for environmental stress factors to play a role in disease may finally account for the largely sporadic nature of these diseases. Identification of modulators of these processes could be exploited to develop targeted approaches for treating TDP- 43 proteinopathies. We hypothesize that TDP-43 aggregation and phase separation are dynamic processes modulated by the cellular environment, and that alterations to these factors from environmental sources may convert the normally dynamic phase-separated structures into amorphous, pathological ones. In this thesis, we took two approaches to probe the linkage between TDP-43 aggregation and its involvement in the formation of phase-separated organelles. As such, it can be divided into two parts, encompassing two primary objectives.

Our first objective focuses on the biophysical characterization of TDP-43 in vitro. In this hypothesis driven approach, we aim to identify factors that affect TDP-43 aggregation using recombinantly expressed TDP-43. The first part of the thesis (Chapter II and Chapter III) presents the findings of this experimental approach. Our first aim was to characterize the aggregation pathway of TDP-43 and identify molecular chaperones that could prevent TDP-43 from entering the aggregated state. This study is presented in Chapter II, which first appeared in 2014 in Biochemistry. It describes the aggregation of wild-type TDP-43 conjugated to an N-terminal YFP tag. This study shows that TDP- 43 converts from its native, dimeric form, into monomeric TDP-43, and subsequently forms amorphous aggregates under physiological salt conditions. This aggregation process could be inhibited at its initial stages by binding to synthetic (poly-TG single-stranded DNA) or natural (TDP-43 mRNA) oligonucleotide targets of TDP-43 at substoichiometric concentrations. This suggests a previously undescribed interplay between TDP-43 solubility and its autoregulation.

Our next aim was to identify the driving forces behind the aggregation of TDP-43. This study is presented in Chapter III, which is being submitted to eLife as of this writing. In this study, we show that electrolytes at physiological concentrations are specific modulators of TDP-43 aggregation. The data in this chapter showcased this electrolyte-dependence of aggregation in solution through turbidity, light scattering, and electron microscopy techniques. We discovered novel properties of TDP-43 aggregation such as reversibility through electrolyte depletion, and specific aggregation sensitivity to Ca2+ ions. We also asked whether these properties of TDP-43 still apply to the protein when it is recruited into a droplet organelle. To characterize TDP-43 in this droplet state, we constructed an artificial stress granule by using pre-formed protein droplets as a scaffold. We find that under these conditions, TDP-43 readily enters these droplet structures and preferentially aggregates at the droplet boundary over time in a physiological ion-dependent process. These effects, caused by submolar concentrations of electrolytes, resemble the electrolyte effects that cause many proteins to phase separate in solution, suggesting a linkage between TDP-43 aggregation and phase-separation.

The second objective, presented in the second part of this thesis (Chapters IV, V, and VI), is to use a discovery-based approach to find direct evidence for the linkage between TDP-43-positive inclusions in patient tissue and stress granules by comparing the proteomic compositions of these structures. To do so, we developed the novel method STOMP (spatially targeted optical microproteomics), which combines photochemistry, fluorescence microscopy, and tandem mass spectrometry, to interrogate the proteomic content of micron-scale features. Our first aim was to validate the efficacy of this method in a well-studied disease model. This study is described in Chapter IV, which first appeared in eLife in 2015, and acts as an introduction to the principles of the STOMP technique. We examine amyloid plaques in an Alzheimer’s disease (AD) mouse model and a post-mortem human AD case as proof of concept cases, and successfully confirm known plaque constituents while discovering new ones. Our next aim was to apply the STOMP method to identify the proteomic composition of ALS/FTD inclusions and stress granules. To do so, many technical challenges were addressed. One problem encountered in the preparation and staining of formalin-fixed aged human tissue for STOMP is the confounding factor of endogenous autofluorescence of the tissue. In Chapter V, we describe a simple and effective method to remove background autofluorescence using a commercially available LED desk lamp. This chapter first appeared in Biochemistry and Cellular Biology in 2016 and later in the Journal of Visualized Experiments in 2017 in video format. The aim of the technique is to acquire high- contrast fluorescence images that accurately defines the target region of interest, free of confounding

autofluorescence. We found that a 48-h photobleaching treatment was sufficient to substantially reduce autofluorescence and generate high quality images for STOMP analysis. Finally, Chapter VI documents the attempts of STOMP analysis on both TDP-43-positive inclusions in FTD tissue and in stress granules generated in HeLa cells. Additional challenges arose from these STOMP attempts and analysis of the overall feasibility of the technique using the current technology, suggested improvements to the method, and the technique’s future potential are discussed. A summary on the impact and outcomes of the findings in this thesis and the future directions of ALS research are discussed at the end of this concluding chapter.

References

Acharya, Kshitish K, Chhabi K Govind, Amy N Shore, Mark H Stoler, and Prabhakara P Reddi. 2006. “Cis-Requirement for the Maintenance of Round Spermatid-Specific Transcription.” Developmental Biology 295 (2): 781–90. doi:10.1016/j.ydbio.2006.04.443.

Anderson, Paul, and Nancy Kedersha. 2009. “RNA Granules: Post-Transcriptional and Epigenetic Modulators of Gene Expression.” Nature Reviews. Molecular Cell Biology 10 (6): 430–36. doi:10.1038/nrm2694.

Ang, JooChuan, Dan Ma, Reidar Lund, Sinan Keten, and Ting Xu. 2016. “Internal Structure of 15 Nm 3-Helix Micelle Revealed by Small-Angle Neutron Scattering and Coarse-Grained MD Simulation.” Biomacromolecules 17 (10): 3262–3267. doi:10.1021/acs.biomac.6b00986.

Arai, Tetsuaki, Masato Hasegawa, Haruhiko Akiyama, Kenji Ikeda, Takashi Nonaka, Hiroshi Mori, David Mann, et al. 2006. “TDP-43 Is a Component of Ubiquitin-Positive Tau-Negative Inclusions in Frontotemporal Lobar Degeneration and Amyotrophic Lateral Sclerosis.” Biochemical and Biophysical Research Communications 351 (3): 602–11. doi:10.1016/j.bbrc.2006.10.093.

Aulas, Anaïs, Stéphanie Stabile, and Christine Vande Velde. 2012. “Endogenous TDP-43, but Not FUS, Contributes to Stress Granule Assembly via G3BP.” Molecular Neurodegeneration 7 (January): 54. doi:10.1186/1750-1326-7-54.

Ayala, Youhna M., Paola Zago, Andrea D’Ambrogio, Ya-Fei Xu, Leonard Petrucelli, Emanuele Buratti, and Francisco E. Baralle. 2008. “Structural Determinants of the Cellular Localization and Shuttling of TDP-43.” Journal of Cell Science 121 (22): 3778–85. doi:10.1242/jcs.038950.

Ayala, Youhna M, Laura De Conti, S Eréndira Avendaño-Vázquez, Ashish Dhir, Maurizio Romano, Andrea D’Ambrogio, James Tollervey, et al. 2011. “TDP-43 Regulates Its mRNA Levels through a Negative Feedback Loop.” The EMBO Journal 30 (2): 277–88. doi:10.1038/emboj.2010.310.

Ayala, Youhna M, Tom Misteli, and Francisco E Baralle. 2008. “TDP-43 Regulates Retinoblastoma Protein Phosphorylation through the Repression of Cyclin-Dependent Kinase 6 Expression.” Proceedings of the National Academy of Sciences 105 (10): 3785–89. doi:10.1073/pnas.0800546105.

Baker, Matt, Ian R Mackenzie, Stuart M Pickering-Brown, Jennifer Gass, Rosa Rademakers, Caroline Lindholm, Julie Snowden, et al. 2006. “Mutations in Progranulin Cause Tau-Negative Frontotemporal Dementia Linked to Chromosome 17.” Nature 442 (7105): 916–19. doi:10.1038/nature05016.

Barmada, Sami J, Gaia Skibinski, Erica Korb, Elizabeth J Rao, Jane Y Wu, and Steven Finkbeiner. 2010. “Cytoplasmic Mislocalization of TDP-43 Is Toxic to Neurons and Enhanced by a Mutation Associated with Familial Amyotrophic Lateral Sclerosis.” The Journal of Neuroscience 30 (2): 639–49. doi:10.1523/JNEUROSCI.4988-09.2010.

Bäumer, D, N Parkinson, and K Talbot. 2009. “TARDBP in Amyotrophic Lateral Sclerosis: Identification of a Novel Variant but Absence of Copy Number Variation.” Journal of Neurology, Neurosurgery, and Psychiatry 80 (11): 1283–85. doi:10.1136/jnnp.2008.166512.

Bembich, Sara, Jeremias S Herzog, Laura De Conti, Cristiana Stuani, S Eréndira Avendaño-Vázquez, Emanuele Buratti, Marco Baralle, and Francisco E Baralle. 2014. “Predominance of Spliceosomal Complex Formation over Polyadenylation Site Selection in TDP-43 Autoregulation.” Nucleic Acids Research 42 (5): 3362–71. doi:10.1093/nar/gkt1343.

Benajiba, Lina, Isabelle Le Ber, Agnès Camuzat, Mathieu Lacoste, Catherine Thomas-Anterion, Philippe Couratier, Solenn Legallic, et al. 2009. “TARDBP Mutations in Motoneuron Disease with Frontotemporal Lobar Degeneration.” Annals of Neurology 65 (4): 470–73. doi:10.1002/ana.21612.

Bentmann, Eva, Christian Haass, and Dorothee Dormann. 2013. “Stress Granules in Neurodegeneration--Lessons Learnt from TAR DNA Binding Protein of 43 kDa and Fused in Sarcoma.” The FEBS Journal 280 (18): 4348–70. doi:10.1111/febs.12287.

Bolton, D C, M P McKinley, and S B Prusiner. 1982. “Identification of a Protein That Purifies with the Scrapie Prion.” Science 218 (4579): 1309–11. doi:10.1017/CBO9781107415324.004.

Borroni, B, S Archetti, R Del Bo, A Papetti, E Buratti, C Bonvicini, C Agosti, et al. 2010. “TARDBP Mutations in Frontotemporal Lobar Degeneration: Frequency, Clinical Features, and Disease Course.” Rejuvenation Research 13 (5): 509–17. doi:10.1089/rej.2010.1017.

Borroni, B, C Bonvicini, A Alberici, E Buratti, C Agosti, S Archetti, A Papetti, et al. 2009. “Mutation within TARDBP Leads to Frontotemporal Dementia without Motor Neuron Disease.” Human Mutation 30 (11): E974-83. doi:10.1002/humu.21100.

Brangwynne, Clifford P, Christian R Eckmann, David S Courson, Agata Rybarska, Carsten Hoege, Jöbin Gharakhani, Frank Jülicher, and Anthony A Hyman. 2009. “Germline P Granules Are Liquid Droplets That Localize by Controlled Dissolution/condensation.” Science 324 (5935): 1729–32. doi:10.1126/science.1172046.

Brooks, B R, R G Miller, M Swash, and T L Munsat. 2000. “El Escorial Revisited: Revised Criteria for the Diagnosis of Amyotrophic Lateral Sclerosis.” Amyotrophic Lateral Sclerosis and Other Motor Neuron Disorders 1 (5): 293–99. doi:DOI 10.1080/146608200300079536.

Buchan, J Ross. 2014. “mRNP Granules. Assembly, Function, and Connections with Disease.” RNA Biology 11 (8): 1019–30. doi:10.4161/15476286.2014.972208.

Buchan, J Ross, Tracy Nissan, and Roy Parker. 2010. Analyzing P-Bodies and Stress Granules in Saccharomyces Cerevisiae. Methods in Enzymology. 2nd ed. Vol. 470. doi:10.1016/S0076- 6879(10)70025-2.

Buchan, J Ross, Je-Hyun Yoon, and Roy Parker. 2011. “Stress-Specific Composition, Assembly and Kinetics of Stress Granules in Saccharomyces Cerevisiae.” Journal of Cell Science 124 (Pt 2): 228–39. doi:10.1242/jcs.078444.

Budini, Mauricio, and Emanuele Buratti. 2011. “TDP-43 Autoregulation: Implications for Disease.” Journal of Molecular Neuroscience 45 (3): 473–79. doi:10.1007/s12031-011-9573-8.

Budini, Mauricio, Emanuele Buratti, Cristiana Stuani, Corrado Guarnaccia, Valentina Romano, Laura De Conti, and Francisco E Baralle. 2012. “Cellular Model of TAR DNA-Binding Protein 43

(TDP-43) Aggregation Based on Its C-Terminal Gln/Asn-Rich Region.” The Journal of Biological Chemistry 287 (10): 7512–25. doi:10.1074/jbc.M111.288720.

Budini, Mauricio, Valentina Romano, Zainuddin Quadri, Emanuele Buratti, and Francisco E Baralle. 2015. “TDP-43 Loss of Cellular Function through Aggregation Requires Additional Structural Determinants beyond Its C-Terminal Q/N Prion-like Domain.” Human Molecular Genetics 24 (1): 9– 20. doi:10.1093/hmg/ddu415.

Buratti, Emanuele, and Francisco E Baralle. 2001. “Characterization and Functional Implications of the RNA Binding Properties of Nuclear Factor TDP-43, a Novel Splicing Regulator of CFTR Exon 9.” The Journal of Biological Chemistry 276 (39): 36337–43. doi:10.1074/jbc.M104236200.

Buratti, Emanuele, Antonia Brindisi, Maurizio Giombi, Sergio Tisminetzky, Youhna M Ayala, and Francisco E Baralle. 2005. “TDP-43 Binds Heterogeneous Nuclear Ribonucleoprotein A/B through Its C-Terminal Tail: An Important Region for the Inhibition of Cystic Fibrosis Transmembrane Conductance Regulator Exon 9 Splicing.” The Journal of Biological Chemistry 280 (45): 37572–84. doi:10.1074/jbc.M505557200.

Buratti, Emanuele, Antonia Brindisi, Franco Pagani, and Francisco E Baralle. 2004. “Nuclear Factor TDP-43 Binds to the Polymorphic TG Repeats in CFTR Intron 8 and Causes Skipping of Exon 9: A Functional Link with Disease Penetrance.” American Journal of Human Genetics 74 (6): 1322–25. doi:10.1086/420978.

Buratti, Emanuele, T Dörk, Elisabetta Zuccato, Franco Pagani, Maurizio Romano, and Francisco E. Baralle. 2001. “Nuclear Factor TDP-43 and SR Proteins Promote in Vitro and in Vivo CFTR Exon 9 Skipping.” The EMBO Journal 20 (7): 1774–84. doi:10.1093/emboj/20.7.1774.

Burke, Kathleen A., Abigail M. Janke, Christy L. Rhine, and Nicolas L. Fawzi. 2015. “Residue-by- Residue View of In Vitro FUS Granules That Bind the C-Terminal Domain of RNA Polymerase II.” Molecular Cell 60 (2): 231–41. doi:10.1016/j.molcel.2015.09.006.

Cairns, N J, R J Perrin, R E Schmidt, A Gru, K G Green, D Carter, L Taylor-Reinwald, J C Morris, M A Gitcho, and R H Baloh. 2010. “TDP-43 Proteinopathy in Familial Motor Neurone Disease with TARDBP A315T Mutation: A Case Report.” Neuropathology and Applied Neurobiology 36 (7): 673– 79. doi:10.1111/j.1365-2990.2010.01121.x.

Capitini, Claudia, Simona Conti, Michele Perni, Francesca Guidi, Roberta Cascella, Angela De Poli, Amanda Penco, Annalisa Relini, Cristina Cecchi, and Fabrizio Chiti. 2014. “TDP-43 Inclusion Bodies Formed in Bacteria Are Structurally Amorphous, Non-Amyloid and Inherently Toxic to Neuroblastoma Cells.” PloS One 9 (1): e86720. doi:10.1371/journal.pone.0086720.

Chang, Chung-ke, Tzong-Huah Wu, Chu-Ya Wu, Ming-hui Chiang, Elsie Khai-Woon Toh, Yin- Chih Hsu, Ku-Feng Lin, Yu-heng Liao, Tai-huang Huang, and Joseph Jen-Tse Huang. 2012. “The N-Terminus of TDP-43 Promotes Its Oligomerization and Enhances DNA Binding Affinity.” Biochemical and Biophysical Research Communications 425 (2): 219–24. doi:10.1016/j.bbrc.2012.07.071.

Chen, Allan K H, Ryan Y Y Lin, Eva Z J Hsieh, Pang Hsien Tu, Rita P Y Chen, Tai Yan Liao, Wenlung Chen, Chih Hsien Wang, and Joseph J T Huang. 2010. “Induction of Amyloid Fibrils by the C-Terminal Fragments of TDP-43 in Amyotrophic Lateral Sclerosis.” Journal of the American Chemical Society 132 (4): 1186–87. doi:10.1021/ja9066207.

Chiang, Chien-Hao, Cédric Grauffel, Lien-Szu Wu, Pan-Hsien Kuo, Lyudmila G Doudeva, Carmay Lim, Che-Kun James Shen, and Hanna S Yuan. 2016. “Structural Analysis of Disease-Related TDP- 43 D169G Mutation: Linking Enhanced Stability and Caspase Cleavage Efficiency to Protein Accumulation.” Scientific Reports 6 (February): 21581. doi:10.1038/srep21581.

Chiang, Huei-Hsin, Peter M Andersen, Ole-Bjørn Tysnes, Ole Gredal, Peter B Christensen, and Caroline Graff. 2012. “Novel TARDBP Mutations in Nordic ALS Patients.” Journal of Human Genetics 57 (5): 316–19. doi:10.1038/jhg.2012.24.

Chiò, Adriano, Giuseppe Borghero, Maura Pugliatti, Anna Ticca, Andrea Calvo, Cristina Moglia, Roberto Mutani, et al. 2011. “Large Proportion of Amyotrophic Lateral Sclerosis Cases in Sardinia due to a Single Founder Mutation of the TARDBP Gene.” Archives of Neurology 68 (5): 594–98. doi:10.1001/archneurol.2010.352.

Chong, P. Andrew, and Julie D. Forman-Kay. 2016. “A New Phase in ALS Research.” Structure 24 (9): 1435–36. doi:10.1016/j.str.2016.08.003.

Cohen, Todd J, Andrew W Hwang, Clark R Restrepo, Chao-Xing Yuan, John Q Trojanowski, and Virginia M Y Lee. 2015. “An Acetylation Switch Controls TDP-43 Function and Aggregation Propensity.” Nature Communications 6: 5845. doi:10.1038/ncomms6845.

Collins, Mahlon, David Riascos, Tina Kovalik, Jiyan An, Kelly Krupa, Kristin Krupa, Brian L. Hood, et al. 2012. “The RNA-Binding Motif 45 (RBM45) Protein Accumulates in Inclusion Bodies in Amyotrophic Lateral Sclerosis (ALS) and Frontotemporal Lobar Degeneration with TDP-43 Inclusions (FTLD-TDP) Patients.” Acta Neuropathologica 124 (5): 717–32. doi:10.1007/s00401-012- 1045-x.

Conforti, F L, W Sproviero, I L Simone, R Mazzei, P Valentino, C Ungaro, A Magariello, et al. 2011. “TARDBP Gene Mutations in South Italian Patients with Amyotrophic Lateral Sclerosis.” Journal of Neurology, Neurosurgery, and Psychiatry 82 (5): 587–88. doi:10.1136/jnnp.2009.198309.

Conicella, Alexander E, Gül H Zerze, Jeetain Mittal, and Nicolas L Fawzi. 2016. “ALS Mutations Disrupt Phase Separation Mediated by α-Helical Structure in the TDP-43 Low-Complexity C- Terminal Domain.” Structure 24 (9): 1537-49. doi:10.1016/j.str.2016.07.007.

Corrado, Lucia, A Ratti, C Gellera, E Buratti, B Castellotti, Y Carlomagno, N Ticozzi, et al. 2009. “High Frequency of TARDBP Gene Mutations in Italian Patients with Amyotrophic Lateral Sclerosis.” Human Mutation 30 (4): 688–94. doi:10.1002/humu.20950.

Cox, B S. 1965. “Ψ, A Cytoplasmic Suppressor of Super-Suppressor in Yeast.” Heredity 20 (4): 505– 21. doi:10.1038/hdy.1965.65.

Cruts, M, I Gijselinck, J van der Zee, S Engelborghs, H Wils, D Pirici, R Rademakers, et al. 2006. “Null Mutations in Progranulin Cause Ubiquitin-Positive Frontotemporal Dementia Linked to Chromosome 17q21.” Nature 442 (7105): 920–24. doi:10.1038/nature05017.

Cruts, Marc, Ilse Gijselinck, Tim Van Langenhove, Julie van der Zee, and Christine Van Broeckhoven. 2013. “Current Insights into the C9orf72 Repeat Expansion Diseases of the FTLD/ALS Spectrum.” Trends in Neurosciences 36 (8): 450–59. doi:10.1016/j.tins.2013.04.010.

D’Ambrogio, Andrea, Emanuele Buratti, Cristiana Stuani, Corrado Guarnaccia, Maurizio Romano, Youhna M. Ayala, and Francisco E. Baralle. 2009. “Functional Mapping of the Interaction between TDP-43 and hnRNP A2 in Vivo.” Nucleic Acids Research 37 (12): 4116–26. doi:10.1093/nar/gkp342.

Dammer, Eric B, Claudia Fallini, Yair M Gozal, Duc M Duong, Wilfried Rossoll, Ping Xu, James J Lah, et al. 2012. “Coaggregation of RNA-Binding Proteins in a Model of TDP-43 Proteinopathy with Selective RGG Motif Methylation and a Role for RRM1 Ubiquitination.” PloS One 7 (6): e38658. doi:10.1371/journal.pone.0038658.

Daoud, H, P N Valdmanis, E Kabashi, P Dion, N Dupré, W Camu, V Meininger, and G A Rouleau. 2009. “Contribution of TARDBP Mutations to Sporadic Amyotrophic Lateral Sclerosis.” Journal of Medical Genetics 46 (2): 112–14. doi:10.1136/jmg.2008.062463.

DeJesus-Hernandez, Mariely, Ian R. R Mackenzie, Bradley F. F Boeve, Adam L. L Boxer, Matt Baker, Nicola J. J Rutherford, Alexandra M. M Nicholson, et al. 2011. “Expanded GGGGCC Hexanucleotide Repeat in Noncoding Region of C9ORF72 Causes Chromosome 9p-Linked FTD and ALS.” Neuron 72 (2): 245–56. doi:10.1016/j.neuron.2011.09.011.

Del Bo, R, S Ghezzi, S Corti, M Pandolfo, M Ranieri, D Santoro, I Ghione, et al. 2009. “TARDBP (TDP-43) Sequence Analysis in Patients with Familial and Sporadic ALS: Identification of Two Novel Mutations.” European Journal of Neurology 16 (6): 727–32. doi:10.1111/j.1468- 1331.2009.02574.x.

Dewey, C. M., B. Cenik, C. F. Sephton, D. R. Dries, P. Mayer, S. K. Good, B. A. Johnson, J. Herz, and G. Yu. 2011. “TDP-43 Is Directed to Stress Granules by Sorbitol, a Novel Physiological Osmotic and Oxidative Stressor.” Molecular and Cellular Biology 31 (5): 1098–1108. doi:10.1128/MCB.01279-10.

Disset, Antoine, Carine Michot, Ann Harris, Emanuele Buratti, Mireille Claustres, and Sylvie Tuffery-Giraud. 2005. “A T3 Allele in the CFTR Gene Exacerbates Exon 9 Skipping in Vas Deferens and Epididymal Cell Lines and Is Associated with Congenital Bilateral Absence of Vas Deferens (CBAVD).” Human Mutation 25 (1): 72–81. doi:10.1002/humu.20115.

Dobson, Christopher M. 2002. “Getting out of Shape.” Nature 418 (6899): 729–30. doi:10.1038/418729a.

Dobson, Christopher M. 2003. “Protein Folding and Misfolding.” Nature 426 (6968): 884–90. doi:10.1038/nature02261.

Dormann, Dorothee, and Christian Haass. 2011. “TDP-43 and FUS: A Nuclear Affair.” Trends in Neurosciences 34 (7): 339–48. doi:10.1016/j.tins.2011.05.002.

Elbaum-Garfinkle, Shana, and Clifford P. Brangwynne. 2015. “Liquids, Fibers, and Gels: The Many Phases of Neurodegeneration.” Developmental Cell 35 (5): 531–32. doi:10.1016/j.devcel.2015.11.014.

Feiler, Marisa S., Benjamin Strobel, Axel Freischmidt, Anika M. Helferich, Julia Kappel, Bryson M. Brewer, Deyu Li, et al. 2015. “TDP-43 Is Intercellularly Transmitted across Axon Terminals.” Journal of Cell Biology 211 (4): 897–911. doi:10.1083/jcb.201504057.

Fujita, Y, M Ikeda, T Yanagisawa, Y Senoo, and K Okamoto. 2011. “Different Clinical and Neuropathologic Phenotypes of Familial ALS with A315E TARDBP Mutation.” Neurology 77 (15): 1427–31. doi:10.1212/WNL.0b013e318232ab87.

Furukawa, Yoshiaki, Kumi Kaneko, Shoji Watanabe, Koji Yamanaka, and Nobuyuki Nukina. 2011. “A Seeding Reaction Recapitulates Intracellular Formation of Sarkosyl-Insoluble Transactivation Response Element (TAR) DNA-Binding Protein-43 Inclusions.” The Journal of Biological Chemistry 286 (21): 18664–72. doi:10.1074/jbc.M111.231209.

Galant, Natalie J., Antoinette Bugyei-Twum, Rishi Rakhit, Patrick Walsh, Simon Sharpe, Pharhad Eli Arslan, Per Westermark, et al. 2016. “Substoichiometric Inhibition of Transthyretin Misfolding by Immune-Targeting Sparsely Populated Misfolding Intermediates: A Potential Diagnostic and Therapeutic for TTR Amyloidoses.” Scientific Reports 6 (April): 25080. doi:10.1038/srep25080.

Gilks, Natalie, Nancy Kedersha, Maranatha Ayodele, Lily Shen, Georg Stoecklin, Laura M Dember, and Paul Anderson. 2004. “Stress Granule Assembly Is Mediated by Prion-like Aggregation of TIA- 1.” Molecular Biology of the Cell 15 (12): 5383–98. doi:10.1091/mbc.E04-08-0715.

Giordana, Maria Teresa, Marco Piccinini, Silvia Grifoni, Giovanni De Marco, Marco Vercellino, Michela Magistrello, Alessia Pellerino, Barbara Buccinnà, Elisa Lupino, and Maria Teresa Rinaudo. 2010. “TDP-43 Redistribution Is an Early Event in Sporadic Amyotrophic Lateral Sclerosis.” Brain Pathology 20 (2): 351–60. doi:10.1111/j.1750-3639.2009.00284.x.

Gitcho, Michael A., Robert H. Baloh, Sumi Chakraverty, Kevin Mayo, Joanne B. Norton, Denise Levitch, Kimmo J. Hatanpaa, et al. 2008. “TDP-43 A315T Mutation in Familial Motor Neuron Disease.” Annals of Neurology 63 (4): 535–38. doi:10.1002/ana.21344.

Gitler, Aaron D, and James Shorter. 2011. “RNA-Binding Proteins with Prion-like Domains in ALS and FTLD-U.” Prion 5 (3): 179–87. doi:10.4161/pri.5.3.17230.

Glenner, George G., and Caine W. Wong. 1984. “Alzheimer’s Disease: Initial Report of the Purification and Characterization of a Novel Cerebrovascular Amyloid Protein.” Biochemical and Biophysical Research Communications 120 (3): 885–90. doi:10.1016/S0006-291X(84)80190-4.

Guerreiro, Rita J, Jennifer C Schymick, Cynthia Crews, Andrew Singleton, John Hardy, and Bryan J Traynor. 2008. “TDP-43 Is Not a Common Cause of Sporadic Amyotrophic Lateral Sclerosis.” PloS One 3 (6): e2450. doi:10.1371/journal.pone.0002450.

Halfmann, Randal, and Susan Lindquist. 2010. “Epigenetics in the Extreme: Prions and the Inheritance of Environmentally Acquired Traits.” Science 330 (6004): 629–32. doi:10.1126/science.1191081.

Henao-Mejia, Jorge, and Johnny J He. 2009. “Sam68 Relocalization into Stress Granules in Response to Oxidative Stress through Complexing with TIA-1.” Experimental Cell Research 315 (19): 3381–95. doi:10.1016/j.yexcr.2009.07.011.

Holmes, Daniel L., Alex K. Lancaster, Susan Lindquist, and Randal Halfmann. 2013. “Heritable Remodeling of Yeast Multicellularity by an Environmentally Responsive Prion.” Cell 153 (1): 153– 65. doi:10.1016/j.cell.2013.02.026.

Huang, Rui, Deng-Fu Fang, Ming-Yi Ma, Xiao-Yan Guo, Bi Zhao, Yan Zeng, Dong Zhou, Yuan Yang, and Hui-Fang Shang. 2012. “TARDBP Gene Mutations among Chinese Patients with Sporadic Amyotrophic Lateral Sclerosis.” Neurobiology of Aging 33 (5): 1015.e1-6. doi:10.1016/j.neurobiolaging.2010.07.007.

Huang, Yi-Chen, Ku-Feng Lin, Ruei-Yu He, Pang-Hsien Tu, Jiri Koubek, Yin-Chih Hsu, and Joseph Jen-Tse Huang. 2013. “Inhibition of TDP-43 Aggregation by Nucleic Acid Binding.” PloS One 8 (5): e64002. doi:10.1371/journal.pone.0064002.

Huey, Edward D, Raffaele Ferrari, Jorge H Moreno, Christopher Jensen, Christopher M Morris, Felix Potocnik, Rajesh N Kalaria, et al. 2012. “FUS and TDP43 Genetic Variability in FTD and CBS.” Neurobiology of Aging 33 (5): 1016.e9-17. doi:10.1016/j.neurobiolaging.2011.08.004.

Hutton, M, C L Lendon, P Rizzu, M Baker, S Froelich, H Houlden, S Pickering-Brown, et al. 1998. “Association of Missense and 5’-splice-Site Mutations in Tau with the Inherited Dementia FTDP- 17.” Nature 393 (6686): 702–5. doi:10.1038/31508.

Igaz, Lionel M, Linda K Kwong, Alice Chen-Plotkin, Matthew J Winton, Travis L Unger, Yan Xu, Manuela Neumann, John Q Trojanowski, and Virginia M-Y Lee. 2009. “Expression of TDP-43 C- Terminal Fragments in Vitro Recapitulates Pathological Features of TDP-43 Proteinopathies.” The Journal of Biological Chemistry 284 (13): 8516–24. doi:10.1074/jbc.M809462200.

Iida, Aritoshi, Tetsumasa Kamei, Motoki Sano, Shuichi Oshima, Torao Tokuda, Yusuke Nakamura, and Shiro Ikegawa. 2012. “Large-Scale Screening of TARDBP Mutation in Amyotrophic Lateral Sclerosis in Japanese.” Neurobiology of Aging 33 (4): 786–90. doi:10.1016/j.neurobiolaging.2010.06.017.

Ishimaru, Daniella, Ana Paula D Ano Bom, Luís Maurício T R Lima, Pablo a Quesado, Marcos F C Oyama, Claudia V de Moura Gallo, Yraima Cordeiro, and Jerson L Silva. 2009. “Cognate DNA Stabilizes the Tumor Suppressor p53 and Prevents Misfolding and Aggregation.” Biochemistry 48 (26): 6126–35. doi:10.1021/bi9003028.

Jain, Saumya, Joshua R. Wheeler, Robert W. Walters, Anurag Agrawal, Anthony Barsic, and Roy Parker. 2016. “ATPase-Modulated Stress Granules Contain a Diverse Proteome and Substructure.” Cell 164 (3): 487–98. doi:10.1016/j.cell.2015.12.038.

Jiang, Lei-Lei, Jian Zhao, Xiao-Fang Yin, Wen-Tian He, Hui Yang, Mei-Xia Che, and Hong-Yu Hu. 2016. “Two Mutations G335D and Q343R within the Amyloidogenic Core Region of TDP-43 Influence Its Aggregation and Inclusion Formation.” Scientific Reports 6 (July): 23928. doi:10.1038/srep23928.

Jiang, Lei Lei, Mei Xia Che, Jian Zhao, Chen Jie Zhou, Mu Yun Xie, Hai Yin Li, Jian Hua He, and Hong Yu Hu. 2013. “Structural Transformation of the Amyloidogenic Core Region of TDP-43 Protein Initiates Its Aggregation and Cytoplasmic Inclusion.” Journal of Biological Chemistry 288 (27): 19614–24. doi:10.1074/jbc.M113.463828.

Johnson, Brian S, David Snead, Jonathan J Lee, J Michael McCaffery, James Shorter, and Aaron D Gitler. 2009. “TDP-43 Is Intrinsically Aggregation-Prone, and Amyotrophic Lateral Sclerosis-Linked Mutations Accelerate Aggregation and Increase Toxicity.” The Journal of Biological Chemistry 284 (30): 20329–39. doi:10.1074/jbc.M109.010264.

Ju, XiaoDong, WenChao Liu, XiaoGang Li, Na Liu, Nan Zhang, Tao Liu, and Min Deng. 2016. “Two Distinct Clinical Features and Cognitive Impairment in Amyotrophic Lateral Sclerosis Patients with TARDBP Gene Mutations in the Chinese Population.” Neurobiology of Aging 38 (February): 216.e1-6. doi:10.1016/j.neurobiolaging.2015.10.032.

Kabashi, Edor, Paul N Valdmanis, Patrick Dion, Dan Spiegelman, Brendan J McConkey, Christine Vande Velde, Jean-Pierre Bouchard, et al. 2008. “TARDBP Mutations in Individuals with Sporadic and Familial Amyotrophic Lateral Sclerosis.” Nature Genetics 40 (5): 572–74. doi:10.1038/ng.132.

Kamada, Masaki, Hirofumi Maruyama, Eiji Tanaka, Hiroyuki Morino, Reika Wate, Hidefumi Ito, Hirofumi Kusaka, et al. 2009. “Screening for TARDBP Mutations in Japanese Familial Amyotrophic Lateral Sclerosis.” Journal of the Neurological Sciences 284 (1–2): 69–71. doi:10.1016/j.jns.2009.04.017.

Kato, Masato, Tina W. Han, Shanhai Xie, Kevin Shi, Xinlin Du, Leeju C. Wu, Hamid Mirzaei, et al. 2012. “Cell-Free Formation of RNA Granules: Low Complexity Sequence Domains Form Dynamic Fibers within Hydrogels.” Cell 149 (4): 753–67. doi:10.1016/j.cell.2012.04.017.

Kedersha, Nancy, Samantha Chen, Natalie Gilks, Wei Li, Ira J Miller, Joachim Stahl, and Paul Anderson. 2002. “Evidence That Ternary Complex (eIF2-GTP-tRNA(i)(Met))-Deficient Preinitiation Complexes Are Core Constituents of Mammalian Stress Granules.” Molecular Biology of the Cell 13 (1): 195–210. doi:10.1091/mbc.01-05-0221.

Kerman, Aaron, Hsueh-Ning Liu, Sidney Croul, Juan Bilbao, Ekaterina Rogaeva, Lorne Zinman, Janice Robertson, and Avijit Chakrabartty. 2010. “Amyotrophic Lateral Sclerosis Is a Non-Amyloid Disease in Which Extensive Misfolding of SOD1 Is Unique to the Familial Form.” Acta Neuropathologica 119 (3): 335–44. doi:10.1007/s00401-010-0646-5.

Kim, Hong Joo, Nam Chul Kim, Yong-Dong Wang, Emily A. Scarborough, Jennifer Moore, Zamia Diaz, Kyle S. MacLea, et al. 2013. “Mutations in Prion-like Domains in hnRNPA2B1 and hnRNPA1 Cause Multisystem Proteinopathy and ALS.” Nature 495 (7442): 467–73. doi:10.1038/nature11922.

Kirby, Janine, Emily F Goodall, William Smith, J Robin Highley, Rudo Masanzu, Judith A Hartley, Rachel Hibberd, et al. 2010. “Broad Clinical Phenotypes Associated with TAR-DNA Binding Protein (TARDBP) Mutations in Amyotrophic Lateral Sclerosis.” Neurogenetics 11 (2): 217–25. doi:10.1007/s10048-009-0218-9.

Kovacs, Gabor G, Jill R Murrell, Sandor Horvath, Laszlo Haraszti, Katalin Majtenyi, Maria J Molnar, Herbert Budka, Bernardino Ghetti, and Salvatore Spina. 2009. “TARDBP Variation Associated with Frontotemporal Dementia, Supranuclear Gaze Palsy, and Chorea.” Movement Disorders 24 (12): 1843–47. doi:10.1002/mds.22697.

Kühnlein, Peter, Anne-Dorte Sperfeld, Ben Vanmassenhove, Vivianna Van Deerlin, Virginia M-Y Lee, John Q Trojanowski, Hans A Kretzschmar, Albert C Ludolph, and Manuela Neumann. 2008. “Two German Kindreds with Familial Amyotrophic Lateral Sclerosis due to TARDBP Mutations.” Archives of Neurology 65 (9): 1185–89. doi:10.1001/archneur.65.9.1185.

Kuo, Pan-Hsien, Lyudmila G Doudeva, Yi-Ting Wang, Che-Kun James Shen, and Hanna S Yuan. 2009. “Structural Insights into TDP-43 in Nucleic-Acid Binding and Domain Interactions.” Nucleic Acids Research 37 (6): 1799–1808. doi:10.1093/nar/gkp013.

Kuo, Pan Hsien, Chien Hao Chiang, Yi Ting Wang, Lyudmila G. Doudeva, and Hanna S. Yuan. 2014. “The Crystal Structure of TDP-43 RRM1-DNA Complex Reveals the Specific Recognition for UG- and TG-Rich Nucleic Acids.” Nucleic Acids Research 42 (7): 4712–22. doi:10.1093/nar/gkt1407.

Kwiatkowski, T J, D A Bosco, a L Leclerc, E Tamrazian, C R Vanderburg, C Russ, A Davis, et al. 2009. “Mutations in the FUS/TLS Gene on Chromosome 16 Cause Familial Amyotrophic Lateral Sclerosis.” Science 323 (5918): 1205–8. doi:10.1126/science.1166066.

la Cour, Tanja, Lars Kiemer, Anne Mølgaard, Ramneek Gupta, Karen Skriver, and Søren Brunak. 2004. “Analysis and Prediction of Leucine-Rich Nuclear Export Signals.” Protein Engineering, Design & Selection 17 (6): 527–36. doi:10.1093/protein/gzh062.

Lagier-Tourenne, Clotilde, Magdalini Polymenidou, and Don W Cleveland. 2010. “TDP-43 and FUS/TLS: Emerging Roles in RNA Processing and Neurodegeneration.” Human Molecular Genetics 19 (R1): R46-64. doi:10.1093/hmg/ddq137.

Langellotti, S., V. Romano, G. Romano, R. Klima, F. Feiguin, L. Cragnaz, M. Romano, and F. E. Baralle. 2016. “A Novel Fly Model of TDP-43 Proteinopathies: N-Terminus Sequences Combined with the Q/N Domain Induce Protein Functional Loss and Locomotion Defects.” Disease Models & Mechanisms 9 (6): 659–69. doi:10.1242/dmm.023382.

Lashley, Tammaryn, Jonathan D. Rohrer, Simon Mead, and Tamas Revesz. 2015. “Review: An Update on Clinical, Genetic and Pathological Aspects of Frontotemporal Lobar Degenerations.” Neuropathology and Applied Neurobiology 41 (7): 858-81. doi:10.1111/nan.12250.

Lee, Edward B, Virginia M-Y Lee, and John Q Trojanowski. 2012. “Gains or Losses: Molecular Mechanisms of TDP43-Mediated Neurodegeneration.” Nature Reviews. Neuroscience 13 (1): 38–50. doi:10.1038/nrn3121.

Lemmens, R, V Race, N Hersmus, G Matthijs, L Van Den Bosch, P Van Damme, B Dubois, S Boonen, A Goris, and W Robberecht. 2009. “TDP-43 M311V Mutation in Familial Amyotrophic Lateral Sclerosis.” Journal of Neurology, Neurosurgery, and Psychiatry 80 (3): 354–55. doi:10.1136/jnnp.2008.157677.

Li, Y. R., O. D. King, J. Shorter, and A. D. Gitler. 2013. “Stress Granules as Crucibles of ALS Pathogenesis.” The Journal of Cell Biology 201 (3): 361–72. doi:10.1083/jcb.201302044.

Lim, Liangzhong, Yuanyuan Wei, Yimei Lu, and Jianxing Song. 2016. “ALS-Causing Mutations Significantly Perturb the Self-Assembly and Interaction with Nucleic Acid of the Intrinsically Disordered Prion-Like Domain of TDP-43.” PLoS Biology 14 (1): e1002338. doi:10.1371/journal.pbio.1002338.

Lin, Yuan, David S W Protter, Michael K. Rosen, and Roy Parker. 2015. “Formation and Maturation of Phase-Separated Liquid Droplets by RNA-Binding Proteins.” Molecular Cell 60 (2): 208–19. doi:10.1016/j.molcel.2015.08.018.

