Molecular Neuropathology in Alzheimer's Disease

Molecular Neuropathology in Alzheimer’s Disease

Dissertation

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy

in the Graduate School of The Ohio State University

Carol Huseby, M.S.

Graduate Program in Biophysics

The Ohio State University

2018

Dissertation Committee:

Jeffrey Kuret, Ph.D., Advisor

Ralf Bundschuh, Ph.D.

Kari Hoyt, Ph.D.

Sherwin Singer, Ph.D.

Copyrighted by

Carol Huseby

2018

Abstract

Alzheimer’s disease affects one in ten Americans and incidences are on the increase while other leading causes of disease are declining. Alzheimer’s disease is defined by two proteinaceous lesions in brain, extracellular Aβ plaques and intracellular tau neurofibrillary tangles. Tau protein is a microtubule-associated protein found abundantly expressed in neurons. Its function is to stabilize and promote microtubules and neurite outgrowth. In

Alzheimer’s disease, tau protein is hyperphosphorylated and dissociates from microtubules to form aggregate filaments inside neurons disrupting the transport along axons within neurons. A remarkable aspect of the AD disease progression is the range of brain regions that are affected in a systematic, sequential manner maintaining a predictable distribution pattern of intraneuronal tau lesions varying little between individuals and an extended prodromal period. The progression of neurofibrillary tau tangle lesions in disease behaves like a prion, propagating neuron-to-neuron recruiting new tau molecules to mis-fold and continue the disease process. The mechanisms of tau protein aggregation are not clear. In this thesis I explore tau protein aggregation using biochemical methods and mathematical models and/or analysis to help clarify the aggregation mechanisms of tau protein leading to potential therapeutic strategies or targets.

Dedication

I dedicate this dissertation work to my two wonderful children I love very much,

John and Emry, who supported my decision to return to school and were my best cheerleaders throughout my doctorate career. I dedicate this work to my family, my magnificent brothers and sisters; Mike, Ian, Shonda, Wade, Kyle, Adam, Jed, Eileen, Peter,

Rebecca, Sam, Audrey, and Susan. I especially dedicate to my parents, who have always let us be ourselves and supported their children’s endeavors. Although my father, Donald

Gene, has passed since I began my doctorate career, my mother, Marian, continues to provide me with encouragement and praise.

iii

Acknowledgements

I would like to acknowledge my advisor and my committee members for their expertise and generous time. A special thanks to my advisor, Dr. Jeff Kuret, for countless hours of training, encouragement and patience throughout my entire doctoral career, as well as, thanks for the support and opportunity to conduct interdisciplinary research in your laboratory. Thank you Dr. Ralf Bundschuh, Dr. Kari Hoyt, and Dr. Sherwin Singer for serving on my committee.

I thank the OSU Campus Microscopy and Imaging Facility for access to electron and fluorescent microscopy resources as well as the OSU Neuroscience Imaging Core for use of the confocal microscopes.

I thank the Targeted Metabolomics Laboratory at The Ohio State University for access to LC–MS/MS equipment (funded by the Translational Plant Sciences Targeted Investment in Excellence). I also thank Jean Christophe Cocuron and Dr. Anna Alonso, at the

BioDiscovery Institute, University of North Texas for LC-MS/MS analysis of samples.

This work was supported by Public Health Service grants NS77441 and AG54018 and also by NIH grant AG14452.

Vita

2012-2018 Graduate Research Associate, Interdisciplinary Biophysics Graduate Program,

Biological Chemistry and Pharmacology, The Ohio State University, Columbus,

OH, Ph.D. Candidate (defense 2018).

2010-2012 Master of Science in Applied Mathematics, University of Washington, Seattle, WA.

2006-2010 Bachelor of Science in Physics, University of Washington, Seattle, WA.

Bachelor of Science in Biology (Physiology), University of Washington, Seattle,

WA.

Publications

1. A liquid chromatography tandem mass spectroscopy approach for quantification of protein

methylation stoichiometry. (2018) G. Cooper-Olson**, C.J. Huseby**, C.N. Chandler, J.C.

Cocuron, A. Alonso, J. Kuret, Analytical Biochemistry Feb 2, 545, 72-77. **equal contribution

2. Structural determinants of Tau aggregation inhibitor potency. (2013) K. Schafer, K. Cisek, C.J.

Huseby, E. Chang, J. Kuret; J. Biol. Chem. 288 (45), 32599-32611.

3. Analyzing tau aggregation with electron microscopy. (2016) C.J. Huseby and J. Kuret; Meth.

Mol. Bio.1345, 101-12.

4. Structure and mechanism of action of tau aggregation inhibitors. (2014) K. Cisek, G. Cooper,

C.J. Huseby, J. Kuret; Current Alzheimer Research. 11(10), 918-27.

Field-of-Study

Major Field: Biophysics

Table of Contents Abstract…………………………………………………….…………………..…………ii Dedication…………………………………………….…………………….….…….…..iii Acknowledgments……………………………………..…………..…………..…...…….iv Vita.…………………………………………………………………………..….………..v List of Tables…………………………………………………………….….…..………vii List of Figures………………………………………………………….……..….……..viii List of Abbreviations…………………………………………………………..…….…...x Chapter 1. Introduction…………………………………………………………..….……1 Chapter 2. Analyzing tau aggregation with electron microscopy……….…….….…..…26 Chapter 3. The role of annealing and fragmentation in tau aggregation dynamics.…...... 43 Chapter 4 Liquid chromatography tandem mass spectroscopy approach……..…….…..72 Chapter 5 LC-MS/MS measure of stoichiometry tau methylation in human samples…..91 Chapter 6 Toward a more robust gene expression signature for AD…………..……..….104 Chapter 7 Conclusions and perspectives……………..…………………….……..….…117 References………………………………………………………………………………121

List of Tables

Table 1 Percent affected with Alzheimer’s dementia in U.S. by age group…….………..7 Table 2 List of diseases associated with tau protein lesions…………………….……....23 Table 3 Model parameters…………………………………….……………….….……..67 Table 4 Mass spec parameters and calibration of amino acids………………….………82 Table 5 Human tissue case demographics…………………………………..….….…..102

vii

List of Figures

Figure 1 Genetic loci of AD risk factors.……………………………….…………………4 Figure 2 Confocal images of Aβ plaques and neurofibrillary tangles……………..….….11 Figure 3 Model of the progression of dynamic biomarkers in AD………………………13 Figure 4 Tau pathology looks like neuron-to-neuron propagation………………………17 Figure 5 Cross-sectional study of disease progression in human brain tissue….……..…19 Figure 6 MAPT gene and tau protein schematics……………………….……………….21 Figure 7 Technique for applying tau samples to grids…………………………….….….31 Figure 8 Electron micrograph of tau filaments binned as function of length……………33 Figure 9 Graphical depiction of parameter estimations…………………….………..…..35 Figure 10 Immunogold labeling of epitope-tagged tau filaments…………..……..……..38 Figure 11 Aggregation model scheme……………………………………………..…….46 Figure 12 Construct comparison of aggregation propensity and morphology…….…….51 Figure 13 Immunogold labeling controls……………………………………….………..52 Figure 14 Direct visualization of tau filament annealing………………………….….….54 Figure 15 Annealing time-course approximates second-order kinetics……..………..….56 Figure 16 mathematical model of tau aggregation…………………………….……..….59 Figure 17 Model fit to protomer concentration and length distributions………..….……61 Figure 18 Global sensitivity analysis of NEAFS model………………………..……..…63 Figure 19 Model fits to protomer concentration time-series…………………..….……..65 Figure 20 Model fits to length distribution time-series………………………………….66 Figure 21 Tau filament annealing fragmentation equilibrium is length dependent…..….70 Figure 22 Ion chromatograms for amino acid standards……………………………..….80 Figure 23 Quantification of Lys methylation stoichiometry in tau protein……………...85 Figure 24 Quantification of methylation stoichiometry in protein standards……………88

viii Figure 25 Western blots and silver stain of human tissue samples………………..……..98 Figure 26 Bulk methyl Lys and methyl Arg in human tau preparations……….……....100 Figure 27 Methylation of tau protein in HEK293 cells…………………….………..…103 Figure 28 Cohort distributions NCBI-GEO database accession number GSE44772…..108 Figure 29 Comparison volcano plots for change of expression in AD…………………110 Figure 30 Fraction of average trends in covariates……………………………………..112 Figure 31 Aβ protease expression changes in AD…………………………….………..113 Figure 32 Methyltransferases and de-methylases expression changes in AD…….……115

List of Abbreviations

1meK, 2meK and 3meK, Nᵋ-(methyl)-, Nᵋ-(dimethyl)-, and Nᵋ-(trimethyl)-L-lysine, respectively; 1meR, NG-Methyl-L-arginine; Aβ, Amyloid-beta; AD, Alzheimer’s disease/dementia; CR, cerebellum; CSF, Cerebral spinal fluid; CJD, Creutzfeldt-Jacob disease; ADMA, NG,NG′-Dimethyl-L-arginine; APOE, Apolipoprotein E; APP, Amyloid precursor protein; DE, differential expression; DLB, Dementia with Lewy bodies; DS,

Down syndrome; ECL, Electrochemiluminescence; EOAD, Early-onset Alzheimer’s disease; HEPES, 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid; HRP, horseradish peroxidase; LC-MS/MS, liquid chromatography–tandem mass spectrometry;

LOAD, Late-onset Alzheimer’s disease; MBP, myelin basic protein; MCI, Mild cognitive impairment; MRI, Magnetic resonance imaging; MRM, multiple reaction monitoring;

PBS, phosphate-buffered saline; PET, positron emission tomography; PFC, pre frontal cortex; PMI, post mortem interval; PSEN1, Presenilin-1; PSEN2, Presenilin-2; PTM, post- translational modification; PVDF, polyvinylidene fluoride; SDMA, NG,NG′-Dimethyl-L- arginine; SNP, Single-nucleotide polymorphism; TEM, transmission electron microscopy;

TR, Thiazin red; UA, uranyl acetate; VC visual cortex

Chapter 1. Introduction

One of the most feared illnesses of the elderly is the fear of losing one’s mind, forgetting the people close to you, being ostracized, and becoming a full-care burden to loved ones. Slowly, and even to this day, unavoidably, in dementia diseases the most important parts of a life deteriorate making future family plans centered around economical costs of care and constantly adjusting behavioral modifying treatments. Those that have family history of dementia are at risk [1]. Betty, at age 80 years, feared that she would develop dementia because her father passed with dementia. She vividly recalls the day he came home from the hardware store and revealed despairingly that he couldn’t remember how to write a check. The loss of cognitive abilities such as writing a check can be a classical symptom of dementia disease signaling clinical onset. Early on a sufferer can display impaired memory or mild cognitive impairment (MCI) which can go undiagnosed as normal aging [2]. Recently, I attended a wedding where many guests, upon hearing that

I was working in Alzheimer’s Disease (AD) research, gathered at my dinner table to find out if there is anything more they should be doing to prevent the onset of this most feared condition. There does not yet exist a proven moderator for the progression of a neurodegenerative dementia disease [3]. The best strategies seem to be targeting the modifiable risk factors and focusing on prevention of the comorbidities of dementia disease, e.g., diabetes, cholesterol levels, high blood pressure, or more simply put, the best recommendation is that of a healthy lifestyle. But people naturally have the tendency to search for ‘hope’ especially those which have a family history of illness [4]. The work

1 presented here in my dissertation is work toward ‘hope’ and trying to find answers to the mechanisms that manifest dementia subsequently leading to treatment strategies for slowing or halting the disease progression.

Prevalence of dementia

Dementia is a group of heterogeneous clinical and pathological characteristics caused by neurological disorders of the brain affecting distinct groups of neurons in specific functional anatomical systems resulting in loss of memory, declining cognitive abilities, and dysfunctional social interactions. Some dementias are reversible like those due to certain vitamin deficiencies, thyroid problems, depression, sepsis, medication side- effects, and excessive use of alcohol. Dementia that is a result of an underlying neurodegenerative disease remains to this day irreversible [3].

There are approximately 70 different progressive neurodegenerative disorders leading to dementia of which Alzheimer’s Disease (AD) is the leading cause accounting for an estimated 60% to 80% of all dementia cases. Lewy body dementia (DLB) ranks next just ahead of vascular dementia, and the remaining explained by dementias like Pick’s disease, Creutzfeldt-Jacob’s Disease (CJD), and Huntington’s disease [3]. Accumulating research recognizes that it is more common that the pathologic changes for different dementias coexist especially AD pathology with that of vascular dementia and/or DLB.

This can give rise to overlapping clinical symptoms termed mixed dementia [5-7]. In fact,

Auguste Deter, the famous dementia patient of Alois Alzheimer, had marked atherosclerosis of the arteries in her brain [8]. Research shows that many cases of probable

AD have mixed pathologies at autopsy and only a fraction of all AD cases can be described as pure AD with only plaques and tangles present in parallel to brain atrophy [5, 9]. Are

2 comorbidities a cause or result of neurodegenerative diseases or are they just representative of overlapping human diseases? If one were to eat healthy, exercise, attenuate blood pressure, diabetes, and cholesterol, then there should be a low risk of a disease such as AD.

If ‘unhealthy’ behaviors are related to the neurodegenerative disease mechanism, then a

‘healthy’ lifestyle should reduce incidence. There exists much debate suggesting this is true

[10, 11], but the bulk of research publications also present evidence that there must exist other more sinister triggers for neurodegenerative diseases. It is true that genetic makeup interacting with the environment plays a role in many human diseases.

With recent work revealing more than 40 genetic variant risks for AD, environmental influences contributing to risk of developing neurodegenerative diseases continues a steady move to the forefront of dementia research. The combination of a genetic profile with that of environmental influences must contribute to risk of developing neurodegenerative diseases (Fig. 1) [12]. What these environmental influences might be is a major research area in dementia. This research proves challenging to identify and measure environmental exposures that may have occurred years before a person begins displaying signs of cognitive decline.

Most degenerative diseases have associated with them a variant such as a gene mutation that makes one more susceptible to developing the disease [13]. For AD, most cases are considered sporadic disease onset or late onset Alzheimer’s disease (LOAD) and only a small subset which comprises less than 1% of the cases are early onset Alzheimer’s disease (EOAD) that are a result of three rare genetic mutations (Fig. 1) [3]. When the disease occurs at an early age, usually before 65 years of age, these cases are usually familial, being inherited in an autosomal dominant fashion.

Fig. 1. Genetic loci of identified risk factors for AD. Each locus

may contain multiple genes associated with AD. The top two risk

factors are considered casual genes being fully penetrant for AD

with PSEN1 having the earliest onset. The remaining single-

nucleotide polymorphism (SNP) variants identified below the top

two risk factors have varying risk associated with them ranked

highest from top to bottom. Environmental influences interacting

with genetic risk factors are believed to be involved [14]. Risk loci

for genome-wide association with AD significance (푝 < 5푥10−5)

from meta-analysis including 89,769 individuals.

The three known genes which result in familial AD are PSEN1, PSEN2, and APP.

Presenilin-1 (PSEN1) is involved in brain and spinal cord development as well as the processing of Amyloid precursor protein (APP) into various smaller fragments. There are

30 families worldwide that have one of the more than 185 mutations known in PSEN1 to result in the outcome to AD. These mutations involve 75% of familial AD cases and age of onset is early at 35-55 years [15].

Presenilin-2 (PSEN2) is involved in cell growth and maturation as well as the processing of APP. There are 13 mutations resulting in AD found in Volga-German families and Italian families worldwide. These mutations are found in 5-10% of familial

AD cases. The age of onset is 40-75 years old with a mean age of 55 years. The PSEN2 4 mutations are not necessarily autosomal dominant and there exists carriers without dementia [16].

Amyloid precursor protein (APP) is the third gene mutation found in familial AD and has the structural characteristics of a cell surface receptor. The APP gene is located on chromosome 21q21 and is ubiquitously expressed as a type1 transmembrane protein having an extracellular domain which may interact with signaling ligands and a cytoplasmic domain which may translocate to the nucleus for end-point signaling after proteolytic cleavage releases it from the membrane [17]. Cleavage of APP by the β-secretase and subsequently γ-secretase results in Aβ peptide associated with the buildup of amyloid plaque in AD. There are 25 mutations (http://www.molgen.ua.ac.be/admutations) known for APP leading to AD and these cases make up 15% of familial AD cases with an age of onset of 40-60 years. But the APP gene is also suspect in Down’s syndrome (DS) which is a genetic disorder coming about as a result of an additional copy of Chromosome 21 or a portion of it. A large percentage of people with DS develop AD in their 50s possibly attributable to excess amyloid precursor protein accumulation [3, 18].

Apolipoprotein gene (APOE), located on chromosome 19, was established as a susceptibility risk gene for LOAD toward the end of the last century [19]. There are three alleles for the APOE locus; e2, e3, and e4. There is an increased risk for AD when carrying one e4 allele that is four-fold higher compared to those with no e4 alleles. With homozygous e4 alleles, the increased risk for AD is up to 12-fold higher. The molecular mechanistic pathway by which APOE increases the disease risk is not well understood. It is thought that the difference in the alleles may have implications on the homeostasis of

5 cholesterol and phospho-lipids, synaptic plasticity, neuroinflammation, amyloid metabolism, and neuronal survival [20-23].

In the previous decade, progress has been made to circumvent deaths due to major chronic diseases, such as, HIV, cancer, cardiovascular disease, and stroke, but deaths due to Alzheimer’s disease have increased as much as 85 percent in the same time period. In the next ten years, it is predicted for 10 percent of people over age 65 to have dementia and that people over the age of 85 years will have a 50 percent chance of developing AD [3].

In 2016, AD was the sixth leading cause of death in the United States. Some researchers believe that AD is grossly under-reported on death certificates since ultimately the death is due to some manifesting infection, such as, pneumonia, after neurodegeneration affects the whole-body systems [3]. As medicine continues to improve, we can expect to live beyond age eighty-five. This makes it more probable that every family will eventually be affected by AD.

It is now estimated that there are 5.7 million Americans living with Alzheimer’s dementia and this number continues to grow rapidly (Table 1 [3, 24]). In 2011, a task force assembled to review the criteria and guidelines created in 1984 for diagnosing AD. The task force proposed that Alzheimer’s disease be dichotomized to include those with clinical biomarkers of disease without cognitive deficit and those with clinical biomarkers and a clinical diagnosis of cognitive deficits will have Alzheimer’s dementia. This change was made in response to the acknowledgement that the defining pathology of AD begins years before dementia onset [3, 25]. Accumulating research shows that many people over age 65 years bear amyloid beta plaques as well as neurofibrillary tangles of the tau protein without major cognitive deficits [26-29].

Table 1. Percent affected with Alzheimer dementia in the United States by age group [24] Greater than 84-years-old 36%

Age 75 – 84 years 44%

Age 65 – 74 years 16%

Less than 65-years-old 4%

With this steady increase of dementia onset in the United States, the financial costs for AD are staggering. Disease management costs of AD reached $226 billion in the United

States in 2017 and $604 billion worldwide [3]. In 2015 the World Health Organization estimated that the total number of people with AD worldwide will rise to 135.5 million by

2050. The global costs will surpass twenty trillion dollars. This figure doesn’t include the unreported amount of unpaid caregiver hours. In 2015 it was estimated that unpaid caregiver hours amounted to eighteen billion hours [30]. The demands of this disease on families, our health care system, taxpayers, and the economy is alarming with the outcome of becoming ill-equipped to handle the need of patients and families touched by AD. There is a need to advance research toward methods of treatment, prevention, support for families affected by dementia and, ultimately, a cure for AD.

History of dementia research leaves us clues

Mental decline and decay with human aging has been documented for centuries as far back as Pythagoras (7th century B.C.), Plato, Hippocrates, and Aristotle. Cicero (1st century B.C.) wrote that he believed an active mental life could prevent or postpone mental failure with age [31]. Careful study of these mental diseases remained limited until the mid- 7 1800s when new histological dyes were developed by scientists such as Santiago Ramón y

Cajal, Camillo Golgi, and Max Bielschowsky. These new staining methods and dyes were then made available to researchers like Alois Alzheimer [32], Oscar Fischer [33], Emil

Redlich [33], and other neurologists or psychiatrists thus acquiring the ability to visualize intracellular and extracellular components such as DNA or proteins from autopsied brain tissue of their demented patients. Early in the 20th century, Alois Alzheimer was pursuing the idea that mental diseases require clinical-histological correlations and thus can be categorized as distinct diseases using defining brain pathologies at autopsy [8]. Alzheimer was an expert neuropathologist and microscopist and documented a peculiar pathology in

1906 from the brain of a 56-year-old woman named Auguste Deter, a housewife married to a railroad clerk in Prussia during the 1800s. Five years earlier, on November 26, 1901,

Auguste was admitted to the Asylum for the Insane and the Epileptic in Frankfurt am Main with clinical symptoms including sudden paranoia and ungrounded suspicion that her husband was having an affair. Her behaviors quickly escalated within a few months to include restlessness, severe anxiety, emotional outbursts manifesting to bang on neighbors’ doors, hiding and losing objects, and errors in housework. Eventually her husband could no longer care for her while maintaining his job and home. It was at the ‘Castle of the

Insane’ where Alois Alzheimer began his observations of her behaviors as her condition progressed. Not too far away at the German University in Prague, Oskar Fischer was practicing under Arnold Pick where he was also documenting his observations of demented patients’ brain tissue by constructing detailed drawings of pathology including plaques external to neurons and intraneuronal tangles of some substance [33] of which he believed neuronal regeneration. But also in Vienna at the Second Psychiatric Clinic of the University

8 of Vienna (directed by Wagner-Jauregg), Emil Redlich described what he named ‘miliare

Sklerose’ (miliary sclerosis) in two cases of senile dementia and was drawing and describing plaques as a location of neuronal destruction which is then replaced by degenerating neuroglial cell [34].

Alois Alzheimer took great interest in the case of Auguste D. as she was much younger than his other cases of which he diagnosed as having senile dementia due to arteriosclerotic brain atrophy. These cases he attributed to the thickening of the blood vessels or atherosclerosis in the brain (Alzheimer, A. (1898) Dementia Senilis and Brain

Diseases Based on Atheromatous Vessel Disease, Monatsschrift fur Psychiatrie.) [32].

At Auguste’s death, Dr. Alzheimer was then practicing at the Royal Ludwig-

Maximilian University in Munich and requested her brain for autopsy. He went about documenting the pathology bearing two principle hallmark lesions today known as the extracellular amyloid plaques and intracellular neurofibrillary tangles accompanied by cortical atrophy due to neuron loss [8, 32]. Alzheimer presented his work ‘A Peculiar

Disease of the Cerebral Cortex’ at the thirty-seventh assembly of the Southwest German

Psychiatrists in November 1906 which was met with seemingly disinterest gauged by only one question posed to him.

Sadly, more than 100 years after Emil Kraepelin published a chapter in his

‘psychiatrie’ textbook, ‘Das senile und präsenile Irresein’, where he coined the eponym

Alzheimer’s Disease after his colleague at The Royal Psychiatric Clinic in Munich [8, 35,

36], we still cannot offer definitive answers to probing questions for hope. The years of research after Alzheimer until now has revealed a great deal including the recognition that

AD begins years before the symptoms of dementia appear [5]. The discovery of the precise

9 biological changes in AD that lead to symptoms has not been made. There does not yet exist a definitive explanation of why these pathological lesions of disease appear [35, 37].

Remarkably, two of the most studied proteins in science are the major components of amyloid plaques and neurofibrillary tangles yet their actual function in brain remains unclear. Amyloid plaques consist of proteinaceous fibrillar deposits made to the exterior of cells surrounded by distended neurites [38] (Fig. 2A). These spherical plaques are found in the hippocampus, amygdala, neocortex, but also the walls of cerebral and meningeal blood vessels. In 1984, it was found that the principle protein in the plaque core is a variation on a peptide consisting of 39 to 43 amino acids [39, 40]. The 42 amino acid peptide is the most predominant form found in amyloid plaques [40]. This peptide, called

Aβ, is a proteolytic cleavage product from a larger single-pass transmembrane protein known as the amyloid precursor protein (APP) which, as mentioned earlier, is a Mendelian gene of AD [41-43]. Interestingly, amyloid β production may actually be a normal part of the brain’s innate immune system [44].

Microtubule associated tau protein was identified in 1985 as the major component of neurofibrillary tangles [45-47]. Tau is normally abundant in neuron axons but it becomes hyperphosphorylated in AD dissociating from microtubules and dislocates to cell soma and dendrites in the aggregated filamentous form [48] (Fig. 2A,B). Tau protein aggregates are found in most dementias (Table 2). I will go into much detail describing tau protein later.

