Deep Phenotyping of ALS and ALS-FTD Mouse Models Using Automated Continuous Behavioral Monitoring Amanda Marie Duffy ScB, Brown University, 2009 THESIS Submitted in partial fulfillment of the requirements for the Degree of Doctor of Philosophy in the Department of Neuroscience at Brown University Providence, Rhode Island May 2018 ©2018 Amanda Marie Duffy This dissertation by Amanda Marie Duffy is accepted in its present form by the Department of Neuroscience as satisfying the dissertation requirements for the degree of Doctor of Philosophy. ___________________ ________________________________________________ Date Dr. Justin Fallon, Advisor Recommended to the Graduate Council ___________________ ________________________________________________ Date Dr. Anne Hart, Reader ___________________ ________________________________________________ Date Dr. Gilad Barnea, Reader ___________________ ________________________________________________ Date Dr. Kevin Bath, Reader ___________________ ________________________________________________ Date Dr. Lawrence Hayward, Outside Reader Approved by the Graduate Council ___________________ ________________________________________________ Date Dr. Andrew G. Campbell, Dean of the Graduate School iii Amanda Marie Duffy Curriculum Vitae Laboratory Address: Amanda Marie Duffy Brown University Neuroscience Department 185 Meeting Street Box GL-N Providence, RI 02912 Home Address: 410 Benefit Street Apartment 1 Providence, RI 02903 Electronic Mail Address: Amanda_Duffy@brown.edu Education Brown University, Providence, RI: Ph.D. Candidate, May 2018 Neuroscience Graduate Program Advisor: Justin Fallon, Ph.D. Brown University, Providence, RI: Sc.M. April 2013 Neuroscience Graduate Program Advisor: Justin Fallon, Ph.D. Brown University, Providence, RI: B.S., Honors, May 2009 Concentration: Neuroscience Academic Advisor: David Sheinberg, Ph.D. Research Experience Graduate Student, Brown University 2012-2018 Department of Neuroscience, Dr. Justin Fallon Identifying cellular and molecular mechanisms that maintain or enhance neuromuscular junction (NMJ) stability. Specifically, I am identifying binding properties of proteins at the NMJ including agrin, lipoprotein receptor related protein 4 (LRP4), Muscle Specific Kinase (MuSK) and biglycan. Characterizing ALS and frontal temporal dementia (FTD) disease progression using an automated video monitoring system and traditional behavioral measures to identify progressive and age-dependent behavioral phenotypes in a variety of ALS-FTD mouse models. Additionally, I am identifying early time windows prior to the onset of behavioral phenotypes during which pharmacological intervention could prolong function. iv Rotation Graduate Student, Brown University 2011–2012 Department of Neuroscience, Drs. Barry Connors, Anne Hart, and Justin Fallon Learned and rehearsed mouse brain dissection and neural recording techniques; Generated a C. elegans model for spinal muscular atrophy and identified neuromotor phenotypes; Conducted immunostaining of specific NMJ proteins on mouse muscle sections to optimize staining procedures and quantify protein level differences. Research Assistant, Massachusetts General Hospital 2009-2011 Division of Neurotherapeutics, Darin Dougherty M.D., Thilo Deckersbach Ph.D. Examined the effects of deep brain stimulation (DBS) in the ventral capsule/ventral striatum on treatment-resistant OCD patients through a double- blinded sham controlled clinical trial. I coordinated major aspects of the trial including patient screening, pre- and post-surgery administration of neuropsychological testing to measure improvements in OCD symptoms, depression, and quality of life on all patients, and administration of pre- and post- operative MRIs (functional and structural) and PET scans. I initiated and conducted avoidance reward conflict and fear extinction tasks on pre- and post- DBS patients to determine whether DBS influenced risk-taking behavior, fear arousal and fear extinction. Coordinated all aspects of the DBS surgeries with the surgical team, Medtronic representatives, attending doctors, and patients, and acted as a liaison between the patients, their families, and the surgeon. Attended surgical DBS surgeries and participated in monthly Psychiatric Neurosurgery Committee meetings to discuss patient selection for neuropsychiatric surgery and discuss their follow-up care. Coordinated, conducted, and analyzed fMRIs on bipolar disorder patients and healthy controls performing a variety of tasks that probe arousal to identify potential differences between cohorts. Performed fMRIs on bipolar patients participating in a mindfulness program to determine the impact of medication on blood flow in different regions in the brain, as well as on self-reported measures of functionality and improvement. Undergraduate Neuroscience Honors Thesis Student, Brown University 2008-2009 Department of Neuroscience, John Donahue M.D. The Relationship between Apolipoprotein E Genotype and -Amyloid Transporter Proteins within the Blood-Brain Barrier in Regard to Alzheimer’s Disease Demonstrated decreased levels of lipoprotein receptor related protein 1 (LRP-1) and increased levels of the receptor for advanced glycation end-products (RAGE), in the prefrontal cortex of Alzheimer’s disease patients. The presence of the Apolipoprotein E4 allele enhanced this ratio. Intern, Boston University School of Medicine Summer Department of Biochemistry, Carmela Abraham Ph.D. 2008 Developed a chemiluminescent tool using secreted alkaline phosphatase (SEAP) to identify and quantify klotho cleaved from cells, and demonstrated that this tool is sensitive enough to detect the effect of insulin increasing the amount of klotho cleaved from Cos7 cells. v Intern, Massachusetts General Hospital Summer Psychiatric and Neurodevelopmental Genetics Unit, David Pauls M.D. 2007 Scored OCD and Tourette syndrome patient questionnaires to identify correlations between the two disorders and conducted literature reviews on Pediatric Autoimmune Neuropsychiatric Disorders Associated with Streptoccocal Infections (PANDAS). Intern, Brown University 2006-2007 Department of Cognitive, Linguistic, and Psychological Sciences, Michael Tarr Ph.D. Helped orchestrate a psychophysics study on visual object recognition to determine whether novel object recognition is mediated by specific features of object shape or by the “holistic” spatial arrangement of the objects. Intern, Harvard Medical School 2006-2009 Department of Neurobiology, David Hubel M.D. Assisted Dr. Hubel in demonstrating that afterimages of the foveal blind spot under scotopic conditions due to rod photoreceptor activation after dark adaptation can be observed and quantified. In cat models of induced amblyopia, measured and compared the diameters of blood vessels between those nourishing the visually deprived eye compared to the visually non-deprived eye and demonstrated no difference in size. Publications Matthew White, Eosu Kim, Amanda Duffy, Robert Adalbert, Benjamin Phillips, Owen Peters, Jodie Stephenson, Sujeong Yang, Francesca Massenzio, Ziqiang Lin, Simon Andrews, Anne Segonds-Pichon, Jake Metterville, Lisa Saksida, Richard Mead, Richard Ribchester, Youssef Barhomi, Thomas Serre, Michael Coleman, Justin Fallon, Timothy Bussey, Robert Brown. (2018). TDP-43 knock-in mouse yields modifiers of cognitive dysfunction in ALS-FTD. Nature Neuroscience, 21(4), 552-563. Dougherty, D. D., Corse, A. K., Chou, T., Duffy, A., Arulpragasam, A. R., Deckersbach, T., Jenike, M. A, Keuthen, N. J. (2015). Open-Label Study of Duloxetine for the Treatment of Obsessive-Compulsive Disorder. International Journal of Neuropsychopharmacology, 18(2), 1-4. Sierra-Mercado, D., Deckersbach, T., Arulpragasam A. R., Chou, T., Rodman, A. M., Duffy, A., McDonald, E. J., Eckhardt, C. A., Corse, A. K., Kaur, N., Eskandar, E. N., Dougherty, D. D. (2015). Decision Making in Avoidance-Reward Conflict: A Paradigm for Non-Human Primates and Humans. Brain Structure and Function, 220(5): 2509-17. Hubel, D. H., Howe, P. D., Duffy, A. M., Hernandez, A. (2009). Scotopic Foveal Afterimages. Perception. 38 (2): 313-316. Presentations Duffy, A.M. (November, 2017). Automated continuous behavioral monitoring reveals early phenotypes in a novel TDP-43 knock-in mouse model of ALS-FTD. Society for Neuroscience, Nanosymposium, Washington, DC vi Duffy, A. M. (March, 2017). Behavioral Phenotyping of an ALS-FTD Mouse Model. Neuroscience Graduate Program In-House Seminar, Brown University, Providence, RI. Duffy, A. M., Fallon, J. (January, 2017). Characterizing new ALS mouse models. ALS@Brown, Brown University, RI Duffy, A.M. (November, 2014). Using Automated Continuous Behavioral Monitoring as a tool to identify and characterize the behavior of an ALS mouse model. ALS@Brown, Brown University, Providence, RI. Duffy, A.M. (April, 2014). Slowing ALS disease progression by targeting synaptic instability at the NMJ. ALS@Brown, Brown University, Providence, RI. Duffy, A.M. (April, 2013). Biglycan and synaptic stability. Neuroscience Graduate Program In- House Seminar, Brown University, Providence, RI. Selected Poster Presentations Duffy, A.M., Fallon, J. (November, 2016). The Role of Biglycan in Regulating Stabilization of the Nerve-muscle Synapse. Society for Neuroscience, San Diego, CA. Duffy, A.M., Li, X., Schmiedel, C., Mentzer, S., Bath, K., Serre, T., Fallon, J. (November, 2015). Automated Continuous Behavioral Monitoring reveals novel neuromotor phenotypes in a mouse model of ALS. Brown-NIH Graduate Partnership Program, Woods Hole, MA. Duffy, A.M., Fallon, J. (October, 2015). Automated Continuous Behavioral Monitoring reveals novel neuromotor phenotypes in a mouse model of ALS. Society for Neuroscience, Chicago, IL. Duffy, A.M., Li, X., Schmiedel, C., Mentzer, S., Bath, K., Serre, T., Fallon, J. (2015, March). Automated Continuous Behavioral Monitoring reveals novel phenotypes in mouse models of ALS. Mind Brain Research Day 2015, Providence, RI. Duffy, A.M., Li, X., Schmiedel, C., Mentzer, S., Bath, K., Serre, T., Fallon, J. (2015, March). Automated Continuous Behavioral Monitoring reveals novel phenotypes in mouse models of ALS and Muscular Dystrophy. MDA 2015 Scientific Conference, Washington, D.C. Duffy, A.M., Fallon, J. (March, 2014). Biglycan as a candidate therapeutic for ALS: Mechanisms underlying synaptic stabilization. Mind Brain Research Day 2014, Providence, RI. Duffy, A.M., Fallon, J. (November, 2013). Biglycan regulates MuSK-LRP4 binding; Implications for synapse stabilization. Neurological and Psychiatric Diseases: Model Systems and Treatments, Providence, RI. Duffy, A.M. Fallon, J. (August, 2013). A new role for biglycan in regulating synapse differentiation: Implications for ALS therapy. Brown University Annual Neuroscience Retreat, Bristol, RI. Duffy, A.M., Fallon, J. (August 2012). Biglycan as a potential ALS therapeutic to promote synaptic stability. Brown University Annual Neuroscience Retreat, Bristol, RI. Peters, A., Duffy, A., Rodman, A., Chou, T., Peckham, A., Ghaznavi, S., Stange, J., Rauch, S., Nierenberg, A., Dougherty, D., Deckersbach, T. (2011). From Cortex to Context – Neural Correlates of Daily Functioning Difficulties in Bipolar Disorder. An fMRI Study. 45th Association for Behavioral and Cognitive Therapies Convention, Toronto, Canada. vii Deckersbach, T., Duffy, A., Chou, T., Rodman, A., Peckham, A., Peters, A., Stange, J., Nierenberg, A., Dougherty, D. (2011). Affective Context and Attention in Bipolar Disorder – Results from fMRI studies. Association for Behavioral and Cognitive Therapies Convention, Toronto, Canada. Duffy, A., Rodman, A., Chou, T., Peckham, A., Hay, A., Peters, A., Ariel, A., Martowski, J., Nierenberg, A., Rauch, S., Dougherty, D., Deckersbach, T. (2011). Emotional Context and Attentional Dysfunction in Bipolar Disorder: An fMRI Study. 64th Annual Meeting of the Massachusetts General Hospital Scientific Advisory Committee, Boston, MA. David J. Grelotti, Amanda Duffy, Tina Chou, Lindsay Carlson, Scott L. Rauch, Darin D. Dougherty, Thilo Deckersbach. (2011). Greater alexithymia in patients with bipolar disorder and its relationship to brain activation during emotion processing. Harvard Psychiatry Research Day Poster Session and Mysell Lecture. Boston, MA. Deckersbach T, Duffy A, Beucke J, Chou T, Dougherty D, Sachs G, Nierenberg A, Rauch S. (2010). Attention and Affect in Bipolar Disorder: An fMRI Study. American College of Neuropsychopharmacology. Miami Beach, FL. Beucke K, Duffy A, Chou T, Rauch S, Dougherty D, Deckersbach T. (2010). Staying Focused: An fMRI Investigation of the Effect of Emotional Distraction and Working Memory. Association for Behavioral and Cognitive Therapies. San Francisco, CA. Deckersbach T, Duffy A, Chou T, Carlson L. (2010). Attention and CBT in Bipolar Disorder: An fMRI Study. Association for Behavioral and Cognitive Therapies. San Francisco, CA. Campos M, Camacho J, Duffy A, Gale J, Niemi S, Dougherty D, Eskandar E. (2010). PET Imaging of Rhesus Monkeys with OCD-Like Behaviors. Society for Neuroscience. San Diego, CA. Donahue J, Duffy A, Lim J, Miller M. (2009). Influence of Apolipoprotein E Genoytpe on RAGE expression in Alzheimer’s Disease. American Association of Neuropathologists Inc. San Antonio, TX. Teaching/Mentoring Experience Guest Lecturer, Physiological Psychology, Rhode Island College, Arielle 2017 Nitenson, Ph.D. Taught two 2-hour course sections on 1.) Neurological diseases - Epilepsy, Parkinson’s disease, Huntington’s disease and Multiple Sclerosis, and 2.) Brain injury, specifically discussing regions of damage and corresponding deficits, as well as opportunities and potentials for neurorehabilitation. Graduate Student Mentor/Supervisor, Brown University, Justin Fallon 2012-2017 Laboratory, Providence, RI Guides and advises graduate, undergraduate, and high school student interns on cellular, molecular, and rodent behavioral laboratory techniques. Teaching Assistant, The Brain: An Introduction to Neuroscience, Brown 2012 University, Providence, RI, Drs. Michael Paradiso and John Stein Taught biweekly sections, ran pre-exam review sections, provided additional one-on-one academic support, and designed practice questions, answers, and review sheets. viii Student and Employee Accessibility Services (SEAS) Note-taker, The 2012 Brain: An Introduction to Neuroscience, Brown University, Providence, RI, Drs. Michael Paradiso and John Stein Wrote comprehensive biweekly course notes for SEAS to distribute to students requesting additional support. Research Assistant/Research Coordinator, Massachusetts General 2009-2011 Hospital, Division of Neurotherapeutics, Drs. Darin Dougherty and Thilo Deckersbach Supervised and guided undergraduate interns on fMRI and neuropsychiatric research. Tutor, Brown University, Providence RI, Perception and Mind, Dr. 2008-2009 Michael Tarr Met individually with students weekly to go over course material and advise on study techniques and projects. Tutor, The Mather Elementary School, Dorchester, MA 2002-2005 Provided weekly after-school homework help, guidance, and mentorship. Brown University Neuroscience Graduate Program Service Neuroscience Graduate Program Mentor 2017-2018 Provides guidance to a first-year neuroscience graduate student on selecting advisors, identifying specific areas of interest, and addressing concerns and questions. Elected Neuroscience Graduate Student Representative 2014-2015 Planned and organized the Neuroscience Graduate Program recruitment, participated as the graduate student admissions committee member, acted as a student-faculty liaison, coordinated the Brown Neuroscience Graduate Program booth at the Society for Neuroscience graduate program fair, and arranged other events, such as the Brown Neuroscience Graduate Program’s Society for Neuroscience social. Student-Selected Speaker Representative 2013-2014 Invited and arranged two scientists to speak at Brown for the “Proseminar Speaker Series” and participated as a graduate student committee member. Social Coordinator 2013-2014 Organized and planned monthly laboratory socials. Professional Development Harvard Medical School Continuing Medical Education 2017 Neurorehabilitation 2017 Course Attended and participated in a three-day neurorehabilitation course on topics including rehabilitative and assistive tools for patients with paralysis, aphasia, visual disorders, and cognitive impairment. ix The Harriet W. Sheridan Center for Teaching and Learning, Brown University Certificate I: The Sheridan Center Teaching Seminar – Reflective Teaching 2012 Learned and rehearsed teaching strategies and techniques to maximize productive and efficient learning by designing syllabi and courses, attending to different learning styles, and providing and receiving teaching feedback. Certificate I Discussion Leader 2013 Led group discussions to facilitate conversation, engage students in group activities and promote sharing of opinions and ideas relating to thoughtful teaching and learning practice. Certificate IV: The Teaching Consultant Program Conducted teaching observations for members of the Brown community and provided written and face-to-face feedback. Fellowships/Awards Neuroscience Scholars Program Fellow, Society for Neuroscience, 2015-2017 Washington D.C. Selected as one of 15 graduate students and post-doctorate scientists to participate in a two-year training program providing a $1,500 professional development travel award, two-year SfN membership and travel reimbursement, and a scientific mentoring committee. Brown Office of Graduate and Post-Doctorate Studies Conference Travel 2017 Award Received travel funding to present my research in a nanosymposium at the Society for Neuroscience in Washington D.C. Brown Neuroscience Graduate Program Annual Retreat Poster Award, 2013 Bristol, RI Presented a poster on synaptic stabilization at the neuromuscular junction and was awarded for my research and clarity of presentation. Brown Internship Award Program, Providence, RI 2008-2009 Awarded funding for my internship at Boston University School of Medicine in Dr. Carmela Abraham’s laboratory. The Winsor School Madras Science Prize, Boston, MA 2005 Selected as one of two seniors for my passion for science and pursuance of scientific opportunities throughout high school. Additional Community Services Boston Walk to Defeat ALS 2012-2017 “Innovate to Innervate” Team Captain and Founder Operation Happy Birthday 2007 Mentored adolescent boys from a group home. x PREFACE Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease of the upper and lower motor neurons resulting in weakness, paralysis and death within two to five years. Frontal temporal dementia (FTD) results in neurodegeneration of the frontal, temporal and insular cortices causing behavioral and cognitive deficits. ALS and FTD exist along a spectrum (ALS-FTD). Characterizing early neurodegenerative disease phenotypes is critical in defining epochs of time during which the nervous system is most plastic and receptive to intervention. Such early events reveal pathways that are deranged prior to the onset of neurodegeneration as well as biomarkers indicative of disease. This information offers insight that allows for earlier diagnosis and the identification of new therapeutic targets, providing an opportunity for early intervention. Because neurodegeneration is a late-stage phenomenon, many contributing pathologies are already present. This is particularly a problem in fast-progressing diseases including ALS and FTD. To identify early phenotypes in ALS and FTD, we used a novel deep behavioral phenotyping technology, Automated Continuous Behavioral Monitoring (ACBM), to identify and quantify the behavior of the transgenic SOD1G93A mouse (B6.Cg-Tg(SOD1*G93A)1Gur/J), as well as the TDP- 43Q331K knock-in mouse, a novel model of ALS-FTD. ACBM collects video data of individually housed mice over five days. Position, velocity, acceleration and motion features are extracted from the data and classified through a Support Vector Machine Hidden Markov Model (SVMHMM) as one of nine pre-defined behaviors. xi We first demonstrated that SOD1G93A mice exhibit neuromotor deficits as early as P30 (Chapter 2). To determine whether such deficits are present in the TDP-43Q331K mouse, we conducted five rounds of ACBM and found TDP-43Q331K deficits as early as P120 (earliest age measured) (Chapter 3). Lastly, to assess earlier phenotypes, we conducted ACBM in a second younger cohort of TDP-43Q331K mice at P30 and P120, and found gender-dependent phenotypes at both ages (Chapter 4). This work presents the earliest behavioral phenotypes identified in the SOD1G93A and TDP-43Q331K mice. ACBM can be used in the future to assess treatment efficacy of pharmacological, genetic or environmental approaches in mouse models of neurodegenerative disease. xii ACKNOWLEDGEMENTS My time in the neuroscience graduate program at Brown has been an invaluable experience for me, and I am incredibly grateful for the support that I have received over the past few years from my forever neuroscience “family”. Firstly, I would like to thank my mentor, Dr. Justin Fallon for his invaluable guidance and care. Ever since first joining his lab (actually, even before graduate school as an undergrad!) I have truly looked forward to my meetings with Dr. Fallon. He has helped me to approach every experiment with an open my mind. He has taught me to explore a myriad of techniques from protein binding assays to higher- level behavioral work. Additionally, he has remained dedicated to me and to my development as a scientist, and I am incredibly grateful to him. I would also like to thank Beth McKechnie for her guidance and advice as well as constant encouragement, support, and unwavering dedication. She has always been there for me, and I will always appreciate everything she has done. Carolyn Ortega and Sarah Killeen have not only taught me many laboratory skills, but they also were able to continue a project for me when I was recovering from surgery. This project ended up turning into a huge portion of my thesis, and I could not have done it without them. They have also become incredible friends. I would also like to thank all other members of the Fallon Lab: Dr. Alison Amenta, Diego Jaime, Lauren Fish, Laura Madigan, and Johnny Page. I am grateful to Dr. Kevin Bath for teaching me multiple rodent behavioral testing techniques and taking the time and effort to help me with data analysis. xiii Dr. Thomas Serre developed the ACBM technology used in my thesis, and I am incredibly grateful for his help with data analysis and developing ideas related to behavioral classifiers. I am also grateful to Vijay Veerabadran and Tarun Sharma – two talented research assistants in Dr. Serre’s laboratory. I would like to thank my committee, Dr. Gilad Barnea, Dr. Kevin Bath, and Dr. Anne Hart for contributing guidance and advice throughout all stages of my graduate work. To Mom, Dad, Grammy, Grampa, Emily and Dan: Thank you for always believing in me no matter what. You have inspired me everyday, and have filled my life with all the wonderful little moments that matter most. And lastly, thank you Max for being my biggest cheerleader and solid rock through all of the excitement and adventures of the past few years. Your constant love and support inspires me and you will always mean the universe to me. xiv Dedication I am dedicating my thesis to my very first mentor, Dr. David Hubel. Dr. Hubel was my neighbor since I was three years old. He was always full of joyful and thoughtful curiosity, and was tremendously passionate about all of the special little things that life has to offer - the wonders of astronomy, the fun that can come from building your own radio, weaving on his loom, and playing Scarlatti and Bach on the piano and flute. He not only instilled in me intense curiosity and a love and passion for science, but also an understanding that every person has an opportunity to have a positive influence on the world. Thank you, Dr. Hubel for teaching me that there is no such thing as growing out of wonder. xv TABLE OF CONTENTS CHAPTER 1: INTRODUCTION p. 2 CHAPTER 2: BEHAVIORAL CHARACTERIZATION OF THE B6.CG- p. 62 TG(SOD1*G93A)1GUR/J ALS MOUSE MODEL THROUGH AUTOMATED CONTINUOUS BEHAVIORAL MONITORING CHAPTER 3: TDP-43 GAINS FUNCTION DUE TO PERTURBED p. 97 AUTOREGULATION IN A TARDBP KNOCK-IN MOUSE MODEL OF ALS-FTD CHAPTER 4: BEHAVIORAL CHARACTERIZATION OF THE NOVEL TDP- p. 156 43Q331K KNOCK-IN MOUSE MODEL OF ALS-FTD THROUGH AUTOMATED CONTINUOUS BEHAVIORAL MONITORING CHAPTER 5: DISCUSSION p. 183 WORKS CITED p. 197 xvi LIST OF TABLES, FIGURES, AND ILLUSTRATIONS CHAPTER 1 Figure 1. Automated Continuous Behavioral Monitoring: p. 41 Mouse Behaviors. Figure 2. Video Pre-Processing Background subtraction. p. 43 Figure 3. Sub-Window Extraction. p. 43 Figure 4. Position, Velocity- and Acceleration-Based Feature p. 45 Extraction. Figure 5. Normalization of Position, Velocity, and p. 46 Acceleration-based features. Figure 6a. Motion Feature Extraction and Template p. 47 Matching. Figure 6b. HMAX Template Matching. p. 48 Figure 7a. SVM-HMM Classifier: Characterizing similarity p. 52 among features within sequences of frames. Figure 7b. SVM-HMM Classifier: Characterizing similarity p. 53 among features between sequences. CHAPTER 2 Figure 1. Schematic of ACBM feature extraction and p. 66 behavioral classification. xvii Figure 2. ACBM detects SOD1G93A early and transient walk p. 69 and translocation deficit at P30 and eating-on-haunches deficit P58. Figure 3. Separating x- and y-translocation demonstrates p. 73 that the overall translocation can be driven by movement in just one direction. Supplementary Figure 1 ACBM detects SOD1G93A early walk p. 89 deficit. Supplementary Figure 2. Rescaled Y-Axis: ACBM detects p. 90 SOD1G93A early and transient walk and translocation deficit at P30 and eating-on-haunches deficit P58. Supplementary Figure 3. SOD1G93A mice exhibit an increase p. 91 in drinking behavior at P58. Supplementary Figure 4. Rescaled Y-Axis: SOD1G93A mice p. 91 exhibit an increase in drinking behavior at P58. Supplementary Figure 5. Eat-from-hopper, hang, and rest p. 92 behaviors did not exhibit a phenotype across all three ages examined. Supplementary Figure 6. Re-scaled Y-Axis: Separating x- and p. 93 y-translocation demonstrates that the overall translocation can be driven by movement in just one direction. Supplementary Figure 7. There is an increase in SOD1G93A y- p. 94 xviii direction translocation at P58 during late hours of darkness. Supplementary Table 1. Summary of SOD1G93A ACBM p. 94 Phenotypes at P30, P58 and P86. CHAPTER 3 Figure 1. CRISPR mutagenesis, ACBM characterization p. 133 and breeding ratios of TDP-43Q331K mice. Figure 2. Motor impairment, hyperphagia and spinal motor p. 134 neuronal transcriptomic changes in mutant mice. Figure 3. Cognitive testing indicates executive dysfunction, p. 135 memory impairment and phenotypic heterogeneity in mutant mice. Figure 4. Perturbed TDP-43 autoregulation and loss of p. 136 parvalbumin interneurons in mutant mice. Figure 5. Splicing analysis indicates TDP-43 p. 137 misregulation, a gain of TDP-43 function and altered Mapt exon 2/3 splicing. Figure 6. TDP-43 misregulation occurs in spinal cords of p. 138 mutant mice, but not in motor neurons. Figure 7. Phenotypic stratification of transcriptomic data p. 139 from mutant mice allows the identification of putative disease modifiers. Figure 8. TDP-43Q331K mice demonstrate age-related p. 140 xix deterioration in cortical transcriptomes with altered expression of multiple ALS-linked genes. Supplementary Figure 1. ACBM walking phenotypes p. 141 Supplementary Figure 2. Neuromuscular investigations p. 142 Supplementary Figure 3. Laser capture and RNA sequencing p. 144 analysis of lumbar spinal motor neurons Supplementary Figure 4. Additional behavioral outcomes p. 145 from five-choice serial reaction time tasks and marble burying assays Supplementary Figure 5. Frontal cortical histology p. 146 Supplementary Figure 6. Frontal cortical RNA-seq, tau p. 147 staining and validation in line 3 mice Supplementary Figure 7. RNA-seq in aged mice p. 149 Supplementary Table 2. Myelination and oligodendrocyte p. 150 associated genes Supplementary Table 3. CRISPR sequences. P .151 Supplementary Table 4. Genotyping protocol p. 152 Supplementary Table 5. Primary antibodies used for p. 153 immunofluorescence and immunoblotting Supplementary Table 6. Primers for quantitative PCR p. 154 CHAPTER 4 Figure 1. ACBM detects a TDP-43Q331K walk deficit at P30 p. 164 xx and a walk and translocation deficit at P120 in females. Figure 2. ACBM detects a TDP-43Q331K walk deficit at P120 p. 165 in males. Figure 3. Separating x- and y-translocation demonstrates p. 168 that the overall decrease in female TDP-43Q331K translocation is driven by movement in both directions. Figure 4. Female TDP-43Q331K mice exhibit a rear deficit at p. 169 p120. Supplementary Figure 1. ACBM detects a TDP-43Q331K walk p. 180 deficit at P30 and a walk and translocation deficit at P120 in females (Rescaled Y-axes). Supplementary Figure 2. ACBM detects a TDP-43Q331K walk p. 181 deficit at P120 in males (Rescaled Y-axes). xxi CHAPTER 1: INTRODUCTION 1 Early Diagnosis of Neurodegenerative Disease Early diagnosis of neurodegenerative disease remains a challenge because early symptoms are often subtle, and patients may not receive a diagnosis or treatment until disease progression is underway. The benefit of early diagnosis is early treatment, which could slow, stop or reverse disease progression, dependent on the disease and disease stage. Early diagnosis would be particularly important for patients with a family history of the neurodegenerative disease as these patients are at risk for disease development. While neurodegenerative diseases are progressive, there have been advances in early diagnosis and treatment for some diseases that have prolonged function. For example, the characterization of mild cognitive impairment (MCI), which is often a precursor to Alzheimer’s disease (progression from MCI to AD is 10-15% pear year), improved diagnostic criteria for dementia [1]. With these criteria, it is now possible to differentiate MCI patients from those with dementia and those who are cognitively normal [2]. Patients who are diagnosed early can then be tested for genetic predispositions to dementia such as the presence of the apolipoprotein-ε4 allele, which is associated with progression of MCI to dementia [3]. Early pre- dementia MCI diagnosis enables earlier intervention and follow-up care. Interventions such as cognitive training, physical activity and healthy diet are protective against MCI as well as some pharmacological interventions, such as piribedil, which can improve cognition [4] [5] [6] [7]. A neurodegenerative disease that does not exhibit symptomology before neurodegeneration has initiated and is underway is amyotrophic lateral sclerosis 2 (ALS). ALS symptom onset and progression in humans is abrupt and rapid - by the time most patients seek treatment and are diagnosed with ALS they are already symptomatic and pharmacological intervention is often too late to slow disease. Therefore, diagnosing ALS early, prior to the onset of obvious symptoms, is critical in the treatment of the disease. Once this early window is identified, early drug intervention targeting specific disease-related biomarkers could be implemented to slow disease progression. Amyotrophic Lateral Sclerosis Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease of cortical (upper) and spinal (lower) motor neurons (UMNs and LMNs, respectively) that results in weakness, paralysis, and eventually death within approximately 2-5 years [8]. The average age of ALS diagnosis is approximately 60 years [9]. There is currently no treatment for ALS, and the primary FDA-approved drug, riluzole, extends life by only a few months. ALS Symptomology Early clinical signs of ALS include mild and generally painless weakness in one or more regions of the body, most commonly identified in the extremities, as well as dysphagia (swallowing difficulties) and dysarthria (slurred speech) [10]. By the time these symptoms have emerged disease progression is underway. As the disease progresses, symptoms worsen, and mixed spastic-flaccid symptoms develop. Spastic symptoms include pseudobulbar palsy (lack of facial 3 movement control resulting from damage to the upper motor neurons), muscle weakness, hyperreflexia, and the Babinski sign. Flaccidity is demonstrated through muscle weakness and atrophy. Mixed spastic-flaccid type dysarthria can result in speech deficits such as effortful speech, including primarily short phrases, inappropriate pauses, hypernasality, strained voice, and decreased loudness and pitch [9]. The majority of ALS cases (70%) are classified as limb-onset. Approximately 25% of cases are classified as bulbar onset, which can result from pathology of upper and lower motor neurons. Symptoms of upper motor neuron bulbar onset ALS include slow speech, jaw jerking movements and slow tongue movements. Symptoms of lower motor neuron bulbar onset ALS include facial weakness, swallowing and speech difficulties due to lower motor degeneration in the brainstem [9, 11]. 90% of ALS is sporadic (sALS) as there is currently no evidence of genetic predisposition. 10% of ALS is familial (fALS), and is linked to specific genes within a family. While ALS can occur at any age, the average age of diagnosis is 60 years [12]. Symptomology presents differentially depending on whether neurodegeneration is occurring primarily in UMNs or LMNs. Degeneration of UMNs typically results in spasticity, weakness and increased reflexes while muscle mass is not affected [13]. Degeneration of LMNs results in twitching, weakness, decreased motor reflexes, and muscle atrophy [13]. Following denervation, compensatory reinnervation of motor neurons occurs, however this reinnervation is unable to sustain function. 4 ALS Diagnosis Diagnosing ALS consists of the identification of electrophysiological changes using the El Escorial and Awaji criteria [14] [15] as well as functional changes using the ALS functional-rating scale [16]. Electrodiagnosis is used to rule out other causes of similar symptomology through nerve conduction studies (NCS), which could detect other disease-associated abnormalities not present in ALS, such as demyelination or abnormalities in the conduction of sensory nerves, which are not affected in ALS [17]. To detect abnormalities specific to ALS, needle electromyography (EMG) is used to detect acute and chronic denervation and compensatory reinnervation [18]. Needle EMG will detect spontaneous muscle fibrillations at rest, which reflect depolarization of denervated muscle fibers. Additionally, needle EMG will identify spontaneous discharge of single motor units, which is considered a hallmark of ALS. Compensatory reinnervation can also be detected with needle EMG, and provides diagnostic information related to abnormal motor unit potentials (MUPs), such as increased MUP duration, which reflects axonal collateral sprouting [18]. In addition to electrodiagnosis, patients are evaluated using the (ALS-FRS), which rates the following measures: speech, salivation, swallowing, handwriting, cutting food and handling utensils, dressing and hygiene, turning in bed and adjusting bed clothes, walking, climbing stairs, and breathing [19] [20]. This scale can be used as a diagnostic measure, and also be used to follow a patient’s disease progression longitudinally. 5 ALS Treatment There are two FDA-approved pharmacological treatments for ALS. Riluzole (Rilutek) extends lifespan for only a few months. Its neuroprotective mechanisms of action are not fully understood. It was originally used as an anti-convulsant, however after the identification of its neuroprotective mechanisms, a clinical trial was performed on 155 ALS patients in France and Belgium, and was approved by the U.S. Food and Drug Administration in 1994 [21]. While no one specific mechanism of action has been identified for riluzole, it likely provides benefit through a number of pathways. One mechanism of action is decreasing excitotoxicity. Riluzole lowers levels of synaptic cleft glutamate and other excitatory neurotransmitters, and it prevents glutamate-induced motor neuron degeneration in vitro (Cheah, 2010). Riluzole may block voltage-gated calcium channels. These channels would naturally release calcium into the pre-synaptic motor neuron, which would cause the release of glutamate into the synaptic cleft. Blocking such channels would therefore decrease overall excitation, and potential excitotoxicity. Additionally, riluzole may block sodium channels, also decreasing neuronal excitation. Riluzole decreased sodium current in an SOD1G93A transgenic mouse model by 46%, suggesting that a mechanism of action of riluzole decreases motor neuron excitation [22] [23]. Despite potential neuroprotective mechanisms, the clinical benefits of riluzole are minimal as life is only extended by 2-3 months, and muscle strength, limb and bulbar function are only minimally improved [24]. Because ALS patients are generally not diagnosed until neurodegeneration is underway, it is possible that 6 earlier riluzole intervention, prior to obvious symptomology, could prolong function. Edaravone (Radicava), primarily used to treat stroke, was recently demonstrated to improve function in Japanese ALS patients over a six-month placebo controlled clinical trial. In comparison to patients receiving placebo, early stage patients who received Edaravone over six months demonstrated a significant improvement on the ALS Functional Rating Scale [25]. Edaravone did not provide benefit to patients with more advanced disease progression demonstrating the importance of early treatment [25]. Gender Differences in ALS There are gender differences in the incidence and prevalence of ALS. Incidence represents the risk of developing ALS in a given period of time. Prevalence represents, at a given time, the number of ALS cases within a population. For example, if a vaccine for a disease is developed in a given year, but does not cure patients who already have the disease, then directly following the vaccine discovery, the disease incidence (risk of developing the disease) decreases, but the prevalence (the proportion or number of people in the population who have the disease at that time) will not drastically change. The annual incidence of ALS in men is approximately 3/100,000 people per year, and the incidence in women is approximately 2.4/100,000 people per year [26]. The lifetime risk of ALS in men is 1 7 in 350, while the lifetime risk in women is 1 in 472 [27] [28]. Across all ALS patients, men have a higher risk of developing ALS [29] [30, 31]. Additionally, men exhibit a younger average age of ALS disease onset than women. The incidence ratio of men to woman is 1.3 and this ratio increases with age after 30 years of age until it peaks during the seventh and eighth decades, likely due to the death rate increasing in male patients over that age [26]. The prevalence reflects not only the incidence but also survival length in patients, and similarly to incidence, the prevalence of ALS is greater in men than women [32]. There are also gender differences in the site of onset (limb and bulbar onset). There is a greater percentage of limb-onset ALS patients overall compared to bulbar onset patients in both men and women. However, within the bulbar onset patient population, there is a larger percentage of women compared to men [32]. Gender has a differential effect on familial ALS (fALS) patients compared to sporadic ALS (sALS) patients. In fALS patients, the gender ratio of males to females is 1:1. Given that many genes associated with ALS are autosomal dominant, this result is predicted [32]. One possible reason explaining the earlier onset and higher prevalence of ALS in males could have to do with neuroprotective estrogen effects. In the most commonly used ALS mouse model, the SOD1G93A overexpression transgenic mouse, estrogen therapy in males reverses phenotypes associated with castration [33] [32]. 8 ALS Disease Progression One of the earliest abnormalities during ALS disease progression is an axonopathy where there is a dying back of the motor neuron from the muscle (denervation), resulting in progressive pathologies [34]. Denervation has been demonstrated in animal models, human patients (through electrophysiological measurements as described previously) as well as in post-mortem human patients [34]. Denervation occurs at different time points depending on the muscle type being innervated. Fast fatigable motor units such as those affiliated with the tibialus anterior (TA) are more susceptible to denervation than slow motor units such as those affiliated with the soleus muscle [35]. In a transgenic SOD1G93A mouse model, a commonly used mouse model to study ALS, Pun et al. (2006) demonstrated early denervation of hind limb muscles [35]. They also found that different motor neuron types exhibit selective vulnerability to denervation, as well as selective compensatory axonal sprouting. Within the lateral gastrocnemius (LGC) and tibialis anterior (TA) muscles, they found that fast fatigable (FF) motor neurons innervating type IIb muscle fibers, are most vulnerable to denervation, followed by type IIa muscle fibers innervated by fast fatigue-resistant (FR) motor neurons. Within the soleus, slow (S) motor neurons innervating type I muscle fibers are resistant to denervation. They demonstrated that the LGC and TA muscles exhibit denervation of the FF motor neurons at P48-P52. Denervated synapses were re-innervated through compensatory sprouting of nearby axon collaterals of FR and S motor neuron axons at P60. Re-innervated muscle fibers are converted to the muscle fiber type of the 9 innervating neuron, and this process results in muscle fiber grouping. These re- innervated synapses again undergo denervation between P80 and P90. Pun et al demonstrated that within the hind limb muscles examined, most of the remaining synapses (within the three muscles studied) contained type I muscle fibers within the soleus muscle [35]. Kaplan et al (2014) demonstrated further corroborating data demonstrating early denervation in an SOD1G93A transgenic mouse model that exhibits a longer lifespan (~p150) than that used by Pun et al. (~P120) [36]. They found that denervation is present within the TA at P50 and within the soleus at P100 again demonstrating that different muscles exhibit different levels of vulnerability related to their innervating motor neurons. Even earlier pathology is demonstrated by Frey et al. (2000) in which SOD1G93A transgenic mice exhibit axonal thinning in the peripheral branches of the medial gastrocnemius muscle as early as P30, followed by dramatic denervation between P45 and P80. Axonal thinning is therefore one of the initiating stages of disease progression in this ALS mouse model [37]. Additionally, Pun et al. (2006) demonstrated early axonal transport abnormalities in FF and FR motor neurons within the gastrocnemius muscle through the identification of synaptic vesicle accumulations present as early as P35- P38 [35]. To further delve into the early disease mechanisms that could result in these behavioral phenotypes, one could probe molecular differences within the SOD1G93A transgenic mice. A specific protein that is present only in motor neurons that 10 denervate early could convey motor neuron susceptibility. Kaplan et al. (2014) demonstrated that metalloproteinase-9 (MMP-9) is absent in S motor neurons of the spinal cord and is selectively expressed in FF motor neurons [36]. Additionally, in the SOD1G93A mice that lack MMP-9, denervation was not detected until p100, suggesting that MMP-9 confers vulnerability [36]. These results suggest that MMP-9 within motor neurons could be a potential target for pharmacological therapy development. During denervation as well as during reinnervation, compensatory motor neuron axonal sprouting occurs in which the nerve terminals that remain in the muscle extend distal sprouts that improve connectivity between the motor neuron and the muscle. Sprouting results when there is a lack of activation from the presynaptic motor neuron indicating partial denervation. Glial cells play an important role in sprouting. Terminal schwann cells (TSCs) present at the NMJ extend “bridges” that grow from the denervated end-plates to the terminal sprouts, and thus guide the sprouts to the denervated muscle resulting in the formation of new nerve terminals. This process occurs following denervation as well as during reinnervation, and is dependent on a lack of pre-synaptic stimulation [38] [39]. However, as the disease progresses, motor neuron sprouting can no longer compensate, and further lack of post-synaptic trophic factors, such as glial cell- derived neurotrophic factor (GDNF), from the muscle, results in eventual motor neuron degeneration [40]. Therefore, disruption of earlier pathological events will likely be most successful in disease treatment. 