Ling, Shuo Chien, Magdalini Polymenidou, and Don W. Cleveland. 2013. “Converging Mechanisms in Als and FTD: Disrupted RNA and Protein Homeostasis.” Neuron 79 (3): 416–38. doi:10.1016/j.neuron.2013.07.033.

Liu-Yesucevitz, Liqun, Aylin Bilgutay, Yong-Jie Zhang, Tara Vanderweyde, Tara Vanderwyde, Allison Citro, Tapan Mehta, et al. 2010. “Tar DNA Binding Protein-43 (TDP-43) Associates with Stress Granules: Analysis of Cultured Cells and Pathological Brain Tissue.” PloS One 5 (10): e13250. doi:10.1371/journal.pone.0013250.

Liu, Gerard Chun-Hao, Bryan Po-Wen Chen, Nancy Ting-Juan Ye, Chih-Hsien Wang, Wenlung Chen, Hsien-Ming Lee, Sunney I Chan, and Joseph Jen-Tse Huang. 2013. “Delineating the Membrane-Disrupting and Seeding Properties of the TDP-43 Amyloidogenic Core.” Chemical Communications 49 (95): 11212–14. doi:10.1039/c3cc46762g.

Lukavsky, Peter J, Dalia Daujotyte, James R Tollervey, Jernej Ule, Cristiana Stuani, Emanuele Buratti, Francisco E Baralle, Fred F Damberger, and Frédéric H-T Allain. 2013. “Molecular Basis of UG-Rich RNA Recognition by the Human Splicing Factor TDP-43.” Nature Structural & Molecular Biology 20 (12): 1443–49. doi:10.1038/nsmb.2698.

Luquin, Natasha, Bing Yu, Rebecca B Saunderson, Ronald J Trent, and Roger Pamphlett. 2009. “Genetic Variants in the Promoter of TARDBP in Sporadic Amyotrophic Lateral Sclerosis.” Neuromuscular Disorders 19 (10): 696–700. doi:10.1016/j.nmd.2009.07.005.

Mackenzie, Ian R. A., and Manuela Neumann. 2016. “Molecular Neuropathology of Frontotemporal Dementia: Insights into Disease Mechanisms from Postmortem Studies.” Journal of Neurochemistry 138 (S1): 54-70 doi:10.1111/jnc.13588.

Mackenzie, Ian R A. 2007. “The Neuropathology of FTD Associated With ALS.” Alzheimer Disease and Associated Disorders 21 (4): S44-9. doi:10.1097/WAD.0b013e31815c3486.

Majoor-Krakauer, D, P J Willems, and a Hofman. 2003. “Genetic Epidemiology of Amyotrophic Lateral Sclerosis.” Clinical Genetics 63 (2): 83–101.

March, Zachary M., Oliver D. King, and James Shorter. 2016. “Prion-like Domains as Epigenetic Regulators, Scaffolds for Subcellular Organization, and Drivers of Neurodegenerative Disease.” Brain Research 1647: 1–14. doi:10.1016/j.brainres.2016.02.037.

Maris, Christophe, Cyril Dominguez, and Frédéric H T Allain. 2005. “The RNA Recognition Motif, a Plastic RNA-Binding Platform to Regulate Post-Transcriptional Gene Expression.” FEBS Journal 272 (9): 2118–31. doi:10.1111/j.1742-4658.2005.04653.x.

McDonald, Karli K, Anaïs Aulas, Laurie Destroismaisons, Sarah Pickles, Evghenia Beleac, William Camu, Guy a Rouleau, and Christine Vande Velde. 2011. “TAR DNA-Binding Protein 43 (TDP-43) Regulates Stress Granule Dynamics via Differential Regulation of G3BP and TIA-1.” Human Molecular Genetics 20 (7): 1400–1410. doi:10.1093/hmg/ddr021.

Mehta, Paul, Wendy Kaye, Leah Bryan, Theodore Larson, Timothy Copeland, Jennifer Wu, Oleg Muravov, and Kevin Horton. 2016. “Prevalence of Amyotrophic Lateral Sclerosis - United States, 2012-2013.” Morbidity and Mortality Weekly Report. Surveillance Summaries 65 (8): 1–12. doi:10.15585/mmwr.ss6508a1.

Mercado, Pablo Arrisi, Youhna M Ayala, Maurizio Romano, Emanuele Buratti, and Francisco E Baralle. 2005. “Depletion of TDP 43 Overrides the Need for Exonic and Intronic Splicing

Enhancers in the Human apoA-II Gene.” Nucleic Acids Research 33 (18): 6000–6010. doi:10.1093/nar/gki897.

Millecamps, Stéphanie, François Salachas, Cécile Cazeneuve, Paul Gordon, Bernard Bricka, Agnès Camuzat, Léna Guillot-Noël, et al. 2010. “SOD1, ANG, VAPB, TARDBP, and FUS Mutations in Familial Amyotrophic Lateral Sclerosis: Genotype-Phenotype Correlations.” Journal of Medical Genetics 47 (8): 554–60. doi:10.1136/jmg.2010.077180.

Mitrea, Diana M, and Richard W Kriwacki. 2016. “Phase Separation in Biology; Functional Organization of a Higher Order.” Cell Communication and Signaling 14 (1): 1. doi:10.1186/s12964-015- 0125-7.

Molliex, Amandine, Jamshid Temirov, Jihun Lee, Maura Coughlin, Anderson P. Kanagaraj, Hong Joo Kim, Tanja Mittag, and J. Paul Taylor. 2015. “Phase Separation by Low Complexity Domains Promotes Stress Granule Assembly and Drives Pathological Fibrillization.” Cell 163 (1): 123–33. doi:10.1016/j.cell.2015.09.015.

Mompeán, Miguel, Emanuele Buratti, Corrado Guarnaccia, Rui M M Brito, Avijit Chakrabartty, Francisco E Baralle, and Douglas V Laurents. 2014. “‘Structural Characterization of the Minimal Segment of TDP-43 Competent for Aggregation’.” Archives of Biochemistry and Biophysics 545 (March): 53–62. doi:10.1016/j.abb.2014.01.007.

Mompeán, Miguel, Rubén Hervás, Yunyao Xu, Timothy H. Tran, Corrado Guarnaccia, Emanuele Buratti, Francisco E Baralle, et al. 2015. “Structural Evidence of Amyloid Fibril Formation in the Putative Aggregation Domain of TDP-43.” Journal of Physical Chemistry Letters 6 (13): 2608–15. doi:10.1021/acs.jpclett.5b00918.

Mompeán, Miguel, Valentina Romano, David Pantoja-Uceda, Cristiana Stuani, Francisco E Baralle, Emanuele Buratti, and Douglas V Laurents. 2016. “The TDP-43 N-Terminal Domain Structure at High Resolution.” The FEBS Journal 283: 1–19. doi:10.1111/febs.13651.

Moreno, Fermin, Gil D Rabinovici, Anna Karydas, Zachary Miller, Sandy Chan Hsu, Andrea Legati, Jamie Fong, et al. 2015. “A Novel Mutation P112H in the TARDBP Gene Associated with Frontotemporal Lobar Degeneration without Motor Neuron Disease and Abundant Neuritic Amyloid Plaques.” Acta Neuropathologica Communications 3 (3–4): 19. doi:10.1186/s40478-015-0190-6.

Murakami, Tetsuro, Seema Qamar, Julie Qiaojin Lin, Gabriele S Kaminski Schierle, Eric Rees, Akinori Miyashita, Ana R. Costa, et al. 2015. “ALS/FTD Mutation-Induced Phase Transition of FUS Liquid Droplets and Reversible Hydrogels into Irreversible Hydrogels Impairs RNP Granule Function.” Neuron 88 (4): 678–90. doi:10.1016/j.neuron.2015.10.030.

Neary, D, J S Snowden, L Gustafson, U Passant, D Stuss, S Black, M Freedman, et al. 1998. “Frontotemporal Lobar Degeneration: A Consensus on Clinical Diagnostic Criteria.” Neurology 51 (6): 1546–54.

Neumann, Manuela, Deepak M Sampathu, Linda K Kwong, Adam C Truax, Matthew C Micsenyi, Thomas T Chou, Jennifer Bruce, et al. 2006. “Ubiquitinated TDP-43 in Frontotemporal Lobar Degeneration and Amyotrophic Lateral Sclerosis.” Science 314 (5796): 130–33. doi:10.1126/science.1134108.

Nishimoto, Yoshinori, Daisuke Ito, Takuya Yagi, Yoshihiro Nihei, Yoshiko Tsunoda, and Norihiro Suzuki. 2010. “Characterization of Alternative Isoforms and Inclusion Body of the TAR DNA- Binding Protein-43.” The Journal of Biological Chemistry 285 (1): 608–19. doi:10.1074/jbc.M109.022012.

Nonaka, Takashi, Masami Masuda-Suzukake, Tetsuaki Arai, Yoko Hasegawa, Hiroyasu Akatsu, Tomokazu Obi, Mari Yoshida, et al. 2013. “Prion-like Properties of Pathological TDP-43 Aggregates from Diseased Brains.” Cell Reports 4 (1): 124–34. doi:10.1016/j.celrep.2013.06.007.

Nozaki, Ichiro, Makoto Arai, Kazuya Takahashi, Tsuyoshi Hamaguchi, Hiroaki Yoshikawa, Toyoteru Muroishi, Moeko Noguchi-Shinohara, et al. 2010. “Familial ALS with G298S Mutation in TARDBP: A Comparison of CSF Tau Protein Levels with Those in Sporadic ALS.” Internal Medicine 49 (12): 1209–12.

Origone, Paola, Claudia Caponnetto, Monica Bandettini Di Poggio, Elisabetta Ghiglione, Emilia Bellone, Giovanna Ferrandes, Giovanni Luigi Mancardi, and Paola Mandich. 2010. “Enlarging Clinical Spectrum of FALS with TARDBP Gene Mutations: S393L Variant in an Italian Family Showing Phenotypic Variability and Relevance for Genetic Counselling.” Amyotrophic Lateral Sclerosis: 11 (1–2): 223–27. doi:10.3109/17482960903165039.

Orrù, S, E Manolakos, N Orrù, H Kokotas, V Mascia, C Carcassi, and M B Petersen. 2012. “High Frequency of the TARDBP Ala382Thr Mutation in Sardinian Patients with Amyotrophic Lateral Sclerosis.” Clinical Genetics 81 (2): 172–78. doi:10.1111/j.1399-0004.2011.01668.x.

Ou, S H, F Wu, D Harrich, L F García-Martínez, and R B Gaynor. 1995. “Cloning and Characterization of a Novel Cellular Protein, TDP-43, That Binds to Human Immunodeficiency Virus Type 1 TAR DNA Sequence Motifs.” Journal of Virology 69 (6): 3584–96.

Pamphlett, R, N Luquin, C McLean, S Kum Jew, and L Adams. 2009. “TDP-43 Neuropathology Is Similar in Sporadic Amyotrophic Lateral Sclerosis with or without TDP-43 Mutations.” Neuropathology and Applied Neurobiology 35 (2): 222–25. doi:10.1111/j.1365-2990.2008.00982.x.

Parker, Sarah J, Jodi Meyerowitz, Janine L James, Jeffrey R Liddell, Peter J Crouch, Katja M Kanninen, and Anthony R White. 2012. “Endogenous TDP-43 Localized to Stress Granules Can Subsequently Form Protein Aggregates.” Neurochemistry International 60 (4): 415–24. doi:10.1016/j.neuint.2012.01.019.

Patel, Avinash, Hyun O. Lee, Louise Jawerth, Shovamayee Maharana, Marcus Jahnel, Marco Y. Hein, Stoyno Stoynov, et al. 2015. “A Liquid-to-Solid Phase Transition of the ALS Protein FUS Accelerated by Disease Mutation.” Cell 162 (5): 1066–77. doi:10.1016/j.cell.2015.07.047.

Pesiridis, G Scott, Kalyan Tripathy, Selçuk Tanik, John Q Trojanowski, and Virginia M-Y Lee. 2011. “A ‘two-Hit’ hypothesis for Inclusion Formation by Carboxyl-Terminal Fragments of TDP-43 Protein Linked to RNA Depletion and Impaired Microtubule-Dependent Transport.” The Journal of Biological Chemistry 286 (21): 18845–55. doi:10.1074/jbc.M111.231118.

Phan, Anh Tuân, Vitaly Kuryavyi, Jennifer C Darnell, Alexander Serganov, Ananya Majumdar, Serge Ilin, Tanya Raslin, et al. 2011. “Structure-Function Studies of FMRP RGG Peptide Recognition of an RNA Duplex-Quadruplex Junction.” Nature Structural & Molecular Biology 18 (7): 796–804. doi:10.1038/nsmb.2064.

Qin, Haina, Liang-zhong Lim, Yuanyuan Wei, and Jianxing Song. 2014. “TDP-43 N Terminus Encodes a Novel Ubiquitin-like Fold and Its Unfolded Form in Equilibrium That Can Be Shifted by Binding to ssDNA.” Proceedings of the National Academy of Sciences 111 (52): 18619–24. doi:10.1073/pnas.1413994112.

Ratti, Antonia, and Emanuele Buratti. 2016. “Physiological Functions and Pathobiology of TDP-43 and FUS/TLS Proteins.” Journal of Neurochemistry 138 (S1): 95-111 doi:10.1111/jnc.13625.

Renton, Alan E, Elisa Majounie, Adrian Waite, Javier Simón-Sánchez, Sara Rollinson, J Raphael Gibbs, Jennifer C Schymick, et al. 2011. “A Hexanucleotide Repeat Expansion in C9ORF72 Is the Cause of Chromosome 9p21-Linked ALS-FTD.” Neuron 72 (2): 257–68. doi:10.1016/j.neuron.2011.09.010.

Rutherford, Nicola J, Yong-Jie Zhang, Matt Baker, Jennifer M Gass, Nicole a Finch, Ya-Fei Xu, Heather Stewart, et al. 2008. “Novel Mutations in TARDBP (TDP-43) in Patients with Familial Amyotrophic Lateral Sclerosis.” PLoS Genetics 4 (9): e1000193. doi:10.1371/journal.pgen.1000193.

Saini, Akash, and Virander Singh Chauhan. 2011. “Delineation of the Core Aggregation Sequences of TDP-43 C-Terminal Fragment.” Chembiochem 12 (16): 2495–2501. doi:10.1002/cbic.201100427.

Schmidt, Hermann Broder, and Rajat Rohatgi. 2016. “In Vivo Formation of Vacuolated Multi-Phase Report In Vivo Formation of Vacuolated Multi-Phase Compartments Lacking Membranes.” Cell Reports 16 (5): 1–9. doi:10.1016/j.celrep.2016.06.088.

Sephton, Chantelle F., Shannon K. Good, Stan Atkin, Colleen M. Dewey, Paul Mayer, Joachim Herz, and Gang Yu. 2010. “TDP-43 Is a Developmentally Regulated Protein Essential for Early Embryonic Development.” The Journal of Biological Chemistry 285 (9): 6826–34. doi:10.1074/jbc.M109.061846.

Seyfried, Nicholas T, Yair M Gozal, Eric B Dammer, Qiangwei Xia, Duc M Duong, Dongmei Cheng, James J Lah, Allan I Levey, and Junmin Peng. 2010. “Multiplex SILAC Analysis of a Cellular TDP-43 Proteinopathy Model Reveals Protein Inclusions Associated with SUMOylation and Diverse Polyubiquitin Chains.” Molecular & Cellular Proteomics 9 (4): 705–18. doi:10.1074/mcp.M800390-MCP200.

Soragni, Alice, Deanna M. Janzen, Lisa M. Johnson, Anne G. Lindgren, Anh Thai-Quynh Nguyen, Ekaterina Tiourin, Angela B. Soriaga, et al. 2016. “A Designed Inhibitor of p53 Aggregation Rescues p53 Tumor Suppression in Ovarian Carcinomas.” Cancer Cell 29 (1): 90–103. doi:10.1016/j.ccell.2015.12.002.

Sreedharan, Jemeen, Ian P Blair, Vineeta B Tripathi, Xun Hu, Caroline Vance, Boris Rogelj, Steven Ackerley, et al. 2008. “TDP-43 Mutations in Familial and Sporadic Amyotrophic Lateral Sclerosis.” Science 319 (5870): 1668–72. doi:10.1126/science.1154584.

Sun, Yulong, Pharhad Eli Arslan, Amy Won, Christopher M Yip, and Avijit Chakrabartty. 2014. “Binding of TDP-43 to the 3’UTR of Its Cognate mRNA Enhances Its Solubility.” Biochemistry 53 (37): 5885–94. doi:10.1021/bi500617x.

Swamy, M. S., and E. C. Abraham. 1987. “Lens Protein Composition, Glycation and High Molecular Weight Aggregation in Aging Rats.” Investigative Ophthalmology and Visual Science 28 (10): 1693–1701.

Tamaoka, Akira, Makoto Arai, Masanari Itokawa, Tetsuaki Arai, Masato Hasegawa, Kuniaki Tsuchiya, Hiroshi Takuma, et al. 2010. “TDP-43 M337V Mutation in Familial Amyotrophic Lateral Sclerosis in Japan.” Internal Medicine 49 (4): 331–34.

Tandan, R, and W G Bradley. 1985. “Amyotrophic Lateral Sclerosis: Part 1. Clinical Features, Pathology, and Ethical Issues in Management.” Annals of Neurology 18 (3): 271–80. doi:10.1002/ana.410180302.

Ticozzi, Nicola, Ashley Lyn LeClerc, Marka van Blitterswijk, Pamela Keagle, Diane M McKenna- Yasek, Peter C Sapp, Vincenzo Silani, Anne-Marie Wills, Robert H Brown, and John E Landers. 2011. “Mutational Analysis of TARDBP in Neurodegenerative Diseases.” Neurobiology of Aging 32 (11): 2096–99. doi:10.1016/j.neurobiolaging.2009.11.018.

Tollervey, James R JR, Tomaž Curk, Boris Rogelj, Michael Briese, Matteo Cereda, J Ule, James R JR Tollervey, Tomaž Curk, and Boris Rogelj. 2011. “Characterising the RNA Targets and Position- Dependent Splicing Regulation by TDP-43; Implications for Neurodegenerative Diseases.” Nature Neuroscience 14 (4): 452–58. doi:10.1038/nn.2778.

Tsai, Ching-Paio, Bing-Wen Soong, Kon-Ping Lin, Pang-Hsien Tu, Jer-Li Lin, and Yi-Chung Lee. 2011. “FUS, TARDBP, and SOD1 Mutations in a Taiwanese Cohort with Familial ALS.” Neurobiology of Aging 32 (3): 553.e13-21. doi:10.1016/j.neurobiolaging.2010.04.009.

Tuite, Mick F, Gemma L Staniforth, and Brian S Cox. 2015. “[PSI(+)] Turns 50.” Prion 9 (5): 318– 32. doi:10.1080/19336896.2015.1111508.

Udan-Johns, Maria, Rocio Bengoechea, Shaughn Bell, Jieya Shao, Marc I. Diamond, Heather L. True, Conrad C. Weihl, and Robert H. Baloh. 2014. “Prion-like Nuclear Aggregation of TDP-43 during Heat Shock Is Regulated by HSP40/70 Chaperones.” Human Molecular Genetics 23 (1): 157–70. doi:10.1093/hmg/ddt408.

Van Blitterswijk, Marka, Michael A. Van Es, Eric A M Hennekam, Dennis Dooijes, Wouter Van Rheenen, Jelena Medic, Pierre R. Bourque, et al. 2012. “Evidence for an Oligogenic Basis of Amyotrophic Lateral Sclerosis.” Human Molecular Genetics 21 (17): 3776–84. doi:10.1093/hmg/dds199.

Van Deerlin, Vivianna M, James B Leverenz, Lynn M Bekris, Thomas D Bird, Wuxing Yuan, Lauren B Elman, Dana Clay, et al. 2008. “TARDBP Mutations in Amyotrophic Lateral Sclerosis with TDP-43 Neuropathology: A Genetic and Histopathological Analysis.” The Lancet. Neurology 7 (5): 409–16. doi:10.1016/S1474-4422(08)70071-1.

Vance, Caroline, Boris Rogelj, Tibor Hortobágyi, Kurt J De Vos, Agnes Lumi Nishimura, Jemeen Sreedharan, Xun Hu, et al. 2009. “Mutations in FUS, an RNA Processing Protein, Cause Familial Amyotrophic Lateral Sclerosis Type 6.” Science 323 (5918): 1208–11. doi:10.1126/science.1165942.

Wang, Hurng-Yi, I-Fan Wang, Jayaramakrishnan Bose, and C.-K. James Shen. 2004. “Structural Diversity and Functional Implications of the Eukaryotic TDP Gene Family.” Genomics 83 (1): 130– 39. doi:10.1016/S0888-7543(03)00214-3.

Wang, I-Fan, Hsiang-Yu Chang, Shin-Chen Hou, Gunn-Guang Liou, Tzong-Der Way, and C-K James Shen. 2012. “The Self-Interaction of Native TDP-43 C Terminus Inhibits Its Degradation

43 and Contributes to Early Proteinopathies.” Nature Communications 3 (3): 766. doi:10.1038/ncomms1766.

Wang, Yi Ting, Pan Hsien Kuo, Chien Hao Chiang, Jhe Ruei Liang, Yun Ru Chen, Shuying Wang, James C K Shen, and Hanna S. Yuan. 2013. “The Truncated C-Terminal RNA Recognition Motif of TDP-43 Protein Plays a Key Role in Forming Proteinaceous Aggregates.” Journal of Biological Chemistry 288 (13): 9049–57. doi:10.1074/jbc.M112.438564.

Weishaupt, Jochen H, Tony Hyman, and Ivan Dikic. 2016. “Common Molecular Pathways in Amyotrophic Lateral Sclerosis and Frontotemporal Dementia.” Trends in Molecular Medicine 22 (9):769-783. doi:10.1016/j.molmed.2016.07.005.

Wickner, R. B., H. K. Edskes, A. Gorkovskiy, E. E. Bezsonov, and E. E. Stroobant. 2016. Yeast and Fungal Prions: Amyloid-Handling Systems, Amyloid Structure, and Prion Biology. Advances in Genetics. Vol. 93. doi:10.1016/bs.adgen.2015.12.003.

Wickner, Reed B., Herman K. Edskes, Dmitry Kryndushkin, Ryan McGlinchey, David Bateman, and Amy Kelly. 2011. “Prion Diseases of Yeast: Amyloid Structure and Biology.” Seminars in Cell and Developmental Biology 22 (5): 469–75. doi:10.1016/j.semcdb.2011.02.021.

Williams, K L, J C Durnall, A D Thoeng, S T Warraich, G A Nicholson, and I P Blair. 2009. “A Novel TARDBP Mutation in an Australian Amyotrophic Lateral Sclerosis Kindred.” Journal of Neurology, Neurosurgery, and Psychiatry 80 (11): 1286–88. doi:10.1136/jnnp.2008.163261.

Winton, Matthew J, Vivianna M Van Deerlin, Linda K Kwong, Wuxing Yuan, Elisabeth McCarty Wood, Chang-En Yu, Gerard D Schellenberg, et al. 2008. “A90V TDP-43 Variant Results in the Aberrant Localization of TDP-43 in Vitro.” FEBS Letters 582 (15): 2252–56. doi:10.1016/j.febslet.2008.05.024.

Wolozin, Benjamin. 2012. “Regulated Protein Aggregation: Stress Granules and Neurodegeneration.” Molecular Neurodegeneration 7 (1): 56. doi:10.1186/1750-1326-7-56.

Wu, Lien-Szu, Wei-Cheng Cheng, Shin-Chen Hou, Yu-Ting Yan, Si-Tse Jiang, and C-K James Shen. 2010. “TDP-43, a Neuro-Pathosignature Factor, Is Essential for Early Mouse Embryogenesis.” Genesis 48 (1): 56–62. doi:10.1002/dvg.20584.

Xiang, Siheng, Masato Kato, Leeju C. Wu, Yi Lin, Ming Ding, Yajie Zhang, Yonghao Yu, and Steven L. McKnight. 2015. “The LC Domain of hnRNPA2 Adopts Similar Conformations in Hydrogel Polymers, Liquid-like Droplets, and Nuclei.” Cell 163 (4): 829–39. doi:10.1016/j.cell.2015.10.040.

Xiao, Shangxi, Teresa Sanelli, Helen Chiang, Yulong Sun, Avijit Chakrabartty, Julia Keith, Ekaterina Rogaeva, Lorne Zinman, and Janice Robertson. 2015. “Low Molecular Weight Species of TDP-43 Generated by Abnormal Splicing Form Inclusions in Amyotrophic Lateral Sclerosis and Result in Motor Neuron Death.” Acta Neuropathologica 130 (1): 49–61. doi:10.1007/s00401-015-1412-5.

Xiao, Shangxi, Teresa Sanelli, Samar Dib, David Sheps, Joseph Findlater, Juan Bilbao, Julia Keith, Lorne Zinman, Ekaterina Rogaeva, and Janice Robertson. 2011. “RNA Targets of TDP-43 Identified by UV-CLIP Are Deregulated in ALS.” Molecular and Cellular Neurosciences 47 (3): 167–80. doi:10.1016/j.mcn.2011.02.013.

Xiong, Hui-Ling, Jin-Yang Wang, Yi-Min Sun, Jian-Jun Wu, Yan Chen, Kai Qiao, Qiao-Juan Zheng, Gui-Xian Zhao, and Zhi-Ying Wu. 2010. “Association between Novel TARDBP Mutations and Chinese Patients with Amyotrophic Lateral Sclerosis.” BMC Medical Genetics 11 (January): 8. doi:10.1186/1471-2350-11-8.

Yokoseki, Akio, Atsushi Shiga, Chun-Feng Tan, Asako Tagawa, Hiroyuki Kaneko, Akihide Koyama, Hiroto Eguchi, et al. 2008. “TDP-43 Mutation in Familial Amyotrophic Lateral Sclerosis.” Annals of Neurology 63 (4): 538–42. doi:10.1002/ana.21392.

Zeineddine, Rafaa, Jay F Pundavela, Lisa Corcoran, Elise M Stewart, Dzung Do-Ha, Monique Bax, Gilles Guillemin, et al. 2015. “SOD1 Protein Aggregates Stimulate Macropinocytosis in Neurons to Facilitate Their Propagation.” Molecular Neurodegeneration 10 (1): 57. doi:10.1186/s13024-015-0053-4.

Zhang, Huaiying, Shana Elbaum-Garfinkle, Erin M. Langdon, Nicole Taylor, Patricia Occhipinti, Andrew A. Bridges, Clifford P. Brangwynne, and Amy S. Gladfelter. 2015. “RNA Controls PolyQ Protein Phase Transitions.” Molecular Cell 60 (2): 220–30. doi:10.1016/j.molcel.2015.09.017.

Zhang, Tao, Patrick C Mullane, Goran Periz, and Jiou Wang. 2011. “TDP-43 Neurotoxicity and Protein Aggregation Modulated by Heat Shock Factor and insulin/IGF-1 Signaling.” Human Molecular Genetics 20 (10): 1952–65. doi:10.1093/hmg/ddr076.

Zhang, Yong-Jie, Ya-Fei Xu, Casey Cook, Tania F Gendron, Paul Roettges, Christopher D Link, Wen-Lang Lin, et al. 2009. “Aberrant Cleavage of TDP-43 Enhances Aggregation and Cellular Toxicity.” Proceedings of the National Academy of Sciences 106 (18): 7607–12. doi:10.1073/pnas.0900688106.

Zhang, Yong-Jie, Ya-fei Xu, Chad a Dickey, Emanuele Buratti, Francisco Baralle, Rachel Bailey, Stuart Pickering-Brown, Dennis Dickson, and Leonard Petrucelli. 2007. “Progranulin Mediates Caspase-Dependent Cleavage of TAR DNA Binding Protein-43.” The Journal of Neuroscience 27 (39): 10530–34. doi:10.1523/JNEUROSCI.3421-07.2007.

Zhang, Yong Jie, Thomas Caulfield, Ya Fei Xu, Tania F. Gendron, Jaime Hubbard, Caroline Stetler, Hiroki Sasaguri, et al. 2013. “The Dual Functions of the Extreme N-Terminus of TDP-43 in Regulating Its Biological Activity and Inclusion Formation.” Human Molecular Genetics 22 (15): 3112– 22. doi:10.1093/hmg/ddt166.

Zou, Zhang-Yu, Yu Peng, Xin-Ning Wang, Ming-Sheng Liu, Xiao-Guang Li, and Li-Ying Cui. 2012. “Screening of the TARDBP Gene in Familial and Sporadic Amyotrophic Lateral Sclerosis Patients of Chinese Origin.” Neurobiology of Aging 33 (9): 2229.e11-2229.e18. doi:10.1016/j.neurobiolaging.2012.03.014.

CHAPTER II BINDING OF TDP-43 TO THE 3’UTR OF ITS COGNATE MRNA ENHANCES ITS SOLUBILITY

This chapter first appeared in Biochemistry as: Y. Sun, P.E. Arslan, A. Won, C.M. Yip, and A. Chakrabartty. (2014). Binding of TDP-43 to the 3’UTR of Its Cognate mRNA Enhances Its Solubility. Biochemistry 53(37):5885–5894. It was written by Y.S with input from A.C. Most experimental work was carried out by Y.S., with some molecular cloning carried out by P.E.A. and atomic force microscopy carried out by A.W. under the supervision of C.M.Y. Some minor revisions have been made to this chapter from the original article and a note has been added to the discussion with regards to the reversibility of TDP-43 aggregation.

45 46

Chapter Abstract

TAR DNA binding protein of 43 kDa (TDP-43) has been implicated in the pathogenesis of a broad range of neurodegenerative diseases termed TDP-43 proteinopathies, which encompass a spectrum of diseases ranging from amyotrophic lateral sclerosis to frontotemporal dementia. Pathologically misfolded and aggregated forms of TDP-43 are found in cytoplasmic inclusion bodies of affected neurons in these diseases. The mechanism by which TDP-43 misfolding causes disease is not well understood. Current hypotheses postulate that TDP-43 aggregation process plays a major role in pathogenesis. We amplify that hypothesis and suggest that binding of cognate ligands to TDP-43 can stabilize the native functional state of the protein and ameliorate aggregation. We expressed recombinant TDP-43 containing an N-terminal Venus yellow fluorescent protein tag in E. coli and induced its aggregation by altering solvent salt concentrations and examined the extent to which various oligonucleotide molecules affects its aggregation in vitro using aggregation-induced turbidity assays. We show that vYFP-TDP-43 binding to its naturally occurring RNA target that comprises of a sequence on the 3’UTR region of its mRNA improves its solubility, suggesting interplay between TDP-43 solubility, oligonucleotide binding, and TDP-43 autoregulation.

Abbreviations used in this chapter

TDP-43, TAR-DNA binding protein of 43 kDa; TAR, Transactive response; ALS, amyotrophic lateral sclerosis; RRM, RNA recognition motif; SG, stress granules; TTR, Transthyretin; DLS, dynamic light scattering; HIV1, human immunodeficiency virus-1; LTR, long terminal repeat; RALS, right angle light scattering; SOD1, superoxide dismutase 1; NMD, nonsense mediated decay; UV-CLIP, UV crosslinking immunoprecipitation; FUS/TLS, Fused in Sarcoma/Translocated in Sarcoma; SDS- PAGE, sodium dodecyl sulfate polyacrylamide gel electrophoresis.

Introduction

A common hallmark of neurodegenerative disease is the aggregation of misfolded proteins inside affected neurons (Schenk et al. 1999; Polymeropoulos 1997; Majoor-Krakauer, Willems, and Hofman 2003). Recent studies have identified TDP-43 as a major component of cytoplasmic aggregates within neurons of patients with amyotrophic lateral sclerosis (ALS; Neumann et al. 2006; Arai et al. 2006) and mutations in the TDP-43 gene (TARDBP) are known to be associated with familial ALS (Kabashi et al. 2008). TDP-43 has since been implicated in a wide range of neurodegenerative diseases that run the gamut from ALS to frontotemporal lobar degeneration (FTLD), which are now coined as TDP- 43 proteinopathies (Lagier-Tourenne, Polymenidou, and Cleveland 2010).

TDP-43 is a 414-amino acid nuclear protein composed of two highly conserved RNA recognition motifs (RRM1 and RRM2) and a C-terminal region (Figure S2.1A). These RRM are involved in binding of RNA/DNA sequences enriched in UG or TG repeats (Buratti and Baralle 2001; Buratti et al. 2004). Additionally, RRM2 mediates dimerization of the protein (Kuo et al. 2009). The C-terminal region is thought to mediate protein-protein interactions and contains yeast prion-like motifs implicated in disease pathology, and this region contains nearly all locations of disease-implicated mutations (Gitler and Shorter 2011; Budini et al. 2012; Lagier-Tourenne, Polymenidou, and Cleveland 2010; Buratti et al. 2005).

Under pathological conditions, TDP-43 is found in cytosolic inclusion bodies, where it is hyperphosphorylated, ubiquitinated and processed into 25 and 35 kDa C-terminal fragments (Neumann et al. 2006; Arai et al. 2006). Expression or introduction of these fragments in cell culture can recapitulate certain pathological features of TDP-43 proteinopathies by sequestration of wild type TDP-43 from the nucleus and inducement of cell death through toxic gain of function (Igaz et al. 2009; Y.-J. Zhang et al. 2009; Nonaka et al. 2013). One possible mechanism for TDP-43 misfolding involves the misfolding of the C-terminal domain into an aberrant structure that may act as a template for recruitment of other, native TDP-43 molecules into this misfolded aggregate. The molecular mechanism of this conversion process is still a matter of debate. Cytoplasmic localization and recruitment into stress granules (SGs) have been proposed as factors contributing to the initiation of TDP-43 aggregation (Barmada et al. 2010; Giordana et al. 2010; McDonald et al. 2011; Parker et al. 2012). Recent findings also suggest that RNA/DNA binding modulates TDP-43 solubility (Pesiridis et al. 2011; Huang et al. 2013).

TDP-43 conducts a variety of RNA processing functions in the cell, such as transcription, RNA regulation, micro-RNA processing, nucleo-cytoplasmic mRNA shuttling and association with stress granules (Ayala et al. 2005; Ou et al. 1995; Buratti and Baralle 2001; Parker et al. 2012; McDonald et al. 2011). Because of its many functions, TDP-43 levels are tightly regulated. A proposed mechanism of regulation is TDP-43 autoregulation by binding to the 3’UTR of its mRNA, leading to nonsense mediated decay (NMD)-independent mRNA degradation and decreases in the level of TDP-43 production (Ayala et al. 2011; Budini and Buratti 2011; Bembich et al. 2014).

To explore the possibility that nucleotide binding might regulate TDP-43 solubility, as well, we investigated the possibility that RNA/DNA binding prevents aggregation of TDP-43 and examined whether naturally occurring sequences, such as the autoregulatory binding region of the 3’UTR of TDP-43’s mRNA, can modulate TDP-43 solubility. We hypothesized that binding of TDP-43 to its natural nucleotide ligands through its RRMs maintains TDP-43 in its soluble functional state, and the loss of this interaction in pathological situations may be an initiating factor for TDP-43 aggregation into inclusion bodies.

Using both natural ligands of TDP-43 and artificially constructed de novo sequences we assess the effect of these compounds on TDP-43 aggregation and discuss the interplay between TDP-43 autoregulation and solubility.

Results Recombinant vYFP-TDP-43 is natively dimeric

TDP-43 is an intrinsically aggregation-prone protein (Johnson et al. 2009). To facilitate purification, enhance solubility, and act as a fluorescent probe, we attached an N-terminal Venus yellow fluorescent protein (vYFP) tag, which is a derivative of green fluorescent protein with mutations to increase its folding rate and brightness (Nagai et al. 2002; Arslan and Chakrabartty 2009). The construct was expressed in E. coli using buffers adapted from previous studies (Johnson et al. 2009).

The quality of the expressed protein was assessed using a variety of assays. Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) of eluted samples confirms the presence of the single band at 76 kDa, corresponding to the expected size of the recombinant vYFP-TDP-43 (Figure 2.1A). The sample was also subjected to dynamic light scattering (DLS) and size exclusion chromatography (SEC) measurements to assess particle size under purification conditions. DLS results indicate a single species of hydrodynamic radius of 4.86 nm, which constitutes 95% of the sample by mass (Figure 2.1B). Conversion of radius to molecular weight yields a weight of 136 kDa, consistent with dimer configuration, which was also observed in recent studies (Kuo et al. 2009). SEC measurements (inset) confirms the presence of a single peak at approximately 158 kDa. Circular dichroism spectroscopy on the eluted protein produced a spectrum indicative of β-structure with a characteristic β-structure signature at 218 nm (Figure 2.1C; Greenfield 2007). This was expected, as vYFP contains a β-barrel structure and contributes to this CD signal (Rekas et al. 2002). The absence of significant α-helix and random coil signals also indicate that TDP-43 is mostly β-structured, which agrees with recent crystallographic data on its RRM2 domain (Kuo et al. 2009). Urea denaturation of the sample monitored by Trp fluorescence and vYFP fluorescence shows cooperative unfolding of the entire construct with a denaturation midpoint of 4.2 M urea. The vYFP fluorescence is maintained at 100% at this urea concentration, suggesting that the observed unfolding curve is representative of the TDP- 43 segment of the fusion protein (Figure 2.1D). The cooperative unfolding of the protein suggests that TDP-43 segment of the recombinant protein is also folded. Collectively, these assays indicate that the protein produced was 95% pure, dimeric, and folded, and that the vYFP does not interfere with these intrinsic properties of TDP-43 under our experimental conditions.

Figure 2.1: Characterization of vYFP-TDP-43 using various biochemical techniques. (A) SDS- PAGE of purified vYFP-TDP-43. Lanes containing crude cell lysate after sonication (lysate), soluble fraction of cell lysate (sup), fraction collected after washing with lysis buffer (wash), and final eluted fractions (eluate) are shown. Approximately 10 μg of samples were applied to a 12% polyacrylamide gel for electrophoresis. The band corresponding to the size of vYFP-TDP-43 is indicated by the arrow. (B) Size distribution by mass of purified vYFP-TDP-43. A protein concentration of 20 μM was used. Measurements of Rh and MW were calculated using appropriate software (Materials and Methods). The size exclusion chromatography peak of purified protein measured by vYFP absorbance and its estimated molecular weight is presented as an inset. (C) Circular dichroism spectrum of vYFP-TDP- 43. Circular dichroism was measured at 25 ºC with 16 s averaging times. The signal for 8.8 µM vYFP- TDP-43 is shown with characteristic β-signal. (D) Urea denaturation of vYFP-TDP-43. The integrated Trp fluorescence from 315 nm to 335 nm using excitation wavelength of 283 nm was measured for samples containing 2.1 µM vYFP-TDP-43 incubated with 0.0 – 7.2 M urea for 12 h. Trp fluorescence was normalized and converted to % folded (●, left axis). vYFP fluorescence of the sample was measured by excitation at 515 nm and emission at 528 nm and normalized (○, right axis).

TG12 inhibits TDP-43 aggregation at sub-stoichiometric concentrations by maintaining dimer configuration

Recent studies have shown that poly-TG compounds can increase solubility of refolded recombinant TDP-43 under aggregation conditions induced by temperature (Huang et al. 2013). Here we investigated the concentration dependent effect of aggregation inhibition by ssDNA consisting of 12

poly TG repeats (TG12) through turbidity measurements. Aggregation of vYFP-TDP-43 was reliably and reproducibly induced by ionic strength reduction using methods adapted from previously established works (Johnson et al. 2009). vYFP-TDP-43 (2 µM) was placed under aggregation

conditions in the presence or absence of 5-fold molar excess of TG12, and the sizes of the particles were monitored using DLS and right angle light scattering (RALS). In the absence of nucleotides, incubation under aggregation conditions for 4 h at room temperature resulted in a majority of aggregates having hydrodynamic radii of 105.6 nm, with a significant secondary population with radii up to 500-700 nm as measured by DLS. The estimated molecular mass of the average particle is

approximately 18300 kDa (Figure 2.2A). In contrast, in the presence of 10 µM TG12 ssDNA, the average particle had a hydrodynamic radius of 5.34 nm, corresponding to a molecular mass of 170 kDa, close to the theoretical size of the dimer (154 kDa) with two strands of TG12 (7.6 kDa) bound (Figure 2.2B). These findings were further validated by measurement of solution turbidity using RALS

on samples incubated with or without 5-fold molar excess of TG12 for 4 h. The level of scattering of

the samples was significantly reduced by the presence of TG12 compound (Figure 2.2C). When the soluble fractions of these samples were analyzed by DLS after centrifugation, untreated samples contained species corresponding to monomeric vYFP-TDP-43 (76 kDa) and soluble oligomers

(~5000 kDa), while samples treated with TG12 remained dimeric (Figure 2.2D). These results indicate

that TG12 is an effective inhibitor of aggregation under these experimental conditions, and that inhibition of aggregation occurs by preservation of the protein’s native dimeric state and prevention of a monomeric state.