10 A B

30 µm 5 µm

Fig. 2. Hallmark lesions of Alzheimer’s Disease. (A) Confocal image (Leica Microsystems

Heidelberg GmbH capturing Aβ plaques, neurophil threads, and neurofibrillary tangles in human brain tissue. Aβ plaques pseudo-colored red were immunolabeled with β-Amyloid antibody #2454 (Cell Signaling Technology) and cy3 flour. Neurofibrillary tangles and neurophil threads appear green immunolabeled using tau5 antibody with Alexa488 conjugate. (B) Updated camera captures the filamentous nature of the neurofibrillary tangles in AD brain tissue stained with ThS. Image taken using Andor Revolution WD

Spinning Disk Confocal with Neo sCMOS camera. (Panel A imaged by Kristin E. Funk.)

What we have learned from clinical trials failing

There is no cure for AD nor proven way to slow the rate of neurodegeneration. At best there are symptomatic treatments available, such as, an acetylcholinesterase inhibitor or a glutamatergic moderator which provides modest benefit for some which is thought to occur through temporary stabilization rather than any significant improvement in memory

11 function. The causes of most neurodegenerative diseases are essentially unknown, and even when identified, the mechanisms by which they initiate the disease remain speculative.

Even more dubious, early clinical identification is challenging as the pathologic processes usually begins decades before symptoms are evident [3]. New evidence in AD shows that

Aβ plaque accumulation in neocortex reaches plateau before clinical function enters a stage of mild-cognitive impairment (MCI) (Fig. 3) [49]. The idea is that the progression of tau protein pathology and brain atrophy in cortex begin in conjunction with cognitive decline.

A recent cross-sectional brain-imaging study using magnetic resonance imaging (MRI) and positron emission tomography (PET) found a correlation between regional brain atrophy, aggregate tau protein and related cognitive decline. Although Aβ positive imaging was also present, it did not correlate with brain atrophy nor regional specific cognitive decline [50].

This suggests that tau pathology may have direct implications for cognitive impairment possibly through a mechanism that includes brain matter loss.

12 Fig. 3. A model of AD and the progression of its dynamic biomarkers in the disease pathological cascade. The model assumes that amyloid-β (Aβ) biomarkers transition to abnormal states long before tau-mediated neuronal injury or changes in brain structure.

At some threshold of these silent changes, mild-cognitive impartment (MCI) and memory decline begin and may progress onto clinical symptoms and dementia. Aβ load can be identified by CSF Aβ42 or PET amyloid imaging. Tau-mediated neuronal injury and dysfunction can be identified by CSF tau or fluorodeoxyglucose-PET. Brain structure is measured by use of structural MRI. (Reprinted from The Lancet Neurology, (Vol 9), Jack, et. al., Hypothetical model of dynamic biomarkers of the Alzheimer’s pathological cascade,

13 The AD drug development pipeline has focused almost exclusively on the hallmark neuropathological markers with very little success. With the acceptance of the Amyloid cascade hypothesis in 1992 [51] – postulating that amyloid β dysfunction starts the disease process – many recent drug discovery strategies are focused on immunotherapy and removing Aβ peptides or plaques from brain. So far these strategies have failed to prove useful in humans (clinicaltrials.gov) [52]. A 2014 study, compiled data gathered from clinicaltrials.gov and reported for the decade 2002 to 2012, that there were 413 clinical trials completed or ongoing targeting AD of which 244 were potential compounds tested for drug approval of which less than 4% targeted tau protein. In that decade, the overall failure of clinical trials for AD was 99.6% with one success, memantine, an NMDA receptor inhibitor (glutamate regulator), which sometimes briefly improves cognitive symptoms [53].

There are currently six drugs approved by the U.S. Food and Drug Administration

(FDA) for the treatment of AD. Three are cholinesterase inhibitors; Aricept (donepezil),

Exelon (rivastigmine), and Razadyne (galantamine), which increase levels of acetylcholine in the brain. Acetylcholine is thought to be involved in thinking, language, judgement, and memory. Namenda (memantine) is an NMDA (N-Methyl-D-aspartate) receptor antagonist which can regulate the activity of glutamate. Glutamate is a chemical neurotransmitter thought to be involved in processing, storing and retrieving information an essential role in learning and memory. Research shows that excess glutamate can lead to damage and

Namenda is thought to help protect cells against excess glutamate by partly blocking receptors [3]. A fifth approved medicine is a combination of donepezil and memantine

14 called Namzaric and there exists a sixth medication, tacrine, that was discontinued in the

U.S. due to safety concerns.

A neurodegenerative disease modifying intervention would presumably achieve maximum benefits if early intervention were achieved thereby arresting neuronal loss [3].

Determining when this intervention should begin and what target(s) is appropriate is a very active research area in AD. Recent advancements in medicine include the development and subsequent approval of radioligands with selectivity to Aβ plaques or pathogenic tau protein antemortem using positron emission tomography (PET) imaging [54]. Trials of

PET-amyloid and PET-tau suggest that temporal tracking of tau pathology is a better diagnostic tool for AD. Using PET-ligands, MRI, and cerebral spinal fluid (CFS) biomarkers, clinicians are gaining ground in staging and monitoring disease progression.

This new information will clarify the contribution of pathologies in AD and contribute greatly to diagnosis and drug development.

It was established decades ago that the pathological progression of tau protein neurofibrillary tangles correlates well with neuronal loss and cognitive decline [55-57]. In fact, Braak staging has long been used by pathologists as guideposts in determining AD stage after a person has died. There is much research evidence supporting neuron-to-neuron transmission and/or trans-synaptic transport of small aggregates of tau protein providing a potential propagation of misfolded tau protein onto the receptor synapse [58-62]. The AD dementia can be divided into stages which correlate well with the advancement of tau pathology beginning in non-thalamic sub-cortical nuclei. (Fig. 4) [55, 57, 58, 63]. A pathological stage 0 is preclinical with no cognitive impairment or memory disorder but amyloid plaques may be present in neocortex along with neurofibrillary tangles in brain

15 stem locus coeruleus. The advancement of the tau pathology to the entorhinal cortex marks stages I-II in which the person may suffer some mild memory decline. It is during stage III

(Fig. 4, 5) that memory problems and symptoms become apparent to family members. The tau lesions are now moving into the limbic and cortical association region of the brain affecting language, planning, and visuospatial skills. In the final stages V-VI, the lesions have invaded the primary functional areas and secondary association areas of the brain severely affecting executive function and require careful observation to remain safe.

Although we know now that the brain maintains some neuroplasticity into adulthood [64], after Braak stage III, neuronal damage may be too advanced for therapeutic interventions.

Halting or perhaps reversing the progression of tau protein neuropathology in the earliest stages of the disease are of great interest to researchers such as myself. In fact, several immunotherapy studies against tau aggregates in mouse models of AD have generated interesting results including pathology clearance and cognitive improvements [65-68].

Fig. 4. Braak staging. The neuropathological spread of tau protein neurofibrillary tangles

(NFT) is used to stage AD at autopsy. The pre-tangle stages begin in the subcortical region of locus coeruleus in the brain stem near the fourth ventricle (stages 1a,b). Locus coeruleus projection neurons are first immunoreactive for hyperphosphorylated tau material (AT8) in the proximal axon (stage 1a). The locus coeruleus neurons (noradrenergic cells) make long axonal projections that terminate in the trans-entorhinal region where next altered neuronal processes appear (brown projections). The pathological process systematically progresses from mesocortical areas of the medial temporal lobe (NFT stages I-II) to neocortical high-order association areas (NFT stages III-IV) and ultimately onto secondary and primary fields of the neocortex (NFT stages V-VI). Cortico-cortical

17 projections that make the ‘return’ pathway (red and orange projections) diffusely synapse with all cortical layers except lamina IV and, thus, bear a resemblance to terminal axons of the diffusely projecting non-thalamic nuclei. Propagation of the disease process would occur via both cortico-cortical projections of the return pathway (red and orange projections) and diffuse projections from subcortical nuclei (brown projections). Apical axons from locus coeruleus also project onto cerebellum (green) which remains for the most part, unaffected by neuropathological processes. Clinical presentation usually occurs in Braak Stage III as the tau pathology moves into the higher order association areas.

(Adapted by permission from: Springer Nature, Acta Neuropathologica, ‘Alzheimer’s pathogenesis: is there neuron-to-neuron propagation?’, Braak and Del Tredici, 2011.)

Figure 5 Brain banking and cross-sectional study of disease progression in human brain tissue. Systematically collected post-mortem human brains from forensic, clinical or prospective studies that belong to the same disease category can provide longitudinal insight into the pathological process. Braak spearheaded this approach by investigating hundreds of Alzheimer’s disease brains and categorizing the progression of neurofibrillary tangles (NFTs) as shown (immunohistochemistry with an antibody against hyperphosphorylated tau). These studies showed the beginning of NFT accumulation in the transentorhinal cortex, its progression into limbic areas in stages III and IV, and its culmination in neocortical and primary sensory areas in stages V and VI. (Figure and legend reproduced, with permission from Journal of Alzheimer’s Disease, Vol 9, Braak et.al., (2006) with permission from IOS Press.)

19 Tau protein

The intraneuronal neurofibrillary tangle in AD affected brain contains a paired- helical filament (PHF) structure that was isolated in 1963 [45]. The molecular structure of

PHF was discovered by immunohistochemical probing in 1985 [47] and by 1992 [69, 70], it was clear that PHF in AD is composed of six tau protein isoforms assembled by alternative splicing of the mRNA product from the MAPT gene on chromosome 17 (Fig.

6). It is now known that the N-terminal region of the tau protein contains two domains translated from exons 2 and 3 in MAPT. The exon 2 or both exons 1 and 2 may be ignored with the result being three different isoform combinations termed 0N, 1N (exon 1 was translated and included) or 2N (includes translation of exons 1 and 2). Toward the C-term of tau protein there is a microtubule binding domain with four very similar amino acid repeats. When all four repeats are present, the isoform is termed 4R and when the repeat translated from exon 10 in MAPT gene is excluded, the result is a 3R isoform.

The tau molecule is divided into four main regions including the N-terminal domain, a proline rich region, the microtubule binding domain with the three or four repeats, and the C-terminal end (Fig. 6). The molecular function of tau protein in brain is still being investigated but it is mainly located in neuronal axons and binds to negative- charged microtubules via its three or four repeat domain. Microtubules provide the transport highways inside neurons constantly moving nuclear products, nutrients, and wastes to and from synapse on other neurons [71]. The N-terminus of tau protein is thought to help maintain spacing between microtubules. The proline rich region is probably involved in cell signaling and interacts with kinases, and lastly, evidence suggests that the

C-terminus may regulate microtubule polymerization [72].

Fig. 6. Microtubules associated protein tau is translated from the MAPT gene located on

Chromosome 17q21.31. Twelve exons are alternatively spliced in transcription resulting in mRNA for six isoforms in human brain. Three isoforms contain three microtubule binding repeats (3R) and three isoforms include exon 10 completing four microtubule binding repeats (4R). Each isoform containing a 3R or 4R repeats may include 0,1,or2 N- terminal expressed splice variants. No scale implied.

Tau is abundantly expressed in nervous system and stabilizes the neuronal microtubules by its three or four repeat domain binding to the tubular surfaces and facilitates neurite outgrowth [73]. The accumulations of fibrillar aggregate tau protein which Alzheimer observed in his dementia patients’ brains are hyperphosphorylated and no longer bind to the negatively charged microtubule surfaces. Dissociation of tau protein

21 from microtubules is thought to lead to microtubule dysfunction [74]. In addition to losing function in the brain and disrupting microtubules, tau aggregates are believed to exhibit properties similar to prions [58]. If a small tau aggregate can be transferred to a neighboring cell via some method such as exocytosis transfer, nanotubes or endosomes, then it may act as a prion by the misfolded protein providing a template which in turn can recruit more tau monomers to misfold and begin spreading the aggregation process. There is increasing evidence that the progression of tau neuropathology in AD resembles neuron-to-neuron propagation (Fig. 4, 5) [58-61, 75]. Understanding the mechanism of aggregation and of the transfer of a pathological species of tau protein is key to strategies for the therapeutic approach to be taken in AD. Intervening with the prion-like behavior of a progressive aggregating protein such as tau will most likely include immunotherapies to remove the pathogenic species or small molecule tau aggregation inhibitors.

The abnormal deposition of tau in the brain is called a tauopathy. Some common tauopathies include pathological deposits of Picks’ disease (PiD), progressive supranuclear palsy (PSP), corticobasal degeneration (CBD), and argyrophilic grain disease (AGD) [76].

In contrast to AD and PiD, the abnormal deposits of PSP, CBD and AGD are found in both nerve cells and glial cells [76]. It is now clear that tau is the most common misfolded protein in human neurodegenerative disease (Table 2) [77]. Tau differs from the extracellular Aβ plaques in that it is found in a variety of dementias (Table 2)[78].

22 Table 2. Diseases associated with tau protein lesions primary tauopathies are in bold type [79, 80] Aging-related tau astrogliopathy Myotonic dystrophy Alzheimer’s disease Niemann-Pick disease, type C Amyotrophic lateral Non-Guamanian motor neuron disease sclerosis/parkinsonism-dementia complex with neurofibrillary tangles Argyrophilic grain disease Pantothenate kinase-associated neurodegeneration Atypical parkinsonism of Guadeloupe Parkinsonism-dementia complex of Guam Chronic traumatic encephalopathy Pick disease Corticobasal degeneration Postencephalitic parkinsonism Diffuse neurofibrillary tangles with Primary age-related tauopathy calcification Down’s syndrome Prion protein cerebral amyloid angiopathy Familial British dementia Progressive subcortical gliosis Familial Danish dementia Progressive supranuclear palsy Frontotemporal dementia and parkinsonism linked to chromosome 17 SLC9A6-related mental retardation caused by MAPT mutations Gerstmann-Straussler-Scheinker disease Subacute sclerosing panencephalitis Globular glial tauopathies White matter tauopathy with globular glial includsions

As explained previously, tau protein is a major constituent of AD as well as many other dementias termed tauopathies. Although Aβ is common in human brain without dementia as a normal part of life, the current drug discovery pipeline is focused on removing the peptide and/or plaques from affected brain. With new PET imaging ligands

23 available for both Aβ plaques and tau protein antemortem, the ability to track the hallmark lesions of AD is improving. The evidence for the tau neuropathology progressing in parallel with cognitive disease stage neuron-to-neuron like a prion is emerging. Because tau protein more closely follows neuron loss and clinical symptoms, more information is needed on the mechanism of tau protein aggregation to clarify potential interventions in the pathway for drug targets.

It follows then that one of the goals of my graduate study was an in-depth study to characterize the aggregation mechanisms of tau protein. First, I describe a new technique to quantify aggregation of an intrinsically disordered molecule such as tau that excludes it from structural studies such as NMR. Characterization of aggregation can be done on recombinant tau protein preparations using high-throughput solution-based detection methods such as thioflavin-dye fluorescence and laser-light scattering spectroscopies but these methods do not offer information on number or size of filaments forming.

Transmission electron microscopy (TEM) is a static imaging tool that complements these approaches by detecting individual tau filaments at nanometer resolution. In doing so, it can provide unique insight into the quality, quantity, and composition of synthetic tau filament populations. I describe here protocols for analysis of tau filament populations by

TEM for purposes of dissecting aggregation mechanism. Using the TEM methods, I compile two data sets characterizing tau aggregation in vitro and then formulate a mathematical model to test the hypothesis of several aggregation mechanisms that may be occurring in vitro. Using this model, I identify a novel association process never described for tau protein and quantify reasonable kinetic parameters for the processes occurring in tau aggregation.

24 The tau protein interacts in vivo with several molecules for structural, signaling, and modifications of function. One such modification of great interest in this lab is the modification of tau protein by methylation on Lysine and Arginine residues. Previous work before me showed that there are methylated sites in vivo for tau protein in brain and that significant methylation of tau protein in vitro can alter the tau aggregation propensity as well as microtubule assembly dynamics [81, 82]. With this new information in hand, we wanted to quantify the amount of tau methylation modifications in AD affected human brain vs non-affected. This task proved difficult due to dimerization of conjugate marker molecules. I conceptualized a novel method of protein methylation quantification using liquid chromatography–tandem mass spectrometry (LC-MS/MS) and show its utility to quantify the differences of tau methylation in non-affected vs AD affected human brain tissue.

Finally, I describe a new approach for differential expression of human brain transcripts and provide examples of the expression changes in AD including that of protein methyltransferases and de-methylators in the AD brain.

Chapter 2. Analyzing tau aggregation with electron microscopy

Reproduced in part with permission from Huseby, C.J. and Kuret, J. (2016)

Analyzing tau aggregation with electron microscopy. Meth. Mol. Bio. 1345, 101-12. The full article is available at https://doi.org/10.1007/978-1-4939-2978-8_7

C.J. Huseby performed all experiments and wrote the initial draft of the manuscript.

J.K. conceived the research and modified the manuscript. This work was supported by the

National Institutes of Health grant AG14452 to J.K.

26 Introduction

One goal of this dissertation is to characterize tau aggregation mechanism in depth.

To approach this problem, powerful tools for characterizing tau aggregate quality, quantity, and composition are required. Here I explain the utility of transmission electron microscopy (TEM)-based methods for assaying the rate, extent, and quality of tau aggregates prepared in vitro. TEM has several advantages for these purposes. First, its ability to capture morphology at nanometer resolution allows one to distinguish mature filaments from amorphous aggregation products, and to determine whether they most closely resemble the paired-helical, straight, or hemi-filament forms isolated from disease tissue [83-85]. Second, TEM captures length information that can be leveraged to assay filament formation and disaggregation. Although low throughput and only semiquantitative in nature, length measurements become a powerful tool for assessing aggregation mechanism when collected as a function of time or protein concentration and subjected to regression analysis. Fundamental aggregation parameters can be estimated by this approach, including the minimal concentration needed to support aggregation, lag times of nucleation dependent reactions, and dissociation rates of mature filaments [86].

Moreover, filament length distributions can be leveraged to detect the presence of secondary aggregation pathways and also to provide an independent check on rate constants deduced from time-dependent evolution of filament mass [87]. Finally, when combined with immunogold labeling methods, TEM provides information on aggregate composition. This approach can be used to confirm that filamentous aggregates contain tau protein [88, 89], to determine whether specific tau epitopes are exposed or sequestered in the aggregated state [89], and to clarify whether tau aggregates associate with heterologous

27 proteins [90].

Below I summarize two protocols for TEM that I use throughout this dissertation to analyze tau aggregation. The basic protocol details adsorption of tau fibrils onto TEM grids and measurement of filament length. It then summarizes methods for analyzing length data to obtain aggregation parameters such as minimal concentration, aggregation rates, and filament dissociation rates. It also describes measurement of length distributions. The second protocol describes immunogold labeling of tau filaments for assessment of composition. Previous reviews of electron microscopy methods applied to amyloid aggregates, including assessment of structure by cryo-electron and scanning transmission electron microscopies, can be consulted for additional approaches [91, 92].

Materials

Tau filaments. All buffers and reagents are made with ultrapure water (18.2 MΩ-cm at

25ºC) and filtered prior to use (pore size ≤0.22 μm).

1. Recombinant tau proteins: These are expressed in E. coli and purified by liquid

chromatography as described previously [93, 94] (see page 39 Note 1).

2. Aggregation inducer: Thiazine Red (TR; also known as Geranine G, Chemical Abstract

Service registry number 2150-33-6; TCI America, Portland, Oregon, USA) (see page

39 Note 2).

3. Assembly buffer: 10 mM HEPES, pH 7.4, 100 mM NaCl, 1 mM DTT.

Transmission electron microscopy

1. Transmission electron microscope. I use a Tecnai G2 Spirit BioTWIN transmission

electron microscope (FEI Company, USA) operated at 80 kV acceleration voltage and

equipped with digital image capture.

28 2. Uranyl acetate (UA) (Electron Microscopy Sciences, Fort Washington, PA, USA).

Prepare 2% (w/v) solution in water (see page 39 Note 3).

3. Copper grids 300-mesh formvar/carbon-coated (Electron Microscopy Sciences Cat.

No. FCF300-CU). These commercial grids are supplied with film laid on the shiny side.

They can be used directly without glow discharging.

4. Hydrophobic laboratory film (e.g., Parafilm, Pechiney Plastic Packaging, Chicago, IL).

Cut into 4 x 4 inch squares for easiest handling.

5. Glutaraldehyde (25% w/v in water; Electron Microscopy Sciences).

6. Fine-tipped forceps for handling grids (110 mm, Structure Probe Inc., West Chester,

PA).

7. Cellulose filter paper (e.g., Whatman Qualitative No. 2 filter paper) for blotting off

excess liquids from grids. Cut into small squares for easiest manipulation.

8. Grid box (Electron Microscopy Sciences) for storage and transport of grids (see page

40 Note 4).

Image Analysis.

1. ImageJ or equivalent image analysis software. ImageJ can be downloaded for free from

the website http//imagej.nih.gov/ij/.

2. Microsoft Excel or equivalent spread sheet software for manipulating filament length

data.

3. SigmaPlot (Systat Software, Inc., San Jose, CA, USA) or equivalent graphics software

for curve fitting and regression analysis.

Immunogold Labeling. This method requires a primary antibody capable of binding recombinant tau proteins with high affinity. Here I illustrate the method using a

29 commercially available rabbit polyclonal antibody raised against the V5 epitope

(GKPIPNPLLGLDST). Tau proteins tagged at the N-terminus with the V5 epitope bind strongly to anti-V5 antibodies in both monomeric and polymeric states.

1. Primary V5 antibody, Rabbit Polyclonal (Bethyl Laboratories, Inc. Cat. No. A190-

120A, Montgomery, TX).

2. Secondary goat anti-rabbit IgG (H + L), 12 nm gold-conjugated, EM grade (Jackson

Immunoresearch Laboratories, Inc. Cat. No. 111-205-144, West Grove, PA).

3. Phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, and

2 mM KH2PO4).

4. Blocking buffer: 1% bovine serum albumin (w/v) in PBS.

5. 96-well, flat bottom, low protein binding assay plate (e.g., Corning Inc, polystyrene

plate #3641).

Methods

Tau aggregation

1. Initiate aggregation of tau protein at 37ºC without agitation in Assembly Buffer

containing 100 μM TR.

2. Stop reactions by gently adding glutaraldehyde to 1% (w/v) final concentration (see

page 40 Note 5).

Basic protocol. All steps can be carried out at room temperature (RT).

1. Dispense 50 μL each of the aggregate sample, two drops water, and two drops UA onto

a sheet of parafilm (Fig. 7).

2. With tweezers, carefully pick the grid up by the edge, being careful not to damage the

grid, and place the grid, shiny-side down, onto the sample drop for 1 min (see page 40

30 Note 6). Do not completely submerge: surface tension will support the grid while fibrils

diffuse and absorb to the grid surface.

3. Again using the tweezers, carefully remove the grid from the sample drop and blot off

excess sample by gently touching filter paper with the grid edge.

4. Wash the grid briefly by dipping it, shiny-side down, on top of the first water drop and

again carefully blot off excess liquid with filter paper.

5. Rinse briefly in the first UA stain drop, blot off excess stain with filter paper, and place

it on the second drop of UA, shiny-side down for 1 min (see page 40 Note 7).

6. Remove grid from the UA, blot with filter paper, wash again on the second water drop,

blot with filter paper, and finally leave grid shiny-side up on filter paper to completely

dry. Stained grids can be stored at room temperature for weeks.

7. Acquire images on transmission electron microscope, and save images for analysis. Be

sure to record the magnification scale of all images (see page 40 Note 8).

Fig. 7. Technique for applying tau samples to grids. Parafilm (grey surface) is tacked down to the bench-top by rubbing a blunt surface at each corner, creating a hydrophobic surface onto which 50 μl aliquots of aggregated sample, water, and UA are placed. Using microforceps, the grid is sequentially contacted to each solution before drying and storage.

Filament length measurement

1. Load TEM images into ImageJ (NIH). Tutorials are available at the ImageJ website.

31 2. Calibrate length scale: In ImageJ, draw a straight line over the TEM calibration scale

bar. Under the analyze tab, use ‘set scale’ to define “units” and “known distance” so

that they match the scale bar marked on each image (Fig. 8A).

3. Measure only the lengths of filaments where both ends are fully resolved in the field

(Fig. 8A). Transfer the length data into Microsoft Excel or similar software for analysis

of replicates and statistics.

Filament length distribution

1. Measure the lengths of a filament population. Typically, three fields or technical

replicates per assay are sufficient for analysis of each assay condition.

2. Choose a bin size for filament lengths, and count the number of filaments per bin.

Typically, bin size is chosen so that each contains multiple observations. Often this

corresponds to 10 – 20 bins per distribution (see page 40 Note 9).

3. Plot the average frequency (i.e., the number of filaments per bin divided by total

number filaments in all bins) as a function of bin number (Fig. 8B).