11 Characterizing such early events offers several advantages including identifying the pathways that are deranged prior to the onset of neurodegeneration. Once specific pathways are identified related to ALS-associated pathology, patients who exhibit pre-symptomatic abnormalities in such pathways could receive an earlier diagnosis. An early diagnosis would enable patients to receive close monitoring, to start existing therapies, and to enroll in clinical trials. Additionally, early diagnosis provides an opportunity to patients and their families to make decisions and future plans earlier than had they waited for obvious symptomology to present. As early deranged pathways are identified and patients are able to receive early diagnoses, efforts can go into identifying the biomarkers associated with such deranged pathways such as the identification of suppressor genes. Perturbation of suppression genes may result in suppression of disease phenotypes. Once these suppressor genes are identified and proven to modify disease phenotypes, they can be pharmacologically targeted early, prior to the onset of symptoms with the goal of prolonging function. Additionally, identification of biomarkers in patients provides an opportunity to treat these patients prophylactically with existing drugs, especially patients with a familial susceptibility. Frontal Temporal Dementia Frontal temporal dementia (FTD), the third most common form of dementia in patients over 65, results from neurodegeneration of the frontal, insular and anterior temporal lobes [41]. As a comparison to Alzheimer’s disease (AD), patients 12 with FTD have a decreased lifespan and faster disease progression. One study demonstrates that FTD patients survive 4.2 years after onset while AD patients survive over 6 years after onset [42]. The prevalence of FTD is 1 in 742, and the incidence is 1.61/100,000 [43]. Symptomology of FTD Variants There are different variants of FTD resulting in specific behavioral and language impairments. Like ALS, by the time patients begin to exhibit symptoms, neurodegeneration is underway and intervention is minimally effective. 50% of FTD patients have a behavioral variant of FTD (bvFTD), which results in progressive social, emotional, personality and conduct abnormalities [41]. bvFTD can result in two separate syndromes. Apathetic-type patients exhibit, decreased motivation, isolation, and loss of socio-emotional awareness. Disinhibited-type patients exhibit hyper-orality, preference for sweet foods, motor stereotypies and perseverative behavior. These patients are often impulsive, disinhibited and socially inappropriate [41] [44]. 20-25% of FTD patients exhibit semantic variant FTD (svFTD) [41]. svFTD patients exhibit fluent language however there is a progressive loss of object knowledge, and they demonstrate paraphasic (adding incorrect syllables and words during speech effort), and semantic (using words with incorrect meanings) errors [41]. These patients may struggle with articulating differences between objects, such as differences between types of trees, followed by loss of knowledge to differentiate between types of plants, followed by loss of understanding of what 13 plants are. Many patients with FTD are also more rigid in their behaviors such as restricting diet to only specific foods and can develop eating disorders [41]. 20-25% of FTD patients exhibit non-fluent variant FTD (nfvFTD), in which patients struggle with motor control related to speech resulting in inappropriate pauses even within utterances [41]. Disease progression can result in patients becoming mute, and it is possible that comprehension may become impaired, although the primary symptom is apraxia of speech. The different variants of FTD are also driven by degeneration of specific brain areas. Patients with bvFTD have bilateral degeneration in their frontal lobe. Patients with svFTD exhibit bilateral degeneration in the anterior temporal lobe, and nfvFTD exhibits degeneration in the left inferior frontal cortex and insular regions. FTD patients also exhibit gender differences within the FTD variants. bvFTD and svPPA are more prominent in males than females, while nfvPPA is more prominent in females [41]. FTD Pathologies There are underlying pathologies associated with FTD syndromes. Transactive response DNA binding protein 43 (TDP-43) is a DNA/RNA binding protein, and its abnormal accumulation accounts for over 50% of FTD cases [41] [45]. TDP-43 associates with ubiquitin, and TDP-43 proteinopathy can result in dysmorphic neurites (DN) and either neuronal cytoplasmic inclusions (NCIs) or neuronal intranuclear inclusions (NIIs) [45]. FTD with TDP-43 inclusions can be characterized based on the type of inclusions and where the inclusions exist – 1.) 14 those lacking NIIs, few NCIs, and containing dystrophic neurites; 2.) those with many NCIs, few DNs, and lacking NCII (intranuclear inclusion that occurs in FTD- MND); 3.) those with many DNs and NCIs, and some NIIs; and 4.) those with many DNs and NIIs, and few NCIs [45]. Tau is a microtubule binding protein, and its abnormal accumulation accounts for approximately 40% of FTD cases, which are referred to as FTLD-tau. These patients do not exhibit TDP-43 aggregates. FTLD-tau cases are due to microtubule associated protein tau (MAPT) mutations, which results in impaired binding of tau microtubules. This impaired binding causes hyperphosphorylated tau aggregates that form neurofibrillary tangles [45]. A minority of FTD cases (10-20%) have fused in sarcoma (FUS) aggregates and do not harbor TDP-43 nor tau aggregates, again suggesting mutually distinct pathological pathways [41]. The majority of bvFTD and svPPA cases are tau-negative and TDP-43 positive, and the majority of nfvFTD cases are tau-positive [41]. As there are currently no treatments that reverse FTD, treatment is directed toward decreasing symptomology, as well as the impact of symptoms on the patients’ and families’ lives. The goals of many interventions are to reduce behavioral as well as executive functioning deficits. Antidepressants, such as SSRIs and trazodone (SARI), have proven useful in modestly improving behavioral deficits. To treat aggression, neuroleptics such as olanzapine have proven to be effective [41]. 15 FTD Disease Progression Like ALS, FTD is progressive. Early symptoms may start with mild behavioral deficits which progress, often until institutional care is necessary. There is a great degree of variability in survival among FTD patients. Those with comorbidities such as those that also have motor neuron disease exhibit the shortest survival [46]. Often, symptoms that start as one specific FTD variant converge with other variants as the disease progresses. For example, over time patients that present with bvFTD may develop other deficits such as motor neuron disease [47]. During end-stage FTD, patients may require full-time care as they will develop more severe hazardous symptoms, such as dysphagia or wandering [44]. ALS-FTD ALS and FTD exist along a spectrum – each disease representing a bookend. 15% of ALS patients are also diagnosed with FTD and 15% of FTD patients are also diagnosed with ALS. However, 50% of ALS patients exhibit cognitive deficits and 40% FTD patients exhibit motor neuron disease [48]. Patients within the ALS-FTD spectrum survive on average for 3-5 years [44]. Ubiquitin positive aggregates that associate with TDP-43 are found in a large proportion of ALS and FTD patients, providing molecular evidence that these diseases exist along a spectrum [45]. 16 Genetic Predisposition There are many genes that are associated with FTD, and approximately one third of FTD patients have relatives who also have the disease. ~10-20% of FTD cases exhibit autosomal dominant inheritance [44, 45]. Genetics of ALS, FTD and ALS-FTD SOD1 One of the most commonly studied genes associated with ALS is the superoxide dismutase 1 gene, SOD1, which is 153 amino acids, and associated with over 170 different mutations that can result in ALS [48]. It translates into the SOD1 protein, which is a 32 kDa homodimer. SOD1 is a cytoplasmic antioxidant enzyme, and each monomer binds one copper and one zinc ion, which bind and metabolize toxic superoxide radicals into hydrogen peroxide and oxygen. The positively charged copper and zinc reaction sites enable interaction with the negatively charged superoxide ions. Hydrophobic interactions stabilize the dimer interactions. SOD1 pathology can include Lewy-body-like hyaline inclusions, ubiquitin, and mutated and wild-type SOD1 inclusions. Additionally, neurofibrillary tangles or SOD1 intracellular inclusions can also be found in cases of fALS patients [11, 44]. While these biomarkers have been identified in ALS patients, they are not consistent among the majority of patients, and the identification of specific SOD1 biomarkers have not been verified for ALS diagnosis. MicroRNA research suggests that specific miRNA levels in ALS patients may indicate disease. One study suggest that miRNA- 206 is increased in the blood of ALS patients, and interestingly this result was 17 observed in both fALS and sALS, suggesting the use of miRNAs as a potential diagnosis criteria, yet more research needs to be done for verification [11]. There are over 150 primarily missense SOD1 mutations, and these account for ~20% of fALS, ~1% sALS, and approximately 2% of all ALS cases [11] [45]. Toxicity from SOD1 mutations can disrupt cellular functions including mitochondrial apoptotic signaling function, protein misfolding, axonal transport, and endoplasmic reticulum stress. ALS patients with SOD1 mutations typically present with weakness in a single limb that can last for months prior to significant functional consequences or muscle atrophy. Most commonly, lower motor neurons are primarily affected with early symptoms including mild calf weakness and loss of Achilles reflex [49]. The mean age of disease onset in patients with SOD1 mutations is 47 +/- 13 years, and extremity onset is more common than bulbar onset [49]. Different SOD1 mutations can impact the length and severity of disease progression. Patients with the SOD1-A4V mutation exhibit markedly fast disease progression and a reduced lifespan of 1+/- 0.4 years after onset in comparison to other mutations, such as patients with the SOD1-D90A, which can exhibit disease duration of over ~13 years [49] [11] [50]. The SOD1G93A missense mutation was the first mutation modeled in mice [51]. In human patients with this mutation, the age of disease onset is approximately 47.9 +/- 17.7 years, and disease duration is 2.2 +/- 1.5 years [52]. 18 TDP-43 The transactive response RNA/DNA binding protein 43 is a 414 amino acid RNA binding protein that regulates transcription (including its own), splicing, miRNA biogenesis and mRNA transport. It contains two RNA recognition motifs as well as a carboxy-terminal glycine-rich domain [53]. It is found primarily in the nucleus and is ubiquitously expressed [54]. Under conditions of stress, it translocates from the nucleus to the cytoplasm. TDP-43 is encoded by TARDBP DNA, and virtually all ALS patients and 40% of FTD patients exhibit cytosoloic TDP-43 aggregates. TDP-43 biomarkers for pathology include elevated TDP-43 plasma and CSF levels in ALS patients, and these levels are comparable between sALS and fALS cases [11]. Over 40 mutations in TARDBP account for 4% of fALS patients, 1% of sALS and 1% of FTD patients [55] [48]. These mutations are located primarily in the C- terminal glycine-rich RNA binding domain, which is important for protein-protein interactions [48]. Mutations in the RNA binding domain of TDP-43 result in a loss of function causing TDP-43 to mislocalize from the nucleus to the cytoplasm. The depletion of TDP-43 from the nucleus can interfere with downstream processes such as splicing, and as TDP-43 levels increase in the cytoplasm, it binds mRNA molecules that are not fully processed, preventing their normal function. As TDP-43 accumulates in the cytoplasm, pathological forms of TDP-43 further propagate its mislocalization and depletion from the nucleus [48]. ALS disease onset in patients with TDP-43 mutations is variable: from 20-77 years old (mean age: 54), with the M337V and G348C mutations exhibiting the 19 earliest disease onset. Symptoms typically manifest in the upper limbs (~60-70% of patients) in these patients [11]. C9orf72 The GGGGCC hexanucleotide repeat in a non-coding region of C9orf72 is of unknown function. It is the most common genetic cause found in ALS, FTD and ALS- FTD patients - 40% of fALS cases, 25% of fFTD cases, 5-25% of sALS and 6% of sFTD cases result from these expansions [48]. Healthy people exhibit up to 30 of these repeats of unknown function, while patients can exhibit hundreds to thousands. High levels of C9orf72 mRNA and protein exist in the brains of ALS patients as well as in ALS mouse models [11]. Interestingly, while there is a correlation between the number of these repeats and disease duration in FTD patients, there does not appear to be any such correlation in ALS patients. There is evidence that there is a higher prevalence of bulbar onset C9ALS patients than non-C9ALS patients as well as a higher prevalence of bulbar onset in FTD patients [11]. Some studies suggest that bulbar onset is more common in C9orf72 ALS patients than ALS patients without the C9orf72 repeat expansion [11]. Despite the fact that C9orf72 is implicated as the most common cause of ALS, little is known about the mechanisms underlying its toxicity. Some studies suggest that ALS/FTD patients with the C9orf72 expansion have lower levels of C9orf72 mRNA implicating a possible haploinsufficiency and loss of function [56]. Additionally, mice lacking C9orf72 function exhibit autoimmune phenotypes however no degeneration of motor neurons [48]. 20 There is also evidence that gain of function resulting from the C9orf72 repeats may result in pathology. It has been demonstrated that accumulation of mRNAs carrying the repeat expansion occurs in nuclear foci, which may represent a potential biomarker. These accumulations interfere with RNA binding proteins and this results in perturbation of mRNA maturation and splicing [48]. Additionally, the discovery of repeat associated non-ATG (RAN) translation may explain C9orf72 toxicity resulting from gain of function. Microsatellite sense and antisense expansions within C9orf72 can interfere with protein synthesis and generate aggregation-prone disruptive dipeptide repeats (DPRs), and these form cytoplasmic inclusion bodies [57] [56]. DPRs have been identified in the brain tissue of ALS/FTD patients [57]. FUS Like TDP-43, fused in sarcoma (FUS) is a primarily nuclear protein that is implicated in transcription, mRNA transport and splicing; however, these proteins do not colocalize and the proteins with which each interacts are distinct, suggesting these two proteins act in distinct pathways. Like TDP-43, FUS has a C-terminal RNA binding domain. It is found in neuronal inclusions in patients with ALS, however those inclusions are negative for SOD1 and TDP-43 [48] [55]. Over 40 FUS mutations have been associated with ALS and three FUS mutations are associated with ALS-FTD or FTD. 4% of fALS patients, 1% of sALS patients, and a smaller percentage of fFTD patients have FUS mutations. These mutations exist primarily in the C-terminal, which contains the nuclear 21 localization signal (NLS), which is necessary to transport into the nucleus. FUS mutations disrupt FUS binding to transportin, which transports FUS to and from the nucleus – explaining its cytoplasmic accumulation. Oligogenic Causes of ALS-FTD While some ALS and FTD mutations exhibit high penetrance, such as mutations in SOD1 and tau, other mutations are more common yet do not exhibit any or a strong disease phenotype. Oligogenic models of ALS/FTD suggest that a combination of mutations result in disease. C9orf72 mutations are the most common cause of ALS and FTD, however on their own they have a low penetrance. Secondary mutations, such as mutations in TDP-43, FUS, SOD1 and Tau, are often associated with C9orf72 mutations. Interestingly, data suggests that C9orf72 mutations in combination with mutations in TDP-43, FUS, and SOD1 result in ALS, while C9orf72 mutations in combination with mutations in tau result in FTD [48] [58]. Environmental Factors In addition to genetic causes of fALS, there are potential environmental causes of sALS, such as repeated head injury, which has been found in cohorts of retired American football and Italian soccer players [59, 60] [61]. However, while some studies suggest a correlation between prior head injuries (as a result of high impact sports such as hockey and football) to ALS, this evidence is not entirely conclusive, as many studies exploring this question indicate varying results [62]. Additionally, some data suggest that the consumption of cyanobacteria- 22 contaminated meat has been associated with ALS-like phenotypes in Guam. Some studies also suggest that inhalation of cyanobacteria in desert dust has been associated with cases of sALS Gulf war veterans, as the cyanobacteria produces neurotoxins [63] [64] [65] [66]. However, the influence of cyanobacteria on the development of ALS has varying levels of support [60]. Models of ALS Accurate animal models are required to study the mechanisms and functions of genes and proteins implicated in disease. Because 20% of fALS patients result from SOD1 mutations, SOD1 mouse models are commonly used to study ALS. SOD1 Knock-Out Model To identify possible SOD1 pathological mechanisms, SOD1 knock-out mice were generated, however these mice do not develop ALS up to six-months after birth and exhibit distal axonopathy, and no motor neuron loss [11]. SOD1-G93A The most commonly used mouse model for ALS is the overexpression SOD1G93A transgenic mouse, which harbors a glycine to alanine point mutation in exon 4 of the human SOD1 gene. 23 The B6SJL-Tg(SOD1*G93A)1Gur/J mouse The B6SJL-Tg(SOD1*G93A)1Gur/J mouse (known as the Gurney mouse model) exhibits a severe and fast-progressing neurodegenerative phenotype, in which mice show signs of hind limb weakness at approximately 3-4 months of age. They develop rapid paralysis and then die 7-10 days later (~P128). Gurney et al. 1994 assessed hind limb weakness by holding the mouse by the base of the tail and determining that these mice extend their hind limbs less [51] [67]. Additionally, motor neuron denervation presents in asymptomatic P60 mice as measured by quantifying glycinergic bouton densities. Interestingly, at P75, these mice exhibit a recovery of glycinergic bouton density, which also corresponds to neurite sprouting [68]. On a cortical level, at P85, these mice also exhibit decreased length, diameter, number and spine density of pyramidal cell basal dendrites within the prefrontal cortex [69]. They exhibit motor neuron degeneration between P90-120 [51] [68]. Additionally, the coats of these mice develop a coarseness suggesting impaired grooming. Pathological phenotypes include a prominent decrease in spinal motor neurons containing choline acetyltransferase (ChAT). Gurney et al. also demonstrate a loss of myelinated, large axons from ventral motor roots, while in comparison dorsal sensory roots are mostly spared. Additionally, to compensate for denervation, motor neurons reinnervate muscle fibers through nodal sprouting, until this mechanism can no longer sustain function. The most pronounced pathology occurred in the ventral spinal cord, while the dorsal spinal cord exhibited less pathology [51]. 24 The transgenic SOD1G93A mouse is useful because the mutation was originally identified in human ALS patients, and in mice it recapitulates many of the progressive phenotypes observed in human ALS patients. One study by Synofzic et al. demonstrate that a small cohort of fALS patients with the SOD1G93A mutation exhibit relatively homogeneous disease onset, symptoms, and progression. These patients presented with ALS between the ages of 55-63 years (mean: 60.3), and the disease course was rapid in which patients died within approximately three years due to pneumonia or respiratory insufficiency. Patients presented with lower motor neuron symptomology consisting of asymmetric limb weakness and paralysis and exhibited both upper and lower motor neuron denervation. They exhibited late onset bulbar dysfunction. The fast disease progression in these SOD1G93A ALS patients as well as the homogeneous symptomology is similar to that which is observed in the SOD1G93A transgenic mice, providing a basis for using the SOD1G93A transgenic mouse as a model for ALS. The B6.Cg-Tg(SOD1*G93A)1Gur/J mouse The B6.Cg-Tg(SOD1*G93A)1Gur/J mouse exhibits a prolonged lifespan (~P157) compared to the Gurney mouse [70]. They exhibit hind limb tremor when suspended by the tail at P142 [70], and motor neuron degeneration is apparent at P112 in the ventral spinal cord [71]. They exhibit motor neuron denervation in the tibialis anterior (TA) as early as P50, in the soleus at P100, and in the diaphragm as early as P120 [36] [72]. Earlier pathology such as synaptic vesicle stalling, 25 accumulation and depletion as well as axonal thinning can initiate as early as P35 [35]. TDP-43 Because the majority of ALS patients and 50% of FTD patients have TDP-43 aggregates, generating mouse models containing these aggregates may recapitulate these diseases. ALS patients exhibit a 1.5-2.5 fold increase in the levels of TDP-43 protein in pathological neurons. While knock-out models are often useful in identifying gene function, total lack of TDP-43 (Tardbp-/-) is embryonically lethal [53] [73] [74] [75]. Transgenic homozygous mice that overexpress wild-type human TDP-43 result in a rapid disease onset, motor deficits and low survival, dying within approximately two months. Pathologically, these mice do not exhibit TDP-43 aggregates as do patients with ALS, suggesting that these mice may not be an accurate model of ALS [11]. To more accurately generate a TDP-43 over-expression mouse model, Picher Martel et al. (2016) used the human endogenous TDP-43 promoter to express wild- type human TDP-43, hTDP-43A315T and hTDP-43G348C at a more moderate level in mice. By about 9 months of age, these mice exhibit rotarod deficits as well as cognitive impairment. Even though survival was not impacted in these mice, these phenotypes could be indicative of ALS-FTD [11]. Double transgenic mice that express hTDP-43WT and hTDP-43Q331K exhibit rapid limb-onset disease progression starting at three weeks of age, and these mice 26 die within 8-10 weeks of age exhibiting spinal cord anterior horn motor neuron loss, muscular atrophy, NMJ loss, neuronal cytoplasmic inclusions, ubiquitin and TDP-43, suggesting that these double-transgenic mice may be useful models for ALS-FTD. Interestingly TDP-43Q331K and TDP-43WT single-transgenic mice survive up to 24 months of age and do not exhibit motor deficits [11]. Another valuable mouse modeling TDP-43 proteinopathy is the transgenic hTDP-43ΔNLS [76]. This mouse harbors the neuron-specific promoter, NEFH, which – in the absence of doxycycline (DOX) – promotes tetracycline transactivator (tTA) protein expression. tTA binds to the tetracycline promotor element, resulting in expression of hTDP-43 without the nuclear localization sequence (hTDP-43ΔNLS). This lack of TDP-43 nuclear localization results in accumulation of TDP-43 in the cytoplasm. Supplementing the mouse diet with doxycycline suppresses hTDP- 43ΔNLS protein synthesis. Doxycycline binds to tTA, preventing it from binding to the tetracycline promotor element [76]. In the presence of a doxycycline diet, hTDP-43ΔNLS is suppressed. Therefore, the nuclear localization of TDP-43 can be controlled and provides an opportunity 1.) to explore the mechanisms that underlie TDP-43 nuclear mislocalization, 2.) to explore the variable vulnerabilities of specific motor neurons to cytoplasmic TDP-43 accumulation, 3.) to explore the potential for phenotype rescue after the re- introduction of DOX into the mouse diet, and 4.) to model the TDP-43 cytosolic aggregation phenotype observed in all ALS patients and 50% of FTD patients. These mice exhibit hypoglossal and lumbar spinal cord motor neuron cell death after 8 weeks of transgene expression [76]. However, there was no significant 27 cell death in the other motor neurons examined (occulomotor, trigeminal or facial motor neurons), demonstrating selective cell death. Additionally, the fast-twitch gastrocnemius and tibialis anterior exhibited denervation after four weeks off DOX while the slow-twitch soleus exhibited denervation after eight weeks off DOX, demonstrating that fast-twitch muscles are more vulnerable to cytoplasmic TDP-43 accumulation than slow-twitch. Because this model recapitulates pathological aspects of ALS (i.e. cytoplasmic accumulation of TDP-43) it may prove to be a useful mouse model in studying neuromotor pathology, specifically ALS disease progression [76]. C9orf72 As 40% of patients with ALS and 25% of patients with FTD have high levels of C9orf72 hexanucleotide repeat expansions, C9orf72 mouse models would be valuable tools to identify underlying pathologies in a large proportion of ALS and FTD patients. Dr. Robert Brown’s laboratory at the University of Massachusetts Medical School developed a transgenic mouse with the human C9orf72 gene containing approximately 300-500 GGGGCC motifs [56]. Importantly, the human C9orf72 transcript levels in this transgenic mouse were comparable to the endogenous C9orf72 transcript levels in the mouse as well as to the endogenous C9orf72 transcript levels in human control and C9ALS/FTD patients. Despite the fact that human patients with these expansions develop ALS/FTD, this transgenic mouse model did not – there was no decrease in lifespan – mice lived beyond two years. These mice did not have evidence of motor neuron degeneration, 28 denervation, or motor and cognitive deficits. These mice did not exhibit neuromotor deficits on rotarod or grip strength testing. Mouse models of FTD often exhibit abnormalities in social interaction [77] [78]. Interestingly, while these mice did not exhibit any overt ALS phenotypes, they exhibited intranuclear C9orf72 sense foci at 3 months of age, and abundant sense foci throughout the CNS at 10 and 24 months of age. Additionally, these mice exhibited RAN translation dipeptide repeat proteins. Therefore, despite the fact that this mouse model does not exhibit overt disease phenotypes, it recapitulates many histopathological ALS/FTD features such as the deposition of DPR proteins demonstrating that this may be a useful model in studying underlying disease mechanisms [56]. Knock-In Mouse Models and CRISPR/Cas 9 Technology Knock-in models provide tools to accurately model human disease due to accurate gene copy number and endogenous gene regulation. The development of CRISPR/Cas9 technology has enabled more effective and efficient generation of knock-in mice. CRISPR/Cas9 technology is based on the endogenous machinery utilized by bacteria as a defense against viral infection. There are three steps to the CRISPR/Cas9 system. During acquisition, upon bacterial exposure to foreign viral DNA, the bacteria will cut the viral genome DNA into short pieces, which are inserted between non-coding regions of the bacterial genome. Therefore, every time a virus infects the bacteria, a piece of its DNA (target DNA) is incorporated into the bacterial genome between spacer DNA, enabling the 29 bacteria to establish a “record” of foreign genetic material, and these sequences of DNA are known as CRISPR loci. Within the CRISPR loci, separating each sequence of target DNA are sequences of crRNA (CRISPR RNA). Additionally, downstream of the CRISPR loci within the bacterial genome are short sequences of DNA known as tracer RNAs (tracrRNA), which are partially complementary to each crRNA upstream from each target DNA sequence. At this point, the CRISPR loci consist of multiple acquired viral DNA sequences separated by crRNA sequences. During the crRNA biogenesis step, tracrRNAs and CRISPR loci are transcribed into separate RNA sequences. Transcribed tracrRNAs then bind to the complementary portions of the transcribed CRISPR loci, now known as pre-crRNA. The tracrRNA is required for the maturation of the pre-crRNA to crRNA. Within the bacterial genome, downstream of the CRISPR loci is the DNA sequence for Cas9, which is transcribed into the Cas9 endonuclease. In the presence of tracrRNA, pre-crRNA is cleaved by Cas9 into separate crRNAs, each encoding a different sequence of RNA complementary to the originally acquired viral RNA. The third step – interference – occurs when the crRNAs and the tracrRNA form a complex with the Cas9 protein. Because the Cas9 is complexed with both the crRNAs and the tracrRNA, site-specific DNA recognition and cleavage is ensured. Viral DNA contains a protospacer adjacent motif (PAM), which is not present in bacterial DNA. The PAM sequence is adjacent to the target DNA and is recognized by both the bacteria and Cas9 enzyme, which increases the specificity of the complex. Cas9 then generates a double-stranded break of the viral DNA that has formed a 30 complex with the guide RNA – inhibiting protein synthesis, and thus protecting the bacteria from viral infection. This process can be harnessed to generate knock-in mice. To generate a knock-in mutation, a repair template is created. The repair template includes a sequence of oligonucleotides complementary to the mutation of interest. To decrease the complexity of CRISPR/Cas9 system, tracrRNA and crRNA can be combined into a single synthetic guide RNA (gRNA). Synthesized Cas9 mRNA can be injected with the gRNA and oligonucleotide into single-cell embryos. The Cas9 will cut the endogenous mRNA resulting in the insertion of the donor oligonucleotides as a mutation. The embryos will then incorporate this mutation into the genome. TDP-43Q331K Homozygous Knock-In Mouse The TDP-43Q331K homozygous knock-in mouse model developed by Dr. Jemeen Sreedharan at UMass has a point mutation of a glutamine to lysine substitution [79]. Within the TDP-43 C-terminus, which is involved in protein- protein interactions, TDP-43Q331K creates a protein kinase A site, which may interfere with phosphorylation. The mutation was identified by sequencing a cohort of 200 sALS patients and a 72-year-old man with limb-onset ALS was identified with the TDP-43Q331K mutation, and the duration of his disease was three years (Sreedharan, 2008). In assessing the functional significance of this mutation Sreedharan et al. expressed TDP-43wt and TDP-43Q331K in Chinese hamster ovary (CHO) cells. Immunoblotting the cytoplasmic fraction for the N-terminal Myc tag showed fragments with molecular weights between ~14 to ~45 kD, with the mutant 31 TDP-43 exhibiting prominent fragments in comparison to WT, suggesting that the TDP-43Q331K mutation increases TDP-43 fragmentation, corroborating the evidence for mutated TDP-43 playing a role in its aggregation in ALS patients. Additionally, the Sreedharan group transfected myc-tagged TDP-43WT and TDP-43Q331K into the spinal cords of stage 14 chick embryos and found that the mutants failed to develop limb and tail buds, and only 5-15% of mutant embryos reached a normal stage of maturation after 24 hours post electroporation, while WT development proceeded normally. Additionally, the mutant embryos exhibited an increase in the number of apoptotic nuclei compared to WT. These data suggest that the TDP-43Q331K point mutation identified in human ALS patients is toxic and could contribute to disease phenotypes like neuromotor deficits [79]. Given these TDP-43Q331K mutation phenotypes, the TDP-43Q331K knock-in mouse provides a useful tool in revealing disease phenotypes. Identifying Behavioral Phenotypes of Disease Models While cellular and molecular techniques provide useful and important information regarding disease mechanisms, animal behavioral techniques provide information on higher-level processes and provide an output for abnormal mechanisms and pathways. Characterizing disease progression through behavioral phenotyping is important for establishing the natural history of specific mouse models. Characterizing this natural history is important because it elucidates the timing of divergence between the WT and disease model. Additionally, different models may 32 exhibit phenotypes at very specific ages, and these phenotypes may resolve, remain constant, ebb and flow, exacerbate, or evolve into different phenotypes altogether. This characterization allows the experimenter to probe the mechanisms underlying these time-dependent phenotypes. Identifying early behavioral phenotypes would provide a behavioral biomarker of disease that could allow for the early testing of new drugs that may be ineffective at a later time. Clinically, ALS patients present with symptoms after disease progression and neurodegeneration is well underway. At this stage, compensation mechanisms such as neuronal sprouting are no longer effective to slow disease. Because ALS and FTD are progressive diseases, the window of time when intervention will be most effective is going to be early, and therefore, identifying this early window is important for testing therapeutics. Behavioral Experimental Design Considerations In conducting and designing behavioral experiments, many factors must be considered. The experimenter must ensure adequate group size and minimize potential sources of variability that would impact results. Careful consideration of group size is important as often subtle to moderate behavioral differences may go undetected due to intrinsic mouse behavioral variability. Larger group sizes can average out possible outliers as well as the influence of home cage dynamics that are dependent on established hierarchy within the cage [80]. Different genders can be grouped if the behaviors between genders are the same. Otherwise, genders should be analyzed as separate groups [80]. 33 Breeding littermates for behavioral experiments controls for potential genetic differences between mice. Heterozygous littermates can be bred to generate homozygous and WT mice. Often a challenge here is generating enough mice of the correct genotype to establish a large enough group size. If age is an important factor, then more mice may be necessary during the initial breeding to ensure that mice run on the paradigm are all the same age. Additionally, if generating enough mice of a specific genotype is challenging and inefficient, multiple rounds of breeding and genotyping can be conducted generating mouse cohorts of different ages. Therefore, multiple cohorts of mice with the same age may be necessary to achieve a large enough number of mice to run the experiment. If multiple cohorts of the same age are used, it will be important to take that into consideration in analyzing the results as a specific cohort may exhibit slightly different behavior due to a time-dependent factor, such as behaviors that are variable depending on the time of year. In running behavioral testing, many sources of variability can arise. Some sources of variability include the time of day the assay is run, the mouse strain, inconsistent smells between different testing rooms or facilities, inter-experimenter technique differences, or inconsistent noise or other environmental factors at play during testing. Behavioral experiments are challenging because controlling these sources of variability is often difficult and often beyond control of the experimenter, for example noise or construction within the mouse breeding facility. Prior to running testing, it is important to assess the overall health of the mouse as an unhealthy or physically variable mouse can skew results. Additionally, specific mice with an obvious observable abnormality such as hydrocephalus should 34 not be included in the experiment as its behavior is not relevant to the experiment [80] [81]. Traditional Behavioral Assays Each behavioral test is designed to measure and detect specific functionalities such as neuromotor ability, cognition and anxiety. In behavioral experimental design, specific assays are chosen to probe these specific areas. It can often be difficult to discern whether a behavioral phenotype is actually indicative of an abnormality that the assay was designed to identify. For example, one may observe and identify a neuromotor phenotype that is actually indicative of a cognitive deficit. For example, a rotarod deficit, which traditionally would indicate a deficit in coordination, may be indicative of the inability to efficiently learn the task – a cognitive deficit. Therefore, a combination of multiple assays that probe different functionalities and deficits is necessary to provide a more comprehensive and accurate interpretation of the mouse behavior. Open Field Open field testing provides a measure of locomotion and exploration. The level of exploration indicates the natural curiosity of the mouse as well as the anxiety associated with open spaces. A single mouse is placed in an arena (40cm x 40 cm x 49 cm) and an overhead infrared camera records its movement. The camera is connected to a computer with video tracking software, which tracks and quantifies the distance traveled by the mouse as well as the amount of time the 35 mouse spends in specific locations within the arena i.e. the center (indicating exploration) and corners (indicating anxiety). Because open field testing involves very little interaction with the experimenter, variability related to mouse handling is minimized. Therefore the distance traveled and the exploratory behavior should not be influenced by the experimenter, and is representative of the mouse’s inherent behavior. A downfall of open field testing is that differentiating between locomotor ability, cognition and anxiety can be challenging when assigning an underlying reason for an open field phenotype. Grip Strength To assess motor function of the forelimbs the experimenter can measure the forelimb strength of the mouse using a grip strength apparatus. The mouse is held by its tail above a horizontal bar. The mouse will grasp the bar, after which the experimenter gently lowers the mouse so that it is horizontal with the ground. The experimenter then gently pulls back on the mouse tail until the mouse releases its grasp. The force is recorded by the apparatus. This is a useful technique to compare the forelimb strength of different genotypes. A downfall of this technique is its necessary involvement with the experimenter. While the experimenter is trained to maintain consistency, for example, pulling on the tail with the same force for every trial, human error cannot truly be held constant. The task must be performed by the same experimenter for every trial. Depending on variable techniques and the innate strength of the 36 experimenter, switching experimenters can result in different results unrelated to the actual strength of the mouse. Additionally, although the experimenter is trained to ensure the mouse has a firm grip on the bar before pulling back on its tail, the grip of the mouse cannot be held constant and is out of the control of the experimenter. This lack of control is a potential drawback to this technique. Wire Hang The wire hang test measures central nervous system motor function and deficit. The mouse is placed on a cage topper and after it has established a grip on the grating, the experimenter slowly turns the cage topper over so the mouse is upside-down. The cage topper is balanced over a large container into which the mouse can fall, approximately 1.5 feet beneath the cage topper. Therefore, the mouse has an incentive not to fall. This task is useful in that it is physically challenging – the mouse needs to hold onto the grating while upside down, otherwise it will fall. Therefore, the task may pick up on subtle deficits that may go undetected in a less demanding task. However, drawbacks to this task include variability in technique even within one experimenter. The paradigm requires careful turning of the cage topper at approximately the same speed over multiple trials across multiple mice, and despite careful attendance, this can be challenging as trials may continue for hours, depending on the number of mice. Additionally, mice can learn to jump off of the cage topper or develop a different strategy to avoid the discomfort of falling. For example, a mouse that hangs on for 7 minutes may learn that the discomfort of 37 holding on outweighs the discomfort of falling, causing the mouse to drop after a few seconds in subsequent trials. This is not a true measure of neuromotor ability, but instead potentially a cognitive phenotype. Rotarod Treadmill The rotarod can assess motor degeneration, motor coordination, motor pattern learning and vestibular function. During rotarod testing, the mouse is placed on a rotating drum, which accelerates from 4 rpm to 20 rpm. Prior to the start of testing, the mouse is acclimated to the drum at 4 rpm. Once the mouse can stay on the drum for 20 seconds without falling, the rotarod accelerates from 4 rpm to 20 rpm over the course of five minutes, after which the speed is held at 20 rpm. A photobeam detects when the mouse falls off, and the amount of time the mouse remains on the drum without falling is automatically recorded. Other than the handling required to place the mouse on the rotarod, this task does not require a great deal of experimenter interaction, which decreases overall variability. One source of variability however is in the mouse’s ability to learn the task. While the purpose of the task is to detect neuromotor phenotypes, it does not take into account cognitive variables that could impact the results. If a mouse has learned how to jump off of the rotarod, it often will do so during subsequent trials and even during re-testing at different ages, potentially skewing results. 38 Marble-burying The marble-burying task can identify anxiety-like or obsessive-compulsive- like (OCD) behavior. A standard shoebox cage is filled to a depth of 5 cm with wood chip or corncob bedding. On top of the bedding, between 10-20 opaque glass marbles (14mm diameter) are placed, evenly spaced. Individual mice are placed in the cage for 30 minutes. The marbles that are at least 2/3 buried at the end of the trial are counted. When mice are placed in a novel cage environment, they exhibit increased digging behavior, and this repetitive behavior qualified as an anxious perseveration, perhaps due to a mouse’s intrinsic desire to escape. Therefore, the number of marbles buried can potentially be correlated with the amount of OCD-like or anxious behavior [82]. In addition to measuring anxious behavior, this task may also be indicative of attention-driven behaviors. Mice that bury few marbles may have an attentional deficit in which they are not able to complete the task due to the distraction of the novel environment. This task also requires minimal handling of the mice, as the mice are left to their own devices during the 30-minute testing period. One potential drawback of the task is that the bedding is generally not cleaned between testing. Despite the fact that this is the same for all mice tested, and therefore the amount of odor that would influence behavior should be relatively similar between mice, this cannot be controlled and could introduce potential error. While these tasks are all useful in identifying overt behavioral phenotypes, they are less useful in the identification of subtle abnormalities. 