Figure 2.2: Size distribution by mass of vYFP-TDP-43 upon aggregation determined by dynamic light scattering and sample turbidity measured by right angle light scattering. (A) vYFP-TDP-43 (2 µM) was placed under aggregation conditions for 4 h at 20 ºC. (B) The same conditions were applied to the sample in the presence of 10 µM TG12. Measurements of Rh and MW were taken using appropriate software (see Methods). (C) Turbidity of samples under condition A and B were measured using right angle light scattering at 400 nm. (D) DLS of the soluble fraction of samples under condition A and B. Identified peak sizes are indicated.

To quantify the potency of aggregation inhibition, a concentration dependency assay was conducted

using TG12 ssDNA monitored by RALS. Under the same conditions that were used for the DLS assays, vYFP-TDP-43 was induced to aggregate in the presence varying concentrations of TG12 ssDNA (0-2 μM) and AC12 ssDNA as negative control (0-10 μM) (Figure 2.3). A concentration

dependence of inhibition was observed with the TG12 ssDNA and the data were fit using non linear

least-squares fitting to the equation (Equation 2.1), where y is the observed normalized scattering signal, x is the compound concentration, and n is the Hill coefficient. An

effective concentration (EC50) of 134 nM TG12 was observed for the aggregation inhibition of a sample containing 2 μM vYFP-TDP-43 (n = 1.76). The maximum effect of inhibition appears to have been

achieved at 1 μM TG12, a substoichiometric 1:2 molar ratio of compound and recombinant protein.

For samples treated with AC12, no effect was observed even at 5-fold molar excess of oligonucleotide.

Figure 2.3: Inhibition of vYFP- TDP-43 aggregation using TG12 monitored by right angle light scattering. vYFP-TDP-43 (2 µM) was incubated with varying concentrations of TG12 (●) and AC12 (○) ssDNA under aggregation conditions for 4 h. Turbidity was determined by right angle light scattering at 400 nm. The inhibition curve of TG12 was fitted by non linear least-squares fitting using Equation 2.1 and the EC50 and n values are listed.

Naturally occurring nucleotide targets reduce the level of TDP-43 aggregation

Although TG12 repeats show strong binding to vYFP-TDP-43, the majority of targets of TDP-43 identified by UV cross-linking immunoprecipitation (UV-CLIP) assays do not have long stretches of consecutive poly-TG and poly-UG residues (Xiao et al. 2011; Tollervey et al. 2011). To assess whether oligonucleotides already present in the cell are capable of inhibiting TDP-43 aggregation, biologically relevant targets of TDP-43 are examined as potential aggregation inhibitors. The first target we selected is the transactive response (TAR) element of human immunodeficiency virus-1 (HIV-1), the first binding target of TDP-43 through which the protein was discovered (Ou et al. 1995). TDP-43 binds to a sequence on the viral element on the long terminal repeat (LTR) sequence of the integrated viral genome, responsible for the TAR element. Although not normally found in healthy cells, this binding sequence holds historical significance and also contains two pyrimidine regions, which are consensus sequences for TDP-43 binding (Xiao et al. 2011). TDP-43 was not only found to bind the double-stranded TAR cDNA, but also has minor affinity for the coding single-strand sequence (Ou et al. 1995). We decided to assess the effect of both the double-stranded and single-stranded HIV1 LTR on TDP-43 aggregation. Another sequence we selected is a stretch of a guanine-rich sequence located at the 3’UTR of TDP-43 mRNA discovered by UV-CLIP, denoted by CLIP34nt (Bhardwaj et al. 2013). We chose this sequence as a candidate RNA sequence for inhibiting TDP-43 aggregation because it is involved in TDP-43 autoregulation. The potential of TDP-43’s autoregulation to be linked to its solubility can be of particular interest in its role in disease.

Using TG12 and AC12 as positive and negative controls, respectively, the aggregation inhibition of vYFP-TDP-43 by CLIP34nt RNA, HIV1LTR ssDNA, and HIV1LTR dsDNA was examined using

RALS using methods and conditions described above (Figure 2.4A). Calculated EC50 and n values (Figure 2.4B) for CLIP34nt RNA and HIV1LTR ssDNA are approximately 2 μM and ~0.5

respectively. Although the potency of inhibition is less than that of the TG12 control (205 nM), they demonstrate that binding of natural ligands of TDP-43 has an inhibitory effect to the aggregation of this protein. The HIV1LTR dsDNA did not have an effect on the aggregation of vYFP-TDP-43 despite its reported ability to bind to this protein determined by previous studies (Ou et al. 1995).

Aggregation inhibition is achieved through RRM1 binding

Previous studies have demonstrated that TG12 fails to inhibit TDP-43 aggregation in the absence of functional RRM1 domain, where phenylalanine resides critical for DNA/RNA binding are mutated (F147L and F149L mutations; Huang et al. 2013). Truncations of RRM1 also demonstrated the same effect. Additionally, F147L and F149L mutations also disrupted TDP-43 autoregulation (Ayala et al. 2011). We thus suspect aggregation inhibition using naturally occurring compounds is also mediated through binding to RRM1. When the right angle light scattering experiments are repeated on DNA/RNA binding-deficient mutants of TDP-43, we observe no change in scattering upon addition of TG12 or CLIP34nt oligonucleotides (Figure 2.4C) compared to the wild-type (wt) protein control. This confirms that aggregation inhibition is accomplished through the binding interaction between a specific oligonucleotide target and the RRM1 domain of TDP-43.

Figure 2.4: Inhibition of wild-type (wt) and F147L/F149L mutant vYFP-TDP-43 aggregation using various natural oligonucleotide binding targets. (A) Protein (2 µM) was incubated under aggregating conditions with varying concentrations of different nucleotides for 4 h. Turbidity was determined by right angle light scattering at 400 nm. (B) Calculated EC50 values, names, and sequences of nucleotides used for native-state stabilization. EC50 and n values were obtained by non linear least-squares fitting using Equation 2.1. (C) Aggregation assays were performed on 2 µM F147L/F149L mutant protein using TG12, AC12, and CLIP34nt and results compared to those for wt protein treated with TG12 as control.

Effect of oligonucleotides on pre-formed vYFP-TDP-43 aggregates

Our data suggest that by binding to the native, dimeric state of vYFP-TDP-43, native-state binding molecules can prevent the protein from entering the aggregation pathway. The following experiments assess whether these compounds have disaggregation properties once the protein has already entered the aggregated state. These results serve a secondary purpose of providing visual validation to the observations in RALS measurements. Samples were observed under a fluorescence microscope after incubation for 20 min under aggregating conditions in the presence or absence of aggregation- inhibiting nucleotides. By visual inspection, samples induced to aggregate in the presence of TG12, showed largely diffuse distributions, with few small particulates (Figure 2.5). Untreated samples and those treated with AC12 exhibited large aggregates of up to 100 µm in diameter. This result is in accordance with the RALS assays previously performed. When samples were subjected to aggregation for 15 min, followed by addition of various ssDNA molecules for 5 more min, no changes in aggregation were observed (Figure 2.5). These samples remain unchanged after 24 h and all samples contained aggregates of similar morphology. These results indicate that although native state binding molecules can reduce the propensity for entry into the aggregated state, pre-formed aggregates cannot be dissolved by the addition of these compounds.

Figure 2.5: Fluorescence microscopy of vYFP-TDP- 43 aggregates. Images were taken at 20 ºC targeting the vYFP tag (λex = 515 nm λem = 528 nm) using 10× objective and 40× lens. vYFP-TDP-43 (2 µM) was incubated for 20 min under aggregating conditions in the presence of various ssDNA (upper row). Various ssDNAs were added to pre-formed aggregates incubating for 15 min and images were taken 5 min after adding ssDNA (lower row). Scale bar = 100 µm.

Morphology of TDP-43 aggregates

Solutions of vYFP-TDP-43 incubated for 20 min at room temperature under aggregating conditions were imaged using AFM with a glass tapping fluid cell to assess overall aggregate morphology on a mica surface. The resulting images show formations of small clusters of globular structures ranging from 10 to 100 nm in diameter (Figure 2.6). The aggregates are non-fibrillar and show no apparent order.

Figure 2.6: Tapping mode atomic force microscopy images of non-fibrillar vYFP-TDP-43 aggregates. Images were acquired in solution consisting of 2 μm vYFP-TDP-43, 170 mM KCl, 36 mM HEPES, 25 mM Imidazole, 18 mM MgCl2, 1% glycerol, 1.8 mM βME, pH 7.4 by tapping mode AFM after deposition onto a mica surface. Two image sizes are provided: 5 µm × 5 µm (top), 2.5 µm × 2.5 µm (bottom), on a 0.5 – 5 µm height scale. Boxes on both panels indicate the same imaged region using different scales. The bottom panel is shown an with increased contrast, and the height scale is the same in both panels.

Discussion

TDP-43 aggregation has been implicated as a key step in ALS and FTD pathogenesis, but the mechanism of aggregation remains enigmatic. Several factors have been proposed to influence TDP- 43 aggregation; including oxidative stress, heat shock, stress granule formation, cleavage into C- terminal fragments, and loss of DNA/RNA targets (T. Zhang et al. 2011; Parker et al. 2012; Furukawa, Kaneko, and Nukina 2011; Furukawa et al. 2011; Pesiridis et al. 2011; Udan-Johns et al. 2014; Liachko, Guthrie, and Kraemer 2010). Binding of TDP-43 to the artificial sequence of poly-TG repeats improve its solubility, and C-terminal fragments of TDP-43 containing RRM2 and C-terminal domains remain soluble in cell culture unless treated with RNAse 1 (Huang et al. 2013; Pesiridis et al. 2011). TDP-43, as an intrinsically aggregation-prone protein, remains soluble in the nucleus even when overexpressed in cell culture (Huang et al. 2013; Johnson et al. 2009). This suggests a crucial role of RNA in TDP- 43 solubility. Our findings support the hypothesis that TDP-43 undergoes a conformational change from its native state to an aggregation-prone state triggered by the loss of its native binding partners, priming it for environmental factors that may lead to its aggregation.

Insights into TDP-43 misfolding and inhibition mechanisms

Our findings confirm previous work on TDP-43 aggregation attenuation through TG12 binding (Huang et al. 2013). In our studies, we measured aggregation using DLS and RALS with recombinant vYFP-TDP-43 that is natively β-structure, folded, and dimeric. Via comparison of the size distribution of aggregates when induced in the presence and absence of 5-fold molar excess of TG12, it is apparent that in the absence of TG12, vYFP-TDP-43 formed large oligomeric species of hundreds of

nanometers in size (Figure 2.2B). When the sample was treated with TG12, 95% of the sample

remained dimeric (Figure 2.2A), suggesting that the presence of TG12 was able to prevent vYFP-TDP- 43 from entering any downstream aggregation-prone pathways. In many cases of protein misfolding such as transthyretin (TTR) and superoxide dismutase 1 (SOD1), the primary step in the misfolding pathway usually involves disassociation of the native oligomer into an aggregation-prone monomer (Miyata et al. 2010; Mulligan et al. 2008; Ip, Mulligan, and Chakrabartty 2011). We have found evidence to suggest that this is also a likely mechanism of TDP-43 aggregation observed under our experimental conditions because of the presence of monomeric and soluble oligomeric TDP-43 in the soluble fraction of samples under aggregating conditions (Figure 2.2D). Our data suggest that maintenance of the native dimer through DNA binding prevents the downstream aggregation processes, which likely involves an initial monomerization event. We observed the same inhibitory effect using RALS and

fluorescence microscopy assays, consistent with our initial findings (Figure 2.2C, Figure 2.5). The fluorescence assay also showed that pre-formed aggregates could not be rescued by the presence of

TG12 compound, consistent with the idea that the aggregation pathway leads to an apparently irreversible end-product, which is a common hallmark in non-native aggregation of proteins detailed in various reviews (Morris, Watzky, and Finke 2009; Roberts 2007). Morphologically, we found that the aggregates, like those observed in ALS neurons, did not share characteristic fibrillar structure of prion aggregates (Figure 2.6; Neumann et al. 2006; Nonaka et al. 2009). However, because of the absence of structural data on TDP-43’s C-terminus under native or aggregating conditions, we cannot confirm that the aggregates observed in our in vitro assays are identical in morphology or behavior to those in disease neurons. Whether oligonucleotides can reverse aggregates in vivo systems remains to

be tested. We generated an inhibition curve by altering the TG12 concentration while measuring

turbidity using RALS and calculated the EC50 value of this inhibition as 134 nM with a Hill coefficient of 1.75 (Figure 2.3). A coefficient >1 in the context of EC50 measurements suggests multiple binding

sites (Shoichet 2006). This is expected, as a single TG12 molecule should be able to bind at least two TDP-43 molecules, potentially serving as a tether to stabilize the native dimer. In fact, we observed

that a 1:2 of TG12-to-protein molar ratio was sufficient to achieve maximum inhibition, while AC12 had no effect. This substoichiometric effect specific to a binding partner in conjunction with the presence of exclusively dimeric species in our DLS results suggests that TG12 inhibits vYFP-TDP-43 aggregation by binding to the native dimer, acting as a chaperone to prevent aggregation initiation (Figure S2.1C). Furthermore, F147L and F149L mutations on RRM1 abolishes this inhibition effect, confirming that the specificity of oligonucleotide targets is achieved through a functional RRM1 domain (Figure 2.4C). There is a general consensus that the C-terminus of TDP-43 is a key player to this aggregation pathway (Igaz et al. 2009; Budini et al. 2012). It is a GQN-rich prion-like domain and has been shown to have prion-like properties (Budini et al. 2012; Gitler and Shorter 2011; Nonaka et al. 2013). Specifically, expression of 12 tandem repeats of the residues 331-369 of the C-terminus containing Q/N-rich sequences is sufficient to form aggregates in which native nuclear TDP-43 can be sequestered (Budini et al. 2012). Seeding of C-terminal fragments from diseased brains into cell culture can also recapitulate disease phenotypes (Nonaka et al. 2013). However, this proposed intrinsically disordered C-terminal region is not required for oligonucleotide interactions (Kuo et al. 2009), yet, we find that native-state binding molecules effectively reduce aggregation caused by this C- terminal region without physical interactions. This suggests that there are likely structural differences with the C-terminal region of TDP-43 when the protein is native dimer versus a non-native, aggregation-prone monomer, and that the retention of TDP-43 in its native state by the binding of

native-state binding compounds prevents the C-terminus from engaging in aggregation-prone interactions with C-termini of non-native TDP-43 in vitro. Recent studies show that the C-terminus is intrinsically disordered, but Q/N-rich segments (341-366) are able to spontaneously form β-rich species, through which aggregation may arise (Mompeán et al. 2014). Co-immunoprecipitation experiments using FLAG-TDP-43 in cells show that FLAG-TDP-43 coprecipitates with full length TDP-43, suggesting TDP-43 self-interactions. The coprecipitation efficiency was reduced drastically when the process was conducted on cells expressing TDP-43 with deletion of the C-terminus (Wang et al. 2012). This suggests that the C-terminus has a role in maintaining TDP-43 as a dimer in the native state, which is affected by cognate DNA/RNA binding. Factors that disrupt these properties of TDP-43, including ALS-causing mutations on the C-terminus or loss of cognate DNA/RNA partners (through mislocalization, cleavage, or RNA binding-deficient mutations), may render TDP- 43 more likely to adopt an aggregation-prone conformation on the C-terminus, increasing its propensity to aggregate.

Implications of TDP-43 aggregation inhibition by naturally occurring targets

TDP-43 has a variety of cellular roles, and the C-terminus has been implicated in many protein-protein interactions such as binding to heterogeneous nuclear ribonucleoproteins (hnRNPs) to participate in mRNA splicing, degradation, stabilization and other various functions (McDonald et al. 2011; Buratti et al. 2005). In cells where these protein-protein interactions are prevalent, many factors likely contribute to TDP-43 solubility. From our findings, cognate binding to its mRNA targets appears to be an additional factor to maintaining TDP-43 solubility under normal conditions. This aggregation inhibitory effect was observed for single-stranded DNA and RNA, but not the double stranded DNA target. We propose that the majority of TDP-43 in the nucleus is bound to a DNA/RNA target, which maintains its solubility in healthy cells. One of the binding partners readily available to newly synthesized TDP-43 is the 3’UTR of its own mRNA. TDP-43 levels in cells are regulated through a negative feedback loop in which TDP-43 binds to the 3’UTR of its own mRNA promoting mRNA instability (Ayala et al. 2011; Polymenidou and Lagier-Tourenne 2011). We find that binding of CLIP34nt to TDP-43 through RRM1 increases TDP-43 solubility (Figure 2.4). Binding of TDP-43 to its own mRNA may have a 2-fold effect: it may increase the solubility of the newly synthesized protein, while simultaneously targeting the mRNA for degradation. Such a finding suggests additional layers of complexity in the spatiotemporal regulation of TDP-43 that previously may not have been considered. Similar mechanisms of native-state stabilization via DNA binding to a tightly regulated protein have also been observed in the tumor suppressor p53, where a consensus DNA sequence

binding to p53’s core domain (p53C) can stabilize the full length protein and prevent amyloid-like aggregation (Ishimaru et al. 2009).

Disruption of interactions between TDP-43 and its own mRNA may promote newly synthesized TDP-43 to adopt a β-rich aggregation-prone fold at its C-terminus, while simultaneously increasing TDP-43 levels due to disruption of the autoregulation pathway, leading to increased concentrations of aggregation-prone TDP-43 species in the cell and making it vulnerable to environmental factors. This serves as the first “hit” in a “two-hit” hypothesis proposed by Persiridis and colleagues (Pesiridis et al. 2011). Cellular stresses such as oxidative damage and heat shock in the presence of the larger population of aggregation-prone species provide the second hit, leading to the potential recruitment of these proteins into SG, creating pathological aggregates and inclusion bodies as seen in ALS neurons. Whereas the C-terminus of TDP-43 normally allows for transient and reversible interactions in SG under non-pathological conditions, other factors proposed to affect TDP-43 aggregation such as C-terminal phosphorylation and cytoplasmic localization may also serve as initial steps to expose the protein to vulnerability by increasing the propensity of the C-terminus to adopt an aggregation- prone fold and effectively reduce the number of locally available native binding partners, respectively. The loss of functional TDP-43 through mis-regulation and increased number of aggregates in the cell may cause ALS-related pathology through both loss of TDP-43 function and gain of TDP-43 aggregate toxic function mechanisms. It is uncertain which mechanism is predominant in disease. Recent studies of TDP-43 mutants such as F147L/F149L in a number of model organisms show that the expression of exogenous wt TDP-43 can generate more neurotoxicity than expression of mutant TDP-43 (Estes et al. 2011; Voigt et al. 2010). We believe that in the context of in vivo systems, introduction of exogenous TDP-43 may be disruptive to the RNA binding activities of the endogenous protein. because of its myriad of roles in RNA processing, exogenous TDP-43 expression may replicate the effect of disrupted TDP-43 autoregulation. The inability of the F147L/F149L mutant to generate toxicity while still forming aggregates in these systems may suggest that depletion of native TDP-43 function (through mutations or aggregation) is a factor in generating neurotoxicity, and that neurotoxicity may not always be dependent on aggregation (Voigt et al. 2010).

The need for neurons to spatially and temporally regulate their transcription dictates the movement of mRNA/RNA-binding protein complexes across long neuronal processes. The long turnover time for these proteins and complexes may render these proteins more susceptible to oxidative damage or other environmental stressors, which may explain why disease phenotypes are specific to neurons and

62 no other cell types. Recently, many RNA binding proteins with intrinsically disordered regions such as Fused in Sarcoma/Translocated in Sarcoma (FUS/TLS) have also been implicated in neurodegenerative diseases and formation of SG, which may cause pathology via a mechanism similar to that of TDP-43 (Baron et al. 2013; Deng et al. 2010; Udan and Baloh 2011).

Herein we determined that TDP-43 mRNA’s 3’UTR can bind TDP-43 as a native-state stabilizer, and as a potential intrinsic folding chaperone for newly synthesized protein and propose a potential pathway through which TDP-43 aggregation and neurotoxicity may occur.

Materials and Methods Protein expression and purification

The pET-30 vector modified with kanamycin resistance gene containing the recombinant wild-type vYFP-TDP-43 protein sequence or F147L/F149L double mutant generated using Q5 Site-directed mutagenesis kit (New England Biolabs; Figure S2.1B), was transformed into BL21-AI competent cells (Invitrogen), and plated onto LB plates containing 35 µg/mL kanamycin (BioBasic) and incubated at 37 °C for 12 h. Colonies were inoculated into 2 mL of LB medium, incubated at 37 °C for 5 h, and diluted into 300 mL of LB medium. When the culture reached an absorbance of 0.4 at 600 nm, the temperature was reduced to 19 °C and induced using 1 mM IPTG and 0.2% Arabinose. Cells were harvested 6 h after induction by centrifugation at 3000 rpm (Sorvall SLA3000 rotor) for 20 min at 4 °C. The cell pellet was resuspended in 50 mL lysis buffer (40 mM HEPES-KCl, 500 mM KCl, 20 mM imidazole, 20 mM MgCl2, 2 mM βME, 10% glycerol, complete EDTA-free protease inhibitor (Roche), pH 7.4; Johnson et al. 2009). Cells were lysed by sonication in 10 mL aliquots at 4 °C for 5 min (5 s pulse, 5 s stop) using a Vibracell sonicator (Sonics). The lysate was centrifuged for 30 min at 4 °C at 15,000 rpm (Sorvall SS-34 rotor) to isolate soluble protein fraction (supernatant). Purification of the

His6-tagged recombinant protein was achieved by Ni-NTA affinity chromatography using Ni-NTA beads (Sigma) at 4 °C. 10 mL of supernatant was applied to 0.5 mL bead volume of Ni-NTA beads. The beads were washed twice with 1 mL lysis buffer and eluted using modified lysis buffer containing 250 mM imidazole and no protease inhibitors to obtain purified protein.

Dynamic Light Scattering Measurements

Dynamic light scattering (DLS) measurements of hydrodynamic radius (Rh) were made at 20 °C with a DynaPro DLS instrument (Protein Solutions) using a 12 µL quartz cuvette. Samples were measured using 10 s averaging time. Ten or more consecutive measurements were used for regularization analysis using DYNAMICS software. Particle translational diffusion coefficients were calculated from decay curves of autocorrelation of light scattering data and converted to hydrodynamic radius (Rh)

with the Stokes-Einstein equation. Histograms of mass versus Rh were calculated using Dynamics data analysis software.

Size exclusion chromatography

Purified vYFP-TPD-43 (40 µL, 20 mM) was injected into Superdex 75 10/300 GL size exclusion column calibrated using YFP, BSA, RFP and catalase, at a flow rate of 0.5 mL/min. The protein signal was detected using Waters™ 486 tunable absorbance detector at 515 nm with AUFS = 0.1.

Urea denaturation

Purified protein samples (2 µM) were unfolded in 0 M – 7.2 M urea and incubated for 30 min at room temperature. Trp fluorescence and vYFP fluorescence of the samples were measured in 1 cm cuvette

using a Photon Technology International QM-1 fluorescence spectrophotometer using λex= 283 nm

and λem= 315 - 335 nm for Trp, and λex = 515, λem = 528 nm for vYFP using 2 nm bandpass. Integrated fluorescence was normalized and converted to percent folded for Trp fluorescence.

Circular dichroism spectroscopy

Purified protein was dialyzed against buffer containing 20 mM MgCl2, 500 mM KCl, 40 mM HEPES (pH 7.4) to remove imidazole for CD spectroscopy. Far UV CD spectra were acquired with an Aviv CD spectrometer (model 62DS) at 25 °C. Spectra were obtained from 209 nm to 320 nm (1 mm pathlength, 1 nm step sizes, 1 nm bandwidth and 16 s averaging time).

In vitro aggregation of vYFP-TDP-43

Aggregation was induced by a 10-fold dilution of stock solutions of TDP-43 (20 µM) dissolved in 40 mM HEPES-KCl, 500 mM KCl, 250 mM imidazole, 20 mM MgCl2, 2 mM βME, 10% glycerol (pH 7.4) to a final concentration of 2 µM TDP-43 in 170 mM KCl, 36 mM HEPES, 25 mM Imidazole, 18

mM MgCl2, 1% glycerol, 1.8 mM βME (pH 7.4), a method adapted from Johnson et al, 2009. For right angle light scattering measurements, samples were incubated for 4 hr. For fluorescent microscopy and atomic force microscopy experiments, samples were incubated for 20 min.

Right angle light scattering

Right angle light scattering (RALS) of aggregated samples were measured in 1 cm quartz cuvette using a Photon Technology International QM-1 fluorescence spectrophotometer using an excitation wavelength of 400 nm and emission wavelength of 400 nm with 2 nm bandpass at room temperature. Emission was collected at 90 º to the excitation source. Sample volumes of 200 µL were mixed before measurements. Twenty measurements were taken for each oligonucleotide concentration and

averaged. Oligonucleotides were synthesized by Integrated DNA Technologies (Coralville, USA), desalted, and lyophilized, and their quality was assessed by mass spectrometry. Oligos were dissolved in DEPC-treated milli-Q water immediately prior to use. Inhibition curves were fitted by non linear least-squares fitting to Equation 2.1 using OriginPro 8.5 software.

Fluorescence microscopy

Protein samples (9.25 µL) induced to aggregate in the presence or absence of oligonucleotides were inserted into a compartment constructed using Secure-Seal Imaging Spacers (Grace) and two no. 2 18 mm circular micro cover glass (VWR) for imaging. Fluorescence images were acquired on an Olympus IX70 (Olympus, Inc.) inverted microscope with a 40× TIRF objective (N.A. 1.45) (Olympus, Inc.), illuminated with a xenon light. Excitation light was reflected by a 485-555-650TBDR dichroic, and emitted light was passed through a 515-600-730TBEM filter (Omega Optical, Inc.). Images were digitized with a cooled Evolution QEi CCD camera (Media Cybernetics, Inc.).

Atomic force microscopy

All atomic force microscopy (AFM) images were acquired a glass tapping mode fluid cell with a Digital Instruments Nanoscope IIIA Multimode scanning probe microscope (Digital Instruments, Santa Barbara, CA). The AFM images were collected using Nanoscope software (version 5.12) and the J scanning head that has a maximum lateral scanning area of 116 μm × 116 μm. V-shaped silicon nitride probes (SNL-10 cantilever D, Bruker AFM Probes, Camarillo, CA) were irradiated under UV for 30 min to remove organic contaminants. Protein sample (10 μL) induced to aggregate was transferred onto freshly cleaved mica and sealed in the liquid cell. All imaging was captured as 512 × 512 scans at tip scan rate between 0.7 and 1.2 Hz with cantilever drive frequency of ~ 8.5 kHz.

Chapter Remarks Acknowledgements

This study was supported by an Alzheimer’s Society of Canada grant to A.C. and the Ontario Government Scholarship to Y.S. The vector containing human TDP-43 sequence was a kind gift from Dr. Janice Robertson. We thank Dr. Douglas V. Laurents, Dr. Fracisco Baralle, and Dr. Emanuele Buratti for their constructive comments and insightful discussions.

Supplemental material

Figure S2.1: Schematic representations of vYFP-TDP-43 constructs and proposed aggregation mechanism. (A) Cartoon representation of wild type and F147L/F149L mutant vYFP-TDP-43. Features of the constructs and functions of each domain are annotated. Two L residues in the mutant construct are indicated in red for the DNA-binding deficient mutation on RRM1. (B) Primer sequences used to generate the F147L/F149L mutation. Primers are designed by NEBaseChanger (http://nebasechanger.neb.com). (C) Schematic representation of proposed aggregation pathway of vYFP-TDP-43. Natively folded proteins are represented as green squares, while misfolded and aggregated forms are drawn as yellow squares or red ovals respectively. Native- state stabilizing ligands are small blue circles.

References

Arslan, Pharhad Eli, and Avijit Chakrabartty. 2009. “Probing Alzheimer Amyloid Peptide Aggregation Using a Cell-Free Fluorescent Protein Refolding Method.” Biochemistry and Cell Biology 87 (4): 631–39. doi:10.1139/o09-038.

Ayala, Youhna M, Sergio Pantano, Andrea D’Ambrogio, Emanuele Buratti, Antonia Brindisi, Caterina Marchetti, Maurizio Romano, and Francisco E Baralle. 2005. “Human, Drosophila, and C. elegans TDP43: Nucleic Acid Binding Properties and Splicing Regulatory Function.” Journal of Molecular Biology 348 (3): 575–88. doi:10.1016/j.jmb.2005.02.038.

Baron, Desiree M, Laura J Kaushansky, Catherine L Ward, Reddy Ranjith K Sama, Ru-Ju Chian, Kristin J Boggio, Alexandre Jc Quaresma, Jeffrey a Nickerson, and Daryl a Bosco. 2013. “Amyotrophic Lateral Sclerosis-Linked FUS/TLS Alters Stress Granule Assembly and Dynamics.” Molecular Neurodegeneration 8 (1): 30. doi:10.1186/1750-1326-8-30.

Bhardwaj, Amit, Michael P Myers, Emanuele Buratti, and Francisco E Baralle. 2013. “Characterizing TDP-43 Interaction with Its RNA Targets.” Nucleic Acids Research 41 (9): 5062–74. doi:10.1093/nar/gkt189.

Budini, Mauricio, and Emanuele Buratti. 2011. “TDP-43 Autoregulation: Implications for Disease.” Journal of Molecular Neuroscience 45 (3): 473–79. doi:10.1007/s12031-011-9573-8.

Budini, Mauricio, Emanuele Buratti, Cristiana Stuani, Corrado Guarnaccia, Valentina Romano, Laura De Conti, and Francisco E Baralle. 2012. “Cellular Model of TAR DNA-Binding Protein 43 (TDP-43) Aggregation Based on Its C-Terminal Gln/Asn-Rich Region.” The Journal of Biological Chemistry 287 (10): 7512–25. doi:10.1074/jbc.M111.288720.

Deng, Han-Xiang, Hong Zhai, Eileen H Bigio, Jianhua Yan, Faisal Fecto, Kaouther Ajroud, Manjari Mishra, et al. 2010. “FUS-Immunoreactive Inclusions Are a Common Feature in Sporadic and Non- SOD1 Familial Amyotrophic Lateral Sclerosis.” Annals of Neurology 67 (6): 739–48. doi:10.1002/ana.22051.

Estes, Patricia S, Ashley Boehringer, Rebecca Zwick, Jonathan E Tang, Brianna Grigsby, and Daniela C Zarnescu. 2011. “Wild-Type and A315T Mutant TDP-43 Exert Differential Neurotoxicity in a Drosophila Model of ALS.” Human Molecular Genetics 20 (12): 2308–21. doi:10.1093/hmg/ddr124.

Furukawa, Yoshiaki, Kumi Kaneko, and Nobuyuki Nukina. 2011. “Molecular Properties of TAR DNA Binding Protein-43 Fragments Are Dependent upon Its Cleavage Site.” Biochimica et Biophysica Acta 1812 (12): 1577–83. doi:10.1016/j.bbadis.2011.09.005.

Gitler, Aaron D, and James Shorter. 2011. “RNA-Binding Proteins with Prion-like Domains in ALS and FTLD-U.” Prion 5 (3): 179–87. doi:10.4161/pri.5.3.17230.

Greenfield, Nj Norma J. 2007. “Using Circular Dichroism Spectra to Estimate Protein Secondary Structure.” Nature Protocols 1 (6): 2876–90. doi:10.1038/nprot.2006.202.

Ip, Philbert, Vikram Khipple Mulligan, and Avijit Chakrabartty. 2011. “ALS-Causing SOD1 Mutations Promote Production of Copper-Deficient Misfolded Species.” Journal of Molecular Biology 409 (5). Elsevier Ltd: 839–52. doi:10.1016/j.jmb.2011.04.027.

Liachko, Nicole F, Chris R Guthrie, and Brian C Kraemer. 2010. “Phosphorylation Promotes Neurotoxicity in a Caenorhabditis Elegans Model of TDP-43 Proteinopathy.” The Journal of Neuroscience 30 (48): 16208–19. doi:10.1523/JNEUROSCI.2911-10.2010.

Majoor-Krakauer, D, P J Willems, and a Hofman. 2003. “Genetic Epidemiology of Amyotrophic Lateral Sclerosis.” Clinical Genetics 63 (2): 83–101.

Miyata, Masanori, Takashi Sato, Miyuki Kugimiya, Misato Sho, Teruya Nakamura, Shinji Ikemizu, Mami Chirifu, et al. 2010. “The Crystal Structure of the Green Tea Polyphenol (-)-Epigallocatechin Gallate-Transthyretin Complex Reveals a Novel Binding Site Distinct from the Thyroxine Binding Site.” Biochemistry 49 (29): 6104–14. doi:10.1021/bi1004409.

Mompeán, Miguel, Emanuele Buratti, Corrado Guarnaccia, Rui M M Brito, Avijit Chakrabartty, Francisco E Baralle, and Douglas V Laurents. 2014. “‘Structural Characterization of the Minimal

Segment of TDP-43 Competent for Aggregation’.” Archives of Biochemistry and Biophysics 545 (March): 53–62. doi:10.1016/j.abb.2014.01.007.

Morris, Aimee M, Murielle a Watzky, and Richard G Finke. 2009. “Protein Aggregation Kinetics, Mechanism, and Curve-Fitting: A Review of the Literature.” Biochimica et Biophysica Acta 1794 (3): 375–97. doi:10.1016/j.bbapap.2008.10.016.

Mulligan, Vikram Khipple, Aaron Kerman, Sylvia Ho, and Avijit Chakrabartty. 2008. “Denaturational Stress Induces Formation of Zinc-Deficient Monomers of Cu, Zn Superoxide Dismutase: Implications for Pathogenesis in Amyotrophic Lateral Sclerosis.” Journal of Molecular Biology 383 (2): 424–36. doi:10.1016/j.jmb.2008.08.024.

Nagai, Takeharu, Keiji Ibata, Eun Sun Park, Mie Kubota, Katsuhiko Mikoshiba, and Atsushi Miyawaki. 2002. “A Variant of Yellow Fluorescent Protein with Fast and Efficient Maturation for Cell-Biological Applications.” Nature Biotechnology 20 (1): 87–90. doi:10.1038/nbt0102-87.

Nonaka, Takashi, Tetsuaki Arai, Emanuele Buratti, Francisco E Baralle, Haruhiko Akiyama, and Masato Hasegawa. 2009. “Phosphorylated and Ubiquitinated TDP-43 Pathological Inclusions in ALS and FTLD-U Are Recapitulated in SH-SY5Y Cells.” FEBS Letters 583 (2): 394–400. doi:10.1016/j.febslet.2008.12.031.

Polymenidou, Magdalini, and C Lagier-Tourenne. 2011. “Long Pre-mRNA Depletion and RNA Missplicing Contribute to Neuronal Vulnerability from Loss of TDP-43.” Nature Neuroscience 14 (4): 459-68. doi:10.1038/nn.2779.

Polymeropoulos, M. H. 1997. “Mutation in the a-Synuclein Gene Identified in Families with Parkinson’s Disease.” Science 276 (5321): 2045–47. doi:10.1126/science.276.5321.2045.

Rekas, Agata, Jean-René Alattia, Takeharu Nagai, Atsushi Miyawaki, and Mitsuhiko Ikura. 2002. “Crystal Structure of Venus, a Yellow Fluorescent Protein with Improved Maturation and Reduced Environmental Sensitivity.” The Journal of Biological Chemistry 277 (52): 50573–78. doi:10.1074/jbc.M209524200.

Roberts, CJ. 2007. “Non-Native Protein Aggregation Kinetics.” Biotechnology and Bioengineering 98 (5): 927–38. doi:10.1002/bit.

Schenk, D, R Barbour, W Dunn, G Gordon, H Grajeda, T Guido, K Hu, et al. 1999. “Immunization with Amyloid-Beta Attenuates Alzheimer-Disease-like Pathology in the PDAPP Mouse.” Nature 400 (6740): 173–77. doi:10.1038/22124.

Shoichet, Brian K. 2006. “Interpreting Steep Dose-Response Curves in Early Inhibitor Discovery.” Journal of Medicinal Chemistry 49 (25): 7274–77. doi:10.1021/jm061103g.

Tollervey, James R JR, Tomaž Curk, Boris Rogelj, Michael Briese, Matteo Cereda, J Ule, James R JR Tollervey, Tomaž Curk, and Boris Rogelj. 2011. “Characterising the RNA Targets and Position- Dependent Splicing Regulation by TDP-43.” Nature Neuroscience 14 (4): 452–58. doi:10.1038/nn.2778.

Udan, Maria, and Robert H. Baloh. 2011. “Implications of the Prion-Related Q/N Domains in TDP-43 and FUS.” Prion 5 (1): 1–5. doi:10.4161/pri.5.1.14265.

Voigt, Aaron, David Herholz, Fabienne C. Fiesel, Kavita Kaur, Daniel M??ller, Peter Karsten, Stephanie S. Weber, et al. 2010. “TDP-43-Mediated Neuron Loss in Vivo Requires RNA-Binding Activity.” Edited by Mel B. Feany. PloS One 5 (8): e12247. doi:10.1371/journal.pone.0012247.

Wang, I-Fan, Hsiang-Yu Chang, Shin-Chen Hou, Gunn-Guang Liou, Tzong-Der Way, and C-K James Shen. 2012. “The Self-Interaction of Native TDP-43 C Terminus Inhibits Its Degradation and Contributes to Early Proteinopathies.” Nature Communications 3 (May 2011): 766. doi:10.1038/ncomms1766.

CHAPTER III PHYSIOLOGICAL ELECTROLYTES AS REGULATORS OF TDP-43 AGGREGATION AND DROPLET-PHASE BEHAVIOR

At the time of this writing, this chapter is being prepared as a manuscript for submission to eLife as Y. Sun, A. Medina-Cruz, K. Hadley, N. Galant, R. Law and A. Chakrabartty (2018). Physiological Electrolytes as Regulators of TDP-43 Aggregation and Droplet-Phase Behavior. Y.S. and K.H. devised the protein production protocol. A.M.C. performed in vitro spectroscopic measurements of TDP-43 aggregation except those involving ammonium acetate (performed by R.L.). N.G. undertook transmission electron microscopy. Y.S. and A.C. designed the experimental plan. Y.S. performed all remaining experiments and analyses. Y.S. wrote the manuscript with input and revisions from A.C.

72 73

Chapter abstract

Physiological electrolytes induced reversible aggregation of TDP-43 into tufted particles 10-50 nm in

+ + + 2+ 2+ diameter. The order of aggregation induction potency was: K

Abbreviations used in this chapter

ALS, amyotrophic lateral sclerosis; sALS, sporadic ALS; fALS, familial ALS; TAR, transactive response element; TDP-43, TAR DNA binding protein of 43 kDa; FTD, frontotemporal dementia; NTD, N-terminal domain; RRM, RNA recognition motif; CTD, C-terminal domain; SG, stress granule; LLPS, liquid-liquid phase separation; CD, circular dichroism; DLS, dynamic light scattering; TEM, transmission electron microscopy; FRAP, fluorescence recovery after photobleaching; FUS/TLS, Fused in Sarcoma/Translocated in Sarcoma; hnRNPA1, heterogeneous nuclear ribonucleoprotein A1; G3BP, GTPase-activating protein-binding protein 1.

Introduction

Amyotrophic lateral sclerosis (ALS) is an incurable neurological disease characterized by the progressive loss of upper and lower motor neurons, leading to fatal paralysis of the diaphragm within 3-5 years of symptom onset (Tandan and Bradley 1985a; Tandan and Bradley 1985b; Majoor- Krakauer, Willems, and Hofman 2003). Ninety percent of ALS cases are sporadic (sALS) with unknown etiology, while 10% of cases are familial (fALS), caused by mutations in genes such as TARDBP, FUS, SOD1, and C9ORF72 (Kwiatkowski et al. 2009; Rosen et al. 1993; DeJesus- Hernandez et al. 2011; Kabashi et al. 2008). ALS possesses the common hallmark of neurogenerative diseases in that it is associated with the abnormal accumulation of protein inclusions within affected neurons. Landmark studies in the past decade have identified transactive response element (TAR) DNA binding protein of 43 kDa (TDP-43) as a major component of such inclusions in ALS patient neurons, and in neurons of patients with frontotemporal dementia (FTD), a related form of neurodegeneration associated with cognitive deficits (Lashley et al. 2015). Remarkably, TDP-43 inclusions are found in 97% of ALS cases and 50% of cases of FTD, suggesting a mechanistic role in disease pathogenesis (Arai et al. 2006; Sampathu et al. 2006).

TDP-43 is a ubiquitously expressed eukaryotic nuclear protein with a variety of RNA processing functions. It shuttles between the nucleus and cytoplasm to fulfill its numerous roles in splicing and RNA processing (Ratti and Buratti 2016). Importantly, TDP-43 is recruited into stress granules (SGs), transient membraneless organelles formed through the physical process of liquid-liquid phase separation, during cellular stress (Colombrita et al. 2009; Parker et al. 2012).