Fig. 8. Electron micrographs of tau filaments. Recombinant human 2N4R tau was

prepared, aggregated (1 μM) in the presence of TR inducer, and subjected to the basic

TEM protocol. (A) This image, captured at 18,500-fold magnification, shows that

filaments are well resolved under these conditions. The lengths of filaments having both

ends visible in the field are quantified using ImageJ. (B) Length distribution calculated

with bin size set to 50 nm, where each bar represents the average of Panel A and two

technical replicates ± S.D. Filaments shorter than 25 nm were not resolved at this

magnification and were not included in the distribution.

Time series

1. Measure total filament length/field as a function of time and constant tau monomer

concentration (see page 41 Note 10).

2. Plot time series in SigmaPlot or other graphics program capable of regression analysis.

33 3. Fit to Gompertz [95], logistic [96], or other growth models appropriate for analyzing

sigmoidal time series. These models yield estimates of ti, the inflection point

corresponding to maximal growth rate, kapp, an estimate of the underlying first order

growth rate constant, and an estimate of the maximum total length/field at reaction

plateau (Fig. 9A).

4. Lag time is when the tangent to the point of maximum growth rate (i.e., ti) intersects

the abscissa of the sigmoidal curve [96]. In Gompertz regression, this time corresponds

to ti – 1/kapp [95].

Tau concentration dependence

1. Measure total filament length/field as a function of tau monomer concentration. For

this experiment, incubation time is held constant and is greater or equal to the time of

reaction plateau.

2. Plot total filament length/field versus tau monomer concentration in SigmaPlot or other

graphics program capable of regression analysis (Fig. 9B).

3. Perform linear regression. Minimal concentration corresponds to abscissa intercept (see

page 41 Note 11).

34 Fig. 9. Graphical depiction of parameter estimation. (A) Hypothetical time series subjected to Gompertz regression, where L(t) corresponds to total filament length/field measured as a function of time t after addition of aggregation inducer. Calculated parameters useful for assessing mechanism include ti, the time of maximal growth rate, kapp, an estimate of the underlying first order growth rate constant, and L∞, an estimate of the maximum total length/field at reaction plateau. Extrapolation from ti to the abscissa intercept (dashed line) yields an estimate of lag time, which for a nucleation dependent aggregation is inversely proportional to nucleation rate. (B) Tau concentration dependence of total filament length/field measured at aggregation plateau (L∞). Linear regression and extrapolation to the abscissa intercept (dashed line) allow quantification of the minimal tau concentration needed to support filament formation. (C) Hypothetical dissociation time series modeled as simple exponential decay, where L(t) corresponds to total filament length/field measured as a function of time t after beginning disaggregation by dilution. Exponential regression analysis yields kapp, the apparent first order decay rate constant for filament length as a function of incubation time. Conversion of kapp to the filament dissociation rate constant, a measure of filament stability, requires knowledge of mass per unit length, the filament length/field at the start of disaggregation (L0), and the number of filament ends at the start of disaggregation. 35 Filament dissociation rate

1. Aggregate tau protein as described above until at least plateau is attained.

2. Dilute the aggregates 10-fold with assembly buffer containing TR and continue

incubation at 37ºC.

3. Withdraw aliquots as a function of time (up to 5 h post dilution), treat with

glutaraldehyde, and subject to TEM as described previously.

4. Measure total length/field (L) and filament number for each sample.

-kapp t 5. Fit to exponential decay function L = L0 e to obtain kapp, the pseudo-first order rate

constant describing the time-dependent decrease in filament length, and L0, the total

filament length at time zero (Fig. 9C). Dissociation rate constant ke- is estimated from

kapp, L0 and the number of filaments at time zero as described previously [86, 97, 98].

(see page 41 Note 12).

Immunogold labeling protocol

1. Dispense 50 μL each of sample and blocking buffer onto a sheet of parafilm.

2. Adsorb the sample to the grid as in steps 2 and 3 of the basic protocol.

3. With tweezers, carefully pick the grid up by the edge, blot off excess sample with filter

paper, and place it shiny-side down onto the drop of blocking buffer. Incubate for 5

min (see page 42 Note 13).

4. Place 50 - 200 μL of primary antibody diluted in blocking buffer in the wells of a 96-

well plate (see page 42 Note 14).

5. Carefully remove grid from the sample, blot off excess blocking buffer with filter

paper, and place the grid sample-side down in primary antibody dispensed in the wells

of the 96-well plate. Incubate 4 h at 4ºC with agitation.

36 6. Carefully remove the grid from the primary antibody, and wash twice by floating on

50 μL drops of blocking buffer interspersed with blotting off of excess liquid.

7. Dispense secondary antibody diluted in blocking buffer into 96-well plate.

8. Carefully place the grid in a well containing gold-labeled secondary antibody. Incubate

with agitation for 2 h at 4ºC.

9. After removing the grid from the well of secondary antibody, wash four times by

floating on successive drops for 5 min each as follows: 2 x PBS, 2 x H2O. Excess liquid

is blotted off with filter paper between transfers.

10. Samples are stained with UA and imaged as described in steps 5–6 of the basic protocol.

11. Subject to TEM, and capture images. An example of immunogold labeling of tau

filaments is shown in Fig. 10.

Fig. 10. Immunogold labeling of epitope-tagged tau filaments. Recombinant human 2N4R tau and V5-tagged 2N4R tau were prepared and aggregated in the presence of TR inducer.

(A) V5-2N4R filaments were analyzed by basic protocol. The filaments retained the length and morphology characteristics of non-tagged 2N4R tau (arrowhead points to one example). (B) V5-2N4R filaments stained with polyclonal anti-V5 primary and 12 nm- labeled secondary antibodies using the immunogold labeling protocol. Extensive decoration of filaments with these antibodies obscures filament morphology, but preserves length and provides clear evidence for the presence of V5-labeled tau (arrow points to one example). (C) 2N4R tau filaments subjected to the same protocol as in Panel B. These filaments lack the V5 epitope, and do not label with the anti-V5/secondary antibody pair.

Morphological information is mostly retained. (D) 2N4R tau filaments were prepared, mixed with V5-2N4R tau, then incubated for an additional 16 h. Products were then subjected to the immunogold protocol as in Panels B and C. This image was captured in negative stain. It shows a filament containing a well resolved, unlabeled central segment 38 corresponding to 2N4R tau (arrowhead) flanked by extensively gold-labeled ends (arrows) composed at least in part of V5-2N4R tau. These data indicate that filament extension in the presence of TR inducer proceeds from both filament ends, and illustrates the utility of gold labeling for detecting tau composition in a filament.

Notes

1. This purification method has been applied to different tau isoforms [86], missense

mutants [99], and postranslational modifications [81, 98]. Additional purification steps,

such as boiling, can be included without changing aggregation or TEM assay

performance [100]. Purified tau is aliquoted and stored at -80ºC.

2. Spontaneous fibrillization of tau protein at physiological concentrations in vitro does

not occur over experimentally tractable time periods [101]. However, tau aggregation

can be accelerated by addition of anionic inducers, such as heparin [89], anionic

surfactants including arachidonic acid or alkyl sulfate detergents [102, 103], and small

molecule dyes, such as Thiazine red (TR) [104]. This protocol leverages TR as tau

aggregation inducer.

3. UA is light sensitive and toxic. The solid material should be stored under the hood in a

dark metal container protected from moisture. Care should be taken to avoid contact

with and exposure to the material and subsequent solutions containing UA. Used UA

and lab disposables that come in contact with UA are disposed of per institutional

guidelines and appropriate Materials Safety Data Sheet.

39 4. Secure the grid box lid with a small piece of labeling tape while transporting grids to

prevent the inadvertent opening of the box and losing and/or mixing of the grids.

5. Glutaraldehyde is toxic and care should be taken not to breathe the fumes or allow

contact with skin. Under the hood with gloves, add a drop of glutaraldehyde to the

inside wall of the tube just above the sample. Carefully allow the drop to mix with the

sample by gentle flicking of the tube. Aggressive mixing or vortexing can cause

clumping of tau filaments.

6. Filament adsorption is time dependent [105]. Therefore, it is important to accurately

maintain constant adsorption time for all samples.

7. UA most frequently interacts with tau filaments to generate a positive staining effect

(Fig. 8), where filaments appear dark against a light background. This staining is

adequate for quantification of filament length. Less frequently, UA fosters negative

staining in certain areas of the grid (Fig. 10). Negative stain is especially valuable for

interrogating filament morphology.

8. Three or more random images from each grid are captured typically at 8,000 to

100,000-fold magnification. High magnification better captures morphological

features, whereas lower magnification is necessary for quantification of filament

length. A typical magnification for quantification of tau filament lengths in the presence

of TR ranges 18,000 to 35,000-fold.

9. Measurement of lengths <20 nm is difficult at lower magnifications (e.g., Fig. 2A). As

a result, frequency measurements can be biased toward higher relative occupancy of

longer lengths. In addition, all length measurements will underestimate total filament

length/field. However, the error is predicted to be modest when tau aggregation

40 proceeds under nucleation-dependent conditions owing to low occupancy of short

length bins [87]. Indeed, aggregation characteristics are similar when quantified by

TEM methods or laser light scattering, a solution-based technique [95].

10. Because inducer-mediated tau aggregation follows a nucleation-elongation mechanism

[106], the dependence of total filament length/field on incubation time at constant tau

monomer concentration is especially informative. In particular, measurement of lag

time is valuable because this parameter is proportional to nucleation rate [107].

Nonetheless, care must be taken when interpreting lag-times because they can reflect

factors other than nucleation under certain conditions [108, 109].

11. The dependence of total filament length/field on tau concentration is linear and

frequently intercepts the abscissa at positive values of tau concentration. Therefore,

inverse prediction methods are used to solve for the intercept [86]. This parameter

represents the minimal concentration of tau monomer needed to support aggregation.

Because tau filament adsorption to grids is sensitive to tau concentration [105], it is

important to employ tau concentrations that vary no more than 2-3 fold above minimal

concentration for this measurement. Extrapolating over large tau concentrations can

yield non-linear plots.

12. Calculation of dissociation rates requires knowledge of filament mass per unit length,

so that changes in length can be related to changes in tau protomer number. Mass per

unit length measurements have been reported for synthetic human TR- [87],

arachidonic acid- [110], and heparin-induced [111] tau filaments (though for only a

limited number of isoforms and filament morphologies).

41 13. When blocking and washing samples on the benchtop, it is prudent to prevent dust

particles in the air from contacting the drops. Covering the drops with a lid from a 96-

well plate works well for this purpose.

14. Volumes of at least 100 μL of antibody solution per well are preferred, because smaller

volumes complicate the placement and removal of grids from 96-well plates.

Chapter 3. The role of annealing and fragmentation in human tau aggregation

dynamics

Reproduced in part from a manuscript by Huseby, C.J., Bundschuh, R. and Kuret,

J. that has been submitted for publication.

C.J. Huseby formulated and coded the mathematical model and its sensitivity analysis. C.J. Huseby performed all experiments and wrote the initial draft of the manuscript. R.B. supervised mathematical methods. J.K. conceived the research and modified the manuscript. This work was supported by the National Institutes of Health grant AG05418 to J.K.

43 Introduction

Tau aggregation is a defining event in the pathogenesis of tauopathic neurodegenerative disorders such as Alzheimer’s disease (AD). It is an established marker for differential disease diagnosis and staging [112], a surrogate marker for neurodegeneration [113, 114], a potential vector for disease propagation [115, 116], and a source of toxicity in biological models [57]. For these reasons, the mechanisms through which tau misfolds and aggregates are of central importance for understanding disease pathogenesis. In vitro, the primary pathway to filamentous cross-β-sheet aggregates leverages a nucleation-elongation mechanism [106], where the rate-limiting nucleation event corresponds to dimerization

[87, 117], and the efficient elongation phase corresponds to addition of monomers to the ends of growing polymers [87, 117] (Fig. 11). A pathway involving isodesmic assembly of small aggregates constitutes an alternative route to these stable structures [117-120]. As observed with other aggregating proteins [121-123], secondary events such as filament breakage and secondary nucleation can increase apparent overall tau aggregation rate by increasing the number of filament ends available for elongation [124]. Secondary nucleation has been predicted to be especially important for creating small, highly diffusible aggregates associated with toxicity [125], whereas filament breakage can seed aggregation and foster spread of misfolding through the nervous system [126]. Each of these alternative or secondary processes increase filament number and therefore decrease average filament length [123]. Yet in the case of tau protein aggregates, intracellular filaments within the neurofibrillary lesions of AD can exceed micrometers in length [127].

Moreover, synthetic tau aggregates prepared in vitro also achieve stable length distributions extending to long lengths, even under aggregation conditions that are claimed

44 to be isodesmic [117, 128]. The length distributions observed in situ and in vitro suggest the existence of a distinct, previously uncharacterized secondary process that opposes filament breakage by promoting increases in average filament length. A candidate for this interaction is end-to-end annealing, which has been observed in linear assemblies of cytoskeletal protein, including tubulin [129], actin [130, 131], intermediate filament proteins [132], and septins [133]. In the case of actin, end-to-end annealing is highly favorable and strongly dependent on length (i.e., annealing efficiency decreases as filaments lengthen [134]). In fact, it is not possible to rationalize actin filament length distribution without incorporating both annealing and breakage terms into its nucleation- dependent assembly mechanism [131]. In the case of vimentin, an intermediate filament protein, modeling studies have shown that end-to-end annealing is obligatory for rationalizing the appearance of long filaments [132]. Because β-sheet edges are especially interaction prone [135], the ends of filamentous cross-β-sheet tau aggregates may be subject to annealing interactions as well.

Here we test this hypothesis with special emphasis on secondary processes that influence tau aggregate size distribution. The results indicate that tau filament ends can anneal, and that their propensity to engage in such homotypic interactions is length dependent. We propose that interactions at tau filament ends are candidate mediators of size-dependent phenomena reported in biological models.

Fig. 11. Aggregation models. In vitro aggregation of 2N4R tau was modeled as beginning with aggregation competent monomer generated by the presence of an inducer. Primary processes include the formation of a dimer, which corresponds to filament nucleation (N), and its elongation (E) to form filaments through endwise addition of monomers. Secondary processes include secondary nucleation (S), filament fragmentation (F), and filament annealing (A). See text for details.

Materials and Methods

Tau Preparations. Recombinant human tau proteins were expressed and purified as described previously [136] with modification. After expression protocol in E. coli cells, the pellet is resuspended in lysis buffer with PMSF and boiled in the presence of beta mercaptoethanol for 20 minutes with agitation every five minutes. The boiled lysate is then centrifuged at 100,000g for 90 minutes to remove cell debris, filtered, and subjected to immobilized metal affinity chromatography (IMAC) followed by size-exclusion 46 chromatography as previously described. Preparations included isoform 2N4R [100] and

2N4R fused to N-terminal tags containing polyhistidine (6His-2N4R; [136]), FLAG tag

(FLAG-2N4R; i.e., MDYKDDDDK-2N4R), and V5 tag (V5-2N4R; i.e.,

MGKPIPNPLLGLDST-2N4R). Expression plasmids for FLAG-2N4R and V5-2N4R were built by inserting synthetic DNA sequence into the NcoI-NdeI sites of vector pT7c-htau40

[136]. Fluorophore-labeled monomeric tau was prepared from 2N4R tau and Alexa Fluor

488-, Cy3-, or Cy5-maleimide as described previously [137]. Interestingly, we find that lack of an N-terminal polyhistidine-tag does not necessarily exclude the use of IMAC for the tau protein purification. We hypothesize that because the intrinsically disordered tau protein contains several indigenous histidines in its primary structure they are immobilized by the Nickel column beads rendering the polyhistidine tag unnecessary in this case.

Immunoblot analysis. Tau proteins were fractionated by SDS-PAGE (8% acrylamide) and subjected to immunoblotting on PVDF membranes as described previously [71] using rabbit polyclonal anti-V5 antibody (Bethyl Laboratories, A190-120A) or mouse monoclonal anti-FLAG (Agilent, 200471) as primary antibodies. Immunoreactivity was detected by the Enhanced Chemiluminescence Western Blotting Analysis System (GE

Healthcare) and captured on film.

Tau Aggregation. Recombinant human tau preparations were incubated (37C without agitation unless stated otherwise) in assembly buffer (10 mM HEPES, pH 7.4, 100 mM

NaCl, 5 mM dithiothreitol) for up to 24 h in the presence of fibrillation inducers Geranine

G (100 M, TCI America; [138]) or octadecyl sulfate (60 M, Lancaster Synthesis; [139]).

Aggregation products were then either analyzed immediately, sheared, or mixed and further incubated in the presence of additional 5 mM dithiothreitol. Shearing was

47 completed by passing aggregation products through a 28G syringe needle four times. Total tau protomer concentration in filaments and relative filament length distributions were estimated as described previously [138].

Transmission Electron Microscopy (TEM). For length measurements, reaction aliquots

(50 μL final volume) were treated with 2% glutaraldehyde (final concentration), mounted on formvar/carbon-coated grids (Electron Microscopy Sciences) and negatively stained with 2% uranyl acetate as described previously [138]. Random fields were viewed with a

Tecnai G2 Spirit BioTWIN transmission electron microscope (FEI) operated at 80 kV and

23000-49000x magnification. Filament lengths were estimated as described previously

[138].

For immunogold labeling experiments [138], reaction products were adsorbed directly onto grids and then incubated with rabbit polyclonal anti-V5 antibody (Bethyl

Laboratories) and/or mouse monoclonal anti-FLAG (Agilent) for 4 h at 4°C. After washing with 1% bovine serum albumin in PBS, grids were incubated with 12-nm gold-conjugated goat anti-rabbit IgG (H + L), EM grade (Jackson Immunoresearch Laboratories) and/or 5- nm gold-conjugated goat anti-mouse IgG (Sigma-Aldrich). Samples were then stained with

2% uranyl acetate for 1 min, and viewed by TEM as described above.

3D-Structured Illumination Microscopy (SIM). Super-Resolution 3D-SIM images were captured on a DeltaVision OMX-SR system (GE Healthcare) equipped with a 60X/1.42

NA oil immersion objective. Tau filaments labeled with Alexa Fluor 488, Cy3 or Cy5 were excited with 488, 568, or 640 nm laser light, respectively, with resulting fluorescence captured with a dedicated CMOS camera for each line. Image stacks (1.25 µm) were acquired with a z-distance of 0.125 μm and computationally reconstructed to generate

48 super-resolution optical serial sections. Images were reconstructed with Softwork software package version 6.5.2 (GE Healthcare). Subsequent image processing was performed using

ImageJ software.

Mathematical modeling. Systems of ordinary differential equations (ODEs) derived from aggregation models (Fig. 20) were coded in MATLAB (2016b, The MathWorks) and converted to C code using MATLAB Coder. Numerical solutions were generated using

ODE15s, a MATLAB function that implements the variable order method for solving stiff

ODEs [140]. The models were fit to recombinant 2N4R aggregation data consisting of total protomer concentration developed over 24 h at 0.4, 0.5, 0.6, 0.8 and 1 μM bulk tau levels, and filament length distributions at every time point [87]. Parameters were allowed to vary except ke+ and ke-, which were constrained within two-fold of their experimentally estimated values [87]. Fitting yielded systematically calculated parameter sets that minimized the root square error for both time series and length distributions at each bulk tau concentration.

Main (Si) and total (STi) sensitivity indices were calculated using the extended Fourier

Amplitude Sensitivity Test (eFAST) [141] and either the quality of fits to protomer concentration or length distribution time series as model output. Parameter space corresponding to best fit parameters ± 15% variation was searched using random phase shift resampling (n = 15 shifts) to ensure stability of the estimates [142]. For two- dimensional data (i.e., protomer concentration vs. time), sensitivity indices were calculated for each time point and expressed as a range, whereas for 3-dimensional data (i.e., length distribution vs. time) each time point represents the average sensitivity index across all bins. Mathematical simulations and analyses were scripted and performed in MATLAB.

49 Results

Tau filaments anneal in vitro. To enable selective detection of tau filament populations, two recombinant human 2N4R tau constructs containing N-terminal V5- or FLAG- epitopes were engineered. Full-length 2N4R was used as the tau isoform for both constructs because it aggregates efficiently at near physiological conditions of ionic strength, pH, and sulfhydryl reducing conditions [86]. On the basis of immunoblot analysis, tagged proteins strongly and selectively bound their cognate anti-V5 and anti-FLAG antibodies (Fig. 12).

To test their aggregation propensity, these proteins along with full-length N-terminally

6His-tagged and non-tagged 2N4R tau were incubated in the presence of Geranine G, after which time the products were subjected to transmission electron microscopy (TEM) imaging. Geranine G was used as anionic aggregation inducer because of its ability to drive aggregation of sub-micromolar full-length 2N4R tau into filaments having mass-per-unit length similar to authentic brain-derived filaments under near-physiological buffer conditions and sub-micromolar tau concentrations [86, 87, 104]. The resulting TEM micrographs showed that the presence of an N-terminal tag did not affect aggregation propensity or ~80 nm filament axial periodicity relative to non-tagged 2N4R tau, regardless of whether the tag was composed of 6His, V5, or FLAG (Fig. 12). In contrast, only filaments prepared from V5- or FLAG-tagged 2N4R monomers bind cognate anti-V5 and anti-FLAG antibodies (Fig. 13). Extensive decoration of filaments with these antibodies obscured filament morphology, but preserved observation of length, and in conjunction with 12-nm or 5-nm gold-conjugated secondary antibodies provided clear evidence for the presence of V5- or FLAG-tagged 2N4R tau protomers, respectively (Fig. 13).

Fig. 12. N-terminal tau fusion constructs display similar aggregation propensity and aggregate morphology. Recombinant 2N4R tau and N-terminal fusion constructs 6His-

2N4R, FLAG-2N4R, and V5-2N4R were expressed in E. coli, purified, and subjected to

SDS-PAGE (500 ng) or immunoblot (10 ng) analyses. (A) Representative SDS-PAGE migration pattern visualized by Coomassie blue stain. The presence of epitope tags changes protein migration on SDS-PAGE proportionate to the size of the tag. (B and C)

Immunoblots probed with polyclonal anti-V5 (B) and monoclonal anti-FLAG (C), respectively. Antibody labelings were specific for their cognate epitopes. (D) Recombinant

2N4R tau and N-terminal fusion constructs (E) FLAG-2N4R, (F) 6His-2N4R, and (G) V5-

2N4R were induced to aggregate in the presence of Geranine G inducer and incubated under aggregation conditions for 24 h at 37ºC. Samples were adsorbed onto grids and imaged by TEM. All three N-term fusion constructs aggregate with similar efficiency and produce filaments of similar morphology as non-fusion 2N4R tau. All scale bars = 100 nm.

Fig. 13. Control labelings with immunogold. Recombinant 2N4R tau and N-terminal fusion constructs 6His-2N4R, V5-2N4R and FLAG-2N4R were induced to aggregate in the presence of Geranine G inducer for 24 h at 37ºC. Samples were then adsorbed onto carbon-coated TEM grids and subjected to immunogold labeling using (A - D) anti-V5 primary/12-nm gold-conjugated secondary antibodies and (E - H) anti-FLAG primary/5- nm gold-conjugated secondary antibodies. Gold labelings using anti-V5 (black arrows) and anti-FLAG (white arrow) antibodies were specific for filaments containing their cognate epitope.

52 To test for tau filament annealing, separate populations of V5- and FLAG-tagged 2N4R filaments were prepared by incubating protein monomers in the presence of Geranine G beyond aggregation plateau (≥16 h; [86, 87]), then mixing them together and incubating for additional time (0 – 24 h) before subjecting products to immunogold labeling with anti-

V5/12-nm gold or anti-FLAG/5-nm gold beads followed by TEM imaging. Immediately after mixing, most fibrils labeled exclusively with 5- or 12-nm gold beads, reflecting the presence of aggregates composed entirely of one construct or the other in the mixture (Fig.

14A, B). However, after 24 h incubation, double labeling with anti-V5/12-nm gold and anti-FLAG/5-nm gold revealed the presence of fibrils with extended alternate segments of

12- or 5-nm gold, consistent with end-to-end annealing of the two populations (Fig. 14C-

F). This experiment was then repeated using filaments prepared from recombinant 2N4R tau covalently labeled with Alexa Fluor 488, Cy3, or Cy5 as substrate, octadecyl sulfate as aggregation inducer [143], and fluorescence microscopy as detection method. When filaments composed of each labeled-tau were mixed and incubated for 24 h, super- resolution fluorescence microscopy recorded the presence of fibrils with extended segments of Alexa Fluor 488, Cy3, or Cy5 fluorescence, again consistent with end-to-end annealing among the three filament populations (Fig. 14G-J). Together, these imaging experiments provide direct evidence for tau filament annealing.