39 Monitoring weight and food consumption also provides information about metabolism and eating behavior. Automated Continuous Behavioral Monitoring Automated continuous behavioral monitoring (ACBM) is a behavioral technique developed by Dr. Thomas Serre at Brown University that uses computer vision and machine learning to characterize mouse behavior, and overcomes many of the limitations indicated above [83]. During ACBM, singularly housed mice are video recorded for five days. After recording, the video data are processed during which the computer assigns one of nine pre-defined behaviors to each frame: walk, drink, eating-from-food hopper, eating-on-haunches, groom, hang, rear, rest, and sniff (Fig. 1). Additionally, to get a more general measure of movement within the cage, translocation characterizes the number of pixels moved by the mouse per hour in either the x- direction, y-direction, or a combination of both x- and y- direction [√(x2 + y2)]. 1.3 million frames/mouse/session are collected from the video data. Therefore, 5 days of recording 24 mice corresponds to 156 million frames of video data. Without automation, it is estimated that it would take approximately 22 person hours to manually score each frame of video data recorded over 1-hour for one animal. Therefore, to manually score five days of video data for 24 animals would require approximately 7 years. In contrast, 5 days of ACBM data can be processed within approximately one week after recording is complete. 40 Figure 1. Automated Continuous Behavioral Monitoring: Mouse Behaviors. 41 ACBM Technology The technology behind ACBM was modeled after the organization of motion processing in the dorsal stream in the primate visual cortex in which low-level features are processed first followed by more complex and invariant features [83]. Video Pre-Processing In ACBM, upon video recording of a single cage, pre-processing occurs for each frame in which there is a background subtraction step that generates a foreground mask specific to the darker pixels within the animal compared to the lighter background (Fig. 2). To identify the video background for subtraction, the median pixel value (corresponding to a median gray value) is calculated for each frame within a video. Subtracting each individual pixel gray value from the median pixel gray value for each frame will generate a region of pixels with higher contrast, and this region corresponds to the mouse, as the mouse is darker in comparison to the background. In other words, background can be separated from mouse because gray values corresponding to background will be significantly different from gray values corresponding to mouse. This allows for establishing clear boundaries between mouse and background, taking into account shadows, or differences in levels of darkness throughout the background. Background subtraction enables analysis to be limited to pixels within the mouse [83]. 42 Figure 2: Video Pre-Processing Background subtraction. Within a frame, a median pixel gray value is calculated, and each pixel’s individual gray value is then subtracted from the median gray value. The resulting differences for the background gray values compared to the mouse gray values will have greatest contrast, and the mouse boundaries can be defined (modified from Serre et al., 2010) [83]. Figure 3: Sub-Window Extraction. Following background subtraction, sub-window extraction occurs in which a rectangular bounding box is generated surrounding the mouse. Video data within the bounding box is processed (modified from Serre et al., 2010) [83]. 43 Sub-window Extraction After background subtraction, there is a sub-window extraction, which crops the video so it is centered on the mouse, generating a bounding box (Fig. 3). The primary purpose of the sub-window extraction is to speed up computation as only the sub-window data are processed [83]. Data Processing and Behavioral Characterization From the video, two streams of information are generated: position, velocity, and acceleration based features, and motion features [83]. Position and Velocity Features Position and velocity-based features characterize the posture of the mouse within the foreground pixels according to its x- y- coordinates (Fig. 4). These ten features provide information such as whether the animal is in a horizontal or vertical position, how close it is to the food hopper, the waterspout, the sides and top of the cage, and its instantaneous velocity and acceleration (Fig. 4) [83]. Because there are variations between cameras in terms of the distance of the camera from the cage and the camera angle, the position- and velocity-based features need to be normalized (Fig. 5). Without normalization, the positions of the mice in different cages will not be comparable. The position of a mouse in one cage videoed by a camera at a slight angle will not be the same as the position of a mouse in a different cage videoed by a camera tilted at a larger angle. For each camera, the normalization occurs by manually defining the specific coordinates of each cage: the 44 water-spout, and the bottom, top, right and left corners of the cage. Therefore, the position of each mouse is always in the context of the defined coordinates within each cage [83] (Fig. 5). Figure 4: Position, Velocity- and Acceleration-Based Feature Extraction. Position, velocity, and acceleration features are extracted based on the mouse’s position in relation to predefined coordinates within the cage (water spout, food hopper, sides of cage, height and width of animal, corners etc. (modified from Serre et al., 2010) [83]. 45 Figure 5: Normalization of Position, Velocity, and Acceleration-based features. For each camera, normalization occurs by manually defining the specific coordinates of each cage: the water spout, and the bottom, top, right and left corners of the cage. Therefore, the position of each mouse is always in the context of the defined coordinates within each cage. Motion Features As mentioned above, processing of motion features is modeled after the motion-sensitive cells in the dorsal stream in the primary visual cortex. Motion features are extracted from the video data in a hierarchical manner. First, through a process known as template matching, a small sequence of nine frames (~1/3 of a second) is compared to a pre-defined template of motion features, and spacio- temporal filters are applied to the data. The filters differentiate between four pre- defined directions of motion (0° [180°], 45° [225°], 90° [270°], 135° [315°]) – a template. A minimum of nine frames captures motion features. For every sequence of frames within each pixel, these spacio-temporal features are extracted, and a 46 “weight” is assigned corresponding to the magnitude of each direction of movement. Like the increasing complexity in the characterization of movement processing in the brain, movement processing in ACBM increases in levels of specificity in a hierarchical manner by combining feature outputs. The combination of these feature outputs results in 1 of 300 previously defined motion features for every pixel [83] (Fig. 5a, b). Figure 6a: Motion Feature Extraction and Template Matching. Nine pixels x nine pixels x nine frames are extracted from each video. A minimum of nine frames is necessary to capture motion features. Each sequence is passed through four spatio-temporal filters (shown) defining four directions of motion. This corresponds to low-level feature extraction. Each input sequence of nine frames is assigned a weight of probability for each of the four directions of motion (modified from Serre et al., 2010) [83]. 47 Figure 6b: HMAX Template Matching Feature extraction continues for every frame of video data (modified from Serre et al., 2010) [83]. A more detailed explanation is as follows: While characterizing each static frame provides significant information, it is not optimized to detect specific subtle features dependent on motion. The temporal context of behaviors is therefore important, and to optimize and accurately characterize spacio-temporal features of behavior, ACBM processing uses HMAX (hierarchical model of object recognition – hierarchical model and X) modeling. 48 HMAX modeling assigns a maximum probability value to a given feature. In our case, motion features are characterized in which previously defined motion features from a separate template dataset are compared to the novel motion features extracted from the video recordings. The template motion features with maximum similarity to the novel motion features are used to characterize and label the novel motion features. To define these features for eventual behavioral characterization, video sequences consisting of a spatio-temporal patch of nine pixels x nine pixels x nine frames are extracted from the video data. First, these sequences are passed through four spatio-temporal filters that define four directions of motion, as discussed previously. This process is known as low-level features extraction and represents a linear stage of processing, in which a small patch of pixels (9 x 9) over an input sequence of nine frames are assigned a weight of probability for each of the four directions of motion. From these four specified directions of motion, a non-linear normalization occurs in which the filter response (the magnitude of directional selectivity for each of the four directions) is divided by the L1 norm. L1 norm is a value that normalizes directional selectivity by excluding less likely predictions for which the majority of the corresponding weights is zero. This further fine-tunes directionality as these more selective predictions produce a more stringent filtered directional output for every input sequence. Therefore, each video sequence harbors information on the magnitude of likelihood for each direction of motion. 49 Next, motion features (based on the combination of weights specifying directionality) within the 9 x 9 pixel patch from the sequence of the first nine frames of collected video data are compared to hard-coded motion features of a previously defined space-time motion template. The degree of similarity between motion features within these nine frames and motion features within the template is defined resulting in a specific movement vector. In other words, features within the template that have maximum similarity to the features within the nine frames define the motion features within those nine frames. To continue to characterize the video frames, the frame sequence moves forward one frame – incorporating the tenth frame and excluding the first. Motion features within the second through the tenth frames are compared to the template motion features, and again a degree of similarity between the tenth frame and the hard-coded motion features of the template is defined resulting in a feature prediction for that frame. This process continues for every frame of collected data – moving forward in the video sequence by one frame. For example, the eleventh frame is then characterized as predicted motion features based on the characterization of the previous 9 frames + the current frame. To relate to biological visual processing – template matching is similar to the hierarchical increasing specificity and invariance to motion features that occurs within the simple, or S1 cells in the primary visual cortex. Therefore, this feature processing step in ACBM is known as S1 processing. After template matching to specific directions of movement, C1 processing (corresponding to the complex cells in the primary visual cortex) occurs in which 50 the data is further filtered, through max pooling, to increase the directional specificity while decreasing over-fitting. During over-fitting, different yet irrelevant features are characterized as their own separate categories. This is a problem because similar features that should be grouped may be different in one attribute, and that difference places the two similar features in different categories. For each patch, the output of S1/C1 processing is a motion feature – consisting of the 4 different directions and corresponding magnitude – feature vectors. In S2/C2 processing, those 4 feature vectors are then combined, further increasing specificity and feature invariance to specific motion features. HMAX modeling was originally designed to model the hierarchical organization of visual processing of object recognition. This hierarchy of increasing specificity in object detection was first described by Drs. David Hubel and Torsten Wiesel. Among their many accomplishments, their discovery of neurons within the visual cortex that exhibit a hierarchical specificity to preferred directions of motion not only enlightened the field of neuroscience, but also spurred machine learning applications to model this elegant biological hierarchical system. Statistical Classifier – Hidden Markov Model Support Vector Machine The output of the position, velocity, acceleration and motion-based features is run through a statistical classifier – a Hidden Markov Model Support Vector Machine (SVMHMM), which incorporates these features into a behavioral characterization. The model works by first taking a sequence of frames with specific motion features and characterized through HMAX modeling, and generates a 51 behavioral “call”, based on pre-defined features. The output of that “call” is then fed forward through the next sequence of frames with corresponding specific and invariant motion features to generate a second behavioral call. Chunks of sequences will therefore likely generate the same behavioral call, as a single behavior will continue over many frames (each frame is 1/30 of a second). This process continues throughout every frame sequence, generating a sequence of behaviors that comprise the complete video (Fig. 7a). This modeling establishes the boundaries between behaviors to classify each behavior as a separate behavioral category, in which motion features that are most similar to each other are characterized together as one behavior [83] (7b). Figure 7a. SVM-HMM Classifier: Characterizing similarity among features within sequences of frames. The SVM-HMM classifier takes a sequence of characterized motion features, and generates a behavioral “call”, based on pre-defined features. The next set of features is then characterized, generating another behavioral call. This process continues for 52 all of the frames, generating a sequence of behaviors that comprise all of the video data. Figure 7b. SVM-HMM Classifier; Characterizing similarity among features between sequences Input features that are similar to each other are grouped to enable behavioral classification. ACBM Development The development of the ACBM system required two databases of annotated behaviors. The “clipped database” consisted of 9,000 short clips, and each frame was manually annotated as a specific behavior with “high stringency” by a group of eight annotators denoted as “Annotator group 1”. To minimize the chance of errors, one annotator went through the annotations again and selected the most un-ambiguous clips that corresponded to 262,360 frames (2.5 hours). These clips corresponded to 12 different videos, which were used to tune and train the system to selectively 53 designate specific behavioral features to specific video frames. The annotations made by annotator group one were used as “ground truth” for training and testing the system since these annotations were made with very high stringency [83]. In order to test the classification module designed to incorporate the above behavioral features, a second database, the “full database” was developed. This database consisted of video clips from 12 unique videos different from those selected for the clipped database. A second group of annotators annotated each frame of this database, but with less stringency than that of the clipped database. While the clipped database consisted of 2.5 hours of video database, the full database consisted of 10 hours of video data. To ensure that the style of annotation was consistent between annotators, an annotator was used to correct any mistakes [83]. ACBM Accuracy To assess the accuracy of the system it was compared to 1.) commercial software (HomeCageScan 2.0 CleverSys Inc.), which classifies mouse home cage behavior, and 2.) human annotators, through a leave-one-out procedure. One of the videos was left out from the training and used to assess the accuracy of the system. The ACBM system exhibited 77.3% agreement with human annotators while the HomeCageScan exhibited 60.9% agreement. To generate these percentages, each system was compared to the annotations from the annotator group 1 (the stringent annotations designated for the cropped database), which was specified as the ground truth [83]. 54 To quantify the accuracy associated with each behavioral designation, confusion matrices were generated. A confusion matrix displays the probabilities of agreement between two comparisons. In this case, they were interested in how comparable the ACBM system is to human annotators and the HomeCageScan system in classifying behaviors. If there is complete agreement between human annotation and the computer system for a specific behavior, than the value within the confusion matrix comparing human annotators to the ACBM system for that behavior would be one, while all other values (for all other behaviors) would be zero. This is generally not the case, as it is likely there will be some variability between human annotators in a portion of their behavioral calls in comparison to the ACBM computer program [83]. As proof of concept, the Serre laboratory characterized home-cage mouse behavior of four different mouse strains in an effort to determine whether ACBM could characterize and quantify the behaviors [83]. They examined the wild-derived CAST/EiJ strain, the BTBR strain (a strain used to model autism) and two inbred wild-type strains - the C57BL/6J and the DBA/2J strains. First in examining just the CAST/EiJ mice, Jhuang et al. (2010) found, as expected, that the mouse hanging and walking activity is greatest during the dark hours. Additionally, ACBM demonstrated that the CAST/EiJ mice exhibited an overall higher level of walking and hanging behavior than the other mouse strains, however, the DBA/2J strain demonstrated a high level of hanging during the first few circadian hours in comparison to the C57BL/6J and the BTBR mice. Because the BTBR mice tend to hypergroom, Jhuang et al. also compared grooming behavior of these mice to the C57BL/6J mice and 55 verified a significant increase in grooming behavior of the BTBR mice. Therefore, ACBM was able to identify behaviors that were specific to each mouse strain, demonstrating the specificity of the system. Storage of ACBM Data and the Benefits of Data Storage After data acquisition, ACBM data are stored indefinitely in the Brown University Common Internet File System. These data are available upon request (through contacting the laboratory of Thomas Serre Ph.D.). Indefinite storage of these data enables re-analysis as additional computational methods become available. Also, data storage allows for the characterization of additional pre-defined behaviors. While in my analyses, I examined nine different pre-defined behaviors using ACBM, it is possible to train the ACBM system to characterize other behaviors through machine learning. For example, there are instances where a behavior is transitioning from one to another – such as walk-to-rear or rear-to-hang. It is difficult to accurately establish when one of these behaviors “stops” and the next “initiates”. Therefore, one could establish a new behavior as “walk-rear” which defines the intermediate behavior between walk and rear. This would further improve accuracy and fine tune specific behaviors through the elimination of data that corresponds to behavioral transitions, and separates those transitions into specific categories, which in themselves can be analyzed. 56 Treatment Development and Testing Using ACBM ACBM has many utilities. One such utility is its potential to be used as a read- out for phenotype change with potential treatments. Given the late onset of disease diagnosis, it is possible that already available treatments like riluzole or radicava would be effective at earlier stages of disease. To test how early riluzole could have an impact on changing the course of disease progression in mouse models of ALS, one could use ACBM to assess behavioral change due to riluzole treatment at different stages of disease. If ACBM identifies subtle behavioral improvements or changes in treated mice compared to untreated mice, then it could detect specific windows of time during which riluzole has a therapeutic effect in these mice. Pathology occurring during or after those windows could be examined to provide insight into what exactly riluzole is targeting. The same concept could be applied to Edaravone. Additionally, with the identification of specific suppressor genes whose perturbation could slow or prevent ALS, one could use ACBM to test behavioral phenotype changes in knock-in mice with a mutated form of that suppressor gene. Additionally, one could generate a knock-out mouse with the ALS suppressor gene removed and run the mouse on ACBM to determine if that mouse exhibits a behavioral difference in comparison to an ALS mouse model with normal expression of the suppressor gene. One may hypothesize that the knock-out mouse may exhibit a complete or partial phenotype rescue in comparison to the ALS mouse model that expresses the suppressor gene. Additionally, to further implicate that gene in disease progression, one could target that gene pharmacologically and test ALS 57 mouse models with and without such pharmacological treatment to determine if there is a rescue in specific behaviors using ACBM. Additional Mouse Behavioral Characterization Systems One caveat to ACBM is it cannot measure social behavior of mice as it can only track singularly housed mice. Giancarlo et al (2013) developed a behavioral monitoring technology that can measure the social behavior of more than one mouse [84]. Through machine learning, behavioral classifications are learned from annotated examples of specific social and non-social behaviors. Differentiating between mice within an arena is important to maintain mouse identity and to discriminate between mice in close proximity to each other. A “heat analysis” is done using a thermal camera that detects the mice and provides dynamic information on where each mouse is in relation to each other. Each mouse is given an identifier so that comparisons can be made between mice in a pairwise fashion. Each mouse (its position and heat analysis) is compared to every other mouse in the arena, and the comparison data is used to assign and quantify the amount of social behavior occurring between each mouse. Within each mouse pair, one mouse is identified as a target mouse (the mouse whose behavior is being characterized) and the other mouse is identified as a reference mouse. The behaviors between each pair were characterized as either social – in which two specific mice are interacting (“Nose2Body”, “Nose2Nose”, “Nose2Genitals”, “Above” [one mouse is above, or mounting, the other], and “Following”), non-social (“StandAlone” and “WalkAlone”), or undefined (“StandTogether”). This technology is interesting because the social 58 context of behaviors provides additional information that can be used to differentiate between mouse type or conditions within an experiment, such as between mouse strains, genders, or treatments [84]. While the ACBM system we used (Jhuang et al., 2012) and the Giancarlo et al., (2013) system are useful in characterizing pre-defined behaviors, Wiltschko et al., (2015) generated an automated machine learning technology that does not rely on previous human annotation. Pre-defined features limits the extent to which one can characterize mice as there are a limited number of categories between which a human can accurately differentiate. Therefore, Wiltschko et al., used machine vision and learning to identify specific behavioral modules, that are defined by the computer and therefore automated over time and between mice. Each module represents a specific stereotyped accumulation of motion and position features and once defined by the computer, these modules or a combination of specific related modules can then be ascribed to specific behaviors post-hoc [85]. This technique is beneficial because it does not confine a complex behavior into a simpler, presumably less accurate, characterization, and more behaviors can be identified using this method. Additionally, a computer system that can separate and characterize specific similar behaviors, eliminates human error in annotating behaviors incorrectly. Conclusion Automated Continuous Behavioral Monitoring provides a rigorous quantitative technique that enables the characterization of mouse behavior through 59 machine learning technology [83]. Our work uses this technology to identify and characterize early behavioral phenotypes of the SOD1G93A transgenic mouse model of ALS as well as the homozygous knock-in TDP-43Q331K mouse model of ALS-FTD. Identifying such early behavioral phenotypes would enable early intervention to test therapeutics, define early pathological pathways and identify potential biomarkers associated with these early phenotypes. 60 CHAPTER 2: BEHAVIORAL CHARACTERIZATION OF THE B6.CG-TG(SOD1*G93A)1GUR/J ALS MOUSE MODEL THROUGH AUTOMATED CONTINUOUS BEHAVIORAL MONITORING Amanda Marie Duffy, Justin R. Fallon Department of Neuroscience, Brown University, Providence, RI 02912 Contributions: I performed all of the experiments and analyses presented here. S. Killeen and C. Ortega collected the ACBM data for the transgenic SOD1G93A mouse model. T. Serre Ph.D., Y. Barhomi, X. Li, V. Veerabadran, and T. Sharma conducted ACBM data processing. 61 ABSTRACT Accurate and reproducible behavioral measurements are essential for characterizing natural history and testing therapeutics in mouse models of neuromotor diseases. Behavioral assessment provides a cumulative output of function, and can provide a window into identifying and characterizing the timing of when and how disease phenotypes present and progress. Behavioral phenotypes that manifest early in disease progression provide insight into the initiating mechanisms that result in eventual loss of function. Often these phenotypes are subtle, prior to the onset of overt symptoms. A measure that detects these subtle early phenotypes would enable the identification of a time window during which therapeutic intervention would be most effective. Automated Continuous Behavioral Monitoring (ACBM) is a novel machine learning technology that can detect, identify and measure specific subtle behavioral phenotypes. ACBM video records individually housed mice continuously and simultaneously for five days in their home cages. ACBM uses a supervised computer algorithm to assign one of nine behaviors to each video frame (thirty frames/sec) (walk, drink, eat-on-haunches, eat-from-hopper, rear, hang, rest, groom, and sniff). Additionally, we characterized a behavior that we call “translocation,” which measures any movement in x- or y-coordinates. The automation, large datasets, and continuous five-day recording periods minimize factors that result in variability. We used ACBM to identify and measure behavioral phenotypes in the transgenic SOD1G93A mouse model of amyotrophic lateral sclerosis (ALS) across three ages (p30, p58, and p86). This mouse model is currently the most commonly 62 used model of ALS (B6.Cg-Tg(SOD1*G93A)1Gur/J). We identified a transient walk deficit at p30 and p58 that disappears by p86. This is the earliest behavioral phenotype identified in this mouse model and parallels the timing of early denervation followed by compensatory sprouting (Pun et al., 2006). We also identified an SOD1G93A deficit in translocation at P30, which resolved by P58. Interestingly, these mice exhibited an increase in translocation at P86. Additionally, they exhibited a decrease in eating-on-haunches behavior at P58 and P86. Therefore, we identified specific behavioral phenotypes in the SOD1G93A mouse model that present at different ages and exhibit specific patterns of progression with age. ACBM enables sensitive, objective, longitudinal and rigorously quantified behavioral phenotyping of disease models, revealing novel features of the natural history and opening a path for testing drugs that could act early in disease progression. 63 Introduction Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease of upper and lower motor neurons (UMNs and LMNs) that results in weakness, paralysis, and eventually death. The average age of ALS diagnosis is in the sixth decade, at which point neurodegeneration is already underway. Patients typically survive for 2-5 years following diagnosis [8] [9]. Therefore, identifying early phenotypes prior to the onset of overt symptoms is critical in earlier diagnosis, treatment, and planning. Mouse models that overexpress human disease mutations provide tools to identify and characterize disease phenotypes and disease progression. Additionally, mouse models can be used to test the efficacy of interventions and identify the timing of when intervention is most effective, specifically during early stages of the disease. Currently, the most commonly used ALS mouse model is the SOD1G93A transgenic mouse, which survives until ~P160 (B6.Cg-Tg(SOD1*G93A)1Gur/J; hereafter referred to as the SOD1G93A mouse). Although early pathology such as synaptic vesicle stalling, axonal thinning, and denervation can occur as early as ~P30, overt symptoms such as hind limb tremors, weakness and paralysis do not occur until as late as P90 due to compensatory mechanisms such as axonal sprouting from neighboring slow or fatigue resistant neurons [34-37] [70] [71] [36] [72] [35, 37]. Accurately characterizing neurodegenerative disease, from initiation through end stage, requires a multi-pronged approach, as many deficits may not be apparent 64 during early stages of the disease, despite underlying pathology. An essential component of this analysis is behavioral phenotyping. Behavioral assessment enables longitudinal tracking of phenotypic changes throughout disease progression. This longitudinal tracking not only improves efficiency in terms of eradicating the need to sacrifice mice at multiple ages to identify age-dependent pathological biomarkers, but also and importantly allows for following the same cohort of mice over time, eliminating problems related to variability between mouse cohorts across multiple ages. Longitudinal assessment of behavior provides insight into the underlying pathology, as well as a tool for drug development in assessing rescue of specific disease phenotypes with and without treatment. Traditional behavioral measures have identified robust late onset phenotypes in the SOD1G93A mice as early as ~p90 [86, 87]. However, the timing of such deficits often exhibits conflicting results due to intrinsic variability [72] [27]. Therefore, behavioral technology to identify early, transient, and later behavioral phenotypes would provide a cumulative output of function as well as a window into characterizing the timing of when and how disease phenotypes present and progress. It is critical to identify early behavioral phenotypes before the onset of neurodegeneration because this early window provides an opportunity for characterizing early initiating mechanisms associated with neurodegeneration as well as intervention. In order to identify early deficits, which are often subtle, behavioral measures that are more sensitive than the current traditional techniques are necessary. 65 To reveal early behavioral phenotypes in the SOD1G93A mouse and to provide sensitive and robust analysis, we used Automated Continuous Behavioral Monitoring (ACBM), a novel machine learning technology that can detect, identify and quantify pre-defined behavioral phenotypes and allows for generating a deep, rich behavioral dataset that is not available through traditional behavioral analyses or manually scoring mouse behavior [83]. In ACBM, individually housed mice are continuously video-recorded on a light/dark cycle in their home cages over five days. Motion, position, velocity, and acceleration-based features are then extracted from the video data (see methods). These features are then fed through a Support Vector Machine Hidden Markov Model classifier, which characterizes them into one of nine pre-defined behaviors (Fig. 1; see introduction and methods for behavior definitions and details on ACBM technology). 66 Figure 1. Schematic of ACBM feature extraction and behavioral classification. The ACBM system and technology are described in introduction and methods. ACBM collects 1.3 million frames/mouse/session from the video data. To manually score five days of video data for 24 animals would require approximately 7 years. In contrast, 5 days of ACBM data can be processed within approximately one week after recording is complete. Using ACBM, we characterized and quantified the behaviors of SOD1G93A mice. To characterize the onset and progression of behavioral phenotypes, we conducted ACBM over three ages (P30, P58, P86), which are prior to the onset of phenotypes identified through traditional behavioral measures. Here, we demonstrated that some behavioral phenotypes were transient in which they exhibit an early deficit that resolves at a later age. Other behaviors exhibit a later onset phenotype. Other behaviors exhibit a more multifaceted age- dependent trajectory in which age impacts the directionality of the phenotype. For clarity and to differentiate between phenotypes that emerge between P30 and P58, we called behavioral phenotypes with an onset after P30, “later onset” phenotypes. The SOD1G93A mice exhibit an early and transient walk deficit, and a later onset eating-on-haunches deficit. We also characterized an additional behavior that measures overall movement, which we termed, “translocation”, which exhibited a switch in directionality over age. While overall translocation exhibited an early onset deficit at P30, this phenotype disappeared at P58 and then increased with age at P86. Translocation can be separated into x- and y-direction movement. Interestingly, we found that translocation in the x- and y-direction contributes to overall translocation differently depending on age. 67 Our characterization of behavioral phenotypes demonstrates that specific behaviors are particularly susceptible to underlying pathology present in the transgenic SOD1G93A mice, and these phenotypes are present early in disease progression. This novel characterization of early behavioral phenotypes allows for identifying an early window during which intervention may be most therapeutic. Results ACBM detects both early and later onset phenotypes in the SOD1G93A mouse. We used ACBM to probe for behavioral phenotypes in the SOD1G93A mouse prior to the onset of differences identified through traditional behavioral techniques (P90) [72]. We chose to investigate behaviors at P30, P58 and P86. As described below, we identified an early walk and translocation phenotype at P30, as well as a later onset eat-on-haunches phenotype at P58 and P86 (definitions of behaviors are in methods) (Fig. 2a, b, c, Supplementary Fig. 1, and 2). 68 Figure 2. ACBM detects SOD1G93A early and transient walk and translocation deficit at P30 and eating-on-haunches deficit P58. The x-coordinate corresponds to circadian hours 0-23. For walk and eat-on- haunches, the y-coordinate corresponds to the average # of seconds/hour that the mice are performing that behavior. For translocation, the y-coordinate corresponds to the average number of pixels/hour that the mouse is moving in the x- and y- directions. Because mice exhibit increased locomotion behavior during the early hours of darkness, we choose to analyze circadian hours 19, 20, 21, 22, and 23 for locomotive behaviors (walk and translocation). For eat-on-haunches, a non-locomotor behavior, we chose to analyze all twelve hours of darkness. Repeated measures ANOVA; G: genotype effect. a. SOD1G93A mice exhibit a decrease in walk behavior at P30 and P58 (seconds/hour [s/hr]) (P30: genotype effect: p=0.042; P58: genotype effect: p=0.022). The walk deficit resolves at P86 (genotype effect: p=0.447). (Circadian hours analyzed: 19, 20, 21, 22, and 23). b. The SOD1G93A mice exhibit a transient decrease in translocation at P30, which reverses by P86 (pixels moved/hour). Translocation corresponds to movement in the x- and y-directions. P30: genotype effect: p=0.014; P58: genotype effect: p=0.901. At P86, the SOD1G93A mice exhibit an increase in translocation. P86: genotype effect: p=0.028. (Circadian hours analyzed: 19, 20, 21, 22, and 23). 69 c. SOD1G93A mice exhibit a later onset decrease in eating-on-haunches at P58 and P86 (s/hr). P30: genotype effect: p=0.296; P58: genotype effect: p=0.018; P86: genotype effect: p=0.025 (twelve hours of darkness analyzed: 19, 20, 21, 22, 23, 0, 1, 2, 3, 4, 5, 6). (See supplementary Figure 2 for re-scaled graphs). The early walk deficit in SOD1G93A mice is transient. As expected, mice exhibit specific behaviors more during the dark (e.g. walking) and specific behaviors more during the light (e.g. resting), as mice are nocturnal. It has been demonstrated that WT mice exhibit more locomotion during the early hours of darkness, and therefore we conducted our analyses on locomotor behavior (walk and translocation) during these hours (see methods) [88, 89]. At P30, we found an SOD1G93A walk deficit that persists at P58 (n=5 SOD1G93A, 6 WT; P30: Genotype effect [G]: p=0.042; p58: G: p=0.022) (Fig. 2a). To characterize these differences we used repeated measures ANOVAs across the early hours of darkness (see methods). To determine whether the SOD1G93A P30 walk deficit persists, we conducted ACBM on these mice at P58, and again found a deficit in walking behavior (P58: G: p=0.022) (Fig. 2a). To further examine potential progression of the walk deficit with age, we analyzed these mice at P86, and interestingly, at this age, the walk phenotype is absent (Fig. 2a). Therefore, the SOD1G93A walking deficit is a transient phenotype. Disease progression reveals a decrease in SOD1G93A eating-on-haunches behavior at P58 and P86. In addition to the P30 early onset phenotypes, we identified a later P58 onset phenotype. At P58 and P86, the SOD1G93A mice exhibit decreased eating-on- 70 haunches behavior (P58: G: p=0.025; P86: G: p=0.018) (Fig. 2c, Supplementary Fig. 2). The appearance of this phenotype at P58 suggests that an eating-on-haunches phenotype may be less susceptible to pathology associated with the SOD1G93A mutation than walk. ACBM also revealed an increase in SOD1G93A drink behavior at P58 (P58: G: p=0.032) (Supplementary Fig. 3, 4). In contrast, ACBM did not detect eating-from- hopper, hanging, or resting behavioral phenotypes at any of the three ages tested (Supplementary Fig. 5). Analysis of translocation reveals a selective increase in vertical movement with age. The early deficit in walk prompted us to examine translocation, which is a broader measure of movement (number of pixels moved per hour). Translocation behavior can be analyzed as the total translocation in both the x- and y-coordinates combined, or translocation in just the x- or the y-coordinate. To compare to walk behavior, we analyzed the same hours of early darkness. Starting at P30, we first analyzed combined x- and y- translocation and found that the SOD1G93A mice exhibit an early P30 deficit in translocation compared to WT mice (G: P=0.014) (Fig. 2b, Supplementary Fig. 2b). To determine whether this deficit persisted, we then conducted the same analysis at P58 and found that the translocation phenotype was absent, suggesting that translocation represents a transient phenotype (G: p=0.901) (Fig. 2b, Supplementary Fig. 2b). Interestingly, 71 when conducting the same analysis at P86, we found that the SOD1G93A mice exhibit an increase in translocation (G: P=0.028) (Fig. 2b, Supplementary Fig. 2b). To explore what could be contributing to this phenotype development, we broke down translocation into its x- and y-component parts since translocation is an accumulation of movement in both the x- and y-coordinates [√(x2+y2]. At P30, similarly to the decrease in translocation in the combined x- and y-coordinates, we found that there was a decrease in translocation in each coordinate separately (X: G: p=0.009; Y: G: p=0.046) (Fig. 3a, Supplementary Fig. 6). At P58, again separating x- and y-coordinate translocation, we found that there was no phenotype in either x- nor y-direction translocation (X: G: p=0.633; Y: G: p=0.411) (Fig. 3b; Supplementary Fig. 6). However, at P86, we identified an increase in translocation specifically in the y-axis, but not the x-axis, suggesting that increased movement in the y-direction is driving the overall SOD1G93A increased translocation behavior (X: G: p=0.156; Y: G: p=0.002) (Fig. 3c, Supplementary Fig. 6). In examining the graphs, we noticed that at P58 while there was no translocation phenotype in the first five hours of darkness, there was an observed increase in SOD1G93A translocation in the later hours of darkness (Supplementary Fig. 7). We decided to follow-up on this observation and analyzed combined x- and y-coordinate translocation in the last five hours of darkness at P58, and found that within these hours there was indeed an increase in SOD1G93A translocation (G: p=0.010). To again determine what direction of translocation could be driving this effect, we broke down x- and y-direction translocation during these late hours of 72 darkness and found a marginal SOD1G93A increase in x-direction translocation and a significant increase in y-direction translocation (X: G: p=0.078; Y: G: p=0.001) (Supplementary Fig. 7). This result demonstrates that the increase in SOD1G93A translocation behavior is actually initiating at P58, driven primarily by Y-direction movement, and this effect is specific to the later hours of darkness. Figure 3. Separating x- and y-translocation demonstrates that the overall translocation can be driven by movement in just one direction. a. At P30, there was an SOD1G93A translocation deficit in x- and y-coordinates combined, the x-coordinate alone, and the y-coordinate alone (X- and Y- translocation combined: G: p=0.014; X-coordinate: G: p=0.009; Y-coordinate: G: p=0.046). b. At P58, there is no translocation phenotype in either the x- and y-coordinates combined, the x-coordinate alone or the y-coordinate alone (X- and Y-translocation combined: G: p=0.901; X-coordinate translocation alone: p=0.633; Y-coordinate translocation alone: G: p=0.411). 