TDP-43 is 414 amino acids in length and is composed of 3 structural domains. It contains the N- terminal domain (NTD;1-102) which is structurally similar to ubiquitin and Axin1; two canonical folded RNA recognition motifs (RRM1; 106-177 & RRM2; 192-259); and a disordered, low- complexity C-terminal domain (CTD; 274-414) which contain short segments that form transient secondary structures composed of α-helices (320-348) and β-strands (341-366) in the unaggregated state (Mompeán et al. 2016; Lukavsky et al. 2013; Kuo et al. 2014; Jiang et al. 2013; Conicella et al. 2016; Mompeán et al. 2015). The NTD plays a role in multimerization of the protein and appears essential in its splicing function, the RRMs bind selectively to UG- and TG-rich sequences for target recognition, and the CTD contains the “amyloidogeneic core” of the protein responsible for TDP-43 aggregation (Jiang et al. 2013; Afroz et al. 2017; Buratti and Baralle 2001). Currently, it is thought that TDP-43 aggregation causes disease through a combination of loss of native function and gain of toxic

75 function, but the mechanism of how TDP-43 converts from its native state to a pathological aggregate remains unclear (Lee, Lee, and Trojanowski 2012).

The CTD is essential for the phase-separating properties of TDP-43 and nearly all ALS-associated mutations are located in this domain. A number of these mutations also enhance the protein’s propensity for aggregation while impairing its phase-separating properties, leading to the speculation of strong interplay between TDP-43 aggregation and phase separation (Johnson et al. 2009; Conicella et al. 2016). Indeed, an emerging theory of ALS/FTD pathogenesis suggests that the formation of SGs and the partitioning of TDP-43 into these phase-separated organelles is a precursor for TDP-43 aggregation and that persistent formation of SGs and impaired SG disassembly may lead to the accumulation and concentration of TDP-43 and the formation of pathological inclusions (Parker et al. 2012). The suggestion that environmental factors such as chronic exposure to stress may be a cause of pathology is tantalizing, as it would explain why most ALS cases arise from non-genetic causes. Carriers of ALS-associated mutations at the CTD may be more sensitive to these environmental stress factors (Pesiridis et al. 2011).

The investigation of TDP-43 aggregation in the context of phase separation has recently gained interest. However, these studies have exclusively utilized heavily modified protein in cell culture or the CTD alone in solution (Schmidt and Rohatgi 2016; Conicella et al. 2016). Here we use recombinant full-length human TDP-43 conjugated to an N-terminal YFP fluorescent tag (yTDP-43) that retains the native dimeric structure, circular dichroism (CD) spectrum, and nucleotide binding specificity (Sun et al. 2014). We assess the effect of physiological electrolytes on yTDP-43 aggregation kinetics in the solution phase and investigate the effect of these electrolytes on yTDP-43 in the droplet phase by using pre-formed protein droplets as scaffolds. Using these solution-phase and droplet-phase experiments, we identify physiological electrolytes as regulators that alter the biophysical and morphological properties of TDP-43 and discuss their physiological and pathological implications. These electrolyte effects were similar to those predicted by Kirkwood’s theory on protein stabilizing interactions of electrolytes with peptide dipoles and charged protein groups (Kirkwood 1943), and they may adequately describe behaviors of intrinsically disordered proteins, such as TDP-43.

Results

TDP-43 aggregation is modulated by electrolyte concentration

TDP-43 is an intrinsically aggregation-prone protein and has historically been purified in vitro under high-salt conditions to improve protein stability (Johnson et al. 2009; Sun et al. 2014). In previous studies, we purified yTDP-43 from the soluble fraction of E. coli lysates. These purified preparations retained their dimeric conformation, as assessed by dynamic light scattering (DLS) and size exclusion chromatography (SEC) and showed aggregation under near-physiological salt conditions of 170 mM KCl (Sun et al. 2014). The normal range for serum sodium concentration is 137-147 mM (Payne and Levell 1968). We found, however, that under the purification conditions of high electrolyte concentrations (250 mM imidazole, 500 mM KCl), the protein was only marginally stable and could easily be aggregated through mild agitation or spin-concentration above 20 µM. In the present study, we have discovered that yTDP-43 remained soluble in 10 mM HEPES for several weeks. Such findings agreed with published works on the CTD of TDP-43, where the fragment remained soluble in 20 mM HEPES, even at concentrations sufficiently high for NMR studies (Conicella et al. 2016). Using a protocol modified from works by Conicella and colleagues, we purified and refolded yTDP- 43 from the insoluble fraction of E. coli lysates. The purified protein had the expected molecular weight of 77 kDa as analyzed by SDS-PAGE (Figure 3.1A), and after buffer exchange into 10 mM HEPES at pH 7.4, it retained its dimeric structure as assessed by DLS (Figure 3.1B). Additionally, yTDP-43 could be induced to aggregate under physiological salt concentrations (130-150 mM NaCl) and this aggregation was inhibited by the addition of native-state stabilizing oligonucleotides with 12 tandem

TG repeats of single stranded DNA, denoted as (TG)12 (Figure 3.1E) in agreement with previous reports (Sun et al. 2014; Huang et al. 2013). This inhibition was not achieved by the non-binding sequence, (AC)12, suggesting binding specificity by the RRM domains. In sum, these results indicated that the purification and refolding process generated properly folded protein that can dimerize (through a folded NTD) and avoid aggregation under low salt conditions or via specific nucleotide binding (through folded RRMs).

Figure 3.1: The effect of NaCl on purified Venus YFP-tagged, full-length human TDP-43 (yTDP-43) aggregation. (A) SDS-PAGE of recombinant yTDP-43 expressed and purified from the inclusion fraction of E. coli under denaturing conditions of 8 M urea and 500 mM NaCl, 10 mM HEPES, pH 7.4. Approximately 10 μg of protein was applied to a 4%-20% tris-glycine gradient gel. The eluted protein appeared as a single band of expected size (77 kDa) when analyzed by SDS-PAGE (arrow). (B) Mass distribution of particles of various hydrodynamic radii (Rh) of yTDP-43 (20 μM) in low salt, non-denaturing conditions. Purified yTDP-43 was buffer-exchanged into 10 mM HEPES, pH 7.4 and measured by dynamic light scattering. Measurements of Rh were converted to MW by the Stoke-Einstein equation using appropriate software. (C) Measurements of yTDP-43 sample turbidity (absorbance at 600 nm) from 1 s to 3600 s after dilution into varying NaCl concentrations. Samples were prepared by diluting stock yTDP-43 in 10 mM HEPES buffer, pH 7.4 into working buffers containing final concentrations of 20 μM protein, 10 mM HEPES, pH 7.4 and varying concentrations of NaCl. Each color curve represents a different NaCl concentration tested, ranging from 0 mM (black) to 200 mM (red). (D) Confocal fluorescence microscopy of yTDP-43 solution immediately after dilution into high NaCl conditions (20 μM protein, 200 mM NaCl, 10 mM HEPES, pH 7.4). Flocculent fluorescent aggregates appeared after several minutes. Scale bar = 50 μm. (E) Relative end- point (1 h) scattering of yTDP-43 samples (2 μM) prepared in the presence (+) or absence (−) of NaCl (150 mM), (TG)12 (10 μM) or (AC)12 (10 μM).

We next assessed the effect of increasing NaCl concentration on the kinetics of aggregation through turbidity measurements. The turbidity did not increase when yTDP-43 was incubated under conditions of 0 mM – 50 mM NaCl but increased in rate and amplitude as the concentration of NaCl was increased to 100 mM and above (Figure 3.1C). When visualized by confocal fluorescence microscopy, the aggregates that form over time appeared initially as small speckles and became flocculent tufts that formed short connecting branches (Figure 3.1D).

We next examined whether this aggregation process was specific to NaCl or a general ionic strength

effect. We measured end-point turbidity of yTDP-43 incubated in various salts (CaCl2, MgCl2, NaCl,

KCl, and NH4OAc) at a range of concentrations/ionic strengths to assess the extent to which these salts caused yTDP-43 aggregation (Figure 3.2A). All of these salts generated sigmoidal aggregation

curves that can be fit to the equation (Figure 3.2A-B). The fitted values showed that yTDP-43 aggregation curves all had Hill coefficients (n) far greater than 1, suggesting a highly cooperative process. For individual electrolytes, salts composed of divalent cations such as

CaCl2 and MgCl2 caused aggregation at very low concentrations, with EC50 values of 2.4 and 3.0 mM

respectively. On the other hand, NaCl and KCl produced aggregation curves with EC50 values of 127 mM and 222 mM respectively. The EC50 of the NH4OAc curve (90 mM) was closer to that of the

NaCl curve (127mM) than all other electrolytes. NH4OAc was tested because it is compatible with electron microscopy, which was used in subsequent experiments. These ion-specific effects on

aggregation were not simply a function of ionic strength. While solutions of NaCl, KCl, and NH4OAc all were of identical ionic strengths, their effects on aggregation were very different.

To ensure that the YFP tag did not interfere with the aggregation properties, we cleaved off the His6- tagged YFP from our construct and performed the same end-point turbidity measurements (Figure

3.2B). The results showed that the presence of His6-YFP tag only marginally affected the NaCl-

dependence of turbidity change, causing a slight shift of EC50 from 127 mM to 115 mM. We continued further experiments using the YFP-tagged recombinant protein due to its ease of purification and fluorescent properties.

Figure 3.2: Aggregation of TDP-43 is induced by various electrolytes. (A) Normalized turbidity of yTDP-43 (20 μM) in 10 mM HEPES, pH 7.4 and varying concentrations of CaCl2 (red), MgCl2 (green), NH4OAc (orange), NaCl (black), and KCl (blue) at equilibrium (1 h) is measured by absorbance at 600 nm. The curves were fit to the logistic function to extract EC50 and n-values. (B) Normalized turbidity of 2 μM YFP-tagged (●) and untagged (○) TDP-43 in 0-200 mM NaCl at 1 h. Curve fitting was done as previously described. (C) Hydrodynamic radii measured by dynamic light scattering (black) and turbidity at 600 nm (orange) of 2 μM yTDP-43 in varying concentrations of NH4OAc. Duplicates were measured per data point. Error bars represent 1 standard deviation. Turbidity curve (orange) was fit using Equation 2.1, while the dotted black line in the Rh curve serves as a guide for the eye. (D) Dynamic light scattering of 2 µM yTDP-43 samples incubated at increasing concentrations of NH4OAc, 10 mM HEPES, pH 7.4. Relative signal intensities (by mass) are plotted against Rh. (E) Uranyl acetate negative-stained transmission electron micrographs of yTDP-43 aggregates induced by various concentrations of NH4OAc. Scale bar = 500 nm.

TDP-43 forms non-fibrillar aggregates

Given that NH4OAc 1) is a salt compatible with electron microscopy and 2) caused aggregation at a

similar concentration as NaCl, we examined the structure of yTDP-43 aggregates caused by NH4OAc by DLS, turbidity, and transmission electron microscopy (TEM). While DLS detected particles with hydrodynamic radii (Rh) >5 nm at NH4OAc concentrations above 40 mM, turbidity measurements only detected an increase in turbidity of the sample at NH4OAc concentrations greater than 75mM, plateauing at 120 mM, at which DLS measurements could not be made due to instrumentation limits

(Figure 3.2C). We detected a gradual increase in the Rh of the major species using DLS when NH4OAc concentration were increased from 0 mM to 80 mM (Figure 3.2D). To visualize these particles detected by spectroscopic methods, we took samples at various NH4OAc concentrations and imaged them using negative stain TEM. The morphology of these particles did not differ from aggregates generated by NaCl, but due to the additional wash steps and lingering NaCl crystals for samples prepared with

NaCl, we conducted TEM and DLS studies on NH4OAc-treated proteins (Figure S3.1). The aggregates formed had similar morphology to those seen at low magnifications using fluorescence microscopy (Figure 3.1D) and resembled short-branched amorphous flocculent tufts that appeared to loosely associate, with no fibrillar characteristics (Figure 3.2E). The size of the particles generally

increased as we increased the NH4OAc concentration, in agreement with DLS measurements.

Pathological protein aggregates are often amyloid fibrils, which are β-structured, unbranched fibrils, ~10 Å in diameter, that bind amyloid dyes like Thioflavin T (Sipe and Cohen 2000). Several experiments we performed demonstrate that the yTDP-43 aggregates observed here are non-amyloid. First, the EM results (Figure 3.2E) demonstrate that the yTDP-43 aggregates are nonfibrillar, and thus non-amyloid. Second, pathological TDP-43 aggregates in ALS/FTD brain tissue and aggregates formed in vitro do not bind Thioflavin S or Thioflavin T, amyloid-specific dyes (Figure 3.3A-B; Kerman et al. 2010). Third, CD measurements demonstrate a loss in CD signal, rather than an increase in β- structure (Figure 3.3C). The loss of CD signal is likely caused by the increased turbidity, which tracks with CD signal loss.

Figure 3.3: Non-amyloid nature of yTDP-43 aggregation. (A) Thioflavin S (ThS) staining of FTD brain tissue with TDP-43-positive inclusions. Tissue was stained using anti-TDP-43 antibodies (red), ThS (green), and DAPI nuclear stain (blue). ThS signals show background staining of the tissue non- specific to TDP-43 inclusions. Scale bar = 50 μm. (B) Thioflavin T (ThT) staining of Aβ42 fibrils, soluble TDP-43, or TDP-43 aggregates. ThT fluorescence intensity was measured by fluorescence (λex = 450 nm, λem= 490 nm). Error bars represent standard deviation. (C) Kinetic measurements of yTDP-43 aggregation by CD spectroscopy. Stock yTDP-43 was diluted to a final concentration of 5 μM protein in high salt (130 mM NaCl, 10 mM HEPES, pH 7.4) to initiate aggregation and immediately measured by CD every 10 min for 6 h. CD at 218 nm was plotted as a function of time to assess changes to β-strand secondary structure (black). Turbidity of the sample was plotted on the same time-scale (orange).

TDP-43 aggregation is reversible

The turbidity changes to protein samples in the presence of 100-200 mM salts resemble the effects of salts on phase separating proteins such as the TDP-43 CTD in previous studies (Conicella et al. 2016). We observed, instead, that under our experimental conditions, the full-length protein formed flocculent aggregates rather than phase separated droplets (Figure 3.1D, 3.2D). Notwithstanding, we tested whether these aggregates had properties similar to phase separating proteins such as disassociation and reversibility. We incubated pre-formed aggregates of yTDP-43 (induced by 200 mM NaCl) in low-NaCl conditions and compared these measurements to freshly purified proteins at the same final protein and electrolyte concentrations (Figure 3.4A). Strikingly, pre-aggregated yTDP-43 samples decreased in turbidity when diluted to lower NaCl concentrations, and a dilution to 50 mM NaCl caused turbidity to reach values similar to unaggregated samples. Although hysteresis is present in the two curves, the reversibility of yTDP-43 aggregation is markedly evident and previously unreported. The hysteresis may be a result of not reaching equilibrium. We confirmed the reversibility process by fluorescence microscopy, where we observed disassociation of these aggregates within 2 minutes of dilution into reduced salt conditions (Figure 3.4B).

Figure 3.4: Reversibility of TDP-43 aggregation. (A) Normalized turbidity of 2 μM yTDP-43 in 0-200 mM NaCl concentrations measured by absorbance at 600 nm. The aggregation curve (●) were measurements of samples 1 h after dilution from 10 mM HEPES, pH 7.4 into various NaCl concentrations. The disaggregation curve (○) were samples of pre-aggregated yTDP-43 in 200 mM NaCl diluted to 50-200 mM NaCl and a final concentration of 2 μM yTDP-43. Data was fit to a logistic function. (B) Confocal fluorescence microscopy of a sample of pre- aggregated yTDP-43 in 200 mM NaCl prior to (T=0s) and following (T>0s) dilution into 50 mM NaCl. Scale bar = 50 µm.

Kinetics properties of yTDP-43 aggregation

After assessing the electrolyte dependence, reversibility, and morphology of yTDP-43 aggregates at end-point, we examined the kinetics of yTDP-43 aggregation by turbidity measurements. The turbidity values of NaCl-treated samples over time showed a growth curve with an initial lag phase followed by a growth phase, which suggested a 2-step process of aggregation (Figure 3.1B). To model this process, we chose a simple 2-step model that has been applied successfully to model the aggregation of a number of proteins involved in protein misfolding diseases such as α-synuclein, amyloid β, and polyglutamate (Morris, Watzky, and Finke 2009; Bentea, Watzky, and Finke 2017). In this simplified model, the soluble protein initially forms a seeding nucleus, then proceeds to grow autocatalytically with continuous nucleation (Figure 3.5G). We first generated data sets by inducing aggregation of yTDP-43 at 200 mM NaCl while varying the concentration of yTDP-43 (Figure 3.5A). We fit this data to extract the nucleation and propagation rates (k1 and k2) and plotted these values as a function of yTDP-43 concentration (Figure 3.5C). While the protein concentration dependence of nucleation rate displayed a linear relationship with a small positive slope, the propagation rate versus protein concentration revealed an inverse hyperbolic relationship that was concentration-independent at protein concentrations > 8 µM. The end-point turbidity plot showed a sigmoidal relationship between turbidity and yTDP-43 concentration (Figure 3.5B). We then performed the same kinetic analysis by varying NaCl concentration and keeping yTDP-43 at 20 µM (Figure 3.5D). Consistent with our equilibrium data, the end point scattering of 20 µM yTDP-43 changed with yTDP-43 concentration in a sigmoidal manner (Figure 3.5E). Additionally, the plot of nucleation and propagation rates (k1 and k2) as a function of NaCl concentration showed an inverse relationship. While nucleation rates increased with [NaCl], propagation rates decreased at higher electrolyte concentrations. With nucleation rates being on par or slightly greater than propagation rates, these results predict the presence of many small aggregates and very few large aggregates at high protein concentrations and [NaCl]. This may explain why full-length TDP-43 does not form typical amyloid (Figure 3.3A; Kerman et al. 2010).

Figure 3.5: Effect of protein concentration and NaCl concentration on yTDP-43 aggregation kinetics. Kinetics of aggregation was modeled using a simple 2-step nucleation-propagation model. (A) Turbidity measurements of varying final concentrations (0-20 mM) of yTDP-43 in 200 mM NaCl over 3600 seconds. Curves were fit to a sigmoidal function using non-linear least square fitting. Curve colors correspond to increasing concentrations of yTDP-43 from 0 µM (black) to 20 µM (red). (B) End-point (T=3600 s) turbidity of yTDP-43 samples as a function of yTDP-43 concentration. Data were fit to a sigmoidal curve as a guide for the eye. (C) Effect of yTDP-43 concentration on nucleation (k1, black) and propagation (k2, red) rates calculated from curve fitting using kinetic data. Trendlines serve as a guide for the eye. (D-F) Kinetic (D) and endpoint (E) turbidity as a function of NaCl concentration. Measurements and curve fitting were performed as described in (A-C) but varying NaCl concentration (0-200 mM) while keeping yTDP-43 concentrations constant (20 µM). Calculated rates were plotted as a function of NaCl (F). (G) Schematic representation of the simple two-step aggregation model used for curve fitting. The model consists of a nucleation step (A → B) and propagation step with continuous nucleation (A + B → 2B). Soluble protein is represented by green dimers while aggregates that contribute to turbidity change are orange oligomers.

Insertion of yTDP-43 into Ddx4N1 droplets

Recent studies have suggested that TDP-43 is able to phase separate under certain experimental conditions. Namely, its CTD fragment forms droplets in vitro and the full length protein has been shown to form nuclear vacuolated droplets in a GFP-tagged overexpression cell culture system (Burke et al. 2015; Conicella et al. 2016; Schmidt and Rohatgi 2016). In experiments using SUMO-tagged full- length TDP-43, the droplet-like structures formed did not have typical liquid-liquid phase separation properties such as the ability to fuse and form a spherical shape (Molliex et al. 2015). We also encountered this ‘atypical droplet’ morphology, but only in freeze-thaw samples, and found them to be unlike liquid droplets as they did not recover after photobleaching (Figure S3.2, S3.3). We suspect that these structures may be an artefact of freeze-thaw treatment.

Under physiological salt conditions with fresh protein, we found that yTDP-43 did not phase separate, but rather formed reversible aggregates. However, a multitude of cell-culture studies have shown that TDP-43 does enter organelles such as SGs (Colombrita et al. 2009; Dewey et al. 2011; Parker et al. 2012). This lead us to speculate that although TDP-43 CTD has phase separating properties, the full length protein may require other agents to form droplets, behaving as a “client” protein that partitions into pre-existing droplets structures, rather than acting as a “scaffold” for phase separation (Banani et al. 2016). To assess how TDP-43 behaves in the droplet state, we mixed yTDP-43 with droplets formed by the N-terminal disordered domain of the DEAD-box helicase Ddx4 (residues 1-236; Ddx4N1). We selected Ddx4 over stress granule proteins because it is a well-behaved and robust protein fragment capable of phase separation in a wide range of salt concentrations (Nott et al. 2015; Brady et al. 2017), including the experimental conditions used here. In addition, Ddx4N1 shows a very high sequence compositional similarity to stress granule proteins, using a dipeptide comparison approach (Vernon et al. 2018). Ddx4N1 is especially similar to the abundant stress granule protein hnRNPA1 (heterogeneous nuclear ribonucleoprotein A1) and to TDP-43 (Table 3.1).

Table 3.1: Compositional imilarity between Ddx4N1 (1-236) and known stress granule proteins with intrinsically disordered regions. Similarity was assessed by a scoring matrix comparing the frequency of each of the 400 possible dipeptide amino acid pairs in Ddx4N1 (1-236) with dipeptide frequencies of all proteins in the human proteome. Scores of reported stress granule proteins with known intrinsically disordered domains (references listed) are rank ordered from most similar to least similar to Ddx4N1. *The single protein with higher similarity score than hnRNPA1 is full-length Ddx4.

Number (percentage) of human SG protein Reference UniProt ID proteins with higher score Molliex et al. 2015; hnRNPA1 Kim et al. 2013; P09651 1 (0.005 %)* Lin et al. 2015 DDX3X Valentin-Vega et al. 2016 O00571 5 (0.024 %) TDP-43 Conicella et al. 2016 Q13148 51 (0.242 %) Kato et al. 2012; hnRNPA2B1 P22626 69 (0.328 %) Kim et al. 2013 Lin et al. 2015; FUS Patel et al. 2015; P35637 207 (0.984 %) Murakami et al. 2015 Blechingberg et al. 2012; TAF15 Q92804 1153 (5.478 %) Schüller and Eick 2016 Blechingberg et al. 2012; EWS Q01844 1795 (8.529 %) Schüller and Eick 2016 Lin et al. 2015; TIA1 P31483 8958 (42.562 %) Kedersha et al. 1999

Under the low salt conditions used (40 mM NaCl, 10 mM HEPES, pH 7.4) Ddx4N1 readily formed droplets, while TDP-43 remained soluble and monodisperse. Mixing of the two proteins resulted in the partitioning of yTDP-43 into Ddx4N1 droplets that varied from 2 to 50 µm in diameter (Figure

3.6A). We calculated the free energy of partitioning (ΔGpartition) using the relative fluorescence in the soluble and droplet phases (Figure 3.6B). We then constructed a number of truncations of full-length

TDP-43 to determine the effect of truncations on ΔGpartition. The construct containing only YFP-tagged CTD (yCTD) showed the highest propensity to enter droplets. The yTDP-43, on average, showed

reduced droplet partitioning compared to yCTD, but had the highest variation of ΔGpartition between individual droplets. Conversely, the ΔGpartition was greatly reduced in constructs where the CTD was deleted (yΔCTD and yNTD), reaching levels comparable to the basal partitioning propensity of YFP (Figure 3.6C). To assess the dynamic properties of full-length (yTDP-43) and its CTD (yCTD) within the droplet (~10 µm diameter), we measured the recovery of YFP signal after photobleaching. The full-length protein recovered up to 65% of its initial pre-bleach fluorescence while yCTD showed near complete recovery, to 90-100% of its pre-bleach levels (Figure 3.6D, 3.6E). Fitting the recovery measurements to an exponential decay curve, we extracted the half-life of recovery and estimated the

. diffusion rate (D) of the fluorescent material within the droplet using the equation: , where /

R is the radius of the bleached area and t1/2 is the half-life of recovery (Soumpasis 1983). We calculated the D and % mobility values in 3 droplets for each construct and the comparison shows that yCTD shows near-complete recovery and has significantly higher diffusion rates compared to yTDP-43 (Figure 3.6F). In sum, our results suggest that the CTD is essential for partitioning into the droplet phase and that yCTD enters droplets much more readily and has improved mobility within. On the other hand, the presence of the NTD and RRMs in the full-length protein appeared to moderate and reduce the diffusion rates of the proteins within droplets and restrict its mobility at room temperature.

Figure 3.6: Insertion of yTDP-43 into pre-formed Ddx4N1 droplets is mediated by the C- terminal domain of TDP-43. (A) Confocal fluorescence microscopy of protein samples in low salt conditions (40 mM NaCl, 10 mM HEPES, 2 mM DTT, pH 7.4). Ddx4N1 alone (30 µM), yTDP-43 alone (2 µM), and mixture of the two proteins are displayed as separate panels. YFP fluorescence (top row) and bright field differential interference contrast (DIC; bottom row) channels of the same field of view are displayed. Scale bar = 50 µm. (B) Calculation of free energy of partition (ΔGpartition) for an individual droplet. The partitioning coefficient (Kpartition) is defined as the ratio of concentration of protein within the droplet and outside the droplet at equilibrium. Scale bar = 10 µm. This was calculated from the average pixel intensity within a droplet (within red dotted lines) and the intensity outside (black). ΔGpartition is defined as the free energy of droplet entry, calculated from Kpartition. (C) Average ΔGpartition of particles containing different YFP-tagged TDP-43 constructs as indicated by the legend. Average ΔGpartition of yTDP-43 (residues 1-414), yCTD (residues 267-414), yΔCTD (residues 1-266), yNTD (residues 1-106) and YFP alone are displayed as bar graphs, with error bars representing 1 standard deviation. The graphic filling each bar are representative fields of view of confocal microscopy images of Ddx4N1 and YFP-tagged TDP-43 construct mixtures. Scale bar = 50 µm. (D) Fluorescence recover after photobleaching (FRAP) of a representative droplet (11 µm diameter) containing yTDP-43. Average fluorescence of the droplet was measured over time after bleaching, normalized to pre-bleaching fluorescence levels and converted to % recovery. An accompanying confocal image of a representative droplet and its recovery is shown (inset). The mobile fraction (% mobile) and diffusion rate D were extracted from FRAP data using an exponential decay function. (E) FRAP curve and confocal images of a representative droplet (10 µm diameter) containing YFP- tagged CTD of TDP-43. Droplet properties are calculated as previously described. (F) Comparison of average (n=3) % mobility (left) and diffusion rates (right) of FL (white) and CTD (grey) YFP-tagged constructs. Error bars represent standard deviation.

TDP-43 behavior in droplets is affected by environmental conditions

In previous studies, it has been suggested that the maturation (aging) of SGs that contain TDP-43 may cause the normally reversible aggregation of the protein inside SGs to become irreversible aggregates (Parker et al. 2012; Li et al. 2013). Here, we allowed the mixture of yTDP-43 and Ddx4N1 droplets to persist from 0.5 h to 24 h and assessed any morphological changes to the yTDP-43 through fluorescence microscopy. In general, the distribution of yTDP-43 fluorescence within the Ddx4N1 droplets - under salt conditions where yTDP-43 should not aggregate - changed from being diffuse at 30 minutes after mixing to progressively more granular after 2 h and at 24 h after mixing (Figure 3.7A). Additionally, FRAP analysis of these droplets (3 per condition) showed that the overall fraction of mobile material and the diffusion rate of fluorescent material within the droplets quickly decrease over time, particularly after 3 hours of droplet persistence, suggesting the conversion from diffuse protein into rigid structures (Figure 3.7B).

Given how physiological ions can affect the aggregation properties of yTDP-43 in solution, we next tested whether these changes to solution electrolyte concentrations can also affect the behavior of yTDP-43 in the droplet phase. While keeping the ionic strength at 150 mM, we altered the ratio of sodium- and potassium- chloride salt concentrations. As expected from our previous experiments we found that yTDP-43 formed aggregates at NaCl >90 mM or KCl >120 mM. However, when mixed with Ddx4N1 droplets, these yTDP-43 aggregates accumulated at the droplet-solution interface (Figure 3.7D). At 3 h, these aggregates appear to grow larger and accumulates at the interface. After 24 h, diffusely distributed fluorescence was no longer present within the droplets, but a ring-shaped arrangement of aggregates appeared for each particle. This behavior is consistent across multiple NaCl-to-KCl ratios, except at 150 mM NaCl, where the ring-shaped aggregates form after only 3 h after mixing. In contrast, at 150 mM KCl, the number of aggregates at end-point were far less numerous and were at reduced fluorescence intensities compared to those generated by samples containing NaCl. This suggested that TDP-43 morphology in Ddx4N1 droplets was more sensitive to NaCl than KCl, agreeing with our solution-phase results.

Divalent cations induced highly cooperative aggregation of yTDP-43 in solution (Figure 3.2A). We tested whether these electrolytes can also alter the behavior of yTDP-43 in the droplet state under physiologically relevant conditions. We incubated yTDP-43 and Ddx4N1 droplets under buffer conditions mimicking the electrolyte concentrations of the cytosol (120 mM KCl, 10 mM NaCl, 0.002 mM CaCl2) and the extracellular space (10 mM KCl, 120 mM NaCl, 2 mM CaCl2). To assess droplet

morphology in detail, we generated orthogonal views of representative droplets by using a series of z- stack confocal images of droplets formed under these electrolyte conditions upon initial mixing of yTDP-43 and Ddx4N1 (Figure 3.7C). Under cytosolic conditions of low Ca2+ and low Na+, yTDP-43 was distributed as speckles within droplets, with some aggregates at the droplet-solution interface and a significant amount of diffuse yTDP-43 within droplets, similar to observations in the high-KCl, low- NaCl condition after 0.5 h mixing. The extracellular condition, however, resulted in a drastically different morphology, where yTDP-43 accumulated in the solution interface and failed to enter the droplets. This morphology resembled that of aged droplets (>3 h) where aggregates accumulated at droplet surfaces. The orthogonal views of these droplets reinforce the notion that while both conditions appeared to generate aggregate-like structures, the distribution of these structures within the droplets were different, and dependent on the electrolyte environment. Whereas yTDP-43 was still able to enter droplets under cytosolic electrolyte concentrations, this entry is impaired under extracellular electrolyte conditions.

Figure 3.7: Alteration to yTDP-43 droplet morphology by droplet persistence and environmental electrolyte concentrations. (A) Confocal fluorescence images of representative fields of yTDP-43 in Ddx4N1 droplets that have persisted for 0.5 h, 3 h, and 24 h. A high- magnification inset of a representative droplet in the region indicated by a dotted box is displayed. Scale bar = 50 µm. (B) FRAP analysis of droplets at varying durations of droplet persistence. Average mobile fraction (black) and diffusion rates (blue) were calculated from 3 droplets for each droplet age as previously described. Error bars represent standard deviation. (C) Ddx4N1 droplets containing yTDP-43 morphology were imaged in 3-dimensions under different physiological electrolyte conditions. Orthogonal cross-sections in the x-y (white), x-z (green) and z-y (yellow) planes were generated from a set of z-stack scans of droplets imaged by YFP fluorescence under conditions mimicking cytosolic (cyto) and extracellular (EC) electrolyte concentrations. Concentrations of electrolytes for each condition is listed. Scale bar = 20 µm. (D) Droplet morphologies assessed by confocal microscopy using YFP fluorescence under varying electrolyte concentrations at 0.5 h, 3 h, and 24 h after sample preparation. Total electrolyte concentration was maintained at 150 mM while varying ratios of NaCl and KCl are used. Scale bar = 30 µm.

Discussion

The purification and in vitro characterization of TDP-43 have been historically difficult due to its intrinsic aggregation-prone properties. Here, we demonstrated that yTDP-43 purified from the

inclusion fraction of E. coli possessed the same dimeric structure and specific binding to TG12 as the construct purified from the soluble fraction in previous reports (Johnson et al. 2009; Sun et al. 2014). Using this purification method, we were able to concentrate protein stocks at concentrations upwards of 200 µM, which provided the potential for future structural-based studies of the full-length protein using methods such as NMR spectroscopy.

TDP-43 aggregation is modulated by physiological electrolytes

When we exposed yTDP-43 to various electrolyte concentrations in vitro, specific electrolytes at different concentrations affected TDP-43 aggregation to different degrees (Figure 3.2A). The

2+ 2+ + + + observed order of aggregation induction potency was: Ca > Mg > NH4 > Na > K , which differs from the Hoffmeister series (Beauchamp and Khajehpour 2012). Furthermore, while Hoffmeister effects are observed at concentrations of ~1 M, the effects reported here for yTDP-43 range from 1- 200 mM, thus the effects of electrolytes we observe on TDP-43 solubility and structure appears not to be caused by the Hoffmeister effect. We observed that divalent cations such as Ca2+ and Mg2+

+ + caused aggregation at EC50 concentrations in the 2-5 mM range, monovalent cations Na and K , as

+ well as NH4 , induced aggregation at EC50 concentrations of 100-200 mM. Remarkably, even though Na+ and K+ ions have the same charge and have generally very similar properties, they induced the

transition of TDP-43 into the aggregated state at significantly different EC50 values (120 mM and 220

mM respectively), suggesting electrolyte specificity. The removal of the His6-YFP tag from the full- length protein, which caused the removal of a local group of positively charged residues, had only marginal effects on TDP-43 aggregation, suggesting that the process is not likely dependent on charge interactions or general ionic effects, but rather a highly ion-specific effect not seen in typical, well- folded proteins. These ion-specific interactions occur in the 1-200 mM range, which are higher than the ion binding site affinities observed with natively folded proteins (which are typically < 1 μM). We propose these ionic effects may represent the effects proposed by Kirkwood for the structure-inducing interactions of ions with protein dipoles and ampholyte groups.

Kirkwood’s theory of electrolyte-protein interactions applied to TDP-43

The electrolyte effects reported here are similar to predictions made by Kirkwood in 1943, where he proposed a theory in which charge-charge and charge-dipole interactions of electrolytes with proteins stabilize protein structure formation (Kirkwood 1943). The protein was considered as a string of peptide dipoles and charges, which interact with electrolytes to induce and stabilize the protein structure. A modern test of Kirkwood’s theory was performed by Scholtz and colleagues (Scholtz et al. 1991). A charge-less partially helical peptide composed chiefly of Ala and Gln was synthesized, and it was shown to gain helical content upon addition of millimolar concentrations of electrolytes (NaCl,

Na2SO4, and CaCl2). The structure formation in the charge-less helical peptide occurred at concentrations around 250 mM and thus is below the range for Hoffmeister effects. The helix- stabilizing property of the electrolytes were thought to be charge-dipole interactions between the electrolytes and the dipoles of the peptide bonds. The effects of electrolytes on the structure of the charge-less helical peptide mirrors the effect of electrolytes on the aggregation of TDP-43 and occurs in a similar millimolar range. These data prompted our speculation that Kirkwood’s theory may be applicable to structural transitions (e.g. aggregation or phase separation) observed with intrinsically disordered proteins, and these ion-specific effects on TDP-43 aggregation and droplet partitioning may have physiological implications (vida infra).

Effect of electrolytes on TDP-43 aggregate morphology

The DLS and turbidity measurements of yTDP-43 aggregates induced by NH4OAc showed that

yTDP-43 shifted from its dimeric native state (~5 nm) at 0 mM NH4OAc into particles of larger radii

(~20 nm) as electrolyte concentrations were increased (80 mM NH4OAc). This increase in the radius of the particles scaled exponentially with electrolyte concentration. This gradual increase in the sizes of the main peaks in the DLS data was not detectable by turbidity measurements and may suggest that the aggregation process involves early stage, small, soluble oligomeric intermediates whose size increase with the concentration of electrolytes. The NTD of TDP-43 has been suggested to form larger, dynamic oligomers, but whether these oligomers are part of the aggregation pathway of the full-length protein remains uncertain (Afroz et al. 2017).

The spectroscopic data are in general agreement with the micrographs produced by EM, where few,

small particles were present in 0 mM NH4OAc and larger particles of ~25 nm radii were present at

40-100 mM NH4OAc. At greater than 120 mM NH4OAc, larger particles of several hundred micron in radii were present, corresponding to the detection of A600 absorbance signals. Morphologically, we

observed that TDP-43 aggregates resemble loose tufts in both immunofluorescence microscopy and by electron microscopy (Figure 3.1D, 3.2E). Previous EM studies have suggested that TDP-43 aggregates form fibrillar/rope-like structures or possess amyloid properties, but these studies were often conducted with TDP-43 fragments such as the truncated RRM domains, where aggregation was caused by the truncation of RRM2, which exposed the β-sheet regions of the RRM fold and lead to cross-β structures. Other studies have artificially introduced aggregates via physical stress such as agitation, which is unlikely to occur under physiological conditions (Chiang et al. 2016; Chang et al. 2013). Earlier findings from this laboratory show that inclusions in SOD1-ALS patients are non- amyloid, as they do not exhibit Congo red birefringence or significant ThS fluorescence (Kerman et al. 2010). In our TEM analysis of TDP-43 aggregates using the full-length protein, we find that these aggregates were amorphous, and also showed no fibrillar morphology (Figure 3.2E). The aggregates were negative for amyloid using Thioflavin T assays, and aggregates in FTD brain tissue also did not stain with ThS (Figure 3.3A-B). These results suggest that while TDP-43 CTD has been shown to form amyloid-like, β-rich structures, and is often described as “prion-like”, aggregates formed by full- lengthTDP-43 do not exhibit morphological signs of amyloid.

Kinetics of electrolyte-induced aggregation correlates with morphological changes

In our kinetic measurements, we show that yTDP-43 aggregation kinetics appeared to be more dependent on NaCl concentration than yTDP-43 concentration, and that the aggregation process is unlike protein aggregations in other pathologies. In many protein misfolding diseases, the hallmark of pathological aggregation is the initial dissociation of the multimeric protein into misfolding-prone monomers, followed by the irreversible accumulation of aggregates (Roberts 2007; Morris, Watzky, and Finke 2009). The kinetics of these aggregation processes typically involves an initial rate-limiting nucleation phase that occurs with slow kinetics, followed by an autocatalytic propagation phase (Bentea, Watzky, and Finke 2017). However, we found that TDP-43 aggregation was atypical in that 1) the aggregates that formed were reversible and readily dissociated upon electrolyte concentration reduction (Figure 3.4), and 2) the nucleation and elongation rate constants in the aggregation kinetics were similar in magnitude (Figure 3.7). In fact, an increased rate of nucleation was coupled with the decrease in propagation as NaCl concentration was increased. These properties are consistent with the morphological analysis of TDP-43 aggregates using fluorescence confocal microscopy (Figure 3.1C) and negative stain TEM (Figure 3.2E), where yTDP-43 formed small nuclei in high numbers rather than long fibrillar structures such as those seen in aggregates formed by α-synuclein or

transthyretin (Karpowicz et al. 2017; Galant et al. 2016). This suggests that as the rate of nucleation increases, the available pool of soluble protein decreases, which hinders the elongation process in this simple two-step model. While larger particles did form in the confocal experiments (Figure 3.1D), we suspect this is due to accumulation of small aggregates sedimenting to the bottom of the imaging

chamber. In our TEM images, the aggregates induced by NH4OAc did not differ in morphology from those induced by NaCl (Figure S3.1), which suggests that the aggregates induced by electrolytes are likely formed through similar mechanisms.

Insights into TDP-43 behavior in a droplet structure

The droplet state of full-length TDP-43 has been difficult to study in vitro due to its complex properties (Molliex et al. 2015). TDP-43 forms very small droplet-like structures (1 µm in diameter) that do not exhibit fusion behavior (Molliex et al. 2015). TDP-43 behaves as a client of stress granules and not a scaffold (Banani et al. 2016). Here, we used Ddx4N1 as a scaffold to characterize yTDP-43 in the droplet phase. Human Ddx4 is a helicase found exclusively in germ cells that plays a role in germ cell development and is not known to interact directly with TDP-43 (Castrillon et al. 2000; Brady et al. 2017). We selected Ddx4 over stress granule proteins like hnRNPA1 because it is a well-behaved and robust protein fragment capable of phase separation under a wide range of salt concentrations at room temperature (Nott et al. 2015; Brady et al. 2017). In addition, Ddx4N1 shows a very high sequence similarity to stress granule proteins, hnRNPA1 and TDP-43 (Table 3.1).

In our two-component system, we showed that the CTD was required for entry into the droplet state. This is consistent with the notion that the CTD is responsible for the phase separating properties of TDP-43. By measuring FRAP of YFP signals within these droplets, we observed relatively fast diffusion and recovery rates of the yCTD construct. In the full-length protein however, entry into the droplet was less favourable and mobility and diffusion rate of yTDP-43 was reduced, suggesting a modulatory effect caused by the NTD and RRMs (Figure 3.6C). The observed reduction of insertion, diffusion and mobility of the full-length protein compared to its CTD may be caused by interactions between TDP-43 CTD and its RRMs. TDP-43 is known to self-associate and bind other RNA binding proteins that contain tandem RRM domains, such as hnRNPA1 (which has a very high sequence similarity with Ddx4N1; Table 3.1), through its CTD (Wang et al. 2012; Buratti et al. 2005). This complementarity of CTDs may reduce the dynamics of the full-length protein within the droplet. Interestingly, the initial diffusion rates that we measured in this system were similar to previous reports measured in cell culture (Schmidt and Rohatgi 2016), suggesting that the model may suitably reflect

TDP-43 behavior in membraneless organelles. Using pre-formed droplets as a scaffold, the effect of ALS mutations or other compounds on TDP-43’s droplet behavior can be explored.