Fig. 14. Direct visualization of tau filament annealing. (A-F) Synthetic filaments composed of V5-2N4R or FLAG-2N4R tau proteins were prepared separately in the presence of aggregation inducer Geranine G (1 µM tau incubated 16 h at 37°C), then mixed together and incubated for an additional 0 – 24 h. (A, B) When mixtures incubated for 0 h were subjected to immunogold labeling using either (A) anti-V5 or (B) anti-FLAG primary antibodies, only homogeneous labeling of single filaments was observed on TEM imaging. Black arrows, V5/12-nm gold immunoreactivity; hollow arrows FLAG/5-nm gold immunoreactivity. (C-F) In contrast, mixtures incubated for 24 h prior to immunogold labeling with both anti-V5 and anti-FLAG antibodies identified filaments containing extended regions of both gold labels, consistent with filament annealing. Arrowheads mark junctions between anti-FLAG and anti-V5 immunoreactivities in annealed filaments. (G-

J) Synthetic filaments composed of 2N4R tau covalently conjugated to Alexa Fluor 488

54 (AF488), Cy3 or Cy5 were prepared separately in the presence of aggregation inducer octadecyl sulfate (1.5 µM tau incubated 5 h at 37°C), then mixed together and incubated for an additional 24 h prior to fluorescence microscopy imaging. (J) Overlay composed of three channels (G-I), where white arrows mark junctions between annealed filaments.

Empirical estimation of the annealing rate constant for tau filament annealing. Because annealing rate depends on the concentration of filament ends [144], annealing kinetics are sensitive to acute perturbation of plateau filament length distributions (e.g., by shearing).

This approach has been used to estimate annealing rates of actin filaments [130]. When tau filaments composed of 6His-tau prepared in the presence of Geranine G for 24 h were incubated for an additional 0-24 h, both median and average length remained constant, consistent with the population attaining aggregation plateau (Fig. 15ABC). However, when this filament population was sheared to transiently increase the concentration of filament ends (by decreasing filament lengths) prior to the 24 h extended incubation (Fig.

15D), a time-dependent increase in both mean and median filament lengths was observed

(Fig. 15E, left axis). Because filament lengths are inversely proportional to filament concentration at aggregation plateau, these data could be converted into reciprocal filament concentration units (Fig. 15E, right axis). On the basis of regression analysis, the correlation between reciprocal filament end concentration and time was positive and linear throughout the time series (Fig. 15E), in conformance with the integrated rate law for a second order homotypic association. Under these conditions, the slope of the correlation approximates the rate constant for the process. Assuming that end-to-end annealing was

55 the principal process responsible for initial velocity yielded preliminary estimates of association rate constant ranging from 103 – 104 M-1 s-1 for this sheared filament population

(Fig. 15E). These data suggest that 2N4R tau annealing rates are robust and measurable.

A B C 200

100

0 0 h 24 h

D 150 E R2 = 0.87 3x108 100 2x108 R2 = 0.94

50 1x108

0 0 0 1 2

Fig. 15. Annealing time course approximates second order kinetics. Synthetic filaments prepared from 6His-2N4R tau protein (1 µM) incubated for 24 h at 37°C were subjected to TEM either (A) immediately or (B) after extending incubation for an additional 24 h at

37°C. (C) Quantification of filament length distributions (triplicate determination ± SD) revealed no significant differences between the two populations with respect to mean-

(blue) or median- (red) lengths, indicating that aggregation processes had plateaued under these conditions. (D) In contrast, shearing of 6His-2N4R filaments aggregated to plateau yielded an artificially shortened length distribution. (E) Sheared filaments incubated an additional 24 h at 37°C resulted in time-dependent changes in length distribution as 56 reflected in the mean (blue) and median (red) lengths of the filament populations (triplicate determination ± SD; left axis). Mean and median lengths were converted into concentrations of filament ends assuming a critical concentration of 200 nM [87] and two active ends per filament (plotted as reciprocals on the right axis). Linear regression analysis (solid lines) yielded slopes of 4.9 x 103 and 1.1 x104 M-1 s-1 for the reciprocal mean and median filament concentration time series, respectively.

Mathematical model of tau fibrillation. To rigorously quantify the contribution of annealing and other secondary processes to tau aggregation kinetics, 2N4R tau aggregation time series were fit by an equilibrium nucleation-elongation scheme [87, 145] modified to include secondary events including secondary nucleation, fragmentation, and end-to-end annealing (Fig. 11). The nucleation component of the primary pathway was constrained to a cluster size of two on the basis of previous rate measurements [87]. Therefore, the smallest stable filament corresponded to a trimer, which also is reported to be the minimal size for spontaneous propagation among cells in biological models [116]. The elongation phase also was constrained by experimental estimation of rate constants ke- and ke+ [86,

87]. In addition, elongation was assumed to proceed by adding or losing one monomer at a time from filament ends and to be governed by rate constants that were insensitive to filament length [146, 147]. The expressions describing the primary nucleation-elongation pathway were extended to model filaments up to N = 900 protomers in length (Fig. 16, black font).

Secondary nucleation (“S”) was added to this model assuming it was governed by the

57 same cluster size as primary nucleation (i.e., dimer), and that secondary nuclei detached immediately from fibrils to join primary nuclei in forming a single bulk population (Fig.

11, green arrow). It also was assumed that the secondary nucleation rate was proportional to filament concentration and length above a threshold length (i ≥ 10, corresponding to ≥2 nm in length, or approximately the length of a fully extended 306VQIVYK311 nucleating sequence motif [148]). The terms describing secondary nucleation are shown in Fig. 16

(green font).

Filament fragmentation (“F”) was modeled as a first order process governed by rate constant kfr and filament concentration ci. The model assumes that all products of breakage are filamentous (i.e., at least 3 protomers in length), and so i must be ≥ 6 before being

′ susceptible to fragmentation (Fig. 11, red arrow). In theory, the fragmentation rate (푘푓푟) for stiff rods depends on both filament length (i) and location of the breakage point [144].

′ Therefore, the model included power exponent β to capture length dependence of 푘푓푟. The exponent β was constrained to range from 0 to 4, its theoretical maximum [144, 149].

′ However, the model made the simplifying assumption that breakage propensity (푘푓푟) was identical between all protomers of each filament [131, 150]. The terms describing tau filament fragmentation are shown in Fig. 16 (red font).

Finally, end-to-end annealing (“A”) was modeled as a second order process governed by rate constant kan and the concentrations of filaments of length i and j (i.e., ci, cj) (Fig.

11, blue arrow). Like fragmentation, annealing involves only filaments, and so it acts only

′ when i and j ≥ 3. In theory, annealing rate (푘푎푛) should vary inversely with reactant lengths i and j, with magnitude dependent on the steric requirements of the annealing interaction

[144, 145]. Therefore, the model included power exponent α to capture the dependence of

58 ′ 푘푎푛 on both reactant lengths. When i = j, the power dependence of annealing rate on length

is 2α. Therefore, 2α was constrained to range from 0 to 3, its theoretical maximum [144].

The terms describing end-to-end annealing are shown in Fig. 16 (blue font).

푁 푁 푑푐 1 = −2(푘 푐2 − 푘 푐 ) − ∑(푘 푐 푐 − 푘 푐 ) − ∑ 2(푘 푖푐 푐2 − 푘 푖푐 푐 ) (S1) 푑푡 푛+ 1 푛− 2 푒+ 1 푖−1 푒− 푖 푛2+ 푖 1 푛2− 푖 2 푖=3 푖=10 푁 푑푐 2 = (푘 푐2 − 푘 푐 ) − (푘 푐 푐 − 푘 푐 ) + ∑(푘 푖푐 푐2 − 푘 푖푐 푐 ) (S2) 푑푡 푛+ 1 푛− 2 푒+ 1 2 푒− 3 푛2+ 푖 1 푛2− 푖 2 푖=10 푁−3 푁 푑푐 3 훼 푖 훼 푖 훽 3 = (푘 푐 푐 − 푘 푐 ) − (푘 푐 푐 − 푘 푐 ) − 푘 ( ) ∑ 2푐 푐 ( ) + 푘 2 ∑ 푐 ( ) (S3) 푑푡 푒+ 1 2 푒− 3 푒+ 1 3 푒− 4 푎푛 푁 3 푖 푁 푓푟 푖 푁 푖=3 푖=6 푁−4 푁 푑푐 4 훼 푖 훼 푖 훽 4 = (푘 푐 푐 − 푘 푐 ) − (푘 푐 푐 − 푘 푐 ) − 푘 ( ) ∑ 2푐 푐 ( ) + 푘 2 ∑ 푐 ( ) (S4) 푑푡 푒+ 1 3 푒− 4 푒+ 1 4 푒− 5 푎푛 푁 4 푖 푁 푓푟 푖 푁 푖=3 푖=7 푁−5 푁 푑푐 5 훼 푖 훼 푖 훽 5 = (푘 푐 푐 − 푘 푐 ) − (푘 푐 푐 − 푘 푐 ) − 푘 ( ) ∑ 2푐 푐 ( ) + 푘 2 ∑ 푐 ( ) (S5) 푑푡 푒+ 1 4 푒− 5 푒+ 1 5 푒− 6 푎푛 푁 5 푖 푁 푓푟 푖 푁 푖=3 푖=8 푗 = 6: 푁 − 4

(푁−3)−(푁−푗) 푁−푗 훼 훼 훼 훼 푁 훽 훽 푑푐푗 푖 푗 − 푖 푖 푗 푖 푗 = (푘 푐 푐 − 푘 푐 ) − (푘 푐 푐 − 푘 푐 ) + 푘 ∑ 푐 푐 ( ) ( ) − 푘 ∑ 2푐 푐 ( ) ( ) + 푘 2 ∑ 푐 ( ) − 푘 ( ) (푗 − 5)푐 푑푡 푒+ 1 푗−1 푒− 푗 푒+ 1 푗 푒− 푗+1 푎푛 푖 푗−푖 푁 푁 푎푛 푗 푖 푁 푁 푓푟 푖 푁 푓푟 푁 푗 (S6) 푖=3 푖=3 푖=푗+3

푑푐 푁−3 = (푘 푐 푐 − 푘 푐 ) − (푘 푐 푐 − 푘 푐 ) 푑푡 푒+ 1 푁−4 푒− 푁−3 푒+ 1 푁−3 푒− 푁−2 (푁−6) 훼 (S7) 푖 훼 (푁 − 3) − 푖 3 훼 푁 − 3 훼 푁 − 3 훽 + 푘 ∑ 푐 푐 ( ) ( ) − 푘 2푐 푐 ( ) ( ) − 푘 ( ) ((푁 − 3) − 5)푐 푎푛 푖 (푁−3)−푖 푁 푁 푎푛 푁−3 3 푁 푁 푓푟 푁 푁−3 푖=3

(푁−5) 푑푐 (푁 − 2) − 푖 훼 푖 훼 푁 − 2 훽 푁−2 = (푘 푐 푐 − 푘 푐 ) − (푘 푐 푐 − 푘 푐 ) + 푘 ∑ 푐 푐 ( ) ( ) − 푘 ( ) ((푁 − 2) − 5)푐 (S8) 푑푡 푒+ 1 푁−2 푒− 푁−1 푒+ 1 푁−1 푒− 푁 푎푛 푖 (푁−2)−푖 푁 푁 푓푟 푁 푁−2 푖=3

(푁−4) 푑푐 푖 훼 (푁 − 1) − 푖 훼 푁 − 1 훽 푁−1 = (푘 푐 푐 − 푘 푐 ) − (푘 푐 푐 − 푘 푐 ) + 푘 ∑ 푐 푐 ( ) ( ) − 푘 ( ) ((푁 − 1) − 5)푐 (S9) 푑푡 푒+ 1 푁−2 푒− 푁−1 푒+ 1 푁−1 푒− 푁 푎푛 푖 (푁−1)−푖 푁 푁 푓푟 푁 푁−1 푖=3

(푁−3) 푑푐 푖 훼 푁 − 푖 훼 푁 훽 푁 = (푘 푐 푐 − 푘 푐 ) + 푘 ∑ 푐 푐 ( ) ( ) − 푘 ( ) (푁 − 5)푐 (S10) 푑푡 푒+ 1 푁−1 푒− 푁 푎푛 푖 푁−푖 푁 푁 푓푟 푁 푁 푖=3

Fig. 16. Mathematical models of tau aggregation. The time-dependent evolution of tau

filaments of protomer length N was modeled with a system of ordinary differential

equations assuming reversible association of monomers. The number of equations was

limited to N = 900. The model included terms for the primary processes of nucleation and

elongation (black font), and the secondary processes of secondary nucleation (green font),

end-to-end annealing (blue font), and fragmentation (red font).

59 Combination of the primary and all three secondary processes created a full model termed “NEAFS”. It, along with the simplest model containing only the primary nucleation-elongation steps (NE), were then fit to aggregation time series collected at different bulk 2N4R tau concentrations (0.4, 0.5, 0.6, 0.8 and 1.0 µM) in the presence of

Geranine G aggregation inducer using quantitative TEM imaging [138]. These assay conditions were used because they allowed quantification of 2N4R fibrillation in terms of both total filamentous protomer concentration (Fig. 17A) and relative filament length distribution (Fig. 17B) at times through fibrillation plateau. The resulting time series were fit by simultaneously minimizing the root square error between observed and modeled evolution of total protomer concentration and filament length distribution. On visual inspection, the simple NE model approximated the evolution of total protomer concentration (Fig. 17A), but not of length distribution at any time point at any bulk tau concentration (Fig. 17B). Specifically, this model was unable to recapitulate the skew toward longer lengths in observed length distributions (Fig. 17B). In contrast, the full

NEAFS model more accurately captured protomer concentrations and length distributions at all tau concentrations and time points (Fig. 17AB). These results indicate that accurate description of 2N4R tau aggregation kinetics requires inclusion of secondary processes.

60 80 A B 0.4 µM 0.5 µM 0.6 µM 0.8 µM 1.0 µM 0.8 60 40 NE ( ) 20 0 0.4 80 Time (h) Time (h) Time (h) Time (h) Time (h) 60 0.25 0.25 0.25 0.25 0.25 1 1 1 1 1 40 2 2 2 2 2 NEAFS ( ) 5 5 5 5 5 20 24 24 24 24 24 0 0 0 5 10 24 0 200 0 200 0 200 0 200 0 200

Fig. 17. Fits of models to protomer concentration and length distribution time series.

Full-length recombinant 2N4R tau (0.4 μM, red; 0.5 μM, blue; 0.6 μM, green; 0.8 μM, brown; 1 μM, black) incubated (37ºC) in the presence of Geranine G (100 μM) inducer was assayed for fibrillation as a function of time (up to 24 h) in terms of (A) protomer concentration and (B) length distribution using TEM methods. Each point represents the average of triplicate determination, whereas dashed and solid lines represent fits to data points by the NE and NEAFS mathematical models, respectively.

Model Evaluation. To clarify the relative contribution of each secondary process to tau aggregation dynamics, the NEAFS model was subjected to global sensitivity analysis

(GSA). GSA quantifies how changes in model input values affect model output, defined here as goodness of fit to the observed protomer concentration and length distribution time series at 1 µM bulk tau concentration. This condition was analyzed because it populated length bins even at the earliest time points (Fig. 17B). The resulting sensitivity indices for each kinetic parameter represents the fraction of output variance resulting from changes in parameter value on a normalized scale of 0 to 1. The first order index (Si) represents model

61 variance explained by the variability of each parameter individually, whereas the total sensitivity index (STi) includes additional impact arising through interactions among parameters. GSA revealed that the quality of fit of the NEAFS model to protomer concentration was strongly sensitive to primary processes (i.e., parameters associated with nucleation and extension) but only weakly sensitive to secondary processes aside from annealing (Fig. 18A). These results were consistent with secondary processes having limited impact on aggregation rate (relative to primary processes) and no effect on aggregation plateau [123]. Moreover, differences between STi and Si were small (Fig. 18A), indicating that the quality of the NEAFS fits were driven directly by individual parameters rather than by interactions among them. In contrast, the quality of fit of the NEAFS model to length distribution data was far more sensitive to secondary processes and to interactions among them (Fig. 18B). With respect to modeling methods, these results indicate that protomer concentration and length distribution data provide complementary information.

With respect to tau aggregation dynamics, the calculated sensitivity indices (Fig. 18AB) predict that annealing and fragmentation, but not secondary nucleation, are the major secondary processes under in vitro 2N4R aggregation conditions.

62 1 Protomer concentration A STi -Si Si

0.5

0 1 B Length Distribution

0.5

0 kn- kn+ ke- ke+ kan  kfr  kn2- kn2+ Nucleation Extension Anneal Fragment Sec nuc Primary Secondary Aggregation process

Fig. 18. Global sensitivity analysis of the NEAFS model. Main (Si) and total sensitivity

(STi) indices were calculated for the NEAFS model fit to the 1 µM starting tau concentration dataset. Indices describe the sensitivity of model output in the forms of (A) protomer concentration or (B) length distribution to changes in NEAFS input parameters. Higher order interactions were then calculated as STi - Si and plotted along with Si to identify sources of sensitivity. Each bar represents the maximum and minimum sensitivity indices for each parameter across (A) all time points or (B) all length bins at all time points.

63 To test this hypothesis, various combinations of the primary and each of the three secondary processes were prepared to yield six additional mathematical models composed of nucleation-elongation steps with either one (NEA, NEF, NES) or two (NEAS, NEFS,

NEAF) secondary processes. Fits of all models to protomer concentration (Fig. 19) and length distribution (Fig. 20) time series were then compared visually with those for NE and

NEAFS. Consistent with GSA, protomer concentrations were only weakly sensitive to inclusion of secondary processes. In fact, models encoding combinations of any two or more secondary processes fit protomer concentration data similarly well (Fig. 19), perhaps reflecting the positive impact of adding free parameters to modeling output. However, only inclusion of both annealing and fragmentation processes could simultaneously capture length distribution (Fig. 20). These data confirmed that secondary nucleation was an unessential variable, and that the simplest model capable of rationalizing both total protomer concentration and filament length distribution time series was NEAF.

64 NE NES NEF NEA 0.8

0.4

0 NEFS NEAS NEAF NEAFS 0.8

0.4

0 0 5 10 24 0 5 10 24 0 5 10 24 0 5 10 24

Fig. 19. Fits of mathematical models to protomer concentration time series.

Mathematical models composed of nucleation-elongation (NE) and either one (NES, NEF,

NEA), two (NEFS, NEAS, NEAF) or three (NEAFS) secondary processes were simultaneously fit to protomer concentration and length distribution time series (i.e., the same data as shown in Fig. 17A). Each point represents protomer concentration as a function of time (0 – 24 h at 37°C) and starting bulk 2N4R tau concentration (0.4 μM, red;

0.5 μM, blue; 0.6 μM, green; 0.8 μM, brown; 1 μM, black), whereas solid lines depict a simulation of the best fit of each model to the data points.

65 Kinetic model NE NES NEF NEA NEFS NEAS NEAF NEAFS Time (h) 60 0.25 1 2 0.4 µM 40 5 24 20 0 Time (h) 60 0.25 1 40 2 0.5 µM 5 20 24 0 Time (h) 60 0.25 1 40 2 0.6 µM 5 20 24 0 Time (h) 60 0.25 1 2 40 5 0.8 µM 24 20 0 Time (h) 60 0.25 1 2 40 5 1.0 µM 24 20 0 0 200 0 200 0 200 0 200 0 200 0 200 0 200 0 200

Fig. 20. Fits of mathematical models to length distribution time series. Mathematical models composed of nucleation-elongation (NE) with either one (NES, NEF, NEA), two

(NEFS, NEAS, NEAF) or three (NEAFS) secondary processes were simultaneously fit to protomer concentration and length distribution time series (i.e., the same data shown in

Fig. 17B). Each set of points represents tau filament length distribution as a function of time (0 – 24 h at 37°C) and starting bulk 2N4R tau concentration (0.4 μM, red; 0.5 μM, blue; 0.6 μM, green; 0.8 μM, brown; 1 μM, black), whereas lines depict a simulation of the best fit of each model to the data points.

66 All rate constants calculated from the final NEAF model are summarized in Table 3.

4 The intrinsic rate constant for 2N4R tau filament annealing (kan) was estimated as 4.9 x10

M-1s-1, confirming that annealing is a secondary process involved in 2N4R tau aggregation kinetics. We conclude that 2N4R tau filament ends are active, that they can efficiently engage in homotypic interactions (i.e., with tau monomers, oligomers, and filaments), and that their engagement in secondary processes takes the form of an equilibrium with filament fragmentation.

Table 3. Model parameters

Symbol Description Units Values

−1 −1 푘푛+ Nucleation association rate constant M s 84

−1 푘푛− Nucleation dissociation rate constant s 0.82

Elongation association – rate constant for 푘 M−1s−1 5.0 x104 e+ monomer addition to both ends

Elongation dissociation – rate constant for 푘 s−1 1.0 x 10-2 e− monomer subtraction from both ends

−1 −1 4 푘푎푛 Annealing rate constant (for trimers) M s 4.9 x10

−1 -5 푘푓푟 Fragmentation rate constant (hexamer) s 2.3 x 10

Annealing power exponent for length 2α none -0.70 dependence

Fragmentation power exponent for length β none 0.50 dependence

67 Discussion

These data clarify the individual contributions of secondary pathways to tau aggregation dynamics. First, end-to-end annealing is a relatively efficient interaction at short filament lengths, with its intrinsic rate constant (kan) being of similar magnitude as that for monomer elongation (ke+) (Table 3). Although our estimate for kan is orders of magnitude slower than that reported for actin annealing [131], the value is typical of diffusion-mediated protein-protein interactions in the absence of long-range attractive forces [151]. As tau filaments lengthen, however, kan changed by the power of parameter

2α = -0.70 (Table 3). The negative sign of 2α indicates that annealing efficiency decreases with increasing length (as predicted by theory [144]), whereas its magnitude suggests that either the steric requirements for annealing were low (as reported for filamentous actin

[149]) or that filament diffusion rates were limited by aggregate flexibility. Indeed, the estimated value of 2α approximates the power exponent for length-dependent flexible linear polymer diffusion (-0.588, [152]), consistent with the persistence length of synthetic

2N4R filaments being far lower than for actin [153].

Second, tau fibrillation dynamics also are governed by fragmentation. The intrinsic rate constant for 2N4R tau filament fragmentation (kfr) estimated here was orders of magnitude slower than for monomer dissociation (ke-; Table 3), consistent with the breakage of additional non-covalent bonds. However, it was more similar in magnitude to the fast rate constants estimated for prion fragmentation [154, 155] than to the constants reported for cytoskeletal assemblies such as actin [131] or for tau over weeks under oxidizing conditions [155]. kfr was length dependent with power exponent β = 0.50 (Table 3). The positive sign of β indicates that fragmentation efficiency increased with increasing length

68 (as predicted by theory [144]), whereas its magnitude was substantially lower than reported for rigid amyloid or cytoskeletal aggregates [131, 156]. These data again are consistent with the flexible nature of synthetic 2N4R tau filaments.

Together, fragmentation and annealing rate constants create an equilibrium at filament ends (Keq) that influences the size distribution of tau aggregates in a population. This effect may be especially impactful for tau aggregate toxicity, which in biological model systems varies inversely with aggregate size [157, 158]. Tau aggregate size distribution is further controlled by power exponents β and 2α, which act together to render Keq length dependent.

For example, Keq was subnanomolar and strongly supportive of annealing when 2N4R tau filaments were short, whereas it became less favorable as filaments lengthened (Fig. 21).

Post-translational modifications that rigidify tau protein (such as phosphorylation [159]), and thereby increase filament persistence length, are predicted to raise power exponents β and 2α still higher [144]. Under conditions where kfr and kan remain constant, increases in power exponents will sharpen the length dependence of Keq (Fig. 21) to favor fragmentation, a process that in authentic prions is associated with propagation efficiency

[126].

69 -2

-3

-4

-5

-6

-7

-8

-9

-10 0 500 1000 Length (i protomers)

Fig. 21. Tau filament annealing/fragmentation equilibrium is length dependent. The solid line represents the equilibrium constant for annealing/fragmentation (Keq) calculated from kan, 2α, kfr, and β values (Table 1) as a function of tau filament length, whereas the dashed line corresponds to Keq calculated under identical conditions except that 2α and β were doubled to -1.4 and 1.0, respectively.

Our modeling results predict that annealing and fragmentation are the major secondary processes governing tau aggregate size and toxicity. In contrast, kinetic modeling of other cross-β-sheet aggregate forming proteins has implicated secondary nucleation as the principal process playing this role [125]. The discrepancy likely stems from the unique structure of tau filaments, which differ from other aggregates in being surrounded by a

“fuzzy coat”. This structure, which has been detected on synthetic 2N4R tau filaments

[160] as well as authentic brain-derived filaments [161], arises from the intrinsically disordered regions of tau protomers that extend away from the cross-β-sheet aggregate core 70 [162]. The fuzzy coat limits access of tau filament surfaces to macromolecules, including the tau monomers necessary to support secondary nucleation [160]. Our kinetic modeling results were consistent with these observations.