73 The absence of an effect in the x-coordinate of translocation demonstrates that translocation does not represent walking behavior, as walking is an x-coordinate behavior, and there is an SOD1G93A walk deficit at this age. c.) At P86, there is an SOD1G93A increase in translocation in the x- and y-coordinates combined as well as in the y-coordinate alone. There is an absence of an SOD1G93A phenotype in the x-direction (x- and y-translocation combined: G: p=0.028; x- coordinate translocation alone: p=0.156; y-coordinate translocation alone: G: p=0.002). The absence of an SOD1G93A phenotype at P86 in the x-coordinate translocation, and the presence of a y-coordinate translocation phenotype demonstrate that the y-coordinate behavior is driving the overall translocation effect. Repeated measures ANOVA, circadian hours: 19, 20, 21, 22, and 23. This study is the first to identify behavioral phenotypes in the transgenic SOD1G93A mice as early as P30, providing insight into the functional properties of early pathologies (Supplementary Table 1). Additionally, this study demonstrates that ACBM technology can be used to identify and quantify subtle behavioral phenotypes that may go undetected using traditional behavioral testing. Discussion Our study demonstrates that behavioral phenotypes are present within the transgenic SOD1G93A mouse as young as P30, the earliest behavioral phenotypes described in this mouse model of ALS. We identified three types of behavioral phenotypes: early and transient, later onset, and phenotype reversal with age. Characterizing early behavioral phenotypes in the SOD1G93A ALS mouse model may shine line on underlying ALS disease mechanisms. There are many early disease pathologies including denervation, axonopathies, synaptic vesicle stalling, and molecular abnormalities that could relate to the described behavioral phenotypes. 74 At P30 and P56, we found that the SOD1G93A mice exhibit an early walk deficit that normalizes by P86 (Fig. 2a). One of the earliest pathologies that occur during ALS disease progression is dying back of the motor neuron from the muscle – denervation. To compensate for this lack of pre-synaptic input, axonal sprouting occurs, in which axon collaterals from neighboring motor neurons sprout to re- innervate the denervated synapses [35-37]. The early walk deficit that we observe at P30 and P58, followed by a recovery at P86 may correspond to early denervation followed by compensatory axonal sprouting. In transgenic SOD1G93A mice, Pun et al. (2006) demonstrated early denervation of hind limb muscles. This denervation may explain the early behavioral deficits that we have identified through ACBM [35]. They found that different motor neuron types exhibit selective vulnerability to denervation, as well as selective compensatory axonal sprouting. They found that fast fatigable (FF) motor neurons are most vulnerable to denervation, followed by fast fatigue- resistant (FR) motor neurons. FF motor neurons that denervate around P50 were reinnervated through compensatory sprouting of nearby axon collaterals of FR and S motor neuron axons at ~P60. These re-innervated synapses again underwent denervation between P80 and P90 [35]. This timing of denervation followed by re- innervation mirrors the SOD1G93A walking phenotypes we observed from P30-P86 (Fig. 2a). Kaplan et al. (2014) demonstrated further corroborating data indicating early denervation in the SOD1G93A mouse [36]. They found that denervation is present within the TA at P50 and within the soleus at P100 demonstrating that 75 different muscles exhibit different levels of vulnerability related to their innervating motor neurons. This early denervation has been observed in both the B6.Cg- Tg(SOD1*G93A)1Gur/J and the B6SJL-Tg(SOD1*G93A)1Gur/J models. Even earlier pathology is demonstrated by Frey et al. (2000) in which SOD1G93A mice exhibit axonal thinning in the peripheral branches of the medial gastrocnemius muscle as early as P30, followed by dramatic denervation between P45 and P80 [37]. In addition to axonal thinning and denervation, Pun et al. (2006) described early axonal transport abnormalities in FF and FR motor neurons through the identification of synaptic vesicle accumulations present as early as P35 -P38 [35]. Because the timing of the walk deficit we identified at p30 and p58 is within the timing of FF motor neuron denervation, axonal thinning, and axonal transport abnormalities, it is possible that this walk deficit is a behavioral corollary of these pathologies. At P86, the walk deficit that we observed is no longer present, suggesting compensatory sprouting of FR and S motor neurons, as demonstrated by Pun et al. (2006). To further delve into early disease mechanisms that could result in these behavioral phenotypes, one could probe molecular differences within the SOD1G93A mice. A specific protein that is present only in motor neurons that denervate early could convey motor neuron susceptibility. Kaplan et al. (2014) demonstrated that metalloproteinase-9 (MMP-9) is absent in slow (S) motor neurons of the spinal cord and is selectively expressed in FF motor neurons [36]. Additionally, in SOD1G93A mice that lack MMP-9, denervation was not detected until p100, suggesting that 76 MMP-9 confers vulnerability [36]. These results suggest that MMP-9 within motor neurons could be a potential target for pharmacological therapy development. Given our behavioral data suggesting that early walking deficits may correlate with pathologies prior to axonal sprouting, ACBM could be used to determine if SOD1G93A mice lacking MMP-9 rescue the walking deficit. Another possible explanation for the presence of early walking deficits in the SOD1G93A mouse is a central nervous system abnormality. Central pattern generators are specific circuits within the CNS that generate the rhythm of motoneuron bursts leading to well-controlled neuromuscular activation that coordinates complex patterns of stereotyped, automatic behaviors in the absence of sensory input, such as walking [90] [91] [92-95] [96]. A CPG can be thought of as a locomotor module within the spinal cord that contains spatio-temporal information to drive specific muscles to perform specific movements. Through measuring EMG activity in 24 different muscles simultaneously during walking motion in neonates, toddlers, preschoolers, and adults, Dominici et al. (2011) demonstrated that CPGs change naturally during development, and that the older the child, the more similar its EMG activity is to that of the adults [97]. They found a similar result in rats [97]. To theorize on the early SOD1G93A walk deficit followed by normalization that we identified, it is possible that this phenotype has to do with differences in CPG development between the SOD1G93A and WT mice, in which the early deficit is a result of an underdeveloped CPG causing a decrease in automaticity. The normalization in walking at P86 could be due to a modified CPG that changed during development to better maintain efficient and more coordinated walk behavior. 77 When the CNS is damaged through stroke, injury or disease, CPGs can become uncoordinated and lose automaticity. Patients who have CNS impairments often exhibit this decrease in automaticity in locomotion, difficulty in performing two tasks simultaneously, increased executive locomotor control, and exhibit an increase in prefrontal cortex activity during walking, again providing evidence that the walking deficit we identified in the SOD1G93A mice could be related to a central nervous system deficit such as abnormal CPGs that regulate walking [96]. While not consistent across the three ages, we also identified an SOD1G93A increase in drinking behavior specifically at P58 (Supplementary Figures. 3, 4). One possibility as to why we are seeing this increased drinking behavior could be due to a swallowing deficit. 25-30% of human patients with ALS exhibit a bulbar onset affecting swallowing, speech and vocal production. Patients with swallowing deficits tend to have increased mealtime due to smaller bites and sips and additional required attention to swallowing [98]. To better characterize swallowing deficits in an animal correlate of ALS, Lever et al. (2009) measured lick and mastication rates at three different ages in the SOD1G93A mouse model – P60, P110, and P140 and demonstrated that the SOD1G93A mice exhibited a decreased lick rate at all three ages measured in comparison to WT mice [99]. Interestingly, despite the decreased lick rate, the SOD1G93A mice and WT mice did not differ in their water consumption until the P140 time point measurement, at which point the SOD1G93A mice consumed less water than WT. This data suggests that while the SOD1G93A mice may be exhibiting a decreased rate of licking during short bouts of time starting at P60, the overall amount of water they 78 drink ad-libitum is comparable to that of wild-type mice, suggesting that they are compensating by spending more time drinking. These data are interesting as the compensation hypothesis mirrors what we found in ACBM drinking behavior of SOD1G93A mice. We identified an overall increase in drinking behavior, and this phenotype could be a compensation for inefficient drinking, as appeared to be the case in the Lever et al paper. A later onset behavioral phenotype that we detected at p58 and p86 was an SOD1G93A eating-on-haunches deficit (Fig. 2c). It is interesting that the eating-on- haunches behavior and the drinking behavior have reversed phenotypes. One possible reason for this result is that while the increased drinking behavior is possibly a compensation mechanism, there may not be any such mechanism for a potential swallowing deficit during eating-on-haunches. Lever et al. (2009) also measured mastication rate and found that the SOD1G93A mice exhibited decreased mastication rate at all three times measured (P60, P110, and P140) in comparison to WT mice [99]. It is possible that eating-on- haunches and rate of mastication may therefore be correlated. Additionally, the decreased SOD1G93A eating-on-haunches behavior could be due to abnormal metabolism. Smittkamp et al. (2014) measured metabolism through quantifying the respiratory exchange ratio of SOD1G93A mice at P30-P40, P65-P75 and P105-P115 in SOD1G93A mice and found that after giving the mice a food pellet, there was no difference in the respiratory exchange ratio between the two genotypes at P30-P40 nor at P105-P115, however the SOD1G93A mice exhibited an increased respiratory exchange ratio at P65-P75 [100]. In other words, in the 79 same amount of time it takes the WT mice to produce approximately 0.90 units of CO2, the SOD1G93A mice produce approximately 0.95 units of CO2, demonstrating that during P65-P75, the SOD1G93A mice exhibit a decreased metabolic efficiency [100]. These data may corroborate the decreased eating-on-haunches behavior that we identified in the SOD1G93A mice within this age range. Perhaps the decreased metabolic efficiency results in less of a drive to eat, decreasing the eating-on- haunches behavior. In addition to the nine ACBM behaviors, we also characterized translocation – a behavior that measures any movement in the x- and y-directions. Translocation provides a more general measure of movement regardless of behavior. Because specific ACBM behaviors are combined in measuring translocation, translocation incorporates a larger amount of data than any individual behavior. We identified a decrease in SOD1G93A translocation at P30, and this phenotype is lost at P58; however at P86, we identified an increase in SOD1G93A translocation (Fig. 2b). In separating the translocation data into movement in the x-and y-directions, we first found an SOD1G93A deficit in both the x- and y-coordinates separately at P30, explaining the overall translocation deficit at this age (X: G: p=0.009; Y: G: p=0.046) (Fig. 3a). Additionally, this x-coordinate translocation deficit mirrors the significant walk deficit at this age, suggesting that walk is driving the x-coordinate translocation phenotype. The loss in a translocation phenotype at P58 would suggest that there is no difference in movement between genotypes. One could then ask: are there specific behavioral phenotypes at P58 that when combined in a measure of overall 80 translocation, cancel out? At P58, SOD1G93A mice are drinking more, however are walking and eating-on-haunches less (Supplementary Fig. 3, 4, and Fig. 2a, c). The varying contributions of each of these behaviors (increase in drinking, and decrease in walking and eating-on-haunches) may result in an absence of a translocation phenotype. At P58, we also noticed that despite the absence of a translocation phenotype during the first five hours of darkness, during later hours of darkness, there was an apparent increase in translocation (Supplementary Fig. 7). Based on this observation, we analyzed the last five hours of darkness using a repeated measures ANOVA, and identified an increase in overall translocation, a marginal increase in x- direction translocation and a significant increase in y-direction translocation. Although we chose to analyze a separate sequence of hours later in the night based on observing the graphs post-hoc, this increase in translocation during later hours of darkness suggests that the increase in translocation phenotype is actually initiating at P58 and expands to incorporate the earlier hours of darkness at P86. Because the deficit in walking behavior is absent by P86, it is likely that the increase in SOD1G93A translocation is due to an increase in movement in the y-direction. To further investigate what behaviors are contributing to the increase in SOD1G93A translocation at P86, we examined specific behaviors that exhibit movement in the y-direction. Of the nine behaviors examined, the ones that exhibit movement primarily in the y-direction are drink, eat-from-hopper, hang and rear suggesting that it is one of these or a combination of these four behaviors that differentiate between SOD1G93A and WT mice and that drive the increased 81 translocation phenotype at P86. Of those four behaviors, the SOD1G93A mice exhibit a trend toward increased drinking, suggesting that the increase in y-coordinate translocation may be driven by a possible increase in drinking as well as contributions from other movement in the y-direction. This increase in y-direction movement and absence in x-direction movement at P86 would explain the translocation phenotype reversal from P30 to P86. There were some behaviors that did not exhibit any phenotype across the three ages measured: eating-from-hopper, hanging, rearing and resting (Supplementary. Fig. 5). An absence in an eating-from-hopper phenotype demonstrates that the SOD1G93A mice approach the food hopper as often as the WT mice. Whether the SOD1G93A mice remain at the food hopper in shorter bouts – that is yet to be determined. The same concept of different bout lengths between genotypes could be applied to the apparent lack of hanging, rearing and resting phenotypes. In studying behavior in mice, steps need to be taken to minimize variability. A major benefit of ACBM over traditional behavioral testing is its inherent low level of variability, requiring minimal human interaction, other than setting up the mice in their respective cages. In comparison, traditional behavioral testing is subject to additional variables such as more handling and an inherent lack of standardization, and these variables can result in conflicting results [101] [72]. Given our findings of early behavioral phenotypes in the transgenic SOD1G93A mouse model of ALS, such as a walking deficit, in our following study (described in subsequent chapters) we chose to use ACBM to detect early behavioral phenotypes 82 in the TDP-43Q331K knock-in mouse model of ALS-FTD, as no early behavioral abnormalities have been previously described in this mouse. Conclusion Our study identified behavioral phenotypes (walk, translocation, eat-on- haunches, and drink) in the transgenic SOD1G93A ALS mouse model as early as P30, prior to the onset of neurodegeneration. This is the earliest that any behavioral phenotypes have been identified in this mouse. This is also the first study to use ACBM to characterize a neurodegenerative disease, and demonstrates the usefulness of ACBM as an effective tool to identify disease onset and progression, as well as specific behavioral abnormalities that provide insight into underlying pathologies. Early behavioral characterization in disease models provides information as to when therapeutic intervention would be most effective. Early behavioral phenotypes also indicate when specific biomarkers may present, and these biomarkers may be targets for potential intervention. ACBM could then be used as a measure to detect resulting rescue that could prolong or potentially normalize function. Methods Mice. Five SOD1G93A transgenic mice (B6.Cg-Tg(SOD1*G93A)1Gur/J; Jax Stock No: 004435) and six wild-type (WT) mice (C57BL/6J; Jax Stock No: 000664) were 83 purchased from The Jackson Laboratory. In comparison to the B6SJL- Tg(SOD1*G93A)1Gur/J mice, these mice exhibit a delayed phenotype onset. Automated Continuous Behavioral Monitoring Technology. We used ACBM to characterize the behavior of five SOD1G93A transgenic and six WT mice at three early ages (P30, P58, and P86) as one of nine different behaviors over five days of recording (see below for details). Definitions of Behaviors. The nine behaviors examined were: walk, hang, rear, eat-from-hopper, eat- on-haunches, drink, groom, rest and sniff. Walking is defined as moving in the x-direction with all four paws on the ground while not sniffing. Hanging is defined as hanging upside down from the food hopper or top of the cage by four paws. Rearing is defined as moving up on its hind limbs reaching toward the top of the cage. Eating-from-hopper is defined as rearing at the food hopper with its snout in contact with the hopper. Eating-on-haunches is defined as the mouse hunched over, holding a food pellet and eating it with its front paws. Drinking is defined as rearing at the waterspout, with its mouth in contact with the spout. Resting is defined as lying immobile at the bottom of the cage. Sniffing is defined as moving its snout in micromovements. To establish a general measure of movement within the cage, we defined “translocation” as the average number of pixels moved by the mouse per hour (for 84 each session) in either the x-direction, y-direction, or a combination of both x- and y- directions [√(x2 + y2)]. ACBM Technology Feature Extraction and Behavioral Classification. At each of the three ages, we video-recorded the eleven individually housed mice over five days. Position and velocity-based features characterize the posture of the mouse within the foreground pixels according to its x- y- coordinates. These ten features provide information such as whether the animal is in a horizontal or vertical position, how close it is to the food hopper, the waterspout, the sides and top of the cage, and its instantaneous velocity and acceleration [83]. Because there are variations between cameras in terms of the distance of the camera from the cage and the camera angle, the position- and velocity-based features need to be normalized. Without normalization, the positions of the mice in different cages will not be comparable. The position of a mouse in one cage videoed by a camera at a slight angle will not be the same as the position of a mouse in a different cage videoed by a camera tilted at a larger angle. For each camera, the normalization occurs by manually defining the specific coordinates of each cage: the water-spout, and the bottom, top, right and left corners of the cage. Therefore, the position of each mouse is always in the context of the defined coordinates within each cage [83]. Motion features are then characterized in which 300 previously defined motion features (templates) are compared to the novel motion features extracted 85 from the video recordings. The template motion features with maximum similarity to the novel motion features are used to characterize the novel motion features. This technology was modeled after the organization of motion processing in the dorsal stream in the primate visual cortex, in which low-level features, such as movement in a single direction, are combined to generate more complex representations of motion [83]. These position, velocity, acceleration, and motion-based features are then fed through a Support Vector Machine Hidden Markov Model (SVMHMM) classifier, which characterizes every frame of video data as one of the nine pre-defined behaviors. Extracted position and motion features that are similar to each other will be grouped, and behavioral classification is based on such groupings. Quantification and Graph Generation. The behavioral classification data are then quantified by binning the average number of seconds/circadian hour that each mouse spends performing one of the nine behaviors over the course of each day, and the data is then averaged over the five days of recording. Each hour of the five days of ACBM data is then averaged for each genotype. Graphs are then generated for each behavior where the x-axis represents the circadian hour (0-23) and the y-axis represents the average number of seconds/hour for each five-day session that the mouse is performing that behavior. The two genotypes are overlaid to visualize differences between genotypes in line graphs [83]. 86 Development and Training of the ACBM System. Data consisting of 20 hours of videos and 40 animals were previously collected at the Brown Rodent Neurodevelopmental Behavioral Testing Facility and used to train the SVMHMM classifier. The video data samples were annotated through human annotation as one of nine behaviors: walk, hang, rear, eat-from- hopper, eat-on-haunches, drink, groom, rest and sniff. Accuracy of the system was assessed by cross-validating the ACBM behavioral characterizations with that of human annotators [83]. SOD1G93A Transgenic Mouse Data Collection. ACBM was conducted at the Brown University Rodent Neurodevelopmental Behavioral Testing Facility, and the Brown University Animal Care and Use Committee approved all procedures. Five SOD1G93A transgenic (B6.Cg-Tg(SOD1*G93A)1Gur/J) and six WT (C57BL6) mice from Jackson Laboratories were individually housed during five days of ACBM. We conducted three rounds of ACBM at P30, P58, and P86, and between rounds, mice were group housed. During ACBM, each cage was video-monitored with a Firefly MV 0.3 MP Mono FireWire 1394a (Micron MT9V022) (https://www.ptgrey.com/firefly-mv-03-mp-mono-firewire-1394a-micron- mt9v022-camera), and each camera had a frame rate of 30 frames per second. All cameras were connected with a firewire card, and were connected to an Ubuntu 14.04 workstation. 87 After five days of continuous recording, the mice were put back into group housing, and the video data underwent processing in which position, velocity, acceleration and motion features were extracted (as described above). ACBM Statistical Analysis. For each behavior, a general linear model repeated measures analysis of variance (ANOVA) was used to identify genotype effects (between-subjects variable). For behaviors where activity is present during the early part of the dark (walk and translocation), repeated measures ANOVAs were conducted across only those five hours (circadian hours 19, 20, 21, 22, and 23). For behaviors where activity is present throughout the course of the night, the analysis is limited to only the twelve hours of darkness (eating-on-haunches and drink). For behaviors that are primarily limited to the twelve ours of light, analysis is limited to those twelve hours. For P58 translocation, we ran an additional repeated measures ANOVA on the last five hours of darkness (circadian hours 2, 3, 4, 5, and 6). Standard error was used to generate error bars, and statistical significance is defined as p≤0.05. Translocation. ACBM characterizes and quantifies translocation by measuring the average number of pixels moved per hour. Graphs are generated in the same manner as they are for the other ACBM behaviors. The y-axis represents the average number of pixels moved/hour/5 days of recording and the x-axis represents the circadian hour 88 (0-23). The data are extracted from x- and y-directions separately, and also combined into a single measurement of movement [√(x2+y2)]. Supplementary Figures. Supplementary Figure 1. ACBM detects SOD1G93A early walk deficit. Scatter plot of walk data from Fig. 2. The x-coordinate corresponds to circadian hours 0-23. For walk, the y-coordinate corresponds to the average # of seconds/hour that the mice are performing that behavior. Because mice exhibit increased locomotion behavior during the early hours of darkness, we choose to analyze circadian hours 19, 20, 21, 22, and 23 for walk. Repeated measures ANOVA; G: genotype effect. SOD1G93A mice exhibit a decrease in walk behavior at P30 and P58 (seconds/hour [s/hr]) (P30: genotype effect: p=0.042; P58: genotype effect: p=0.022). The walk deficit resolves at P86 (genotype effect: p=0.447). (Circadian hours analyzed: 19, 20, 21, 22, and 23). 89 Supplementary Figure 2. Re-scaled Y-Axis: ACBM detects SOD1G93A early and transient walk and translocation deficit at P30 and eating-on-haunches deficit P58. The x-coordinate corresponds to circadian hours 0-23. For walk and eat-on- haunches, the y-coordinate corresponds to the average # of seconds/hour that the mice are performing that behavior. For translocation, the y-coordinate corresponds to the average number of pixels/hour that the mouse is moving in the x- and y- directions. Because mice exhibit increased locomotion behavior during the early hours of darkness, we choose to analyze circadian hours 19, 20, 21, 22, and 23 for locomotive behaviors (walk and translocation). For eat by hand, a non-locomotor behavior, we chose to analyze all twelve hours of darkness. Repeated measures ANOVA; G: genotype effect. a. SOD1G93A mice exhibit a decrease in walk behavior at P30 and P58 (seconds/hour [s/hr]) (P30: genotype effect: p=0.042; P58: genotype effect: p=0.022). The walk deficit resolves at P86 (genotype effect: p=0.447). (Circadian hours analyzed: 19, 20, 21, 22, and 23). b. The SOD1G93A mice exhibit a transient decrease in translocation at P30, which reverses by P86 (pixels moved/hour). Translocation corresponds to movement in the x- and y-directions. P30: genotype effect: p=0.014; P58: genotype effect: p=0.901. At P86, the SOD1G93A mice exhibit an increase in translocation. P86: genotype effect: p=0.028. (Circadian hours analyzed: 19, 20, 21, 22, and 23). 90 c. SOD1G93A mice exhibit a later onset decrease in eating-on-haunches at P58 and P86 (s/hr). P30: genotype effect: p=0.296; P58: genotype effect: p=0.018; P86: genotype effect: p=0.025 (twelve hours of darkness analyzed: 19, 20, 21, 22, 23, 0, 1, 2, 3, 4, 5, 6). Supplementary Figure 3. SOD1G93A mice exhibit an increase in drinking behavior at P58. The increase in drinking behavior could be related to a potential swallowing deficit. P30: genotype effect: p=0.170; P58: genotype effect: p=0.032; P86: genotype effect: p=0.364, Repeated measures ANOVA; Twelve hours of darkness (19, 20, 21, 22, 23, 0, 1, 2, 3, 4, 5, 6). Supplementary Figure 4. Rescaled Y-Axis: SOD1G93A mice exhibit an increase in drinking behavior at P58. The increase in drinking behavior could be related to a potential swallowing deficit. P30: genotype effect: p=0.170; P58: genotype effect: p=0.032; P86: genotype effect: 91 p=0.364, Repeated measures ANOVA; Twelve hours of darkness (19, 20, 21, 22, 23, 0, 1, 2, 3, 4, 5, 6). Supplementary Fig. 5. Eat-from-hopper, hang, rear, and rest behaviors did not exhibit a phenotype across all three ages examined. ACBM does not detect SOD1G93A phenotypes for behaviors indicated. Eat-from-hopper: P30: genotype effect: p=0.0.896; P58: genotype effect: p=0.093; P86: genotype effect: p=0.280. Hang: P30: genotype effect: p=0.0.247; P58: genotype effect: p=0.257; P86: genotype effect: p=0.094. Rear: P30: genotype effect: p=0.348; P58: genotype effect: p=0.185; P86: genotype effect: p=0.765 Rest: P30: genotype effect: p=0.990; P58: genotype effect: p=0.254; P86: genotype effect: p=0.341. For eat-from-hopper, hang and rear, repeated measures ANOVA analyses were conducted across dark hours, since these behaviors are associated with active periods. For rest, repeated measures ANOVAs were conducted during the light hours, since high levels of rest occur during the light. 92 Supplementary Figure 6. Re-scaled Y-Axis: Separating x- and y-translocation demonstrates that the overall translocation can be driven by movement in just one direction. a. At P30, there is an SOD1G93A translocation deficit in x- and y-coordinates combined, the x-coordinate alone, and the y-coordinate alone (x- and y translocation combined: G: p=0.014; x-coordinate: G: p=0.009; y-coordinate: G: p=0.046). b. At P58, there is no translocation phenotype in either the x- and y-coordinates combined, the x-coordinate alone or the y-coordinate alone (x- and y-translocation combined: G: p=0.901; x-coordinate translocation alone: p=0.633; y-coordinate translocation alone: G: p=0.411). The absence of an effect in the x-coordinate of translocation demonstrates that translocation does not represent walking behavior, as walking is an x-coordinate behavior, and there is an SOD1G93A walk deficit at this age. c.) At P86, there is an SOD1G93A increase in translocation in the x- and y-coordinates combined as well as in the y-coordinate alone. There is an absence of an SOD1G93A phenotype in the x-direction (x- and y-translocation combined: G: p=0.028; x- coordinate translocation alone: p=0.156; y-coordinate translocation alone: G: p=0.002). The absence of an SOD1G93A phenotype at P86 in the x-coordinate translocation, and the presence of a y-coordinate translocation phenotype demonstrates that the y-coordinate behavior is driving the overall translocation effect. Repeated measures ANOVA, circadian hours: 19, 20, 21, 22, and 23. 93 Supplementary Fig. 7. There is an increase in SOD1G93A y-direction translocation at P58 during late hours of darkness. Despite the absence of a translocation deficit at P58 in the x- and y-coordinates, the x-coordinate and the y- coordinate during circadian hours 19, 20, 21, 22, and 23, SOD1G93A mice exhibit an increase in x-and y-direction translocation combined, a marginal increase in x- direction translocation, and an increase in y-direction translocation during hours 2, 3, 4, 5, and 6 (x- and y-coordinates combined: G: p=0.010; x-coordinate alone: G: p=0.078, y-coordinate alone: G: p=0.001) (hours analyzed indicated by black bar beneath x-axis). This result demonstrates that an SOD1G93A deficit is present at P58 during the later hours of darkness, despite an absence of this phenotype during early hours of darkness. Supplementary Table 1. Summary of SOD1G93A ACBM Phenotypes at P30, P58 and P86. 94 CHAPTER 3: TDP-43 GAINS FUNCTION DUE TO PERTURBED AUTOREGULATION IN A TARDBP KNOCK-IN MOUSE MODEL OF ALS-FTD Matthew A. White1,2,14, Eosu Kim3,4, Amanda Duffy5, Robert Adalbert6, Benjamin U. Phillips3, Owen M. Peters7,15, Jodie Stephenson8,16, Sujeong Yang6, Francesca Massenzio1,2, Ziqiang Lin1,2, Simon Andrews1, Anne Segonds-Pichon1, Jake Metterville9, Lisa M. Saksida3,10,11, Richard Mead8, Richard R Ribchester12, Youssef Barhomi13, Thomas Serre13, Michael P. Coleman1,6, Justin R. Fallon5, Timothy J. Bussey3,10,11, Robert H. Brown Jr9 and Jemeen Sreedharan 1,2,14* 1 The Babraham Institute, Cambridge, UK. 2Maurice Wohl Clinical Neuroscience Institute, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, UK. 3Department of Psychology and MRC/Wellcome Trust Behavioural and Clinical Neuroscience Institute, University of Cambridge, Cambridge, UK. 4Department of Psychiatry, Institute of Behavioral Science in Medicine, Brain Korea 21 Plus Project for Medical Sciences, Yonsei University College of Medicine, Seoul, Republic of Korea. 5Department of Neuroscience, Brown University, Providence, RI, USA. 6John van Geest Centre for Brain Repair, University of Cambridge, Cambridge, UK. 7The Vollum Institute, Oregon Health & Science University, Portland, OR, USA. 8Sheffield Institute for Translational Neuroscience, University of Sheffield, Sheffield, UK. 9Department of Neurology, UMass Medical School, Worcester, MA, USA. 10Molecular Medicine Research Group, Robarts Research Institute & Department of Physiology and Pharmacology, Schulich School of Medicine
& Dentistry, Western University, London, ON, Canada. 11The Brain and Mind Institute, Western University, London, ON, Canada. 12SBMS, University of Edinburgh, Edinburgh, UK. 13Department of Cognitive, Linguistic and Psychological Sciences, Brown University, Providence, RI, USA. Present address: 14Maurice Wohl Clinical Neuroscience Institute, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, UK. 15School of Biosciences, Dementia Research Institute, Cardiff University, Cardiff, UK. 16Centre for Neuroscience and Trauma, Blizard Institute, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, UK. 17These authors contributed equally: Matthew A. White and Eosu Kim. *e-mail: jemeen.sreedharan@kcl.ac.uk Contributions [J. Sreedharan, M.A.W., M.P.C., R.H.B., T.J.B., J.F., R.M. and L.M.S. designed experiments. M.A.W. and J. Sreedharan performed studies on cohort 1 mice including behavioral assessments, histology and transcriptomics. E.K. performed touchscreen studies on cohort 2 mice with assistance from B.U.P. A.D. collated ACBM data and quantified NMJ innervation. R.A. performed spinal cord dissections for laser capture and histology. O.M.P. and J.M. conducted histological studies and image analysis. J. Stephenson performed motor behavioral studies. S.Y. and E.K. performed the OR assay. F.M. quantified motor neurons and western blots. Z.L. performed sequencing to exclude off- target mutagenesis events. S.A. and A.S.-P. assisted with analysis of RNA-seq data and statistical analyses, respectively. R.R.R. performed neuromuscular electrophysiological studies. Y.B. and T.S. developed ACBM software and analyzed ACBM data. J. Sreedharan wrote the manuscript with contributions from all authors.] [As published]. 95 Details of A. Duffy Contributions: I collected and analyzed ACBM TDP-43Q331K knock-in and wild-type data comprised in Figure 1c and d, and Supplementary Figure 1b. Additionally, I quantified and analyzed neuromuscular junction innervation and denervation of 5-month-old (eight mice/genotype) and 18 through 23-month-old (three mice/genotype) TDP-43Q331K and wild-type male mice (Supplementary Figures 2c, d, e). I generated figures corresponding to my data collection and analyses. 96 TDP-43 gains function due to perturbed autoregulation in a Tardbp knock-in mouse model of ALS-FTD Matthew A. White1,2§, Eosu Kim3,4§, Amanda Duffy5, Robert Adalbert6, Benjamin U. Phillips3, Owen M. Peters7, Jodie Stephenson8, Sujeong Yang6, Francesca Massenzio1,2, Ziqiang Lin1,2, Simon Andrews1, Anne Segonds-Pichon1, Jake Metterville9, Lisa M. Saksida3,10,11, Richard Mead8, Richard R Ribchester12, Youssef Barhomi13, Thomas Serre13, Michael P. Coleman1,6, Justin R. Fallon5, Timothy J. Bussey3,10,11, Robert H. Brown Jr9, Jemeen Sreedharan1,2* 1- The Babraham Institute, Cambridge, UK 2- Maurice Wohl Clinical Neuroscience Institute, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, UK 3- Department of Psychology and MRC/Wellcome Trust Behavioural and Clinical Neuroscience Institute, University of Cambridge, UK 4- Department of Psychiatry, Institute of Behavioral Science in Medicine, Brain Korea 21 Plus Project for Medical Sciences, Yonsei University College of Medicine, Seoul, Republic of Korea 5- Department of Neuroscience, Brown University, Providence, RI, USA 6- John van Geest Centre for Brain Repair, University of Cambridge, UK 7-The Vollum Institute, Oregon Health & Science University, Ohio, USA 8- Sheffield Institute for Translational Neuroscience, University of Sheffield, UK 9- Department of Neurology, UMass Medical School, Worcester, MA, USA 10- Molecular Medicine Research Group, Robarts Research Institute & Department of Physiology and Pharmacology, Schulich School of Medicine & Dentistry, Western University, London, ON, Canada 11- The Brain and Mind Institute, Western University, London, ON, Canada. 12- SBMS, University of Edinburgh, UK 13- Dept. of Cognitive, Linguistic and Psychological Sciences, Brown University, RI, USA Present addresses Matthew A. White: Maurice Wohl Clinical Neuroscience Institute, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, UK Jemeen Sreedharan: Maurice Wohl Clinical Neuroscience Institute, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, UK Owen M Peters: School of Biosciences, Dementia Research Institute, Cardiff University Jodie Stephenson: Centre for Neuroscience and Trauma, Blizard Institute, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, 4 Newark Street, London E1 2AT § these authors contributed equally 97 * corresponding author: jemeen.sreedharan@kcl.ac.uk Amyotrophic lateral sclerosis-frontotemporal dementia (ALS-FTD) constitutes a devastating disease spectrum characterised by TDP-43 pathology. Understanding how TDP-43 contributes to neurodegeneration will help direct therapeutic efforts. Here, we have created a novel TDP-43 knock-in mouse with a human-equivalent mutation in the endogenous mouse Tardbp gene. TDP- 43Q331K mice demonstrate cognitive dysfunction and a paucity of parvalbumin interneurons. Critically, TDP-43 autoregulation is perturbed leading to a gain of TDP-43 function, and altered splicing of Mapt, another pivotal dementia gene. Furthermore, a novel approach to stratify transcriptomic data by phenotype in differentially affected mutant mice reveals 471 changes linked with improved behaviour. These changes include downregulation of two known modifiers of neurodegeneration, Atxn2 and Arid4a, and upregulation of myelination and translation genes. With one base change in murine Tardbp, this study identifies TDP-43 misregulation as a pathogenic mechanism that may underpin ALS-FTD, and exploits phenotypic heterogeneity to yield candidate suppressors of neurodegenerative disease. Amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) are destructive neurodegenerative diseases that exist on a clinicopathological spectrum (ALS-FTD)1. ALS is characterised by motor impairment and FTD by executive dysfunction, language impairment and behavioural changes. Nearly all cases of ALS, half of FTD cases, and most hereditary forms of ALS and FTD are characterised by cytoplasmic mislocalisation and aggregation of the 43kDa TAR DNA-binding protein (TDP-43)2,3. Significantly, the identification of mutations in the gene encoding TDP-43 (TARDBP) as a cause of ALS and FTD confirmed that TDP-43 plays a mechanistic role in neurodegeneration4,5. This role remains undefined. 98 TDP-43 is a conserved RNA-binding protein with critical roles in splicing in the nervous system6. TDP-43 also demonstrates exquisite autoregulation by binding to its transcript, triggering alternative splicing of intron 7 within the TARDBP 3’-untranslated region (UTR) and destruction of its mRNA7. Experimentally increasing or decreasing TDP-43 levels both cause neuronal loss, but whether human neurodegeneration is caused by a gain or loss of TDP-43 function remains unclear. Modelling of mutant TDP-43 in vivo has relied on variable degrees of transgenic overexpression of TDP- 43 to replicate pathological changes seen in post-mortem human tissues8. However, TDP-43 transgenic mouse models have demonstrated that TDP-43 aggregation is not necessary to cause neurodegeneration9, and whether TDP-43 aggregation is causally linked to disease onset is unclear. A caveat of transgenic TDP-43 mouse models is that phenotypes may partly be artefacts of overexpression. Furthermore, the cell-type specific expression of single TDP-43 splice forms in transgenic models using neuronal promoters, and temporally- triggered expression of transgenes in adulthood do not reflect ubiquitous expression and alternative splicing of Tardbp, including during embryonic development10. To unravel the role of mutant TDP-43 in the disease pathogenesis we created a knock-in mouse harbouring only a human-equivalent point mutation in the endogenous mouse Tardbp gene. This model replicates the human mutant state as closely as possible, retaining the endogenous gene structure including promoters and autoregulatory 3’UTR, and maintaining the ubiquitous expression of TDP-43 during embryonic development and in adulthood. By avoiding deliberate manipulation of TDP-43 expression, this model helps elucidate both mediators and modifiers of cognitive dysfunction in ALS-FTD. Results TDP-43Q331K causes behavioural phenotypes and disproportionately affects male 99 mice Over 50 TARDBP mutations at conserved sites have been identified in ALS-FTD11. We chose to introduce the n.991C>A (p.Q331K) mutation into murine Tardbp because TDP-43Q331K is a particularly toxic species in vitro and in vivo4,9,12,13. Mutagenesis was performed using CRISPR/CAS9 methodology yielding four founders with the Q331K mutation (Fig. 1a). Mutagenesis events at predicted off- target regions and in the remainder of Tardbp were excluded by Sanger sequencing. Founder #52 was outcrossed to F4 to remove other potential off-target mutagenesis events. Heterozygous (TDP-43Q331K/+) F4 animals were intercrossed to generate mutant and wild-type mice. Homozygotes (TDP-43Q331K/Q331K) were viable (Fig. 1b, Supplementary Fig. 1a) and appeared superficially normal as juveniles. Since TDP- 43 transgenic mice have not been shown to rescue TDP-43 knockout mice, TDP- 43Q331K/Q331K knock-in mice represent a unique opportunity to study mutant TDP-43 in vivo in the absence of wild-type TDP-43. We initially screened for phenotypes in a small group of wild-type and TDP- 43Q331K/Q331K mice using automated continuous behavioural monitoring (ACBM)14. At ~4 months of age TDP-43Q331K/Q331K male and female mice demonstrated reduced walking and hanging, and increased rearing and eating-by-hand, but no alterations in circadian rhythmicity (Fig. 1c). The most consistent phenotype was reduced walking in males (Fig. 1d and Supplementary Fig. 1b). Further breeding revealed an under representation of male mutants, yet females were present at Mendelian ratios, further suggesting that males are more susceptible to deleterious effects of TDP-43Q331K (Fig. 1e). This is notable as sporadic ALS is more common in men, and TDP-43 mutations demonstrate greater penetrance in men than women15. We therefore focussed on males in subsequent studies, breeding two cohorts of mice: Cohort 1 for 100 motor, pathological and transcriptomic studies; Cohort 2 for cognitive studies. TDP-43Q331K mice demonstrate no significant motor impairment, weight gain due to hyperphagia, and transcriptomic changes in spinal motor neurons To identify ALS-like motor deficits we measured Rotarod performance in Cohort 1 mice. From ~6 months of age TDP-43Q331K/+ and TDP-43Q331K/Q331K mice demonstrated reduced Rotarod latencies (Fig. 2a). Interestingly, mutants demonstrated hyperphagia, a feature of FTD16, and gained more weight than wild-types (Fig. 2b,c). Increased weight could contribute to impaired Rotarod performance, so we tested Cohort 2 mice, which were weight-matched due to dietary control (Supplementary Fig. 2a). Weight-matched mutants performed similarly to wild-types up to 16 months of age (Fig. 2d), suggesting that mutant mice do not have significant impairment of motor coordination. To determine if mutant mice demonstrated lower motor neuron degeneration we examined spinal cords from 5-month-old mice to identify early pathological changes. Motor neurons demonstrated normal morphology and numbers with no TDP-43 aggregation or mislocalisation in TDP-43Q331K/Q331K mice (Fig. 2e,f Supplementary Fig. 2b). Quantification of neuromuscular junctions (NMJs) and succinate dehydrogenase staining in gastrocnemius muscles were normal in TDP-43Q331K/Q331K mice, suggesting no significant denervation (Supplementary Fig. 2c,d,f). Examination of 18 to 23-month-old mice similarly found no evidence of denervation (Supplementary Fig. 2e) and no electrophysiological evidence of motor unit loss (Fig. 2g, Supplementary Fig. 2g-o). Collectively, these data indicated a remarkable resilience of neuromuscular units to TDP-43Q331K. We hypothesised that gene expression changes occurring in motor 101 neurons of mutant mice could elucidate how these cells respond to cellular stress caused by TDP-43Q331K. We isolated RNA from laser-captured lumbar motor neurons from 5-month-old mice and performed RNASeq (Supplementary Fig. 3a,b). This yielded 31 significant expression and splicing differences between wild-type and TDP- 43Q331K/Q331K mice (Fig. 2h,i Supplementary Fig. 3c-e, Supplementary Table. 