Phase separating properties of TDP-43 CTD as a driver for aggregation

An emerging theory of ALS pathogenesis suggests an intimate linkage between the formation of droplet organelles and aggregation. Indeed, phase separated droplets composed of purified proteins such as FUS/TLS (Fused in Sarcoma/Translocated in Sarcoma), hnRNPA1, and TDP-43 CTD all show intrinsic propensities to aggregate over time (Bentmann, Haass, and Dormann 2013; Li et al. 2013; Conicella et al. 2016). In the cell, SGs gradually separate into stable “core” structures surrounded by a mobile, liquid “shell”. Whereas proteins in the liquid shell rapidly exchange with the surrounding solution, the core undergoes interconversion with the shell phase in a slower, ATP-dependent process (Jain et al. 2016).

In our simple two-component system composed of yTDP-43 and Ddx4N1, we recapitulated the maturation of yTDP-43 within the droplet state. Over time, the mobility and diffusion rates of yTDP- 43 decreased while speckled aggregates formed within the droplets (Figure 3.7B). These speckled aggregates, like those formed in solution, did not recover after photobleaching (Figure S3.4), consistent with the notion that persistence of TDP-43 in the droplet state leads to the generation of rigid and stable aggregates. Our kinetic results show that the main contributor to yTDP-43 aggregation is environmental electrolyte concentrations. However, yTDP-43 aggregates formed within the Ddx4N1 droplets even under conditions where the yTDP-43 should not aggregate in solution (40 mM NaCl). It is unlikely that the aggregation is driven solely by the local concentration increase of yTDP- 43, since no aggregation occurred when the protein was stored in high concentration stock solutions under low salt conditions. This seems to suggest that the droplet phase may provide a site for aggregation by directly altering TDP-43 structure or by manipulating the local electrolyte concentration or water content. Under physiological salt concentrations, the aggregates that readily form in the solution phase have a strong affinity for the droplet-solution interface, suggesting a certain degree of surface complementarity between aggregates and droplets.

Although TDP-43 phase separation and aggregation appear as separate or distinct phenomenon, we showed that these processes have similar properties. They both appear to be 1) induced by physiological salt concentrations, 2) are readily reversible, and 3) dependent on the CTD of TDP-43. In our in vitro studies in the droplet and solution phases, we showed that the CTD is the key component

100

of the protein which is required for both solution phase aggregation and partitioning into droplet scaffolds (Figure 3.6C, S3.5). The effect of NaCl on the turbidity of a sample of full length protein is

very similar to that of the CTD construct reported in literature, both having an EC50 value around 150 mM NaCl. TDP-43 aggregates do not resemble phase separated droplets and show no recovery after FRAP (Figure S3.3). However, their capacity to fully return from aggregates to a non-turbid state after diluting into low-salt conditions has not been previously reported in TDP-43 aggregation and sets these aggregates apart from other protein aggregates seen in protein misfolding disease. This dynamic and reversible conversion between the two states is dependent on the CTD, as constructs without it did not aggregate. Moreover, the results suggest that the structural changes that must occur during aggregation are also dynamic and readily reversed, similar to the structural changes that occur during CTD phase separation. The similarities in the properties of full length aggregates and CTD droplets, the observation that both are driven by changes in electrolyte concentration, and the strong affinity between aggregates and phase-separated droplets may suggest that the two processes are occurring concurrently, or that the protein’s aggregation behavior is a byproduct of TDP-43’s phase separating properties. Overall, our results support the SG precursor hypotheses of TDP-43 aggregation.

Role of electrolytes in physiological conditions

Our results have identified a novel driver for TDP-43 aggregation, in that the aggregation and droplet- phase morphology of TDP-43 are affected by a unique order of electrolytes

+ + + 2+ 2+ (K

such as CaCl2 also had more drastic effects on yTDP-43 in the droplet state than electrolytes that induced aggregation at higher concentrations such as KCl. Na+ and K+ had markedly different effects on TDP-43 aggregation in solution, and similar but subtler effects on aggregation in the droplet phase. The presence of 150 mM NaCl appeared to increase the rate at which yTDP-43 formed aggregates in the droplet state. Notably, Ca2+, which induced aggregation at a mere 2 mM, caused the accumulation of yTDP-43 aggregates at the Ddx4N1 droplet-solution interface. This can either be attributed to

failure of yTDP-43 entry into droplets in the presence of 2 mM CaCl2, or rapid exclusion of yTDP- 43 to the droplet surface. In both cases, Ca2+ must have caused some structural change to the CTD of yTDP-43, which is responsible for both aggregation and droplet entry. We speculate that the structural changes are caused by charge-dipole interactions between electrolytes, peptide bond dipoles, and protein charged groups as predicted by the Kirkwood theory (Kirkwood 1943).

101

This exquisite control of yTDP-43 transition from solution to aggregates by electrolytes may have physiological implications, both in the function of TDP-43 and in its role in disease. Our salt- dependence assays of yTDP-43 aggregation suggests that TDP-43 would aggregate under cytosolic concentrations of salts, but the protein does not form obvious aggregates in healthy neurons. We proposed previously that TDP-43 is normally bound to nucleotide targets, which improves its solubility and prevents it from entering the aggregate state when in solution (Sun et al. 2014). Within the context of a droplet organelle such as SGs, we believe the same principle of native-state stabilization would apply, as RNA is also a natural component of SGs and is known to bind TDP-43 with specificity. In membraneless organelles, various other components (protein or otherwise) may also modulate TDP-43 behavior. ATP, for instance, is known to modulate the conversion from stable SG “cores” to fluid SG “shells” (Jain et al. 2016). The role of RNA/DNA binding targets on TDP- 43 droplet behavior were attempted unsuccessfully, due to undesirable interactions between the scaffolding protein and oligonucleotides. Specifically, Ddx4N1 droplets selectively uptake ssDNA and exclude dsDNA in its intrinsically disordered protein (IDP) matrix (Nott, Craggs, and Baldwin 2016). The addition of fluorescent-labeled RNA/DNA to the Ddx4N1 droplets complicated the analysis, as oligonucleotide bound yTDP-43 could not be differentiated from oligonucleotide bound Ddx4N1. Other proteins that have been suggested to form droplets such as Ras GTPase-activating protein- binding protein 1 (G3BP1), a component of the SGs to which TDP-43 directly interacts, may provide another suitable system for identifying modulators for TDP-43 behavior in the droplet phase in future studies (McDonald et al. 2011).

The effect of electrolytes on TDP-43 aggregation and phase behavior may also be relevant to its function and its role in disease propagation. In order to perform its various RNA processing roles, TDP-43 is shuttled between the nucleus and cytosol, where the levels of these ions are actively maintained in the neuron. The maintenance of a Na+/K+ gradient across the cell membrane is crucial for action potentials. But this gradient is also maintained between the cytosol and nuclear environments by Na+/K+ ATPases within the nuclear envelope (Fedorenko and Marchenko 2014; Palmer and Civan 1977). Although the exact levels of sodium and potassium in the two cellular compartments are unclear, numerous historical studies have proposed differences in ion levels between the two mediums of 40-100 mM (Century, Fenichel, and Horowitz 1970; Dick 1978; Moore and Morrill 1976; Paine et al. 1981). The sensitivity of TDP-43 to these ions poses a possible mechanism of TDP-43 structural and perhaps functional modulation through the protein’s unique electrolyte environments in different subcellular compartments.

102

The effect of electrolytes on TDP-43 may also provide insight into cellular spread of ALS/FTD. Based on our in vitro data, the electrolyte environment in the extracellular space containing high Na+ and Ca2+ would cause free TDP-43 to immediately aggregate. Upon neuron death, cytosolic and nuclearTDP- 43 within a cell can enter the extracellular space and aggregate under these conditions. This provides a local environment of aggregated TDP-43 that may interact and potentially induce TDP-43 aggregation within neighbouring, healthy neurons. This local spread of TDP-43 misfolding and formation of TDP-43 aggregates from extracellular TDP-43 has been well-documented in cell culture models (Furukawa et al. 2011; Smethurst, Sidle, and Hardy 2015). The precise mechanism of this spreading process and what structural changes must occur to the protein for cellular entry remains to be explored.

While current studies have focused on discerning the cellular functions of TDP-43, the precise mechanism of TDP-43 aggregation as seen in affected ALS/FTD neurons remains unclear. Here, we posit that the electrolyte environment of TDP-43 is a significant and physiologically relevant modulator of TDP-43 function, aggregation, phase separation, and potentially cell-to-cell spread.

103

Materials and Methods Protein expression and purification

Sequences encoding Venus yellow fluorescent protein (vYFP) and human TDP-43 were inserted into a pET-30 vector modified with kanamycin resistance gene using restriction enzymes according to

manufacturer’s instructions (New England Biolabs). The vector was constructed to contain a His6 tag, vYFP, a TEV-cleavage site, and the full length human TDP-43. Truncation to the full-length protein were generated using Q5 Site-directed mutagenesis kit (New England Biolabs) with appropriate primers. Plasmids were transformed into BL21-AI Star (DE3) competent cells (Invitrogen), and plated onto LB plates containing 35 µg/mL kanamycin (BioBasic) and incubated at 37 °C for 12 h. Colonies were inoculated into 50 mL of LB medium, incubated at 37 °C for 5 h and diluted into 1 L of LB medium. When the culture reached an absorbance of 0.4-0.6 at 600 nm the temperature was reduced to 19 °C and the culture was induced using 1 mM IPTG and 0.2% Arabinose. Cells were harvested 12 h after induction by centrifugation at 3000 rpm (Sorvall SLA3000 rotor) for 30 min at 4 °C. The cell pellet was resuspended in 50 mL Lysis Buffer (8 M urea, 500 mM KCl, 20 mM imidazole, 2 mM βME, 10% glycerol, 20 mM HEPES-KCl, pH 7.4) supplemented with complete EDTA-free protease inhibitor (Roche). Cells were lysed by sonication in 10 mL aliquots at 4 °C for 5 min (5 s pulse, 5 s stop) using a Vibracell sonicator (Sonics). The lysate was centrifuged for 30 min at 4 °C at 15,000 rpm

(Sorvall SS-34 rotor) to pellet cellular debris. Purification of the His6-tagged recombinant protein was achieved by Ni-NTA affinity chromatography using Ni-NTA beads (Sigma) at room temperature. The supernatant was applied to Ni-NTA beads at a ratio of 20:1 supernatant to bead volume. The beads were washed twice with 2 bead volumes of Lysis Buffer and eluted using 2 bead volumes of modified Lysis Buffer containing 250 mM imidazole and no protease inhibitors. The eluted protein in 8 M urea was flash frozen in liquid N2 in aliquots for storage. Immediately prior to experiments, the eluted protein was first dialyzed overnight against buffer containing 8 M urea, 10 mM HEPES, pH 7.4 to remove salts, followed by a second overnight dialysis against 10 mM HEPES, pH 7.4 to acquire protein stocks. Dialysis was performed at room temperature using Pur-A-Lyzer Dialysis Kits with

molecular weight cut-off of 6000 kDa (Sigma). To cleave the vYFP and His6-tag, the His6-tagged TEV protease was mixed with purified yTDP-43 protein at a 1:6 ratio of TEV to TDP-43 in 10 mM

HEPES, pH 7.4 and incubated for 24 h at room temperature. The His6-tagged TEV was replenished every 2 h in the first 12 h of the reaction. The mixture was reapplied to Ni-NTA beads to remove the

cleaved His6-tagged YFP and the remaining His6-tagged TEV protease. The cleaved TDP-43 protein was collected from the flow-through.

104

Dynamic light scattering measurements

Dynamic light scattering (DLS) measurements of hydrodynamic radius (Rh) were made at 20°C with a DynaPro DLS instrument (Protein Solutions) using a 12 µL quartz cuvette. Samples were measured using 5 s averaging time. Ten or more consecutive measurements were used for regularization analysis using DYNAMICS software. Particle translational diffusion coefficients were calculated from decay

curves of autocorrelation of light scattering data and converted to Rh with the Stokes-Einstein

equation. Histograms of mass versus Rh were plotted using Dynamics data analysis software. For conditions containing molar excess of oligonucleotides with 12 tandem TG or AC repeats, oligos were synthesized by Integrated DNA Technologies (Coralville, USA), desalted, lyophilized, and verified by mass spectrometry. Oligos were dissolved in DEPC-treated milli-Q water immediately prior to use.

In vitro aggregation of vYFP-TDP-43

Stock yTDP-43 was diluted to final concentrations of 2-20 μM protein in 10 mM HEPES, pH 7.4 and various concentrations of electrolytes. The turbidity of the sample was measured by absorbance at 600 nm immediately after dilution for kinetic measurements, and up to 1 day after dilution for equilibrium measurements using a SpectraMax M5 plate reader. 200 μL samples were prepared in sealed 96-well clear-bottom plates and agitated in the plate reader for 5 s prior to measurements. For kinetic data, measurements were taken at 15 s intervals. For aggregation reversibility assays, pre-formed aggregates were prepared by diluting stock yTDP-43 to aggregating conditions (20 μM yTDP-43 in 200 mM NaCl, 10 mM HEPES, pH 7.4) for 1 h. These pre-formed aggregates were diluted to reduced NaCl concentrations and the turbidity was measured after 1 h.

Curve fitting of aggregation kinetics

For equilibrium measurements of the effect of electrolytes on sample turbidity, raw turbidity values

were normalized and the curves were fit to the equation (Equation 2.1) by non linear least-squares fitting using OriginPro 8.5 software to extract the EC50 and Hill coefficients (n). For fitting of kinetic aggregation data, curves were fit to a simple two-step model of protein

aggregation consisting of an initial nucleation step ( →) followed by a second propagation step

with continuous nucleation ( →), where k1 and k2 are respective rates of each step (Morris et al. 2008; Bentea, Watzky, and Finke 2017). The change in aggregation as a function of time is represented

by the equation 1 (Equation 3.2), where [B]t is the aggregated ⋅

105

species measured by turbidity, and [A] is the initial concentration of soluble protein. Fitting the kinetic data to this equation yields apparent rates of nucleation and propagation, k1 and k2.

Transmission electron microscopy

Ammonium acetate-induced yTDP-43 aggregates were imaged using a Hitachi microscope operating at 80 kV. Samples of 2 μM protein in 0-150 mM NH4OAc were resuspended and deposited on fresh continuous carbon films prepared from copper rhodium grids (Electron Microscopy Sciences). Grids were charged using a glow discharger for 5 s at 3.5 mA negative discharge before sample addition. Aggregate solutions (2 μM) were adsorbed to grids for 60 s and blotted using No. 2 Whatman Filter paper and stained with freshly filtered 2% uranyl acetate for 8 s.

Immunofluorescence staining of formalin-fixed human brain tissue

Formalin-fixed brain blocks of the orbitofrontal gyri in a case of frontotemporal dementia with TDP- 43-positive inclusions (~2 days fixation) were run through 10%, 20%, and 30% sucrose gradients, infiltrated with optimal cutting temperature (OCT) compound, frozen, and cut into 10 μm thick sections and attached to glass microscope slides. Slides were photobleached for 48 h using a commercial LED lamp as previously described (Sun, Ip, and Chakrabartty 2017; Sun and Chakrabartty 2016). The slides were treated in antigen retrieval buffer (10 mM Citric Acid, 2 mM EDTA, 0.05% Tween 20; pH 6.2) for 30 min at 90 °C, then washed twice for 5 min in TBS plus 0.025% Triton X-

100 (TBS-Triton). The slides were then stained with 1% ThS in ddH2O for 5 min and washed 2 × 5 min with 70% ethanol to remove excess dye. The slides were then blocked in 10% normal goat serum (Aurion, Wageningen, the Netherlands) and 1% bovine serum albumin (BSA; ThermoFisher Scientific, Waltham, Massachusetts, USA) in TBS-Triton for 2 h at room temperature. Primary rabbit anti-TDP-43 antibody (1:50; ProteinTech) in 1% BSA–TBS was applied overnight at 4 °C in a humidified chamber. The slides were rinsed 2 × 5 min with TBS-Triton and Texas Red goat anti- rabbit (1:100; ThermoFisher Scientific) secondary antibody in 1% BSA–TBS was applied for 1 h at room temperature in the dark. The slides were rinsed 2 × 5min in TBS and stained with DAPI (0.25 μg/mL; ThermoFisher Scientific) for 10 min. Finally, the sections were washed 2 × 5 min with TBS and coverslipped using Immu-mount aqueous mounting medium (ThermoFisher). The sections were visualized using LSM710 confocal microscope (Zeiss) with ZenBlack software, using the following excitation sources for each fluorophore; ThS (λex =488 nm, λem = 493–605 nm); Texas Red (λex = 561

nm, λem = 566–689 nm); DAPI (λex = 405 nm; diode 405 laser, λem = 410–507 nm).

106

In vitro Thioflavin T fluorescence

To generate Aβ fibrils, lyophilized Aβ42 (0.1 mg) was dissolved in 100 µL of HFIP overnight and

resuspended step-wise, in 20 µL of DMSO, followed by 480 μL of 10 mM NaPO4, pH 7.6. The sample was sonicated for 10 min and agitated in a shaker at 900 rpm for 72 hours at room temperature. For ThT measurements, Aβ42 fibrils (5 µM), soluble TDP-43 (5 µM, 10 µM HEPES, pH 7.4), and TDP- 43 aggregates (5 µM, 200 mM NaCl, 10 µM HEPES, pH 7.4, 1 h incubation) were mixed with Thioflavin T (final concentration of 20 μM) for 20 min in a clear bottom 96-well plate. The fluorescence intensity was measured after 20 min using excitation and emission wavelengths of 450 nm and 490 nm, respectively, by a SpectraMax M5 plate reader. Measurements were made by averaging 6 reads per well. Relative fluorescence was calculated by subtracting a ThT-only blank (20 µM in 10 mM HEPES buffer). All samples were prepared in duplicate.

Circular Dichroism Spectroscopy

Purified yTDP-43 was diluted to a final concentration of 5 μM protein in high salt (130 mM NaCl, 10 mM HEPES, pH 7.4) to initiate aggregation and immediately measured by CD. Far-UV CD spectra were recorded with an Aviv CD spectrometer (model 62DS) at 25 °C every 10 min for 6 h in a 2 mm quartz cuvette. Spectra were obtained from 208 to 280 nm (1 mm path length, 1 nm step size, 1 nm bandwidth, and 16 s averaging time).

Fluorescence microscopy

For real-time visualization of yTDP-43 aggregation or disaggregation, protein samples (200 μL) were prepared in 8-well chambered coverglass (ThermoFisher) pre-mounted to the microscope stand. Dilution of protein was performed in the wells by gentle pipetting and the sample was imaged immediately using a LSM700 inverted fluorescence microscope (Zeiss) with a 20x Plan-Apochromat 20x/0.8 NA objective. YFP fluorescence was acquired by excitation at 488 nm using a solid-state laser. Images were acquired through a time-series at 1 s intervals and processed using LSM Zen 2012 software. To visualize interactions of yTDP-43 with pre-formed Ddx4N1 droplets, Ddx4N1 was expressed in E. coli according to established methods (Brady et al. 2017) and prepared by overnight dialysis into 10 mM HEPES, 20 μM βME, pH 7.4. This stock of Ddx4N1 was flash-frozen in 10 μL aliquots for storage and thawed immediately prior to imaging. For each 10 μL sample of Ddx4N1 + yTDP-43 mixture under varying buffer conditions, stock 300 μM Ddx4N1 was mixed with 8 μL of buffer solution by pipetting to emulsify the mixture and allowed to equilibrate at room temperature

107

for 15 min. 1 μL of 20 μM yTDP-43 was then added to the Ddx4N1 droplets and mixed by pipetting. 5 min after mixing, the sample was mounted on a glass microscope slide using Secure-Seal Imaging Spacers (Grace) and a no. 2 18 mm square cover glass (VWR) and imaged immediately using LSM700 inverted confocal microscope (Zeiss) as described above. In addition to YFP fluorescence, simultaneous brightfield images were obtained through differential interference contrast (DIC) microscopy using a brightfield photomultiplier tube (PMT). Fluorescence levels and gain settings across multiple experiments were kept constant using Zen 2012 software. For orthogonal views of droplet morphology, a 3-dimensional image of droplets under varying buffer conditions was generated by a set of z-stack scans of droplets imaged by YFP fluorescence with Zen 2012 software.

Quantitation of yTDP-43 insertion into Ddx4 droplets

To determine free energy of partitioning (ΔGpartition) within a droplet, confocal fluorescence images of yTDP-43 and Ddx4N1 droplets were analyzed using ImageJ (Schneider, Rasband, and Eliceiri 2012) to determine average pixel intensity within a droplet and compared against the solution phase

background. Partitioning coefficients (Kpartition) was defined in accordance with previous studies as

, which was calculated from the ratio of , and the free energy of

partitioning was calculated using the following equation: ∆ (Equation 3.3; Nott et al. 2015). Particles greater than 10 μm were selected using particle picking algorithm in ImageJ to negate artifacts associated with imaging particles too similar in size to the focal plane depth.

Average ΔGpartition was calculated for each yTDP-43 construct.

Fluorescence recovery after photobleaching (FRAP)

FRAP experiments were carried out on particles of 10-12 μm in diameter by photobleaching a circular region that encompasses the droplet using a LSM700 inverted fluorescence confocal microscope (Zeiss). Bleaching was done using a 488 nm solid state laser at 100% power for 50 iterations at 0.1 s bleaching time per iteration to completely bleach the particle. The particle was imaged over time by a time-series at 1 s intervals, including 5 s prior to bleaching, and up to 5 min after bleaching. The resulting images were analyzed using ImageJ to quantify the average fluorescence within the droplet as a function of time. The data was normalized to the initial unbleached fluorescence intensity values and fitted to the exponential decay function 1 (Equation 3.4) where A is the mobile

. fraction and k is the decay rate. The diffusion rate D is calculated by the Equation 3.1( ) /

108

where R is the radius of the bleached region and t1/2 is the half time of decay calculated from k (Soumpasis 1983). The average and standard deviation of diffusion rate and mobile fraction were calculated from three particles under each experimental condition.

109

Chapter Remarks Acknowledgement

The authors thank Dr. Jacob Brady from Dr. Lewis Kay’s group for generously providing a sample of purified Ddx4N1 for use in our experiments. We thank Drs. J. Brady, N.R. Cashman, J. Forman-Kay, D.V. Laurents, and S. Padmanabhan Iyer for their critical evaluation of this work. This work was supported by a Hudson Translational Team Grant, the Canadian Consortium of Neurodegeneration and Aging (CCNA), Canada: ALS Transformational Research grant from the Weston Brain Institute, and the Alzheimer Society of Canada.

Supplemental material

Figure S3.1: Negative stain EM of TDP-43 aggregate induced by NaCl and NH4OAc. Aggregates of yTDP-43 (2 µM) induced for 1h by 80 mM NaCl or 80 mM NH4OAc were deposited on fresh continuous carbon films prepared from copper rhodium grids and imaged by negative stain EM using uranyl acetate. For NaCl- treated yTDP-43 samples, excess NaCl was washed off by washing the grids several times in Milli-Q water. The aggregates produced are morphologically similar. Scale bar = 500 nm.

110

Figure S3.2: Complex yTDP-43 structures induced by freeze-thaw treatment. Purified yTDP- 43 stocks in 10 mM HEPES, pH 7.4 was induced to aggregate immediately (untreated) or flash frozen in dry-ice ethanol bath for storage and thawed prior to use (freeze-thaw). Aggregates formed within 1 h under conditions of 10 mM yTDP-43 in 100 µM NaCl, 10 mM HEPES, pH 7.4. The freeze-thaw step induced yTDP-43 to form droplet-like structures that appeared rigid and did not fuse, while fresh yTDP-43 formed aggregates that resembled loose tufts. Scale bar = 50 µm.

Figure S3.3: Aggregates of yTDP-43 do not recover after photobleaching. Aggregates of yTDP-43 induced by 100 mM NaCl as previously described from fresh or freeze-thawed samples were subjected to FRAP. The yellow dotted region represents region of bleaching. In both cases, no recovery of fluorescence was observed. Scale bar = 10 µm

111

Figure S3.4: Aggregates of yTDP-43 within Ddx4N1 droplets do not recover after photobleaching. Photobleaching of yTDP-43 aggregates within a Ddx4N1 droplet that has persisted for 3 h under high salt conditions (120 mM NaCl, 30 mM KCl, 10 mM HEPES, pH 7.4). The entire droplet was bleached as described in experimental procedures. Recovery of the fluorescence within the droplets was marginal. The brightly fluorescent aggregates near the droplet-solution interface did not regain its initial fluorescence, but the signal within the droplet recovered after 1-5 min as a diffuse, evenly distributed fraction, suggesting continuous exchange of soluble yTDP-43 into the Ddx4N1 droplet phase.

Figure S3.5: TDP-43 aggregation is dependent on the CTD. Constructs consisting of YFP-tagged TDP-43 fragments and truncations (Figure 3.6C) were induced to aggregate as described in the main text using 150 mM NaCl and 20 µM protein. Constructs that did not contain the CTD failed to produce any aggregates within 1 h. Scale = 50 µm.

112

References

Afroz, Tariq, Eva-Maria Hock, Patrick Ernst, Chiara Foglieni, Melanie Jambeau, Larissa A. B. Gilhespy, Florent Laferriere, et al. 2017. “Functional and Dynamic Polymerization of the ALS- Linked Protein TDP-43 Antagonizes Its Pathologic Aggregation.” Nature Communications 8 (1): 45. doi:10.1038/s41467-017-00062-0.

Banani, Salman F., Allyson M. Rice, William B. Peeples, Yuan Lin, Saumya Jain, Roy Parker, and Michael K. Rosen. 2016. “Compositional Control of Phase-Separated Cellular Bodies.” Cell 166 (3): 651–663. doi:10.1016/j.cell.2016.06.010.

Beauchamp, David L., and Mazdak Khajehpour. 2012. “Studying Salt Effects on Protein Stability Using Ribonuclease t1 as a Model System.” Biophysical Chemistry 161: 29–38. doi:10.1016/j.bpc.2011.11.004.

Bentea, Lucian, Murielle A. Watzky, and Richard G. Finke. 2017. “Sigmoidal Nucleation and Growth Curves Across Nature Fit by the Finke-Watzky Model of Slow Continuous Nucleation and Autocatalytic Growth: Explicit Formulas for the Lag and Growth Times Plus Other Key Insights.” Journal of Physical Chemistry C 121 (9): 5302–5312. doi:10.1021/acs.jpcc.6b12021.

Blechingberg, Jenny, Yonglun Luo, Lars Bolund, Christian Kroun Damgaard, and Anders Lade Nielsen. 2012. “Gene Expression Responses to FUS, EWS, and TAF15 Reduction and Stress Granule Sequestration Analyses Identifies FET-Protein Non-Redundant Functions.” PLoS ONE 7 (9). doi:10.1371/journal.pone.0046251.

Brady, Jacob P., Patrick J. Farber, Ashok Sekhar, Yi-Hsuan Lin, Rui Huang, Alaji Bah, Timothy J. Nott, et al. 2017. “Structural and Hydrodynamic Properties of an Intrinsically Disordered Region of a Germ Cell-Specific Protein on Phase Separation.” Proceedings of the National Academy of Sciences 114 (39): E8194–8203. doi:10.1073/pnas.1706197114.

113

Castrillon, D. H., B. J. Quade, T. Y. Wang, C. Quigley, and C. P. Crum. 2000. “The Human VASA Gene Is Specifically Expressed in the Germ Cell Lineage.” Proceedings of the National Academy of Sciences 97 (17): 9585–9590. doi:10.1073/pnas.160274797.

Century, T J, I R Fenichel, and S B Horowitz. 1970. “The Concentrations of Water, Sodium and Potassium in the Nucleus and Cytoplasm of Amphibian Oocytes.” Journal of Cell Science 7 (1): 5–13.

Chang, Chung Ke, Ming Hui Chiang, Elsie Khai Woon Toh, Chi Fon Chang, and Tai Huang Huang. 2013. “Molecular Mechanism of Oxidation-Induced TDP-43 RRM1 Aggregation and Loss of Function.” FEBS Letters 587 (6): 575–582. doi:10.1016/j.febslet.2013.01.038.

Colombrita, Claudia, Eleonora Zennaro, Claudia Fallini, Markus Weber, Andreas Sommacal, Emanuele Buratti, Vincenzo Silani, and Antonia Ratti. 2009. “TDP-43 Is Recruited to Stress Granules in Conditions of Oxidative Insult.” Journal of Neurochemistry 111 (4): 1051–1061. doi:10.1111/j.1471-4159.2009.06383.x.

Dick, D A. 1978. “The Distribution of Sodium, Potassium and Chloride in the Nucleus and Cytoplasm of Bufo Bufo Oocytes Measured by Electron Microprobe Analysis.” The Journal of Physiology 284 (6465): 37–53. doi:10.1038/367745a0.

Fedorenko, Olena A., and Sergey M. Marchenko. 2014. “Ion Channels of the Nuclear Membrane of Hippocampal Neurons.” Hippocampus 24 (7): 869–876. doi:10.1002/hipo.22276.

114

Karpowicz, Richard J, Conor M Haney, Tiberiu S Mihaila, Raizel M Sandler, E James Petersson, and Virginia M-Y Lee. 2017. “Selective Imaging of Internalized Proteopathic α-Synuclein Seeds in Primary Neurons Reveals Mechanistic Insight into Transmission of Synucleinopathies.” The Journal of Biological Chemistry 292: jbc.M117.780296. doi:10.1074/jbc.M117.780296.

Kedersha, Nancy L, Mita Gupta, Wei Li, Ira Miller, and Paul Anderson. 1999. “RNA-Binding Proteins TIA-1 and TIAR Link the Phosphorylation of eIF-2 Alpha to the Assembly of Mammalian Stress Granules.” The Journal of Cell Biology 147 (7): 1431–42.

115

Kirkwood, J. G. 1943. Proteins, amino acids and peptides as ions and dipolar ions. New York:Reinhold; pp 586-622.

Li, Y. R., O. D. King, J. Shorter, and A. D. Gitler. 2013. “Stress Granules as Crucibles of ALS Pathogenesis.” The Journal of Cell Biology 201 (3): 361–72. doi:10.1083/jcb.201302044.

Lukavsky, Peter J, Dalia Daujotyte, James R Tollervey, Jernej Ule, Cristiana Stuani, Emanuele Buratti, Francisco E Baralle, Fred F Damberger, and Frédéric H-T Allain. 2013. “Molecular Basis of UG-Rich RNA Recognition by the Human Splicing Factor TDP-43.” Nat Struct Mol Biol. 20 (12): 1443–1449. doi:10.1038/nsmb.2698.

Majoor-Krakauer, D, P J Willems, and a Hofman. 2003. “Genetic Epidemiology of Amyotrophic Lateral Sclerosis.” Clinical Genetics 63 (2): 83–101.

116

Moore, R. D., and G. A. Morrill. 1976. “A Possible Mechanism for Concentrating Sodium and Potassium in the Cell Nucleus.” Biophysical Journal 16 (5): 527–533. doi:10.1016/S0006- 3495(76)85707-4.

Morris, Aimee M., Murielle A. Watzky, Jeffrey N. Agar, and Richard G. Finke. 2008. “Fitting Neurological Protein Aggregation Kinetic Data via a 2-Step, minimal/‘Ockham’s Razor’ model: The Finke-Watzky Mechanism of Nucleation Followed by Autocatalytic Surface Growth.” Biochemistry 47 (8): 2413–2247. doi:10.1021/bi701899y.

Nott, Timothy J., Timothy D. Craggs, and Andrew J. Baldwin. 2016. “Membraneless Organelles Can Melt Nucleic Acid Duplexes and Act as Biomolecular Filters.” Nature Chemistry 8 (6): 569–575. doi:10.1038/nchem.2519.

Nott, Timothy J., Evangelia Petsalaki, Patrick Farber, Dylan Jervis, Eden Fussner, Anne Plochowietz, Timothy D. Craggs, et al. 2015. “Phase Transition of a Disordered Nuage Protein Generates Environmentally Responsive Membraneless Organelles.” Molecular Cell 57 (5): 936–947. doi:10.1016/j.molcel.2015.01.013.

Paine, Philip L, Terry W Pearson, L J Tluczek, and Samual B Horowitz. 1981. “Nuclear Sodium and Potassium.” Nature 291 (5812): 258–259. doi:10.1007/s13398-014-0173-7.2.

Palmer, Lawrence G, and Mortimer M Civan. 1977. “Distribution of Na +, K + and C1-between Nucleus and Cytoplasm in Chironomus Salivary Gland Cells.” J. Membrane Biol 33: 41–61.

Payne, R. B., and M. J. Levell. 1968. “Redefinition of the Normal Range for Serum Sodium.” Clinical Chemistry 14 (2): 172–78.

117

Ratti, Antonia, and Emanuele Buratti. 2016. “Physiological Functions and Pathobiology of TDP-43 and FUS/TLS Proteins.” Journal of Neurochemistry 138 (S1):95–111. doi:10.1111/jnc.13625.

Payne, R. B., and M. J. Levell. 1968. “Redefinition of the Normal Range for Serum Sodium.” Clinical Chemistry 14 (2): 172–78.

Roberts, CJ. 2007. “Non-Native Protein Aggregation Kinetics.” Biotechnology and Bioengineering 98 (5): 927–38. doi:10.1002/bit.

Rosen, Daniel R DR R, Teepu Seddique, David Patterson, Robert H. Brown Jr, T Siddique, David Patterson, D A Figlewicz, et al. 1993. “Mutations in Cu/Zn Superoxide Dismutase Gene Are Associated with Familial Amyotrophic Lateral Sclerosis.” Nature 362 (6415): 59–62. doi:10.1038/362059a0.

Sampathu, D. M., M. Neumann, L. K. Kwong, T. T. Chou, M. Micsenyi, A. Truax, J. Bruce, M. Grossman, J. Q. Trojanowski, and V. M. Lee. 2006. “Pathological Heterogeneity of Frontotemporal Lobar Degeneration with Ubiquitin-Positive Inclusions Delineated by Ubiquitin Immunohistochemistry and Novel Monoclonal Antibodies.” The American Journal of Pathology 169 (4): 1343–1352. doi:10.2353/ajpath.2006.060438.

Schmidt, Hermann Broder, and Rajat Rohatgi. 2016. “In Vivo Formation of Vacuolated Multi-Phase Report In Vivo Formation of Vacuolated Multi-Phase Compartments Lacking Membranes.” CellReports 16 (5): 1–9. doi:10.1016/j.celrep.2016.06.088.

Schneider, Caroline a, Wayne S Rasband, and Kevin W Eliceiri. 2012. “NIH Image to ImageJ: 25 Years of Image Analysis.” Nature Methods 9 (7): 671–675. doi:10.1038/nmeth.2089.

Scholtz, M.J., Eunice J. York, John M. Stewart, and Robert L. Baldwin. 1991. “A Neutral, Water- Soluble, α-Helical Peptide: The Effect of Ionic Strength on the Helix-Coil Equilibrium.” Journal of the American Chemical Society 113 (13): 5102–4. doi:10.1021/ja00013a079.

Schüller, Roland, and Dirk Eick. 2016. “Getting Access to Low-Complexity Domain Modifications.” Trends in Biochemical Sciences 41 (11): 894–97. doi:10.1016/j.tibs.2016.05.010.

Sipe, Jean D., and Alan S. Cohen. 2000. “Review: History of the Amyloid Fibril.” Journal of Structural Biology 130 (2–3): 88–98. doi:10.1006/jsbi.2000.4221.

Smethurst, Phillip, Katie Claire Louise Sidle, and John Hardy. 2015. “Review: Prion-like Mechanisms of Transactive Response DNA Binding Protein of 43 kDa (TDP-43) in Amyotrophic Lateral Sclerosis (ALS).” Neuropathology and Applied Neurobiology 41 (5): 578–597. doi:10.1111/nan.12206.

Soumpasis, D.M. 1983. “Theoretical Analysis of Fluorescence Photobleaching Recovery Experiments.” Biophysical Journal 41 (1): 95–97. doi:10.1016/S0006-3495(83)84410-5.

118

Sun, Yulong, and Avi Chakrabartty. 2016. “Cost-Effective Elimination of Lipofuscin Fluorescence from Formalin-Fixed Brain Tissue by White Phosphor Light Emitting Diode Array.” Biochemistry and Cell Biology = Biochimie et Biologie Cellulaire 94 (6): 545–50. doi:10.1139/bcb-2016-0125.

Sun, Yulong, Philbert Ip, and Avijit Chakrabartty. 2017. “Simple Elimination of Background Fluorescence in Formalin-Fixed Human Brain Tissue for Immunofluorescence Microscopy.” Journal of Visualized Experiments 127 (September). doi:10.3791/56188.

Tandan, R, and W G Bradley. 1985a. “Amyotrophic Lateral Sclerosis: Part 1. Clinical Features, Pathology, and Ethical Issues in Management.” Annals of Neurology 18 (3): 271–280. doi:10.1002/ana.410180302.

Tandan, R, and W G Bradley. 1985b. “Amyotrophic Lateral Sclerosis: Part 2. Etiopathogenesis.” Annals of Neurology 18 (4): 419–431. doi:10.1002/ana.410180402.

Valentin-Vega, Yasmine A., Yong Dong Wang, Matthew Parker, Deanna M. Patmore, Anderson Kanagaraj, Jennifer Moore, Michael Rusch, et al. 2016. “Cancer-Associated DDX3X Mutations Drive Stress Granule Assembly and Impair Global Translation.” Scientific Reports 6 (December 2015): 1–16. doi:10.1038/srep25996.

Vernon, Robert McCoy, Paul Andrew Chong, Brian Tsang, Tae Hun Kim, Alaji Bah, Patrick Farber, Hong Lin, and Julie Deborah Forman-Kay. 2018. “Pi-Pi Contacts Are an Overlooked Protein Feature Relevant to Phase Separation.” eLife 7 (February): 1–90. doi:10.7554/eLife.31486.

CHAPTER IV DETERMINING COMPOSITION OF MICRON-SCALE PROTEIN DEPOSITS IN NEURODEGENERATIVE DISEASE BY SPATIALLY TARGETED OPTICAL MICROPROTEOMICS (STOMP)

This chapter first appeared in eLife as: K.C. Hadley, R. Rakhit, H. Guo, Y. Sun, J.E.N. Jonkman, J. McLaurin, L.-N. Hazrati, A. Emili, and A. Chakrabartty. (2015). Determining Composition of Micron- Scale Protein Deposits in Neurodegenerative Disease by Spatially Targeted Optical Microproteomics. eLife 4 (September): 1–21. It was written by K.H. with revisions from A.C. The STOMP technique was conceptualized by K.H. and R.R. and developed by K.H. over the span of several years. K.H. developed the STOMP macro (based on early work by J.J.), mass spectrometry was performed by H.G. and human immunohistochemistry was performed by L.H. Y.S performed STOMP analysis on archived human tissue and validated the STOMP hits by immunofluorescence microscopy. This chapter serves as an introduction to the STOMP technique and showcases its potential in neurodegenerative disease research. Subsequent chapters will discuss the refinement of STOMP and its use as a tool to unravel the mechanisms of ALS/FTD proteinopathies.

119 120

Chapter abstract

Spatially targeted optical microproteomics (STOMP) is a novel proteomics technique for interrogating micron-scale regions of interest (ROIs) in mammalian tissue, with no requirement for genetic manipulation. Methanol or formalin-fixed specimens are stained with fluorescent dyes or antibodies to visualize ROIs, then soaked in solutions containing the photo-tag: 4-benzoylbenzyl-glycyl- hexahistidine. Confocal imaging along with two-photon excitation are used to covalently couple photo-tags to all proteins within each ROI, to a resolution of 0.67 μm in the xy-plane and 1.48 μm axially. After tissue solubilization, photo-tagged proteins are isolated and identified by mass spectrometry. As a test case, we examined amyloid plaques in an Alzheimer’s disease (AD) mouse model and a post-mortem AD case, confirming known plaque constituents and discovering new ones. STOMP can be applied to various biological samples including cell lines, primary cell cultures, ex vivo specimens, biopsy samples, and fixed post-mortem tissue.

Abbreviations used in this chapter

STOMP, spatially targeted optical microproteomics; ROI, region of interest; 6HisBP, 4- benzoylbenzyl-glycyl-hexahistidine; AD, Alzheimer’s; ThS, Thioflavin S; Aβ, amyloid β peptide; TDP- 43, transactive response DNA binding protein 43 kDa; FTD, frontotemporal dementia; ALS, amyotrophic lateral sclerosis; LCM, laser capture microdissection; DEPC, diethyl pyrocarbonate; βME, β-mercaptoethanol; FWHM, full width at half-maximum; SAINT, Significance Analysis of INTeractome; VAMP2, vesicle-associated membrane protein 2; EAAT, excitatory amino acid transporter.