In summary, these data identify tau filament ends as mediating a range of homotypic interactions that include monomers and other tau aggregates. Indeed, it has been argued that exposed β-sheet ends are fraught with danger, and as a result are blocked to lessen risk of interaction [135]. It remains to be seen whether tau filament ends also engage in heterotypic interactions potentially associated with toxicity.

Chapter 4. A liquid chromatography tandem mass spectroscopy approach for

quantification of protein methylation stoichiometry

Reproduced in part with permission from G.L. Cooper, C.J. Huseby, C.N.

Chandler, J.C. Cocuron, A. Alonso, J. Kuret (2018) A liquid chromatography tandem mass spectroscopy approach for quantification of protein methylation stoichiometry. Analytical

Biochemistry 545, 72-77. The full article is available at: https://doi.org/10.1016/j.ab.2018.01.018.

C.J. Huseby conceived the research and organized the collaboration with Drs.

Cocuron and Alonso, who performed all LC/MS-MS measurements. G.C. and C.N.C. performed biochemical isolations of proteins. C.J.H. and G.C. wrote the initial draft of the manuscript and share first-authorship of the published manuscript. J.K. modified the manuscript. This work was supported by the National Institutes of Health grant AG14452 to J.K. C.N.C. was supported by training grant GM118291.

72 Introduction

In the preceding chapter I disclosed the aggregation mechanism of synthetic full- length human 2N4R tau protein under defined conditions. But unlike recombinant tau produced in our laboratory, authentic human tau protein isolated from human brain contains many PTMs [163]. As noted in Chapter 3, some of these, such as phosphorylation on Ser, Thr and Tyr residues [164], are expected to alter the rate and extent of aggregation.

Recently, tissue-derived human tau was reported to also contain modifications on Lys residues in the form of acetylation [165] and ubiquitylation [166, 167]. Our laboratory went on to discover a third Lys modification, methylation, that is capable of strongly depressing tau aggregation propensity [81, 82]. As a result, this modification may be an important regulator of tau lesion formation in AD. Following our discovery, others reported that methylation of tau protein extends to Arg residues as well. But the significance of any of these PTMs depends on both the sites modified and their occupancy. To identify sites of modification, our laboratory applied “bottom up” proteomic methods to human-derived samples, and identified ≥17 modified Lys and Arg residues [81, 168]. The methylation pattern differed with disease state, indicating that this PTM is positioned to modulate tau aggregation propensity and turnover [81, 82]. However, the full functional implications of these observations were unknown in part because biological effects are mediated by modification stoichiometry, and bottom-up approaches alone capture site distribution but not occupancy. Characterization of complex methylation substrates such as tau would benefit from a “top down” method [169] capable of capturing overall stoichiometry in terms of quantity (mol methyl equivalents per mol of protein) and quality (chemical form)

73 of methylation under various biological conditions. A second goal of this dissertation is to deliver this methodology.

The classic top-down approach for characterizing bulk methylation stoichiometry from an ensemble of intact proteins is amino acid analysis, which leverages the unique stability of Lys [170], Arg [170, 171], and their methylated derivatives 1meK, 2meK, 3mek, 1meR and 2meR (composed of both ADMA, NG,NG′-Dimethyl-L-arginine; and SDMA, NG,NG′-

Dimethyl-L-arginine) upon acid hydrolysis [172, 173]. In contrast, other Lys and Arg derivatives are unstable and reduce to the parent amino acids under these conditions. For example, the most prevalent stable mammalian Lys modifications yield amide linkages through acylation (e.g., acetylation, etc; [174]) and isopeptide linkages through conjugation with ubiquitin-like proteins [175], both of which readily hydrolyze in parallel with peptide bonds. Similarly, ADP-ribosylation of Arg, which is mediated by N,O-acetal linkages, also is acid labile. As a result, it has been possible to estimate Lys and Arg methylation stoichiometries for specific proteins including calmodulin [176] and myelin basic protein

(MBP) [177-180]. Nonetheless, limitations of amino acid analysis have hampered its general application. First, simultaneous separation of Lys, Arg, and all of their methylated derivatives by liquid chromatography is problematic [173], and so methylation stoichiometry has been reported primarily for proteins of low methylation complexity such as calmodulin and MBP (calmodulin contains 3meK and no other methylated residue [176], whereas MBP methylation is limited to 1meR and 2meR [177-180]). Second, amino acid analysis requires derivitization for quantification, which yields poor sensitivity with classic colorimetric agents such as ninhydrin [181]. Replacement with fluorometric detection greatly improves sensitivity [172, 173], but even mid-pmol quantification is not adequate

74 for characterizing proteins isolated in low abundance from tissues. Moreover, common derivitization agents such as o-pthalaldehyde (OPA) yield Lys adducts that dimerize and quench [182], thereby lowering detection sensitivity for this amino acid still further.

Alternative top-down approaches are confined to samples where the complexity of methylation is modest [183].

In this chapter I introduce a targeted mass spectrometric approach for quantifying protein Lys and Arg methylation stoichiometry, and reduce it to practice. The method leverages the established stability of methylated amino acids to acid hydrolysis, but then uses LC-MS/MS to quantify analytes with high accuracy and sensitivity. Because targeted precursor/product ion pairs are detected by a triple-quadrupole mass spectrometer operated in multiple reaction monitoring (MRM) mode, the approach yields simultaneous and sensitive detection of even complex analyte mixtures. The final assay successfully quantified the methylation stoichiometry of multiply modified, tissue-derived proteins.

Materials and Methods.

Materials. Recombinant human 2N4R tau was prepared as described previously [81]. All other reagents were obtained from commercial vendors, including Sigma-Aldrich (St

Louis, MO) for calmodulin from bovine testes (P1431), MBP from bovine brain (M1891), unfractionated whole histone from calf thymus (H9250), PVDF membrane (IPVH00005),

1meR (M7033), ADMA (D4268), SDMA (D0390), and all unmodified L-amino acids, and

Chem-Impex Intl (Wood Dale, IL) for amino acids 1meK, 2meK and 3meK.

Reductive tau methylation. Recombinant human 2N4R tau was reductively methylated with NaBH3CN and formaldehyde as described previously [81, 184]. Reactions were quenched by addition of glycine after 0, 7, 15, 30 and 60 min incubation. Non-methylated

75 controls were processed identically, except that formaldehyde was omitted from the reaction.

SDS PAGE and blotting. Tau proteins were electrophoresed through 8% acrylamide gels, then transferred (100 V for 1 h at 4C) to PVDF membranes in Transfer Buffer A (25 mM

Tris, 0.2 mM glycine, 10% methanol). PVDF membranes were then stained (50%

Methanol, 7% Acetic Acid, 0.1% Coomassie Brilliant Blue) for 30 min at room temperature with agitation, and then destained (50% methanol, 7% acetic acid) for ~10 min at room temperature with agitation until the staining pattern was visible. After membranes dried at room temperature, Coomassie-blue stained bands were excised using a razor blade

(2 x 6 mm slices) and stored at -20°C until used.

MBP, unfractionated histones, and calmodulin were subjected to SDS-PAGE as above, except that electrophoresis was performed through 15% acrylamide gels. MBP and unfractionated histones were then transferred (100 V for 1 h at 4°C) to PVDF membranes in Transfer Buffer B (25 mM Tris-HCl, 0.2 mM glycine, 20% methanol), whereas calmodulin was transferred (100 V for 1 h at 4°C) to PVDF in Transfer Buffer C (25 mM

Tris, 0.2 mM glycine, 2 mM CaCl2, 20% methanol) as reported previously [185]. Staining and band excision were performed in the same way as described above for tau proteins.

Acid hydrolysis. Excised PVDF membranes were placed in glass tubes containing 1 mL

6 M HCl and purged with nitrogen as described previously [186]. Samples were then sealed and incubated for 24 h at 125C. Following hydrolysis, samples were dried under nitrogen at 60C and then stored at -20C until LC-MS/MS analysis.

LC-MS/MS quantification. Dried hydrolysates were resuspended in 250 μL of 10 mM

HCl, vortexed, filtered (0.2 μm Pall nanosep Mf operated at 14,000g for 10 min at 4°C),

76 and finally prepared for LC by 10-fold dilution with water in glass vials. LC separations were performed on a Hypercarb column (100 × 2.1 mm, 5 µm pore; Thermo Fisher

Scientific) operated at 0.2 mL/min using a mobile phase prepared from mixtures of Solvent

A (acetonitrile containing 0.1% formic acid) and Solvent B (water containing 0.1% formic acid). The gradient expressed in terms of %-Solvent A was: 0–1 min, 0%; 1–1.1 min, 20%;

1.1–2.7 min, 45%; 2.7–5 min, 60%; 5–5.1 min, 90%; 5.1–7 min, 90%; 7–7.1 min, 0%; 7.1–

10 min, 0% [187].

Mass spectra were acquired on a triple-quadrupole QTRAP 5500 (AB Sciex) using turbo spray ionization at 2.5 kV in positive ion mode and MRM of parent and characteristic product ions (the transitions monitored are listed in Table 1). The curtain gas (nitrogen) and the collision-activated dissociation were set to 30 psi and medium, respectively. The

MS was set to have a dwell time of 35 ms. Analyst 1.6.1 software was used to acquire and process all data.

Analytical methods. To capture modification quality, the relative proportion ( P ) of all K X peptidyl-Lys residues in the form of 1meK, 2meK or 3meK (XmeK) was calculated ratiometrically by the equation:

XmeK P = K X 3 (1) Lys +  XmeK X =1

The stoichiometry of total peptidyl-Lys methylation (i.e., mol methyl groups per mol protein; SK) was then calculated by summing the relative proportion of each methyl-Lys species ( XP ) and multiplying it by the number of peptidyl-Lys residues (N) expressed K X in the target analyte:

77 3 SK = N XP (2)  K X X =1

Similarly, the relative proportion ( P ) of Arg methylation was calculated ratiometrically RX from mol amounts of all methyl-Arg species (XmeR) by the equation:

XmeR P = RX 2 (3) Arg +  XmeR X =1

The stoichiometry of peptidyl-Arg methylation (i.e., mol methyl groups per mol protein) was calculated by summing the relative proportion of each methyl-Arg species ( XP ) and RX multiplying it by the number of peptidyl-Arg residues (N) expressed in the target analyte:

2 SR = N XP (4)  R X X =1

Methylation time series were modeled as a simple exponential growth to maximum using the function:

−kappt yt = ymax (1− e ) (5) where kapp and ymax are the rate constant and maximum stoichiometry of methylation, respectively.

Statistics. Stoichiometry data were calculated as the mean ± SD of three biological replicates unless otherwise stated. Groups were compared with Student's t-test for single values and one-sample t-test for relative values. The probability (p) of differences between estimated parameters (kapp, and ymax) was assessed by z-test:

x − x z = 1 2 (6) 2 2 (sx1 ) + (sx2 ) where x1  sx1 and x2  sx2 are the pair of estimates  SE being compared, and z is the 1-

78 point of the standard normal distribution using and JMP 13.1 (SAS Institute, Cary, NC).

The null hypothesis was rejected for all statistical tests at p < 0.05.

Results

LC-MS/MS quantification of methylated amino acids. To establish assay conditions, unmodified Arg, Lys, the Lys isobar Gln, and methylated amino acids 1meK, 2meK,

3meK, 1meR, ADMR and SDMR were mixed together and subjected to LC-MS/MS. LC separation leveraged a porous graphitic carbon stationary phase developed with an acetonitrile gradient in 0.1% formic acid as the mobile phase. These conditions were used because they had previously been optimized for separation of the 20 naturally-occurring

L-amino acids [188]. All nine test amino acids eluted within the gradient (Fig. 22) where they were resolved by a combination of retention time and fragmentation pattern (Table

4). For example, Lys and its isobar, Gln, generated identical precursor/product transitions, but were easily resolved from each other on the basis of retention time (reported in [189] and confirmed in Fig. 22). In contrast, Arg and its methylated derivatives migrated identically during LC, but were resolved from each other by their distinct precursor/product transitions (Fig. 22). As a result, it was possible to simultaneously detect Lys, Arg, and their methylated derivatives in the mixture. However, as reported previously [190], isobaric dimethylated arginine derivatives ADMA and SDMA were not distinguishable under these conditions. For this reason, dimethyl arginine (i.e., 2meR) was captured and reported as the sum of SDMA and ADMA.

Fig. 22. Ion chromatograms for amino acid standards detected by multiple reaction monitoring. (A) Total ion count of mixed amino acid standards Lys (3.75 pmol), Gln (3.75 pmol), methyl-Lys (1meK, 2meK, 3meK; 1.5 pmol each), Arg (750 fmol), and methyl-Arg

(1meR, 1.5 pmol; 2meR, 0.75 pmol each of ADMA and SDMA), where each color represents the measured ion intensity of an individual transition precursor/product ion pair (see Table 1). (B) LC-MS/MS chromatograms of individual amino acid standards extracted from Panel A. The separation of all amino acids standards was completed within

10 min.

80 To determine detection sensitivity [191], calibration graphs relating total area under the

LC curve (i.e., signal) to the amount of each analyte were generated. These showed excellent linearity from mid fmol to low pmol levels, with correlation coefficients >0.995 and with corresponding limits of detection ranging from 0.2 – 1.3 fmol and limits of quantification ranging between 0.8 – 4.4 fmol (Table 4). Together these data demonstrate the feasibility of quantifying mixtures of Lys, Arg, and their methylated derivatives in the mid-fmol – low pmol range by LC/MS-MS.

81 Table 4. Mass spectrometry parameters and calibration of amino acids. Amino Precursor Product Retention Linear DPb EPb CEb CXPb r2 LoDc LoQc Acid ion ion timea range (mass) (mass) (min ± SD) (volts) (volts) (volts) (volts) (fmol) (fmol) (fmol) Lysine and derivatives Lys 147 84 1.16 ± 0.01 40 10 24 12 20 - 5000 0.995 1.30 4.35 1meK 161 84 1.27 ± 0.02 70 10 23 10 20 - 5000 0.998 0.56 1.85 2meK 175 84 1.39 ± 0.07 70 10 25 10 20 - 5000 0.999 0.51 1.69 3meK 189 84 1.46 ± 0.08 70 10 29 10 20 - 5000 0.999 0.42 1.41 Arginine and derivatives Arg 175 70 3.27 ± 0.01 60 10 33 10 20 - 5000 0.999 0.20 0.65 1meR 189 70 3.27 ± 0.01 50 10 31 8 20 - 5000 0.999 0.56 1.85 2meR 203 70 3.26 ± 0.01 46 10 33 10 20 - 5000 0.999 0.24 0.80 aMean ± SD of eight biological replicates collected over a six-month period bDP; declustering potential; EP, entrance potential; CE, collision energy; CSP, collision cell exit potential as optimized by Analyst 1.6.1 software. cLimits of detection (LoD) and quantification (LoQ) were extrapolated from linear regression analysis as signal-to-background ratios of 3:1 and 10:1, respectively [191].

82 Method validation with tau protein. Application of the above detection method to protein hydrolysates provides a route toward estimating relative molar stoichiometries of peptidyl

Lys and Arg methylation. To validate this approach, proteins with established methylation stoichiometries were separated by SDS-PAGE, immobilized by blotting onto PVDF filters, hydrolyzed to amino acids with concentrated HCl, then assayed for methylated amino acid content by the LC/MS-MS method described above. First we examined human recombinant 2N4R tau, which we previously showed could be reductively methylated on

Lys residues with 14C-formaldehyde to yield estimates of methylation stoichiometry [81].

When quantified over reaction time course, 2N4R tau methylation stoichiometry was found to increase monotonically to ~22 mol methyl/mol tau protein by 60 min of incubation [81].

By replacing 14C-formaldehyde with unlabeled formaldehyde and subjecting the samples to bottom-up proteomic analysis, it was shown that all detectable methyl incorporation at all analyzed time points was in the form of 1meK and 2meK [81]. We used these unlabeled samples (taken after 0, 7, 15, 30 and 60 min reductive methylation) to validate the ratiometric LC/MS-MS approach. When subjected to SDS-PAGE, all five samples migrated identically with intact 2N4R tau (Fig. 23A). Following transfer to PVDF membrane and acid hydrolysis, only Lys, 1meK, 2meK and Arg were identified in the samples by LC/MS-MS, consistent with reported proteomic results [81] and the established selectivity of reductive methylation for Lys residues [184]. The stoichiometry of Lys methylation (SK) was then calculated from measured quantities of Lys, 1meK and 2 meK, the known Lys composition of human 2N4R tau (44 Lys residues; [192]) and eqns. 1 and

2. Comparison of these estimated stoichiometries with those determined previously by 14C-

83 labeling showed excellent concordance between the two methods (Fig. 23B), with the null hypothesis being accepted at all time points (triplicate determinations).

These data indicate that LC/MS-MS could replace 14C-labeling for quantifying tau methylation stoichiometry. However, the approach provided additional information by disaggregating methylation stoichiometry into its qualitative 1meK and 2meK components.

These components appeared together early in the time series, but differed in their subsequent kinetics, with 1meK plateauing early while 2meK continued to increase over time (Fig. 23C). By 60 min, most methylation stoichiometry was in the form of 2meK

(compare Fig. 23 panels B and C). To quantify these observations, stoichiometry data were modeled as simple first-order approaches to plateau by eqn. 5. Both 14C-labeling and

LC/MS-MS stoichiometry data produced statistically similar estimates of pseudo-first order rate constant kapp and total plateau methylation stoichiometry ymax (Fig. 23D), demonstrating the utility of the LC/MS-MS assay for kinetic analysis. Disaggregating these total stoichiometry data into their 1meK and 2meK components revealed that the kapp for

2N4R tau mono-methylation was faster than that for dimethylation, but plateaued at lower levels (Fig. 23CD). As a result, 2meK was the predominant contributor to total stoichiometry at later time points (Fig. 23D). These data were consistent with the established kinetics of reductive methylation, where monomethylation must precede addition of a second methyl group [184]. Together these data illustrate the utility of

LC/MS-MS detection for quantification of Lys protein methylation stoichiometry and its dissociation into elementary components.

Fig. 23. Quantification of Lys methylation stoichiometry in tau protein. (A) recombinant human 2N4R tau that was reductively methylated for 0 (lane 1), 7 (lane 2), 15 (lane 3), 30

(lane 4) or 60 (lane 5) min was subjected to SDS-PAGE and Coomassie blue staining.

Bands excised for hydrolysis and LC-MS/MS analysis are marked by asterisks. (B) Time 85 course of methylation stoichiometry. Solid circles represent total methylation stoichiometry (in units of mol methyl/mol tau) as a function of time (yt) determined through radiolabeling with [14C]formaldehyde (n = 3) whereas hollow circles represent total methylation stoichiometry determined using LC-MS/MS (n = 3). Solid lines represent best fit of each time series with Eqn. 5. (C) Time course of methylation stoichiometry, where methylation stoichiometry determined by LC-MS/MS in Panel B was disaggregated into its

1meK (solid squares) and 2meK (hollow squares) components. Solid lines represent best fit of each time series with Eqn. 5. (D) Replot of data shown in Panels B and C, where ymax and kapp correspond to the calculated plateau stoichiometry (in units of mol methyl/mol tau) and the pseudo-first order rate constant of each time series, respectively. Time series are labeled as 14C ([14C] formaldehyde labeling), Total (total methylation stoichiometry determined by LC/MS-MS), 1meK (1meK component of LC/MS-MS stoichiometry), and

2meK (2meK component of LC/MS-MS stoichiometry). *, p < 0.05; ***; p < 0.001; n.s., p

> 0.05 on z-test (eqn. 6).

Application to tissue-derived proteins. Because of its sensitive and label-free method of detection, the LC/MS-MS approach is ideally suited for analysis of low abundance, tissue- derived samples. To test this utility, three tissue-derived methyl-protein samples were analyzed. The first of these was bovine calmodulin, which when isolated from beef brain or testes is reported to contain one 3meK residue and no other Lys or Arg methylations

[176]. When subjected to SDS-PAGE, bovine calmodulin electrophoresed as a 14 kDa species (Fig. 24A). After gel-purified calmodulin was acid hydrolyzed and assayed by LC- 86

MS/MS, only 3meK was detected (Fig. 24B). On the basis of the amino acid composition of calmodulin [176], a stoichiometry of 3.72 ± 0.57 mol methyl/mol calmodulin was calculated (Fig. 24B). This value corresponded to 1.24 ± 0.19 mol 3meK/mol calmodulin, consistent with the single 3meK modification site established by protein chemistry methods (null hypothesis accepted, p = 0.16).

The second sample was MBP, which reportedly contains 1meR and 2meR (but not methyl Lys) when isolated from bovine brain [177-180]. Upon SDS-PAGE separation, bovine MBP electrophoresed as a ~18 kDa species (Fig. 24A). After acid hydrolysis and

LC-MS/MS analysis, 1meR and 2meR were the only detectable methyl amino acids (Fig.

24B). On the basis of the amino acid composition of MBP, a stoichiometry of 0.65 ± 0.14 mol methyl/mol MBP was calculated (Fig. 24B) composed of ~0.2 mol each of 1meR and

2meR per mol MBP. These values were consistent with previous estimates made on the basis of amino acid analysis and colorimetric detection (0.18 – 0.80 mol 1meR/mol MBP and 0.12 – 0.31 mol 2meR/mol MBP; [177-180]), demonstrating that the utility of the ratiometric approach for estimating protein methylation stoichiometry also extends to methyl-Arg.

Finally, mixed histones were assayed to gauge the feasibility of detecting broad- spectrum methylation occurring within one sample. Histones are subject to diverse post- translational modifications, and have been reported to contain 1meK, 2meK and 3meK as well as 1meR and 2meR residues [193, 194]. When analyzed by SDS-PAGE, mixed histones prepared from calf thymus separated into constituent core H2A, H2B, H3, and H4 components (Fig. 24A). After H2A, H2B, and H3 were excised as a mixture and subjected to acid hydrolysis, LC-MS/MS analysis successfully confirmed 1meK, 2meK, 3meK, 87

1meR and 2meR as being simultaneously present in the sample (Fig. 24B). Estimates of stoichiometry in this case were averaged over the composition of the mixture (H2A, H2B, and H3) and therefore not reflective of any single species. However, the observed rank order stoichiometry of 2meK > 1meK was consistent with previously reported characterizations of arginine-rich histones by amino acid analysis [195, 196]. These results indicate that the LC-MS/MS method is able to detect and quantify both Lys and Arg methylation when present simultaneously in a qualitatively complex sample.

Fig. 24. Quantification of methylation stoichiometry in protein standards. (A) Coomassie- blue stained SDS-PAGE gel of input protein standards calmodulin (lane 1, 1 μg), MBP

(lane 2, 2 μg) and mixed histones (lane 3, 3 μg). Bands excised for hydrolysis and LC-

MS/MS analysis are marked by asterisks. The excised histone bands consisted of core histones H2A, H2B, and H3. (B) Methylation stoichiometry determined by LC-MS/MS and 88 eqns. 1-4. Bar height indicates contributions of 1meK, 2meK, 3meK, 1meR and 2meR to mol methyl/mol protein in each sample (triplicate determination) whereas error bars represent the standard deviation.

Discussion

These data indicate that an LC-MS/MS approach can reliably detect mid-fmol quantities of methylated amino acids and quantify methylation stoichiometries in proteins. It’s overall precision and accuracy are sufficient to replace 14C-labeling in kinetic analyses of protein methylation. However, its greatest utility is its compatibility with a broad range of low abundance, tissue-derived protein samples, its ability to disaggregate bulk methylation stoichiometry into its individual mono-, di-, and tri-methylated residue components, and its compatibility with diverse sample sources owing to its leveraging of SDS-PAGE for final sample preparation. Despite these advantages, accurate ratiometric quantification of tissue protein methylation stoichiometry will depend on three conditions. First, as with other methods, the analyte must be of high purity. While not attainable for all protein samples, the incorporation of an SDS-PAGE isolation step provides a convenient and powerful means of maximizing input sample purity. Second, ratiometric analysis requires full-length protein or at least precise knowledge of Lys and Arg composition. In the case of tau protein, this was easily achieved by using recombinant protein of defined isoform composition. Full-length, post-translationally modified tau proteins of known isoform composition also can be isolated from mammalian brain tissue and separated by SDS-

PAGE [197, 198], and so these will likely be amenable to analysis as well. In contrast, the 89 tau proteins that accumulate in human cerebral spinal fluid can become extensively proteolyzed, resulting in complex peptide mixtures ([199, 200]). Accurate quantification of methylation stoichiometry in this pool would be challenging. Finally, the ratiometric approach requires that diverse post-translational modifications other than methylation on

Lys and Arg residues break down to unmodified amino acids after acid hydrolysis. This condition is largely met for stable Lys modifications, which apart from specific hydroxylated derivatives associated with collagen [201], and certain acid-stable advanced glycation end products associated with aging and metabolic disease [202], are dominated by acid-sensitive acylations. Arg too is subjected to hydrolysable PTMs. However, Arg also can be modified by peptidyl Arg deiminases that convert peptidyl-Arg residues to citrulline [203], which upon hydrolysis yields ornithine rather than Arg. As a result, the denominator for eqn. 3 will be underestimated when applied to citrullinated analytes, resulting in overestimation of methylation stoichiometry. However, the stoichiometry of citrullination in vivo is low [204]. Indeed, we found that ratiometric quantification of bovine MBP methyl-Arg stotichiometry was consistent with previously reported analyses despite the presence of five citrullination sites on this analyte [205]. Because ornithine also is amenable to LC-MS/MS analysis [189], it may be possible to create a ratiometric assay for quantification of citrullination stoichiometry in future.