1). A notable change was upregulation of Agrin. Agrin is secreted by neurons and functions through muscle specific kinase to cluster acetylcholine receptors at NMJs17. Agrin upregulation may therefore promote NMJ function in TDP-43Q331K/Q331K mice. Interestingly, the largest gene expression change was a three-fold increase in expression of aldehyde oxidase 1 (Aox1). Little is known about the neurobiological functions of AOX1 although its transcript has been observed in the anterior horn of the spinal cord18. AOX1 catalyses the conversion of retinaldehyde to retinoic acid (RA)19, which functions in neuronal maintenance in the adult nervous system and following axon injury. Thus, Aox1 upregulation may benefit motor neurons in TDP- 43Q331K/Q331K mice. Immunostaining revealed expression of AOX1 in spinal motor neurons (Fig. 2j), but no difference in expression between TDP-43Q331K/Q331K and wild- type mice (Fig. 2k, Supplementary Fig. 3f). This could be because upregulated AOX1 is transported into peripheral motor axons, as we found abundant expression of AOX1 in motor axons (Fig. 2k). TDP-43Q331K mice display executive dysfunction, memory impairment and phenotypic heterogeneity In parallel with motor studies, to determine if TDP-43Q331K causes FTD-like cognitive dysfunction we performed neuropsychological assessments on Cohort 2 mice using touchscreen operant technology. To test if mice exhibited FTD-related deficits we conducted the 5-choice serial reaction time task (5-CSRTT; Fig. 3a), which measures frontal/executive function including attention, perseveration, impulsivity, and 102 psychomotor speed22. At 4 months of age the number of training sessions required to reach performance criteria for probe testing was higher in TDP-43Q331K/Q331K mice than wild-types (Fig. 3b), indicating learning deficits in mutants. Following training, animals underwent probe testing at 6 and 12 months of age. Accuracy (Fig. 3c,d insets) and omission percentage were comparable between genotypes at 6 months of age (Fig. 3c). However, at 12 months of age, while accuracy remained normal, omission percentage was greater in TDP-43Q331K/+ and TDP-43Q331K/Q331K mice (Fig. 3d), suggesting attentional deficits and cognitive decline in mutants. Reward collection and response latencies, and premature and perseverative response rates were similar between genotypes (Supplementary Fig. 4a-h), arguing against visual, motivational, or significant motor deficits as causes for increased omissions. We also measured motivation using fixed (FR) and progressive-ratio (PR) schedules. No significant differences were found between genotypes, further suggesting that increased omissions in mutants were not due to motivational deficits (Fig. 3e,f). Collectively, these data indicate an inattention phenotype in TDP-43Q331K/+ and TDP- 43Q331K/Q331K mice, which is consistent with frontal/executive dysfunction. Next, to explore temporal lobe-dependent function, we conducted the spontaneous object recognition task, a test of declarative memory. Initial exploratory times did not differ between genotypes (Fig. 3g), but in the choice phase a deficit emerged in TDP- 43Q331K/+ and TDP-43Q331K/Q331K mice (Fig. 3h), indicating memory impairment. The combination of executive dysfunction and memory impairment, together with hyperphagia in free-fed Cohort 1 mice led us to conclude that TDP-43Q331K/+ and TDP- 43Q331K/Q331K mice recapitulate FTD at the behavioural level. 103 During touchscreen analyses we noted that some Cohort 2 mutant mice demonstrated consistently worse performance than other mutants (Fig. 3b, Supplementary Fig. 4i). This phenotypic heterogeneity was intriguing given that the mutant mice were genetically homogeneous. Furthermore, ALS-FTD is a remarkably heterogeneous disease in which patients display varying phenotypic severity and different rates of disease progression. Indeed, TARDBP mutation carriers demonstrate variable penetrance even with homozygous mutations15. We therefore looked for further evidence of phenotypic heterogeneity by examining Cohort 1 mice using the marble-burying assay, a measure of innate digging behaviour23. From 5 to 18 months of age, wild-type mice buried ~80% of marbles. Mutants demonstrated a range of digging behaviours, with some animals behaving similarly to wild-types, but others demonstrating attenuated digging behaviour (Fig. 3i, Supplementary Fig. 4j). These observations confirm the presence of phenotypic heterogeneity in genetically homogeneous groups of mutant mice, and suggest that some mutants were relatively resistant to behavioural deficits caused by TDP-43Q331K. TDP-43Q331K/Q331K mice demonstrate perturbed TDP-43 autoregulation and reduced parvalbumin-positive neurons To obtain mechanistic insight into the cognitive dysfunction caused by TDP-43Q331K we sacrificed 5-month-old mice for pathological and transcriptomic studies. Prior to sacrifice we performed the marble-burying assay to identify animals with a range of different behaviours (Fig. 4a). Analysis of frontal cortices from wild-type and TDP- 43Q331K/Q331K mice demonstrated no significant reduction in cortical thickness or cellular density in mutants (Fig. 4b, Supplementary Fig. 5a-c), and no nuclear clearing or cytoplasmic aggregation of TDP-43 (Fig. 4c). However, subcellular fractionation and immunoblotting demonstrated a ~45% increase in nuclear TDP-43 in TDP-43Q331K/Q331K compared to wild-type mice (Fig. 4d,e, Supplementary Fig. 5d). 104 TDP-43 has critical roles in RNA processing, which may be disturbed in disease. We therefore performed transcriptomic analyses using RNASeq of frontal cortices from six wild-type, six TDP-43Q331K/+, and eight TDP-43Q331K/Q331K mice (Supplementary Fig 6a). We identified 171 genes that were upregulated and 233 that were downregulated in TDP-43Q331K/Q331K mice relative to wild-type (Fig. 4f,g). TDP-43Q331K/+ mice demonstrated changes that trended in the same direction as TDP-43Q331K/Q331K mice, suggesting a dose-dependent effect of the mutation. In particular, we noted a 14% increase in expression of Tardbp in TDP-43Q331K/Q331K mice (Fig. 4h). As nuclear TDP-43 protein expression was also raised in mutants, we conclude that the Q331K mutation disturbs TDP-43 autoregulation. One notable gene that was downregulated in mutant mice was Nek1. This change is consistent with human data indicating that loss-of-function mutations in NEK1 cause ALS24,25. Another downregulated gene was Pvalb, which encodes the calcium buffering protein parvalbumin. Reduced parvalbumin immunopositivity is observed in patients with ALS and is linked with selective cellular vulnerability in ALS26. We therefore immunostained for parvalbumin and found a ~25% reduction in parvalbumin- positive cells in the frontal cortex of TDP-43Q331K/Q331K mice (Fig. 4i,j). Co-staining for TDP-43 in this affected subset of cortical neurons did not demonstrate TDP-43 mislocalisation (Fig. 4k,l). Notably, fast-spiking parvalbumin interneurons are GABAergic inhibitory cells that play a direct role in the control of attention27. We therefore conclude that a paucity of parvalbumin interneurons may be responsible for the attentional impairment of TDP-43Q331K/Q331K mice. Splicing analysis indicates a gain-of-function of TDP-43Q331K and links aberrant TDP-43 homeostasis with altered splicing of Mapt 105 TDP-43 plays key roles in alternative splicing. We therefore interrogated the cortical transcriptomic dataset further for splicing differences between mutant and wild-type mice and identified 138 splicing changes in 106 genes (Fig. 5a,b, Supplementary Fig. 6b). This included an ~80% increase in retention of Tardbp intron 7 in TDP- 43Q331K/Q331K mice (Fig. 5c,d), which will promote the production of stable mRNA species7. This confirms that TDP-43 autoregulation is perturbed in mutant mice. Another prominent change was a 2.4-fold increase in exclusion of Sort1 exon 17b, a known splicing target of TDP-43 (Fig. 5e,f). This change is consistent with a gain of function of TDP-4328. We also noted altered splicing of exons 2 and 3 of Mapt, which encodes the microtubule associated protein tau and is mutated in FTD with Parkinsonism29. We detected increased inclusion of Mapt exons 2 and 3 in TDP-43Q31K/Q331K mice (Fig. 5g- i). This is notable as inclusion of exons 2 and 3 of Mapt is associated with increased somatodendritic localization and aggregation of tau30. We immunostained wild-type and mutant frontal cortices for total tau but found no difference in the localization or aggregation of tau (Supplementary Fig. 6c). Analysis of iCLIP databases (http://icount.biolab.si/groups.html) revealed that TDP-43 binds to an intronic sequence upstream of Mapt exon 2 (Fig. 5g). This confirmed that Mapt exons 2 and 3 are likely splicing targets of TDP-43. The identification of this novel splicing effect of TDP-43 on Mapt mechanistically links these two major dementia genes. Next, to determine if TDP-43 misregulation could be responsible for temporal lobe- dependent functions we analysed hippocampal RNA extracts from male mice. We also examined hippocampi from female mice to determine if TDP-43 misregulation was restricted to male mice. Splicing analyses for Tardbp, Sort1 and Mapt were consistent with a gain of function of TDP-43 in mutant mice of both genders (Fig. 5j,k). This indicates that TDP-43 misregulation occurs beyond the frontal cortex, and 106 in both male and female mice. Finally, to confirm that our behavioural and transcriptomic observations were caused by mutant TDP-43 and not off-target CRISPR mutagenesis effects we performed the marble-burying assay in a second line of Tardbp Q331K knock-in mice, line #3, and found a similar impairment of digging behavior to line #52 mice (Supplementary Fig. 6d). We also analysed RNA from line #3 mice and observed an increase in Tardbp expression and altered splicing of Tardbp and Sort1, which is consistent with perturbed autoregulation and a gain of function of TDP-43 (Supplementary Fig. 6e). Furthermore, line #3 TDP-43Q331K/Q331K mice also demonstrated increased inclusion of exons 2 and 3 of Mapt, and a paucity of parvalbumin-positive neurons relative to wild- type mice, replicating key splicing and pathological observations made in line #52 mice (Supplementary Fig. 6e,f), TDP-43 misregulation in lumbar spinal cords of mutant mice further implicates interneurons in ALS-FTD pathogenesis Our transcriptomic profiling of frontal cortices and hippocampi elucidated a gain of function of TDP-43 in the brains of mutant mice. By contrast, spinal motor neurons from mutants did not demonstrate TDP-43 misregulation as Tardbp, Sort1 and Mapt were not differentially expressed or spliced in these cells (Fig. 6b). However, TDP-43 misregulation could occur in other cells of the spinal cord, namely glia or interneurons. We therefore analysed RNA from homogenates of lumbar spinal cord from the mice from which we had laser captured spinal motor neurons (Fig. 6a). Interestingly, spinal cord homogenates demonstrated increased expression of Tardbp, and altered splicing of Tardbp and Sort1 consistent with a gain of function of TDP-43 in mutant mice (Fig. 6c). Furthermore, spinal cords from mutant mice also 107 demonstrated increased inclusion of Mapt exon 2 (Fig. 6d). Given that Mapt expression is predominantly neuronal rather than glial this suggests that a gain of TDP-43 function occurs in interneurons of the spinal cord. Stratification of transcriptomic data from TDP-43Q331K/Q331K mice by phenotype identifies novel expression and splicing changes As stated earlier, some mutant mice appeared relatively resistant to the cognitive effects of the Q331K mutation. We wished to exploit this phenotypic heterogeneity in TDP-43Q331K/Q331K mice to identify potential modifiers of cognitive dysfunction. For this purpose we divided the frontal cortical transcriptomic data from the eight TDP- 43Q331K/Q331K mice into two subsets according to their antemortem marble-burying behaviour. We designated this the ‘MB+/-’ comparison. TDP-43Q331K/Q331K mice that dug consistently well were designated MB+, and those that dug consistently poorly were designated MB- (Fig. 7a,b). We hypothesised that transcriptomic differences between these two genotypically homogeneous groups would indicate molecular pathways that influenced the risk of developing cognitive impairment. Using this strategy we found 410 gene-expression and 61 splicing differences between MB+ and MB- groups (Fig. 7c, Supplementary Fig. 6g,h), which were entirely different to those seen in the earlier comparison with wild-type mice when all eight TDP- 43Q331K/Q331K mice were considered as one group (Fig. 4g, 5b). Interestingly, for 78% of these genes MB+ and MB- mice demonstrated opposing expression changes relative to wild-type (Fig. 7c, Supplementary Table 2 and MB+/- sections in Supplementary Table. 1). Effectively, for these genes an expression change in one direction is associated with a poor behavioural phenotype, yet an expression change in the opposite direction is associated with improved behavior. Furthermore, there was no difference in TDP-43 expression or the degree of TDP-43 gain of function as evidenced by Sort1 splicing between MB+ and MB- groups. Taken together, these 108 data indicate that the MB+/- comparison genes could be metastable modulators of TDP-43-mediated cognitive dysfunction. Significantly, two of the genes from the MB+/- comparison have previously been linked with suppression of neurodegeneration: Atxn2 and Arid4a. Compared to wild- type mice, MB+ mice demonstrated reduced Atxn2 expression, while MB- mice demonstrated increased Atxn2 expression. This is in keeping with previous observations that Atxn2 knockdown suppresses TDP-43 toxicity in yeast, Drosophila and mouse31,32. Furthermore, intermediate expansions of Atxn2 CAG repeat length is associated with ALS disease risk in humans31. Similarly, reduced expression of the chromatin-modeling gene Arid4a in MB+ mice is notable, as we previously found that loss of function mutations in hat-trick, the Drosophila orthologue of Arid4a, suppress TDP-43-mediated neurodegeneration in flies12. It is therefore likely that reduced levels of Atxn2 and Arid4a are similarly neuroprotective in TDP-43Q331K/Q331K MB+ mice. To identify the most significant pathways linked with phenotypic heterogeneity in the MB+/- comparison we cross-referenced the differential gene expression list with the Gene Ontology database for biological processes (Fig. 7c). Genes downregulated in MB+ mice were enriched for biological processes involving transcription, DNA methylation and chromatin modification. Genes upregulated in MB+ mice were enriched for processes involving protein translation and myelination, including the myelin repair gene Olig1, and Mbp, which encodes myelin basic protein (Supplementary Table 2). Furthermore, examination of the splicing gene list also identified Mbp as a candidate (Fig. 7d,e). Specifically, MB- mice demonstrated a significantly increased expression of a specific splice form, which is predicted to encode Golli-Mbp, in which three additional exons upstream of classical Mbp are 109 normally expressed in non-myelinating cells including neurons, and in immature oligodendrocytes33. Collectively, this Gene Ontology analysis identifies an association between the upregulation of protein translation and oligodendrocyte genes and improved behaviour in TDP-43Q331K/Q331K mice, and suggests that the promotion of myelin repair pathways by oligodendrocytes in a mature state contributes to improved cognition. To confirm the validity of MB+/- hits we deliberately swapped data from the worst performing MB+ mouse with that of the best performing MB- mouse. This resulted in all transcriptomic hits disappearing from the analysis (Fig. 7f). We also compared only the three best performing MB+ mice with the three worst performing MB- mice and found a diminished hit list, but which largely overlapped with the genes from the complete MB+/- comparison. Furthermore, we found two TDP-43Q331K/Q331K mice that were littermates yet demonstrated contrasting digging behaviour on repeated assessment (Fig. 7a,b). This indicated that transcriptomic differences between MB+ and MB- groups were not due to a genetic founder effect within our breeding program. Collectively, these data indicate that the MB+/- transcriptomic differences were genuinely reflective of two phenotypic subsets of young TDP-43Q331K/Q331K mice. TDP-43Q331K mice demonstrate age-related deterioration of cortical transcriptomes with altered expression and splicing of other ALS-linked genes Ageing is the greatest known risk factor for sporadic ALS-FTD. To determine the effects of ageing on TDP-43Q331K mice we performed a frontal cortical RNASeq study in 20-month-old mice (Fig. 8a,b,e,f, Supplementary Fig. 7a,b). Comparison of wild- type and mutant mice revealed transcriptomic differences that partly overlapped with the 5-month-old dataset (Fig. 8c,d,g,h). Significantly, aged mutant mice still demonstrated a gain of function of TDP-43, increased retention of Mapt exons 2 and 110 3, and reduced Nek1 and Pvalb expression. However, a broader range of transcriptomic changes was seen, further implicating inhibitory interneuronal disturbances, including downregulation of Sirt1 and Pgc-1α, which encode proteins involved in Pvalb transcription, and downregulation of GAD1/GAD67, which encodes the GABA synthetic protein glutamate decarboxylase (Supplementary Table 1). Aged mice also demonstrated downregulation of Tbk1 (encoding Tank binding protein kinase 1) (Fig. 8d), loss of function mutations of which cause ALS and FTD34,35. Several other ALS-FTD-linked genes also demonstrated significant downregulation, including Chmp2b, mutations of which cause FTD36, Erbb4, mutations of which cause ALS37, the ALS risk-linked gene Epha4a38, and the TDP-43 nuclear import factor Kpnb140. We also observed altered splicing of ALS-linked genes Matr341 (decreased exclusion of exon 14, which encodes a zinc finger domain), and Sqstm142 (Fig. 8h-j, Supplementary Fig. 7f,g). For Sqstm1 two splice variants (major and minor) were detected in wild-type and mutant mice, but a third variant was present only in mutants. This TDP-43Q331K-specific variant comprises a truncated 7th exon and a 2bp frameshift in exon 8 of Sqstm1, which is predicted to introduce a premature stop codon with loss of the C-terminal ubiquitin-associated domain of sequestosome 1 (Fig. 8j). Furthermore, Gene Ontology and pathway analysis of the RNASeq dataset in 20-month-old mice revealed many more significant networks than had been identified in 5-month-old TDP-43Q331K mice. Aged mutants demonstrated changes in processes classically linked to neurodegeneration, including protein ubiquitination, autophagy, and glutamate receptor activity, while KEGG pathway analysis highlighted ‘ALS’ and immune pathways (Fig. 8b). These pathways were not invoked in young mice (Fig. 4g). Collectively, these observations in aged mutant mice validate key transcriptomic findings in young mutants, link aberrant TDP-43 homeostasis with other key ALS-FTD-linked genes, and indicate age-related progressive changes in the cortical transcriptomes of TDP-43Q331K mice. 111 Finally, to identify transcriptomic differences associated with long-term resistance to cognitive impairment we performed an MB+/- comparison in aged mice. As most aged TDP-43Q331K/Q331K mice had progressed to an MB- state by 20 months, we compared TDP-43Q331K/+ mice, which we were able to stratify into MB+ and MB- groups. This comparison yielded only 21 differentially expressed genes, and 45 splicing differences between TDP-43Q331K/+ MB+ and MB- mice, which did not overlap with those genes identified in the MB+/- comparison of 5-month-old TDP-43Q331K/Q331K mice (Supplementary Fig. 7c-e). This suggests that aged TDP-43Q331K/+ mice are not amenable to stratification in the same way as young TDP-43Q331K/Q331K mice, and further suggests that modulation of MB+/- genes early in life has the potential to influence longer-term susceptibility to cognitive impairment secondary to aberrant TDP-43 homeostasis. Discussion Here, we show that with a single human disease-linked base change in murine Tardbp it is possible to replicate behavioural, pathological and transcriptomic features of the ALS-FTD spectrum. Significantly, by creating a model that mimics the human mutant state as closely as possible and in the absence of exogenous expression we elucidated that the Q331K mutation perturbs TDP-43 autoregulation. This leads to an increase in TDP-43 expression (effectively a gain of function defect). Interestingly, spinal cords from sporadic ALS patients and from TARDBP mutation carriers demonstrate increased TDP-43 mRNA expression, as do human stem cell-derived motor neurons with TARDBP mutations43,44. This indicates that TDP-43 misregulation could underpin the human disease state. Interestingly, lumbar motor neurons of TDP-43Q331K/Q331K mice demonstrated 112 upregulation of genes that may confer neuroprotection and did not demonstrate TDP- 43 misregulation, both of which might explain why mutant mice did not demonstrate significant neuromuscular phenotypes. By contrast, the FTD-like phenotypes in mutant mice were more significant. The identification of reduced parvalbumin expression as a possible cause for cognitive impairment in ALS-FTD is intriguing as parvalbumin interneuron loss has been observed in sporadic ALS and FTD26. As parvalbumin interneurons are GABAergic a reduction in their number could increase activity of cortical projection neurons with excitotoxic consequences. Early interneuronal dysfunction may have analogous consequences in the spinal cord and is suggested by our observation that TDP-43 autoregulation is perturbed in the spinal cord, but not in motor neurons. That TDP-43Q331K mice demonstrate a specific increase in inclusion of Mapt exons 2 and 3 is of great interest as 2N tau oligomers appear to have a greater ability to provoke tau aggregation than 0N and 1N isoforms30, and inclusion of exon 2 and 3 influence subcellular localisation and protein-protein interactions of tau45. Furthermore, in humans the H2 Mapt haplotype is associated with a greater inclusion of Mapt exon 3 and is associated with an earlier age of onset in FTD46,47. Although we did not observe clear disturbances of total tau localisation in TDP-43Q331K mice, more detailed analyses to identify specific tau isoforms are warranted. Our identification of a mechanistic link between TDP-43 and Mapt adds to growing evidence that ALS- FTD is characterised by both TDP-43 and tau pathology48. Furthermore, transcriptomic analysis of aged TDP-43Q331K mice elucidated changes in other ALS- FTD linked genes. Collectively, these findings emphasise a central role for TDP-43 in neurodegeneration. Finally, we observed phenotypic heterogeneity among mutant mice with the same 113 genotype and identified distinct transcriptomic profiles corresponding to differing phenotypes. This transcriptomic dataset contains genes already implicated in neurodegeneration, including Arid4a12, and Atxn231. The unbiased discovery of Atxn2 downregulation as a hit in our model is consistent with observations validating Atxn2 knockdown as a therapeutic approach for ALS-FTD32. Our data suggest a delicate balance in the transcriptome of the brain, which is metastable and can influence disease onset or progression. Identifying the environmental factors that influence this balance is a priority in future studies. Indeed, the strong representation of DNA methylation and chromatin modelling genes in the MB+/- comparison suggests a critical role for epigenetic influences in determining disease susceptibility. Genes with roles in protein translation and oligodendrocyte biology including myelination also feature in our list of putative disease modifiers, and it is encouraging that both these pathways have roles in neurodegenerative disease49,50. Our wider list of potential modifiers of disease is composed of over 450 gene-expression and splicing changes that are associated with improved behaviour in TDP-43Q331K/Q331K mice. We conclude that this list contains additional novel suppressors of neurodegeneration that will help direct efforts towards developing treatments for ALS-FTD. Accession code RNASeq data were deposited in the NCBI GEO database, number GSE99354 Acknowledgements We thank Babraham Institute Experimental Unit staff for technical assistance, Alexandra Weiss for technical assistance at UMMS, Michael Brodsky for assistance with CRISPR mutagenesis, the DERC morphology core at UMMS for assistance with histological preparations, and Sam Hilton for assistance with OR testing. We thank MPC lab members and J. Gallo for helpful discussions. EK is supported by a grant from the Korean Health Technology R&D Project, Korea-UK AD Collaborative Project (HI14C2173), Ministry of Health and Welfare, Republic of Korea.. SY is supported by an ARUK grant (RF-2016A-1). RHB gratefully acknowledges support from the ALS 114 Association, Project ALS, Target ALS, ALS-One, ALS Finding A Cure, and NIH grants RO1NS088689, RO1FD004127, RO1NS065847, and RO1 NS073873. JS is funded by the Motor Neuron Disease Association, the Medical Research Council UK, the Lady Edith Wolfson Fellowship Fund, and the van Geest Foundation. Author contributions JS, MAW, MPC, RHB, TB, JF, RM, and LS designed experiments. MAW and JS performed studies on Cohort 1 mice including behavioural assessments, histology and transcriptomics, EK performed touchscreen studies on Cohort 2 mice with assistance from BUP, AD collated ACBM data and quantified NMJ innervation, RA performed spinal cord dissections for laser capture and histology, OMP and JM conducted histological studies and image analysis, JoS performed motor behavioural studies, SY and EK performed the OR assay, FM quantified motor neurons and western blots, ZL performed sequencing to exclude off-target mutagenesis events, SA and ASP assisted with analysis of RNASeq data and statistical analyses respectively, RRR performed neuromuscular electrophysiological studies, YB and TS developed ACBM software and analysed ACBM data, JS wrote the manuscript with contributions from all authors. Competing Financial Interests The authors declare no competing financial interests References 1 Burrell, J. R. et al. The frontotemporal dementia-motor neuron disease continuum. Lancet 388, 919-931, doi:10.1016/S0140-6736(16)00737-6 (2016). 2 Neumann, M. et al. Ubiquitinated TDP-43 in frontotemporal lobar degeneration and amyotrophic lateral sclerosis. Science 314, 130-133 (2006). 3 Arai, T. et al. TDP-43 is a component of ubiquitin-positive tau-negative inclusions in frontotemporal lobar degeneration and amyotrophic lateral sclerosis. Biochemical and biophysical research communications 351, 602- 611 (2006). 4 Sreedharan, J. et al. TDP-43 mutations in familial and sporadic amyotrophic lateral sclerosis. Science 319, 1668-1672 (2008). 5 Benajiba, L. et al. TARDBP mutations in motoneuron disease with frontotemporal lobar degeneration. Ann Neurol 65, 470-473, doi:10.1002/ana.21612 (2009). 6 Tollervey, J. R. et al. Characterizing the RNA targets and position- dependent splicing regulation by TDP-43. Nat Neurosci 14, 452-458, 115 doi:10.1038/nn.2778 (2011). 7 Ayala, Y. M. et al. TDP-43 regulates its mRNA levels through a negative feedback loop. EMBO J 30, 277-288, doi:10.1038/emboj.2010.310 (2011). 8 Philips, T. & Rothstein, J. D. Rodent Models of Amyotrophic Lateral Sclerosis. Current protocols in pharmacology 69, 5 67 61-21, doi:10.1002/0471141755.ph0567s69 (2015). 9 Arnold, E. S. et al. ALS-linked TDP-43 mutations produce aberrant RNA splicing and adult-onset motor neuron disease without aggregation or loss of nuclear TDP-43. Proc Natl Acad Sci U S A 110, E736-745, doi:10.1073/pnas.1222809110 (2013). 10 Wu, L. S. et al. TDP-43, a neuro-pathosignature factor, is essential for early mouse embryogenesis. Genesis 48, 56-62, doi:10.1002/dvg.20584 (2010). 11 Buratti, E. Functional Significance of TDP-43 Mutations in Disease. Advances in genetics 91, 1-53, doi:10.1016/bs.adgen.2015.07.001 (2015). 12 Sreedharan, J., Neukomm, L. J., Brown, R. H., Jr. & Freeman, M. R. Age- Dependent TDP-43-Mediated Motor Neuron Degeneration Requires GSK3, hat-trick, and xmas-2. Current biology : CB, doi:10.1016/j.cub.2015.06.045 (2015). 13 Johnson, B. S. et al. TDP-43 is intrinsically aggregation-prone, and amyotrophic lateral sclerosis-linked mutations accelerate aggregation and increase toxicity. J Biol Chem 284, 20329-20339, doi:10.1074/jbc.M109.010264 (2009). 14 Jhuang, H. et al. Automated home-cage behavioural phenotyping of mice. Nature communications 1, 68, doi:10.1038/ncomms1064 (2010). 15 Borghero, G. et al. Genetic architecture of ALS in Sardinia. Neurobiol Aging 35, 2882 e2887-2882 e2812, doi:10.1016/j.neurobiolaging.2014.07.012 (2014). 16 Ahmed, R. M. et al. Assessment of Eating Behavior Disturbance and Associated Neural Networks in Frontotemporal Dementia. JAMA Neurol 73, 282-290, doi:10.1001/jamaneurol.2015.4478 (2016). 17 Burden, S. J., Yumoto, N. & Zhang, W. The role of MuSK in synapse formation and neuromuscular disease. Cold Spring Harb Perspect Biol 5, a009167, doi:10.1101/cshperspect.a009167 (2013). 18 Berger, R. et al. Analysis of aldehyde oxidase and xanthine dehydrogenase as possible candidate genes for autosomal recessive familial amyotrophic lateral sclerosis. Human Molec Genetics 21, 121-131 (1995). 19 Garattini, E., Fratelli, M. & Terao, M. The mammalian aldehyde oxidase gene family. Human genomics 4, 119-130 (2009). 20 Jiang, Y. M. et al. Gene expression profile of spinal motor neurons in sporadic amyotrophic lateral sclerosis. Ann Neurol 57, 236-251, doi:10.1002/ana.20379 (2005). 21 Kolarcik, C. L. & Bowser, R. Retinoid signaling alterations in amyotrophic lateral sclerosis. American journal of neurodegenerative disease 1, 130- 145 (2012). 22 Mar, A. C. et al. The touchscreen operant platform for assessing executive function in rats and mice. Nature protocols 8, 1985-2005, doi:10.1038/nprot.2013.123 (2013). 23 Thomas, A. et al. Marble burying reflects a repetitive and perseverative behavior more than novelty-induced anxiety. Psychopharmacology 204, 361-373, doi:10.1007/s00213-009-1466-y (2009). 116 24 Kenna, K. P. et al. NEK1 variants confer susceptibility to amyotrophic lateral sclerosis. Nat Genet 48, 1037-1042, doi:10.1038/ng.3626 (2016). 25 Brenner, D. et al. NEK1 mutations in familial amyotrophic lateral sclerosis. Brain 139, e28, doi:10.1093/brain/aww033 (2016). 26 Nihei, K., McKee, A. C. & Kowall, N. W. Patterns of neuronal degeneration in the motor cortex of amyotrophic lateral sclerosis patients. Acta Neuropathologica 86, 55-61 (1993). 27 Kim, H., Ahrlund-Richter, S., Wang, X., Deisseroth, K. & Carlen, M. Prefrontal Parvalbumin Neurons in Control of Attention. Cell 164, 208- 218, doi:10.1016/j.cell.2015.11.038 (2016). 28 Polymenidou, M. et al. Long pre-mRNA depletion and RNA missplicing contribute to neuronal vulnerability from loss of TDP-43. Nature neuroscience 14, 459-468, doi:10.1038/nn.2779 (2011). 29 Hutton, M. et al. Association of missense and 5'-splice-site mutations in tau with the inherited dementia FTDP-17. Nature 393, 702-705, doi:10.1038/31508 (1998). 30 Swanson, E. et al. Extracellular Tau Oligomers Induce Invasion of Endogenous Tau into the Somatodendritic Compartment and Axonal Transport Dysfunction. J Alzheimers Dis 58, 803-820, doi:10.3233/JAD- 170168 (2017). 31 Elden, A. C. et al. Ataxin-2 intermediate-length polyglutamine expansions are associated with increased risk for ALS. Nature 466, 1069-1075, doi:10.1038/nature09320 (2010). 32 Becker, L. A. et al. Therapeutic reduction of ataxin-2 extends lifespan and reduces pathology in TDP-43 mice. Nature 544, 367-371, doi:10.1038/nature22038 (2017). 33 Harauz, G. & Boggs, J. M. Myelin management by the 18.5-kDa and 21.5- kDa classic myelin basic protein isoforms. J Neurochem 125, 334-361, doi:10.1111/jnc.12195 (2013). 34 Freischmidt, A. et al. Haploinsufficiency of TBK1 causes familial ALS and fronto-temporal dementia. Nat Neurosci 18, 631-636, doi:10.1038/nn.4000 (2015). 35 Cirulli, E. T. et al. Exome sequencing in amyotrophic lateral sclerosis identifies risk genes and pathways. Science 347, 1436, doi:10.1126/science.aaa3650 (2015). 36 Skibinski, G. et al. Mutations in the endosomal ESCRTIII-complex subunit CHMP2B in frontotemporal dementia. Nat Genet 37, 806-808, doi:10.1038/ng1609 (2005). 37 Takahashi, Y. et al. ERBB4 mutations that disrupt the neuregulin-ErbB4 pathway cause amyotrophic lateral sclerosis type 19. Am J Hum Genet 93, 900-905, doi:10.1016/j.ajhg.2013.09.008 (2013). 38 Van Hoecke, A. et al. EPHA4 is a disease modifier of amyotrophic lateral sclerosis in animal models and in humans. Nat Med 18, 1418-1422, doi:nm.2901 [pii] 10.1038/nm.2901 (2012). 39 Landers, J. E. et al. Reduced expression of the Kinesin-Associated Protein 3 (KIFAP3) gene increases survival in sporadic amyotrophic lateral sclerosis. Proc Natl Acad Sci U S A, doi:0812937106 [pii] 10.1073/pnas.0812937106 (2009). 40 Nishimura, A. L. et al. Nuclear import impairment causes cytoplasmic 117 trans-activation response DNA-binding protein accumulation and is associated with frontotemporal lobar degeneration. Brain 133, 1763- 1771, doi:10.1093/brain/awq111 (2010). 41 Johnson, J. O. et al. Mutations in the Matrin 3 gene cause familial amyotrophic lateral sclerosis. Nature Neuroscience 17, 664, doi:10.1038/nn.3688 (2014). 42 Fecto, F. et al. SQSTM1 mutations in familial and sporadic amyotrophic lateral sclerosis. Archives of neurology 68, 1440-1446, doi:10.1001/archneurol.2011.250 (2011). 43 Koyama, A. et al. Increased cytoplasmic TARDBP mRNA in affected spinal motor neurons in ALS caused by abnormal autoregulation of TDP-43. Nucleic acids research 44, 5820-5836, doi:10.1093/nar/gkw499 (2016). 44 Egawa, N. et al. Drug screening for ALS using patient-specific induced pluripotent stem cells. Science translational medicine 4, 145ra104, doi:10.1126/scitranslmed.3004052 (2012). 45 Liu, C., Song, X., Nisbet, R. & Gotz, J. Co-immunoprecipitation with Tau Isoform-specific Antibodies Reveals Distinct Protein Interactions and Highlights a Putative Role for 2N Tau in Disease. J Biol Chem 291, 8173- 8188, doi:10.1074/jbc.M115.641902 (2016). 46 Trabzuni, D. et al. MAPT expression and splicing is differentially regulated by brain region: relation to genotype and implication for tauopathies. Hum Mol Genet 21, 4094-4103, doi:10.1093/hmg/dds238 (2012). 47 Borroni, B. et al. Association between tau H2 haplotype and age at onset in frontotemporal dementia. Archives of neurology 62, 1419-1422, doi:10.1001/archneur.62.9.1419 (2005). 48 Behrouzi, R. et al. Pathological tau deposition in Motor Neurone Disease and frontotemporal lobar degeneration associated with TDP-43 proteinopathy. Acta neuropathologica communications 4, 33, doi:10.1186/s40478-016-0301-z (2016). 49 Moreno, J. A. et al. Oral treatment targeting the unfolded protein response prevents neurodegeneration and clinical disease in prion-infected mice. Science translational medicine 5, 206ra138, doi:10.1126/scitranslmed.3006767 (2013). 50 Kang, S. H. et al. Degeneration and impaired regeneration of gray matter oligodendrocytes in amyotrophic lateral sclerosis. Nat Neurosci 16, 571- 579, doi:nn.3357 [pii] 10.1038/nn.3357 (2013). 118 Figure 1. CRISPR mutagenesis, ACBM characterisation and breeding ratios of TDP-43Q331K mice (a) Chromatograms from the patient originally identified with the Q331K mutation and CRISPR/CAS9 knock-in founder mouse #52. Bases are given above the chromatograms and amino acids coded are given below. The mutation is highlighted with the red arrow. (b) SapI restriction enzyme digestion of 1000 bp PCR products across the mutation site from representative genotyping of wild-type, TDP-43Q331K/Q331K, and TDP-43Q331K/+ mice. (c) Automated continuous behavioural monitoring (ACBM) of 4-month-old mice (n = 10 mice per genotype; 5 males and 5 females). Significantly altered behaviours are displayed: walking: interaction P<0.0001; hanging: interaction P=0.002; rearing: interaction P=0.038; eating-by-hand: genotype P=0.008; repeated measures two- way ANOVA. (d) Walking behaviour as assessed by ACBM in 7.5-month-old male and female mice (n = 5 mice per genotype). Walking male: interaction P<0.0001; walking female: interaction P=0.334; repeated measures two- way ANOVA. (e) Ratios of mice genotyped at 10 days (all of which were successfully weaned) broken down by gender. Female (χ2=2.311, d.f.=2, P=0.315), Male (χ2=7.612, d.f.=2, P=0.022); Chi square test. Error bars represent mean ± s.e.m. Figure 2. Motor impairment, hyperphagia and spinal motor neuronal transcriptomic changes in mutant mice (a) Rotarod and (b) weights of Cohort 1 mice (n = 14 wild-type, 13 TDP-43Q331K/+ and 13 TDP-43Q331K/Q331K mice). (a) Pairwise comparisons: wild-type vs. TDP-43Q331K/+: P=0.014 (*); wild-type vs. TDP-43Q331K/Q331K: P=0.0024 (**). (b) Pairwise comparisons: wild-type vs. TDP-43Q331K/+: P=0.002 (**); wild-type vs. TDP- 43Q331K/Q331K: P=0.0002 (***). (c) Weekly food consumption over 9 weeks (n = 2 cages per genotype). Comparison: Genotype: P=0.047(*). (d) Rotarod of weight-matched Cohort 2 mice (n = 16 wild-type, 13 TDP-43Q331K/+ and 15 TDP-43Q331K/Q331K mice). For (a-d) repeated measures two-way ANOVA followed by Holm-Sidak post-hoc test for pairwise comparisons. (e) Nissl-stained lumbar motor neurons of 5-month-old mice. Representative images shown. Scale bar, 40 μm. (f) Quantification of lumbar motor neurons (n = 4 mice per genotype). Comparison: P=0.089 (ns); unpaired t test. (g) Examples of isometric twitch force recordings during graded nerve stimulation of FDB muscles from representative wild-type and TDP-43Q331K/Q331K mice. Each increment corresponds to recruitment of motor units of successively higher electrical threshold (n = 5 mice per genotype). (h) MA plot and (i) hierarchical clustering of significantly differentially expressed genes (DEGs) in laser- captured motor neurons. In (h) blue dots indicate significant changes, red dots indicate intensity hits. In (i) Genes Aox1 and Agrin are labelled. Comparison: DESeq2 wild-type v TDP-43Q331K/Q331K (j) Immunohistochemistry for AOX1. Representative images from a 5-month-old wild-type mouse shown. Scale bars, 10μm motor neuron, 100μm ventral root. (k) AOX1 immunofluorescence in lumbar motor neurons. Comparison: P=0.433 (ns); unpaired t test. For (h-k) n = 4 mice per genotype. All error bars denote mean ± s.e.m. 119 Figure 3. Cognitive testing indicates executive dysfunction, memory impairment and phenotypic heterogeneity in mutant mice (a) Schematic for the 5-choice serial reaction time task (5-CSRTT). (b) Sessions required to reach performance criteria for 5-CSRTT (n = 16 per genotype). Pairwise comparisons: wild-type vs. TDP-43Q331K/+: P=0.083 (ns); wild-type vs. TDP-43Q331K/Q331K: P=0.004 (**). (c) 5-CSRTT at 6 months of age (n = 15 wild-type, 16 TDP-43Q331K/+, 15 TDP-43Q331K/Q331K mice). Baseline session genotype effects: accuracy: P=0.109; omission: P=0.283). Stimulus duration (SD) probe test genotype effects: accuracy: P=0.833; omission: P=0.077 (ns); SD effect: accuracy and omission: P<0.001; Mixed- effects model. (d) 5-CSRTT at 12 months of age (n = 15 wild-type, 16 TDP-43Q331K/+, 16 TDP-43Q331K/Q331K mice). Baseline session genotype effects: accuracy: P=0.487; omission: P=0.120. SD probe test genotype effects: accuracy: P=0.880; omission: P=0.044 (*); SD effect: accuracy: P<0.0001; omission: P<0.0001; genotype by SD interaction: accuracy: P=0.081; omission: P=0.271; Mixed-effects model. (e) Mean trials completed on an unrestricted fixed-ratio schedule (n = 16 per genotype). (f) Mean breakpoint on a progressive-ratio schedule (response increment per trial = 4; n = 16 per genotype). (g) Novel object recognition sample and (h) choice phases (n = 8 wild-type, 9 TDP-43Q331K/+, 8 TDP- 43Q331K/Q331K mice). For (h) 1 min delay pairwise comparisons: wild-type vs. TDP-43Q331K/+: P=0.158 (ns); wild- type vs. TDP-43Q331K/Q331K: P=0.158 (ns); 3 hour delay pairwise comparisons: wild-type vs. TDP-43Q331K/+: P=0.014 (*); wild-type vs. TDP-43Q331K/Q331K: P=0.009 (**). For (b,e,f) one-way ANOVA and (g,h) two-way ANOVA, all followed by Holm-Sidak post-hoc tests for pairwise comparisons. (i) Marbles buried in Cohort 1 at 18 months of age (n = 15 wild-type, 13 TDP-43Q331K/+, 14 TDP-43Q331K/Q331K mice). Pairwise comparisons: wild-type vs. TDP-43Q331K/+: P=0.009 (**); wild-type vs. TDP-43Q331K/Q331K: P<0.0001 (****); Kruskal-Wallis followed by Dunn’s test for pairwise comparisons. Error bars denote s.e.m. for (c) to (h) and median and interquartile range for (b) and (i). Figure 4. Perturbed TDP-43 autoregulation and loss of parvalbumin interneurons in mutant mice (a) Marbles buried by 5-month-old mice. Coloured dots indicate animals used for RNASeq analysis. Yellow dots indicate TDP-43Q331K/Q331K littermates (n = 19 wild-type, 19 TDP-43Q331K/+, 17 TDP-43Q331K/Q331K mice). Pairwise comparisons: wild-type vs. TDP-43Q331K/+: P=0.028 (*); wild-type vs. TDP-43Q331K/Q331K: P=0.013 (*); Kruskal-Wallis followed by Dunn’s test for pairwise comparisons. Error bars represent median and interquartile range. (b) Representative Nissl staining of frontal cortex (layers indicated) (n = 5 wild-type, 6 TDP-43Q331K/Q331K mice). Scale bar, 500μm. (c) Immunohistochemistry for TDP-43 in pyramidal neurons of motor cortex layer V. Representative images shown (n = 4 mice per genotype). Scale bar, 20μm. (d) Immunoblot of fractionated frontal cortical tissue from 5-month-old mice (two biological replicates shown, uncropped in Supplementary Fig. 5). (e) Immunoblot band intensity quantification (n = 4 mice per genotype). Comparison: P=0.007 (**); unpaired t test. Error bars denote s.e.m. (f) MA plot and (g) hierarchical clustering of DEGs (n = 6 wild-type, 6 TDP-43Q331K/+, 8 TDP-43Q331K/Q331K mice) in frontal cortex. For (f) blue dots indicate significant changes, red dots indicate intensity hits. Comparison: DESeq2 wild-type v TDP-43Q331K/Q331K. For (g) gene ontology (GO) biological process and KEGG pathway enriched terms are displayed. (h) Expression changes for parvalbumin and ALS-FTD linked genes identified by RNASeq. (i) Immunohistochemistry for parvalbumin in cortices of 5-month-old mice. Representative images shown. Scale bar, 250μm. (j) Quantification of parvalbumin-positive neurons (n = 3 mice per genotype). Comparison: P=0.0003 (***); unpaired t test. Error bars denote s.e.m. (k) Immunohistochemistry for TDP-43 in parvalbumin-positive cells. Representative images shown. Scale bar, 5μm. (l) TDP-43 expression in parvalbumin-positive cells (n=5 mice per genotype). Comparison by two-way ANOVA. Error bars denote s.e.m. 120 Figure 5. Splicing analysis indicates TDP-43 misregulation, a gain of TDP-43 function and altered Mapt exon 2/3 splicing (a) MA plot and (b) hierarchical clustering of frontal cortical alternative splice events (n = 6 wild-type, 6 TDP- 43Q331K/+, 8 TDP-43Q331K/Q331K mice). Comparison: DESeq2 wild-type v TDP-43Q331K/Q331K. (c) Schematic of altered splicing in the 3’UTR of Tardbp. Arrow indicates reduced exclusion of intron 7 of the Tardbp transcript in TDP-43Q331K/Q331K relative to wild-type mice. (d) Quantitative PCR (qPCR) of splicing changes in Tardbp intron 7 (n = 6 wild-type, 6 TDP-43Q331K/+, 8 TDP- 43Q331K/Q331K mice). (e) Schematic of exon 17b inclusion/exclusion in Sort1. Arrows indicate reduced inclusion of exon 17b in TDP- 43Q331K/Q331K relative to wild-type mice. (f) qPCR of splicing changes in Sort1 exon 17b (n = 6 wild-type, 6 TDP-43Q331K/+, 8 TDP-43Q331K/Q331K mice). (g) Schematic of altered splicing of exons 2 and 3 of Mapt. Arrows indicate increased inclusion of exons 2 and 3 in the Mapt transcripts of TDP-43Q331K/Q331K relative to wild-type mice. The expanded view of exon 1 to exon 2 includes a site of TDP-43 binding as detected by iCLIP (iCount pipeline; TDP-43_CLIP_E18-brain). (h) Schematic of N-terminal Mapt splice variants (0N, 1N and 2N). (i) qPCR of splicing changes in Mapt exons 2 and 3 (n = 6 wild-type, 6 TDP-43Q331K/+, 8 TDP-43Q331K/Q331K mice). 