121

Introduction

Pathological protein deposits have a long history as hallmarks of neurodegenerative disease. Early methods used to identify these deposits include the silver stain introduced by Golgi (1873), and Virchow’s iodine–sulfuric acid stain for starch (Virchow 1854) that led to the coining of the term amyloid (Virchow 1855). More specialized stains and labels have emerged which have begun to probe the structure and composition of pathological protein deposits. The stains Congo red and thioflavin S (ThS) were discovered later and remain in use (reviewed by Sipe and Westermark 2005, Tanskanen 2013), specifically identifying deposits containing a particular structural motif: amyloid β-pleated sheets. Beyond mere detection lies comprehensive identification of the protein components of these deposits; biochemical analyses of the deposits have produced transformative results in the field of neurodegeneration research. Two classic examples include the discovery of the prion protein and formulation of the protein-only hypothesis of prion disease (Bolton et al. 1982), and the discovery of the Alzheimer amyloid peptide (Aβ; Glenner and Wong 1984) and formulation of the amyloid cascade hypothesis of Alzheimer’s disease (AD; reviewed by Musiek and Holtzman 2015). More recently, there was the discovery that the RNA/DNA binding protein TDP-43 (transactive response DNA binding protein 43 kDa) is a significant component of ubiquitin-positive intraneuronal inclusions in certain cases of frontotemporal dementia (FTD) and amyotrophic lateral sclerosis (ALS) (Neumann et al., 2006). This discovery helped establish these diseases as spectrum disorders (Mackenzie, 2007). It led to the identification of TDP-43 mutations that are causative for ALS and FTD (Sreedharan et al., 2008) and established a role for RNA metabolism in ALS/FTD pathogenesis (Lagier-Tourenne and Cleveland 2009; Mackenzie et al. 2010).

Identification of the protein components of pathological deposits has traditionally involved the partial biochemical purification of detergent-insoluble proteins present in a tissue specimen followed by protein sequencing. Immunohistochemical methods are then used to confirm that the identified proteins are bona fide constituents of the pathological deposits. These methods require large amounts of sometimes-scarce pathological tissue and identification of novel protein components can require increasingly elaborate, time-consuming, and costly protocols. For example, the discovery of TDP-43 in ubiquitin-positive intraneuronal inclusions required the generation of ∼1000 monoclonal antibodies, which were used to screen thousands of tissue sections by immunohistochemistry, to obtain a single monoclonal antibody that specifically labeled ubiquitin-positive intraneuronal inclusions (Neumann et al. 2006). That antibody was used for the proteomic identification of TDP-

122

43 from the detergent-insoluble fraction of pathological tissue homogenates, and confirmatory staining with commercial TDP-43 antibodies was required for validation. In addition to being costly and labor intensive, another major limitation of biochemical purification of protein deposits is that soluble protein components associated with the core aggregates are likely to be lost during fractionation. Identifying deposit-associated soluble proteins could further our understanding of disease mechanisms.

A different strategy that can preserve some of the deposit-associated soluble proteins is laser capture microdissection (LCM), in which protein deposits are lifted intact out of sections of the fixed specimen; their protein compositions are then determined by mass spectrometry (Gozal et al., 2006). The resolution of LCM is ∼10 μm in the horizontal plane and captures the entire thickness of the tissue section along the vertical axis. Thus, for features smaller than ∼10 μm—such as inclusion bodies, small amyloid plaques, narrow fibers, and other irregularly shaped structures—LCM analysis permits enrichment but not complete isolation of target proteins; samples are diluted with extraneous surrounding material. For example, when capturing even a relatively large—3 μm in diameter— inclusion body with a single 10 μm LCM spot, only ∼10% of the recovered protein is from the pathology and 90% from the surrounding cellular milieu. LCM has nevertheless proven useful, however, in the proteomic analysis of systemic amyloidosis, where the amyloid deposits are very large and the feature size is typically ∼50,000 μm2 in area (Sethi et al. 2012, 2013). Mass spectrometry imaging and its variants can also provide spatially resolved mass spectrometry analysis by directly coupling imaging with laser ablation and achieving 1 mm–10 μm resolution (Stoeckli et al. 2001; Wiseman et al. 2006; Wucher et al. 2007), but it is usually used to look for specific targets rather than in discovery mode.

Recently, proximity labeling techniques such as Bio-ID and APEX have provided high spatial resolution to mass spectrometry, and these techniques have been used to elucidate the composition of several difficult to purify organelles (Roux et al. 2012; Rhee et al. 2013; Firat-Karalar et al. 2014; Hung et al. 2014). However, proximity labeling relies on genetic manipulations to express an exogenous fusion protein to label adjacent components; this protein must be specifically and accurately targeted into a particular biological structure. Imperfect localization, especially in rare structures, results in selective labeling of neighbors outside of the target region yielding false positive associations. Many of these approaches have been brought to bear on amyloid (senile) plaques in AD brain tissue, in an attempt to identify core components, if any, in addition to Aβ. Early LCM studies

123 utilizing 2D gel electrophoresis coupled with mass spectrometry have detected a modest number of plaque-associated proteins, including an assortment of proteins associated with cell signaling, chaperone function, membrane trafficking, and proteolysis (Liao et al. 2004). A more recent LCM study has suggested that amyloid plaques in AD are highly homogenous structures that are composed almost exclusively of Aβ (Söderberg et al. 2006). The latter study, however, employed sample washing steps with 1% SDS, which may strip away soluble plaque-associated proteins. Finally, a recent attempt to capture more plaque-associated proteins potentially lost to LCM examined the sarkosyl-insoluble fraction of brain samples from patients with AD and appropriate controls (Gozal et al. 2009). Altered levels of eleven specific proteins were identified by this approach but colocalization with amyloid plaques was difficult to validate in many cases—not all insoluble proteins within the brain will necessarily originate from senile plaques. A recent proteomic analysis of fractions obtained from differential centrifugation procedures that yield fractions enriched in post-synaptic proteins have demonstrated that the post-synaptic protein IRSp53 was highly down-regulated in AD (Zhou et al. 2013). While extant LCM and biochemical fractionation studies have identified candidate plaque- associated proteins, significant mechanistic insights have not been produced.

We have developed a semi-automated technique called spatially targeted optical microproteomics (STOMP), which combines two-photon laser scanning microscopy with photochemical affinity labeling and mass spectrometry. While two-photon excitation has previously been used to drive microscopic, 3D-resolved photochemistry (Lee et al. 2008), STOMP represents the first application of this technique to affinity photolabeling, offering single-micron-scale three-dimensional resolution that is an order of magnitude better than the current state of the art LCM. In addition to providing clues to the etiology of neurodegenerative disease, this new proteomic technique also has great potential to advance research in cell biology by yielding the composition of small features that are not amenable to biochemical purification. We sought to create an unbiased discovery technique that would: (1) have high spatial resolution, (2) target any identifiable or user-defined spatial region in a biological sample, (3) forgo the need for genetic modification, (4) have sufficient sensitivity to detect minor species within the sample, and (5) have sufficient specificity such that most hits can be readily validated as being enriched in the biological structure of interest. Because STOMP does not require genetic manipulation, this technique is applicable to a large variety of biological samples that run the gamut from cell lines, primary cell cultures, ex vivo specimens, human biopsy samples to fixed post- mortem human tissue.

124

Results and discussion The STOMP technique

In STOMP, laser light from the microscope is used not only to image the fluorescently stained specimen, but also to photochemically crosslink affinity purification ligands to protein components within pathological deposits (Figure 4.1A). The photo-affinity ligand or photo-tag used here is a peptide of sequence: 4-benzoylbenzyl-Gly-His-His-His-His-His-His-CONH2 (6HisBP) (Figure 4.1B). Benzoylbenzene is a photoreactive group that forms covalent bonds with C–H and N–H groups upon excitation. The 6HisBP photo-tag molecule (molecular mass, 1105 Da) is of similar size to Congo red (molecular mass, 697 Da) and should have a similar ability to penetrate biological specimens. The specimen is fixed and permeabilized in methanol, then protein deposits in the specimen are selectively stained using antibodies or specific fluorescent dyes. The specimen is then soaked in a solution of 6HisBP to saturate the subcellular compartments with the photo-tag, and imaged using conventional confocal microscopy. The wavelength of the confocal laser is in the visible range and does not excite the UV-absorbing photo-tag. Optical sectioning by confocal microscopy is used to collect a 3D image of the protein deposits in the specimen. This image in turn is used to generate a list of spatial coordinates—a ‘mask’ file—identifying every point in the tissue containing pathological protein deposits (Figure 4.1C). The photo-tagging procedure uses two-photon microscopy (Skoch et al., 2006) rather than conventional confocal microscopy. Two-photon microscopy can excite a small volume (<1 μm3, Figures 4.1A and Figure 4.2) anywhere in the specimen, while conventional confocal microscopy excites a continuous cone that spans the entire thickness of the specimen, which would result in the covalent attachment of the photo-tag to off-target regions. A custom software macro was written that reads the mask file and selectively delivers two-photon excitation light to each protein deposit present, while leaving the remainder of the specimen untouched by excitation light. The STOMP macro provides a semi-automated tool for photo-tagging proteins in pathological deposits. Importantly, unlike LCM, no manual tracing of features is required with STOMP. We rely on reversible binding of photo-tagged proteins using nickel affinity chromatography. Histidine-rich proteins ubiquitous in all mammalian tissues represent potential contaminants that could bind the nickel affinity beads and confound the STOMP analysis. To circumvent this problem, the first step of the procedure involves limited treatment of the tissue sections with dilute diethyl pyrocarbonate (DEPC). DEPC covalently modifies histidyl residues that abolishes their ability to bind nickel (Wallis and Holbrook 1973). After completion of the photo-tagging step, the specimen is solubilized in buffer containing 2% SDS, 8 M urea, and β-mercaptoethanol (βME). The photo-tagged proteins are purified

125 using nickel affinity beads that bind the hexahistidyl moiety of 6HisBP. The purified proteins are then analyzed by gel electrophoresis (Figure 4.1D) and identified by database searching spectra from peptides resulting from digestion and liquid-chromatography and tandem mass spectrometry (LC- MS/MS).

Figure 4.1: Overview of STOMP technology. (A) Schematic diagram comparing volumes selected for excision by laser capture microdissection (LCM) (yellow cylinder) and spatially targeted optical microproteomics (STOMP) (blue volume). The STOMP excitation volume is limited approximately to the point spread function of a high numerical aperture lens (<1 μm in the xy plane, ∼1.0 μm axially) compared to ∼10 μm for LCM. (B) Structure of the bifunctional photocrosslinker used in this study; the affinity purification hexahistidine peptide is highlighted in blue, the photocrosslinker benzophenone group is highlighted in yellow. (C) Overview of the STOMP protocol. Tissue sections or cells are first fixed in cold methanol and stained using specific dyes or antibodies. (1) A guide image is acquired at a wavelength that does not activate the photocrosslinker. (2) The guide image is converted to a digital mask that directs the two-photon laser. (3) Two-photon laser selectively illuminates region of interest. The final image shown is anti-His tag immunofluorescence corresponding to areas of photo-tag crosslinking. (D) Silver stain of SDS-PAGE of material retrieved from selective STOMP of CRND8 mouse plaques (left panel) or non-plaque regions of an adjacent brain section (right panel). In each case the control (‘con’) is non-specific binding to Ni2+-NTA agarose beads. Load is 1% of total protein after solubilization of tissue. Note that the predominant protein photo-tagged in the STOMP sample is Aβ (arrow), which is not visible in the non-plaque experiment because of its low overall abundance.

126

The resolution of STOMP

To establish the spatial resolution limits of our photo-tagging procedure, we tagged single-voxel spots within sections of methanol-fixed murine brain tissue. By subsequently staining the tissue with an anti- hexahistidine antibody, we were able to use confocal immunofluorescence imaging to directly visualize and measure the extent of the photo-tagged volume. We measured the diameter (full width at half- maximum, FWHM) of the phototagged volume at 0.67 μm along the x and y axes, and 1.48 μm along the z axis (Figure 4.2). Taking the excited region to be an ellipsoid, the total volume of a single spot is 0.38 μm3.

Figure 4.2: The smallest photo-tagging volume is less than 0.5 μm3. A single voxel in a TgCRND8 tissue section was photo-tagged. The volume of photo-excitation was measured by confocal fluorescence imaging of the resulting photo-tagged spot. Maximum-intensity projections of the xy- (A, B) and xz- (C) planes are shown. Scale bar 1 μm. Fluorescence profiles of the indicated regions (shaded yellow) are shown (D–F). The width of the peaks is 0.67 μm in x and y, and 1.48 μm along the z axis, corresponding to an ellipsoidal excited volume of 0.38 μm3.

127

STOMP analysis of amyloid plaques in a transgenic mouse model of AD

We used TgCRND8 mice, a well-characterized transgenic mouse model of AD (Chishti et al. 2001), as a model system for the development of the STOMP technique. These mice express a human form of the amyloid precursor protein carrying two mutations associated with familial AD, and they produce amyloid plaques and exhibit spatial learning impairments by 3 months of age. This study used frozen sections (post-fixed in methanol) of the brains known to contain plaques, from TgCRND8 mice of 8 months of age. Sets of serial sections on separate slides were treated with DEPC, stained with ThS, and soaked in a solution of 6HisBP. Slides were imaged by confocal microscopy to identify ThS- positive amyloid deposits. Confocal images of ThS-positive amyloid deposits (Figure 4.1C1) were used to construct individual masks (Figure 4.1C2). Because our technique hinges on selective photolabeling and purification, we needed to assess the extent of non-specific labeling of 6HisBP in ambient light and under immunofluorescence excitation, as well as non-specific binding to affinity purification beads. Adjacent sections were put aside as ‘dark’ controls used to assess the extent of non-specific labeling of 6HisBP to proteins caused by confocal laser light (488 nm) exposure or other handling, and to assess nonspecific binding of proteins to nickel affinity beads. The STOMP macro was used to deliver two-photon excitation light to regions of the specimen corresponding to each pixel in the mask image. This excitation light has two effects. First, and most importantly, it photo-activates 6HisBP molecules that are in the amyloid deposits causing photo-tagging of constituent proteins. Second, it serendipitously photo-bleaches the ThS fluorophores present in regions of the specimen targeted by the mask. Immunofluorescence staining of the tissue section with anti-His6 antibody after photoactivation superimposes on the digital mask, thus highlighting the very high accuracy of targeting of the two-photon laser (Figure 4.1C3). STOMP combines microscopy with selective photo-labeling to accurately resolve, capture, and affinity-label highly irregular shaped micron-scale structures, via a semi-automated procedure. After solubilization of the specimen, the photo-tagged proteins were bound to nickel affinity beads. Each sample was divided into two portions: one used for mass spectrometry (Table 4.1) and one for gel electrophoresis and silver staining (Figure 4.1D). The dark control sample, which—aside from two-photon excitation—was treated identically to the STOMP sample, was run alongside the STOMP sample. It shows very few bands in the silver-stain gel from material bound to the nickel-nitrilotriacetic acid (Ni-NTA) beads compared to the STOMP sample, confirming that nonspecific photo-tagging and nonspecific binding of proteins to the nickel affinity beads is minimal. In addition to a number of proteins ranging in molecular weight from 20 kDa to >250 kDa, the STOMP sample contains large amounts of a low molecular weight protein that was

128 subsequently identified as Aβ (4.5 kDa) (Figure 4.1D). As an additional control, an entire brain section fixed in methanol and soaked with photo-tag was photo-activated by exposure to 365 nm ultraviolet light. Section-wide photo-activation of this specimen caused indiscriminate photo-tagging of proteins in the specimen. Gel electrophoresis of the indiscriminately photo-tagged proteins reveals a very different pattern of protein bands compared to the specimen in which amyloid plaques were specifically targeted for STOMP analysis (Figure 4.1D).

Identification of photo-tagged amyloid plaque proteins by mass spectrometry

The total volume of tissue that needs to be photo-tagged to obtain sufficient material for mass spectrometry analysis is a subjective matter and also depends on instrument sensitivity, fixation method used, sample complexity and perhaps other factors. Greater amounts of tissue will increase sensitivity and enable identification of lower abundance proteins, particularly in very complex samples containing a large number of unique proteins. In this study, we were able to identify proteins present in relative abundance in the pathological feature being examined using a total photo-tagged volume of approximately 2 × 106 μm3. (This corresponds to a photo-tagged region of 500 μm × 500 μm in an 8-μm thick tissue section or a cube of photo-tagged tissue 130 μm on a side.) Lowering the stringency will increase the number of hits but may introduce more false positives. Two frozen sections of TgCRND8 brain tissue were sufficient to produce the necessary volume of photo-tagged tissue required for one STOMP experiment, and the process required a total of 12–16 h of microscope time per sample (6–8 h per section). Based on our determination of the resolution of STOMP, regions of interest (ROIs) as small as 0.38 μm can be analyzed (Figure 4.2). To analyze such small features, however, sufficient numbers of ROIs must be captured, and this may necessitate pooling photo-tagged material from several tissue sections. Protein identification via mass spectrometry involved on-bead digestion of the photo-tagged proteins with endoproteinase Lys-C followed by trypsin and LC- MS/MS analysis. We performed technical and biological replicates of the proteomic analyses of our STOMP experiment. The technical replicates (Figure S4.1A) demonstrate a very high degree of correlation (r2 = 0.98), and the biological replicates demonstrate significant overlap in proteins identified (Figure S4.1B). The replicate studies illustrate the high reproducibility of the technique. The list of all identified proteins in amyloid plaques of TgCRND8 mice based on our high stringency criteria is summarized in Table 4.1. The aim of STOMP is to provide a compendium of annotated proteins present within the ROI under study—whether or not they are enriched or depleted relative to the surrounding tissue. As regards the quantitative nature of the data, Table 4.1 collects and summarizes our high-confidence mass spectrometry protein identifications, sorted by molecular-

129 weight-normalized spectral counts: a semi-quantitative proxy for protein abundance. As expected, the most abundant protein found in STOMP from TgCRND8 mouse plaques was Aβ (Figure 4.1D, Table 4.1). In addition to Aβ, we identified 454 proteins in at least one of two biological replicates. Combining our biological and technical replicates with our negative controls (‘dark’ samples and replicates), we used the Significance Analysis of INTeractome (SAINT) algorithm (Choi et al. 2011), which uses Bayesian inference to assign a probability of a true interaction, using Aβ as the bait protein. We found 62 proteins with probability >0.8 of being found associated with the plaques, and 146 with probability >0.5. Among these, synaptosomal proteins were highly enriched, including SNAP25, vesicle-associated membrane protein 2 (VAMP2), vesicular proton pump ATPase (v-ATPase), and synapsins 1 and 2. Apolipoprotein E, which is sometimes associated with senile plaques (Thal et al. 1997), was also detected, albeit at a SAINT score of 0.70. Certain common synaptic markers— including chromogranin A/C, synaptophysin, synaptogyrin I, synaptoporin, α-synuclein, and VGLUT1 and 2—were not detected in the plaques. Interestingly, the plaques were also devoid of uniquely post-synaptic proteins. We also identified many mitochondrial proteins; although the reason for this is unclear, it may be that the intrinsic hydrophobicity of mitochondrial membrane proteins may predispose them to associate with amyloid plaques. Comparison with proteins found from STOMP analysis of non-plaque areas revealed only limited overlap; VAMP2, synapsin 1, and dynamin 1 are present in both plaque and non-plaque STOMP analysis but are significantly greater enriched in the plaque STOMP samples.

Figure S4.1: Reproducibility of STOMP technique across technical and biological replicates. (A) Cross correlation of spectral counts for all protein identifications are shown for duplicate runs of STOMP analysis on human amyloid plaque material. (B) Biological replicates of STOMP analysis of amyloid plaque material from TgCRND8 mice show good repeatability. Using our thresholding criteria (at least three spectral counts for each identified protein, at least threefold excess over corresponding dark control) one sample run (P1) identified fewer total proteins than the other (P2), however most proteins identified in P1 were also detected in P2.

130

Table 4.1: Proteins statistically significantly enriched in the amyloid plaques of TgCRND8 mouse brain identified and retrieved by STOMP. Table is sorted in descending order of abundance, by normalized spectral counts. Protein Uniprot ID STOMP SAINT Previous reports of counts score enrichment in plaques (Söderberg et al., 2006) Amyloid beta A4 protein P12023 378 0.83 Detected, enriched Synaptosomal-associated protein 25 P60879 88 1.00 Detected, not enriched Cytochrome c1 Q9D0M3 70 0.98 Detected, not enriched Excitatory amino acid transporter 2 P43006 64 1.00 Detected, not enriched V-type proton ATPase subunit B P62814 56 0.80 Detected, enriched Vesicle-associated membrane protein 2 P63044 52 0.87 Detected, not enriched Pyruvate kinase isozymes M1/M2 P52480 49 0.98 Detected, not enriched Fructose-bisphosphate aldolase A P05064 47 1.00 Previously unreported Cytochrome b-c1 complex subunit 1 Q9CZ13 47 0.92 Previously unreported Guanine nucleotide-binding protein G(o) subunit P18872 41 0.98 Previously unreported alpha Cytochrome c oxidase subunit 2 P00405 40 0.97 Detected, not enriched Tubulin alpha-1B chain P05213 39 0.98 Detected, not enriched 4-aminobutyrate aminotransferase P61922 39 0.81 Previously unreported Tubulin beta-3 chain Q9ERD7 38 1.00 Detected, not enriched V-type proton ATPase catalytic subunit A P50516 37 0.99 Detected, enriched Aralar 1 Q8BH59 37 1.00 Previously unreported Citrate synthase, mitochondrial Q9CZU6 37 1.00 Detected, not enriched Clathrin heavy chain 1 Q68FD5 35 1.00 Detected, enriched Alpha-internexin P46660 30 0.98 Previously unreported NADH dehydrogenase 1 alpha subcomplex Q9DC69 30 0.98 Detected, not enriched subunit 9, mitochondrial Fructose-bisphosphate aldolase C P05063 28 0.91 Detected, not enriched Synapsin-2 Q64332 27 0.89 Detected, not enriched Dynamin-1 P39053 26 0.99 Detected, not enriched NADH dehydrogenase 1 alpha subcomplex Q99LC3 24 0.89 Detected, not enriched subunit 10, mitochondrial Spectrin alpha chain, brain P16546 22 1.00 Detected, not enriched Succinate dehydrogenase flavoprotein subunit, Q8K2B3 21 0.99 Detected, not enriched mitochondrial Synapsin-1 O88935 20 0.97 Detected, not enriched Rab GDP dissociation inhibitor alpha P50396 20 0.94 Detected, not enriched V-type proton ATPase 116 kDa subunit a Q9Z1G4 19 0.98 Detected, enriched Hexokinase-1 P17710 18 1.00 Detected, not enriched Heat shock protein HSP 90-alpha P07901 18 0.93 Detected, enriched Ubiquitin thioesterase OTUB1 Q7TQI3 18 0.85 Previously unreported Vesicle-fusing ATPase P46460 17 0.88 Previously unreported Spectrin beta chain, brain 1 Q62261 17 1.00 Detected, not enriched NADH-ubiquinone oxidoreductase 75 kDa Q91VD9 16 0.95 Detected, not enriched subunit, mitochondrial Neurofilament light polypeptide P08551 15 0.85 Detected, not enriched Neurochondrin Q9Z0E0 15 0.97 Detected, not enriched Heat shock protein HSP 90-beta P11499 14 0.98 Detected, enriched Na/K-transporting ATPase subunit alpha-2 Q6PIE5 14 0.95 Detected, not enriched Excitatory amino acid transporter 1 P56564 14 0.93 Detected, not enriched Microtubule-associated protein 6 Q7TSJ2 13 0.87 Detected, not enriched Serine/threonine-protein phosphatase 2A 65 kDa Q76MZ3 11 0.80 Detected, not enriched regulatory subunit A alpha isoform Catenin beta-1 Q02248 11 0.83 Detected, not enriched Tenascin-R Q8BYI9 10 0.85 Detected, not enriched

131

Validation of the STOMP results in TgCRND8 mice with immunofluorescence and immunohistochemistry

Using confocal microscopy, we examined colocalization of ThS-positive plaques with immunfluorescently labeled Aβ, SNAP25, VAMP2, v-ATPase, synapsin 1, and ApoE. All were found to be present and are apparently genuine components of the amyloid plaque (Figure 4.3). Optical sectioning demonstrates the presence of each of these proteins in the core of the amyloid plaques suggesting they were incorporated at the earliest stages of plaque formation. Staining of SNAP25, VAMP2, and v-ATPase reveal an immunopositive halo surrounding each of the plaques indicating an accumulation of these proteins not only in the amyloid core but also in the vicinity of the plaques; interestingly, synapsin 1 immunostaining did not have this halo. As mentioned above, certain common synaptic markers were not detected by the STOMP analysis. Immunostaining of some of these markers, syntaxin 1, synaptophysin, and α-synuclein, confirmed that these proteins are not concentrated in the amyloid plaques (Figure 4.4). α-Synuclein has been reported to be associated with senile plaques in human patients with AD; however, TgCRND8 mice are known to not accumulate α-synuclein in their plaques (Xu et al. 2002). It is not clear why the amyloid deposits contain some pre- synaptic proteins, but not others, and are devoid of post-synaptic proteins. One possibility is that damaged dystrophic neurites that are known to surround the plaques (Chishti et al. 2001) are spilling their contents and the physical properties of certain pre-synaptic proteins cause them to bind with especially high affinity to amyloid plaques. In this regard, it is noteworthy that many of these proteins are membrane proteins with coiled coil domains (Takamori et al. 2006). Another possibility, which is not mutually exclusive and provides an explanation for the presence of pre-synaptic proteins in plaques, is that Aβ fibrils and/or oligomers bind directly to synaptic vesicles and interfere with synaptic function. Evidence for a damaging effect of Aβ on synapses and dendrites has been available for some time (Mattson et al. 1998). Diffusible Aβ oligomers added to cultured primary neurons have been shown to associate with synaptosomes and a concomitant synaptic deterioration is observed (Lacor et al., 2007). In post-mortem AD tissue and in transgenic AD mice, an inverse correlation between levels of Aβ oligomers and levels of synaptic proteins was noted (Pham et al. 2010). The reduction in the levels of detectable synaptic protein in AD tissue may be the result of the aggregation and entrapment of the synaptic proteins in the amyloid plaques. In sum, the STOMP technique has provided clear evidence of interaction of amyloid plaques with pre-synaptic proteins.

132

Figure 4.3: Immunofluorescent confirmation of synaptic proteins in amyloid plaques of TgCRND8 mice. Colocalization of ThS-stained plaques with beta-amyloid positive control (A, G, M) and the synaptic proteins ATP6V1B2 (B, H, N), SNAP25 (C, I, O), VAMP2 (D, J, P), synapsin 1 (E, K, Q), and ApoE (F, L, R). ATP6V1B2, SNAP25, and VAMP2 (N–P) all show a halo of elevated protein concentration in the region surrounding the dense core of the plaque. Scale bar = 50 µm, 20 µm in the ApoE panels.

Figure 4.4: Common synaptic or disease-associated proteins in plaques of TgCRND8 mice not detected by STOMP are also absent by immunofluorescence. The SNARE protein syntaxin-1 (A), synaptic marker synaptophysin (B), the neuronal protein alpha-synuclein (C) are absent from ThS-positive plaques (G, H, I, M, N, O). Staining of the glial protein GFAP (D) surrounds the ThS-positive plaques but does not infiltrate the plaque core (J, P). Scale bar = 50 µm.

133

STOMP analysis of senile plaques from post-mortem AD brain

After developing the STOMP method using brain sections of TgCRND8 mice as the test subject, we applied the technique to senile plaques from a case of AD. Human tissue is inherently more difficult to work with than animal tissue because of biochemical changes associated with post-mortem degradation. Furthermore, formalin, the fixative of choice for human tissue, causes covalent modification of proteins and confounds mass spectrometry analysis. These limitations notwithstanding, we find the STOMP technique is suitable for the analysis of formalin-fixed post- mortem human tissue. Senile plaques from brain sections of a severe case of AD were visualized with ThS and photo-tagged using a procedure identical to that used for the TgCRND8 tissue. Mass spectrometry analysis of the protein composition of these senile plaques identified 60 proteins that were statistically significantly enriched, with Aβ being the most abundant protein component (Table 4.2). Synaptic proteins and proteins involved in synaptic vesicle transport comprised 15% of the abundant protein component of the plaques. For validation of the STOMP analysis of senile plaques, immunofluorescence staining was employed, revealing that GFAP and SNAP25 were detected by both immunofluorescence, while tau and α-synuclein were not detected by either technique (Figure 4.5).

Gliosis is a well-known feature of AD, and reactive glial cells are often found surrounding amyloid plaques (Mandybur and Chuirazzi, 1990; Sofroniew and Vinters, 2010). Glial proteins, GFAP and excitatory amino acid transporter (EAAT) (Anderson and Swanson, 2000), were both detected in (human) senile plaques (Table 4.2); however, only EAAT was detected in TgCRND8 amyloid plaques (Table 4.1). This apparent discrepancy is likely due to astrocytes being more intimately associated with amyloid plaques in human tissue (Figure 4.5A, E, I) compared to plaques in the TgCRND mouse (Figure 4.4D, J, P). Tau protein was also not among the abundant plaque proteins. Immunofluorescence staining shows that tau, which is contained in dystrophic neurites that surround the plaques, do not occupy the same physical space as the plaques and were hence not detected in the STOMP analysis (Figure 4.3B, H, N). SNAP25, on the other hand, was detected in the STOMP analysis and was seen to colocalize with senile plaques (Figure 4.5D, H, L) and amyloid plaques from TgCRND8 mice (Figure 4.3C, I, O). α-synuclein is part of the NAC (non-amyloid components) of plaques—with a possible role in fibril formation but not part of the plaque core (Culvenor et al. 1999; Wirths and Bayer 2003). These proteins were found not to be enriched in the plaques of this case of AD as detected by both STOMP analysis (Table 4.2) or by immunofluorescent staining (Figure 4.5C, G, K).

134

Table 4.2: Proteins statistically significantly enriched in senile plaques from a patient with AD identified and retrieved by STOMP. Table is sorted in descending order of abundance, by normalized spectral counts. Protein Uniprot IDSTOMP Dark Previous reports of counts counts enrichment in plaques (Söderberg et al., 2006) Amyloid beta A4 protein P05067 898 0.00 Detected, enriched Tubulin alpha-1A chain Q71U36 39.4 1.99 Detected, not enriched Tubulin beta-2A chain Q13885 23.0 3.51 Detected, not enriched Actin, cytoplasmic 1 P60709 13.8 0.00 Detected, not enriched Glial fibrillary acidic protein P14136 11.0 0.00 Detected, enriched Alpha-internexin Q16352 10.8 0.90 Previously unreported Synaptosomal-associated protein 25 P60880 9.65 0.00 Detected, not enriched Carbonyl reductase [NADPH] 1 P16152 9.05 0.00 Detected, not enriched ATP synthase subunit beta, mitochondrial P06576 9.28 0.00 Detected, enriched Tubulin polymerization-promoting protein O94811 6.33 0.00 Previously unreported Tubulin beta-3 chain Q13509 5.95 0.00 Previously unreported Cytochrome b-c1 complex subunit 8 O14949 5.05 0.00 Previously unreported Calcium/calmodulin-dependent protein kinase I Q9UQM7 5.08 0.00 Previously unreported Immunoglobulin superfamily member Q969P0 4.23 0.00 Previously unreported V-type proton ATPase subunit B, kidney isoform P15313 3.96 0.00 Detected, enriched Vesicle-associated membrane protein 2 P63027 3.95 0.00 Detected, not enriched Pyruvate dehydrogenase E1 component subunit beta P11177 3.82 0.00 Detected, not enriched Fructose-bisphosphate aldolase C P09972 3.80 0.00 Detected, not enriched Ferritin light chain P02792 3.75 0.00 Detected, not enriched Neurofilament heavy polypeptide P12036 3.33 0.89 Detected, not enriched Cytochrome c1, heme protein, mitochondrial P08574 2.82 0.00 Detected, not enriched Peptidyl-prolyl cis-trans isomerase Q13526 2.74 0.00 Detected, not enriched Phosphoglycerate mutase 1 P18669 2.60 0.00 Detected, not enriched Beta-actin-like protein 2 Q562R1 2.38 0.00 Detected, not enriched Creatine kinase B-type P12277 2.34 0.00 Detected, not enriched Transketolase P29401 2.21 0.00 Detected, not enriched Alpha-enolase P06733 2.12 0.00 Detected, not enriched Excitatory amino acid transporter 1 P43003 2.10 0.00 Previously unreported 40S ribosomal protein S8 P62241 2.07 0.00 Previously unreported 60 kDa heat shock protein, mitochondrial P10809 2.05 0.00 Detected, not enriched Stress-70 protein, mitochondrial P38646 2.04 0.00 Detected, not enriched Protein SERCA1 Q96JX3 2.02 0.00 Previously unreported Spectrin beta chain, brain 1 Q01082 1.82 0.00 Detected, enriched Fascin Q16658 1.83 0.00 Detected, not enriched 6-phosphogluconolactonase O95336 1.82 0.00 Previously unreported Thioredoxin-dependent peroxide reductase P30048 1.81 0.00 Detected, not enriched Glutamine synthetase P15104 1.78 0.00 Previously unreported Clathrin heavy chain 1 Q00610 1.70 0.00 Detected, not enriched Elongation factor Tu, mitochondrial P49411 1.51 0.00 Detected, not enriched Tubulin alpha-4A chain P68366 1.50 0.00 Detected, not enriched Methionine adenosyltransferase 2 subunit beta Q9NZL9 1.33 0.00 Previously unreported Dihydropyrimidinase-related protein 4 i O14531 1.21 0.00 Previously unreported Tenascin-R Q92752 1.17 0.00 Detected, not enriched Microtubule-associated protein 6 Q96JE9 1.16 0.29 Detected, not enriched Prelamin-A/C P02545 1.01 0.00 Previously unreported Neurofascin O94856 1.00 0.00 Detected, not enriched Protein kinase C and casein kinase substrate Q9BY111 0.98 0.00 Detected, not enriched Hexokinase-1 P19367 0.98 0.00 Detected, not enriched NADH-ubiquinone oxidoreductase 75 kDa subunit P283311 0.94 0.00 Detected, not enriched

135

Figure 4.5: Immunofluorescent confirmation of results of STOMP analysis of senile plaques in a case of AD. GFAP (A), a marker of glial cells, surrounds the ThS-positive plaques and partially infiltrates the plaque core (E, I). Tau in dystrophic neurites (B) surrounds the ThS-positive plaques but does not infiltrate the plaque core (F, J). α-synuclein (C) is absent from ThS-positive plaques (G, K). SNAP25 colocalizes with ThS-positive plaques in this case of AD (D, H, L). Scale bar is 20 μm.

Immunohistochemical staining of pre-synaptic proteins, SNAP25, VAMP2, and synaptophysin in senile plaques from human patients with AD was similar to the immunofluorescence staining in the AD mice, with some notable differences (Figure 4.6). Senile plaques in the human AD tissue were positive for both SNAP25 and VAMP2. However, synaptophysin staining, which was absent in the amyloid deposits of AD mice, displayed a unique punctate lobular profile that decorated the periphery of the plaque in the human tissue (Figure 4.6). These STOMP analyses demonstrate the molecular differences of amyloid deposits in AD mice compared to senile plaques in AD patients, but confirm the segregation of these pre-synaptic proteins in and around amyloid plaques and point to their potential importance in AD pathophysiology.

136

Figure 4.6: Microphotographs of Synaptophysin (A), VAMP2 (B) and SNAP25 (C) immunohistochemistry on the brain of human Alzheimer's disease cases. The amyloid plaques (arrow) are positive for VAMP2 and SNAP25, whereas synaptophysin does not stain the core of the plaques and punctate lobular profiles decorating the plaque are positive for synaptophysin. Scale bar is 30 µm. Comparison with previously published data

Some proteins identified by STOMP (including v-ATPase, dynamin, SNAP25, VAMP2, synapsin, clathrin, ankyrin, and so on; see Tables 4.1, 4.2) were detected in previous LCM-based studies; however, with the exception of v-ATPase and dynamin, these proteins were not previously shown to be significantly or sufficiently enriched compared to the control samples (Liao et al. 2004). The highly specific nature of STOMP protein labeling, and recovery permits very sensitive detection of proteins present in plaques, even if those proteins are also abundant in the remainder of the brain.

Conclusions

We have shown that the STOMP technique can be used to determine the proteomic composition of micron-scale microscopic features. The procedure does not require large quantities of tissue and can be completed in a reasonable amount of time. Using amyloid plaques in TgCRND8 mice as a test subject, we demonstrated that STOMP is capable of determining the protein composition of pathological deposits in neurodegenerative disease. Our discovery that there is an accumulation of certain pre-synaptic proteins in AD plaques and in TgCRND8 mice provides direct evidence for an association between amyloid plaque formation and synaptic function. While our initial work in characterizing amyloid deposits in mouse brain used methanol fixed tissues to minimize problems with solubilizing tissue, our experience with human AD pathological tissue shows that STOMP of formalin-fixed archival tissue is possible. The STOMP technique is ideally suited for determination of the compositions of intracellular protein inclusions seen in ALS and FTD. While the principle components of these inclusions, such as TDP-439, FUS (Kwiatkowski et al. 2009; Vance et al. 2009), and dipeptide repeat proteins (Mori et al., 2013) are known, determination of the secondary

137 components of the inclusions could shed light on disease etiology. In developing the STOMP technique, we have focused on identification of photo-tagged proteins, however, the nonspecific nature of benzophenone photochemistry should cause photo-tagging of nonprotein components as well. In this regard, the STOMP technique can easily be modified to identify the RNA composition of biological features. RNA-containing subcellular structures such as P-bodies and stress granules have been implicated in neurodegenerative disease (Li et al. 2013); these structures are of suitable size for STOMP, and there is a realistic possibility of determining jointly their proteomic and transcriptomic composition with this technique. The application of this technique to many other aspects of cell biology research is also conceivable.

138

Materials and Methods

Unless otherwise specified, all dry reagents are from Sigma–Aldrich, St. Louis, MO. All organic solvents are from Caledon Laboratories, Georgetown, ON.

Photo-tag synthesis

The 6HisBP peptide (4-benzoyl-benzamidyl-Gly-His-His-His-His-His-His-amide, Figure 4.1) was synthesized using FMOC-His(Trt)-OH, FMOC-Gly-OH (Advanced Chemtech, Louisville, KY) and 4-benzoylbenzoic acid; all synthesis steps were carried out in N,N-dimethylformamide (DMF). FMOC cleavage employed 2% 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU); activation and coupling used 0.5 M N,N-diisopropylethylamine (DIPEA) and o-(7-azabenzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium hexafluorophosphate (HATU, Applied Biosystems, Waltham, MA). The peptide was cleaved from the resin by incubating in 5 mL of a solution containing trifluoroacetic acid, thioanisole, ethanedithiol, and anisole in the volume ratio of 90:5:3:2, for 2 h at room temperature, then precipitated by dropwise addition to 40 mL −70 ˚C diethyl ether. The crude 6HisBP was washed five times with cold diethyl ether to remove residual cleavage reagents, dried under nitrogen, redissolved in Milli-Q water, frozen, and lyophilized (Freezone 4.5, Labconco, Kansas City, MO) to remove any remaining volatiles. The presence of 6HisBP was confirmed by ESI-MS.

Murine tissue sectioning and preparation

Transgenic CRND8 (TgCRND8) mice express a transgene that encodes human amyloid precursor protein (APP695) harboring both the ‘Swedish’ double mutation KM670/671/NL and the additional ‘Indiana’ point mutation V717F; they develop dense-cored amyloid plaques and neuritic pathology by 5 months of age (Chishti et al. 2001). Animals were sacrificed at 8 months and their brains were embedded in OCT medium for frozen sectioning onto glass slides. Sections were post-fixed by immersion for 10 min in −20 ˚C methanol, air dried at room temperature, and stored at −70 ˚C. Brain sections were treated with DEPC to covalently modify endogenous histidyl residues and abolish their capacity to bind to nickel. Sections were rehydrated by immersion in DEPC buffer (10 mM NaHCO3,

10 mM NaH2PO4, pH adjusted to 6.0 with hydrochloric acid). An initial working solution of 1 M DEPC in methanol was prepared immediately before use and diluted in DEPC buffer to 10 mM DEPC. Tissue sections were immersed in DEPC solution for 3 h at room temperature, then washed twice with PBS. Sections were stained with 1% ThS. Sections were soaked in two changes of 6HisBP working solution (12 mM 6HisBP, 10 mM sodium phosphate, pH 7.6), then coverslipped. For each

139

STOMP sample, a ‘dark’ control was prepared from an adjacent section. All experiments were performed in accordance with the guidelines imposed by the University of Toronto Animal Care Committee and Canadian Council for Animal Care.