In summary, the method disclosed herein will complement existing bottom-up proteomic identification of protein methylation sites by yielding estimates of overall methylation stoichiometry and quality. The approach will facilitate study of methylation kinetics as well as placing observations of in vivo methylation into a more quantitative context. 90

Chapter 5. Protein methylation stoichiometry of tissue-derived human tau protein

Reproduced in part from work completed in collaboration with G.L. Cooper, C.N.

Chandler, K. Funk, J.C. Cocuron, A. Alonso, and J. Kuret

C.J. Huseby conceived the research and organized the collaboration with Drs.

Cocuron and Alonso, who performed all LC/MS-MS measurements. C.J. Huseby, G.C.,

C.N.C., and K.F. performed biochemical isolations of proteins. This work was supported by the National Institutes of Health grant AG05418 to J.K. C.N.C. was supported by training grant GM118291.

Introduction

The previous chapter demonstrated the feasibility of determining methylation stoichiometry in diverse tissue derived protein samples. In this chapter, I apply these methods to samples of tau protein isolated from human brain.

The first identified and best studied tau PTM is phosphorylation (reviewed in

[164]). Remarkably, more than half of the ~80 Ser and Thr residues in tau protein are phosphorylated in human brain, with new sites still being reported in the literature. The broad accessibility of hydroxyamino acid sidechains likely reflects the intrinsically disordered structure of monomeric tau, which exposes most residue sidechains to solvent.

On average, each site contains only 0.05 – 0.1 mol phosphate/mol. Despite these low occupancies of individual sites, two key findings indicate that phosphorylation plays an important role in tau biology. The first is the bulk stoichiometry of tau phosphorylation, which sums over all sites to achieve 2-3 mol phosphate/mol tau under basal conditions in normal human brain. The second is that bulk stoichiometry increases in AD brain to 8-9 mol/mol. Disease levels of bulk phosphorylation are sufficient to interfere with the ability of tau to interact with microtubules, its physiological binding partner [206, 207]. As a result, phosphorylation is positioned to act as gatekeeper in the cascade of events leading to tau aggregation.

Here, I investigate tau methylation stoichiometry with two objectives in mind.

Because tau methylation strongly depresses aggregation propensity, I first seek to determine whether tau is methylated in normal individuals, and whether those levels decrease with aging. A positive result would identify a potential age-related driver of tau

92 lesion formation. A second objective is to test methylation levels with disease, to determine whether the tau that accumulates in AD varies in methylation status with normal tau.

Materials and Methods

Materials. This study used only archival, de-identified post mortem human brain tissue samples from autopsies performed with informed consent of each patient or relative via procedures approved by the relevant institutional committees. The tau protein ladder composed of all six human isoforms was purchased from rPeptide (#T-1007-1).

SilverQuest silver staining kit was purchased from ThermoFisher Scientific.

Brain-derived tau protein. Soluble brain-derived tau was enriched using methods detailed previously [208]. Gray matter was homogenized in 5 v/w of homogenization buffer (20 mM MES, pH 6.8, 80 mM NaCl, 1 mM MgCl2, 2 mM EGTA, 0.1 mM EDTA,

1 mM PMSF) containing inhibitors of phosphoprotein phosphatases (10 mM sodium pyrophosphate, 20 mM NaF, 1 mM Na3VO4) and deacetylases (2 µM trichostatin A, 10 mM nicotinamde). After centrifugation (20 min at 27,000g), the supernatant was collected, adjusted to 0.5 M NaCl and 2% 2-mercaptoethanol, boiled (10 min), and recentrifuged (20 min at 27,000g). The resulting heat-stable supernatant fraction was treated with 2.5% (final concentration) of perchloric acid. After isolating the acid-soluble fraction by centrifugation

(20 min at 27,000g), protein was concentrated by TCA precipitation (20% w/v) followed by two washes in cold acetone. Dry protein pellets were stored at -20ºC until solubilized for LC-MS/MS.

Authentic human tau filaments were purified from autopsy samples of AD parietal cortex using a modification of two procedures [209, 210]. Briefly, after cutting away and discarding the white matter, the gray matter was minced and homogenized with a blender 93 in 5 volumes of a buffer containing 20 mM MES/NaOH, pH 6.8, 80 mM NaCl, 1 mM

MgCl2, 2 mM EGTA, 0.1 mM EDTA and 1 mM PMSF. The mixture was further homogenized in a Teflon/glass homogenizer with a Fisher Scientific model LR400C stirrer and centrifuged at 27,000 x g for 20 min at 4°C. After discarding the supernatant, the pellet was resuspended in 10 volumes of 0.8 M NaCl, 10% sucrose, 10 mM MES/NaOH, pH 7.4,

1 mM EGTA and 0.1 mM PMSF, and rehomogenized and centrifuged as above. After discarding the pellet, the supernatant was incubated with 1% N-laurylsarcosine (w/v) for 1 h at room temperature on an orbital shaker. The mixture was then centrifuged at 87,000 x g for 35 min at 4°C and the supernatant was discarded. The pellet was resuspended in 10 mM MES/NaOH, pH 7.0 (0.2 mL/g tissue) and added (9.25 mL/tube) to 40 mL, thin-walled polyallomer tubes containing a discontinuous sucrose gradient constructed from 9.25 mL layers of 1 M, 1.5 M, and 2 M sucrose in 10 mM MES/NaOH pH 7.0. The tubes were centrifuged at 112,600 x g for 2 h and 25 min at 4°C and the layers were removed from the top and examined by TEM. Layers 4 and 5, which contained the majority of tau filaments, were then stored in aliquots at -80°C until needed.

Solubilization of human tissue derived PHF.

To solubilize PHF for further analysis, layers 4 and 5 extracted by sucrose gradient centrifugation where incubated at 37C in 2M guanidinium thiocyanate (GndSCN) for 1h and centrifuged at 100,000g for 40 min. The supernatant (S/N) was removed and injected into dialysis tubing (Spectra 25225 2mL/cm) and dialyzed for 30 hours in 25mM

Ammonium Acetate pH 7.4 at 4C changing dialysis buffer 4X. The dialyzed samples were centrifuged at100,000g for 40 min at 4C and remove S/N to SpeedVac dry. Resuspend in water for further protein isolation and analysis. 94

SDS PAGE and immunoblots. Tau proteins were subjected to SDS-PAGE on 8% acrylamide gels, then visualized through three approaches. First, for direct detection of tau protein, gels were silver stained. Second, for detection of tau immunoreactivity, gel contents were transferred (100 V for 1 h at 4C) to PVDF membranes in Transfer Buffer

A (25 mM Tris, 0.2 mM glycine, 10% methanol). The resulting membranes were incubated with blocking buffer (4% non-fat dry milk, 10mM Tris, 150 mM NaCl, and 0.05% Tween-

20) for 1 hour at room temperature before incubation with Tau5 primary antibody at a

1:5000 dilution in blocking buffer for 2 hours at room temperature. Membranes were then incubated with goat-anti-mouse secondary antibody conjugated with HRP diluted 1:5000 in blocking buffer for 1 hour at room temperature. Both antibodies were purchased from

Thermo Fisher Scientific. Secondary antibody was detected using the Amersham ECl

Western Blotting Detection kit from GE Healthcare Life Sciences. Finally, for LC/MS-MS analysis, PVDF transfers were then stained (50% Methanol, 7% Acetic Acid, 0.1%

Coomassie Brilliant Blue) for 30 min at room temperature with agitation, and then destained (50% methanol, 7% acetic acid) for ~10 min at room temperature with agitation until the staining pattern was visible. After membranes dried at room temperature,

Coomassie-blue stained bands were excised using a razor blade (2 x 6 mm slices) and stored at -20°C until LC/MS-MS analysis.

LC-MS/MS quantification. Excised PVDF membranes containing tau protein were hydrolyzed in 6 M HCl under nitrogen as described in Chapter 4. Dried hydrolysates were resuspended and subjected to LC-MS/MS under conditions described in Chapter 4.

Analytical methods. The relative proportion ( P ) of all peptidyl-Lys residues in the form K X of 1meK, 2meK or 3meK (XmeK) was calculated ratiometrically by the equation:

XmeK P = K X 3 (1) Lys +  XmeK X =1

3 SK = N XP (2)  K X X =1

Similarly, the relative proportion ( P ) of Arg methylation was calculated ratiometrically RX from mol amounts of all methyl-Arg species (XmeR) by the equation:

XmeR P = RX 2 (3) Arg +  XmeR X =1

2 SR = N XP (4)  R X X =1

Results and Discussion

Preparation of human tau proteins. To prepare material for pilot analysis, tau was isolated from the brains of four cognitively normal humans aged 55 ± 1 yrs (Table 5).

These cases were chosen to establish baseline levels of tau methylation in normal, middle aged individuals. In addition to meeting behavioral criteria, these cases were pathologically normal with respect to plaque and tangle lesion counts. A second cohort of four cognitively normal humans aged 86 ± 2 yrs also was investigated (Table 5). These cases, which also met behavioral and pathological criteria to serve as normal controls, were selected to capture the variable of aging. Finally, a third cohort of three AD cases aged 77 ± 13 yrs was investigated. This population was used to capture the effect of disease state and aggregation on tau methylation stoichiometry. Together, the cohorts spanned three neocortical regions associated with AD vulnerability and were derived from three independent brain banks.

Soluble tau was isolated from the 55- and 86-yr old cohorts as described in

Materials and Methods. These preparations were subjected to SDS-PAGE and immunoblot analysis with the Tau5 primary antibody under conditions where all six human isoforms could be resolved (Fig. 25A,C). The tau isoform expression pattern approximated the reported ratios of 1:1 for 4R to 3R tau isoforms, and 9:54:37 for 2N:1N:0N splice variants

[80]. On the basis of silver staining, the tau immunoreactive species represented the major proteins in the preparation (Fig. 25B,D). In contrast, PHF tau immunoreactivity prepared from the AD cases migrated more slowly on SDS-PAGE (Fig. 25E), consistent with established behavior [80]. Silver staining confirmed that the tau-immunoreactive species were the principal proteins present in the preparation (Fig. 25F). Together these data show 97 that brain-derived tau proteins were prepared at levels of purity required for analysis by targeted metabolomic methods.

To quantify levels of methylated amino acids, samples were subjected to SDS-

PAGE, blotted onto PDF, and excised as mixtures of all six tau isoforms. PVDF filters were then acid hydrolyzed and assayed by LC-MS/MS as described in Chapter 4.

Stoichiometries were calculated from equations 2 and 4 assuming the number of peptidyl

Lys residues represented the canonical ratios of six tau splices variants, such that N = 40.2 for Lys and N = 14 for Arg. Final bulk stoichiometries are shown in Fig. 26. Values ranged from 0.1 – 0.2 mol methyl-Lys/mol tau and 0.04 – 0.1 mol methyl-Arg/mol tau with no statistical differences among cohorts (p < 0.05). These data indicated that methylation is a minor physiological modification of tau protein in both normal and disease cases.

A rTau 1 2 3 4 B rTau 1 2 3 4 2N4R 2N3R 1N4R 1N3R 0N4R 0N3R

C rTau 5 6 7 8 D rTau 5 6 7 8 2N4R 2N3R 1N4R 1N3R 0N4R 0N3R

E rTau 9 10 11 F rTau 9 10 11

2N4R 2N3R 1N4R 1N3R 0N4R 0N3R

Fig. 25. Western blots and silver stain analysis of human brain tissue samples. Lane numbers correspond with sample demographics in Table 5. (A, C, E) Western blots isolate solubilized paired-helical filament (PHF) material for acid hydrolysis and LC-MS/MS 98 analysis. (B, D, F) Silver stain analysis of PHF in human brain tissue samples. Sample description in Table 5.

Together these results reveal that while detectable by LC/MS-MS methods, the levels of methyl-Lys do not rise to the levels we found associated with depression of tau aggregation propensity. I conclude that under normal physiological conditions, tau methylation is probably not contributing to its proteostasis. Nonetheless, artificial elevation of tau methylation levels could have beneficial effects. This may be done through pharmacological inhibition of tau demethylases, for example. If demethylases had sufficient selectivity for tau protein, they may serve as targets of drug discovery.

meK

meR

Age (years)

Fig. 26. Bulk methyl-Lys (meK; solid symbols) and methyl-Arg (meR; hollow symbols) stoichiometries of normal (blue) and AD (red) human tau preparations. Tau methylation stoichiometries were uniformly low across all samples.

The identification of tau methyltransferases and demethylases requires a model system capable of recapitulating the pattern seen in human brain. Therefore, I tested whether HEK293 cells could serve this purpose. Normal HEK293 cells are devoid of tau expression, so a tau-stable cell line previously prepared in this laboratory [211]. This cell 100 line expresses high levels of 2N4R tau, which simplifies analysis. Application of the purification methods outline in the Materials and Methods section to these cells yielded a highly purified preparation of 2N4R tau protein (Fig. 27A). When subjected to targeted metabolic analysis, HEK293 tau was found to be methylated on both Lys and Arg residues at similar stoichiometries as observed in human brain tissue (Fig. 27B). These data suggest that the HEK293 system offers a route toward identifying tau methyltransferases and demethylases, and for determining whether tau demethylases are potential targets for drug discovery efforts.

101

Table 5. Case demographics

Case Diag Age Sex Racea PMIb Sourcec Identifier Neocortical Commentsd (#) (yr) (h) Region HASCVD; History of Diabetes, 1 Normal 55 M AA 13 NICHD 5172 Temporal cocaine abuse & HBP Acute Bronchopneumonia/ASCVD; 2 Normal 57 M C 5 NICHD 5117 Temporal Smoker, history of alcohol abuse, HASCVD; PTSS, No history of drug, 3 Normal 54 M AA 10 NICHD 4923 Temporal alcohol abuse 4 Normal 54 M C --- OSU Parietal 5 Normal 88 F C 5.2 UCI 15-02 Frontal NFT/Aβ stage 2A; ApoE 3/3 6 Normal 86 M C 2.9 UCI 24-11 Frontal NFT/Aβ stage 3A; ApoE 3/3 7 Normal 86 M C 4.4 UCI 11-12 Frontal NFT/Aβ stage 2-; ApoE 3/3 8 Normal 83 M C 1.8 UCI 25-01 Frontal NFT/Aβ stage 2-; ApoE 3/3 9 AD 88 M C --- OSU 394 Parietal 10 AD 81 F C --- OSU 2571 Parietal 11 AD 63 M C --- OSU 212 Parietal aAA, African American; C, Caucasian. b Post-mortem interval. c NICHD, NICHD Brain and Tissue Bank for Developmental Disorders; UCI, University of California, Irvine dASCVD, atherosclerotic cardiovascular disease; HASCVD, hypertensive arteriosclerotic cardiovascular disease; PTSS, Post Traumatic Stress Syndrome.

102

Fig. 27. Methylation of tau protein in HEK293 cells. Tau from a 2N4R tau-stable cell line was isolated and subjected to targeted metabolomics to quantify methylation stoichiometries. (A) Coomassie-blue stained 2N4R tau after purification from cells. (B) methylation stoichiometries for all analytes estimated by targeted metabolomics. Tau is methylated in HEK293 cells.

103

Chapter 6. Toward a more robust gene expression signature for Alzheimer’s Disease

Reproduced in part from work completed in collaboration with C. Wagner, and J.

Kuret

J. Kuret and C.J. Huseby conceived the research. C.J. Huseby, and C.W. performed the computational bioinformatics research. This work was supported by the National

Institutes of Health grant AG05418 to J.K.

104

Introduction

Alzheimer’s disease is a neurological disorder in which vulnerable neuron populations in the central nervous system are lost. It is defined by the appearance and accumulation of proteinaceous lesions both exterior and interior to neurons (amyloid plaques and neurofibrillary tangles, NFT) in vulnerable regions of the brain. The damage is associated with irreversible and progressive neurodegeneration and cognitive impairment causing significant burden on individuals, their families, and society. The pathological progression is resultant from complex genetics interacting with unknown environmental factors. The molecular networks that mediate their formation and associated changes in neuronal and glial function remain unsettled. As discussed previously, the failure of large-scale clinical trials for drug therapies is partly due to the lack of success in translating preclinical therapies to humans and leaves clinicians without effective treatment options for AD. Additionally, the lack of our understanding of AD pathophysiology is a major obstacle to the development of targeted therapies.

Differential expression (DE) analysis is widely used to identify genes that show different expression levels across different conditions [212]. The motivation being that the differentially expressed genes may have roles in disease phenotypes and hence studying these genes may reveal underlying biological mechanisms. The DE analysis can identify dysregulated genes that may be candidates for drug treatment or point to environmental toxins. Combined with a systems biology approach the DE can elucidate coordinated molecular processes underlying the pathophysiology of a complex disease. By examining the changes in a genome-wide gene expression during the course of AD from human brain, gene regulatory network signatures can be assembled to improve our understanding of the 105 molecular mechanisms involved in AD pathogenesis, provide new opportunities in the diagnosis, early detection, and tracking of this disorder, and provide novel targets for the discovery of interventions to treat and prevent this disorder.

One approach to identifying these pathways involves analysis of gene expression in disease dominated by sporadic onset (late-onset AD; LOAD) relative to non-demented control cases (Eq. 1) for i genes and j samples where BA9 is Brodmann Area 9 in prefrontal cortex, AD is Alzheimer’s disease affected, and NA is non-affected.

퐴퐷 푁퐴 푙표푔퐹퐶 = 〈푙표푔(퐵퐴9푖푗 )〉푖 − 〈푙표푔(퐵퐴9푖푗 )〉푖 Eq 1.

But AD progresses in a stereotypical fashion that involves some brain regions while sparing others. The pattern suggests cell-to-cell spread of a pathogen combined with selective vulnerability (Fig 4). Similarly, with the discovery that misfolded protein components have prion-like activity offers a potential mechanism for AD progression and spread, which proceeds stereotypically among specific brain regions [60]. The spread of AD lesions into neocortical areas is consistent with misfolded proteins being transmitted through connectivity networks, thereby templating new aggregate formation. (Fig. 4). In addition to marking brain regions undergoing neurodegeneration, tau misfolding and aggregation has been proposed to contribute directly to the process of neurodegeneration [50, 59]. The brain regions undergoing protein misfolding at a given time is used clinically to deduce disease “stage” (Fig. 5). Certain brain regions, such as cerebellum (CR), are invulnerable to AD pathogenesis despite being in an affected connectivity network and being exposed to misfolded protein aggregates throughout the course of disease (Fig 4) [58]. Investigation of selective vulnerability may provide novel insight into the AD phenotype.

106

Specifically, we propose that the expression phenotype of invulnerable brain regions can reveal molecular signatures associated with AD neuroprotection. In so doing, it provides information orthogonal to current practice, which focuses exclusively on vulnerable brain regions. By normalizing first to the invulnerable region, CR, the DE can now be described as containing in-animal control features (Eq. 2).

퐴퐷 푁퐴 퐵퐴9푖푗 퐵퐴9푖푗 푙표푔퐹퐶 = 〈푙표푔 ( 퐴퐷 )〉푖 − 〈푙표푔 ( 푁퐴 )〉푖 Eq 2. 퐶푅푖푗 퐶푅푖푗

We hypothesize that the in-animal DE approach is refined by first analyzing expression data on the basis of constituent regions of interest, and there by better controlling for the variables of age, sex, post-mortem interval (pmi), and RNA integrity (Fig 28). To test this hypothesis, large multi-tissue gene expression data sets of autopsied brain tissue from the

NCBI GEO database are available for analysis which leverage most recent datasets that are large (hundreds of cases), that include affected (pre-frontal cortex), somewhat spared

(Brodmann area 17), completely spared (cerebellum) regions of interest, and that contain transcripts from all cell types.

107

Figure 28. Cohort distributions NCBI GEO database accession number GSE44772 for post-mortem interval (PMI) vs age left panel and sex cohort distributions right panel.

Methods

RNA expression data was obtained from the Gene Expression Omnibus (GEO) database

GEO accession number GSE44772 (raw data sets GSE44770, GSE44771, and GSE44768), containing data from non-affected and late-onset Alzheimer’s disease (LOAD) human autopsy tissue from dorsolateral prefrontal cortex (PFC [BA9]), visual cortex (VC

[BA17]), and cerebellum (CR) compiled using Rosetta/Merck Human 44k 1.1 microarray

[213]. Starting with the raw datasets and after normalization, differential expression statistical methods were coded and calculated using MATLAB to make comparisons between AD affected and non-affected groups, for the regions of interest, and yielded a list of statistically reliable (FDR corrected p-value < 0.01) genes associated with AD, ranked by their log-fold change (Eq. 1) (Figure 29A) [214]. Additionally, the novel ratio-of-ratios method was treated similarly (Eq. 2) yielding similar results for the comparisons of AD

108 affected and non-affected groups including the in-animal control feature and ranked by their log fold-change (Figure 29B). A covariate analysis was made on the datasets with and without the ratio-of-ratios method by ordering each dataset by covariate and making a linear fit on each gene. The slopes of the fits, or trends attributable to covariate manipulations, were averaged for comparison (Figure 30A,B).

109

Figure 29. Volcano plots of adjusted p-value vs log fold-change of expression in AD. (A)

The standard method of comparing expression changes in a vulnerable region between AD affected tissue and non-affected tissue. (B) The ratio-of-ratios feature first compares tissue

110 to an invulnerable region in brain, making an in-animal control for global covariates.

Transcripts meeting statistical significance (FDR corrected for multiple testing, p-value <

0.01) log fold-change between non-affected subjects and AD affected cohort appear as blue ( >1.5 fold-change) and red markers( >2 fold-change). Transcripts are re-ranked in a ratio-of-ratios method in favor of transcript changes directly associated with the disease process.

Results

The results yielded a novel method for prioritizing genes associated with AD neurodegeneration. Those genes that experience global changes (due to inflammation for example) are deprioritized while genetic changes associated with neurodegeneration and neuroprotection are prioritized (Figure 29A,B). By considering the transcript changes in the invulnerable region, cerebellum, in parallel to the transcript changes in the vulnerable region, prefrontal cortex, the genes are re-prioritized based on the global expression changes happening. This results in a reduction of transcripts that meet statistical significance. This method also captures changes seen in AD vs. control comparisons. The results identified highly robust down-regulation of transcripts associated with cortical interneurons, and up-regulation of regulatory networks. Both up- and down-regulated transcripts varied with region of interest, indicating the existence of “local” and “global” changes associated with disease.

111

Figure 30. A measurement of trends in data due to differences in age, post-mortem interval

(pmi), sex imbalances in cohorts, and disease show that data subjected to a method with in-animal features, improves the variations attributable to covariates. (A) A ratio-of-ratio method with in-animal control features has a slight decrease in each trend measurement created by the covariates age, pmi, and sex compared to (B) standard differential analysis method.

Aβ protease changes

Novel results have yielded an unmasking of two Aβ proteases, MME and MMP3, as implicated in the molecular signature for Alzheimer’s disease. When first controlling for co-variants using a normalization to the invulnerable region, CR, thus removing changes associated with brain-wide inflammation, these Aβ proteases are found to be dramatically down in expression in AD. Both genes are implicated in clearance of Aβ in brain, a system hypothesized to be dysregulated in disease [215, 216]. MME is a type II metalloendopeptidase transmembrane protein considered an important Aβ degrading enzyme. MMP3 is a matrix metalloproteinase endopeptidase which is thought to be

112 important in AD neurodegeneration by activating other MMPs directly involved in Aβ degradation (Figure 31).

0.3 Abeta proteases

0.2

0.1

0.0 log2FC

-0.1

-0.2

-0.3

Figure 31. Aβ proteases. The expression changes in AD using a standard expression analysis blue bars. A ratio/ratio method of comparing transcript changes in AD when considering the vulnerable region PFC (BA9) to the invulnerable region CR gray bars.