2N/0N pairwise comparisons: wild-type vs. TDP-43Q331K/+: P=0.047 (*); wild-type vs. TDP-43Q331K/Q331K: P=0.0001 (***); TDP-43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.013 (*). (j-k) qPCR of hippocampal splicing changes (n = 4 wild-type, 3 TDP-43Q331K/+, 4 TDP-43Q331K/Q331K mice per gender). Pairwise comparisons: Tardbp intron 7 exclusion, male: wild-type vs. TDP-43Q331K/+: P=0.043 (*); TDP- 43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.002 (**); female: wild-type vs. TDP-43Q331K/+: P=0.013 (*); TDP-43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.0002 (***); Mapt: 0N, male: wild-type vs. TDP-43Q331K/+: P=0.023 (*); wild-type vs. TDP- 43Q331K/Q331K: P=0.023 (*); TDP-43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.877 (ns); female: wild-type vs. TDP-43Q331K/+: P=0.365 (ns); wild-type vs. TDP-43Q331K/Q331K: P=0.324 (ns); TDP-43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.858 (ns); 1N/0N, male: wild-type vs. TDP-43Q331K/+: P=0.008 (**); TDP-43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.008 (**); female: wild-type vs. TDP-43Q331K/+: P=0.077 (ns); TDP-43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.002 (**); 2N/0N, male: wild-type vs. TDP-43Q331K/+: P=0.002 (**); wild-type vs. TDP-43Q331K/Q331K: P=0.0001 (***); TDP-43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.151 (ns); female: wild-type vs. TDP-43Q331K/+: P=0.202 (ns). For (d,f,i-k) P<0.0001 (****). For (d,f,i) one-way and (j,k) two-way ANOVA, all followed by Holm-Sidak post- hoc tests for pairwise comparisons. Error bars denote s.e.m. Figure 6. TDP-43 misregulation occurs in spinal cords of mutant mice, but not in motor neurons (a) Schematic detailing lumbar spinal cord (LSC) processing for transcriptomic analysis (LCM, laser capture microdissection). (b) MA plots of lumbar motor neuronal differentially expressed and spliced genes (n = 4 mice per genotype). Comparison: DESeq2 wild-type v TDP-43Q331K/Q331K. Blue and red dots indicate significant changes. Green dots highlight Tardbp expression, Tardbp intron 7 exclusion and Sort1 exon 17b inclusion, which are not significant changes. (c-d) Quantitative PCR of homogenised lumbar spinal cord (n = 4 wild-type, 4 TDP-43Q331K/+, 4 TDP- 43Q331K/Q331K mice). Comparisons as follows: (c) Tardbp expression: wild-type vs. TDP-43Q331K/+: P=0.103 (ns); wild-type vs. TDP-43Q331K/Q331K: P=0.0008 (***); TDP-43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.007 (**). Tardbp intron 7 exclusion: wild-type vs. TDP-43Q331K/+: P=0.001 (***); wild-type vs. TDP-43Q331K/Q331K: P>0.0001 (****); TDP-43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.002 (**). Sort1 exon 17b inclusion: P<0.0001 (****). (d) 0N Mapt. 1N Mapt: wild-type vs. TDP-43Q331K/+: P=0.640 (ns); wild-type vs. TDP-43Q331K/Q331K: P=0.02 (*); TDP-43Q331K/+ vs. TDP-43Q331K/Q331K: P=0.03 (*). 2N Mapt. (c-d) Comparisons by one-way ANOVA followed by Holm-Sidak post-hoc tests. Error bars denote s.e.m. 121 Figure 7: Phenotypic stratification of transcriptomic data from mutant mice allows the identification of putative disease modifiers (a) Marble-burying in 5-month-old mice prior to sacrifice. MB+ mice bury at or above the median number of marbles for the group, and MB- mice bury fewer. Yellow dots indicate TDP-43Q331K/Q331K littermates. (b) Marble burying activity of TDP-43Q331K/Q331K littermates as described in (a). (c) Hierarchical clustering of DEGs in frontal cortices comparing MB+ and MB- TDP-43 Q331K/Q331K mice. Genes Atxn2 and Arid4a are highlighted (n = 6 wild-type, 4 MB+ TDP-43Q331K/Q331K and 4 MB- TDP-43Q331K/Q331K mice). Comparison: DESeq2 MB+ v MB-. Gene ontology (GO) biological processes and KEGG pathway enriched terms are displayed. (d) Graphical representation of altered splicing of Mbp. Arrows indicate the altered pattern of splicing in MB+ relative to MB- TDP-43Q331K/Q331K mice. (e) qPCR of the ratio of Mbp Basic to Mbp Golli (n = 6 wild-type, 4 TDP-43Q331K/+, 4 TDP-43Q331K/Q331K mice). Pairwise comparisons: wild-type vs. MB+: P=0.005 (**); wild-type vs. MB-: P=0.024 (*); MB+ vs. MB-: P=0.0003 (***); one-way ANOVA followed by Holm-Sidak post-hoc tests. Error bars denote s.e.m. (f) Representative marble burying analyses: 4:4, original analysis; 3:3, comparing the three best MB+ and three worst MB- mice; 4v4 mixed, one MB- mouse swapped with one MB+ mouse. Number of DEGs identified by DESeq2 comparison of MB+ v MB- mice for each comparison is given below. For 3:3, hits common to the 4:4 stratification are shown in brackets. Figure 8. TDP-43Q331K mice demonstrate age-related deterioration in cortical transcriptomes with altered expression of multiple ALS-linked genes (a) MA plot and (b) hierarchical clustering of DEGs in frontal cortices at 20 months of age (n = 8 wild-type, 10 TDP-43Q331K/+, 10 TDP-43Q331K/Q331K mice). For (a) blue dots indicate significant changes, red dots indicate intensity hits. Comparison: DESeq2 wild-type v TDP-43Q331K/Q331K. For (b) gene ontology (GO) biological processes and KEGG pathway enriched terms are displayed. (c) Venn diagram highlighting DEGs between wild-type v TDP-43Q331K/Q331K mice that were common to analyses in 5 and 20-month-old mice. Known ALS-FTD linked genes within this common subset are highlighted in (d). (e) MA plot and (f) hierarchical clustering of frontal cortical alternative splice events at 20 months of age (n = 8 wild-type, 10 TDP-43Q331K/+, 10 TDP-43Q331K/Q331K mice). Blue dots indicate significant changes, red dots indicate intensity hits. Comparison: DESeq2 wild-type v TDP-43Q331K/Q331K. For (a,b,e,f) n = 8 wild-type, 10 TDP-43Q331K/+, 10 TDP-43Q331K/Q331K mice. (g) Venn diagram highlighting alternative splice events between wild-type v TDP-43Q331K/Q331K mice that are common to analyses in 5 and 20-month-old mice. Known ALS-FTD linked genes within this common subset are highlighted in (h). (i) Schematic of Matr3 exon 14 inclusion/exclusion. Arrows indicate increased inclusion of exon 14 in TDP- 43Q331K/Q331K relative to wild type mice. (j) Schematic of Sqstm1 transcript splice variants. Percentages given indicate the relative amount of each variant in TDP-43Q331K/Q331K mice. The TDP-43Q331K-specific variant is undetectable in wild-type mice. 122 Online methods CRISPR/CAS9 mutagenesis to introduce Q331K mutation Nucleases were designed to be close to/overlap the desired point mutation. Three CRISPR- Cas9 nucleases were tested for activity using a GFP reporter plasmid. A 121 bp single- stranded DNA (ssDNA) oligonucleotide with the point mutation at the mid-point was used as a repair template. Guide RNA (gRNA) and a capped Cas9 mRNA were synthesised and injected with the donor oligonucleotide into 270 single-cell C57Bl/6J embryos. For sequences see Supplementary Table 3. Off-targets were predicted using CRISPRseek51. Mouse breeding and maintenance Mouse founder #52 was outcrossed with wild-type C57Bl/6J mice through to the F3 generation. Three F3 male siblings were bred to wild-type C57Bl/6J mice to generate F4 TDP- 43Q331K/+ mutants, which were intercrossed to generate animals for study. Power calculations were based on historical rotarod and touchscreen data of wild-type mice. This indicated required group sizes of 15 animals per genotype to identify a ~20% difference in performance between genotypes. Animals were only excluded from analyses if specified in the following methods. Mouse breeding was carried out in the UK and USA. ACBM was carried out at the Brown University Rodent Neurodevelopment Behaviour Testing Facility. All procedures were approved by the Brown University Animal Care and Use Committee. Touchscreen analysis; marble burying; object recognition; motor behaviour; food intake and weight measurement; pathology; electrophysiology and RNA sequencing all took place in the UK. All experiments were conducted in accordance with the United Kingdom Animals (Scientific Procedures) Act (1986) and the United Kingdom Animals (Scientific Procedures) Act (1986) Amendment Regulations 2012. Animals were housed in cages of up to five animals under a 12 hr light/dark cycle. Genotyping The Q331K mutation coincidentally introduces a SapI/EarI restriction site, which facilitates genotyping (see Supplementary Table 4). Automated continuous behavioural monitoring Ten TDP43Q331K/Q331K and 10 wild-type animals (5 female, 5 male of each genotype) from the same breeding campaign were obtained from the animal care facility at the University of Massachusetts Medical School. Animals were group housed between sessions, but housed individually during the 5-day ACBM recording sessions. Cages were monitored with a Firefly MV 0.3 MP Mono FireWire 1394a (Micron MT9V022) at 30 frames/s. Cameras were connected to a workstation with Ubuntu 14.04 with a firewire card to connect to all cameras. For processing by the computer vision system, all videos were down-sampled to 320×240 pixels. The system used for ACBM was modified from that previously described and was re- implemented in Python and NVIDIA’s CUDNN to speed video analysis subroutines. All video analyses were conducted using the Brown University high-performance computer cluster. The system was retrained using data collected at the Brown Rodent Neuro-Developmental Behaviour Testing facility (~20 h of video and 40 animals total). Data were annotated by hand for 8 behaviours as previously described (drink, eat, groom, hang, rear, rest, sniff, walk). Accuracy was evaluated using by cross-validation. The average agreement with human 123 annotations was 78% for individual behaviour and 83% overall for individual frames. Evaluation of the system was also run on a subset of the data collected for the present study, which found an overall mean agreement of 71% for individual behaviours and 82% over all video frames. Rotarod Motor testing was performed using Rotarod (Ugo Basile, Model 7650, Varese, Italy). At least 24 h prior to testing mice were first trained for 5 min at the slowest speed and then 7 min with acceleration. During testing mice were subjected to 7 min trials with acceleration from 3.5 to 35 rpm. In each session mice were tested 3 times with a trial separation of 30 min. The latency to fall (maximum 420 s) for each mouse was recorded and mean values for each mouse calculated. An individual mouse recording was excluded if it fell off the rod while moving backwards, accidentally slipped or jumped off at slow speed. Two consecutive passive rotations were counted as a fall and the time recorded as the end point for that mouse. Mouse weights were recorded immediately after completion of rotarod testing. All testing was conducted by operators who were blind to genotype and in a randomised order. Feeding Cages containing either two or three mice of the same genotype were topped up with 400g of food on Monday mornings. The following Monday the surplus food in the hopper together with any obvious lumps of food in the cage was removed and weighed. The difference from 400g was calculated and recorded as the total food consumed in seven days. This was normalised to the number of mice in a given cage. Weekly consumption was calculated for 9 consecutive weeks. Mice were 12 months of age when recording commenced. All testing was conducted while blind to genotype and in a randomised order. Touchscreen studies 48 male mice (n = 16 per genotype) were housed in groups of 2-5 per cage under a 12 hr light/dark cycle (lights on at 7:00pm). Testing was conducted during the dark phase. To ensure sufficient levels of motivation, animals were food-restricted to ~85-90% of free-fed weights by daily provision of standard laboratory chow pellets (RM 3; Special Diet Services, Essex, UK). Drinking water was available ad libitum. Experiments were performed in standard mouse Bussey-Saksida touchscreen chambers (Campden Instruments Ltd, Loughborough, UK). The reward for each correct trial was delivery of 20 μL of milkshake (Yazoo Strawberry milkshake®; FrieslandCampina UK, Horsham, UK). The chambers are equipped with infrared activity beams (rear beam = 3 cm from magazine port and front beam = 6 cm from screen) to monitor locomotor activity. Following two days of habituation to touchscreen chambers, mice underwent pretraining and training. Briefly, mice were first trained to touch the correctly lit stimulus in return for a food reward, and to initiate a trial by poking and removing their nose from the magazine. Finally, mice were discouraged from making responses at non-illuminated apertures by a 5 s time-out period during which the chamber was illuminated. Investigators were blind to genotype. 5-choice serial reaction time task (5-CSRTT) Upon completion of training at 2 s stimulus duration (baseline), mice were tested on 4 sessions of decreasing stimulus durations (2.0 s, 1.5 s, 1.0 s, 0.5 s) pseudo randomly within a session. Animals that had not reached the criterion (> 80% accuracy, < 20% omissions in two consecutive sessions in baseline training before entering the probe test, N = 1 in the first probe test) or whose body weights were below 80% of free-feeding weight (N = 1 in the first, and N = 1 in the second probe test) were excluded. Fixed-ratio (FR) and progressive-ratio (PR) schedule 124 FR and PR were conducted as described elsewhere52. When performance stabilised on FR5 (completion of 30 trials within 20 min), all mice were tested on two sessions of an unrestricted FR5, which allowed an unlimited number of trials in 60 min. Next, animals underwent 3 sessions of PR4, in which animals should emit a progressively increasing number of responses (i.e. 1, 5, 9, 13, …) in each subsequent trial to obtain a single reward. PR session terminated following either 60 min or 5 min of inactivity. Breakpoint, the number of responses made to obtain the reward in the last completed trial, was recorded as an index of motivation. Object recognition The novel object recognition task was conducted as described elsewhere53 in a randomised order with the operator blind to genotype and under dimmed white light. Six-month-old male mice (n = 8-9 per genotype) were randomly chosen from Cohort 2. Mice were habituated to a Y-maze for 5 min. One day later mice were reintroduced to the Y-maze, which now contained two identical objects in each arm. Exploration time for each object over a 5 min period was recorded (sample phase). Mice were then removed from the maze and one of the objects replaced with a novel object. After a delay of 1 min or 3 h mice were reintroduced to the maze (choice phase) and the time spent exploring each object over a 5 min period was recorded. The memory for the familiar object was expressed as a discrimination ratio (difference in exploration of the novel and familiar objects divided by the total object exploration time). Marble burying All testing was conducted in the morning and blind to genotype. Cages of size 39.1cm x 19.9cm x 16.0cm height (Tecniplast) were used. Fresh bedding material (Datesand, grade 6) was placed into each cage to a height of ~6cm. Ten glass marbles (1cm) were placed evenly across the bedding. Ten cages were prepared in a single round. One mouse was placed in each of the cages and the lids replaced. Mice were left undisturbed for 30 min under white light. Mice were then removed and the number of marbles buried by at least two thirds was scored. Cages were reset using the same bedding material to test another 10 mice. In stratifying mice prior to frontal cortical RNAseq, animals were tested twice, three days apart to identify those that consistently buried high or low numbers of marbles. Repeat behavioural studies Cohort 1 mice underwent rotarod, weight, feeding and marble testing all under a standard light/dark cycle (lights on at 7:00am for 12h). Cohort 2 mice underwent all touchscreen, object recognition and rotarod studies under a reverse light/dark cycle. Pathological studies Mice were culled by cervical dislocation, decapitated and tissues processed as follows. Brains Right hemispheres were processed for RNA and/or protein extraction (see below). Left hemispheres were immersion fixed in 4% paraformaldehyde (PFA) at 4°C for 24 h, washed in PBS, cryoprotected in 30% sucrose in PBS at 4°C, embedded and frozen in M1 matrix (Thermo Fisher Scientific) on dry ice and sectioned coronally at 16 μm thickness on a cryostat (Leica Biosystems). Sections were mounted on Superfrost Plus charged slides (Thermo Fisher Scientific), allowed to dry overnight and stored at -80°C. Spinal cords Vertebral columns were dissected from culled mice, immersion fixed in 4% PFA at 4°C for 48 h, washed in PBS and dissected to extract spinal cords and nerve roots. The lumbar enlargement was sub dissected, cryoprotected in 30% sucrose at 4°C, embedded in M1 matrix in a silicon mould, frozen on dry ice and sectioned at 16 μm thickness onto charged slides, briefly air dried and stored at -80°C. 125 Antigen retrieval and immunostaining Sections were thawed at R/T and briefly rinsed in distilled water. Antigen retrieval was performed by heating slides for 20 min at 95°C in antigen unmasking solution, Tris-based (Vector laboratories). Sections were cooled to R/T, washed in distilled water, and blocked and permeabilised in a solution containing 5% bovine serum albumin (BSA), 0.1% Triton X- 100 and 5% serum (specific to secondary antibody species used) for 1 h at R/T. Slides were incubated with primary antibody for 2 h at R/T or 4°C overnight in 5-fold diluted blocking buffer. Secondary antibodies were applied for 1 h at R/T (Alexa Fluor conjugated, Thermo Fisher Scientific; 1:500 in diluted block). Sections were counterstained and mounted with VECTASHIELD with DAPI (Vector labs) hard-set. Alexa Fluor 568 conjugated secondary antibodies were false coloured magenta (ImageJ 1.15j). To quantify parvalbumin-positive neurons, parvalbumin stained sections were imaged on a Nikon Ti-E live cell imager. Images were acquired using a Plan Apo lambda 10x objective with a final image dimension of 4608 x 4608 with 2x2 binning, stitched (NIS-Elements) and analysed (ImageJ 1.15j) blind to genotype. For each mouse, matching sections through the frontal cortex from Bregma 2.8 mm to 0.74 mm were analysed with a total of 10 sections quantified for 3 wild-type and TDP43Q331K/Q331K mice. Images were converted to greyscale and thresholded to produce a binary image. Consistent regions of interest were drawn around the cortex using the polygon selection tool and the ‘analyse particle’ function used to count cells. To investigate TDP-43 in parvalbumin-positive neurons, sections were costained with antibodies against TDP-43 and parvalbumin and imaged using a Zeiss LSM 780, AxioObserver with a Plan-Apochromat 63x/1.40 Oil DIC M27 objective running Zen system software. Data analysis (ImageJ 1.15j) and imaging was carried out blind to genotype. For each cell, a maximum intensity projection of Z stacks was created and regions of interest were drawn around the nucleus and the cytoplasm using the polygon selection tool. Area, integrated density and mean grey value measurements were taken for the cytoplasm and nucleus, together with a background reading. Corrected total fluorescence for a region of interest was calculated as: CTF = Integrated Density - (Area region of interest x background fluorescence) Corrected fluorescence was recorded for at least 10 cells per mouse in matched sections corresponding to Bregma 1.18 mm (The Mouse Brain, compact third edition, Franklin and Paxinos). To quantify AOX1 fluorescence in lumbar motor neurons, sections were costained with antibodies against AOX1 and neurofilament heavy and imaged on a Nikon Ti-E live cell imager with a Plan Apo VC 20x DIC N2 objective with a final image dimension of 1024 x 1022 pixels and 2x2 binning. Data analysis (ImageJ 1.15j) and imaging were carried out blind to genotype. Corrected fluorescence was recorded for at least 29 cells per mouse. TDP-43 immunostaining in spinal cord and brain were imaged using a Nikon Ti-E live cell imager and a Plan Apo VC 100x Oil objective with a final image dimension of 1024 x 1024 pixels with 2x2 binning. Images are a maximum intensity z-stack created using ImageJ 1.15j with a z-step of 0.2µm. Tau immunostaining in cortex was imaged using a Zeiss LSM 780, AxioObserver with a Plan- Apochromat 63x/1.40 Oil DIC M27 objective running Zen system software. Images are a maximum intensity z-stack created using ImageJ 1.15j. For list of primary antibodies see Supplementary Table 5. 126 Nissl staining of spinal cord and brain Sections were thawed at R/T, washed in distilled water then stained with cresyl etch violet (Abcam) for 5 min, briefly washed in distilled water, dehydrated in 100% ethanol, cleared in xylene, mounted (Permount, Fisher) and dried overnight at R/T. Images were taken on a Zeiss Axio Observer.Z1 running Axiovision SE64 release 4.8.3 software. Cortical images were taken with an EC Plan-Neofluar 5x/0.16 M27 objective with a total area of 4020 x 2277 pixels auto stitched within the software. Spinal cord images were acquired with an LD Plan- Neofluar 20x/0.4 korr M27 objective with an image size of 1388 x 1040 pixels. Lumbar spinal motor neuron quantification Motor neurons were quantified as described elsewhere54. Briefly, large motor neurons (diameter >20 μm) in the ventral horn were counted blind to genotype in 18 sections from the lumbar L3-5 levels of each animal. Cellular quantification in brain Data analysis using ImageJ 1.15j and imaging was carried out blind to genotype. For total frontal cortical area, matching sections through the frontal cortex from Bregma 2.8 mm to 0.74 mm were selected with a total of 10 sections quantified for six wild-type and six TDP43Q331K/Q331K mice. Matching regions of interest were drawn around the cortex and the area quantified using the measure function. To count cells within cortical sub regions, matching sections based on Bregma references were identified. Images were converted to greyscale and thresholded to produce a binary image. Consistent regions of interest were drawn around the cortex and the ‘analyse particle’ function used to count cells. A minimum size of 10 pixel units ensured that intact cells were counted and results were displayed with the overlay option selected. Western blotting Brain tissues were weighed to ensure equal amounts of starting material between samples, thawed on ice and processed using a modified fractional protocol55. Briefly, tissue was sequentially homogenised and centrifuged using buffers A [NaCL 150 mM, HEPES (pH 7.4) 50mM, digitonin (Sigma, D141) 25 μg/mL, Hexylene glycol (Sigma, 112100) 1 M, protease inhibitor cocktail (Sigma, P8340), 1% v:v] and B [same as buffer A except Igepal (Sigma, I7771) 1% v:v is used in place of digitonin] to extract cytoplasmic and membrane fractions respectively. The subsequent pellet was sonicated in 1% sarkosyl buffer containing 10μM Tris-Cl (pH 7.5), 10μM EDTA, 1M NaCl and centrifuged (14,000g for 30min at 4°C). The supernatant was taken as the nuclear fraction. Protein lysates were quantified (bicinchoninic acid protein assay, Pierce), electrophoresed in 4-12% or 12% SDS polyacrylamide gels, wet transferred to PVDF membranes, blocked with a 50:50 mixture of Odyssey PBS blocking buffer and PBS with 0.1% Tween20 for 1 h at R/T and then probed with primary antibodies at 4°C overnight. Secondary antibodies were either fluorescently tagged for Odyssey imaging, or HRP tagged for ECL visualisation. Western blot band intensities were quantified using Fiji (ImageJ; Version 2.0.0-rc-54/1.51h; Build: 26f53fffab) using the programs gel analysis menu option in 8-bit greyscale. Quantification was carried out by an independent user blind to genotype. For list of primary antibodies see Supplementary Table 5. Muscle histology The right gastrocnemius was dissected, fixed in 4% PFA at R/T, washed in PBS for 10 min (x2) and cryoprotected and stored in 30% sucrose with 0.1% azide. Tissues were placed in a silicone mould with M1 matrix,and frozen on dry ice. Longitudinal cryosections (50 μm) were mounted onto slides (Superfrost Plus), air dried at R/T for 5 min and stored at -80°C. 127 To stain neuromuscular junctions (NMJs), slides were brought up to R/T and incubated in blocking solution (2% BSA, 0.2% Triton X-100, 0.1% sodium azide) for 1 h. Primary antibodies against βIII-tubulin (rabbit polyclonal, Sigma T2200) and synaptophysin (mouse monoclonal, Abcam ab8049) were applied at 1:200 dilution in blocking solution. Sections were incubated at R/T overnight. Sections were washed in PBS (x3) and incubated for 90 min with mouse and rabbit Alexa488-conjugated secondary antibodies (Thermo Fisher Scientific) diluted 1:500 in blocking solution together with TRITC-conjugated alpha bungarotoxin (Sigma, T0195) 10 μg/ml. Sections were washed in PBS and coverslipped (VECTASHIELD hardset). Confocal Z-stacks were obtained using a Zeiss LSM 780, AxioObserver with a Plan-Apochromat 20x/0.8 M27 objective running Zen system software blind to genotype. For succinate dehydrogenase (SDH) staining, the left gastrocnemius was dissected, flash frozen in isopentane in liquid nitrogen and stored at -80°C until use. Frozen sections of 12 μm were prepared and stained using a modified version of a previously described method56. Briefly, sections were stained with freshly prepared SDH staining solution at 37°C for 3 min, washed through saline, acetone and ethanol solutions, cleared in xylene and mounted (Permount). Images were taken using an Olympus BX41 light microscope (10x objective) with Q Capture Pro 6.0. Quantification of NMJ Innervation NMJs from flattened z-stacks of muscle were analysed (ImageJ; Version 2.0.0-rc-54/1.51h; Build: 26f53fffab) blind to genotype. Brightness and contrast thresholds were set to optimise the signal-to-noise ratio of the presynaptic staining (anti-tubulin and anti-synaptophysin). Innervated NMJs were defined as having observed overlap of staining for pre- and post- synaptic elements. Denervated NMJs were defined as alpha-bungarotoxin signal in the absence of pre-synaptic staining. A small percentage (~5% in each genotype) of NMJs could not be scored and were excluded from this analysis. Neuromuscular electrophysiology Isolated FDB-tibial nerve preparations were mounted in an organ bath in HEPES-buffered MPS of the following composition (mM): Na+ (158); K+ (5); Ca2+ (2); Mg2+ (1); Cl- (169); glucose (5); HEPES (5); pH 7.2-7.4, and bubbled with air or 100% O2 for at least 20 min. The distal tendons were pinned to the base of a Sylgard-lined recording chamber and the proximal tendon connected by 6/0 silk suture to an MLT0202 force transducer (AD Instruments, Oxford, UK). The tibial nerve was aspirated into a glass suction electrode and stimuli (0.1-0.2 ms duration, nominally up to 10V) were delivered via a DS2 stimulator (Digitimer, Welwyn Garden City, UK) triggered and gated by an AD Instruments Powerlab 26T interface. Force recordings were captured and digitised at 1 kHz using the Powerlab interface and measured using Scope 4 and Labchart 7 software (AD Instruments) running on PC or Macintosh computers. For motor unit recordings, the stimulating voltage was carefully graded from threshold to saturation, to evoke the maximum number of steps in the twitch tension record. Motor unit number estimation (MUNE) was performed by inspection, counting the number of reproducible tensions steps, and by extrapolation between the average twitch tension of the four lowest threshold motor units and the maximum twitch tension. For tetanic stimulation, trains of stimuli, 1-5 s in duration were delivered at frequencies of 2-50 Hz. To measure muscle fatigue, 50 Hz stimulus trains, 1 s in duration were delivered every five seconds for about a minute. A fatigue index was calculated as the time constant of the best fitting single exponential to the decline of the maxmimum tetanic force. Brain RNA isolation Frontal cortices and hippocampi were subdissected in RNase free conditions (RNaseZap, Sigma Aldrich) from right hemispheres of freshly culled mice and flash frozen until further use. For RNA extraction tissue was thawed directly in TRIsure reagent (Bioline) and RNA 128 isolated following manufacturer’s instructions. RNA was purified (RNeasy kit, Qiagen) with on- column DNase treatment and analysed on an Agilent 2100 Bioanalyzer. Spinal motor neuron laser capture microdissection Mice were culled by cervical dislocation and decapitation. Lumber spinal cord was rapidly dissected taking care to avoid RNase-exposure, embedded in pre-cooled M1 embedding matrix (Thermo) in a silicone mould and flash frozen in isopentane on dry ice. Samples were stored at -80°C until use. Transverse cryosections (14 μm) were taken through the lumbar enlargement and placed onto PEN membrane glass slides (Zeiss) that were kept at -20°C during sectioning. One spinal cord was processed at a time. ~50 sections were taken per mouse and placed onto two PEN slides. Slides were immediately stained in the following RNase-free, ice-cold solutions (each for 1 min): 70% ethanol, water (with gentle agitation), 1% cresyl violet in 50% ethanol, 70% ethanol, 100% ethanol (with gentle agitation), 100% ethanol (with gentle agitation). Slides were dabbed onto tissue paper to remove excess ethanol, air-dried for 1 min and taken for immediate microdissection (Zeiss PALM Microbeam). Cells were cut at x40 magnification, keeping laser power to a minimum. Motor neurons were identified by location and diameter >30 μm. ~120 cells were captured per mouse into Adhesive Cap 500 tubes (Zeiss). RNA was extracted using the Arcturus PicoPure kit (Thermofisher). 1 ul of RNA was run on an RNA 6000 Pico chip on an Agilent 2100 Bioanalyzer to evaluate RNA quality. 1ng of RNA was used as input for cDNA library preparation. Spinal motor neuron cDNA and library preparation Library preparation for sequencing on an Illumina HiSeq2500 sequencer was carried out using the SMART-seq v4 Ultra low Input RNA kit (Clontech) following the manufacturer’s instructions. All steps were carried out on ice unless otherwise specified. Reverse transcription, PCR cycles and incubation steps utilised a BioRad T100 Thermal Cycler. Amplification of cDNA by LC PCR used a 10-cycle protocol. After bead purification, cDNA library concentration was measured (High Sensitivity DNA kit, Agilent Technologies). Sequencing libraries were generated using the Nextera XT DNA Library Prep Kit (Illumina) using 150 pg cDNA as input following the manufacturer’s instructions with the following modification. Following library amplification and bead purification the final fragment size was analysed and libraries quantified using the Universal KAPA Library Quantification kit (Kapa Biosystems) and a Bio-Rad C100 thermal cycler. An equal amount of cDNA was used to pool up to four samples, which were sequenced in one lane. Sequencing was carried out to a depth of 50 million 100 bp paired-end reads per library. Frontal cortex RNAseq library preparation Only RNA samples with RIN >8 were used for sequencing. Libraries were prepared using the TruSeq Stranded mRNA kit (Illumina) following the manufacturer’s low sample protocol with the following modification. RNA fragmentation time was reduced to 3 min at 94°C to increase median insert length. Final libraries were analysed, quantified and sequenced as above. Bioinformatics pipeline and statistics FastQ files were trimmed with trim galore v0.4.3 using default settings then aligned against the mouse GRCm38 genome assembly using hisat2 v2.0.5 using options --no-mixed and --no- discordant. Mapped positions with MAPQ values of <20 were discarded. Gene expression was quantitated using the RNA-Seq quantitation pipeline in SeqMonk v1.37.0 in opposing strand specific (frontal cortex) or unstranded (motor neuron) library mode using gene models from Ensembl v67. For count based statistics, raw read counts over exons in each gene were used. For visualisation and other statistics log2 RPM (reads per million reads of library) expression values were used. 129 Differentially expressed genes were selected using pairwise comparisons with DESeq2 with a cut-off of P<0.05 following multiple testing correction. Differential splice junction usage was detected by quantitating the raw observation counts for each unique splice donor/acceptor combination in all samples. Initial candidates were selected using DESeq2 with a cut-off of P<0.05 following multiple testing correction. To focus on splicing specific events hits were filtered to retain junctions whose expression change was >1.5 fold different to the overall expression change for the gene from which they derived, or which showed a significant (logistic regression P<0.05 after multiple testing correction) change in observation to another junction with the same start or end position. A secondary intensity filter was applied to DESeq2 hits akin to a dynamic fold-change filter. DESeq2 comparisons were between wild-type and TDP43Q331K/Q331K mice or between MB+ and MB- mice. Significant expression and splicing changes between wild-type and TDP43Q331K/Q331K were used to generate hierarchical cluster plots including TDP43Q331K/+ mice to identify patterns of changes across replicates. Significant expression and splicing changes between MB+ and MB- mice were used to generate hierarchical cluster plots including wild- type mice. GO, KEGG enrichment analysis The Database for Annotation, Visualization and Integrated Discovery (DAVID) v6.8 was used for functional annotation of gene expression data in addition to the Functional Enrichment Analysis tool (FunRich v3.0) (available at:http://funrich.org). Gene ontology (GO) biological process (BP) and KEGG pathway enrichment analysis was conducted using DAVID and FunRich with a threshold Benjamini-corrected p-value≤0.05. Spinal cord RNA extraction for qPCR Tissues were briefly washed in ice cold PBS to remove mounting media, homogenised and RNA was extracted as described above for frontal cortices and hippocampi. Quantatitive PCR 500 ng of RNA was reverse transcribed (QuantiTect Reverse transcription kit, Qiagen) and the output volume of 20 μL diluted 10-fold in nuclease free water (Promega). Real-time PCR was performed using Brilliant-III Ultra-Fast SYBR (Agilent Technologies) on a Bio-Rad CFX96 instrument with cycle conditions based on Agilent’s quick reference guide (publication number 5990-3057, Agilent Technologies). Reaction specificity was confirmed by melt curve analysis and normalised expression (ΔΔCq) calculated using CFX Manager software 3.1 with at least four reference genes. For qPCR primer sequence see Supplementary Table 6. Reference genes used were: Ywhaz, Pgk1, Gapdh and Hprt1. KiCqStart SYBR Green primers for these reference genes were purchased from Sigma-Aldrich in addition to Tardbp. Statistical analyses Statistical analyses were conducted using Prism 6.05 (GraphPad). Graphs were plotted using Graphpad or Python. Use of parametric tests required data to be sampled from a Gaussian distribution. Homogeneity of variance between experimental groups was confirmed by the Browne-Forsythe test for ANOVA and F test for unpaired t-tests. For comparisons between genotypes or experimental groups two-tailed, unpaired t-tests or one-way ANOVA were used when comparing two or three groups respectively. Multiple comparisons by ANOVA were corrected using the Holm-Sidak test. Where the assumptions of one-way ANOVA were violated the non-parametric Kruskal-Wallis test was performed followed by Dunn’s multiple 130 comparison test. All statistical comparisons are based on biological replicates unless stated otherwise. Where technical replication of experiments occurs, this is highlighted in the respective method. Analyses of Rotarod performance, weights and food intake utilised repeated measures two- way ANOVA. Mice lacking measurements at any timepoint were excluded from analyses. Multiple comparisons by two-way ANOVA were corrected using the Holm-Sidak test. TDP-43 fluorescence in the nuclear and cytoplasmic compartments of parvalbumin positive cells and cell counts in multiple regions of the cortex were compared using multiple t-tests. Multiple comparisons were corrected using the Holm-Sidak test (alpha = 5%) without assuming consistent standard deviation. Statistical Analysis: ACBM The ACBM system characterized each behaviour for every frame of recording and quantified the amount of time the mouse was performing a given behaviour for each hour (0-23). These data were averaged across five days of recording within each animal and then subject to statistical comparison for within-day and between-group analyses. Statistical analysis to compare the average time spent performing a given behaviour between TDP43Q331K/Q331K and wild-type mice was conducted using repeated measures two-way ANOVA, in which the between-subjects variable was genotype and the within-subjects variable was circadian hour (0-23). We report main effects of genotype and genotype x circadian hour interactions. All statistics were calculated using IBM SPSS Statistics 24, alpha = 0.05. Statistical analyses: Touchscreens Data analyses for touchscreen and object recognition tasks were conducted using R version 3.3.1. Mixed-effects models were used to identify the main effects of genotype or task conditions (i.e., stimulus duration in 5-CSRTT or delay in object recognition task) and interactions between these factors. Between-genotype differences in sessions to criteria, FR, and PR outcomes were analysed by one-way ANOVA with Holm-Sidak post hoc test. Additional statistical information See Supplementary Table 7. Randomisation The order and genotype of animals and samples tested was randomized by one operator before subsequent experimental studies were conducted by a second investigator. Reproducibility Life Science Reporting Summary is available online. Data availability The authors will make all data available to readers upon request. RNAseq data have been deposited are available at https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?accGSE99354. Online methods references 51 Zhu, L. J., Holmes, B. R., Aronin, N. & Brodsky, M. H. CRISPRseek: a bioconductor package to identify target-specific guide RNAs for CRISPR- Cas9 genome-editing systems. PLoS One 9, e108424, doi:10.1371/journal.pone.0108424 (2014). 52 Heath, C. J., Bussey, T. J. & Saksida, L. M. Motivational assessment of mice 131 using the touchscreen operant testing system: effects of dopaminergic drugs. Psychopharmacology 232, 4043-4057, doi:10.1007/s00213-015- 4009-8 (2015). 53 Romberg, C. et al. Depletion of perineuronal nets enhances recognition memory and long-term depression in the perirhinal cortex. J Neurosci 33, 7057-7065, doi:10.1523/JNEUROSCI.6267-11.2013 (2013). 54 Wu, L. S., Cheng, W. C. & Shen, C. K. Targeted depletion of TDP-43 expression in the spinal cord motor neurons leads to the development of amyotrophic lateral sclerosis-like phenotypes in mice. J Biol Chem 287, 27335-27344, doi:10.1074/jbc.M112.359000 (2012). 55 Baghirova, S., Hughes, B. G., Hendzel, M. J. & Schulz, R. Sequential fractionation and isolation of subcellular proteins from tissue or cultured cells. MethodsX 2, 440-445, doi:10.1016/j.mex.2015.11.001 (2015). 56 Kalmar, B., Blanco, G. & Greensmith, L. Determination of Muscle Fiber Type in Rodents. Current protocols in mouse biology 2, 231-243, doi:10.1002/9780470942390.mo110229 (2012). 132 Type of file: figure Label: 1 Filename: figure_1.tif 133 Type of file: figure Label: 2 Filename: figure_2.tif 134 Type of file: figure Label: 3 Filename: figure_3.tif 135 Type of file: figure Label: 4 Filename: figure_4.tif 136 Type of file: figure Label: 5 Filename: figure_5.tif 137 Type of file: figure Label: 6 Filename: figure_6.tif 138 Type of file: figure Label: 7 Filename: figure_7.tif 139 Type of file: figure Label: 8 Filename: figure_8.tif 140 Supplementary Figure 1 ACBM walking phenotypes (a) Uncropped agarose gel of genotyping PCR used in figure 1b. (b) Automated continuous behavioral monitoring of walking behavior from 4 to 11.5 months of age separated by sex (n = 5 mice per genotype). Male mice; 4 months: interaction P=0.001; 6 months:  interaction P=0.003; 7.5 months: interaction P<0.0001; ∗ ∗ ∗ 10 months: interaction P< 0.0001; 11.5 months: interaction P< 0.0001. ∗ ∗ Female mice; 4 months: interaction P=0.034; 6 months: n.s. interaction P=0.138; 7.5 months: n.s. interaction P=0.334; 10 months: ∗ interaction P< 0.0001; 11.5 month: interaction P< 0.0001;  repeated measures two−way ANOVA. ∗ ∗ Error bars represent mean ± s.e.m. 141 Supplementary Figure 2 Neuromuscular investigations Q331K/+ Q331K/Q331K (a) Weights of psychology cohort 2 mice (n = 16 wild-type, 12 TDP−43 , 14 TDP-43 mice). Comparison by repeated measures two−way ANOVA. (b) TDP−43 immunohistochemistry in motor neurons of 5−month−old mice (n = 5 mice per genotype). Representative images shown. Scale bar, 10µm. (c) Representative examples of NMJ immunostaining in gastrocnemius muscles (scale bar, 20µm) with (d) quantification of innervation Q331K/Q331K in wild−type and TDP43 mice at 5 months of age (n = 8 mice per genotype). Comparisons: Innervated: P=0.528 (ns); denervated: P=0.127 (ns). Q331K/Q331K (e) Innervation/denervation analysis of gastrocnemius muscles in 18 to 23−month−old wild−type and TDP43 mice (n = 3 mice per genotype). Comparisons: Innervated: P=0.678 (ns); denervated: P=0.801 (ns). For (d) and (e) unpaired t test with correction using the Holm−Sidak method. (f) Succinate dehydrogenase enzymatic activity in gastrocnemius muscles from 5−month−old mice (n = 8 mice per genotype). Representative images shown. Scale bar, 200µm.  Error bars for (a,d,e) denote s.e.m.  (g) Tetanic force responses to repetitive stimulation at 5, 10 and 20Hz from FDB-tibial nerve preparations. Representative Q331K/Q331K wild−type and TDP43 mice are shown. Q331K/Q331K (h) Motor unit number estimates in wild−type and TDP43 mice based on inspection (Insp) of traces such as those shown in main Fig. 2g, and extrapolation between average motor unit size of the first four units relative to maximum muscle twitch tension Q331K/Q331K (MUNE) (n = 5 wild−type and 5 TDP-43 mice). Comparison: Kruskal−Wallis. Q331K/Q331K Q331K/Q331K (i) Maximum tetanic force in TDP43 mice compared with controls (50Hz stimulation) (n = 5 wild−type and 5 TDP−43 mice). Comparison: P=0.150 (ns); two-tailed Mann−Whitney. Q331K/Q331K (j) Tetanus−twitch ratio (expressed as a percentage of maximum force) (n = 4 wild−type and 5 TDP−43 mice). Comparison: P>0.999 (ns); two-tailed Mann−Whitney. 142 (k,l) Fatigue profiles showing decline in maximum tetanic force with repeated stimulation (50Hz for 1 s every 5 s) in representative Q331K/Q331K TDP43 and wild−type mouse FDB preparations. Q331K/Q331K (m) Time constant of decay of tetanic force comparing the muscles tested (n = 4 wild−type and 5 TDP−43 mice). Comparison: P>0.999 (ns); two-tailed Mann−Whitney. (n,o) Contractile force measurements with progressively graded stimulation (arbitrary units) from representative wild−type and Q331/Q331K TDP−43 mice (n = 3 mice per genotype). Error bars for (h,i,j,m) represent median and interquartile range. 143 Supplementary Figure 3 Laser capture and RNA sequencing analysis of lumbar spinal motor neurons (a) Nissl staining of lumbar spinal cord showing anterior horn motor neurons before and after laser capture (n = 4 wild-type, 4 Q331K/+ Q331K/Q331K TDP−43 , 4 TDP−43 mice). Scale bar, 300µm. (b) Quality control measures of laser−capture RNA sequencing data. A mean of 49.9 million reads−per−mouse (range 43.7−57.3million) were obtained. Displayed is the percentage of reads that are mapped within genes and exons; reads from ribosomal or mitochondrial RNA; percentage of annotated genes measured and the percentage of reads that are mapped to the sense strand. Colored dots represent individual libraries prepared from each mouse sequenced. (c) Filtering of lumbar spinal cord DESeq2 alternative splice events that are significantly different between wild−type and Q331K/Q331K TDP−43 mice. Non−expression hits reflect changes in splice junction usage that exceed a 1.5−fold change relative to expression of the gene from which they are derived. Log Reg hits include splice junctions whose usage changes relative to another junction with the same start or end position.  