Microscopy and photoactivation

Microscopy employed Zeiss LSM 510 and LSM 710 confocal microscopes, equipped with two-photon infrared (Coherent Chameleon, tuned to 720 nm) laser sources. ThS fluorescence was excited at 488 nm and imaged confocally through 505 nm longpass or 500–530 nm bandpass filters. The nominal pixel spacing was 0.44 × 0.44 × 0.44 μm using a 20× 1.20 NA water immersion objective lens. Masks containing only amyloid plaques were created using the following four-step process in ImageJ (Schneider et al. 2012). High-frequency noise in the ThS fluorescence image was removed using ImageJ’s ‘Smooth’ function, and vascular amyloid deposits—if any—were manually excluded from each fluorescence image. Dense-cored plaques were highlighted with the ‘Threshold’ tool. Finally, the ‘Erode’ method was used to trim a single-pixel (0.44 μm) margin from the mask edge, to ensure that only material definitely within the plaque would be photo-tagged. The resulting mask image was saved as a ‘text image’. Using a custom macro for controlling the two-photon microscope (Pham et al. 2007) each pixel in the mask image was irradiated six times, for 4 ms on each pass. A detailed description of the STOMP macro interface available in Source code 1. The total volume photo-activated was approximately 2 × 106 μm3 per sample (1 × 106 μm3 per section).

Solubilization and affinity purification

Each section was lifted from the slide using a clean, sterile razor blade and transferred into a 1.5-mL polypropylene microcentrifuge tube. For methanol-fixed murine tissue, 200 μL of solubilization buffer containing 2% sodium dodecyl sulfate (SDS, Bio-Rad, Hercules, CA), 8 M urea, 250 mM NaCl, 10 mM sodium phosphate buffer (pH 8.2), 10% (wt/vol) glycerol, and 2% βME was added to each tube. Tubes were immersed in a boiling water bath for 5 min, sonicated for 5 min in a Branson 1210 bath sonicator (Branson Ultrasonic Corporation, Danbury, CT), incubated at room temperature for 45 min, and returned to boiling water for an additional 5 min to fully solubilize the tissue sample. For solubilization of formalin-fixed human tissue, 400 μL of a modified solubilization buffer containing 200 mM Tris (in addition to the ingredients noted above, as a formaldehyde scavenger) was used. Formalin-fixed sections were held at 90 ˚C for 60 min. Each sample was transferred to a 15-mL polypropylene centrifuge tube and combined with a 20-fold excess of room-temperature dilution buffer (8 M urea, 250 mM NaCl, 10 mM sodium phosphate buffer, 1 mM imidazole) to dilute the

140

SDS, βME, and Tris to concentrations compatible with Ni-NTA agarose bead binding. 80 μL of Ni- NTA agarose bead slurry (Qiagen, Valencia, CA) was added, and the mixture was tumbled overnight at room temperature to allow the His6-tagged proteins to bind to the beads. After binding, the beads were pelleted by centrifugation at 500×g for 30 s. The supernatant was discarded, and the beads were resuspended in 1 mL wash buffer (6 M urea, 0.1% SDS, 250 mM NaCl, 10 mM phosphate at pH 8.2, 0.1% βME, 1 mM imidazole) and tumbled for 2 min. This wash was repeated three times. During the final wash step, each sample was divided into two portions: one reserved for mass spectrometry, and one for gel electrophoresis and silver staining

Mass spectrometry

After recovery, the proteins captured on the surface of the beads for each sample were solubilized in 8 M urea with 50 mM Tris at pH 8, reduced by 5 mM DTT for 1 hr, and alkylated with 10 mM iodoacetamide for 45 min in darkness at room temperature. Proteins were digested at RT with Endoproteinase Lys-C (Roche) for 6 h first, and then diluted ninefold into ammonium bicarbonate buffer before adding sequencing grade trypsin (Promega) overnight. After acidification to 1% formic acid (FA), mixtures were desalted using disposable Toptip C-18 columns (Glygen) and the eluted peptides lyophilized to dryness. Peptides were loaded and separated on sequential reverse phase micro- capillary liquid trap and analytical columns using an EASY-nLC nanoflow pump system (Proxeon). The micro-capillary trap column was constructed in a 25 mm × 75 mm silica capillary packed with 5 μm Luna C18 stationary phase (Phenomenex). The analytical column was constructed in a 100mm × 75 μm silica capillary, with a fine tip pulled with a column puller (Sutter Instruments), packed with 3 μm Luna C18 stationary phase. Separation was performed in 105 min using an organic gradient w consisting of buffer A (5% acetonitrile with 0.1% FA) to buffer B (95% acetonitrile with 0.1% FA) at a flow rate of 300 nL/min starting with a gradient of 2–6% buffer B in 1 min, followed by 6–24% in 74 min, 24–90% in 16 min, then 90% buffer B for 5 min and 90–0% in 1 min and finally 0% buffer B in 8 min. Eluted peptides were electrosprayed from a nanospray ion source (Proxeon) directly into a high-performance Orbitrap Velos hybrid tandem mass spectrometer (ThermoFisher Scientific). 10 data-dependent collision-induced dissociation (CID) ion trap scans (centroid mode) were automatically acquired simultaneously with one high resolution (60,000 FWHM resolution) full scan (profile mode) mass spectra. A dynamic exclusion list was enabled to exclude a maximum of 500 ion targets for 22.5 s. RAW data files were extracted with the ReAdW program and submitted to database search using SEQUEST (v2.7) against UniProt/SwissProt protein sequence FASTA file containing 22,491 human proteins as well as an equivalent number of reversed decoy proteins (to estimate the

141 empirical false discovery rate). Search parameters were set to allow for one missed cleavage site and one fixed modification (+57 on cysteine) using a precursor tolerance of 3 m/z. Matched peptides were further filtered at the precursor ion mass accuracy level using a 20-ppm cut-off, while protein hits were compiled and filtered using the StatQuest program with a minimum confidence threshold of 99%.

Gel electrophoresis and silver staining

Silver staining employed an adaptation of the method of Shevchenko et al. (1996). Gels were fixed in 6% formaldehyde and 25% ethanol for 1 h and washed with water. Gels were sensitized for 45 min in 50 ppm sodium thiosulfate solution, soaked in 0.1% silver nitrate for 45 min, rinsed with Milli-Q water, and developed for approximately 1 min in a solution of 2% sodium carbonate with 0.037% formaldehyde. Development was stopped with 50 mM EDTA. Silver-stained gels were photographed with an EOS 550D digital camera (Canon, Tokyo, Japan). Immunofluorescence and immunohistochemistry TgCRND8 brain sections were stained with 1%ThS and blocked overnight with 10%fetal bovine serum (FCS) in PBS. Sections were incubated with primary antibodies against β-amyloid peptide, SNAP25, synapsin 1, synaptophysin, syntaxin 1, α-synuclein, V-type proton ATPase subunit B (ATP6V1B2) or VAMP2 at the dilutions indicated in Supplementary file 4, and then with the appropriate Texas Red- labeled secondary. Formalin-fixed, paraffin-embedded sections of human brain from deceased AD patients were incubated with synaptophysin, SNAP25, and VAMP2 antibodies at the dilutions indicated in Supplementary file 4. After incubation with secondary antibodies, immunoreactivity was developed using diaminobenzidine and counterstained with hematoxylin. Tissues from AD cases were collected with patient consent and handled under protocols approved by the University Health Network.

Photo-tagging volume measurement

TgCRND8 brain sections (frozen sectioned, methanol post-fixed, DEPC treated) were soaked in 6HisBP solution, and a single pixel spot was photo-tagged using the STOMP macro as described above. Unbound photo-tag was then washed out of the section with several changes of PBS. Sections were blocked with 5% bovine serum albumin, incubated with an anti-hexahistidine antibody, and fluorescently labeled with an Alexa 488-conjugated secondary antibody. A high-resolution confocal z stack of immunofluorescence images was collected using a 63× 1.4 NA oil-immersion objective. Maximum intensity projections through the resulting z stacks were used to quantify the extent of the photo-tagged spot. Fluorescence intensity profiles through the center of the spot in x, y, and z

142 directions were extracted using ImageJ and fitted with a Gaussian curve using OriginPro 8.5.0 (OriginLab, Northampton MA) to determine the width (FWHM) of the spot in each dimension.

143

References

Anderson CM, Swanson RA. 2000. “Astrocyte glutamate transport: review of properties, regulation, and physiological functions.” Glia 32: 1–14. doi:10.1002/1098-1136.

Bolton DC, McKinley MP, Prusiner SB. 1982. “Identification of a protein that purifies with the scrapie prion.” Science 218: 1309–1311.

Chishti MA, Yang DS, Janus C, Phinney AL, Horne P, Pearson J, Strome R, Zuker N, Loukides J, French J, Turner S, Lozza G, Grilli M, Kunicki S, Morissette C, Paquette J, Gervais F, Bergeron C, Fraser PE, Carlson GA, George- Hyslop PS, Westaway D. 2001. “Early-onset amyloid deposition and cognitive deficits in transgenic mice expressing a double mutant form of amyloid precursor protein.” The Journal of Biological Chemistry 276: 21562–21570. doi:10.1074/jbc.M100710200.

Choi H, Larsen B, Lin ZY, Breitkreutz A, Mellacheruvu D, Fermin D, Qin ZS, Tyers M, Gingras AC, Nesvizhskii AI. 2011. “SAINT: probabilistic scoring of affinity purification-mass spectrometry data.” Nature Methods 8: 70–73. doi:10.1038/nmeth.1541.

Culvenor JG, McLean CA, Cutt S, Campbell BC, Maher F, Jäkälä P, Hartmann T, Beyreuther K, Masters CL, Li QX. 1999. “Non-Abeta component of Alzheimer’s disease amyloid (NAC) revisited. NAC and alpha-synuclein are not associated with Abeta amyloid.” The American Journal of Pathology 155: 1173–1181.

Firat-Karalar EN, Rauniyar N, Yates JR, Stearns T. 2014. “Proximity interactions among centrosome components identify regulators of centriole duplication.” Current Biology 24: 664–670. doi:10.1016/j.cub.2014.01.067.

Glenner GG, Wong CW. 1984. “Alzheimer’s disease: initial report of the purification and characterization of a novel cerebrovascular amyloid protein.” Biochemical and Biophysical Research Communications 120: 885–890.

Golgi C. 1873. Sulla struttura della sostanza grigia del cervello. Gazzetta Medica Italiana Lombardia 33:244–246. Gozal YM, Cheng D, Duong DM, Lah JJ, Levey AI, Peng J. 2006. “Merger of laser capture microdissection and mass spectrometry: a window into the amyloid plaque proteome.” Methods in Enzymology 412: 77–93.

Gozal YM, Duong DM, Gearing M, Cheng D, Hanfelt JJ, Funderburk C, Peng J, Lah JJ, Levey AI. 2009. “Proteomics analysis reveals novel components in the detergent-insoluble subproteome in Alzheimer’s disease.” Journal of Proteome Research 8: 5069–5079. doi:10.1021/pr900474t.

Hung V, Zou P, Rhee HW, Udeshi ND, Cracan V, Svinkina T, Carr SA, Mootha VK, Ting AY. 2014. “Proteomic mapping of the human mitochondrial intermembrane space in live cells via ratiometric APEX tagging.” Molecular Cell 55: 332–341. doi:10.1016/j.molcel.2014.06.003.

Kwiatkowski TJ Jr, Bosco DA, Leclerc AL, Tamrazian E, Vanderburg CR, Russ C, Davis A, Gilchrist J, Kasarskis EJ, Munsat T, Valdmanis P, Rouleau GA, Hosler BA, Cortelli P, de Jong PJ, Yoshinaga Y, Haines JL, Pericak-Vance MA, Yan J, Ticozzi N, Siddique T, McKenna-Yasek D, Sapp PC, Horvitz HR, Landers JE, Brown RH Jr. 2009. “Mutations in the FUS/TLS gene on

144 chromosome 16 cause familial amyotrophic lateral sclerosis.” Science 323: 1205–1208. doi:10.1126/science.1166066.

Lacor PN, Buniel MC, Furlow PW, Clemente AS, Velasco PT, Wood M, Viola KL, Klein WL. 2007. “Abeta oligomer-induced aberrations in synapse composition, shape, and density provide a molecular basis for loss of connectivity in Alzheimer’s disease.” The Journal of Neuroscience 27: 796– 807. doi:10.1523/JNEUROSCI.3501-06.2007.

Lagier-Tourenne C, Cleveland DW. 2009. “Rethinking ALS: the FUS about TDP-43.” Cell 136: 1001–1004. doi:10. 1016/j.cell.2009.03.006.

Lee S-H, Moon JJ, West JL. 2008. “Three-dimensional micropatterning of bioactive hydrogels via two-photon laser scanning photolithography for guided 3D cell migration.” Biomaterials 29: 2962– 2968. doi:10.1016/j.biomaterials.2008.04.004.

Li YR, King OD, Shorter J, Gitler AD. 2013. “Stress granules as crucibles of ALS pathogenesis.” The Journal of Cell Biology 201: 361–372. doi:10.1083/jcb.201302044.

Liao L, Cheng D, Wang J, Duong DM, Losik TG, Gearing M, Rees HD, Lah JJ, Levey AI, Peng J. 2004. “Proteomic characterization of postmortem amyloid plaques isolated by laser capture microdissection.” The Journal of Biological Chemistry 279: 37061–37068. doi:10.1074/jbc.M403672200.

Mackenzie IR, Rademakers R, Neumann M. 2010. “TDP-43 and FUS in amyotrophic lateral sclerosis and frontotemporal dementia.” Lancet Neurology 9: 995–1007. doi:10.1016/S1474- 4422(10)70195-2

Mackenzie IRA. 2007. “The neuropathology of FTD associated with ALS.” Alzheimer Disease and Associated Disorders 21: S44–S49. doi:10.1097/WAD.0b013e31815c3486.

Mandybur TI, Chuirazzi CC. 1990. “Astrocytes and the plaques of Alzheimer’s disease.” Neurology 40: 635–639. doi:10.1212/WNL.40.4.635.

Mattson MP, Partin J, Begley JG. 1998. “Amyloid beta-peptide induces apoptosis-related events in synapses and dendrites.” Brain Research 807: 167–176. doi:10.1016/S0006-8993(98)00763-X.

Mori K, Weng SM, Arzberger T, May S, Rentzsch K, Kremmer E, Schmid B, Kretzschmar HA, Cruts M, Van Broeckhoven C, Haass C, Edbauer D. 2013. “The C9orf72 GGGGCC repeat is translated into aggregating dipeptide-repeat proteins in FTLD/ALS.” Science 339: 1335–1338. doi:10.1126/science.1232927.

Musiek ES, Holtzman DM. 2015. “Three dimensions of the amyloid hypothesis: time, space and ‘wingmen’.” Nature Neuroscience 18: 800–806. doi:10.1038/nn.4018.

Neumann M, Sampathu DM, Kwong LK, Truax AC, Micsenyi MC, Chou TT, Bruce J, Schuck T, Grossman M, Clark CM, McCluskey LF, Miller BL, Masliah E, Mackenzie IR, Feldman H, Feiden W, Kretzschmar HA, Trojanowski JQ, Lee VM. 2006. “Ubiquitinated TDP-43 in frontotemporal lobar degeneration and amyotrophic lateral sclerosis.” Science 314: 130–133. doi:10.1126/science.1134108.

145

Pham HH, Gourevich I, Jonkman JEN, Kumacheva E. 2007. “Polymer nanostructured material for the recording of biometric features.” Journal of Materials Chemistry 17: 523. doi:10.1039/B614491H.

Pham E, Crews L, Ubhi K, Hansen L, Adame A, Cartier A, Salmon D, Galasko D, Michael S, Savas JN, Yates JR, Glabe C, Masliah E. 2010. “Progressive accumulation of amyloid-beta oligomers in Alzheimer’s disease and in amyloid precursor protein transgenic mice is accompanied by selective alterations in synaptic scaffold proteins.” The FEBS Journal 277: 3051–3067. doi:10.1111/j.1742- 4658.2010.07719.x.

Rhee HW, Zou P, Udeshi ND, Martell JD, Mootha VK, Carr SA, Ting AY. 2013. “Proteomic mapping of mitochondria in living cells via spatially restricted enzymatic tagging.” Science 339: 1328– 1331. doi:10.1126/science.1230593.

Roux KJ, Kim DI, Raida M, Burke B. 2012. “A promiscuous biotin ligase fusion protein identifies proximal and interacting proteins in mammalian cells.” The Journal of Cell Biology 196: 801–810. doi:10.1083/jcb.201112098.

Schneider CA, Rasband WS, Eliceiri KW. 2012. “NIH Image to ImageJ: 25 years of image analysis.” Nature Methods 9: 671–675. doi:10.1038/nmeth.2089.

Sethi S, Vrana JA, Theis JD, Leung N, Sethi A, Nasr SH, Fervenza FC, Cornell LD, Fidler ME, Dogan A. 2012. “Laser microdissection and mass spectrometry-based proteomics aids the diagnosis and typing of renal amyloidosis.” Kidney International 82: 226–234. doi:10.1038/ki.2012.108.

Sethi S, Theis JD, Vrana JA, Fervenza FC, Sethi A, Qian Q, Quint P, Leung N, Dogan A, Nasr SH. 2013. “Laser microdissection and proteomic analysis of amyloidosis, cryoglobulinemic GN, fibrillary GN, and immunotactoid glomerulopathy.” Clinical Journal of the American Society of Nephrology 8: 915– 921. doi:10.2215/CJN.07030712.

Shevchenko A, Wilm M, Vorm O, Mann M. 1996. “Mass spectrometric sequencing of proteins silver-stained polyacrylamide gels.” Analytical Chemistry 68: 850–858. doi:10.1021/ac950914h.

Sipe JD, Westermark P. 2005. “Amyloid proteins: the beta sheet conformation and disease.” In: Sipe JD, editor. (Wiley). p. 2–27. doi: 10.1002/9783527619344.

Skoch J, Hyman BT, Bacskai BJ. 2006. “Preclinical characterization of amyloid imaging probes with multiphoton microscopy.” Journal of Alzheimer’s Disease 9: 401–407.

Söderberg L, Bogdanovic N, Axelsson B, Winblad B, N¨aslund J, Tjernberg LO. 2006. “Analysis of single Alzheimer solid plaque cores by laser capture microscopy and nanoelectrospray/tandem mass spectrometry.” Biochemistry 45: 9849–9856. doi:10.1021/bi060331+.

Sofroniew MV, Vinters HV. 2010. “Astrocytes: biology and pathology.” Acta Neuropathologica 119: 7– 35. doi:10. 1007/s00401-009-0619-8.

Sreedharan J, Blair IP, Tripathi VB, Hu X, Vance C, Rogelj B, Ackerley S, Durnall JC, Williams KL, Buratti E, Baralle F, de Belleroche J, Mitchell JD, Leigh PN, Al-Chalabi A, Miller CC, Nicholson G, Shaw CE. 2008. “TDP-43 mutations in familial and sporadic amyotrophic lateral sclerosis.” Science 319: 1668–1672. doi:10.1126/science.1154584.

146

Stoeckli M, Chaurand P, Hallahan DE, Caprioli RM. 2001. “Imaging mass spectrometry: a new technology for the analysis of protein expression in mammalian tissues.” Nature Medicine 7: 493–496. doi:10.1038/86573.

Takamori S, Holt M, Stenius K, Lemke EA, Grønborg M, Riedel D, Urlaub H, Schenck S, Brugger B, Ringler P, Muller SA, Rammner B, Grater F, Hub JS, De Groot BL, Mieskes G, Moriyama Y, Klingauf J, Grubmuller H, Heuser J, Wieland F, Jahn R. 2006. “Molecular anatomy of a trafficking organelle.” Cell 127: 831–846. doi:10.1016/j.cell.2006.10.030.

Tanskanen M. 2013. Amyloidosis. (InTech). doi: 10.5772/46140. Thal DR, Glas A, Schneider W, Schober R. 1997. “Differential pattern of beta-amyloid, amyloid precursor protein and apolipoprotein E expression in cortical senile plaques.” Acta Neuropathologica 94: 255–265. doi:10.1007/s004010050701.

Vance C, Rogelj B, Hortobagyi T, De Vos KJ, Nishimura AL, Sreedharan J, Hu X, Smith B, Ruddy D, Wright P, Ganesalingam J, Williams KL, Tripathi V, Al-Saraj S, Al-Chalabi A, Leigh PN, Blair IP, Nicholson G, de Belleroche J, Gallo JM, Miller CC, Shaw CE. 2009. “Mutations in FUS, an RNA processing protein, cause familial amyotrophic lateral sclerosis type 6.” Science 323: 1208–1211. doi:10.1126/science.1165942.

Virchow R. 1854. Zur Cellulose-Frage. Virchows Archciv Pathological Anatomy and Histology 6:416–426. Virchow R. 1855. “Ueber den Gang der amyloiden Degeneration.” Virchows Archciv Pathological Anatomy and Histology 8:364–368.

Wallis RB, Holbrook JJ. 1973. “The reaction of a histidine residue in glutamate dehydrogenase with diethyl pyrocarbonate.” The Biochemical Journal 133: 183–187. doi:10.1042/bj1330183

Wirths O, Bayer TA. 2003. “Alpha-synuclein, Abeta and Alzheimer’s disease.” Progress in Neuro- Psychopharmacology & Biological Psychiatry 27: 103–108. doi: 10.1016/S0278-5846(02)00339-1.

Wiseman JM, Ifa DR, Song Q, Cooks RG. 2006. “Tissue imaging at atmospheric pressure using desorption electrospray ionization (DESI) mass spectrometry.” Angewandte Chemie 45: 7188–7192. doi:10.1002/anie.200602449.

Wucher A, Cheng J, Winograd N. 2007. “Protocols for three-dimensional molecular imaging using mass spectrometry.” Analytical Chemistry 79: 5529–5539. doi:10.1021/ac070692a.

Xu G, Gonzales V, Borchelt DR. 2002. “Rapid detection of protein aggregates in the brains of Alzheimer patients and transgenic mouse models of amyloidosis.” Alzheimer Disease and Associated Disorders 16: 191–195.

Zhou J, Jones DR, Duong DM, Levey AI, Lah JJ, Peng J. 2013. “Proteomic analysis of postsynaptic density in Alzheimer’s disease.” Clinica Chimica Acta 420: 62–68. doi:10.1016/j.cca.2013.03.016.

CHAPTER V COST-EFFECTIVE ELIMINATION OF LIPOFUSCIN FLUORESCENCE FROM FORMALIN-FIXED BRAIN TISSUE BY WHITE PHOSPHOR LIGHT EMITTING DIODE ARRAY

This chapter discusses the elimination of autofluorescence in human brain tissue as a prerequisite for the acquisition of high quality mask images for the STOMP technique. This chapter first appeared in Biochemistry and Cell Biology as: Y. Sun, and A. Chakrabartty. (2016). Cost-Effective Elimination of Lipofuscin Fluorescence from Formalin-Fixed Brain Tissue by White Phosphor Light Emitting Diode Array. Biochemistry and Cell Biology 94 (6): 545–550. It was written by Y.S with input from A.C. Tissue sectioning was performed by core facilities at Princess Margaret Cancer Research Tower. All remaining experimental work was carried out by Y.S. The same method was published in video format in the Journal of Visualized Experiments as Y. Sun, P. Ip, and A. Chakrabartty. (2017). Simple Elimination of Background Fluorescence in Formalin-Fixed Human Brain Tissue for Immunofluorescence Microscopy. Journal of Visualized Experiments 127 (September). doi:10.3791/56188, where P.I. and Y.S. produced the video.

147 148

Chapter abstract

Autofluorescence of aldehyde-fixed tissues greatly hinders fluorescence microscopy. In particular, lipofuscin, an autofluorescent component of aged brain tissue, complicates fluorescence imaging of tissue in neurodegenerative diseases. Background and lipofuscin fluorescence can be reduced by greater than 90% through photobleaching using white phosphor light emitting diode arrays prior to treatment with fluorescent probes. We compared the effect of photobleaching versus established chemical quenchers on the quality of fluorescent staining in formalin-fixed brain tissue of frontotemporal dementia with tau-positive inclusions. Unlike chemical quenchers, which reduced fluorescent probe signals as well as background, photobleaching treatment had no effect on probe fluorescence intensity while it effectively reduced background and lipofuscin fluorescence. The advantages and versatility of photobleaching over established methods are discussed.

Abbreviations used in this chapter

STOMP, spatially targeted optical microproteomic; ThS, Thioflavin S; AD, Alzheimer’s disease; ALS, amyotrophic lateral sclerosis; FTD, frontotemporal dementia; TDP-43, TAR DNA binding protein of 43 kDa; FTLD, frontotemporal lobar degeneration; LED, light emitting diode; PB, photobleaching; TB, TrueBlack™; EBT, Eriochrome Black T; OCT, optimal cutting temperature.

149

Introduction

The unambiguous fluorescence staining of features of interest within a cell or tissue is required to generate an accurate mask file for the spatially targeted optical microproteomic (STOMP) technique. Non-specific signals confound the mass spectrometry analysis by introducing surrounding proteins outside of the regions of interest and reduce replicability of the method. In the previous chapter, Thioflavin S (ThS) staining of brain tissues with Alzheimer’s disease (AD) was simple to perform and highly specific for the cross-β features in the amyloid plaques of interest (Hadley et al. 2015). However, to apply STOMP to inclusions seen in amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) requires the immunostaining of these tissues with antibodies against a known protein within these inclusions such as TAR DNA binding protein of 43 kDa (TDP-43). These additional steps introduce non-specific background signals, and the intensity of fluorescence is usually far lower than that of ThS, making them difficult to distinguish from tissue autofluorescence. Indeed, the successful visualization of fluorescent markers in tissue is often complicated by autofluorescence. Many sources of endogenous autofluorescence exist in mammalian tissues, such as lipofuscin, collagen, and elastin (Banerjee et al. 1999, Ottis et al. 2012). Formalin fixation, a common method of sample preparation, also generates autofluorescence. Lipofuscin, whose presence is a common hallmark of aging, is a particular complication in fluorescence imaging of specimens of neurodegenerative diseases, where patients are often of advanced age (Ottis et al. 2012). Lipofuscin’s broad fluorescence emission greatly hinders the application of common fluorescent labels like FITC and Texas Red. Lipofuscin signals often need to be manually labeled in fluorescence images to aid image interpretation (Xiao et al. 2015).

A number of treatment options have been suggested to reduce lipofuscin fluorescence. These include chemical treatment of the specimen with CuSO4 in ammonium acetate or quenching dyes such as Sudan Black B and Eriochrome Black T (Davis et al. 2014). Digital image processing by spectral un- mixing has also been used, but the process is difficult and has the disadvantage of generating false positives and artifacts (Zimmermann et al. 2003). Previously, photobleaching using a fluorescent tube lamp has been performed, although certain lipofuscin structures still remain unaffected after 48 h in brain tumour sections (Neumann and Gabel 2002). It has also been reported previously that RGB- based, multispectral light emitting diode (LED) arrays are effective at reducing autofluorescence, though the apparatus may be cost-prohibitive (Duong and Han 2013). The most thorough photobleaching can be achieved by irradiating the sample under a fluorescence microscope using a

150 high-powered laser source through an objective (Schnell et al. 1999, Neumann and Gabel 2002). However, this method is impractical due to the limited field of view of the objective compared to the size of the sample. Prolonged exposure will also cause sample damage.

Recently, white phosphor light emitting diodes (LEDs) used in household lighting have emerged as an alternative light source for fluorescence microscopy over halogen lamps due to their wide emission spectrum, high luminosity, and very low manufacturing cost (Albeanu et al. 2008, Robertson et al. 2009). Phosphor-based LEDs generate white light through a single blue LED surrounded by yellow phosphors, and are less costly than their RGB-based counterparts. Given the viability of white LEDs as a fluorescence excitation source, we propose a simple, effective, and widely applicable method to reduce autofluorescence from formalin-fixed tissue by photobleaching the specimen using a commercially available LED lamp prior to immunostaining and fluorescence imaging. In the study by Duong and Han, an apparatus was constructed using custom scaffolds, slide holders, and an array of high power multispectral LEDs in an enclosed refrigeration unit. The cost of raw materials, fabrication and construction for this apparatus may be upwards of 1000 USD. In the present study, we show that an equivalent result can be achieved using off-the-shelf components for less than 1% of the cost.

We quantified the efficacy of white phosphor LED photobleaching in formalin-fixed human brain tissue from a patient with frontotemporal lobar degeneration with tau-positive inclusions (FTLD-T). We stained the tissue with a number of fluorescent probes and compared and quantified the quality of staining in sections that have been pre-treated with photobleaching or treated with chemical quenchers TrueBlack™ (a proprietary formula) or Eriochrome Black T. We demonstrate that photobleaching, unlike chemical quenchers, does not suppress fluorescent probe signals.

151

Results and Discussion

Photobleaching significantly reduces autofluorescence

Autofluorescence of unstained brain tissue from a case of FLTD-T was measured before photobleaching and at 12-h intervals of LED photobleaching. The emission peaks of the lamp at 450 nm and 500-600 nm suggests that it can effectively target the excitation wavelengths of a broad range of visible-light chromophores (Figure 5.1B). While previous studies targeted individual emission peaks with multispectral LEDs (Duong and Han 2013), we found that white phosphor LEDs could produce a similar emission spectrum. It was evident from the photographs of the sections that the majority of background fluorescence was eliminated after 24 h (Figure 5.1C), while the fluorescent, lipofuscin-like speckles were nearly all photobleached after 72 h. Using the exponential decay function for quantification, we calculated the half-life of fluorescent intensity to be 4.6 and 7.1 h for Alexa 488 and Texas Red channels respectively (Figure 5.2), however, particles with reduced intensities were still generally visible in the photographs at this stage (Figure 5.1C). Using the decay constants, we calculated that 95% of the fluorescence should be removed after 20 and 31 h for the Alexa 488 and Texas Red channels respectively. Visually, we found that after 36 h photobleaching, the intensities of lipofuscin particles became sufficiently low compared to intended fluorescent signals, allowing for generation of immunofluorescence images with little to no background. We used 48 h as a convenient and effective photobleaching duration for subsequent experiments.

152

Figure 5.1: Time-dependent photobleaching of formalin- fixed brain tissue using a white phosphor LED array. (A) Photograph of photobleaching apparatus indicating the LED array light source, plastic platform, and sample chamber where a mock slide is placed in buffer. A reflective dome is constructed using aluminum foil to cover the apparatus during photobleaching. (B) Emission spectrum of LED array consists of a sharp peak at 450 nm and a broad peak centered at 550 nm and spanning 500-600 nm (arbitrary fluorescence units). No significant emission below 400 nm or above 650 nm is observed. (C) Photographs of 2 fields of view of unstained 10 µm thick brain section of grey matter in a case of FTLD-T imaged at Alexa 488 (λex = 488 nm, λem = 493-570 nm) and Texas Red (λex = 561 nm, λem = 601-635 nm) channels 0 to 72 h after photobleaching. Distinct fluorescent particles resembling lipofuscin are observed and their signal intensities are reduced after 72 h of photobleaching, although a few dim fluorescent bodies remain. Scale bar = 100 µm.

153

Figure 5.2: Quantification of LED-induced signal intensity reduction of lipofuscin fluorescence in two fields of view. Each dot represents the signal intensity of one lipofuscin particle present in a field of view at Alexa 488 (left) and Texas Red (middle) channels. Particles from two 425 µm × 425 µm fields of view are quantified and plotted. The population of fluorescent particles shifts to reduced intensity levels after photobleaching. The mean and standard deviation of the population is plotted (right) and fit to the exponential decay function , where y is the intensity, y0 is the baseline intensity, A is the initial maximum intensity, x is the photobleaching duration, and T is the decay constant. The fit parameters for each channel is listed.

Immunostaining of tau-positive inclusions

Immunostaining for hyperphosphorylated tau was performed to visualize pathological tau inclusions in the FTLD-T case (Figure 5.3). We tested the efficacy of 48-h photobleaching (PB) compared to a commercially available lipofuscin autofluorescence quencher TrueBlack™ (TB) and a previously reported treatment protocol using Eriochrome Black T (EBT; Davis et al. 2014). The tissues were stained with AT8 antibody, followed by commonly used Alexa 488- and Texas Red-conjugated fluorescent secondary antibodies and counterstained with the nuclear stain DAPI. In the untreated sample, autofluorescence was ubiquitous in both the Texas Red and Alexa 488 channels, significantly compromising the quality of the image. Signal profiling of the image showed that the tau inclusions were stained with relatively high intensity for both secondary antibodies. However, the lipofuscin intensity in the Texas Red channel was comparable to the intensity of the Texas Red secondary fluorescence, leading to complications in image interpretation. EBT-treated samples showed over-

154

quenching and complete loss of the DAPI signal. The dye also generated an undesirable basal level of fluorescence in the Texas Red channel. EBT appeared to be effective in reducing Alexa 488 channel autofluorescence, but indiscriminately reduced the secondary antibody fluorescence as well, limiting its utility. Both 48-h PB and TB treatment produced images with minimal background florescence. However, in the 48-h PB treatment, intensities of tau inclusions labeled by Alexa 488 and Texas Red fluorescence were relatively similar based on their signal profiles, whereas the samples treated with TB had significantly reduced Texas Red and DAPI signals.

Figure 5.3: Immunofluorescence imaging of phospho-tau stained FTLD-T formalin-fixed brain tissue. 10 µm thick sections were stained with AT8 primary antibody and Texas Red and Alexa 488 conjugated secondary antibodies. Slides were counterstained with DAPI. The signals from Alexa 488 (λex = 488 nm, λem = 493-570 nm), Texas Red (λex = 561 nm, λem = 601-635 nm), DAPI (λex = 405 nm, λem = 410-507 nm) and a merged image are displayed. Multiple autofluorescence reduction methods including 48-h photobleaching, TrueBlack™, Eriochrome Black T, and non-treatment are tested. Signal profiles at the cross section indicated by the dotted lines on the merged panels are quantified. Signal peaks representing Tau inclusions (T), nuclei (N) or lipofuscin particles (L) are labeled. Scale bar = 100 µm.

155

We then investigated the effect of PB and TB treatments on more sensitive stains such as Nissl, which can be quenched by BSA and other serum proteins (Figure 5.4). We stained the tissue with AT8 primary antibody followed by Alexa 488 secondary as before and applied NeuroTrace® 515/630 Nissl stain which labeled neurons within the tissue. The specimen was counterstained with DAPI. Similar to our previous results, PB treatment did not appear to weaken the fluorescence of the Nissl stain in the 570-690 nm emission range, and the cell bodies of the neurons were readily visible and distinct from the DAPI nuclear staining. In contrast, the TB treated samples showed reduced Nissl staining which largely overlapped with the DAPI signal (Figure 5.4). It is possible that the TB treatment quenched the weaker staining of Nissl bodies in the endoplasmic reticulum, but did not penetrate the nucleus, which Nissl also stains (Kádár et al. 2009).

Figure 5.4: Immunofluorescence imaging of phospho-tau and Nissl stained FTLD-T formalin fixed brain tissue using photobleaching and TrueBlack™ treatments. 10 µm thick sections were stained with AT8 primary antibody and Alexa 488 conjugated secondary antibody, NeuroTrace® fluorescent Nissl stain, and DAPI. The signals from Alexa 488 (λex = 488 nm, λem = 493-570 nm), Nissl (λex = 561 nm, λem = 601-635 nm), DAPI (λex = 405 nm, λem = 410-507 nm) and a composite channel are displayed. The sections are either photobleached prior to staining, or treated with TrueBlack™ after staining. Scale bar = 100 µm.

156

Overall, we find that irradiation of formalin-fixed brain tissue with a white phosphor LED lamp is an effective and low-cost method of reducing autofluorescence. Photobleaching does not interfere with the fluorescence intensity of probes due to the fact that it can be applied prior to sample staining. This allows for exclusive bleaching of unwanted autofluorescence, whereas commercial dyes and quenchers such as TrueBlack™ are applied prior to mounting the sample, allowing for undesirable quenching of probe signals. Additionally, photobleaching does not introduce any exogenous material to the specimen that potentially interferes with subsequent handling. Quenchers such as TrueBlack™ must also be applied in 70% ethanol, which will precipitate proteins. Compared to the commercial quencher TrueBlack™, our photobleaching protocol, which uses a commercial lamp, requires more processing time. Our LED array is also not expected to photobleach fluorescence excitable near UV range since the lamp has no emission at wavelengths below 405 nm. However, LED with higher power and intensity can reduce the processing time, and as previous studies have described, custom LEDs with different emission wavelengths can be used to direct photobleaching to specific chromophores (Duong and Han 2013). For targeting broad-spectrum autofluorescence, however, it appears that white phosphor LEDs is sufficiently effective.

We demonstrate that photobleaching treatment followed by immunostaining generally produces signals that are brighter and more intense at certain wavelengths compared to post-staining autofluorescence suppression approaches, while producing little to no background. The generation of clean images for immunofluorescence microscopy is not only important for the accurate visualization of features of interest, but it is also a prerequisite for the application of advanced microscopy techniques such as STOMP (Hadley et al. 2015). While previous methods have involved intricately assemblies, we present an extremely simple, inexpensive, but effective method for autofluorescence removal that is accessible to all investigators. Overall, photobleaching using white phosphor LED prior to staining is a versatile treatment that we expect to be amiable to a broad range of specimens.

157

Materials and Methods

Photobleaching apparatus

A 6-watt, 800 lux LED array desk lamp with flexible arm and flat lighting surface (DBpower, Shenzhen, Guangdong, China) was inverted such that the lamp could be placed under a square petri dish (Sarstedt, Nümbrecht, Germany) containing sterile buffer. The dish was elevated by a plastic support to avoid sample heating (Figure 5.1A). Samples on microscope slides were photobleached by submerging the slides in sterile TBS and 0.05% sodium azide at 4 ºC in a cold room. A reflective dome covered the petri dish containing the sample for the duration of photobleaching. The emission spectrum (400-800 nm) of the lamp was measured using a QM-1 fluorescence spectrophotometer (Photon Technology International, Edison, NJ, USA).

Sample preparation and Immunofluorescence

Formalin-fixed brain blocks of the orbitofrontal gyri in a case of FTLD-T (~2 days fixation) were run through 10%, 20%, and 30% sucrose gradients, infiltrated with optimal cutting temperature (OCT) compound, frozen, and cut into 10 µm thick sections and attached to glass microscope slides. For samples undergoing photobleaching treatment, slides were photobleached for 48 h in the photobleaching apparatus prior to staining. The slides were treated in antigen retrieval buffer (10 mM Citric Acid, 2 mM EDTA, 0.05% Tween 20, pH 6.2) for 30 min at 90 °C, then washed twice for 5 min in TBS plus 0.025% Triton X-100 (TBS-Triton). The slides were blocked in 10% normal goat serum (Aurion, Wageningen, The Netherlands) and 1% BSA (ThermoFisher Scientific, Waltham, MA, USA) in TBS-Triton for 2 h at room temperature. Primary mouse anti-phospho-PHF-tau pSer202+Thr205 (AT8) antibody (1:50, ThermoFisher Scientific, Waltham, MA, USA) in 1% BSA/TBS was applied overnight at 4 ºC in a humidified chamber. The slides were rinsed 2 × 5 min with TBS-Triton and a mix of secondary antibodies consisting of Alexa 488 goat anti-mouse (1:100, ThermoFisher Scientific, Waltham, MA, USA) and Texas red goat anti-mouse (1:100, ThermoFisher Scientific, Waltham, MA, USA) in 1% BSA/TBS was applied for 1 h at room temperature in the dark. The slides were rinsed 2 × 5 min in TBS and stained with DAPI (0.25 µg/mL, ThermoFisher Scientific, Waltham, MA, USA) for 10 min. Slides not receiving photobleaching treatment were then either treated with TrueBlack™ lipofuscin autofluorescence quencher (Biotium, Hayward, CA, USA) according to manufacturer’s instructions or with 1.65% Eriochrome Black T (Sigma-Aldrich, St. Louis,

158

MO, USA) in ddH2O for 5 min according to established protocols (Davis et al. 2014). The slides were rinsed 3 × 5 min in TBS to remove excess quencher and mounted with Immu-mount aqueous mounting medium (ThermoFisher Scientific, Waltham, MA, USA) for fluorescence microscopy. For Nissl staining, the above protocol was altered such that the blocking and antibody diluting buffers contained 0.5% fish gelatin (Electron Microscopy Sciences, Hatfield, PA, USA) in TBS instead of BSA or serum, and only Alexa 488 conjugated antibody was used for secondary antibody staining. Following DAPI staining and prior to mounting, NeuroTrace® 530/615 red fluorescent Nissl stain (ThermoFisher Scientific, Waltham, MA, USA) was applied to the slides according to manufacturer’s instructions.