Two proteases, MME and MMP3) are exposed as transcript changes due to disease processes rather than brain-wide inflammation.

Protein methylation processes

It is unclear which enzymes are responsible for catalyzing tau methylation and demethylation. The first histone lysine methyltransferase and demethylase were identified in 2000 and 2004 respectively [217, 218]. There are now 208 putative human

113 methyltransferases identified of which 57 of them contain SET domains associated with protein lysine methyltransferase activity [219]. Nuclear enzymes owing to histone modifications have been studied most aggressively, however, lysine methyltransferases have been found in the cytoplasm [220]. It remains to be seen which methyltransferases will be identified as active on tau protein.

Consistent with the previous chapter, quantifying methylation of tau protein in AD, a systems biology strategy was taken to examine the expression changes in AD for methyltransferases and demethylases in human brain. There are a variety of methyltransferases for adding mono-, di-, or tri- methylation moieties to DNA, RNA or protein in human cells (Fig. 32) [221, 222]. Much of the methylation is directed to histones which regulate DNA transcription but tau protein is also found to be methylated at Lysine and Arginine [81]. Applying our new approach, we find no changes in AD for demethylase transcripts, but for the methyltransferases, two transcripts in the SMYD family of histone methylators, one NSD type histone methyltransferase transcript, a PRDM family member, and three PRMT transcripts in the arginine methyltransferase family appear to decrease expression in AD. PRMT8 is protein arginine methyltransferase but also a phospholipase that locates to neuron synapses [223]. Only one methyltransferase of the PRDM family transcript appear to be up in AD - PRDM6.

114

Figure 32. (A) Methyltransferases and (B) de-methylases expression changes (PFC) in AD using the ratio/ratio method of comparison. Green no change in AD. Blue down. Red up.

No information is available for genes in white.

Comparison of data from spared regions of interest with those regions affected by the defining pathology within AD cases (Eq. 2) will better resolve transcripts associated with active disease from confounds related to age, sex, PMI, brain-wide neuroinflammation, and disease stage. Removal of systemic variation can expose and highlight gene expression changes associated with the molecular signature for Alzheimer’s disease focuses comparisons on disease severity and identified transcriptional changes potentially associated with active degeneration as well as active neuroprotection.

Results show consistency with known genetic factors for AD. We conclude that data analysis can help identify expression changes associated with locally-active neurodegeneration as well as neuroprotection thus unmasking changes in expression of 115 transcripts related to Aβ degradation. Proteases involved in the clearance of Aβ plaques are found to be down-regulated in AD.

The expression signature for the methyltransferases are found to be consistent with the previous work quantifying the differences in methylation of tau protein in non-affected vs AD affected human brain tissue (see previous chapter) (Fig. 32). The transcript expression for enzymes involved in methylation of tau protein do not seem to be changed in AD but a few of the histone methyltransferases are found to be down probably affecting transcription regulation at DNA. However, the one histone methyltransferase (PRDM6) that acts as a transcriptional repressor of smooth muscle gene expression and repressor of endothelial cell proliferation, survival and differentiation, appears to be up in AD.

In the staging of AD, the visual cortex is one of the last regions to be affected perhaps because it is positioned at the end of the vulnerable neuron pathway (Fig. 5).

Additional analysis on this NCBI GEO data set GSE44772 using PFC (BA9) vs VC

(BA17) comparisons can potentially identify transcriptional changes associated with active degeneration. We conclude that data analysis can help identify expression changes associated with locally-active neurodegeneration.

116

Chapter 7. Conclusions and perspectives

Conversion of monomeric tau protein into filamentous aggregates is a defining event in the pathogenesis of Alzheimer’s disease. To gain insight into disease pathogenesis, the mechanisms that trigger and mediate tau aggregation are under intense investigation.

Here I first describe protocols for analyzing tau filament population by TEM. Transmission electron microscopy is a static imaging tool that can detect individual tau filaments at nanometer resolution. In doing so, I provide unique insight into the quality, quantity, and composition of synthetic tau filament populations for purposes of dissecting aggregation mechanism. With this method I compiled data sets of in vitro tau filament aggregation for additional mathematical simulation analysis.

The conversion of monomeric tau protein into prion-like filamentous aggregates in

AD is at least partially dependent on aggregate size and the processes that underlie filament formation and size distribution are of special interest. However, the major processes controlling tau aggregate formation and size distribution have been incompletely characterized. I used a computational approach and in vitro data sets for the investigation of human 2N4R tau fibrillation dynamics. I report that tau filaments engage in a previously uncharacterized secondary process involving end-to-end annealing, and that rationalization of empirical aggregation data composed of total protomer concentrations and fibril length distributions requires inclusion of end-to-end annealing that acts to increase filament lengths in opposition to filament fragmentation. Annealing of 2N4R tau filaments is robust, 117 with an intrinsic association rate constant of similar magnitude to that mediating monomer addition. In contrast, secondary nucleation on the surface of tau filaments makes no detectable contribution to tau aggregation dynamics. My data demonstrates that tau filament ends engage in a range of homotypic interactions involving monomers, oligomers and filaments. They further indicate that in the case of tau protein, annealing and fragmentation along with primary nucleation and elongation are the major processes controlling filament size distribution. This finding links equilibria at filament ends to control of tau filament length-distribution and identifies interactions that may influence the toxicity and propagation of tau pathology.

Here I validated a mathematical model of tau aggregation as modeled in vitro. With this model, simulation trials can be formulated to investigate therapeutic interventions at particular points in the tau aggregation pathway (Fig. 11). This is a future direction planned to aid in assessing and designing tau aggregation inhibitors. By probing the parameters for the individual aggregation processes, the output variation can be used to direct the development of small molecule interventions which will generate the most favorable output for a potential tau protein aggregation inhibitor drug.

Additionally, previous work in the lab identified post-translational sites on the tau protein and showed that the methylation of tau protein can alter tau aggregation kinetics as well as microtubule assembly. Post-translational modifications are biologically important and wide-spread modulators of protein function. Although methods for detecting the presence of specific modifications are becoming established, approaches for quantifying their mol modification/mol protein stoichiometry are less well developed. With a bit of research and inquiry I found a solution to quantify the tau methylation stoichiometry in 118 human brain tissue using an established facility directly above the lab. Here I introduced a ratiometric, label-free, targeted liquid chromatography tandem mass spectroscopy-based method for estimating Lys and Arg methylation stoichiometry on post-translationally modified proteins such as tau protein. Methylated amino acids were detected with limits of quantification at low fmol and with linearity extending from 20 – 5000 fmol. This level of sensitivity allowed estimation of methylation stoichiometry from microgram quantities of various proteins, including those derived from either recombinant or tissue sources. The method also disaggregated total methylation stoichiometry into its elementary mono-, di-, and tri-methylated residue components. In addition to being compatible with kinetic experiments of protein methylation, the approach will be especially useful for characterizing methylation states of proteins isolated from cells and tissues.

We validated the utility and accuracy of the method by replicating stoichiometries of proteins with established mol/mol quantities for both mono-, di-, tri- Lys and mono- and di- Arg. Our analysis of human brain tissue revealed the stoichiometry of methylated tau protein is low and remains constant with age and/or disease status.

Lastly, I used differential expression (DE) analysis to identify dysregulated genes in AD brain that may be candidates for drug treatment or direct us toward toxic environmental inputs. I reported here a new method of DE in which I first normalize the transcript expressions to an invulnerable region of brain in AD, cerebellum, thereby reducing the influences of biases such as age or sex. In examining the changes in a genome- wide gene expression during the course of AD from human brain and combining it with systems biology gene networks, I report novel expression changes in AD such as Aβ protease MME and verified the unchanged transcript expression for methyltransferases and 119 demethylators. This information can be assembled to improve our understanding of the molecular mechanisms involved in AD pathogenesis, provide new opportunities in the diagnosis, early detection, and tracking of this disorder, and provide novel targets for the discovery of interventions to treat and prevent this disorder.

It remains to be seen if this new approach of differential expression analysis method of ratio-of-ratios is more robust for adjusting unwanted variables as compared to a standard linear regression model adjustment. Further analysis is required to justify the usefulness of this method or an optimized variation of it.

We learn more about AD every day and we may in fact see a pharmacological treatment sometime in the future. I hope that my research here has provided a small step in that direction but in fact it is apparent that it is a multidimensional disease with at its core glucose and lipid dysregulation, the hallmark pathological markers of plaques and tangles, inflammation, oxidation, and neurodegeneration. Beyond the identification of the mutations in three genes causing AD, there are a variety of hypothesis leading to the onset of AD including nutrient deficiencies, environmental toxicities, sleep apnea, hypertension, diabetes, traumatic brain injury, and genetic variant risk factors. Perhaps in the end, AD will require a personalized medical treatment customized for each individual’s genetic and molecular makeup.

120

References

1. Tchalla, A.E., et al., Predictors of Rapid Cognitive Decline in Patients with Mild-to- Moderate Alzheimer Disease: A Prospective Cohort Study with 12-Month Follow-Up Performed in Memory Clinics. Dement Geriatr Cogn Disord, 2018. 45(1-2): p. 56-65. 2. McKhann, G.M., et al., The diagnosis of dementia due to Alzheimer's disease: Recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers & Dementia, 2011. 7(3): p. 263-269. 3. Association, A.s., 2018 Alzheimer's disease facts and figures. Alzheimers & Dementia, 2018. 14(3): p. 367-425. 4. Perissinotto, C.M. and V. Valcour, Preventing dementia: is there hope for progress? Am Fam Physician, 2011. 83(9): p. 1034-5. 5. Schneider, J.A., et al., Mixed brain pathologies account for most dementia cases in community-dwelling older persons. Neurology, 2007. 69(24): p. 2197-204. 6. James, B.D., et al., Dementia from Alzheimer disease and mixed pathologies in the oldest old. JAMA, 2012. 307(17): p. 1798-800. 7. Barnes, L.L., et al., Mixed pathology is more likely in black than white decedents with Alzheimer dementia. Neurology, 2015. 85(6): p. 528-34. 8. Bick, K.L., et al., The early story of Alzheimer's disease : translation of the historical papers by Alois Alzheimer, Oskar Fischer, Francesco Bonfiglio, Emil Kraepelin, Gaetano Perusini. 1987, Padova New York: Liviana Press ; distributed internationally by Raven Press. xi, 147 p., 14 p. of plates. 9. Schneider, J.A., et al., The neuropathology of probable Alzheimer disease and mild cognitive impairment. Ann Neurol, 2009. 66(2): p. 200-8. 10. Sherzai, D. and A. Sherzai, The Alzheimer's solution : a breakthrough program to prevent and reverse the symptoms of cognitive decline at every age. First edition. ed. 352 pages. 11. Groot, C., et al., The effect of physical activity on cognitive function in patients with dementia: A meta-analysis of randomized control trials. Ageing Res Rev, 2016. 25: p. 13- 23. 12. Williams, J., Genetics is more important than ever in directing our understanding of Alzheimer's Disease. Alzheimer's and Dementia, The Journal of the Alzheimer's Association, 2018. 14(7): p. P605. 13. Bertram, L. and R.E. Tanzi, The genetic epidemiology of neurodegenerative disease. J Clin Invest, 2005. 115(6): p. 1449-57. 14. Lambert, J.C., et al., Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer's disease. Nat Genet, 2013. 45(12): p. 1452-8. 15. Guzman-Velez, E., et al., A Three-Factor Structure of Cognitive Functioning Among Unimpaired Carriers and Non-Carriers of Autosomal-Dominant Alzheimer's Disease. J Alzheimers Dis, 2018. 65(1): p. 107-115. 16. Tedde, A., et al., Identification of new presenilin gene mutations in early-onset familial Alzheimer disease. Arch Neurol, 2003. 60(11): p. 1541-4. 121

17. Pimplikar, S.W., et al., Amyloid-independent mechanisms in Alzheimer's disease pathogenesis. J Neurosci, 2010. 30(45): p. 14946-54. 18. Head, E., et al., Aging in Down Syndrome and the Development of Alzheimer's Disease Neuropathology. Curr Alzheimer Res, 2016. 13(1): p. 18-29. 19. Corder, E.H., et al., Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer's disease in late onset families. Science, 1993. 261(5123): p. 921-3. 20. Nathan, B.P., et al., Differential effects of apolipoproteins E3 and E4 on neuronal growth in vitro. Science, 1994. 264(5160): p. 850-2. 21. Poirier, J., Apolipoprotein E and cholesterol metabolism in the pathogenesis and treatment of Alzheimer's disease. Trends Mol Med, 2003. 9(3): p. 94-101. 22. Strittmatter, W.J., et al., Hypothesis: microtubule instability and paired helical filament formation in the Alzheimer disease brain are related to apolipoprotein E genotype. Exp Neurol, 1994. 125(2): p. 163-71; discussion 172-4. 23. Strittmatter, W.J., et al., Binding of human apolipoprotein E to synthetic amyloid beta peptide: isoform-specific effects and implications for late-onset Alzheimer disease. Proc Natl Acad Sci U S A, 1993. 90(17): p. 8098-102. 24. Hebert, L.E., et al., Alzheimer disease in the United States (2010-2050) estimated using the 2010 census. Neurology, 2013. 80(19): p. 1778-83. 25. Jack, C.R., Jr., et al., Introduction to the recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement, 2011. 7(3): p. 257-62. 26. Beach, T.G., et al., Theoretical impact of Florbetapir (18F) amyloid imaging on diagnosis of alzheimer dementia and detection of preclinical cortical amyloid. J Neuropathol Exp Neurol, 2014. 73(10): p. 948-53. 27. Wolf, D.S., et al., Progression of regional neuropathology in Alzheimer disease and normal elderly: findings from the Nun study. Alzheimer Dis Assoc Disord, 1999. 13(4): p. 226-31. 28. Sperling, R.A., et al., Toward defining the preclinical stages of Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement, 2011. 7(3): p. 280-92. 29. Bennett, D.A., et al., Neuropathology of older persons without cognitive impairment from two community-based studies. Neurology, 2006. 66(12): p. 1837-44. 30. Wolff, J.L., et al., A National Profile of Family and Unpaid Caregivers Who Assist Older Adults With Health Care Activities. JAMA Intern Med, 2016. 176(3): p. 372-9. 31. Berchtold, N.C. and C.W. Cotman, Evolution in the conceptualization of dementia and Alzheimer's disease: Greco-Roman period to the 1960s. Neurobiol Aging, 1998. 19(3): p. 173-89. 32. Maurer, K. and U. Maurer, Alzheimer : the life of a physician and the career of a disease. 2003, New York: Columbia University Press. x, 270 p. 33. Goedert, M., Oskar Fischer and the study of dementia. Brain, 2009. 132(Pt 4): p. 1102- 11. 34. Savvaidou, N.E. and L.C. Triarhou, Sesquicentennial Tribute to Emil Redlich (1866-1930), The 'Embodied Conscience of Neurology'. Eur Neurol, 2016. 76(5-6): p. 267-277. 35. Lock, M.M., The Alzheimer conundrum : entanglements of dementia and aging. x, 310 pages.

122

36. Kraepelin, E., Psychiatrie; ein Lehrbuch f©*r Studierende und ©*rzte. 8., vollst©Þndig umgearb. Aufl. ed. 1909, Leipzig: Barth. 37. Kapsambelis, N., The inheritance : a family on the front lines of the battle against Alzheimer's disease. First Simon & Schuster hardcover edition. ed. xviii, 344 pages. 38. Fischer, O., Miliare Necrosis with Glandular Vegatations of Neurofibres, a Regular Change of the Cortex in Senile Dementia. Monatsschrift Fur Psychiatrie Und Neurologie, 1907. 22(4): p. 361-372. 39. Glenner, G.G. and C.W. Wong, Alzheimer's disease: initial report of the purification and characterization of a novel cerebrovascular amyloid protein. Biochem Biophys Res Commun, 1984. 120(3): p. 885-90. 40. Wild-Bode, C., et al., Intracellular generation and accumulation of amyloid beta-peptide terminating at amino acid 42. J Biol Chem, 1997. 272(26): p. 16085-8. 41. Kang, J., et al., The precursor of Alzheimer's disease amyloid A4 protein resembles a cell- surface receptor. Nature, 1987. 325(6106): p. 733-6. 42. Robakis, N.K., et al., Chromosome 21q21 sublocalisation of gene encoding beta-amyloid peptide in cerebral vessels and neuritic (senile) plaques of people with Alzheimer disease and Down syndrome. Lancet, 1987. 1(8529): p. 384-5. 43. Goldgaber, D., et al., Isolation, characterization, and chromosomal localization of human brain cDNA clones coding for the precursor of the amyloid of brain in Alzheimer's disease, Down's syndrome and aging. J Neural Transm Suppl, 1987. 24: p. 23-8. 44. Kumar, D.K.V., et al., Alzheimer's disease: the potential therapeutic role of the natural antibiotic amyloid-beta peptide. Neurodegenerative Disease Management, 2016. 6(5): p. 345-348. 45. Kidd, M., Paired helical filaments in electron microscopy of Alzheimer's disease. Nature, 1963. 197: p. 192-3. 46. Terry, R.D., The Fine Structure of Neurofibrillary Tangles in Alzheimer's Disease. J Neuropathol Exp Neurol, 1963. 22: p. 629-42. 47. Brion, J.P., et al., Immunological Detection of Tau Protein in Neurofibrillary Tangles of Alzheimers-Disease. Archives of Biology, 1985. 96(2): p. 229-235. 48. Honson, N.S. and J. Kuret, Tau aggregation and toxicity in tauopathic neurodegenerative diseases. J Alzheimers Dis, 2008. 14(4): p. 417-22. 49. Jack, C.R., Jr., et al., Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade. Lancet Neurol, 2010. 9(1): p. 119-28. 50. Bejanin, A., et al., Tau pathology and neurodegeneration contribute to cognitive impairment in Alzheimer's disease. Brain, 2017. 140(12): p. 3286-3300. 51. Hardy, J.A. and G.A. Higgins, Alzheimer's disease: the amyloid cascade hypothesis. Science, 1992. 256(5054): p. 184-5. 52. Piton, M., et al., Alzheimer's Disease: Advances in Drug Development. J Alzheimers Dis, 2018. 65(1): p. 3-13. 53. Cummings, J.L., T. Morstorf, and K. Zhong, Alzheimer's disease drug-development pipeline: few candidates, frequent failures. Alzheimers Research & Therapy, 2014. 6(4). 54. Koychev, I., et.al., PET Tau and Amyloid-beta Burden in Mild Alzheimer's Disease: Divergent Relationship with Age, Cognition, and Cerebrospinal Fluid Biomarkers. J Alzheimers Dis, 2018. 63(1): p. 407. 55. Gomez-Isla, T., et al., Neuronal loss correlates with but exceeds neurofibrillary tangles in Alzheimer's disease. Ann Neurol, 1997. 41(1): p. 17-24.

123

56. Gomez-Isla, T., et al., Profound loss of layer II entorhinal cortex neurons occurs in very mild Alzheimer's disease. J Neurosci, 1996. 16(14): p. 4491-500. 57. Braak, H. and E. Braak, Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol, 1991. 82(4): p. 239-59. 58. Braak, H. and K. Del Tredici, Alzheimer's pathogenesis: is there neuron-to-neuron propagation? Acta Neuropathol, 2011. 121(5): p. 589-95. 59. Mudher, A., et al., What is the evidence that tau pathology spreads through prion-like propagation? Acta Neuropathol Commun, 2017. 5(1): p. 99. 60. Holmes, B.B. and M.I. Diamond, Prion-like Properties of tau protein: The importance of extracellular tau as a therapeutic target. J Biol Chem, 2014. 61. Morris, M., et al., The many faces of tau. Neuron, 2011. 70(3): p. 410-26. 62. Mirbaha, H., et al., Tau Trimers Are the Minimal Propagation Unit Spontaneously Internalized to Seed Intracellular Aggregation. Journal of Biological Chemistry, 2015. 290(24): p. 14893-14903. 63. Nelson, P.T., H. Braak, and W.R. Markesbery, Neuropathology and cognitive impairment in Alzheimer disease: a complex but coherent relationship. J Neuropathol Exp Neurol, 2009. 68(1): p. 1-14. 64. Teter, B. and J.W. Ashford, Neuroplasticity in Alzheimer's disease. J Neurosci Res, 2002. 70(3): p. 402-37. 65. Boutajangout, A. and T. Wisniewski, Tau-based therapeutic approaches for Alzheimer's disease - a mini-review. Gerontology, 2014. 60(5): p. 381-5. 66. Boutajangout, A., D. Quartermain, and E.M. Sigurdsson, Immunotherapy targeting pathological tau prevents cognitive decline in a new tangle mouse model. J Neurosci, 2010. 30(49): p. 16559-66. 67. Asuni, A.A., et al., Immunotherapy targeting pathological tau conformers in a tangle mouse model reduces brain pathology with associated functional improvements. J Neurosci, 2007. 27(34): p. 9115-29. 68. Boutajangout, A., et al., Passive immunization targeting pathological phospho-tau protein in a mouse model reduces functional decline and clears tau aggregates from the brain. J Neurochem, 2011. 118(4): p. 658-67. 69. Andreadis, A., W.M. Brown, and K.S. Kosik, Structure and novel exons of the human tau gene. Biochemistry, 1992. 31(43): p. 10626-33. 70. Goedert, M., et al., Tau proteins of Alzheimer paired helical filaments: abnormal phosphorylation of all six brain isoforms. Neuron, 1992. 8(1): p. 159-68. 71. Li, G., H. Yin, and J. Kuret, Casein kinase 1 delta phosphorylates Tau and disrupts its binding to microtubules. J. Biol. Chem., 2004. 279(16): p. 15938-15945. 72. Mandelkow, E.M., et al., Tau domains, phosphorylation, and interactions with microtubules. Neurobiol Aging, 1995. 16(3): p. 355-62; discussion 362-3. 73. Wang, Y. and E. Mandelkow, Tau in physiology and pathology. Nat Rev Neurosci, 2016. 17(1): p. 5-21. 74. Congdon, E.E. and K.E. Duff, Is tau aggregation toxic or protective? J Alzheimers Dis, 2008. 14(4): p. 453-7. 75. Braak, H., et al., Vulnerability of cortical neurons to Alzheimer's and Parkinson's diseases. J Alzheimers Dis, 2006. 9(3 Suppl): p. 35-44. 76. Gotz, J., G. Halliday, and R.M. Nisbet, Molecular Pathogenesis of the Tauopathies. Annu Rev Pathol, 2018.

124

77. Kovacs, G.G., Invited review: Neuropathology of tauopathies: principles and practice. Neuropathol Appl Neurobiol, 2015. 41(1): p. 3-23. 78. Ballatore, C., V.M. Lee, and J.Q. Trojanowski, Tau-mediated neurodegeneration in Alzheimer's disease and related disorders. Nat Rev Neurosci, 2007. 8(9): p. 663-72. 79. Williams, D.R., Tauopathies: classification and clinical update on neurodegenerative diseases associated with microtubule-associated protein tau. Internal Medicine Journal, 2006. 36(10): p. 652-660. 80. Lee, V.M., M. Goedert, and J.Q. Trojanowski, Neurodegenerative tauopathies. Annu Rev Neurosci, 2001. 24: p. 1121-59. 81. Funk, K.E., et al., Lysine methylation is an endogenous post-translational modification of tau protein in human brain and a modulator of aggregation propensity. Biochem J, 2014. 462(1): p. 77-88. 82. Thomas, S.N., et al., Dual modification of Alzheimer's disease PHF-tau protein by lysine methylation and ubiquitylation: a mass spectrometry approach. Acta Neuropathol, 2012. 123(1): p. 105-17. 83. Crowther, R.A., et al., The microtubule binding repeats of tau protein assemble into filaments like those found in Alzheimer's disease. FEBS Lett, 1992. 309(2): p. 199-202. 84. Barghorn, S., P. Davies, and E. Mandelkow, Tau paired helical filaments from Alzheimer's disease brain and assembled in vitro are based on beta-structure in the core domain. Biochemistry, 2004. 43(6): p. 1694-1703. 85. King, M.E., et al., Ligand-dependent tau filament formation: implications for Alzheimer's disease progression. Biochemistry, 1999. 38(45): p. 14851-14859. 86. Zhong, Q., et al., Tau isoform composition influences rate and extent of filament formation. J Biol Chem, 2012. 287(24): p. 20711-9. 87. Congdon, E.E., et al., Nucleation-dependent tau filament formation: the importance of dimerization and an estimation of elementary rate constants. J. Biol. Chem., 2008. 283(20): p. 13806-13816. 88. Montejo de Garcini, E. and J. Avila, In vitro conditions for the self-polymerization of the microtubule-associated protein, tau factor. J Biochem, 1987. 102(6): p. 1415-21. 89. Goedert, M., et al., Assembly of microtubule-associated protein tau into Alzheimer-like filaments induced by sulphated glycosaminoglycans. Nature, 1996. 383(6600): p. 550-3. 90. Giasson, B.I., et al., Initiation and synergistic fibrillization of tau and -synuclein. Science, 2003. 300(5619): p. 636-640. 91. Gras, S.L., L.J. Waddington, and K.N. Goldie, Transmission electron microscopy of amyloid fibrils. Methods Mol Biol, 2011. 752: p. 197-214. 92. Sousa, A.A. and R.D. Leapman, Mass mapping of amyloid fibrils in the electron microscope using STEM imaging. Methods Mol Biol, 2013. 950: p. 195-207. 93. Carmel, G., et al., The structural basis of monoclonal antibody Alz50's selectivity for Alzheimer's disease pathology. J Biol Chem, 1996. 271(51): p. 32789-95. 94. Carmel, G., et al., Expression, purification, crystallization, and preliminary x-ray analysis of casein kinase-1 from Schizosaccharomyces pombe. J Biol Chem, 1994. 269(10): p. 7304-9. 95. Necula, M. and J. Kuret, A static laser light scattering assay for surfactant-induced tau fibrillization. Anal. Biochem., 2004. 333(2): p. 205-215.