Q331K/Q331K (d) MA plot and (e) hierarchical clustering of alternative splice events in the lumbar spinal cord of wild−type and TDP−43 mice Q331K/+ Q331K/Q331K Q331K/Q331K (n = 4 wild-type, 4 TDP−43 , 4 TDP−43 mice). Comparison: DESeq2 wild−type v TDP−43 . Q331K/Q331K (f) Immunohistochemistry for AOX1 in lumbar motor neurons of 5−month−old mice (n = 4 wild-type, 4 TDP−43 mice). Representative images shown. Scale bar, 100µm. 144 Supplementary Figure 4 Additional behavioural outcomes from 5−choice serial reaction time tasks and marble burying assays (a−b) Response latencies. Genotype comparison: P=0.448 in 6−month−old; P=0.181 12−month−old mice. (c−d) Reward collection latencies. Genotype comparison: P=0.662 in 6−month−old; P=0.821 12−month−old mice. (e−f) Premature responses. Genotype comparison: P=0.068 in 6−month−old; P=0.194 12−month−old mice. (g−h) Perseverative responses after correct response. Genotype comparison: P=0.312 in 6−month−old; P=0.291 12−month−old mice. Q331K/+ Q331K/Q331K At 6 months of age, n = 15 wild−type, 16 TDP−43 , 15 TDP−43 mice, and at 12 months, n = 15 wild−type, 16 Q331K/+ Q331K/Q331K TDP−43 , 16 TDP−43 mice. For (a−h) mixed−effects model was used and error bars denote s.e.m. Q331K/+ (i) Correlation between omissions at 6 and 12 months of age in individual mice (n = 15 wild−type, 16 TDP−43 , 15 Q331K/Q331K Q331K/+ Q331K/Q331K TDP−43 mice). Comparisons: wild-type: P=0.600 (ns); TDP−43 : P=0.392 (ns); TDP−43 : P=0.046 (∗); Pearson’s analysis. Q331K/+ (j) Progressive marble burying of cohort 1 mice. Pairwise comparisons: 5 months (n = 19 wild−type, 19 TDP−43 , 17 Q331K/Q331K Q331K/+ Q331K/Q331K TDP−43 mice); wild−type vs. TDP−43 : P=0.03 (∗); wild−type vs. TDP−43 : P=0.013 (∗); 8 months (n = 16 Q331K/+ Q331K/Q331K Q331K/+ Q331K/Q331K wild−type, 11 TDP−43 , 15 TDP−43 mice); wild−type vs. TDP−43 : P>0.999 (ns); wild−type vs. TDP−43 : Q331K/+ Q331K/Q331K Q331K/+ P=0.02 (∗); 10 months (n = 16 wild−type, 14 TDP−43 , 15 TDP−43 mice); wild−type vs. TDP−43 : P=0.03 (∗); Q331K/Q331K Q331K/+ Q331K/Q331K wild−type vs. TDP−43 : P=0.03 (∗); 12 months (n = 16 wild−type, 13 TDP−43 , 15 TDP−43 mice); wild−type vs. Q331K/+ Q331K/Q331K Q331K/+ TDP−43 : P>0.999 (ns); wild−type vs. TDP−43 : P=0.001 (∗∗); 14 months (n = 15 wild−type, 13 TDP−43 , 14 Q331K/Q331K Q331K/+ Q331K/Q331K TDP−43 mice); wild−type vs. TDP−43 : P=0.396 (ns); wild−type vs. TDP−43 : P=0.003 (∗∗); 18 months (n = 15 Q331K/+ Q331K/Q331K Q331K/+ Q331K/Q331K wild−type, 13 TDP−43 , 14 TDP−43 mice); wild−type vs. TDP−43 : P=0.009 (∗∗); wild−type vs. TDP−43 : P<0.0001 (∗∗∗∗); Kruskal−Wallis followed by Dunn’s test for pairwise comparisons. Error bars for (j) represent median and interquartile range. 145 Supplementary Figure 5 Frontal cortical histology (a) Nissl−stained coronal sections of frontal cortex. Representative images shown. Cortex scale bar, 1mm; motor, cingulate and somatosensory cortex scale bars, 500µm; motor cortex layer V scale bar, 500µm. (b) Quantification of cells in cortical sub regions (n = 5 mice per genotype). Comparisons: Prefrontal: P=0.477(ns); Motor: P=0.931(ns); Motor layer V: P=0.897(ns); Cingulate: P=0.734(ns); Somatosensory: P=0.150(ns); multiple t tests with correction using the Holm−Sidak method.  Q331K/Q331K (c) Quantification of frontal cortical area in TDP−43 mice relative to wild type (n = 6 mice per genotype). Comparison: P=0.701 (ns); unpaired t test. Error bars denote s.e.m. Q331K/Q331K (d) Uncropped western blots of frontal cortical tissue from four wild−type and four TDP−43 mice used in the quantification of TDP−43 in figure 4e. 146 Supplementary Figure 6 Frontal cortical RNAseq, tau staining and validation in line #3 mice (a) Quality control measures of sequencing reads from 20 frontal cortical libraries from 5-month-old mice (coloured dots represent libraries for individual mice). A mean of 58.7m reads−per−mouse (range 42.8−76.3m) were obtained. (b) Filtering of frontal cortex DESeq2 alternative splice events that are significantly different between 5−month−old wild−type and Q331K/Q331K TDP−43 mice. Non−expression hits reflect changes in splice junction usage exceeding a 1.5−fold change relative to expression of the gene from which they are derived. Log Reg hits include splice junctions whose usage changes relative to another junction with the same start or end position. (c) Immunostaining for tau in the cortices of 20−month−old mice. Neuronal cells have been stained with NeuN (n = 3 mice per genotype). Representative images shown. Scale bar, 25µm. Q331K/+ Q331K/Q331K (d) Marbles buried by line #3 mice (n = 4 wild−type, 8 TDP−43 , 5 TDP−43 mice). Pairwise comparisons: wild−type vs. Q331K/+ Q331K/Q331K TDP−43 : P=0.691 (ns); wild−type vs. TDP−43 : P=0.020 (∗); Kruskal−Wallis followed by Dunn’s test. Error bars represent median and interquartile range. Q331K/+ Q331K/Q331K (e) qPCR of expression and splicing changes in line #3 mice (n = 5 wild−type, 5 TDP−43 , 5 TDP−43 mice). Pairwise Q331K/+ Q331K/Q331K comparisons: Tardbp expression: wild−type vs. TDP−43 : P=0.076 (ns); wild−type vs. TDP−43 : P=0.0016 (∗∗); Q331K/+ Q331K/Q331K Q331K/+ Q331K/Q331K TDP−43 vs. TDP−43 : P=0.036 (∗); 0N: wild−type vs. TDP−43 : P=0.072 (ns); wild−type vs. TDP−43 : Q331K/+ Q331K/Q331K Q331K/+ P=0.495 (ns); TDP−43 vs. TDP−43 : P=0.03 (∗); 2N/0N: wild−type vs. TDP−43 : P=0.877 (ns); wild−type vs. Q331K/Q331K Q331K/+ Q331K/Q331K TDP−43 : P=0.002 (∗∗); TDP−43 vs. TDP−43 : P=0.002 (∗∗); P<0.0001 (∗∗∗∗); one−way ANOVA followed by Holm−Sidak post−hoc tests for pairwise comparisons. Error bars denote s.e.m. (f) Immunohistochemistry for parvalbumin in cortices of line #3 mice. Representative images shown. Scale bar, 250µm and quantification of parvalbumin−positive neurons (n = 4 mice per genotype). Comparison: P=0.006 (∗∗); unpaired t test. Error bars denote s.e.m. (g) Filtering of 5−month−old frontal cortex DESeq2 alternative splice events that are significantly different between MB+ and MB− Q331K/Q331K TDP−43 mice. Refer to subfigure (b). Q331K/Q331K (h) Hierarchical clustering of alternative splice events in 5−month−old frontal cortices comparing MB+ and MB− TDP−43 mice 147 Q331K/Q331K Q331K/Q331K (n = 6 wild−type, 4 MB+ TDP−43 and 4 MB− TDP−43 mice); Comparison: DESeq2 MB+ vs MB−. 148 Supplementary Figure 7 RNASeq in aged mice (a) Quality control measures of sequencing reads from 28 frontal cortical libraries from 20−month−old mice (coloured dots represent libraries for individual mice). (b) Filtering of 20−month−old frontal cortex DESeq2 alternative splice events that are significantly different between wild−type and Q331K/Q331K TDP−43 mice. Non−expression hits reflect changes in splice junction usage exceeding a 1.5−fold change relative to expression of the gene from which they are derived. Log Reg hits include splice junctions whose usage changes relative to another junction with the same start or end position. Q331K/+ (c) Hierarchical clustering of differentially expressed genes in 20−month−old frontal cortices comparing MB+ and MB− TDP−43 Q331K/+ Q331K/+ mice (n = 8 wild−type, 5 MB+ TDP−43 and 5 MB− TDP−43 mice); Comparison: DESeq2 MB+ vs MB−. (d) Filtering of 20−month−old frontal cortex DESeq2 alternative splice events that are significantly different between MB+ and MB− Q331K/+ TDP−43 mice. Refer to subfigure (b). Q331K/+ (e) Hierarchical clustering of alternative splice events in 20−month−old frontal cortices comparing MB+ and MB− TDP−43 mice (n Q331K/+ Q331K/+ = 8 wild−type, 5 MB+ TDP−43 and 5 MB− TDP−43 mice); Comparison: DESeq2 MB+ vs MB−. Q331K/+ (f) qPCR of splicing changes in Matr3 exon 14. Pairwise comparisons: wild−type vs. TDP−43 : P=0.07 (ns); wild−type vs. Q331K/Q331K Q331K/+ Q331K/Q331K TDP−43 : P=0.0014 (∗∗); TDP−43 vs. TDP−43 : P=0.033 (∗). Q331K/+ (g) qPCR of splicing changes in Sqstm1. Pairwise comparisons: wild−type vs. TDP−43 : P=0.003 (∗); wild−type vs. Q331K/Q331K Q331K/+ Q331K/Q331K TDP−43 : P<0.0001 (∗∗∗∗); TDP−43 vs. TDP−43 : P=0.005 (∗∗). (f−g) For qPCR, n = 5 mice per genotype; one−way ANOVA followed by Holm−Sidak post−hoc tests. Error bars denote s.e.m. 149 Supplementary Table 2 Myelination and oligodendrocyte associated genes 150 CRISPR/Cas9 PAM sequences Name target site (spacer+PAM) gRNAf5_msTDP43_ptmutStart66End88 aggcagcgttgcagagc|agttgg gRNAf6_msTDP43_ptmutStart67End89 ggcagcgttgcagagca|gttggg gRNAf7_msTDP43_ptmutStart68End90 gcagcgttgcagagcag|ttgggg single-stranded DNA (ssDNA) oligonucleotide ssDNA donor oligonucleotide_ TDP43msPtmut1 atgaactttggtgcttttagcattaacccagcgatgatggctg (mutation highlighted in red) cggctcaggcagcgttgAagagcagttggggtatgatgggcat gttagccagccagcagaaccagtcgggcccatctg Supplementary Table 3 CRISPR sequences 151 Preparation of DNA for genotyping Extract DNA from ear biopsies at 10 days of age using QuickExtract (Epicentre). Heat sample to 65°C for 15 minutes, briefly vortex and heat to 98°C for 2 minutes. 1µl was subsequently used for PCR. Genotyping primers Name Sequence Forward TTGTTCAGCAGATTGCCACC Reverse CAGCAGTTCACTTTCACCCA PCR cycling parameters Temperature Time 94°C 2 minutes 94°C 30 seconds 64°C 30 seconds 72°C 60 seconds 72°C 7 minutes 30x cycle PCR Restriction digestion and gel electrophoresis The Q331K mutation introduces a SapI/EarI restriction site, which facilitates genotyping 1) Column purify PCR products (Qiagen) 2) Digest with SapI at 37°C for 1 h 3) Resolve on a 1.5% agarose gel 4) A wild-type allele is indicated by a ~1 kb band. A mutant allele is indicated by a ~500 bp band (due to the overlap of two 500 bp digestion products) Supplementary Table 4 Genotyping protocol 152 Primary antibodies for immunofluorescence Antigen (code) Antibody Dilution Source TDP-43 (ab41881) Rabbit polyclonal 1:200 Abcam Neurofilament heavy (ab4680) Chicken polyclonal 1:200 Abcam Parvalbumin (ab11427) Rabbit polyclonal 1:200 Abcam MAP2 (ab5392) Chicken polyclonal 1:200 Abcam Tau (A0024) Rabbit polyclonal 1:500 Dako NeuN (ab104224) Mouse monoclonal 1:500 Abcam AOX1 (PA5-36922) Rabbit polyclonal 1:200 ThermoFisher Scientific Synaptophysin (ab8049) Mouse monoclonal 1:200 Abcam βIII-tubulin (T2200) Rabbit polyclonal 1:200 Sigma Aldrich Primary antibodies for western blotting Antigen (code) Antibody Dilution Source TDP-43 (ab41881) Rabbit polyclonal 1:1000 Abcam Cyclophilin (ab41684) Rabbit polyclonal 1:1000 Abcam Nucleoporin (mab414) Mouse monoclonal 1:500 Abcam Supplementary Table 5 Primary antibodies used for immunofluorescence and immunoblotting 153 Gene Forward Reverse Tardbp Intron 7 TTCATCTCATTTCAAATGTTTATGGAAG ATTAACTGCTATGAATTCTTTGCATTCAG exclusion Sort1 Exon 17b AACCCCACAAAGCAGGACT CTGCTACGACTGTGACAAGC inclusion Sort1 Exon 17b ATCCCAGGAGACAAATGCCA TGGAATTCTGCTTTGTGGGG excluded 0N Mapt TTAAAAGCCGAAGAAGCAGGC CTGGAGGAGTCTTAGGGCTG 1N Mapt CGCCAGGAGTTTGACACAAT CCTGCTTCTTCGGCTTCAG 2N Mapt CATGGCTTAAAAGAGTCTCCCC CCTGCTTCTTCGGCTGTAAT Mbp (Basic) GCCCTCTGCCCTCTCATG TCTTGAAGAAATGGACTACTGGG Mbp (Golli) ATTGGGTCGCCATGGGAAA CGCTTCTCTTCTTTCCAGCC Matr3 GCACTTTGGTTTCAGGGGAG CTGTTAGGAATCCGCAGCAC Matr3 (Exon 14) AGCAGCCTTCCTCATTATCAGA TTCCTTCTTCTGCCTCCGTT Sqstm1 CCAGTGATGAGGAGCTGACA CCGGCACTCCTTCTTCTCTT Sqstm1 (TDP-43 ACCCATCTACAGGCTGATCC GTCTGTAGGAGCCTGGTGAG Q331K variant) Supplementary Table 6 Primers for quantitative PCR 154 CHAPTER 4: BEHAVIORAL CHARACTERIZATION OF THE NOVEL TDP-43Q331K KNOCK-IN MOUSE MODEL OF ALS-FTD THROUGH AUTOMATED CONTINUOUS BEHAVIORAL MONITORING Amanda Marie Duffy, Justin R. Fallon Department of Neuroscience, Brown University, Providence, RI 02912 Contributions: I performed all of the experiments and analyses presented here. T. Serre Ph.D., V. Veerabadran, and T. Sharma conducted ACBM data processing. 155 ABSTRACT Characterizing disease progression in multiple mouse models provides the opportunity to identify similar phenotypes between models that may relate to the same underlying mechanism. It is of particular interest to characterize early phenotypes, as these are more likely to reflect the initial events in disease progression - the preferred targets for therapeutic interventions. Automated Continuous Behavioral Monitoring (ACBM) provides an opportunity to identify such early phenotypes. Using ACBM, mice are video- recorded continuously at thirty frames/sec in their home cages for five days (total of ~1.3 x 107 frames/mouse/session). Behavioral assessment is then performed using a supervised, machine learning-based, computer algorithm to assign one of nine designated behaviors to each frame. Having identified early P30 and P58 neuromotor deficits in the transgenic SOD1G93A mouse model of ALS using ACBM, here we used ACBM to assess early behavioral phenotypes in two separate cohorts of a novel knock-in mouse model of ALS-FTD (amyotrophic lateral sclerosis-frontal temporal dementia) that harbors the human mutation of TDP-43Q331K. Mutations in the RNA binding protein TDP-43 can cause familial ALS-FTD. However, the mechanisms by which these mutations cause disease are poorly understood. Knock-in animal models are particularly useful as they express mutant alleles under the control of endogenous regulatory machinery. Using ACBM, we conducted five sessions of ACBM on a cohort of wild-type (WT) and TDP-43Q331K knock-in (KI) mice at P120, P180, P225, P300, and P345 (5 156 males, 5 females per genotype) (each session lasts five days). The KI mice exhibited a walk deficit at P120 in both males and females. We also revealed gender differences in which male KI mice exhibit decreased walking in all rounds, while females exhibit decreased walking in three out of five rounds. To assess earlier ages and to evaluate reproducibility of our P120 data, we conducted ACBM on a second cohort of mice at P30 and P120 (10 males, 10 females per genotype). At P30, the male knock-in mice did not exhibit a walk deficit, however they did at P120. The P30 knock-in females, however, exhibit a walk phenotype that persists at P120 demonstrating a gender difference in walk phenotype onset. Our results demonstrate that TDP-43Q331K knock-in mice exhibit behavioral abnormalities as early as P30 that progress with age. ACBM represents a general approach for rigorous, quantitative and unbiased behavioral assessment of both identifying early phenotypes and characterizing natural history of ALS, ALS-FTD and potentially other models for neurological disease. 157 Introduction Amyotrophic lateral sclerosis – (ALS) is a progressive neurodegenerative disease of upper and lower motor neurons (UMNs and LMNs) that results in weakness, paralysis, and eventually death within approximately 2-5 years [8]. The average age of ALS diagnosis is within the sixth decade [9]. There is currently no treatment for ALS, and the primary FDA-approved drug, riluzole, extends life by only a few months. Frontal temporal dementia (FTD) is the third most common form of dementia in patients over 65, and results from neurodegeneration of the frontal and anterior temporal lobes [41]. Compared to Alzheimer’s disease (AD), patients with FTD have a decreased lifespan and faster disease progression. One study demonstrates that FTD patients survive approximately four years after disease onset while AD patients survive six years after onset [42]. There are three variants of FTD. 50% of FTD patients have a behavioral variant (bvFTD), which results in progressive social, emotional, personality and conduct abnormalities [41]. These patients can exhibit apathetic-type behaviors such as decreased motivation, isolation, and loss of socio-emotional awareness, or disinhibited-type behaviors such as hyper-orality, preference for sweet foods, motor stereotypies and perseverative behavior. These patients are often impulsive, disinhibited and socially inappropriate [41] [44]. 20-25% of FTD patients exhibit semantic variant FTD (svFTD) [41]. svFTD patients exhibit fluent language however there is a progressive loss of object knowledge, and they demonstrate paraphasic errors (adding incorrect syllables and 158 words during speech effort), and semantic errors (using words with incorrect meanings) [41]. 20-25% of FTD patients exhibit non-fluent variant FTD (nfvFTD), in which patients struggle with motor control related to speech resulting in inappropriate pauses even within utterances [41]. ALS and FTD exist along a spectrum, as 15% of ALS patients are also diagnosed with FTD and 15% of FTD patients are also diagnosed with ALS. However, 50% of ALS patients exhibit cognitive deficits and 40% FTD patients exhibit motor neuron disease [48]. While FTD patients can survive for over a decade, patients within the ALS-FTD spectrum have a shorter lifespan, surviving on average for 3-5 years [44]. Virtually all ALS patients and 40% of FTD patients exhibit cytosolic transactive-response RNA/DNA binding protein (TDP-43) aggregates implicating TDP-43 as a potential target for further investigation and possible intervention. TDP-43 is an RNA/DNA binding protein that regulates transcription (including its own), splicing, miRNA biogenesis and mRNA transport. It contains two RNA recognition motifs as well as a carboxy-terminal glycine-rich domain [53]. It is found primarily in the nucleus and is ubiquitously expressed [54]. Under conditions of stress, it translocates from the nucleus to the cytoplasm. TDP-43 biomarkers for pathology include elevated TDP-43 plasma and CSF levels in ALS patients, and these levels are comparable between sALS and fALS cases [11]. Over 40 mutations in the TDP-43 gene (TARDBP) account for 4% of fALS patients, 1% of sALS and 1% of FTD patients [55] [48]. These mutations are located 159 primarily in the C-terminal glycine-rich RNA binding domain, which is important for protein-protein interactions [48]. Mutations in the RNA binding domain of TDP-43 result in a loss of function causing TDP-43 to mislocalize from the nucleus to the cytoplasm. The depletion of TDP-43 from the nucleus can interfere with downstream processes such as splicing, and as TDP-43 levels increase in the cytoplasm, it binds mRNA molecules that are not fully processed, preventing their normal function. As TDP-43 accumulates in the cytoplasm, pathological forms of TDP-43 further propagate its mislocalization and depletion from the nucleus [48]. Knock-in models are especially useful tools to model human disease due to accurate gene copy number and endogenous gene regulation. Therefore, knock-in mouse models with a mutation found in human patients would provide targeted information as to the function of the mutated gene and the resulting protein. The development of CRISPR/Cas9 technology has enabled more effective and efficient generation of knock-in mice. CRISPR/Cas9 technology is based on the endogenous machinery utilized by bacteria as a defense against viral infection. The TDP-43Q331K homozygous knock-in mouse model developed by Dr. Jemeen Sreedharan harbors a point mutation of a glutamine to lysine substitution [79]. Through sequencing a cohort of 200 sALS patients, the mutation was identified in a 72-year old man with limb-onset ALS. The duration of his disease was three years [79]. Within the TDP-43 C-terminus (involved in protein-protein interactions), TDP-43Q331K creates a protein kinase A site, which may interfere with phosphorylation. In assessing the functional significance of this mutation, Sreedharan et al. expressed TDP-43wt and TDP-43Q331K in Chinese hamster ovary 160 (CHO) cells. Immunoblotting the cytoplasmic fraction for the N-terminal Myc tag showed fragments with molecular weights between ~14 to ~45 kD, with the mutant TDP-43 exhibiting prominent fragments in comparison to WT, suggesting that the TDP-43Q331K mutation increases TDP-43 fragmentation, corroborating the evidence for mutated TDP-43 playing a role in its aggregation in ALS patients. Additionally, the Sreedharan group transfected myc-tagged TDP-43WT and TDP-43Q331K into the spinal cords of stage 14 chick embryos and found that the mutants failed to develop limb and tail buds, and only 5-15% of mutant embryos reached a normal stage of maturation after 24 hours post electroporation, while WT development proceeded normally. Additionally, the mutant embryos exhibited an increase in the number of apoptotic nuclei compared to WT [79]. These data suggest that the TDP-43Q331K point mutation identified in human ALS patients is toxic and could contribute to disease phenotypes. Behavioral phenotyping is critical to understanding how mechanistic abnormalities in disease models result in functional abnormalities. To reveal behavioral phenotypes in the TDP-43Q331K mouse, we conducted Automated Continuous Behavioral Monitoring (ACBM), as described in previous chapters. Through ACBM, we had previously revealed a walk deficit in male and female TDP-43Q331K knock-in mice (referred to hereafter as KI mice [knock-in]) as early as P120 using ACBM [82]. In that study, we assessed these mice at five ages starting at P120 (see previous chapter) [82]. Here, we wanted to probe behavioral phenotypes in a separate cohort of mice at an earlier age to reveal an age of possible phenotype onset. Therefore, we conducted ACBM on a larger cohort of KI and WT mice (male 161 and female) at P30 and P120. The purpose of conducting ACBM at P120 was to characterize reproducibility or potential differences between cohorts. We found that the KI walk deficit is absent at P30 in males, yet there is a slight walk deficit in females at this age. At P120, both male and female KI mice exhibit this walk deficit, and this result replicates our previous findings. We also characterized translocation and did not identify a phenotype in males nor females at P30, yet identified a decrease in translocation in females at P120. The presence of a small translocation deficit in KI females and an absence of a translocation phenotype in males at P120 also replicate what we found previously. To further probe contributing factors to the P120 female translocation deficit, we separated translocation into its x- and y-coordinates and found that translocation in a specific direction can drive overall translocation and provide information as to what behaviors may be contributing to such an effect. As ALS-FTD is a neurodegenerative disease, identifying early phenotypes prior to the onset of pathology contributing to neurodegeneration provides an opportunity for early intervention. Early behavioral deficits could indicate underlying or developing pathology. 162 Results ACBM detects early and late onset gender-dependent phenotypes in the novel TDP-43Q331K knock-in mouse. Our previous study demonstrated KI behavioral phenotypes as early as P120, and this is the only behavioral data existing for this mouse model [82]. Here, we wanted to probe for phenotypes earlier than P120 to identify even earlier phenotypes or to determine an age of phenotype onset. Therefore, we chose to conduct ACBM on a larger cohort of male and female KI mice at P30 and P120 (20 KI, 20 WT; 10/gender). We identified an early KI female walk deficit at P30, which progressed at P120 (Fig. 1, Supplementary Fig. 1). We did not identify a walk phenotype in the males until P120 (Fig. 2, Supplementary Fig. 2). Additionally, at P120, we found a rear and translocation deficit in the KI females, which we did not identify in the males (Fig. 1, 2, 3, 4, Supplementary Fig. 1, 2). These data, described below, demonstrates that the KI mice exhibit behavioral phenotypes as early as P30, and there is a difference in disease progression and phenotypes between male and female mice. 163 Figure 1. ACBM detects a TDP-43Q331K walk deficit at P30 and a walk and translocation deficit at P120 in females. We conducted our analyses across the twenty-four circadian hours. Repeated Measures ANOVA; G: genotype effect; G-I: genotype x hour interaction. a. TDP-43Q331K female mice exhibit a walk deficit at P30 and P120 (P30: G: P=0.0.432; G-I: P=0.040; P120: G: P=0.012; G-I: p=2.378E-13). b. TDP-43Q331K female mice do not exhibit a translocation phenotype at P30 but exhibit a translocation deficit P120 (P30: G: p=0.0.778; G-I: p=0.179; P120: G: p=0.114; G-I: P=0.000001). Male KI mice develop a walk deficit at P120. In our previous study, we demonstrated a walk deficit in male KI mice at P120. Here, we conducted ACBM on a larger cohort of mice starting at P30 in order to determine the age of onset of this phenotype. At P30, we found that the KI male mice did not exhibit a walk phenotype (P30: Genotype effect [G]: p=0.302, Genotype x hour interaction [G-I]: p=0.552) (Fig. 2a, Supplementary Fig. 2a). To characterize these differences we used repeated 164 measures ANOVAs across all twenty-four hours (see methods). The lack of an effect at P30 demonstrates that the KI pathologies have either not initiated at this age or do not impact walking behavior. To determine whether the KI male mice develop a walk phenotype at P120, which would corroborate our results from our previous experiment, we conducted a second round of ACBM on these same mice at P120. At this age, we found that the KI male mice exhibit a walk deficit, suggesting that pathology associated with this genotype drives a change in walking behavior between P30 and P120 (p120: G: p=0.050, G-I: p=0.003) (Fig. 2a, Supplementary Fig. 2a). Therefore, the KI male mice do not exhibit an early P30 walk deficit, and the walk deficit at P120 corroborates our previous findings [82]. Figure 2. ACBM detects a TDP-43Q331K walk deficit at P120 in males. 165 The x-coordinate corresponds to circadian hours 0-23. For walk, the y-coordinate corresponds to the average number of seconds/hour averaged over the five recording days that the mice are performing walking. For translocation, the y- coordinate corresponds to the average number of pixels/hour over the five recording days that the mouse is moving in the x- and y-directions. We conducted our analyses across the twenty-four circadian hours. Repeated Measures ANOVA; G: genotype effect; G-I: genotype x hour interaction. a. TDP-43Q331K male mice do not exhibit a walk phenotype at P30, but exhibit a decrease in walking at P120 (P30: G: P=0.0.302; G-I: P=0.552; P120: G: P=0.050; G- I: P=0.003). b. TDP-43Q331K male mice do not exhibit a translocation phenotype at P30 or P120 (P30: G: P=0.0.542; G-I: P=0.287; P120: G: P=0.135; G-I: P=0.283). Female KI mice exhibit an early and progressive walk deficit at P30. We also conducted ACBM on female KI and WT mice at P30 and P120 to identify phenotypes as well as assess any gender differences. We found that the female KI mice exhibit a small walk phenotype at P30: (G: p=0.432, G-I: p=0.040) (Fig. 1a, Supplementary Fig. 1a). This early deficit suggests that the KI females are more susceptible at this age to early pathological mechanisms influencing the walk phenotype than males. To determine if the KI walk deficit persists, we conducted a second round of ACBM at P120, and found a significant deficit in KI walking behavior (G: p=0.012, G- I: p=0.2.37E-13) (Fig. 1a, Supplementary Fig. 1a). Therefore, the KI female mice exhibit an early and progressive walk deficit starting at P30. Female KI mice exhibit a translocation deficit at p120. The early deficit in walk prompted us to examine translocation, which is a broader measure of movement (number of pixels moved per hour). Translocation behavior can be analyzed as the total translocation in both the x- and y-coordinates 166 combined, translocation in just the x-coordinate and translocation in just the y- coordinate. To determine whether there is a translocation phenotype in the KI mice as well as to determine an age of onset, we analyzed translocation data on P30 and P120 in both male and female mice. Interestingly, at P30, the KI mice did not exhibit a translocation phenotype deficit in either males or females (males: G: p=0.542, GI: p=0.287; females: G: p=0.778, GI: p=0.179) (Fig. 2b, 1b, Supplementary Fig. 2b, 1b). This result demonstrates that translocation is not a measurement of overall walk behavior, as the females exhibited a walking deficit at this age. To determine whether disease progression results in the development of a translocation phenotype, we analyzed translocation in both males and females at P120, and found that the KI females translocate less than WT females (G: p=0.114, GI: p=0.000001) (Fig. 1b, Supplementary Fig. 1b). We found no translocation phenotype in the KI males (G: p=0.135, GI: p=0.283) (Fig. 2b, Supplementary Fig. 2b). In our previous cohort of KI and WT male and female mice, we also observed a translocation deficit in the females, and did not in the males at P120. Therefore, in repeating the experiment in a separate cohort of KI and WT mice at P120, we replicated the gender-dependent translocation deficit in the KI mice. Therefore, in females the translocation phenotype is a later onset phenotype than the earlier P30 walk deficit. 167 Analysis of translocation reveals a marginal decrease in X-coordinate translocation in males and a decrease in both X- and Y-coordinate translocation in females. To determine whether movement in either the x- and/or y-direction influences translocation at P120, we separated translocation at this age into its x- and y-component parts. KI males exhibited a marginal decrease in x-coordinate translocation, and did not exhibit a y-coordinate translocation effect (X: G: p=0.082; GI: p=0.204; Y: G: p=0.238; GI: p=0.260) (Fig. 3a). KI females, however, exhibited a deficit in both x- and y-coordinate translocation (X: G: p=0.092; GI: p=8.5314E-7; Y: G: p=0.143; GI: p=0.000002) (Fig. 3b). Figure 3. Separating x- and y-translocation demonstrates that the overall decrease in female TDP-43Q331K translocation is driven by movement in both directions at P120. 168 a. In males, there is no TDP-43Q331K translocation phenotype in the x- and y- coordinates combined, the x-direction alone (marginal), nor the y-direction alone (x- and y-combined: G: p=0.135; G-I: p=0.283; x-coordinate: G: p=0.082, G-I: p=0.204; y-coordinate: G: p=0.238; G-I: p=0.260). b. In females, there is a decrease in TDP-43Q331K translocation in the x- and y- coordinates combined, in the x-direction alone, and in the y-direction alone (x- and y-combined: G: p=0.114; G-I: p=0.000001; x-coordinate: G: p=0.092, G-I: p=8.5314E-7; y-coordinate: G: p=0.143; G-I: p=0.000002). Female TDP-43Q331K mice exhibit a rearing deficit at P120. In addition to KI walking and translocation deficits at P120 in females, we also identified a significant deficit in rearing. Interestingly, this rearing deficit was specific to the KI females, as KI males did not exhibit a rearing phenotype. This result demonstrates that gender influences rearing behavior in the KI mice (females: G: p=0.074, G-I: p=0.001; males: G: p=0.296, G-I: p=0.159) (Fig. 4). 169 Figure 4. Female TDP-43Q331K mice exhibit a rear deficit at p120. While the male TDP-43Q331K do not exhibit a rear phenotype, the female TDP-43Q331K mice exhibit a decrease in rear behavior (males: G: p=0.296, G-I: p=0.159; females: G: p=0.074, G-I: p=0.001). Discussion Using ACBM, we found that the TDP-43Q331K knock-in mouse model exhibits behavioral phenotypes that emerge in females as early as P30 and in males as early as P120. In our previous TDP-43Q331K mouse study, we identified an early walk deficit in males and females at P120 [82]. To determine the onset of this phenotype, we repeated our experiment with a separate cohort of mice starting at P30 [82]. While we did not observe a walk phenotype in the males at P30, we did observe a small walk deficit in the females at this age (Fig. 1, 2, Supplementary Fig. 1, 2). At P120, both males and females exhibited a walk deficit, replicating the results from our original cohort of KI mice, and demonstrating that the female mice are more susceptible to a walk deficit, and the onset of the deficit in males occurs between P30 and P120 (Fig. 1, 2, Supplementary Fig. 1, 2). We also identified a translocation deficit that was present at P120 in females but not males, again potentially demonstrating a higher susceptibility in females for developing behavioral (locomotor or cognitive) abnormalities at this age. We also identified a deficit in rearing behavior in females at P120 (Fig. 4). The decrease in walking and rearing behavior in KI females could contribute to the overall decrease in translocation. When we separated translocation into its x- and y-component parts we found that there was a significant deficit in x-coordinate 170 translocation and in y-coordinate translocation (Fig. 3). Walking contributes to translocation in the x-coordinate and rearing contributes to translocation in the y- coordinate. Therefore, it is not surprising that a decrease in both walk and rearing in females results in a substantial overall translocation deficit, as well as a deficit in x- and y-translocation separately. ALS-FTD patients exhibit both neuromotor and cognitive deficits. In characterizing the behavior of an ALS-FTD mouse model, one must consider both cognitive and neuromotor aspects of the disease. The most obvious potential neuromotor deficit that we identified in the knock-in mice was the walk deficit in both males and females, and this deficit was consistent at P120 across both of our studies. To determine if this phenotype is driven by denervation, one of the earliest pathologies associated with ALS, we quantified innervated and denervated neuromuscular junctions in the gastrocnemius muscle of 5-month and 18-23-month-old male mice [82]. We found that neither 5-month nor 18-23-month-old mice exhibit differences in the percent of innervated nor denervated neuromuscular junctions, suggesting that behavioral phenotypes occurring during these ages may not be due to denervation (Chapter 3) [82]. Cortical deficits may be driving the behavioral phenotypes of this knock-in mouse. FTD patients exhibit cortical degeneration in the frontal, temporal and insular regions of the brain. Depending on where degeneration is occurring results in differences in symptomology. The different variants of FTD are also driven by degeneration of specific brain areas. Patients with bvFTD have bilateral 171 degeneration in their frontal lobe. Patients with svFTD exhibit bilateral degeneration in the anterior temporal lobe, and nfvFTD exhibits degeneration in the left inferior frontal cortex and insular regions. While one cannot transpose the different variants of human FTD onto mouse models, it is interesting that different variants exist, suggesting that the KI mouse model of ALS-FTD may represent just one aspect of FTD. To investigate cortical phenotypes in the KI mouse, White et al. (2018) quantified the amount of parvalbumin-positive interneurons in the frontal cortex and found that there was a 25% decrease in these neurons. Parvalbumin is a calcium binding protein, and therefore by binding calcium, reduces calcium induced excitatory neurotransmitter release. Cortical and spinal motor neurons are especially susceptible to excessive excitation because they lack parvalbumin [102]. A decrease in parvalbumin positive interneurons in the frontal cortex of the KI mouse could result in an increase in excitation. Parvalbumin interneurons synapse on excitatory glutamatergic neurons and therefore, regulate excitation. The decrease in parvalbumin-positive interneurons would therefore result in a potential imbalance of excitation to inhibition and therefore a loss in the regulation of coordination among brain networks [103]. The walking deficit that we observe in the KI mice may relate to this decrease in parvalbumin interneurons and may represent a deficit in coordinated cortical circuitry. Further evidence suggesting that parvalbumin may be neuroprotective and therefore result in behavioral abnormalities, was demonstrated by Alexianu et al. (1994). They demonstrated that ALS autopsy specimens exhibit an absence of 172 parvalbumin specifically in cortical and spinal motor neurons, which are lost early in disease progression in comparison to motor neurons that exhibit resistance to degeneration such as oculomotor neurons [102]. This decrease in parvalbumin in the human cortex and spinal cord in ALS patients provides additional evidence that parvalbumin may be protective. Relating an increase in excitability to abnormal motor function, Burrell et al. (2011) demonstrated that FTD patients, specifically with progressive non-fluent aphasia (PNFA) exhibit reduction in short-interval cortical inhibition, as well as a prolonged central motor conduction time, which is a measurement of the amount of time it takes for a signal from the motor cortex to reach the spinal or bulbar neurons [104]. This increase in conduction time could also relate to the decrease in walking behavior that we observed in the KI mouse. While it was not examined, it is possible that there is a decrease in parvalbumin in the muscle of the KI animals as well, and a decrease in parvalbumin could therefore influence muscle function, potentially driving the walk deficit that we identified. Zebra fish that overexpress parvalbumin exhibit an increase in velocity and sustained locomotion and velocity than fish that do not overexpress it, providing some evidence that parvalbumin is important for efficient locomotion [105]. It is also possible that cognitive deficits contribute to the behavioral phenotypes we identified in the KI mice. If the walking deficit that we observe is related to a cortical disease phenotype, then it is possible that the walk deficit, and 173 potentially the rear deficit in females, is modeling apathy, which can also be a feature of FTD in humans [11]. White et al. (2018) conducted a series of behavioral experiments on the KI mice that may corroborate this theory. White et al. conducted a marble-burying task in which the KI mice bury fewer marbles than the WT mice [82]. The marble- burying task is used as a measure of repetitive and compulsive-like behavior, which is a natural response in mice to anxiety and fear [106]. The decrease in KI marble- burying may suggest a decrease in fear or anxiety, which may be modeling apathy. Additionally, the decrease in walking behavior that we observe in the KI mice may also be due to a general decrease in drive, motivation or interest in their surroundings, which may represent apathy. In addition to apathy, it is also possible that the KI mice are burying fewer marbles because the task is cognitively demanding, and requires sustained attention that these mice may lack. Patients with FTD, specifically those with frontal lobe atrophy often exhibit attentional deficits. It is possible that the KI marble-burying deficit indicates decreased attention, similar to that observed in FTD patients [107]. Additionally, White et al. used Touchscreen analysis as a measure of cognition in these mice. Touchscreen analysis is a paradigm in which the mouse is trained to identify a stimulus displayed on a screen by touching the stimulus with its nose. When the mouse has successfully completed the trial, it receives a food pellet. In this study, five stimuli are displayed on a screen and one lights up. The mouse must choose the lit stimulus in order to receive the food reward. White et al., 174 demonstrated that at 12-months of age the KI mice exhibited a greater number of errors than WT mice, demonstrating a cognitive deficit [82]. Rodents show a preference for novelties in their environment [108]. The deficit in KI walking in both males and females and decreased rearing behavior in females could be due to a loss of novelty seeking within their environment. We also found that the KI female mice, but not males, exhibit a decrease in translocation at P120, again suggesting a loss in exploratory behavior. Therefore, this decrease in exploratory behavior could represent a potential cognitive deficit, which may mirror the cognitive deficits exhibited by FTD patients. Another theory to explain decreased walking behavior in the KI mice may have to do with differences in the way the KI mice and WT mice adjust to single- housing, having been previously group-housed. Wistar rats exhibit an increase in locomotor activity after social isolation [109]. Perhaps WT mice exhibit the same phenomenon and increase their level of locomotion (walk) after being placed in single housing during ACBM. Therefore, relative to WT, the KI mice may not experience this same isolation-induced stress, and therefore their overall locomotion behavior relative to WT mice may be less. Interestingly, patients with FTD often exhibit social withdrawal, perhaps indicative of a decreased drive to engage in social behaviors [110]. We identified gender-based behavioral differences in the KI mice. The KI females exhibited walking deficits at P30, and P120. Male KI mice did not exhibit a phenotype at P30, and did at P120 (Females: P30: G: p=0.432, GI: p=0.040; P120: G: p=0.012; GI: p=2.3788E-13; Males: P30: G: p=0.302; GI: p=0.552; P120: G: p=0.050; 175 GI: p=0.003). This data suggests that while both males and females exhibit a walk deficit at P120, the females exhibit a more significant effect and therefore may be more susceptible to this phenotype deficit than males. Additionally, while neither KI males nor females exhibited a translocation phenotype at P30, at P120 the KI females exhibited a significant translocation deficit and the males did not (Females: P30: G: p=0.778, GI: p=0.179; P120: G: p=0.114; GI: p=0.000001; Males: P30: G: p=0.542; GI: p=0.287; P120: G: p=0.135; GI: p=0.283). Because we identified a significant translocation effect in the KI females at p120, we decided to separate out the contributing x- and y-coordinate translocation. We found that a P30, neither the x- nor the y-coordinate translocation exhibited a significant effect. However, at P120, we found that there was a significant decrease in the KI x-coordinate and y-coordinate translocation separately. This result corroborates our finding that the female KI mice exhibit a decrease in walking behavior, which is an x-coordinate behavior, as well as a decrease in rearing, which is a y-coordinate behavior. At P120, decreased walking in both males and females as well as decreased rearing and translocation in females could suggest a neuromotor or cognitive deficit that is specific to the KI females. As the KI females exhibit a decrease in walking, translocation and rearing at P120 and the males exhibit a decrease in only walking, it is possible that the KI female mice have an intrinsic level of vulnerability that is impacting their deficits. A possible factor contributing to this specific vulnerability in female mice may be related to an increase in estrogen levels. While studies have demonstrated that women are twice as likely to develop Alzheimer’s disease (AD) than men, and those 176 who undergo estrogen replacement therapy have a lower risk for developing AD, no such evidence exists for FTD [111]. While some studies indicate that estrogen replacement therapy (ERT) puts individuals at a decreased risk for developing AD, interestingly, one study has shown that female patients who undergo ERT may place them at a higher risk for developing FTD [111] [112]. Levine et al. demonstrated that 70% of women diagnosed with FTD (UCSF-ADRC) were undergoing ERT when they were evaluated. To verify that this result was not driven by a sampling bias, Levine et al. also found that 20% of AD patients within the same facility were undergoing ERT at the time of diagnosis, and this percentage is comparable with the percentage of healthy women who also undergo ERT (24%), demonstrating that ERT is not increasing the susceptibility of these AD patients [111]. Additionally, men and women both have the same risk of developing FTD, indicating that estrogen loss during menopause does not result in an increase in FTD susceptibility [111]. Relating the potential correlation between FTD development in women and estrogen use to our results, perhaps the behavioral deficits we observe specifically in females relates to an increase in estrogen. Additionally, Frey et al. (1995) demonstrated that seven-month-old estrus and diestrus females performed worse on the morris water maze than males in which they took longer and swam longer distances to identify the platform [113]. This could suggest a possible decrease in spatial acquisition in female rodents compared to males, which could relate to the behavioral deficits we observed in females compared to males. 