Fluorescence microscopy and image quantitation

For detection of autofluorescence, several unstained sections were photobleached using the photobleaching apparatus for up to 72 h. At various time points, the slides were removed from the photobleaching apparatus, coverslipped in sterile TBS and imaged using a Zeiss LSM 710 confocal microscope (Zeiss, Oberkochen, Germany) with a Plan Apochromat 10× 0.45 NA objective (Zeiss, Oberkochen, Germany). For the Alexa 488 channel, the sample was excited at 488 nm (argon laser) and detected at 493-570 nm. For the Texas Red channel, the sample was excited at 561 nm (DPSS

561 nm laser) and detected at 601-635 nm. For stained slides, fluorescence at Alexa 488 (λex = 488 nm,

λem = 493-605 nm), Texas Red/Nissl (λex = 561 nm, λem = 566-689 nm), and DAPI (λex = 405 nm;

Diode 405 laser, λem = 410–507 nm) channels were measured. Laser power and gain settings were kept identical using Zen software for each set of comparative images. Lipofuscin fluorescence was quantified by particle analysis in ImageJ (Schneider et al. 2012), plotted, and fitted to a single exponential decay function (Equation 5.1) using Origin 8.5 software (OriginLab, Northampton, MA, USA), where y is the intensity, x is the photobleaching duration and T is the decay constant. Fluorescence RGB profiling was measured using ImageJ and the ‘RGBProfiler’ ImageJ plugin (http://rsb.info.nih.gov/ij/plugins/rgb-profiler.html).

159

Chapter Remarks

Acknowledgements:

This study was supported by the Canadian Consortium of Neurodegeneration and Aging (CCNA). The authors would like to thank Sultan Darvesh, and Andrew Reid from the Maritime Brain Tissue Bank for providing the FTLD brain tissues. Milan Ganguly from Spatio-temporal Targeting and Amplification of Radiation Response (STTARR) imaging center is thanked for preparing and sectioning the tissues. We also thank Kevin C. Hadley for helpful discussions and editing the manuscript.

Statement of Ethics

The work presented was performed in compliance with recognized international standards, including the International Conference on Harmonization (ICH), the Council for International Organizations of Medical Sciences (CIOMS) and the principles of the Declaration of Helsinki. Use of human tissue was in accordance with the University Health Network Research Ethic Board. The human brain samples were collected as a part of the Maritime Brain Tissue Bank. At the time of collection, informed consent was obtained.

160

References

Albeanu, D.F., Soucy, E., Sato, T.F., Meister, M., and Murthy, V.N. 2008. “LED arrays as cost effective and efficient light sources for widefield microscopy.” PLoS One 3 (5): 1–7.

Banerjee, B., Miedema, B.E., and Chandrasekhar, H.R. 1999. “Role of basement membrane collagen and elastin in the autofluorescence spectra of the colon.” J. Investig. Med. 47 (6): 326–32.

Davis, A.S., Richter, A., Becker, S., Moyer, J.E., Sandouk, A., Skinner, J., and Taubenberger, J.K. 2014. “Characterizing and Diminishing Autofluorescence in Formalin-fixed Paraffin-embedded Human Respiratory Tissue.” J. Histochem. Cytochem. 62 (6): 405–423.

Duong, H., and Han, M. 2013. “A multispectral LED array for the reduction of background autofluorescence in brain tissue.” J. Neurosci. Methods 220 (1): 46–54.

Hadley, K.C., Rakhit, R., Guo, H., Sun, Y., Jonkman, J.E., McLaurin, J., Hazrati, L.-N., Emili, A., and Chakrabartty, A. 2015. “Determining composition of micron-scale protein deposits in neurodegenerative disease by spatially targeted optical microproteomics.” eLife 4: 1–21.

Kádár, A., Wittmann, G., Liposits, Z., and Fekete, C. 2009. “Improved method for combination of immunocytochemistry and Nissl staining.” J. Neurosci. Methods 184 (1): 115–118.

Neumann, M., and Gabel, D. 2002. “Simple method for reduction of autofluorescence in fluorescence microscopy.” J. Histochem. Cytochem. 50 (3): 437–9.

Ottis, P., Koppe, K., Onisko, B., Dynin, I., Arzberger, T., Kretzschmar, H., Requena, J.R., Silva, C.J., Huston, J.P., and Korth, C. 2012. “Human and rat brain lipofuscin proteome.” Proteomics 12 (15-16): 2445–2454.

Robertson, J.B., Zhang, Y., and Johnson, C.H. 2009. “Light-emitting diode flashlights as effective and inexpensive light sources for fluorescence microscopy.” J. Microsc. 236 (1): 1–4.

Schneider, C. a, Rasband, W.S., and Eliceiri, K.W. 2012. “NIH Image to ImageJ: 25 years of image analysis.” Nat. Methods 9 (7): 671–675.

Schnell, S.A., Staines, W.A., and Wessendorf, M.W. 1999. “Reduction of Lipofuscin-like Autofluorescence in Fluorescently Labeled Tissue.” J. Histochem. Cytochem. 47 (6): 719–730.

Xiao, S., Sanelli, T., Chiang, H., Sun, Y., Chakrabartty, A., Keith, J., Rogaeva, E., Zinman, L., and Robertson, J. 2015. “Low molecular weight species of TDP-43 generated by abnormal splicing form inclusions in amyotrophic lateral sclerosis and result in motor neuron death.” Acta Neuropathol. 130 (1): 49–61.

Zimmermann, T., Rietdorf, J., and Pepperkok, R. 2003. “Spectral imaging and its applications in live cell microscopy.” FEBS Lett. 546 (1): 87–92.

CHAPTER VI PROBING TDP-43 DISEASE MECHANISMS USING STOMP TECHNOLOGY: CHALLENGES AND FUTURE DIRECTIONS

The contents of this chapter have not been previously published. Kevin Hadley performed initial SG- staining optimization (Figure 6.2), and Hongbo Guo performed tryptic digestion and MS/MS analysis on the initial processed SG sample (Figure 6.3B). The remaining experiments were conducted by Y.S.

161 162

Chapter abstract

The development of the STOMP technique and the use of photobleaching to overcome autofluorescence in aged tissue allow for the proteomic analysis of TDP-43-positive inclusions of archived patient samples. This data can be compared to the proteomic composition of stress granules (SGs) in cultured tissue to assess whether there are common proteins to infer any linkage between the formation of these two cellular features. However, the application of STOMP in either structure was challenging. While lightly fixed cultured cells containing SGs are easy to solubilize, the size of SGs are at the limits of STOMP technology, thus requiring impractical amounts of sample acquisition time. The inclusions in ALS/FTD, on the other hand, are larger but are still relatively small compared to amyloid plaques, resulting also in long processing time. Additionally, the solubilization protocol used for amyloid plaques were not amenable for use in these inclusions and resulted in sample loss. A similar solubilization problem was encountered in the application of STOMP to tau-positive inclusions in FTLD-T cases, suggesting that an optimization procedure may be required for each individual target of interest. Further advances in mass spectrometry technology and the development of fully automated STOMP algorithms would make this technique more widely applicable in the future. In the meantime, the biophysical characterization of TDP-43 performed in this work lays the foundation for the development of targeted approaches for the early detection of misfolded TDP-43 in patients suspected of an ALS diagnosis. The strategic approach for these future studies are discussed.

Abbreviations used in this chapter

TDP-43, TAR-DNA binding protein of 43 kD; ALS, amyotrophic lateral sclerosis; FTD, frontotemporal dementia; SG, stress granule; LLPS, liquid-liquid phase separation; LCM, laser-capture microdissection; ROI, region of interest; DEPC, diethylpyrocarbonate; 7AAD, 7-Aminoactinomycin D; FTLD-U, frontotemporal lobar degeneration with ubiquitin-positive inclusions; FTLD-T, frontotemporal lobar degeneration with tau-positive inclusions; CSF, cerebrospinal fluid; TTR, transthyretin.

163

Introduction

TDP-43 (TAR-DNA binding protein of 43 kDa) is a ubiquitous, nuclear, dimeric RNA binding protein involved in a number of cellular functions including pre-mRNA processing, splicing, microRNA processing and regulation, lncRNA and ncRNA expression, mRNA transport, stability, translation, and entry into stress granules (SGs) during the cellular stress response (Ratti and Buratti 2016). Despite these important RNA-processing functions, pathological accumulation of TDP-43 within motor and cortical neurons is associated with amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD), respectively. TDP-43 positive inclusions are found in 97% of all ALS cases, and nearly 50% of all FTD cases, suggesting a common disease pathway (Lagier-Tourenne, Polymenidou, and Cleveland 2010).

ALS is an incurable neurodegenerative disease caused by the death of motor neurons and subsequent fatal paralysis of the diaphragm within 3-5 years of disease onset. The cause of ALS is predominantly (90%) sporadic (sALS), whereas 10% of ALS have genetic causes (fALS), through mutations to genes encoding for proteins such as TDP-43, SOD1, FUS/TLS, and dipeptide repeats. However, their clinical symptoms are indistinguishable (Sun and Chakrabartty 2017). FTD, on the other hand, causes behavioral, speech, or cognitive deficits through the death of cortical neurons. 30-50% of FTD show family history with mutations to 3 major genes (MAPT, GRN, and C9ORF72), with no clear correlation between clinical, pathological, and genetic presentations (Lashley et al. 2015). A commonality between ALS and FTD is the presence of TDP-43 within inclusions of the affected neurons. This has led to the idea that the two diseases belong to a broader spectrum of neurodegenerative disorders coined TDP-43 proteinopathies (Forman, Trojanowski, and Lee 2007).

An emerging theory of TDP-43 pathogenesis suggests environmental stress (e. g. heat, toxins, or trauma) as a factor for the onset of disease. During cellular stress, polysomes translating mRNAs into non-essential proteins are arrested and, along with other RNA-binding proteins, coacervated into a 0.5-5 nm membraneless organelle called a stress granule (SG), through the physical process of liquid- liquid phase separation (LLPS). SGs are formed transiently and disassociate after the stress has passed (Molliex et al. 2015). Since TDP-43 is recruited into SGs during the stress response, it is thought that persistence of SGs or dysfunction of SG disassembly leads to the conversion of SGs from reversible droplet structures into, pathological aggregates as seen in disease (Li et al. 2013). The lack of cellular turnover and their large size may render neurons particularly susceptible to TDP-43 aggregation and

164

accumulation. To find links between SGs and TDP-43 positive inclusions, we sought to identify and compare the proteomic composition of these two cellular features.

The isolation of SGs for mass spectrometry analysis has been historically difficult due to the transient nature of these structures. Centrifugation has been used in yeast, but this method only pelleted the more rigid “core” structures of SGs and not their liquid “shells” (Wheeler et al. 2017; Jain et al. 2016). Laser-capture microdissection (LCM) methods, on the other hand, generally introduced considerable amounts of surrounding tissue due to its low resolution of ~10 microns (Emde et al. 2015). As such, the comprehensive proteomic content of SGs has not yet been determined.

Current methods that can be used to identify the proteomic content of inclusions of TDP-43 in patient tissue largely included the isolation of the detergent-soluble fraction of the entire sample, which would miss nearly all loosely associated proteins and those surrounding the inclusion and introduce unwanted proteins not specific to the inclusions (Seyfried et al. 2012; Martins-de-Souza et al. 2012; Schweitzer et al. 2006). Other methods include the production of thousands of monoclonal antibodies against specific post-translational modifications of these proteins (used previously to identify TDP-43 as a major component), but such a technique can be prohibitively costly (required 1000 monoclonal antibodies; Neumann et al. 2006).

We’ve developed a novel method called spatially targeted optical microproteomics (STOMP) that appears suitable in both applications, as it does not require a large amount of often-scarce patient sample material for TDP-43-positive inclusions and provides highly specific tagging of proteins in small structures such as SGs as the technique has a resolution of 1 µm2 (Hadley et al. 2015). In STOMP, a sample containing a region of interest (ROI) is first treated with diethylpyrocarbonate (DEPC) to modify endogenous histidines. It is then stained with a ROI-specific fluorescent label (dyes or fluorescently-labeled antibodies) to visualize the ROI. The sample is then saturated with a solution containing a phototag (photo-activatable benzophenone crosslinker conjugated to a His6 affinity tag; 6HisBP) and imaged using a confocal microscope capable of two-photon excitation. First an image is acquired through the ROI-specific label. This image is then converted into a binary mask and used to guide the two-photon laser to excite only the ROI. This excites the 6HisBP and crosslinks the affinity handle to all proteins within the ROI. The sample is then solubilized, purified via the His6 affinity tag, and subjected to tandem mass spectrometry (MS/MS) for proteomic identification. The use of a two-photo excitation allows for specific excitation of the phototag only at the focal plane, which avoids tagging the of material through the entire thickness of the sample,

165 allowing for high spatial resolution. Given the success of the proof of principle study using amyloid plaques in Alzheimer’s brain tissues, we attempted to perform STOMP on both SGs in cultured cells and on TDP-43-positive inclusions of FTD patients to compare their proteomic content and assess whether the SG precursor model of ALS pathogenesis is plausible.

166

Results and Discussion Application of STOMP to SGs in cultured cells

The first target chosen for STOMP analysis was SGs, as no previous attempts have been made for the comprehensive analysis of its proteome. The initial step in STOMP analysis requires a high-quality mask that allows the precise targeting of the ROI. To obtain a mask of perinuclear stress granules (SGs) in cell culture, our strategy was to induce cellular stress by sodium arsenite using established methods, lightly fix the cells to preserve SG structure, treat these fixed cells with DEPC to modify endogenous histidines, and stain the structures with an antibody against the SG protein HuR (Human antigen R, Parker et al. 2012). Since HuR is also a nuclear protein, we needed to modify the STOMP mask to exclude the nuclei. To do so, we used the DNA-specific dye 7-Aminoactinomycin D (7AAD) to stain the nucleus such that we can later produce a nuclei exclusion mask during imaging. The sample preparation was completed by using the phototag as mounting media (Figure 6.1).

Figure 6.1: Modified STOMP work-flow for stress granule analysis. Confluent cells are induced to form SGs by sodium arsenite. The SGs are lightly fixed to preserve morphology, treated with DEPC to modify endogenous histidines, stained by immunofluorescence using antibodies against SG-specific proteins (HuR), and co-stained with 7AAD. The stained cells are saturated with phototag in the mounting media in preparation for STOMP.

167

In the preliminary imaging, we successfully generated the final STOMP mask by subtracting the nuclei exclusion mask from the HuR mask. We then used the final mask in a round of targeted excitation with the STOMP macro and successfully photobleached the ROI while leaving the nuclei unbleached, suggesting specific targeting of SG ROI (Figure 6.2).

Figure 6.2: Preliminary photobleaching of SGs in microscope slides containing fixed HeLa cells under arsenite-induced stress. (A) The nuclei and perinuclear SGs are visualized by immunofluorescence using anti-HuR antibody. (B) The same field of view can be imaged by 7AAD nuclei fluorescence and converted into a binary, exclusion mask. (C) The final mask is generated by subtracting the exclusion mask from the HuR signal. (D) After targeted photobleaching using the final mask, the immunofluorescence of HuR was re-assessed to show photobleaching and specific excitation of the SGs.

We then tested whether a 2-week sample acquisition time was sufficient for MS/MS analysis by performing STOMP on a monolayer of stressed HeLa cells covering approximately half of an 18 ×

25 mm coverslip. To determine whether there was sufficient material we ran the purified, His6-tagged material from the photolabeled tissues on SDS-PAGE gel followed by silver staining and compared the band intensities with eluted material with a 6HisBP-soaked but non-photoexcited (dark) control (Figure 6.3A). The eluted fraction of the STOMP-treated sample showed a visually discernable increase in the intensity of bands compared to the dark control. Such a difference in a silver-stained gel typically suggests that there is sufficient material for MS/MS analysis, as we have used such a benchmark in our proof of principle study (Hadley et al. 2015). However, when 1.5× the amount of material was photolabeled and subjected to MS/MS analysis, no significant differences in proteomic composition was found between the dark and STOMP lanes, with the most abundant proteins as human keratin, suggesting background contaminants (Figure 6.3B).

168

Figure 6.3: STOMP analysis of SGs in HeLa cell culture. (A) Prepared, stressed HeLa cells on glass coverslips were subjected to two-photon excitation (STOMP) or untreated (dark) and solubilized and subjected to Ni-NTA affinity purification. The samples were eluted off the affinity column and loaded to a 12% SDS-PAGE gel and stained by silver staining. (B) A replicate of the samples loaded onto the SDS-PAGE gel, though with 50% more material, was subjected to MS/MS. The detected proteins are listed in table format.

The lack of hits in the MS/MS results can be caused by a number of issues at various steps in the STOMP protocol, and requires troubleshooting at individual steps of the multi-step STOMP process to determine whether the problem occurred at sample preparation, at imaging/photolabeling, or at the final purification and MS/MS stages. We initially performed troubleshooting procedures to verify

6HisBP integrity and the successful attachment of the His6 tag to the targeted ROI during microscopy. We did this by marking a region on the coverslip after STOMP and re-staining the tissue using anti-

His6 antibodies. These results suggest the successful attachment of His6 tag within proteins of the ROI after STOMP. The degree of attachment, such as number of tags per protein molecule, was difficult to verify.

Figure 6.4: Verification of phototag performance. (A) ESI-MS spectrum of synthesized 6HisBP indicating the relative abundance of the expected MW species at 1104 Da. (B) Verification of His6- tagging in regions specified by the mask file.

169

We don’t believe solubilization process was an issue, since the solubilization of lightly fixed cultured cells is considerably simpler than formalin-fixed brain tissues. We used the same solubilization conditions of 2% SDS and 8 M urea as the previous studies (Hadley et al. 2015). We next estimated whether the quantity of tagged material was truly accurately gauged by the silver stain gel. Previously, 2 × 106 µm3 of material was required for the analysis of amyloid plaques. However, due to the large size of the plaque, a 10× objective was sufficient for mask generation. Thus, the volume of material (in µm3) roughly correlated 1:1 with the number of pixels marked by the mask file. To photolabel this volume of material, only 16 h of two-photon beam time (~1 week of sample acquisition) was required (Hadley et al. 2015). However, due to the smaller size of SGs, the magnification must be increased in order to generate images of sufficient resolution to visualize the SGs and generate the STOMP mask. This meant that a 20× objective was needed, which decreased the volume-to-pixel ratio of tagged material to 1:4. This meant that the 3-week sample preparation of SGs for the initial analysis, which labeled approximately 1.2 × 106 pixels, only tagged an effective volume of 0.3 × 106 µm3. To tag the 2 × 106 µm3 volume used in the amyloid plaque study, the process would take approximately 4 months just to acquire sufficient material. Testing different parameters in the STOMP technique such as varying laser intensities or iteration numbers, or attempts to vary different solubilization or elution techniques, considering the sample preparation time, makes the optimization of this assay rather challenging. Due to the semi-automatic nature of the STOMP method, the current version of the technology to determine SG proteomics may not yet be feasible. Full automation of the method and better sensitivity of MS/MS technology to allow for the use of fewer amounts of material will be required to realize this project in the future.

170

Application of STOMP to FTLD-U inclusions with TDP-43 in archived brain tissue

Since SGs were near the technological limits of the STOMP technique, we next attempted to analyze the proteomics of larger targets such as the inclusions found in formalin fixed brains of FTD patients with ubiquitin positive inclusions (FTLD-U pathology). The samples we acquired for this type of tissue is similar in post-mortem preparation to those seen in the previous amyloid plaques study, and may be amenable to the same solubilization and purification protocols. Using the photobleaching method developed in previous chapters to remove background fluorescence, we generated clean immunofluorescence images of these inclusions co-stained with nuclear stain DAPI, and antibodies against both ubiquitin and TDP-43 (Figure 6.5). The morphology of oval and crescentic neuronal intracellular inclusions was similar to the inclusions visualized by immunohistochemistry in the literature (Reviewed by Mackenzie and Neumann 2016; Sun and Chakrabartty 2016). Interestingly, TDP-43-positive inclusions almost always co-stained with ubiquitin, but some ubiquitin-positive inclusions did not contain TDP-43 staining (Figure 6.5A). This suggests that other proteins may form ubiquitin-positive inclusions that do not contain TDP-43. We thus chose to use the anti-ubiquitin antibody to generate the STOMP mask in order to identify these components.

To select the initial STOMP laser settings of intensity and iteration number, we excited arbitrary regions in untreated tissue while varying laser power and iteration number, and confirmed the attachment of His6 tags within inclusions by anti-His6 immunofluorescence (Figure 6.5B, C). We found

that attachment of His6-tags occurred efficiently at 15% laser power and in only required 1-3 iterations, and that further repeated excitation (up to 8 iterations) did not result in a corresponding increase of

His6-tag immunofluorescence, suggesting saturation of labeling. We thus selected a setting of 15% laser power at 2 iterations, which ensured at least 1 extra round of photolabeling, while keeping the processing time to 50 h per 2×106 µm3 of tagged sample volume. We processed this sample volume for SDS-PAGE as previously described (Hadley et al. 2015).

171

Figure 6.5: Application of STOMP to FTLD-U inclusions. (A) Immunofluorescence images of an archived brain section a FTLD-U brain. The section was co-stained with anti-TDP-43, and anti- ubiquitin antibodies and counterstained with DAPI, as indicated by their respective colors. A merge image shows the colocalization of TDP-43 and ubiquitin staining in perinuclear inclusions. Scale bar = 100 µm. (B) Optimization of iteration number of laser power for STOMP of FTLD-U tissue section. A hand-drawn line of 50 µm was used as a mask file to excite an arbitrary region on a tissue section soaked in phototag. Various combinations of laser power and iterations were used at each region, and the slide was then immunostained using anti-His6 antibody to visualize degree of photolabeling. The intensity of fluorescence corresponds to increased His6 phototag attachment. Scale bar = 50 µm. (C) STOMP analysis in 2 fields of a tissue section stained with anti-ubiquitin antibodies, photolabeled at 15% laser power using 1 or 8 iterations. The tissue was then stained with anti-His6 antibodies to visualize attachment of His6 in inclusions. Inclusion fluorescence was measured after STOMP to check for photobleaching. Scale = 100 µm.

172

For this attempt, we split the processed sample and analyzed one half of the sample by western blotting and the other half with SDS-PAGE followed by silver staining (Figure 6.6). Since the major component of these inclusions is TDP-43, western blotting would be a sensitive indicator of TDP-43 enrichment in the photolabeled STOMP sample compared to the dark control. While eluate lanes of dark and STOMP appeared to contain material, no discernable difference in banding pattern or intensity was observed in the silver stain gel (Figure 6.6A). In the western blot, TDP-43 was present in the raw homogenate of both STOMP and dark lanes, but this band was lost after Ni-NTA purification (Figure 6.6B).

Figure 6.6: Silver stain and western blot of brain homogenate and Ni-NTA-eluted fractions of a FTLD-U tissue section analyzed by STOMP. (A) Silver-stain of the SDS-PAGE indicates similar amounts of material in homogenate and eluate fractions in both the STOMP-treated tissue and its dark control. (B) Western blot of the corresponding gel shows the presence of TDP-43 in the lysate fraction (arrow) but its absence in the eluate fraction in both STOMP and dark lanes. 20 ng of yTDP-43 was loaded as positive control.

173

We then assessed whether DEPC treatment or the Ni-NTA purification process was at fault. To do this, we irradiated a whole STOMP-ready tissue section with 360 nm UV light to non-specifically attach His6-tags to all proteins in the section. We then performed the same solubilization and Ni-NTA purification steps as before and compared the amount of His6-tagged material purified, using a dark, unirradiated section as control. The silver stained gel suggests a that the homogenates of both UV- irradiated and dark slides contain similar material. However, only Ni-NTA eluates from UV-treated slides contained enriched material, suggesting very low levels of background Ni-NTA binding. This suggests that the DEPC treatment of endogenous histidines was effective in removing non-specific binding, and that Ni-NTA purification should yield an enrichment of His6-tagged proteins if sufficient material was photo-labeled.

Figure 6.7: Enrichment of His6-tagged material from UV-irradiated tissue (UV) compared to untreated (dark) control. Homogenates and eluates from 2 brain sections soaked in phototag (UV- irradiate and dark control) were analyzed by SDS-PAGE and silver staining. Enrichment of eluate is observed only in the UV-irradiated sample.

174

Together, these results seem to suggest that His6 tags were being attached to proteins within these inclusions, and that DEPC pre-treatment and Ni-NTA binding should result in enrichment of protein samples if sufficient material was tagged. This led us to believe that the solubilization of these inclusions was the problem, and that not enough tagged material was being extracted. A few changes to the solubilization protocol were attempted following more STOMP trails, such as using only Tris- SDS without urea, but these attempts still did not generate a significant difference in the silver stain gel lanes between STOMP’ed and control samples. Additionally, further attempts were made on FTLD-T tissues stained for pathological tau tangles in the hope of having more material per field of view to decrease labour time, but despite the relatively high amounts of tangles present in the samples, a similar problem with solubilizing the tau inclusions arose, where the majority of homogenate material remained in the stacking gel, suggesting the inability of our solubilization protocols to dissociate the tau tangles effectively. As the attempts increased, it also became difficult to judge whether there was enough material being tagged, or how much tagging would result in sufficient material. As a result, the use of the technique has been paused until a more user-friendly method of tagging the sample, such as fully automating the technique in a slide-scanner-like fashion, should be developed.

The feasibility of STOMP analysis on TDP-43-positive cellular structures

The attempts at identifying the proteomic compositions of SG and FTLD-U inclusion using the STOMP technique were largely unsuccessful. Despite this, the technique itself appears logically sound and has shown to be successful in its application on AD tissues. It appears that in the current state, STOMP can be optimized for a single target of interest (such as amyloid plaques in AD) but requires substantial number of trials just to assay one phototagging or solubilization condition, making optimizations time-consuming and impractical until full-automation can be achieved. For small inclusions in ALS/FTD or smaller sub-micron sized bodies such as SGs, the processing time for these targets almost necessitates a fully-automated system.

In terms of identifying linkages between SG formation and TDP-43 inclusions, the evidence for this theory continues to grow in literature. Although the comprehensive, proteomic approach has stalled, this ultimately diverted our research focus towards more hypothesis-driven approaches (Chapter III), where we identified electrolytes as a key controller of TDP-43 aggregation and behavior in the droplet phase.

175

Materials and Methods Induction and visualization of stress granules

HeLa cells were grown on 18 mm x 25 mm glass coverslips placed in a 6-well cell culture plate (Nunc) in Dulbecco’s Modified Eagle Medium supplemented with 10% fetal bovine serum (DMEM-FBS). Cells were stressed by replacing media with fresh DMEM-FBS and 1 mM sodium arsenite and incubated for 1 h at 37 ºC. Stressed cells were washed twice with PBS and fixed for 15 min in 1% formaldehyde. The cells were treated with 2.5 mM diethylpyrocarbonate for 15 min to modify endogenous histidines. Cells were then permeabilized with 0.1% Triton-x 100 and blocked with 10% BSA in PBS for 2 h at RT, incubated with 1:200 dilution of primary rabbit anti HuR antibody (Millipore) at 4 ºC, and incubated with 1:200 dilution of Alexa488-conjugated goat anti-rabbit secondary antibody. The cells were then stained with 7AAD nuclear stain 1:1000 stock concentration for 10 min. The slide was then mounted to a glass microscope slide using 5 mM 6HisBP photo-tag as the mounting medium. The coverslips are then sealed with nail polish and stored at 4 ºC for STOMP analysis.

Immunofluorescence staining of FTLD-U cases

Formalin-fixed brain blocks of the orbitofrontal gyri in a case of FTLD-U (~2 days fixation) were run through 10%, 20%, and 30% sucrose gradients, infiltrated with optimal cutting temperature (OCT) compound, frozen, and cut into 10 µm thick sections and attached to glass microscope slides. Slides were photobleached for 48 h in the photobleaching apparatus prior to staining. The slides were treated in antigen retrieval buffer (10 mM Citric Acid, 2 mM EDTA, 0.05% Tween 20, pH 6.2) for 30 min at 90 °C, then washed twice for 5 min in TBS plus 0.025% Triton X-100 (TBS-Triton). The slides were blocked in 10% normal goat serum (Aurion, Wageningen, The Netherlands) and 1% BSA (ThermoFisher Scientific, Waltham, MA, USA) in TBS-Triton for 2 h at room temperature. For ubiquitin and TDP-43 double labeling, a mixture containing primary mouse anti-ubiquitin antibody (1:50, Abcam) and primary rabbit anti-TDP-43 (ProteinTech, 1:200) in 1% BSA/TBS was applied overnight at 4 ºC in a humidified chamber. The slides were rinsed 2 × 5 min with TBS-Triton and a mix of secondary antibodies consisting of Alexa 488 goat anti-rabbit (1:100, ThermoFisher Scientific, Waltham, MA, USA) and Texas red goat anti-mouse (1:100, ThermoFisher Scientific, Waltham, MA, USA) in 1% BSA/TBS was applied for 1 h at room temperature in the dark. The slides were rinsed 2 × 5 min in TBS and stained with DAPI (0.25 µg/mL, ThermoFisher Scientific, Waltham, MA, USA) for 10 min. For visualization, the slides were mounted in Immu-mount aqueous mounting medium

176

(ThermoFisher Scientific, Waltham, MA, USA). For samples to be photolabeled using STOMP (or its associated dark control), only ubiquitin antibody and Alexa488 secondary antibody was used for tissue staining. The stained tissue is the mounted in 6HisBP phototag (5 mM) and sealed with coverslip. To

visualize His6 attachment to the regions of interest, samples that have undergone 2-photon excitation

were detached from the coverslips, re-stained using primary rabbit anti-His6 antibodies (1:100) and goat anti-mouse Texas Red secondary antibodies (1:200). All images were taken using a LSM710 two- photon confocal microscope using Zen Black software (Zeiss).

STOMP analysis and purification

Samples undergoing STOMP was carried out in the same manner as previously described (Chapter V, (Hadley et al. 2015)). The parameters used for STOMP of stress granules was 9.0% laser power, 4 iterations, the STOMP settings for FTLD-U inclusions varied depending on the conditions tested. Purification, SDS-PAGE, and silver staining followed the same protocols mentioned in the previous chapter. For western blots, proteins were transferred to nitrocellulose membrane, blocked in 1% BSA, incubated with TDP-43 primary antibodies (ProteinTech 1:500) and Li-Cor secondary antibody (1:2000). Images of western blots were generated from inverting the raw Li-Cor output (fluorescent bands on a dark field).

177

Concluding Remarks and Future Directions Impact and significance

The studies presented in this thesis have revealed unique and previously unidentified properties of TDP-43 aggregation and phase separation, and produced novel imaging methods to visualize and analyze structural features in human tissue. Although the exploratory approach to identify common proteins shared between TDP-43-positive pathological inclusions and stress granules was not successful, the validation of the STOMP method in AD tissue implies that the automation of this technique combined with future sensitivity improvements to mass spectrometry may help realize the full potential of this technology such that this comparison can eventually be made. In optimizing the sample preparation protocols to produce STOMP masks, an inexpensive and simple method for the removal of background autofluorescence in aged human tissue was developed. This technique is inexpensive, can be implemented in any laboratory, and has a wide range of applications, since tissue autofluorescence has always complicated the immunostaining of post-mitotic samples such as liver, heart, and brain tissues.

The most prominent findings in this thesis centers around the in vitro characterization of TDP-43 aggregation using purified protein. Historically, TDP-43 has been notoriously difficult to purify, but after identifying the underlying electrolyte-dependence of aggregation of TDP-43, we have developed a suitable method to produce full-length protein at high concentrations, which allows for the protein’s structural characterization by NMR spectroscopy. In characterizing TDP-43, we have discovered that unlike most protein misfolding diseases, where the pathological protein forms ordered, stable structures, aggregates of TDP-43 appear far more complex and behave differently than typical protein aggregates. These aggregates do not exhibit cross-β structural features, and unlike Aβ plaques or PrP fibrils, possess the remarkable property of aggregation reversibility. This reversibility has not been reported previously in aggregates implicated in neurodegeneration, and is highly atypical of general protein misfolding, where the typical end-product is an irreversible protein aggregate.

The sub-molar, electrolyte-specific induction of TDP-43 aggregation and the ability for TDP-43 to enter protein droplet scaffolds are both dependent on its CTD, which is capable of forming phase- separated droplets in isolation. This provides further evidence for the suggestion that formation of these aggregates and the assembly of protein into the droplet state are driven by the same electrolyte

178

effects and are likely occurring concurrently. Interestingly, the properties of these droplet-forming proteins may have been inadvertently described in 1941, when Kirkwood thought of proteins in general as long chains of dipoles. The models developed during that era to incorrectly characterize stable, folded proteins may in fact be a suitable and well-studied theory to predict the behavior of intrinsically disordered proteins. The use of Ddx4N1 scaffold to characterize TDP-43 aggregation in the droplet state could also be applied to future experiments to assess the effects of ALS/FTD- associated mutations on the electrolyte dependence of TDP-43 aggregation.

The data presented in this thesis challenges the concept that ALS is a typical neurodegenerative disease caused by the age-related failure of the protein maintenance machinery. We found that TDP- 43 aggregation can be induced by a variety of changes to its cellular environment including the availability of cognate RNA binding partners, prolonged or repeated association and disassociation from stress granules during the stress response, or sudden changes in local electrolyte concentrations. The lack of ordered aggregate structures, the reversibility of aggregation, and the sensitivity to numerous factors that modulate TDP-43 aggregation both in solution and in the droplet-state suggests a multi-hit model of disease where protein aggregation and its associated pathology can be triggered by any combination of environmental or genetic factors. The reversibility of TDP-43 aggregation also suggests that the interactions between TDP-43 aggregates are not tight, but are rather labile, weaker interactions that can be disrupted. This implies that compounds that can dissociate these aggregates may function as drugs to prevent pathology caused by the TDP-43 aggregates’ toxic gain of function. The implementation of such treatments, however, need to be applied to patients before significant damage has been done to the effected neurons, which brings into focus the current unmet need for an effective clinical diagnosis of ALS/FTD.

Misfolded TDP-43 as a biomarker and therapeutic target for ALS

Current studies of TDP-43 proteinopathies such as ALS and FTD are focused on the pathological or the physiological functions of the protein as it relates to disease pathogenesis. Understanding the early molecular signs of ALS and identifying biomarkers for the disease would allow for earlier diagnoses to be made and drug trials to be carried out at a stage of the disease where neuron death can still be prevented. Currently, drugs for the treatment of ALS such as Riluzole have largely been ineffective, because these treatments have always been administered to patients with full-blown disease. It is generally agreed upon that the sooner a neurodegenerative disease is diagnosed, the higher the probability that it can be successfully treated. Unfortunately, ALS is typically diagnosed

179

approximately 2 years from symptom onset (Brooks 1999). One reason for the lengthy time required for diagnosis is the absence of a diagnostic laboratory test. Currently, ALS is an exclusion diagnosis that relies on techniques such as electromyography, nerve conduction studies, and magnetic resonance imaging to rule out other more treatable diseases that show similar symptoms to ALS (Brooks 1994).

Due to the lack of diagnostic tests for sporadic ALS, the search for diagnostic biomarkers of ALS is currently in full force. We are focused on TDP-43 as our diagnostic biomarker because TDP-43 pathology is seen in 97% of all cases of ALS (Ling et al. 2013), and from the work presented here, the amount misfolded species of the protein in cerebrospinal fluid (CSF) or plasma of patients may be a good indicator of pathology. Others groups have attempted to quantify total TDP-43 in the CSF of ALS cases and pathological controls using ELISA-based detection methods. Unfortunately, these strategies were not sensitive enough to be used for diagnostic purposes because the true biomarker for ALS is misfolded TDP-43 and not total TDP-43 (Noto et al. 2009; Kasai et al. 2011).

Currently, we are developing antibodies that specifically bind to the pathological misfolded form of TDP-43 but not to the natural, natively folded form of the protein. The properties of TDP-43 identified in this thesis allows us to effectively manipulate the folded or aggregated state of TDP-43, through addition of electrolytes or poly-TG single-strand DNA to assay the specificity of our antibodies. The Chakrabartty lab has had a successful history of using this misfolding-specific antibody approach to produce antibodies for the treatment, diagnosis or research of protein misfolding diseases. The most recent project being a misfolding-specific antibody against transthyretin (TTR) for the treatment of TTR cardiac amyloidosis (Galant et al. 2016). This project has led to industrial partnerships and a phase 2 clinical trial scheduled in early 2018.

We used the same strategy to produce antibodies against misfolded TDP-43, and we believe that early diagnosis is a key step in finding an effective treatment for ALS. We’ve established partnerships with ALS clinicians to acquire patient samples to validate antibody specificity in tissue and to determine whether antibody-based detection techniques can be used on CSF to differentiate ALS cases from pathological controls. Our ultimate goal is to enable clinicians to diagnose ALS early, track disease progression, and provide an earlier opportunity for patients to enroll in clinical trials, when the most promising therapeutics can have optimal effect.

180

References

Brooks BR. 1999. “Diagnostic dilemmas in amyotrophic lateral sclerosis.” J Neurol Sci. S1: S1-9.

Brooks BR. 1994 “El Escorial World Federation of Neurology criteria for the diagnosis of amyotrophic lateral sclerosis. Subcommittee on Motor Neuron Diseases/Amyotrophic Lateral Sclerosis of the World Federation of Neurology Research Group on Neuromuscular Diseases and the El Escorial "Clinical limits of amyotrophic lateral sclerosis" workshop contributors.” J Neurol Sci. 124 Suppl: 96-107.

Emde, A., C. Eitan, L.-L. Liou, R. T. Libby, N. Rivkin, I. Magen, I. Reichenstein, et al. 2015. “Dysregulated miRNA Biogenesis Downstream of Cellular Stress and ALS-Causing Mutations: A New Mechanism for ALS.” The EMBO Journal 34 (21): 2633–51. doi:10.15252/embj.201490493.

Forman, Mark S, John Q Trojanowski, and Virginia M-Y Lee. 2007. “TDP-43: A Novel Neurodegenerative Proteinopathy.” Current Opinion in Neurobiology 17 (5): 548–55. doi:10.1016/j.conb.2007.08.005.

Hadley, Kevin C, Rishi Rakhit, Hongbo Guo, Yulong Sun, James E N Jonkman, Joanne McLaurin, Lili-Naz Hazrati, Andrew Emili, and Avijit Chakrabartty. 2015. “Determining Composition of Micron-Scale Protein Deposits in Neurodegenerative Disease by Spatially Targeted Optical Microproteomics.” eLife 4 (September): 1–21. doi:10.7554/eLife.09579.

Kasai, Takashi, Takahiko Tokuda, Noriko Ishigami, Hiroshi Sasayama, Penelope Foulds, Douglas J. Mitchell, David M A Mann, David Allsop, and Masanori Nakagawa. 2009. “Increased TDP-43 Protein in Cerebrospinal Fluid of Patients with Amyotrophic Lateral Sclerosis.” Acta Neuropathologica 117 (1): 55–62. doi:10.1007/s00401-008-0456-1.

Li, Y. R., O. D. King, J. Shorter, and A. D. Gitler. 2013. “Stress Granules as Crucibles of ALS Pathogenesis.” The Journal of Cell Biology 201 (3): 361–72. doi:10.1083/jcb.201302044.

Ling SC, Polymenidou M, Cleveland DW. 2013. “Converging mechanisms in ALS and FTD: disrupted RNA and protein homeostasis.” Neuron 79 (3): 416-38.

181

Martins-de-Souza, D, P C Guest, D M Mann, S Roeber, H Rahmoune, C Bauder, H Kretzschmar, B Volk, a Baborie, and S Bahn. 2012. “Proteomic Analysis Identifies Dysfunction in Cellular Transport, Energy, and Protein Metabolism in Different Brain Regions of Atypical Frontotemporal Lobar Degeneration.” J Proteome Res 11 (4): 2533–43. doi:10.1021/pr2012279.

Ratti, Antonia, and Emanuele Buratti. 2016. “Physiological Functions and Pathobiology of TDP-43 and FUS/TLS Proteins.” Journal of Neurochemistry, 138 (S1):95–111. doi:10.1111/jnc.13625..

Schweitzer, Kelly, Emily Decker, Liping Zhu, Richard E Miller, Suzanne S Mirra, Salvatore Spina, Bernardino Ghetti, Mu Wang, and Jill Murrell. 2006. “Aberrantly Regulated Proteins in Frontotemporal Dementia.” Biochemical and Biophysical Research Communications 348 (2): 465–72. doi:10.1016/j.bbrc.2006.07.113.

Seyfried, Nicholas T, Yair M Gozal, Laura E Donovan, Jeremy H Herskowitz, Eric B Dammer, Qiangwei Xia, Li Ku, et al. 2012. “Quantitative Analysis of the Detergent-Insoluble Brain Proteome in Frontotemporal Lobar Degeneration Using SILAC Internal Standards.” Journal of Proteome Research 11 (5): 2721–38. doi:10.1021/pr2010814.

Sun, Yulong, and Avijit Chakrabartty. 2017. “Phase to Phase with TDP-43.” Biochemistry 56 (6): 809– 23. doi:10.1021/acs.biochem.6b01088.

Wheeler, Joshua R., Saumya Jain, Anthony Khong, and Roy Parker. 2017. “Isolation of Yeast and Mammalian Stress Granule Cores.” Methods 126: 12–17. doi:10.1016/j.ymeth.2017.04.020.