125

96. Nielsen, L., et al., Effect of environmental factors on the kinetics of insulin fibril formation: elucidation of the molecular mechanism. Biochemistry, 2001. 40(20): p. 6036- 6046. 97. Kristofferson, D., T.L. Karr, and D.L. Purich, Dynamics of linear protein polymer disassembly. J. Biol. Chem., 1980. 255(18): p. 8567-8572. 98. Necula, M. and J. Kuret, Site-specific pseudophosphorylation modulates the rate of tau filament dissociation. FEBS Lett., 2005. 579(6): p. 1453-1457. 99. Chang, E., et al., Pathogenic missense MAPT mutations differentially modulate tau aggregation propensity at nucleation and extension steps. J. Neurochem., 2008. 107(4): p. 1113-1123. 100. Schafer, K.N., et al., Structural determinants of Tau aggregation inhibitor potency. J Biol Chem, 2013. 288(45): p. 32599-611. 101. Jarrett, J.T. and P.T. Lansbury, Jr., Seeding "one-dimensional crystallization" of amyloid: a pathogenic mechanism in Alzheimer's disease and scrapie? Cell, 1993. 73(6): p. 1055-8. 102. Chang, E., et al., Structure-activity relationship of cyanine tau aggregation inhibitors. J Med Chem, 2009. 52(11): p. 3539-47. 103. Chirita, C.N., M. Necula, and J. Kuret, Anionic micelles and vesicles induce tau fibrillization in vitro. J Biol Chem, 2003. 278(28): p. 25644-50. 104. Chirita, C.N., et al., Triggers of full-length tau aggregation: a role for partially folded intermediates. Biochemistry, 2005. 44(15): p. 5862-5872. 105. Necula, M. and J. Kuret, Electron microscopy as a quantitative method for investigating tau fibrillization. Anal. Biochem., 2004. 329(2): p. 238-246. 106. Friedhoff, P., et al., A nucleated assembly mechanism of Alzheimer paired helical filaments. Proc. Natl. Acad. Sci. U.S.A., 1998. 95(26): p. 15712-15717. 107. Zhao, D. and J.S. Moore, Nucleation-elongation: a mechanism for cooperative supramolecular polymerization. Org. Biomol. Chem., 2003. 1(20): p. 3471-3491. 108. Pappu, R.V., et al., A polymer physics perspective on driving forces and mechanisms for protein aggregation. Arch Biochem Biophys, 2008. 469(1): p. 132-41. 109. Ferrone, F., Analysis of protein aggregation kinetics. Methods Enzymol, 1999. 309: p. 256-74. 110. King, M.E., et al., Structural analysis of Pick's disease-derived and in vitro-assembled tau filaments. Am. J. Pathol., 2001. 158(4): p. 1481-1490. 111. von Bergen, M., et al., The core of tau-paired helical filaments studied by scanning transmission electron microscopy and limited proteolysis. Biochemistry, 2006. 45(20): p. 6446-6457. 112. Braak, H., et al., Stages of the pathologic process in Alzheimer disease: age categories from 1 to 100 years. J Neuropathol Exp Neurol, 2011. 70(11): p. 960-9. 113. Gomez-Isla, T., et al., Profound loss of layer II entorhinal cortex neurons occurs in very mild Alzheimer's disease. J. Neurosci., 1996. 16(14): p. 4491-4500. 114. Ghoshal, N., et al., Tau conformational changes correspond to impairments of episodic memory in mild cognitive impairment and Alzheimer's disease. Exp. Neurol., 2002. 177(2): p. 475-493. 115. Wu, J.W., et al., Small misfolded Tau species are internalized via bulk endocytosis and anterogradely and retrogradely transported in neurons. J Biol Chem, 2013. 288(3): p. 1856-70.

126

116. Mirbaha, H., et al., Tau Trimers Are the Minimal Propagation Unit Spontaneously Internalized to Seed Intracellular Aggregation. J Biol Chem, 2015. 290(24): p. 14893-903. 117. Ramachandran, G. and J.B. Udgaonkar, Understanding the kinetic roles of the inducer heparin and of rod-like protofibrils during amyloid fibril formation by Tau protein. J Biol Chem, 2011. 286(45): p. 38948-59. 118. Maeda, S., et al., Granular tau oligomers as intermediates of tau filaments. Biochemistry, 2007. 46(12): p. 3856-61. 119. Pastor, M.T., et al., Amyloid toxicity is independent of polypeptide sequence, length and chirality. J Mol Biol, 2008. 375(3): p. 695-707. 120. Pavlova, A., et al., Protein structural and surface water rearrangement constitute major events in the earliest aggregation stages of tau. Proc Natl Acad Sci U S A, 2016. 113(2): p. E127-36. 121. Ferrone, F.A., J. Hofrichter, and W.A. Eaton, Kinetics of sickle hemoglobin polymerization. II. A double nucleation mechanism. J. Mol. Biol., 1985. 183(4): p. 611- 631. 122. Cohen, S.I., et al., Proliferation of amyloid-42 aggregates occurs through a secondary nucleation mechanism. Proc Natl Acad Sci U S A, 2013. 110(24): p. 9758-63. 123. Cohen, S.I., et al., Nucleated polymerization with secondary pathways. I. Time evolution of the principal moments. J Chem Phys, 2011. 135(6): p. 065105. 124. Ramachandran, G. and J.B. Udgaonkar, Evidence for the existence of a secondary pathway for fibril growth during the aggregation of tau. J Mol Biol, 2012. 421(2-3): p. 296-314. 125. Tornquist, M., et al., Secondary nucleation in amyloid formation. Chem Commun (Camb), 2018. 54(63): p. 8667-8684. 126. Tanaka, M., et al., The physical basis of how prion conformations determine strain phenotypes. Nature, 2006. 442(7102): p. 585-9. 127. Gray, E.G., M. Paula-Barbosa, and A. Roher, Alzheimer's disease: paired helical filaments and cytomembranes. Neuropathol. Appl. Neurobiol., 1987. 13(2): p. 91-110. 128. Carlson, S.W., et al., A complex mechanism for inducer mediated tau polymerization. Biochemistry, 2007. 46(30): p. 8838-49. 129. Rothwell, S.W., W.A. Grasser, and D.B. Murphy, End-to-end annealing of microtubules in vitro. J Cell Biol, 1986. 102(2): p. 619-27. 130. Andrianantoandro, E., et al., Kinetic mechanism of end-to-end annealing of actin filaments. J Mol Biol, 2001. 312(4): p. 721-30. 131. Sept, D., et al., Annealing accounts for the length of actin filaments formed by spontaneous polymerization. Biophys J, 1999. 77(6): p. 2911-9. 132. Kirmse, R., et al., A quantitative kinetic model for the in vitro assembly of intermediate filaments from tetrameric vimentin. J Biol Chem, 2007. 282(25): p. 18563-72. 133. Bridges, A.A., et al., Septin assemblies form by diffusion-driven annealing on membranes. Proc Natl Acad Sci U S A, 2014. 111(6): p. 2146-51. 134. Murphy, D.B., et al., Direct demonstration of actin filament annealing in vitro. J Cell Biol, 1988. 106(6): p. 1947-54. 135. Richardson, J.S. and D.C. Richardson, Natural -sheet proteins use negative design to avoid edge-to-edge aggregation. Proc Natl Acad Sci U S A, 2002. 99(5): p. 2754-9. 136. Carmel, G., et al., The structural basis of monoclonal antibody Alz50's selectivity for Alzheimer's disease pathology. J. Biol. Chem., 1996. 271(51): p. 32789-32795.

127

137. Elbaum-Garfinkle, S. and E. Rhoades, Identification of an aggregation-prone structure of tau. J Am Chem Soc, 2012. 134(40): p. 16607-13. 138. Huseby, C.J. and J. Kuret, Analyzing Tau Aggregation with Electron Microscopy. Methods Mol Biol, 2016. 1345: p. 101-12. 139. Nanavaty, N., et al., Detection and Quantification Methods for Fibrillar Products of In Vitro Tau Aggregation Assays. Methods Mol Biol, 2017. 1523: p. 101-111. 140. Segel, L.A. and M. Slemrod, The Quasi-Steady-State Assumption - a Case-Study in Perturbation. Siam Review, 1989. 31(3): p. 446-477. 141. Saltelli, A., et al., Update 1 of: Sensitivity analysis for chemical models. Chemical Reviews (Washington, D. C.), 2012. 112(5): p. PR1-21. 142. Saltelli, A., S. Tarantola, and K.P.S. Chan, A quantitative model-independent method for global sensitivity analysis of model output. Technometrics, 1999. 41(1): p. 39-56. 143. Chirita, C.N. and J. Kuret, Evidence for an intermediate in tau filament formation. Biochemistry, 2004. 43(6): p. 1704-14. 144. Hill, T.L., Length dependence of rate constants for end-to-end association and dissociation of equilibrium linear aggregates. Biophys J, 1983. 44(2): p. 285-8. 145. Wegner, A. and J. Engel, Kinetics of the cooperative association of actin to actin filaments. Biophys Chem, 1975. 3(3): p. 215-25. 146. Hill, T.L., Bioenergetic aspects and polymer length distribution in steady-state head-to- tail polymerization of actin or microtubules. Proc Natl Acad Sci U S A, 1980. 77(8): p. 4803-7. 147. Hill, T.L., Steady-state head-to-tail polymerization of actin or microtubules. II. Two-state and three-state kinetic cycles. Biophys J, 1981. 33(3): p. 353-71. 148. von Bergen, M., et al., Assembly of tau protein into Alzheimer paired helical filaments depends on a local sequence motif (306VQIVYK311) forming  structure. Proc. Natl. Acad. Sci. U.S.A., 2000. 97(10): p. 5129-5134. 149. Wegner, A. and P. Savko, Fragmentation of actin filaments. Biochemistry, 1982. 21(8): p. 1909-13. 150. Knowles, T.P., et al., An analytical solution to the kinetics of breakable filament assembly. Science, 2009. 326(5959): p. 1533-7. 151. Schreiber, G., G. Haran, and H.X. Zhou, Fundamental aspects of protein-protein association kinetics. Chem Rev, 2009. 109(3): p. 839-60. 152. Doi, M. and S. Edwards, The Theory of Polymber Dynamics. 1986, Oxford: Clarendon. 153. Wegmann, S., et al., Human Tau isoforms assemble into ribbon-like fibrils that display polymorphic structure and stability. J Biol Chem, 2010. 285(35): p. 27302-13. 154. Zampieri, M., G. Legname, and C. Altafini, Investigating the conformational stability of prion strains through a kinetic replication model. PLoS Comput Biol, 2009. 5(7): p. e1000420. 155. Kundel, F., et al., Measurement of Tau Filament Fragmentation Provides Insights into Prion-like Spreading. ACS Chem Neurosci, 2018. 9(6): p. 1276-1282. 156. Xue, W.F. and S.E. Radford, An imaging and systems modeling approach to fibril breakage enables prediction of amyloid behavior. Biophys J, 2013. 105(12): p. 2811-9. 157. Berger, Z., et al., Accumulation of pathological tau species and memory loss in a conditional model of tauopathy. J. Neurosci., 2007. 27(14): p. 3650-3662. 158. Lasagna-Reeves, C.A., et al., Tau oligomers impair memory and induce synaptic and mitochondrial dysfunction in wild-type mice. Mol Neurodegener, 2011. 6: p. 39.

128

159. Hagestedt, T., et al., Tau protein becomes long and stiff upon phosphorylation: correlation between paracrystalline structure and degree of phosphorylation. J Cell Biol, 1989. 109(4 Pt 1): p. 1643-51. 160. Wegmann, S., et al., The fuzzy coat of pathological human Tau fibrils is a two-layered polyelectrolyte brush. Proc Natl Acad Sci U S A, 2013. 110(4): p. E313-21. 161. Wischik, C.M., et al., Structural characterization of the core of the paired helical filament of Alzheimer disease. Proc Natl Acad Sci U S A, 1988. 85(13): p. 4884-8. 162. Fitzpatrick, A.W.P., et al., Cryo-EM structures of tau filaments from Alzheimer's disease. Nature, 2017. 547(7662): p. 185-190. 163. Martin, L., X. Latypova, and F. Terro, Post-translational modifications of tau protein: implications for Alzheimer's disease. Neurochem Int, 2011. 58(4): p. 458-71. 164. Hanger, D.P., B.H. Anderton, and W. Noble, Tau phosphorylation: the therapeutic challenge for neurodegenerative disease. Trends Mol Med, 2009. 15(3): p. 112-9. 165. Cook, C., et al., Acetylation of the KXGS motifs in tau is a critical determinant in modulation of tau aggregation and clearance. Hum Mol Genet, 2014. 23(1): p. 104-16. 166. Petrucelli, L., et al., CHIP and Hsp70 regulate tau ubiquitination, degradation and aggregation. Hum. Mol. Genet., 2004. 13(7): p. 703-714. 167. Shimura, H., et al., CHIP-Hsc70 complex ubiquitinates phosphorylated tau and enhances cell survival. J. Biol. Chem., 2004. 279(6): p. 4869-4876. 168. Morris, M., et al., Tau post-translational modifications in wild-type and human amyloid precursor protein transgenic mice. Nat Neurosci, 2015. 18(8): p. 1183-9. 169. Coorssen, J.R. and A.L. Yergey, Proteomics Is Analytical Chemistry: Fitness-for-Purpose in the Application of Top-Down and Bottom-Up Analyses. Proteomes, 2015. 3(4): p. 440- 453. 170. Moore, S. and W.H. Stein, Chromatographic determination of amino acids by the use of automatic recording equipment. Methods in Enzymology, 1963. 6: p. 819-831. 171. Murray, K., et al., The Hydrolysis of Arginine. J Biol Chem, 1965. 240: p. 705-9. 172. Tsiboli, P., et al., Identification of post-translational modified amino acids. Amino Acids, 1997. 13(1): p. 13-23. 173. Yan, J.X., et al., Method for identification and quantitative analysis of protein lysine methylation using matrix-assisted laser desorption/ionization--time-of-flight mass spectrometry and amino acid analysis. Electrophoresis, 1999. 20(4-5): p. 749-54. 174. Hirschey, M.D. and Y. Zhao, Metabolic Regulation by Lysine Malonylation, Succinylation, and Glutarylation. Mol Cell Proteomics, 2015. 14(9): p. 2308-15. 175. Streich, F.C., Jr. and C.D. Lima, Structural and functional insights to ubiquitin-like protein conjugation. Annu Rev Biophys, 2014. 43: p. 357-79. 176. Watterson, D.M., P.A. Mendel, and T.C. Vanaman, Comparison of calcium-modulated proteins from vertebrate brains. Biochemistry, 1980. 19(12): p. 2672-6. 177. Young, P.R. and F. Grynspan, Analysis of methylated amino acids by high-performance liquid chromatography: methylation of myelin basic protein. J Chromatogr, 1987. 421(1): p. 130-5. 178. Brostoff, S. and E.H. Eylar, Localization of methylated arginine in the A1 protein from myelin. Proc Natl Acad Sci U S A, 1971. 68(4): p. 765-9. 179. Deibler, G.E. and R.E. Martenson, Determination of methylated basic amino acids with the amino acid analyzer. Application to total acid hydrolyzates of myelin basic proteins. J Biol Chem, 1973. 248(7): p. 2387-91.

129

180. Kakimoto, Y., et al., Methylated amino acid residues of proteins of brain and other organs. J Neurochem, 1975. 24(5): p. 893-902. 181. Friedman, M., Applications of the ninhydrin reaction for analysis of amino acids, peptides, and proteins to agricultural and biomedical sciences. J Agric Food Chem, 2004. 52(3): p. 385-406. 182. García Alvarez-Coque, M.C., et al., Formation and instability of o-phthalaldehyde derivatives of amino acids. Anal Biochem, 1989. 178(1): p. 1-7. 183. Richardson, S.L., et al., A direct, ratiometric, and quantitative MALDI-MS assay for protein methyltransferases and acetyltransferases. Anal Biochem, 2015. 478: p. 59-64. 184. Jentoft, N. and D.G. Dearborn, Labeling of proteins by reductive methylation using sodium cyanoborohydride. J Biol Chem, 1979. 254(11): p. 4359-65. 185. McKeon, T.A. and M.L. Lyman, Calcium ion improves electrophoretic transfer of calmodulin and other small proteins. Anal Biochem, 1991. 193(1): p. 125-30. 186. McClure, T., et al., Impact of Environment on the Biomass Composition of Soybean (Glycine max) seeds. J Agric Food Chem, 2017. 65(32): p. 6753-6761. 187. Matsuo, E.S., et al., Biopsy-derived adult human brain tau is phosphorylated at many of the same sites as Alzheimer's disease paired helical filament tau. Neuron, 1994. 13(4): p. 989-1002. 188. Cocuron, J.C., E. Tsogtbaatar, and A.P. Alonso, High-throughput quantification of the levels and labeling abundance of free amino acids by liquid chromatography tandem mass spectrometry. J Chromatogr A, 2017. 1490: p. 148-155. 189. Cocuron, J.C., et al., Targeted metabolomics of Physaria fendleri, an industrial crop producing hydroxy fatty acids. Plant Cell Physiol, 2014. 55(3): p. 620-33. 190. Wang, Z., et al., Targeted metabolomic evaluation of arginine methylation and cardiovascular risks: potential mechanisms beyond nitric oxide synthase inhibition. Arterioscler Thromb Vasc Biol, 2009. 29(9): p. 1383-91. 191. Vial, J. and A. Jardy, Experimental comparison of the different approaches to estimate LOD and LOQ of an HPLC method. Analytical Chemistry, 1999. 71(14): p. 2672-2677. 192. Goedert, M., et al., Multiple isoforms of human microtubule-associated protein tau: sequences and localization in neurofibrillary tangles of Alzheimer's disease. Neuron, 1989. 3(4): p. 519-526. 193. Bernstein, B.E., et al., Genomic maps and comparative analysis of histone modifications in human and mouse. Cell, 2005. 120(2): p. 169-81. 194. Di Lorenzo, A. and M.T. Bedford, Histone arginine methylation. FEBS Lett, 2011. 585(13): p. 2024-31. 195. Paik, W.K. and S. Kim, Epsilon-N-Dimethyllysine in Histones. Biochemical and Biophysical Research Communications, 1967. 27(4): p. 479-483. 196. Delange, R.J., E.L. Smith, and J. Bonner, Calf Thymus Histone 3 - Sequences of Amino- Terminal and Carboxyl-Terminal Regions and of Regions Containing Lysyl Residues Modified by Acetylation and Methylation. Biochemical and Biophysical Research Communications, 1970. 40(4): p. 989-993. 197. Goedert, M. and R. Jakes, Expression of separate isoforms of human tau protein: correlation with the tau pattern in brain and effects on tubulin polymerization. EMBO J, 1990. 9(13): p. 4225-30. 198. McMillan, P., et al., Tau isoform regulation is region- and cell-specific in mouse brain. J Comp Neurol, 2008. 511(6): p. 788-803.

130

199. Johnson, G.V., et al., The tau protein in human cerebrospinal fluid in Alzheimer's disease consists of proteolytically derived fragments. J Neurochem, 1997. 68(1): p. 430-3. 200. Meredith, J.E., Jr., et al., Characterization of novel CSF Tau and ptau biomarkers for Alzheimer's disease. PLoS One, 2013. 8(10): p. e76523. 201. Yamauchi, M. and M. Shiiba, Lysine hydroxylation and cross-linking of collagen. Methods Mol Biol, 2008. 446: p. 95-108. 202. Rabbani, N., A. Ashour, and P.J. Thornalley, Mass spectrometric determination of early and advanced glycation in biology. Glycoconj J, 2016. 33(4): p. 553-68. 203. Amin, B. and W. Voelter, Human Deiminases: Isoforms, Substrate Specificities, Kinetics, and Detection. Prog Chem Org Nat Prod, 2017. 106: p. 203-240. 204. Tutturen, A.E., B. Fleckenstein, and G.A. de Souza, Assessing the citrullinome in rheumatoid arthritis synovial fluid with and without enrichment of citrullinated peptides. J Proteome Res, 2014. 13(6): p. 2867-73. 205. Jin, Z., et al., Identification and characterization of citrulline-modified brain proteins by combining HCD and CID fragmentation. Proteomics, 2013. 13(17): p. 2682-91. 206. Bramblett, G.T., J.Q. Trojanowski, and V.M. Lee, Regions with abundant neurofibrillary pathology in human brain exhibit a selective reduction in levels of binding-competent tau and accumulation of abnormal tau-isoforms (A68 proteins). Lab. Invest., 1992. 66(2): p. 212-222. 207. Biernat, J., et al., Phosphorylation of Ser262 strongly reduces binding of tau to microtubules: distinction between PHF-like immunoreactivity and microtubule binding. Neuron, 1993. 11(1): p. 153-163. 208. Ksiezak-Reding, H., W.K. Liu, and S.H. Yen, Phosphate analysis and dephosphorylation of modified tau associated with paired helical filaments. Brain Res., 1992. 597(2): p. 209- 219. 209. Greenberg, S.G. and P. Davies, A preparation of Alzheimer paired helical filaments that displays distinct tau proteins by polyacrylamide gel electrophoresis. Proc Natl Acad Sci U S A, 1990. 87(15): p. 5827-31. 210. Ksiezak-Reding, H. and J.S. Wall, Mass and physical dimensions of two distinct populations of paired helical filaments. Neurobiol Aging, 1994. 15(1): p. 11-9. 211. Bandyopadhyay, B., et al., Tau aggregation and toxicity in a cell culture model of tauopathy. J. Biol. Chem., 2007. 282(22): p. 16454-16464. 212. Smyth, G.K., Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol, 2004. 3: p. Article3. 213. Zhang, B., et al., Integrated systems approach identifies genetic nodes and networks in late-onset Alzheimer's disease. Cell, 2013. 153(3): p. 707-20. 214. Stafford, P., Methods in microarray normalization. Drug discovery series. 2008, Boca Raton: CRC Press. xvi, 284 p. 215. Miners, S., et al., Genetic variation in MME in relation to neprilysin protein and enzyme activity, Abeta levels, and Alzheimer's disease risk. Int J Mol Epidemiol Genet, 2012. 3(1): p. 30-8. 216. Reitz, C., et al., Matrix metalloproteinase 3 haplotypes and dementia and Alzheimer's disease. The Rotterdam Study. Neurobiol Aging, 2008. 29(6): p. 874-81. 217. Shi, Y., et al., Histone demethylation mediated by the nuclear amine oxidase homolog LSD1. Cell, 2004. 119(7): p. 941-53.

131

218. Rea, S., et al., Regulation of chromatin structure by site-specific histone H3 methyltransferases. Nature, 2000. 406(6796): p. 593-9. 219. Petrossian, T.C. and S.G. Clarke, Uncovering the human methyltransferasome. Mol Cell Proteomics, 2011. 10(1): p. M110 000976. 220. Su, I.H., et al., Polycomb group protein ezh2 controls actin polymerization and cell signaling. Cell, 2005. 121(3): p. 425-36. 221. Hirano, T., S. Mori, and H. Kagechika, Recent Advances in Chemical Tools for the Regulation and Study of Protein Lysine Methyltransferases. Chem Rec, 2018. 222. Gowher, H. and A. Jeltsch, Mammalian DNA methyltransferases: new discoveries and open questions. Biochem Soc Trans, 2018. 46(5): p. 1191-1202. 223. Penney, J., et al., Loss of Protein Arginine Methyltransferase 8 Alters Synapse Composition and Function, Resulting in Behavioral Defects. J Neurosci, 2017. 37(36): p. 8655-8666.

132