177 Conclusion ALS-FTD is a complex disease that results in both neuromotor and cognitive deficits. The TDP-43Q331K walk deficit we identified could therefore be related to abnormalities stemming from neuromotor abnormalities, cognitive dysfunction or both. The behavioral deficits that we identified at P30 and P120 demonstrate that pathology within these mice occurs early. Early intervention may prolong function, and ACBM could be used as a tool to assess possible future interventions within this novel model. Methods Mice. Twenty TDP-43Q331K knock-in mice (10 males, 10 females) and 20 WT mice (10 males, 10 females) were bred within the Animal Care Facility at Brown University. Approximately 1-2-month old females were set up with 2-3-month-old males in ten breeding trio cages (one male, two females per cage). Pups were weaned at P20 and littermates remained housed together (genders separated). During breeding, we generated four groups of mice, separating by differences in their birthdates. Group 1 mice were born on 6/16/17, group 2 mice were born on 6/23/17, group 3 mice were born on 7/12/17, and group 4 mice were born on 7/26/17). The purpose of having four groups of mice was to generate enough mice of each gender and genotype (20 KI, 20 WT [10/gender]). Therefore, we staggered 178 our ACBM experiments accordingly to ensure that all of the mice were being tested at the same age. Automated Continuous Behavioral Monitoring Technology. We used ACBM to characterize the behavior of 20 TDP-43Q331K (10/gender) and 20 WT littermate controls (10/gender) at two different ages (P30 and P120) (see methods from chapter one and two for an in depth explanation of ACBM technology). ACBM Statistical Analysis. Regardless of groups (1-4; based on the mouse birthdates), data from mice of the same genotype and gender were combined in statistical analyses. For each behavior, as well as for translocation, a general linear model repeated measures analysis of variance (ANOVA) was used to identify genotype effects (between-subjects variable), and genotype x hour interaction effects (within- subjects variable). We conducted these analyses across all twenty-four circadian hours. Standard error was used to generate error bars, and statistical significance is defined as p≤0.05. 179 Supplementary Figures. Supplementary Figure 1. ACBM detects a TDP-43Q331K walk deficit at P30 and a walk and translocation deficit at P120 in females (Rescaled Y-axes). We conducted our analyses across the twenty-four circadian hours. Repeated Measures ANOVA; G: genotype effect; G-I: genotype x hour interaction. a. TDP-43Q331K female mice exhibit a walk deficit at P30 and P120 (P30: G: P=0.0.432; G-I: P=0.040; P120: G: P=0.012; G-I: p=2.378E-13). b. TDP-43Q331K female mice do not exhibit a translocation phenotype at P30 but exhibit a translocation deficit P120 (P30: G: p=0.0.778; G-I: p=0.179; P120: G: p=0.114; G-I: P=0.000001). 180 Supplementary Figure 2. ACBM detects a TDP-43Q331K walk deficit at P120 in males (Rescaled Y-axes). The x-coordinate corresponds to circadian hours 0-23. For walk, the y-coordinate corresponds to the average number of seconds/hour averaged over the five recording days that the mice are performing walking. For translocation, the y- coordinate corresponds to the average number of pixels/hour over the five recording days that the mouse is moving in the x- and y-directions. We conducted our analyses across the twenty-four circadian hours. Repeated Measures ANOVA; G: genotype effect; G-I: genotype x hour interaction. a. TDP-43Q331K male mice do not exhibit a walk phenotype at P30, but exhibit a decrease in walking at P120 (P30: G: P=0.0.302; G-I: P=0.552; P120: G: P=0.050; G-I: P=0.003). b. TDP-43Q331K male mice do not exhibit a translocation phenotype at P30 or P120 (P30: G: P=0.0.542; G-I: P=0.287; P120: G: P=0.135; G-I: P=0.283). 181 CHAPTER 5: DISCUSSION 182 Overall Conclusions The focus of this dissertation was to identify early behavioral phenotypes that are present prior to the onset of neurodegeneration in the transgenic SOD1G93A ALS mouse model and the novel knock-in TDP-43Q331K ALS-FTD mouse model. While traditional behavioral techniques, such as grip strength, rotarod, and wire hang, can detect many overt, later-onset behavioral phenotypes, the detection of more subtle phenotypes may be inconsistent or not detected at all due to uncontrollable variability as well as short testing periods. Therefore, we used Automated Continuous Behavioral Monitoring (ACBM), a novel, highly quantitative, and sensitive deep behavioral phenotyping technique to detect, identify and quantify the onset and progression of specific phenotypes within these mouse models. The SOD1G93A mouse model of ALS exhibits early pathology such as synaptic vesicle stalling and axonal thinning at approximately P30, as well as later pathology such as denervation and degeneration [34-37] [70] [71] [36] [72] [35, 37]. However, behavioral abnormalities in these mice are not detected until ~P90, suggesting that either these pathologies do not result in behavioral change until then, or more likely, the current technology is not sensitive enough to detect change [86, 87]. As this mouse is extensively studied for ALS research, we sought to fill in the early gap of its behavioral characterization. In addition to the SOD1G93A mouse model of ALS, we also characterized early behavioral phenotypes in the TDP-43Q331K knock-in mouse. As this is a novel mouse model of ALS-FTD, behavioral characterization had yet to be explored. White et al. (2018) conducted rotarod, marble-burying and the five-choice serial reaction time 183 task (5-CSRTT) [82]. They identified a rotarod deficit at six months, a reduction in reaction time on the CSRTT at four months, and a marble-burying deficit between 5- 18 months (the ages of mice tested varied) [82]. These data may suggest cognitive or neuromotor deficits. Therefore, we used ACBM to further probe the behavior of each of these mouse models with the goal of identifying earlier and more specific phenotypes. In Chapter 2, we conducted ACBM on the SOD1G93A mice across three ages (P30, P58 and P86) and identified specific behavioral phenotypes as early as P30. In Chapter 3, we conducted ACBM on the TDP-43Q331K mouse across five ages (P120, P180, P225, P300 and P345) and characterized early and longitudinal behavioral phenotypes. Because we identified TDP-43Q331K behavioral phenotypes at P120, we conducted ACBM on a second cohort of TDP-43Q331K mice at P30 and P120 (Chapter 4). We identified a female-specific deficit in the TDP-43Q331K knock-in mouse at P30 and deficits in both genders at P120. These data not only provide a deeper characterization of behavior within these mouse models but also opens the doors to further applying behavioral technology to phenotyping different mouse models of disease. Just as the field of genomics has enabled the characterization, quantification and comparison of genes and their resulting proteins and pathways within and across species, ACBM and further improved behavioral technologies, can provide an additional biological database to generate a more complete phenotypic picture of a disease model. 184 Early and Longitudinal Behavioral Phenotypes in the Transgenic SOD1G93A ALS mouse model: Implications and Future Directions We used ACBM to identify early behavioral phenotypes in the SOD1G93A mouse at P30, P58, and P86 (Chapter 2). We chose these ages because they represent time points prior to previously identified behavioral deficits within these mice [72]. Using ACBM, we characterized three types of phenotypes that change throughout early disease progression: early and transient, later onset (initiating at P58), and phenotypes that switch their directionality with disease progression. SOD1G93A Walking Phenotype The SOD1G93A mice exhibited an early walk deficit at P30 that sustained at P58 but disappeared at P86 (Chapter 2: Fig. 2a). This early and transient phenotype potentially demonstrates a neuromotor deficit and interestingly, this age represents a time when motor neuron denervation is occurring. This deficit could also represent other pathologies such as synaptic vesicle stalling and axonal thinning that are occurring during this same point in disease progression. SOD1G93A Eating-on-Haunches Phenotype The SOD1G93A mice also exhibit a later onset eating-on-haunches deficit at P58 and P86 (Chapter 2: Fig. 2c). Some evidence suggests that these mice exhibit a decrease in mastication rate, and the decrease in eating-on-haunches behavior could be modeling this phenotype [99]. It is also possible that the SOD1G93A mice have a 185 decreased metabolism resulting in a decreased drive to eat, which could also explain their deficit in eating-on-haunches behavior [100]. SOD1G93A Translocation Phenotypes The SOD1G93A mice also exhibit an early translocation deficit at P30 (Chapter 2: Fig. 2b). In separating translocation into x- and y-coordinate movement at this age, we found that there is a decrease in both x-coordinate and y-coordinate translocation (Chapter 2: Fig. 3a). As walking is an x-coordinate behavior, the decrease in x-coordinate translocation is likely modeling the walk deficit. It is interesting that despite the significant SOD1G93A deficit in y-coordinate translocation, there are no significant deficits in behaviors that one may consider as y-coordinate behaviors, specifically rearing and hanging (Chapter 2: Supplementary Fig. 5). However, looking at the graphs, specifically the shapes of the curves, hanging and rearing mimic the shape of the y-coordinate translocation effect. Although not significant, there appears to be a trend toward an SOD1G93A deficit in rear and hanging during the same hours as the y-coordinate translocation deficit. Because translocation incorporates all movement and is not limited to specific behavioral classifications, it can incorporate a larger amount of data than any individual behavior. Therefore, the y-coordinate translocation may be detecting a potential deficit in rearing and hanging behavior combined, and this combination of y- coordinate behaviors increases the power of the measurement, which may result in a significant SOD1G93A y-coordinate deficit. 186 At P58, the translocation deficit is gone (Chapter 2: Fig. 2b). This result may have to do with a cancellation effect in which the walking and eating-on-haunches deficit may cancel out with increased SOD1G93A drinking behavior (Chapter 2: Fig. 2a, c, Supplementary Fig. 3, 4). In separating translocation into x- and y-coordinate movement, walk would be characterized as an x-coordinate movement and drink would be characterized as a y-coordinate movement as drinking behavior consists of the mice rearing in the y-coordinate toward the drinking spout. Eating-on- haunches, however, consists of micromovements in both the x- and y-coordinates, as the mouse is hunched over performing quick movements in both directions. At P58, the mice spend approximately ten times as much time eating-on-haunches as walking and drinking. Therefore, the eating-on-haunches behavior, a non- directional behavior, could wash out the SOD1G93A increase in walking and decrease in drinking behavior, potentially explaining the absence of not only the combined x- and y-coordinate translocation, but also the loss of the x-coordinate and y- coordinate translocation phenotypes separately. Interestingly, at P86, the SOD1G93A mice exhibit an increase in translocation (Chapter 2: Fig. 2b). This is likely a result of the absence of the walk deficit (Chapter 2: Fig. 2a). While not significant, we also observed an increase in drinking behavior at P86, also potentially contributing to the increase in translocation at this age. Again separating out translocation into x- and y-coordinate movement, we found a significant increase in y-coordinate translocation but not in x-coordinate translocation (Chapter 2: Fig. 3c). This is not surprising due to the lack of x- 187 coordinate walking behavior and the trend toward an increase in SOD1G93A drinking, hanging and rearing behavior at this age. Transgenic SOD1G93A ACBM Future Directions To follow-up on our results in this study, a future direction would be to increase the number of mice in this experiment. Here, our group sizes were relatively small (five SOD1G93A mice and six WT). Increasing the group size to ten SOD1G93A and ten WT may allow us to discern specific phenotypes that with a small group size, currently appear to be trending. For example, while SOD1G93A mice exhibited an increase in drinking behavior at P58, the genotype effect was not present at P30 and P86, despite the trending increase at these ages during specific hours (Chapter 2: Supplementary Fig. 3, 4). Increasing the number of mice may elucidate these potential effects, and therefore provide a more accurate assessment of drinking phenotypes within this mouse. While we chose to focus on only males in this study, a future direction would be to repeat the experiment with female SOD1G93A and WT mice. Human females have a lower risk of developing ALS [29] [30, 31]. We could use ACBM to determine whether this lower risk extends to SOD1G93A mice. Early and Longitudinal Behavioral Phenotypes in the Knock-In TDP-43Q331K ALS-FTD mouse model: Implications and Future Directions We conducted two separate studies on the novel knock-in TDP-43Q331K mouse model of ALS-FTD and identified early behavioral phenotypes in both 188 studies. In our first study (Chapter 3), we conducted five rounds of ACBM on the same cohort of mice in which the youngest age was P120. These mice exhibited a walking deficit at P120 in both males and females (Chapter 3: Fig. 2c, d, Supplementary Fig. 1b). Interestingly, the walk deficit sustains in males across all five ages, yet in females resolves at P180 and P225 prior to a progressive walk deficit at P300 and P345 (Chapter 3: Supplementary Fig. 1b). This is interesting because it suggests that during these ages, the TDP-43Q331K females may exhibit a recovery period while the males do not. Bargsted et al. (2017) demonstrated that in a different transgenic TDP- 43A315T mouse model of ALS-FTD with an abbreviated lifespan, male mice exhibit a shorter survival than female mice (males: ~P90; females: ~ P140) [114]. Both genders exhibit a much more aggressive form of ALS-FTD than the TDP-43Q331K mice. They exhibit spinal motor neuron degeneration, TDP-43 aggregates in the cortex and spinal cord, hind limb clasping, hanging task deficits, and a hunched posture throughout disease progression. Regarding gender differences within these mice, the TDP-43A315T females perform better than males on a hanging task however exhibit increased levels of neuronal loss in caudal regions of spinal cord tissue in comparison to the males. These data suggest that behavioral deficits do not always correlate with spinal motor neuron pathology. Perhaps the female decrease in spinal motor neurons may be related to ALS pathology, while the male decrease in hanging performance may be related to FTD pathology in the TDP-43A315T mouse. It is possible that this gender difference corroborates our behavioral data. We found that despite the fact that TDP-43Q331K males do not exhibit 189 neurodegeneration (P150) nor denervation at P150 or P540-P690 (Chapter 3: Supplementary Fig. 2c, d, e), males exhibit a walk deficit at P120, P180, P225, P300, and P345 (Chapter 3: Supplementary Fig. 1b). However, females exhibit a walking deficit at only P120, P300 and P345 (Supplementary Fig. 1b). Because we did not detect denervation in either genders, it is plausible that the increased walking deficit in males may instead be related to the “FTD” symptomology – not neuromotor deficits. Therefore, in our study, the TDP-43Q331K females may be less vulnerable to underlying pathology than the TDP-43Q331K males. There may be different mechanisms that underlie the behavioral abnormalities in ALS-FTD mouse models [114] [82]. A future direction would be to analyze neurodegeneration and denervation in female TDP-43Q331K mice longitudinally. Because we observed an early walking deficit in the TDP-43Q331K male and female mice at P120, we conducted a separate study on a larger cohort of male and female TDP-43Q331K and WT mice (10 TDP-43Q331K and 10 WT [10/gender]) to determine if the walking deficit we observed at P120 was present at P30, as well as to replicate our P120 results (Chapter 4). At P30, the TDP-43Q331K males did not exhibit a walk deficit however did at P120 (Chapter 4: Fig. 2a). This result suggests that young P30 TDP-43Q331K males are potentially protected from developing this deficit, yet older mice are not. Female TDP-43Q331K mice, however, exhibit an early P30 walk deficit that progresses at P120 (Chapter 4: Fig. 1a). This data again suggests a difference in disease/phenotype progression between the male and female TDP-43Q331K mice. 190 We also identified a translocation deficit in the TDP-43Q331K female mice that presents at P120 (Chapter 4: Fig. 1b). There was no translocation phenotype at P30 or P120 in the males again revealing gender differences within these mice (Chapter 4: Fig. 2b). This further demonstrates that translocation is not a direct correlate of walking behavior, as the males exhibit a walking deficit at P120 while females exhibit a walking deficit at P30 (Chapter 4: Fig. 1a, b, 2a, b). TDP-43Q331K Cohort 1 and TDP-43Q331K Cohort 2: A Comparison In comparing TDP-43Q331K cohort 1 (Chapter 3) and cohort 2 (Chapter 4), at P120, we found and replicated a walk deficit in both males and females. However, we identified some interesting differences at this age. Cohort 1 males exhibit about twice as much walking behavior in both TDP-43Q331K and WT mice than they do in cohort 2 (Chapter 3: Supplementary Fig. 1b; Chapter 4: Fig. 2a). A few possibilities could explain this result. One possibility is that both WT and TDP-43Q331K cohort 1 mice truly exhibit increased walking behavior compared to cohort 2 mice. This could be due to different environmental factors that influenced the behavior of both genotypes. Both cohorts had different breeding environments. Cohort 1 mice were bred at the University of Massachusetts Medical School and shipped to our animal facility at Brown University where they remained in quarantine for sixty days. Cohort 2 mice were bred at Brown and therefore did not go into quarantine but instead went into ACBM at P30. Additionally, the cohort 1 mice were not maintained as littermates, however the cohort 2 mice were. Perhaps these differences in 191 breeding and rearing resulted in a consistent difference in male walking behavior between cohort 1 and cohort 2. Alternatively, something environmental that would affect both genotypes may have changed between cohort 1 and cohort 2. While no one enters the ACBM room during an experiment, outside uncontrollable noise could disrupt the mouse behavior. If there was a difference in noise and activity level outside of the ACBM room for just one of the cohorts, that would impact the mouse behavior. Additionally, as approximately 1.5 years elapsed between cohort 1 and cohort 2 data collection, it is possible that differences in lighting or smell could have changed, and this would also result in behavioral differences between cohorts. A change in lighting between cohorts, could also affect the ability of ACBM technology to make consistent behavioral calls. Additionally, while the female TDP-43Q331K mice in cohort 1 and cohort 2 both exhibited a walk deficit at P120, the deficit was more significant in cohort 2 (Chapter 3: Supplementary Fig. 1b; Chapter 4: Fig. 1a). Specifically, between cohort 1 and 2, while the WT females maintained approximately the same amount of walking between cohorts, the TDP-43Q331K females exhibited a larger walk deficit in cohort 2, resulting in a greater genotype difference. Therefore, between cohorts, the walking behavior of the TDP-43Q331K females changes relative to that of the WT females, demonstrating that the change in significance in female walking behavior may be genotype-dependent. It is possible that the doubling in sample size from cohort 1 to cohort 2 could result in this change. 192 Additionally, the female TDP-43Q331K mice may be more sensitive to potential differences in environmental change, such as those listed above, than the male TDP- 43Q331 mice. If such environmental factors were different between cohorts, perhaps these had a larger impact on mouse behavior in females than in males. Comparing Underlying Cortical Pathology of the SOD1G93A Transgenic mouse model of ALS to the TDP-43Q331K Mouse Model of ALS-FTD While both the SOD1G93A transgenic and the TDP-43Q331K mice exhibit behavioral abnormalities, it is possible that these behavioral phenotypes are driven by discrete pathological mechanisms. The SOD1G93A mice exhibit early motor neuron denervation and degeneration similar to what human ALS patients exhibit, and this mouse model is therefore likely modeling such pathologies present in ALS patients. While most research on SOD1G93A mice demonstrate spinal cord and peripheral nervous system abnormalities, few studies demonstrate any cortical pathology. Alexander et al. (2000) demonstrate an absence of pathology in the sensorimotor cortex of the SOD1G93A mice and normal pyramidal neurons [115]. However, they identified an increase in the extracellular fluid levels of glutamate and aspartate, both excitatory amino acids, within the sensorimotor cortices of these mice [115]. These increased levels may result in an increase in excitotoxicity within the cortex of these mice, and provide evidence that SOD1G93A mice may exhibit early cortical pathology, likely contributing to disease progression. The TDP-43Q33K mouse may also be susceptible to cortical excitability. As described in Chapter 4, White et al. (2018) demonstrated a decrease in 193 parvalbumin-positive interneurons in the frontal cortices of these mice [82]. Because parvalbumin is a calcium binding protein, its decrease could potentially result in a decrease in calcium buffering and an increase in excitatory activity. While increased excitatory activity has not been found through electrophysiological recordings of the flexor digitorum brevis muscles during graded nerve stimulation in this mouse, cortical recordings have yet to be performed [82]. Characterizing both upper and lower motor neuron abnormalities in both the SOD1G93A transgenic and TDP-43Q331K knock-in mouse models would provide a larger picture of pathology and may enable the identification of shared pathways. Future Uses for Automated Continuous Behavioral Monitoring ACBM has provided a remarkable tool to identify early and longitudinal behavioral phenotypes in the SOD1G93A transgenic and TDP-43Q331K knock-in mouse models of ALS, and ALS-FTD, respectively. Not only can ACBM be used for identifying phenotypes that may not be detected through traditional behavioral measures, but also could be used to develop large behavioral databases of disease mouse models. Fully characterizing disease requires multiple experimental methods. Rigorous, quantifiable, and reproducible behavioral data should be part of this characterization. Treatment efficacy in mice could be measured through ACBM. Because ACBM enables longitudinal assessment, the effects of treatment can be measured over time. Given the late onset of disease diagnosis, it is possible that already available treatments like riluzole or radicava would be effective at earlier stages of disease. To 194 test how early riluzole could have an impact on changing the course of disease progression in mouse models of ALS, one could use ACBM to assess behavioral change due to riluzole treatment at different stages of disease. If ACBM identifies subtle behavioral improvements or changes in treated mice compared to untreated mice, then ACBM could detect specific windows of time during which riluzole has a therapeutic effect in these mice. Pathology occurring during or after those windows could be examined to provide insight into what exactly riluzole is targeting. The same concept could be applied to Edaravone. Additionally, with the identification of specific suppressor genes whose perturbation could slow or prevent ALS, one could use ACBM to test behavioral phenotype changes in knock-in mice with a mutated form of that suppressor gene. Additionally, one could generate a knock-out mouse with the ALS suppressor gene removed and run the mouse on ACBM to determine if that mouse exhibits a behavioral difference in comparison to an ALS mouse model with normal expression of the suppressor gene. One may hypothesize that the knock-out mouse may exhibit a complete or partial phenotype rescue in comparison to the ALS mouse model that expresses the suppressor gene. Additionally, to further implicate that gene in disease progression, one could target that gene pharmacologically and test ALS mouse models with and without such pharmacological treatment to determine if there is a rescue in specific behaviors using ACBM. Longitudinal behavioral analysis would provide important information on when intervention is most therapeutic, how long therapeutic benefits last, what effects do varying dosages have on efficacy, does efficacy change with age, and does efficacy very between males and females – 195 to name just a few benefits. This information could improve efficiency in therapy development through more targeted approaches to animal testing. 196 WORKS CITED 197 1. Petersen, R.C., et al., Mild cognitive impairment: clinical characterization and outcome. Arch Neurol, 1999. 56(3): p. 303-8. 2. Tampi, R.R., et al., Mild cognitive impairment: A comprehensive review. Healthy Aging Research, 2015. 4: p. 1-11. 3. Corder, E.H. and M.A. Woodbury, Genetic heterogeneity in Alzheimer's disease: a grade of membership analysis. Genet Epidemiol, 1993. 10(6): p. 495-9. 4. Kane, R.L., et al., AHRQ Comparative Effectiveness Reviews, in Interventions to Prevent Age-Related Cognitive Decline, Mild Cognitive Impairment, and Clinical Alzheimer's-Type Dementia. 2017, Agency for Healthcare Research and Quality (US): Rockville (MD). 5. Kinsella, G.J., et al., Early intervention for mild cognitive impairment: a randomised controlled trial. J Neurol Neurosurg Psychiatry, 2009. 80(7): p. 730-6. 6. Cooper, C., et al., Treatment for mild cognitive impairment: systematic review. Br J Psychiatry, 2013. 203(3): p. 255-64. 7. Nagaraja, D. and S. Jayashree, Randomized study of the dopamine receptor agonist piribedil in the treatment of mild cognitive impairment. Am J Psychiatry, 2001. 158(9): p. 1517-9. 8. Wijesekera, L.C. and P.N. Leigh, Amyotrophic lateral sclerosis. Orphanet J Rare Dis, 2009. 4: p. 3. 9. Watts, C.R. and M. Vanryckeghem, Laryngeal dysfunction in Amyotrophic Lateral Sclerosis: a review and case report. BMC Ear Nose Throat Disord, 2001. 1(1): p. 1. 10. Swinnen, B. and W. Robberecht, The phenotypic variability of amyotrophic lateral sclerosis. Nat Rev Neurol, 2014. 10(11): p. 661-70. 11. Picher-Martel, V., et al., From animal models to human disease: a genetic approach for personalized medicine in ALS. Acta Neuropathol Commun, 2016. 4(1): p. 70. 12. Chio, A., et al., Global epidemiology of amyotrophic lateral sclerosis: a systematic review of the published literature. Neuroepidemiology, 2013. 41(2): p. 118-30. 198 13. Huynh, W., et al., Assessment of the upper motor neuron in amyotrophic lateral sclerosis. Clin Neurophysiol, 2016. 127(7): p. 2643-60. 14. de Carvalho, M., et al., Electrodiagnostic criteria for diagnosis of ALS. Clin Neurophysiol, 2008. 119(3): p. 497-503. 15. Okita, T., et al., Can Awaji ALS criteria provide earlier diagnosis than the revised El Escorial criteria? J Neurol Sci, 2011. 302(1-2): p. 29-32. 16. Cedarbaum, J.M. and N. Stambler, Performance of the Amyotrophic Lateral Sclerosis Functional Rating Scale (ALSFRS) in multicenter clinical trials. J Neurol Sci, 1997. 152 Suppl 1: p. S1-9. 17. Duleep, A. and J. Shefner, Electrodiagnosis of motor neuron disease. Phys Med Rehabil Clin N Am, 2013. 24(1): p. 139-51. 18. Joyce, N.C. and G.T. Carter, Electrodiagnosis in persons with amyotrophic lateral sclerosis. Pm r, 2013. 5(5 Suppl): p. S89-95. 19. Buzgova, R., et al., [The review of questionnaires and scales evaluating patients with amyotrophic lateral sclerosis]. Cas Lek Cesk, 2018. 157(1): p. 41-45. 20. Cedarbaum, J.M., et al., The ALSFRS-R: a revised ALS functional rating scale that incorporates assessments of respiratory function. BDNF ALS Study Group (Phase III). J Neurol Sci, 1999. 169(1-2): p. 13-21. 21. Bensimon, G., L. Lacomblez, and V. Meininger, A controlled trial of riluzole in amyotrophic lateral sclerosis. ALS/Riluzole Study Group. N Engl J Med, 1994. 330(9): p. 585-91. 22. Bellingham, M.C., A review of the neural mechanisms of action and clinical efficiency of riluzole in treating amyotrophic lateral sclerosis: what have we learned in the last decade? CNS Neurosci Ther, 2011. 17(1): p. 4-31. 23. Cheah, B.C., et al., Riluzole, neuroprotection and amyotrophic lateral sclerosis. Curr Med Chem, 2010. 17(18): p. 1942-199. 24. Miller, R.G., et al., Riluzole for amyotrophic lateral sclerosis (ALS)/motor neuron disease (MND). Cochrane Database Syst Rev, 2002(2): p. Cd001447. 25. Cruz, M.P., Edaravone (Radicava): A Novel Neuroprotective Agent for the Treatment of Amyotrophic Lateral Sclerosis. P t, 2018. 43(1): p. 25-28. 26. Logroscino, G., et al., Incidence of amyotrophic lateral sclerosis in Europe. J Neurol Neurosurg Psychiatry, 2010. 81(4): p. 385-90. 199 27. Kim, H.J., et al., Therapeutic modulation of eIF2alpha phosphorylation rescues TDP-43 toxicity in amyotrophic lateral sclerosis disease models. Nat Genet, 2014. 46(2): p. 152-60. 28. Alonso, A., et al., Incidence and lifetime risk of motor neuron disease in the United Kingdom: a population-based study. Eur J Neurol, 2009. 16(6): p. 745- 51. 29. Manjaly, Z.R., et al., The sex ratio in amyotrophic lateral sclerosis: A population based study. Amyotroph Lateral Scler, 2010. 11(5): p. 439-42. 30. Rosen, A.D., Amyotrophic lateral sclerosis. Clinical features and prognosis. Arch Neurol, 1978. 35(10): p. 638-42. 31. Pegoraro, V., A. Merico, and C. Angelini, Micro-RNAs in ALS muscle: Differences in gender, age at onset and disease duration. J Neurol Sci, 2017. 380: p. 58-63. 32. McCombe, P.A. and R.D. Henderson, Effects of gender in amyotrophic lateral sclerosis. Gend Med, 2010. 7(6): p. 557-70. 33. Choi, C.I., et al., Effects of estrogen on lifespan and motor functions in female hSOD1 G93A transgenic mice. J Neurol Sci, 2008. 268(1-2): p. 40-7. 34. Fischer, L.R., et al., Amyotrophic lateral sclerosis is a distal axonopathy: evidence in mice and man. Exp Neurol, 2004. 185(2): p. 232-40. 35. Pun, S., et al., Selective vulnerability and pruning of phasic motoneuron axons in motoneuron disease alleviated by CNTF. Nat Neurosci, 2006. 9(3): p. 408- 19. 36. Kaplan, A., et al., Neuronal matrix metalloproteinase-9 is a determinant of selective neurodegeneration. Neuron, 2014. 81(2): p. 333-48. 37. Frey, D., et al., Early and selective loss of neuromuscular synapse subtypes with low sprouting competence in motoneuron diseases. J Neurosci, 2000. 20(7): p. 2534-42. 38. Lubischer, J.L. and W.J. Thompson, Neonatal partial denervation results in nodal but not terminal sprouting and a decrease in efficacy of remaining neuromuscular junctions in rat soleus muscle. J Neurosci, 1999. 19(20): p. 8931-44. 39. Love, F.M., Y.J. Son, and W.J. Thompson, Activity alters muscle reinnervation and terminal sprouting by reducing the number of Schwann cell pathways that grow to link synaptic sites. J Neurobiol, 2003. 54(4): p. 566-76. 200 40. Tovar, Y.R.L.B., et al., Trophic factors as modulators of motor neuron physiology and survival: implications for ALS therapy. Front Cell Neurosci, 2014. 8: p. 61. 41. Bott, N.T., et al., Frontotemporal dementia: diagnosis, deficits and management. Neurodegener Dis Manag, 2014. 4(6): p. 439-54. 42. Rascovsky, K., et al., Rate of progression differs in frontotemporal dementia and Alzheimer disease. Neurology, 2005. 65(3): p. 397-403. 43. Coyle-Gilchrist, I.T., et al., Prevalence, characteristics, and survival of frontotemporal lobar degeneration syndromes. Neurology, 2016. 86(18): p. 1736-43. 44. Warren, J.D., J.D. Rohrer, and M.N. Rossor, Clinical review. Frontotemporal dementia. Bmj, 2013. 347: p. f4827. 45. Ferrari, R., et al., FTD and ALS: a tale of two diseases. Curr Alzheimer Res, 2011. 8(3): p. 273-94. 46. Onyike, C.U. and J. Diehl-Schmid, The epidemiology of frontotemporal dementia. Int Rev Psychiatry, 2013. 25(2): p. 130-7. 47. Kertesz, A., et al., The evolution and pathology of frontotemporal dementia. Brain, 2005. 128(Pt 9): p. 1996-2005. 48. Ji, A.L., et al., Genetics insight into the amyotrophic lateral sclerosis/frontotemporal dementia spectrum. J Med Genet, 2017. 54(3): p. 145-154. 49. Gaudette, M., M. Hirano, and T. Siddique, Current status of SOD1 mutations in familial amyotrophic lateral sclerosis. Amyotroph Lateral Scler Other Motor Neuron Disord, 2000. 1(2): p. 83-9. 50. Andersen, P.M., et al., Amyotrophic lateral sclerosis associated with homozygosity for an Asp90Ala mutation in CuZn-superoxide dismutase. Nat Genet, 1995. 10(1): p. 61-6. 51. Gurney, M.E., et al., Motor neuron degeneration in mice that express a human Cu,Zn superoxide dismutase mutation. Science, 1994. 264(5166): p. 1772-5. 52. Cudkowicz, M.E., et al., Epidemiology of mutations in superoxide dismutase in amyotrophic lateral sclerosis. Ann Neurol, 1997. 41(2): p. 210-21. 53. Roberson, E.D., Mouse models of frontotemporal dementia. Ann Neurol, 2012. 72(6): p. 837-49. 201 54. Kuo, P.H., et al., Structural insights into TDP-43 in nucleic-acid binding and domain interactions. Nucleic Acids Res, 2009. 37(6): p. 1799-808. 55. Van Langenhove, T., J. van der Zee, and C. Van Broeckhoven, The molecular basis of the frontotemporal lobar degeneration-amyotrophic lateral sclerosis spectrum. Ann Med, 2012. 44(8): p. 817-28. 56. Peters, O.M., et al., Human C9ORF72 Hexanucleotide Expansion Reproduces RNA Foci and Dipeptide Repeat Proteins but Not Neurodegeneration in BAC Transgenic Mice. Neuron, 2015. 88(5): p. 902-909. 57. Cleary, J.D. and L.P. Ranum, Repeat-associated non-ATG (RAN) translation in neurological disease. Hum Mol Genet, 2013. 22(R1): p. R45-51. 58. Lattante, S., et al., Defining the genetic connection linking amyotrophic lateral sclerosis (ALS) with frontotemporal dementia (FTD). Trends Genet, 2015. 31(5): p. 263-73. 59. Chio, A., et al., Severely increased risk of amyotrophic lateral sclerosis among Italian professional football players. Brain, 2005. 128(Pt 3): p. 472-6. 60. Brown, R.H., Jr. and A. Al-Chalabi, Amyotrophic Lateral Sclerosis. N Engl J Med, 2017. 377(16): p. 1602. 61. Lehman, E.J., et al., Neurodegenerative causes of death among retired National Football League players. Neurology, 2012. 79(19): p. 1970-4. 62. Feddermann-Demont, N., et al., Prevalence of potential sports-associated risk factors in Swiss amyotrophic lateral sclerosis patients. Brain Behav, 2017. 7(4): p. e00630. 63. Tabata, R.C., et al., Chronic exposure to dietary sterol glucosides is neurotoxic to motor neurons and induces an ALS-PDC phenotype. Neuromolecular Med, 2008. 10(1): p. 24-39. 64. Bradley, W.G., et al., Is exposure to cyanobacteria an environmental risk factor for amyotrophic lateral sclerosis and other neurodegenerative diseases? Amyotroph Lateral Scler Frontotemporal Degener, 2013. 14(5-6): p. 325-33. 65. de Munck, E., et al., beta-N-methylamino-l-alanine causes neurological and pathological phenotypes mimicking Amyotrophic Lateral Sclerosis (ALS): the first step towards an experimental model for sporadic ALS. Environ Toxicol Pharmacol, 2013. 36(2): p. 243-55. 202 66. Cox, P.A., et al., Cyanobacteria and BMAA exposure from desert dust: a possible link to sporadic ALS among Gulf War veterans. Amyotroph Lateral Scler, 2009. 10 Suppl 2: p. 109-17. 67. Geracitano, R., et al., Altered long-term corticostriatal synaptic plasticity in transgenic mice overexpressing human CU/ZN superoxide dismutase (GLY(93)- ->ALA) mutation. Neuroscience, 2003. 118(2): p. 399-408. 68. Chang, Q. and L.J. Martin, Glycinergic innervation of motoneurons is deficient in amyotrophic lateral sclerosis mice: a quantitative confocal analysis. Am J Pathol, 2009. 174(2): p. 574-85. 69. Sgobio, C., et al., Abnormal medial prefrontal cortex connectivity and defective fear extinction in the presymptomatic G93A SOD1 mouse model of ALS. Genes Brain Behav, 2008. 7(4): p. 427-34. 70. Wooley, C.M., et al., Gait analysis detects early changes in transgenic SOD1(G93A) mice. Muscle Nerve, 2005. 32(1): p. 43-50. 71. Dobrowolny, G., et al., Muscle expression of a local Igf-1 isoform protects motor neurons in an ALS mouse model. J Cell Biol, 2005. 168(2): p. 193-9. 72. Perez-Garcia, M.J. and S.J. Burden, Increasing MuSK activity delays denervation and improves motor function in ALS mice. Cell Rep, 2012. 2(3): p. 497-502. 73. Wu, L.S., et al., TDP-43, a neuro-pathosignature factor, is essential for early mouse embryogenesis. Genesis, 2010. 48(1): p. 56-62. 74. Sephton, C.F., et al., TDP-43 is a developmentally regulated protein essential for early embryonic development. J Biol Chem, 2010. 285(9): p. 6826-34. 75. Kraemer, B.C., et al., Loss of murine TDP-43 disrupts motor function and plays an essential role in embryogenesis. Acta Neuropathol, 2010. 119(4): p. 409- 19. 76. Spiller, K.J., et al., Progression of motor neuron disease is accelerated and the ability to recover is compromised with advanced age in rNLS8 mice. Acta Neuropathol Commun, 2016. 4(1): p. 105. 77. Alfieri, J.A., N.S. Pino, and L.M. Igaz, Reversible behavioral phenotypes in a conditional mouse model of TDP-43 proteinopathies. J Neurosci, 2014. 34(46): p. 15244-59. 78. Gascon, E., et al., Alterations in microRNA-124 and AMPA receptors contribute to social behavioral deficits in frontotemporal dementia. Nat Med, 2014. 20(12): p. 1444-51. 203 79. Sreedharan, J., et al., TDP-43 mutations in familial and sporadic amyotrophic lateral sclerosis. Science, 2008. 319(5870): p. 1668-72. 80. Crawley, J.N., Behavioral phenotyping strategies for mutant mice. Neuron, 2008. 57(6): p. 809-18. 81. Crawley, J.N. and R. Paylor, A proposed test battery and constellations of specific behavioral paradigms to investigate the behavioral phenotypes of transgenic and knockout mice. Horm Behav, 1997. 31(3): p. 197-211. 82. White, M.A., et al., TDP-43 gains function due to perturbed autoregulation in a Tardbp knock-in mouse model of ALS-FTD. Nat Neurosci, 2018. 21(4): p. 552- 563. 83. Jhuang, H., et al., Automated home-cage behavioural phenotyping of mice. Nat Commun, 2010. 1: p. 68. 84. Giancardo, L., et al., Automatic visual tracking and social behaviour analysis with multiple mice. PLoS One, 2013. 8(9): p. e74557. 85. Wiltschko, A.B., et al., Mapping Sub-Second Structure in Mouse Behavior. Neuron, 2015. 88(6): p. 1121-35. 86. Olivan, S., et al., Comparative study of behavioural tests in the SOD1G93A mouse model of amyotrophic lateral sclerosis. Exp Anim, 2015. 64(2): p. 147- 53. 87. Alves, C.J., et al., Early motor and electrophysiological changes in transgenic mouse model of amyotrophic lateral sclerosis and gender differences on clinical outcome. Brain Res, 2011. 1394: p. 90-104. 88. Fujita, M., et al., Light/dark phase-dependent spontaneous activity is maintained in dopamine-deficient mice. Molecular Brain, 2017. 10(1): p. 49. 89. Oyanedel, C.N., et al., Peripheral and central blockade of interleukin-6 trans- signaling differentially affects sleep architecture. Brain Behav Immun, 2015. 50: p. 178-185. 90. Guertin, P.A., Preclinical evidence supporting the clinical development of central pattern generator-modulating therapies for chronic spinal cord-injured patients. Front Hum Neurosci, 2014. 8: p. 272. 91. Zehr, E.P. and J. Duysens, Regulation of arm and leg movement during human locomotion. Neuroscientist, 2004. 10(4): p. 347-61. 204 92. The intrinsic factors in the act of progression in the mammal. Proceedings of the Royal Society of London. Series B, Containing Papers of a Biological Character, 1911. 84(572): p. 308-319. 93. Brown, T.G., On the nature of the fundamental activity of the nervous centres; together with an analysis of the conditioning of rhythmic activity in progression, and a theory of the evolution of function in the nervous system. J Physiol, 1914. 48(1): p. 18-46. 94. Knikou, M. and B.A. Conway, Effects of electrically induced muscle contraction on flexion reflex in human spinal cord injury. Spinal Cord, 2005. 43(11): p. 640-8. 95. Lundberg, A., Multisensory control of spinal reflex pathways. Prog Brain Res, 1979. 50: p. 11-28. 96. Clark, D.J., Automaticity of walking: functional significance, mechanisms, measurement and rehabilitation strategies. Frontiers in Human Neuroscience, 2015. 9(246). 97. Dominici, N., et al., Locomotor primitives in newborn babies and their development. Science, 2011. 334(6058): p. 997-9. 98. Strand, E.A., et al., Management of oral-pharyngeal dysphagia symptoms in amyotrophic lateral sclerosis. Dysphagia, 1996. 11(2): p. 129-39. 99. Lever, T.E., et al., An animal model of oral dysphagia in amyotrophic lateral sclerosis. Dysphagia, 2009. 24(2): p. 180-95. 100. Smittkamp, S.E., et al., SOD1-G93A mice exhibit muscle fiber type-specific decreases in glucose uptake in the absence of whole body changes in metabolism. Neurodegener Dis, 2014. 13(1): p. 29-37. 101. Mead, R.J., et al., Optimised and rapid pre-clinical screening in the SOD1(G93A) transgenic mouse model of amyotrophic lateral sclerosis (ALS). PLoS One, 2011. 6(8): p. e23244. 102. Alexianu, M.E., et al., The role of calcium-binding proteins in selective motoneuron vulnerability in amyotrophic lateral sclerosis. Ann Neurol, 1994. 36(6): p. 846-58. 103. Lodge, D.J., M.M. Behrens, and A.A. Grace, A loss of parvalbumin-containing interneurons is associated with diminished oscillatory activity in an animal model of schizophrenia. J Neurosci, 2009. 29(8): p. 2344-54. 205 104. Burrell, J.R., et al., Motor neuron dysfunction in frontotemporal dementia. Brain, 2011. 134(Pt 9): p. 2582-94. 105. Seebacher, F. and I. Walter, Differences in locomotor performance between individuals: importance of parvalbumin, calcium handling and metabolism. J Exp Biol, 2012. 215(Pt 4): p. 663-70. 106. Angoa-Perez, M., et al., Marble burying and nestlet shredding as tests of repetitive, compulsive-like behaviors in mice. J Vis Exp, 2013(82): p. 50978. 107. Perry, R.J. and J.R. Hodges, Differentiating frontal and temporal variant frontotemporal dementia from Alzheimer's disease. Neurology, 2000. 54(12): p. 2277-84. 108. Hughes, R.N., Intrinsic exploration in animals: motives and measurement. Behav Processes, 1997. 41(3): p. 213-26. 109. Sestakova, N., et al., Determination of motor activity and anxiety-related behaviour in rodents: methodological aspects and role of nitric oxide. Interdiscip Toxicol, 2013. 6(3): p. 126-35. 110. Gossink, F.T., et al., Psychosis in behavioral variant frontotemporal dementia. Neuropsychiatr Dis Treat, 2017. 13: p. 1099-1106. 111. Levine, A.J. and L. Hewett, Estrogen replacement therapy and frontotemporal dementia. Maturitas, 2003. 45(2): p. 83-8. 112. Dye, R.V., et al., Hormone replacement therapy and risk for neurodegenerative diseases. Int J Alzheimers Dis, 2012. 2012: p. 258454. 113. Frye, C.A., Estrus-associated decrements in a water maze task are limited to acquisition. Physiol Behav, 1995. 57(1): p. 5-14. 114. Bargsted, L., et al., Disulfide cross-linked multimers of TDP-43 and spinal motoneuron loss in a TDP-43(A315T) ALS/FTD mouse model. Sci Rep, 2017. 7(1): p. 14266. 115. Alexander, G.M., et al., Elevated cortical extracellular fluid glutamate in transgenic mice expressing human mutant (G93A) Cu/Zn superoxide dismutase. J Neurochem, 2000. 74(4): p. 1666-73. 206